EPA/600/A-93/053
Evaluation of Portable Gas Chromatographs
Richard E. Berkley
US Environmental Protection Agency
Research Triangle Park, NC 27711
Michael Miller
IIT Research Institute
Chicago, IL 60616
Joseph C. Chang
Battelle Columbus Laboratories
Columbus, OH 43201
Karen Oliver, Christopher Fortune, and Jeffrey Adams
ManTech Environmental Services
Research Triangle Park, NC 27709
ABSTRACT
Limits of detection, linearity of responses, and stability of
response factors and retention times for five commercially-available
portable gas chromatographs (PGC) were determined during laboratory
evaluation. The PGCs were also operated at the French Limited
Superfund site near Houston, TX during startup of bioremediation.
Concentrations of volatile organic compounds (VOC) at the site were
slightly above ambient background levels. Concurrent collocated grab
samples were collected periodically in canisters and analyzed by
Method TO-14 using a mass-selective detector. Canister data were
taken to indicate correct concentrations and were used to assess the
accuracy of PGC data. Durability, reliability, and complexity of
operation of PGCs were also evaluated. The principal goal of the
study was to determine the best way to use each instrument as a
monitor for airborne VOCs.
This paper has been reviewed in accordance with the U.S. Environ-
mental Protection Agency's peer and administrative review policies and
approved for presentation and publication. Mention of trade names or
commercial products does not constitute endorsement or recommendation
for use.
INTRODUCTION
Analysis of volatile organic compounds (VOC) in ambient air is
usually done by collecting samples at field sites and returning them
to a laboratory [1], Laboratory analysis can be done with state-of-
the-art procedures, but data are delayed and sample integrity could
become compromised during collection or storage. A portable gas
chromatograph (PGC) can produce data immediately and can deliver more
information at less cost than laboratory-based methods [2], We have
evaluated five commercially-available PGCs as monitors for VOC in air.
During the evaluation a field test was conducted at the French Limited
Superfund Site near Crosby, TX in January 1992.
-------
EXPERIMENTAL
PGCs were borrowed from manufacturers. They were evaluated in a
laboratory and then tested at the field site. The laboratory
evaluation was repeated for some instruments because they were
modified at the field site by factory personnel.
Portable Gas Chromatographs
All PGCs were microprocessor-controlled instruments capable of
unattended autosampling. Specific features and capabilities of each
PGC are summarized in Table 1. Information specific to individual
instruments, which was not included in Table 1 is given below.
The MSI 301 is an organic vapor monitor which operates on the
principle of a gas chromatograph. It can be adjusted to detect any
three compounds from a large group of potential analytes.
The SRI 8610 is a field-deployable laboratory chromatograph. It
can accommodate up to three detectors operating simultaneously. Other
available detectors are thermal conductivity, flame ionization,
nitrogen-phosphorus, thermionic ionization, electron capture, and
flame photometric. Carrier gas flow is regulated by a digital flow
controller, with which preselected flow rates can be set.
Standards
Calibration standards were prepared by dynamic dilution of
commercially-prepared 10 ppm mixtures of analytes in nitrogen
(Alfagaz, Scott) using humid zero air.
Laboratory Evaluation
Laboratory evaluation consisted of 1) learning how to operate
each unit as a monitor for VOCs in air, 2) calibrating it with
standard mixtures of VOCs in zero air, 3) evaluating the calibration
plots, and 4) calculating detection limits from calibration data.
Each unit was first set up and operated according to manufacturer's
specifications. After operating procedures had been learned and
proper function of the instrument had been verified, calibration was
performed by analyzing standard mixtures at five levels. Reliability,
durability, convenience of operation, and clarity and relevance of the
operator's manual were also considered.
Field Evaluation
The instruments were shipped to the hazardous waste site and
operated there. Data from the PGCs were compared with data obtained
from concurrent collocated whole-air grab samples collected in
evacuated passivated canisters. Accuracy, precision, ease of
operation, and reliability were determined.
Field Site
The French Limited Superfund site was an abandoned flooded sand
pit into which refinery waste had been dumped. Before remediation,
twenty-five feet of water overlay ten feet of sludge covering an area
of seven acres. The water above the sludge was clean enough to
support aquatic life, but volatile solvents were leaching from the
sludge into the surface aquifer. Bioremediation at the site was
performed using selectively-nurtured indigenous microorganisms. Pond
fluid was pumped out of the pond, injected with oxygen gas and
nutrients, then pumped back into the pond. Dredges loosened sludge
from the bottom and high-speed stirrers mixed it with the water. The
-------
bacteria fed on the sludge, converting it into biomass which
eventually decayed into carbon dioxide and water. Agitation of the
pond resulted in a slight increase in atmospheric concentrations of
organic vapors due to release of solvents dissolved in the sludge.
