£EPA
United States
Environmental Protection
Agency
Environmental Monitoring
Systems Laboratory
P.O. Box 93478
Us Vegas NV 89193-3478
EPA/600/4-89/025
June 1989
Research and Development
GC/FT-IR and
GC/FT-IR/MS
Techniques for
Routine Evironmental
Analysis
Project
Report
-------
GC/FT-IR AND GC/FT-IR/MS TECHNIQUES FOR
ROUTINE ENVIRONMENTAL ANALYSIS
by
Peter R. Griffiths and Charles L. Wilkina
University of California, Riverside
Riverside, California 92521-0403
Cooperative Agreement Number CR-811730-03
Project Officer
Donald F. Gurka
Quality Assurance and Methods Development Division
Environmental Monitoring Systems Laboratory
P.O. Box 93478
Las Vegas, NV 89193-3478
ENVIRONMENTAL MONITORING SYSTEMS LABORATORY
OFFICE OF RESEARCH AND DEVELOPMENT
U.S. ENVIRONMENTAL PROTECTION AGENCY
P.O. BOX 93478
LAS VEGAS, NEVADA 89193-3478
-------
NOTICE
The information in this document has been funded wholly or in part by the
United States Environmental Protection Agency (U.S. EPA) under Cooperative
Agreement Number CR-811730-03 to the University of California, Riverside,
Riverside, California. It has been subject to the Agency's peer and admini-
strative review, and it has been approved for publication as a U.S. EPA docu-
ment. Mention of trade names or commercial products does not constitute
endorsement or recommendation for use.
ii
-------
ABSTRACT
This report documents progress on three tasks related to the design and
testing of procedures and techniques for analyzing volatile and semivolatile
components of environmental samples. The tasks include:
(1) Develop and test a procedure to use infrared spectra to estimate the
approximate quantity of environmental contaminants injected into a gas
chromatograph.
(2) Prepare and test computer software to use for the on-tne-fly analysis
(both qualitative and quantitative) of mixtures of volatile components by
direct-linked gas chromatography/Fourier transform infrared/mass spectrometry.
(3) Develop, test, and construct a high-sensitivity interface between a
gas chromatography and a Fourier transform infrared spectrometer, and retrofit
the Fourier transform infrared spectrometer at the U.S. Environmental
Protection Agency's Environmental Monitoring Systems Laboratory In Las Vegas,
Nevada (EMSL-LV) with this device.
In response to Task 1, a method to estimate the quantity of each com-
ponent separated by gas chromatography for which an infrared spectrum can be
obtained waa developed. The method involves the calculation of a sample quan-
tity based on the top matches of a spectral search program, with the estimated
quantity calculated by weighting the sample quantity according to the hit
quality index of the best matches*
In response to Task 2, several interfaces between a gas chromatograph, a
Fourier transform spectrometer and a mass spectrometer have been developed.
Software to permit the identification of chromatographlcally separated com-
ponents by the combined information afforded by their infrared and mass
spectra has been written.
In response to Task 3, two methods of improving the sensitivity of the
interface between a gas chromatograph and a Fourier transform spectrometer
have been developed. Several approaches where the chromatographlc effluent is
passed through a heated light-pipe gas cell were constructed, and an optimized
configuration was delivered to EMSL-LV. In the second type of interface, the
eluites are trapped as small spots on a moving window held below ambient tem-
perature. The Infrared spectrum of each spot is then measured using
microscope optics. Two prototype systems, one in which the window is cooled
thermoelectrically and the other where liquid nitrogen cooling ia employed,
were constructed, and the feasibility of measuring identifiable spectra from
subnanogram compound quantities was demonstrated.
iii
-------
CONTENTS
Abstract Ill
Figures * vl
Tables ix
Abbreviations and Symbols x
1. Introduction 1
2. Project Description 3
3. Conclusions and Recommendations 4
Task 1 4
Task 2 4
Task 3 5
4. Quantification by GC/FT-IR without the Need for Identification
(Task 1) 6
5. Linked GC/FT-IR/MS (Task 2) 15
Introduction 15
CC/FT-IR/FTMS 15
Gas chromatography/Fourler transform mass infrared/
mass spectrometry using a mass selective
detector 29
Combined low-cost infrared and mass spectrometry for
GC/FT-IR/MS analysis 33
Maximum absorbance algorithm for reconstruction of
gas chromatograms from gas chromatography/
infrared spectrometry data 54
6. Improved Sensitivity for the GC/FT-IR Interface 56
Introduction 56
Light-pipe temperature and other factors affecting
signal In gas chromatography/Fourier transform
infrared spectrometry 57
Optimized GC/FT-IR system based on the use of light-pipes. . 59
Decomposition of certain compounds during GC/FT-IR
Analysis 60
Sample-trapping GC/FT-IR interfaces. .... 61
7. Review Papers * 70
8. Related Manuscripts 71
References 72
Appendices
A. Gas Chromatography/Fourler Transform Infrared/Mass Spectrometry
Using a Mass Selective Detector 79
iv
-------
B. Combined Infrared and Mass Spectral Library for GC/FT-IR/MS
Analysis 80
C. Maximum Absorbance Algorithm for Reconstruction of Gas
Chromatograms from Gas Chromatography/Infrared Spectrometry
Data 96
D. Llghtpipe Temperature and Other Factors Affecting Signal
In Gas Chromatography/Fourier Transform Infrared
Spectrometry 101
E. Optimizing the Optical Configuration for Light-Pipe Gas
Chromatography/Fourier Transform Infrared
Spectrometry 106
F. Evaluation of an Improved Single-Beam Gas Chromatography/Fourier
Transform Infrared Interface for Environmental Analysis . . . . U2
G. Analysis of the Metal-Catalyzed Decomposition of T- and
6-Haloesters by GC/FT-IR 120
H. Gas Chromatography/Fourier Transform Infrared Spectrometry
Under a Microscope. 126
I. Capillary Gas Chromatography/Fourier Transform Infrared
Microspectrometry at Subamblent Temperature 129
J. The Hyphenation of Chromatography & FT-IR Spectrometry 135
K. Coupled Gas Chromatography and Fourier Transform Infrared
Spectrometry 143
L. Linked Gas Chromatography Infrared Mass Spectrometry 172
-------
FIGURES
Number Page
1 A plot of the number of carbon atoms In straight chain
aliphatic carboxyllc acids vs. the ratio of the optical
density for OH stretching/optical density for symmetric
CH2 stretching 7
2 (Above) Functional group chromatograms and (below) Gram-Schmidt
reconstructed chromatogram of a synthetic mixture. The
peaks marked A, B, and C are isobutyl methacrylate,
nitrobenzene, and naphthalene, respectively 11
3 Search results for naphthalene 12
4 Search results for isobutyl methacrylate ,.... 13
5 GC/FT-IR/MS parallel split 16
6 Pulse sequence used for (above) EI-FTMS and (below) CI-FTMS . . . . . 17
7 Flame ionization detector trace of SOCAL fuel additive chromatogram . 18
8 FT-IR integrated absorbance reconstruction (2995-2870 cm'1) for
the SOCAL fuel additive chromatogram 20
9 Chromatogram reconstructed from electron Impact data for the
GC/FT-IR/EI-CI FT-MS analysis of the SOCAL fuel additive 21
10 Mass spectra from El (above) and CI (below) data for aliphatic
component of SOCAL fuel additive 22
11 Gram-Schmidt infrared reconstruction of chromatogram of peppermint
oil sample 23
12 FTMS electron impact data reconstruction of chromatogram of
peppermint oil sample 24
13 (Below) Infrared spectrum of first peak in Figure 11. (Above)
EPALIB reference spectrum of o-pinene 25
14 Representation of parallel chromatographic interface for
GC/FT-1R/MSD (95:5 split ratio for infrared light-pipe:MSD) .... 30
vi
-------
Number Page
15 Data communication for GC/MSD and GC/FT-IR. MSD data transfer
(post-run) to Nicolet 1280 computer via RS-232 32
16 Features of search strategies , 34
17 Logic of combined IR-MS search algorithm based upon availability
of spectral data 35
18 GC/FT-IR reconstructed chromatogram generated via Absorbance
Maximum algorithm 38
19 GC/MSD reconstructed chromatogram generated from total ion current. . 39
20 Ordinate expansion of Figure 18 demonstrating complexity of
sample 40
21 Ordinate expansion of Figure 19 41
22 GC/FT-IR and GC/MSD reconstructed chromatograms from unleaded
gasoline. (Latter half of chromatograms shown). Indicated
peaks identified as indane 44
23 (Above) Infrared library spectrum of indane. (Below) FT-IR
spectrum retrieved from file indicated in Figure 22 45
24 MSD spectrum retrieved from file indicated in Figure 22 and
identified as indane (molecular weight 118 amu) 46
25 GC/MSD and GC/IRD reconstructed chromatograms from a mixture of
nonane, isobutyl/methacrylate and tridecane at 125 ng/ul
concentration 47
26 GC/MSD and GC/IRD spectra from 10 ng injected of tridecane
(molecular weight 184) . 48
27 GC/MSD and GC/IRD spectra of Isobutyl methacrylate (12 ng injected) . 49
28 Reconstructed chromatograms for a mixture of 14 phenolic compounds
using GC/IRD/MSD system 51
29 Third component of phenol mixture. MSD and IRD spectra from
125 ng injected 52
30 Eighth component of phenol mixture. MSD and IRD spectra from
80 ng injected. • 53
vii
-------
Number Pagj^
31 Liquid nitrogen cooled stage and transfer line positioner used
in this study: (A) 50x25x2 mo ZnSe window, (B) brass cooling
plate; (C) coiled thin wall stainless steel tubing; (D) phenolic
insulating plate; (E) vertical aluminum plate; (F) adjustment
screws; (G) tension springs; (H) aluminum mounting plate;
(I) translational stage; (J) fiber optic (transfer line)
positioner 65
32 Optically scanned photograph of 12 ng acenaphthenequinone
deposition 68
33 Profile of 12 ng acenaphthenequinone deposition by absorbance
measurements measured with a 16 ym aperture 68
viii
-------
TABLES
Number Page
1 Data for EPALIB spectrum of 0.80 mg of phenol, search against
all other spectra in the database 10
2 Data for GC/FT-IR spectrum of naphthalene (150 ng injected) 10
3 Data for GC/FT-IR spectrum of isobutylmethacrylate (150 ng
injected) 14
4 FT-IR search results for aliphatic component of SOCAL fuel
additive 26
5 FT-IR search results for first peak of peppermint oil sample. .... 27
6 FT-IR library search results for peak identified as menthone
(peppermint oil) 28
7 Comparison of various AMM algorithms for the 45 "unknown"
compounds «.... 29
8 Result of IB./MS search for isobutylmethacrylate 36
9 Result of IR/MS search for 2-chloronaphthalene 37
10 Components returning 1 match on IR/MS search 37
11 Correlation results for two of the samples studied (the 23-component
mixture was injected at two volumes) 42
12 Result of IR/MS search for component identified as indane 43
13 Phenols in order of elution 50
14 Percentage of signal retained on heating light-pipe from ambient
to 250°C. ............ 59
ix
-------
LIST OF ABBREVIATIONS AND SYMBOLS
ABBREVIATIONS
ADC
AMAX
AMM
AQ
CAS
CI
El
EMSL-LV
EPALIB
FID
FT
FT-IR
FTMS
FWHH
GC
GC/FT-IR
GC/FT-IR/FTMS
GC/FT-IR/MS
GC/FTMS
GC/MS
HQI
H?
ILS
IR
IRD
1R-MS
KBr
LN2
MCT
MI
MIQ
M.Pt.
MS
MSD
NIH
NITS
Rxn
S.I.
SNR
TSCA
analog-to-digltal converter
chromatogram reconstruction based on maximum absorbance In
spectrum
accurate mass measurement
acenaphthenequlnone
Chemical Abstracts Service
chemical lonizatlon
electron impact
Environmental Monitoring Systems Laboratory, Las Vegas
EPA Library of Vapor Phase Infrared Spectra
flame lonizatlon detector
Fourier transform
Fourier transform infrared
Fourier transform mass spectrometry
full width at half height
gas chromatography
coupled GC and FT-IR spectrometry
coupled GC, FT-IR, and FTMS
coupled GC, FT-IR, and MS
coupled GC and FTMS
coupled GC and mass spectrometry
hit quality index
Hewlett Packard Corporation
instrument line shape
infrared
HP Infrared Detector
combined infrared and mass spectrometry
potassium bromide
liquid nitrogen
mercury cadmium tellurlde
matrix isolation
minimum identifiable quantity
melting point
mass spectrometry
HP Mass Selective Detector
National Institutes of Health
National Institute of Standards and Technology
reaction
search index
signal-to-noise ratio
Toxic Substance Control Act
-------
ABBREVIATIONS (continued)
UCR
U.S. EPA
ZnSe
University of California, Riverside
U.S. Environmental Protection Agency
zinc selenlde
SYMBOLS
a(\»)
A(v)
AD
Af
AR(VO)
AS
As(v0)
amu
b
C
cm
D*
f/
I
Ii
l.d.
K
mAU
mg
ms
ng
o.d.
°C
Pg
ppm
psi
2
Q
Qi
Qif
S.I.
sr
t
T
Ui
U(v,T)
absorptivity at wavenumber, v
absorbance at wavenumber, v
area of detector element (cm^)
area of beam at a focus (cm^)
peak absorbance of band In reference spectrum at wavenumber, v0
area of sample (cra^)
peak absorbance of band In sample spectrum at wavenumber, v0
atomic mass units
path length (cm)
concentration
concentration of reference
concentration of sample
centimeters
specific detectivity (cm Hzl£ W"1)
focal ratio
hit quality index
hit quality index for compound, i
internal diameter
degrees Kelvin
path length of reference cell
path length of sample cell
ith data point of library spectrum
milllabsorbance units
milligrams
milliseconds
nanograms
outside diameter
degrees Celsius
picograms
parts per million
pounds per square inch
sample quantity
calculated weighted average sample quantity
quantity of analyte i injected
quantity injected into GC calculated by probability-based
matching
search index
steradians
measurement time
temperature (K)
ith data point of spectrum of unknown
spectral energy density at wavenumber v for source of
temperature T (u/sr cm* cm'*)
xi
-------
SYMBOLS (continued)
— volume of light-pipe
W — watts
x — fraction of peak present in light-pipe
xmax — maximum fraction of peak present in light-pipe
Av — spectral resolution (cm"*)
Ug — micrograma
urn " micrometer
\iL — microliter
8 — optical throughput (cm2 sr)
£ —• optical efficiency
tlf — solid angle (ar) of beam at focus
xli
-------
SECTION 1
INTRODUCTION
Currently the U.S. Environmental Protection Agency (U.S. EPA) screens the
gas chroraatographicable portion of sample extracts for a few hundred target
organic compounds. Since over sixty thousand manufactured chemicals are
currently regulated under the Toxic Substance Control Act (TSCA), it is
apparent that many regulated compounds are not identified (let alone deter-
mined) in environmental samples. The present protocol for the analysis of
volatile and semivolatile compounds by the U.S. EPA involves a separation by
gas chromatography (GC) and measurement of the mass spectrum of each separated
component. Mass spectrometry (MS) is a fast, sensitive instrumental
analytical technique, and the linkage of GC and MS (GC/MS) represents a good
first step towards the total characterization of environmental samples.
