United States Environmental Protection Agency

   Office of Water Regulations and Standards
        Industrial Technology Division
 EIGHTH ANNUAL ANALYTICAL SYMPOSIUM


               Norfolk, Virginia

               April 3-*, 1985

-------

-------
                                FOREWORD
      The  Industrial  Technology  Division of  the USEPA  Office  of Water
Regulations  and Standards  sponsors the  Annual  Analytical  Symposium  to
provide a forum where scientists and other interested parties can present new
ideas and advances in methodology for the analysis of Priority Pollutants. The
topics addressed during the Eighth Annual Analytical Symposium have deviated
from  strict analytical water  chemistry; we  have considered  the  role  that
quality assured biomonitoring  techniques will play in compliance monitoring,
as well as advances  in high performance liquid chromatography/mass  spec-
trometry analysis of environmental  samples.   In  concert with the  Division's
responsibility to promulgate  industrial effluent regulations, we feel  that  it is
in the common interest to promote new technologies and  ideas that  will serve
as analytical tools and assist  us in  this increasingly complex task.
                                          W. A. Telliard

-------

-------
                   EIGHTH ANNUAL ANALYTICAL SYMPOSIUM

                     Office of Water Regulations and Standards
                          Industrial Technology Division

                                 April 3-4, 1985
                                Norfolk, Virginia
                                    INDEX

                                  April 3, 1985
Presentation/Speaker
WELCOME AND INTRODUCTION .
William A. Telliard, Chief
Energy and Mining Branch
          and
Devereaux Barnes, Deputy Director
Industrial Technology Division
WATER QUALITY BASED TOXICS CONTROL  ........
Stephen L. Bugbee
USEPA Office of Water Enforcement and Permits

TOXICITY TESTS AND BEST AVAILABLE TECHNOLOGY (BAT)
DETERMINATIONS FOR DISCHARGE FROM OFFSHORE
OIL AND GAS PLATFORMS	 .'  .
Thomas W. Duke
USEPA Environmental Research Laboratory - Sabine Island
Page

   1
BIOLOGICAL ANALYSES OF COMPLEX EFFLUENTS
Teresa Norberg-King
USEPA Environmental Research Laboratory - Duluth

MAGIC-LC/MS: A POWERFUL NEW TOOL FOR
ENVIRONMENTAL ANALYSIS
Richard F. Browner
School of Chemistry, Georgia Institute of Technology
(Paper will be available in open technical literature.)

APPLICATIONS OF THERMOSPRAY HPLC/MS
FOR MONITORING THE ENVIRONMENT  .  . . ...
Robert D. Voyksner
Research Triangle Institute

RECENT ENVIRONMENTAL APPLICATIONS OF
THERMOSPRAY LC/MS	
Marvin Vestal
Chemistry Department, University of Houston
  23
  38
  79
 122
                                       11

-------
                                   INDEX

                                April 3, 1985
Presentation/Speaker
Paee
DETERMINATION OF DYES BY THERMOSPRAY
IONIZATION AND MS/MS	
John M. Ballard
Lockheed Engineering and Management Services Co., Inc.

PROBLEM SOLVING WITH MASS SPECTROMETRY AND FTIR	   19*
Walter M. Shackelford
USEPA Environmental Research Laboratory - Athens

PROGRESS REPORT ON DMR QA STUDIES: QUALITY ASSURANCE
PROGRAM FOR NPDES SELF-MONITORING DATA
Samuel To	   232
USEPA Office of Water Enforcement and Permits
         and
Paul Britton	   242
USEPA Environmental Monitoring and Support Laboratory, ORD

INTER- AND INTRA-LABORATORY ASSESSMENT OF SELECTED
SW-8*6 METHODS FOR ANALYSIS OF APPENDIX VIII COMPOUNDS
IN GROUNDWATER	   282
George H. Stanko
Shell Development Company
                                    ui

-------
                                    INDEX

                                  April *, 1985
Presentation/Speaker
Paee
UTILITY ROUND ROBIN RESULTS FOR THE DETERMINATION
OF ARSENIC AND SELENIUM BY GRAPHITE FURNACE AAS  .........  355
Judith Scott
TRW Energy Development Group

COMPARISON OF METHODS FOR ANALYSIS OF PCBs	..'.....  384
Mitchell D. Erickson
Midwest Research Institute

ANALYSIS OF VOLATILE WATER SOLUBLE COMPOUNDS  ............  436
Denis C.K. Lin
Environmental Testing and Certification Corporation

DETERMINATION OF FIVE-DAY CARBONACEOUS BOD
IN WASTEWATER	  448
James C. Young
Department of Civil Engineering, University of Arkansas


The Battleship USS IOWA	  449


ANALYSIS OF PRIORITY POLLUTANTS BELOW FIVE
NANOGRAMS (ON COLUMN) IN MARINE SEDIMENTS
BY ISOTOPE DILUTION GCMS		471
Peggy Knight
Scientific Services, Weyerhaeuser Company

USES OF ION CHROMATOGRAPHY FOR INORGANIC
ANALYTES IN WATER	  495
John D. Pfaff
USEPA Environmental Monitoring and Support Laboratory, ORD

DIRECT ANALYSIS OF PHENOLS BY HPLC	  519
Suzanne Lesage
Environmental Canada, Environmental Protection, Wastewater
Technology Centre


CLOSING REMARKS	  546


Roster of Attendees	  547
                                      IV

-------

-------
               P R OCEEDINGS
                          MR. TELLIARD:  Good
morning.  My name is Bill Telliard.  I'm from EPA
and I'm here to help you.  Welcome to the Eighth
Annual...boy, that's old...Analytical Meeting of
the Office of Water.  It's the first one for the
Industrial Technology Division or the eighth one
for the Effluent Guidelines Division, depending
on how people score this one.
     I'd like to open this morning's session with
a few words from Dev Barnes, who I don't see.
Dev?  Dev is the Deputy Division Director for the
Industrial Technology Division and I'd like to have
him say a few words.
                          MR. BARNES:  You  found
my hiding place, Bill.  I'd like to welcome every-
body here.  It's a pleasure for the division to
sponsor this every year.  We look forward to this
annual event and we think that it's paid a  lot of
dividends in terms of just getting everybody together
and ironing out differences in various meetings that
take place concurrently with this, dealing with
things that our division directly deals with in
terms of establishing guidelines.  So I think this

-------
meeting has a great deal of value to us.  We hope
that you also feel that way and we look forward to
your continuing participation in it.  So, I hope
you all have a nice time today and tomorrow and
we'll look forward to seeing you again next year,
hopefully.  Thank you.
                          MR. TELLIARD:  A couple
of announcements.  Concurrent with last year's open
sea   disaster, we're repeating it again this year.
We're going to try to get everyone out of here on
time to catch the boat.
     As the tradition holds, we have Tonie and
Sandy from the County Court Reporters taking down
every gem of knowledge that you're about  to drop on
us today.  When you go to the microphone, again,
please  state your name,  otherwise you will be beaten
brutally about the knees; all except George Stanko
because...just tradition.
     In the back of  the  room you'll find  some copies
of previous years proceedings for all of  you who
really  want to remember, and some copies  of the new
1625, 1624 GCMS methodology with the QC stuff in
it.  They're available in the back of the room.
There's also copies  of 304(h); you remember that.
     Breaking with  tradition, our first speaker

-------
this year is not Bob Medz.  Bob didn't come because
he's home being sued, but then again, it just shows
he.did a real good job.
     To show that we're an open minded group this
year, this morning will be spent discussing critters,
both crawling and floating types, and most analytical
chemists will find this a great eye opener.  I
particularly think it will make you feel warm and
fuzzy when you look at standard deviations among
star measurements.and some of these guys.
     So our first speaker this morning is Steve
Bugbee from the Permits Office and Steve is going
to talk about water quality type things.  Steve.

-------
                 STEPHEN L. BUGBEE

   UNITED STATES ENVIRONMENTAL PROTECTION AGENCY
      OFFICE OF WATER ENFORCEMENT AND PERMITS
         WATER QUALITY BASED TOXICS CONTROL


                          MR. BUGBEE:  Good morning.

The use of biological techniques to identify and

control toxicity is not new, but as you will hear,

it has been a very long and tortuous road to get

where we are today.

     The control of toxic pollutants is one of the

Nation's most critical remaining water quality

problems.  For example, localized water quality

standard violations or impairments of water uses

have been widely reported by states in their

biennial water quality reports to EPA*  Because of

the tendency of some toxics to accumulate in fish

tissue, fishing bans and fish consumption warnings

are in effect in many of the Nation's waters.

     For a number of reasons, however, implementing

controls for toxic pollutants has always been a

difficult problem.  These reasons include the sheer

number of toxic compounds which are involved;

insufficient monitoring data because of the high

cost of laboratory analyses for toxics; and

-------
uncertainties about the fate and effects of toxic
substances and uncertainties about the fate and
effects of these chemicals in the aquatic environ-
ment.  EPA has developed a new approach which/
together with advances in the field of biological
monitoring, should help us deal more effectively
with toxic pollution problems.
     The Clean Water Act established two types of
regulatory requirements to control pollutant dis-
charges: technology-based effluent limitations
which reflect the best controls available, consider-
ing the technical and economic achievability of those
controls; and water quality-based effluent limita-
tions which must be met so discharge permits reflect
the more stringent of the two whenever there are
differences.  So, we have on one hand the technology-
based approach which you're most familiar with, and
then we have the water quality-based approach.
     The technology-based requirements for discharges
are currently being issued and will have a substantial
affect in reducing toxic discharges.  However, in
some cases these controls will not be sufficient to
eliminate water quality impacts and enable water
quality standards to be met.  In these cases,  water
quality-based controls are needed.  Two technical

-------
approaches are available for developing water
quality-based effluent limits, the pollutant-specific
approach and the biomonitoring approach.
     EPA and the States have traditionally used the
pollutant-specific approach.  Pollutant-specific
techniques are best used where effluents contain a
few well quantified pollutants and the interactions
and the effects of these pollutants are known.
Thus/ they have worked well for pollutants such as
oxygen-demanding loads and nutrients and heavy
metals.  In addition, pollutant-specific techniques
must be used where health hazards are a concern
and/or if bioaccumulation is suspected.
     In the case of toxic pollutants and complex
effluents, however, it may be difficult in some
cases to determine the attainment or the nonattain-
raent of specific water quality use or water quality
standards and, to set the appropriate limits because
of the complex chemical interactions which affect
the fate and ultimate impact of these toxic sub-
stances in the receiving water.  In many cases,
chemical methods cannot easily be used to identify
all potential toxic pollutants.  Developing numeri-
cal water quality criteria and determining allowable
loadings for all the wide variety of pollutants

-------
found  in effluents  is also very time-consuming and
resource intensive.  In such situations, it would
be very desirable to examine overall toxicity
directly without identifying and analyzing every
pollutant individually.
     Biological methods, based on the direct
measurement of the  impact of whole effluents on
biological test organisms and communities/ provide
this capability.  Therefore, EPA is integrating
this whole effluent biomonitoring into our programs
to control toxic pollutants.  The result is a two-
fold approach.  In  certain situations, an example
would be where potential human health impacts are a
concern, we must rely upon the chemical-specific
approach, measuring the individual toxicants and
evaluating their specific toxic properties.  In
other situations, however, especially where we have
complex effluents, multiple discharges, it is more
appropriate to examine the harmful effects, or
toxicity, of the whole effluent rather than trying
to attempt to individually identify all the poten-
tial toxicants and understand the inherent chemical
reactions.  In other words, we're looking at both
the whole effluent toxicity and the receiving water
body toxicity.

-------
     This toxicity testing approach relies on newly
developed methods and laboratory testing procedures
while the two-fold approach employs both the chemi-
cal-specific and biological monitoring.  The biologi-
cal component deserves a special focus because it is
the area that is developing most rapidly, both in
scientific and programmatic terms.  In fact/ the
scientific basis for using biological techniques
has advanced so significantly in recent years that
it is now an important aspect of the water quality-
based approach for controlling toxic pollutants  in
the EPA's permit program.
     The importance of biological monitoring, both
as an ambient monitoring tool and as a potential
regulatory tool/ has long  been recognized.  We can
point to a long history of field studies that have
employed some type of biomonitoring activity.  How-
ever, the success of biomonitoring as a regulatory
tool has until now been limited because  it  has
often proved difficult to  translate the biological
instream effects  into effluent requirements  that
will be incorporated into  a permit.
     This  is  a very  important point.   Up until
approximately last year, the Chemical  Manufacturers
Association,  among others, had been very critical

-------
of our toxicity-based approach in NPDES permitting.
One of their major criticisms was the fact that we
were unable to make the transition from the end of
the pipe into the receiving water.  In other words/
if you limit the effluent using some type of toxici-
ty limit/ that limit would be very difficult to
enforce because of defining receiving water im-
pacts.
     So to deal with this problem/ we've conducted
research over the last few years which has provided
us with more effective biological monitoring tools
that can help us screen for problems/ assess impacts/
set toxicity-based limits/ and help identify ways
to reduce toxicity.  These toxicity tests are an
outgrowth of those used in the laboratory to estab-
lish aquatic life criteria for specific chemicals.
These tests include acute toxicity tests which
measure short term exposure effects and chronic
toxicity tests which measure long term exposure
effects.
     Through the work of EPA's Office of Water
Regulations and Standards/ the Office of Water
Enforcement Programs, the Office of Research and
Development/ the EPA Regional Offices/ the States
and academia, we have taken these methods into the

-------
field, employed them in site-specific situations,
especially those involving complex effluents. We're
primarily addressing the fresh water environs at
the present time.  However, we do have a similar
program being developed which will deal with
marine and estuarine toxicity.
     The results of these applications have been
promising.  Not only do we have a better understand-
ing of the behavior of toxics in the ambient con-
ditions, but we also have learned that the various
methods can be used in concert to complete the pic-
ture of the water quality impact.  It now appears
that effluent toxicity can be evaluated in conjunc-
tion with chemical and ecological data and can be
very useful in developing regulatory requirements.
We intend to continue these field applications of
biological methods in FY86 and translate their
results into useful guidance for the States to
help in issuing water quality-based permits.
     We are committed to a balanced and integrated
approach; that is, one that applies both the chemical
and biological techniques.  This isn't to alarm you
that we're suddenly shifting over to just the
biological approach.  I think one way of better
illustration is the policy EPA published in the
                         10

-------
Federal Register in March, 1984, that essentially



underwrites this whole water quality-based approach



for toxics control.  So I'll run through a few of



the basics of the policy for you.



SLIDE 1                                ,



     The policy was published in the Federal Register



March 9th, 1984.



SLIDE 2



     Basically, the policy states, one, that we're



going to be controlling pollutants beyond BAT, and



EPA will use an integrated strategy consisting of



biological and chemical methods.  Secondly, through



Section 308 of the Clean Water Act, EPA or a State



may require a discharger to provide chemical,



toxological, or ambient biological data to assure



compliance with standards.



     Third, through Section 402 of the Clean Water



Act, EPA may develop NPDES permit limits based on



effluent toxicity and require a toxicity reduction



evaluation to eliminate unacceptable toxicity.



The fourth point is that the EPA Regional Admini-



strators will assure that each region has the



capability to conduct both chemical, biological



assessments and provide technical assistance to the



States.  This is an ambitious goal but we have been
                        11

-------
Q
lU
PQ
I
LU
u.
o
1U


D.

O


UJ


UJ
      X
      o
      cr
      o
      u.
oc
o
o


o
      cr
      UJ
        12

-------
u.
o
     o

     CD


     I-
     CO
 < CO CO
 CQ Z Q
    O O
 a cj x
 Z    I-
 O >- LU
 >- CD Z
 UJ UJ
 CQ h- «|

 CO OS O
 h- H- ~
 Z CO £
 <    UJ
 I- Q X
 Z) LU O
 -J I-
 _l < Q
 o ac z
 Q. CD <
    UJ
 -I J- -I
 O Z <
 OS •-* CJ
 I-    ~
 Z Z CD
 O < O
 CJ    «J
    UJ O
 O CO ^
I— 3 CQ
                            Qu
                            UJ O
                            H- OS
                            O UJ
                                CD
                            a: <
                            UJ X
                            »- o
                            < co
                               UJ
                               ac
                            UJ =3
                            x o
                            I- UJ
                               a?
                            u.
                            o >-
UJ X
~ I-
CQ •-•

<
   LU
QC CJ
O Z
   <
 ^ ^^

< Q.
CJ £
^ O
CD CJ
O
_J LU
O Q£
CJ Z)
•-* CO
x co
O <
                       H» < LU
                       ^^^     ^^^
                       ^^     pMB
                       £ UJ <
                       as o: z
                       UJ 1-4 •-•
                       a. => £
                          a ^
                      co LU _i
                      LU a: LU
                    UJ
                    cc o
                    =5 Z
                    CO O
                    co o
                                                                         U co
                           oo
                               UJ
                            CD CO

                            O <
       UJ
       X —I  -^
       o < co
          o a
       LU ~ cr
       o CD <
       »- O Q
       > -J Z
       o o <
       QZ •-*« h-
       CL CQ CO
                                                  a.  a o
                       o
                       -j
                       UJ
                       LU CJ 3
                       Q ~ -J
                          X <  -^
                       >- o > >-
                       < I-UJ I-
                          H- Z O
                          z o ~
                          LU « x
                          ID I— O
                          -J O h-
CM U. O UJ
C=> LU LU _J
      ^5^ CO
    z    •<
    O >• I-
X    I- Q.
CD CO ^ UJ
3 I— CJ O
O ^ *- CJ
£K Z X <
X ^ O Z
                    CO -J
                    01 ^-*
                    O OQ

                    < a.
                    or <
                    I- O
                    CO
                    M UJ
       UJ
       £  •
       CO  LU
       co  o
       UJ  Z
       CO  <
       CO  I-
       <  CO

       -J  CO
           CO
                    O CO
       CD  -J
       O  <

       O  ~
       ^^—  ^w
       ^^*  •••
       CQ  X
           CJ
       Q  Uj


           UJ
 Z O — O
 o ~ z: a:
 M CD UJ Qu
 CD UJ X
       CJ O

-------
building this into our work plans and budgets, and
we've seen considerable success and progress  in the
Regions as well as the States in terms of developing
toxicity testing capability.
SLIDE 3
     Other issues addressed in the policy; one,
the importance of whole effluent toxicity is  stressed
as an evaluation and control parameter/ particular-
ly for complex effluent permitting situations.  In
other words, toxicity can be used as a parameter
just as BOD is used in NPDES permits.  Second,
effluent toxicity can provide a valid indication
of receiving water impact.
     The policy then goes on to respond to some of
the issues raised in the NPDES permit regulations
regarding the toxicity-based limits, in addition
to outlining the benefits and the disadvantages
of the chemical-specific versus the whole effluent
approach.  Obviously, when you're dealing with
complex mixtures of chemicals, the whole effluent
toxicity-based approach may be the way to go.  The
policy also stresses the  importance of using  toxi-
city testing when evaluating large publicly owned
sewage treatment plants.  Some of those effluents
are extremely toxic.

-------
             OS. -^
             UJ CO

             UJ O
                         UJ <
                         O O.
          UJ Q- I-

          -J -J CO
          U- O
          u. o: e>
          UJ O I—
          -J O I-
          o    «
          I Q Z
          ^ z cr
             < UJ
          U.    O.
          O Z
             O X
          UJ ~ UJ
          O I- -J
          Z < Q.
                z cr
                < I-
                O CO

                >• CO
                I— Z
                X  UJ
                O  l>
                         UJ O
                         U. O
                         u. *••
                             >- h-
                             H- O
             3  O
             O  U.

             at:  co
             UJ  ~
             I-  CO
             <  < .
             ^  CO CO
                    I-
             *   UJ -^
             z  H- x
             p  <~
             oc  ~ _j
             u.  o:
                 a. >-
             UJ  O H-
             UJ  QC «
             cr  a. o
             * -

             UJ
                                   X
                                   O
CO
CO
UJ
CO
UJ


CO
CO
         UJ     >-
         I  CO J
 CO >- J
 UI I- D
 CO ~- O
 CO O ~
 UJ — I—
 «r x oc
 h- O <
GO I— Q.
                                CO
                             CO
                             UJ
                         co
                         UJ Q
CO
                                      UJ
                                      Q
CO O
UJ

-------
     The policy also asserts that toxicity testing
methods are currently available.  EPA is confident
that there are adequate test methods which are
reliable and reproducible.  With these test
methods available, we don't see any reason why'
this should prevent regulatory agencies from not
using this approach.
     In addition to the policy, we also have put
out a technical support document which essentially
supports the policy and provides technical guidance
to the permit writers and water quality specialists.
     Thus  it is clear that in addition to the more
traditional chemical-specific approach to evaluat-
ing water  quality, whole effluent toxicity tests
and other  biological measures must now be systema-
tically employed throughout the various steps of
the regulatory process such as screening, impact
assessment, setting limits and measuring compliance.
These tests and measures must be related to instream
factors such as zones of initial dilution, flow
variability and sediment contribution so that the
instream behavior of toxics can be fully understood
as part of the decision-making process.  Finally,
the selection of tests and their application must
be tailored to the specific site in question.
                        16

-------
     I would like to conclude with this observation.



The history of the Nation's chemical and biological



monitoring activities has experienced several sig-



nificant shifts in emphasis.  As you recall, in



the early 60's the emphasis was on biological



monitoring.  We were looking at water quality



standards, water quality wasteload allocation



models, dealing with more of the traditional pol-



lutants such as BOD and dissolved oxygen.  Then



in the '70's, with the event of the Clean Water



Act, the shift was from out of the stream to the



end of the pipe and we were following a technology-



based approach which is essentially where we are



today.



       Now we are beginning to realize the real



advantages of biomonitoring in dealing with complex



effluents.  However, chemical monitoring continues



to have many advantages over biomonitoring, such as



controlling human health hazards and in providing



direct measures of operating treating process and



evaluating compliance.  Therefore, we must continue



to view both chemical and biological monitoring not



as separate entities but rather as integrated



components of a balanced and effective water quality-



based toxicity control program.  Thank you.
                        17

-------
                          MR. TELLIARD:  We're  going



to do questions now.  We're not going to let you off,



Any questions?
                        18

-------
            QUESTION AND ANSWER SESSION








                          MR. DELLINGER:  Bob



Bellinger, Industrial Technology Division.  It



seems to me that probably the most critical thing



in biomonitoring is selection of a test species or



organism.  I'm familiar with pulp and paper effluents



and there's a wide range of effect depending on the



test species chosen.  Some are more sensitive than



others.  What would be the policy in that regard?



Do we pick a very sensitive species or a relatively



insensitive species, one in the middle, or how...



                          MR. BUGBEE: There really



is no such thing as the most sensitive species.



What we try to do is use several species to bracket



the range of sensitivity which can occur.



     The puip and paper example is a good one



because right now they're running acute tests



using rainbow trout and people would think that



rainbow trout would be the most sensitive species.



Well, it turns out that these rainbow trout really



do quite well in the kraft mill effluent.



     The States of Oregon and Washington are now



looking for more tests to use in order to regulate



this toxicity.  There is a chronic toxicity
                        19

-------
associated with the kraft mill effluent, but the
rainbow trout in the tests that they're using,
does not pick this up.  So if you would happen to
user let's say, an oyster-spat or a Ceriodaphnia
test, these tests would show toxicity.
     This is one of the reasons we're promoting
the use of more than one species, preferably three
to five species, and perhaps using both the acute
and the chronic test combined.
                          MR. DELLINGBR:  Thanks.
                          MR. BUGBEE:  Yes, sir.
Jim?
                          MR. RICE:  Jim Rice.  I
wanted to ask, Steve, you made some mention about
part 136, I think.  My recollection is that none of
the tox tests methods that you're using are in part
136 now.  I was wondering what your plans were in
terms of publishing for notice and comment and the
like, for these tox methods that you have under
development now?
                          MR. BUGBEE:  I'm glad you
asked that question, Jim.  We started in 1978, trying
to get the acute toxicity tests for fathead minnows
and daphnia into part 136, and the latest information
I have, it will be March of 1986 when they propose
                        20

-------
to get these acute methods into 136.
     Corny Weber of ORD, Cincinnati, has put out
the third edition of his Biological Methods Manual,
which is a very fine manual for acute tests.  He's
going to also put out a separate manual just for
some of the chronic tests.  But that's what we're
using right now, and also, I believe, in Standard
Methods, there are some routine acute tests that
can be used.
                          MR. TELLIARD:  Thank you,
Steve.
                          MR. BUGBEE:  Your welcome.
                        21

-------
                          MR. TELLIARDs  Our next
speaker, Dr. Duke, is going to talk a little bit
about critters that I like better, which are shrimp.
     Due to the fact that Industrial Technology
Division always like to stay ahead or behind,
depending on which view you take, we are, this year,
going to for the first time, establish an effluent
limitation guideline based on an LC50 killing
critter number.  This is for the offshore oil and
gas industry.  For those industries feeling slighted,
we will do our best to get you an LC50 number as
soon as possible.
     Much of the work on our offshore oil and gas
program has been supported by our agency laboratory,
Gulf Breeze, and particularly by Dr. Duke.  So Tom,
this morning, is going to talk to you something
about the biological monitoring program that
they're conducting.  Tom.
                        22

-------
1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25
               THOMAS W. DUKE,  PH.D.

  UNITED  STATES ENVIRONMENTAL  PROTECTION AGENCY
        ENVIRONMENTAL RESEARCH LABORATORY
TOXICITY TESTS AND BEST AVAILABLE TECHNOLOGY  (BAT)
            DETERMINATIONS FOR DISCHARGE
       FROM OFFSHORE OIL AND  GAS  PLATFORMS
         (Revised presentation  submitted.)
                                  23

-------
               Toxicity Tests and Best Available Technology
                    Determinations for Discharges from
                 Offshore Oil and Gas Drilling Platforms
                        T.W. Duke and P.R. Parrish

     The Environmental Protection Agency is charged with regulating the
discharge of materials from offshore oil and gas drilling platforms.
This is accomplished through Best Available Technology (BAT)  and through
Section 403(c) of the Clean Water Act.  BAT is promulgated on a national
basis and Section 403(c) includes guidelines for testing indigenous
species and for more complex chronic and community tests.  The proposed
BAT guidelines for drilling fluid discharge from offshore platforms
contain a requirement for a toxicity test to estimate potential adverse
biological effect of a particular drilling fluid.
     This paper discusses the toxicity test specified in the  proposed
guidelines and how test results can be applied to regulatory  activities.

Drilling Fluid Used and Characteristics  '
     Before discussing the toxicity test, it is important to  describe
drilling fluids (muds) and how the fluids are used before discharge.
Drilling fluids are a complex mixture of chemicals and clays  that are
forced through the drill pipe, through the rotating bit, and  are returned
to the surface through the space between the drill string and casing.
During this process, drilling fluids lubricate the bit, coat  the bore
hole with an impermeable cake to prevent fluid loss, reduce corrosion,
transmit hydraulic power to the bit, and remove cuttings (Ayers, 1981).
     Four basic components — barite, bentonite, lignite, and lignosulfonate
comprise about 90 percent of the materials used in drilling fluids (Table 1),

-------
Barite, the most common weighting agent, is used to increase the
density of the fluid.  Bentonite clay is used to thicken the drilling
fluid; lignite and lignosulfonates are used to ensure that the fluid
does not become too viscous.  Speciality additives, including diesel or
other oils to increase lubricity and biocides to control microorganisms,
also may be added as the drilling process requires (Perricone, 1980).

Discharge of Fluids
     After a drilling fluid is pumped from the well to the drilling
platform, it is passed through solids-control equipment and the separated
solids (cuttings) are discharged into the sea.  This discharge occurs
continually, as long as drilling is in progress.  Fluid passing through
the solids-control equipment is  returned to a holding tank for possible
treatment and recirculation into the well.  When the fluid becomes too viscous
for use or if it  no  longer  functions as desired, a portion is intermittently
discharged into the  sea.  All of the used drilling fluid is discharged at
the completion of drilling  of exploratory wells.  The volume of solids
continuously discharged varies from 3,000 to  6,000 barrels per well  (one
barrel equals 42  U.S. gallons),  and the volume  of used  drilling fluid
intermittently discharged varies from 5,000 to  30,000 barrels per well
 (National Research Council, 1983).
      The  fate of  drilling fluids discharged  into the sea is determined by
physical, chemical,  and biological processes.   A portion of discharged
drilling  fluids contains  larger  particles  that  sink to  the  bottom relatively
 near  the  well site,  depending upon their density and upon current velocitites
 and other environmental factors.  A  lesser amount  of solids,  as well  as
 soluble  components,  remains in  the water column and can be  transported
 away  from the well  by  ambient currents  (Ayers et  a!.,  1980a;  Ayers  et
                                              25

-------
al., 1980b; Brandsma et al., 1980).  Thus, there is potential  interaction
between discharged drilling fluids and biota in the water column and
on the bottom.

Toxicity Test
     The proposed guidelines for BAT include a toxicity test,  the purpose
of which is to obtain a general idea of the toxicity of a specific drilling
fluid or fluid component to marine organisms.  This static, acute toxicity
test may lead to more complex tests, but it is not intended to provide
detailed toxiciological data as may be required under 403(c).   Rather, the
test is intended to provide a basis for comparing the toxicity of drilling
fluids.
     In order to be an effective test for BAT purposes, methodology should
be simple, the test organism should be relatively sensitive and readily
available and transportable, and results should be amenable to statistical
analysis.  For these and other reasons, a 96-hour test with mysids,
Mysidopsis bahia, was chosen as the BAT toxicity test.  Mysids have been
used as a reliable and sensitive test organism for test materials other
than drilling fluids and a substantial data base exists.  In addition,
the mysid test has been used to evaluate the toxicity of many  drilling
fluids; therefore, new results can be compared with the existing drilling
fluid data base.  Although the test described here is only four days in
duration, mysids can be used for long-term tests to study the  effects of
materials on partial or full life-cycle stages (Nimmo et al.,  1977).
     Mysids can be cultured by those performing the test or purchased
from a supplier.  All mysids used in the drilling fluid test should be b+1
days old at the beginning of the test and fed Artemia salina nauplii
during culture and testing.
                                             26

-------
     Details of the test method are documented by Petrazzuolo (1983)  and
Duke et al. (1984).  Natural or artifical  seawater can be used to prepare
the suspended participate phase (SPP)  of the fluids.  The SPP is prepared
by adding one part fluid to nine parts seawater (volume/volume) and stirring.
The mixture is allowed to settle for one hour and the material that
remains in suspension, the SPP, is the test material.  It is added to
seawater  (volume/volume) to prepare the test solutions.  Dissolved oxygen
(DO) and  pH of- the SPP are controlled during preparation.  For range
finding tests, 10 mysids are added to each of four concentrations —
100%, 50%, 10% and 1% SPP and a seawater control, none of which is
replicated.  For definitive tests, 20 mysids are added to a seawater
control and each of five concentrations that are based on the results of
the range-finding tests.  Three replications give a total of 60 animals
per treatment.  All treatments are aerated during the 96-hour exposure.
Water quality  (DO, pH, and and salinity) are measured at 24-hour intervals;
temperature is measured continuously.
     After 96  hours, the number of live organisms is determined in each
drilling  fluid concentration and in the control.  Mortality data from the
drilling  fluid test and a reference toxicant test that must be conducted   •
at the  same time are subjected to statistical analyses.  A 96-hour LC50
(the concentration lethal to 50% of the test animals  after 96  hours  of
exposure) and  its  95% confidence limits are calculated for each drilling
fluids  (if the mortality data  are amendable) by  using  probit  analysis
(Finney,  1971) or  other suitable methods  (Stephan,  1977).

Application  of Toxicity tests  Results
     The  purpose  of the toxicity test  is  to obtain  general  information on
the  comparative toxicity of  a  specific drilling  fluid  or component.  The
                                              27

-------
test is considered a first-tier test because other tests may be required
to determine the sensitivity of other organisms or communities if the
toxicity of the tested material warrants.
     An example of how the tiered system works and how the results of
toxicity tests can be used to protect the environment through product
substitution resulted from a recent evaluation of drilling fluids by the
Environmental Resarch Laboratory at Gulf Breeze (ERL/GB).  A cooperative
program (Duke and Parrish, 1984) was conducted to determine the effects
of 11 used drilling fluids on selected marine organisms.  The fluids were
collected from operating platforms that were drilling wells of various
depths at different locations in the Gulf of Mexico.  Drilling fluids
were tested at ERL/GB to determine their effect on mysids and were analyzed
for specific metals and hydrocarbon content by other laboratories.  The
tests with mysids indicated that several of the 11 drilling fluids were
more toxic than fluids tested in the past (Petrazzuolo, 1981 and Ayers et
al., 1983).  Toxicity of whole muds to mysids ranged from 26 to >1,500 pl/1
(ppm) and toxicity of the SPP phase from 726 to >50,000 ppm (Gaetz et
al., in press).
     Subsamples of the fluids were provided to other laboratories to
determine toxic effects on several marine organisms.  The toxicity pattern
of whole muds to grass shrimp (Palaemonetes pugio) was similar to that for
mysids.  Grass shrimp were not as sensitive as mysids, but the relative
sensitivity of grass shrimp and mysids to the 11 fluids was similar.  The
96-hr LCSOs for grass shrimp ranged from 142 to >100,000 ppm  (Conklin and
Rao, 1984).
     The results of the mysid and grass shrimp tests suggested the need
to test other organisms to confirm the effect of the petroleum hydrocarbons
                                             28

-------
in the used fluids.  Fertilized eggs of hard clams (Mercenaria mercenan'a)
were exposed to the SPP phase of the fluids for 96 hr.  The number of
clams exposed to the drilling fluids that reached straight-hinge or "D"
stage larvae was compared with the number of clams that attained this stage
that were maintained in sea water without the fluids.  The 96-hr EC50s
(the effective concentration that prevented 50 percent of the clams from
reaching the straight-hinge stage) were from 64 to >3,000 ppm (New England
Aquarium, 1984).
     A comparison of the diesel content of the 11 used drilling fluids
and their toxicity to mysids, grass shrimp, and clam embryos indicated
that the higher the diesel content, the greater the toxicity.  The
correlation between toxicity and diesel content was significant, according
to the Spearman rank order method (Simpson et al., 1960).  Subsequent
results obtained at the University of West Florida by testing the toxicity
of drilling fluids to larval grass shrimp, P. intermedius. before and
after addition of diesel oil and mineral oil also confirmed that addition
of petroleum hydrocarbons increased toxicity of the muds tested (Gonklin
and Rao, 1984).
     Thus, the results of toxicity tests that began with a modification
of the first-tier mysid test described in this paper  revealed-a component
in the used drilling fluids that contributed greatly  to the toxicity of
the fluids to marine organisms.  Under use conditions, results such as
these would justify excluding the toxic material  from a drilling fluid
and adding a less-toxic substitute that would also perform the necessary
function.
                                              29

-------
TABLE 1.  Some chemical ingredients in drilling fluids1
Ingredient
           Use
Barite
Bentonite
Attapulgite

Sodium Tetraphosphate
Modified Tannin
Chromium Lignosulfonate
Calcium Lignosulfonate
Lignite

Starch
Cellulose

Detergents
Non-ionic Emulsifier
Processed Hydrocarbons,
  including Diesel Oil

Aluminum Stearate
Paraformaldehyde

Sodium Chromate
Sodium Hydroxide
Potassium Hydroxide
Weighting Agents and Viscosifiers
Dispersants and Thinners
Fluid Loss Reducers
Lubricants and Emulsifiers
Defoamers, Bactericides
Corrosion Inhibitor
pH Control
pH Stability
1 After Perricone (1980)
                                             30

-------
                                REFERENCES
Ayers, R.C., Jr., T.C. Sauer, Jr., D.O. Steubner, and R. P. Meek.  1980a.
     An environmental- study to assess the effect of drilling fluids on
     water quality parameters during high rate, high volume discharges to
     the ocean.  In:  Proceedings of Symposium on Research on Environmental
     Fate and Effects of Drilling Fluids and Cuttings, Vol. I, pp. 351-
     391.  American Petroleum Institute, Washington, D.C.

_ __.  T.C. Sauer, Jr., R.P. Meek, and G. Bowers.  1980b.
     An environmental study to assess the impact of drilling discharges
     in the mid-Atlantic.  I.  Quantity and fate of discharges.  In:
     Proceedings of a Symposium on Research on Environmental Fate and
     Effects of Drilling Fluids and Cuttings.  Vol. I, pp. 382-418,
     American Petroleum Institute, Washington, D.C.

_ .  1981.  Fate and effects of drilling discharges in the
     marine environment.  Proposed North Atlantic DCS Oil and Gas Lease
     Sale 52.  Statement delivered at public hearing, Boston, MA, November
     19, 1981.  Bureau of Land Management, U.S. Department of the Interior.
           .   .C. Sauer,   ., an
mud concept for offshore drilling for
    ling Conference Proceedings,
Drilling
11399.
                   pp
 NPDES.
327-330.
            1983.
T.C. Sauer, JR., and P. Anderson.      .
                                In:  IADC/SPE 1983
                                                             The generic
                                                             PE 1983
                                                    Paper No. IADC/SPE
Brandsma, M.G., L.R. Davis, R.C. Ayers, Jr., and T.C. Sauer, Jr.  1980.
     A computer model to predict the short-term fate of drilling discharges
     in the marine environment.  In:  Proceedings of a Symposium on
     Research on Environmental Fate and Effects of Drilling Fluids and
     Cuttings, Vol.  II., pp. 588-610, American Petroleum Institute,
     Washington, D.C.

Conklin, P.J. and K. R. Rao.   1984.  Comparative toxicity of offshore and
     oil -added drilling muds to larvae of Palaemonetes intermedius.
     Archives of Environmental Contamination and Toxicology 13:685-690.

Duke, T.W. and P.R.  Parrish.   1984.  Results of the drilling fluids
     research program sponsored by the Gulf Breeze Environmental Research
     Laboratory, 1976-1984, and their application to hazard assessment.
     EPA-600/4-84-055, Environmental Research Laboratory, Gulf Breeze,
     FL.  94 pp plus appendices.

_ , P.R. Parrish,  R.M. Montgomery, S.D. Macauley, J.M. Macauley, and
     G.M. Cripe.  1984.  Acute toxicity of eight generic drilling fluids
     to mysids  (Mysidopsis bahia).  EPA-600/ 3-84-067, Environmental Research
     Laboratory, Gulf Breeze,  FL.  11 pp.

Finney, D.J. 1971.  Probit Analysis, 3rd Ed. Cambridge University Press,
     London.  333 pp.
                                              31

-------
Gaetz, C.T., R.M. Montgomery, and T.W. Duke.  In Press.  Toxicity of
     component phases of used drilling fluids to mysids (Mysidopsis
     bahia).  Environmental Toxicology and Chemistry.

National Research council (U.S.).  1983.  Drilling discharges in the
     marine environment.  Panel on Assessment of Fates and Effects of
     Drilling Fluids and Cuttings in the Marine Environment.  National
     Academy Press, Washington, D.C.  192 pp.

New England Aquarium (Edgerton Research Laboratory).  1984.  A survey of
     the toxicity and chemical composition of used drilling fluids.   EPA-
     600/3-84-071, Environmental Research Laboratory, Gulf Breeze, FL. 109 pp.

Nirreno, D.R., L.H. Banner, R.A. Rigby, J.M. Sheppard and A.J. Wilson, Jr.
     1977.  Mysidopsis bahia:  an estuarine species suitable for life-cycle
     toxicity tests to determine the effects of a pollutant.  In:  Aquatic
     Toxicology and Hazard Evaluation, F.L. Mayer and J.L. Hamelink, Eds.,
     pp. 109-116.  ASTM STP 634, American Society for Testing and Materials,
     Philadelphia, PA.

Perricone, C.  1980.  Major drilling fluid additives.  In:  Proceedings
     of a Symposium on Research on Environmental Fate and Effects of
     Drilling Fluids and Cuttings, Vol. I., pp. 15-29.  American Petroleum
      Institute, Washington, D.C.

Petrazzuolo, G. 1981.  Preliminary report on environmental assessment of
     drilling fluids and cuttings released onto the outer continental
      shelf.  Vol. 1:  Technical assessment.  Vol 2:  Tables, figures and
     Appendix A.  Prepared for  Industrial Permits Branch, Office of Water
     Enforcement  and Ocean Programs Branch, Office of Water and Waste
     Managenent,  U.S. Environmental Protection Agency, Washington, D.C.

_ .   1983.   Proposed methodology:  Drilling fluids toxicity
      test for offshore subcategory; oil and gas extraction industry.
      May 19, 1983.  45pp.
Simpson, G.G., A. Roe, and R.C. Lewontin.  1960
     revised edition.  Harcourt, Brace & World,
                                                  Quantitative Zoology,
                                                 Inc., New York.  440 pp
 Stephan,  C.E.   1977.   Methods  for  Calculating  an LC50.   In:  Aquatic
      Toxicity  and  Hazard  Evaluation,  F.L. Mayer and J.L.   Hamelink, Eds.,
      pp.  65-84.  ASTM STP 634,  American Society for Testing  and Materials,
      Philadelphia, PA.
                                              32

-------
                INTRALABORATORY VARIATION
SAMPLE
   1
   2
   3
96-HOUR LC5Q
   2,9 SPP
   2.5%
   1,3%
   2,1%
95% CONFIDENCE LIMITS
    2,4-3,5% SPP
    2,3-2,3%
    0,5-2,2%
    1,7-2,4%
    2,1-2,;
                                33

-------
   SOURCES OF VARIATION

1,  MYSID CONDITION
2,  TREATMENT OF SPP
    A,  DEFINITION
    B,  SEPARATION FROM SOLID PHASE
    C,  AERATION TO INCREASE DO

-------
         BASICS OF METHOD
   *• 1:9 DILUTION OF DRILLING FLUID
  **  PH ADJUSTMENT
 ***
****
SUSPENDED PARTICULATE PHASE
POSITIVE CONTROL INCLUDED
                     35

-------
                  SENSITIVITY  OF MYSIDS
                       TQXICITYA  (96-HOUR  I C
GENERIC
MUD

n

MUD
ALONE

51,6
29,3
MINERAL
1
13,5
7,1
OIL ADDED «.
5
1,8
0,90

10
0,49
0,76
ACONCENTRATIONS GIVEN AS PERCENT  (v/v) OF SUSPENDED
 PARTICULATE PHASE IN S.EAWATER,
                                36

-------
                    INTERLAB RESULTS
TEST
LAB
GULF BREEZE

NARRAGANSETT
GENERIC
MUD •
  #5
  #1
  #5
TOXICITY
(96-HOUR LC5Q)
  2,7% SPP
No MEDIAN EFFECT
  2,3% SPP
No MEDIAN EFFECT
95% CONFIDENCE
LIMITS
 2,5-2,9% SPP
 2,5-3,0% SPP
                                 37

-------
                TERESA NORBERG-KING

   UNITED STATES ENVIRONMENTAL PROTECTION AGENCY
         ENVIRONMENTAL RESEARCH  LABORATORY
      BIOLOGICAL ANALYSES OF COMPLEX EFFLUENTS



                          MRS. NORBERG-KING:

Actually, Norb told me to pass a message on to

Bill.  He told me that you owe him one.  I got

to come here today and explain some ideas of bio-

logical methods to chemists.

     When Dale Rushneck asked me to present the

paper on fresh water toxicity tests in relation to

the testing approach that's being used for the

permitting process that's currently underway, I

said a tentative yes and it depends on who can pay

my travel.  Then he explained it was for an ana-

lytical symposium and I was pretty sure he wouldn't

get me to go.  Finally he said I didn't have to

discuss policy because Steve Bugbee was going to,

and in that case, I said sure.  I don't like to

talk about policy.  Biologists never like to.

     First, what I want to discuss is how EPA got

involved in the toxicity testing developed for

complex effluents.  Then quickly'describe the test
                        38

-------
methods that we're using, and then to take some of
our test results and compare them to the biological
survey results.
     First I just want to give you some quick back-
ground that explains and again reiterates some of
what Steve Bugbee was saying.  Under the Public Law
92-500, the procedures for regulating pollutants
were laid out and a national pollutant discharge
system, the NPDES Permit System was introduced.
Permits were originally written to incorporate
standards for toxicants.  Yet these were not success-
full and EPA chose to establish effluent limitations
for the "priority pollutants."  The limitations
were to incorporate the Best Available Technology
(BAT) that was economically feasible and to do so
on a chemical-by-chemical basis.
     Toxicity information on many compounds is
limited, if nonexistent, and the interactions of
mixtures are poorly understood.  Consequently, a
move to effluent toxicity testing in order to assess
the overall toxicity of effluents has been under
way.                               ,
     The priority pollutant approach has not been
useful  in effluent testing because they are just
not a concern to aquatic organisms.  With the
                        39

-------
exception of a few metals and pesticides, barely
any priority pollutants are in effluents above the
detection level.  In fact, a recent article by
Staples et al (1985) addressed the issue of priority
pollutants in effluents.  The authors evaluated
EPA's STORET computerized data base of water quality
information that's collected by the states and the
regions.  Their review pointed out that 80 percent
of organic priority pollutants are not detectable
in industrial effluents and that 78 percent of the
priority pollutants are not found in the average
waterway.
     Effluent testing for aquatic organisms has
been going on for years, but it's only recently
with the water quality-based permit policy (EPA,
1985) that the renewed interest has been displayed.
With this new-found interest, the emphasis for sub-
lethal or subchronic tests to supplement the short-
term acute toxicity has increased.
     This is where the research efforts at the
Environmental Research Laboratory at Duluth (ERL-
Duluth) Minnesota, have been expended for the last
three to four years.  The goals of the complex
effluent program are to validate the ability of the
laboratory toxicity tests to predict the community

-------
impact of the industrial and municipal effluents.
To do so, we developed two short-term, subchronic
tests which are both run for seven days.  We use a
cladoceran, Ceriodaphnia dubia, and a small fish,
the fathead minnow, Pimephales promelas.  We run a
set of various effluent dilutions and we run samples
of the receiving water.  Both tests that we run are
static with daily renewal of the test solutions.
Most often, these tests are conducted on-site.
     The Ceriodaphnia test end-points are reproduc-
tion and survival to estimate the effect of the
level of the effluents.  The fathead minnow test
end-points are growth and survival to estimate the
effluent effect levels.  We have generally found
the mean number of young produced per female for
the Ceriodaphnia and that the mean weights for the
fathead minnows are more sensitive end-points than
survival in each seven day test.
     Briefly, the Ceriodaphnia test uses newborn
young  (0-4 hours old) to start the test.  The young
are placed in individual test cups containing
15 mis. test solution each.  As an aside, most
people are familiar with Daphnia magna.  So, the
first thing people always ask me is how small is
the Ceriodaphnia compared to the Daphnia.  I can

-------
see it with my naked eye; some people like to use a



microscope/ and we don't have to use electromicro-



scopy to read these results.  Ceriodaphnia produce



three broods in seven days which is an end-point



for toxicity tests following the OECD guidelines.
                  •


They produce an average young of 18 to 30, depending



on what source of water is used in the test.



     The fathead minnow test is a daily renewal



static test and is initiated using less than 24



hour old post-hatch larvae.  In seven days we have



found they typically increase their weight five to



seven times that of their initial weight.  This is



a quick overview of what the test methods are.



If you want more information on the methods, see



Mount and Norberg, 1984; and Norberg and Mount,



1985.



     In addition to the effluent dilution tests, a



set of tests that we have termed the ambient toxicity



tests, are run (Mount et al, 1984).  These are run



with water that is collected directly from the



stream.  Typically the locations of these stations



are established above and below each discharger and



the water collected for the ambient tests allows us



to evaluate what's actually happening when the



effluent and the receiving water mix.  The results

-------
of the Ceriodaphnia and fathead minnow ambient
tests are then used for comparisons with the stream
community data.
     For these ambient tests, the concentrations
of effluents in the stream are not important, but
rather whether the water in the stream has an
observable effect in the toxicity tests and on the
stream community.  So what you are actually getting
with these ambient tests is a measure of toxicity
of various stream stations which we can then use
to compare to the condition of the biological
community at the same stations.
     Steve Bugbee mentioned the work that has been
going on for these sites.  It can be intensive
laboratory and field work.  We have completed eight
site studies around the country covering Connecticut
(Mount et al, in press), Alabama (Mount et.al, 1985),
Oklahoma (Norberg-King and Mount, in press), Mary-
land (Mount et al, in press), West Virginia (Mount
and Norberg-King, in press; and Mount et al, in press),
and Ohio (Mount et al, 1984; Mount and Norberg-King,
1985).
     Along with the ambient tests, the field surveys
of the biological and hydrological conditions are
done.  These consist of fish collection, benthic

-------
sampling, periphyton and zooplankton sampling or
artifical substrates where appropriate.
     What I want to discuss now, in the remainder
of this talk, is how these biological analyses of
the effluents and the ambient•toxicity tests compare.
This discussion will present the analysis of effluents
by means of the ambient toxicity tests compared to
the stream community for just three out of eight
field sites.
     The first site I want to discuss was conducted
on a very small prairie stream in Oklahoma called
Skeleton Creek.  It has a very low gradient and the
stretch of river studied covered approximately 27
kilometers.  There are three dischargers that are
located on this small creek directly.  They are a
Publicly Owned Treatment Works (POTW), a refinery
and a fertilizer plant.
SLIDE 1
     What you see on this slide, along the bottom,
are the stream stations that we plotted, and along
the left hand side is the mean fathead minnow
weights.  The data is the mean individual weight at
the end of the seven day test, which is a dry weight.
Please note these stations were not equal distances
apart on the stream as they are depicted here.

-------
     Again, just  to quickly  explain what  the  ambient
tests are, they are the daily renewal  tests with  a
grab or  a composite sample from each river station,
that are tested to determine if chronic toxicity
occurs instream.  Note with  this site, we saw
^substantial toxicity observed at Station  5.
SLIDE 2
     Next are the results of the Ceriodaphnia and
the fathead minnow ambient tests plotted  together.
These show that there were similar responses  of
each species up to Station 5 and yet there was a
dramatic drop in  the Ceriodaphnia young production
evident  at Stations 7 and 8.  Irrigation  flow
entered  the creek at various points, but  the  first
known entering above Station 7.  In fact, the water
from Stations 7 and 8 had chronic activity levels
that were three times those of upstream stations.
We have  data that show the Ceriodaphnia to be much
more sensitive to salinity than the fathead minnows,
which may account for this.  You may also note that
the slight inhibition of young production and
weight values at  the most upstream station, which
was Station 2, was removed after the POTW and the
refinery effluents were mixed in the stream.

-------
SLIDE 3
     Next is the fathead minnow weight data and the
dotted line is the number of benthic taxa.  The
results of the benthic taxa show us they were
unaffected over the entire study area.  If toxicity
observed in the ambient tests was due to ammonia
from the fertilizer plant, or from one discharger—
I'm trying not to point fingers here—the benthic
population may not have been affected due to their
lack of sensitivity to the ammonia.
SLIDE 4
     Next, is the dash line, is the number of the
fish taxa plotted against the mean fathead minnow
weights, and these data match extremely well*  You
can see that the number of fish taxa show a response
at Stations 7 and 8 that were not displayed in the
fathead minnow ambient test.
SLIDE 5
     But when we plot the Ceriodaphnia data together
with the fish taxa as the dash line, you can readily
see more similar responses.  For this site, an
invertebrate test organism is much more predictive
of the stream fish population as the Ceriodaphnia
test results correlate extremely well with the
number of fish taxa collected in the biological

-------
survey.  This all is aimed at trying to answer the
question of which species to use and whether it's a
sensitive species, or how do we choose which species
we're testing for effluents.
     The second site I want to discuss is Five Mile
Creek in Alabama.  It is a foothill, spring-fed
creek with a high gradient in the upper end.  The
study area covered a 42 kilometer stretch of river.
In this area there are two coke plant dischargers
in the upper stretch and a POTW discharge farther
downstream, and a few tributaries entered the
stream.
SLIDE  6
     What you  see here are  the mean weights  of
the  fathead minnows plotted against the stream
stations for Five Mile Creek.  Substantial toxicity
was  observed at Station  5.  Again, to reiterate,
what I'm showing you  are chronic  test results and
it's important to keep that in mind.  We feel we
are  able to make better  predictive value judgments
based  on chronic data than  we can on acute data.
SLIDE  7
     Here  is one  of my  favorite  slides.  Tom had  one
of  his favorite  slides and  I have one of mine.   I'd
 like to just leave  this  one up  for  a while and  figure

-------
out something else to talk about so you can stare
at this one because this is one of our better
slides.  The values for the fathead minnow weights
and the number of the fish taxa at every station,
show the response is the same, with the sharp
decline of the taxa correlating extremely well with
the measurable toxicity of the fathead minnow
ambient toxicity test.  Even the slight upstream
effect that was observed at Station 1, the ambient
toxicity test, was also observed in the number of
fish taxa.  This is better agreement than we had
expected or we had hoped for in our comparisons.
SLIDE 8
     So, if we take the same fathead minnow ambient.
test data and we plot the number of benthic inver-
tebrates, there's good agreement.  At the upper
three stations the agreement was not as good, but
because of those stations at the upper end, what
actually happens is that we would under-predict the
impact at those stations.
SLIDE 9
     The Ceriodaphnia toxicity test results also
show some of that same inhibition at the upper end,
yet when the tributary flow entered at Station 3,
just after 2A, and after the coke plant's effluents

-------
entered, what we saw was increased young production.
Only after the POTW discharge entered the stream did
we see additional toxicity in Ceriodaphnia young
production.
     These dramatic result differences between the
fathead minnow weights and the Ceriodaphnia young
production indicates the value of running toxicity
tests with more than one test species.  We may have
missed something with just running one species, but
we gained a lot with running two species.  Now, how
to make them mesh is another question.
SLIDE 10
     We took that same Ceriodaphnia data and we
plotted the zooplankton density in the stream with
the mean number of young per female for the Cerio-
daphnia at each ambient station.  We  shifted the
zooplankton response in the stream upstream to
match the upstream stations and allow for the drift
that would be occuring in the stream.  There is
really good agreement and the same decrease in the
upper reaches is  there with the zooplankton in the
stream that we saw with the Ceriodaphnia ambient
test.
     So with this site, the Five Mile Creek site,
we can sum the results of the toxicity testing and

-------
the  field  survey as  follows.   The  fathead minnow
tests and  the fish test results correlate extremely
well.  My  favorite slide pointed that out.   The
Ceriodaphnia and the zooplankton tests correlate
extremely  well when  one allows for the drift.  The
benthic taxa does not agree well with the fathead
minnow tests at a few stations and the reasons for
this are too numerous to postulate  here.  However/
completely different responses of  the two laboratory
toxicity tests again points to the  value of  testing
more than  one species.
     The last site I am going  to discuss is  the
Naugatuck  River in Connecticut.  It is a very high
gradient stream with large rubble  boulders in the
upper reaches.  This was a complex  site and  it
covered 64 kilometers of river.  There were  also
numerous tributary inputs that occurred.  A  total
of 27 discharge outfalls are located on the  stream
and on the tributaries.  Most  of the industrial
discharges are metal plating operations with similar
plant designs and the effluents contain mostly the
same heavy metals.
SLIDE 11
     Again, please note on this slide, I've plotted
where the  tributary input enters the stream across
                        50

-------
the top as well as the POTW inputs.  Again, on this
slide as on the other one, the stream stations are
not equal distances apart as the slide appears to
indicate.  Most of the plating operations and the
other industrial discharges went directly into two
main tributaries.  Also, there was a reservoir
above Station 4.
     This same slide shows the results of the fat-
head minnow ambient test and this site, being so
much more complex, it's difficult to interpret.
The toxicity at Stations 10 and 11 was not chronic
but it was rather known to be due to a spill in the
river that occurred early during the testing.  When
such an event occurs, it results in an overestimate
of the toxicity at those two stations.  We are not
able to start the test again after a spill happens
because then the animals are older or a different
age group or something.
     The two single data points shown by  the arrows
are the results of fathead minnow ambient tests
with the source of that test water being  collected
directly in the mouth of the tributary with the
large amount of dischargers on it.  Essentially
what was measured was an estimate of the  combined
effluent effects  in each  tributary prior  to mixing
                        51

-------
with  the river.  We saw  toxicity  in the tributary
sample above Station 6.  We saw some decrease in
growth in ambient Station 6 water.  We also saw
toxicity in terms of reduced growth above Station
8 with the tributary sample, and we saw a sharp
decline in the ambient toxicity test weights of the
fathead minnows at Station 8.
SLIDE 12
     If you take the periphyton diversity and plot
it against the fathead minnow test results, there's
good agreement on these  two curves.  There is a
problem here in that periphyton diversity was used
rather than taxa, but this is done because periphyton
have not been identified as to species as well as
many other groups such as the fish or the benthic
raacroinvertebrates.
SLIDE 13
     The plot of the number of fish taxa with the
fathead minnow ambient test data does not show any
definite trend except possibly for lesser numbers
of fish taxa downstream.  We feel we could not
discern any clear area of impact.
SLIDE 14
     The benthic invertebrate taxa plotted against
the fathead minnow ambient test shows a general

-------
decline of taxa from the headwaters with a substantial
and consistent drop over Stations 3 through 7.  Except
it appears to recover at Station 9 as the number of
benthic taxa increase, and increased weights were
obtained at Station 9 in the ambient toxicity test.
SLIDE 15
     The Ceriodaphnia ambient toxicity data is
plotted in dotted line, and the dash line is the
zooplankton density in the stream.  When one corrects
for the drift of the zooplankton and moves the
response of the Ceriodaphnia upstream to account
for drift, again we see good agreement.
     So, what this all sums up  for the Naugatuck
River is that there is a good agreement from the
periphyton, the benthic macroinvertebrates, the
zooplankton to the laboratory ambient toxicity test
data, while the fish  taxa  data  did not show any
real clear trend of impact with the toxicity
tests  in  this  instance.
     From these studies and from this approach  to
complex effluent testing,  the following conclusions
can  be made.   One,  that in every study  the  toxicity
was  shown  by one or both test species which correlated
with an adverse  impact in  the community.  This  says
that no one test species represents one group of

-------
 the  community,  but  rather  some  group  in  the  commun-
 ity, albeit  fish, benthic  macroinvertebrates  or  the
 periphyton,  and that  the impacted  component  or com-
 ponents of the  community correlates with the  toxicity
 found  in  the lab.   Two laboratory  test results
 differ from  site to site.  Therefore, the test
 species cannot  be considered a  surrogate for  any
 one component of the  community.
     Finally, to say  that  this  has been a brief
 summary of the  lab  to field data.  We have other
 projects  we  are considering now such as negative
 interactions, species sensitivity, and the fre-
 quency and cause of reverse toxicity curves.  This
 data is only for three out of the eight sites.   We
 have reports on all of the sites in near final
 form.  It appears that toxicity tests will be used
 as permits limits soon.  We know of at least  two
 states that already have implemented toxicity
 testing into their  state policies.  Also, we may
need to start looking for the toxic components in
 an effluent  in  order  to advise  industries where
 their problem is.   However, when that happens, we
 feel as the biologists, that we really must in-
corporate toxicity  testing in that the analytical
methodology cannot  provide the whole picture.

-------
                        REFERENCES








Environmental Protection Agency. 1985.  Technical



Support Document for Water Quality - based Toxics Control,



Office of Water, Washington, B.C. 20460








Mount, D.I. and  T.J. Norberg, 1984.  A Seven day Life-



Cycle Cladocenan Toxicty Test.  Environment Toxicol &



Chemisty - Vol. 3 : 425-434







Mount, D., N. Thomas, M. Harbour, T. Norberg, T. Roush,



and R. Brandes. 1984.   Effluent and Ambient Toxicity



Testing and Instream Community Response on the Ottawa




River, Lima, Ohio.



EPA -  600/2-84-080. August







Norberg, T.J. and D.I.  Mount, 1985.  A new Fathead



Minnow (Pimephales promelas)  Subchronic  Toxicity Test.



Environment Toxicol & Chemistry - Vol. 4  : 711-718







Norberg, T.J.  and D.I.  Mount,  (editors).  (In  press).



Validity of  Effluent and Ambient  Toxicity Tests  for



Predicting Biological  Impact,  Skeleton Creek,  Enid,



Oklahoma.   U.S. Environmental  Protection  Agency,  EPA



 600/in preparation.
                            55

-------
Staples/ C.A., A. Frances Werner/ and Thomas  I, Hoogheerru
1985.  Assessment of Priority Pollutant Concentrations
in the United States Using STORET Database.   Environment/
Toxicol & Chemistry - Vol. 4 : 131-142

Mount/ D.I. and T.J. Norberg. (editors). (In  press).
Validity of Effluent and Ambient Toxicity Test for
Predicting Biological Impact, Kanawha River/  Charleston/
West Virginia.  U.S. Environmental Protection Agency/
EPA 600/in preparation.

Mount/ D.I. and T.J. Norberg. (editors). 1985.  Validity
of Effluent and Ambient Toxicity Tests for Predicting
Biological Impact, Scippo Creek, Circleville, Ohio.
U.S. Environmental Protection Agency, EPA/600/3-85/044/
June, 1985.

Mount, D.I., et al.  (editors).  (In press).   Validity of
Effluent and Ambient Toxicity Tests for Predicting
Biological Impact, Back River,  Baltimore Harbor, Maryland.
U.S. Environmental Protection Agency, EPA 600/in preparation,

Mount, D.I. et al. (editors). (In press).   Validity of
Effluent and Ambient Toxicity Tests for Predicting
Biological Impact, Five Mile Creek,  Birmingham,  Alabama.
                            56

-------
U.S. Environmental Protection Agency, EPA 600/8-85/015.

Mount, D.I., et al. (editors). (In press).  Validity of
Effluent and Ambient Toxicity Tests for Predicting
Biological Impact, Naugatuck River, Waterbury, Con-
necticut.  U.S. Environmental Protection Agency, EPA
600/in preparation.