Concentrations were lower than anticipated from a pilot study in which
air had been used to provide oxygen. Diesel exhaust from dredges and
agitators contributed significantly to airborne VOC levels'. The
evaluation was conducted at a power-control shed which was located 20
feet south and 15 feet above the edge of the pond. Initial attempts
to conduct the study on an outdoor deck were defeated by inclement
weather, and all units were moved inside the shed where temperature
was maintained at 20°C. All units were connected to 110 volt AC
power. Each unit used its own sample pump to import outside air
through 1/8 inch OD stainless steel tubing. Calibrations were
performed periodically. Grab sampling was done by opening the valve
of an evacuated passivated canister while holding it within three feet
of the ends of the intake tubes as the PGCs sampled simultaneously.
Grab samples were returned to the EPA laboratory in Research Triangle
Park where they were analyzed by GC/MSD according to Method TO-14.
Canister data were assumed to accurately reflect concentrations of
compounds in the air sampled by the PGCs.
RESULTS AND DISCUSSION
The ability of an instrument to reliably analyze VOCs in ambient
air depends upon stability of retention times, adequate
chromatographic resolution, and the sensitivity, selectivity, and
stability of the detector. Stable performance is a significant
problem for any sensitive instrument which travels, and it was one of
two principal subjects of investigation in this study. The other
subject of investigation was accuracy, which was determined by
comparing PGC data with data from simultaneous canister grab samples.
Retention Time Stability
Retention time stability is crucial to accurate identification of
compounds. It was assessed in terms of the standard deviations of
retention times. Retention times from each calibrated compound were
averaged, and the standard deviation was calculated. Average
retention times are shown as horizontal bars in Figure 1, and standard
deviations are indicated at the end of each bar. They were usually
small but discernible. During the Scentograph field calibration, the
carrier flow rate was changed, producing two distinct groups of
retention times and a large standard deviation. This was probably the
result of human error and not due to a fault in the instrument.
Standard deviations in all other cases were small, but they needed to
be even smaller. The importance of retention time stability to
compound identification cannot be over-emphasized. It is vital to be
able to set carrier flow quickly and accurately. Column temperature
stability is even more important, because a shift in temperature
changes relative retention times and invalidates the entire
calibration table. Ideally, the standard deviations in Fig 1 should
have been almost invisible at the scale of the drawing.
Detector Stability
Stable detector response is crucial to accurate guantitation. It
was evaluated by observing the scatter in the calibration plots
(Figures 2-4) and by noting how closely the correlation coefficients
of the plots approached unity (Table 2). Significant scatter was
-------
taken to indicate an unstable response or possibly random sample loss.
The calibration plots showed some scatter, but all of the instruments
were able to report approximately correct data within the range
calibrated. In several cases precision was mediocre.
Sensitivity.
Sensitivity is indicated by the detection limit (TabLe 3), which
was taken to be the concentration corresponding to a response equal to
four times the baseline noise adjacent to the peak. Baseline noise
was difficult to evaluate in some cases. Neither the Scentograph nor
the Photovac sent a real baseline signal to external devices, and the
baseline of the MSI 301 was difficult to locate. Estimates of
baseline noise were made by visual inspection in those cases. The
lower the correlation coefficient, the less credible are calculated
detection limits. Detection limits were usually higher for the field
calibrations, reflecting the stress of field operation.
Resolution and Selectivity
Resolution of PGCs is generally lower than for laboratory
instruments because 1) shorter columns are used to get quicker
results, 2) autosampling schemes tend to introduce dead volume into
the system and thus broaden injection bandwidths, 3) field operation
tends to degrade performance because instruments are vulnerable to
vibration, rough handling, dust, and fluctuating ambient temperatures.
Iven the best resolution achievable cannot separate all of the
compounds present at background levels in ambient air. Typical
chromatograms obtained in laboratory analyses display as many as 200
peaks with up to four compounds coeluting in each peak. Detector
selectivity must be considered together with resolution when assessing
the ability of an instrument to distinguish target compounds from
other compounds in authentic field samples. To some extent,
selectivity substitutes for resolution in a PGC. The resolution of
toluene from benzene was calculated for each PGC in the study to be
twice the difference in retention times divided by the sum of the peak
widths at base height. These figures are listed in Table 4 with a
summary of the compounds to which each detector responds.