Nevertheless, GC/MS does have several drawbacks.
First, the mass spectra of many isomeric compounds are very similar, yet
the toxicity and/or carclnogenicity of the individual Isomers may be very
different. For certain compounds It may be difficult to produce a discernable
molecular ion when lonizatlon is initiated by electron impact (El), and even
when chemical ionization (CI) methods are used the molecular Ion (M*) of cer-
tain compounds may be present at a level considerably below the concentration
of other Ions in the mass spectrum. Under these circumstances, a priori
interpretation of the mass spectrum may be difficult if not impossible.
Finally, even a semiquantltative determination of all peaks in the gas
chromatograms of complex environmental samples is usually Impossible without
an unambiguous assignment of each peak and the availability of calibration
factors for each component.
The application of Fourier transform infrared (FT-1R) spectrometry both in
place of, and In addition to, MS for the identification of components
separated by GC (GC/FT-IR and GC/FT-IR/MS, respectively) has several times
been proposed as an alternative to GC/MS for environmental analysis. Although
GC/FT-IR measurements can never be expected to have superior sensitivity to
the corresponding GC/MS measurements, the capability of infrared spectrometry
to distinguish between isomers and the reproducibility of GC/FT-IR spectra
from one instrument to another should enhance the potential for unambiguous
structural assignments of components of complex mixtures separated by GC.
Because of the greater repeatability of infrared spectra, GC/FT-IR also has
the potential to improve the capability of obtaining estimates of the quantity
of each component elutlng from the chromatographlc column, whether or not this
molecule has been unequivocally identified or has merely been assigned to a
certain chemical class.
-------
Although GC/FT-IR was first shown to be feasible as early as 1967, it did
not become a truly useful technique until the early 1980's when the sensi-
tivity of GC/FT-IR measurements was increased to the point that nanogram quan-
tities (typically 10-100 ng) of many components of complex mixtures separated
by capillary GC now yield identifiable infrared spectra. By reducing the
minimum identifiable quantity (MIQ) of GC/FT-IR measurements by two orders of
magnitude, the practical detection limits of GC/FT-IR would be comparable to
those of GC/MS for complex environmental samples.
The availability of a viable GC/FT-IR/MS system should greatly increase
the chemical information from environmental samples, thereby allowing better
risk assessment which in turn should lead to more efficient cleanup and dis-
posal measures at chemical waste sites. Several aspects of the development of
such a system are being addressed in this project.
-------
SECTION 2
PROJECT DESCRIPTION
Three casks were Co be performed under Chis CooperaCive Agreement:
e In Task 1, methods of obtaining semiquantitatlve estimates (±50 percent)
of Che concentraCion of each component separated by capillary GC for which
an infrared spectrum could be measured were Co be evaluated.
• In Task 2, methods of linking FT-IR and MS for identifying coraponencs of
mixtures separated by GC were to be investigated, with particular emphasis
on software development.
0 In Task 3, methods of improving the sensitivity of the GC/FT-IR interface
were to be studied.
Significant progress was made in all three tasks. Descriptions of
several aspects of Chis work has already appeared in the refereed literature.
Twelve of these papers are appended as Appendices A through L. In this
Report, the experimental methodology and important results of Che work per-
formed under each task will be discussed. When Che work is described in
detail in a published paper, only a summary will be given in the body of the
Report.
-------
SECTION 3
CONCLUSIONS AND RECOMMENDATIONS
TASK 1
a. Measurement of the absorbance of the most intense peak in the
spectrum or that of specific functional groups does not permit the quantity of
a given component present in the light-pipe to be accurately determined
without prior knowledge of its identity and the preparation and measurement of
calibration standards.
b. A method based on the weighted average of the quantities of the
samples given as the top few matches by spectral searching routines gives
semi-quantitative (±25 percent) estimates of sample quantities in many cases.
c. It is recommended that this technique be tested rigorously in cases
where the signal-to-noise ratio of the spectrum is low.
d* It is further recommended that the quantitative accuracy of presently
available GC/FT-IR reference spectra be tested, especially in cases where the
estimated quantity was found to be In error by more than ±50 percent.
TASK 2
a. The capability of measuring the masses of ions in a mass spectrum to
an accuracy of better than ±4 ppn through the use of a Fourier transform mass
spectrometer greatly increases the probability of correctly Identifying the
components eluting from a gas chromatograph, but requires expensive equipment.
b. An algorithm to correlate GC/FT-IR and GC/MS data was developed.
c. A search Index, utilizing prefiltered data from low-cost GC/FT-IR and
GC/MS interfaces allows many GC eluites to be identified by library searching.
d. Gas chromatograms can be reconstructed from stored GC/FT-IR spectra
at high sensitivity by plotting the maximum absorbance of each spectrum.
e. It Is recommended that both an improved and expanded FT-IR data base
and a larger combined MS-IR data base containing both types of high-quality
spectra should be developed for at least 10,000 compounds.
f. GC-IR sensitivity should be further improved to yield optimum results
in GC-IR-MS applications.
-------
g. It Is further recommended that investigations of the scope of
environmental applications of the Hewlett-Packard (HP) Infrared Detector (IRD)
system should be carried out.
h. GC-IR-MS interpretation software capable of making structural
inferences in those situations where either MS, IR or both types of spectra
for compounds are not found in libraries should be developed.
TASK 3
a. Two methods of reducing the effect of detector or preamplifier non-
linearity due to unmodulated radiation emitted by the hot light-pipe were
developed.
b. One of these, involving the use of a cooled light-pipe mounted imme-
diately after the hot light-pipe, was installed in a commercial GC/FT-IR
interface and improved the sensitivity by at least a factor of two relative to
ambient when the light-pipe temperature was raised above 200°C.
c. A detailed study of the characteristics of the beam transmitted
through a light-pipe showed that the solid angle of the output beam was much
less than that of the input beam.
d. The results of this study allowed an optimized GC/FT-IR interface to
be built which gave an approximately tenfold increase in sensitivity when com-
pared with a commercial GC/FT-IR using the same source and interferometer.
e. Methods of trapping eluites from a gas chromatograph as small spots
were developed.
f. Two interfaces where the eluites were trapped on a moving cooled win-
dow were built to demonstrate the feasibility of obtaining identifiable
infrared spectra from subnanogram quantities of GC eluites.
g. The first of these, where the window was thermoelectrically cooled,
showed the concept tc be feasible but did not allow very volatile or very non-
volatile molecules to be trapped efficiently.
h. A second Interface, where the window was cooled with liquid nitrogen
and the entire unit was mounted In a vacuum, enabled the major problems of the
previous Interface to be overcome.
i. It ia recommended that a more versatile version of this interface, in
which the window is mounted in an x-y stage be built to allow deposition of
all the components of a complex gas chromatogram to be deposited.
-------
SECTION 4
QUANTIFICATION BY GC/FT-IR WITHOUT THE NEED FOR IDENTIFICATION (TASK 1)
Several ways have been investigated by which infrared reference spectra
may be used to quantify eluites from capillary gas chromatographs, even though
the identity of each eluite may not have been unequivocally assigned. It was
recognized that the integrated absorbance of all bands in the infrared
spectrum of a given quantity of an analyte can vary by more than a factor of
ten. However, for molecules In a homologous series, the molar absorptivity of
a given vibrational mode Is fairly constant. For example, Nyquist [1] has
demonstrated a monotonic decrease in the ratio of the Intensity of the
0-H stretching mode to that of the symmetric C-H stretching mode of methylene
groups for a homologous series of linear alphatic carboxylic acids, see
Figure 1. From the data in this figure, it can be shown that if the intensity
ratio is divided by the number of CH2 groups in the molecule, the quotient is
5.5±0.2, indicating that the molar absorptivity for both the 0-H stretching
mode and the symmetric C-H stretching mode of each CH2 group is constant.
The absorptivity and molar absorptivity of vibrations associated with
different functional groups were then calculated for vapor-phase reference
spectra in the U.S. EPA Reference Spectral Library (EPALIB). For each entry
in this database, the dimensions of the cell were known, the sample quantity
injected into the cell is given in the header information, and the density
could be found from the Handbook of Chemistry and Physics [2]. It was
believed that the accuracy of the quantitative data given in the header of
each entry in EPALIB was quite good for the first half of Che samples to be
measured for this collection, but the accuracy decreased for reference spectra
that were measured about one year after the start of data collection. Thus we
evaluated the constancy of molar absorptivity using only the first
1000 entries in EPALIB, which contains a total of about 3300 entries.
In light of the observation that the molar absorptivity of the
0-H stretching mode is constant for a homologous series of alphatic carboxylic
acids (vide supra), the absorptivity and molar absorptivity of 110 alcohols in
this group were examined* The average value for the molar absorptivity was
1081 (arbitrary units) with a standard deviation of 1646. The range of these
values was obviously far too great to be applied in any quantitative technique
in which the absorptivity is assumed to be constant. The absorptivity of the
C"0 stretching mode of 116 carbonyl compounds In the group of 1000 reference
spectra was 12,600 (arbitrary units) with a standard deviation of 8010.
Although the relative standard deviation of the absorptivity of this mode was
a little less than that of the 0-H stretching mode of alcohols, the range was
still far too great to permit even a semiquantitative estimate of the quantity
of any component eluting from a gas chromatograph to be made. From this study
it was recognized that vapor-phase absorptivities of vibrational modes of
-------
o
6 8 10 12 U 16
Number of Carbon Atoms
18 20 22 24
Figure 1* A plot of the number of carbon atoms In straight chain aliphatic
carboxyllc acids vs. the ratio of the optical density for OH
stretching/optical density for symmetric CH2-
-------
molecules containing a certain functional group were only constant for
molecules in very limited chemical classes, but not for molecules containing
that chromophore in a wide variety of chemical environments.
It is well known that if a GC/FT-IR spectrum is compared to all entries
in the EPALIB database for a ^u_ali_ta_tiye_ identification of the elulte, the
closest matches are usually compounds whose chemical structure is similar to
that of the elulte* In most spectral searching programs, the measure of simi-
larity (or Hit Quality Index [HQI], I) is represented by the Euclidean
distance between the spectrum of the eluite and the reference spectrum. If
several compounds are listed with similar values of I, these compounds usually
have similar structures (unless the value of I is very large, in which case
none of them is chemically similar to the eluite)* If the HQI for the best
match is much less than that for the other entries, then there is a good
probability that that compound has been correctly identified. If several
entries have low values of I, then there is a definite ambiguity about the
true identity of the eluite but it is certain that its structure is similar to
the structures of all of the top entries in the list of best matches.
When the amount of a given component approaches the minimum identifiable
quantity (MIQ), the level of uncertainty in distinguishing between the
reference spectra of similar compounds increases. Nevertheless, the best
matches will usually be compounds with a similar structure to that of the
eluite, and the true identity of the eluite will generally appear in the top
four or five "hits" if its reference spectrum is in the database. Even if the
reference spectrum of the eluite is not in the library, the top hits usually
have structures which are in the same compound class as that of the unknown.
In most cases, the absorptivities of many bands in the spectrum of the eluite
will be similar to the absorptivities of the corresponding bands in the
spectra of the best matches found on library searching.
In view of this hypothesis, a method for the semlquantitative analysis of
compounds separated by GG and characterized by GC/FT-IR was devised. From a
knowledge of the flow-rate of the carrier gas and the dimensions of the light-
pipe, the true profile of the GC peak can be calculated in terms of the rela-
tive retention volume (rather than retention time). The^ fraction, x, of the
peak present in the light-pipe at any Instant can then be calculated. When
the eluite is present at its maximum concentration, let this fraction be xmax.
From the quantitative data In the header of each of the top N "hits" of the
spectral search, the apparent quantity of each of these N compounds can then
be calculated at Its maximum concentration in the light-pipe simply by calcu-
lating the ratio of the strongest band(s) in the spectrum of the eluite and of
the reference spectrum.
Let Ag(v0) and A^(v0) be the peak absorbancea of the most intense band in
the sample and reference spectra (at wavenumber v0), respectively. Similarly,
let G£ be the concentration of the compound when the refere~-:e spectrum was
measured (In units of g/mL), IR be the pathlength of the cell used for the
acquisition of the EPALIB database, and £5 be the pathlength of the light-pipe
used for the GC/FT-IR measurement. Then, by Beer's Law, the apparent concen-
tration of the sample in the light-pipe is:
-------
The quantity of analyte i injected, Q^, la therefore:
Qt - V* (2)
xmax
where V^p is the volume of the light-pipe.
Cg can be calculated for each of the top N hits of the library search,
giving N estimates of the concentration, Qj, Q2, . .., Qi(..., Qfl, Of these, QI
has the greatest probability of being the correct quantity, and QN will be the
poorest.
A weighted estimate based on the HQI, Ij , for each of the _top N hits is
then made to give the most likely estimate of sample quantity, Q. When the
HQI is given in terms of the Euclidean distance, as Is most commonly found in
commercial GC/FT-IR software:
N n N i
' Q- I £ / * if (3)
i-1 li i-1 Li
If the HQI is calculated using a probability-based matching scheme, so that
the best hit has the highest index, I'i, then:
_ N N
Q - £ Qtl'i/ I I't (4)
This approach can also be used for estimating the quantity of compounds
whose reference spectrum is not present In the database. Its potential
accuracy was tested by searching a given entry In EPALIB against all the other
reference spectra in the database. Including only a small number of the top
hits in this procedure (i.e., small N) does not adequately allow for errors in
the quantitative information in the headers of the reference spectra. Making
N too large, however, will result in compounds with significantly different
chemical structures to be included in the calculation of Q, so that the
accuracy will again be degraded. An adequate compromise was found when N ~ 5.