Mount, D.I., et al. (editors). (In press).  Validity of
Effluent and Ambient Toxicity Tests for Predicting
Biological Impact, Ohio River, Wheeling, West Virginia.
U.S. Environmental Protection Agency, EPA 600/in pre-
paration.
                           57

-------
  MEAN FATHEAD MINNOW WEIGHT (mg)
m
s
3
(A
ss
0>
<2.  oo
p
KD
                       p
                       b>
                                   p
                                   bo
             T
                                 POTVV a
                                  Refinery

                                 Fertilizer
                                  Plant

-------
  MEAN FATHEAD MINNOW WEIGHT (mg)
3D
m
ro
C/3
3
c£

<§.   oo
P
ro
            T
                      P
                      o>
                                 P
                                 CD
                           POTVV a

                            Refinery

                           Fertilizer
                            Plant
             CERIODAPHNIA


        MEAN YOUNG/FEMALE
                        59

-------
  MEAN FATHEAD MINNOW WEIGHT (mg)
;0
m
C/5
IX)
*  52
<2.   oo
            p
            ro
                      p
                      0)
                    CD
                    m
                   ro
                   o
                      OJ
                      o
                            poTwa
                             Refinery

                            Fertilizer
                             Plant
       NUMBER OF BENTHIC TAXA
                        60

-------
CO
m
CO

5!
5
3T

CO
5
rf
         MEAN YOUNG/FEMALE
   ro -
   iff
         3
         X

         NUMBER OF FISH TAX A
                                  porwa
                                   Refinery
                                  Fertilizer
                                   Plant
                                 oo
                        61

-------
  MEAN FATHEAD MINNOW WEIGHT (mg)

            P      P      p      o
                          O)      00
;0
m
CO
    O)
2.
S-
            T
T
           T	
            s
            x
                 •

                X
     POTVV a
      Refinery

     Fertilizer
      Plant
            ro
      oo
         NUMBER OF FISH TAXA
                        62

-------
  MEAN FATHEAD MINNOW WEIGHT (mg)

           o     °      °      °
           •
           O
CO
H
ffl
V)
ro
ro
3
ss.
3  °°
0°
                                COKE PLANT #1
                                COKE PLANT .#2
                                POTW
                     63

-------
MEAN FATHEAD MINNOW WEIGHT (mg)
               pop

               8
          o
          •
          o
OJ
o
   ro

1  £


9
£  *
10  °>


3

OJ,

2  CX)
4^
O
~i
                             COKE PLANT
                              COKE PLANT #2
                              POTW
                            ro
                            O
      NUMBER OF FISH TAXA

-------
  MEAN FATHEAD MINNOW WEIGHT (mg)
rn
I
en
a"
p
5
                8
         NUMBER OF BENTHIC
         INVERTEBRATE TAXA
o
•
o
                    COKE PLANT #1
                               COKE PLANT #2
                               POTW
                    65

-------
           CERIODAPHNIA


        MEAN YOUNG/FEMALE


          58     8  '
m
>
   ro

   >
CO
ID


?
                        COKE PLANT*I
                        COKE PLANT #2
                        POTW
                     66

-------
    CERIODAPHNIA

MEAN YOUNG/FEWALE

         8
                CU
                a
                       COKE PLANT
                       COKE PLANT #2
ZOOPLANKTON SAMPLES

    NUMBER/LITER
              67

-------
MEAN FATHEAD MINNOW WEIGHT (mg)
     o     o     p      p     p
           8     fc
                                TRIBUTARY
                                POTW
                     68

-------
MEAN FATHEAD MINNOW WEIGHT (mg)
           o     p      P     P
           '8     8      &     8
3)
•
   ro
          o
          5
                  ro .

         PERIPHYTON
                                     TRIBUTARY
                    DIVERSITY
                         69

-------
MEAN FATHEAD MINNOW WEIGHT (mg)
                                 TRIBUTARY
                                 TRIBUTARY
                                 TRIBUTARY





                                 TRIBUTARY
                                 POTW
    NUMBER OF FISH SPECIES
                       70

-------
   MEAN FATHEAD MINNOW WEIGHT (mg)
       p      p     p     p     p
       5      b     5S  .  .fe     &
ro
                                TRIBUTARY
         NUMBER OF BENTHIC
         INVERTEBRATE TAXA
                        71

-------
  ro -
i
m
en
c/>
3
o
   CD
              CERIODAPHNIA
           MEAN YOUNG/FEMALE
           5.8      8   .__

                                      TRIBUTARY
•r
 N

•POTW


•DAM
•POTW

-TRIBUTARY

-TRIBUTARY

•TRIBUTARY


-POTW


-POTW
   ro
            ZOOPLANKTON DENSITY
                (NUMBER/m3)
                                      o
                                      en
                           72

-------
           .QUESTION AND ANSWER SESSION








                          MR. PRESCOTT:  I'm Bill



Prescott.  Could you amplify just a little bit what



this correction for drift business was?  Why and how?



                          MRS. NORBERG-KING:  Sure.



One of the feelings we have is that since the



zooplankton in the stream are drifting organisms,



we don't expect them to be stationary.  If you



collect them at Station 5, they're probably reflect-



ing the population response of an upstream station



as they move downstream with the water.  What we



are trying to do is predict how fast they're moving



with the water and then move the response of the



animals upstream to where they would have been in



terms of what exposure they would have received.



                          MR. PRESCOTT:  Thank you.



                          MR. HUNT:  I'm Gardner



Hunt with the main department of Environmental



Protection.  Have you had any experience with the



diversity index?



                          MRS. NORBERG-KING:  Yes,



we've done a number of things with the diversity



indices.  Right now we present indices in our



reports, but we don't feel that the biological
                        73

-------
indices are the best measure for us.  They're not



giving us the actual number of taxa present in the



stream.  We may find that in one species there



may not be a true representative of the species



at that station.



     I'm not the one to explain the diversity



indices.  We have a contractor that does all the



work for our biological field work, doing the



biological indices.  Don Mount, for instance, is a



much better person to decide what we're using about



the biological indices as he does the integration



chapter in our report.  I do a lot of the field



work, a lot of the toxicity testing, and am just



starting to try making some heads or tails out of



integrating the approaches.  I'm sorry I don't have.



a better answer.



                          MR. TUROSKI:  Vic Turoski,



James River.  What two states have instituted



biomonitoring with respect to permitting?



                          MRS. NORBERG-KING:  New



Jersey and Virginia.



                          MR. TELLIARD:  Any other



questions?  I have one, Teresa.



                          MRS. NORBERG-KINGs  Yes.



                          MR. TELLIARD:  If I was

-------
fortunate enough to have one of these permits with


biological monitoring requirements, what would it

cost me to run these...


                          MRS. NORBERG-KING;  For a


Ceriodaphnia test?

                          MR. TELLIARD:  Yes.

                          MRS. NORBERG-KING:  Given


that you have the laboratory...or say that you

would have to go after a contractor to run one of

these tests, our estimates for a Ceriodaphnia test

are anywhere from $500 to $1,000 to run one seven-

day Ceriodaphnia test.  That's extraordinarily low.

The fathead minnow test probably runs up to $1,000

to $1,500.  It's just a little more labor-intensive,

                          MR. TELLIARD:  Thank you.


I'd like to thank this morning's speakers—Steve,

Teresa and Tom.  This is our first time talking

about critters and I think it was very good.  Can

we have a hand for them?

     We're going to break now for coffee and please
                                    ?
get back in here in 15 minutes.  We've got a long

day ahead of us.  Thank you.

(WHEREUPON, a 15 minute break was taken.)

-------
                          MR. TELLIARD:  For those
people who struggled in late, there's a lot of room
in the front pews and they don't cost any more.
     Continuing with this morning's session, we
have two speakers that are going to be
addressing, .'.probably if you had a theme for a
meeting, we had critters and probably LC/MS.  LC/MS
is a new initiative for the industrial technology
division.  We, of course, feel that it's only fair
to stay ahead of API and CME by grabbing new
analytical tools when they're not looking and
running off and using them.
     We have, this year, started looking seriously
at LC/MS which gives us a lot more flexibility in
some of our areas of concern.  The folks at our
Athens Laboratory, the industrial technology
division's lab, has been working hard at some LC/MS
work.  John McGuire has been working on some analysis
of some of our pharmaceutical samples and so forth.
We see down the road this is going to be a valuable
tool both for investigating compounds that we know
we can no longer beat through a standard GC.
     So, with that as a brief introduction of why
we're concerned about it, in that it is a new tool
and it does cover new compounds, and we want to keep
                        76

-------
your life interesting, particularly the industries
who feel somehow we've slighted them by missing
certain compounds.
     We'd like to start this morning's presentation
and we're going to talk about LC/MS as a new and
powerful analytical tool.       .
                       77

-------
          RICHARD F.  BROWNER

   GEORGIA INSTITUTE OF TECHNOLOGY
         SCHOOL OF CHEMISTRY
    MAGIC-LC/MS:   A POWERFUL NEW
   TOOL FOR ENVIRONMENTAL ANALYSIS
(Dr.  Browner's paper will be available
    in open technical literature.)
                   78

-------
             ROBERT D. VOYKSNER, PH.D.
            RESEARCH TRIANGLE INSTITUTE
        APPLICATIONS OF THERMOSPRAY HPLC/MS
           FOR MONITORING THE ENVIRONMENT
                          DR. VOYKSNER:  Thank you,
Bill.  My name is Robert Voyksner and I'd like to
talk about thermospray LC/MS application to environ-
mental analysis.
     In today's presentation, I'd like to talk a
little bit about what thermospray is, what the
interface consists of, talk about optimization and
going to some applications and analysis of pesticides,
dyes and mycotoxins.
     First of all, I'll talk a little bit about the
type of samples we like to analyze by LC/MS.  The
technique is still not routine and if you can do it
by GCMS or GC, it's still best to do it that way.
Samples can be derivitized or if you don't want to
run by GC/MS, HPLC serves as a good method for
separation.  We use mass spec as a detector.  The
conventional LC detector doesn't offer the speci-
ficity or the sensitivity.
     Of the combined HPLC techniques available,
                        79

-------
thermospray is an ideal technique for a couple of
reasons.  First of all, the interface can accept
the full HPLC flow rate, so conditions that sepa-
ration chemists developed don't have to be changed
and it's very easy to transfer from a normal LC
detector to a mass spec detection.  Secondly,
there's no reverse phase solvent restrictions.  If
we're using thermospray ionization, we can use any
reverse phase solvent.  More recently, the thermo-
spray interface is being equipped with a filament
so we can do essentially water type CI and in that
case...or solvent CI, excuse me.  In that case, we
really have no salt restrictions at all.  Finally,
in either ionization case, both techniques, or
either ionization soft providing level of the
weight information.
SLIDE 1
     The schematic of interface is shown here.
The HPLC effluent enters this directly heated coil
(A) to form an aerosol, which we see in area F.
This aerosol contains the analyte as well as some
of the solvent.  The aerosol droplets shield it
from the heating to prevent any thermal decomposi-
tion.  To ionize the compounds I'm looking at,
primarily we get gas phase ionization with the
                        80

-------
volatile buffer,  in my case, ammonium acetate.   So
we're effectively getting ammonia CI type spectra.
For more ionic compounds where ions are actually
formed in solution, the integrity of the ion is
maintained in this aerosol and by means of the
repeller (L) and first lens, the ions are extracted
through this ion exit cone (E) into the quadrupole
rods.  The neutral gases or solvent basically goes
out into auxiliary pumping and liquid nitrate in
cold trap (I).
     Probably the main critical parameter in
thermospray is the monitoring of temperatures or
controlling the temperatures, and we monitor the
temperatures of the source block (D), the aerosol
(F), slightly beyond ion exit cone as well as a
vaporizer (B).
SLIDE 2
     First of all I'd like to talk about parameters
that affect the thermospray results.  In order to
develop a good analytical scheme, we should have a
good understanding on the factors influencing
thermospray spectra and sensitivity.  Most of these
parameters listed do affect thermospray sensitivity
and a few affect the type of spectra recorded.
I'd like to go through each parameter and  show you

-------
briefly their optimization and their effects.
     The first thing I'm going to talk about is
the aerosol temperature.  You control the aerosol
temperature primarily by the vaporizer temperature
and the source block temperature.  But on the aerosol
temperature that's being read just beyond ion exit
cone, effects analyte sensitivity.
SLIDE 3
     You can see that the optimal aerosol tempera-
ture for this triazine is between 120 degrees C to
130 degrees C. If you're off by 10, 20 degrees you
lose a significant amount of sensitivity.  Also I
should point out this plot here is very dependent
on solvent composition.  As you increase the percent
water, your optimal aerosol temperature also in-
creases.
SLIDE 4
     Next, I'd like to show you the effect of
changing the percent of water, which you do in the
gradient dilution type HPLC analysis, would have
effect on sensitivity.  You can see here, I looked
at three different percentages of water in methanol.
As we increase the percent water, we gain signifi-
cant factors in sensitivity.  If we compare the
high and the low end, we gain about, close to two
                        82

-------
orders of magnitude in sensitivity.
SLIDE 5, SLIDE 6
     To compensate for this effect, we developed a
scheme to add water post-column and this way, even
though you're making a dilution of your sample,
you can gain over a factor of four in sensitivity.
     It's very simple to do.  We're using a coaxial
team which we can add our water or buffer post-
column and mix it with your HPLC effluent going into
the detector and minimize band broadening, in effect
enhancing our sensitivity.
SLIDE 7
     Also, the percentage of water effects slightly
the type of spectra we generate.  As I show you
here, we do see different types of ions either
(M+H)+ or (M+solvent)+ that are cluster ions, for
a percentage of water.  The lower percentage of
water in methanol, we see more cluster and more
acetate addition.   In a high percentage of water,
we see more pronation.  The same roughly holds
true for acetonitrile in water.  In general, though,
the change in solvent composition doesn't drastic-
ally affect the spectra and really there's no gain
or loss as one goes from one extreme to another.
                       S3

-------
                                                             I,'1'::'1 iff .'Si.11.,1..:11,,'!! f,.
SLIDE 8
     Another factor that plays an important role  in
thermospray is the selection of a buffer.  For ther-
mospray ionization to occur, a volatile buffer
should be present.  We evaluated five volatile buf
fers:  triethylamine, ammonium carbonate/ bicarbon-
ates, formates, and acetate.  For this organo-
phosphorous pesticides, you can see that ammonium
acetate produced the best response.  In some cases,
ammonium formate was comparable to ammonium acetate,
but for most of the work I've been using acetate be-
cause it gave the best response.
SLIDE 9
     As for concentration level of the acetate, we
found doing a plot of concentration versus (M+H)+
intensity, we see it plateaus near about  .08 M and
any addition really gains very little in sensi-
tivity.  For people who worry that the acetate will
degrade the separation, an ammonium acetate solution
can be added post-column.  For example, you can
add  .3 M solution post-column so that your final
concentration might come out to be .1 M.  So actual-
ly the acetates or the buffer requirements for
thermospray doesn't really affect analysis from the
viewpoint of a chromatographer.

-------
SLIDE 10



     Next, I'd like to show some examples of analy-



sis of pesticides, primarily carbonate pesticides



in the soil and water by thermospray.  Now that we



have a handle on optimal conditions where we hope



to give us the best sensitivity in thermospray, we



tried to design an analysis scheme keeping these



points I showed you in mind, to get the best possi-



ble results both chromatographly and by mass spec-



trometry.



     For sample preparation for water samples, we



stuck with literature-type extraction techniques.



We compared two extraction techniques, sep pak and



methylene chloride extractions.  We found out



that methylene chloride extraction came out the



best with a 98 percent recovery and sep pak was



reasonable with 50 percent recovery.



     One advantage to the LC/MS technique would be



to inject the entire volume of sample on column,



which is something you can't do with GC/MS unless



you dried down to a very low volume.



SLIDE 11



     The conditions we used are developed for the



carbonate pesticides that are shown here, Zorbax



ODS  (25 cm.).  We're doing a gradient/multi-

-------
gradient, solvent 50 prograrrio  You can see we're
working on normal flow rates (1.2 mL/min).  I have
a UV detector in line for comparison purposes»
Mass spec was scanning (m/z 160-600).  Primarily I
do most of my work in positive ion.  The vaporizer,
as I showed you before, is temperature dependent.
We compensate for the loss or change in sensitivity
of optimal temperature by programming our vaporizer,
so this way we can always maintain optimal sensiti-
vity with varying solvent composition.  The vapori-
zer is programmed down in sequence with the gradient,
SLIDE 12
     Here's an analysis or one example where you
have the HPLC UV trace and I selected ion chroma-
tograms for about five pesticides spiked into lake
water.  This water sample contained one part per
billion of each pesticide.  This is under full
scan conditions.  You can see here in the selected
ion chromatograms, each peak was easily detected
with very little noise, very little other ion inter-
ferences.  If you look in the UV chromatogram,
they're very difficult to pick up where these
peaks would be detected.
     We can go much lower.  We can see per part per
billion there is very little noise in most of these
                        86

-------
channels. .Due to detection limits on these types
of samples, we found for most of the carbonates we
could go close to about 10:1 part per tray*
SLIDE 13
     Typically the thermospray spectra for these
samples are very simple, consisting of an (M+H)+
or (M+NH4)+ ion, or a combination of both.  This
is one of the examples, propoxur, which showed in
the selected ion chromatogram.  Very simple spectra.
Fro most of the spectra, I usually just map out
the most or base peak for each of the compounds.
The big disadvantage for these simple spectra is
there's really no structural information gained
from fragmentation so it's very difficult to make
a qualitative identification of an unknown.
SLIDE 14
     In summary, for the carbamates we worked in
the concentration range from 100 parts per billion
to 10 parts per trillion and it proved quite linear
over that range.  We looked at about 15 pesticides
and they did show a significant variation in
detection limits.  For the ones that were able to
go down to 10 parts per trillion, we used for
determining the linear coefficient (0.99).  So
for about 8 or 10 that did go down to about 10

-------
parts per trillion, we did have a linear range of
about 10 to the fourth.
SLIDE 15
     This method worked very well for soil analysis.,
A 100 parts per trillion of pesticides were spiked
into soil, and as observed in the HPLC UV, it's
pretty hopeless.  I'm not sure you can make any
sense out of that chromatogramf however, in the
ion chromatograms you can pick up the pesticides
quite easily, looking at the base peak for each
representative pesticide.  Again, most of them
show very little interference.  I think propoxur
shows the most channel interference with some
other components showing common ion interference.
But in most cases, thermospray seems to be trans-
parent to a number of biological or environmental
matrix interferences.
     Next, I'd like to talk a little about another
example, analysis of dyes.  In this example, we're
looking at dyes in gasoline.  Dyes in general are
very difficult.  There's really no mass spec methods
available.  Dyes are added to gasoline for a few
reasons.  One, they're a fingerprint for the
manufacturers.  If you have four service stations
and there's a ditch nearby with gasoline leaking
                        88

-------
into it, somebody could point a finger at the
proper service station that is leaking.  Also they
use dyes to differentiate between leaded and un-
leaded gasoline.
     Typically, TLC has been the main way to analyze
dyes in gas.  TLC shows that the supposedly pure
dye component contains a multitude of components.
What is bought from a manufacturer with a given
structure is not very pure.  The only thing else
we know about dyes is that they're spiked into gas
about the part per million level.  Since the USA
uses millions of gallons of gas, we are talking
about kilograms of dye being used or thrown into
the environment after burning.
     To get an idea of what degradation of product
are formed,,we have to have a good handle on what
the initial products are to begin with.  So we try
to characterize some of the dyes initially just to
see what's  in the parent dye to get a better handle
on what will come out after burning.
SLIDE 16                              .             '
     An HPLC and a TIC trace for a commercial red
dye is shown here.  There's really not too much in
the UV— multitude of peaks, very broad peaks,
indicating probably a multitude of components.  In
                        89

-------
thermospray, quite a few peaks there.
SLIDE 17
     We knew the dye had this basic structure with
R-group as being a methyl group.  When we looked
at the selected ion plots, we found that-R-groups,
ranging from H all the way down to CIO H21 were
present and in many case, many isoraers for each
R-group were present.  So this supposedly pure
dye, or being sold as a pure dye, is a multitude
of alkyl homologs of about 20 to 30 components.
SLIDE 18
     Not only that, we also detected some other
colored dyes in the sample.  We determined the
structure of this dye by purchasing the standard.
I think the standard was the methyl R-group again
and showed the same type of spectra as the sample.
This orange pigment was found in this commercial
red dye and again, a whole range of alkyl substi-
tutes were found.
     So in this one dye, we had to deal with on
the upwards of 30, 40 components, making detection
of a specific component very difficult.  We are
looking at concentrations in the lower parts per
billion now, instead of in the parts per million
level.
                        90

-------
SLIDE 19
     To show a real example of a gasoline extract,
this was supposedly spiked with one part per million
red dye.  We did a sep pak silica gel extract.  There
is nothing in the HPLC UV trace where the dye
should elute.  However, we looked at HPLC selected
ion chromatograms.  The red pigment (m/z 381) and
the orange pigment (m/z 249) were detected.  You
can see the two very neat peaks way above noise
level as detected.  This we estimated, assuming
that each alkyl substitution has the same response,
was about 10 to 20 parts per billion of dye  in
gasoline.
     So in this way, if we wanted to characterize
the red dye, we might choose two or three of the
major components found in the red dye, measuring
the area ratios as a fingerprint for a particular
dye.  The multitude of components in the dye make
it very difficult to detect each with adequate
sensitivity.
     I'd like to go through one other example.
This is a study of mycotoxins and this was a big
issue last fall with Afghanistan and the Yellow
Rain issue.  I'd also like to talk about highly
toxic contaminates in feeds and grains.
                         91

-------
     The methodology for these toxins is rather
limited, primarily doing GC/MS techniques using
sample derivisation.  However, GC/MS techniques
have a few major disadvantages.  In El conditions
one detects low mass fragments below m/z 200.  That
mass range is very easy to have interferences from
blood and serum samples.  NCI overcomes some of
this problem by providing molecular anion infor-
mation.  However, to get reasonable sensitivity,
we have to do at least two derivatives to analyze
these mycotoxins.  There are extreme differences in
sensitivity of the different toxins with deriva-
tives.
     On the other hand, thermospray offers some
advantages.  First of all, there's no sample
derivisation.  We can bypass that step completely
so your samples don't have a short shelf-life time.
Thermospray does provide molecular weight information
for all components.  Primarily the spectra are con-
sistent of  (M+H)+ or  (M+NH4)+  ion with very  few
fragments.  If there are fragments, they're  very
simple  fragments like the loss of water.
     The common interferences  found in blood and
urine are pretty well transparent to thermospray
ionization, so the examples  I'll show you, there's
                         92

-------
very little backdrop problems.  Again, like I said
before/ there's no splitting HPLC effluents, so we
can inject an entire sample into HPLC and maintain
good sensitivity.
SLIDE 20
     For those not familiar with the structures of
buffing toxins, they're shown here.  Primarily T2,
HT2, DAS, T2 tetrol have this basic structure that's
varying Rl through R5 groups.  F2 has the structure
shown below.  With the weight ranges, you're dealing
from 296 on up to 466.
SLIDE 21
     The first example I have here is a variety of
toxins in the urine.  We have a quick isocratic
HPLC analysis. In about 10 minutes we can do a
complete assay of these four or five toxins.  Again,
we have good sensitivity for these toxins (two
nanograms injected.) You can see in the chromato-
gram there's a.high background, making F2 a little
bit more difficult to detect.
SLIDE 22
     Here's another example of detecting for HT2
in serum.  HPLC UV chromatogram shows the retention
time of HT2.  (See arrow.)  Here's the actual
thermospray selected ion chromatogram for the

-------
standard and sample.  The standard represents
one nanogram of HT2.  One can see detection limits
below one nanogram for HT2 are achievable.  It
also shows you the advantage of specificity of
the mass spectrometry over HPLC UV.
SLIDE 23  .
     One more example, the detection of T2 in
serum.  The serum sample again shows very few
interferences.  An one nanogram standard chroma-
togram is also given.  The mycotoxin work proved
to be quite linear and about from 0.1 nanograms
up to about 50 nanograms.
     One other thing I should mention.  Our detec-
tion limits can be reduced by about another factor
of 2 or 3f by performing post-column buffer addi-
tion.  We did some experiments where we added an
aqueous solution post-column (0.3 mLs/min), of
100 percent water.  This increased our overall
percentage of water because we were operating with
a high percentage of methanol in a range where
thermospray sensitivity is not optimal.
SLIDE 24
     This is an older slide comparing detection
limits for toxins using various methods.

-------
Recently, we've been able to reduce thermospray
sensitivity down to 0.05 to 0.01 nanograms.  You
can see how it compares with other techniques,
both NCI, GC, El, HPLC, UV.  I believe  HPLC/MS is
the method of choice for analysis of toxins.
     In conclusion, I think thermospray pretty
well revolutionalized the technique of LC/MS.  It
offers a new combination of sensitivity and
specificity and is very versatile for a variety of
compounds.  I think all these reasons have made it
one of the most popular techniques for LC/MS.
It's at the point now where it's just becoming to
be routine for some assays.
     I'd like to thank you for your attention and
EPA Cincinnati, Ohio, for their support.
                          MR. TELLIARD:  Questions?

-------
    t
0.006" I. D. Stainless
Steel Tubing
    HPLC In
0.8 to 2 mL/min
Schematic of Thermospray Interface
A  Directly heated vaporizer
B  Vaporizer thermocouple
C  Jet chamber
D  Source thermocouple
E  Ion exit cone
F  Aerosol thermocouple
G  Lenses
H  Quadrupole assembly
I   Liquid nitrogen trap and forepump
J  Source block heater
K  Vaporizer heater
L  Repeller
                                      96

-------
 o
 CO £
•CO   CD
-Q CO
 0) -C

SSf
 5 to
w "
|a
I?
CO CO
1
Q.
 O
 O

•S
 CD
           s
           CO
       o
       o
       "o
       o
 taoe
 g
V;
'w
 O
 Q.


 O
 O
              O
              o
             'o
       i '  i    ^^

       C  C  u.
   _  >  >
   9r  O  O
   X CO  CO
                      O

                      CD
                        .2
                        v»
                         CD
CD
O

O
O

CD
              CO QQ
CO •-
»;: co
Q. C
co o
o co
                   97

-------
                                                     o
                                                     o
                                                     OJ
                                                     o
                                                     CO
                                                     o
                                                     CO
                                                     O
                                                     OJ
                                                     CD
                                                    •O
                                                   __o
                                                     CO
                                                   4_o
                                                   J.o
                                                   (  CM
O
CD
O -
CO
.O
o
CO
                            o
o
ro
O-
cvj
                     Oil %
                           98

-------

-------
i  *
S1
8 *

_
.O
* T
s_ .
^fl) in
^^^^~ 00 i '
Z3 d
J3
C
^
i
j
i
i
i
bj


"o
9
to
£ ,
|2




r

1
in
CM
(0
O
>l
•n
I
I i
A





^^
M 	
• , 	












1 i
x Si
5 sll
                                ,0
                                +5



                                I
       100

-------
                                       I

                                      1
                                      Q.

                                      T3
                                      
-------
    The Effect Solvent Composition
      Has On Thermospray Spectra
100
     20%    50%   80%
       % Water in Methanol
   [M+H]+
   [M+Acetate]+
     or
   [M+H+CH3CN]
20%   50%    80%
 % Acetonitrile in Water
                        102

-------
        COMPARISON OF BASE PEAK INTENSITIES FOR
                 SfMAZINE IN EACH BUFFER
  50,000
  40,000
  30,000
CO

o>
  20,000
   10,000-
             0
                                         52,000
9,100
                         7,200  ,
                   360
                (NH4)2C03NH4HCOOHNH4HC03  NH4COOCH;
                           103

-------
   I
 c-J
 o —
 o 2
 c c
 o e
055
 3  CL
28
    s
   o
LU
CO

o
  2


x 8
O C
  1
                                             "E
                                             o
                     Ai|suaiu|

-------
                                 CO
 O


1
 (0
 Q.

 £
CL

_0
 Q.


 (0
 0)

i
                                                           CM
CO
                                                           1
                                                           DC
                                                           CO
                                                           CD
                                    105

-------
Column:
Solvents:
Gradient:
Flow:
UV:
MS:
Thermospray
Interface:
HPCL/MS Analysis

Zorbax C18 25 cm x 4.6 mm
Methanol/water (0.1 m AA)
50/50
       15 min
70/30(5)  — 85/15(10)
        Smin
1.2 mL/min
254 nm
Scan 160-600 in 2.0 s (+ ion det.j
Vaporizer 106° C
Jet 270° C
         96° C
            106

-------
                          Analysis for Pesticides in Water (1  ppb)
         HPLC/UV
   100.0 -i
   100.0 -i
   100.0 i
a>
CC
                 BPMC
   100.0 -i
                 Linuron
   100.0 -i
                                    m/z 266
                 Benzopropethyl
                                                      m/z 366
5:00      10:00
                                  15:00     20:00     25:00

                                         Time (min)
                                              30:00     35:00
                                                107

-------
                                     o
                                     • o
 ••••



 i
 o

Q.

*O


 S
 o
 0)
      I
                                    hS
                                     CM
O


CM
      CM
                                     CM
Q_

{2
                  CM
                                    J-8
                                     CM
                                    K§
            x

            a
            o
                                     o
                                    -CD
        O

        o
        o
                     p

                     S
                              108

-------
       Q.
       Q.


.2     ?
 CO     -Q
       Q.
       Q.

       O
       O

           Q.
           Q.

           O

           JQ
 0)
CD
D)

50  o
*—  .—
"co     -g
•    8®
                  o
       CD  »= O •«
       O  C     
-------
                         Analysis of Pesticides in Soil 100 ppt
         HPLC/UV
   100.0
       I/
         HPLC/ms
              Carbaryl        I   m/z 219

        ^^^JJ^^^
   100.0 T
   100.0 n
i
   100.0 n
   100.0 n
              BPMC
     m/z 225
              Propoxur     L m/z 227
               5:00       10:00
15:00     20:00
   Time (min)
25:00     30:00      35:00
                                               110

-------
            Commercial Diazo Red Dye
     HPLC/UV
100-1
     HPLC/MS
   *U"**N^^
         6:50    13:40
- 1 - 1 - 1
  20:30     27:20
 Time (min)
                                       34:10
                                111

-------
       HPLC/MS Ion Chromatograms for the Various
       Alkyl-Substituted Azo Benzene-Azo Naphthols
1001
100 n
iMn
100i
100
100 n
1001
100 -,
100-,
100 n
lOO-i
mli   Rd)

353   H
                                      OH
        II'
                      13:40
                           20:30

                          Tim* (min)
                                          27:20
                                                    34:10
                            112

-------
       Various Alkyl-Substituted Phenyl-Azo-Naphenols
           (orange pigments) Found in the Red Dye
100-
m/z  R.

249  H
                                 OH
                       \jy
                         R
                             N =N
100-
    263   CH3
lOO-i


        I	1	'     I
            6:50
                      13:40
                            20:30

                          Time (min)
                                          27:20
                                                    34:10
                          113

-------
       HPLC/UV
  100-r
                       Analysis of 1 ppm of Red Dye
                                  in Gasoline
  lOO-i
                                                                m/z 381
                                                                Red pigment
1 100 n
                                                                m/z 249
                                                                Orange pigment
6:50
                            13:40
   I
 20:30
Time (min)
27:20
                                                                  34:10

-------
 CH3
H  H
                     --R1
                    ~ H
            CH3    R2
Compound
T2
HT2
DAS
(a) DON
TZfetraol
R1
OH
OH
OH
OH
OH
R2
OAc
OH
OAc
H
OH
R3
OAc
OAc
OAc
OH
OH
R4
H
H
H
OH
H
R5
i-C4H9CO2
i-C4HgC02
H
=0(a)
OH
Molecular
Weight
466
424
366
296
298
                   CH3
F2
                            Mol. Wt. 318
                            115

-------
                                 Spiked Urine
 76,0
                             DON<2 ng)
                                 .«<^<-*VUv^n/>/'•>*•*''W*V^^
 47.0-
       mix 319
       m/z384
JL
                                   DAS(2ng)
 28.6-
180.9
01/1484
Total Ion Currant

A!"


             1:30      3:00      4:30      6:00       7:30       9:00      10:30
                                      TInw (mini
                                             116

-------
                                                                      f

                                                                      09
                                                                     .8
                                                                      


                                                                                    I   ^
                                                                                    .?   Q.
                                                                                    a.   x
                                                          117

-------
 E

J
 §
                                                                    -9.
                                                                     (0
u> 	
                                                                     10
                                                                     •t-
                                                                     n
          1
                                                                                E
                                                                                o
          p
          •5
          •
          1
          I
          I
                                       -    1
                          i     i           i      i      i      i
                                 A)iniaui| OAiniay
          a.
          ta
                                                    118

-------
119

-------
            QUESTION AND ANSWER SESSION

                          MRS. LAURIN:  Any idea
what kind of sensitivity moving belt LC/MS has
for carbamates?
                          DR. VOYKSNER:  From other
comparisons, I would say it's probably not too
good.  I think the closest comparison is carbonate
pesticides analyzed by moving belt LC/MS, was about
a factor of 10 to 20, less sensitive than thermo-
spray HPLC/MS.
                          MR. TELLIARD:  It's lunch
break.  Lunch is being served next door.  We're due
back here at 1:30.  I'd like to have you all come
in for the afternoon session on time so that we can
get out on the HMS Sinkfast at our due date.  Thank
you very much.
(WHEREUPON, a lunch break was taken.)
                        120

-------
                          MR. TELLIARDs  Our first
speaker this afternoon is not Bruce Hidy, but Marv
is going to move up and fill in.  Again, continuing
the discussion that we started before lunch, he'll
be discussing LG/MS and its application for
environmental measurement.  Marv.
                       121

-------
                   MARVIN VESTAL
               UNIVERSITY OF HOUSTON
                CHEMISTRY DEPARTMENT
         RECENT ENVIRONMENTAL APPLICATIONS
                OF THERMOSPRAY LC/MS
                          MR. VESTAL:  I'm afraid
my title may be a little bit misleading because I
had thought we might get around to actually doing
some real environmental applications.  We've been
involved in developing the thermospray technique
for about 10 years, but the definition of an
academic analytical chemist probably applies to
me even though I don't really consider myself.one.
The definition I heard is that an academic analytical
chemist is one who never runs any real samples.
     What I would like to do is to tell you a
little bit about the thermospray technique, how and
why it works, without going into all the details.
I could spend a couple of hours telling you every-
thing I know and some things I suspect, but I'm
not sure that would be productive.  I'll just give
a little bit of an overview of how the technique
works, some of the instrumentation that's involved,
and then give some examples of the various
                        122

-------
alternatives that you have available.
     Bob Voyksner gave/ I thought, an excellent
talk on thermospray this morning and with some
real examples.  I'm envious of some of his results.
However, you have to realize that he was doing all
that on one leg.  He was using the technique which
is sometimes called thermospray ionization, some-
times called direct ion evaporation, in which the
primary ions are produced directly by ion evapora-
tion from charged liquid droplets.
     This technique requires that you have a
substantial concentration of ions in solution,
on the order of .1 M/L for it to work well.  That
doesn't mean your sample has to be in anything like
that concentration, but you have to have something
that's ionized in solution, and in many cases of
real liquid chromatography that is not what you
have.  Furthermore, if you're using something like
ammonium acetate, that technique requires that your
sample have a higher proton affinity, for example,
in the positive ion mode than ammonia, and there
are a great many environmental samples for which
that is not true.
     So, from the beginning in our thermospray
systems, we have had an alternative method of
                        123

-------
ionization, such as a hot filament, which all of you
doing mass spectrometry are very familiar with.
It's a nuisance, but it does provide a very effi-
cient way of producing ions when you don't have
them by some surprising new technique.  We find
that external ionization is extremely important for
many applications as is the ability to do both posi-
tive and negative ions.  I will try to give some
examples of where one of these modes works and
others do not.
     First of all, I'd like to tell you a little
bit about what thermospray is, and since we made up
the word, I guess our definition should count.
If you put the words thermo and spray together, you
get something like the production of a jet of fine
liquid particles by heating.  That's basically
what we're doing.  We force the liquid through a
capillary tube, typically on the order of 100 to
150 microns in internal diameter, and supply enough
heat to that tube to vaporize almost all of the
liquid.  It is important that we not quite reach
100 percent vaporization.  Out of this capillary we
produce a supersonic jet of vapor which contains
some entrained liquid particles or solid droplets.
Since  the samples that we are interested in are

-------
generally somewhat less volatile than the solvent,
these sample molecules tend to be contained in the
droplets that survive this initial process.
     Now, it is also true that if one has ions
present in solution, then you can produce ions by
ion evaporation from the charged liquid droplets or
solid particles.  This process is in fact quite
analogous to ordinary field desorption.  The
difference is we don't apply any electrical field.
The electrical field is self-generated simply by
the charge on the particles and their small size.
If you start off with particles on the order of one
micron in diameter, and put something on the order
     5
of 10  elemental charges on it, you can calculate
that the electrical field at the surface is in
  7
10  volts per meter, which is the range you need
for field desorption of ions.  So the mechanism of
this ion production, to say it very quickly and
very simply, is field-assisted evaporation of ions
from these charged liquid droplets.
     As we discovered some years ago, if the fila-
ment burns out in your thermospray mass spectrometer,
you still have ions present with no visible means
of producing ions.  When we first discovered this,
we were quite excited about it.  We're still quite
                       125

-------
excited about it, but it's not the only way of
using thermospray.  The very nice thing about this
ionization mechanism is it is an extremely soft
ionization technique and it does not require a
neutral in the gas phase as an intermediate between
the sample in solution and the sample in the mass
spectrometer.
     That is a very important distinction because
you can define a non-volatile molecule essentially
as one which does not have a vapor pressure.  That
means it doesn't exist in the gas phase as a neutral,
Basically, if a sample doesn't have any significant
vapor pressure, then you're not going to be able to
analyze it by conventional electron impact no
matter what sort of interface or other technique
you use, because that requires having molecules in
the gas phase.  Thermospray does, in fact, work for
many kinds of molecules, for example, peptides,
nucleotides and so forth, which do not have any
appreciable vapor pressure; that is, they do not
exist as neutral molecules in the gas phase.  We're
producing ions in the gas phase directly from ions
in the solution and this allows a whole host of
things to be done by LC/MS and by thermospray
LC/MS that you couldn't do otherwise.
                        126

-------
      But thermospray LC/MS is also applicable to a
 whole range of molecules which are not nearly as
 demanding as this/ which may be either neutral or
 ionic in solution, but are smaller, more volatile,
 et cetera.   Those are the kinds of cases that I'm
 going to talk about primarily today.  I will not
 show any slides of peptides even though that has
 been one of our favorite systems for studying by
 this technique and has turned out to be quite
 fruitful.
      The last point on the previous slide was that
 thermospray is not the same thing as direct liquid
 introduction.  This slide shows what the jet (very
'similar to  what Rich Browner showed this morning)
 looks like  if you don't heat the liquid.  There's
 actually a  solid jet here which is not visible
 because of  the way the lighting was done and it
 breaks up into droplets by the processes that he
 discussed this morning.
      This is not the way we do thermospray, but
 rather we increase the heat so that we get  a super-
 sonic jet of vapor containing droplets.   When you
 reach the transition point,  you can see  a rather
 dramatic difference.  The particle sizes are much
 smaller and determined by the aerodynamics  now
                        127

-------
rather than the size of the nozzle.  This is at
something like 50 percent vaporized.  If we increase
the heat still further, we get over to a jet
containing fine droplets on the order of one micron
or smaller in diameter.  We still don't have good
measurements of the actual distribution, although I
think it is fairly close to a monodisperse aerosol
as well.
     This slide shows a schematic of a thermospray
system.  This is somewhat different than what Bob
Voyksner showed this morning, but the principles
are very similar.  Many of the details are different.
As he mentioned, it's very important to monitor the
appropriate temperatures in order to control the
degree of vaporization.  The vaporizer is a capillary
tube mounted into a probe which can go through, for
example, the conventional solids probe inlet that
is available on many mass spectrometers.  The
capillary is heated by passing current through the
capillary.  The power input to that capillary is
controlled by a thermocouple attached near the
entrance end.  By doing that we can compensate for
flow rate fluctuations by automatically adjusting
the power in the feedback system to keep this tem-
perature constant.  We also monitor the temperature
                        128

-------
at the exit of the vaporizer and downstream of the
sampling point, as well as monitoring the temperature
of the block itself.  In fact, we use two feedback
systems; one feedback system to control the power
into the vaporizer, and another one which controls
the heater input to the block.
     If we reproduce the two temperatures and
flow rates, then we can quite reliably reproduce
conditions and obtain the same results from one day
to the next.  It now is quite routine to set this
system up and run it every day, all day and all
night, for that matter, if you want to.
     The other thing that I want to mention is the
position of the electron beam.  What we normally do
in all the systems that I've had anything to do
with, is to supply an electron beam perpendicular
to the direction of the jet in the region, well
upstream from the normal position for an electron
beam in an ion source.
     We find that the sampling efficiencies in this
ion source are very similar to what you get in a
conventional chemical ionization ion source, even
though we have a very large flow superimposed on
this.  After all, we're vaporizing up to two milli-
liters a minute of liquid, which in the case of
                        129

-------
water corresponds to something like two standard
liters per minute going directly into the ion
source.  The reason we're able to accomodate this?
as Bob mentioned this morning, is we pump on the
other side of this ion source off to a cold trap
and a mechanical vacuum pump.
     All that's required to convert a conventional
mass spectrometer to a thermospray system is to
replace the ion source assembly with one designed
for thermospray.  We bring the vaporizer probe
in through a probe lock or equivalent, depending on
what's available on the instrument, and we provide
an additional mechanical pump and trap to pump away
the excess vapor.  Otherwise, conventional GC/MS
systems can be used with no modification and the
interfaces we've developed don't require any pertu-
bation of the GC connections at all.  The GC is not
connected to the ion source when you're doing
thermospray, but it's sitting there so that when
you put your conventional source back in, it's
ready  to do GC/MS or what other kinds of measure-
ments  you want.
     This slide shows a schematic diagram of the
control system where we sense the temperature, feed
it back to a triac control power supply and supply
                        130

-------
current to the capillary here for the vaporization.
It's all quite straightforward, and now works quite
routinely.
     This slide shows an ion source for a Hewlett-
Packard instrument being installed.  This assembly
goes in place of the standard ion source and the
optics are identical to the ones that are normally
present on the mass spectrometer.  The only part
that's really changed is the source block.  The
control system and pumping system are in the back-
ground.
     This slide shows the probe which has the vapor-
izer in it, going in through the standard probe
lock into the ion source.
     This slide shows the whole system including the
control for automatic recycling of the cold trap.
The vacuum pump is inside the cabinet.  The refrig-
erator is outside.  Of course, you can use dry
ice or liquid nitrogen if you like to fiddle around
filling traps every half an hour or so, but this
system can run unattended for up to 16 hours and
will recycle itself overnight while you're sleeping.
     I mentioned our excitement over this new ioniza-
tion technique and I'm not going to say much more,
but just show one example of the reasons for getting
                        131

-------
excited.  This slide shows some results on adenosine
back in the early days when we first discovered
this effect.  With the filament on, the total ion
current shows a rather high background.  Two
injections of one micogram each of adenosine show
a reasonable molecular ion, MH+, and significant
amount of a fragment ion on top of some background
at 136.  We then turned off the filament and did
the same injections again.
     There are two things that are initially surpris-
ing about the results.  First, the molecular ion
leapt up by nearly an order of magnitude when the
filament was turned off total area, and second,
the fragment ion almost disappeared along with most
of the background.  Now, that's not always the
case.  Some people have concluded from this that
there's no reason to have a filament present; in
fact, it even may be a disadvantage.
     Of course, everybody knows what a nuisance it
is to keep a filament running, particularly in an
atmosphere of water vapor and methanol and nasty
things like that.  But in fact, if you use a filament
that's designed for hostile environments such as the
thoriated irridium filaments used in non-burnout
ionization gauges, this is not a serious problem.
                       132

-------
Also/ you must put enough voltage on the filament
to get the ions to penetrate into this high pressure
region; we typically use a kilovolt for that purpose,
     The characteristics of the direct ionization/
without the use of a filament, are summarized in
this slide.  If you have multiply charged ions in
solution/ you very often observe the multiply
charged ions in the gas phase/ particularly in the
positive ion mode.  We've seen up to quadruply
protonated peptides.  You don't see these for small
ions because presumably they're too reactive in the
gas phase.  But we do produce abundant molecular
ions/ generally with little fragmentation.  This
can be good or bad depending on whether you're
interested in structure or in getting the maximum
intensity at a particular ion.  Of course/ the
performance depends rather critically on the
temperature, the capillary dimensions/ flow rate
and so forth.  We now understand all of those
factors in much more detail and can now control
them rather well/ but it still is rather more
critical than running with the filament present.
     One of the questions that always comes up
concerns calibration of the mass spectrometer.
                          \
With polar solvents such as water and methanol/
                        133

-------
PFK or perfluorotributlamine do not give good
spectraf so you need some other calibration com-
pound.  We have found that the standard mixtures of
compounds such as polypropylene glycol and poly-
ethylene glycol both work extremely well by thermo-
spray and give very reproducible regular spectra
for calibration.  This slide shows a mixture of two
cuts of polypropylene glycol.  This is a mixture of.
1,000 molecular weight and 2,000 molecular weight.
Paul Goodley from HP provided me with this slide.
It was run on one of the new HPs with the 2,000
mass range, and you can see that we have two over-
lapping Gaussian envelopes, one centered near
2,000 and one at about 1,000.  Unfortunately," the
tail of the higher mass one is off the scale of the
mass spectrometer, but I'm quite sure those ions
are there if you look for them.
     This gives you a regular sequence of ions
spaced 58 units apart.  You can use any molecular
weight cut of these glycols and they give a pre-
dictable spectrum which you can use to calibrate
the mass spectrometer.  The ions observed are all
ammonium adducts to the individual molecules of
the polypropylene glycol present.
     The next slide shows a few examples of some
                        134

-------
compounds which have some environmental significance



and are of considerable biological importance.



These were all done in the filament-off mode and are



typical of the kind of spectra one sees.  The MH+



ion is observed if the proton affinity of the



molecule is higher than ammonia; if not, you may



see the M+ ammonia.  Cortisol is a good test case



for a thermally labile compound because it is quite



sensitive to conditions, and people who have done



DLI have sometimes had difficulty getting molecular



ions on this.  In this spectrum obtained on 50 ng.



the molecular ion is the base peak in the spectrum.



     One interesting thing that you can do with an



electron beam that you can't do without it, is



electron capture CI, which may be a very powerful



tool both in ordinary CI analysis and in thermo-



spray.  If you have compounds present that capture



electrons, then electron capture can give you quite



high sensitivity and specificity for these kinds of



compounds.  In the direct thermospray ionization,



there are apparently no free electrons produced, so



one does not see M- ions in thermospray by the



direct  ionization.  Only if you have the electrons



present using the filament do you actually produce



these.  There are many examples of cases which work
                        135

-------
extremely well with the filament on by electron

capture and don't work particularly well by some of

the other ionization techniques.  Chlorinated

benzophenone is one example, and another is this

rather large floppy sugar, avermectin, molecular

weight 874, which gives the M- ion as the base peak

in the spectrum, with some structurally significant

fragments.  The masses around the 500 region corres-

pond to breaking off the major pieces around the

central branch.

     Another interesting case is the organic acids.

I know Bob Voyksner and others have worked on these.

They don't work well in positive ions and they don't

work particularly well in negative ions filament-off,

but with filament-on in the negative ions, they give

quite high sensitivity.  In one example, in the

positive ion mode, 10 micrograms of this sample was

injected and not detected.  On the other hand, two

nanograms was injected in the negative ion filament-

on mode, and one sees a very nice response, many,

many times noise.  I think we could probably detect
    *
thi'fe^at around the one picogram level in this case,

in selected ion monitoring.

     You can also do hydrocarbons, and I mention

that because some people think you can't do hydro-
                       136

-------
 carbons  by  thermospray.   You  can't do hydrocarbons



 by thermospray  if your solvent has a higher proton



 affinity than the sample.   If you use methanol or



 water, you  probably wouldn't do hydrocarbons anyhow



 for other obvious reasons,  but we've done quite a



 bit of normal phase work recently using hexane as



 a solvent.  So, you can do  hydrocarbons, filament-on.



 Filament-off, you don't do  hydrocarbons because



 they just aren't ionized.



     This slide shows some  of the advantages of



 having the  filament.  This  is a pyrethrin pesticide



 mixture.  The upper trace is  the LC/MS total ion



 current  trace for an injection of a mixture of five



 of these at the 50 ng. level.  As you can see for



 the first three, we get quite a nice response, but



 these other two, you would  not even know were there



 in the filament-off mode.   On the other hand, at an



 order of magnitude lower concentration with the



 filament-on, we see all five components.



     One reason that the sensitivity is low for the



 filament-off mode in this case is that this was a



gradient from 70 percent acetonitrile to 90 percent



acetonitrile.  At the higher concentration of organic



modifier, the direct ionization does tend to fall



off rather  rapidly.  So when you get up to about
                        137

-------
90 percent acetonitrile, the filament-off ionization

is not working very well at all.

     I just wanted to show one final example from

some work that Al Yergy and Dan Liberatto are doing

at NIH.  Al generously loaned me these next few

slides.  The question comes up quite a bit, "Well,

it's okay for qualitative work but can you do

quantitation?"  In fact, yes, I think the answer is

now that you can do quantitation and you can begin

to approach what you can do by GC/MS.  That may be

a rather bold statement because there certainly

hasn't been even .001 percent as much work done as

yet, so we still have some things to learn.

     This particular example is a mixture of sugars.

They're interested in monitoring glucose in plasma

samples from patients.  This is the LC/MS chromato-
                                      ,.          ' ''• , i '"
gram of these five sugars, done by the technique

that Bob mentioned this morning of using post-column

addition.  The separation for these sugars involves

pure water.  In this case they were using 0.6 mL/min,

of water in the column and added 0.4 mL/min. of

ammonium acetate downstream of the column in order

to get ionization without the filament.

     The spectrum of glucose shows mainly M+ ammonia

and loss of water; a very simple spectrum.  They're
                       138

-------
interested in doing isotope dilution experiments in
patients, so they wanted to determine how well they
could measure the isotope ratios for something like
this on column.
     This slide shows their standard dilution curve.
You can see the sort of scatter they have, but you
have to realize this is down at the .1 percent
level so that in fact, they really are doing quite
well in terms of the standard curve.  They are now
applying this to actual measurements of glucose
metabolism in patients at NIH and I'm sure will be
reporting on this work before very much longer.
This is a real application although not really an
environmental one.
     I would like to summarize by giving you an
update as to where the technique stands, at least
in my view.  Detection limits, of course, depend a
great deal on what kind of sample and what kind of
conditions and even what kind of mass spectrometer,
but our experience with a wide range of not too
difficult samples is that we can usually detect
them in the range of 1 to 10 picograms per second.
I put the per second in there because, of course,
in terms of the absolute amount, it depends very
much on what kind of chromatography you're doing
                       139

-------
and how wide your peaks  are,  but  if you use  some  of


these short columns with small particles, getting
                          i,      ,          I' nil1
peaks a second or two wide  is not all that hard.


In these cases, detecting on  the order of 1  to 10


picograms of a great many samples is quite feasible.


Getting full spectra on one nanogram is also quite


feasible for a very wide range of compounds.
                                          ,:;/'  4

     The linear dynamic range is at least 10 .


We can accomodate virtually any liquid flow, but


typically for the pumping systems that are available,


we work in the .5 to 2 milliliters per minute range.


Going to lower flows is no problem if you make some


minor changes in the vaporizer, but the standard


commercial systems that are available work best in


this sort of flow rate range.  The time constant  in


fact is considerably less than one second, so there


really is no detectible peak  broadening unless


there's something seriously wrong with the way the


system is being operated.


     We can handle any buffer.  That is, the thermo-


spray vaporizer will vaporize anything, including


non-volatile buffers in solution.  However,  these


tend to get deposited somewhere, often in your ion


source or in your pumping line or whatever,  so we


really don't recommend using non-volatile buffers.

-------
The other thing they do is clutter up the mass



spectrum by giving you lots of high mass cluster



ions which are a nuisance.  Buffers like ammonium



acetate, ammonium formate, trifluorocetic acid,



HCL all can be used to produce a buffer at almost



any Ph you want, I think, without using things



like phosphates and sodium salts.  But it may take



some minor changes in your chromatographic proce-



dures.  If you can't do without them, you can in



fact run non-volatile buffers and pay the price by



having to clean the system a little more frequently.



     Thermospray can analyze any sample.  When I



say any, that's obviously a bit of an exaggeration



because we haven't tried everything yet, but I



would say that a very high percentage of the samples



that we have tried do in fact work by one ionization



technique or another, provided you have the right



match between solvent and sample.  You obviously



can't do nonpolar samples of low proton affinity in



polar solvents of high proton affinity if you're



looking at positive ions.  It's the same as chemical



ionization in that respect.  There's nothing myster-



ious about it.  Everything as far as what works and



what doesn't work, is understood in principle,



although, of course, in> practice we may not know

-------
enough about the individual molecule to predict



what's going to happen, but we can explain anything.



     Finally, thermospray is gradient compatible,



and now we do have an automatic control system



which automatically compensates for both flow



fluctuations and solvent composition and maintains



the fraction vaporized constant throughout the




whole range.



     This slide of remaining problems is one that I




made a year or two ago.  Some of them are still



with us, I think, although many of these have in



fact been solved, either in principle or in practice.



The first one I just mentioned, I think now is in



fact finished.  We do know how to control the



vaporizer to routinely maintain the proper operating



conditions.  If we want to go to lower flows and



maintain the same kind of performance, we do need



to go to smaller diameter capillaries.  In principle



that's easy, in practice it presents some problems.



     In particular, the smaller the diameter of the



capillary, the more prone it is to plugging, par-



ticularly if you're not careful in keeping the
                                          'In, .


dirt out of your system.  I'm afraid we're not very



careful because with the 150 micron diameter capil-



laries we're using, it generally is not a problem
                       142

-------
and we are really quite sloppy in terms of worrying



about this.  We don't put filters in line because



they plug up much faster than the vaporizers them-



selves plug up, so they tend to be more of a nuisance



than a help.



     Item #3 is something that we're still studying.



We really do need to have a better understanding of



the downstream vaporization process and how to



maximize the performance for really non-volatile,



really labile molecules.  We can do it fairly



routinely, but as we get to the larger things like



the peptides, the sensitivity is quite a lot lower



than it is for smaller molecules.  Of course, you're



taking a beating here in many ways, some of which



you have no control over, but I feel that we still



have the potential for improving that.  Now we



require on the order of a nanomole, for example,



of a peptide of molecular weight, 1,000 or so, which



of course is about a microgram, to get a really good



spectrum from that molecule, while we may require



three orders of magnitude less for a smaller molecule,



I feel that there's still the potential for getting



the sensitivity down into the picomole or even



lower range for these larger molecules, but that



still remains to be worked out in detail.
                       143

-------
     Thermospray is now available for magnetic ana-



lyzers.  It's lagging a little behind the quadrupoles



because it is somewhat more difficult, but the basic



problems of dealing with the high voltage, et cetera,



are all well in hand now.  Kratos has a system



commercially available for magnetic instruments, and



systems will be commercially available, I think,



for all the standard magnetic instruments very



soon.



     Obtaining a better fundamental understanding of



the mechanism of thermospray is something I plan on



spending at least the next 15 years studying.  At



that point I probably will retire, and I suspect



there will still be some things to be done.  That's



the state of the art as it stands now.  Thank you



very much.



                          MR. TELLIARD:  Any



questions?

-------
            QUESTION AND ANSWER SESSION








                          MR. KROCHTA:  Bill



Krochta of PPG Industries.  You mentioned plugging




due to dirt in your system and dealing with these



small orifices.  How about the fact that you have a




lot of salts in that system and you're constantly




vaporizing them, do you have a problem during the



course of a day with this?




                          MR. VESTAL:  No, provided



the vaporizer is properly controlled.  That is, if



one runs for an extended period of time with complete



vaporization before the liquid exits from the capil-



lary, then of course you will get plugging rather



quickly if you have non-volatile salts present.



However, the way that the system is normally control-




led is to keep to say only 95 percent vaporized as



it comes out of that capillary.  The non-volatile



salts then are carried away in the droplets.  We



have run, for example, with sodium salts of ion



pairing reagents and things like this for days, in



fact, until we plugged up the ion source.  The



vaporizer would not plug, but in fact they would




sit down either in the ion source or the pumping



line.  I've actually filled an ion source

-------
completely full of solid material, say five grams



or so.  The vaporizer was still fine, but the ion



source was no longer working.  In fact, you could



see it happen.  Everything is working and all of a



sudden, the ion beam quits and the pressure in the



vacuum system goes down.  What happens is that you



actually plugged up the ion exit from the ion



source.  In some cases, you couldn't even see



across the ion source, the whole thing was full of



solid.  But you just take it out, hold it under the



water faucet, and it's all water soluble so it goes



away, put it back in and it runs.



                          MR. KROCHTA:  If you



don't completely vaporize, would this give you a



problem when you do quantitative work unless you



use internal standards?



                          MR. VESTAL:  It can.



If you really want to get quantitative performance,



it is important, I think, to completely vaporize



before you get to the ion sampling point.  This is



why I was talking about the downstream heating.



It is essential that you not completely vaporize



within the capillary tube itself, but it's certainly



desirable for getting the best performance to



complete the process between there and the sampling

-------
orifice, and as Rich Browner mentioned this morning,



that is not a trivial problem at reduced pressures



in the very short residence time.  We do fairly



well at doing that in the present versions because



we've learned in an artistic way how to accomplish



that, although I have to admit there are some



details about it that we don't really understand



yet.



     But you're right.  It is important, if you



really want to do the best quantitative work, to



get the sample completely or essentially completely



vaporized.  I think that's one of the sources of



variation that have been observed by some people



where some samples worked and some didn't.  I think,



in fact, the way they were operating, that the ones



that didn't work, weren't actually getting vaporized,



That's partly a matter of the design of the ion



source.



                          MR. TELLIARD:  Thank you,



Marv.
                       1*7,

-------
                          MR. TELLIARD:  Our next



speaker is John Ballard.  John is going to tell you



how to get two for the price of one, then you can



put a thermospray on it.  Everyone should have one



of these.  It probably costs $30, $40, I imagine.



John.
                        148

-------
               JOHN M. BALLARD, PH.D.

              LOCKHEED ENGINEERING AND
         MANAGEMENT SERVICES COMPANY, INC,
               DETERMINATION OF DYES
         BY THEMOSPRAY IONIZATION AND MS/MS
                          DR. BALLARD: Dyestuffs

are of environmental interest because of their

large scale production and their potentially hazard-

ous synthetic intermediates and by-products.  Just

to give you some idea of the scale and complexity

of the dye manufacturing industry in this country,

the annual production runs at about 350 million

pounds—I think that's 200,000 tons or thereabouts,

at something like 87 different manufacturing loca-

tions.  This results in a product line of about

1,000 commercially available products.  Forty-eight

to fifty Chemical processes are needed to adequately

describe the chemical inputs into this industry,

and the resulting dyes can be classified into 24

different structural groups according to the Colour

Index, which tends to be the Bible of the dye and

organic pigment industry.

     The toxicity and/or carcinogenicity of dye-

stuffs and some of their synthetic intermediates

-------
has been well documented.  Compounds such as 2-



naphthylamine, 4-aminobiphenyl, benzidine, Basic



Violet 14, Solvent Yellow #2 and Citrus Red #2 have



all been identified as actual or potential carcino-



gens, but there is some evidence that in fact meta-



bolites of some of these compounds are the true



carcinogenic agents.  The EPA and the government



have acted to reduce the number of colorants that



can be used as additives in the food, drug and



cosmetic industries in recent years, and of the



compounds that are still permitted, the permissible



amounts have been greatly reduced.



     Because of the need for sensitive methods of



detection or identification of these materials in a



variety of matrices, HPLC coupled with UV, visible



or spectrofluorometric detection has been widely



used.  Up to the present time, mass spectrometry



has not been routinely applied because of the



volatility requirement for generating electron



impact or chemical ionization spectra.  With the



recent developments over the last 5 to 10 years of



the desorption techniques such as californium-252



particle-induced desorption, fast atom bombardment,



secondary ion mass spectrometry and field desorp-



tion, a whole new range of molecules, initially
                       150

-------
the biomolecules e.g. carbohydrates, peptides,



proteins, nucleosides, have become accessible to



mass spectrometry.  We have now applied thermo-



spray ionization to the analysis of dyestuffs.



     I will also mention briefly in passing, that



the organic pigments, a specialized group of dyes,



so far have not been accessible to us.  We've



obtained about a dozen of these organic pigments



designed for use in plastics and textiles, and



they're generally intended to be impregnated into



the substrate rather than applied in solution.



We've yet to find a suitable solvent that will



enable us to look at these.  Of the dozen or so



pigments we've examined, only three could be solu-



bilized in DMSO; they were two sulfonated azo com-



pounds and a polychlorophthalocyanine.  Having



achieved solubilization, we thought we were okay,



but in our hands, DMSO was not a good solvent for



generating thermospray spectra, either in the positive



or negative ionization modes.



     As a generalization, it became apparent to us



that in contrast to the normal analytical situation



using standards of very high purity, for the dye-



stuffs we examined that situation was the exception



rather than the rule.  Even with the dye "standards"
                       151

-------
we're usingf it's obvious that there's a lot more
in there than just the stated dye.  Sometimes even
the dye itself isn't present, and I'll show you
an example of that.
     In general terms, we do see a lot of peaks
other than the protonated molecular ion and these
other peaks are variously due to solvents, dis-
persants and other compounds which the manufacturers
put in.  It's very difficult to find out what these
are because the industry tends to be somewhat
secretive and protective.  So we just have to accept
these materials, which are formulations designed to
meet a specific color and use application.
SLIDE 1
     The first slide is just to give you some idea
of the equipment we use.  The most expensive item
is the Finnigan TSQ instrument, which has the
capability of doing MS/MS analyses.  It's been
modified for thermospray ionization by Marvin Vestal
here, using the Vestec thermospray probe.  We use a
Waters 6000 pump as a delivery system just to get
the sample and solvent into the mass spectrometer.
We use the Rheodyne injector valve with a 10 micro-
liter loop.  At the moment, we're not doing any
chromatography, but we do have a short chromatographic
                        152

-------
column in-line between the pump and the injector



to damp out pulsations of the pump which, if not



removed, cause severe fluctuations in ionization



in the ion source.



