lies
;PA
lM.iliiiii.il
ulate
DC 20234
United States
Environmental Protection
Agency
:arch
W.isl 1C 20460
EPA 600 7 81 010
M.iy
Research and Development
A Procedure for
Establishing
Traceability of Gas
Mixtures to Certain
National Bureau of
Standards SRMs
-------
Errata
Cover Title: "A Procedure for Establishing Traceability of
Gas Mixtures to Certain National Bureau of Standards
SRM's"
Page 4 All (+) in the 3 equations should read (-).
Page 7 Line 8. "Cyliners" should read "Cylinders"
Page 20 Par. B, diagram, "Internal Response Standard"
should read "Internal Reference Standard"
•Co
Page 21 Equation 6 should read: t?i=Ri 5 So
Page 24 Equation should read:
K ~ 42K+ °ci2
-------
EPA-600/7-81-010
January
A PROCEDURE FOR ESTABLISHING TRACEABILITY OF GAS MIXTURES TO CERTAIN
NATIONAL BUREAU OF STANDARDS STANDARD REFERENCE MATERIALS
by
E. E. Hughes
Center for Analytical Chemistry
and
J. Mandel
National Measurement Laboratory
National Bureau of Standards
Washington, D.C. 20234
Contract No. EPA-IAG-D8-E684
Project Officer
D. Von Lehmden
Quality Assurance Division
Environmental Monitoring Systems Laboratory
Research Triangle Park, NC 27711
ENVIRONMENTAL MONITORING SYSTEMS LABORATORY
U.S. ENVIRONMENTAL PROTECTION AGENCY
RESEARCH TRIANGLE PARK, NC 27711
NBSIR 81-2227
-------
DISCLAIMER
This report has been reviewed by the Environmental Monitoring Systems
Laboratory, U.S. Environmental Protection Agency, and approved for publica-
tion. Approval does not signify that the contents necessarily reflects the
views and policies of the U.S. Environmental Protection Agency, nor does
mention of trade names or commercial products constitute endorsement or
recommendation for use.
11
-------
ABSTRACT
A procedure is described by which the concentration of commercially
produced gas mixtures may be related to certain Standard Reference Materials
(SRM's) currently offered by the National Bureau of Standards. The concen-
tration of the gas mixture, the Certified Reference Material (CRM) must be
close to the concentration of a particular SRM in order to reduce the error
involved in the comparative analysis. Statistical treatment of the trace-
ability process is given by an example as is the process by which a body
requiring traceability can evaluate the quality of the CRM.
iii
-------
CONTENTS
Disclaimer ii
Abstract ill
1. Introduction 1
2. Certified Reference Materials 2
Description 2
Preparation 2
Analysis 2
3. Suggested Analytical Procedure 5
Description 5
First Analysis 6
Second Analysis 6
Stability and Homogeneity 7
Estimation of the Uncertainty of the CRM 8
4. Audit of the CRM 9
Introduction 9
Environmental Protection Agency Audit Program 9
Selection of Samples 9
Analysis 10
Report of Results 10
Examination of Results 11
Certification 11
Appendices
A. Statistical Analysis of Data 12
B. Report Forms 25
C. Carbon Monoxide 30
iv
-------
FOREWORD
The role of the National Bureau of Standards in the Interagency
Energy/Environment R&D Program, coordinated by the Office of Research
and Development of the U. S. Environmental Protection Agency, is to
provide those services necessary to assure data quality in measurements
being made by the Federal, state, local, and industrial laboratories
participating in the interagency program. The work at NBS is coordinated
by the Office of Environmental Measurements and is conducted in the
Center for Analytical Chemistry, the Center for Radiation Research,
and the Center for Thermodynamics and Molecular Science. NBS activities
form part of the Characterization, Measurement, and Monitoring Program
Category and address the data quality assurance needs of air and water
monitoring programs. NBS efforts in support of data quality assurance
include:
0 Studies of the feasibility of production of
Standard Reference Materials which could be
used for the verification of performance audit
samples for quality control programs or used
for the calibration of field and laboratory
instruments.
0 The development and demonstration of new or
improved measurement methods, particularly when
needed for the certification of Standard Reference
Materials.
0 The evaluation and dissemination of data on the
physical and chemical properties of effluents,
products and raw materials of environmental
significance in energy production.
0 The provision of reference materials for the
evaluation and validation of monitoring methods.
This report is one of the Interagency Energy/Environment Research and
Development Series Reports prepared to provide detailed information
on the development of an NBS measurement standard or method.
WILLIAM H. KIRCHHOFF, ChMef
Office of Environmental Measurements
National Bureau of Standards
-------
1. INTRODUCTION
This procedure is intended to increase the availability of accurate
gas standards by the creation of a series of secondary standards, to be
referred to herein as "Certified Reference Materials" (CRM's). The CRM's
prepared according to this procedure will be related, within known limits of
uncertainty, to specific gaseous Standard Reference Materials (SRM's) offered
by the National Bureau of Standards. The CRM's are intended to supplement
the supply of existing SRM's and are not intended to offer traceability for
gas mixtures at concentrations other than those of the complementary SRM's
nor of gas mixtures containing components other than those in the SRM's.
The procedures described are based on experience at NBS relative to the
production and certification of gaseous SRM's and are intended to assure
reliable CRM's with reasonable effort but are not intended to reduce the
effort to a minimum. The acceptance of the CRM's by regulatory agencies, by
industry, and by other users, will depend on the reliability of the CRM which
in turn will depend on the integrity of the gas supplier relative to the
methods, conditions, and limitations of the procedure described here and on
the reliability and extent of any complementary audit program.
-------
2. CERTIFIED REFERENCE MATERIALS
2.1 Description
The CRMs will consist of compressed gas samples in cylinders prepared
in lots of ten or more of identical concentration. The average concentration
for any lot must lie within ±1.0 percent relative of the concentration of a
specific SRM, but preferably the concentration should be as close as techni-
cally possible to the SRM in order to reduce to a minimum any errors which
might arise in subsequent analyses of the CRMs. The concentration for a
specific CRM will be determined by analysis with a specific SRM and the
uncertainty assigned to the concentration of the CRM will consist of the
uncertainty of the SRM and the added uncertainty resulting from the inter-
comparison of the CRM to the SRM. The period of time during which the
concentration of the CRM is certified, and the presence of any impurities of
consequence will be reported by the supplier.