A typical result, found when the spectrum of phenol was searched against the
other spectra in EPALIB, is summarized in Table 1. The amount of phenol used
for the generation of the reference spectrum was 0*800 mg. It can be seen
that the first six Compounds in the list are phenols. It is not unexpected
that the quantity, Q, calculated to be in the cell is usually greater than
that of the unsubstituted phenol because of the added mass of the
substltuent(s) . Nevertheless the calculated value of Q is quite close to the
true value after the weighted average of the first five to seven compounds is
determined. The greater the number of compounds in the database^ with similar
structures to the target compound, the closer was the value of Q to the quan-
tity listed in the header of the reference spectrum of that target compound.
-------
TABLE 1. DATA FOR EPALIB SPECTRUM OF 0.80 MG OF PHENOL, SEARCHED AGAINST
ALL OTHER SPECTRA IN THE DATABASE
Match, i
1
2
3
4
5
6
7
*1
290
356
372
396
405
406
407
Phenol ,
Phenol,
p-Cresol
Phenol ,
Phenol,
Phenol,
Benzene,
Name
4(4*-tsopropylidene ,di-
p-phenyl
, crphenyl
m-phenyl
p-bromo
p-hydroxy
fluoro
Q(mg>
1.137
1,012
0.898
0.922
0.798
0.864
0.782
This approach therefore appears to have the potential for yielding semiquan-
tltatlve estimates of the Injected quantity of components separated by GC and
characterized by GC/FT-IR.
The success of this technique for practical GC analyses may be
Illustrated by a separation for which the functional group and Gram-Schmidt
reconstructed chromatograms are shown In Figure 2, (The sinusoidal variation
In the baseline of the Gram-Schmidt chromatogram was caused by one tower In
the heatless air drier being defective, so that the level of atmospheric CO2
In the instrument was varying throughout the run). A typical result, for
naphthalene, Is summarized in Table 2 and Figure 3. It can be seen that the
TABLE 2. DATA FOR GC/FT-IR SPECTRUM OF NAPHTHALENE (150 ng Injected)
Match, 1
1
2
3
4
5
6
li
0.27
0.44
0.46
0.54
0.58
0.60
Name
Naphthalene
Naphthalene, 1 -methyl
1-Naphthalenesulfonlc acid, dihydrate
Naphthalene, 1-chloromethyl
1-Naphthalenecarbonitrlle
Naphthalene, 1-ethyl
QCog)
130
126
98
107
140
134
10
-------
top five hits are all substituted naphthalenes, with naphthalene itself as the
#1 match. The weighted average for the first five hits is 140 ng, which com-
pares favorably with the actual injected quantity of 150 ng.
. 5-21-67
Figure 2. (Above) Functional group chromatograms and (below) Gram-Schmidt
reconstructed chromatogram of a synthetic mixture* The peaks
marked A, B, and C are isobutyl methacrylate, nitrobenzene, and
naphthalene, respectively.
11
-------
5 =? -L I5c'. 2=! 7} 2.55 =5-52-5 -V.3^ r-rLENEI^3C\ I •" :_£
£.=-.u:55:553) 2.=* 35-52-2
3 :PALl3eCU95) 2.^5 35-^7-2
2 i?AL32C7Se> 3.44 53-12-3
=?AL38( 1357) 0.27 91-20-3
TIC ES .3IM1N =1S. 15
1233
Figure 3. Search results for naphthalene.
12
-------
5 =?ALS8C256: 2.32 '26-31-2
5 EP-LI39CZZ45) 2.23 3?-53-2
^ .£?-_:52Ci-3} 3.23 : 2575-^3-1
3 i?aL38C:354; 2.22 3773-92-4
2 £=-L33C2^4) 3. IS 97-98-1
I E?ALI58C=S5) 0.29 97-8S-9
.^OACRYLIC ACID
XC8) RETENTICN TINC IN .aiMZN -10.33
izez
Figure 4. Search results for isobutyl aethacrylate.
13
-------
The best result was found for Isobutylmethacrylate, for which all the top
hits were methactylate esters, see Table 3 and Figure 4. In this case, the
weighted average for the top five hits was 134 ng, which may be compared with
the actual injected quantity of 150 ng. It is perhaps noteworthy that the
quantity calculated solely on the basis of the first hit for both naphthalene
and isobutylmethacrylate was 13 percent low, indicating the possibility of a
systematic error. The probable cause of this error is our assumption of a
triangular GC peak. The actual shape of the peaks should be Gaussian, so that
the peak concentration is less than that at the peak of a triangle by a factor
of about 18 percent.
TABLE 3. DATA FOR GC/FT-IR SPECTRUM OF ISOBUTYL METHACRYLATE
(150 ng injected)
Match, i
Name
Q(mg)
1 0.09
2 0.16
3 0.22
4 0.29
5 0.29
6 0.32
Methacrylic acid,
Methacrylic acid,
Methacrylic acid,
Methacrylic acid,
Methacrylic acid,
Methacrylie acid,
Isobutyl ester
butyl ester
2-tert-butylamino
2-hydroxyethyl ester
ethyl ester
2, 3-epoxypropyl ester
130
126
140
134
134
137
Other results were less impressive. For example, the value of Q calcu-
lated for a 130 ng injection of nitrobenzene was 82 ng, and even the value
based on the first hit alone (i.e., nitrobenzene) was equal to 80 ng* The
reason for this error is unknown and should be investigated further. It was
not due to an injection error, but could have been caused by an error in the
quantity of sample injected in the gas cell for the measurement of the _
reference spectrum. In all cases tested, however, the calculated value of Q
was within a factor of 2 of the actual injected quantity.
14
-------
SECTION 5
LINKED GC/FT-IR/MS (TASK 2)
INTRODUCTION
The main thrust of this task was the development of software for linking
low-cost GC/FT-IR and GC/MS interfaces. At the start of the project, neither
the FT-IR nor the mass spectrometer installed in Wilkins1 laboratory could be
considered to be low cost units, however. The FT-IR spectrometer was a top of
the line Nicolet 7199 (original list price, ~$110,000) and the mass spectro-
meter was a Nicolet FTMS-1000 3 Tesla Fourier transform mass spectrometer
(FT-MS) with an original list price of about $320,000. During the course of
this project, less expensive equipment was either purchased or acquired on an
extended loan agreement with the manufacturer.
In the first year of this project, a HP Model 5970 Mass Selective
Detector (MSD) was purchased. In the second year, a Nicolet 20-SX FT-IR
spectrometer was obtained via a loan from Nicolet and largely replaced the
original Nicolet 7199 FT-IR spectrometer for this project. In the third year,
an HP IRD was acquired; this instrument incorporates a high performance, low
cost Michelson interferometer. By the end of the project, the MSD and IRD had
been combined to form a low-cost GC/FT-IR/MS system.
The nature of the work performed under Task 2 in each year reflects, to a
large extent, the availability of the various instruments. Thus, for example,
most mass spectra acquired in the first year were measured by FTHS, since the
MSD was only delivered at the end of this year. Each phase of the work
performed under Task 2 will be described below.
GC/FT-IR/FTMS
For GC/FT-IR/MS, two configurations are possible, a series or a parallel
interface. The series Interface was not considered appropriate for the FTMS
measurements for two reasons* Since the GC effluent has to traverse the FT-IR
gas cell first, some "dead volume" is introduced. Therefore the interface
slightly degrades chromatographic resolution before the mass spectral analy-
sis. More importantly, a very low pressure is required for FTMS. Since the
relative sensitivities of FT-IR spectrometers is at least an order of magni-
tude less than that of FTMS, a parallel split of the GC effluent was favored
for GC/FT-IR/FTMS (Figure 5). Retention times measured by FT-IR and FTMS
correspond very closely using this parallel configuration.
The parallel splitter consists of a low dead volume heated tee in which
the effluent path to the mass spectrometer is placed 90* to that of the
15
-------
GC
Splitter
FT-IR
Parallel
MS
Figure 5. GC/FT-IR/MS parallel split.
original GC effluent path. This junction has a 2-cm piece of 10-ym internal
diameter (i.d.) fused silica capillary tubing to act as a ca. 200:1 splitter
and as a restrictor to maintain low (10~® torr) pressure in the mass spectro-
meter vacuum system. The flow path to the FT-IR light-pipe is virtually
unobstructed.
There are several advantages to using a FT-MS in this system. Besides
the high split ratio which allows almost all of the effluent to pass to the
less sensitive FT-IR, several types of mass spectral data can be obtained
during one chromatographic run. Due in part to the rapid data acquisition of
the FTMS, El and CI mass spectra can be acquired alternately. To prevent CI
reagent gas (methane) from interfering in El acquisition, the gas is pulsed
into the FT-MS only during the CI experiment. The respective pulse sequences
are shown in Figure 6.
To demonstrate that the interface is amenable to the study of complex
mixtures, a SOCAL fuel additive was separated via capillary GC and analyzed
with the GC/FT-IR/EI-CI FTMS system. The flame lonization detector (FID) has
a very low dead volume and the chromatogram measured using this detector may
be considered to exhibit the best attainable resolution (Figure 7). The
16
-------
5ms
_n_
5ms
3m>
n_
30ms
-
.QUENCH
•DELAY
ELECTRON SEAM
DELAY
ION EXCITE a
DETECT
5ms
_TL_
2m»
100 mi
r—L
Stnl
_700m»
30ms
_T"L
QUcNCH
TRIGGER VALVE
DELAY
BEAM
RXN DELAY
EXCITE &
DETECT
Figure 6. Pulse sequence used for (above) EI-FTMS and (below) CI-FTMS.
17
-------
SOCAL fuel •d.lltlv*
CC coliuui: 40 • X 0.33 mm id OB-i
0.3 ul Injection
TCBB. proirMi 40 C for 5 Mtnutei
KleccroMtcri 32 H Ifl
He HIM prct«ur«: 2? p«l
•"' -."T | .:i- "Ti' "^^ .^r-l-r-l-Hr-l.::?.-i-: -
i-L---- I I.I>-Lli'J±.ll-:L.LkJ-L
f.-....'•
Figure 7. Fla«e ionization detector trace of SOCAL fuel additive chromatogram.
-------
GC/FT-IR chromatogram reconstruction (Figure 8) closely approached the chroma-
tographic resolution of the FID trace. Each FT-IR data file was acquired with
a time resolution of one second (32 scans coadded). The alternate El-CI data
acquisition for the FTMS results in slightly degraded chromatographic resolu-
tion. Fifty El scans coadded plus a CI acquisition result in a two second
time resolution. Despite the longer time acquisition, many peaks are still
resolved in the chromatogram reconstruction made from the EI-FTMS data
(Figure 9).
Most of the components of the SOCAL mix are aliphatic hydrocarbons. One
example of an identification of an eluite using FT-IR, EI-FTMS and CI-FTMS
will be given. A hydrocarbon that eluted after 12 minutes in the MS
reconstruction and file 564 in the FT-IR reconstruction is identified as a
branched hexane. No molecular ion is seen in the mass spectrum (Figure 10)
which might correspond to the mass spectral and infrared spectral search
results. The number one hits for the FTMS and FT-IR, however, are identical
(2,2,5-trimethylhexane) and give a molecular weight of 128. Although alkanes
are known to give [M-H]+ ions via chemical lonlzation (CI) (hydride Ion
abstraction), no peak at mass 127 is observed. The lack of the [M-H]"1" ion
could be due to the significant branching of the alkane compound. As the
extent of branching increases, the [M-H]* intensity can drop markedly and has
been observed to account for as little as 0.5 percent of total additive
lonization [3]. The peak.at 113 amu corresponds to loss of a methyl group
(15 amu). The infrared search results (Table 4) indicate similar branched
hydrocarbons as well in the list.
Another sample which was analyzed by GC/FT-IR/MS was peppermint oil.
This sample also contains potentially difficult compounds to identify. This
oil, however, has been characterized by a chromatographic supplier and can be
considered as a "known" sample. Peppermint oil contains Isomers of cyclic
hydrocarbons as well as isomers of cyclic ketones. The dynamic range was also
a concern since 95 percent of the sample la represented by only a few com-
ponents* Figures 11 and 12 show the Gram-Schmidt IR reconstruction and
EI-FTMS reconstruction, respectively, for peppermint oil. Certain components
can be seen in one reconstruction and not In the other, depending on quan-
tities present, molar absorptlvitles, etc.
Figure 13 shows the Infrared spectrum of the first peak in Figure 11.
Despite the low signal-to-noise ratio (SNR), the FT-IR results do give the
correct identification of alpha plnene (Table 5). The mass spectral El
results give a molecular ion of 136 amu and a strong CI (M+H) ion at 137. The
successful infrared search results may be partially attributed to the rather
unusual cyclic structure of alpha plnene.
The differentiation of menthone (12.3 minutes) from isomenthone
(12.7 minutes) was hampered by the partial coelutlon of Isomenthone with other
later eluting components. The Infrared spectra of these isomers are nearly
identical. The mass spectra are also not much help in differentiation. The
menthone peak is identified, however, by the infrared spectral search
(Table 6) and supported by mass spectral molecular weight and CI (M+H) data.
The isomenthone peak could not have been identified by IR data alone. With
molecular weight information obtained from the mass spectrum, the isomer was
readily differentiated (since menthone had already been identified).
19
-------
Is)
O
"520 600" 840
OflTfl POINTS
l'»QO
Figure 8. PT-IR Integrated absorbance reconstruction (2995-2870 cm"1) for the SOCAL fuel additive
chroMtogra*.
-------
K)
•10 12'. iT' '"juT 16 '20 r! 2-i 2
C. C. RUN TIU£ IN U1NUTES
Figure 9. Chromatogran reconstructed from electron Impact data for the GC/PT-IR/Ei-CI PT-MS analysis
of the SOCAL fuel additive.
-------
El
71
1 I
50
• 113
60 70 80 90 100 110 120 130
MASS IN A. M. U.