SLIDE 2



     The operating conditions are pretty standard,



i.e., a small percentage of methanol in a large amount



of 0.1M ammonium acetate.  We find a flow rate of



about 1.3 mL per minute is optimum for us with our



trapping system.  As Marvin and others have



mentioned, the control of various temperatures,



particularly the tip and vaporizer temperatures,



are critical.  We have put a range up there because



no two days tend to be the same, depending on



which analyte we're looking at.



     For mass calibration, we use one of the poly-



ethylene glycol mixes, typically that of average



molecular weight 400, which in combination with



the lower mass thermospray ions gives us good



calibration from about m/z 30 up to say m/z 500-



600, which is adequate for our work.  As "standard"



conditions for the collision-activated dissociation,



we start with a collision energy of 20 volts and



argon is the collision gas at one millitorr pressure.



The term "standard" just refers to what we do in
                       153

-------
our laboratory.  We find that those conditions



usually generate good CAD spectra.



SLIDE 3



     This is yet another view of a typical interface



design.  I think maybe my colleague lifted this from



one of Marvin's publications.  He may recognize it.



The pump-out line at the bottom here, is at right



angles to the line of flow of the solvent and in



our system it goes to a large cold-finger trap



which is cooled with an FTS cryo-refrigerant unit.



We get a temperature of about -80 degrees C to -90



degrees C using methanol in the trap.  From just



doing rough measurements and calculations of what



goes in and what comes out of the trap at the end



of the day, we think that about 95 percent of the



solvent gets caught in the trap.  You may wonder



where the other five percent goes, and the answer,



of course, is into the mechanical pump on the



other side of the trap.  What comes out of the



mechanical pump when you change the oil is really



a hazardous waste in its own right.



SLIDE 4



     This is to refresh people's knowledge of what



the triple quadrupole mass spectrometer is and how



we can use it.  In comparison to gas-chromatography
                       151-

-------
where you separate the components of a mixture and
ionize them separately, with the triple quadrupole
you can ionize all constituents of a mixture simul-
taneously and then use the first mass filter to do
a separation.  Suppose you select a particular
ion, it may be a molecular ion or a fragment ion,
you don't know which it is before you start.  You
allow it to pass into the collision cell which is
pressurized with an inert gas, in this case argon,
at something like one millitorr.  The molecular ion
is fragmented and typically you generate a spectrum
which is qualitatively similar to a standard El
mass spectrum and you analyze the fragments that
are formed using the third quadrupole mass filter.
SLIDE 5
     This is just to give you an idea of what
actually comes out of the mass spectrometer in
terms of printout.  You see there are quite a lot
of spikes in this RIC trace on the bottom.  Those
are due to the pulsations of the pump which were
particularly bad when this was recorded.  This is
quite a high loading of material in there, and so
you can see that there's a reasonable amount of
spreading from the injector loop and through the
flow line.  With our configuration, we have a flow
                        155

-------
time of about 12 seconds from the injector into the



ion source.



     I'm going to divide my material dealing with



dyes into three areas.  The first'will be generaliza-



tions on what you might expect to see in the spectrum,



together with detection limits that we've established



for some of these materials.  The second area will



be characterization of two or three dye wastes.



The third area will be use of the TSQ to do the



structural determination of an unknown dye.



SLIDE 6



     The figures here were selected from a table of



about 16 dyes which we worked up; these were



deliberately chosen to show respectively the best



case (at top), a typical case in the middle (Dis-



perse Orange 13), and the worst case I could find



(Solvent Red 23), which has a large number of



peaks other than the protonated molecular ion in



the spectrum.  As I say, that's a typical case,



the Disperse Orange 13.  You might also notice



that these two dyes have the same molecular weight



and obviously the same MH+ ion.  The question is-



What ARE all of these ions here?-and it's a good



question.



     From our observations of the purest dyes that
                        156

-------
we've looked at, I would tend to say that they're



not fragment ions.  This is based on our observa-



tions and those of others published in the litera-



ture, that therraospray is a soft ionization process,



Also, we noticed that typically the spectrum we



generate from the molecular ion when we do CAD



experiments, does not contain any or all of the



lower mass ions that are seen in the full-scan



spectrum of the dye sample.



SLIDE 7



     We wanted to look at this in a quantitative



way and these are results obtained with dye



standards from a well known chemical supplier.



These .are the percent dye content as stated on the



label, tabulated with the percentage of the total



ion current that was actually carried by the pro-



tonated molecular ion.  This slide shows that, in



general, there's a reasonably good correlation.



For the Disperse Orange 3, I bracketed the 100



percent simply because no information was given



and so reading the manufacturer's handbook, you



were led to believe that the material is of 99



percent or better purity.
                       157

-------
SLIDE 8



     This second slide is a continuation of the



first.  You can see here that we've gone from



reasonable correlation to virtually no correlation,



especially in the case of Solvent Red 23, in which



there's very little of the molecular ion compared



with what's supposed to be in there.  Solvent



Red 3 is good.  For the first sample of Basic



Yellow 11 we looked at, also from this well known



supplier, the dye content was listed as about 60



percent.  So we ran its mass spectrum.  Even



expecting the worst, we were a little surprised to



find that there was absolutely none of the expected



molecular ion in there.  Experiments showed that



it's not even of the correct dye class.  It does



have the correct color, it is yellow, but that's



just about all you can say for it.  So this is one



of the pitfalls for the unwary.  Don't believe



everything you read on the label.



SLIDE 9



     Each one of these seven dyes is from a different



chemical class.  We've had a look at detection



limits and we find, like others have reported for



pesticides and other compounds, that even within



the same chemical class, detection limits can vary
                        158

-------
somewhat widely, say over an order of magnitude.



The detection limits were obtained by scanning the



mass spectrometer from about m/z 150 to just above



the molecular ion region.  We weren't doing selected



ion monitoring because in a real life situation,



if you're scanning a sample and you don't know



what's in there, you've got to scan for every



possibility, you can't just go looking for one



particular ion.



     The amount we get in the best case, I think,



is 15-20 ng.  Those figures are comparable to



numbers I've seen in the literature for pesticides



and other compounds.  The boxes that are highlighted



here refer to negative ion sensitivity and you can



see that the detection limit is an order of magni-



tude higher than for the same dye in the positive



ion mode.



     Another parameter the EPA uses is called the



method sensitivity, and it gives you a feel for



what you get out for what you put in.  Essentially,



it's the slope of the straight line obtained by



plotting the area counts versus the weight of



material.  One other point I should make about



this, of course taking the best case, is that if



we can see 15 ng., that's on a 10 microliter
                        159

-------
injection, then for this particular compound, Basic



Green 4, we can detect it in solutions of 1.5 ng/



ML.  I don't care to do the mental gymnastics



to convert that to ppb or ppm, but I think it's



pretty reasonable.



SLIDE 10



     Moving along to the analysis of waste streams



from dye manufacture, we obtained some waste stream



samples.  These were not chemically pretreated.



The only thing we did to them was to filter the



solutions to make sure that we didn't get anything



solid into the system.  I imagine from what Marvin



said, that it wouldn't destroy the LC pump, rather



it would block the vaporizer.  From a knowledge of



the manufacturing process, you can look up the



relevant books, find out what the expected inter-



mediates are, and then you can look for them.  In



only one case did we find the expected compound.



That's good and bad, I guess.  It means that maybe



the manufacturer is doing a good cleanup job and



he's getting most of the impurities out, especially



the dichlorobenzidine.  I think most people are



aware of the carcinogenic properties of these



benzidine compounds.



     Pigment Yellow 12.  We didn't see any of the
                        160

-------
expected intermediates in the second sample and we



didn't see the same by-product that we'd found in



the first sample, but we did find these acylated



toluidine derivatives and again, from the literature,



you can deduce that in fact this was a waste stream



from Pigment Yellow 14 or another pigment, not



from Pigment Yellow 12.




     In some cases, again the expected intermediates



are not present in the waste stream, but there are



certainly some organics present.  Again, if I just



put the number there, it means we couldn't identify



them and that is partially because with something



like 1,000 chemical intermediates used in the dye



industry, it is rather difficult to get hold of



all of these and run standards and generate CAD



spectra of all of them.  We can make guesses, but



we'd rather not.



     Acid Red 114.  We did have a sample of the



authentic dye standard and it was interesting that



the dye itself also contained the two by-products



that we detected in the dye waste.  These are



condensation products formed at one stage of



the synthesis.  P-Toluenesulfonyl chloride reacts



with phenol which is left over from a previous



stage.  Some of these compounds go back two or three
                        161

-------
stages in the synthesis.  If you're prepared to put



a fair amount of time and a little imagination into



itf you can track some of these reaction sequences



backwards.



SLIDE 11



     The picture here doesn't look quite as rosy.



These two direct dyes were in our waste streams.



You can see that/ again, they don't contain any of



the expected intermediates, but they do contain a



large number of other by-products, solvents, et



cetera.  It's very difficult to work out what some



of these are, even with the CAD capability.



SLIDE 12



     Now, you may remember in one of the first



slides, the typical and worst cases of dyes.  Two



of them had the same molecular weights.  Well,



here they are with their structures.  You can see



that they're very similar.  If you only have one



single stage of mass spectrometric analysis, and



because of the soft ionization, then assuming a



pure compound, all you see is 353, the MH+ ion.



The MS/MS capabilities of triple quadrupole mass



spectrometry allow you to differentiate these two.



You can see that although they do have some ions



in common at very low mass and the mid-mass range,
                        162

-------
these two molecules are clearly differentiated by
their CAD spectra which were recorded under identi-
cal conditions.
SLIDE 13
     The third area that we've looked at using the
thermospray and triple quadrupole combination are
some structural determinations on unknown dyes.
This slide shows the full scan thermospray spectrum
from m/z 120 up to 400, of Basic Red 14.  You'll
notice the presumed molecular ion at m/z 344.
There's another peak of about 40 percent abundance
at 346 and also these two ions at m/z 189, 174.
SLIDE 14
     This spectrum shows the daughter ions generated
by CAD of the 344 ion.  The important ions...well,
they're all important usually, but some of the more
important ions are those at 329, 314, 303, 289, 274.
If you look at 344 and wonder what goes to 289,
you'd probably go crazy trying to work out what the
loss of 55 is.  Well, it turns out that it's not
55, it's a loss of 40 and 15;  289 is the loss of 40
from 329.  These ions here, 173 and 158, are also
important and I'll come back to those in a minute.
SLIDE 15
     The ion at 173 was tentatively identified from
                       163

-------
its CAD spectrum by comparison with the El spectrum



which we had in an old publication from Analytical



Chemistry/ of a compound called Fischer's base, which



has this structure.  It's a trimethylindoline.  It



happens that this compound is a very well known



intermediate.  It's used a lot in the dye industry



in the manufacture of cationic polymethine dyes



which are used for dying polyester and other syn-



thetic fabrics.  We confirmed that this was the



structure by obtaining an authentic specimen of



this and running the CAD of its M+l ion.  That was



identical with the previously generated CAD spectrum.



SLIDE 16



     This is the daughter ion spectrum of the  ion



at 189.  You see there's a loss of 41 to 148 and



other subsequent losses.  The loss of 41 and/or 40



which we observed previously in the daughters  of



344, is very characteristic of a  2-cyanoethyl



group, and  the structure of the dye starts to  become



more apparent when you  look at the literature  and



you see that they can make these  polymethine cationic



dyes by condensing Fischer's base with,  usually,



substituted aldehydes.



SLIDE 17



     So if  you get out  your calculator  and add up

-------
the numbers, this has the right molecular weight,



188.  We obtained a sample of that compound and



generated its M+l ion in thermospray interface,



did the CAD and again obtained an identical CAD



spectrum, therefore confirming our guess that



this was the second half of the synthetic process.



SLIDE 18



     So we have the two molecules there, and you do



the condensation and get this quatetnized nitrogen-



containing molecule.  There are probably hundreds



of these with different atyl moieties and some



with an extended carbon-carbon linkage as well.



SLIDE 19



     Going back to the 344 ion.  That's it's struc-



ture, and it is the molecular cation.  Now the



losses of 15 and 40 are well explained by consecu-



tive losses of methyl groups, losses of the CH2CN



and/or CH3CN, and you can get all the way down to



274.  Considering the 173 and 158 ions, this mole-



cule is mass symmetric about that carbon-carbon



double bond and initially it wasn't obvious to



us which half retained the charge and formed the



m/z 173 ion.  In fact, it's the right hand side,



and 158, I think, is due to loss of the methyl



from the 173.
                        165

-------
SLIDE 20
     These are the daughter ions of the 346 peak
that we saw associated with 344 in the material when
we first looked at it.  From the approximately 3:1
ratio, we wondered at first if this was an indication
of one chlorine being present.  However, none of
the CAD spectra we looked at showed any losses
corresponding to elimination of small chlorine-
containing compounds and so the possibility of
chlorine being present was ruled out.  In fact,
there are some similarities in the low mass end of
the spectrum, but the high mass end is very, very
different from that of the 344 CAD spectrum which
again is evidence that we're not just looking at
an isotopic variant.  We think that the 346 is a
product that's formed by reduction of the 344 ion
somewhere along the line in the manufacturing
process.  Either the indoline ring or maybe the
carbon-carbon double bond linkage is reduced.
SLIDE 21
     We looked at another one of these cationic
dyes of known structure.  This one is Basic
Orange 21 which has a molecular weight of 315.
We wanted to see if we could be sure that the 173
ion which we said was derived from the right hand
                        166

-------
side of the Basic Red 14 molecule, was in fact



correctly ascribed.  This is the full scan spectrum.



This is one of the higher purity dyes, although I



imagine that m/z 315 is probably only about 30 to



40 percent of the total ion current.  We obtained



CAD spectra of these ions at 180 and 198, but we



can't identify them.  We think they're probably



solvents or dispersants that are used in these



proprietary formulations. A lot of the information



on these dyes is contained in the patents literature



which is pretty notorious for being ambiguous and



vague and sometimes misleading, so we really could



not identify these.



SLIDE 22



     This is the CAD spectrum of m/z 315 from Basic



Orange 21.  That's the molecular ion, and the ions



at 300, 285, 270 are due to the expected consecutive



losses of methyl groups from that molecule.   144



is the only other ion at low mass.  If you do the



numbers game, you'll find that it's cleavage at the



carbon-carbon double bond, with a hydrogen transfer



and charge retention on the right hand side of the



molecule.  So, we conclude that our assumption in



assigning the 173 peak from Basic Red 14 is correct.



     I've run out of slides, so that's more or less
                       167

-------
itf except that I would just like to make a few



conclusions.  I think for the analysis of dyes,



you've got to be sure that you're looking at the



dye that you think you're looking at.  If possible,



you'd like to have pure dye standards and that's



rather difficult.  At the moment we're not in a



position to do quantitative analyses of dye wastes.



It means we've got to get our lab coats on and



actually do some recrystallizations or chromato-



graphy to purify these.  We have taken delivery of



a brand new, nice, complicated Spectrophysics LC



recently, which somebody is leaning how to use,



and we eventually- want to get on line with the



mass spectrometer and do some chromatographic



separations followed by MS/MS experiments.



     In theory, the ability to use the first



quadrupole of a triple quadrupole instrument for



separation of ionized species means that you can



ignore chromatography, but I think you probably



would do that at your peril.  It's better to take



advantage of all the parameters and cleanups



available.



     That more or less ties it up.  I would like



to acknowledge Don Betowski of EPA, who is my



co-worker on this project.  Thank you.
                       168

-------
                          MR. TELLIARD:  Questions?



No questions.  Thank you very much.
                        169

-------
                                  SLIDE  CAPTIONS



  1.   Instrumentation




  2.   Operating Conditions




  3.   Schematic of  Thermospray  Interface




  4.   Mixture Analysis  by Triple Quadrupole Mass Spectrometry




  5.   Thermospray LC/MS Elution Profile




  6.   Typical Thermospray Mass  Spectra of Dyes




  7.   Dye Content vs. % Total Ion  Current carried by  (M+H)+



  8.   Continuation  of Slide 7




  9.   Detection Limits/Sensitivities for Dyes




10.   Characterization  of Dye Manufacture Waste Streams




11.   Continuation  of Slide 10




12.   Differentiation of Isomeric Azo Dyes by CAD of  their (M+H)+ ions



13.   Full-scan Mass  Spectrum of Basic Red 14




14.   CAD Mass  Spectrum of m/z  344 from  Basic Red 14



15.   Structure of  Fischer's Base




16.   CAD Mass  Spectrum of m/z  189 from Basic Red 14




17.   Structures of Synthetic Precursors of Basic Red 14




18.   Synthesis of Cationic Methine Dyes




19.   Structure of Basic Red 14 and CAD Mass Spectrum of m/z 344




20.   CAD Mass Spectrum of m/z 346 from Basic Red 14




21,  Full-scan Mass spectrum of Basic Orange 21




22.  CAD Mass Spectrum of m/z 315 from Basic Orange 21
                                            170

-------


c
,g
IS
c
rume
**
C/)
c
•••


rupole(TSQ
"O
(Q
3
0
0)
(0

0
sQ.
*ZZ
+-•
tO

c
(0
'E
c
iZ





>meter
S
^^j
T*
o
0
Q.
CO
CO
to
(0


0)
o
•g
0
_c

ospray
E
k.
o
•£
Vestec



Q.
E
D
Q.
O
0
o
o
<0
Waters



"co
0
o
'E*
LO
CM
t-
^
C
•o
o
DC
171

-------
(ft
c
acetate/methanol (97:3 v/v
HH     £
                O O
                o  o
                O O
                CO 
                CM r-

                "oT

                g

                ||

                H H*
         O  O
         o   o
         O  O
         CO  CO
         t-  CM
          i   i
         in  ID
"O
c
0
O
0
c
^^^J
£
05
Q.
O
mJ
VI Ammon
^i^M
6



1
C
E
E
CO
r-


r- CM
0
_N
'C
O
Q. *r?
S-&
H H



O
O
0)
Q)
0.



0
O
CM
D)
S
C
LU

T5
••
o
E
q
r"
C
o
o>
^

^^
                         c   * <

                        •S   !«
                         2
     (U
 L*   OC

£   5
^   ^
DO   U.
                         (0    C  «J
                        O    o  o
                             O Q
             172

-------
3
H
 .2
 TJ
  C
O 0)

•si
                                         a


                                        a.
                                        
-------
 I
 ti>
 O
 U)
 (0
 a> »-
 II
 •o
 CD
 (3
 c
 o

 1
 "o
 O
 w
 V)
 CO


 0) k.

 o£
 0.=
 gu.

•o
 (0
St
W S
o i
8
                                   .
                                   JE <0
                                 £ a "5.
                                 (O
                   r*
                                JS

                                o
                                  I
                                  0)
                                  C
                                  o
                                  O
                     17*

-------
00

-------


1


CO ^
CD <
Q *
H- 2^
0 -o
m ®
2 £
+•* 0
O (A
O -Q
a o
c
CO o
CO —
'W


J2
CD

"HH
"co
O t"
0- §
Xj* •
£
Q,
CO
O
C
r"
CD
JZ






















«•».
o
o
' T-
5
CM
AN!
So
r"
5"
Kj1
CM


CM
CM

CO

0
CD
C
CO
6

0
(0

0
a.
.£2
O

00
CO
00
^
CM


p
T—
5T
CM



?
CO
CO
in
fY%

CO
T-
<5*
in
CO


CM
IT)
CO

CO

0
CO
c
CD
o

0
(0
k.
0
a.
.12
Q

CO
r-
co
co
^


o
r-
0
00
r-


^_^
O
O
r-
O)
O)
mmm»

CO
r-
O
O
CM




















176
00
r-
^
'SJ1
CM


CM"
£.
0?
CO



5"
SI
55"
LO
Oft
* #
O
CM
5F
in
CO


CM
ID
CO




CO
CM
•o
0


4^
c
0

"o
CO

O)
00
00
en
^


^
00
en
O)



5"
r-
O
O
CM
» x
fs»
"«t
CO
•SJ"
CM





















cn
CN.
co
fx
t—


0
o

^r
00



p
r-
CM
00


5*
CM
O)
00
t~





















5"
T-
co
ID

-------




c
JD

__
CO
'T^
o
xo
(/)
>
CI
0
c
o
O
o
Q















+



M
^^^P
~
.Q
"O
.2
CO
O

c
o
3
O











C
fc»
3
O i-
c 5
™ $
JM ^***
CO


xp


c

?
O
0
>»
O
S5















fN








_

O
CM
CD
O


CO
0
3

5
0
w
0
Q.
CO

Q



O
CO









o
CO
CD
o


LO
O
ZZ
fli
>-
cy
(0
0
Q.
CO

Q



(O
0)









0
0
«^—


CO
0
o
c
CD
k.
O
0
c/>
0
Q.
(/}

O



r-
CO



t





LO
CD
O


CO
0
0)
c
(C
i_
O
0
V)
k^
fl^
Q.
(0

O



CO
CO









ID
CD
O


CO

^3
0
DC
0
CO
0
Q.
CO

Q
177

-------
 0)
 c n:

*i
 eo —

£
vP
^«
 c
 o
O
 0)
o  rs  if)
r-  CM  O
O (O
in  if)  ~T.  o
00  O)  O  If)
     .  o
CD  CD  t~  00
o  o  •—  o
8-1
CO
o
       00  O)
       CM  f-  00

       T3  T5  T3
       CD    -8
       O  O  O  (0
       CO  CO  CO  CD
,_
c
o
k.
CD
r-
g
_O
41%
(1)
< CD
0) 0)
a. a
E E
(0 (0
CO CO
                o
                "35
                CD
                CD
                 178

-------







CO
0
^^*
O
j5
C/>
0
"•M
[5
'w
c
0
0)
C/)
4-*
£
c
o
o

0
O





£
B>
]P
"35
CD

T3

-*•*
0)


c

• —
o
CD
0)
G




^j




fl^
Q




i
D)
C
CO
c
D
O
JH
(0
0
k.
5.


*r«
c
*—•
1


*N"
X
E.















LO
CO





O
O
CM
i
O
LO
r"



CO
LO
CO


CO
*""
D)
C
fam
cu
O
0)
s
Q.
.J2
Q





r^
r<








0
LO



CO
LO
CO




CO
T3
0)
DC
C
CD
>
O
CO






(O
r-





O
LO
i
LO
CM



0
0)
CO

CD
C
+rf
^^^
00
c
CD
O
CO
CD
O (O
5 CO
LJL CM






O)
CM





O
(O
i
O
CO



CO
ft




o
r-
0)
to
o
"55
CO
00



(O

-------




V)
O
3
|
^^
CO




•






















CD
^5
U
£
f^


"P
0
O



Intermediate



E
CO
CD

4"'
CO
+2
(0
CO






0
'E
CO
o
^

w
0




.cetoacetanilide
<
<
CM
1"
>
£

"0

'c
0
D)
'£












o
Z

0
ac
"5

,3'-Dichlorobenz
CO














0
T3
T3
0 1
."2 2
•«» «v
2 o
3 Q)
1 S
*^ CO
O O
^ ^J
0 0
0 0
^ ^

o o
z z




§
o
_Q
CO
(A
**•
ffl





















ID

O
Z




-Naphthol
CM


CO.
ID

T3'
0
DC
C
0
E
D)
CL














i
Q.
i
-Amino-5-chloro
CM
























O
Z


T5

)luenesulfonic ac
\*












CO
§
*^
3
if\
Vv
0
C
0
3
O
CL
^«
C
0
a.






-Toluenesulfonyl
Q.




^«
^—
^^
^
0
DC
"o
^

0
._
E
CO
o
it
3
(A
0
C
0

o
1
Q.

O
Z




0
,0
o
























o o
z z

0
[2
ik.i
,3'-Dimethylbem
henol
co a.
























o
z

1
£

°R '1
CO CO
fr-s
ll
1 • •»•
CM T3












180

-------

o
D
13
g
a.
i
00






















•o
0

0
tS
O

•o
0
£
0
-Q
O


c/)
0
Intermedial


E
(0
0






00





o
Z

1
CD
C
CD
C
CD
.Q
O
N
CO
i
CO '5

E
•••i
O














CD
C
"co
.c
£
Q.
CO
C
1
cJ
^
o
•S
>•
1
CO


















o
Z






TJ
'5
CO
.0
'E
o
H^
3
(0
'•D








181

-------
                       CO
                       LC
                       CO
                  •in
                   CO
                                  o
                                 ho
                                  CO
                           o
                           
-------
                                                                                     m
                                                                                    CM

                                                                                    CM
                                                                                    <0

                                                                                    CM,


                                                                                    00


                                                                                    CM

                                                                                    (O
                                                                                    CO
                                                                                    CM
                                                                                   cat
                                                                                   fffi

                                                                                   CM
                                        00
                                                   g
                                                   D

                                              °   .s?
                                            "~CM   fe
                   en
                   oo-
                                                                                         O
                                                                                         o
                                                                                         CM
                                                                                         O
                                                                                         to
                                                                    CO
                                                                    co-
o
O
O
in
                                                                                         UJ
                                                    183

-------
                          co
                         o>
                         CM-
                         n
                                                                             n
                                                                        n
                                                                        o-
                                                                        co
                                           O
                                        J-O
                                           co
60.
CM
                                                                            CM
                                                                           rt-
                                                                           CM
                                                                            O>
                                    CO
                                    r-.
                                            eo
                                                                       CO
                                                                       co-
                                                                                   O
                                                                                 -to
                                                                                   CO
                                                                                  CM
                                           O
                                          •O
                                           CM

                                                                                  O
                                                                                -in
                                                                                  o
                                                                               h-o
 o
 o
o
d
u>
                                                                                  LLJ

-------
o
      185

-------
                                                                                                             '"T; 'j'f'i'ilili1, "ii "I-"
                                                                                  O)

                                                                                  CO-i
CO
'fr-
                                                                       O
                                                                       CM-
                                                                          (O

                                                                          O-
                                                                                O)
                                                                                   to
                                                                                  CM
                                                                                   O
                                                                                   CO
                                                                                       o
                                                                                      —CO
                                                                                       o
                                                                                      —
                                             «-    ^7
                                                                                       •s
                                                                                      _o
 o
 o
•o

'6
10
ut

5
                                                       186

-------
z
o
 CM
X
o
 CM
X
(J
 co
X
o
                                                          o
                                   o
       o
       o
                          187

-------
            en
          ,1
           o
o
           o
            188

-------
                                                                                       —m
                          
-------
                                                                     <0
                                                                     <*-
                                                                     CO
                                       O
                                     I-to
                                       CO
                                                                   ,	Ill'
                                                                         in

                                                                        .8-

                                                                         g-
                                                                         CM
                                                                          10- -10
                                                                          CM
                                                                          CO
                                                                          or
                                                                          CM
                                                      S3-
         n
o
to-
                                                             m
                                                             co-
                                        CO
                                                                               01
                                                                               O
                                                                              •O
                                                                               CM
                                                                               O
                                                                              —to
                                                                               O
                                                                             f—o
 q
 6
 o
o
10
111

5
                                              190

-------
                                                                                    ov
                                                                                    CO
                                                                                    
-------
                             in

                             co
o

8-
                                                                      in
                                                                      CO-
                                                                      CM
                                                                        o
                                                                        I—
                                                                        CM
                                                                             
                                                                                —o
                                        _o
                                          in
LU
                                             192

-------
                          MR. TELLIARD:  Our next




speaker is Walt Shackelford from our Athens



laboratory.  He is going to talk about problem



solving and since he's from R&D, this ought to be



novel.  Walter.
                         193

-------
               WALTER M. SHACKELFORD

   UNITED STATES ENVIRONMENTAL PROTECTION AGENCY
      ENVIRONMENTAL RESEARCH LABORATORY-ATHENS
  PROBLEM SOLVING WITH MASS SPECTROMETRY AND FTIR


                          MR. SHACKELFORD:  Bill,

I'm glad you put us back in R&D.  You had us in

your division just a few minutes ago, and a few

long minutes ago, I'll tell you.

     What I'd like to talk about today is essentially

an extension of work that we've carried on in Athens

for the past several years.  It has to do primarily

with understanding or characterizing industrial

effluents.  This sort of work has gone on in Athens

for at least the last 12 years.  Initially, indus-

trial effluents were studied one plant at a time,

and work was done with kraft paper mills, organo-

phosphorus pesticide manufacturing and in textile

dye works.

     With the advent of the consent decree, indus-

trial wastewater studies took on a much different

profile in that a nationwide study was undertaken

to look at a specific group of compounds in 21

industrial classes plus the publicly-owned
                       194

-------
treatment works.



     Our contribution to that in Athens has been,



first of all, to make use of our experience in



earlier work and help determine simply through data



that we had in data base management systems, which



compounds among the priority pollutants were worthy



of study in resolving the original court list of



compound classes.  Since then, we have aided the



Effluent Guidelines Division by matching spectra



in reference libraries to EGD contractor's GC/



mass spec analyses, to determine what compounds



other than priority pollutants might be in indus-



trial wastewaters that at some later time might



become important.



     The third part of this study has been to try



to determine some of those compounds that were not



determined by spectrum matching and to perhaps see



the problems or design a protocol for identifying



unknown compounds in some sort of efficient manner.



I'm sure that if we checked in the audience, every-



one would have a good idea about how to get an



identification of a compound for which there was



no standard nor any reference spectrum, and



probably you'd come up with something very similar



to what we came up with.  If you look in the
                        195

-------
literature, that's exactly what you'll see.



However, in making this efficient, we seemed to



have had to learn by trial and error.



     Before I go any further, I would like to



acknowledge that some of the data that will be



presented here was acquired at Research Triangle



Institute.  Dr. Joan Bursey was the project officer.



The GC high resolution mass spec work was done by



Mike Harvey, at that time at Harvey Laboratories.



I would also like to acknowledge the influence of



Leo Azarraga at our laboratory, who first made a



good interface for a light pipe, making GC/FTIR a



possibility; and John McGuire and Don Betowski who



off and on worked on one of the unknowns I will



show you, for several years.



SLIDE 1



     This is an overall picture of the frequency



of occurrence of the most frequently occurring 50



compounds that we saw in our computer survey of



the screening phase of the Effluent Guidelines



work.  As you notice here, the priority pollutants



are certainly some of the most frequently occurring



compounds.  However, if you'll look again, you'll



see—and I'll do this for you—that only 12 out of



the first 50 compounds are priority pollutants.
                        196

-------
This has something to say about the need/ perhaps,



for surveying and characterizing industrial effluents



beyond just looking at priority pollutants.  I



don't think it's any surprise to anyone that's



done any analysis that priority pollutants aren't



the whole story.



SLIDE 2



     This second slide gives you an indication of



just how many possible compounds are there that we



don't identify.  In our data base management system,



we logged approximately 3,000 spectra that we



believe we observed five times or more within a



tenth of a retention time unit.  What this means



is, at least as far as we're concerned, there are



a lot more compounds that do not match our spectrum



library than do.  Of these, we chose some 55 for



further study to try to identify them.



SLIDE 3



     This brings us to the subject of the talk:



problem solving with GC/MS and GC/FTIR.  I recall



here at this group in the last eight years we!ve



had only one talk that concerned itself with GC/



FTIR.  I believe it was done by Jim Brasch of



Battelle.



     As we've seen from the earlier slides,
                        197

-------
 characterization  of  any  sample  is  going  to  require



 identification of unknowns.  Next,  if spectrum-



 matching  fails, barring  any  secondary information,



 we're going  to have  to do manual interpretation.



 The gist  of  this  talk will be that  the use  of



 alternate instrumental techniques may lower your



 overall cost and may even increase  your  success



 rate.  There are no  promises here,  but that seems



 like a reasonable tact to take.



 SLIDE 4



     To give you some of the limitations of spectrum-



matching, in a study that was done  by Bill Budde



 and Steve Heller of  EPA  several years ago,  it was



 found that if you matched the compounds  known to be



 manufactured in industry with the compounds in the



 EPA/NIH reference spectrum library, you'd find less



 than a 20 percent overlap.  This means that there



 are a lot of compounds manufactured that we do



not have  reference spectra for.  Also, if you



 total up  the gas phase infrared spectra  that are



 in libraries, you probably will not reach the



10,000 level.  One thing to be said for  the present



collection of gas phase  infrared spectra is that



they are quite pertinent to water analyses.  They



were chosen  for that reason.
                       198

-------
     Finally, and this is an important point as far



as we're concerned, if you don't have some sort of



corroborative data, your confidence levels for



everything but the best spectrum match is going to



be very poor.



SLIDE 5



     I'd like to illustrate with this slide.  This



is some work we did in evaluating the efficiency



of spectrum-matching.  The bottom curve here is



data that we took from the literature.  It came



out of Fred McLafferty's laboratory at Cornell.



The X axis here is the match quality increasing,



and along the Y axis, the number of hits that were



confirmed.  As you can see, at low levels of match



quality, we have low levels of confidence and if



we have a high degree of match quality, we're



going to see something on the order of 80 percent



confidence in our match.



     The two upper traces, however, involve data



that we collected  in Athens.  This data all has



retention time corroboration.  Not only did the



spectrum match, but the retention data matched



with that of our data base management system.  As



you see, even down for very low levels of K, we're



talking about something on the order of 60 percent
                        199

-------
 confidence we could have in a match.
      This  point  is  even  further  confirmed  when  you
 see  the  diversion of these  two curves.   The  bottom
 curve  is for  all compounds  and the  top curve  is
 that of  the compounds  except carboxylic  acids.
 When we  first plotted  the data and  saw the downward
 curve, we  felt like there was something  wrong.
      It  turns out that our  retention data  for
 carboxylic acids of chain length greater than
 14—this is all packed column data—was  so impre-
 cise,  that we really were only guessing  as to
 whether or not a carboxylic  acid was within a two
 carbon length.  The mass spectra of all  these
 are  fairly close.   When  we  take  the carboxylic
 acids  out,  we're left  with  essentially the same
 sort of confidence  at  the high end  of matching
 parameters as the earlier study  in  McLafferty's
 lab.
     If our spectrum-matching program hasn't  helped
 us,  the next  thing  we  probably will try  is spectrum
 interpretation.  The only problem is that  in  most
 laboratories,  this  is  going  to be very expensive
 because it  requires  time from your  most  expert
people.  I  realize  that in some industries, espe-
 cially where  there's a lot of  synthesis  being done,
                       200

-------
it probably is cost effective to keep one or more
people in the laboratory who do interpretation
regularly.  In most of our laboratories doing
water analysis, however, we're so busy trying to
get the samples through that we really can't spare
any of this expert help to sit down and interpret
spectra.  Again, unless there's some confirming
evidence, the confidence levels even for a spectrum
interpretation, are going to be fairly low.
     The solution to this cost problem, to us at
least, seems to target for identification only
those compounds that are deemed to be high priority,
Another thing is to build a data base management
system of identifications for later reference.
Our experience has been that a local library or a
historical library, comes into play very, very
often in determining the identification of com-
pounds.
     Another solution is to acquire data from more
than one independent technique.  This may be, in
the long run, cheaper than having a person on board
to do interpretation.  Our experience was that we
could contract out high resolution mass spec runs,
for instance, much cheaper than we could buy a
high resolution mass spectrometer for ourselves.
                       201

-------
     The criterion we use to determine whether or



not a compound is of high priority is frequency



of occurrence.  Again, you're not going to know



that if you don't keep your local data base and



record compounds that you have identified.  The



second criterion would be high concentration.



This is a type situation that would cause most



perturbation to the environment.  Finally, does



the compound have an adverse effect on the system.



This is the case where a compound may only occur



once and there might not be much o'f it, but it



may cause easily identified problems in the



environment.  In this case, you've got to be there



with an identification.



     At the present time, we have been working with



the Industrial Technology Division on a case of an



industrial effluent that had an abnormally high COD



but yet the effluent was passing all of the priority



pollutant standards.  This involves a series of



compounds of very high concentration but found



only in one effluent.  They're not occurring with



great frequency.  They are in high concentration



and they certainly have an adverse affect on the



COD from that particular industrial effluent.



That work, incidentally, will be reported on later.
                        202

-------
     The criteria for compounds that we chose for



study all had a frequency of occurrence greater



than five.  This was an arbitrary cutoff but one



that we felt, from our experience, gave us enough



retention data corroboration that we felt like we



indeed had a compound present.  The apparent con-



centration had to be greater than 50 parts per



billion.  The apparent concentration was determined



by comparing a response of the unknown with the



internal standard.  Now, since we didn't have a



standard for the unknowns, the apparent concentra-



tions could be off by as much as a factor of 100



from the actual concentration.  You can even look



at the priority pollutant phenols and see that



much variation in response factor.



     Finally, as far as effects on an environmental



system, since these compounds were found in indus-



trial effluents, their treatability is in question.



     The additional techniques that we believe to



be important to get a determination were high



resolution GC/high resolution mass spec to develop



a possible chemical formula; chemical ionization



mass spec in both positive and negative ion mode



for molecular weight confirmation, when that might



be needed; and finally, GC/FTIR for functional
                        203

-------
group determination.
SLIDE 6
     This  is a flow chart that we worked out for
identification of unknowns.  The first step is to
rerun the  sample extract using fused silica capil-
lary column GC/MS and check again for spectrum
match.  In the case that we might have had two
closely eluting compounds that in none of the
samples that we looked at were ever resolved, we
would see  the sum of those two spectra consistent-
ly.  So, if we could separate them'by a fused
silica capillary column, then of course we could
separate them into their components rather than
trying to  waste our time determining a mixture.
     If indeed we did find a spectrum match due to
our greater resolution, we then checked with our
data base management system to see if we have
found anything like that before.  If we have, and
if it matches the characteristics of something
that is already in our data base, we then look for
the standard to try to confirm the identification.
That did not happen but twice in the ones we looked
at.  If there's no spectrum match or if the data
base management system does not bring up a match
for that compound, we then have to decide on
                        204

-------
additional data needed.




     Some of the inputs to this are whether one has



a good idea if the molecular ion is present in the



sample for that particular compound, how closely



related are compounds that we found in the spectrum



matching program, and finally, do we think that we



really need to go the expense of high resolution




mass spec to try to determine that molecular formu-



la.  If indeed the case is that we need all of



them, then we get all that data, try to propose a




structure, and get a standard.



SLIDE 7




     Our confirmation criteria was that the reten-



tion with a co-injected standard must be identical



on the capillary column and the mass spectrum



identical with that of the standard.  Also, if one



had FTIR available, you would certainly want to



get spectral confirmation there as well.



SLIDE 8




     This is the first example of one compound that



was identified.  This is a good example of why not



to trust your spectrum matching program or data



base unless you have some confirming information



along with it.  If you take a look at this spectrum,



we have a peak at 199 and then a loss of 15 and a
                        205

-------
 loss of 44.  Also we have a big peak at 91, indi-
 cating that perhaps we have alkyl group on a phenyl
 ring.   We did not get a match in the reference
 library,  although the library indicated that there
 were compounds of similar spectrum that were
 alkylphenylsulfonamides.
 SLIDE  9
     The  first thing to do was check and see if  we
 indeed had  the correct molecular ion.   We  ran  iso-
 butane CI for this and found the pseudo-molecular
 molecular ion at  200.   Thus,  we  felt like  we had
 the  correct molecular weight.   Then comparing  the
 actual spectra that are listed in the EPA/NIH
 library,  we found  that the  two sulfonamides  which
 are  actually isomeric  with  this  compound,  had  very
 poor spectra.   As  a matter  of  fact,  they only  had
 six  peaks  in the whole spectrum.   So the matching
 program could  not  generate  a good  answer.  Never-
 theless, when  we checked  the  retention  time  of
 one of those compounds  we were able  to  get as  a
 standard, it was incorrect.  We  believed this  to
be a different  isomer  from  those listed  in the
 reference.
SLIDE 10
     The second unknown that I'd like to go  through
                        206

-------
that  illustrates some of our points is this com-




pound.  It had a molecular weight of 164, which was



confirmed by chemical ionization.  This compound



was not only seen in the effluent guidelines study



but was seen in an effluent study some three or



four years before that.  In that study, we got




GC/FTIR data and made what we thought to be a




reasonable determination of the structure.  It



turns out, after getting high resolution data,



that we were completely wrong in the earlier



study.



SLIDE 11




     The next step, after determining our highest



mass was indeed the molecular weight, was to get



high resolution data.  Of course, the computer will



calculate every possible formula that it can, given



the elements.  The more elements you give it, the



more possibilities it can come up with.  But if you



look at this one closely and have the benefit of



examining the spectrum closely, you'll see that



these two that have sulfur in them are probably



incorrect because we don't observe any sulfur



isotopic peaks.  These two that both contain fluor-



ine were later eliminated with the FTIR data.



This one with phosphorus is fine, but you can't
                       207

-------
find an analogue to it in any of the daughter ions
down at 149 or 131.  It turns out that our proposed
molecular formula for that is C10H12O2.  We took
our infrared data, compared it to the infrared li-
brary and there was no exact match for it.  Never-
theless/ a close match, or one that seemed to
be close to us, was a dihydrobenzofuran.
SLIDE 12, SLIDE 13
     Take a look at the actual FTIR spectrum of
the unknown.  You can see similarities in the
region at and in the region at.  The region has
some extra alkyl character that we felt belonged
in an ethyl group which, as you look at the mass
spectrum, you see the loss of an M -29.   This
peak, at about 3,600 wavenumbers, is due  to the OH
stretch, but the OH, to give a sharp peak like
this in the gas phase, cannot be intramolecularly
bonded.  That is why we have decided the  oxygen
would be on the opposite side of the ring from  the
hydroxyl group.
     We were unable  to find  a standard to confirm
this compound.   In a later edition  of  the NIH/EPA
spectrum library,  there  is an isomer of this  com-
pound.   It  is a  dimethyl  isomer  rather than being
the ethyl, however, and  it has the  hydroxyl and
                        208

-------
the oxygen in the furan ring adjacent.  We believe



this to be incompatible with the infrared spectrum



showing the free hydroxyl group.



SLIDE 14



     Here's a summary of some selected analysis



results and to give you some idea of what I feel is



the power of the added information from FTIR.  The



only unknown in this group that we were able to get



infrared data on is one of two that we were able to



assign a structure to.  The high resolution MS data



gave us the ability to come up a reasonable formula,



but only with the FTIR were we able to assign a



good structure.



     As you can see from that table, we didn't get



FTIR data on every unknown compound.  One of the



problems with FTIR is that only a small reference



library is available.  The lack of sensitivity,



though, in our case, seemed to be our biggest  •'



problem.  The way we were able to get around that



in the few cases where we were successful with



FTIR,  was to use the Unicon sample concentrator as



a front end for the GCIR.  This way we were able to



inject up to 50 microliter injections and thus



increase by a factor of 10 to 50, the amount of



injected material.
                       209

-------
SLIDE 15



     There are quite a few interesting references



in the recent literature combining GC/mass spec



with GC/FTIR and also some applications that I



offer to you here.  Interestingly enough, in all



the recent applications of FTIR, you'll always see



it either combined with GC/mass spec in the same



instrument or at least the data combined with



independently taken GC/mass spec data.  The reason



for this is that the two forms of data are



so complementary that summed together, they form



a much bigger total information package than taken



separately.



     Wilkens was really the first person to inter-



face a GC/mass spec/FTIR system.  He was at Nebras-



ka at that time.  He has since gone to UC Riverside



and has built a GC FT/MS FTIR system.  Richard



Crawford at Lawrence Livermore had a paper in



which he did identify in the same type scheme



we're talking about here, some unknown compounds



using GC/mass spec and GC/FTIR.  Gurka and Betowski



did some work with packed column GC/mass spec, pack-



ed column GC/FTIR and then later, Shafer and co-



workers at Battelle did the same samples using



capillary columns.  In that work they found FTIR
                       210

-------
to be superior to mass spec in identifying compounds,



However, the results are somewhat skewed since



they were looking at alkyl benzenes in which IR,



of course, is much stronger confirmatory tool than



mass spec would be.  Finally, the report from which



some of this data is taken, Joan Bursey of RTI in



the EPA report dated 1984.



                          MR. TELLIARD:  Questions?
                        211

-------
                                  Overall
1500
                                                         Priority Pollutant
                                  Compound
                                       212

-------
         UNKNOWN COMPOUNDS
   3000 SPECTRA SEEN 5 OR MORE TIMES WITHIN
±0.1 RRT UNIT

55 CHOSEN FOR FURTHER STUDY
                       213

-------
         PROBLEM SOLVING WITH GC/MS AND GC/FTIR






0 CHARACTERIZATION OF  A SAMPLE REQUIRES  THE IDENTIFICATION



   UNKNOWN COMPOUNDS.






0 IF  SPECTRUM MATCHING FAILS,  MANUAL INTERPRETATION IS



   REQUIRED.






0 THE USE OF  ALTERNATE  INSTRUMENTAL TECHNIQUES  MAY LOWER



   OVERALL COSTS AND MAY INCREASE SUCCESS RATE.

-------
           SPECTRUM MATCHING  LIMITATIONS








0 LESS THAN  20% OF  MANUFACTURED COMPOUNDS ARE  IN



   COLLECTIONS OF REFERENCE MASS  SPECTRA  (BUDDE AND HELLER).






0 ONLY ABOUT 10000 COMPOUNDS  ARE IN COLLECTIONS  OF GAS-



   PHASE IR SPECTRA.






0 WITHOUT CORROBORATIVE DATA,  CONFIDENCE LEVELS FOR ALL BUT




   THE BEST SPECTRUM MATCHES  IS POOR.
                                 215

-------
    lOO-i
     90-
     80-
     70-
 
-------
                          BEGIN
GC/HRMS
                            V
FSCC
GC/MS
\
/

SPECTRUM
MATCH
9
\
NO
/
YES N
s
DECIDE ON U
ADDITIONAL
DATA NEEDED \

\
/
GC/FTIR

/

                         PROPOSE
                        STRUCTURE
                             V
                            GET
                            STD
                                                DBMS
                                                MATCH
               GET
               STD
                                                ^
CIMS
                                     217

-------
            CONFIRMATION CRITERIA
1.   RETENTION WITH CO-INJECTEU STANDARD IDENTICAL
     ON CAPILLARY COLUMN

2.   MASS SPECTRUM IDENTICAL WITH STANDARD
                            218

-------
r
g«
ss
cviw

Ii  s
     g
     s
     r>j
gsg-
a^-e

Sla§
iil-
                                                                              a
                                                                               "
                                                                                        0)
                                                                                        C

                                                                                        0)
                                                                                        CO
                                                                                        CD
                                                                                        G1
                                                                                        vO
                                                                                        m
                                                                                       o c-i
                                                                                       « o
                                                                                       i-i c-l
                                                                                       V  G
                                                                                          OJ
                                                                                        • O
                                                                                       4J  OJ
                                                                                       CO
                                                                                       "
                                                                                       a  i-
                                                                                       € 4J
                                                                                       O  GO
                                                                                       O  3
                                                                                      Cu ^
                                                                                      W
                                                                                         0!
                                                                                      m fH
                                                                                      o te
                                                                                      3  e
                                                                                      M  0)
                                                                                      •U J=
                                                                                      O  O
                                                                                      0)
                                                                                      CU  O
                                                                                      CO -H
                                                                                         C
                                                                                      co  cd
                                                                                      w  M.
                                                                                     CO

                                                                                     PQ
                                                                                      60
                                          219

-------
SB
                                                                                           K-
4ANSS*
                   I          I
                                               ->	r
                                                                                                          -S
                                                                                                                                E
                                                                                                                                O
                                                                                                                                0)
                                                                                                                                c.
                                                                                                                                to

                                                                                                                                w
                                                                                                                                CO
                                                                                                                                o
                                                                                                                                N rH
                                                                                                                               •H CO
                                                                                                                                C O
                                                                                                                                O CM
                                                                                                                               tH  (3
                                                                                                                               n)  cd
                                                                                                                               o  u
 Q)   «
j:  AJ
 u  co
  ,,  a)
 0)  !-<
 d  a)
 Cfl  4J
•u  C
 3  -H
J2
 O  H-l
 in  o
                                                                                                                               OJ
   T3

 >  §
•H  O
4-1  Q.
 03  g
4J  O
 C  O
 0)
 co 
-------
                                                                                                       o
                                                                                                    r-co
                                                                                                     -  o
                                                                                                     -to
                                                                                                     -  o
                                                                                                     -Tf
                                                                                                    -C\2
                                                                                                    - O
                                                                                                    -O
O
o
o
o
                                                                                                                -o
                                                                                                                 o
                                                                                                                 n
                                                                                                                u-i
                                                                                                                00
                                                                                                                m
               Q)  O
                  o

              4J  CN
               tO =fe
               cu
               v-  c
               Q)  CO
              4_1  O
               c  to
              •a  to
               c  3
               3  -o

               c. -H

               o  to
               u  c
                                                                                                                    o
                                                                                                                u-i  cfl
                                                                                                                O 14-1

                                                                                                                E  C
                                                                                                                3  cd
                                                                                                                u  a)
                                                                                                                a) -a
                                                                                                                c. -H
                                                                                                                to  u

                                                                                                                co u
                                                                                                                to  to
                                                                                                                03  Q)
                                                                                                                S  a.
                                                                                                                cu
                                                                                                                u
                                                                                                                3
                                                 221

-------
Figure 5.160.  Accurate mass measurement data, extract 5855
            (acid), pesticides manufacturing industry.
      High Resolution Mass Data
M/E
164





149








131







C
10
7
8
7
7
7
9
6
7
6
4
9
6
6
3
9
6
4
6
3
6
3
3
H
12
13
11
16
17
17
9
10
8
13
12
10
11
14
15
7
8
7
11
12
9
10
13
16Q
2
3
0
2
2
0
2
3
0
2
0
0
1
0
1
1
2
0
1
2
0
1
0
F
0
1
3
0
0
0
0
1
3
0
3
0
1
0
1
0
1
4
0
,1
1
2
1
d*g
0
0
0
1
0
1
0
0
0
1
1
0
0
1
1
0
0
0
1
1
0
0
1
p
0
0
0
0
1
1
0
0
0
0
0
1
1
1
1
0
0
0
0
0
1
1
1
MMU
-4.8
-36
-7.2
-1.4
8.1
-9.6
3.9
5.1
1.5
7.3
4.9
-4.3
-3.2
-0.9
0.2
1.1
2.2
-0.2
4.5
5.6
-6.0
-4.9
-2.6
                                         164.0885
                                         149.0563
                                         131.0486
                         222

-------
enzofuran
                                                    CM

                                                     O
                                                     )-i
                                                    T3


                                                    •H
                                                    u-i
                                                     O

                                                     e

                                                     i-i
                                                    4J
                                                     CJ
                                                     0)
                                                     O.
                                                    CO

                                                    Pi
                                                    M

                                                     0)
                                                     CO
                                                     CO
                                                     O
                                                     a.
                                                    PM
                                                    w

                                                    -a
                                                     1-1
                                                     «
                                                    •a
                                                     d
                                                    in
                                                     
-------
sotreqjosqv
                                          to
                                          m
                                          CO
                                           u
                                           ca
                                          4-1
                                          X
                                          
                                             O
                                           p3 CO

                                           1 3

                                           E co

                                           £E
                                           cj >-a
                                           M t-l
                                           CO O
                                           r-l -H
                                           rH 4-1
                                           •H CO
                                           O. 0)
                                           cd O<
                                           O

                                           C '^^
                                           o ^o
                                           CJ -H
                                           ca o
                                           C «
                                           u->

                                            cu
                                            60
       22^

-------



























to
P
rH
3
to
01
K

to
•H
to .
r^"1
.-1
C
rf
0)
P
CJ.
CD
r-i
0)
w

IM
o

t-,
^|
(rt
£5
E
3
CO
CM

V
r- 1
J2 '
IS













-O
(1)
iH 03
IB D H
4J H
C 03
0> 0> VH
§3
O1
Vl -H
P C
0) .C
c u
H 0) CO
EH




•O
O
4->
fO -
r-l P
3 U
P 3
01 i
O •!
ft CO





*O *•*
0) W
-P •-'
IB (8
r-l i— 1
3 3
P E
01 >-i
O O
ft b<




to
<4-l CD
O 0
c
(-1 CD
CD H
<"J }•
E 3
3 U
a u
o





^•t
M
C
a
4-
(0
u

ra
•H
P
W
3
C
HI
=3= C
o
1 t .^J
18 tj
V4 (0
•P ^
5^









w w
y • S
K PM
tX hC










0) 0)
c a
o o
a z







H
CJ
tJ M O
U O T
n 03 Z '
O *~ oo

co in CTI
CJ CJ CJ









en o
T^







O
•H
4J
0)
•H
4->
^

^4 *O
0) C
Da (8
(0
ft 03
4J
'O CM
C (1) -O
t8 E O
0) O
CU 03 CJ3
•3 i
ft <
-~ 2
M •^
O ffl
i , """
*3" VO
CO «*1





M
CJ

w
CO
s
&
s










0)
o
• z





co
CM
o
ft
CM
i-H
CJ

?_.
cn
CJ









VO











0)
P •
0
3

o
^4
ft
r-l
0
•r-l
C
.c
u
M
CJ
5
o
CO
CM


a,
M
- f
EH
b

%
03 CO CO
S 23 TS1
K K K
S ffi »C




,

,^\. .
*r ^s
\ ^
\ J
C C **4 ' >>^*V\
a a ° ~\()/

v ^ /
5^ x*^
O f^
>,
t^

CM
fSJ Jj
CMZ ^" CM
m O 'O O C5
O VO VO VD -
CJ CJ U CJ CJ









cr\ in m





„


03
U
-H
4J
0)
5
G

W

TJ
C
(8 01
WE 13
03 -H
•i-l C CJ
P -H .H
03 E P
(03 01
rH iH CU
CJ \ HH
< CQ O
VO CO
ro in in
^ CM • ^
CTl r- CO
r- CM in









W

s
33










0)
I









CM
Cu
o

"tb
o









cr*
CM
















W
0)
'o
Jj
01
M
U

CO
en























0)
o












' 0)
o
a .









m











01
i— i
18
O

'g
CD
CJ

ganic
j^
o
H
3
m
VO
ro
in





H
o

^
C/}
*SJ
9
JTJ


1 - S3
53- K CM
U 0 0
(M U
/-. n
^CM ^-v

O S
^^ u
33 V
U Q










CD1*
VO
r-
E
CO
U%









m














^

-------
                         REFERENCES






A.  COMBINING GC/MS WITH GC/FTIR



     CHARLES WILKINS, ET AL, UC RIVERSIDE, ANALYTICAL CHEMISTRY,



               1981, 113; 1982, 2260; 1984, 1163.






     RICHARD CRAWFORD, LAWRENCE LIVERMORE, ANALYTICAL CHEMISTRY,



               1982, 817.



B.  APPLICATIONS






     DON GURKA AND DON BETOWSKI , EMSL-LV ,  ANALYTICAL CHEMISTRY,



               1981, 1819.






     KEN SHAFER, ET AL, BATTELLE COLUMBUS, ANALYTICAL CHEMISTRY,



                   , 237.
     JOAN BURSEY, RTI , EPA-600/S4-84-072 , 1984.
                                       226

-------
            QUESTION AND ANSWER SESSION








                          MR. GORE:  Bill Gore,



American Cyanamid.  There have been some recent



reports of dramatically improving PTIR's sensitivity




and resolution with cryogenic trapping techniques.



Have you had a chance to do some experiments there




and do you have any experience to draw on or comment?



                          MR. SHACKELFORD:  No, I




have no experience whatsoever with cryogenic



trapping.  I would say that the sensitivity that



has been reported in the literature has been much



better than what we've experienced, but it could be



that in many of our cases we're talking about a



small peak and a large background and that is



perhaps interference rather than sensitivity problem.



                          MR. YOUNG:  I'm Jim Young



from the University of Arkansas.  A few years ago



we were working with complex samples, or samples of



complex mixtures, and we found that the relative



retention time of a compound would shift if you



pass a sample through treatment so that you removed




a lot of the junk.  Did you find this to be a



situation in your samples?  I think you mentioned




that you used relative retention time as a key
                        227

-------
identifier, but it was a problem for us because it
would shift significantly with treatment.
                          MR. SHACKELFORD:  One of
the things that we did do was to go through some
cleanup stages to try to get a clean peak to go to
the GC/FTIR and yes, after cleanup especially, when
we're talking about same compound but a much
different matrix injected, we did have significant
changes in the relative retention times.
                          MR. YOUNG:  Do you see
that as a major problem in using this particular
confirmation technique?
                          MR. SHACKELFORD:  In the
way we use it, no, because our samples in our data
base management system would be kept separate from
that of any later run acquired under different
conditions.
                          DR. LESAGE:  Suzanne
Lesage, Environment Canada.  You're talking about
your data base management system.  Is there any way
of sharing with the world this kind of information?
I think everybody doing industrial work has the
same problem of the unknown spectrum and maybe the
MBS library should contain unknowns rather than
knowns.
                        228

-------
                          MR. SHACKELFORD:  Our




system is built on a commercially available data



base management system called Inform.  That's not



the most up-to-date data base management system.



One of our problems is that so much of our data is



packed column.  Fortunately, there seems to be a



consensus now on fused silica capillaries and




perhaps in building a new data base, that would be



more appropriate.  But I certainly do think it



would be worthwhile to perhaps discuss that, but I




don't know that you would ever be able to get some



kind of uniform system that everyone would feel



like was appropriate for their own laboratory.



                          DR. ERICKSON:  Mitch



Erickson, Midwest Research.  A kind of follow-up to



the first question.  Walt, as you know from this



study, the throughput on GCIR tends to be maybe



half to less, maybe a quarter, of the throughput



you see on GCMS just because the data systems aren't



nearly as facile as what we're experiencing on



GCMS.  With the interface on the cryogenic trapping,




my understanding from their salespeople is that the



throughput goes way down.  The second problem with



trying it out is that the interface alone is



$115,000, so you have to want to invest in it.
                        229

-------
                          MR. SHACKELFORD:  If Bill
had not moved us back into R&D, we'd probably have
that kind of money, but...
                          DR. ERICKSON:  I can
write you a proposal.
                          MR. TELLIARD:  Thank you,
gentlemen.  It's time for our school break.  There's
cookies and milk in the hallway.  We're running
late, as usual, so to get to the HMS Sinkfast,
would you please get your cookie and so forth and
get back in here in the allotted time?  Thank you.
(WHEREUPON, a break was taken.)
                        230

-------
                          MR. TELLIARD:  Our third
speaker, due to the departure of the HMS Sinkfast,
will open tomorrow morning's session and in so
doing, we'd like to start 15 minutes early which is
a quarter to nine for those of you who can tell time.
     One quick announcement.  The HMS Sinkfast will
be located at Waterside, which is down near Phillips,
and be ready to leave at about 5 o'clock, plus or
minus a couple of minutes.  We will then break
here.  It's definitely a dress-down affair.  The
issuance of life rafts and so forth will then take
place and you can then get on the boat.  The best
way to get to the boat would be to go downstairs
either through the lobby bar or around behind it
and out by the swimming pool and then just walk
down the side.  Captain John will be there.  He'll
be the derelict standing almost in front of the
boat.
     Now, to get our afternoon session going, I'd
like to have Samuel To open with his presentation.
                        231

-------
                   SAMUEL TO,

UNITED STATES ENVIRONMENTAL PROTECTION AGENCY
   OFFICE OF WATER ENFORCEMENT AND PERMITS
  PROGRESS REPORT  ON  DMR QA STUDIES: QUALITY
 ASSURANCE PROGRAM FOR NPDES SELF-MONITORING
       (Revised  presentation submitted.)
                      232

-------
              Progress Report on DMR QA Studies
  A Quality Assurance Program for NPDES Self-Monitoring Data

                          Samuel To

     Municipal and industrial wastewater treatment facilities are
regulated under the NPDES as mandated by the Clean Water Act.
Each direct discharger is usually regulated by a unique NPDES
permit, which specifies certain limits on pollutants in the discharge,
The permittees are required to routinely sample and analyze their
discharges and report the data in Discharge Monitoring Reports
(DMR),  The validity of the NPDES Program hinges on the quality of
these DMR's.

Description of the Program
                                            *
     Through EMSL-Cincinnati, the Office of Water Enforcement and
Permits has been conducting a Quality Assurance (QA) program to
assure the quality of NPDES self-monitoring data.  The program is
designed to evaluate the major NPDES permittee laboratories' ability
to analyze and report accurate NPDES self-monitoring data.

     In 1979, DMR QA pilot studies were conducted in two States.
Responses were good.  As a result, a national study was initiated
to include all 7500 major permittees.  Since 1980, four national
studies have been completed.

     Major permittees under NPDES are sent performance evaluation
samples containing constituents normally found in industrial and
municipal wastewaters.  The samples are then anlayzed using the
methods employed for reporting NPDES self-monitoring data.
Responding permittees subsequently receive an evaluation of their
data, and are given guidance for checking error sources, advice
for taking voluntary remedial action, and requests to communicate
corrective actions in writing to the State for EPA Regional QA
Coordinators.

Evaluation of Data Quality

     This program has provided valuable data in assessing the
quality of DMR's.   As illustrated in Figures 1 and 2, improvements
in the DMR QA data had been significant.

     Since permittees in each study were largely the same, we can
conclude that the  quality of NPDES self-monitoring data has improved.
Data analyses at the permittee level also made possible the
identification of  about 3000 permittees that have participated in
all four studies.   This group showed slightly higher success rates
than the general population.  (Figures 3 and 4)
                                      233

-------
                            - 2 -
     Identification of the Standard Industrial Classification (SIC)
code made possible the tracking of improvements by individual
industries.  Tables 1 and 2 summarizes the performance of major
industries (i.e., with over 10 permittees in Study 4).

Other Uses of the Data
     Besides measuring the quality of DMR data, this program also
encourage of proper OA procedures.  When permittees receive evaluation
reports with unacceptable data, they are asked to check for sources
for errors.  An examination of follow-up information indicated that
a substantial portions of errors were due to data management problems,
such as transcription, calculation, erroneous units, or misplaced
decimals.  In Study 2, almost half of the errors were of this type.
(Figure 5)

     Understanding the source of errors is the first step to improve-
ment.  Data management errors are relatively easy to correct compared
to analytical problems.  It is, however, a very important, yet often
neglected, part of a OA program.  Such errors can be minimized by
instituting proper data handling procedures.  A comparison of sources
of errors between (Figure 5) studies confirms this, as there was a
much smaller percentage of data management errors in Study 3.