Field Performance
Ambient levels of analytes found during the field study were
scarcely above calculated detection limits, but PGC results generally
resembled data from the corresponding grab sample. Some discrepancies
were probably caused by poor mixing of air close downwind of the
source. Agreement between methods was not exact, but it was close
enough to show that all of the PGCs provided reasonable estimates of
the concentrations of compounds for which they were calibrated. In
Table 5 these data are summarized by the average of the averages for
the individual compounds and the average of the standard deviations
for the individual compounds. The ranges shown in Table 5 are the
difference between the best and worst agreement between paired
canister and PGC results among all of the compounds. These data are a
further condensation of data which were tabulated previously [3]. A
small average difference with a smaller standard deviation and a
narrow range would indicate close agreement between the two methods.
A large average difference with a small standard deviation would
indicate good precision but substantial disagreement due to systematic
error. A small average difference with a standard deviation of about
the same magnitude as the average difference would indicate data of
reasonable accuracy but mediocre precision, which would be expected
-------
when analyzing concentrations near detection limits. That is
generally what is seen in Table 5. A somewhat larger average
difference and a standard deviation larger than the average difference
with a broad range indicates some outliers among the data. A very
large average difference with a standard deviation as large as the
range would indicate little or no agreement between methods, and that
is not seen in Table 5.
CONCLUSIONS
All of the PGCs performed substantially as expected. It would be
futile to try to rank them on the basis of performance, since they are
not uniformly applicable to the same compounds. All instruments
evaluated in this study performed substantially as claimed. Each was
able to detect compounds at the levels encountered, generally with
reasonable accuracy. A variety of instruments and detectors may be
required to match capability with need. Choice of an instrument for a
particular application should be based upon consideration of its
particular features and capabilities. All of the instruments in this
study were seriously handicapped by exposure to cold wind, and they
all required shelter in order to function properly.
REFERENCES
1. Compendium of Methods for the Determination of Toxic Organic
Compounds in Ambient Air. Environmental Protection Agency,
Atmospheric Research and Exposure Ass
-------
TABLE 1. SUMMARY OF INSTRUMENT FEATURES AND CAPABILITIES
Power
Column
oc
Detector
Injector
Back-
flush
Carrier
Data
report
HNU 311
110 V AC
isothermal
to 200
PID
injection
loop or
syringe
port
yes
helium,
& others
printer or
computer
MSI 301
110 V AC
12 V DC
isothermal
65
SAW
precon-
centrate
& desorb
no
scrubbed
ambient
air
internal
processor,
external
computer,
or printer
Photovac
10S+
110 V AC
12 V DC
isothermal
40, 50
PID
injection
loop or
syringe
port
yes
ultra-
zero
air
magnetic
card,
external
printer,
or computer
Scento-
graph
110 V AC
12 V DC
isothermal
to 180
AID
precon-
centrate
& desorb
or syringe
port
no
argon
helium
computer
disk
SRI 8610
110 V AC
12 V DC
program
to 250
PID + ELCD
precon-
centrate
& desorb
or syringe
port
no
helium,
nitrogen,
scrubbed
ambient
air
external
printer or
computer
PID photoionization detector
SAW surface acoustic wave detector
TCD thermal conductivity detector
ECD electron capture detector
AID argon ionization detector
ILCD electrolytic conductivity
detector
-------
TABLE 2. CORRELATION COEFFICIENTS OF CALIBRATION PLOTS
HNU 311
MSI 301
Photovac 10S+
Scentograph
SRI 8610 PID
SRI 8610 ELCD
Lab
Field
Lab
Field
Lab
Field
Lab
Field
Lab
Field
Lab
Field
Benzene
0.958
0.995
0.985
0.971
0,997
0.994
0.939
0.964
0.990
0.963
NA
NA
Toluene
0.958
0.993
0.964
0.971
0.998
0.996
0.836
0.954
0,991
0.955
NA
NA
Tetra-
chloro-
ethylene
0.976
0.996
NA
NA
0.999
0.999
0.955
0.992
0.993
0.977
0.965
0.845
Chloro-
benzene
0.977
0.976
NA
NA
0.997
0.990
0.817
0.870
0.981
0.959
0.903
0.850
o-Xylene
0.996
0.966
0.922
0.979
0.996
0.988
ND
0.919
0.977
0.972
NA
NA
TABLE 3. DETECTION LIMITS CALCULATED FROM CALIBRATION PLOTS
(parts per billion by volume)
Benzene
HNU 311
Lab
Field
MSI 301
Lab
Field
Photovac 10S+
Lab
Field
Scentograph
Lab
Field
SRI 8610 PID
Lab
Field
SRI 8610 ELCD
Lab
Field
5C = sss s= ss ~ s as ssssa sr ssssis: =
1.9
2.5
2.4
9.0
1.1
0.9
5.1
6.7
3.4
4.9
NA
NA
Toluene
2.4
3.2
9.1
10.3
2.6
2.0
4.4
7.5
2.7
3.4
NA
NA
Tetra-
chloro- Chloro-
ethylene benzene
3.1
4.6
NA
NA
2.1
1.3
5.9
3.4
2.9
3,4
1.9
2.2
2.0
4.1
NA
NA
3.6
3.3
5.5
6.6
2.0
2.2
3.5
8.8
o-Xylene
5.3
9.3
4.5
10.1
14.4
20.6
27,9
46.6
1.1
3.0
NA
NA
ND Not detected. NA Detector does not respond.