CI
J7
71
Jl
"
50 60 70 80 90 100 110 120 130
MASS IN A. U. U.
Figure 10* Mass spectra froa El (above) and CI (below) data for aliphatic
component of SOCAL fuel additive*
22
-------
190 430 670 910
DATA POINTS
1150 1390
Figure 11. Gram-Schaidt Infrared reconstruction of chromatograa of peppermint
oil sample*
23
-------
G.C. RUN TIME IN MINUTES
Figure 12. FTMS electron Impact data reconstruction of chromatogram of
peppermint oil sample.
24
-------
FINENE
in
CT
a
a
03
(O
a
a
UJ
c
o -
CD a
<£O
a
a
a
o
tOOO 3S3Q 3QSQ
2530 2120
WAVENUMSERS
1650 11SC
710
Figure 13. (Below) Infrared spectrum of first peak in Figure 11. (Above)
EPALIB reference spectrum of a-pinene.
25
-------
TABLE 4. FT-IR SEARCH RESULTS FOR ALIPHATIC COMPONENT OF SOCAL FUEL ADDITIVE
EPALIB
Number
43
2491
381
452
802
2494
699
38
994
993
HQI
51
102
116
121
125
127
129
141
161
164
Identification
Hexane, 2,2,5-trimethyl
Hexane, 2,5-dimethyl
Isooctane (so-called)
Heptane, 2,2,4, 6,6-pentamethyl
C12H26
Hexane, 2,2,4-trimethyl
C9H20
Heptane, 2,2-dimethyl
Isopentyl disulfide
ClOH22s2
Pentane, 2,4-dimethyl
^7^16
3-Heptene, 2, 2, 4, 6,6-pentamethyl
C 12^24
4-Penten-2-ol, 2-methyl
CAS
Numbe r
3522-94-9
592-13-2
540-84-1
13475-82-6
16747-26-5
1071-26-7
2051-04-9
108-08-7
123-48-8
624-97-5
The spectral library search results from the GC/FT-IR/MS experiment are
very Informative in themselves. With both sets of data, a very powerful iden-
tification tool la possible. One may ask, however, whether one set of results
should be weighted against the other. An algorithm to accept or reject search
result combinations was published by Laude et al. 14]. The algorithm utilizes
accurate mass measurement (AMM) FTMS results to establish the most accurate
molecular formula possible. The mass error for such a determination was typi-
cally less than 10 ppm at mass 250 amu. An error of less than 10 ppm has been
found to lead to unambiguous determinations of molecular formula.
Forty-five model compounds of various group-types and polarities were
used at concentrations which might typically be found in a thick film
capillary GC separation. Infrared and mass spectral search results were
incorporated such that if the calculated molecular formula is not represented
by any of the first five MS search results, then the MS search results are not
used. If none of the first five matches of the IR search results coincide
26
-------
TABLE 5. FT-IR LIBRARY SEARCH RESULTS FOR FIRST PEAK OF PEPPERMINT OIL SAMPLE
EPAL1B
Number
HQI
Identification
CAS
Number
2244
1318 Blcyclo/3.1.1/hept-2-ene,
2,6,6'-trimethyl
ClOH16
80-56-8
3303*
2433
3013
2925
3304*
327
121
457
1319
1338
1339
1352
1353
1355
1360
1363
Alpha pinene
ClOH16
Pinene, 2/10/
ClOH16
Bicyclo/3. 1. l/hept-2-ene-2-ethano
Ethane
Beta-plnene
ClOH16
Cyclohexane » 1 , 4-dimethyl
C8H16
Cyclohexane, cls-l,3-dlmethyl
C&HI&
2-Butene, 2,3-dlmethyl
80-56-8
18172-67-3
128-50-7
74-84-0
127-91-3
589-90-2
633-04-0
563-79-1
* Entries with an EPALIB Number greater than 3300 have been added to the data
base at UCR. The similarity of the HQI values given by first two hits
(both a-plnene but measured at different locations) is noteworthy.
with the determined molecular formula, then the compound is considered uniden-
tified. No compounds were incorrectly identified, such that the MS sampling
conditions appear to be adequate. The performance of the various algorithm
combinations possible with the data collected is summarized In Table 7. The
accurate mass IR-MS combination shows no incorrect identification and more
correct identifications than IR-MS without accurate mass data.
Although this computerized interpretation can have great utility in
sample component Identification, the disparity of the two search libraries can
still be a limiting parameter. The infrared data base contains 3400 spectra,
and the mass spectral library contains many more (38,000). Even though both
libraries contain erroneous entries, the redundancies In the mass spectral
library minimize the effect of these errors. Such, unfortunately, is not the
27
-------
TABLE 6. FT-IR LIBRARY SEARCH RESULTS FOR PEAK IDENTIFIED AS MENTHONE
(PEPPERMINT OIL)
EPALIB
Number
3306
3307
732
985
285
622
734
621
HQI
84
155
429
452
460
544
607
653
Identification
Menthone
C10H18°1
Isomenthone
C10H18°1
4-Decanone
ClOH16°I
4-Octanone
CgHigOi
6-Undecanone
3-Nonanone , 2-methyl
3-Undecanone
3-Decanone, 2-methyl
CAS
Number
89-80-5
491-07-6
624-16-8
589-63-9
927-49-1
5445-31-8
14476-37-0
*
984 654 2-Heptanone, 3-butyl *
CllH22°l
2133 706 4-Heptanone, 2-methyl 626-33-5
* CAS Number not given In EPALIB header.
case for the vapor-phase infrared data base. Nevertheless, for environmental
samples the limitations of the GC/FT-IR data base may be less in practice than
might be construed from the above discussion. For example, it is estimated
that only 6,000 of the mass spectra in this data base are those of manufac-
tured chemicals. In addition, none of the 38,000 mass spectra were measured
on the fly but were bled in via the probe and none were FTMS data. This
suggests the possibility that many of these compounds will not pass through
the interface lines without condensation.
28
-------
TABLE 7. COMPARISON OF VARIOUS AMM ALGORITHMS FOR THE 45 "UNKNOWN" COMPOUNDS
Algorithm
IR
IR/AMM
MS
MS/AMM
IR/MS
AMM IR/MS
Number
• Identified
Correctly
32
33
26
30
32
35
Number
Identified
Incorrectly
13
8
19
12
0
0
Number
Eliminated
0
4
0
3
13
10
GC/FT-IR/MS USING A MASS SELECTIVE DETECTOR
(See also Appendix A: Cooper, J. R., I. C. Bowater and C.L. Wilkins.
Anal. Chera., 58:2791, 1986.)
The potential of a low cost combined system was examined both with
respect to the MS and the infrared spectroscopy. Aa noted in the
Introduction, two systems comprising spectrometers of lower cost than the
Nicolet 7199 FTIR and FTMS-1000 became available in Wilkins' laboratory and
were used to assess the potential of GC/FT-IR/MS as a technique which may be
more accessible in general.
The first GC/FT-IR/MS system utilized a Nicolet 7199 FT-IR (later
upgraded to a lower cost 20SX), a Varian 3700 GC and a Hewlett Packard 5970
Mass Selective Detector (HP 5970 MSD). The MSD is a low cost quadrupole mass
spectrometer capable of unit mass resolution and a high mass limit of 800 amu.
The second system incorporated an MSD and a HP 5965 FT-IR detector capable of
up to 4 cm""* resolution spectra.
Initial work on a linked GC/FT-IR/MS system using the MSD and flicolet
7199 FT-IR spectrometer was concerned with development of a parallel interface
to allow simultaneous detection of GC eluents. Due to the sensitivity of the
MSD as compared to the FT-IR spectrometer, a ca. 95:5 splitter was constructed
with fused silica tubing and a Vespel low-dead volume fitting. The split was
normalized to achieve approximately equal SNR's in the respective instruments
(Figure 14).
Software to compare the large amount of data generated from even one
capillary GC run was developed. The Initial scheme called for a comparison of
Chemical Abstracts Service (CAS) registry numbers for those reference library
29
-------
INFRARED LIGHTFIFE
FUSED SILICA
TEE JUNCTION
MSD REPELLER
50 MICRON ID.
RESTRICTOfl
O
O
b
VflHlflN 3700 GC
Figure 14. Representation of parallel chromatographic interface for
GC/PT-IR/MD (95:5 split ratio for infrared light-pipe:MSD).
30
-------
entries returned from the respective FT-IR and MSD library search algorithms.
The MSD reference data base consisted of 38,791 National Institute of
Standards and Technology (NIST) spectra which were reduced to a format of the
ten most significant ions. Significance was determined by the product of the
mass multiplied by abundance in the original data base. The infrared library
consisted of the U.S. EPA vapor phase data base of 3300 compounds plus 100
additional locally measured spectra.
A terminal emulation program resident on the MSD Model 216 computer
enabled search results to be transferred to the Nicolet 1280 computer and to
hard disk. The CAS numbers from the transferred MS results were extracted and
stored for later use by a comparison program. The communication scheme is
illustrated in Figure 15.
Two synthetic mixtures were prepared to test the system. A lA-component
mixture was used to represent a sample where all components are detected by
both instruments. A 23-component mixture represented a less ideal sample in
which several components were not detectable by one or the other instrument.
The components in each mixture are listed in Table IV and V of Appendix A.
Of primary importance to the analysis of samples by GC/PT-IR/MS is the
requirement of correlation between both sets of data. For samples of the
complexity of the 14-component mixture, visual inspection or correlation by
sequential component peak apex spectra may be straightforward. However, for
more realistic samples in which complexity or detectivity are more
challenging, an automated correlation algorithm is preferred. A flow chart of
the algorithm to correlate mass and FT-IR spectra is shown in Figure 1 of
Appendix A. Initial ideas revolved around the premise that comparison of
absolute retention times from the moment of injection would be possible.
Unfortunately, exact synchronization between start of data collection from the
FT-IR spectrometer and the MSD was not possible. Secondly, due to the fact
that each Instrument collects data at separate sampling frequencies, the
eluting component may not be represented by the data to the same extent*
Thirdly, it was recognized that the reproducibility of scan rates for both
systems was varied enough so that absolute retention times were often ambi-
guous *
In the correlation algorithm described in Appendix A, it Is assumed only
that the first component detected by FT-IR spectrometry will also be repre-
sented in the MSD total Ion chromatogram. This requirement is reasonable con-
sidering the superior sensitivity of the MSD over FT-IR. The algorithm then
normalized all FT-IR and MSD peak retention times based on the assumption that
the first, then second, etc., MSD peak corresponds to the first FT-IR peak.
Several more MSD detections are possible for the reason stated above. The
algorithm, however, only assumes that two normalized peak retention times are
coincident if the difference between them is less than the sum of IR and MSD
acquisition/scan times I.00 + 0.55 seconds). An addition this algorithm now
takes into account the possibility of "missing" coincidence by searching for
MSD matches a maximum of 10 peaks away from the FT-IR peak in question.
When the proper correlation Is complete, a search result comparison
program attempts to locate pairs of results representing analysis of the same
component. Figure 4 of Appendix A illustrates the GC/FT-IR/MS software output
31
-------
NICSLET
7122 FTIFI
H.T. 5270
MflSS SEL. DET.
38721 COMPOUND
NBS OflTfi BffSE
flS-232-C I/O
MSC CCMPUTEH
StOO COMPOUND
INp-RflRED
DflTfl BASE
HOST COMPUTER
NICOLET 1280
Figure 15. Data communication for GC/MSD and GC/FT-IR. MSD data transfer
(post-run) to Nicolet 1280 computer via RS-232.
32
-------
for component 13 of the 14-component mixture. In many cases for these analy-
ses, only one library entry is coincident in both lists. Here, p-chlorophenol
is returned as the coincident match between both lists. The modified (0-100)
match values are included as an aid in assessing the performance of the
search. Normalization of the respective search routine match values is
necessary due to the different scales of measurement involved. These conver-
sions are described in detail in Appendix A. Confirmation of identification
can come in two ways. Table V of Appendix A lists the components confirmed by
comparison of the FT-IR and MSD search results and by comparison of molecular
weight formed from the MSD data with the FT-IR results. In this work, the
molecular ion was taken as the highest mass ion observed above the noise
threshold. A tolerance of ±2 amu was used to match potential FT-IR search
results. No false positives were encountered for this sample. Even for
electron impact spectra, the molecular ion is often represented as for many of
the compounds shown.
COMBINED LOW-COST INFRARED AND MASS SPECTROMETRY FOR GC/FT-IR/MS ANALYSIS
(See also Appendix B: J.R. Cooper and C.L. Wllkins, Anal. Chem., 1987,
submitted.)
The acquisition of the Nicolet 20SX FT-IR spectrometer enabled better
sensitivity to be achieved for the GC/FT-IR/MSD low cost system than when the
Nicolet 7199 FT-IR spectrometer was utilized. The overall spectral noise
level of spectra measured with the 20SX was found to be about a factor of 4
less than spectra measured with the older spectrometer. An additional update
to the system utilizes a new data base of 5,010 IR Vapor Phase spectra from
Nicolet Instrument Corp./Aldrich Chemical Co. 117 other spectra were sub-
sequently added to make 5,127 16 coT* resolution spectra. The mass spectral
library to be used for IR-HS development was changed to the 31,579 compound
U.S. EPA/National Institutes of Health (NIH) mass spectral data base. A
reduced format was used, with the 64 most intense mass spectral ions saved to
each entry. The MS data would be used for further work.
A comparison of the new infrared and mass spectral libraries determined
that 2,487 compounds shared the same CAS registry numbers. It was of interest
to determine whether a combined 1R-MS data base could provide any potential
advantages over the dual search strategy. The features of each type of
strategy are listed in Figure 16. Although the combined library is smaller
than the old IR data base, upcoming acquisition of a larger mass spectral data
base from HIST, U.S. EPA and the United Kingdom Mass Spectral Data Centre
should improve that situation. A detailed description of the combined library
format is given in Appendix B. The organization of the spectral file is such
that rapid acquisition is possible with the Nicolet assembly-level
subroutines. Each combined library entry occupies 256 20-bit computer words.