     The DMR OA data were also useful in planning inspections and
directing other follow-up.  Since there are a large number of permittees,
this program enables EPA and States to concentrate corrective actions
on permittees with more needs.  This results in increase efficiency
of NP-DES compliance monitoring.
                                       234

-------
o
o
§    g
                aiavidaoov SSSAIVNV
                                                        CM   Q
                                                             I I I

                                                        -    &

                                    235

-------
O
CO
o
(O
    g


saaumaad
o
C9
                                             I
                                         K)
                                             S
                                             fe
                                             I
                                             CV2
                                   o    o
                      236

-------
3 O
2 "
o e
to r
3 O
CO
g
s
o
n
o
01
o
o
aiaviaaoov
                  237

-------
S33JJJWH3d
                232.

-------
           >-  uj in

           SS5
                 (O
               : uj
               i en
 a
l-
i- o <
in z z
SIC

ESCRIPTION
         u  a
         M  O
         (0  (J
                                                                    .  —  —  " .  •  .  i,   - "—.  -_-~  -— -- -—  —rf -^* —v  "-v «^j >^-  i-i f-i  r i  r"i r
-------
           «3- CO H o <
               i r~ o i
               • in o i
                                                                                                                                       >  co
                                                                                                                                       i  in
           «^ o* *~4 o i
           a> -< CM IT- •
           CM
           CJ O O tO I
            .
            CJ I
     • co CM i
     I CM &• i
                               IUT|OinCM«-*C7*O»-4»-4i-tOjrO3-cocvir^KiooooincO'-(-oo,cMotoror^•
                                                                                                                                         t-
                                    >    E E 3
                                    D; >-  UJ aj O
                                    Ul —I  I- t- UJ
                                    tn a.  to in tn
                                       Q.  V >-
                                    o 3  tn in -i
                                    M en       *t


                                    CJ UJ  <£ 3 M
   Kt (> -^ —I I
   O O CM IO •

-------
3
       a:
       O
       O
       O
       DC:
       e
CM
o
a:
2
Q
                                                                             CO
                                                                             g
1*4
O
CO

8

g
O
CO
                                                                            §
                                                                            CO
I
 I
                                                                            10


-------
                    PAUL BRITTON

   UNITED STATES ENVIRONMENTAL PROTECTION AGENCY
  ENVIRONMENTAL MONITORING AND SUPPORT LABORATORY
    PROGRESS REPORT ON DMR QA STUDIES:  QUALITY
  ASSURANCE PROGRAM FOR NPDES SELF-MONITORING DATA
                          MR. BRITTON:  During my

time here today, I want to cover two major topics.

First/ some special summary results that provide a

better basis for direct comparison from one study

to another; then some simulation results allowing

comparison of several statistical estimation pro-

cedures when applied to data with known character-

istics.  Limits in DMR QA studies have been and

will continue to be based upon a statistical

estimation procedure equal or similar to one of

these.

     On the study summaries, Sam To gave the official

study results and the percentage of acceptable data

for each study.  However, the basis for limits in

each one of these studies differ.  We wondered

whether going to a standardized basis for limits

and reevaluating the study results would lead to

some difference in the percentage of acceptable

data from study to study.  Certainly it would lead
                        242

-------
to percentages which could be more validly compared



from one study to another.



     For those who are not familiar with the basis



for our limits, they are developed from two different



kinds of information.  First, regressions based on



statistics from the six previous studies are de-



veloped directly from data generated by EPA and



state laboratories in these studies on samples



just like the ones in the DMR QA study.  Statisti-



cal estimates are then developed directly from the



study data; again, data produced by EPA and state



laboratories analyzing the DMR QA samples.



     By comparing limits developed directly from the



study statistics against limits developed from re-



gression-based statistical estimates, the limits to



be used for each study are selected.  In essence,



the two limits are compared and if the limits from



study statistics are completely enclosed within the



regression-based limits, then the regression-based



limits, the broader of the two, are used.  If in



any way, the regression-based limits fail to enclose



the study statistics-based limits, then the study



statistics-based limits are used directly, because



there's no question about the relevancy of those



limits  to the study samples.
                       243

-------
     Obviously/ from one DMR QA study to another,

the six most recent studies would have changed and

therefore, the regressions would have changed.  In

comparing and deciding on the broader of either the

regression-based limits or the study statistics-

based limits, there would also be a considerable

change from one study to another in how the limits

were developed.  Consequently, we decided to go

back and reevaluate all the study results using one

set of fixed regressions to establish all appro-

priate limits for each of the studies.  These

regressions are based upon the six most recent
                             .,!'•             i "
studies which are related to the DMR QA samples.

     Overhead 1 provides an example of one set of

fixed regressions.  It happens to be the set for

aluminum, the first analyte, but gives you some

idea of what the regressions are like.  Each set

includes relationships between the true concen-

tration (T) of a sample and the mean (X) arid

standard deviation (S) that can be expected for  •

analytical responses from multiple laboratories

operating within control.  The R2 values for

these two fits, R2 being the percent of the total

variability in the data that is explained by the

regression, are also given.  For the relationship

-------
between X and T, R2 = 99.7 which is very high,



showing that this relationship fits the data very



well.  Standard deviation being a little more



imprecise in nature, only 89.5 percent of the



total variability in the data were explained by



the regression, but that is still a solid linear



relationship and a reasonable basis for estimation.




     The columns of residuals in Overhead 1 show



how the estimates from the regressions relate to



the observed statistics, sample by sample, again




indicating that the standard deviation is more



variable and that there are a few individual sta-



tistical estimates which vary somewhat from that



line, but none which vary to any significant extent.



This suggests good relationships and is character-



istic of these regressions.




     There are 26 analytes and two concentrations



involved in each DMR QA study, so there were 26



sets of relationships like this which were applied



to two different concentrations for each study.



There was a total of four, studies to go back and



look at.



OVERHEAD 2




     Using limits for each study that were strictly



based upon the fixed regressions, we obtained the

-------
results that are given in Overhead 2 for the DMR QA



studies.  The second column shows the time the study



was actually run—basically an estimate of when the



analysis was done.  The next column shows the total



number of reported values received in each study/



which has been fairly consistent.  There were more



respondents in DMR QA Study 4 than in any of the



other studies—probably the main reason why the



total number of reported values was highest.



     In the fourth column are the percentages of



acceptable data using the special fixed regression



limits.  Surprisingly, they are not that much



different from the official percent of acceptable



data results, but the potential is certainly there



for a difference.  This apparently shows that the



regression relationships have been fairly consistent



for several years; not changing as much from one



year to another as they did in the early years.



The next column shows the change in percentages



using the first study as a base, and that in DMR



QA studies, the percentage of acceptable data has



increased by 10.4 percent between the first and



fourth studies.



     The last column attempts to show the percentage



of improvement in performance since the first study,
                       246

-------
showing the DMR QA participants have reduced the




amount of unacceptable data they reported by about



40 percent over these four studies.  The EPA and



state laboratory data from analyses of the same



samples was similarly analyzed to give some idea



of relative performance by these groups.  For EPA



laboratories, the percentage of data acceptable



was 94 percent in the first study related to the



DMR QA Study 1, and 96 percent for Study 4; improve-



ment not unlike the 40 percent seen for the DMR




QA studies.  For state laboratories, the percent-



age starts out at 86.2 and goes to 91.4; a



similar 37.7 percent improvement.



OVERHEAD 3




     Overhead 3 is a graphic representation of the



performance over the three and a half year period



by those three laboratory groups.  Again, it's



fairly obvious that all show improvement with time,



which is pretty much what was expected, and that



they are all improving at about the same relative



rate, which was somewhat surprising.



     The second topic I want to discuss has to do



with some simulations done on a series of different



statistical estimation procedures, any." of which



could potentially have been used as the basis for
                       247

-------
the background statistics on data from EPAand

state laboratories, that become the basis for all

limits.

OVERHEAD 4

     Overhead 4 attempts to briefly depict the major

difference between the four techniques.  If the X-
                   f ;   .  '  ;:"•   ;          i        , '••:•.'
axis is the order of the ranked data, then the first
                       ' .. .  , '•'"„•'    i   i   	     	 ,-, <':•

procedure, the traditional calculation without any

outlier removal, would allow every observation to

have an equal weight of one in the statistical cal-

culation.  Where you have traditional estimation

but do some outlier testing before you do the

calculation itself, some of the observations, the

extreme observations on either end, would have a

weight of zero in the calculation.  All other
                                   , ... '   , '!•• ' "i  , i' ' ' '
observations would have a weight of one.

     If you use a robust estimation procedure—a

procedure that does not use all of the observations

in a data set in order to develop estimates of the

mean and standard deviation of that set—the calcu-

lations are different, compensating for the fact

that not all of the data are used.

     The robust estimation procedure we are using

involves outlier testing first, then a robust

calculation on the middle 70 percent of the
                                        	i	?•	•!' ,: v
                        2*8

-------
retained data.  In other words, any observations




beyond the 15th and 85th percentiles of the retained



data have weight zero, while data within this inter-



val have full weight.  The influence of observations




transcending either percentile is prorated on the



portion falling within the interval.



     The fourth procedure is called the biweight



estimator.  It involves development of a weighting



factor for every observation such that more extreme



observations have less weight and, therefore, less



influence on the biweight estimates.



     We are currently investigating the effect of




these estimation techniques on data from studies



like ours and will probably continue to look for



and investigate alternative ways in the future.



This investigation is still in progress, so final




conclusions are not available on whether we will



change from the 70 percent robust procedure current-



ly in use, however, some early results are presented



in the following overheads.



OVERHEAD 5



     The first set of simulation results in Over-



head 5 involves a thousand samples taken from a



normal population with mean 100, standard deviation



10, and 20 observations in each sample.  The mean
                       249

-------
and standard deviation of the underlying population



were estimated from each of the thousand samples



using the four estimation techniques.



     The two columns under "EST. OF MEAN" show the



traditional arithmetic mean and estimate of var-



iability (the standard deviation) to give a general



basis for comparison of the 1000 estimates of the



population mean obtained using each of the four dif-



ferent estimation procedures.  In general, the only



difference among means produced by the four proce-



dures involved their variability; as expected, the



traditional estimates had slightly less variability.



Estimates from the traditional procedure should



always have the least variability when the popula-



tion is normal.  For practical purposes, you get



the same estimate by all these procedures.



     The last two columns characterize estimates



of the standard deviation developed using the four



procedures.  As before, traditional has less varia-



bility, but there is really little practical differ-



ence, except for an apparent 5 percent high bias in



the biweight estimates.  Origins of the bias in bi-



weight estimates of the standard deviation are un-



clear, but if used in its current form when the un-



derlying population is a normal, it would slightly
                        250

-------
overestimate the standard deviation and therefore




lead to slightly broader limits.



     Overhead 5 also contains results from 1,000



samples of size 50.  As you would expect with a



larger sample size/ the variability of estimates is



reduced for all procedures.  The only real differ-



ence seems to be the bias on the estimate of the




standard deviation from the biweight procedure;



as before, it's about five percent high.



OVERHEAD 6




     Okay, that shows what happens with normal dis-



tributions.  What happens when you do not have a




perfect underlying normal population?  Certainly,




we do not always see data from a perfect normal



distribution during our studies.  Overhead 6 shows



results for simulations involving 1,000 samples



again, but the data is a mixture of two underlying



populations.  For the first set of results, each



sample of 50 contained 40 observations from a nor-



mal population with mean 100 and standard deviation



10.  The other 10 observations in each sample came



from another normal, with mean 130 and standard



deviation 10.  So, 20 percent of each sample re-



flected another distribution with a 30 percent



higher mean.
                       251

-------
     We are really interested in estimating the



characteristics of the 80 percent of each sample



which represents the good data, and we want to



look past the bad data which is represented by



the data with a 30 percent bias. Therefore we



would prefer our estimation procedure to give mean



estimates of 100 and standard deviation estimates



of 10, which reflect the capability represented in



the good data here.



     The mean estimate from traditional calculation



techniques is 106.8 and is somewhat better with the



outlier testing.  Mean estimates from the two



robust procedures are slightly better yet,although



the difference is not dramatic.  The estimates of



the standard deviation are all similar; instead of



being 10, they are all over 15.  As a result,



where we should have estimates of the mean of 100



and estimates of the standard deviation of 10,



we've got estimates of the mean that run 105 or



better, and estimates of the standard deviation



that exceed 15.  Obviously, evaluation limits




developed from these estimates will be much broader



than they should be to represent the bulk of the



data.	



     The second set of results in Overhead 6 show
                       252

-------
the effect of slightly increasing the bias of the




20 percent bad data in our simulation.  The estimate



of the mean from the traditional calculation has



gone up to 112, whereas the biweight has already



recognized all the bad data and has dropped the



mean estimate down to 100.7.  The other two proce-



dures have both partially recognized the bad data,



but haven't successfully ignored all of it.  The



standard deviation for the traditional calculation



is over 26, for the next two is over 16, and for



the biweight is down to 13.  They should all be 10.



     In the last set of results in Overhead 6, the



bias of the bad data has been increased to 90



percent and the estimate of the mean from the tra-



ditional calculatipn has increased to 118.  The



other procedures have recognized virtually all 20



percent of the bad data and are estimating 100.



Obviously, there are some residual effects on the



mean estimates of the standard deviation, but by



far the worst is the 38 produced by the traditional



calculation as a mean estimate of the standard



deviation.



     A tremendous mixture of data comes in during



these studies.  As could be expected when anywhere



from 15 to 25 percent of the data are considered
                       253

-------
not acceptable, errors involving factors of 10, 100



or 1,000 are not unusual.  Perhaps five percent of



the data will reflect a decimal placement error of



this kind.  Factors of five, two and somewhat



lower biases are also quite common.  Although 20



percent of a particular kind of biased data is not



a frequent occurrence in our studies, these simu-



lations certainly demonstrate the effect of such



data when you are really interested in the charac-



teristics of the 80 percent that represent good



performance.



OVERHEAD 7



     We have seen what happens when the bias



changes, but what happens when the standard



deviation changes?  In Overhead 7, 80 percent of



each sample comes from a normal distribution with



mean 100 and standard deviation 10, while for the



first set of results, the remainder comes from a



normal with mean 100, but a standard deviation of



30.  The mean estimates, of course, are not influ-



enced one way or the other, but this shows how the



standard deviation estimates are affected.  The



estimate of the standard deviation from the tradi-



tional calculation is up to 16.2.  For the others



it is better, but still varies between 11.5 and
                       254

-------
12.8.            .    ,         •    -     •   .


     Of course, you can have an alternative distri-


bution which represents a very, very high quality


group of performers.  Under such circumstances, we


were concerned that statistical estimates might be


dramatically affected,  causing evaluation limits to


be much too narrow.  Therefore, we looked at the


effect when 20 percent of the data represented a


normal distribution with a much smaller standard


deviation of 3.3.  The mean of the standard devia-
           i

tion estimates from all procedures are between 8.5


and 9.1, which is reasonably close to 10, the


standard deviation of the general population repre-


senting average performance.


     We plan to make comparisons of these procedures


using real data.  The disadvantage of using real


study data, of course,  is that its true nature is


unknown.  Its advantage, however, is in showing how


much results from the estimation procedures would


differ in actual use.  We also plan to look at


simulations using other mixtures of data, perhaps


some more complex, but  the ultimate is going to be


how it performs on the  real study data.


     I'm not at all unhappy with results for the


70 percent robust procedure, the one we currently
                       255

-------
use to produce the statistical estimates used as a

                                           »i ' • ' •
basis for acceptance limits in DMR QA studies.  If


we do decide to change our statistical estimation


procedure, it seems unlikely there would be a


noticeable effect on limits in DMR QA studies.


     That's basically the end of my presentation.


I guess it's time for questions for Sam or I.
                        256

-------
REGRESSION EQUATIONS FOR  ALUMINUM
X - . 982 T * 43. 1




     - .997
            S - . 0527 T * 20. 4
            R*CS> - .
TRUE
CONC.

I960
1310
1260
938
930
662
607
503
340
98.
94.
43.
NUMBER
OF DBS.
REPORTED
46
46
45
37
43
42
43
46
35
1 38
6 35
0 32
MEAN
RECOVERY
CX>
1830
1293
1186
1003
925. 1
697.3
614.7
487.3
397.5
134. 3
145.4
82. 6
RESIDUAL
FOR EST.
OF X
- 3.6
-11.2
-69.7
27.9
-13.0
17. 1
-12.6
-39.9
27.2
- 3.2
11.3
- 1.8
STD.
DEV.

128.3
95.5
105.9
82.7
46.7
57.0
43.5
51.0
50.3
22.9
21.4
23.9
RESIDUAL
FOR EST.
OF S
9.90
6.07
19.87
11.26
-22.69
1.68
- 8.93
4.08
11.94
- 2.75
- 4.02
1.17
                    257

-------
PERFORMANCE IN DMR-QA
 AND RELATED STUDIES




STUDY TIME
DMR-1 10-11/80
DMR-2 5-6/82
DMR-3 4-5/83
DMR-4 4-5/84
FOR EPA LABSa
WP006 10/80
WP008 3/82
WP010 3/83
WP012 3/84
FOR STATE LABSi
WP006 10/80
VP008 3/82
WP010 3/83
WP012 3/84
TOTAL

kiA f*^
NO. OF
RESPONSES
35.056
33,208
33. 730
39. 324

515
590
582
595

3.312
2.952
2.960
2.872




ACCEPT.
74.3
78.0
83.9
84.7

94.0
96.6
96.0
96.0

86.2
90.3
91.3
91.4

:,!,-: I',* "'»• A1 .
4%A * A & ^rf^^W 4& • ^ A^^^M
CHANGE SINCE
FIRST STUDY
...
3.7
9.6
10.4
ii<
...
2.6
2.0
2.0

—
4.1
5.1
5.2



IMPROVEMEft
SINCE FIRST SI
-nttm-m
14.4
37.4
40.5
•.IK ii,
...
43.3
33.3
33.3

...
29.7
37.0
37.7
             258

-------
CO
CO

Q
LU
      CO

      t-t

      *-•
      _J

      UI
      CL
      UI
                                                                                       W)
                                                                                       U.
                                                                                       o

                                                                                       UJ
                                                                                v-C
i  i i  i  i  t i i  i  I  i  i	•
                                                 259

-------

                                   K
                                   o
                                   UJ
                                   y.
                                    0
260

-------
   SIMULATIONS  -  EACH INVOLVING  1000 SAMPLES
POP.
N
NC100,  10)  20
N<100.  10)  50
EST.
PROCED.
TRAD.
W/0. T.
70% R.
SIWEIGHT
TRAD.
W/0. T.
70% R.
BIWEIGHT
EST.
MEAN
100.1
100. 1
100. 1
100.1
100.0
100.0
99. 9
100.0
OF MEAN
VAR.
2.21
2.2)3
2.27
2.30
1.48
1.46
1.53
1.52
EST. OF
MEAN
10.0
10.0
10.0
10.5
10.1
9.9
10.0
10.5
STD. DE\
VAR.
1.66
1.73
2.08
1.95
1.03
1.09
1.37
1. 18
                            261

-------
   SIMULATIONS  - EACH  INVOLVING  1000 SAMPLES
POP.

NC100, 10)
NQ30. 10)
NCI 00. 10)
NC160, 10)
NCI 00, 10)
NC190. 10)

N
40
10


40
10


40
10


EST.
PROCED.
TRAD.
W/0. T.
70% R.
BIWEIGHT
TRAD.
W/0. T.
70% R.
BIWEIGHT
TRAD.
W/0. T.
70% R.
BIWEIGHT
EST. OF
MEAN
106.0
105.7
105.1
104.4
112.0
104.7
103.5
100.7
118.0
100. 1
100.1
100.0
MEAN
VAR.
1.40
1.54
1.65
1.71
1.44
4.64
4.27
2.25
1.40
1.80
1.76
1.59
EST. OF
(CAN
15.9
15.3
15.8
15.9
26.5
16. 8
16.3
12.9
38.1
10.0
10. 1
11.5
sm PEV,
VAFt
1. 30
1.53
2.27
1. 65
1.38
6.67
8.43
3.37
1.38
2.02
2. 24
1. 45
                              262

-------
    SIMULATIONS -  EACH INVOLVING 1000 SAMPLES
             N
NU00,  10>    40



NC100,  30>    10
NU00.  10)   40



NU00.  3U3>  10
 •^^^*1aV*^hk*^B^BV
 PROCEDL



  TRAD.



 W/Q.T.



 702 R.



BZWEIOHT





  TRAD.



 W/O.T.



 702 R.



SIWEIGHT
EST.  OF MEAN

MEAN    VAR.
                                              EST. OF S1BL QEV«
100.1



100.0



100.0



100.0





100.0



100.0



100.0



100.0
        2.30
        1.
        1*78





        1.30



        1.34



        1*38



        1,
MEAN
12.2
11*5
12.8
8.1
8.7
8.5
81,0
VAIL
1.81
i.m
1.72
8,88
1*11
1*38
1.18
                                 263

-------
            QUESTION AND ANSWER SESSION




                          MR. RICE:  Jim Rice.


Paul, all of those examples that you put up there


could just as well have represented two different


analytical methods, which is the case, for example,


for where there are alternative methods given say
                                     ,  I •'. 11 Jr  , ,    : , •,

for chromium or copper or whatever, one of them


flame and one of them furnace.  For instance,


these two procedures or methods have considerably


different precision statements when operating on


the same metal sample.  I mean, you could construct


them so that everything you said there would work


for that.  Why not recognize that?  Have you attemp-


ted to look and separate out and examine each method


by the appropriate statistics on the data base
                                       , 1" i £ I;",,::'',
separated into its components as you can identify


them.


                          MR. BRITTON:  Yes.  You


know that we did collect the information on the


method that was used when these studies were run,


and you also know that we do generate statistics by


method.  There would certainly be a policy  issue


involved in setting limits method-by-method, but I


really think that's the only alternative, assuming
                       264

-------
that indeed there is a significant mixture of



methods and significantly different statistical



characteristics for the various methods.  I think




that in general, we're not seeing that great a



mixture of methods, and in general, we're not seeing



tremendously different statistical results either.




     I think one major exception is the group of



nutrients where in the DMR QA studies, we observed



that the percentage of unacceptable data in the



nutrient group was characteristically about twice



the average unacceptable data for the other groups.



We started considering that issue and decided that



it was probably caused by the fact that two-thirds



of the EPA and state laboratories use automated



methods for nutrients.  We looked at the DMR QA



group, and they use manual methods about 2:1.  So



we decided we had to make an adjustment, and from



now on, in fact I think with Study 4, the limits



for nutrients are to be based on statistics from



all labs rather than the EPA and state laboratory



group.  This change has brought the percentage of



unacceptable data for nutrients in DMR QA studies



right down in line with all the other analyte



groups, so things are balanced better now than



they were.
                       265

-------
     In general, I don't anticipate looking at




methodology and somehow compensating for statistical



differences from one method to another.  We really



haven't had a chance to study the difference between



methods in a rigorous sense, and I'm not real sure



what I would do with the information.  EPA policy,



as I understand it, is that approved methods are



all equivalent methods and therefore should be



judged against the same general criteria.



                          MR. YOUNG:  Jim Young



from the University of Arkansas.  I'm a little



concerned that this approach might tend to cause



people to think that this is the test precision



that you are measuring, when in fact the precision



within a particular laboratory should be much better



than from the data you have collected and are



analyzing.  Is there any plan to look at intra-



laboratory precision versus interlaboratory



precision in this program?



                          MR. BRITTON:  No, there



isn't any plan to look at within-lab standard devia-



tion in this program.  That would require partici-



pants to do replicate analyses on the same sample,



or analyze two samples that were identical or nearly



so.  But to do that, we would probably have to give
                        266

-------
up having samples at two concentrations,  in order



to keep the analytical effort from increasing.  I



believe having information at two concentrations is



more helpful for identifying where potential problems



may be.  If there appears to be similar bias at both



concentration levels, follow-up personnel can look



for a systematic error.  If there are dramatically




different errors at the two concentrations, then



some kind of a precision problem is more  likely.



In general, we have not tried to get into within-



laboratory standard deviation estimates in these



studies.  There are other sources for those esti-



mates.




                          MR. STANKO:  It would be



difficult for me to look at the DMR QA program and



criticize it on the fact that it has improved the



quality of the data.  It certainly has improved the



quality of the data for the NPDES permit  labora-



tories that are generating these data.




     Last year at this conference, I gave a paper



that was titled, Industry's Experience with the



DMR QA program.  The very specific point  that I



made last year was that the criteria acceptance



limits identifies too many industrial and contract



laboratories as bad performance.  I would like to
                        267

-------
show one slide for DMR QA Study 4, if I might.


SLIDE 1


     This slide takes the EPA laboratory data for


DMR QA Study 4, for the 52 parameters.  The second


column is the true value, the third column is the


acceptance criteria limit that was used oh industry


for Study 4.  The next column in identifies the


number of federal and state laboratories that are


using the program.  The column after that is the


number of outliers that were identified by the EPA


statistical procedures.  The next column identifies


the additional number of EPA laboratoriesthat fell


outside of the 99 percent criteria.  The last


column identifies the percent of EPA laboratories


that fall out of the 99 percent confidence interval.


Statistics dictate that only approximately one


percent of the laboratories should fall out of the


99 percent confidence limits.


     If you will look at some of the bottom values,


for total kjeldahlnitrogen, there was one identified


outlier out of 53 observations, there were seven
                               ••  ' ' ''/'I .- " '•.,• -li'i1' •	.';,.' • • •; ;;,-' .•<, "•

additional EPA laboratories that fell out of EPA


criteria, for a total of 13.2 percent of EPA's own


laboratories falling out of the 99 percent con-


fidence interval.  If I work my way up that column,
                        268

-------
and what this column should be is 1 percent, we



get 6.8, 9.1, 13.2, 9.4, 10.6, 6.1, 8.1, 8.1 and



so on.



     There is something wrong with the statistical



procedures.  I really don't care if you Winsorize,



mesmerize, or hypnotize, but when you apply your



criteria limit to your data set, only one percent




of that data should be excluded.  Thank you.



                          MR. BRITTON:  You're



right, George, something is wrong.  The problem is



your assumption that only one percent of the EPA



and state data should fail our acceptance limits.



We know that some of these data are not good and



our limits are supposed to pass 99 percent of all



good data.  We would like all the bad data to



fail, and to make this more likely, we are willing



to pay by accepting rejection of one percent of



the good data as well.  Clearly, EPA and state



laboratories are not perfect yet.  They did generate



some bad data in these studies.  If they hadn't,



then the failure rates you calculated would indeed



have been one percent.
                       269

-------
Revised Presentation
          270

-------
         STATISTICAL BASIS FOR  LABORATORY PERFORMANCE EVALUATION LIMITS

                    Paul  W.  Britton *  and Daniel F. Lewis **
1.  I PRODUCTION

The effectiveness of the water  regulations enforced by the U.S.  Environmental
Protection Agency (USEPA) depends upon the quality of data generated  by USEPA,
state, local government, industrial, commercial and other non-USEPA
laboratories.  In an effort  to  improve the quality of data, USEPA conducts
collaborative studies evaluating the ability of laboratories to analyze water
samples and to produce data  within specific evaluation limits.   At  present,
five formal studies  are conducted each year for USEPA by the Environmental
Monitoring and Support Laboratory at Cincinnati, two for the drinking water
laboratory certification program, two for point and non-point source  discharge
monitoring, and one  for major NPDES permit dischargers.  In addition, special
performance evaluation studies  are conducted for the Suparfund and  Solid Waste
activities, and other USEPA  programs involving contract laboratories.

Acceptable performance during these studies demonstrates a well-managed
laboratory operating competently, while unacceptable performance indicates a
laboratory which is  likely to be having problems generating quality data,
either for a particular analyte or in general.  Unacceptable performance
results in an investigation  of  the circumstances, and if appropriate, remedial
action by USEPA.   It should  be  mentioned, however, that even the best of
laboratories will periodically  develop analytical  problems.  It is  perfectly
reasonable for a  quality laboratory to occasionally produce analytical  data
that are outside of  evaluation  limits, due either to an unfortunate result of
randan chance or to  an actual analytical error.

During these performance evaluation (PE) studies, the Qua! ity Assurance Branch
of the Environmental Monitoring and Support Laboratory - Cincinnati sends
participating laboratories a set of stable sample concentrates  in sealed glass
ampules, a data reporting form  and appropriate instructions.   Each  laboratory
produces the study samples by diluting a measured quantity of specific
concentrates to volume with  reagent water, then analyzes them using its
routine procedures.   The completed form is sent to USEPA for evaluation and a
fully-detailed report is returned to each laboratory.   The responsible state
or USEPA office follows up with laboratories that demonstrate potential
pro bl ems .

The effectiveness of these studies depends upon two things:   first, whether
the evaluation limits  properly  indicate data resulting from substandard
analytical work;  and second,  the commitment of all  participants  to  ensure that
laboratories performing poorly  receive appropriate follow-up work so  that
substandard laboratories  either improve their performance or are excluded from
producing further data.  It should be noted that these studies serve only as an
indicator of potential  substandard ability and should not be  used as  the sole
basis for  a final  decision as to the quality of a laboratory.
**
USEPA Quality  Assurance Branch,
  Laboratory- Cincinnati
Computer Sciences  Corporation
                                  Environmental Monitoring  and  Support
                                                          Draft  1/11/85
                                           271

-------
2.   DEVELOPMENT OF EVALUATION LIMITS

EPA's objective in developing evaluation  limits  is  to  be  able to distinguish
between data from analytical  systems  operating within  state-of-the-art
capabilities and data representing substandard performance.  A statistical
prediction interval  at an appropriate significance  level  is clearly suitable
if acceptable data have a normal  distribution and if proper estimates of the
arithmetic mean and standard  deviation are  available.   (Note that a prediction
interval is similar to a confidence interval except that  where a confidence
interval uses an assumed mean, a  prediction interval uses an estimated mean.
A prediction interval, therefore, is  calculated  the same  as a confidence
interval except that it requires  an estimate of  the standard deviation that
has been adjusted to account  for  the  variabil ity of this  estimated mean.)
Since there are no independent sources of such information, it is necessary to
investigate the data from the past or present studies  themselves to establish
a basis for the desired limits.  Since even the  best of data sets will contain
an unpredictable and often substantial portion resulting  from substandard
analyses, it is necessary to  use  outlier  tests to exclude extreme data points
from the statistical process  before producing estimates of the mean and
standard deviation that will  be robust to the effects  of  any outliers that may
remain.

2.1 Estimating the Study Statistics Using Data From the Current Study

For most analytes in a sample, basic  descriptive statistics are estimated from
the data submitted by the participating USEPA and state laboratories, which
usually number over 100.  The statistics  are estimated from these data because
USEPA is more familiar with these laboratories and  is  confident that they will
produce results generally representative  of the  current analytical procedures
when properly applied.  The statistical report for  each analyte includes the
following:

1.  Identification of obvious outliers in the data  set;

2.  Traditional and robust estimates  of the mean and standard deviation from
    the retained data;

3.  Tests of the normal ity for the retained data;

4.  A histogram of the retained data;  and

5.  A complete listing of the USEPA and state data  with the obvious outliers
    noted.
                                            272

-------
 Outliers  are identified by a two-stage procedure.  The first stage involves
 the  rejection of values larger than 500 percent or less than 20 percent  of the
 true value  and is intended to remove all  data resulting from a decimal
 placement error.  Data retained through this first stage are used to calculate
 traditional estimates of the mean and standard deviation for the second  stage
 tests-.   If  there are 25 or fewer values left, the second stage involves
 successive  Grubbs1 tests at the two-tailed 1 percent significance level  until
 a suspected value is retained.  As described in Grubbs (1950 and 1969) and
 Standard Practice E-178 of the American Society for Testing and Materials,
 this test involves ordering the data, identifying the largest or smallest
 value as the suspected X — the value farthest from the mean of all  the
 data --  and calculating a statistic T = (X - X)/S, where X is the traditional
 estimate of the mean and S is the traditional estimate of the standard
 deviation.  The suspected value is discarded as an outlier if the absolute
 value of T  exceeds the appropriate critical  value, then X and S are
 recalculated from the remaining data prior to testing the next most  extreme
 vatlue.  For sets of more than 25 retained values, any value beyond a 99
 percent confidence interval,
 out! i er.
                      generated using the Student's  t,  is  considered
an
Final estimates of the mean and standard deviation are calculated from the n
values retained through both stages of outlier testing.   The  traditional
estimates are:
 (1)   X = ( I  X.)/n
(2)
      S =
         /
        V
where X^
                is the ith retained  value.
Robust estimates are calculated from  an  ordered set of the retained data,
using a 15 percent Winsorizing procedure to  estimate the mean as described by
Dixon and Massey (1969),  and  standard normal deviates to estimate the standard
deviation.
                                           273

-------
After the retained data have been arranged in ascending order,  the  15th  and
85th percentile are calculated as:


(3)   PIB   -Xj  - (3 - .15n- .5)(Xj  - XM)

      where j = 1 + (the greatest integer portion of (.15n + .5)),  and
(4)   P85   = Xk  - (k - .85n - .5)(Xk  - X^)

      where k = 1 + (the greatest integer portion of (.85n + .5)).

After the percentiles have been calculated, the Winsorized estimate of the
mean is calculated as:
(5)
      h(X )  + (n+2-k)(X    )
         h            k-1
                                      k-2
                                       Z    X
                                     i=h+l   i
                          n

      where h ~ j - (the greatest integer portion of (j - .15n - .5)).

      Note that Xn is the smallest observation that is greater than or equal

      to PIS, and Xk_i is the largest observation that is less than or

      equal to
The  robust estimate of the standard deviation is calculated from the
distance across the middle 70 percent of the ordered retained data.
Assuming an  underlying normal distribution, this calculation is:
 (6)
    P85 "  P15

S =   2.0729


where 2.0729 is the number of standard deviations between the
15th and 85th percentiles of a normally distributed population,
This value is obtained using an inverse normal  function.
                                             274

-------
 In order to produce  prediction  intervals, this estimate for standard deviation
 must be adjusted for the  variability in the estimate of the mean, as follows;
 (7)
'adj
• /s2
V
= s
* 4
/__—
S2/n
        where Sadj.  is the adjusted estimate of standard deviation to be used
        to  construct a prediction interval, S is estimated using equation 6, and
        Sy  is an estimate of the standard deviation of the mean and is equal to

        the square  root of S /n for means calculated from n observations.



To  provide information regarding normality, Kolmogorov-Smirnov (K-S) and
Anderson-Darling (A-D) statistics are calculated for the full  set of retained
data  and a K-S statistic is calculated for the middle 70 percent of the
retained data.  The K-S statistic is used because it is a highly regarded
distance statistic that can be used to test normality of the middle 70 percent
of  the  retained data as well as the full set.  The A-D statistic is calculated
for the full set of retained data because it is more sensitive to deviations
from  normality in the distribution tails, and was found to be  one of the most '
powerful and easily calculated distribution tests studied by Stevens (1974)
and Green  and Hegazy (1976).  Suitable critical  values are provided -at 1 and 5
percent significance levels using the approximations provided  by Stevens
For analytes with rugged analytical procedures or with which the laboratories
have considerable experience, there is no practical  difference  between  the
traditional and robust estimates of the mean and standard deviation.  However
there are analytes for which a nunber of the USEPA and state laboratories
still do not have their system fully under control,  and so are  generating data
that do not represent competent performance.  Under  such conditions, the
robust estimates are clearly superior to the traditional  estimates  since the
robust estimates are not as likely to be influenced  by persistent outliers and
the outlier scheme will fail to identify all the outliers whenever  analytical
problems become too common.  Because they are universally appropriate,  the
robust estimates of the mean and standard deviation  are used to characterize
acceptable performance.

The remainder of the statistical report is a histogram and an ordered listing
of all the data from USEPA and state laboratories  with outliers  followed by an
asterisk.  The histogram provides a convenient visual  impression of how the
retained data are distributed and the ordered listing  is  necessary to document
the actual data that the statistics were developed from.   Figure 1 i s an
example of the report.
                                            275

-------
2.2  Estimating Study Statistics Using Regressions on Previous Study Statistics

When appropriate background statistics are available, estimates of the mean
and standard deviation for the current study can  be made from linear
regressions generated from historical  data.   If these historically based
estimates are universally beneficial  to the  participants,  and they appear to
be reasonable, they will  be used as  a basis  for evaluation limits instead of
the estimates generated from the current study.


These linear regressions, for X~ versus true  value (TV) and S versus TV, are
achieved by using the customary least-squares algorithm to fit:
(8)

(9)
           :=F£)    and
w  •
                        d
After the coefficients have been estimated,  these equations may be simplified
by multiplying through by TV, giving the desired forms:
(10)

(11)
 X  = a(TV) + b    and

 S  = c(TV) + d
(Note that S is an estimate of standard deviation before it  has  been  adjusted
for the variability in the estimate of the mean.   Which  means, of  course, that
the historical data used to generate the regression  equation for S will  be
unadjusted standard deviations from past studies.)

Using this regression technique leads to a minimization  of the sum of the
squares of the residuals as a percentage of the true value associated with
each residual.

The chosen regression procedure was selected after a comparison  of results
with traditional linear, quadratic, cubic and power  curve fits to  the
historical statistics.  The nonlinear alternatives were  rejected because they
were more complex and did not generally produce regressions  that fit the data
better.  The traditional linear least-squares alternative, i.e., directly
fitting a linear relationship between X or S and  TV, was rejected  because the
statistics for samples with high true values would tend  to dominate the
regressions, causing regressions which might provide misleading  estimates of
the mean and standard deviation for samples with  low true values.  The
propriety of the chosen regression procedure is also consistent  with
theoretical expectations since yariabil ity in results tends  to increase  for
most analytes in direct proportion to increasing  concentration,and this
regression model was designed to deal with that problem. Currently, residual
randomness tests are performed and coefficients of determination are
calculated for every regression equation as it is produced.  In  this way the
regressions are continually monitored to assure they are appropriate  and
effecti ve.
                                           276

-------
 To produce the prediction intervals defined in the introduction to Section 2
 the estimates of the standard deviation must be adjusted to account for the
 variability of the estimate of the mean.  This is done as follows:
 (12)
'adj
s
SX
        where Sadj.  is the adjusted estimate of standard deviation to be used
        to construct a prediction interval, S is the regression estimate
        of standard deviation  from equation 9, and % is the standard
        deviation of the  estimate of  X  obtained fromAequation 8.

 2.3  Determining the Evaluation Limits

 The first step  in  determining evaluation  limits for an analyte is to calculate
 trial  95  percent confidence limits using  the estimates from the  current study
 as  well as the  historical  estimates, both after the stated  adjustments.
 Wherever  the limits from the  historical estimates do  not  completely encompass
 those  from the  current study  estimates, the current estimates  are used  as  a
 basis  for calculating the  final  evaluation limits.  Otherwise, the historical
 estimates are used as a  basis for final limits.   In recent  studies  limits
 from current and historical statistics are generally  quite  similar,  however
 limits based upon  the historical  statistics usually predominate.

 There  are several  reasons  why the estimates  produced  from regressions on
 historical  statistics might be inappropriate.  The true value for  an analyte
 may  be outside  of  the concentration range  from which  the  regression  was
 developed,  the  true value  for the analyte  may  be  in error,  there  may have  been
 a significant change in  the way  the current  sample  was  made, or one  of  the
 regressions may not represent the true relationship between the true value  and
 the  estimates of a  statistic  generated from  past  studies.  Where  there  are  no
 historically based estimates  available for  use in setting limits, either
 because the  samples for previous  studies were made  differently or did not
 include particular analytes,  limits can only be calculated using estimates
made from  the data of the current study.

To insure  that these or other problems do not find their way into actual
 laboratory evaluations, the final limits  as well as the estimates and
procedures that lead up to them are inspected manually before any laboratories
are evaluated.  Suspected problems are thoroughly investigated.
                                           277

-------
3.  PERFORMANCE EVALUATION  REPORTS

Once appropriate standard deviations  and mean recoveries are available,  it is
a simple matter to  develop  the  95 and 99 percent prediction limits (explained
in the introduction to Section  2) to  be used as evaluation limits, and proceed
with evaluation of  the study data.  An individual report is generated for each
participating laboratory showing the  data reported, the related true values
and limits, and an  evaluation judgment for each reported value based upon its
relationship to the appropriate limits.  Each laboratory receives a copy of
its report and a copy is sent to the  responsible USEPA or state office for
follow-up contact,  as necessary, to resolve unacceptable performance.

4.  DISCUSSION AND  CONCLUSIONS

To date, USEPA has  successfully completed 30 studies involving a total of 108
water analytes and  thousands of participating laboratories.  The results of
these studies have  been well received by the participating laboratories, the
USEPA regi-orial and  participating  program offices, and the states which depend
on these reports to highlight laboratories requiring priority attention.  When
states have a parallel interest in  an environmental regulation, they often
request that their laboratories be  included in the USEPA studies rather than
developing a system of their own.
                                            278

-------
             fM O
             » O
             •O *•
                                                                                                                                    ooooooooooao
                                                                                                                                    oaoooooooooo
      >- 0 —O
      Z 0 ^ »-
      t=l O «0 -O
                                                                                                                                                                                     Q.

                                                                                                                                                                                     (C
      O, rv)
      a.
      3
                                                                       •  •
                                                                      «> u,
                                                                                                                                    ooooooooooooooa
                                                                                                                                    aooooaoooooaooo
         fv M* ;A
         o e> »*>,

UJ    Q) c1* ^>» f^

7    V3  O  0  9

<       O» 0* f*i
c
o
         o a o
         •o o o
                                    0

                                      •
                                    it.
                                                                                                                                  O
                                                                                                                                  R3
                                                                                                                                 UJ
                                          <
                                 •O O    »-
                                 o-o    <
                                                                      -I  -I
 ar oc    o       ee ac
 O O             00
 Z Z    «•.       Z Z
 I  I     O       II
 z z             z ±
 o o    _j       o o
.z z    <  ,     z z
                                                   Z Z    Z
                                                   >— »•    o
                                                  . o 
                                          •e    o
                                          o    «c
                                          o    o
                                          K    M
                                          ^
                                          (0    UJ
                                                                              z
                                                                              o
                                                            l»    O
                                                            z    c
                                                                                                            279

-------
                          REFERENCES

Grubbs, Frank E.,  1950.   Sanple Criteria for Testing Outlying
Observations, Annals  of  Mathematical Statistics, Vol. 21, pp. 27-55.

Grubbs, Frank E.,  1969.   Procedures for Detecting Outlying
Observations in Samples, Technometrics, Vol. 11, No. 1, Feb.,
pp. 1-21.

American Society for  Testing and Materials  (ASTM), Subcommittee
£-11.  E178-Recommended  Practice for Dealing with Outlying
Observations, Annual  Book of ASTM Standards, Part 41.

Dixon, W. J. and F.  J. Massey, Jr., 1969.   Introduction to
Statistical Analysis, Third Ed. McGraw-Hill Book Co., New York, NY,
pp. 330-331.

Stevens, M. A., 1974.  EOF Statistics  for Goodness of Fit and Some
Comparisons, Journal  of  the American Statistical Association, Vol.
69, No. 347, Sept.,  pp.- 730-737.

Green, J. R. and Y.  A.  S. Hegazy, 1976.  Powerful Modified EOF
Goodness-of-Fit Tests,  Journal of the  American  Statistical
Association, Vol.  71, No. 353, March,  pp. 204-209.
                                   280

-------
                          MR. TELLIARD:  Thank you,



Paul.  Our next speaker is George Stanko.  George



is going to talk about somebody else's methods for



a change.
                        281

-------
                  GEORGE H. STANKO

             SHELL DEVELOPMENT COMPANY
 INTER- AND INTRALABORATORY ASSESSMENT OF SELECTED
    SW-846 METHODS FOR ANALYSIS OF APPENDIX VIII
              COMPOUNDS IN GROUNDWATER
                          MR. STANKO:  I'd like to

share with you the results of a Chemical

Manufacturers Association study that evaluted some

selected SW-846 methods for the analysis of Appendix

VIII compounds in groundwater.  I'd also like to

point out that this is really a Chemical Manufactur-

ers Association paper or project, and that I am co-

author with Peter Fortini of the American Cyanamid

Company.

     On October 1st, 1984, the EPA proposed in the

Federal Register to make the use of SW-846 methods

mandatory under Subtitle C, 40 C.F.R., Parts 260

through 271.  In 1983, they had published the second

edition of SW-846.  In May of 1983, the Chemical

Manufacturers Association hired a contract labora-

tory, Environmental Testing and Certification

Corporation, to do an evaluation of the effective-

ness of SW-846 as a methods manual.  The American

Petroleum Institute also hired the Radian
                       282

-------
Corporation to conduct a similar study on the same




document.  Both of these reports revealed that SW-



846 was not adequate to guide an analytical lab due



to the lack of sufficient information and details,



technical inaccuracies and inconsistencies, and



also, pointed to numerous problems with the com-




pounds on the Appendix VIII list.




     To verify the findings of the CMA study on the



second edition, a project was developed in which



three prominent laboratories participated.  The



prime objective of the project was an assessment of



the inter- and intralaboratory precision and bias



of some selected methods out of SW-846.  A second



objective was to determine how well the resulting



data from this study would in fact define the



contamination problem.



     The study was designed to look at a simulated



situation, a real environmental situation.  We have



an abandoned waste site.  Up gradient from this



waste site is a service station that's been closed.



Several years earlier there was a known leak of



product, gasoline, and it would be reasonable to



expect that it made its way into the groundwater



and compounds such as benzene, isobutyl alcohol,



toluene, might be expected to be found in the
                       283

-------
groundwater, up gradient of the dump site.



     Also up gradient from this dump site is a



subdivision that uses septic tanks.  They have been



using liquid drain cleaners, 1,1,1-trichloroethane



primarily, and one characteristic of this simulated



environmental situation is that all the organics



that went into this dump site are absolutely known.



It is reasonable to expect after 25 years that all



of the drums would have been leaking and the



compounds we know we put there probably showed up




in the water table.  For a person who would have



to assess this environmental situation, it would



be reasonable to put one up gradient well and



three down gradient wells, collect water samples



and analyze these water samples.



     CMA environmental monitoring task group pre-



pared a list of some 34 compounds from the Appendix



VIII list, and also prepared a list of distribution



across these wells in this environmental situation



I just described.  CMA sent requests for proposals



to four prominent laboratories and each of the



laboratories were asked to maintain the integrity of



the list and the samples.  CMA task group suggested



some nine methods that were listed in SW-846 as



possible candidates to define this environmental

-------
situation.  Three bids were  received   from the  four



laboratories and all three of those laboratories



concurred  that method 8240,  8270, which  are GC/MS



methods, and 8330, which  is  an HPLC method, would



be suitable to identify all  the compounds  on  the



list.  This was not surprising since the list was




prepared, more or less, with that in mind.  Many of




us wanted  to test the best procedures  we thought



were  in SW-846 and we arbitrarily picked one  HPLC



procedure  that no one had any experience with HPLC



for the Appendix VIII list type compounds.



      CMA accepted proposals  from all three  of these



laboratories and they selected one of  these



laboratories to prepare the  simulated  environmental



situation  for analysis by all laboratories.   The



list  that we originally sent out of the RFP was




altered—some of it.  Also,  the concentrations  that



were  initially listed had been changed a little



bit.'  Each lab was advised to use the  three methods




that  they themselves indicated would be suitable,



and each laboratory was also advised that  there



were  some Appendix VIII compounds not listed in the



two methods, but amenable to the procedures,



according to the EPA.  Three laboratories analyzed



these samples independently  and submitted the data
                       285

-------
to CMA.  One of the laboratories was selected to


collect the data, compile it and do statistical


analysis.


     The actual water used for this sample was a


groundwater sample collected in the coastal plain


region of Texas.  It was provided by one of the


member companies of the CMA environmental monitoring


task group.


SLIDE 1


     This particular slide identifies the physical


and chemical properties of that groundwater.  We


also analyzed a portion of this water to verify


that none of the organics of interest for this


study was indeed present in the sample.  It turns


out that the water in this well had no organics.


     The study had four samples, and because we
                                          • ii ,

were interested in inter- and intralaboratory


precision, bias, we elected to run these samples as


blind triplicates.  We used a random number generator,


my wife.  We assigned a number to each one of the


four samples, and each of the three laboratories


received 12 samples with 12 different numbers.  The


only identification on the sample was a number.
                                   < . .  •  •  f   "

     The samples that were prepared for this study


followed procedures similar to those described in
                        286

-------
Section 7, "Calibration" of Method 624.  Sample 1



was a background and contained only four volatile



organics; benzene, toluene, 1,1,1-trichloroethane—



and the slide will tell us what the next one is.



SLIDE 2




     Isobutyl alcohol.  The three other samples were



prepared to actually simulate what might be expected




in the down gradient wells.  The slide shows the



compounds that were used for the volatile list.



Sample 1 contained four volatiles in a concentration



that varied from 25 parts per billion to 50 parts per



billion.  Sample 2 contained approximately half of



the volatiles.  Sample 3 contained the other half



of the volatiles, and Sample 4 was the one that was



loaded up and contained each one of the volatile



organic compounds.



SLIDE 3




     This is a list of the semi-volatile compounds



or the compounds that were amenable to Method 8270.



As you can see, Sample 1 was the background.  It



only contained volatile materials.  Sample 2




contained the lowest concentration and about half



of the semi-volatile compounds.  Sample 3 contained



about the other half, and and here again, Sample 4



contained all of the compounds.  The concentrations
                        287

-------
were highest in Sample 4.


SLIDE 4


     This slide identifies the three compounds that

                                       1  jpf   -,    ;L 	h
were taken off of the 8330 list, more or less amen-
                                         ' !'!' :  i .

able to this procedure.  Sample 1, again, did not


contain any.  Samples 2, 3 and 4 contained various


concentrations ranging from 100 Mg/L to 500 Mg/L.


     The data from the study was statistically


analyzed to assess precision and bias.  The assess-


ment of bias procedure was taken from the quality


assurance section of SW-846, Methods 8240, 8270.


In this procedure, one calculates the percent


recovery of spiked compounds; you calculate the


standard deviation of the percent recovery, assuming


a normal distribution; and you calculate the percent


relative standard deviation.  Mean percent recovery


and percent relative standard deviation define bias


and precision as stated by the EPA for a given


compound.


     SW-846 Method 8240 and 8270 also have the


performance criteria to be met in Section 8.2.4.


The mean recovery must be greater than 20 percent


for all compounds to be measured, greater than 60


percent for all surrogate compounds, and the percent


relative standard deviation must be less than 20
                       288

-------
percent for all compounds to be measured, including



the surrogates.




     Section 8.1.1 of the quality assurance section



states, before performing any analyses, the analyst



must demonstrate ability to generate acceptable



accuracy and precision with this method.  Ability is



established in Section 8.2 that I just described.



Calculations from the study were made for each



of the laboratories and each of the compounds.



SLIDE 5




     This is a summary table that was prepared



because you wouldn't be able to read all the small



numbers.  What this does is an assessment of how



well these prominent laboratories met EPA criteria



stated in Method 8240 and 8270.  For Method 8240,



there were 17 compounds.  Laboratory A did not meet



them for 29 percent, B for 29 percent.  Laboratory C



did not meet them for 94 percent of the compounds.



For 8270—31, 69 and 81.  For Method 8330, there were



three compounds studied.  The laboratories, all three,



100 percent, did not meet the criteria.



     It's difficult to explain why such a high




percentage of the data did not meet the EPA criteria



since these were experienced labs, experienced



operators and useed very good equipment.  I'd also
                        289

-------
like to point out that there is no guidance listed in

either one of the analytical procedures to tell you

what to do when you do not meet the criteria.

     An assessment of precision that we selected to

use was different from the EPA.  We did not

use percent relative standard deviation.  We used

lognormal distribution theory to assess the precision.

As a measure of a precision, we selected to use

variability factors and repeatability factors.

These define 95 percent confidence intervals.  This

defines variability factor, Vu, and as an example,

if you have a variability factor of 1.68 for an organic

compound and the true or the mean concentration is
                                           j • ',
known to be 100 parts per billion, the 95 percent

confidence interval of all observations should fall
                                          , il.lf;''',. i',
between 60 parts per billion and 168 parts per

billion, using these calculations.

     When you don't know the true value or the mean

value, and you have a single observation, and you're

trying to define the distribution of a second

observation, you use a repeatability factor.  Here

again, I use the same example.  The range is expanded;

208 parts per billion, 48 parts per billion.  That

describes, more or less, how variability and

repeatability factors are used to define precision
                        290

-------
of the method.




     In our statistical analysis of the data, we



considered looking at the nondetected observation—



we put the stuff in the samples and it should have



had a value—and we elected not to include non-



detects in our data.  It would have made things



look a lot worse if we had.




     There were two unusual values for diphenylamine



on different samples by one of the laboratories,




and a single value for methyl ethyl ketone by a



different laboratory.  We excluded three outliers



out of somewhere between 500 to 600 observations.



I don't think we overdid it.  The standard deviations



from the triplicate analyses were pooled over the



laboratories—samples, compounds and specific basis.



     I have a number of copies available of this paper



and there is a Table 9 in there.  If you're interested



in finding out for each one of the compounds, it's



in there.  I will only show a summary of the intra-



laboratory data on this slide.



SLIDE 6



     This is a summary based on the Method 8240,



8270, for the three particular laboratories.  You



can look at the Vu values or the Ru, whatever you



care to, the number of degrees of freedom are
                       291

-------
listed.  You can see that for the different com-
                                     (•  i  '; ILL,' •' ', "  %  •'   i i

pounds, the laboratories did not perform equally


as well.  For 8240, Laboratory B had the smallest


Vu value or variability factor, which means it has


the best precision.  For the 8270 methodology,


Laboratory A had better precision.   If you want to


look at 8330, one laboratory didn't  see anything,


and I would really disregard all the results, the


Vu and Ru; there's just not enough values  there.


     We also took the data and handled it on the


inter-laboratory basis.  In other words, we pooled


all the data and these are the results.  Here again,


in a copy of the paper you can look  at individual


compounds on a laboratory basis.  This is a summary


of values in the paper itself.  You  can see for


Method 8240, the variability factor  is 2.31; for


8270, 2.40, exclusive of a couple of outliers—and


there's not a whole lot of difference between the


precision of these two methods on an average.   I


would like to direct your attention  to the Ru


range.  For an Ru value of 3.27, the actual range


of Ru values is 1.70 to 9.01.  It is very sensitive


to laboratories as well as compound  specific.


     There were other problems with the SN-846


methods.  There were 11 compounds which one or
                       292

-------
more  laboratories  failed  to  detect  in  any  of  the



samples.




      This  brings up my  favorite  topic—false



positive,  false negative  observations.   These are



two important properties  to  actually describe the



qualitative aspects of  an analytical method.  The



CMA study  was designed  to allow  a very good assess-



ment  of false positive  and false negative  observa-



tions.  We knew what compounds we had  spiked  into



water.  We looked  at the  data.



SLIDE 7




      These are the false  positive,  false negative



observations for Method 8240 volatile compounds.



For Sample 1, we had seven false positive  observa-



tions by two laboratories.   It's interesting, if you



look  at the actual raw  data, which  is also included



with  the report, that the total concentrations of



false positives exceeded what we put in the sample



initially.




      The false negative observations from  the list



of compounds that we used for the study, are listed



on this slide.   Three laboratories could not find



isobutyl alcohol at a concentration range between



50 and 250 mg/L in all 12 of the samples.  That's a



false  negative.   Three laboratories also could not
                       293

-------
detect the presence of 1,4-dioxane in the six



samples that it had been spiked in.  The concentra-



tion level here was 200 to 400 parts per billion.



Here again, I'm not going to go through all these,



but of the 17 compounds that were spiked into two



or more samples, there were seven cases of false



negatives.



SLIDE 8



     This  is the tally sheet of false positives and



false negatives for Method 8270.  There are eight



false positives by one or more of the laboratories.



There was  an interesting situation where two of the



laboratories had a false positive of N-nitrosodi-



phenylamine in six samples.  I would hate to put



somebody  in jail on that kind of evidence and



knowing that compound was not spiked in the sample



and not present.  Two laboratories confirmed it was.



     One  interesting thing was noted in the data.



One laboratory did not experience any false negatives



with Method 8270.  The tally sheet shows what the



other two  laboratories did on a compound specific



basis.  Some had problems with methyl methacrylate,



acetophenone.  The one surprising  thing, one of



these is  a priority pollutant, chlorobenzene.  One



laboratory could not find it in six samples at a
                        294

-------
concentration between 100 and 200 Mg/L.




     A second objective of the study was how well



the resulting data defined the groundwater situation



that I described earlier.  The results from the



study were reviewed from the perspective of those



conducting an environmental assessment and not from



an analytical point of view.  Because the true



concentration and identity was known, it was possible



to evaluate the data on how well groundwater



contamination at the hypothetical situation was



really defined.




     The analytical results were arbitrarily clas-



sified for the spiked components only and did not



deal with false positives.  You either failed to



detect the compound, which is a false negative...



If you detected a compound, we had two categories.



You detected it and it was within a factor of two



of what was actually spiked into the samples, or



you detected it and there was a factor greater than



two than the actual spiked amount.



SLIDE 9




     Table 9 in the handout lists this information



for all of the compounds.  This is a summary table



right here.  What this tells you is that we do



better for some compounds than we do for others.
                       295

-------
This actually identifies agreement between the



analysis run by two laboratories for Samples 2, 3



and 4 only.  Percent of the compounds detected in



the first laboratory will be found by* the second



laboratory.



     For the 8240 compounds, the non-chlorinated,




you can see that for 49 percent...the bulb went?



I'll have to ad lib from here on out without slides.



Essentially what this says, is that if a first



laboratory found a compound, the chances of a second



laboratory...you have one out of two chances that a



second laboratory would indeed find it.  I can't go



through all of the data on that basis.  It is



included in the report.



     I think I'll just drop down to the conclusions



and the recommendations from CMA.   The results from



the CMA SW-846 Assessment Study revealed that even



the best GC/MS methods contained in SW-846 are



somewhat inadequate for the analysis of compounds




reportedly amenable by Method 8240 and 8270 from



the Appendix VIII list.  Method 8330 was found to



be completely inadequate for the three compounds



included in the study.  This observation was



confirmed by all three laboratories.  The statement



that SW-846 contains analytical methods for all
                       296

-------
375 compounds from the Appendix VIII list simply is



not true.



     Review of the results from the assessment of



bias indicated that the three laboratories were not



able to meet the EPA criteria to demonstrate one's



ability to generate acceptable accuracy and precision



for a large percentage of the compounds included in



the study.  This observation was unexpected since



three prominent laboratories—and these were



experienced laboratories—took part in the study,



and the samples contained only pure compounds in a



relatively simple groundwater matrix.



     There were a large number of false positive



and negative observations by all laboratories to



varying degrees.  This observation is of particular




concern since GC/MS methodology was employed for




rather simple samples in a rather simple groundwater



matrix.  With less specific detectors such as FID



(GC) or UV (HPLC), the problem of false positive



and negative observations would be expected to be



even more severe.  The resulting data for Method



8330 supports this conclusion.



     The calculated Vu and Ru values, repeatability



and variability factors, which express intra- and



interlaboratory basis, revealed precision varies
                       297

-------
from laboratory to laboratory for different methods



and for specific compounds.  These facts must be


considered when analytical data are being interpreted



to assess an environmental situation.



     It is reasonable to conclude that the numerous



problems observed for Method 8240, 8270 and 8330


for the samples included in the CMA SW-846 Assessment



Study, may be expected for other methods in SW-846



not evaluated by SW-846.  SW-846 certainly has not



reached the level of development whereby its use



should be mandated by law.  Until all of the methods



contained in SW-846 are adequately validated, SW-


846 is nothing more than a collection of methods



that may or may not work.  It is only suitable as a



reference document.  Even as a reference document,


SW-846 has limitations.



     If one reviews the resulting data from the CMA



study from a data user's perspective, one would


conclude that the analytical community could indeed



identify groundwater contamination, but not very



well both qualitatively and quantitatively.  The


problem of false positive and false negative



observations would be of major concern in determining
                                   i      , • 'in         • ,,	


intermediate corrective action, as well as the


potential for future problems.
                       298

-------
     CMA recommends that the EPA should not



promulgate the mandatory use of SW-846 for testing



for Appendix VIII compounds under Subtitle C of




RCRA.  The methods have only been validated for a




few of the Appendix VIII list and further-validation



is required before promulgation.




     The EPA should concentrate its efforts initially



to the validation and development of GC/MS Methods



8240 and 8270.  The list of Appendix VIII compounds



amenable to these procedures needs to be established,



along with supporting precision and bias data.



     For Appendix VIII compounds not amenable to



8240 and 8270, EPA needs to validate other methods



in SW-846 to confirm they actually work for some



matrix with a reasonable expectation that they may



work in other matrices.




     Multilaboratory studies of all methods similar



to the CMA SW-846 Assessment Study would be advisable



prior to including a method into a manual such as



SW-846, and prior to considering promulgation of



a method for mandatory compliance testing.  Results



from such studies should be subjected to the analyti-



cal peer review process and appropriate statistical



evaluation, both within and outside the Agency.



     I'd like to give an acknowledgement to the CMA
                       299

-------
Environmental Monitoring Task Group who contributed




greatly to this paper, and particularly to CMA



Staff Excecutive, Sharon Kneiss, who happens to be



in the audience, and to Becky Wilson, the CMA



Administrative Assistant.



     Thank you.