2.2 Preparation
The CRN's will be prepared in lots of at least ten (10) cylinders all
filled from an homogeneous bulk mixture prepared in a single container
or in several containers ganged together in such a manner that delivery occurs
simultaneously from the several containers or by use of a dynamic blending
system. The reagent gases with which the bulk mixture is prepared must be
analyzed for the major component(s) and for minor constituents of interest.
The concentration of the component of interest in the bulk mixture must lie
within ± 1 percent relative of the particular SRM which is being duplicated.
The cylinders into which the bulk mixture is transferred and which will become
the CRM's must be clean, of known compatibility and history, and must be
treated to assure stability of the mixtures which will be contained. The
cylinders will be equipped with valves of appropriate material which conform
to Compressed Gas Association (CGA) recommendations for the particular gas
mixture.
After filling, an "incubation" period must be allowed before final
analysis. This period will vary depending on the nature of the gas mixture
and will be described in the appendices.
2.3 Analysis
The analysis consists of a comparison of each CRM to the appropriate
SRM. The requirement that the CRM have a concentration within ± 1 percent
relative of the SRM reduces the error arising from the intercomparison and
simplifies the calibration procedure for the analytical instruments. In
most cases the characteristics of the analyzer are such that the response of
-------
the instrument over a small concentration range (1 percent relative or less)
can be described by the linear equation
y = mx + b
where y is the signal, m is the instrument sensitivity in signal units per
unit of concentration, x is the concentration and b is a constant in signal
units. Consequently, it is only necessary to determine that the sensitivity
does not change significantly over the interval between the concentration of
the CRM and the SRM in order to compare the two signals directly. It follows
then that the closer in concentration are the CRM and SRM, the less possibil-
ity exists of error in assigning a value to the CRM.
The general characteristics of the analyzer response relative to concen-
tration is determined by constructing a calibration curve using the SRM
against which the CRM is to be compared as one calibration point and at least
two other gas mixtures for a minimum of three calibration points. These
additional calibration gases should be SRM's, one higher in concentration and
one lower in concentration as compared to the SRM which is being used to
analyze the CRM. Obviously this is not possible when the CRM is being
compared to either the highest or lowest concentration SRM in a particular
series. In this case, the CRM should preferably have a concentration slightly
lower than the SRM, if it is the highest SRM in a series, or slightly higher
if the SRM is the lowest in a series. If the SRM is the lowest in a series,
the use of a "zero" gas may be feasible. However, either the use of a zero
gas or extrapolation beyond the calibration curve established with SRM's
should be done with caution.
The ideal case where the instrument response over a wide range is
directly proportional to concentration can be readily recognized by measuring
the signal generated by the three SRM calibrating gases and dividing each
by concentration to obtain the sensitivity (m) at the three concentrations.
If the response is linear and if the calibration line passes through the
origin, the value for M will be the same at all three concentrations and the
equation reduces to y = mx. If m is constant over a wide range of concentra-
tions, then it follows that it will be constant over a smaller range. In
this case, the concentration of the CRM is simply
°r
m
where I is the signal generated by either the SRM or the CRM and C is
concentration.
-------
In the somewhat less ideal case where the response is linear but does
not pass through the origin, the value of the signal divided by the concen-
tration will not be equal at the three concentrations but a straight line of
slope m and intercept b can be fitted to the data. It again follows that
if m is constant over a wide range of concentrations, it will be constant
over a small interval and the concentration of the CRM is expressed as
c - c
CRM I_nvr + b SRM or
oKrl
CCRM = - m -
m =
CSRM
A third case is that of an analytical instrument with non-linear response.
In this situation a calibration curve is constructed using at least three
SRM's and the concentrations are determined using the derived equation for
the curve. In most cases the component in a CRM which is being analyzed will
be measured with an instrument which is essentially linear over small ranges
of concentration. In other words, the sensitivity of the instrument — the
response in signal units per unit of concentration — will not change within
the limits of precision of the instrument over short intervals of concentra-
tion. If the sensitivity is found to be essentially constant over the
internal of ±1% of the concentration of the SRM which is to be reproduced by
the CRM then the sensitivity measured for the SRM can be used to calculate
the concentration of the CRM. In this case, the closer the CRM is in concen-
tration to the SRM the smaller will be the error introduced by assuming short
range linearity.
It is essential in analyzing the CRM's that the analyses be made with
the highest degree of precision possible in order that the total uncertainty
of the concentration of the CRM be not appreciably larger than the total
uncertainty of the concentration of the SRM with which it is compared.
The principal elements which contribute to the imprecision of this
comparison include sensitivity and instrument instability. The sensitivity
of the instrument should be high enough so that differences of concentration
of 0.1 percent relative can be measured. Instruments are available which are
capable of measuring all current SRM's with this sensitivity. However, most
instruments exhibit drift to one degree or another. The drift has two
components, short term drift or noise and a longer term drift characterized
by a slowly changing signal under constant operating conditions. The effect
of short term drift can be minimized by a number of techniques including
the multiple analyses of each sample, and signal averaging. The effect of
long term drift can be minimized by frequent calibration of the instrument.
The characteristics of each analytical system should be considered
individually to determine the optimum frequency of calibration and the
-------
approach to compensation or control of the effect of noise. However, it
should be noted that the effort expended in minimizing these effects will be
reflected in the precision of the analyses relating the CRM to the SRM. In
general, if duplicate analysis of single samples consistently agree within
± 0.1 percent relative, then the effect of noise is at an acceptable level
and if repeated analysis of the same sample performed at intervals during the
analysis of a lot agree within the same limits, then the long term drift is
at an acceptable level.
3. SUGGESTED ANALYTICAL PROCEDURE
3.1 Description
The analytical procedure recommended here is considered to be the
minimum effort necessary to adequately reduce the possibility of accepting
an unstable CRM. All CRMs must be stable and each batch should be homoge-
nous. An unstable batch must be rejected while an inhomogenous batch may
be acceptable providing that the reason for the inhomogeneity is known and
further providing that the cause of the inhomogeneity will not result in a
future change in concentration not revealed by the test for stability.