The original MSD library and search algorithm were abandoned in favor of
the above MS library and an absolute difference search algorithm. The abso-
lute difference algorithm has been used successfully for Infrared spectral
searching and was retained for the combined library. The absolute difference
result is normalized in the manner illustrated below. The final value for the
33
-------
DUAL SEARCH:
A) SOFTWARE ALREADY DEVELOPED
B) ACCESS FULL- SEPARATE SPECTRAL LIBRARIES
C) CORRELATE RESULTS BASED ON CAS NUMBER COMPARISONS
D) MS LIBRARY MAY NOT REPRESENT GC-ABLE TSCA COMPOUNDS
E) FULL POTENTIAL OF COMBINED DATA NOT REALISED
COMBINED SEARCH:
A) SAME COMPUTER SYSTEM: SOFTWARE DEVELOPMENT STRAIGHTFORWARD
B) INFORMATION IN SAME FORMAT
C) SPECIFIC FREQUENCY AND MASS AXIS INFORMATION ACCESSIBLE
D) LIBRARY TAILORED TO GC-ABLE ANALYTES
E) DISCRETE FT-IR AND MS SEARCHES STILL POSSIBLE
Figure 16. Features of search strategies.
IR-MS similarity index was the average of IR and MS indices since both data
sets were considered to contain equivalent information. The similarity index
(S.I.) was computed as follows:
S.I. -I-
N
E |
1"0
N N
I Ui + £ Lt
i-o i»o
Range - 1.0000-0.0000
where U^ - 1th data point of spectrum of unknown;
LI • ith data point of library spectrum;
N - number of data points used (typically 384).
The correlation of Information between FT-IR and MSD data is achieved as
described in the previous section. Instead of using separate files of search
results as was done previously, however, both infrared and mass spectra con-
sidered coincident are utilized by the combined search algorithm. Built into
the system is the provision of performing an IR-only, MS-only, or IR-MS search
based on the Information available. Figure 17 illustrates the logic of the
program based on correlated retention times.
To Increase the speed of the combined search, pref liter information was
compiled for each library member Into a separate file. The primary presearch
requires that the unknown spectrum (IR and MS) contain an absorption band or
Ion that corresponds to the most intense frequency (IR) or ion (MS) in the
library spectrum. However, this data point need not also be the most intense
data point in the unknown. Secondary presearch information in the form of the
second most intense IR peak apex data point or mass spectral ion was also
Included. The minimum allowable intensity of those data points can be spe-
cified by the user in the form of the percentage of maximum (normalized)
-------
FTIR-MSD R.T. COMPARISONS
FTIR
MSD
FTIR-MSD
Figure 17. Logic of combined IR-MS search algorithm based upon availability
of spectral data,
intensity. This tolerance level can be varied from 0 percent (all library
entries used) to 100 percent (very restrictive) but was varied between 5 and
75 percent for this work. Figure 1 of Appendix B shows the FT-IR AMAX
(jfide jjifra., in the following section) and MSD total ion chromatograms
generated for a 23-component mixture similar to the one used previously. In
the FT-IR reconstruction, the tenth component appears not to be detected.
Figures 2 and 3 of Appendix B illustrate the varying sensitivity for both IR
and MSD In this sample* Table 1 of Appendix B lists the 23 components in
order of elution and the quantities injected on-column. For 50 ng of iso-
butylmethacrylate injected, there Is abundant IR signal but a relatively low
MS signal. In contrast, 125 ng of 2-chloronaphthalene produces an intense MSD
signal but relatively poor infrared response. Table 2 of Appendix B shows the
effectiveness of the IR-MS prefllter to screen library entries from further
consideration by the search algorithm* A relatively low threshold tolerance
is preferred so that the compound of interest or similar compounds are not
excluded. Empirically, a threshold value of 20 percent appears a safe minimum
requirement. In the worst case, with the four-level prefliter for IR-MS as
much as 94 percent of the library (144/2487) is eliminated at 20 percent
threshold. In Table 3 of Appendix B, the effectiveness of each data type as a
prefiltering feature is summarized for several compounds. In almost all of
the cases, more entries are retained for the IR-only mode as for the MS-only
search mode* This is not surprising considering that infrared spectra exhibit
various bandwldths* The last column Indicates that a maximum total search
time of less than 5 seconds was achieved.
The performance of the search software for this sample is shown in
Table 4 of Appendix B with respect to type of search. Although the infrared
results performed well, the incorporation of mass spectral information may
35
-------
be required to assure detection of all components (recall the lack, of FT-IR
detection for component 10).
Table 8 shows the IR-MS search results for tsobutylmethacrylate. la this
example, the class of compound is represented in 6 of the top 10 returns. For
TABLE 6. RESULT OF IR/MS SEARCH FOR ISOBUTYLMETHACRYLATE
EPALIB
Number
1664
1603
1853
967
700
CAS Molecular
S.I. Compound Name Number Weight
0.8496 Methacrylic Acid, Isobutyl Ester 97-86-9
C8H1402
0.8222 Methacrylic Acid, Butyl Ester 97-88-1
C8H1402
0.7427 Ethyl Methacrylate, 99Z 97-63-2
C6H10°2
0.6999 Allyl Methacrylate, 98Z 96-05-9
C7H1002
0.6711 Methyl Methacrylate, 99X 80-62-6
142
142
114
126
100
997 0.6271 Ethylene Glycol Dimethacrylate 97-90-5 198
ClOH14°4
2029 0.5856 Formic Acid, Butyl Eater 592-84-7 102
C5Hi002
1149 0.5784 (+/-)-3-Methylcyclohexanone, 978Z 591-24-2 112
790 0.5684 (R)-(+)-3-Methylcyclohexanone, 981 13368-65-5 112
C7H1201
2126 0.5668 Azelaic Acid, Dibutyl Ester 2917-73-9 300
the last component, however, only 3 reference spectra fulfill the prefilter
criteria (at a 20 percent IR-MS tolerance level) (see Table 9). In this case,
had 2-chloronaphthalene not been represented In the library, the nearest match
would have provided a poor similarity index. Generally, a similarity index
much less than 0.6 has been found to be unreliable for identification pur-
poses. Seven of the 23 components of the mixture had search returns con-
taining only.l match. In each case, the single match was the correct match.
these compounds are listed in Table 10.
36
-------
TABLE 9. RESULT OF IR/MS SEARCH FOR 2-CHLORONAPHTHALENE
EPALIB
Number
1980
944
2325
S.I.
0.7763
0.4498
0.1243
Compound Name
2-Chloronaphthalene
3,5-Dichlorophenol, 99Z
CgH^C^O}
Sulfur Hexafluoride
CAS
Number
91-58-7
591-35-5
2551-62-4
Molecular
Weight
163
163
146
TABLE 10. COMPONENTS RETURNING 1 MATCH ON IR/MS SEARCH
Compound S.I.
4-Chlorotoluene 0.7303
4-Chlorobenzene 0.8002
Indene 0.8580
1,3,5-Trichlorobenzene 0.8465
1,2,4-Trichlorobenzene 0.7890
a,'a,a-Trichlorotoluene 0.8379
Bipheayl 0.8371
A second sample was analyzed using the combined library and correlation
software. To test that correlation algorithm fully, a very complex sample was
sought that would represent a possible limit in complexity generated by a gas
chromatographic separation as well as varying detectivities of components. A
sample of unleaded gasoline was separated on a 60 meter x 0.32 mm l.d. DB-5
(5 percent phenyl methyl polyslloxane) column, with a 1 urn film of the
stationary phase. 0.2 yL of pure gasoline was injected using on-column injec-
tion. Figures 1ft and 19 represent the reconstructed chromatograms generated
by the AMAX algorithm and total ion current, respectively. An expansion of
the ordinate of both traces reveals a very complex mixture of extreme chemical
dynamic range. The expansions are shown In Figures 20 and 21.
A correlation of this sample was performed as follows. The individual
FT-IR and MSD software peak finding routines were set so that approximately
37
-------
00
20
UNLEADED GASOLINE
0.2 UL ON-COLUMN
60 1780 2000
FTIR FILE
Figure 18. GC/PT-IR reconstructed chromatogram generated via Absorbance Maximum algorithm.
-------
UNLEADED GASOLINE
O.e UL ON-COLUMN
MSD TOTAL ION SIGNAL
lA80 If'iO 1800
MSD SCANS
Z160
2880
Figure 19. GC/MSD reconstructed chroaatogram generated from total ion current.
-------
• 111 nil •
ii ii
UNLEADED GASOLINE
O.e UL ON-COLUMN
FTIR FILES
zo So4eo
680 $00 lIZQ 1&*0 1560 1^80 2&00
FTIR FILES
Figure 20. Ordlnate expansion of Figure 18 demonstrating coaplexity of sample.
-------
II III I
UNLEADED GASOLINE
Q.Z UL ON-COLUMN
USD TOTAL ION CHROM
deo i£eo 1600
MSD SCANS
Figure 21. Ordlnate expansion of Figure 19.
-------
100 components would be detected in each reconstructed chromatogram.
Specifically, 100 FT-IR detections and 101 MSD detections were produced. No
prior knowledge about actual coincidences was available. The correlation
algorithm generated 84 pairs of detections with retention times within the sum
of IR and MSD data acquisition rates. A summary of the correlation is shown
in Table 11 along with the correlation results for two dilutions of the
23-component mixture discussed earlier. Although the first MSD component
detected aligns with the first FT-IR detection for both these dilutions, the
TABLE 11. CORRELATION RESULTS FOR TWO OF THE SAMPLES STUDIED (the
23-component mixture was injected at two volumes)
23-Component Mixture
0.8 uL
MSD $ FT-IR
Peak i Matched
1 23*
2 2
3 1
4 2
5 4
6 0
7 3
23-Component Mixture
0.4 \H
MSD 0 FT-IR
Peak # Matched
1 22*
2 1
3 I
4 5
5 2
6 2
7 2
Unleaded
MSD
Peak 9
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
Gasoline
# FT-IR
Matched
7
84*
22
19
6
29
28
11
7
27
24
15
20
31
11
4
42
-------
gasoline mixture has many noncoincident detections. Here, the first FT-IR
peak aligns with the second MSD peak detected. Due to the complexity of the
sample, other alignments were also possible. However, those overlaps were
incorrect. Again, in this list, up to the 16th MSD peak was assumed to
overlap with the first FT-IR peak in the correlation.
The automated IR-MS search software was then run to generate IR-only,
MS-only and IR-MS search results where appropriate. The results generated
alkane, substituted aromatic and substituted naphthalenic Identifications.
One IR-MS result was the identification of indane, a partially reduced poly-
cyclic hydrocarbon. Figure 22 shows the latter half of each reconstructed
chroaatogram with the stronger bands off-scale. The indicated peaks were
identified as indane. For this relatively minor component of the sample, an
infrared spectrum with good SNR was achieved (Figure 23). The match to the
reference compound is nearly identical. The full mass spectrum at scan number
1923 confirms a compound of molecular weight 118 (Figure 24). The IR-MS
search results generated with these spectra are shown in Table 12. Due to the
uniqueness of these spectra, the prefliters, which were run at 20 percent
tolerance, could find only 3 reference spectra fulfilling the criteria.
TABLE 12. RESULT OF IR/MS SEARCH FOR COMPONENT IDENTIFIED AS 1NDAM
EPALIB
Number
2475
1299
2399
S.I.
0.8667
0.7023
0.5723
Compound Name
Indane
C9H10
B-Methylstyrene
C9H10
Cyclopropylbenzene
CAS
Number
496-11-7
637-50-3
873-49-4
Molecular
Weight
118
118
118
C9H10
The second low cost GC/FT-IR/MSD unit in Wilkins1 laboratory incorporates
a HP 5890 GC, 5965 IRD and 5970 MSD. The instruments are linked In series
with the chroaatographic effluent passing through the 1 mm 1.4. x 12 CD gold-
coated light-pipe of the IRD and then to the MSD transfer line* One feature
of the IRD IB Its very low infrared spectral noise level compared to the
Nicolet 60SX, 7199, and 20SX FT-IR spectrometers. In practice, the peak-to-
peak noise level in the 3800-3400 cnT1 region is typically 0.2-0.3 milli-
absorbance units (nAU), affording the possibility of very low detection limits
and a normalization of detectivity between infrared and mass apectrometry.
HP's specification for the IRD is that of about 1 mAU per 5 nanograms of
dodecane. Figure 25 shows the reconstructed chromatograms for a mixture of
nonane, isobutylmethacrylate and tridecane at the 125 ng/yL concentration
(1 uL introduced with splitleas Injection). Figure 26 shows the MSD and IRD
spectra of 10 ng of tridecane in which reasonable SNR is encountered for both
spectra. Often, papers in the literature will quote GC/FT-1R sensitivity
using isobutylmethacrylate as the analyte. Figure 27 indicates that the
43
-------
FTIR SIGNAL
1110 l&OO l£901$801*70 1560 1650 mO 1830 1920
MSD SIGNAL
VI
J
1710 1810 1970 2100 2^30 2^60 2*90 2620 2750
GC/FT-IR and GC/MSD reconstructed chromatograas from unleaded gasoline. (Latter half of
chroaatograas shown). Indicated peaks Identified as Indane.
-------
INOftN
oo
CO
Ol
FTIfi FILE L340
3620 3£«to 2650 zieo 2 loo
WAVENUMBEflS
1340 §eo Seo
Figure 23. (Above) Infrared library spec trim of Indane.
file Indicated in Figure 22.
(Below) FT-IR spectrum retrieved from
-------
Scan 1923 (21.073 mln)
1500-
1000-
500-
,ll. 1,1,
i-i'» . • r »'.'.'llt
30 40 SB 60 70 B0 90 100 110 120
Figure 24. USD Spectrun retrieved fro* file indicated in Figure 22 and identified as indane (molecular
weight 118 ami).
-------
MRSS SPEC RESPONSE
8 . 0E+5-
6.0E+5-
4.0E+5-
2 . 0E+5:
6.0
B.0 10.0
Tlme Cm In.)
12.0
TNFRRRED RESPONSE
300:1
200:
100-
6.0
8.0 10.0
TIME (min)
12.0
Figure 25. GC/MSD and GC/IRO reconstructed chromatograas fro* a mixture of nonane, isobutyl/
•ethacrylate and tridecane at 125 ng^ 1 concentration.
-------
10000
5000-3
43 N
/ d?
1 B'5
.. |. II Illl . |--u|'l I.I.
1 12
. ... .1 . ll 1 1 u ll
127 '41 '58
1 .1 .1 .. .1.1. II. . . 1 .