                          MR. TELLIARD:  I'm glad




you're behind us again, George.  Any questions?
                        300

-------
 Inter- and Sntralaboratory Assessment of
 Selected SW-846 Methods for Analysis of
Appendix VIII Compounds in Groundwater
       Chemical Manufacturers Association
             2501 M Street NW
            Washington, DC 20037
     Authors: George H. Stanko
             Shell Development Company

             Peter E. Fortini
             American Cyanamid Company
                                            08611-1
                       301

-------
Physical and Chemical Properties of Groundwater
                 CMA Assessment Study
          Property
   Appearance
   pH
   Total Suspended Matter
   Total Dissolved Solids*
   Chloride, Cl
   Hardness, as CaCOs
        Results
Very Turbid, Sandy Solids
         6.9
      1060 mg/L
       580 mg/L
        70 mg/L
       318 mg/L
   *Filtered thru 0.45ft, membrane.
                                                   08611-2
                              302

-------
 Sample Code Used for
CMA Assessment Study
Samples
Sample 1
Sample 2
Sample 3
Sample 4
Assigned Numbers
10,7,6
8,11,2
5,1,9
3,12,4
                     08611-3
             303

-------
CMA SW-846 Assessment Study (8240)
 Sample Identity and Concentration Information

Compounds
Acetone
Acetonitrile
Benzene
Carbon Tetrachloride
Chloroform
1,1-Dichloroethane
1,2-Dichloroethane
1,2-Dichloropropane
Ethylbenzene
Tetrachloroethane
1,1,1-Trichloroethane
Trichloroethene
Toluene
1,1,2,2-Tetrachloroethane
Methyl Ethyl Ketone
Isobutyl Alcohol
1,4-Dioxane
Samples f/xg/U
1
—
-
50
—
-
-
—
—
—
-
25
—
25
—
—
50
—
2
200
200
50
—
25
25
100
100
—
—
25
—
25
-
200
50
200
3
—
-
50
200
-.
—
—
—
200
200
25
200
25
200
-
250
—
4
400
400
100
300
50
50
200
200
200
300
325
300
75
300
400
250
400
                                           08611-4
                      30*

-------
CMA SW-846 Assessment Study (8270)
 Sample identity and Concentration Information

Compounds
Aniline
p-Chloro-m-Cresol
2-Chlorophenol
Di-n-Octyl Phthalate
Diphenyiamine
Methylmethacrylate
Naphthalene
4-Nitropheno!
Phenacetin
2,4,6-TrichIorophenol
2A5-Trichlorophenol
4-Ch!orophenol
Acetophenone
1,3-Dkhlorobenzene
Chlorobenzene
Samples (/*glL)
1
-
—
-
—
-
-
_
-
—
—
_
-
—
-
—
2
25
100
-
-
100
-
100
—
—
-
—
—
25
25
100
3
- ,
—
300
300
—
300
—
300
300
300
300
300
-
—
-
4
250
200
350
500
250
500
250
500
500
350
350
350
250
200
200
                                         08611-5
                      305

-------
CMA SW-846 Assessment Study (8330)
 Sample Identity and Concentration Information

Compounds
Thiourea
N-Phenylthiourea
1-Acetyl-2-Thiourea
Samples (fJLgIL)
1
-
—
-
2
—
100
100
3
300
300
300
4
500
500
500
                                          08611-6
                      306

-------
Percent of Recovery Observations
     Not Meeting EPA Criteria

SW-846 Method
8240(17)*
8270(16)*
8330 (3)*
Laboratory
A
29
31
100
B
29
69
100
C
94
81
100
*Total number of compounds included in calculations.
                                    08611-7
                    307

-------
      Variability Factor
Vu= Exp (2 S) for the Upper Limit

VI = Exp (-2 S) for the Lower Limit

Example
   Factor       =  1.68
   True or Mean  =  100 ppb

Vu = 100 ppb* 1.68  = 168 ppb

VI  a 100ppb/1.68   =  60 ppb
                               08611-8
                     308

-------
      Repeatability Factor
Ru = Exp (2 • VFS) for the Upper Limit
Rl  = Exp (-2- VTS) for the Lower Limit
Example
  Factor       =  2.08
  First Analysis  =  100 ppb
Ru  =  100ppb* 2.08  = 208 ppb
Rl   =  100ppb/2.08   =  48 ppb
                               08611-9
                    309

-------
         CMA SW-846 Assessment Study
Intralaboratory Precision: Within Compound and Sample Pooled
Method
8240


8270


8330


Laboratory
A
B
C
A
B
C
A
B
C
Degrees of
Freedom
70
60
53
64
43
50
3
6
0
Vu
1.29
1.21
2.14
1.63
3.34
2.40
1.63
2.43
-
Ru
1.44
1.32
2.94
2.00
5.50
3.45
2.00
3.52
-
                                                 08611-10
                             310

-------
 CMA SW-846 Assessment Study
   Intel-laboratory Precision, Combined*
Method
8240
8270
8330
Vu
2.31
2.40**
-
Ru
3.27
3.46**
-
Ru Range
1.70-9.01
1.63-7.63
-
**
* Based on average variance.
  Excluding 4-nitrophenol, 2,3,5-trichlorophenol.
                                     08611-11
                    311

-------
 False Positive and Negative Observations
Method 8240 (Volatiles)
  7 False Positives, Sample 1,2 Labs
  False Negatives
     3 Labs, Isobutyl Alcohol
     3 Labs, 1,4-Dioxane
     2 Labs, Acetone
     2 Labs, Acetonitrile
     2 Labs, Methyl Ethyl Ketone
     1 Lab, 1,1,1-Trichloroethane
 50-250>g/L (12)
200-400 /xg/L  (6)
200-400//g/L  (6)
200-400 jug/L  (6)
200-400 //g/L  (6)
   25   wg/L  (9)
   Of 17 Compounds Spiked into 2 or More Samples,
        There Were 7 Cases of False Negatives
                                               08611-12
                            312

-------
 False Positive and Negative Observations
Method 8270 (Semi-Volatiles)
  8 False Positives, 1 or More Labs
  2 Labs N-Nitrosodiphenylamine in Same 6 Samples
     1 Lab, No False Negatives
     2 Labs, Methylmethacrylate  300-500 /xg/L   (6)
     2 Labs, Acetophenone        25-250 /xg/L   (5)
     1 Lab, Diphenylamine       100-250 jUg/L   (6)
     1 Lab, Phenacetin           300-500 /xg/L   (6)
     1 Lab, Chlorobenzene       100-200 ftg/L   (6)
                                              08611-13
                          313

-------
                                                          Jy,.>	l:i
C

O
  _
*>>>
o

C
l/l
vn

(0




0)
h.
1
V)
C
Q
*g
«J
"5
Q
M-
o
s-











o
'•5
0)
01
Q
Present
Within X2
3 CX
.alEs-
•M±i o

0) 5 O
tri ^^ 4l*
£ o v
Q- C ,,

0)
(Q 'w?
U. O
O.


ut
(O
.U
T3
C
3
O
Q.
E
o
u




in «* oo (N  ^Q ^



*
*
.. o» en o
TS .E T3 .£ .£
QIC OJ C C
to oj nj 'jg 
-------
 a?
 °s *>«
 2  o
 *f *j
 c "5.

 3  E
^   <
O)
E3
    £^
    a.
(U
 O

 2
 o
JQ
 to
01
Wl
 O
 a.


 o
u

*o

S?
         -U. Q


        §.Il
                         So
 c+: c
+4 tO (JJ

 3 « C



"X3 flj 
                              T-  00
                              m  io
                                       *
                                       *
                  
-------
        Acknowledgements
CMA Environmental Monitoring Task Group
Sharon H. Kneiss, CMA Staff Executive
Becky Wilson, CMA Administrative Assistant
                                    08611-16
                       316

-------
             QUESTION AND ANSWER SESSION








                           MR.  BRITTON:   What  were



 the outlier  criteria that you  used  in  that  study,



 George?




                           MR.  STANKO:   It  failed




 all of them.   Grubbs was the primary one.   Those




 three values were  so far out that they were out by



 a  factor  of  10 to  15 from the  rest; very isolated



 cases.  We didn't  chase  down what caused the  outlier,



                           MR.  BRITTON:   So  you



 basically ignored  only those observations which




 were roughly a factor of 10 larger  than they  should



 have been?




                           MR.  STANKO:   We applied



 the Grubbs test, but that's how  far out they  were,




 to give you  some idea of the magnitude  of how far



 out they  were.




                           MR.  BRITTON:   Of  course



 it depends on  the,confidence level.




                           MR.  STANKO:   Right.



                           MR.  BRITTON:   That's one



of the reasons why there'd be  a great difference



between results from your  analysis and  results from



our analysis, even on the same  data.  We have a bit
                        317

-------
of a difference in philosophy, I think, as to



whether we are concerned with characterizing the




data that would reflect in general use of the



method, or whether we're interested in character-



izing how the method data would lookwhen performed



properly.  That philosophical difference is



tremendous.  It's one of the reasons why the log



transformation...log transformation is appropriate



if you're trying to encompass virtually all the



data that would be generated by use of that method




in practice but would basically accomodate all of



the data that represents inadequate performance.



But if you are interested in what  the method would



look like, what data from the method would look



like when performed properly, then you have to try



and look past that data.  It's a very big trick



cause none of us know exactly which numbers are  bad



and which numbers aren't.  We want to look past



those bad numbers.



                          MR. TELLIARD:  John.



                          MR. McGUIRE:  John McGuire,




EPA.  George, one question.   You said...probably I




wasn't paying attention.  You said that you had



filtered  the samples.   Was that before or after




spiking.
                        318

-------
                          MR.  STANKO:   No,  the



samples were not filtered.




                          MR.  McGUIRE:   I thought



the groundwater was filtered.




                          MR.  STANKO:   No.  That



groundwater sample contained solids.  But we looked




at the data to see if the solids had any affect on




the analysis of the compounds  and in all honesty, I



could not detect any irreversible absorption of the



compounds we used in this study.




                          MR.  McGUIRE:  Okay.



There was no attempt made to extract the solid



itself, then?




                          MR.  STANKO:   The solids



were extracted, yes.




                          MR.  McGUIRE:  They were?



                          MR.  STANKO:   With the



water, in situ.




                          MR.  McGUIRE:  Okay, but



I mean totally extracted?




                          MR.  STANKO:   Total



extraction.




                          MR.  McGUIRE:  Thank you.



I certainly hope I can get a copy of that.




                          MR.  TELLIARD:  In the back.
                       319

-------
Any other questions?  All right, folks, thank you
very much.
     Point of interest.  Tomorrow at a quarter to
nine here.  The boat will be leaving about 5:30-ish.
That gives you about 20 minutes to get your snuggies
on and get out there.  Waterside, down there by
Phillips.  Thank you.
(WHEREUPON, the 4-3-85 session was adjourned.)
                        320

-------
Revised Presentation
         321

-------
INTER- AND INTRALABORATORY ASSESSMENT OF
 SELECTED SW-846 METHODS FOR ANALYSIS OF
 APPENDIX VIII COMPOUNDS IN GROUNDWATER
   CHEMICAL MANUFACTURERS ASSOCIATION
           2501  M Street,  N.W.
         Washington,  DC  20037
 Authors:
            George tf. Stanko
        Shell Development Company

            Peter E. Fortini
        American Cyanamid Company
               Presented at
        U.S. EPA Symposium on the
Analysis of Pollutants  in the Environment
            Norfolk, Virginia
             April 3-4,  1985
                         322

-------
                                  ABSTRACT
    EPA  Methods  8240,   8270  (GC/MS),  and  8330  (HPLC)  from  SW-846  were
evaluated at three prominent  laboratories  for  a select list of  36 compounds
that  were  spiked  into  a  relatively  simple  groundwater  matrix.    Sample
preparation was designed to simulate  a  groundwater contamination problem at
an abandoned  waste site and  samples  were prepared as  blind  triplicates to
facilitate  statistical  analyses  for  precision  and bias.    Results  from the
assessment  of   bias   (accuracy,   recovery)   indicated   that   the   three
laboratories  that took  part  in  the study  were not  able  to meet  EPA's
quality  assurance criteria  for  a  large percentage  of  the  36  compounds
included  in the  study.    The assessment  for  precision  revealed  precision
varies  from  laboratory  to  laboratory,   for   different  methods,  and  for
specific  compounds.     A  large  number  of  false  positive  and  negative
observations resulted at all  laboratories to varying  degrees.   The list of
Appendix  VIII  compounds reportedly  amenable to  Methods  8240  and  8270 was
found to  be somewhat less  than  reported  by  EPA.  Method  8330  was found to
be completely inadequate for  the detection  and quantification  of the three
compounds included in the  study.   The statement by EPA that SW-846 contains
the  analytical  methods  for  all  (375)  Appendix VIII  compounds  (excluding
exotics and water reactive  compounds) simply is not true.
 RLL8506605
                                           323

-------
      INTER- AND INTRALABORATORY ASSESSMENT OF SELECTED SW-846 METHODS
      	FOR ANALYSIS OF APPENDIX VIII COMPOUNDS IN GROUNDWATER
                              1.  INTRODUCTION

     The U.S. Environmental Protection Agency  (EPA)  published "Test Methods
for  Evaluating  Solid  Wastes,  Physical/Chemical Methods"  (SW-846)  to serve
as   a   methods  manual  for  the  sampling  of  groundwater  or  leachate  and
analyses of 375 parameters listed in Appendix VIII of Title 40  Part 261 of
the  code of Federal Regulations  (40  C.F.R.  261).   On October 1, 1984 (Fed_.
Ree    pp.   38786-38809),  the  EPA  proposed  to  amend  its hazardous  waste
"regulations under Subtitle C  of  the  Resource Conservation and Recovery Act
(RCRA)  to  make  SW-846  methods  mandatory for  all testing  and monitoring
activities required under Subtitle C,  as  specified in 40 C.F.R. Parts 260-
271.  The  EPA  identified  the  principal  issue  of the proposed rule as  how
testing under  RCRA should  be  carried  out,  and the  proposed  rule specif-
ically addresses  SW-846  to be  used  to,  "ensure  accurate,  consistent,  and
comparable testing results—year to  year,  facility to facility, and region
to region".

     In May  1983,  Environmental  Testing  and Certification Corporation (ETC)
prepared a report1  for the  Chemical Manufacturers  Association (CMA)  from
evaluating the effectiveness of  SW-846  (2nd Edition)  in  assuring that users
would  be guided to select consistent analytical and  sampling procedures for.
each  parameter.   The American  Petroleum Institute  had  a similar  report
prepared by Radian Corporation.  Both of  these  reports revealed that SW-846
was  not adequate  to  properly guide an analytical laboratory due to  lack  of
sufficient  information and  details,  technical  inaccuracies, and  incon-
sistencies.   The evaluations  by ETC  and Radian  also  pointed to numerous
problems associated  with  compounds  identified  by the  Appendix VIII list.
To  verify  these findings,  CMA conducted a limited study  at  three  prominent
laboratories involved in  environmental analyses.    The  prime  objective  of
the CMA study  was an  assessment  of inter- and intralaboratory precision and
bias  (accuracy)  of   selected  analytical  methods  from  SW-846.   A second
objective  was  to  determine  how  well  the  resulting  data  would define
groundwater contamination where  the  identity  and  levels of  all  pollutants
were known.

                               2.   STUDY DESIGN

     The  CMA   limited  study  was  designed  to  simulate  an  environmental
 situation  that could very  well exist,  and someone  would  have to  install
wells;  collect  and   analyze  groundwater  samples;  and  finally assess  the
nature  and magnitude of  the problem.  The  environmental situation was  as
 follows:

     There  is  an old waste disposal  site  where groundwater contamination is
     suspected.   The  direction  of  flow of  the grourdwater  is known and a
     decision was  made to  install four monitoring wells.  One of these wells
     is  located as a  background  well and three are down gradient  from the
     disposal  site.    Up  gradient  from the  background  well is  a gasoline
     service  station   that experienced  a  leaking  storage tank  in  the  past;
 RLL8506605
                                            324

-------
     however,  groundwater contamination from  the  leak was never assessed.  A
     small  subdivision up  gradient  from  the  waste site  uses septic tanks.
     Chlorinated  drain  cleaner  (1,1,1-trichloroethane)  has  been  used  for
     years.  All  of the organic compounds  that  were disposed  of at the site
     (unlined)  are  known.   It  is reasonable  to assume that  the drums have
     been  leaking  for sometime, and  it would be  reasonable  to believe that
     all of the compounds have  migrated into  the groundwater.

     Four  samples  were   designed  for   the   CMA  study   to  simulate   the
 environmental  situation  described.   A list  of  34 organics of interest from
 Appendix  VIII  was  developed  and  the  distribution and approximate concen-
 trations  for the four samples  was prepared for a request for proposal  (RFP)
 from four prominent  contract laboratories.  The laboratories were contacted
 about  the CMA RFP  and advised  to  maintain  the  integrity of  the compound
 list as well  as  the sample list.   As part of  the  RFP, nine  SW-846 methods
 were suggested to  be used to conduct  the  analyses.   CMA  received bids from
 three  laboratories  for  the  study.  All  three laboratories  concurred and
 recommended that SW-846  methods 8240  (GC/MS  for volatiles),  8270 (GC/MS for
 semi-volatiles),   and  8330  (HPLC  method,  sulfur  compounds)   be  used  to
 analyze for the  select  list of  compounds.   This concurrence was  not sur-
 prising since  the  compound list  was  originally designed to  test  the GC/MS
 methods which were considered  the best and most reliable  in SW-846.  Method
 8330 was  randomly  selected for evaluation of an  HPLC procedure  from SW-846
 and compounds  listed in the method were included in the compound list.

     CMA accepted proposals from  three  laboratories to analyze  the samples
 for the  study  and selected  one  laboratory  to prepare the   samples  and  to
 prepare a  report  which  included  data from all laboratories.   To  maintain
 the integrity of the study,  the CMA list  of  compounds  was slightly altered
 with respect  to  compounds  listed  as well  as  concentrations  in  various
 samples.   In the case of the laboratory that prepared  the samples,  special
 precautions  were taken  to make  sure  the  confidential information was not
 available  to the analysts.   Each  of the laboratories was instructed  by CMA
 to  analyze each sample  by SW-846 Methods 8240, 8270,  and  8330.   The  lab-
 oratories   were  also  instructed  that   there  were   other   Appendix  VIII
 compounds  not  listed  in  Methods  8240  and  8270, but were  amenable  to  these
 methods according  to EPA.    Three  laboratories  performed  the  analyses
 independently  and  submitted the data  to CMA.   When all data  were  reported,
 CMA sent  the  data  to  the one  laboratory selected  to  prepare a  report.
 Statistical  analysis  of all the data  was performed to  determine inter- and
 intralaboratory  precision  and  bias.    These  statistical results  from the
 contract laboratory were  included  and  were the basis for this  report.
                              3.  EXPERIMENTAL
3.1 Groundwater Matrix
    Groundwater  for  the study  was  provided by  a CMA member  company.   The
water  was  collected  from  a monitoring well  located  in  the  coastal  plain
region  of  Texas.  Some  physical  and chemical properties  for  the water  are
listed in Table  1.
RLL8506605
                                          325

-------
                                  Table  1
   Physical and Chemical Properties of Groundwater - CMA Assessment Study
    Property
    Appearance
       PH
Total Suspended Matter *
Total Dissolved Solids3'
    Chloride, Cl
  Hardness, as
                                                        Results
                                               Very turbid, sandy solids
                                                          6.9
                                                       1,060 mg/L
                                                         580 mg/L
                                                          70 mg/L
                                                         318 mg/L
 a) Filtered thru 0.45u membrane.

A  portion of  the groundwater  was also  analyzed prior  to preparation  of
samples  to establish  the presence  or absence  of organics.   The  results
demonstrated  the well sample  to be  free  of organics and  specifically the
list  of  organics of  interest  from the  Appendix VIII  list.   Since  the
groundwater  was free  of  organics, method  bias  (recovery,  accuracy)  could
easily be assessed from the spiking levels  (true  value - spiking value).

3.2 Sample Identification

    Four  samples were used for  the CMA  study.   To obtain  the  necessary
data,  these  four  samples were   analyzed  as  blind  triplicates.    Sample
numbers  were  randomly generated  and  assigned   to  the  samples.    Table  2
identifies the  sample  numbers employed.
                                   Table 2
                  Sample Code Used for CMA Assessment Study
Samples
Sample 1
Sample 2
Sample 3
Sample 4
                                                       Assigned Numbers
                                                            10,  7,  6
                                                             8,  11,  2
                                                             5,  1,  9
                                                             3,  12,  4
Each  laboratory  received  12  samples  identified  only by  a  single number.
For  each sample  number  they received two  one-liter  containers and two VOA
vials.  The  study was coordinated  in a way  that analysis  of samples was
initiated  on the  same day.
 3.3  Sample  Preparation

     Table 3 reveals the identity and  spiking  levels  of  compounds  in each of
 the  four  samples.   Complete  details  for  the sample  preparation have  not
 been included in this  report.   The groundwater matrix was  spiked following
 procedures  described  in  Section 7,  "Calibration"  of EPA  Method  624.   The
 samples were  prepared  to simulate  what  might  be  expected  for the  four
 hypothetical monitoring  wells  described  in  the  enviornmental  situation.
 Sample 1 (identified by numbers 6,  7,  and  10) was  prepared to represent the


 RLL8506605
                                          326

-------
                                                 I      I      II
                                                                 in
                                                                 fM  I
                                                                                                  I      m      III      ||
CO
                            O
                      i   I   in  i
I   I  in  i
                                       I   I
                                                                        in
                                                                        fN
                                                                 in
                                                                 fM  I
                                                                 in
                                                                 fM  1
                                                                                                  .      s
                                                                                       o
                                                                                 i      in
                                                                       i   i
                                                                                                                                 l   i
                                                                                                                                                  i   i
                    o  o  o
                    o  o  in
                    fM  (M
                ITl IT
                fM CM
O
0
                                      O
                                      0
                                         ||
   in
   (N      1
                                                                      in
                                                                      fM  1
                                                 0
                                                 0
                                                 fM
                                                                                       c
                                                                                       IT,
                 O
                 o
                 fM
                                                                                                in o
                                                                                                fM O
                                                                                                                                   IT-
                                                                                                                                I   fM
                                                                                                                                c
                                                                                                                                c
00
                    O O  O
                    O o  in
                    (N (N
                          O
                          0
       O
       0
                                         l|
                                                                     in
                                                                     fM  1
                           C
                           o
                           fM
                                                                                       C
                                                                                       m
                                                              O
                                                              O
                                                              IN)
                                                                                               in  o
                                                                                               fM  O
                                                                                                                                   in
                                                                                                                                1   fM
                                                                                                                                C

                                                                                                                                •— '  1
    O  O  O
fM  o  o  in
  IfM  fM
                                     in in     o
                                     (N (M     0
                                 o
                                 0
                                                              l|
                                      IT
                                      fM   1
                                                                               0
                                                                               o
                                                                               fM
                                  0
                                  m
                                                                                        O
                                                                                        o
                                                                                        fM
                                                                                                                    in  o
                                                                                                                    fM  O
                                                                                                                  in
                                                                                                                  fM
                                                                                                                                                    i
         §
                                                            °  0
                                                            00
                                                            (N  (N
                                                 in
                                                 fN)
                              o
                              O
                              CM
                                                                        0
                                                                    in  o
                                                                    fM  fM
                                 o
                                 in
                                 fM
                                                                                                     1   1
                                                                                                             O
                                                                                                             O
                                                                                                             ro
                                           o
                                           o
                                                                                                                                  o
                                                                                                                                  o
                                                                                                                                  ro
 EH 2
 CO O
    M
 EH EH
 2 <
 W K
 E EH
 CO 2
 CO U
 W u
 W 2
 CO O
 < U

VO Q
"*  2


IS
CO  EH
                         SI
               •    '
                                       00
                                       
                                                                                                                       0
                                                                                                                       O
                                                                                                                       rn
                                                                                                                                  0
                                                                                                                               1   O
                                                     O
                                                     O
       Q
       H

       W
              CM
              O O  O O
              O O  O O
              V T  t—l m
                                   O O
                        O
                        0
    O
    o
    CM
                                           O O
                                           oo
                                           (M er,
                                                     in
                                                     fM
        o
        O
                                                                   tn  o
                                                                   !-•  0
                                                                             0
                                                                             O
        0
        in
        fM
                                                                                           O
                                                                                           0
                                                                                           rr
 00     00     O     00
 ifi O     m o     o     m  o
 fM CM     m fM     m     fM  in
       CO
O  O O  O
O  O O  O
*»  ** r-l  CO
                                             8
                              g
                              H
rH O
£ r-(
u x:
o
o
o
2
•H C
a 10
i x:
- i
& 5
o

o
u
(0
o>
1 1
(1) (00) O
C >H C _, 0
fO 4J 10 V? rH 0)
i x: i oi x: ? < c
•H 4J O EH 4-> i (0
lj -J 1 Q) S '-H X
0)E-«OO< O
Cl tHr-HCC -VjrHC4-> -rH
(OrHOxiaJaicMo^iOs Q
£ »rHCJX:D «rHX:4JXirH|
4->rHX:-H4JrHrHX:4J 0) O O^
a) -uk-ioio --cjaii^tnx:'-
<-H EH EH t-H E M rH
                                                                                                                            O  I
                                                                                                                            c  o
                                                                                                                            (D  V!
                                                                                                                            rC  O
                                                                                                                            & rH
                                                                                                                     O rH  o x:  01
ne
-
                                                                                                                  o>  kj  o  IH   u  c
                                                                                                                  c  O  in  O -H
                                                                                                                 •H  rH  0)  rH O
                                                                                                                 •H  x:  »-i  x:  i
                                                                                                                 •H  U U  U , 4J  ra  4J  a)
                                                                                                                4J   10 rH  0,' 4-1
                                                                                                                 U  rH  >, E  10
                                                                                                              a'  o
                                                                                                              N
a
                                                                                              C rH  rH
                                                                                       i   x:  o>  >i  >i
                                                                                   c  c  4J x: x:   >j
                                                                                       i   x:  & 4J   o
                                                                                      •H  Hi -H  O>   10
                                                                                      C     Q E
             CO
             §
             1
                00
                                                                                            (M
                                                                                            cc
                                                                                   327

-------
                  I    I   I   I       I
                                                   II       I       II       I
             r- II    i   i   i       i       i       ii       i
•a
 o
 o
                                                   ii       i        ii       i
 CO

 g
             i   i
                                      in LH     o
                                      CM CM     O
                                                     O
                                                  I   O
                                                         o
                                                         o
                10
             CO  O  I
                                      in LO     o
                                      CM CM     o
                                                     O
                                                  I   O
                                                         o
                                                         o
                 o
             CM  O  I   I    I
                                      in in     o         o     o
                               I       CM CM     O      I   O     O
             a>
        ooo
        o  o o
        ro  ro ro
                O
                O
                ro
                   O
                   O
                   ro
                                      O  O
                                      O  O
                                      ro  ro
           O
           O
           ro
U
        ooo
        ooo
        ro  ro ro
                O
                O
                ro
                   O
                   o
                   ro
                                                   i    i
                                      O  O
                                      O  O
                                      ro  ro
           O
           O
           ro
a
                     OOO
                     OOO
                     ro co  ro
                       o
                       o
                       ro
                       O
                       O
                       ro
                            l   i
                                      O  O
                                      O  O
                                      ro  ro
           O
           O
           ro
1-1  r-1
EH
Z
U
O
l-t
                 O O O  O
                 in o o  in
                 CM in in  ro
                       O
                       in
                       ro
                       o
                       in
                       ro
                           O  O
                           in  o
                           CVJ  (N
                              O
                              O
                              CN
O  O
O  O
in  in
O
O
in
                 O O  O  O
                 m o  o  in
                 (N in  in  ro
                       o
                       in
                       ro
                       o
                       in
                       ro
                           O O
                           in o
                           (N (N
                              o
                              o
                              H
 0)  EH
 U   I
 10  UJ
 c   -
 01  ^r

o<  CM
                               di M
                               O  EH  O
                                      •§,  A
1-1     T-t              Id)
 O      O          01  O      C
 C      C          C  U      0)
 0)   I   01          O  O      N
                   C r-1      C
                   01 S^   01  01
                   f.  O   C  X!
 ...       o  <-<  & -H   01  o
 OinOr-lOOQNtH
,H   .r-IJ=C4J|   CO
.CT.CU
-------
background  well  and was  spiked  with benzene,  toluene,  isobutyl alcohol
(leaking   tank),   and  1,1,1-trichloroethane   (septic  tank).  These  same
compounds were  spiked in all samples.  Samples 2, 3,  and 4 were prepared  to
simulated the three  down gradient wells at the hypothetical waste  site.   As
noted  in  Table  3, samples 2 and  3 contained  approximately half of the list
of  36  compounds in  each sample.    Sample  4 contained  all  compounds  at the
highest  concentration levels.   The  study  was designed  so each  of  the  36
selected compounds was present  in  a minimum of  two  samples.


3.4 Analytical  Results

    The  analytical   results   from the three  contract  laboratories  were
summarized  into the  five tables included in Appendix A.   The data were put
in  a  format  to  facilitate  statistical   analysis.    The  tables identify
laboratories  by code  (A,  B, C),  the coded samples,  compounds spiked into
samples,  and the  reported  values for  the  study.  Reviewers  of Appendix A
are  cautioned  that   the purpose   for  the study was not  to  compare  the
performance of  a  given  laboratory with respect  to other laboratories, but
to  assess  analytical  methods contained   in   SW-846 for  the  analysis   of
Appendix  VIII compounds in  a real groundwater matrix.   It  should also  be
noted  that  the  three laboratories are prominent  laboratories  that  employ
modern instrumentation,  have  years  of experience with  GC/MS methodology,
and  certainly  have   qualified  and  experienced  analyst.   There  is  every
reason to  believe  that  the  analytical   results  obtained  for  the  study
represent  the current state-of-the-art with  respect  to  low level environ-
mental analyses.

3.5 Statistical Analysis of  Data

    The analytical data were statistically analyzed  to assess the two most
important  criteria  of  analytical  methods, precision and  bias  (accuracy,
recovery).    There are numerous  acceptable ways  that this  could be done.
One of the procedures employed was taken  from procedures  described  in the
EPA methods as part  of  the quality  assurance sections.   Every effort was
made  not  to  distort the  statistical  results,  but  to be  conservative and
reasonable  where arbitrary decisions  or assignments were  required.

3.5.1  Assessment of Bias (Recovery, Accuracy)

    The  statistical   procedures used  to assess  bias  were taken  from the
quality assurance  section  of SW-846 Methods 8240 and 8270.  One is directed
to calculate  percent recovery of  spiked  compounds,  then calculate standard
deviation for the  percent  recoveries  assuming  a normal distribution for the
recovery values.   The resulting  standard deviation  is  then used  to calcu-
late  a  percent  relative   standard  deviation  (ZRSD).    The  mean percent
recovery  and %RSD of the  percent recovery  defines  the bias  for a given
compound by a specific method.   For  the CMA  study,  these calculations were
done for  each laboratory and  for  all  laboratories combined.   It should  be
noted  that  the  %RSD  defines the  dispersion  of values  of  percent recovery
with respect  to  the  simple  mean percent recovery.  The %RSD also represents
one measure for precision  as well.  While  the %RSD  values  were  calculated


RLL8506605
                                          329

-------

                                                     c ^ r  r  c. -T   cc «» — i-i c. —   T  c
                                                     c^^cr-c  * ir r- c — r^ C i—'

        if
                    p-irseccvcriNCcccr-
1
    01     •    ....  *                 «                 C
   r-IKJ i i r- u-. ic c ,/• _, u- c c f-   rv r,     a- £ "•"



   I

   * -[

   |i|=0S?l~2^i = HE?S   ~°°     ??-S:£SS=sgcsss;£-c    c





3



  s


  £



            O       in  a\  *r   M                                  v o


  "IIs = - * = s 2 "' 2 "' - "' 2;''   ~ '  '      ssrssss^ayz-'-'s'ss



   e



                              f*"1   ^^OO     V'^\CC'Mifr^cC'ff^i'*lf**t.'"\ct^f1*    CC


   e cl


2 ,

  i



  I



  I.






      OQOOCCOO   omomo   coo     occcoccccooooooc1
  _,   COCOU^U*. OO   Q f*t C r* &   OiTC     ifCiT-CCC "L"	C iT'COtTiiT'inii^O

   [cj*  i  i  i't  i  i  ici  i  i  i  i    i  iT     T'i7'fi47'irri'^virirTirv'i<'i*
  c. u=. oooiAtnoorsctrciT. o   ooe     w-ooi.-t.-coooooocotro
      C.CtT'OiNiNCC   O^CINC   Oi/*C     M C C "•< «v C C Cf C O C O C C" *N C»
      '•**'*   rj    F-W   IN   rs*  rg   f>*  (N       —,—     f-^i-i^-f-|f-if-,f-i?i   „
                         «»
                         C
                         «
                         2.
                                   g
                                   AJ M
                                   a. c
                                                        fc « -H

                                                        N « flj

                                                        C,,, C £
.|
                                                              & U   «
                                 .

                         OO.wt. S-
              ^;

              C C C

              C t) C


                                              UCCCC-CU   C   — ^, c
                                              t  ^ w t-   •-< ID «. C   ££4

                                              ft£.2»Lif f i; £ !: !!i

                                             « £  £f tt xl1; &-t fi: £,
                                             c c-  c - « o c « £ fc. u i  i  r

                                             SSS^^fci^StiS^'.s

                                             s vv ". •"..: s-1 s- f i *.'. v
                                             « O.r%» — ^^oojCZvO.rsi
                                                      330

-------
and  Included in the  report (Table 4), an  alternate procedure was employed
to assess precision.

    Table  4 was prepared  to summarize all  the statistical information for
bias  (accuracy)  following the procedures specified  by EPA in Section 8.2.4
of SW-846  Methods  8240 and  8270.   Table 4  shows  the bias results for each
compound included  in the  CMA study by  laboratory,  by method,  and for all
three  laboratories combined.  The number of  observations included in each
recovery calculation was  also  listed.    Some  discretion was  used in prep-
aration  of  Table   4.    No  mean  recovery  value  for  all  laboratories  was
calculated  when  only one laboratory reported  values for a compound.  There
were  several instances  where a laboratory  reported values for  one set  of
samples  but did not  find  the compound in  a second set.   For these cases,
the  zero percent  recovery observations  were  not included  in calculations
since  they  would   have  influenced bias  in  a  way  that  did   not  appear
justified nor useful.

    The  performance  criteria set  forth in Section  8.2.4 of  the Quality
Control  section  of  SW-846  Methods  8240 and 8270 is identical.  "The average
percent  recovery  must   be greater  than 20 (%)  for all  compounds  to   be
measured and greater than 60 (%)  for all surrogate  compounds.  The percent
relative  standard  deviation  (ZRSD)  must   be   less  than  20  (%)  for  all
compounds  to be measured  and  all surrogate  compounds."    Section  8.1.1
states,  "Before  performing  any  analyses, the  analyst must demonstrate the
ability  to  generate acceptable  accuracy  and  precision  with  this  method
(8240,  8270).    This  ability is established as described  in  Section 8.2."
Table 4  was reviewed to determined the  "ability"  of the three laboratories
to  "generate acceptable  accuracy  and  precision" using the  EPA  criteria.
The results are  summarized in Table 5.

    The  data summarized in  Table  5 indicate  that these three laboratories
did  not  meet the  criteria  for  a  large  percentage  of  compounds to demon-
strate  their ability  to  generate  acceptable  data.   Considering that the
three  laboratories  are  prominent  laboratories  with  considerable  GC/MS
experience,  it   is  doubtful  that any laboratory can meet  the criteria for
all  compounds  amenable  to  the  three  SW-846  Methods.   It should also   be
noted  that  the   methods  give no guidance or instructions on  what is to  be
done when the criteria are not met.

                                   TABLE  5

          Percent of Recovery Observations NOT  Meeting EPA Criteria
         SW-846 Method
         8240 (17)
                  a)
         U^,v, V*Wv
         8330 (3)
                 i)
 Laboratory
 A.    JL    £
 29   29   94
 31   69   81
100  100  100
         a) Total number of compounds  included  in calculations.
RLL8506605
                                          331

-------
 3.5.2 Assessment of Precision

     The resulting data from the CMA study were  statistically  analyzed  using
 lognormal distribution  theory  to assess  precision for  each  method  (i.e.,
 8240,  8270,  8330).   As  measures of  intralaboratory  and interlaboratory
 precision,  variability  factors   (V)  and  repeatability  factors  (R)  were
 calculated.

     Upper and  lower variability  factors  are defined  by equations (1)  and
 (2)   where  S  is  a  pooled   standard  of  logarithms  of  amounts  found  by
 analysis.

     (1)  Vu * exp(2S) for  the upper  limit

     (2)  VI ~ exp(-2S) for the  lower  limit

 These variability factors define  95%  confidence limits for the  ratio  of an
 observed  to  a known mean or  true  value.   If  the  true or  mean value is
 expressed as  (X),  then 95%  confidence   limits  on  an  observed  value  are
 calculated using equations (3)  and (4).

     (3)  X o  Vu « upper  limit

     (4)  X/Vu - lower limit

 For  example,  for  a mean or  true  value of 100 ppb and  a Vu value of  1.68,
 95%  of the  results for a  sample would fall  in a range of 60  ppb  (100/1.68)
 to 168 ppb  (100 o 1.68).

     Upper and lower  repeatability factors are defined by equations (5)  and
 (6).    These  define  95% limits for  an  observation  when the  true or mean
 value is not  known  but only  estimated  from  a  single previous  observation
 (x).

     (5)  Ru - exp (2 /2  s) for  the upper  limit

     (6)  Rl «  exp (-2 /2 s)  for the lower  limit

 The  95% prediction limits  for the second  observation  are given by equations
 (7)  and (8).

     (7)  X o  Ru - upper  limit

     (8)  X/Ru - lower limit

 Thus,  if  the  first analysis  of a sample gives a  value of  100  ppb and  the Ru
 value  is 2.08,  then with  95%  probability,  the  results  for  the  second
 analysis  will  fall  in  the  range  of 48  ppb  (100/2.08)  to 208  ppb  (100  o
 2.08).

    Variability  and  repeatability   factors   expressing  intralaboratory
 precision for  each  compound  in  the  CMA  study  are  given in Table 6.
RLL8506605
                                          332

-------
 Reported  results were transformed  to  the standard deviations of log values
 (amount  found) were  calculated  for each laboratory,  compound,  and sample
 from  results  reported for  the  (up  to)  three triplicates.   Non-detected
 observations  were omitted  from  statistical  analysis.   Two  unusual values
 for  diphenylamine on  different  samples  reported  by laboratory  B,  and  the
 single value  for methyl ethyl ketone  by  laboratory  B were also excluded as
 outliers.   Standard  deviations from triplicate analyses samples were pooled
 over  laboratories and sampled  on compound-specific basis  and  Vu  and Ru
 calculated  from  equations  (1)  and  (5)  using the  pooled  standard devi-
 ations.   Also included in Table 6 are  the  number of  samples  in which  the
 compound  was  detected, divided  by the  number of  samples in  which it  was
 present;  the  degrees  of freedom  in the estimate of (S) used to calculate Vu
 and Ru;  the geometric mean level;  and the percent recovery.   The geometric
 mean levels were calculated using equation (9).

     (9)  Geometric X - exp (1/nSlog X)

 The  percent  recovery  reported  is  this geometric mean divided  by   the
 geometric mean of  actual  levels   in  the samples  where  the  compound  was
 detected.

     The Vu  and Ru calculated  in Table 6  represent  somewhat of  an  average
 within laboratory variability  for the  three  laboratories.  These  Vu and Ru
 values can  be used to  estimate  the range of values one  might expect  (95%
 confidence interval) for  any  of  the laboratories with  respect  to  a  true or
 mean value  (Vu)  or first  observation  (Ru).    For  example, the value of Ru
 listed in Table 6 for  1,2  dichloroethane  is  2.02.   If a first determination
 for that  dichloroethane was  100  ppb,  then the range  of  values  for a second
 observation  would be 50 ppb  (100/2.02) to 202 ppb (100.2.02) when the  same
 laboratory  analyzed  the  sample  a  second  time.    Precision does vary  from
 compound   to  compound  and that  this   variability  must  be considered  when
 interpreting the analytical data.

    The Vu  and Ru values  shown in Table 7  quantify the  variability  each
 laboratory  experienced  for  the   group   of   compounds using  a   specific
 method.    Standard  deviations   for triplicate  samples  were  pooled  over
 compounds  detected by each method for  each of the  three laboratories and Vu
 and Ru values  calculated for each method  and  laboratory using equations (1)
 and (5).   Using  Ru values  it  can  be  seen that Laboratory B had the  best
 precision with Method  8240 and the poorest for Method  8270.    Laboratory A
 had the  best  precision  for  Method   8270.    The  precision  observed   for
 Laboratory C  was  fairly   consistent   for  both  Methods   8240 and 8270.
 Although  the  values  for  Ru  and Vu were calculated  for  Method  8330  and
 included  in  Table 7,  they  are of very  limited value due to the low  rate of
 detection  for  all  laboratories.

    Interlaboratory  estimates of precision were also computed  for  the  CMA
 Study  and  expresssed as Vu and  Ru  values  in Table 8.   This  was  done on a
 method  and   compound  specific   basis.     Actual   reported   values   were
 transformed  to the  logarithm  scale.    For  each  compound,  the  root mean
 square  different  (in  this  scale) between   determinations  of  different
 laboratories on equivalent  samples was  calculated using  equation (10).
RLL8506605
                                           333

-------
r
M ''it1'
            VO
                         c.
                         E
                         I
                         a



                         c
                         *t
                         n
                         s     °
                         •"     c
                         e     <•
                         «     »<
                         £     &
                         £     £
                        l.ii
                         1    I
                                                                       o cc c c    rt
                                                                                             r- c -• ^ ***
                                                                                                                                                  ml CC


                                                                                                                                                  rsi cc
                                                                                             cctrccccccccoccccccrccttacccccc

                                                                                             X X. V. "V \ \ V X X.\X.XtX,Nfc-xN.i
                                                                                             co "** a i\£ccccf>icccc
                                        i   i  t  i
                                          §c
                                          o
                                       r>j rsi
                                                                                                   occc
                                                                                                                             ooocc











O
V
(N
CC


fc
•»•*
w
*J
II

o u
* <




o
o c
•o « re
.^ a. a- c o> £
1. C C 1C C 4-.
C 10 10 C* Of Of
— £ £ O £ O It
O 6}SjC.DQtOIL'
(0 COGCC~*£
4JECCOKOL.il

C, luUUUICUE-O
C C 0 •-. — •« 1 -<
M .Q O 1 1 1 >• ^ - O

muo««rtiat--iE-




4J
a
*
10
u

H
l
0) CM
C -
3 -
O -





U
C
o
4J *-t
C 0
^8
£ <
4J
>.

£*£
Of IT.








C

X
o

1
fc
-*






0)
o
c
£ &
O. C
4J ~*
o c
< t

o
r-
CM
(D
^4
C
tr.
(L
U C
E K
1 C
£1
0 0
£ C
V jr
a u






c
c
£
a
c
c
£
O
1
(N






O
c
Q>

O
O
.C
L-'
^




QJ
C
C
fj
c
Or
C
c

u

c
°i
^




a. a
C 4-
c m
M —
C (C
Q.' £
X< •*-*
£ •§.
c •-<

o u
'5 ?
t C
- ..-*
-H Q




a-
jj
>.
a- u
c ra
E -w

£ z:
c ^
QJ >s
£ x:
.^ G.1
O E






C
Q> C
c a,
a x:
*-* a.
*D O

£ -tl
n >








c

a
c
c
XI
a




•H (-4
o c
c c
It
c c
o c
u u

H H
1 1
\c ir^
^ "^
(M fM




f-
a
g
X?

(C ~*
a> >•
t- C
r> &•
O x:
• *4 t
JC 1
H Z

o
fl
m
CC
                                                                                                                                                                      i

                                                                                                                                                                      L'
                                                                                                         334

-------
                                                O  C
                                                C  i.".
                      O
                      O
                       rsi  IN  —

                       r+  —  .-N
                                                ^-  f-  IN
                C '
                k.;
               t-l
                :•'
                C .
                              in
e-
7--

jr
rr

1 c
•—
u.

v.i c
t:
L".

^

Jj
cc
1
^~
*
^
.5








t.'.

•^
C
*"-
•c
c
|

F"
,c
c'
'_c
«_
—
y


C
m
a

^
•—

*t z: it —. tf
V "~" w •— ^^ **• ^ CT
C — C •« «* r= T •
*"" *™^ *"*" ^" * O IN L"

Ixl P* rr 
-------
                                                                      TABLE 8
                           CMA SW-846 ASSESSMENT  STUDY
Interlaboratory Precision Of Methods 8240, 8270,
ppb Labs
Method, Compound Spike Range Reporting

8240
Acetone
Acetonitrile
Benzene
Carbon Tetrachloride
Chloroform
1 , 1-Dichloroethane
1 ,2-Dichloroethane
1 , 2-Bichloroprcpar.e
Ethyl Benzene
Tetra chloroethene
1 ,1 ,1-Trichloroethar.e
Tri chloroethene
Toluer.e
1,1,2, 2-Tetr achl cr oethane
Methyl Ethyl Ketone
Isobutyl Alcchol
1 ,4-Dioxane
Acetophenone
Aniline
p-Chlcrc- ir-Crescl
Chlorobenzene
2-Chlcropher.cl
4-Chloropher.ol
1 ,2-Dichlorobenzene
1 , 3-Dichlorobenzene
Di-n-Octylphthalate
Diphenylair.ine
Methyl Methacrylate
Naphthalene
4-Nitrophenoi
Phenacetin
2,4 ,6-Trichlorophenol
2,4, 5-Trichlorophenol
B33C
Thiourea
N-Phenylthiourea
l-Acetyl-2-Thiourea


200-400
200-400
50-100
200-300
25- 50
25- 50
100-200
100-200
200
200-300
25-325
200-300
25- 75
200-300
200-4CO
50-250
200-4CO
25-250
25-250
10C-2CC
100-20D
300-350
300-350
25-200
25-200
300-500
100-250
300-500
100-250
300-500
300-500
300-350
300-350

300-500
100-500
100-500
.1 i, ;'!" , r, ,1, ,,i

A.C
• A
A.B'.C
A',B' 	
A.B.C
A.B.C
A.B.C
A,E,C
A.B'.C
A.B.C
A.B.C
A.B.C
A.B.'C
A.B.C
A.C
-
C
A.B
A.B.C
A.B.C
A,C
A.B.C
A.B.C
A.B.C
A.B.C
A.B.C
A,B
A
A.B.C
A.B.C
A,C 	
A'.B/d
A.B.C "

-
B
A '"
8330
Vu


4.73
-
1.51
2.63
1.72
2.12
2.06
1.73
2.80
1.6B
2.05
1.70
1 .'45
3.3E
2.59
"

4.21
1.61
1.75
3.4?
1.39
2.07
2.54
2.58
2.65
2.20
™
1.65
22.33
3.22
1.41
7.30

—
—
—
Ru


9.01
" -
1.79
3.92
2.15
2.69
2.77
2.17
4.30
2.09
2.75
2.11
1.7C
5.61
3.E4
~

7.63
2.31
2.21
5.67
1.59
2.e:
3.74
3.83
' 3.97
3. 05
"
2.02
80.84
5.22
1.63
16.64

—
-
—
8240   Combined ,  based  on  average variance




8270   Combined,  excluding  4-nitrophenol,  2,4,5  TCP




8330   Combined
2.31




2.40
3.27




3.46
                                             336

-------
     (10)  S,
                                    ^2
               M
where  the  summation  is  over

     0    samples  containing  the compound

     0    laboratories i and  i'  « A,  B,  C with
          triplicates  j,  j1  - 1,  2,  3

and M is the number of pairs (X.^.,, X
laboratories on equivalent  samples.
                                        j') of nonzero  values  by  independent
                                                                     o-Z2)]1/2
    Equation   (10)   is  equivalent  to  the  estimate   of   [2(a +
derivable using  standard analysis of variance techniques when no values  are
missing.    Here  a ,  and  cr2   are  respectively  components  of  between  and
within  laboratory  variance  of  J.og  (concentration  found).    S d  is   an
unbiased  estimated  of 2( cr +  o2 )  regardless  of the  imbalance  caused  by
missing values.

    For  three 8240 compounds, one 8270  compound and all 8330 compounds,  no
laboratory  or  only  one  laboratory  reported  values   on  any  sample.
Therefore,  no  estimates  for   between  laboratory  variability  could   be
derived.   The single  value for  methyl  ethyl ketone  from Laboratory B  was
excluded  from the  analysis.  Two unusually low values for diphenylamine on
different samples by Laboratory B were also excluded  as  outliers.

    The  interlaboratory  variability  (Vu)   and  repeatability  (Ru)  factors
shown in Table 8 were  calculated  as  follows using equations  (11) and (12).

    (11) Ru - exp (2Sd)

    (12) Vu » exp 2Sd//2)

The  Vu  and  Ru  values  are used as  previously  illustrated,  but  the  95%
confidence  intervals  are  for  an  observation  from one  laboratory with
respect to a  second observation at a different  laboratory.

    Ru values  for 8240 and  8270 compounds indicate that, for most compounds
analyzed  by  these  methods, results from  independent laboratories  can  be
expected  to agree  to  within  factors of  two to four.   Problems with  these
methods  are indicated by  the 11  compounds which one or  more  laboratories
failed  to  detect  in  any  samples  and   by unusually  high  Ru values  for
4-nitrophenol  and 2,4,5-trichlorophenol.   These appear to be due to factors
influencing all  determinations  within laboratories and  not  to isolated  bad
analyses.
RLL8506605
                                           337

-------
3.6 False Positive and Negative Observations

    Calculations for precision  and bias are used to define the quantitative
aspects of  an  analytical method.  There are  two  important properties of an
analytical  method  which  are used  to assess  the qualitative attributes of a
method.    These are  the  number  of  false  positive  and  number  of  false
negative observations  that result when a method  is used.   The experimental
design  for  the CMA study  allows  for  such an assessment  since the identity
of every compound  in every  sample  was known.

    The following  summarizes  the qualitative properties  for the analytical
methods with respect  to  the compounds  included in the CMA study and for the
groundwater matrix used  for all samples.

3.6.1 Method 8240  (Volatiles)

    Only  four  compounds were  spiked  into  Sample  1;  however,  two  of the
laboratories found a  total of  seven additional compounds  in one or more of
the   blind   triplicates   (6,   7,   10).     These   are  all  false  positive
observations.   This is  a  particular problem since  Sample 1 was prepared to
represent  an  up-gradient  well  that  was  supposed  to  establish background
contamination  levels  both quantitatively and qualitatively.   Review of the
data  reported  in  Appendix  A,   Table  A-l  also   indicates  that one  gets a
different   picture  for  background  contamination  depending  upon  which
laboratory  analyzed the  sample  and what sample from  the blind triplicate is
used.   Methylene  chloride was reported to be present in most  samples by all
three laboratories.   For Sample 1 (6, 7, 10), Table  A-2 in Appendix  A  shows
the  level  of  methylene  chloride exceeded  the   total  of all other  (four)
compounds  spiked   into the sample.  It  would be reasonable  to assume  that
methylene  chloride was  present as a  contaminant  even though the compound
was   not  spiked   into   the  sample.      Methylene  chloride  would  not be
considered  a  false positive observation.   The  compound was not included in
the  precision  and bias  calculations  since the  true concentration was  not
known.

    There  were a  number  of  false negative  observations  with Method  8240.
None  of  the   laboratories were  able   to  detect  the  presence  of  isobutyl
alcohol  in  any   of  12   actual  samples   (4  samples  in  triplicate)  at
concentrations that ranged from 50 to  250  ug/A.  The same was true  for 1,4-
dioxane in six samples  where the  concentration range was  200-400  \ig/l.  Two
of  the  laboratories  did  not  detect  the  presence  (6  samples)  of  acetone
(200-400  ug/A), acetonitrile (200-400 ug/A), methyl ethyl ketone (200-400
ug/Jl),  and 1,1,1-trichloroethane  (25  ug/Jl).    This  represents  a  serious
problem of false  negative results.   Of the 17 compounds  spiked into two or
more  samples,   there were  7 cases of  false negative  observations  for one or
more  laboratories.   This indicates that  the list of  compounds reported by
EPA   to be  amenable  for  analysis  by Method  8240  is  somewhat  less  than
reported.
 RLL8506605
                                           338

-------
3.6.2. Method  8270  (Semi-volatiles)

    There were a total of eight false positive observations reported by one
or more  laboratories.   In one  instance,  two laboratories reported the same
false  positive  observation  for  N-nitrosodiphenylamine and  reported  it
present  in  the  same  six  samples  analyzed.    This   observation  is  an
interesting  situation  when  two  independent  laboratories   concur on  the
presence of a  component not actually present in samples.

    There  were  16  compounds spiked  into  the  various  samples.    Sample  4
contained all  16;  Sample  3  contained  8;  and,  Sample 2  contained  8.   Only
one  laboratory  was  able to   detect  and  quantify  all  compounds in  all
samples.   Two  of  the laboratories  did  not  detect  the  presence  of  methyl
methacrylate (300-500  ug/A)  in  any of  six samples, nor acetophenone (25-250
ug/A)  in five of  six  samples.   Additional  false negative  observations
included  diphenylamine   (100-250  ug/A),  phenacetin  (300-500  ng/A),  and
chlorobenzene  (100-200 ng/A).  These false  negative observations  indicate
the list of  compounds  reported  by  EPA to be amenable for analysis by Method
8270 is somewhat less than reported.


3.6.3. Method  8330  (HPLC.  Sulfur Compounds)

    False  positive  observations  were   a  common  problem  at  all  three
laboratories when  the  method   was  followed.   One  laboratory  used a  dual
wavelength detector and  employed absorbance  ratios  to  reduce the number of
false   positive    observations   (interferences);    however,   even   this
modification did not solve the  problem completely.

    None of  the laboratories were able  to detect the presence  of thiourea
in  any  sample.    There  were   some  reported  values   for   the  other  two
compounds, but  for  all practical purposes the quantification was so poor it
would be better to consider these observations as false  negatives,  as well.


                                4.  DISCUSSION

    A  second objective  for  the CMA study  was  to  determine how  well  the
resulting data would define  groundwater  contamination.   The  results  from
the study were,  therefore, reviewed  from the perspective of those using the
data  for  an  environmental  assessment.    Because  the  identity  and  true
concentration   level  of  every   organic   contaminant   (except   methylene
chloride) is known  for each  sample,  it  is possible to evaluate how well the
resulting  data  defined   groundwater  contamination  for  this  hypothetical
situation.   This  evaluation  is  strictly applicable  only   to  groundwater
samples  resembling  those  used  in  the  study;  however,  the  evaluation
provides the  basis  for what might be  expected in other similar situations
where  the  true concentration levels  and  identity are  not known  and  other
available information  are factored  into  the overall assessment.   In  most
situations,  the  qualitative  and quantitative quality of the  data base will
be much  poorer  since  one does  not  usually have blind  triplicate analyses
done at three laboratories. .
RLL8506605
                                          339

-------
    In  Table 9,  analytical results  for  all  the  spiked compounds  for all
four samples are  classified.   For each compound spiked, a laboratory either
failed  to  detect  its   presence   (generated   a  false  negative result)  or
detected  the  compound.    Detections  are  further  classified  according  to
whether  the  reported value is within  a factor of  two of the spiked amount,
our minimal standard for quantitative  accuracy.

    Of  the background compounds  (also EPA priority pollutants), benzene and
toluene  were  quantified in  all  samples  by  all  laboratories,  1,1,1-tri-
chloroethane by  most,  and  isobutyl  alcohol  (Appendix  VIII  compound)  by no
one.    Methods   8240   and   8270  detected  and  quantified  non-background
compounds  not  containing  chlorine  in  about  half  of the  analyses;  they
detected  and quantified compounds  containing  chlorine in  most  analyses.
Method  8330  for  sulfur-containing organic compounds performed  poorly on the
three  spiked compounds; no reported  values  fell within a  factor  of two of
the true value.                           	^ J"   i '    i;  ^

    The  user of  these  data is less concerned with how well the analytical
method  performs  on compound which may or  may not be present as groundwater
contaminants  than  with the  validity of  the  particular list of   compound
names and concentrations reported for his sample.  From this point  of view,
reported results  for the simulated down gradient samples (Samples 2, 3, and
4) have been tabulated  in Table  10.

    A  total  of 531  analytical detections other than background species or
methylene  chloride are  reported  for  these samples.    63%  (336 of   531) are
for  species  actually   spiked  at  between  half  and  double  the   reported
amount.   An additional  19% of reported values  represent compounds present
but  at  levels   not  within  a  factor  of  2  (non-quantitative)   of   those
reported.   Most  of  these  reported values  are underestimates.   18% of the
reported  values  are  for compounds  not  spiked^ into  the samples.    Table 10
also  is broken  down by analytical  method  as  well as  chlorined   and non-
chlorinated   compounds.    Using  methods  8240   and   8270,  results  for
chlorinated  compounds  are  generally  trustworthy,  while results   for non-
chlorinated  compounds  include 16% and 25% false detections.  Three  quarters
of detections by  method  8330  are  false positives.

    An  evaluation of the  degree of  agreement between  independent  analyses
is  also of  interest  to a user  of  the  data.   Table 11  was prepared by
examining  all  possible pairs  of  samples of the  same type  submitted to
different  laboratories.    The  number  of such pairs is  162  (9  x 3 = 27
choices for  the  first analysis,  2x3-6 for the second)  For each set of
values  reported by  the  first laboratory,  results  of  the second laboratory
were classified as either  (1) not confirming  the presence  of  the  compounds,
(2) confirming  the  presence  of  the  compound by  an amount  differing  by  a
factor  of  2  or  more   from  the  original amount,  or (3) confirming with
quantitative agreement.

    Based  on the tabulation in  Table 11,  the  submitter  of samples to  two
laboratories  can  expect that  results reported from  the  first  laboratory
will  be confirmed  quantitatively (to within  a factor  of  2)   by the  second
RLL8506605
                                          340

-------
I
                                              O O  O CO
                                              o o     r-
                                                                      en M
O CO •*  o
   co -3-  m
                      ON O «n O O 0*
                      eo o oo o o oe
                                             o o o
                                                                      r-  o
    f»  «H eo  r-
    _<  ,* CM  i—I
                                                                                                  O O O O •-<
                                             O O O  CN
                                                    O  —i
                                                                      
*H  O .C
 C  1-1
 O  CJ t-l

 «  sr x
 u   « AJ
<  —i W
           o
           CM
           CM
                                                                                        O
                                                                                        o
          o
          o
          CM
                                                                                                                g
                                                                                               c
                                                                                               o
                                                                                               CM
                                                                                                •h
                                                                                               o
                                                                                               0
/-* o
o  o
O  CM
CM   •>
 • O
O  O
O  «-i
                                                                                        W    v^,s_,
              O  6
              «n  «
           u
           to
           »-
                                                                                            g ^

                                                                                            O J=
                                                                                           IM  O
                                                                                            O i-<
                                                                                            W" -O
                                                                                            O  I
                                                                                                      «
                                                                                                   w  c
                                                                                                   c  «
5  e
 o  o»
 o  o
 ^  tl
 o  o
       CM
o     >—
O *-s
mow
  • o  c
O C^  03
o   -.c
CM O  *J
s-' O  »
   CM  O
 W *--  »J
 C      O
 •I  « i-H

 *g-g
j: IH  a
4-1X1-1
 5 jc  w
 o  4-1  a>
 t*  W  ^J
 o  o
i-H  I-l   •>
    O CM
                     O  O
                     1-1 «rt  I)
                     •o -o  os
                     I   I  w
                     CM CM
                      •   • W  i-   •
           fi CM
           O   •
                                                                                           -C   ...«)-   .
                                                                                           t_) i-H i—i  •—i 4J  4J 1-1

-------
                                                           o o o
vot--r-«-iOCMvr>ooocovocooor->oco
   _!_,,_<     CM m CM in CM    CM CM  co    en
                                                              CM co
                                                              CM co
vococovoi-»oococooo--OOOO
in  CM  en    «c       co co        *->
O co r-
o r~ \o
 o

 CM
 in
 CM
 4)
 O
      O O
      o o
      m m

      o o
   X-NO O    x-s
   o co co    o
   in ^^X-KO
   CM       o m x-*
    *  V  w in   • o
   O  4J 4J CM O O
   o  o  03   «o m
X-M-I r-l iH O CO  •>
O *-^  B  >%O '-x O
in    .E  H •-<     O
CM  I)  4J  O ^ ?-< CO
 •  c .e  «     o <^
m -H  P..C  01  B
CM  B I-H w  B  v B
s_^  CO  >> V  •) f -H
   l-l  4J  6 lH  O.W
 «  >. U     0  O •)
 B  B  O *H iC  Vi U
•rt  If  T  ^ 4J  4J «
rH rfT  B f. £1 -H B
•H  B. T W  D. B *

£232£^£
    o
    w
    Kl
       s
                                                               T^  ^
                                                               fig
                                                             « iH iH
                                                             tl  >. >N
                                                             M  4J  B
                                                             E
                                                                u .c
                                                                «  o-
                                          342

-------
















o
*d
s
1









































«— «.
i

*
••a-
4k
co
m
CM

«A
W
IA
^
H«d
H
t^
W
G

(b
°
9S
O
M
S
»^
fc
•"H
(A
CJ







































1
B
M
i
M
H
U
(d
tH
O

(L,
^


























i
M
|3
£tU
1
CM
X
2
M
tH
S
t
w
(A
W
eti
P-


WITHIN
1
CM
t^ {^
&
CO Pb
O
z B:
w o
tA H



E
E™1
(H
(A
g

W
(A
iJ
25.








(A
tA
^3
u
£
g

1

tn C?
l«O CO
~c




fc*l? b^
O vD
\O CT*








•* sr
CM












VC O
^•4



*
*
60
B
•H
B
* «H
•O CO
0) *J
co o
B U
•H
Id Ol
0 B
T*"4 *rt
j: i-i
S5
S£

o
CM
CO

CO CM CM*
co sr ^0
N_^ >— '



0)
»< M B
IT) O O
m t^ 2








c \o sr
CM CM CM












m sr \o
CM r»-





60
B
•H 60
B B
•H -H
•on B
W 4J - co
CM CO
CO 00
£
CO
trt

.
•o
w
•o
t-l

CO B
so £

4J
o

0)
B
O
4J
•a o
« o>
•O 4J
3 V
tH -0
U
B oi
•H CO
tH
 01 *•>
•a S-
01 N «
B C —
•H •! •
1 V
U * *
343

-------
























vH
r->
w
1








































»>-\
1
2
o

en 2
Id
M H
OS Z
O W
SB
2 (x,
§5
3°

^3 *
la
P^
>- £
PQ <
C/3
§ w
*§
en CO
W
tn pc
;- H
Is
£
E 
pa H-<
« o <
Z H
S5 M
O H
£*§
W &
f-i o*
Q

en
p?j
^3
o >-
£ pa
o w 2
c^
°H§
K W 0
e K
en
g
fz





«
CR
c/:
4
G
i
I
1
i














*
O VO 1-1 OO «-"
en r* en vD








*
<_l VO Is- CM »-l
CM -^ CM CM — «













O\ 00 CM O OO
 CM CM en
, 00 00 00














•o
A)
w
T5
M Z3
5 o
01
0)
to
R)
4J
•o

0)
•0
• H
o
1— 1
M 0)
s S
4-J
0)
s
•O W
C 
• r4
*T3 !_'
c -^
3 to
0 0
W 0-
60
^ 0)
U CO
to .-<
X> tg
^-J
CO •*
3 J-*
»
^ 3 E
"2 « >H
* 0) uj
.£ 4J C
« z cS
§ *
" * *
3**

-------
 laboratory  for  49%  of  the  compounds,  confirmed  qualitatively  for  an
 additional 20%,  and  not confirmed at all for 31%.   One  in 200 quantitative
 confirmations  and  one  in 11  qualitative confirmations  (with quantitative
 disagreement) will be for compounds not actually present in the sample.