The object of the analysis is to compare a lot of at least ten samples
of presumably identical concentration to an SRM and to express the concentra-
tion of each sample in the lot with error limits defined by the uncertainty
of the SRM and the uncertainty added by the analytical procedure. Obviously
the simplest procedure would involve direct analysis of the CRM with the
SRM but with most analytical methods this approach would be unduly extrav-
agant in terms of SRM consumption. Consequently, it is recommended that the
concentration be determined on a relative basis by comparing one sample from
the lot to all others in the lot. Thus only the one sample is compared to
the SRM and the concentration of all others in the lot can then be determined
from the measured ratio of each to the one sample. An additional uncertainty
is introduced in this procedure but the size of this uncertainty can be
reasonably small and the final uncertainty in the CRM will not be increased
significantly.
The sample from the lot with which the rest of the lot is to be compared,
the internal standard, is selected at random from the lot. The signal
generated by the internal standard is compared to the signal generated by
each sample. The internal standard is repeated throughout the sequence of
analysis so that any long term instrument drift can be compensated for
if necessary.
The resulting data can be expressed as the ratio of the signal generated
by the Individual CRM1s to that generated by the internal standard. The
ratio in all cases should be very close to 1.000 if the lot is homogenous.
-------
3.2 First Analysis
An initial analysis of the batch of CRM's should be performed as soon as
practical after the preparation of the CRM's. A sample is selected at random
from the lot to serve as the internal standard to which a number of samples
from the lot are compared. At least ten samples selected at random from each
lot should be analyzed during this first analysis. However, if the lot is
small (< 20 samples) it is recommended that all samples be analyzed.
The internal standard should be analyzed during the first analysis to
confirm that the concentration lies within ± 1 percent relative of the
appropriate SRM and to establish a reference for later analyses intended to
define the stability of the lot. The internal standard should be analyzed
at least ten times using SRM's to calibrate the instrument. Each of the
remaining nine or more samples selected is then analyzed in duplicate, using
the internal standard as a reference.
Some preliminary conclusions regarding the homogeneity and stability of
the batch may be drawn from the results of the first analyses. For instance,
if the batch was homogenous when transferred and if no subsequent reactions
have occurred randomly in the cylinders, the concentration of all samples
analyzed should be the same within the limits of precision of the analytical
method. The appropriate statistical treatment at this step is given in the
appendices. If the statistical test shows lack of homogeneity between the
samples, it is not possible at this time to determine whether such inhomoge-
neity is due to improper preparation or to reaction in the cylinders. If the
test shows reasonable closeness among the results for all samples, then
there is some assurance that the batch is homogenous.
If the analyses at this point indicate that no serious problem exists,
the entire batch is set aside for the required "incubation" period. If,
however, the batch appears to be somewhat inhomogeneous the analyses should
be continued to include all samples in the batch. An analysis of each sample
at this time may become critical later in differentiating inhomogeneity
resulting from preparation and inhomogeneity arising from instability.
3.3 Second Analysis
After the required incubation period has passed, all of the samples in
the lot are analyzed including those analyzed during the first analysis. The
first analysis served to define the concentration of the CRM and to test the
stability and homogeneity on a qualitative basis. Final decisions regarding
these factors depend on the second and more extensive analysis. Sufficient
time will have elapsed so that reactions in individual cylinders which
were of too low a rate to be recognized by the distribution of concentrations
in the initial sample lot, will be revealed by this subsequent analysis.
Decisions regarding homogeneity which may have been based on analysis of a
portion of the batch will now be based on results obtained on the entire
batch.
-------
3.3.1 Stability and Homogeneity
The stability of the lot is assumed if the concentration of each
CRM included in the first analysis has not changed appreciably at the time
of the second analysis (see Appendix A). An inspection of the data of
the first and second analyses may reveal gross instability and if such
is found the lot should be discarded.
In a lot consisting of cylinders prepared by transfer of a single
bulk mixture to a number of smaller containers, the composition of the
gas in all the smaller cyliners should be identical. Differences observed
between cylinders should reflect only the random errors in the analytical
process. Differences greater than this random error arise from two sources.
First, there are the differences which result from faulty mixing or from
dilution by residual gases left in the cylinder prior to transfer of the bulk
mixture. The second and more critical source arises from reactions,
chemical or physical, which occur in the cylinder after transfer.
Because these reactions are largely due to wall effects and because no two
cylinder walls are alike, the reactions generally proceed at different rates
in different cylinders. The two effects may be differentiated one from the
other by sets of measurements made over a period of time sufficiently long so
that the extent of reaction is greater than the error in the analysis.
For instance, if a lot is analyzed immediately after preparation and
individual samples are found to have concentrations that vary more than can
be accounted for statistically by analytical error, then the mixture is
either inhomogenous or some or all samples in the lot are unstable. If the
lot is reanalyzed after some time and each sample has retained essentially
its previous value then the samples are probably stable but the lot is
inhomogenous. If the concentration of one or more of the samples has changed
significantly, then it is plausible to conclude that the lot is unstable.
An inhomogenous lot need not necessarily be discarded but an unstable
lot must be discarded. An inhomogenous but stable lot in which the range of
values does not exceed ± 1 percent of the value of the SRM with which it is
to be compared may be used providing the reason for the inhomogeneity is
determined and that the cause does not otherwise affect the integrity of the
CRM. A lot prepared by transfer from a single bulk container should be
homogenous. If the bulk container consists of several large cylinders ganged
together to allow simultaneous delivery, then some inhomogeneity may arise
if the concentrations in the several bulk containers differ slightly. In the
special case of large scale dynamic dilution systems where two components are
mixed under constant flow conditions and are then compressed into the
cylinders in the lot, some inhomogeneity may result depending on the design
of the delivery system. If variations in concentration are observed in a
lot prepared by this method, the variation may be related to the position
of individual cylinders on the manifold and it is therefore important to
record the position of each cylinder for possible future reference.