184
i
40 60 80 100 120
Mass/"Charge
140
160
180
200
00
-------
ij 8000:1
tf 6000-
•o
C 4000-
a! 2000:
4,1 69
i
III.. 1. . (Ill i .,
87
/
100 V49
ii II. ill. ... | . nl
190
/
i i
40 60 80 100 120
MassXChargo
140
160
180 200
ISOBUTYL METHnCRYLRTE 12 ng
-------
alkane may be the better choice for establishing Impressive detection limits
both with respect to IR response and MSD response. However, the spectra of
the n-alkanes are so similar that unequivocal identification may not be
possible even though the spectrum is measured with a SNR greater than 10.
An application to environmental contaminants was pursued with the
integrated system. Fourteen phenolic compounds were successfully separated in
18 minutes with a 30 meter x 0.32 mm l.d. DB-5 capillary column. Figure 28
Illustrates the detection achieved for each instrument and the separation of
all components. In each chromatogram, one component is barely detected.
Table 13 lists the compounds analyzed in order of elution. The concentration
range was 80-300 nanograms per component injected via splltless sampling. A
TABLE 13. PHENOLS IN ORDER OF ELUTION
1 Phenol
2 o-Nitro Phenol
3 2,4-Dlmethyl Phenol
4 2,4-Dichloro Phenol
5 p-Chloro Phenol
6 4-Chloro 3-Methyl Phenol
7 2,4,6-Trichloro Phenol
8 2,4-Dinitro Phenol
9 p-Nitro Phenol
10 4,6-Oinitro 2-Methyl Phenol
11 Pentachloro Phenol
three-level temperature program enabled a rapid separation to be achieved.
The oven was held at 100*C for 3 min, then programmed at 4°C/min to 120°C, and
ramped at 15°C/min to 220°C. The injection temperature was 250°C.
Figures 29 and 30 demonstrate excellent SNR for the spectra of the third and
eighth components, 2,4-dimethylphenol and 2,4-dlnitrophenol, respectively.
These infrared spectra are representative of the range of molar absorptivities
for a set of phenolic compounds.
50
-------
MR55 SPEC RE5PON5L
B
10
12
14
16
16
INFRRRED RESPONSE
300:
6
10
TIME
12
(min)
14
16
18
Figure 28. Reconstructed chromatogram* for a mixture of U phenolic compounds using
GC/IRD/MSD system.
-------
j 8000-
0 6000-j
c 4000:
•J •
# 2000:
107
39
/
77
.Ji ..in . ki,ll
22
139
180
/
244
50
100 150
MassXCharge
200
2,4-DIMETHYL PHENOL 125 ng
1 .5:
1.0:
3000 2000
WRVENUMBER Ccm-1)
1000
Figure 29. Third component of phenol mixture. MSD and IRD spectra from 125 ng injected.
-------
Ul
Scan 979 (14.205 mtn) of
o
o 6000-
c
•0 4000-
C
3
A 2000-
(E
DRTR: PHENOL. D
SUBTRRCTED ,
53 63 91
\ / /
i 1
id 1 i n l?0
L Jll Al 1 1 J . ( 1
1. LlliltAJ 1 .11 11, II 1 1 .L I
154
/
1 .
ll .
34
jj r* ^
221 c°^
1 i i
50 100 150 200 250
Mass/Charge
2,4-DINITRO PHENOL 80 ng
4.0-i
i 2.0-i
E :
'•"1 ^
0DI- **.*. __« ~. . jfj* Xu%/*^*V^VlAA(W^NX\^Ww^
» *^ ] ™ V^*F WAff PT '"YTT
3000
NRVENUMBER <
I
,
k L
IV UA A
'Wvuu, ,A/
. t n ^J u\A.vvv
2000 1000
cm— 1 )
Figure 30. Eighth component of phenol mixture. MSD and IRD spectra from 80 ng Injected.
-------
MAXIMUM ABSORBANCE ALGORITHM FOR RECONSTRUCTION OF GAS CHROMATOGRAMS FROM
GAS CHROMATOGRAPHY/INFRARED SPECTROMETRY DATA
(See also Appendix C: Bowater, I. C., R. S. Brown, J, R. Cooper, and
C. L. Wilkins. Anal. Chem., 58:2195 (1986).)
In order Co retrieve infrared spectra of gas chromatographtc eluents
detected In real time with GC/FT-IR, the infrared signals are usually manipu-
lated to form a reconstructed chromatographic trace. This chromatogram of
infrared responses can then be used to locate the file spectra of interest.
The two algorithms most widely used to reconstruct chromatograms in
GC/FT-IR are based on the Gram-Schmidt and integrated absorbance methods.
Although the Gram-Schmidt method is often preferred for universal reconstruc-
tions due to computational speed and better SNR, the integrated absorbance
method can be used to create selective frequency chromatograms which may be
characteristic of a particular functional group.
The algorithm developed for GC/FT-IR here has the capability of both
Gram-Schmidt and integrated absorbance methods. The method, a maximum absor-
bance reconstruction, is based upon the magnitude of frequency-domain data
values within a selected frequency window. This window, however, can include
the entire mid-infrared spectral, region being utilized. This approach has
been applied in liquid chromatography/ultraviolet spectrometry [7] and shares
the same fundamental concept as that used in the algorithm described earlier
by Krishnan et^ al. [8].
Two sample mixtures were prepared and represented various functional
group classes such as aliphatic and aromatic hydrocarbons, esters, and
ketones. Figure 1 of Appendix C shows reconstructed chromatograms generated
using a universal algorithm covering much of the mid-infrared and selective
wavelength regions including C-H stretching, C-0 and C-0 stretching frequen-
cies. These examples illustrate the selectivity indicative of integrated
absorbance methods. As a universal detection method, the SNR can be manipu-
lated depending upon the resolution of the infrared spectra being used. A
thirteen-component mixture was used to generate signal-to-nolse ratios in the
universal mode at various infrared spectral resolutions. The original 2048
point interferograms were truncated and Fourier transformed to yield 8, 16,
and 32 cm"* resolution spectra. Table II of Appendix C lists the SNR values
for each component detected. Overall, the chromatogram generated by using
16 cm'1 spectra retained the best overall SNR. A comparison was performed
between the absorbance maximum and integrated absorbance methods using a
spectral frequency range of 771-677 cm'^ and 16 cm"* resolution spectra with
the 4-component sample.
The absorbance maximum (AMAX) method produces a chromatogram with better
SNR for all components. Since characteristic infrared bands are known to vary
on the wavenumber axis, the AMAX method is considered to be more reproducible
for narrow regions of interest that the Integrated absorbance method.
A comparison with the Gram-Schmidt method was also performed. Table V of
Appendix C summarizes the comparison of Gram-Schmidt using single and double-
sided Interferogram vectors and the AMAX method. Depending upon the com-
54
-------
ponent, the AMAX method was better for some and and worse for others. With
the Gram-Schmidt method, however, the SNR depends upon the spectral features
of the component and the location of the vectors. For example, hydrocarbons
gave the best SNR with single-sided vectors whereas esters and ketones gave
better SNR with double-sided vectors.
In general, the AMAX algorithm is useful for both universal and selective
detection of GC/FT-IR eluents. The method is substantially more sensitive
than the integrated absorbance method and comparable or complementary to the
Gram-Schmidt method for universal detection.
55
-------
SECTION 6
IMPROVED SENSITIVITY FOR THE GC/FT-IR INTERFACE
INTRODUCTION
By far the most commonly used GC/FT-IR interface, and the one used for
all measurements described so far in this Report, involves passing the GC
effluent through a heated light-pipe. Several factors are known to affect the
sensitivity of measurements made in this manner. These include the specific
components Incorporated in the spectrometer (e.g. the material from which the
source Is fabricated and the specific detectivity of the detector), the way in
which the light-pipe is made, the optical design, and the signal-processing
electronics.
Efforts were made in both Griffiths' and Wilklns' laboratories to develop
methods of increasing the sensitivity of GC/FT-IR measurements made using
light-pipes. Both groups developed techniques for minimizing the loss in
interferogram signal which has been observed to occur when the temperature of
the light-pipe is raised. The technique developed in Wilkins1 laboratory
involved the addition of a heat shield to existing commercial GC/FT-IR inter-
faces. This work will be described in the next section, A fundamental
investigation of the parameters affecting GC/FT-IR sensitivity was made in
Griffiths' laboratory. This work resulted in the construction of an optimized
GC/FT-IR interface which was subsequently installed at Dr. Donald Gurka's
laboratory at EMSL-LV* A description of this system is given later in this
section. This system was evaluated for its application Co environmental
analysis by Dr. Gurka, and found to be at least an order of magnitude more
sensitive than a commercial GC/FT-IR interface with the same spectrometer [9],
Identifiable spectra of low nanogram sample quantities can be measured
either with this system interfaced to a Dlgllab FTS-60 FT-IR spectrometer or
with the HP IRD, but It is rarely possible to measure a spectrum that can be
useful for identification purposes If less than 1 ng of the component of
interest is injected In the GC. For subnanogram quantities, sample trapping
techniques must be employed if an identifiable spectrum is to be measured.
The most successful method for elulte trapping in GC/FT-IR Involves
adding a small quantity of argon to the helium carrier gas and passing the
mixture of these two gases, together with any component which is eluting from
the column, onto a rotating reflective metal cylinder that is held at a tem-
perature between 10 and 15K. At such a temperature, helium is still above its
boiling point, and does not condense. The argon is condensed onto the
metallic substrate, trapping any eluite(s) which are present in the column
effluent. The train of trapped eluites in the argon matrix is then rotated
56
-------
until 1C passes through the focused beam of an FT-IR spectrometer, where the
reflection-absorption spectrum Is measured*
A GC/FT-IR Interface based on the principle of matrix Isolation (MI) has
been Introduced commercially by Mattson Instruments [10-12], and has been suc-
cessfully applied to the analysis of dloxlns [12-14]. The biggest problem
with this Instrument is Its mechanical complexity (including the need for a
very high vacuum) and hence its relatively high cost. Other problems include
the low sample throughput, the need for a new database, and an unproven quan-
titative accuracy. Since this Cooperative Agreement was designed to investi-
gate lower cost alternatives to existing technology for GC/FT-IR, work was
initiated in Griffiths' laboratory to study the feasibility of trapping GC
eluites at temperatures below ambient but well above 15K.
Two methods of cooling were investigated. The first involved the use of
thermoelectric (Peltier effect) cooling. This technique was very simple
mechanically, and allowed the feasibility of the technique to be verified.
However, the lowest temperature accessible by Peltier-effect coolers was
-40°C, so that relatively volatile molecules were not condensed*
The optimum sensitivity of this device is found when the eluites are
trapped as very small spots, and techniques were developed to trap each eluite
as a spot of less than 0.25 mm diameter, necessitating the use of an. FT-IR
microscope for the spectral measurement. In view of the physical size of
FT-IR microscopes, the Interface could not be evacuated; instead it was simply
purged with dry air. The presence of an atmosphere surrounding this interface
led to one other disadvantage, in that the tip of the transfer line was con-
vectively cooled by the cold substrate* Thus eluites of relatively low vola-
tility condensed in the transfer line before reaching the substrate.
In summary, therefore, the first Interface to be constructed demonstrated
the principle of sample trapping, but could only be used for components that
were volatile enough that they did not condense in the transfer line but
non-volatile enough to be trapped on the substrate.
The most recent work has Involved the construction of an analogous inter-
face operating in a vacuum with the substrate at liquid nitrogen temperature.
A prototype unit was constructed and has been used to demonstrate that the
disadvantages of the Peltier-cooled Interface can be overcome. The early
results obtained with this system are described later In this section.
LIGHT-PIPE TEMPERATURE AND OTHER FACTORS AFFECTING SIGNAL IN GC/FT-IR
SPECTROMETRY
(See also Appendix D: Brown, R. S., J. R. Cooper and C. L. Wilkins.
Anal. Chenu, 57:2275 (1985).)
It has been observed by many workers in the field of GC/FT-IR that heating
of the light-pipe to 200-300°C results in a loss of infrared interferogram
signal and a subsequent decrease in SNR. A variety of explanations have been
presented in the literature to account for this phenomenon. Work in Griffiths1
laboratory indicated that the effect can be explained by saturation of the
57
-------
mercury cadmium tellurlde (MCT) detector or its preamplifier by unmodulated
radiation emitted from the hot light-pipe. The nonlinear response would then
reduce the apparent signal at the detector and decrease the SNR of GC/FT-IR
spectra. Approximately 70 percent of infrared signal was lost upon heating
the light-pipe from 25°C to 300'C.
The first results in this study were obtained in Wllklns' laboratory
using a Nlcolet 60SX FT-IR spectrometer, the optical geometry of which is
shown in Figure 1 of Appendix D. With these optics, the 1 mm diameter image
of the light-pipe end should be focused to an image of slightly less than
0.3 mm diameter on the 1 mm MCT detector chip in the absence of any aberra-
tions. The remainder of the area of the detector is filled with the image of
the (hot) end of the light-pipe. Table 1 of Appendix D summarized the loss of
signal with increasing temperature.
A water-cooled shield was constructed in an attempt to limit the amount
of unmodulated heat emanating from the light-pipe without reducing infrared
signal measured at ambient temperature. The heat shield, shown in Figure 2 of
Appendix 0> was designed so that the Infrared beam could be picked up by a
small length of gold-coated 1.47 mm i.d. glass tubing. This device enabled a
considerable amount of signal to be retained upon heating of the light-pipe to
300°C. Table 11 of Appendix D shows that only 23 percent of original signal
was lost. Measurements performed with the heat shield prior to inclusion of
the short light-pipe addition resulted in a fortuitous increase in infrared
signal even at ambient temperature. Presumably, the original heat shield
design incorporating a conical opening redirected a significant component of
the scattered beam back onto the collection optics {9.3" focal length). A
series of polished metallic cones was constructed to determine the optimum
angle of Incidence. The measurements with these cones are presented in
Figure 6 of Appendix D. Table III of Appendix D reflects the increase in
throughput with a cone of 13° Included angle. A similar increase was observed
when these cones were added to the end of a Nicolet 7199 FT-IR light-pipe
assembly. A 38 percent increase resulted from using a 19° cone and collection
optics similar to those of the Nicolet 60SX GC/FT-IR accessory.