                                5. CONCLUSIONS
     The results from  the  CMA SW-846 Assessment Study revealed that even the
 best  (GC/MS)  of  analytical  procedures contained  in  SW-846 are  somewhat
 inadequate  for  the  analysis  of  compounds  reportedly  (EPA)  amenable  to
 Methods 8240 and  8270 from the Appendix VIII list.   Method  8330  was found
 to  be  completely  inadequate  for  the three  compounds  included  in  the
 study.   This observation was  confirmed  by  all three  laboratories.   The
 statement  that  SW-846  contains  the  analytical  methods  for  all  (375)
 Appendix VIII  compounds  (excluding exotics  and water  reactive  compounds)
 simply is not true.

     Review  of  results  from  the  assessment  of  bias  (accuracy,  recovery)
 indicated  that   the  three  laboratories  were  not  able  to  meet  the  EPA
 criteria to demonstrate one's  ability to generate  "acceptable  accuracy and
 precision"  for a large percentage  of  compounds included in the  study.  This
 observation  was    unexpected   since   three   prominent  and   experienced
 laboratories took  part  in the  study,  and  the samples contained  only  pure
 compounds in a rather simple groundwater matrix.

     There were a  large  number of  false positive and  negative  observations
 by all laboratories  to  varying degrees.  This  observation is of  particular
 concern since GC/MS methodology was employed  for rather  simple  samples  in a
 rather simple groundwater matrix.  With less  specific  detectors such as FID
 (GC) or  UV (HPLC),  the  problem  with  false  positive  and negative  obser-
 vations would be expected to be even more  severe.   The resulting data for
 Method 8330 supports this  conclusion.

     The  calculated  Vu   and  Ru   values,   which   express   intra-   and
 interlaboratory  precision,  revealed  precision  varies  from  laboratory  to
 laboratory, for  different  methods, and  for  specific  compounds.  These facts
 must be considered when analytical  data are being interpreted to  assess  an
 environmental  situation.

     It  is  reasonable  to  conclude  that  the numerous problems observed  for
 Methods  8240,  8270,  and 8330  for  the  samples included in  the CMA  SW-846
 Assessment  Study  may be expected for other methods in  SW-846  not  evaluated.
 SW-846  certainly  has  not  reached  the level of development whereby its  use
 should  be mandated by law.   Until all  the  methods  contained  in SW-846  are
 adequately  validated,  SW-846 is  nothing more  than  a collection of methods
 that may  or may  not  work,  and is  only suitable as a reference  document.
 Even as a reference  source, SW-846  has limitations.

     If  one  reviews  the resulting data from the  CMA study from a data  user's
 perspective, one  would conclude that  the analytical community could  indeed
 identify  groundwater  contamination,  but not-  very well  both qualitatively


RLL8506605

-------
and quantitatively.   The  problem with  false  positive and negative  obser-
vation  would  be  of  major  concern  in  determining  immediate  corrective
action, as well as the potential for future problems.

                             6. RECOMMENDATIONS

    The EPA  should not promulgate  the  mandatory use  of  SW-846  for testing

                                                -
validation is required before promulgation.

    The  EPA  should concentrate its efforts  initially to the validation and
development  of  GC/MS Methods 8240  and  8270.   The  list of Appendix VIII
compounds amenable to these procedures  needs to  be established, along with
supporting precision and bias data.

    For  Appendix VIII compounds not amenable to  Methods 8240 and 8270, EPA
needs  to validate other methods in SW-846 to confirm they actually work for
some  matrix  with  a  reasonable  expectation that  they  may work  in  other
matrices.

    Multilaboratory  studies  of   aU  methods  similar  to  the  CMA  SW-846
Assessment  Study  would  be  advisiEli  prior  to  including  a method  into  a
manual such  as SW-846 and prior to considering promulgation of  a method  for
mandatory compliance  testing.    Results from  such  studies  should  be sub-
jected  to the  analytical peer review  process and  appropriate  statistical
evaluation,  both within and  outside the Agency.
 Acknowledgement:
The authors gratefully acknowledge the work of CMA's
Environment Monitoring Task Group,

Sharon H. Kneiss, Staff Executive, and
Becky Wilson, Administrative Assistant
                                 7.  REFERENCES

 1.  Environmental  Testing and Certification Corporation  1983  "Evaluation
     of  EPA  SW-846  Test  Methods  for  Evaluating  Solid  Waste  Physical/
     Chemical  Methods  for Appendix VIII  Parameters",  Report  to  Chemical
     Manufacturers Association, Washington, D.C.


 2.  Radian  Corporation,  1983,  "Evaluation  of the  Feasibility of Analyzing
     Groundwater  For the  Appendix  VIII  Hazardous  Constituents  ,  Report  to
     American Petroleum Institute,  Washington, D.C.
 RLL8506605
                                           3*6

-------
                                8. APPENDIX A
                       Summary  of  Data  for  CMA SW-846
                              Assessment Study
RLL8506605
                                          3*7

-------

f

bl
I
t- _»
"c.
E
in


t rsi
a
E

wi
c
"&.
E
t

Z

^
££
c.
-*
E
tr.
a

1
I
8
O
2»

g
1



>.
s

&J
.
i
to
a!
1

1
t
±
S
<


„
s.
*->4

p"»
c


-c.
e •
c
i:
0. S
E I*/
ut








rsi

&
"S
e
o




»

&
s
t
1
(C
tr.

e

E
5
c

Z
E
10
"0








-
H
c.
ra


e
*£_

£
e.
E
f3
i".

1
!
in





4,
>
*£


•*
°c.
in















I
i




i
„
t
c
z


c
.
I

r\j
C
2



C-
— "*

^

"
£


c
z

§--•000 \c\rcc*<2CZ o)r-2P»r- ^vCZ^rcr
—. rs ^t m p^ JM


2 g i g i S?i55 S S i S S
— IN i-> " ,-. rsi
gSiSg SsiiSS £5i§f
_i ry •— "*> ^^ *"
g*^u?o^ circin cifCin
^ u- ' (N c r- ir js o P- jf; pJ
£ — 2 - C-CiC c*--C

roK.CCCr~' ,-imQinC \fr-CCO
^y^i2'-(CC m2i-»CC inrM2(NvO v«-«22^
r^t
Q O
022
2 *"• *^
xcc-^vD^^ mmoocc O^-TCD^T
^rsjc»-tCC L*"r*jinn*jc m«-(in2\c
in rg (N
IN — —

n^Qf^-rM O^Cff>in r- r- D D f"'




ccr*Cinr-i r*«"D^T»- r-r"CCiC
TT--2'-'r>> IT o* 2 rsi r^ r-iN22C
f-H


^-ir^DOxT n* *N C r- f- CTCC.CO
xcrM2f^O tr rg z *M \e L"fNZ2rn





(TCCCvm ^t£OCV ^CCCDC
f*i^-*2^-rw mr*j2f*''CC cc'^>22r*"
n

— • (N o cc co fM*rcr-cc r-cCdo
lf»lcC'*'>C
vCrv2fMiN ifitNZmC* mfM2—*CC
^ ^ ?
&
3
1C
>
•o
a-
4J


^
v

fl


>a
u
t3
C
Oiromo omomo o m o tr. c '^
m*Niro* ifiNinrj irtNintN
C JJ
a. T
E &
1= C-
u: c
re
C £
(M- Q^. Q>— 4-1 U
cc cm cio c
10 (0 
^c/ £&. .not or--*
jj*n *j*c aj*o •Oi-'a.
_
CO>-i OOk- OC^ JilVlV
£1.0 .e ^ o £i.c •S*J"<
co~* oo« oc« D.a>
^ u«-i.cca u^^u u-^js tn*Da'
^ -> H t 4J6Jfc>*Ha; AJOfC^HO'
HJCCiTl^ *DCC4JI^ TOCCAJI^ III
w»aa-3«H>- i^a.QJSi-«>» L*Ofap'-'>-
OK7£»X O M 3 ^ - JZ CN;:'£'*'C *""
i C — • O »-* *J ,fl C »-* O *^ 4J £C^HO«-*4J
3a.otn-4i «ajOtn»(y 
-------
s
to
                      c.
                      E
                         i
O O O
z 2 r~
D D D
Z 2 2
                                     D C
                                     2 Z
                      c.
                      E
                                  o
                                  z
                                                           c c
                                                           z o-
                               »-> IN  z  —i03"i^
                      lo
                      K Z
         o
         z
                               D C C D
                               z z z z
                      (N        —i O
                      -H        V 2
                         1
         CM *T  O
         irt pg  2
                                               C O V *->
                                               2 Z T IN
                         o
                         z
                      C        -< O
                      Z        NJ 2
                      1  o
                                                                        O        -10
                                                                        z        v z
      I
                      a
                      E
         D C C  C  C  C
         z z z  z  z  z
                                                              c
                                                           c z
Ic
IB Z
                                                                        Q
                                                                        z
                                               c
                                           I-  Z
                      •i  o
                         Z
                               VCCICDCWC
                               vCiNZrMiNZZZZ
                      a
                      V)
                                            C  O C O C C! O
                      •D

                      I
Q,
Benzer
a-
TOlUP!
tyl Alcohol
3
O
in
-Trichloroethai
1-1
Lene Chloride
£
0;

C
11
o
<
(0
I
i
1
(N
lylbenzene
*H
u"
o>
c
D
1
(jit
0 C
ro
Alk
-------
s
o.
        id
I:
li
si

I"
y
K
in z
f -
              — x v: — — o 01
                              z ™ — z
              Z Z r» —
                                      .B
                       [COOCrrir-DODC D
                       : — i^ezoirrZMZz z
                                                           Z f*l ~* C. ZZZ
                                                                             C C C DC C AT  c
                                                                             Z Z Z ZZ •» ..-  §
                                                                             c TT c ~a c -
                                                  VCCr-CCC — CCG'-lC
                                                  ^•Z^rrMrH^xCZZS;'*^
                                                           IN     cm
                                                     — (NOOr-c
                                                      -
                                                                                   §i
                                                                             c c c QC, a —
      CC"
      Li^rt-
                      — — etc
                              r- — CCCCDr.
                              — n IP z z z z z
              CCC—
        _   ,
        •S Cl  o, —
        IT. Zl
                                                                     ceo cc c c
                                                                     zzz zz z ^
                                                                             DCC CC C C.
                                                                             zzz zz z z
                                                                             C C C CC C C.  o .
                                                                             ZZZ ZZ ZZ  y.
          x   e
        c E   o


        1.1  "
        K  1
                                if. O C O O O O
                                                                                   oc c o  c c

— 0>
ro c
e. m
a » c o> £
C C 1C >O 4J
J(C C. -4 0.
£ O ^ C
4J 4J U O W
c c. a -< o
e e o £ —
W W O U £
E _0 _C *• ^ £
C £ £ £ C k.
w o Li O Of H fl;
k. D D D >• -i I>
O 1 I I £ - 3
^ r* IN, Ol +* ^ ^
£ * M. . 6f « o
o ~* - -. z -i £





c;
c
o
4^ t-t
il O
-J U
£ <
U -H

£-5
JJ O
a; w
c *-•


^
"0

>.
X
o
£
4J
' f
a. 1-1 -<
c n -
ID— C O/

C U 01
D C TJ O
1 fO 10 ^
V J£ *J ft-
- ^H U 1
- < 0 ~>





tn
•O
u
u
»-^ a:
m c
u c
4J C
3 a.
U £
V 0
01 4J
m a>
So
<


o. «' ' c c S . _ -
= £ — £ « — t C C (0=>>»a 4^~ ^ C n 2 %
WCC CX££ — OJCCI-KTBi ?v *
40101 C.XC.C C b •» c 9 0 Ol c —
k.£J30< 0;£0' U ~H 3££C£0<| rf C
UOCC C a £ — — 10 Ot-U— — Tlci ir '
i n ^ -M a a, — < — >. < **•-•-* i o £—-«.« s, c i£ ~
EOOECK'O>.£ c: £oii.(-4-Ev.? .- T -
l^^iCOCO£^u o 4JIC KT.rir ire"
C££ — -(O.W4JO'— U lci"
0)fc-uo>.ic-j2OWE*J o>xx£c^-Q.a'E. CL"*
co^'«4C£Ofc-Oc'" A:TCT
.-— ,QOo«-fc**jrg^£ 33ft»Olt-i£iD;;y .«r*t
rt£ii££c-«-oa u-iO£oox^OD.i ait.
•H u 01 1-1 a &. — z « « — -H •- a. < — £ c .- it _, !T t. — L.
Ct - * — « 10 £ 1 — 1 £ *-i 3£lll4J.-*£ r'c

-------

a


tn
         zr
         e
         13  C
         V.  Z
IL
(0
in
         1  c
         tn  z
         I"!
         f  C
         V.  Z
           d
         V. Z
         c
         ~« i-i
         c.
         E  •
         1C C
         in z
         1 6
         w z
         I i
         a

         I c
         tn z
                S-iiSiiSSSSg?      i § & 5  S § ='"E § § § g §
                      CCC-CCCC C'WCCC
                      ZTCT- TO-L">TZ->ZZZ
                        (N f   ^- .^ ry  rsj
                                                                                       — cccccoccc
                                                                                       Z. ZZfNZC-CZttZ
                                                                                              =   —  
                                                    z >?•
                      is
                               D O
                               Z CC
' O
 z
                                                                       C D T C C
                                                                       Z Z " IN Z
        O  C D Ov 1-1
        r-  Z Z    C1
O  C
^  Z
                  CCQCC'-COGOCCOC
                  ^-i2«-• C 1C 1C
•-• — ' £ 0. X ft' — 1C ft.
c c 4J 4-' c c f a ~ £
tn il i c c £ 10 £ t * £ * c 5 —
•0 c ID £ £_ _ reu^ m t.^ci' ">.*
o-i -S f-o" x &i£ -S f ??£t^ 5
^ B o f - £ o o S= .2 T £ .5 S Z 1 T i ' % fc
^££ti 2S££af2g 5 Iff? 3 St.-
W4J£-.CCUO£>,CC - it C £, o Si o>.u)ScSj
SE2'!'S'-iri:^0^x~'cu um-.-ii£.:u;. i.
S-.fe££8TlrSlf.2_,-S.2 *,t&2re£zi£t
^1-S tS g'c.'n.S2?g-55 Sgii^ffSig-
S t V i f i * » V 5 » £ g S ?2tfTfi|flJ
*ziNQTra
-------
j
              li
           Sh  c
           2i3*
              _

                  s
                                           m
8
                                                                                      gSSS§Sil?gS2g|?i8§g    iii
                                                                                      ^4^-tfMrH^in       ,H  (N  iH  ^ m M       v£)
OPPPvDPDQPPPPPOroDP   Q g

^H^iiNfH     &        ,-1 CM CM ro (N  o CD co co tn  co
                                                                                      ,H ^i 01 r-t f-i fM  rg
                                                                               «H  CM  f-4' iH QP    PQQ
                                                                                                                 n^rzz^  Z«N    ^ZZ
                                                                                                                                                            o  o o     a  c  o  a  o
                                                                                                                                                            ZZ-H     z  n  \c  z  o
                                                                                                                                                                  O        fN  C     fl
                                                                                                                                                             Q  O  "1     DOCVO
                                                                                                                                                             Z  Z  C^     Z r-  P -. C
                                                                                                                                                                            \O  m    (E
                                                                                                                                                         ODDCD
                                                                                                                                                         2  Z  Z  ^T
                                                                                                                                                                    O O O O
                                                                                                                                                                    Z ^ Z Z
                                                                                                                                                                         D O C Q  D    —£
                                                                                                                                                                         Z Z Z Z  Z     C  •*

                                                                                                                                                                                           1^
                                                                                                 n Q Q Q     c —
                                                                                                 z z z z    •;  =.
                                                                                                                                                                      01

                                                                                                                                                                     3aoo     occac
                                                                                                                                                                     C^HZZ     ZZZZZ
                                                                                                                                                                    o o  \D
                                                                                                                                                                                    TT C C Q O
                                                                                                                                                                                    ro Z Z Z Z
                                                                                                                                                             Or-t\C    fnOQCO
                                                                                                                                                             Z -H       m  z  Z  Z  Z
                                                                               Q d fi ON    i-HOCQQ
                                                                               ZZ-*       rozZZZ
                           gppppppppppinpmppoo
                           ooPtninooP    ptMpp*oomo
                        ^•^^itn        CMC*™    nmm     (n*a*o     a;
                     §£  C  O :** "•< *•* JO *^  Gl*-•* >  3 O     \
                 «^  M  w*  O t  I   I   >H £  ki   » O  ^^   *JC.QI      1*
                                                                                                                                                          P P  P  P     O P
                                                                                                                                                          PPP
                                                                                                                                                          in in  m
                                                                                          WCCIO-
                                                                                                               •it          5
                                                                                                               00^        C
                                                                                                        --*     o c  o  a*    ^
                                                                                                         c     £ !c  a;  a-  c •-<  c
                                                                                                       -J
                                                                                       a>k,uuou>,Eigoa:t*t-i<.
                                                                                                                                                                           •H  a1     t-  -
                                                                                                                                                                        of
                                                                                   >- >.
                                                                                   C  V

                                                                                  £  8
                                                                                  a  <
                                                                                         •S-   i£*?'
                                                                                         Our   i  - -  i  c
                                                                                         ».     —  ^ E •>! c
                                                                                         0  -3  >. f. -  I  -•

                                                                                                                    352

-------
                           MR.  TELLIARD:   This



morning we have a number of speakers.  The  way



we're going to work this program  to get you out



of here relatively on time is  two very important



things.  One, we have a boat break and the  boat



break will take precedence over your other  break.



The boat break happens to be that we found  out last




night that the USS Iowa is coming up the channel



this morning, and they're ain't many of us  who have



ever seen a real battleship, so we will take our



coffee break when the battleship comes up the



channel.  In between times, if you have other needs,



you're on your own.  We can't  train you to  do



everything.
                       353

-------
                          MR. TELLIARD:  Our first




speaker this morning is yesterday's speaker.  Judith



is going to talk to us a little bit about the



analysis that EPRI has been working on in their



round robin on metals analysis.  Judith.
                         354

-------
                    JUDITH SCOTT

            TRW ENERGY DEVELOPMENT GROUP
 UTILITY ROUND ROBIN RESULTS FOR THE DETERMINATION
  OF ARSENIC AND SELENIUM BY GRAPHITE FURNACE AAS
                          MRS. SCOTT:  Good morning.

Neither Winston Chow, Program Manager at EPRI, nor

Ray Maddalone, our Program Manager at TRW, could be

here today.  Ray was hoping as late as last Friday

that he'd be able to make it, but he's being detained

in Los Angeles against his will.  He is foreman of

a jury on a major trial which has kept him more

than a month longer on jury duty than anyone ever

dreamed.

     My name is Judy Scott and I have been with the

project team since the beginning of the Analytical

Methods Qualification Project.  Contrary to rumor,

I did not bribe a juror to remain a holdout, but I

am happy to be here.

     EPRI is sponsoring the Analytical Methods

Qualification Project at TRW to examine the methods

available to determine priority pollutants in

utility discharge streams.  We are concentrating

primarily on the trace metals at this point.
                       355

-------
     There has been a trend  in  recent  years  toward

basing discharge limits on water quality  criteria,

and there is concern over whether  or not  the methods

at hand are capable of determining these  low levels
         •              ,       . " , " '•  JV-'/iii ,Y  • ;I,IL , 3!1./1' if'.' "  \'.
in utility matrices.  Therefore, we are working

right now on a round robin using real  world  spiked

and split samples  to measure precision and bias of

the analytical methods in utility  streams.

     The Analytical Methods  Qualification Project
                          '!; ' '     "    ' '''I':    'i' 'I'!, ',!" 1,' •'   ". 'i ''
has four phases.   We are now in the first phase

which has two  rounds, and we have  just completed

the first round.   In the first  round we are deter-

mining arsenic and selenium  by  graphite furnace

atomic absorption.  In the second  round,  we will be

looking at  the determination of copper, nickel,

chromium and lead  by graphite furnace.  Today,  I will
                         •	  ,". :.:?J'I;J	(:.'^:':'-\'L
be giving you  a  progress report on round one.

FIGURE 1

     Our title was prepared  at a  time  when we were

very optimistic  about the  length  of time  it would

take for the participating  utilities to return
                                 " '  't  	.;  ' '''"   ' "
their results.   Unfortunately,  we  were overly

optimistic  and the results  are just now coming in

and being collated.  Although I will not have

any hard and  fast  precision  and bias data to give
                        356

-------
 you,  I  will  present  an  overview of our activities



 to date.



 FIGURE  2




      Initially we  sent  out  198  requests to  utilities



 to participate in  the project and  as  their  responses



 came  back, we screened  them.  We were looking  for



 utilities which would be representative of  the  steam




 electric  industry.   We  also were looking for utilities



 that  had positive  experiences with DMR-QA samples  in



 the determination  of these  elements.   In addition,




 we wanted utilities  that routinely use  graphite



 furnace analytical techniques.




      Our goal was  to have two groups  of 30  labs



 each, one group being fresh water  matrices  primarily,



 and one group being  saltwater.  We were successful



 on the  freshwater  labs.  However,  on  the saltwater



 side, we did not have as many labs responding.  In



 addition, we found that the saltwater labs  are



 using a greater diversity of methods  to  handle



 their problems.  Therefore, in  round  one  for the



 arsenic and selenium determination, we  had  labora-



 tories using three different methods: graphite



 furnace with matrix modification or matrix match-



 ing,  graphite furnace with  Zeeman  background correc-




tion,  and gaseous hydride.   We realize that for the
                       357

-------
seawater labs, we'll have at best a qualitative




comparison when the results are all in.



     Just a brief word about the matrices.  You'll



notice here, the first two matrices on the freshwater



side, river water and ashpond overflow, represent



an influent and effluent at a coal-fired utility.



Similarly on the seawater side, the first two



matrices, seawater and then seawater plus fireside



wash, represent also influent and effluent at an



oil-fired utility on the west coast.



     The next matrix, treated chemical metal cleaning



waste (TCMCW), was selected because it's a matrix



with disposal problems and a matrix which has



typically been difficult to analyze.



     The reagent grade water is included so that we



can compare the results of this study with those of



previous studies and it also is a measure of what



utilities can do in an interference-free matrix.



     Then,  finally, the QA/QC sample. We were fortu-



nate in obtaining vials of concentrate  from EPA/EMSL,



Cincinnati, which will be used as a measure of abso-



lute bias.



FIGURE 3



     The activities have  fallen  into  the  categories



shown.  Once we decided on the matrices and found
                        358

-------
the  laboratories, we  then  collected  the  test



samples.  The first two samples, riverwater and




ashpond overflow, were acquired at an east coast



utility, sampled by utility personnel in bottles



that we had pre-cleaned and sent to  the utility and



then they returned to us at TRW.  The other three



matrices listed were acquired by TRW personnel at



a utility on the west coast.



     For the spiking and disbursement activity, we



used a churn splitter, which was an  original design



of the USGS scaled up to meet the project needs.



We acquired six different churn splitters to handle



each of the matrices.




     Finally, in the process of anticipating the



kind of data we would want from this study, we



devised a laboratory data reporting  form, which



I'll be showing you in a few minutes, as well as a



QA/QC questionnaire to help us develop a laboratory



profile so that we would have some way,  perhaps,



when we discovered outliers, of examining labora-



tory practices in effect during analysis of these



samples.




     Finally, the analysis of the data will follow



the ASTM 2777-77 and we will be deriving estimates



of limits of detection and limits of quantitation.
                       359

-------
FIGURE 4



     This is a picture of the 25 liter polyethylene



carboys used to acquire the samples and the shipping



crate used.  The cleaning that was used for the



carboys is the same as that which was used for the



bottles on the project; 10 percent hydrochloric




acid, soak for 48 hours; 10 percent nitric acid,



soak for 48 hours.  Then the bottles were filled



with reagent grade water until just before use.



This is similar to an NBS procedure which we had



established would be the best for our purposes



during an earlier literature survey.  The cleaning



method is efficient in removing any trace element



contamination, and at the same time, doesn't overly



sensitize the surface so that later on you get



adsorption of the trace elements that you're trying




to add.



FIGURE 5



     This is a picture of our disassembled churn



splitter which we used on the project.  You see the



container, the churn disc and then the cover with a



hole for the shaft.  You'll notice that the spigot



is at the bottom.  It's modeled after the USGS 14




liter design.  The one you're looking at here is



120 liters.  We had them fabricated by Bell Arts.
                        360

-------
      The  mixer  is  extremely  efficient.   You  get  a




strong suction  downward and  then upwelling,  and



USGS  knew that  the  14  liter  size mixing  was  very



efficient.  Our next step was to determine whether



or not our scaled-up model would be  just as



efficient.



FIGURE 6




      The  next slide shows you the  result of  a  test



to determine mixing efficiency.  We  used  a water



matrix and we spiked into it a manganese solution



which we  knew would have an ultimate mixed concen-



tration of five parts  per million.   The  churn  handle




was operated at about  seven inches a second.   The



first minute, samples  were withdrawn every 10



seconds and then after that, every 30 seconds  until



the end of the  test.   Initially, you can  see that



there's actually a very strong downward  surge,



because remember, we're withdrawing  the  samples  at



the bottom of the churn mixer.  Mixing is essentially



complete within two to three minutes.  We selected



five minutes as the mix time for use on the project.



FIGURE 7




     Once the mixing efficiency was verified,  we



then proceeded to the spiking and disbursement of



actual samples.  This is a picture of the churn
                       361

-------
splitter on the digital scale, which is accurate




to a tenth of a kilogram.  We added the contents



of the carboys to the churn splitter, mixed them



for five minutes, and then withdrew samples for



density measurements.  By measuring both weight and



density very accurately, we were able to determine



volume at any given point within a few tenths of a



percent.



FIGURE 8



     This slide shows the volumetric flasks used for



measuring density in triplicate.



FIGURE 9



     This is a copy of our spiking data worksheet.



On this we kept all the relevant data in one place,



including the density data and target spiking



levels.  The target spiking levels were chosen



based on the background  levels in the samples as



well as the methods' recommended use range.  We



sent each participating  laboratory four samples of



each matrix: the background,  plus three spike levels



to cover the recommended  use  range.  At the bottom



of the sheet are the formulas used for calculating



the amounts of subsequent spike additions.



FIGURE 10



     The next  slide  shows the actual preparation  of
                        362

-------
the spiking solution.  The spikes themselves were




1,000 ppm certified standards which we in turn



checked against an independent standard.  We used a



syringe to deliver the approximate spiking solution



that we desired, anywhere from say 2 to 10 milli-



liters, and then we made an accurate weighing of



that amount.  We had determined earlier that in




the short time spent on the balance, evaporation



was not a problem.  Then by knowing the weight of



the solution and the density, we were able to



calculate the exact amount which had been added.



     For each sample, we added all six test elements-



copper, nickel, chromium and lead and the arsenic



and selenium.  This was done so that the two rounds



would be consistent in terms of any potential inter-



elemental interferences which could bias the results,



FIGURE 11




     The standards that were prepared were then



combined in a teflon beaker for addition to the




churn mixer.  You see here the teflon beaker and



top of the disc.  The mixing action tips it and



rinses it and disperses the spike throughout the



sample.  The churn was operated five minutes before



any samples were withdrawn and then continuously



during withdrawal of the samples.
                       363

-------
                                                              ,,„!,;',!„! i|. 'i	ii"!,:iSr,i'ij if1	.„'.I!I!HI:",'
FIGURE 12


     The next slide is our churning  team  in  action.


Sample preparation takes place in an  isolated


laboratory off-limits to non-project  personnel.


The bottles being filled with samples are  500


millimeter low-density polyethylene  which, as  I


mentioned, were stored with reagent  grade  water  and


emptied just prior to filling.   I should  mention


that the churn master you see on the  right reported


a marked improvement in his golf game after  this


activity, undoubtedly due to the large  number  of


bottles that we filled.


     We filled approximately 640 bottles  for shipping


to participating utilities.  We  have another couple
                                      ,'!'"' 'I!!'  » „",',  .1 '' ,

hundred for spares and QA purpose,  such as the


measurements being made to determine that indeed


the samples are stable in the concentrations that


we prepared.


FIGURE 13


     The bottles shown here have what we  call the

                           ;  •     ,   I;,.'*  ,'   : '•
pretest labels.  These labels were  designed  to


minimize the amount of clerical  work needed  during


the filling operation.  When you're  talking  about


this many bottles, you want to streamline the


filling activity as much as possible.  The labels,
                       '364

-------
 which  were  prepared  in  advance,  contain a code




 indicating  the matrix,  the  concentration level and



 the  filling order  for later traceability.  Bottles



 were removed from  storage in  groups,  a  particular



 spiking  level filled, and then  the  bottles returned



 to their places on the  shelves  as a group.   They



 were then stored in  locked  cabinets until prepara-




 tion of  the boxes  to go to  the  laboratories.



 FIGURE 14




     This is a bottle with  the  final  test label




 attached.   These labels were  also prepared  in



 advance  with the exception  of the fill  order number,



 FIGURE 15




     The code of the label  has  several  parts.   I



 can tell from the  label the lab  number,  the  round



 number,  the matrix identification,  concentration



 level, and  fill order number.




     The bottles were then  bagged,  boxed in  card-



 board  boxes,  and shipped via  two day  air to  the



 participating utilities.  With  the bottles went a



 set of instructions, a  copy of the procedures  from



 the 1983 EPA  Methods for Chemical Analysis of  Water



 and Wastes,  and a  lab survey questionnaire designed



 to determine  the overall capability of  a  partici-



pating laboratory.
                       365

-------
     I should note that we used special packaging



for the reference samples.  Those come in, I believe,



five percent nitric acid which is considered a



hazardous material, and EMSL has a waiver from DOT



for this kind of shipment.  I was told I probably



wouldn't live long enough to get a waiver for this



project since EMSL had had so much work getting




theirs.  The vials were shipped separately as



hazardous material, and I was pleased to discover



that they, too, usually arrived within two to



three days.



FIGURE 16



     I have here a copy of our laboratory data



reporting form which was a four-part carbonless



package.  There's actually perforation on the top



of the first page, which doesn't show here, which



allowed us to use this top part as a label for the



bottle.  The second and third pages went to the



utilities for recording results and then one sheet



was returned to us; the final page was kept at



TRW for our records.  You notice here, that we have



places to record all the requisite analytical data.



In addition, we asked for other  information about



the conditions of the analysis.  Again, we're



trying to anticipate our data analysis needs
                        366

-------
downstream when certain data points turn out to




be outliers.  Our goal is to be able to examine



factors which might have biased the results, so



we've requested furnace conditions, temperatures,



times, et cetera.



FIGURE 17




     This is a copy of the first page of our general



laboratory questionnaire in which we ask for infor-



mation about the equipment used, sources of standards,



the level of operator experience, and calibration



ranges that were used for the tests.



FIGURE 18




     I want to just briefly summarize what our



approach to the data analysis will be.  Once the data



are collated, we will first perform a screening



test in which we look for any values which are



less than one-fifth or greater than five times the



average value calculated by our laboratory.  At



that point, these values will be flagged and we



will contact the participant to investigate the



possibility of transcriptional or decimal errors.



I was very interested in what Samuel To had to say



yesterday when he reported that as much as 50



percent of DMR-QA errors were due to trascriptional



or decimal errors.
                       367

-------
     If we find that an error is due to what was



referred to as a data management error, that value



will be corrected for purposes of this data analy-



sis and will be included in its corrected form in



the data collection.  Any data points which we can't



reconcile by that analysis will remain in the



data base but will be flagged.



     The next analysis of the data will be for



systematic errors and we will be using the ASTM



five percent two-tail ranking, which is called out



in D2777-77.  The remaining data will then be



examined for outliers using a one percent T-test.



I believe the 1985 revision of ASTM D-2777 permits



one repeat cycle through this test and that will



be the maximum that will be used.  I think earlier



versions did not specify how many times you could



go through this kind of an outlier process.



FIGURE 19



     After rejection of outliers, we will then



calculate both single operator and overall precision



and bias for each element and each matrix.  For



the percent recovery calculation, the accepted



background concentration will be the mean of all



reported concentrations after outliers are removed.



Other concentration levels will then be calculated
                       368

-------
 from the background plus added spikes.




      Precision data will be plotted against




 concentration and linear regression used to extra-



 polate back to zero concentration.  This value



 then,  will enable us to develop an estimate of the



 limit  of detection using the standard definition




 of three times the precision at zero concentration.




 In the real world, the  precision tends to flatten



 out as you approach zero so, if anything, we'll be



 coming up with a  bit of an  optimistic estimate of



 the method's capability.  Finally, any observed



 biases will be tested for significance at the  one



 percent level.




     Currently the data that we've collected are



 being  entered  in  a dBase III program so that we



 can spin out appropriate files  as needed for the



 data analysis.  The  summary  of  the results  from



 round  one will be  reviewed with EPRI in the begin-



 ning of  May, and  if  all  goes  well,  we  will  be



 preparing samples  for round  two in May.   The summary



 report of  these two  rounds will be  prepared  and



 issued during  the  August/September  time  frame.



     A brief word  about  phase two  for  the



Analytical Methods Qualifications Project.   We



are tentatively scheduled to look at arsenic and
                       369

-------
selenium by gaseous hydride, cadmium by graphite
furnace, mercury by cold vapor, and iron by flame
AAS.  Again, I apologize for not having hard data
for this presentation, but I thank you for your
attention.
                          MR. TELLIARD:  Any
questions?  Thank you.
                        370

-------
Analytical Methods
Qualification
         Utility Round Robin
         Results for the
         Determination of
         Arsenic and Selenium
         by Graphite Furnace AAS
Raymond F. Maddalone, Ph.
Judith W. Scott (TRW Inc.)
Winston Chow, P.E. (EPRI)
                              (TRW Inc.)
                         371

-------
  Utility Round Robin
  Results for the
  Determination of
  Arsenic and Selenium
  by Graphite Furnace AAS
                     Figure 1
   Raymond F. Maddalone, Ph.D {TRW Inc.)
   Judith W. Scon (TRW Inc.)
   Winston Chow, P.E. (EPRI)
     Figure 2
      "Freshwater"
        (30 Labs)

         "i
        GFAAS
      .4
    Matrices
    River water
    Ash pond overflow
    TCMCW
    Reagent grade water
    QA/QC
                                              AMQ-I, Round 1
                                             Four Concentration
                                                   Levels
                                                                         "Seawater"
                                                                           (13 Laos)


                                                                          Z-GFAAS      GHAAS
                                                                      Matrices
                                                                      Seawater
                                                                      Seawater + fireside wash
                                                                      TCMCW
                                                                      Reagent grade water
                                                                      QA/QC
AMQ-I Round 1  Effort
Collect
Test
Samples


Spike
and
Disburse


Collate
Data


Analyze
Results
River water
Ash pond overflow
TCMCW
Fireside wash
Seawater
Churn splitter
LDRF
QA/QC
S0.St,x
Regression
equations
LOD, LOQ
                                                 Figure 3
                                                  372

-------
 Figure 4
Figure 5
             373

-------
      USGS CHURN-SPLITTER MIXING  EFFICIENCY  TEST
EC
C_D

CD
CJ
                     ^ Mn CONCENTRATION WITH TIME
                 1 TIME AFTER ADDITION,  min.
                          Figure 6
                                37*

-------
                                               Figure 7
Figure 8
                                        375

-------
             Figure 9
            Figure 10
376

-------
                                        Figure 11
Figure 12
                             377

-------
                                    " :•.•,'„(I I,!,;'
                                             Figure 13
Figure 14
                                 37B

-------
                       ANALYTICAL METHODS QUALIFICATION
                     UTILITY AQUEOUS DISCHARGE MONITORING

                          ,XXX,-XX -  XX.-.XXX.
CODE:
       LABORATORY
       IDENTIFICATION
       NUMBER
       (THREE DIGITS)
TEST ROUND
IDENTIFIER
(TWO DIGITS)
MATRIX
IDENTIFICATION
(TWO DIGITS)
BOTTLE CODE
NUMBER
(THREE DIGITS)
USE:
       IDENTICAL ON
       ALL SAMPLES
       RECEIVED BY
       PARTICIPATING
       LABORATORY:
       USED ON DATA
       REPORTING
       FORMS
IDENTIFICATION
OF STORED
SAMPLES BY
TEST ROUND
PERMITS
LABORATORY
IDENTIFICATION
OF
INTERFERENCES
SPECIFIC TO
MATRIX
NUMBER USED
AS IDENTIFIER OF
CONCENTRATION
LEVELAND
BOTTLE FILLING
ORDER
                             Figure 15
                                          379

-------
                                         EPRI  AMQ-I, Round1
                                         AMQ-l/1
                                               in-
                                         Mitriifypt:   Freshwater (Olf Ash Pond Overflow (02).
                                                     Saawiter (03). Seewater with Metal/Wastes (04),
                                                     Reagent Gride Water (05). QA/QC Ampule (06).
                                                     treated Chemical Metel Cleaning Wastes (07)
                                                              Fireact Ceadtttons, *C. Sec
                 pHelTaM     p>ta     Wn Standard
                                                            Char
                                                                          Atotnization
Aatrrtt   tt Free.   Piep.  AM!.  AddiUoa Used?  Taaip   Tmt   Temp    Ta»a    Tamp   Time
  At
  Se
                   Measured Alipat
Ekmett  Rapieete      Ceoc(^/L)
                              Dihtiex Fecter
  Si
1
2
3
1
2
3
                                                         Calcriated CMceMratian
                                                          at Ori|. Sampte
Saftple CwMilie* M Anital:
CaeMMits/PraWem:	
                                                                                                                . <(	I
                                           Blu* • PwticipMlne LatnrMorv'i Copy
                                                                           •ufl Ccrd • DuburMnwnt Copy
                                        Figure  16
                                                    380

-------
               GENERAL LABORATORY EQUIPMENT AND PRACTICE  SURVEY
LABORATORY
KEY CONTACT/PHONE:
MATRICES ANALYZED:
Apparatus
         1.    Atomic Absorption Spectrometer  (AAS):

               Instrumentation Laboratory _^^
               Perkln Elmer	
               Varlan	
               Jarren-Ash
               Fisher Scientific	
               Beckman 	
               Hitachi
               Sargent-Helen 	
               Other:
         2.
Model Number:

Type of AAS:

Single beam __
Double beam
        3.    Type of background correction used during AMQ project:

              None	
              Deuterium arc	
              Smlth-Hefltje ^^^
              Secondary line	
              Zeeman 	
              Other
        4.    Graphite Furnace manufactured by:

              Instrumentation Laboratory _____
              Balrd Atomic	
              Perk1n-E1mer
              Varfan
              Jarren-Ash
              Beckman
              Sargent-Welch	
              Other	
              Model W.   ''
                           Figure  17
                                      3S1

-------
Highlights of Data Analysis Approach
          Follows ASTM D2777-77 (1985 version)

          Screens for transcriptional errors
                   • Does not remove data unless it is
                     a reporting error

          Ranks laboratories (5% two tail limit)

          Tests for individual outliers (1% t-test)
                   • Permits one repeat of t-test
                                  Figure 18
                                 Anilflical Method!
                                 Qyiliiicition
                                 Expected Data Outputs
           Figure 19
By matrix for each element
         • Mean, S0, St, BSD (S0, St)
                   - Distribution plots

         • % recovery
                   - % recovery versus concentration

         • Regression (linear) analysis
                   - "x versus S0
                   - 7 versus S(

         • Bias (tested for significance at 1 % level)
                                                382

-------
                          MR. TELLIARD:  Our next



speaker is going to talk a little bit about one



of our favorite subject, pesticides.  Many of you



here are living off of them.  I understand Mr.



Whitlock's story where he has this little can and



goes around every Thursday and spots a small piece



of ground and then calls for the contamination.



     I forgot to announce earlier that there are



some limited copies of Paul Britton's talk in the



back on the table with the other publications.  If



you'd like one, feel free to pick one up.
                       383

-------
            MITCHELL D. ERICKSON, Ph.D.



             MIDWEST RESEARCH INSTITUTE



               KANSAS CITY, MISSOURI






     COMPARISON OF METHODS FOR ANALYSIS OF PCBs








                          MR. ERICKSON: I would



like to present my talk in three segments this morn-



ing.  The first segment reviews why we are concerned



about PCBs.  Second, I will talk about philosophy



of methods that one would use for analysis.  Finally,



I will present one answer which is what we've been



doing which we think has many advantages  over some



of the other methods that are available.  I'm going



to focus more  on  the broader scope of  method develop-



ment and less  on  the data, although  I  should empha-



size that  there is  a substantial body  of  data back-




ing up all of  the work that we've done.   I'm going



to be comparing different  methods and  talking about




them today.



SLIDE 1  &  2



     Just  as  a quick review,  that's  a  PCB.   there



are  209  PCBs.  The  group  as  a  whole, or  the  individ-



ual  members,  are  called  congeners.   If you  refer to



them in  terms of  the degrees  of chlorination,  you
                        38*

-------
have 10 different homologs.  For  instance,  tetra-



chlorobiphenyl is one homolog.  Within each homolog



group, there are anywhere from 1  to 46 isomers.



SLIDE 3




     The commercial production of PCBs.  Through



1976, 90 percent of the world's PCBs were manufac-



tured by Monsanto, sold under the trade name of




Aroclor, with a number associated with it, 1221



through 1260.  The percentage composition of these



aroclors is listed by homolog.  The point to be made




here is that these are complex mixtures.  They range



from 30 to 60 individual congeners per mixture.



That really is what separates PCBs from most other



analytes.




     PCBs are treated generally as a class, they



are regulated as a class, and analysts are asked to




do an analysis and then report on a class of com-



pounds rather than on a specific compound.  It



would be much analogous to being asked to do an ICP



analysis and then report total metals, or to report



total PAH's.  PCBs are complex mixtures in the



environment and in virtually any sample.  This



creates all sorts of analytical problems, which is



why PCBs are a little bit different in terms of



the approach to a method.
                       385

-------
SLIDE 4


     The reason that we are concerned about PCBs is


that they are present in virtually any environmental


compartment.  The one sample that I can think of


right off hand that I can be reasonably sure that


would not have PCBs in it, would be the Hope Diamond,


but anything else—you, me, anyone else—has PCBs


in them.


SLIDE 5

     As I have summarized here, PCBs are stable com-

                                       ii,  MP
pounds, and they do biomagnify through the environ-


ment, so they are of concern from that respect.


They are ubiquitous; they are found at least 5,000


meters  depth in the ocean; they are found  in the


upper atmosphere; they are found in the arctic snow.


They are found everywhere.  They are toxic, according


to some of  the old data that prompted  their inclusion


in TSCA, yet there's a  lot of evidence that they're


nontoxic.   It's a very  controversial area.  Regard-

                                         '"' V1,,1,'1:"1, • :'     :
less, TSCA  essentially  banned their production use,


manufacture, et cetera, in 1976.


     However, EPA has  seen fit  for a number of


uses to continue and so we are  still,  at this


date,  faced with a  lot  of  analytical problems.


There are a lot of  samples to be analyzed  because
                        386

-------
 of disposal by the utilities, because of environmental
 contamination, because of spills from PCB capacitors
 and transformers in your backyard that rupture, and
 we're also faced,  as I mentioned earlier, with a
 complex analysis.
 SLIDE 6
      Basically,  when one develops a method, one has
 to keep in mind  what the objectives of the various
 aspects are.   For  instance,  with the extraction,
 one wants  to  remove the PCBs from the sample matrix
 into a solvent.  There are several general consider-
 ations in  selecting an extraction technique.  Like-
 wise,  with cleanup, you're removing interferences.
 Another way to look at it  is that you are trying to
 enrich the PCB content at  the expense of  the other
 things and there are  several different things  that
 one wants  to  consider.   There are many excellent
 cleanup  techniques  out  there that probably  would
 remove the PCBs.  You  need to have  good recovery.
 SLIDE  7
     There  are three categories of  cleanup.  There's
 a gross  interference removal.  For example,  removing
 fat.   One  example of a way to  do  that  is  with GPC.
 Removing trace interferences like organochlorine
pesticides can be done with  column chromatography.
                       387

-------
Some of them can also be selectivelychemically


degraded with chromium trioxide.  The last category


is not widely applied by people who are simply


looking for total PCBs in the sample, but it  is


very important for people who are trying to do


congener-specific analysis.  In this category, one


fractionates the PCB groups  by some of their  subtle


functional differences.  You can use carbon columns


to separate these compounds  according to the  number


of ortho-chlorines.  So, there are different  levels


of cleanup.


SLIDE 8
                                      •  ','• "l;fi"'lM

     These are the three major  instrumental  tech-


niques that are used to determine PCB content of a


sample: packed column  GC/ECD, high  resolution


GC/ECD, and high  resolution  GC/EIMS.  There  are


many, many others, of  course.   Quantitation  is


where we  get  into the  real  difference  between


regular pesticide analysis  and  PCB  analysis.   The


difference  is  that we're  trying to  get  a single


number  out of  a  very complex chromatographic pro-


 file.   It depends on which instrumentation you1re


using how complex that profile  gets,  but the Webb-

                       •    ' ,    . >'•    !i 	 f :' 	1'1"'1 "j !	  I.

 McCall  and the total areas and the  Aroclor are the


 standard  way s  of doing pa eked col_umn quan tl tat ion.
                        388

-------
 The  Aroclor quantitation is,  to my way of thinking,




 the  worst  choice  of  all  because you're trying  to



 say  that the sample  contains  a  commercial product



 and  you're  trying  to say that it  contains no other



 PCBs  and that in  fact  the product hasn't  changed.



 That  works  okay if you've got a fresh  transformer



 oil,  but it falls  down very rapidly when  you start




 having weathering  of the samples.  There's a lot of



 data  that  indicates  that you  can  get as much as  an



 order of magnitude error by trying to  quantitate as



 a specific  Aroclor and ignore peaks that  might be



 from another  Aroclor or  might be  from  other PCB



 sources or  where you don't take into account the



 degradation.




     Webb-McCall  is  tried and true.  It's been



 studied in many round robins and  proven time and



 again to give very decent data.   With  GCY EIMS,



one wants to utilize the  information available.



 For instance, one can report the  data  by homologs.



One can also do congener-specific analysis if there



 is a need to know, for instance,  the presence or



absence of  say 3,3',4,4'-tetrachlorobiphenyl,



which is purportedly the most toxic of the PCBs.



So there are a variety of ways  to quantitate.
                        389

-------
SLIDE 9



     Another thing that one has to consider when



they're doing a method development is what kinds of



QC is going to be required.  This is just a list of



topic headings, any one of which has a dozen differ-



ent items to consider when planning QC.  For instance,



under method execution, one wants to have provisions



in there for blanks, duplicates, maybe standard



addition.  One wants to have proper data recording



in notebooks and so forth.  All of those kinds of



things would be under just that one heading of



method execution. •



SLIDE 10



     Now, given that background, we see that there



are many methods available for PCBs.  In fact, this



is only a partial listing.  I have a table with 45



different methods.  One might call them standard



methods.  In some of the cases, that's a pretty far



stretch of the imagination, since the methods have



not been endorsed by any standard-issuing organiza-



tion.  Buried  in this list and not explicitly brought



out are several methods that are near and dear to



this group's heart.  For instance, in the water



category, there are the methods that we've been



talking about  off and on here, 608, 625.  Under the
                        390

-------
 solid  waste  category  are  8080,  8250,  8270.   Under




 the sludge category,  one  of  them  is  625S.   So  there



 are a  number of methods that this group  is  familiar




 with,  but others probably not.  The methods  differ



 widely in terms of  the specificity.   Some of them



 are highly specific in terms  of what  options are



 allowed; some of them are quite general.  With some




 of them, the PCBs are piggy-backed with  say, prior-



 ity pollutants or organo-chlorine pesticides.  In



 other  cases, the PCBs are the only analyte  that's



 addressed.   Some of these have been highly  tested



 and validated with round robins and so forth,  and



 others basically just got written up  and have  no



 validation data behind them.



     So that is the background.  The  overall assess-



ment at this point is that there are  a lot of meth-



 ods out there for PCBs; not  enough to do every



matrix, though.  There are several matrices  that are



 not covered and time  and again I've seen people



 trying to do an analysis and shoehorn in a method



 that doesn't exactly  fit the matrix and  was  not



validated for that matrix.  That's part of where we



were coming from and why we developed our method.



SLIDE 11




     Stepping back for a second, I want  to talk a
                       391

-------
little bit about the philosophy of writing a method.

There are two general approaches that one can take.

One is to specify all of the procedures throughout

the method.  An example of that is the EPA 600 series

where there is a fairly explicit set of directions.

One of the positive aspects of this is that it gives

you an off-the-shelf method.  You can go to the

manual, grab it, send somebody to the lab and they

can do the analysis.  Some of the negative aspects

are shown there.  Basically it means that if you

have a method that's validated for a given matrix

and a new matrix comes along, you have to make

revisions and so forth.  Also, it doesn't allow a

lot of flexibility for laboratory preferences.

     The other option is to specify that the method

meets certain performance criteria.  An example of

that is the method I'll show you  in a minute.  There

are other methods out there that also utilize this

general approach to  a greater or  lesser extent.

The options allowed  can be numerous and you can

cover more analytes  in some cases.  You can certain-

ly cover more matrices.  The key  thing  is that when

you  specify  that the method must  meet  certain perfor-
                                  '';	   	,'li"'"::1:	3 < "•  '"  	" "'
mance criteria on a  per-sample basis,  you end up

with known data quality on  each  sample.  The  downside
                        392

-------
of  this  option  is  that when you go  through  the  list



of options that might be presented, you have to



write an in-house  protocol and that does  require




some up-front work on the part of an individual



laboratory.  So those are the two extremes  of



method philosophy.



SLIDE 12




     Now, changing to what we've done.  We've got



in draft  form at this point, a method that's titled,



"Analysis of PCBs'  in Liquids and Solids."   The



analytes  that can be measured are any mixture of



PCBs.  The matrices include water, solids,  liquids



—you name it.  It does utilize GC/EIMS.  It has



options  for packed or capillary column GC.  The quan-



titation  is based upon 10 individual congeners, one



for each homolog.  They were selected after an



extensive study of all of the available PCB congeners



which determined response factors, retention times.



We selected the congeners from each homolog class



that were averages.  The QC that's involved is that



on every sample, or four 1 ISC-labeled PCBs are added



to the sample and the recoveries  of those PCBs are



measured and reported along with the values for the



native PCBs.   There are also instrumental perform-



ance criteria, blanks,  and so forth, as with any
                       393

-------
other method, say with 625.
SLIDE 13
     The four 13C PCBs that were selected are shown
                                  , ' mi      "H i i ,
In the bottom half of this slide.  The asterisks in
the center of the ring denote the 13C labeling.
The exact isomers that were picked were picked for
expediency in synthesis.  We were given six weeks
to generate a gram of each of these.  Incidentally,
an interesting story aside, buying 70 grams of 13C
benzene turned out to be quite a trick and somewhat
expensive also.  That constituted the starting
material and was the entire U.S. supply at that
time.
     The rationale for choosing  these recovery sur-
rogates is shown at the top.  We wanted compounds
that are chemically similar so that there would be
no real question about the accuracy of the recovery
assessment.  We also wanted something that could be
differentiated from the analyte.  Clearly, these
compounds are no good for  ECD work because they
have no different ECD response than the native
ECDs, but in mass spectrometry,  a separation  of 12
mass units gives you a very nice clear channel to
work with.  They were not  commercially available,
so we did a custom synthesis.  The four compounds
                        39*

-------
 are now available in mixture solutions from EMSL,
 Cincinnati.   I don't have any specifics but I
 understand that there are at least three commercial
 labeled isotope manufacturers that are in some
 stage of preparation of these compounds for sale,
 so  that if for some  reason you don't want to avail
 yourself of  the free ones through Cincinnati,  you
 will  also be  able to purchase them.
 SLIDE 14
      The general flow scheme of the  method is  not a
 lot different than in many methodsi   One  physically
 homogenizes  the sample—breaks up chunks,  stirs  it,
 whatever necessary.   Then one adds the  surrogates
 before doing  any kind of  chemical processing which
 could  change  the PCS content.   At that  point,  the
 analyst has a decision to make,  which extraction  and
 cleanup  should  be  used.   Pretty  much  any  technique
 that gives decent  recoveries  is  allowed.   We have
 provided  some examples, and  I'll  get  to those  in  a
 minute.
     After the  sample  is  ready for GC/EIMS,  you add
 an internal standard.  The original internal standard
was dS-tetrachlorobiphenyl, but  that's no  longer
commercially available, so we've also added d!2-
chrysene and  iodonaphthylene as options, depending
                       395

-------
upon laboratory preference.  Conduct the GC/EIMS




analysis, do qualitative and quantitative data



reduction and report as is required by the client,



either by horaologs or by individual peaks, or just



the total number.  There is flexibility in the re-




porting.



SLIDE 15



     I mentioned that we allowed a number of extrac-



tion techniques.  These are just the general categor-



ies that are mentioned.  Literature references are



supplied.  The choice of the technique depends upon



the sample.  For instance, with a  chlorinated



organic still bottom, your only option might be



dilution.  In another case, evaporative concentra-



tion may be appropriate.   If your  sample  is a vola-



tile solvent and you're trying to  find  the PCB



content, you can simply put it on  a rotary evaporator



or  Kuderna-Danish  and achieve say  a 8500:1 concentra-




tion.



     Liquid-liquid extraction is applicable  for  water



samples  and many other  samples.  Liquid-solid extrac-



tion  is  used  for soils, et cetera, and  others.   Matrix




destruction might  be utilized for  a commercial product



such  as  an acid chloride  where you hydrolize  it  to



form  the acid  and  then  rinse away  the acid with  water.
                        396

-------
 SLIDE 16
      We've also recommended the specific extraction
 techniques listed  here  for a  variety of matrices.
 Most  of these  should  be familiar.   The  reference to
 Watts  for  the  blood and the adipose is  the  old  HERL
 pesticide  manual which  was formally authored  by
 Randy  Watts.   "MOG" is  Mills  Onley  Gaither  from .the
 classic JAOAG  publication  in  1963.   The AOAC  refers
 to their Official  Methods  of  Analysis manual.   FDA
 refers  to  the  FDA  Pesticide Analytical  Manual.
 Those methods  are  given as  specific  options.  If
 people  want to use those,  that's fine.
 SLIDE 17
     Cleanup options  are listed here.   Many of
 these are  taken from  the oil method  of  Bellar and
 Lichtenberg.
 SLIDE 18
     Quality control  is  the heart of the method.
 There are laboratory  QC requirements.  One must
 achieve certain performance criteria.   It's not
precisely specified in our method, but the labora-
 tory has to demonstrate  to  the client or the
regulatory agency or whatever, that they're achieving
performance.
                       397

-------
SLIDE 19

     GC and MS performance are similar to what's

found in 625.  In addition, one must perform routine

qualitative and quantitative calculation checks to

make sure that calculation transcription errors are

not occurring, much along the line of what was

mentioned yesterday by Samuel To.  Sample QC—the

recoveries must meet specified percent recoveries.

The acceptable recovery is something one sets up

with a  client before you start up.  For instance,

you might specify that any recovery between 50 percent

and 150 percent is acceptable.  There is some guidance

on mass spectral data quality and  it's also important

to see  that you have an internal standard  response

that's  consistent from sample to sample.   Last' but

not least, the typical blanks, replicates,  and

standard addition are all mentioned in the  QC

section.

SLIDE 20

     Now I'd  like to present  some  application  slides.
                                           i1 ''    . „' "'
This is a  fish oil.  I believe it's bluefish from

this area, the Chesapeake  Bay.   This  is  the total

ion chromatogram.  PCB region  is labeled.   They

don't really  stand out.   The  big peak at  the right

is labeled just because everybody  wants  to know
                        398

-------
what  it  is,  and  it's good  old  cholesterol.



SLIDE  21



      Same  fish sample.   These  are  ion  plots  for



tetra, penta, hexa, and  heptachlorobiphenyls.  You



can see  that  the PCBs are  standing out nicely.   I



don't  have a  slide, but  recoveries of  the surrogates



were  adequate and worked very  nicely.   This  is a




typical  routine type of  sample.  People have been



determining PCBs in fish since maybe 1969.   No big



deal.



SLIDE  22




      Now,  let's move on  to something that's  a



little bit more complex  and a  little tougher, and



that's a by-product PCB  problem.  By-product PCBs



are regulated under TSCA and there are a couple of



rules  out  in  the Federal Register.



      This  is what one sees.  This is a chlorinated



organic still bottom.  We ran  it; several other



labs ran the same sample in a CMA round robin.  The




top two traces are the M+2 and the M+4 for octa-



chlorobiphenyl, and you see we have a good match



for six isomers.  The third trace is the 13C octa-




surrogate that had, I believe, in this case, 67



percent recovery.  This  is an extremely complex



sample.  There's everything from mono- through
                       399

-------
deca-chlorpbiphenyls in it; somewhere on the order




of 80 congeners total in the sample; somewhere



around 300 to 500 parts per million aggregate PCBs



in the sample.  Very, very complex.  There are lots



of other chlorinated compounds.  An ECD chromatogram



of this sample just totally obscured the PCBs.



In fact, you can see from the RIG the complexity



of this particular sample.  A tough sample.



SLIDE 23



     We have some performance data.  This is a



table that's in the method writeup.  I really don't



want to direct your attention so much to the top--



that 's a tabulation of the possible range of LOQs.



Right here we have some intralaboratory data that



we generated in our lab.  We feel that with good



samples and good technique, you should be getting



recoveries close to 90 percent.  Depending upon the



type of sample, we've seen recoveries as low as 22



percent.  Those are worst cases.  Typically, though,



we tend to see 70 to 130.  Some of  that error is



from the MS, which has its own errors associated



with it.



     An interlaboratory study was conducted about a



year and a half ago.  The recoveries observed by



the  different laboratories ranged quite widely, as

-------
you  can  see.   There  were  some  very,  very  bad  data



in there.  As  was mentioned yesterday,  some of  this



was  clerical errors.-   Some of  it we  patched up.  For



example, when  decimal  places were wrong,  and  so



forth, we  fixed.  In the  worst case  column, with



the  690 percent recovery, one laboratory  failed to



correctly  identify the  internal standard.  If you




don't identify your  internal standard,  everything



goes downhill  rapidly.  They were given a chance to




change their report  and took a look  at  it and said,



"Yes, those are the  numbers we believe,"  so we  had



to take  it.




     With  throwing out  some of those really terrible



results, we feel like +/-60 percent  precision is



achievable, and this is for the very low  levels in



very complex samples.   One of the samples in  this



sample set was in fact  that chlorinated benzene



waste that I talked  about earlier.



SLIDE 24




     In  summary, the advantages of this type of a



method are presented on this slide.  As with isotope



dilution, although not  quite as all-encompassing,



we have data of known quality on each sample.



That's important.  The  surrogates, we feel, can



reduce the validation time.  The idea is  where you
                       401

-------
have a matrix that you've not encountered before,



you really don't know the precise way to handle it



in the extraction and cleanup, you simply try your




best guess.  If the surrogate recoveries are good,



you report it.  If they're not good, you know you



did something wrong and you can go back and do



something else, but you don't have to do a $50,000




validation project before you get started.



     We eliminated Aroclor as a calibration mixture.




It eliminates the temptation to try to make every



sample with PCBs in it be an Aroclor.  Most of the



PCBs in the environment did start out as Aroclors,



but they're weathering and as the years go by, the



profiles are going to look less and less like the



original Monsanto products.  We are quantitating



all the way from mono through deca.  We have incor-




porated flexibility that accomodates laboratory



preferences.



SLIDE 25



     It's not a perfect method.  The surrogates must



be incorporated into the matrix.  If you've got a hard



plastic, for instance, and you throw the PCBs on  the



top and don't get them down inside, you're going  to



get good surrogate recoveries, but you really haven't



incorporated them into the matrix.
                       402

-------
     We are extrapolating from 77 to 209 congeners



and there's some error associated with that.  We



have complex matrices, complex analytes and we are



having a lot of sources of error.  That holds true



for PCBs regardless of which method you're talking




about.  It's always going to have inherently more



error than a single analyte method.  Finally, keep




in mind that any method is only as good as the



person who's performing the work.



     Thank you.  Are there questions?
                       403

-------
co

CO
            0
          c
CO
O  0)
O QQ
X O
I- 0-
UJ —
LL. LL
CO
          O to
O

   O  O
LJ-I £= CQ

—•01-
Q co  0


"0 DC "5
        co
          O CO ^
         ^ 0 10
        O

        CO
               OJ  55
            p^  g
            ^    a
O

-------
405

-------





W
V
-H
l-i
O
60
OJ
4J

O

0)
3

0)

U
d
(U
1
CQ
U
Oi
w
>
O
oo
(U
Jj
ra
CJ










VO
O\ O •
-------



e
o
•H
4-1
•H
CO
O
&
S
o to
o
VI rH
re u
rH O
3 IH
U <
(U
i— 1 CD
0 E
S3 0
to
60 tw
TO O
i-J
QJ ^~N
^ ^S
<^ •
4->
CM

QJ
«
H




















1-1
o
rH
U
O

*c














O
*»o
CM
rH
CM
rH



CO

CM
rH


CM
*3"
CM
rH



, VO
rH
O
rH


03
CM
fO
rH
rH
CM
CM
rH

/— ^
to
60 1)
o a
rH -H
O VI
E 0
O rH

CM
rH
rn ID en
rH IT)





rH CM  r--
>* CM




rH fO ID rH O
rH ^f CO rH





CM Ch t-~ CM
rH ID CM





VD ON "^ ID
CM CM CM rH

O O ID -* rH
rH ID CO






O rH CM CO ""^ ID


CO
                        0)
                        9  CO
                            0)

                        (y p ,
                        U
                        a rH

                        o  u
                        <1!  OJ

                            o
                           •H
                           n
                    •t-l OO
                            OJ
                           •H
                    
                    c
                    O)
                   •o
                   'H
                    c


                   *J

                    0)
                    u
 
 TO J3
 £  00
•A* •!"(
                           a
                    QJ  PQ  O
                    pt     D

                    (U  ..   ~
                    >  aj  d
                   'HOC
                   •a*  Vl  -H
                       S  to
                       O
                    «  to

-------






CO
••J
LU
*^
LU
_J
3!
1—
J2
LU
•s
C
CO
p£
1_
*-
CO
5



fc.
<




-J
"S
c



o

1
CO
(
o



co
"S>
c

in
0
•
CM
o
•
O
o
10 o
nr- 10
1 •
O T-
*•""




1O CM
i •

T~ T"
O O
o
m
CM
•
o
^™




CO
I

10
o










CM
CO
I
o
V
O>   CO    •
in    •   &
     o   -i-
O
DC
Z
LU

^
O
0.