-------
It should be reemphasized at this point that all samples in a batch
which is not entirely homogenous should be analyzed both during the first and
second analysis. It is essential that in cases where the samples are stable
but not entirely identical in terms of concentration, each sample be certified
separately as an individual.
The nature of the distribution of the concentration of samples in a lot
may give significant information concerning both the homogeneity and
stability of the lot. A lot which has concentrations that vary no more than
can be accounted for by the errors of the analytical method, is probably
homogenous and stable. A lot in which the range of values exceeds
significantly what might be expected on the basis of analytical error and in
which several samples are considerably below the average, probably is unstable,
A lot in which the range of values is large but all concentrations are essen-
tially symmetrically distributed around the average, was probably not prepared
or transferred in such a way that homogeneity was assured. A lot may be
found with concentrations grouped around two or more average values (bimodal
or multimodal distributions). This usually results from errors in the gas
blending or transfer operations. It should be noted, however, that the
detection of such situations is, except for very large numbers of cylinders,
statistically very uncertain. Occasionally a single sample in a lot will be
found with a concentration appreciably below the average for the lot. A
single such sample probably represents a fault in the preparation or
filling of the cylinder and may be omitted from the lot providing the lot is
otherwise stable and homogenous. The presence of more than one outlier in a
lot or the occurrence of a single outlier in several lots should be cause for
concern and in the absence of a reasonable explanation for the multiple out-
liers, the lots should be remade.
All of the evidence at this point must support the assumption that the
lot is stable, reasonably homogenous and that it meets all other requirements
for a CRM. If there is any suspicion that the lot may be unstable or
inhomogenous, then further analytical work should be conducted before, a
final decision is made.
3.3.2 Estimation of the Uncertainty of the CRM
The estimated uncertainty of the CRM is composed of at least three
components: the uncertainty of the SRM with which the CRM is analyzed,
the imprecision of analysis of the internal standard with the SRM, and the
imprecision of analysis of the CRM's with the internal standard. A dis-
cussion of the statistical procedure for estimating and combining these
three sources of error will be found in the appendices. Here we merely
mention that the second component is evaluated on the basis of the ten or
more measurements of the concentration of the internal standard, and that the
third component is evaluated from the variability of the results of all
samples in the lot. From the combined variability of these three
sources, an upper limit for the uncertainty of the concentration can be cal-
culated for each individual sample, for use in the certificate accompanying
the sample.
-------
4. AUDIT OF THE CRM
4.1 Introduction
At this point the producer should have sufficient evidence to make a
decision as to whether or not the batch will qualify as a CRM. This
decision must be based on the measured stability and homogeneity and on
the fact that the measured concentration of any sample in the batch lies
within ± 1 percent relative of the complementary SRM. The evidence for
compliance with the requirements for a CRM may be so overwhelming that a
visual examination of the data may convince the producer that all is
well. However, the final decision must await the results of an independent
examination of the data and of an independent analysis of at least some
of the samples. This independent analysis will be referred to as the
"audit program". Upon its completion, the data of the producer and
those resulting from the audit will be intercompared, using appropriate
statistical procedures of data analysis which are described in the
appendices.
4.2 Environmental Protection Agency Audit Program
The purpose of the Audit Program is to determine whether or not the
CRM's in a particular lot are of the concentration claimed by the producer.
This will be determined by an analysis of two or more samples from the
lot. The analysis will be performed by a laboratory chosen by the
Environmental Protection Agency (EPA). Analytical results obtained by
the auditor and the producer will be submitted to the National Bureau of
Standards for evaluation. If the results of the analysis by the auditor
confirm the analysis claimed by the producer, the EPA will be so informed.
If the results do not agree, either the auditor or the producer (or
perhaps both), will be asked to repeat the analysis. If this additional
analysis demonstrates the validity of the producer's value, the EPA will
be so informed. If, however, there is still disagreement, the samples
will be submitted to NBS for analysis. The auditor may be required to
analyze the CRM's for substances other than that for which it is certified
depending on the requirements specified in the appendix concerning the
particular substance.
4.2.1 Selection of Samples
The producer will submit a list of samples identified by sample
number and cylinder number to the auditor. The auditor will select two
samples at random which the producer will then send to the auditor. The pro-
ducer will inform the auditor of the particular SRM which the CRM is intended
to duplicate. This SRM will be identified by SRM number, sample number,
cylinder number and date of purchase.
-------
4.2.2 Analysis
The analysis of the CRM will be performed with an instrument
whose response characteristics are known at the concentration of the
CRM. The auditor will calibrate the instrument using SRM's, one of
which must be of the same nominal concentration as the CRM. The number
of other SRM's used in the calibration will depend on the characteristics
of the instrument but should include at least two others. The auditor
must be able to demonstrate the validity of the calibration of the
instrument in the concentration range between the SRM and the CRM.
Each sample will be analyzed ten times with the calibrated instrument.
The procedure and sequence of analyses will be planned by the auditor and
should be such that a high degree of precision is obtained.
The average concentration for each sample and the standard deviation
are to be reported. In general, the precision of the methods used in
the analysis of gas mixtures of a type represented by the CRM should yield
standard deviations which do not exceed 0.5 percent relative. If this value
is exceeded, then the procedure should be carefully examined for sources of
imprecision and the analyses should be repeated. In calculating the average
and the standard deviation of the ten measurements for each sample, it is
not permissible to omit or change any values (possible "outliers") except
where experimental conditions were known to be inappropriate during the
measuring process.
4.2.3. Report of Results
Both the producer and the auditor will submit to NBS a comprehensive
report of their analyses. Each individual analysis result for each of
the samples analyzed shall be reported. The producer will also report
the SRM's, identified by number, sample number, cylinder number, concentration
and purchase date, which were used in the calibration of the instrument.
A brief description of the method, the calibration procedure and the
measured instrument sensitivities at the calibration points will also be
submitted.
The auditor will submit the individual result of each analysis on
each of the two samples, with information concerning the SRM's used, the
instrument employed, the calibration procedure and any other analytical
results obtained as may be required for the particular mixture.
The report may be made on the forms shown in the appendices or by
letter containing the same information.