A comparison was also performed with two other FT-IR systems of varying
optical geometries Installed in Wilkins* laboratory. The optics of a Nicolet
7199 GC-IR interface were replaced with optics which could refocua the
infrared beam onto an iris capable of being reduced to a diameter of 1 mm. A
Nicolet 20SX FT-IR spectrometer was used as the source of modulated infrared
radiation. By choice of optics which would condense the image of a 1.33 mm
i.d. light-pipe to approximately 1 mm, it was believed that minimal unmodu-
lated radiation would be collected. The effective beam condensing chosen
would result in a 0.95-min image on the 1-rnm2 square detector chip in the
absence of aberrations.
This type of geometry was also represented by the interface used in a
HP S96SA IRD. For this instrument, the 1-mm diameter light-pipe image is
refocused by two identical parabolic mirrors onto a 1-mm2 chip. The data from
heating the light-pipe are shown in Table 14. Compared to these data, the
additional signal retained by the shield described above may lend itself to
maximizing the SNR ratio of GC/FT-IR measurements.
58
-------
TABLE 14. PERCENTAGE OF SIGNAL RETAINED ON HEATING LIGHT-PIPE FROM AMBIENT
TO 2 50*0
FT-IR 7, Ambient signal
60 SX (original) 37
60 SX (modified) 81
H.P. IRO (original) 64
20 SX (modified) 66
OPTIMIZED GC/FT-IR SYSTEM BASED ON THE USE OF LIGHT-PIPES
(See also Appendix E: Henry, D. E., A. Giorgetti, A. M. Haefner, P. R.
Griffiths and D. F. Gurka. Anal. Chem., 59; 2356-2361 (1987); Appendix F:
Gurka, D. F., R. Titus, P. R. Griffiths, D. Henry and A. Giorgetti. Anal.
Chem., 59:2362-2369 (1987).)
A detailed investigation designed to determine the optimum optical con-
figuration for GC/FT-IR interfaces incorporating light-pipes was carried out
in Griffiths' laboratory. The goal of this part of the oroject was to design
an optical system to collect most of the modulated radiation transmitted
through a light-pipe while discriminating against unmodulated radiation
emitted from the end of the hot light-pipe, and hence to maximize the sen-
sitivity at any light-pipe temperature.
Optics were constructed in which the f/ number of the collection optics
could be varied by repositioning a KBr lens. It was found that the use of f/1
collection optics allowed the maximum percentage of the modulated radiation to
be measured when the light-pipe is at ambient temperature (see Table IV of
Appendix E), but it was also found that the signal loss on heating the light-
pipe was maximal when f/1 optics were employed.
Several optical configurations were studied to test their capability of
discriminating against radiation emitted by the light-pipe while maintaining a
high collection efficiency for the beam transmitted by the light-pipe. One
system, shown as Figure 5 of Appendix E, involves simply filling the detector
with the image of the bore of the light-pipe in the manner described at the
end of the previous section. This configuration gave reasonable discrimina-
tion against unmodulated radiation but was very susceptible to the effects of
optical misalignment* A second design, shown as Figure 6 of Appendix E, gave
superior sensitivity, equal loss in signal on heating the light-pipe, and was
less susceptible to optical misalignment. For a light-pipe of 300'C, the
interferogram signal was still 80 percent of the signal at ambient temperature
for both of these optical designs.
59
-------
A GC/FT-IR Interface using the optical design shown In Figure 6 of
Appendix E was constructed In Griffiths' laboratory and installed In
Dr. Gurka's laboratory at EMSL-LV, In accordance with the "Deliverables" of
the Cooperative Agreement. This system was tested for Its performance for
environmental analysis, and the results are described In detail In Appendix F,
The MIQ of 52 typical environmental contaminants was found to be an average of
13 times better than the corresponding values measured on the same spectro-
meter with a commercial Interface purchased In 1982 (see Table 1 of
Appendix F).
Subsequently even better sensitivity was obtained at EMSL-LV using an
HP IRQ (D. F. Gurka, U.S. EPA, EMSL-LV, personal communication, 1987) which
has an optical configuration similar to that of Figure 5 In Appendix E. This
improvement is not believed to be caused by a significant improved optical
design, as the optical throughputs of the IRD and the system used for the
measurements described in Appendices E and F are comparable and the light-pipe
dimensions in both systems are approximately equal (10 cm long, 1 mm i.d.).
The noise equivalent power of the narrow-range 1-mm MCT detector installed In
the IRD Is also approximately equal to that of the intermediate-range 0.5 mm
MCT detector in the interface delivered to EMSL-LV.
Over the last twelve years, there has been an Increase in the performance
of commercial FT-IR spectrometers (by about an order of magnitude) even though
there have been no significant changes made to the optical components of these
Instruments. These improvements may largely be assigned to improved signal-
processing electronics, and we believe that the electronics of the HP IRD are
as good (and probably better) than those of any other commercial FT-IR
spectrometer. It Is probable that If the signal from the detector installed
on the GC/FT-IR interface delivered to EMSL-LV under this Cooperative
Agreement were processed by the electronics of the HP IRD, the sensitivity
would be at least equivalent to that of the IRD.
DECOMPOSITION OF CERTAIN COMPOUNDS DURING GC/FT-IR ANALYSIS
(See also Appendix G: Henry, D. E., S. L. Pentoney, R. W. Kondrat and
P. R. Griffiths. Chromatographia, 23:547-552 (1987).)
During the course of this work, a report was published by Warthen and
McGovern 115] in which decomposition of certain compounds (y- and 6-haloesters)
In GC/FT-IR light-pipes was reported. It was suggested that the decomposition
might have been catalyzed by the gold light-pipe at fairly low temperature
«250°C). We had not been aware of the decomposition of any compounds on a
gold surface at temperatures below 350*C, and it became apparent that an
understanding of the decomposition of compounds being studied by GC/FT-IR was
Important if this technique was to be generally applicable to a wide range of
analytes.
The separations of Warthen and McGovern were carried out using a commer-
cial GC/FT-IR interface (Nicolet Analytical Instruments). These separations
were repeated using a light-pipe constructed in Griffiths' laboratory in which
the fused silica transfer line was passed directly into the light-pipe through
60
-------
a groove cut Into the end of the glass tubing from which the light-pipe was
made, and no decomposition of y- and 6-haloesters was observed. It was there-
fore apparent that the decomposition of the analytes was not occurring on the
gold surface.
The Nicolet GC/FT-IR interface contains a short (-2 mm long) brass fix-
ture allowing the convenient union of the light-pipe, infrared window and
transfer line (see Figure 1 of Appendix G). Further experiments in which GC
effluents were passed through short lengths of heated metal tubes confirmed
that decomposition of haloesters can occur rapidly on these surfaces by a
mechanism shown in Figure 6 of Appendix G. Other compounds, including chlori-
nated pesticides, decompose immediately on contact with hot stainless steel.
It was because of this decomposition that all-glass injection ports are incor-
porated on most gas chroma tographs.
In light of these results, it is apparent that GC eluites should not come
into contact with any metal except gold in a GC/FT-IR interface. In par-
ticular, copper, brass or steel fittings should never be used. To eliminate
the need for fixtures of the type used by several of the commercial vendors of
GC/FT-IR interfaces, the transfer line must be led directly through the light-
pipe in a manner similar to the construction of the light-pipe used in this
study.
SAMPLE-TRAPPING GC/FT-IR INTERFACES
(See also Appendix H: Shafer, K. H. , P. R. Griffiths and R. Fuoco. J.
High Res. Chromatogr. Chromatogr. Commun., 9: 124-126 (1986) and Appendix I:
R. Fuoco, K. H. Shafer and P. R. Griffiths. Anal. Chem., 58: 3245-3254
(1986).)
As noted in the Introduction to this section, the MTQ achieved using
light-pipe based GC/FT-IR interfaces incorporating light-pipes can rarely be
reduced below 1 ng per component. In this section we report on the feasi-
bility of trapping GC eluites on thermoelectrlcally cooled substrates and
measuring the spectrum of the deposited spot. The increased sensitivity of
interfaces in which the sample is trapped on a cooled substrate Is derived
from (a) the decreased area in which the sample is retained In comparison to
the cross-sectional area of a light-pipe and (b) the increased time allowed
for measurement of the spectrum. The effect of all the factors affecting the
SWR of GC/FT-IR spectra was discussed by Griffiths [16] in terms of the
general expression for the SNR of an FT-IR spectrum given by Griffiths and
de Haseth [17]:
U(v.T).Av.9.D*.tI/Z.C
where U(v,T) is the spectral energy density at wavenumber, v cm'1, of a source
of temperature, T (in units of W/sr cm2 cm"1); Av is the spectral resolution
(cm"*); 0 is the optical throughput (cm2 sr), i.e. the product of the solid
angle of the .beam at a given focus, Of (sr), and the area of the beam at this
61
-------
focus, Af(cm2); D* is the specific detectivity of the detector (cm Hz^Z W"1);
t is the measurement time (seconds); c is the efficiency; AQ is the area of
the detector (cm^).
The root-mean-square noise level on the baseline of a spectrum plotted
linear in absorbance units is therefore:
Noise- ^ (7)
By Beer's Law, the absorbance, A, of any chromatographically separated
component present in the infrared beam at wavenumber, v, is given by:
A(v) - a(v) b C (8)
where a(v) is the absorptivity at v, b is the pathlength, and C is the
concentration of the analyte. Since:
C - Q/V (9)
where Q is the sample quantity and V is the volume which it occupies, and V -
Ag b (where AS is the cross sectional area of the sample), it follows that:
A(v).SLstvl no)
As
Since A(v) is the analytical signal at v, and 9 - OfAf, the SNK Is:
(In 10).Q.a( v).U(v,T).Av.nf.Af.D*.t1^.C (11)
SNK m — — - - - - - •
A0Vz AS
The effect of each parameter in this equation is often not obvious. For
example, to obtain the highest possible SNR in any measurement of a small
sample quantity (Q), it appears obvious at first glance that only the
strongest bands should be observed in order that their absorptivity a(v) is as
large as possible. However some of the weaker bands In the spectrum can give
good clues as to the detailed chemical structure of the analyte, and it is
usually necessary to detect bands with small a(v) along with the stronger
bands in the spectrum. Similarly it would appear from Equation (11) that the
most intense source, i.e., the source with the largest U(v,T), and the most
sensitive detector, i.e., with the largest possible D*, should be used. In
many cases this is true, but if the sample area la large, the SNR can be
limited by the dynamic range of the analog-to-digltal converter (ADC) rather
than by the intensity of the source or the sensitivity of the detector; in
this case, increasing U(v,T) or D* will not lead to an improved SNR,
The effect of digitization noise can always be decreased by reducing the
signal at the detector. For GC/FT-IR measurements, the most productive method
of reducing the signal is by reducing the area of the beam at the sample
62
-------
focus. In any well-designed measurement of small samples, the if number of
the optics should be as large as possible, Indicating that the beat results
will be found using a microscope, for which the f/ number Is as large as
possible. The solid angle of the beam at the sample and at the detector
should also be both equal, necessitating very fast collection optics at the
detector. Since the optical throughput is about equal at the sample and
detector, Af ~ As - AD. Thus:
SNR
m (In 10).Q.a(v).U(v,lQ.&v.aE.D*.t.c (12)
It is important to note that the diameter of the detector must be matched to
that of the sample, in order to minimize the denominator of this equation and
to maximize the SNR,
Resolution is a parameter that is rarely given adequate consideration
when the relative sensitivities of GC/FT-IR interfaces are compared. From
Equation (12) it can be Inferred that the lower the resolution, the higher the
SNR of the measurement. This conclusion is true provided that the full width
at half height (FWHH) of the bands in the spectrum of each analyte is greater
than Av. However it should be realized that during the measurement of a
spectrum the true spectrum Is convolved with the Instrument line shape (ILS)
function of the spectrometer. If the FWHH of the bands Is greater than iv,
the effect of this convolution on the measured spectrum Is negligible. On the
other hand, if the FWHH of the bands is less than Av, the effect of the ILS
function Is to reduce the apparent value of the peak absorbance of each band
[17].
In GC/FT-IR measurements made using light-pipes, it is customary to make
the measurement at a nominal resolution of 8 cm"1. Since the FWHH of bands
in vapor-phase infrared spectra measured under these conditions is on the
order of 8 cm'1, the apparent peak absorbance is only slightly less than the
true absorbance, and the resolution at which the measurement is made is nearly
optimal in terms of both sensitivity and spectral information content* la
matrix isolation GC/FT-IR measurements, however, the width of most bands is
less than 1 cm"*. In this case the peak absorbance of a band measured at,
say, 4 cm"* resolutoa would be about one-half of the peak absorbance of the
same band measured at a resolution of 2 cm'1. On the other hand, the noise
level of a 2 cm"1 resolution measurement would be twice that of a 4 en."1
measurement of the spectrum of the same sample made in the same measurement
time, so that the SNR of the two measurements would be about equal. Thus the
principal reason for measuring GC/MI-FT-IR spectra at high resolution is the
increased specificity, and not the increased sensitivity, of the measurement*
As noted in the Introduction, the increased measurement time, t, attained
by sample trapping is probably the single most important factor in causing
increased sensitivity of any chromatography/FT-IR interface Involving
elimination of the mobile phase prior to the spectral measurement of the
trapped eluite.
The final parameter in the equation for SNR Is the optical efficiency, ?.
Although it is intuitively obvious that the optical efficiency of any
spectroscopic measurement should be as great as possible, one cause of
63
-------
decreased signal at the detector caused by reflection losses In the light-pipe
has the effect of reducing the SNR of the Interferogram so that It becomes
limited by detector noise rather than by digitization noise when an MCT detec-
tor is used. In fact, the length-to-diameter ratio of a well designed light-
pipe should be large enough that the SNR of the Interferogram at the
centerburst is about one bit leas than the dynamic range of the ADC. For
reflection losses occurring in a light-pipe, £ is approximately independent of
wavenumber, so that the SNR is reduced by a uniform factor across the entire
spectrum. This is contrast to FT-1R measurements of components separated by
liquid chromatography, where £ is governed by absorption by the solvent which
is strongly dependent on wavenumber.