C

o




s
0
CD
DC



CO
O
CO

O






— . C
C^S CD
3 "S
DC 2






O
O
•o
c


To
5
"co
£
cS
o





CO
CO
K ^a
^^^
O
55



"co
5

OL
3
o
O

-------
£
"5
yz   CO
                                        jtn
                                         CO
                               o>
              ^   =        *»  T—  V*   ^
              &   Q       B*j       Q)   J"j
              O) **        *<  T3  co   ®
• • •       ^Ik  ^^  mm^        ^S   21  V^  ai^
iu       g>  co   3   ,.   o   ©  ""^   o
         J3  £   CT  *••  "c   ^  ~   E
O      JS  .2  !5   o   o   to  S   o
         co  m  3   i-  z  m  co  o

-------
Herring gull
3,530,000

i
Cotfish
11,580

i ,
Plonkton
1,830
               ffi
410

-------
O
OC

X
LU


LU
O
LU
3
0
"c
CD
"5
CO
0
"JI
to
CD
D.
E
CO
CO
2
CO
m
o
a.
emove
cc

• •
CO
0
^r
CONSIDER.
cc
LU
LU
0


C
CO
a.
•a
o
o
0
T—



Contact
•o
o
o
O
CM

Iterferences (''Cleanup
&
.2,
"cF
"o
"«
o
DC
CO

"55
15
c
15
•»-»
CO
CD
CC
«*•»
"i
.CD
JD
to
Q.
O
O
"c
CD
"o
CO
f!t

es, Equipment)
Q.
Q.
3
CO
*\
0
CO
W
o
o
U)


-------
UJ

O
            ^        Q
                            ^      Q- ^
             2*

             CO     (A
                    OW*
                    e
             0)     m

             fe    p
             0>    ™
             *•*     co
                        CO  ^
                        CD  ^
                        £  3
                       £  &>
                            "
                                     o   5.
                           .£   a)
O
0)

                   o
g     o
a     §
       O
               S  S       5  ^

               s=  "2      °   §"
                       «
                 "
                     >
                       o
                        0)

                       °-   C  S
               _  -   Q.  
                              412

-------
UJ

o

LL
O

CO

LLJ

UJ
          0)
          3
          O
          CD
      O
      Q.
      O
              E
             _3
              o
             o
                      coo
                  o
                 £
                  CO
                 O
         E
         CO
         X
         LU
 CO
      CO
      CD
      O
      C
      CD
                      CO
                      o
                      •o
             CO
             
-------
O
DC

O
CL

LU
LLJ
        CO
        I
        CO LU
 IS   O    CM

^  OC  Z
              CO
              CO
              o
              o
              o

             CO

              II

              CD
o   §.  I  S
     CO   O
            >  T5   T3
         6   «*
        QQ   .-=
                     CO

-------
 O
          0
—^   o    .w
§   -Q "5  P

O
         1-  <
 .o
 M—
 "o
 0
                     §
                                ,o
                                vt—
                                "o
                                0
                                Q.
                                CO
            E c
            0 °
           I O
                                   O •*=;
o
<
      Q
      O
      LU

      O
      O
      Q.
Q
O
LLJ

O
O
DC
X
                            CO
                            LU

                            O
                            O
                            DC
                            X
LU

-------
CO
Ill
Q CO
CO CO
z >:

8<
                   CO
                   "CO
                   >»  0)
                   "co  .£
                   c  -c
                   <  o
                   CO
                              CO
                      CO
                       CD
                       a:
                              CD
   OQ
         CO
        !3
CD
                o
                0)
            Q.
            CO
                O
           CO   O
           Q. ^
               o
    0)  W OL  -5
    c  -D  m   ®
   ^  O -sr  OC
               3  -^
                           CO
                    E  JS 1
                    CO  CO  O
                   CO  Q O
                               CO

-------
         o
         W
        O
        (3
                    W


                    ED
                  Q
                  o
         ^
     o9ooooo1_uj,_o
     LUU.UJLUUJUJUJOXOUJ
                                   0  i
                                   o  1
                                   UJ  UJ
                           o
                           o
                           UJ
CO
UJ
DC

o
UJ
o
o
DC
o_

o
cr
CO
_x
™

to
            =5     Ji
 c  t
 0)  «  .£

 £  «  S
=5  «  w
 ®  «  -^
 co  5  2,
 1^  "5  tr

^  Q.  *-

^  O  ^
    l
    (0
    o
CO  CO
                       l
                       T3
                        o
                       CO
                                                    ,n
                                                 N W
                                             jo  « »

                                             "
s
                    CO
                    2
                    to

                    §

                    o
                    CO
                                   -   Q.
           o>


           'o
           o
           CO

           kT
           0)

           to
        0)

        3
        •o
        0)
        o
        o
                CO  CM     **
     — S  S  ^  "-^     *•"
     W H  H-  <  < <  <
     2 CO  W  Q.  Q. Q.  Q.
     < <  <  UJ  UJ LLJ  HI
                                                       •  CO  *5
                  a.
                  LU
      a.
      UJ
       3
<  
-------
13
 CD


_O


<

 CO  X
                    CO
                    CD
                X  Q.
                    CO
                    0
                        cr
                        0)
                        QC
CL

O
CO
O
             O CD -^
                        O
                        o
                    _   o
             > O)  nr  •*=!
             O .£  CD   2
             LL CO tr   Q.
X
Q.

a
o
X

HI
                 o

                 ®

                £ E_
                 CO "TZ
             o _Q  w   co _0
             'S J? •*-   § .9-
               co     co
               a)     =s
               *r co rx
               >• CD ^
               CO O  CO
                      CO
                     Q
                            c *c
                   P
           '•5^50
Q
or
h-
co
             0

             CD  CD
             Q. CO

             ^o

             2,9
                 D)
                         O  O)
                        S .2
                        'o  o
                           O
                         o
                        t  o
                         CD "—•
                        D.

-------
ND SOLIDS
rf
J LIQUIDS
£.

m
o
Q_
LL
O
CO
CO
>-
rf

<--K(
eo
o>
75
o>
(Erickson (






CO
DO
O
Q.
'o
0
J5
."x
C
CO
0
">,
"co
c


"l3
Zi
o
"o
CO
CO
"co
•4—1
CO
0
.O
to
2


CO
UJ
o
o
c
.0
to
c
E"

"0
Q


geners
o
O
o
o
'"£
"c
CO
o
rmance)
3
i_
0
CL
0
Q.
CO
00
O
Q_
6
m
I

O
0


formance
trumental Per
CO
JZ






o
°0
1
CO
m




-------
RECOVERY SURROGATES

Criteria
  1. Chemical similarity (identical liquid chromatography,
     identical behavior under acid cleanup)
  2. Able to be differentiated from analyte

Best Choice

  13C labeled PCB
  But: Not commercially available

Solution:  Custom Synthesis
  (Enough for thousands of analyses)
                     Cl    Cl
                                420

-------
     ANALYTICAL METHOD
  Obtain Sample
       I
     Physical
   Preparation
 Add Surrogates
 Extract, Cleanup,
   Concentrate
       I
   Add Internal
    Standard
      I
   GC/EIMS
    Analysis
      I
  Analyst
 Decisions
  RRFs*
RT Windows
  Identify PCBs'
   Quantitate"
 Report by Peaks,
Homologs or Total
                    *QC criteria provided
                421

-------




CO
LLJ
ID
o
^s
O
LLJ
h-
0
^_
O
^£
DC
H
X
LLJ






c:
.g
+•*
J-?
tw
•s
LU
32
"o
CO
1

"5
.C7












c
.g
i«— >
_2
Q
.g
^4— <
1
x
LU

Column
-t— '
c
CD
^
b
CO









C
"o
Q^
c
^™ ""
•4—'
QJ
i_
b

c
o
"7^
D.
b
CO
CD
Q
"CO
E
o
h-



c
,g
2
•j-
0
o
g
_>
CO
Iw.
O
Q.
LU


C
n
estructi<
Q
_x
•s




c
.g
"o
CO
i_
X
"5
C7
ZJ
t
T3
'ZJ
C7
1_J

-------
CO
UJ


O
O
LU
    CO
    LU
    O
25
HO
X  co
LU       iJ,

O
o.
               0
               O)
CO
<
a.
UJ^

0
4— >
§




.— »
l£>
CM
CD
CO
0
0
•s
>
CO
Z5
O
"E
CO
a
X
0"
O)
ZJ
CO
O)
CO
0
CO

^~
1-
co
"^- ^"
'6
CO
•D
CO
0
0
CO





£>
^— '
1
T3
O
0
DO
                            CO
                            •+—>
                            -(—•
                            CO
                                    Q
                                    LL


                                    O
                                    O
                      0
                      ZJ
                      CO
                      CO

                      h-
     S?
     0
    -Q
     C
    _0


     O
                       ~  C5
                                     c:   rr*  -Q
                           O  —  «  O   c
                           ^  O  Q_  <   CO
                           £, <  _,  O   .=
                            0
                            CO
                            0
                            a.
                                CO
                      c
                      CO

                      T3
                      O
                      O
rr   ®
0  CD
Q.  ^~"
CO  -S
                               •^^   ^^   '•*'   ^^

                               LL   LL  CL   O

-------
      c
      O
      CO
      D_
          C
          g

         to
      CD
      •,   c:
      ;   a
i  s  g
      S   «   0
      0   >r   CO
      o   _g   co
      <   LL  en
              3
              DL
a.
O
<


O



c
ZJ
o
O
"O 'il
"o o
< LL



C
1
^o
O
CO
c
"E
_TJ
<£


C
E
"o
O
"0
O
CO
o

(^
-omatograp
^i
o
c
_g
*i
CO
0
E
0
Q.
"0
O

-------
ONTROL
O
QUALITY




o
o
CO
_J
£
cc
0
-Q
CO
CO
C
.g
to
.0
ormance Cert
0
QL

0
o
CO

0
D_
CO
•D
O
O

•+-<
CO
O
^H^
~o
"co
a
o
0
6



o
a
CD
D.
CO
CO
b
ID
1
o
D>
0
CO
.0
0
O
0
0?
0
to
0>
0
1—
CO


CO
o
"o
0
Q.
CO
CO
.3

Itandard Response
UJ
13
0
•4— >
JC
I
CO
"co
c
o
O



O
0
CO

-------
                                  en +  i—
                                  oo
                                  CM  CO
            r—    CM
           D      i—
            CO   O
ea®

§8
                                                       CM

                                                      0~
            i
                                                                            S
                                                                                                          CO CM
                                                                             10    ;
                                                                         + t->    o
                                                                           CM    ^^^
§

-------
CC
LU


UJ
CD

"3


"ee
o
"5L

H"



®

(0
O
  10
IO CM
q q
o 6
                       To 1
                  o CM
                 I CD co

                  +1+1
10 10  ;±
CC   «£


Q.

Ill
O

Z   o
•s   £°°
o
8§5s
    O to
    t  --
o
o
CM

A
                         1 I
                             o
                             O5
                CM
                CM
        !  I
•^i»
EC
O
• _
LL
CC
LLI
0.
Q
0
X
LU
^>

•^w
(0
O
0)
Q>







rf^
Measuremeni
IO
o






U
^
O)
c
Instrumental LOQ >
Full Scan
IO
o
T 0
o 6






.c
e: -2
"
o o
C05
CO .-
So
^•o
"S~
2 g
E-|
Deo
t
O
o
q
T—




J5
"o
CO
LL
c
_o
luS
Sample Concentra
o
o





nr
1
1
0
_J
Matrix Interference
o






I
CD
|
"o
>
.g
•*-*
o
CD
jET
0
Q.
cc
co
IO
O
•I—







'5
o
3
•o
o
•*— '
OJ
IO
+ 1
o
O5








Intralaboratory
Recovery (%)
Precision (%)
i









Accuracy
I









Interlaboratory
Recovery (%)
i i









Precision (%)
Accuracy
                     427

-------
                                                                  IE LU
                                                                  •cc E:
                                                                  O i—i
                                                                   -
e?
                      0)
                     U
                     0)
                                                                     C-4
                              428

-------
• CM CD . CM CD . CM CD . CM Gi
§ s®. s gg &j s®, i; cgp: i^
l-'-1 • • u .' • i •— i . « co • ', r •( Jm
cn® ~ CM*3 -* S3® " ffic' l7"-
™* "* CDi!






/.w
w**=.!.zC



L' -!•"» I ^

»™
>"*
*d
'
~CC
^=
~f



l/>
"x *••«•
c "*
0>
"•

•
•

-i
*


" jff



~

-"•»»«



— f
«rt
">- - T-
—

-
•

.
•




•*i*

—
•1
„,

w • "^
X
h » ^"*
•-Jl— •
.9- -"
15
(u ri. jr
_c -»C O
1 I ?
o r o
fci —•.««, X
_O ill 	 if 0)
_C "*
u
0
-§. ~^*J.g- c -^
.— -—«»«• (U
J2 f~ CU
2 4
— — 1-
"« V
s H-
-fc L
«... — «vdE
t— -^-31,
i'^-Sg-
•




X















P _ in ..

•


. •"'••MM-MI
•
"^2
— 4M
~
.
•JCS
	 — «••«••
««
X
c
0)
_c — <•
.S-
15
8
o
1 |
"o.
X














—




i_ 1 1 i tilt i i ; i • i — j — i — i — *-
M-' (ID i"--j r-.
• CM .!£, .CD . -rf
8 a i; s s H; 2 S
co r--
- »-i CM
GS G?
~ CO ID
• •.- 1 CM
.
CD
£D CD
~ CM u"
• '-i C-4
•
CD
CD CD
«• CD *B
CM •«-
. T-I CM
_

• «- Gl
CS> d'
_ ITj ..
— < CO
. *-< C-4


CD
Gi CD
~ ^H CM
•^ CM



»- Gl

_ in ..
.^J CM


i^
GI s
- 'S «•
GI CD
«-« CM
IS
CD 'S'
'Tl Tl
•^H

-------
430

-------







C/>
LLJ
C5
<
1-
Z.

>
O

•*— '
"co
cr
g
o
.§
H-
3
"co
T5
CD
1
Q.
to
"CO
O>
O

CO


CD
E
c
,o
•*— •
CO

CD
0
•D
CD
C.
CO
o
CO
_CD
"co
c?
k_
CO

CD
3
E
g
"ro
J5
"co
0
CO
CO
k.
o
o
o
CO
_CD

I
UJ
type mixtures
j_
o
o
o
"co
•— t
o
c
to"
00
£
CD
"cO
^t^
"c
CO
13
cr
c
CO
O
matrices
•*—*
1

_o
c
~
ermits applica
D.
^
5
"x
CD
U_
ences
CD
"CD
Q.
Q.
C
co

13
ccommodates
CO
£•
S
"x
CD
U.

-------
r


ROBLEMS
CL
orated. If not, surrogate recovery £
E-
o
Surrogate must be inc
•


analyte recovery.

m 77 of the 209 congeners available for
o
Must extrapolate RRF
•
CD
O)
CD
CO
CD
CD
-Q
-»-*
O
1
DC
DC
CD
CD
T5
CD
measurement. Measur

CO
•a
H—
O
CO
CD
3
O
CO
CO
T5
.CD
CO
"CO
CO
X
CD
a.
E
8
Complex matrices and
•


interpretation error.

CO
CO
o
8
TJ
O
0
O)
CO
cc
o
•o
o
"CD
•
                                             432

-------
            QUESTION AND ANSWER SESSION








                          MR. KROCHTA;  Bill Krochta,



PPG Industries.  Mitch, you mentioned that the PCBs



were stable, however, there are reports that some



of the congeners are extremely sensitive or degrade




with radiation or light.  Do you take any precautions



or have you noticed this to be a problem?



                          MR. ERICKSON:  Well,



there's no question that PCBs do have a half-life.




Some of the monos have a very short biological half-



life in, say, aquatic systems.  The one thing that



I've heard where people have worried over the years



about degradation is a concentrated sulfuric acid



cleanup.  There are some reports in the literature



that the lower PCBs can be degraded in a sulfuric



acid cleanup.  Those reports are for hot sulfuric



acid, say 50 degrees for 15 minutes.  There are



also many reports in the literature where people




have done sulfuric acid at room temperature with



monochloro- and dichlorobiphenyls and had no



losses.




     That's the one thing I know of where people



have actually done some studies.  There's some



conflicting literature, but basically if you don't
                        433

-------
heat it up, it seems like you can get away with




sulfuric acid treatment.



     With regard to photochemical degradation in the



laboratory, I'd say that's a minimal problem.  I



don't know of any data to support that.



                          MR. TELLIARD:  Any other




questions?  Thank you.

-------
                          MR. TELLIARD:  Our next



speaker is Denis Lin.  Denis Lin is from some



organization whose letters I can never remember.



Denis is going to talk about some volatile analysis
                       435

-------
               DENIS C. K. LIN, PH.D.



ENVIRONMENTAL TESTING AND CERTIFICATION CORPORATION
    ANALYSIS OF VOLATILE WATER SOLUBLE COMPOUNDS






                          DR. LIN:  Good morning,



ladies and gentleman.  By all means, get up, get



a cup of coffee; you probably need it to hear me



speak.  Before I start getting into this particular
                                        1 ' hi i!1  ' . i . „.     ' ":•

presentation, contemplating what  transpired yesterday



in terms of the presentations and statistics, et



cetera, perhaps I should change my title to, In
                                         " i,   ],!"''' i ,   ,' '    ,i


Search Of Our Experience in a Preliminary  Study of



Analysis of Volatile Water Soluble Compounds.



Before I really start again, I'd  like to acknowledge



my co-workers, Faith Dees and Dave Lessing,  who



have worked hard in generating the data  and  some



of these viewgraphs.


     As you have heard of, this  list  called Appendix



VIII compounds, we as an analytical service  lab,



always have some kinds that  come  in to  request  for



something somewhat different, like dioxin  in combat



army boots, folate  in baby plastic panties and  all



Appendix VIII compounds  in groundwater.  We have



our approaches  to Appendix VIII  compounds  and

-------
today,  I'm not going to talk about all of  them.



I'm specifically interested in a certain segment of



it, which seemingly very simple.  You say, how could




someone analyze formic acid in water or isobutanol



in water.  The first thing you think of is, I drink,



I worry about alcohol in blood, so you think, how do



you analyze ethanol in blood.  Clinical tests have




shown that when you directly inject the serum, you



get very good results.  You always heard John




McGuire preaching about direct aqueous injection on



some of these so-called water soluble compounds.



Certainly direct aqueous injection is an approach



to doing these kinds of analyses.



     However, some of the regulators don't like



the detection limits because they are somewhat



high.   In that respect, there has been some work




done through, I believe, and contracted by Athens



                , the master analytical scheme.



There is this animal called heated purge and trap




which has been suggested to us to tackle some of



these Appendix VIII compounds.  So today, what I'm



going to talk about is our experience of doing



these compounds via that method compared to say,



perhaps direct aqueous injection.




     The objective is to make sure that heated purge
                        437

-------
and trap worked in those compounds of interest, and


also, in the same process, we would like to know
                                       11 mi, ' • ','V.v1!  .. '

the linear range of sensitivity and somewhat of an


off-the-cuff error range, not a very rigorous


winsorized, whatever, statistic approach.


     In this particular study, what we have done  is,


because we want to compare it with direct aqueous
                               '•;     , ., '!  ,r  IE; r ,- •;•  .':'	i

injection, we have deliberately chosen columns that


can be used for direct aqueous injection.  If


everything works, the heated purge and trap technique


should have a thousand-fold better sensitivity,


mainly because in direct aqueous injection normally


you  inject micrometer quantities and  in  our case,

                                          '"i *  ' 'i,.,  '     L,
we inject about 5 microliters.  When  you do heated


purge and trap, you use  5 mil but  the absolute


amount of analyte is the same.  So the concentration


factor if one considers  it that way,  is  a thousand-


fold.


     The data I'm going  to show is going to talk


about absolute amount.   So whenever the  heated


purge and trap technique works, you can  factor  in a


thousand-fold better sensitivity.  You have to


compare apples with apples,  I guess.


SLIDE 2


     As George said, there are 375 compounds listed

-------
 in  the Appendix VIII  list.   Some of these are water
 soluble compounds  and  here's  the list.   The  approach
 that  we tried  to tackle  first is either direct aqueous
 injection  or heated purge and trap.   The absolute
 amount range that  we  looked  into is between  5 ng.
 to  5,000 ng.   We don't know what's  the  right number
 so  we explore  the  whole  range.   The column,  as
 stated,  is down at the bottom.   There's  a .spelling
 error in the nitrosopyrolidine.
 SLIDE 3
      The next  list...since we're doing  this, we
 said  we  might  as well  take a  look at  other things
 because other  things are of  interest  also.   These
 are very commonly  found alcohols.   Methanol  to
 butanol.   If you are in clinical analysis, you know
 that  every time  you do blood  alcohol  you  use methanol
 and isobutyl alcohol just to  check  up.   The
 column again is  at the bottom.
 SLIDE 4
      The third  list we have is a list of  C1-C4
 alcule acids.   The clinical chemists  that have
 done  this for a  while...I listen to them  and  they
always seem to have very good results.  I'd  like to
 try it and if you look at a            catalog,
they always also show very good results.

-------
When you get into the nitty-gritty, things don't



seem to be as rosy as the literature says.



So this is the list of acids we looked at and the




column we used, which was recommended by



direct aqueous injection for C1-C8 acids.



SLIDE 5



     I'm going to make the corrections on the first



one.  The rest I'm not going to make any corrections



there.  Dave Lessy is a young man who's the technician



in our group and who made all these slides and I



just didn't have the heart to make him go back and



do it all over again.  The Y axis is actually the



absolute area per nanogram, so it is not relative



response factor.  It is response factor.  The black



asterisk is the average reponse factor of triplicate




analysis.  We bootlegged this project.  We didn't



have the time or the luxury to do 7 out of 10.



We used all the data in triplicate.  The redmarks



are basically the standard deviation error range



you want to consider that way.  At the bottom of



the X axis is the absolute amount analyzed.  In



other words, 5 mL contain 100 ng. of the analyte or



100 ng. of the analyte  in 5 mils of the water,



assuming that you have  100 percent purge efficiency.
                        440

-------
     In the  ideal situation,  if  the world  is perfect,




it should be a straight line  across the graph.  As



I said, this is not, so, there are situations where



one thing is better than others, and in this particu-



lar case, this is the direct  aqueous injection of 1-



4 dioxane compared to the next one which is the



heated purge and trap technique.




     In almost all the cases, the 5,000 ng. absolute



amount is just off range as linearity goes.  Depends



on the compound, for those that are successful, it



is somewhere between 50 to 5,000 ng. absolute



amount.  We did not use any internal standard in



this study.  I'm sure you are probably thinking



what C13 internal standard can you use in this



study to improve it.



SLIDE 6




     We'll move on to the next one that will be



fairly repetitive now,  in terms of some of these



charts.  The next one,  if I remember correctly,



should be ethylcyanide  gas.  This is the direct



aqueous injection of ethylcyanide.



SLIDE 7




     The next one is the heated purge and trap.



That may be an old photo of that chart.  It




looks very straight.  It looks as though it's
                       441

-------
almost the idea case for the heated purge and  trap
          ;•                    :        ' ';'p  hi ;• ,  j;;1;!1!111;;;;;;,;1;';	 '
as far as linearity goes, from 50 to about  5,000

ng. absolute amount.
                                            p  "r •'•••." i1
SLIDE 8

     This is aereol alcohol.  It looks  like 50 ng.

doesn't seem to be appropriate.  It should  start

somewhere higher for direct aqueous injection.

The heated purge and trap of this alcohol seems  to

be better in terms of  linearity compared to the

direct aqueous injection.

SLIDE 9

     As you remember,  George demonstrated amply

yesterday, isobutyl alcohol if you use  8240 as a  way

to test without any modification is a miserable

failure.  We have experienced that and  we do  recom-

mend either use direct aqueous injection, which

gives a better result,  or mix heated purge  and

trap for  isobutyl alcohol.

SLIDE 10

     Then we move on to the next series of  the

compounds as a continuation in terms of alcohol.

Methanol.  Next one is the heated purge and trap

for methanol.  I believe that there are couple more

alcohols  that we have  done.  This  is one.
                        442

-------
 SLIDE 11




      The next one is butynol using heated purge and




 trap.   It's  very linear  in  terms  of response  factor



 across the  range.   The heated purge and trap.   I



 believe  the  next set is  the acids.



 SLIDE  12




      I started with reversed order,  somewhat,  for



 the acids.   Acetic  acid  was the only one  that  we



 have  some success  in using  the heated purge and




 trap technique.   The other  acids  that we  look  into



 simply fail.   We did not have any response, but



 that doesn't  deter  us from  looking  into the direct




 aqueous  injection  technique,  because you  have  to



 have some way,  no matter how  poor,  in some sense  of



 sensitivity.   If you are pushed,  you at least  can



 report some detection limit for the  acids.  The  next



 three  or four are all acids,  direct  aqueous injection.



 I'm not  impressed with the  failure,  but at least



 we got a response.



 SLIDES 13, 14,  15




     This is  acetic  acid with direct  aqueous



 injection, and there's a couple more.   Direct acid,



 crotonic acid.   That's all  the acids.   If I recall



correctly, formic acid is in Appendix VIII list and

-------
                                                                s;; •	jit	I;-:
our experience with formic acid has been miserable




in terms of getting any consistent result.  We



tried direct aqueous, we tried to extract with



ether.  Every now and then we get response and



then all of a sudden, for no reason you don't have



any response at all.  So, the formic acid, if you



look into the biomedical community where they worry




about the acid in urine, there seems to have



literature assess that  it can be done, and I



always thought water is simpler than urine, for



whatever reason.  Maybe the urine is essential to



have successful formic  analysis to do something




with water.



SLIDE 16



      In our  opinion, as I said—this  is  a  pre-



liminary study, using the heated purge and trap



technique,  the  range that we  think  is reason-



able  is represented  in  the second column  from the



right.  The  percentage  standard deviation is



calculated  basically by forming the  so called,



the one outside range,  an average  response factor.



Maybe George and  Paul can argue the  validity  on



using this.   As I said, this is  just some estimation



how good,  bad or  indifferent the method  is.



      The  third column,  just  for  interest's sake,

-------
 what we did is assume the direct aqueous injection




 is  100 percent recovery and  compare absolute response



 of  the heated purge and trap to the direct aqueous



 injection.




      It looks very  easy indeed.   The direct aqueous



 injection represents  100  percent  recovery.   These




 are for the Appendix  VIII compounds.   I  don't know




 why Dave didn't have  the  chance  to  calculate  out



 the nitrosalpyrolidine.



 SLIDE  17




     These  are for  the  alcohols  in  terms of heated



 purge  and trap.  Quickly,  go down and  we don't know




 why for two propanol, the  purge efficiency  is 159




 percent, but  I believe  in  reporting what you  see.



 Maybe  somebody else may not, maybe  John  McGuire



 has  a  lot of  experience in these functions  here.



 So  we  don't know why  it was 159 percent, but  if...



 other  than  that, things look very normal.



     How can  we improve this?  This  is a bootleg



project and we still have to look further.  There



 are  a  number  of things.  Obviously, if you  decide



heated purge and trap is a viable way to go,  then



we  should really dig deep  into the perhaps master



analytical  scheme which has a section on heated



purge and trap, and try to optimize the material

-------
or the purge and trap conditions.  I'm sure addition
of internal standard, perhaps the CIS label ones,
will improve the precision and accuracy.  The
chromatographic conditions, as I said, because we
want to compare apples with apples, we did not
think in terms of...for those series of compounds
we are thinking in  terms,  how can we have comparable
results and how can  we improve the purging efficiency.
     The three  items that  we have thought of  that
                                        1 ,:,,,! Wiiii	'  . , ,,. , i
might just be the change,  and I'd welcome any other
ones that you can help me  on, is  solvent Ph  adjustment
and saponification.
     As  far  as  the  preliminary  study  goes,  that's
all  I have to say,  and  I  welcome  any  suggestions
and  questions.
                           MR.  TELLIARD:   Any
questions?   You going to  let him off?

-------
                          MR. TELLIARD:  Now a



completely different subject.  We'd like to keep




this meeting still somewhat ressembling industrial




waste chemistry, so our next speaker is going to



talk on one of our...this is for all you old shit-




chemists, so many of you people out there with



GC/MS don't even know what one of these things is




let alone what the letters mean.  But our next



speaker is going to talk about the basic shit



chemistry, commonly called a BOD.
                       447

-------
1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25
                  JAMES C. YOUNG

              UNIVERSITY OF ARKANSAS
         DEPARTMENT OF CIVIL ENGINEERING
    DETERMINATION OF FIVE-DAY CARBONACEOUS BOD
                  IN WASTEWATER
                          MR. YOUNG:  Good morning.

I feel a little bit out ofplace...

                          VOICE FROM AUDIENCE:   The

ship is coming!

(WHEREUPON, a 10 minute break was  taken  to view the
battleship.)

                          MR. YOUNG:  This  is  the

first time I've ever been in competition with  a

battleship.  I guess I should have known who'd win.

When you stop and think about it,  that battleship is

probably older than most of  us  here.   I'm not  sure

when it was first built; it  was a  few years  ago.

Any historians here that know?   I  expect it was

built in the late 30's or early 40's.
         (Revised presentation submitted.)

-------
449

-------
          BIOCHEMICAL OXYGEN DEMAND:




        MEASUREMENT AND INTERPRETATION
                      By
                James C. Young




    Professor and Head of Civil Engineering



University of Arkansas, Fayetteville, AR  72701
       Prepared for presentation at the








   Eighth Annual Analytical Symposium on the



  Analysis of Pollutants  in the Environment,




                  Norfolk, VA
                April  3-4,  1985
                             450

-------
                         BIOCHEMICAL OXYGEN DEMAND:




                       MEASUREMENT AND  INTERPRETATION




                                     By




                               James C.  Young




                   Professor  and Head of Civil  Engineering




              University of  Arkansas,  Fayetteville,  AR   72701








                                INTRODUCTION








    The Biochemical Oxygen Demand  (BOD)  test is probably the least




understood and most frequently misused  of  any  analytical measurement  in  the




field of water pollution control.   Yet  the proper measurement and




interpretation of BOD data is highly essential to the evaluation of the




performance of wastewater treatment plants and to determining the impact of



waste loads on receiving streams.




    Not only must the wastewater analyst know how to measure BOD properly,




but plant operators and others involved in pollution control must know how




to use the results of BOD tests.  They must have a basic knowledge of the




factors affecting the accuracy and precision of the measurement technique




and must understand the basic factors causing the test results to deviate



from anticipated or normal  values.




    The purpose of this paper is to review the BOD test procedure,  to




discuss factors affecting its accuracy  and precision, and to present recent




developments in analytical  procedures,  specifically nitrification control,




that affect the interpretation and use  of BOD  data.

-------
                             THE BOD REACTION



    The basics of the BOD test are simple: ' organic materials are



decomposed in the presence of oxygen to form carbon dioxide and water
                                                      1'   I'll ,ni!,';":'


(energy and products) and to synthesize new cells (Figure 1).  These



biological cell solids in turn require "oxygen for respiration and can decay



and release free organic matter  that  is  recycled  through  the



oxidation/systhesis  process.   This biological process  is termed



heterotrophic growth and the resultant oxygen uptake represents Carbonaeous



Biochemical Oxygen Demand (CBOD).



    Reduced inorganic—specifically, ammonia (N!^), nitrite   (IK^), and



sulfide  (S= )—also are oxidized biologically to create  energy for



synthesis of carbon dioxide to biological  cell solids (Figure 2).  This



latter,  or autotrophic, reaction can occur simultaneously with the



carbonaceous reaction, separate from the  carbonaceous reaction, or may not



occur  at all  in  a given sample. The BOD test provides a  measure of the



amount of oxygen consumed in  these biochemical reactions and  can, if
                  ii


measured properly, provide  a  reasonably accurate  indication  of the amount



of  biodegradable organic matter and reduced  inorganics present in the water



sample at the  beginning of  a  test.



    The  oxygen uptake  reaction  is  time dependent  as illustrated  in Figure 3



and the  accuracy of  BOD measurement is a  function of  the time at which  the



measurement  (reading)  is  taken  and the rate  and  extent  of completion  of  the



two major biological  reactions. The challenge  of  BOD  measurements  is  to



control  or manage  the  factors contributing to  the rate  and amount  of  oxygen



consumed up  to the  time  the oxygen uptake is measured,  and to make sure



that  tests conducted at  different  times  have a common basis of  comparison.
                                            452

-------

Organic
Materials

New
Syn

Cell
thesis
L ^
i
Energy
1
1
1
1
1
1 *
Oxidation
Biological
Cell
Solids


Energy
Fnd-
Products




Endogenous
Respiration
and
Decay
Organic
  Material
H-
                        C02  + -H20  +  Cell Solids
Figure 1.  Schematic diagram and  equation  expressing  components
           of heterotrophic growth.

-------
Reduced
Inorganic
NH3, NC-2, S~"

Carbon
Dioxide
a) NH3 + 3/1
b) HN02 + I
Net: NH
Energy 	
\
1
'«• i .'' 	 'i
1
i
i
i
i
T 	 _^—
Synthesis
11
It o ^- TTNO
12. 0_ 	 ^ 3
+ f) r\ 	 \Rar- HNH -f-
1. (J ^^ tllNU^ ^
Oxidized
Inorganic
NOl . NO- , SO,
2 3 4
'' j ;• , ' ^
vr / , , •' I11:
Biological
Cell
Solids
,. !!» i! 	 <
H2°
Figure 2.  Schematic diagram and equations experssing
           components of autotrophic growth, specifically
           for the nitrification reaction.

-------
I?
o
o
                                        Ultimate Oxygen Demand  (UOD)
                                                                  NOD
                                                  Nitrogenous
                                                    Demand
                                                                 CBOD
                                                                     u
                                                  Carbonaceous
                                                    Demand
1 - 1 - 1
                                i     t    i
                                                   10
                                  TIME,  days
   Figure  3 .   Schematic representation of a typical BOD curve showing
   carbonaceous and nitrogenous oxygen demand reactions
                                                              15
                                             455

-------
                           THE STANDARD BOD TEST
                                   •'       •      : :   'i: "-''Vd1 iir ' i- .•':•' .<•   '••''•"•:•  ; '
    The standard (traditional) procedure  for  measuring BOD is  the  dilution
                                                    lil,u, • «, i',,,!, i|||,     ,   ', ,     ,' |i

method (Standard Methods. 1985).   In  this method,  a sample of  wastewater is

diluted with a standard mixture of distilled  water and known amounts of

nutrients and buffering agents.   The  purpose  of dilution  is to reduce the

total concentration of oxygen  demanding material  in the diluted sample to

below about 10 mg/1 so that  the amount of dissolved oxygen exceeds the

5-day oxygen uptake capacity of the diluted sample. The diluted sample is

then placed in a "standard", usually 300  mL bottle, sealed or  capped, and
                                                         ' "Ijjj    '     '. ',, 	., ,;,   ' " '"
incubated at 20°C  in  a dark  room  for five-days.  The reduction of dissolved

oxygen concentration  in  the  five  day-test period  is then multiplied by the

dilution ratio  to  give a measure  of 5-day biochemical oxygen demand  (BOD).
                                        i        ,      .11',',.        I '      ,' '!, |	!,

    The plague  of  the BOD test is that the largest sources of  variability

are not analytical.   Almost  any experienced analyst can prepare ah

acceptable  dilution water,  can make accurate dilutions, can measure

dissolved  oxygen relatively precisely and accurately, and  can multiply  by  a

dilution ratio.   The  problem is that the major sources of  variation  are

biological  and largely  uncontrolable by  the analyst.  And  while these

variations  are natural,  that is they represent inherent characteristics  of

 the  sample,  it is difficult to compare the results of one  test to the

 results  of  another.   The need, then, is  to understand  the  factors

 contributing to variability so that  the  resulting test data are more easily

 understood and provide  a more consistent measure  of a known biological

 reaction (see Young,  1984).

     The  goal of committees assigned  to improve the standard BOD  test has
                ji         "               '  i       '   ; ;  ?;., ',: '• :. .  .,.'..'  ^ •.  ;','.•
 been to  establish procedures that minimize variability and maximize the
                                            456

-------
 uniformity between test procedures.  Over the last 40 years (about 1940 to




 1980) the Standard Methods BOD Task Group has made relatively minor




 refinements in the method of preparing dilution water, in the buffer




 formulation,  dilution procedure,  the seeding procedure and the precise time




 of starting and ending a test (Young,  McDermott and Jenkins, 1982).




 However,  the  largest source of test variability — the nitrogenous oxygen




 demand (NOD)  reaction — has been largely ignored.  The effect of the




 nitrification reaction was acknowledged in the 1930's in studies of the




 impact of waste loads on receiving streams,  and its impact on BOD reactions




 was well  documented in the late  1940's  (Hurwitz,  et al,  1947).   Methods




 such as acidification or pasteurization were even proposed for  the




 inhibition of  the  nitrogenous oxygen demand  reaction in BOD tests (Sawyer




 and Bradney,  1946).




     The impact  of  nitrification on the  variability of test data is




 illustrated in  Figure  4.   Shown here is  a  correlation of  5-day  BOD as




 measured  by electrolytic respirometer (EBOD)  and  5-day dilution BOD.   In




 Plant  A,  nitrification was  inhibited in  both  test  procedures; at  Plant  M,




 nitrification control was not  used in either  test.   A major  part  of the




 variability was attributed  to  the  fact that the nitrification reaction  was




 at  different stages of completion  in each  test at  the  end  of the  5-day  test



 period.




    Failure to separate carbonaceous and nitrogenous BOD can lead  to gross




 errors in calculating the impact of wastewater discharges  to receiving




 streams.  For  example, consider Figure 5. Shown here is a BOD reaction




exhibiting both carbonaceous and nitrogenous oxygen demands.  With




nitrification  inhibited,  the carbonaceous BOD reaction would exhibit a
                                           457

-------
   450
8
s
       O RAW INFLUENT
       • AERATOR EFFLUENT
       • PRIMARY EFFLUENT
       £ FILTER EFFLUENT
       o FINAL EFFLUENT
         SLOPE = 1.17
            •7.89
300 -
   150 -
   450
           O RAW INFLUENT
           • PRIMARY EFaUENT
           o FINAL EFFLUENT
              SLOPE =1.06
              B =18.41
              R°= 92.44
                                                      PLANT A

                                                    (With Nitrification
                                                     Control)
                                                    PLANT M

                                                 (No Nitrification
                                                  Control)
       0       80
                                 DILUTION BOOj, ma/I
        Figure 4.   Comparison of electrolytic and  dilution  BOD
                    measurements showing  the effect of nitrification
                    on the variability  of test data.
                                              458

-------
   200
I?
Q
O
PQ
  100 -
                                      TIME, days
    Figure 5.  Comparison of measured and calculated BOD curves and

               reaction coefficients for a wastewater sample with and
               without nitrification inhibition.
                                             459

-------
curve represented by the solid line which would have a relatively high


first-order rate and a reasonably low ultimate carbonaceous BOD (CBODU).


With nitrification, the best-fit, first-order equation (dashed line) would


show an apparent decrease in rate coefficient and an increase in ultimate


oxygen demand (CBODU + NOD).  Obviously the use of these two different sets


of reaction coefficients to represent the BOD reaction for the same


wastewater would have a dramatic effect on the results of an oxygen


depletion model using uptake rate and measured 5-day BOD as input


parameters. The impact of nitrification in BOD tests on treatment plant


compliance  is well-documented by Hall and Foxen  (1983).


     The  reasons  for not  including nitrification  control.in BOD tests  in the


past seem to  have  been  both technical and non-technical.   The methods


available were  cumbersome  and did  not improve the accuracy of  the  test


 since the carbonaceous  BOD was  adversely  affected by pasteurization and


 acidification,  and reseeding was required.   This changed  both the  chemical


 and biochemical nature  of  the test sample.   But  perhaps the  largest reason


 for not separating carbonaceous and nitrogenous  BOD in tests  is  related to


 the purposes for which BOD was  used, especially in treated effluents.


 Prior to 1972,  when major  amendments to the Clean Water Act  (PL-92-500)


 were passed,  levels of treatment were based largely on process technology.


 That is, communities were required to have primary or secondary treatment
                                                   1 .1', '! H

 based on size and location of the city, and secondary treatment was  defined


 as a "biological process".  Hence, trickling filters, activated sludge,


 rotating biological contactors, lagoons, etc.—were essentially equivalent


 and the choice of one over the  other was largely based on economics  and


 operability.  PL-92-500 led to  the definition of secondary treatment in


 terms of effluent quality, that is, 30 mg/1  BOD  and suspended solids on  a


 monthly  average basis,  45  mg/1  for  a peak  7-day period.   Too  few people



                                             460

-------
 realized at the time that so called "secondary processes" could not




 consistently provide "secondary effluent" quality unless the measure of




 effluent quality was carbonaceous BOD.




     However, the standard method for measuring BOD—as developed by




 Standard Methods and accepted by the U.S. Environmental Protection




 Agency—did not allow for separation of the carbonaceous and nitrogenous




 components of the BOD reaction.  Thus began a 10-year debate of the merits




 of and need for basing treatment performance on carbonaceous BOD alone or




 carbonaceous plus nitrogenous BOD.   The arguments were based largely on the




 premise that nitrogenous oxygen  uptake is in fact biochemical oxygen demand




 and the fact that the multi-decade  base of BOD data did not include only




 carbonaceous BOD measurements; and  decisions for  granting discharge permits




 were based almost entirely on calculations using  past  BOD records.




     After  considerable debate, numerous lawsuits,  and  significant




 expenditures of  funds for  construction of treatment  plants  to provide




 "secondary"  or better quality effluent,  EPA  has accepted  the  use of




 carbonaceous BOD  as a measure of permit compliance,  although  in restricted




 situations and subject to  state approval.  Consequently, we now are  where we




 should  have  been  10 to 12  years ago  in  BOD measurement  and application



 technology.




    The challenge now is to precede as  rapidly as possibly to using




carbonaceous and nitrogenous BOD as independent performance references for




biological treatment plants, thus requiring a thorough understanding of how




to conduct BOD tests involving separate measurements of carbonaceous and



nitrogenous BOD.
                                           461

-------
                       NITRIFICATION CONTROL METHODS

    The accepted method for measuring carbonaceous and nitrogenous BOD is

to inhibit the nitrogenous oxygen uptake reaction in one set of samples

thereby  leaving only the measurement of carbonaceous demand. The

nitrogenous oxygen demand is measured by setting up a parallel set of

samples so that nitrogenous oxygen uptake is the difference between

inhibited and uninhibited samples.  An alternate method is to measure the

amount of reduced nitrogen in a sample—either as ammonia-nitrogen or Total

Kjeldhal Nitrogen (TKN)—and calculating the anticipated NOD using the

equations stated in Figure 2.  This latter method is considered to be as
                                                      •  . '!'ii': ''
accurate as direct measurement in BOD bottles and much  simpler and faster.

    The presently accepted method of inhibiting the nitrogenous reaction  in

BOD tests is  to use chemical inhibitors of  the nitrification reaction.

While  pasteurization or  acidification have  been used sporadically in  the

past,  these methods are  cumbersome  and  are  not as reliable as  is  chemical

inhibition.   Two chemicals  can be used:   allythiourea   (ATU)  or

trichloromethyl  pyridine (TCMP).  Allylthiourea seems  to be the favored

chemical  in Europe, and  while effective it  does  have  some  disadvantages.

First, it  is  biodegradable and  its  oxygen demand can  be a significant part

of the total  BOD of  samples having  large dilution ratios.   Secondly,  since

 it is biodegradable,  it  loses  its effectiveness  after a few days and must

 be replenished periodically in  long term tests.

     The favored chemical in U.S.  methods is 2-chloro-6-(trichloromethyl)

 pyridine.   This chemical is highly specific in inhibiting the ammonia to

 nitrite nitrogen conversion and does not adversely affect carbonaceous

 demand reactions (Young, 1973,  1983).   It is quite stable in water and it

 is effective for up to 30 days in BOD tests. One disadvantage of TCMP is

 that  it is not soluble in water and must be added to samples as  a powder.


                                            462

-------
 This is not a major problem,  however,  because commercial dispensers are




 available to provide accurate doses to individual BOD bottles.




     The effectiveness of  TCMP as a nitrification inhibitor has  been




 demonstrated in both dilution and respirometer tests (Young,  1973,  1983).




 Typical results are shown in  Figure 6,  and nitrogen and oxygen  balances are




 shown  in Table 1  for a number of test  cases.  In all cases,  TCMP has been




 shown  effective for nitrification control  and there is no evidence  that it




 interferes  with the carbonaceous BOD reaction.








                ACCURACY  AND  PRECISION  OF  BOD AND NOD TESTS




     One  problem facing BOD measurements  is  that it  is not possible  to  know




 the  stoichiometry or  completeness  of the reaction at any time so  that  there




 is no  standard for  establishing  accuracy.   Standard Methods (APHA,"  1980)




 gives  a  procedure for  checking the  dilution procedure and for identifying



 problems with  seeding  or measurement technique.   This involves  adding  5 ml




 of a solution  containing  50 mg/1 each of glucose  and glutamic acid  to




 seeded dilution water  and measuring  the 5-day BOD.   The  theoretical oxygen




 demand of this mixture  is 308 mg/1.  If the measured  BOD5  falls  within  a




 range of 220 + 20 mg/1 after seed correction, the analyst can feel with




 some confidence that his procedure  is correct and that there is no toxic




 substance in the dilution water.   This, however, does not establish a



 method for determining the accuracy of the test.




    Test precision reflects the ability to repeat a measurement in a set of




 replicates or among tests conducted under a given set of conditions.   It is



 important when establishing precision that a common or known basis of




 reference is used.  For eample, it should be made clear whether the




analysis of precision includes sampling, seeding, and transfer errors or




simply reflects the variability between replicates after samples have been

-------
                                           FINAL EFaUENT
                                            FINAL EFFLUENT
                                             TCMJ
                                            PRIMARY
                                            EFFLUENT
                                            PRIMARY
                                            EFaUENT -f KMF
                                            RAW INFLUENT
                                            RAW INFLUENT
                                            + TCMP
                          TIME, days
100
Figure 6.  Examples of BOD measurements with and without
          nitrification inhibition (From Young, 1972)

-------
Table 1.  Measured total and carbonaceous BOD and measured and calculated
nitrogenous oxygen demand (NOD).
          (From Young,  1973).
Sample

•
Primary
effluent
Trickling filter
effluent
Activated
sludge effluent
Trickling filter
effluent
Trickling filter
effluent
Trickling filter
effluent
BOD
Total


328

80

75

297

100

135
i mg/1
Carbon-
aceous

224

25

16

95

38

84
^Measured NOD = Total BOD-Carbonaceous
"fal 1-11 1 af o^ Mnn — / MU . _«_ Mn — s .. / o •> .
NOD
Measured3


104

55

59

202

62

51
BOD
, mg/1
Calculated5


96

56

56

224

58

51

                                         465

-------
transfered, seeded and diluted.  Measurement precision for a given analyst


can be consistently below 7 percent if sampling, dilution and transfer


errors are eliminated and nitrification is controlled.  This verification


would include natural biological variability between samples plus errors in


measuring dissolved oxygen concentration.  Another contributing factor is


that the error in DO measurement is multiplied by the dilution ratio so


that considerable error may be introduced when measuring the BOD of a


high-strength sample that requires a large dilution.

    A common technique for monitoring BOD test methodology is to conduct


interlaboratory analyses of mixtures containing known amounts of organic


materials  (typically glucose-glutamic acid). Such samples are sent to a


number of  different analyst who are asked to measure  the dilution BOD using


their normal analytical procedure at 2,  3, 5, 6-day or  similar intervals by


using a  stated method  for seeding.  Typical results of  this  type of

analysis are shown  in  Tables  2 and 3.  While interlaboratory testing does
                                                        ' • 'I •   ,   "  ', '  ,

not give a good  indicator of  precision of BOD measurements  in a given


laboratory or treatment plant environment,  it does give an  indication of


the variability  of  analyses  from plant to plant.




                        THE  FUTURE OF THE BOD TEST


    The  BOD test has  been  in use  in  essentially its  present form  for  about


60 years.   Throughout this  period,  there has been considerable  criticism


about the  soundness of the  test  procedure and  its use as a pollution

control  measure.  So  why does the  test  continue to  be used?  Basically,  the


 reason is  that  no acceptable substitute  test has been developed that


 responds to and provides a measure of the amount of biodegradable matter


 present in a wastewater.
                                            466

-------
 Table  2.   Precision  of  BOD5  measurements,
 Parameter
                                          Reference  Source
                      Method Research  Study  3,   Ballinger  and      Young and   ~
                      EPA,  1971  (Dilution Test)    Lishka,  1962     Baumann,  1976

                                                   (Dilution test)   (Respirometer)
                      Low-level     High level
Theoretical
value, mg/L
Mean of measured
BOD5 values, mg/L
Recovery, %
Coefficient of
2.2 (BOD5)
2.12
96
33.2
194 (BOD5)
175
90
15.0
308 (THOD)
214
69
19.5
350 (THOD)
296
85
5.8
No. of analyses/     74/56
 No. of laboratories
73/56
34/34
                                   4/4
                                          467

-------
Table 3.  Intel-laboratory analysis of 5-day BOD as conducted  by  the  U.S.
Environmental Protection Agency (From Britton, 1985).
Added Glucose/ EPA & Stat
Glutamic Acid Laboratori
Study (50/50) in mg/L Reporting
•J UMMT \ w v
WP003
WP004
WP005
WP006
WP007
WP008
WP009
WP010
WP011
a Coef. of
From these
X = 0
S = 0
37.0
340
6.0
325
4.0
176
101.3
24.2
6.0
144.0
39.1
89.6
4.0
231.0
114.0
161.0
5.0
192.0
Var. = Std. Dev./X
data (WP007-WP011)
.665 (added level)
.0998 (added level)
98
102
85
85
102
102
90
95
94
95
68
70
65
68
69
69
74
75
, %
e . Statistics
es Mean
(X. mg/L)
25.47
229.9
4.270
218.7
2.734
118.8
68.97
16.81
4.09
100.0
26.7
61.5
2.91
146.0
75.4
103.6
3.56
125.2

the following linear
+ 0.225;
+ 0.430
with an R2 =
2
; with an R
Std. Coef .dof
Dev. (S. mg/L) Var.
• "; 'lii'" '. 	 I*
4.299
37.06
.8710
33.36
.6672
17.05
9. 785
2.131
1.0662
12.603
3.377
6.609
0.9528
27.257
10.408
17.850
0.5500
14.714

relationships
0.99+
= 0.99+
16.9
16.1
20.4
15.3
24.4
14.4
14.2
12.7
26.1
12.6
12.6
10.7
32.7
18.3
13.8
17.2
15.4
11.8

may be calculated:


                                             468

-------
    Consequently, until a better and more acceptable biochemical-based test

is developed, we no doubt will have to continue to use the BOD test as, one

measure of water quality. Our objective, then, should be to learn as much

as possible about the factors affecting the BOD test and to learn to

control those factors that interfere with proper analysis, interpretation

and application.  Presently, the factors contributing to nitrification in

BOD tests are known and methods for nitrification control are available.

We should use this technology!



                                REFERENCES

Britton, P. W.,  Unpublished results of EPA Laboratory Performance
Evaluation Studies, Analytical Quality Control Laboratory, U.S.
Environmental Protection Agency, Environmental Monitoring and Support
Laboratory, Cincinnati, Ohio (March 1985).

Hall, J.C. and Foxen, R.J., "Nitrification in BODg Test Increases POTW
Noncompliance," Journal Water Pollution Control Federati.on. 55, 1461-1469
(Dec. 1983).

Hurwitz, E., Barnett, G.R., Beaudoin, R.E. and Kramer, H.P., "Nitrification
and BOD," Sewage Works Journal. 19. 995-999 (1947).

Sawyer, C.N. and Bradney, L., "Modernization of the BOD Test for
Determining the Efficacy of Sewage Treatment Processes," Sewage Works
Journal. 18. 1113-1120 (1946).

Standard Methods for the Examination of Water and Wastewater. 16th Edition.
American Public Health Association, New York (1985).

U. S. Environmental Protection Agency, "Method Research Study 3. Oxygen
Demand Analysis." Analytical Quality Control Laboratory, Cincinnati, Ohio
(1971).

Young, J.C., "McDermott, G.N. and Jenkins, D., "Alterations in the BOD
Procedure for the 15th Edition of 'Standards Methods for the Examination of
Water and Wastewater," Journal Water Pollution Control. 53. 1253-1259 (July
1981).

Young, J.C., "The Electrolytic Respirometer," Water Research. 1.0.
1031-1040, 1141-1149, (1976).

Young, J.C., "Chemical Methods for Nitrification Control," Journal Water
Pollution Control Federation. 45. 637-646 (1973).

Young, J.C., "Comparison of Three Forms of Trichloromethyl Pyridine for
                                           469

-------
Nitrification Control," Journal Water Pollution Control Federation. 55.
415-416 (April 1983).

Young, J.C., "Waste Strength and Water Pollution Parameters,"
in Water Analysis - Vol 3.;  Organic Species, edited by
R. Minear and Keith.  Academic Press, 1984.
                                           470

-------
                 PEGGY  KNIGHT,  PH.D.

      WEYERHAEUSER ANALYTICAL  TESTING  SERVICES
          ANALYSIS OF  PRIORITY  POLLUTANTS
          BELOW FIVE NANOGRAMS  (ON COLUMN)
   IN MARINE SEDIMENTS BY  ISOTOPE DILUTION  GC/MS
                          DR. KNIGHT:  I've been

spending a lot of  time the  last  few weeks  trying  to

figure out why I should give this talk.  When Dale

Rushneck called last year to ask  if I could give  a

talk about this project, I  was, of course, flattered.

I told Dale that there were other labs, much larger

labs, out there which had used the method  and who

had participated in the development of the technique,

which we had not.  That they also had better

precision and accuracy, and I was certain  they had better

QA/QC practices.   Dale asked anyway and I  guess I'm

here.

     I think probably the advantage to my  giving

this talk is to show that a small lab used to

handling a variety of analyses can effectively use

the isotope dilution technique for the analysis of

priority pollutants in complex matrices.

     Our lab is relatively  small.  Our chromatography

group consists of four full time people and one
                        471

-------
part time consultant.  We have one Pinnegan GC/MS

and a battery of GCs.  Like many industrial labs,
                                    1     I i:1:;" '".
we analyze a plethora of compounds in about as

many matrices as we can come up with.  Seldom do we

see very many samples which develop  into what might

be called routine analysis.

     The hundred or so samples  in this particular

project consisted of marine sediment samples to be

analyzed ultimately for the Washington DOE  through

a contractor for the Port of Tacoma., Our lab was.

subcontracted for these analyses.  The sediments

were from the Blair, and Milwaukee waterways which

impinge on  Commencement Bay  in  the Puget Sound.

Commencement Bay, as many of you  know, is  a Superfund

site.  Part of  the requirements for  the  analysis

were the detection limits should  be  as low  as

practical and at.least  five  micrograms compound per

extract.

SLIDE  1                                ;

     The method was  based,; on' the EPA. isotope dilu-

tion method 1925 for priority  pollutants.   It  was

modified  to use soxhlet extraction  of the wet

sediment and add few other  procedures to clean up

 the sample  extract  before  analysis.   This medication

was developed  jointly by  Tetratech  and California
                         472

-------
 Analytical Laboratories for the Superfund Project



 analysis of Commencement Bay sediment.



      One hundred grams of wet sediment  were weighed



 directly into  a  soxhlet extraction  thimble.  The



 soxhlet had been previously cleaned by  detergent



 washing,  extensively  rinsing it with water, rinsing



 it  with solvent  and then running the thimble and




 the glassware  overnight with the extraction solvent.



 Five micograms of each label were spiked  into the




 sediment,  and  the sediment stirred  and  the  extraction



 started.   The  sediment was stirred  after  the first



 hour and  twice more within the  next eight hours  in



 attempt to minimize channeling.




      The  extract  was  transferred to a separatory



 funnel  and half-saturated  aqueous sodium sulfate



 added.   The  aqueous phase  was acidified and then



 extracted.   The  aqueous phase was then made basic



 and  and re-extracted with methylene  chloride.  We



 had  a great  problem here with emulsion formation



 and  it would have been  a heck of  a  lot easier if we



 had  been able  to  simply  ignore the  bases, but they



 wouldn't let us get away with it.




     Acid/neutral and base extracts were combined



and Kuderna-Danish evaporated with a steam bath.



Recoveries of all compounds were  good to this stage.
                        473

-------
We had taken blank spikes through the procedures to
find out where the recovery losses would be.
     The sample was shaken with mercury to remove
sulfur.  We seemed to lose benzidene at this stage
at the five microgram level.  To get rid of the
mercury sulfide fines, which were extensive, the
extract was repeatedly centrifuged and decanted.
Fines which remained collected at the top of the
GPC column.
     Gel used for the GPC was Biobeads SX-3 from
Biorad.  The columns were nitrogen pressurized and
manually operated.  The gel was optimized for
separation of corn oil lipid from phthalate and
recovery of pentachlorophenol, and the phthalate.
The columns were reused, but if an exceedingly
sloppy sample was analyzed, the column was  discarded
or the column was washed.
     Part  of  the extract was split out for  pesti-
cides, but since it's not an isotope dilution
technique,  I  will not be  discussing  them.   I under-
stand  now  that  several of the  labelled pesticides
are  available.
     The  solvent was  exchanged and  the  extract  run
through with  solvent  washings,  a  disposable C-18
solid  phase  extraction  column.

-------
      As  some  of  the  higher  labeled  PAHs  fluoresce
very  well,  we  watched the fluorescence throughout
the entire  procedure.   This  gave  an excellent  visual
check of  the progress.  By this aid,  it  seemed  that
we lost  a bit  of higher PAHs  on the SPE  column,
however,  they  were not  removed even with methylene
chloride.   The extract  was concentrated  again,  the
injection internal standard,  dichlorobiphenyl added,
and sample  injected  on  the GC/MS.
     The  instrument  used was  a Finnegan  4000 GC/MS
with  a Nova 3  data system.  We used standard columns,
DB-5.  It could be useful to  say  that a  standard of
4ug/mL of Phenanthrene  gave an area of about 50,000.
At these  instrument  settings  we found that we some-
times had a lot of problems with  disc space.  Some-
times the more concentrated samples  would run out
off of the end of the disc, which would  be discon-
certing.
     Numbers were developed from  a  program developed
by Dale Rushneck and Joel Karnofski, made available
modified, from Finnegan which reverse-searches for
labels.   After finding these, it searches for the
corresponding non-labeled compounds using a window
around a linear least squares fit of current times
compared to reference retention times.
                       475

-------
     The lower limits were calculated on a compound-


by-corapound basis for each individual sample using


instrument detection limits and adjusting this by


the sample weight, recovery of the corresponding
                                     , '•  . '  >.	i	•,'    , '  ,i
label, and the variability of the area of the


injection internal standard, DFB.
                                        1 i  • V,!' ' ( M .1! ,
     The calibration curve consisted of five points


ranging nominally from 0.5ug/mL to 25 ug/mL.  For

the phenols, the range was usually slightly higher.


I've picked out a few to show.  They are  fairly

typical.  Some of them were as straight as strings.


SLIDES 2&3

     The precision and accuracy data are  based on


duplicate, spike pairs; that is, a sample was

taken in triplicate, analyzed in duplicate, and

and the remaining replicate spiked with five micro-


grams of the non-labeled materials.  It was carried

through the entire procedure.  This was done on


approximately 10 percent of the samples.


SLIDE 4

     Again, I've picked out a few  to show as a

complete table would be absolutely indigestible.


These precision and accuracy figures are  in percent


recovery of the spike.  These are  fairly  typical.

In general, the PAHs are a  little  better  and the

-------
 phenols,  of course,  are worse.   Butylbenzylphthalate



 has  no  label standard  to recover.




      Somehow, when I proofed the slides,  I managed



 to miss the big  blank  area  there.   It  doesn't  mean



 that they weren't  recovered except  for 4-nitrophenol



 which usually is not.   I count  it among those



 compounds which  don't  really work.




      In comparison with Table 8  in  Method  1625,



 most of these figures  correspond fairly well.  The



 accuracy  figures are usually well within the ranges.



 The  standard deviations sometimes are  a bit higher



 and  sometimes within the  ranges.  The  recovery of



 the  labels  over  on this side usually were  within




 the  range.   The  label  recovery figures  represent



 a summary of the data  over  all samples  as  well as



 the  duplicates and spikes.




     To take a few of  these  examples,  2-chlorophenol



 from 1625,  the range for accuracy is 79-135 and the



 standard  deviation is +/-13  in 1625.  Cabel recovery



 ranges from  23 to  255 percent.   Dichlorobenzene is



 63-201, the  standard deviation is 43, and  a lower



 limit for the recovery  is not specified although it



 should be present,  and up to 550 percent.



     To fill in these numbers that I'm missing, our



recoveries here for acenaphthene were somewhere
                       477

-------
between about 30 and 150 percent.  4-Nitrophenol



was not recovered in the sample much of the time.