10
-------
4.3 Examination of Results
The analytical data submitted by the producer will be examined by
NBS to assess the homogeneity and stability of the batch. The results
reported by the auditor will be used to confirm or reject the concentration
claimed by the producer. It is assumed that an unstable lot would be
readily evident to the producer and would not have been retained for use
as CKM's. It is further assumed that an inhomogeneous lot would be
recognized and would not be retained except in the very special case
where the reason for the inhomogeneity was known and where the stability
was unequivocally demonstrated. Such batches will be individually
considered and further analyses may be requested of the producer.
4.4 Certification
If the particular batch of CRM's is reasonably homogenous and
definitely stable and if there is no significant difference between the
concentration claimed by the producer and that found by the auditor,
both the EPA and the producer will be notified. The producer may then
certify the samples in that particular batch as having met the requirements
of a Certified Reference Material.
In order to assure the purchaser of a CRM that it has been prepared
according to the procedures described herein, and that the results of
the EPA audit and the NBS analysis of the analytical data confirm the
producers concentration within acceptable limits, the producer will
supply the purchaser with the following:
1. A copy of the letter from NBS to EPA and to the producer stating
that the particular lot meets the requirements for CRM's. The
letter will list each cylinder in the lot identified by cylinder
and sample number and will show the producers value for the con-
centration of each sample in the lot. In addition the letter will
identify the samples audited by EPA and will give the concentration
determined by the audit.
2. A producers "Certificate of Analysis" which will include as a
minimum the cylinder and sample number, the concentration for the
particular sample and the uncertainty assigned to that concentration.
In addition, the period of certification and the date of the beginning
of the period will be given. Finally, the certificate will identify
the NBS-SRM used by the producers to establish traceability of the
CRM.
11
-------
APPENDIX A
Statistical Analysis of Data
I. Introduction
The data submitted to NBS by the producer and the auditor will be
examined in detail to determine that the candidate lot of CRMs is of the con-
centration claimed and that the lot is stable and homogenous. The statistical
treatment is illustrated in this appendix by means of an example based for the
most part on real experimental data. However, it was necessary to synthesize
some data, particularly in describing the auditor results, to produce an
example illustrative of the whole process. It should be noted that this
example does not define the degree of measurement precision required of a CRM
and in many cases the precision shown would not be attainable because of such
factors as instrument sensitivity and the chemical and physical properties of
the gases involved.
II. Illustrative^ Example^ jmd^ Statistical Analysis
1. Producer's Calibration Data:
The producer provided the following data on the calibration of his
instrument :
_ TABLE 1.
SRM No. Concentration* Signal or Sensitivity
1677
1678
1679
9.
44.
97.
67
9
1
+
+
+
.09
.5
.9
ppm
ppm
ppm
11.
51
54.10
114.
86
mV
mV
mV
(signal)
(signal)
(signal)
* The ± values are the total uncertainties given in the SRM certificates.
12
-------
A linear regression of signal vs. concentration provides the equation:
signal = .464 + 1.181 (concentration)
The fit to the straight line is good, and is only minimally affected by
making the intercept zero. This leads to the modified equation:
signal = 1.187 (concentration)
The fit provided by this equation is shown in the following table where the
fitted values in the last column should be compared with the observed values
of signal:
TABLE 2
Signal
Observed Fitted
11.51 11.48
54.10 53.30
114.86 115.26
We conclude that the instrument gives essentially a linear response and is
in good state of calibration. However, since the samples to be analyzed have
a concentration close to that of SRM 1678 (about 45 ppm), a sensitivity value
can be adopted that is based on a calibration line going through the origin
and through the point whose abscissa is the certified value of SRM 1678 and
whose ordinate is the measured value for this SRM. This gives the calibra-
tion equation which will be used for all further analyses:
signal = 1.205 (concentration)
13
-------
2. Analysis of Internal Reference Standard
One of the cylinders was selected as an Internal Reference Standard.
The measurements,* in mV are shown in the second column of Table 3. The
third column shows the concentration values, in ppm, obtained by using the
sensitivity 1.205 mV/ppm for the conversion.
TABLE 3
Measure-
ment No.
1
2
3
4
5
Signal Concentration
(mV) (ppm)
53.7623 44.616
53.7382 44.596
53.7454 44.602
53.7948 44.643
53.7719 44.627
Measure- Signal Concentration
ment No. (mV) (ppm)
6 53.7683 44
7 53.7454 44
8 53.7237 44
9 53.7707 44
10 53.7767 44
.621
.602
.584
.623
.628
The average concentration for the Internal Reference Standard is:
x = 44.6139 ppm
with a standard deviation among single replicate measurements of
s = .018 ppm
x
The standard error of the average value for this cylinder is:
.018
s- =
X
\/10
.0057 ppm
* Note: It is not required that either producer or auditor provide the actual
measured value of the signal. However, the value of the concentration
calculated from the signal should be expressed with sufficient digits to
reflect the magnitude of the signal. In other words, don't round off the
calculated concentration when submitting the data for evaluation.
14
-------
3. Analysis of All Samples
The Internal Reference Standard is now used by the producer to analyze
all the cylinders in the lot. This is generally accomplished by measuring
the ratio of the signal for the sample to that for the internal standard.
Table 4 shows the signal ratios, denoted R., and the corresponding calculated
concentrations, denoted C^, using the value 44.6139 ppm for the internal
standard. Thus, each concentration is obtained by the equation:
C = R (44.6139)
where
R =
signal for sample
signal for internal standard
15
-------
TABLE 4
Sample
5
6
11
13
15
21
23
29
30
31
33
34
35
39
45
46
48
49
50
51
52
Average
Std. dev.
First Analysis
R C
1.000177
1.000102
1.000904
.999906
.999350
.999520
.999563
.999514
.999881
.999452
44.622
44.618
44.654
44.610
44.585
44.592
44.594
44.592
44.609
44.589
44. 6065
.021
Second Analysis
R C
.999601
1.000592
1.000592
1.000406
1.000392
.999482
. 999807
1.000438
1.000368
1.000353
1.000307
1.000050
1.000480
1.000210
1.000050
. 999415
.999168
.999800
.999304
.999060
.999321
44.596
44.592
44.640
44.632
44.631
44.591
44.605
44.633
44.630
44.630
44.628
44.616
44.635
44.623
44.616
44.588
44.577
44.605
44.583
44.572
44.584
44.6099
.022
Difference
between
Duplicates
.026
-.014
.049
-.020
-.043
-.024
-.029
-.024
.026
.005
-.0048
*
.030
* Since the numbers in this column are differences of two measurements,
the standard deviation found for this column is /2 times that for
single measurements. Consequently, the standard deviation for single
measurements derived from the differences is (.030)//2~ = .021. This
value does not include possible variability between samples.