-------
TABLE 4. CHRQMATOGRAPHIC RESOLUTION AND DETECTOR SELECTIVITY
Resolution of benzene from toluene was taken to be twice the
difference in retention times divided by the sum of the peak widths
at base height.
HNU 311
MSI 301
Photovac 10+
Scentograph
SRI 8610
SRI 8610
Resolution Detector
2.58 PID
1.09 SAW
7.77 PID
7.23 AID
5.32 PID
5.32 ELCD
Compounds Detected
Aromatics, halocarbons
Benzene, toluene, xylene
Aromatics, halocarbons
Universal
Aromatics, halocarbons
Halocarbons
TABLE 5. AVERAGE ABSOLUTE DIFFERENCES BETWEEN
TO-14 REFERENCE METHOD AND PORTABLE GAS CHROMATOGRAPH DATA
Average differences between paired canister and PGC results for the
five compounds were averaged to summarize performance of the
instrument. Standard deviations for individual compounds were also
averaged. Ranges are the differences between the largest and smallest
discrepancies among all five compounds.
Average of
per Compound
Average
Differences
4.80
HNU 311
15 Runs, 5 Compounds
MSI 301 1.75
13 Runs, 2 Compounds
Photovac 10+ 0.63
14 Runs, 5 Compounds
Scentograph 1.08
13 Runs, 5 Compounds
SRI 8610 PID 1.18
13 Runs, 5 Compounds
SRI 8610 ELCD 1.30
13 Runs, 2 Compounds
Average of
per Compound
Standard
Deviations
15.13
0.60
0.73
1.00
Maximum
Range of All
Differences
61.15
7.20
2.20
2.58
3.98
5.30
-------
hNU 31 ! P1D
Leb
HMD 311 PID
F,Ğld
PV 10S* PID
Leb
PV IBS* PID
Fi.ld
Lob
AID
FiĞld AID
SRI 8610 PID
Lob
SRI 8610 PID
r,.id
SRI 8610 ELCD
Lab
SRI B610 ELCD
Fi.ld
MS! 301 SAW
Lab
MSI 301 SAW
Fi.ld
TeLf~ecH 1
CH3 or
O~Xy 1 ene
ea
s
Ret.ent.lon Time
Figure 1. Retention time stabilities. Standard deviation of
retention time for each compound is represented by a shaded bar
centered on the end of an unshaded bar which represents average
retention time.
-------
0)
01
c
0
D.
Ul
(D
o:
MSI 301 Colibroiions
5000 ,
2500 .
5000 ,
500
Lab
-------
HNU 311 Col 1 brat ions
500000
400000 .
(D
w 300000 .
0
D- 100000 .
W
0)
o:
200000 ,
100000 .
Lob
IS
s
(9
(S
Photovoc 10S+ Co 1 i brat, i ons
IBB,
CD
CD
C
0
CL
0)
QJ
cr
50 .
Lob
s
s
V
ConcentroLion (ppbv) ConcenLroLion (ppb1
Benzene * TeLrach1oroeLhy1ene ° o~Xylene
Tol
uene
Chi
or-obenzene
Figure 3. Calibration plots from laboratory and field studies for
HNU 311 and Photovac 10S+.
,d
-------
1000,
SRI 8610 PID Calibrations
1000.
03
CD
C
0
a
CD
0)
see.
Lab
rv
<9
es>
s
cs
s
SI
SRI 8610 ELCD Calibration:
3000 ,
(L)
0)
C
0
CL
0)
(U
o:
1500
2B00
Lab
cs
IS
M U)
Ğ s
s s
(9
t~"
ro
ts
IS
ts
Concentration (ppbv) Concentration (ppb
v
Benzene
To 1uene
Tetrochloroethylene
Ch1orobenzene
o o~Xy1ene
Figure 4. Calibration plots from laboratory and field studies for
SRI 8610 photoionization and electrolytic conductivity detectors.
------- |