It just happened to be for some reason with our



spiked samples.  Fluorene ranged from 40 to 180.



     Most of the compounds that are analyzed by



this method are really very similar to this.  If



the PAHs seem more highly represented than warrants,




it's because those compounds were the compounds



with which I was dealing.  I should point out that



in comparison with the precision and recovery samples



in 1625, 1625 figures were developed using what



Dale referred to as a "real world reagent water



spike," whereas these are actual sediment samples



and, of course, have to be corrected for the amount



that was in  the sample  itself.



     There were a few compounds  for which the method



seemed  ultimately to fail, failure  being determined



by non-recovery of the  label.  As was mentioned



earlier, benzidene was  lost even on a  blank  spike



at five micrograms at the mercury stage.  Other



compounds  did  not seem  to survive the  procedure,



also at this level;  2,4-Dinitrotoluene,  2-Methyl




4,6-Dinitrophenol and 4-Nitrophenol.
      Hexachlorocyclopentadiene may or may
                                          not




survive the procedure.  We seemed to have problems
                        478

-------
 getting that compound through the column.   On a
 fresh  column or  a  regenerated column,  there are no
 difficulties.  However,  after a  few of the  sediment
 extracts have  been  injected on the  column,  the
 response rapidly degrades.  A rather similar effect
 was exhibited  with  acenaphthene.  On a new  or
 regenerated  column,  the  response  was twice  what it
 would  have been  on  a  column which had  been  exposed
 to sediment  extracts.  Consequently, after  the  sedi-
ment extracts, we reexamined  the  calibration curve.
 SLIDE  6
     This is  a pictorial representation of  the
 replication  of numbers.  A two sample  plot.   The
 amount  at 50'would  represent  five micrograms of
 compound.  The PAHs are also.
 SLIDE  7
     This is  just blowing up  that particular slide.
 The phenols are,  as expected, more scattered.
 SLIDE  8
     These are the people in  the  lab that also
participated in the study: Ed  Barnes,  Candy  McFaul,
Jim Leong and Mike Grove.  Two of them  have  spent
a lot of time doing extractions and...well,  three
spent a lot of time doing extractions  and the rest
of us were  involved in data manipulation and keeping
                        479

-------
the lab going.  You learn a lot about your lab when



you have a project like this; how well the group



functions as a unit, how well the normal workload



is handled with the additional sample load, and



where some of the faults in normal practices are.



     We had two major faults which came to light



during the project, one caused by the other.  All



the data had to be manually transferred at least



twice before the final report.  Sometimes this can




act as a prompt to .double check the data.  In our



case, we did have transcription errors which



developed and it also took so much time to transfer



the data that we allowed some false positives



through.  The compounds which come to mind are



benzidene and N-nitrosodiphenylamine.  Fortunately,



a QA/QC review practice outside caught these errors



before the data was entered  into the data base.



     With a small number of  samples, the  errors



would not have occurred, but with the bulk of work--



the sheer number of numbers— they did occur.  As  a



result, one of the highest priorities in  our lab-



oratory has become  the computerized transfer of



data, and I might add that no report is issued



until it has  been doubly rechecked, no matter how



loudly the client screams or the lab manager.



     Are there any  guestions?
                        480

-------
                           Sample
Developed jointly
by Tetratech and
CAL for the
Superfund Analysis of
Commencement
Bay Sediment
                                       CH2OH/CH2CI2
                     481

-------
o

-------
                                      Fluorene

                                   Ref: D10-Fluorene
 o


<


M-^
 CP
 o
 (U
2.2-


2.0-


1.8-


1.6-


1.4-


1.2-


1.0-


0.8-


0.6-


0.4-


0.2-


0.0-
              0
                    X 10% error limits
10    12    14    16



    Amount
                                                     —r~
                                                      18
—T~

 20
—r—

 22
                                                                                 24
                                Benzo(ghi)perylene

                                Ref: D12-Benzo(ghi)perylene
 D

<


%
o
cu
              01  23456789 10
                                       *83

-------
COMPOUND
PRECISION/ACCURACY
                                                   LABEL RECOVERY
2-Chlorophenol
1 , 3-Di chlorobenzene
Acenaphthene
4-Nitrophenol
Fluorene
Pen tachlor ophenol
Phenanthrene
Fluoranthene
Butylbenzylphthalate
Chrysene
Benzo(a)pyrene
Benzo(ghi)perylene
94 +/- 9.8
128 +/- 56
104 +/- 20
95 +/- 27
80 +/- 38
107 +/- 45
106 +/- 28
105 +/- 26
nn i_ / ic
80 +/- 15
100 +/- 30
91 +/- 46
102 +/- 41
16-75
2-63
42-83
0-250
47-154
18-180
13-147
32-150
40-250
12-300
14-400

-------
COMPOUNDS GENERALLY NOT RECOVERED

•  Benzidene
• *HexachlorocycIopentadiene
•  2,4 - Dinitrotoluene
•  4 - Nitrophenol
•  2 - Methyl - 4,6 - Dinitrophenol
                   485

-------
I
         a
         PU
          CO

          CD
          cd
          o
          CD
                                                                          m
                                                 486

-------
ffl-
PH
 CO
 CD
 ctf
 O
 CD

-------
 C/3
i—i
 O


 CD



PH
CD
cd
o
CD
                                                                •g
                                                                       QQ
                                                                 o
                                                                 CO
                                                                 o
                                                                 CN
                                                               - O
                                    
-------
             QUESTION AND ANSWER SESSION









                           MRS. VOLLMERHAUSEN:



 Jill  Vollmerhausen,  Martin Marietta.   You said you




 had a lot of trouble with emulsions.   I was wondering




 what  method  you  used to handle the emulsions?




                           MRS. KNIGHT:   We used



 centrifugation.




                           MRS. VOLLMERHAUSEN:  Did



 that  usually take  care  of it?




                           MRS. KNIGHT:   Well, it



 got us to  about  80 percent of  recovery  of the




 extract after repeat centrifugation,  but it




 sometimes  took a long time.




                           MRS. VOLLMERHAUSEN:  I



 also  notice  you use  soxhlet and  sep funnel  extraction.




                           MRS. KNIGHT:   That's right.




                          .MRS. VOLLMERHAUSEN:   What



 was the reason for that?




                           MRS. KNIGHT:   The methods



were dictated by a client  and  previous  studies  had



 indicated  that this  was necessary...




                           MR.  TAYLOR:   The  answer



for that  is when you extract 100 grams  of wet




sediment—and methanols are used in this—you  have
                        489

-------
a horrendous amount of water which you then have to
separate.  That's also the reason for the sodium
sulfate.
                          MR. STANKO:  George
Stanko/ Shell Development.  In your spiking procedure,
you spiked the labelled compounds into the wet
sediment?
                          MRS. KNIGHT:  That's right.
                          MR. STANKO:  Did you
conduct any experiments on taking the sediment,
drying  it first  and comparing your spiking of
labelled compounds onto a dry sediment, then add
the water and do your extraction, versus  spiking
labelled compounds into a wet sediment?
                          MRS. KNIGHT:  No, we
didrt't  undertake that study.  The methods, as  I said,
were  totally  specified.   Whether  that study has
been  done, perhaps Paul would like to comment.
                          MR. BARRICK:   I have  an
answer  for that.  The major  reason that wet sediments
were  used  is  because  there  have  been several  studies
that  show with marine sediments  and  other kinds of
 things  like  this,  that  if you do a drying procedure
on them, unless  it's  very careful freeze  drying,
you get development  of  artifacts, active  sites,
                        490

-------
 things like that.   When we were dealing with having
 to recover acids,  bases,  neutrals,  pesticides,
 everything all over the place,  you  simply couldn't
 afford to do a modification of  the  matrix through
 drying,  so we did  a wet sediment extract and then
 removal  of the water through the separatory  technique,
      There are so  many  different phase  associations
 of all these chemicals  that you  can't expect the
 isotope  dilution spike  to mimic  the recovery out
 of the matrix of all those  things.   There  are
 surface  cuttings,  things  that are engulfed and
 locked up  in matrices and  clay lattices  and  things
 like  this.   So the  real purpose  of  this  was  to get
 your  spikes  in there, get  them mixed up, get  them
 as equilibrated as  possible,  and then pull them out
 and primarily  use the isotopes as an analytical
 recovery  spike so you could  see  what was being lost
 in  the lab and what  wasn't.
      You were  making the  tacit assumption  that you
were  starting  out with a given amount of extract
 from  your matrix, but they weren't geared  in  a true
sense of isotope dilution to say that it was mimick-
 ing the total  recovery of every  single compound out
of every single one of the sum matrices  within the
samples.
                        491

-------
                          MR. TELLIARD:  Bob, would


you tell the ladies who you are?

                          MR. BARRICK:   I'm sorry.


Bob Barrick, Tetra Tech.

                          MR. STANKO:   I have one

final question on that.  When you actually calculated

the concentrations of  the analyte,  did  you use  the

isotope dilution calculation or  did you use the

isotope dilution compounds, or were you spiking

only to give you an indication of quality assurance?

In other words, how did you use  the labelled  spiking

data?  Was  it  incorporated into  calculations?

                          MRS.  KNIGHT:   Yes,  it
         i             •     •          ;i ,  ,  ,• ,;,' ;;;- ;.
was.   That  was also specified  in the method  that we

were  to use.

                           MR.  STANKO:,  I think I

would  have  some concerns  based  on the  information

you  just presented.
                                          \
                           MRS.  LESAGE:   Suzanne

Lesage, Canada.   I  would  like  to know  if you had to

look  for  other unknown compounds in those marine

sediments,  and if  so,  did the  addition of all

those labelled compounds  make  the chromatograms

so complicated as  you  would  have a hard time.

                           MRS.  KNIGHT:  We did not
                        •492

-------
have  to  look  for other unknown  compounds.   It  would
make  the identification of those materials  perhaps
more  complex, but as has been noted 'before,  the
masses for the labelled compounds are fairly unique
that  we're using, and I believe it would have  helped
quite a bit in making sure that these were  the labels
Also, we knew what the labels were supposed to be
and what labels were involved and where they
occurred.
                          MRS. LESAGE:  Except in
coellution problems.
                          MRS. KNIGHT:  Well,
coellution problems happen.
                          MR. TELLIARD:  Thank you,
Peggy.
                       493

-------
                          MR. TELLIARD:  Our next



speaker is going to talk about some ion chromato-



graphy.  Jack is from Cincinnati—that other lab.



Jack.
                         494

-------
                   JOHN  D.  PPAFF

   UNITED STATES ENVIRONMENTAL PROTECTION AGENCY
  ENVIRONMENTAL MONITORING  AND SUPPORT  LABORATORY
             USES OF ION CHROMATOGRAPHY
          FOR INORGANIC ANALYTES  IN WATER
                          MR. PFAFF:  We've heard

so many talks throughout the last day and a half

on HPLC, I thought maybe what I ought to do is

back up a little bit here and tell you what the

difference between ion chromatography and HPLC is.

     Basically, it's very simple.  HPLC came first,

1C or ion chromatography was a later addition to

HPLC.  The original design for ion chromatography

came from Dow Chemistry in the early 70's and it

was originally designed to do inorganic ions

using aqueous solutions and using conductivity

detectors.  Throughout the years, it has grown up

to a certain extent, and now it can be used for

some of the smaller organic materials and it cer-

tainly is using more different types of eluants

than just the aqueous types.

SLIDE 1

     What I've tried to do here is give you a very

basic schematic of an ion chromatographic system.
                       495

-------
There are basically two types of ion chromatography,



if you want to call it that, and this is due to



patents more than anything else.  When Dionex came



out with the first commercial type of ion chromato-



graph, they used what's known as a suppressor—I'll



get into that in just a minute—and they patented



the use of the term suppressor and the use of a



suppressor column following the separator column.



Consequently, anyone else who wanted to get into



the field, could not use a suppressor column.  So



now you have what is called a suppressed and non-



suppressed, or as some of them like to call it,



electronically suppressed pieces of equipment.



      If you start at the top of the slide where



the eluent is coming down, you get to the injec-



tion port.  The injection port is a simple straight-



forward thing except when you try to modify it and



put what is called a preconcentrator column in



there.  In the Dionex instrument, which most of my



work has been done on, the injection port goes



into  a replaceable sample loop so that you can



change the sample loop to put in any size from



approximately 10 microliters to somewhere in the



neighborhood of one milliliter.  Of course, you



get corresponding peak broadening depending on
                        496

-------
 the size that you use.




      Another way of injecting larger amounts of



 samples into the instrument is to put in a concen-




 trator column,  which is very similar to the separa-



 tor column  that is  being used.  It is placed behind



 the sample  injection port and up to 100 millimeters




 can be injected through this concentrator column.



 The ions of interest are held on the concentrator



 column while the matrix is  allowed to go to waste.



 The instrument  is then  switched  over and the eluent



 is  passed through the concentrator column and  goes



 through the separation  mode.   So when you start



 talking about detection limits,  you  usually have



 to  specify  what size sample loop you're using



 and/or whether  you're using preconcentration of



 the  sample.




     The next thing  in  the  line  is usually  what's



 called  a guard  column.   The  guard  column  is  simply



 put  there to protect the  investment  that  you put



 into the separator column.   You  simply  don't want



 to  ruin  your separator  column, so  you put something



that can be  sacrificed at the head of the column.



Let me  say here  that the  instrument  itself  cannot



tolerate solids.  Depending on whether you're



talking about the eluent or the sample, you're
                       497

-------
talking about particles in the neighborhood of



about 0.02 microns that have to be taken out.



They are taken out to a certain extent by the guard



column, but then you have to replace this more



often.  The separator column, obviously, as in any



kind of chromatography, is the heart of the system.



It does the separation so that you isolate one of




your components from the other.   I want to discuss



a little bit later the advantages of suppressed



and non-suppressed, so let's just continue on




down the slide.



     The dash lines that I've put in here are simply



to differentiate between the suppressed and the non-



suppressed  pieces of equipment.   Anything between



the dashed  lines then, would be suppressed and if



you sort of surgically remove  that section of the



slide,  you  would have., in essence, a non-suppressed



piece of equipment.  So  in  the  Dionex  equipment,



you go  through  a suppressor.



     Now,  what's  the sense  of  the suppressor?  The



suppressor  is going to take the eluent, which for



anions  is  usually  some  ratio of carbonate/bicar-



bonate  solution,  and change it into  a  non-conduct-



 ing  species so  that your background  is extremely




low,  and  in turn,  it changes your anions  to  the
                         498

-------
 acid  form so  you're  looking  at  things  like  HCL  or



 HF, which are highly conductive.



      Now, this  does  one  thing for  the  Dionex




 equipment.  It  does  increase the sensitivity levels



 which you can go down  to.  Usually a rule of thumb



 is that you're  talking somewhere in the neighborhood




 of one order of magnitude difference in sensitivity




 between suppressed and non-suppressed  equipment.



      So,  you go through  the  suppressor.  If you'll



 notice, you have two outlets on your suppressor.



 You have  a regenerative  going in and out as well



 as an eluent in and  out.  This is a relatively  new



 piece of  equipment.  The original suppressors were



 packed bed suppressors.  I don't know  how many  of



 you have  ever used ion chromatography, particularly



 with packed bed suppressors.  The packed bed sup-



 pressors  were good for let's say 12 hours of use



 before they just simply quit on you.   This neces-



 sitated approximately two hours of regeneration



 before you could put the instrument back into



 operation.




     The  introduction of the fiber suppressors



did away with the drawbacks of packed bed suppres-



 sion completely.  The suppressor is not a packed



column but rather it's more of a hollow fiber
                       499

-------
packed with beads, simply to cut down on the dead



volume, and it is constantly being regenerated by



a flow of sulfuric acid counter-current to the



flow of the sample.



     After the suppressor, the sample enters the



detector; whichever kind of detector you're using.



Probably the most used type of detector is the



conductivity detector, but as I'll show you later,



you have many different choices now that you can




use.



SLIDE  2



     This  is a separation of the common anions.



For those of you who can't read the slide, fluoride



chloride nitrate, not  total, but orthophosphate,



bromate nitrate and sulfate are separated.  The



eight  minute separation time is going  to be very



dependent on the type  of separator column you're



using  and your eluent.  It can vary anywhere  from



8 to let's say 20 minutes, depending on the column.



This particular column here  is the proprietary



column for Dionex that is referred to  as an AS-4



column.  I know  it's a matter of  choice, however,



I don't particularly care for the AS-4 column.   It's



very easily  overloaded and,  secondly,  if you  change



any of these ratios of solutions  that  you see here,
                        500

-------
you tend to get a lot of interference.  My personal



preference is that I would rather stretch it out



and have a larger, concentration that I can look



at.



     We did most of our work with what's called the




AS-3 column.  It's a little older column, it's more



loosely packed and consequently, you can put more




material on it, but your separation times are



going to be at least three times as long.  For



sulfate, we're talking somewhere in the neighbor-



hood of let's say about 20 minutes for elution.



Sulfate is your last elutor from this column.



     I mentioned that I would later differentiate



between suppressed and non-suppressed pieces of



equipment.  Sensitivity is better for the suppressed



equipment.  What are we talking about?  As I



mentioned, about an order of magnitude.  The cost



is significantly different.  Your non-suppressed



pieces of equipment are going to be considerably



cheaper than your suppressed pieces of equipment.



     Eluents and columns that can be used with the



two different kinds of pieces of equipment tend to



be more numerous for the non-suppressed pieces of



equipment.  Why?  Number one, you have many more



companies actually involved in the production of
                        501

-------
non-suppressed  equipment.   You  even have a  few



people who are  producing columns now that really



don't produce the equipment  themselves and  I think



this is a healthy situation.  It's the way  I think



gas chromatography went in  its  infancy.  If you do



tend to get these people, I  think you're going to



get away from proprietary columns and there's going



to be more competition and  I think there's going to



be a more successful introduction of columns.



     I mentioned the fact that our lab has a Dionex

                                            1 • I ,,'n. "

piece of equipment, a 2120,  which is the latest in
                                         1 i ill.


their series.   Let me point out one thing here,



when you start  looking at people's slides.  If you



notice this negative peak right here, this is what's



known as the water dip.  This is the thing that



causes a lot of gray hair in people who are trying



to do anionic separation on  ion chromatography.  If



you see the water dip coming out in your standard



anion separation after fluoride, you can sort of



guess that the chromatogram that you're looking



at was done on a packed bed suppressor.  When you



get to the fiber suppressor, the negative peak



occurs in front of fluoride, and then it starts



to cause more and more interferences in the



separation of fluoride.
                       502

-------
     One other thing with packed bed suppressors.



The retention times changed drastically as they



tended to exhaust themselves, and you really had



to watch what you were doing because even the water



dip would migrate into an area where it would become



an interference.



SLIDE 3




     Talking about the migration of retention times,



one of the drawbacks of ion chromatography seems to



be the fact that the retention times can vary



considerably with concentration.  This work was



some early work that we did.  I picked out the



worst and best possible cases, obviously.  Fluoride



is an early eluter, so as you can see, if you



jump from 100 ppm to .5 ppm, you're only talking



two-hundredths of a minute change, which you might



as well forget.  There's really no difference there.



For nitrate using that same ratio of 100 parts to



one-half a part, you'll notice now that you're



talking in the neighborhood of two minutes variation



with concentration.  In sulfate, which is the



worst, going from 500 ppm to 1, you're changing in



the neighborhood of four minutes in retention time.



     I think the one thing here that you've got to



say right at the beginning of any kind of a
                       503

-------
methodology using  ion chromatography  is, you've



got to do an awful lot of standard injection into



the instrument so  that you can get some kind of an




idea of what it is that you're actually looking



at.  There's just  no way around it.



SLIDE 4



     This is a slide that I borrowed  as a figure



from Small who published a paper in Analytical



Chemistry and that paper introduced the hollow



fiber suppressor for Dionex.  If you're looking at



the first one...the only difference between the



first and the second is the size of the packed bed



suppressor...you'11 notice the two dips; number one,



the water dip and  secondly, the carbonate dip.  The



second one shows that as the column gets longer,



the two peaks seem to change in retention time.




The third one actually shows that with the hollow



fiber suppressor, they migrate together and you



think you only have one peak, but in  essence you



don't.



     Let's treat them separately.  The water dip ,



peak,  although it is something that gives you a



problem, is easily gotten around and  that is simply



by matching the ionic strength of your sample to



the eluent that you're using.   That sounds
                       504

-------
difficult, but if you take and make up an eluent



at 100 times the concentration that you're using



in the instrument, and simply add 1 mL of that



to 100 mL of your sample, you really haven't changed



much in the form of concentration, but your water



dip will level out and you don't have to worry



about interference from the water dip.




     The carbonate dip, although Small showed it



as a negative dip, this negative dip shows that you




have depletion of the carbonate ion; that it's lower



in concentration than the eluent.  What happens if




you have carbonate present in your sample at a



higher level?  What happens is you get a positive



peak.  For a long time with the system that we



were using, we were trying to determine fluoride



and the reproducibility of it was not good.



Eventually we weakened the eluent so much trying



to determine it that our fluoride peak in natural



water samples evolved into two peaks.  What happened



was that carbonate elutes at the same time as fluo-



ride and is a severe interference as far as fluoride



determination is concerned.



     The carbonate is one of the replaceable ions



in the separator column.  It is the ion that the



anion replace throughout the separation.  So if
                        505

-------
you have a sample that is very high in concentration
of total ionic strength—add everything together,
you tend to get a rather high carbonate peak due
to the separator column, and somehow or another,
you can't quite do without that.  So you're always
going to have it.  We originally started off look-
ing at this as hardness, but later on we found out
that it has been coming off the column.
     How did we get around it?  Actually, we
got a column made by another company, other than
Dionex, and put it into the suppressed equipment.
It will actually separate the carbonate from the
fluoride, so we're hoping that we can now do this
quantitatively.
SLIDE 5
     I mentioned the fact that there are other
detectors, and so I included this slide so that
you could see that you can do things other than
the standard anions.  What you're looking at, the
bottom chromatograph, is the common six anions
done by a conductivity detector while the top
chromatogram—this is the separation here--is done
by an electrochemical detector.
     If you'll notice the same sample here, you've
got your separation down in here of your standard
                        506

-------
anions/ but now we're getting sulfite, cyanide and




bromate coming out on the electrochemical detector.



These detectors can be run in series since neither



are destructive, and you can get considerable



separation.  Since cyanide is one of the materials



that is most commonly sought after for separation,



we're hoping that this particular separation can




be used to do some of the separations for cyanide



in place of some of the wet methodology that is



used at the current time.



     In addition to electrochemical detectors,



ultraviolet detectors, fluorescence detectors can



be used and there's even an adjunct, if you will,



detector known as a post column reactor in which a



chemical is pumped into the system to react with



the eluent and the anions, or whatever it is you're



trying to  separate, that will usually give you a



reaction that can be determined using ultraviolet



detectors.



     Getting back to some of the  interferences for



the common anions, aluminum interferes, obviously,



with the fluoride, but that has nothing to do with




the ion chromatograph.  With fluoride, which is



first  off  the column, you're really  interfering



with what  we call the garbage peak—everything
                        507
                        509

-------
have trouble in zeroing the baseline.  This  is

part of the reason for this order of magnitude

difference.

     In way of summary, let me say that I think

ion chromatography has a place in the analytical

scheme of things.  I want to say one other thing,

and that's to caution you to run a lot of standards

and spikes because of this migration of retention

time, in order to be sure of what you're doing.  I

think it is probably one of the better and more

promising techniques to come along, in inorganic

chemistry anyway, since the advent of atomic

absorption.

     Thank you.  Any questions?

                          MR. TELLIARD:  Any

questions?  Thank you.
NOTE:  Figures  are  reprinted with  the permission
       of the Dionex Corporation.
                        510

-------
511

-------
512

-------
513

-------
6 a-
—- Q.
 ^ a
                      dip J8ie/v\
                         H *0 UO|ID3(U|
  CM OL
dip J31R/NA

-------
515

-------
516

-------
517

-------
                          MR. TELLIARD:  Our last




speaker this morning is Suzanne with Environment,



Canada.  She's going to talk about some HPLC with



electroconductivity detection for phenols.
                        518

-------
                SUZANNE LESAGE, PH.D.


             WASTEWATER TECHNOLOGY CENTRE
                 ENVIRONMENTAL CANADA
           ENVIRONMENTAL PROTECTION SERVICE


          DIRECT ANALYSIS OF PHENOLS BY HPLC
                           DR.  LESAGE:   Thank you

 for inviting me here.   The little ship break this

 morning reminded me much of the situation that we

 have.   Our  offices  are  located on Lake Ontario

 between Toronto and Niagara Falls,  right on the

 lake and beside a lift  bridge, very similar to

 the one that is here.   Whenever my  guests stop

 listening to me,  I'll turn around and  inevitably

 there's a ship  going by.   Fortunately,  we've got

 winter  when  we  manage to  work  because  then  there

 are no  ships going  through.

     Before  I start, I  will acknowledge  the  people

 I work  with.  The data  that I  will  present  today

was mostly produced  by  Mrs. Sharon  Hay-Pole  and

Mrs. Sandra Abbott.   Mr. Ken Conn is the  supervisor

of the  lab and  is the person who  allowed  me  to do

this kind of work, and  Mr. Fowlie is our  guality

control chemist.  Maybe I should  say the  data
                       519

-------
didn't quite pass his criteria, but I told him,
"It's organic analysis, you have to tie understand-
ing."
     I work for Environment Canada, the Canadian
Federal Department of the Environment, more speci-
fically for the Environmental Protection Service
in the Wastewater Technology Center.  We are a
small research facility, mostly staffed with en-
gineers, and our role  is to study  industrial dis-
charge problems and  try  to solve them by developing
appropriate control  technology.
      Our  laboratory  works  in  support  of  those
research  programs, and also acts as a national
center  for GCMS  analysis of  industrial  wastes.
We've done all kinds of wastes—tanneries,  textiles,
wood preservation,  steel mills, coal conversion
wastewater,  resin manufacturing, pulp and  paper.
We're only one lab,  so we do  everything.
      Whenever possible, we use the EPA methods.
 To us it only makes good sense; why reinvent the
 wheel?   Well, of course, like everybody else,  we
 get the urge once in a while to try, and our
 attempts at reinventing the wheel  for phenol
 analysis  is what I will be describing now.
      This program was started because we needed
                         520

-------
 to turn out a lot of phenol results in coal conver-




 sion wastewater.  We have a GCMS, which we prefer



 to use where we can.  Unfortunately, we only have




 one, and when you have several pilot plants running



 continuously, we have to be'able to provide data



 relatively quickly,  and the engineers  with whom we



 work are more used to the BOD, TOCs type of turn-




 around time and they don't have the patience to



 wait for six months  for their samples  to be analyzed



 by GCMS.   Of course, their needs are to run the pilot



 plants,  thus they require results much faster.



      The  situation we had here,  was we were looking



 at the effluent from a pilot  plant, consisting  of a



 biologically active  fluidized  bed treating  coal



 conversion wastewater.   As  an  aside, the  coal



 conversion wastewater used  actually was  imported



 from the  H-coal  Pilot Plant in Catlettsburg,



 Kentucky.   We don't  seem  to have  enough Canadian



 problems,  so we  import  some.




     We looked at  the alternate method 604, the GC



method for  phenols,  so  we could  turn out  results  a



little bit  faster  than  with GCMS.   Unfortunately,



coal conversion wastewater contains a  lot of other



chemicals.   Typically it's full of heterocyclic



nitrogenous  compounds—aniline, quinoline, carbazole
                        521

-------
and the like.  So we got a tremendously complex



chromatogram.  The only way really we could handle



it was GCMS.  The other problem we had with this



as with all GCMS analysis, is that typically with



this kind of matrix, our recoveries from phenol



ran from very poor to simply pathetic.  That means



30 down to about 5 percent, if we found any.  We



knew from the conventional phenol, 4-amino-



antipyrene method, that there was a substantial



amount of phenols in there, the only problem is that




the GCMS just couldn't find it.




SLIDE 1



     So we  turned ourselves to HPLC as  a possible



substitute.  The method we tried  is in  essence very



simple.  We  used  a  five micron ultrasphere ODS



column that  can be  purchased  anywhere,  and the



mobile phase is acetonitrile  and  water  with



acetic acid  and phosphoric acid as modifiers at



1.5 mL/min.  We used a  BAS amperometric detector



running at  1.2 mV,  and  20 mA  current.   That is



really middle  range for that  particular detector,



but we  find that  although the label  says it goes



down  to .1, the  actual life  situation is different.



We also  have to  use a UV detector to follow the




 response of nitrophenols.
                         522
                                                             1 '•''fc!'IIL: !**:;. IK'

-------
     For PCP analysis, the percentage of aceto-



nitrile has to be increased to 60 percent, other-



wise it takes something like an hour and a half to




elute PCP with detection that would be very poor.



We find that usually it's a lot more practical to




do PCPs one day and the rest of them the other



day.  In coal conversion, PCP was of no concern




whatsoever.  We were actually looking more at



phenol, methylated phenols and dihydric phenols,



and when we do PCP, it's usually for wood preser-



vation plants where tetra and pentachlorophenol



are most important.  It would be possible to do



all of priority pollutant phenols by solvent



programming, but the electrochemical detector does



not like any kind of programming whatsoever.  The



response of the detector depends on the mobile phase,




on its electroconductivity.  When you increase the



percent of acetonitrile, the baseline drifts down



and the detector does not recover fast enough to



enable you to do a lot of analysis in a day, so



it's easier to change mobile phase over night.



SLIDES 2&3




     This is a typical standard chromatogram.



There's not really much to see looking at this



standard chromatogram,  other than everything is
                       523

-------
nicely separated.  There is response to dinitro-
                                ,	.,",	Si ;"'i I"' ,! A'lH'i	>.    ••  .;•  ,
phenol, and 2- and 4-nitrophenol on the electro-

chemical detector.  The amount  injected on column
                                  '• ''  '  ':;' ':M J"11:     ;  •  ;'•
here is representative of 100 to 500 micrograms per

liter solution.  That represents 2 to  10 ng.  injected
                             1 .  •  ,• . i	•:   ' :  i •»*.•'.   .« .'...,
in a 20 microliter loop.  We also followed the

analysis with the U.V. and visually checked the

presence of nitros from both detectors.  You  can

see that for 4-nitrophenol one  actually  is better

off using the electrochemical detector rather than

the U.V.  detector.  For the 2,4 nitrophenol  and

4,6 dinitrocresol, the response is not sufficient

in the electrochemical detector.  But  as  I'll

show you a little bit later, we have done  some work

and it is possible to enhance  the  response by

addition of specific buffers.   Unfortunately, it

doesn't really  quite work.

SLIDE 4

     These are  the retention  times  of  allthe

compounds of interest.  As you  can  see here,  we can

complete analysis within 25 minutes.   One  thing I

might not have  stressed enough  here  is that  we're

doing direct analysis.  We get  a  sample,  filter it

through to a  .45 micron filter  and  inject.   That  is

the end of sample preparation.   This  is  not quite
                        52*

-------
 the  type  of  work  that  we're  used  to.   In  a  normal



 organic lab,  if you haven't  done  three  extractions



 and  five  cleanups, you haven't  done analysis.   But



 it has the advantage of  being able to do  duplicates,



 triplicates,  spikes and  all.



 SLIDE 5




     We analyzed  the standards  in mobile  phase  and



 we did five  replicate  analyses.   This is  the typical



 kind of recovery  we got.   Now,  looking  at 97.8



 percent recovery  for phenol, and  really,  I  don't



 know that anybody can  honestly  report more  than 50



 percent, when you're doing extractions, I got quite



 excited, especially with really rather  low  standard



 deviation, nothing above  10  percent.  This  is not



 the kind of  numbers I'm  used to in the  lab.



     The nitrophenols  and dinitrophenols  can be done




 either way.  For  the 2-  and  the 4-dinitrophenols, we



 usually get slightly better  standard deviation  using



 the electrochemical rather than the U.V.  detector.



 SLIDE 6




     To prove that this  really worked,  we used  the



 standards which are most generously provided by the



 EPA for quality control.  These standards are pro-



vided for quality control under method 604.   That's



 the GC method with or without derivatization.  The
                        525

-------
first column is the true value of the standard.




These are the results obtained with Method 604



and provided with the standards.  This is the kind



of number that we got in our method.  Again, where



we got most excited was finding that with a true



value of 75 microgram per liter of phenol, we were



actually reporting 78 rather than 34.  To us that



is most crucial.  The other thing is that this



really isn't very much work.  It's just basically




get the sample, filter and inject.



     For the other phenols, we think that the results



either are equivalent or better than ones obtained



with Method 604, with the exception of PCP.  PCP




at 70 micrograms per liter, that's below the



detection limit.  Now, I must stress that is for



direct analysis.  We could get better by extracting



the sample and  injecting the extract, which  is



obviously a very easy way to go.  We had very little



interest in actually doing that because when we



have to do PCPs, we get samples that are high



rather than low.  We get levels that would be



running around  50 or 60 milligrams per liter, so



dilution is more of a problem than actual detection



limits.
                        526

-------
 SLIDE  7




     Now,  in  this  next  slide we  decided  to  try  a



method in  real  life and to do so, we selected two



 typical, or maybe  not very so typical wastewaters



 that we deal  with.  These are not nice wastewaters.



 They were  very  yellow,  fairly high solids content.




 I don't have  the exact TOC's, but it's fairly high.




 The other  reason we chose those  is that  we  had  GCMS




data to back  up our analysis and we knew roughly



how much phenol was in there to  start with.  For




 instance,  the conversion wastewaters I'm talking



about here, by  GCMS we were reporting a  figure  of



 6 micrograms  per liter.  My supervisor got  a bit



upset because the  4-AA method was reporting some-



thing to the  order of 400 micrograms per liter  in



total phenols,  and he said well, what are these



other phenols?  I said, I don't know.  When we ran



this, phenol  itself accounted for over 100 micro-



grams of the  total and there were a substantial



number of methyl phenols and other analytes.  So



this method usually gives a better correlation with



the total phenols.




     These numbers here are not the measured values



but actually the percentage  of recovery.   The water



data is basically what was presented in  the previous
                       527

-------
slide.  Everything is in fairly good standing and
would pass except for some of the nitros.  We all
know, of course, problem with 2-nitrophenoI.
     I must say that we did not throw away any data
whatsoever.  So, what happens here when you see
193 +/-111 on triplicate, we found it in two samples
and didn't find it in the other.  At 150 micrograms
per liter with 2-nitrophenol, that's basically at
the detection limit, so it is not that surprising
that there is such a variation.
SLIDE 8
     We repeated the experiment at twice the level to
see whether the method is linear.  The percentage of
recovery  in water again is very similar  to what we
had obtained at the lower level and in the two
samples,  everything is more or less above 50 percent
recovery  except for the nitrophenols, and I think
that's one of the nightmares of the analytical
chemists.
OVERHEAD  1
     Briefly I'll touch on the effect of adding
buffers,  one interesting fact we found here.  At
the  top part here,  that's the chromatogram  that I
showed you in the slides previously (Slide  2).
That's our normal standardization chromatogram.
                        528

-------
 Here,  below, is a chromatogram we obtained where we



 added  .01M sodium acetate buffer to the solution.



 If you look at the 4-nitrophenol response here



 compared  to the 4-nitrophenol  here, you can see



 that there's a substantial difference.   The same



 case with the 2-nitrophenol which is substantially



 enhanced  by the addition of buffer.




     You  will ask probably why on earth didn't we



 use the buffer all the  time.   Well, the problem is




 that not  only is  the  2  and the 4-nitrophenol



 response  enhanced,  but  so is  the 2, 4-dinitrophenol



 and the 4,  6-dinitrocresol.  Their response is



 enhanced  and the  retention time shifts.   What  we



 think  is  happening  in this case is that  the species



 we're  actually measuring is not nitrophenol itself,



 but  an  ion-pair.



 OVERHEAD  2




     There  are the  typical detection  limits that



 one gets, based on  20 microliters  analyzed  directly.




 It  isn't  great for  dinitrophenol,  but neither  is



 the GC method.  If  there  is a need  for lower detec-



 tion limits,  I  think one  would  have to go back




 to the  extraction with all the problems it  entails.



 It  isn't  as good as one would want, maybe,  for



some clean effluents, but  is usually sufficient in
                       529

-------
a real life situation for the monitoring of an



industrial discharge.



SLIDE 9



     Finally/ here are what we believe are the



advantages of the method.  It has all of the advan-



tages of direct analysis.  There is no sample



preparation.  The method would be very easily



automated because there is no need for an operator.



It's low cost.  We estimate that getting a single



pump with the BAS detector and a column would cost



you less than $10,000 U.S.  For us, of course, it's



about $20K Canadian.



     It is a very rapid method and the reproduci-



bility of standard deviation is usually better



than one would obtain having to extract samples,



especially very dirty samples.  It's also fairly



free of interferences.  The coal conversion waste-




water that we used,  for instance, we really could



not use anything else but GCMS and this provided



us an alternative.   So we believe this is ideally



suited for routine monitoring of effluents when



the samples are already characterized.  Thank you.



                          MR. TELLIARD:  Any




questions?
                        530
                                                                .:1*	1.

-------
      ANALYTICAL CONDITIONS










 COLUMN: 5  urn ULTRASPHERE ODS







 MOBILE PHASE:  CH3CN: WATER:ACETIC  ACID:PHOSPHORIC ACID




                  40:  59.7 :     0.1    s       0.2







 FLOW RATE:  1.5  ML/MIN





 DETECTOR A:  BAS AMPEROMETRIC




               1.2 VOLT





               20 uA CURRENT







DETECTOR B:  FIXED  U.V. AT 254 nm




   ATTENUATION:  0.005 A.U.










PCP:  CHANGE  TO  60% ACETONITRILE
                                  531

-------
STANDARDS by ELECTROCHEMICAL DETECTION
   (2 to 10 ng per component  (20 >il loop)
         2,4,6-trichlorophenol
                        2 nitrophenol  2^ aimethylphenol
                                 -chloro-m-cresol
                       2,4-dichlorophenol
                               532
                               ''	"l 1

-------
4,6 dinitrocresol
STANDARDS 10 ng each
U.V.    254 nm
  2-nitrophe"nol
     2,4 dinitrophenol
     4-nitrophenol
                 533

-------
                      COMPOUNDS  OF INTEREST
                                                  R
             PHENOL




             4-NITROPHENOL




             2-CHLOROPHENOL




             2,4-DINITROPHENOL




             2-NITROPHENOL




             2,4-DIMETHYLPHENOL




             P-CHLORO-M-CRESOL





             2,4-DICHLOROPHENOL




             4,6-DINITROCRESOL




             2,4,6-TRICHLOROPHENOL




             PENTACHLOROPHENOL
 3 .77




 4.46




 5 .87




 6.63




 7 .54




 8.24




 9 .56




1 1 .7




14 .6




22 .2




13.3*
* USING  60% ACETONITRILE -  40% WATER WITH ACETIC  ACID
                                 534

-------
ANALYSIS OF  STANDARDS  IN MOBILE PHASE ON  5  REPLICATES
TRUE
VALUE
COMPOUND ug/L

BCD
PHENOL
4
2
2
2
P
2
2

4
2
2
4
-NITROPHENOL
-CHLOROPHENOL
-NITROPHENOL
, 4-DIMETHYLPHENOL
-CHLORO-M-CRESOL
,4-DICHLOROPHENOL
,4 ,6-TRICHLOROPHENOL
UV
-NITROPHENOL
,4 DINITROPHENOL
, NITROPHENOL
,6 DINITROCRESOL

100
500
100
500
250
250
250
500

500
500
500
500
MEAN
X

97.8
451
95.5
474
238
244
246
516

427
534
619
351
STD.
DEV.
n

2
23
5
25
9
13
6
35

31
28
86
20

.6
.1
.0
.7
.2
.9
.3
.8

.9
.2
.0
.6
REL.
STD .
DEV.

2
5
5
5
3
5
2
7

7
5
13
5

.7
. 1
.2
.4
.9
.7
.6
.0

.5
.3
.9
.9
                               535

-------
                                                                                                                      11  '!   '.'i,,i1:"1'!1;!,, Hi',IKI,.,'•!'' 'L"1'1:"''"''!
                                                                                                                                                                             ''IPS  <:"''!-'  v1 J1 -"•	!!' "',[ ' ill1'.;,!,:1!"!!!!!1!!
                                                                                              EPA   STANDARD
                                                                                  TRUE
METHOD    604
OUR   METHOI
PARAMETER
                                                                                  VALUE
                                                                                                                          x
                                                                                                                                                                                        x
PHENOL
2,4-DIMETHYLPHENOL
2-CHLOROPHENOL
4-CHLORO-3-METHYLPHENOL
2 , 4-DICHLOROPHENOL
2,4, 6-TRICHLOROPHENOL
PENTACHLOROPHENOL
2-NITROPHENOL
4-NITROPHENOL
2,4-DINITROPHENOL
75.0
100
75.0
125
100
150
70.0
150
160
250
34 .1
61 .4
60.3
1 06
79.6
124
60.1
120
73.0
205
11
21
13
19
17
20
13
22
31
67
.1
.2
.6
.0
.0
.4
.9
• 8
.2
.8
78.1
92 .0
73.8
125
99.4
170

176
1 18
221

'1
1
1 J
41
-1
	 1
\
4 J
ia
                                                                                                                  536

-------
                   RECOVERY OF  SPIKED  EPA STD





                           HIGH  LEVEL
COMPOUND
BCD
PHENOL
4-NITROPHENOL
2-CHLOROPHENOL
2-NITROPHENOL
2 , 4-DIMETHYLPHENOL
P-CHLORO-M-CRESOL
2 ,4-DICHLOROPHENOL
2 ,4 ,6-TRICHLOROPHENOL
TRUE
VALUE
ug/L

150
320
150
300
200
250
200
300
WATER
% + S.D .

94±1 .0
84±1 .7
89±3.3
107±12
99±10
107±18
130±35
144±17
SAMPLE A
% + S.D.

158±4.0
40±44
60±3.3
62±0 .5
85±2.0
86±2.7
100±2 .9
69±7.5
SAMPLE B
% + S.D.

109±0.9
124±17
103±2 .0
71±1 .2
82±6.2
89±18
1 1 9±27
1 10±20
   uv
2,4 DINITROPHENOL




4,6 DINITROCRESOL
500       86±36        93±3.5




NOT  PRESENT  IN  STD .
92±0.7
                                   537

-------
                     RECOVERY OF  SPIKED  EPA STD




                        USED  FOR  METHOD  604
  COMPOUND
TRUE



VALUE



ug/L
WATER      SAMPLE  A  SAMPLE B



%+S.D.   %+S.D.  %+S.D.
    BCD





 PHENOL




 4-NITROPHENOL




 2-CHLOROPHENOL




 2-NITROPHENOL




 2,4-DIMETHYLPHENOL




 P-CHLORO-M-CRESOL




 2,4-DICHLOROPHENOL




 2,4,6-TRICHLOROPHENOL  150
75
160
75
150
100
125
100
150
104±.3
74±33
86±9.1
1 17±27
9 2 ± 1 6
82±13
86±45
1 1 9±63
155±14.
12±11
39±34
193±1 1 1
92±6 .5
86±2 .6
1 10±4 .5
1 18±61
     uv




 2,4  DINITROPHENOL




 4,6  DINITROCRESOL
 250        89±8.2      61±33




 NOT  PRESENT IN  STD.
                                   221±192





                                    84±13




                                   191±17




                                    86±2 .2




                                    91±2.9




                                   101±4.9




                                    79±34
                        90±14
SAMPLE A:   COAL  CONVERSION WASTEWATER





SAMPLE B:  EFFLUENT  FROM  A WOOD  PRESERVING PLANT
                                                                         1 
-------
      STANDARDS by ELECTROCHEMICAL DETECTION
                                      .phenol
                    nitrophenol
                    —       2 chlorophenol
                    =rr==2.nitrophenol  2/4 dinethylphenol
                                   -p-chloro-m-cresol
                           2,4-dichlorophenol
              2,4,6-trichlorophenol
INJECT  01/29/35 11:22:22

              II  0
      fe&. 82

       22. 2222' 44
            24. 66
                            ICH
                                                          O-o/M
                                  539

-------
                   DETECTION  LIMIT




           Based on  20  ul  analyzed directly
                                               ug/L
                                                             •Id ' ljl" ill'I'I?ft1
Phenol




4 Nitrophenol




2 Chlorophenol




2 Nitrophenol




2,4 dimethylphenol




p-Cl-m-cresol




2,4-dichlorophenol




2,4,6 trichlorophenol




pentachlorophenol
 10




200




 20




100




 25




 25




 25




 50




150
U.V.




4 Nitrophenol




2 Nitrophenol




4 ,6-dinitrophenol
300




300




300
                               5*0

-------
                  ADVANTAGES  OF THE  METHOD










DIRECT  ANALYSIS







        NO SAMPLE PREPARATION





        EASILY  AUTOMATED





        LOW COST





        RAPID





        IMPROVED REPRODUCIBILITY








IDEALLY SUITED FOR  ROUTINE  MONITORING OF  EFFLUENTS

-------
                                                  ':;	I" ft 'i'1:: i>"i' ilii1
            QUESTION AND ANSWER SESSION




                          MR. STANKO:  George

Stanko, Shell Development.  I have one question.

Can you explain the difference in your procedure

and Method 604 with respect to why the 604 method

only has approximately 50 percent recovery of

phenol itself and yours gets just about 100 percent

recovery?  Do you have an explanation or a reason?

                          DR. LESAGE:  We do not

extract.  That's why.  Method 604, I believe, is a

methylene chloride extraction, and typically phenol

doesn't seem to extract out of water.  In this

case, we just take a sample and inject it directly.

So these recoveries are not really recoveries.  The
                                       .it    i1'. ']h r
measurement is a direct measurement.

                          MR. STANKO:  Thenit's

really nothing in the chromatography?  In other

words, there's nothing unique to your column or

your chromatographic system, it is all in the

extraction versus no extraction?

                          DR. LESAGE:  I would tend

to believe so, yet the chromatography on the GC

column would obviously be better to get more

theoretical plates than one gets with HPLC.  The
                        542

-------
 detector  is  also more  specific,  so  that  there  are



 fewer  interference problems.




                          MR. McMAHON:   Wayne



 McMahon,  Martin Marietta.  Did you  do any comparison




 studies to compare the data from the 4-AAP method




 to your HPLC data to see how well there  was cor-



 relation  between the two methods?




                          DR. LESAGE:  We always



 compared  the data.  One of the problems  with com-



 parison with the 4-AAP method is  that of course,



 it is for total phenols, whichever  ones  are present,



 the color is the total.  To be able to compare, it



 is necessary to know which species  are present and



 what is their respective response for the 4-amino-



 antipyrine because as you know, that method is



 based on  total phenols as phenol.   The color response




 from other phenols is not equivalent.  It would be



 something like 40 to 80 percent depending on the



 phenol.  Anything that, is substituted in the para



 position will not respond to the 4-aminoantipyrine.



 So when we correct for that and in samples that are



 not typically too bad, that is they don't have a



 lot of other interferences,  the correlation is



 really quite good.   Typically, we can usually



account for about 80 percent of the total phenols.
                        543

-------
There was a case where there was a lot of pehta-



chlorophenol in the sample where the correlation



fell through completely.



     Now, when we try to do the same correlation



with GCMS, there is usually absolutely no correlation,



it never makes any sense.  So in this case, it's a



little bit better—quite a bit better, actually.  At



least the curves follow the same slopes.



                          MR. EDELMAN:  Dave




Edelman, Crown Zellerbach.  When you do your



percent recoveries on your real world effluent



samples, how do you know you had positive inter-



ferences beforehand when you make your  corrections



for what's  initially in there before you spike  the



samples?



                          DR. LESAGE:   We did first




an analysis of the sample as is, then spiked at the



two levels.   I just reported here the percent of



recovery, not the amount that was found.  There



were some phenols that were  not found at all and



typically,  phenol itself was running about  100



micrograms  a  liter as well as a few others  that




were present  in there, and we corrected to



percentage  recovery.
                        5*4

-------
                          MR. EDELMAN:  Right.  I



understand that, but how do you know you don't have



positive interferences initially that peak your




chlorophenol before you spike it, as phenol,



not something else, but one of the other peaks



is...




                          DR. LESAGE:  Yes, that's




what I said earlier.  A method like this has the



same problem as Method 604.  For instance, it's



like any GC or HPLC method.  You have to be able



to correcterize your sample by an alternate method



and check for those kind of possible interferences.



We usually do not have very much problem with that



and we have run very, very dirty samples.  There is



usually very few peaks in the electrochemical



chromatogram.




     One thing is that, basically because you're



running a very specific buffer,  something like



aniline will not be in the same ionization state.




It will be a positive ion in the solution, and



be easily separated so you don't have a problem.



It is more specific because you only elute in



certain window.  So it's not completely free of



interferences, I'm not going to say that, but if
                        545

-------
you have something that you have correcterized, you



can fairly safely use it.




                          MR. TELLIARD;  I'd like



to thank Susan and Peggy and Jack for their



presentations this morning.  I'd also like to  thank



Dale Rushneck and Todd Fielding for working so much



in getting this thing together, and, of course,



Whitescarver  Associates and County Court Reporters} inc.



for their work here.



     This ends the eighth.  We hope to see you for



the ninth.  Until we figure out where else we're



going to have it, we might assume it's going to



be here, unless somebody comes up with a real



winner.  It's rough to follow up on a battleship,



but we'll try to figure out something.  Thank  you



very much for coming; enjoyed having you here.
                        546

-------
ROSTER OF ATTENDEES
         547

-------
J. H. Alexander
Supv. Chem i st
Norfolk Naval Shipyard
Cod 134.2, Bldg. 184
Portsmouth VA 23709
804-396-3373
B. A. All en
Manager of Environmental
Dow Chemical U.S.A.
Building B-1226
Freeport TX 77541
409-238-5208
                         Services
Sue Apter
United States Testing Co.,  Inc.
1415 Park Avenue
Hoboken NJ 07030
201-792-2400
Gerald Babsk i
Environmental  Chemist
Westchester County Dept.
Hammond House Road
Valhalla NY 10595
914-347-3155
                         of Labs.
John M. Ballard, Ph. D.
Lockheed EMSCO
944 E. Harmon Ave.
Las Vegas NV 89114
702-798-2230
Devereaux Barnes
Deputy Dir., Ind. Tech. Division
USEPA -• Washington, DC
401 M Street, SW, 
Washington DC 20460
202-382-7120
Robert C. Barrick
Senior Chemist
Tetra Tech, Inc.
11820 Northup Way, NE, Suite 100
Bellevue WA 98005
206-822-9596
                                548

-------
Richard F. Browner
School of Chemistry
Georgia Institute of Technology
Athens GA 30332
404-894-4002
Stephen L. Bugbee
OWEP
USEPA - Washington, DC
401 M Street, SW, (EN-336)
Washington DC 20460
202-382-5596
Keith L. Cherryholmes, Ph.D.
Coordinator of Environ. Studies
Hyg. Lab., The University of  Iowa
Oakdale Campus
Iowa City IA 52242
319-353-5990
         Colby
         Chemi stry
Bruce N.
Manager,
S-CUBED
Box 1620
LaJolla CA 92038
619-453-0060
Carol Colclough
IT Corporation
312 Directors Drive
Knoxvi1le TN 37923
615-690-321 1
Donald A. Cooper
Techn i c 5 an
E. I. du Pont de Nemours
Engineering Test Center
Wilmington DE 19898-7104
302-366-4674
                         & Co.,  Inc
Thomas F. Cull en,  Jr.
Div. Manager Analytical  Services
Environmental Research  Group,  Inc,
117 North First Street
Ann Arbor MI 48104
313-662-3104
                                 549

-------
Mary Lou Daniel
Laboratory  D i rector
S. Florida  Water Management  Dist.
P. O. Box V, 3301  Gun  Club Road
West Palm Beach FL 33402
305-686-8800
Susan de Nagy
Project Officer,  Ind. Tech.  Div.
USEPA - Washington, DC
401 M Street, SW,  
-------
Mitchell D. Erlckson, Ph.D,
Principal Chemist
Midwest Research Institute
425 Volker Boulevard
Kansas City MO 64110
816-753-7600
Barrett P. Eynon
Stat i st i ci an
SRI International
333 Ravenswood Avenue
MenTo Park CA 94025
415-859-5239
Paul Farrow
VG Masslab
Tudor Road
Altrincham, England WA14 5RZ
44 61 969 9222
Thomas E. Fielding
Chem i st
USEPA - Washington, DC
401 M Street, SW, 
-------
                                                       ", '  JJ;;,IHF!	Ill	Sf-'fil'"
 Gail  S.  Goldberg
 PermIts  D i v i s i on
 USEPA -  Washington, DC
 401 M Street, SW, 
-------
Earl M. Hansen
Laboratory Manager
Weston
256 Welsh Pool Road
Lionvilie PA 19353
215-525-0180
J. Ronald Mass, Ph.D.
Presi dent
Triangle Laboratories  Inc.
P. 0. Box 13485
Research Triangle Park NC 27709
919-544-5729
John C. Hendricks
Chem 5 st
American Electric
1 Riverside Plaza
Columbus OH 43215
614-223-1238
Power Serv. Corp,
Frank H. Hund
Industrial Technology  Divsion
USEPA - Washington,  DC
401 M Street,  SW,  
-------
 Sharon H. Kneiss
 Chemical  Manufacturers Association
 2501  M Street
 Washington DC 20037
 202-887-1180
 Margaret M.  Knight,  Ph.D.
 Weyerhaeuser Company
 WTC 2B25
 Tacoma WA 98477
 206-924-6002
 Wll11am G.  Krochta
 Manager,  Analytical  Research
 PPG  Industries
 Box  31
 Barberton OH 44203
 216-848-4161  Ext.  505
Marcia  Kuehl
Donohue and Associates
4738  N  40th St.
Sheboygan WI  53081
414-458-8711
Barbara A. Larka
Mass Spectroscop i st
Twin City Testing & Engineering Co.
662 Cromwel1 Ave.
St. Paul MN 55114
612-645-3601
Suzanne Lesage, Ph.D.
Organic Chemist
EPS/Wastewater Technology Center
867 Lakeshore Rd., P. O. Box 5050
Burlington, Ontario CD L7R 4A6
416-637-4504
Denis C.K. Lin, Ph.D.
Envr. Testing and Certification
284 Raritan Center Parkway
Edison NJ 08837
201-225-6707
                                                              '1	1

-------
John M» McGu1 re
USERA - Region IV
College Station Road
Athens GA 30613
404-546-3185
L. Wayne McMahon
Prog. Mgr. - Environmental Analysis
Martin Marietta Energy .Syst.-ORGDP
P. 0. Box P, MS 443
Oak Ridge TN 37922
615-574-9701
Ronald A. Michaud
Principal Chemist
Connect 5 cut Hea1th
P. O. Box 1689
Hartford CT 06101
203-566-3802
Laboratory
Raymond F. Mindrup
Market Development Specialist
SUPELCO,  Inc.
Supelco Park
Bellefonte PA 16823
814-359-3441
Lee Myers
CompuChem Laborator i es
P. 0. Box 12652
Research Triangle Park NC  27709
919-549-8263
Robert N i cho1 son
Laboratory Manager
Cal Lab East
P. O. Box 11106
Richmond VA 23230
804-359-1900
Teresa Norbert-K i ng
Env.  Research  Laboratory/ORD
USEPA -  Region V
6201  Congdon Boulevard
Duluth MN  55804
218-727-6692   Ext. 528
                                 555

-------
 John Do  Pfaff
 Monitoring and Support Lab., ORD
 USEPA -  Region V
 26 W St.  Clair St.
 Cincinnati  OH 45268
 513-684-7372
 Joseph A.  Poland
 Sen i or Chem 5 st
 CT State Dept. of
 10 C1 5nton St.
 Hartford CT 06106
 203-566-2787
Health Services
 Richard Posner
 Vice President - Metals Division
 United States Testing Company,  Inc.
 1415 Park  Avenue
 Hoboken NJ 07030
 201-792-2400   Ext.  320
David  H.  Powel1
Manager GC/MS
Environmental  Science
P. 0.  Box ESE
Gainesville FL 32602
904-332-3318
    & Engineering
William B. Prescott
Consultant
724 Hawthorne Avenue
Bound Brook NJ 08805
201-469-1198
James K. R 5 ce
James K. R i ce Chartered
17415 Satchel 1ors Forest
Olney MD 20832
301-774-2210
       Rd,
Russ Roegner
USEPA - Washington, DC
401 M Street, SW, 
Washington DC 20460
202-382-5410
                                556

-------
Richard Ronan
V i ce Pres i dent
VERSAR, INC.
6850 Versar Center
Springfield VA 22151
703-750-3000
Ann E. Rosecrance
Laboratory Director
JTC Environmental  Consultants,
4 Research Place
Rockville MD 20850
301-921-9790
                               Inc.
Dale R. Rushneck
Interface,  Inc.
P. O. Box 297
Ft. Collins CO 80522-0297
303-223-2013
Sam R.  Sax
Un 5 on Camp  Corporat i on
'Technical Center
Frank 1 in  VA 23851
804-569-4614
 Robert  B.  Schaffer
 Vice  President
 CENTEC  Corporation
 11260 Roger  Bacon Drive
 Reston  VA  22090-5281
 703-471-6300
Will? am Schnute
Environmental Marketing
Finnegan Mat
355 River Oaks Parkway
San Jose CA 951 34
408-946-4848
                         Manager
 Noel  Schwartz
 Vice  President
 United  States Testing
 1415  Park  Avenue
 Hoboken NJ 07030
 201-792-2400
                      Company,  Inc,
                                 557

-------
 Judy  Scott
 TRW
 One Space Park  01/2030
 Radondo Beach CA 90278
 213-536-2451
                                                              i&i i'"::;1
 Walter  M.  Shackel ford
 ERL/ORD
 USEPA - Region IV
 College Station  Road
 Athens  GA  30613
 404-546-3186
John C.  Sheats
Head of  Envir.  Sciences  Lab.
NC State Lab. of  Public  Health
P. O. Box 28047
Raleigh  NC 27611
919-733-7308
Nannette S i mon
Chemi st
Occidental Chemical
Long Road
Grand Island NY  14072
716-773-8655
James S. Smith
Director Weston Analytics
Roy F. Weston, Inc.
Weston Way
West Chester PA 19380
215-692-3030
George H. Stanko
Staff Research Chemist
Shell Development Company
P. 0. Box 1380
Houston TX 77001
713-493-7702
Edward Stigall
Chief, Inorganic Chem. & Ser. Br.
USEPA - Washington, DC
401 M Street, SW, (WH-552)
Washington DC 20460
202-382-7124
                                55B

-------
Paul Taylor
Calfornia Analytical Labs.,  Inc.
2544 Industry Blvd.
W. Sacramento CA 95691
916-372-1393
W 511i am A. Tel 1 5ard
Chief, E & M  Ind. Branch
USEPA - Washington, DC
401 M Street, SW, 
Washington DC 20460
202-382-7131
Kurt Theurer
AH i ed Corporat i on
Box 1021 R-CRL
Morristown NJ 07960
201-455-2141
Samuel To
Quality Assurance Officer
USEPA - Washington, DC
401 M Street, SW, (EN-338)
Washington DC 20460
202-475-8319
David F. Tompkins
Director of Analytical
CENTEC Corporation
2160 Industrial  Drive
Salem VA 24153
703-387-3995
Serv i ces
A11 an Tord 5 n i
Asst. Vice President  -  Metals
United States  Testing Company,
1415 Park Avenue
Hoboken  NJ 07030
201-792-2400   Ext.
       D5 v.
         Inc,
 V 5 ctor  Turosk i
 Supervisor  of  Analytical  Laboratory
 James River Corporation
 P.  O. Box 899
 Neenah  WI 54956
 414-729-8064
                                 559

-------
 Alexandra Uleeka
 Chem! st
 Rahway  Valley Sewerage
 Foot  of East Hazel wood
 Rahway  NJ 07065
 201-388-0033
Author i ty
Avenue
Marvin  Vests!
Chem i stry  Department
University of  Houston
Houston TX 77004
713-749-2675
Joseph  F.  Viar,  Jr.
Presldent
Viar and Company,  Inc.
300 North  Lee  St., Suite
Alexandria VA  22314
703-683-0885
  200
Jill Vol1merhausen
Martin Marietta  Environmental  Syst,
9200 Rumsey Road
Columbia MD 21045
301-964-9200
Robert D. Voyksner, Ph.D.
Research Triangle  Institute
P. 0. Box 12194
Research Triangle  Park NC 27709
919-541-6697
Dr. Dal las Wait
Laboratory Director
ERCO/Division of ENSECO
205 Alewife Brook Parkway
Cambridge MA 02138
617-661-3111
Ton i e Wai 1 ace
President
County Court Reporters,
27 E. Loudoun Street
Leesburg VA 22075
703-777-8645
 Inc.
                                560

-------
Bruce Wai 1 i n
Tech. Din.  Environmental
E. C. Jordan Company
92 Oak Street
Portland ME 04101
202-775-5401  Ext. 467
Lab. Svcs.
John P. Whitescarver
Presi dent
Wh i tescarver Assoc 5 ates ,
P. 0. Box 17088,  IAD
Washington DC 20041
703-661-8800
Inc.
Stuart Whitlock
Environmental  Science  &  Engineering
P. O. Box ESE
Gainsvi11e  FL  32602
904-332-3318
 Bruce  E.  W5 Ikes
 Project  Sc 5 ent i st
 Union  Carbide Corporation
 P.  O.  Box 8361,  Technical Center
 South  Charleston WV 25303
 304-747-4463
 Ed Winstead
 Chem i st
 The B i onet 5 cs Corporat 5 on
 20 Research Drive
 Hampton VA 23666
 804-865-2686
 Hugh E.  W 5 se
 Industrial  Technology Division
 USEPA -  Washington, DC
 401  M Street, SW, 
-------
Lauren Yel1e
Mass Spectrometrist
Arthur D. Little, Inc.
^5 Acorn Park
Cambridge MA 02140
617-864-5770  Ext. 2586
James C. Young, Ph.D., P.E.
Professor and Head, Dept. Civil
University of Arkansas
340 Engineering Bldg.
FayetteviUe AR 72701
501-575-4954
Eng
Jerry Zwe 5 genbaum
Eastman Kodak Company
B-34 Kodak  Park
Rochester NY 14650
716-722-3205
                               562

-------