16
-------
The last column of Table 4 is used for two purposes : a) to obtain an addi-
tional estimate of the standard deviation among replicates, and b) to test
whether a significant systematic shift has occurred between the two sets
("first analysis" and "second analysis") . The estimate of the standard
deviation, converted to a single measurement basis is .021 and is consistent
with that obtained previously (s = .018). As to a possible shift, there is
no evidence for such an occurrence. (A test of significance can be carried
out as follows:
.0307/10'
This value is not significant, when compared with the critical value of
Student's t^, for 10-1 = 9 degrees of freedom.)
The standard deviations, .021 and .022, for the two sets of values in
Table 4 are mutually consistent. Moreover, since they are of the same order
of magnitude as the measurement error (as derived from replicate measurements
on the same sample), it may be concluded that no measurable heterogeneity
exists between the cylinders of this lot.
The best average value for the concentration of the lot is:
(44.6065 x 10) + (44.6099 x 21) .. ,nOB
- = 44 . OUoO
10 + 21
The standard error of this overall average is :
.022
/31
= .0040
This standard error does not include calibration error, errors in the
value of the SRM and in the value of internal reference standard.
17
-------
4. Auditor's Calibration Data
Table 5 shows the auditor's calibration results.
TABLE 5
SRM No.
1677
1678
1679
1680
Concentration*
9.79 ± .09
45.3 ± .5
97.1 ± .8
476. ± .4
Signal (counts)
19,688
91,642
197,990
979,130
* The ± values are the total uncertainties given in the SRM certificates.
The sensitivities (signal/concentration) are successively:
2011, 2023, 2039 and 2057 counts/ppm.
These values indicate a trend due, either to the presence of a blank or
to curvature, or to both.
A regression analysis shows a slight amount of curvature, but otherwise
the calibration data appear satisfactory. Since the samples are of the order
of magnitude of SRM 1678, the latter will be used for conversion of signal to
concentration, through the equation:
signal = 2023 (concentration).
18
-------
5. Auditor's Sample Measurements
The auditor made 10 replicate analysis on each of two cylinders,
no. 48 and no. 29. The results are shown in Table 6.
TABLE 6
Signal
89866
90044
89495
89570
90024
89997
89568
89708
89742
89852
Average
Std. dev.
Cylinder 48
Concentration
44.422
44.510
44.239
44.276
44.500
44.487
44.275
44.344
44.361
44.415
44.3729
0.099
Signal
89781
90121
89874
90050
89558
89874
90226
89888
89987
89655
Cylinder 29
Concentration
44.380
44.548
44.426
44.270
44.270
44.426
44.600
44.433
44.482
44.318
44.4396
0.101
19
-------
6. Evaluation of Uncertainties and Intercomparison of Results
a. Internal Comparison of Auditor's Results
The standard deviation of a single measurement made by the auditor
is 0.10. The results for cylinders 48 and 29 may be compared by Student's
_ 44.3829 - 44.4396
t ~~ ~~
_ . JL_
D 10
The _t-value is not significant. There is therefore no evidence of
heterogeneity between the two cylinders.
b. Total Uncertainty of Producer's Values
The value of each sample obtained by the producer is obtained by a
procedure represented by the following diagram:
Calibration Internal Response
SRM Factor Standard Sample
^^ ^** ^* ^h ^* ^T*
Calibration A B
The total error in the sample value is composed of four parts:
1) the uncertainty in the SRM value
2) the uncertainty in the calibration experiment
3) the uncertainty due to comparison A
4) the uncertainty due to comparison B
We use the rule (derived from the law of propagation of errors)
that the square of the relative error of the final value is equal to the sum
of the squares of the relative errors of the components.
More specifically, if "c", is the final concentration value obtained for
a particular sample (denoted by the subscript _i) in the lot, we have (from
step B):
-i=Ri CRef (1)
Where R. is the ratio of signals for sample ±_ to the reference sample, and
C r is^the concentration attached to the reference sample. But C - is
Ref Kel
obtained in step A by averaging ten values obtained each as
Signal
k
20
-------
where k is the calibration value derived from the calibration experiment.
In our case, k = 1,205. The average of the ten measurements may be described
by
CRef = f (3>
where S is the average of ten replicate signal values. Combining (1) and
(3) gives
c. - R. f (4)
The value of k is obtained experimentally from a single signal value
divided by the concentration of the SRM. Thus we may write:
k = f8- (5)
^o
where C0 is the concentration attached to the SRM by the certificate
and S0 is the signal corresponding to it.
Combining (4) and (5), we obtain finally:
t± - »4 S £ (6)
The law of propagation of errors gives:
(7)
The first term of the right side represents step B; the second term step
A and the third term the uncertainty of the calibration experiment itself.
The last term represents the uncertainty of the SRM used for calibration.
We now estimate these four components.
.
— (step B) is obtained from the last column of Table 4:
Ri
V
R± 44.61
21
-------
S (step A) is obtained from the calculations derived from Table 3:
3) aS
° (calibration experiment) has not been measured, but we can
t>0
assume the same precision as in step A for a single measurement:
"c m R — it
-IT- $§2 =4-03x10 (10)
4) ac
— —*• (uncertainty of SRM) is derived from the uncertainty given in
Co
the certificate. We will assume that this stated uncertainty is
equal to two standard deviations. Thus:
- f T7^ = 55.68 x 10 * (11)
Co 2 44.9
Adding the squares we have:
/a~ \2
[—- | = [(4.71)2 + (1.28)2 + (4.03)2 + (55.68)2Jx 10~8
V -c-. / L R
x ' = 3140 x 10
Hence: ,
a^±fc± • 56 x 10~
Since all c" are approximately the same and equal to
c\ = 44.62, we obtain
^ = 56 x 10~4x 44.62 = 0.25 (12)
We see that the predominant component of uncertainty, in
this case, is that of the SRM.
22
-------
c. Comparison of Producer's and Auditor's Values
By a calculation similar to that above, we obtain, for the auditor's
value for a particular cylinder, say c*:
ct
Step A is no_t_present for the auditor's data. The symbol ct represents
a count and ct an average of ten counts; C£ represents the value given by
the certificate for the SRM used in the calculation of the calibration
factor. We have:
22
/.1Q \ (.25 \
44.4 ) + 144.4J + Us.3J
222
= (7.12xlO~4) + (22.52xlO~4) + (55.19xlO~4)
2
= 3604xlO~8 = (60xlO~4)
Since c* = 44.4 for both samples analyzed by the auditor, we have:
°c* = 60xlO~4x44.4 = .27
We now obtain the following results (Table 7):
TABLE 7
Sample
48 29
Producer 44.58 ± .25 44.63 ± .25
Auditor 44.38 ± .27 44.44 ± .27
* The ± values represent standard errors in this table.
23
-------
It is apparent that the result obtained by the auditor for each sample is not
significantly different from that of the producer for the same sample. Thus,
the auditor's value in this case, substantiate those provided by the producer.
In general it may be assumed that there is no significant difference
between the concentration claimed by the producer and that found by the
auditor if the/following expression is satisfied.
- c* < 2
24
-------
APPENDIX B
REPORT FORMS
The following forms, A and B, are suggested report forms which will be
made available to both the producer and the auditor.
25
-------
FORM A. Report of Analysis of Certified Reference Materials
(Producer)
Prepared by; (Company name & address)
(Individual preparing report
& date)
Description;
Composition:
Corresponding SRM No.:
Date Blended:
Date Transferred:
Analysis
SRM's Used as Calibrants:
No. Sample No. Cylinder No. Concentration Purchase Date
1.
2.
3.
4.
5.
26
-------
FORM A (Continued)
Brief Description of Methods:
Brief Description of Calibration Procedure:
Calibration Data;
SRM No. Signal or Sensitivity (Identify)
1.
2.
3.
4.
5.
Results
Analysis of Internal Standard (if used)
Sample No. Cylinder No.
Concentration Date of Analysis
1.
2.
3.
4.
5.
6.
7.
8.
9.
10.
Average s.d.
27
-------
FORM A (Continued)
Analysis of CRM's
Concentration on Date Shown
Sample No. Cylinder No.
Average s.d._
28
-------
FORM B. Report of Analysis of Certified Reference Materials
(Auditor)
Prepared by:
Description:
Analys is;
SRM's Used as Calibrants:
No. Sample No. Cylinder No. Concentration Purchase Date
1.
2.
3.
4.
5.
SRM No.
1.
2.
3.
4.
5.
Brief Description of Method:
Brief Description of Calibration Procedure:
Calibration Data:
Signal or Sensitivity (Identify)
29
-------
FORM B (Continued)
Analysis of CEM
Sample No. Sample No.
Cylinder No. Concentration Cylinder No. Concentration
Ave. Ave.
s. d. s ..d.
30
-------
APPENDIX C
CARBON MONOXIDE
A. SRM's
The available SRM's are as follows:
CO in N,
SRM No.
1677
2635
1678
1679
2636
1680
1681
2637
2638
2639
2640
2641
2642
Nominal Concentration
10 pptn
25 ppm
50 ppm
100 ppm
250 ppm
500 ppm
1000 ppm
2500 ppm
5000 ppm
1 %
2 %
4 %
8 %
SRM No.
2612
2613
2616
CO in Air
Nominal Concentration
10 ppm
18 ppm
42 ppm
31
-------
The concentrations shown are in parts per million (ppm) by mole and in mole
percent. The actual concentration of a particular SRM may differ by more
than several percent relative from the nominal value and it is therefore
essential that the exact concentration of the SRM to be duplicated be known
before attempting the preparation of the CRM.
The maximum permissible levels of impurities in the CRM are as follows:
Water Vapor 10 ppm
Methane 1 ppm (CO in N )
Methane 5 ppm (CO in air)
Carbon Dioxide 5 ppm (CO in N )
Carbon Dioxide 400 ppm (CO in air)
B. Cylinders
CRM's must be packaged in new aluminum cylinders or in used aluminum
cylinders which have been used exclusively for mixtures containing carbon
monoxide in nitrogen or air. The history of any cylinder to be reused must
be known and confirmation must be obtained by analysis that the contents have
not been diluted or contaminated. Cylinders to be reused must contain
sufficient gas from the last filling to allow an analysis. The concentration
must lie within ± 1 percent relative of the original analysis. The cylinder
must be evacuated before refilling and the concentration of the CRM to be
added must lie within 2 orders of magnitude (± 100% relative) of the concen-
tration of the CRM previously contained.
C. Preparation of Analysis
Prepare a lot of at least 10 cylinders all at identical concentra'tion.
The concentration of carbon monoxide must lie within ± 1 percent relative of
the concentration of the specific SRM with which the lot is to be compared.
Analyze a representative number of cylinders by direct comparison to the SRM
or by comparing cylinders in the lot to a single sample selected at random
from the lot (internal standard from the lot). Analyze each sample at least
two times. Calculate the average of all results for all samples and calculate
the standard deviation of the average.
The lot should be put aside to "incubate" for at least one month after
which time all samples in the lot are analyzed. The results for the first and
second analyses should be statistically identical. If the two analyses are
the same, then it may be assumed that the lot is homogenous and stable. If
the analysis was performed by direct comparison to the SRM, then the analytical
value for the sample will be the certified value. If the analyses were
performed by comparison to an internal standard for the lot, then it will be
necessary to determine the concentration of the internal standard. This is
done by repeated comparison of the internal standard to the SRM. The concen-
tration measured ratio of each sample to the internal standard to obtain the
concentration of each sample.
32
-------
The uncertainty of the CRM is composed of the uncertainty of the SRM,
and any random errors introduced by the analysis of the CRM. The error of the
SRM, E, is given on the certificate for the particular cylinder and is defined
as the estimated upper limit of the total uncertainty. The estimated upper
limit of the uncertainty for the CRM should be calculated as follows:
E2, (Total Error) = 2^/[ ^SRM ] + a2 + b2 + c2
where E is the uncertainty of the CRM; a is the imprecision of intercompar-
ison of the internal standard with the lot; b is the imprecision of inter-
comparison of the internal standard and the SRM; and c is the error of the
calibration experiment if applicable (see Appendix A).
The following table lists the methods of analysis which are applicable
to the analysis of CRM's of carbon monoxide in nitrogen or air. The method
chosen is not necessarily limited to those shown providing that the method
chosen is of adequate sensitivity, precision and linearity. The precision
and sensitivity are evaluated during the analysis described earlier and are
reflected directly in the "imprecision of intercomparison". The linearity
must be defined in the concentration range between the CRM and the SRM.
TABLE 1. METHODS FOR THE INTERCOMPARISON OF CRM'S AND SRM'S OF CARBON
MONOXIDE IN AIR OR NITROGEN
Method Concentration of CO
Gas Chromatography - TC Detector 8% - 0.5%
Gas Chromatography - Ultrasonic Detector 1% - 0.05%
Gas Chromatography followed by catalytic 1% - 10 ppm
reduction to methane with flame ioniza-
tion detector
Non-dispersive Infrared 8% - 10 ppm
33
-------
D. Period cif Certification
Experience at NBS indicates that carbon monoxide mixture contained in
aluminum cylinders are stable for periods of three years or more. If no
change in concentration of the CRM's is indicated by the results of the
producer's first and second analysis, then stability for a period of two
years from the date of the second analysis is assured with a high degree of
probability. Any samples not sold within one year of the second analysis
should be reanalyzed and the period of certification should be appropriately
extended.
E. Recertification
CRM's of carbon monoxide in nitrogen or air which have exceeded the
period of certification may be recertified by the original supplier according
to the following provisions.
1. The pressure remaining in the cylinder must be greater
than 500 psi (3.4 kPa).
2. The ratio of the signal generated by the CRM when
compared to the original internal standard must be
the same as when originally measured within the limits
of uncertainty generated by both sets of measurements.
3. The concentration of the internal standard, determined
by comparison to the SRM with which it was originally
compared, must be the same within the limits of
precision of the comparison.
4. The concentration of the SRM must be known either by
comparison with a new SRM or the original SRM must have
been recently recertified by NBS.
5. If either the original SRM or the internal standard no
longer exists then no recertification can be made.
6. The uncertainty to be assigned to the recertified
value must include the added uncertainty, if any, of
the recertification process.
34
-------
TECHNICAL REPORT DATA
(Please read Instructions on the reverse before completing)
1. REPORT NO.
EPA-600/7-81-010
3. RECIPIENT'S ACCESSION NO.
4. TITLE AND SUBTITLE
A PROCEDURE FOR ESTABLISHING TRACEABILITY OF GAS
MIXTURES TO CERTAIN NATIONAL BUREAU OF STANDARDS
STANDARD REFERENCE MATERIALS
5 REPORT DA.TC
January 1981
6. PERFORMING ORGANIZATION CODE
7. AUTHOR(S)
8. PERFORMING ORGANIZATION REPORT NO.
Ernest Hughes and John Manclel
NBSIR 81-2227
9. PERFORMING ORGANIZATION NAME AND ADDRESS
National Bureau of Standards, Washington, DC 20234
U.S. Environmental Protection Agency
Office of Research and Development, EMSL
Research Triangle Park, NC 27711
10. PROGRAM ELEMENT NO.
6AP-1AAF60E1QN
11. CONTRACT/GRANT NO.
EPA-IAG-D8-E684
12. SPONSORING AGENCY NAME AND ADDRESS
Jointly Sponsored by:
National Bureau of Standards, Gaithersburg, MD
U.S. Environmental Protection Agency, EMSL, RTPNC
13. TYPE OF REPORT AND PERIOD COVERED
Final
14. SPONSORING AGENCY CODE
EPA/ORD/17
15. SUPPLEMENTARY NOTES
This project is part of the EPA-planned and coordinated Federal Interagency
Energy/Environment Research and Development Program.
16. ABSTRACT
This procedure includes the specifications and requirements that must be followed
by gas manufacturers during the preparation of compressed cylinder gas Certified Refer-
ence Materials (CRM). A CRM is a certified gas standard prepared at a concentration
that does not exceed + 1 percent of currently available National Bureau of Standards
Standard Reference Material (SRM) cylinder gases. The procedure includes specifica-
tions and requirements for: (1) preparation of compressed gas samples in cylinders
prepared in lots of ten or more of identical concentration with the average concen-
tration for the lot within + 1.0 percent relative to the concentration of a specific
SRM; (2) tests to verify compressed gas samples stability and within lot homogeniety;
(3) simultaneous submission by the gas manufacturer of analysis results to NBS and
cylinder gas numbers to USEPA (without analysis results); (4) random selection by
USEPA of two cylinders per lot for an USEPA performance audit analysis; (5) submission
by USEPA of audit results to NBS, and (6) decision by NBS whether to allow the gas
manufacturer to sell the lot of cylinders as CRM. A procedure for CRM for CO in N2
or air is described as Appendix C. Future appendices will be added for other CRM
including NO in N2> S02 in N2, C02 in N2, and 02 in N2-
17.
KEY WORDS AND DOCUMENT ANALYSIS
DESCRIPTORS
b.IDENTIFIERS/OPEN ENDED TERMS
c. COSATl Field/Group
Traceability, Certified Gas Standards,
Nitric Oxide, Sulfur Dioxide, Carbon
Dioxide, and Oxygen
18. DISTRIBUTION STATEMENT
RELEASE TO PUBLIC
19. SECURITY CLASS (This Report)
UNCLASSIFIED
21. NO. OF PAGES
42
20. SECURITY CLASS (This page)
UNCLASSIFIFD
22. PRICE
.$6.50
EPA Form 2220-1 (R»v. 4-77) PREVIOUS EDITION is OBSOLETE
------- |