The first interface based on these principles was constructed in
Griffiths' laboratory and is described in Appendices R and I. The output of
the GC was passed through a heated 250-ym fused silica transfer line into a
short length of 50-um i.d. fused silica. The end of this tube was held about
100 um from the surface of a thermoelectrically cooled ZnSe window which could
be moved in a straight line, allowing the eluites to be trapped as a train of
"spots". These spots travelled into the beam of an Analect AQM-515 microscope
and spectra were measured using an Analect fX-6200 FT-IR spectrometer.
Identifiable spectra of submicrogram quantities of nonvolatile polar analytes
such as acenaphthenequinone (AQ) were measured, see Figure 2C of Appendix H,
and excellent spectra of a broad range of environmental contaminants at the
low nanogram level were obtained, see Figure 7 of Appendix I. Chromatographic
resolution was retained, even for peaks separated by only 6s, see Figure 5 of
Appendix I*
Besides having high sensitivity (despite the fact that spectra were
measured on a five-year old low-coat FT-IR spectrometer)» this system had the
advantage of providing spectra that could be compared qualitatively to stan-
dard reference spectra of materials prepared as KBr disks. It did, however,
have several disadvantages. The minimum temperature obtained using a Peltier
cooler was only -40°C, so that relatively volatile molecules were not trapped.
The system was purged with dry air, and the tip of the 50-ym tube was convec-
tlvely cooled because of its proximity to the cold ZnSe window. Thus molecu-
les of low volatility, such as AQ, condensed in the transfer line unless the
temperature of the window was increased. Thus molecules with only a rather
small range in volatility could be studied with this interface.
Finally, it was observed that when the quantity of a given elulte
increased, the width of the track also increased. This behavior was assigned
to the effect of the low thermal conductivity of the solid organic eluites on
the surface of the window leading to a steep temperature gradient across the
sample. The eluites did not condense on the relatively hot sample immediately
adjacent to the tip of the transfer line, but were retained in the vapor phase
until they encountered a cooler surface.
All the above problems could be eliminated if the window temperature
could be reduced by over 100°C to allow condensation of volatile components
and to keep the temperature of the upper surface of the deposited eluites low
enough that the material elutlng subsequently from the column is condensed
-------
B
Figure 31. Liquid nitrogen cooled stage and transfer line positioner used in
this study: (A) 50 x 25 x 2 mm ZnSe window; (B) brass cooling
plate; (C) coiled thin wall stainless steel tubing; (D) phenolic
insulating plate; (E) vertical aluminum plate; (F) adjustment
screws; (G) tension springs; (H) aluminum mounting plate;
(I) translation stage; (J) fiber optic (transfer line) positioner.
immediately. For low temperature operation, the entire interface must be held
in a vacuum, which should eliminate convective cooling of the transfer line.
To address these problems, a prototype system in which the window is cooled
with liquid nitrogen (LN2) was constructed.
The design of this interface is shown schematically in Figure 31. The
ZnSe window (A) is held against a brass plate (B) through which LN2 is passed.
LN2 is passed through thin-wall 1.6-mm o.d. stainless steel tubing (C) which
was coiled (50 mm diameter) to permit the window mount to travel at least
50 mm without undue torque. The cooled metal plate was itself attached
rigidly to a phenolic insulating plate (D) and the entire unit was held
tightly to a vertical aluminum plate (E) by three ball-end adjustment screws
(F) and three springs (G). Using these screws, the plane of the window could
be adjusted to coincide with the infrared beam focus. The temperature of the
window was estimated at 90 K by taping a thermocouple to its surface. The
vertical aluminum plate (E) was bolted to a horizontal plate (H) that could be
65
-------
translated smoothly over a linear bearing (I) (New England Applied Technologies,
Lawrence, MA). Linear motion over a distance of 90 mm was achieved, allowing
up to about 40 eluites to be condensed. Position control was achieved using a
stepping motor with 200 steps per revolution and a 2:1 jjear reduction onto a
28 threads per inch screw. The scan speed could be varied between 0 and
8 mm/min.
A 60-m long, 0.21-yta i.d. DB-5 (5 percent phenyl methyl polyslloxane)
column (J&W Scientific, Folsom, CA) was mounted in a Varian Model 3700 GC and
used for all separations. The end of the column was attached to about 2 m of
uncoated deactivated 250-pm i.d. aluminum-clad fused silica tubing (Quadrex,
New Haven, CT), which was passed through an equal length of 1.6 mm o.d. thin-
wall stainless steel tubing. The two tubes, a thermocouple, and a 1-mm i.d.
Aerorod heater (ARi, Adason, 1L) were inserted into a 4.8-mm o.d. copper tube,
that served as an oven for even heating of the transfer line. The 250-ym
fused silica tubing was attached a short length (about 4 cm) of 50-von i.d.
fused silica tubing that was heated by conduction through a short length of
stainless steel tubing soldered to a butt connector. This tubing allowed the
50-ym fused silica tubing to be heated to within 2 mm of its end and addi-
tionally held the tip rigid. The temperature of the transfer line was
controlled using a variable transformer. The copper tube holding the transfer
line, heater and thermocouple was mounted through phenolic Insulating ferrules
into a fiber optic positioner (J) (Newport Corp., Fountain Valley, CA) to give
fine control in three dimensions. Using this device the track of deposited
components could be adjusted to be at the same height as the infrared beam.
The fiber optic positioner was mounted on a bracket attached to the base plate
of the optical bench» allowing coarse positional control.
For optimum sensitivity, the effective beam diameter at both the sample
focus and the detector must be about 100 ym and the solid angle of the beam at
each of these foci should be as large as possible 116]. As noted above, the
ideal optical arrangement to achieve these goals is that of an FT-IR
microscope. However, the size of the vacuum chamber required to enclose a
typical microscope would be inconveniently large* An optical arrangement
similar to that of a rudimentary microscope, but in a horizontal plane with no
facility for viewing the sample, was constructed* The collimated beam
emerging from the interferometer of a evacuable Digilab FTS-60 FT-IR spectro-
meter (Digilab Division of Blo-Rad Labs., Cambridge, MA) was focused by a 90°
off-axis paraboloidal mirror with an effective focal length of about 50 mm.
The cooled ZnSe window was mounted so that the surface on which the GC eluites
were condensed could be moved into this focus. The beam was then passed into
a Cassegrain objective and refocused at a plane where an aperture used to
control the effective beam diameter' was located. The beam was then focused on
a downward-looking lOO-jin MCT detector using Cassegrain optics Identical to
the ones mounted after the sample focus.
All the optics, together with the window unit and the transfer line posi-
tioner, were mounted on a free-standing 12-tua thick aluminum plate with five
feet attached to its underside. The chamber of the FTS-60 spectrometer was
divided into two compartments, one for the source unit and Interferometer.
The plate with the GC/FT-IR Interface mounted on it was placed in the second
compartment. . Because this plate was not rigidly attached to the base of the
66
-------
vacuum chamber, no distortion or optical misalignment was observed when the
spectrometer was evacuated.
The following three questions were addressed in this preliminary study;
a. Is the vapor pressure of very volatile solids so high at 90 K that
the rate of evaporation is too great to permit them to be trapped for
long enough to allow their spectra to be measured?
b. Will operation in a vacuum reduce convective cooling of the tip of
the transfer line so that molecules of low volatility will not con-
dense before reaching the window?
c. Will operation in a vacuum allow the diameter of the deposited spot
to be reduced below the size found in our previous study (see
Appendix I) and potentially allow the sensitivity to be increased?.
The first of these questions was addressed by injecting pure dich-
loromethane (M.Pt., -95.1°C). No decrease in the amount of CH2C12 was
observed visually 2 hours after deposition. The second question was addressed
by injecting a solution of acenaphthenequinone in dichloromethane. In our
previous study with a dry air atmosphere » AQ had been observed to condense in
the transfer line when the window temperature was reduced below -10°C. In the
present study, AQ did not condense in the transfer line even when the window
temperature was below 100 K. It should be noted that this result was not
surprising in view of the fact that cooling has not been observed in the
GC/FT-IR Interface based on the principle of matrix Isolation (S. Bourne,
Digilab Division of Blo-Rad, Cambridge, MA, personal communication, 1987).
Thus It appears that operation In vacuo with a LN2~cooled window allows all
gas chromatographable molecules except the permanent gases, H2, 02, ^2» He » Ar
and Xe, none of which have infrared-active absorption bands, CO, CH$ and
(probably)
The width of the deposited spot was measured using AQ as the probe mole-
cule because of the low volatility of AQ and because it was used In our pre-
vious study so that results could be compared directly. Since the deposited
eluites could not be photographed jji sjltu due to the rather primitive optical
system employed, the window had to be removed from the interface for charac-
terization of the deposited eluites. Thus the analyte had to be sufficiently
nonvolatile to withstand slow heating to ambient temperature In vacuo. 0. 1 pL
of a 0.012 percent solution of AQ in CH2C12 was injected (12 ng of AQ). The
window was moved at a rate of 45 iWs; this rate allowed the track to be
easily photographed and characterized spectrophotometrlcally* Because of the
low volatility of AQ, the window could be allowed to return to ambient tem-
perature under vacuum simply by stopping the flow of LN2. The window could
then be removed from the interface for characterization of the deposited spots
by conventional photography and by FT-IR microscopy.
Figure 3i shows the result of transcribing a photograph of the spot using
a 'ThunderScan1 optical scanner (Thunderware , Orinda, CA). The scale was
added subsequently using the graphics software of an Apple Macintosh computer.
Visible interference fringes could be seen In the original color photograph
67
-------
100
100
200
300
400
500
DISTANCE IN MICRONS
Figure 32. Optically scanned photograph of 12 ng acenaphthenequlnone deposi-
tion.
0.09
-60
-40
-20
0
Figure 33. Profile of 12 ng acenaphthenequinone deposition by absorbance
measurements measured with a 16 urn aperture.
and can just be perceived in Figure 32, indicating the thickness of the sample
to be about I urn. The thickness profile was also characterized spectrophoto-
metrically using an IR Plan microscope (Spectra-Tech Inc., Stamford, CT) with
a 16 urn aperture installed in a Perkin-Elmer Model 1800 FT-IR spectrometer.
68
-------
The result of scanning across the spot is shown in Figure 33. Both these
traces show that the entire track is contained In a width of less than 100 urn,
which is slightly less than one half of the width measured in a similar manner
in our previous study. The' full width at half height of the track, is about
50 pm, which is equal to the internal diameter of the transfer line, and
therefore cannot be expected to be reduced without installing a narrower tube.
These results indicate that the principal drawbacks of the interface that
we described earlier have been overcome. A second generation interface in
which the window is mounted in a computer-controlled x-y translation stage has
been designed in our laboratory. Calculations indicate that the MIQ for polar
molecules measured on these systems should be less than 100 pg when an inter-
mediate range MCT detector (peak D* of 2 x 10^ cm Hz * W~*, minimum wavenumber
of 600 cm~*) with a 100 urn element ia installed.
69
-------
SECTION 7
REVIEW PAPERS
Three review papers were published In the work supported by this
Cooperative Agreement. These are attached to this report as Appendices J-L.
Appendix J: Griffiths, P. R., S. L. Pentoney, A. Giorgetti, and K. H.
Shafer. The Hyphenation of Chromatography and Infrared Spectrometry. Anal.
Chen., 58:1349A-1362A, 1986.
Appendix K: Griffiths, P. R. and D. E. Henry. Coupled Gas
Chromatography and Fourier Transform Infrared Spectrometry. Progr. Analyt.
Spectrosc., 9:455-482, 1986.
Appendix L: Wilkins, C. L. Linked Gas Chromatography-Infrared-Mass
Spectrometry. Anal. Chem., 59:571A-581A, 1987.
70
-------
SECTION 8
RELATED MANUSCRIPTS
Several other papers were published or have been submitted for publica-
tion which refer in part to work supported under this Cooperative Agreement.
These papers have not been Included with this report as Appendices in the
interest of conserving space. The full references are listed below:
Laude, D. A., Jr., J. R. Cooper and C. L. Wilkins. 1986. Analysis of Gas
Chromatography/Mass Spectrometry Library Search Results with' Edited and
Quantitative ^-*C Nuclear Magnetic Resonance Spectra. Anal* Chen).
58:1213-1217.
Qian, C. and C. L. Wilkins. 1987. A Data Access Subroutine for Spectrometric
Data Bases. J. Chem. Infor. and Comp. Sci. In press.
71
-------
REFERENCES
1. Nyquist, R. A. The Interpretation of Vapor-Phase Infrared Spectra.
Volume 1. Sadtler Research, Philadelphia, Philadelphia, 1984.
2. Weast, R. C. CRC Handbook of Chemistry and Physics, 56th Edition.
CRC Press, Inc., Cleveland, Ohio, 1975.
3. Harrison, A. G. Chemical lonization Mass Spectrometry* CRC Press, Inc.,
Cleveland, Ohio, 1983.
4. Laude, D. A., C. L. Johlman, J. R. Cooper and C. L. Wilkins. Accurate
Mass Measurement in the Absence of Calibrant for Capillary Column Gas
Chromatography/Fourier Transform- Mass Spectromctry. Anal. Chem.,
57:1044, 1985.
5. de Haseth, J. A. and T. L. Isenhour. Reconstruction of Gas Chroraatograms
from Interferometric Gas Chromatography/Infrared Spectrometry Data.
Anal. Chem., 49:1977, 1977.
6, Coffey, P. J., D. R. Mattson and J. C. Wright. A Programmable GC/FT-IR
System. Araer. Lab., 10(5):126, 1978.
7. Drouen, A. C. J. H., H. A. H. Billiet and L. De Galan. Multiwavelength
Absorbance Detection for Solute Recognition in Liquid Chromatography.
Anal. Chem., 57:962, 1985.
8. Krishnan, K., R. Curbelo, P. China and R. C. Noonan. Design and
Applications of a High Sensitivity Gas Chromatographic Fourier Transform
Infrared System. J. Chromatogr. Sci., 17:413, 1979.
9. Gurka, D. F., R. Titus, P. R. Griffiths, D. E. Henry and A. Giorgetti.
Evaluation of an Improved Single-Beam Gas Chromatography/Fourier
Transform Infrared Interface for Environmental Analysis. Anal. Chem.,
59: 2362, 1987.
10. Bourne, S., G. T. Reedy, P. Coffey and D. Mattson. Amer. Lab., 16(6): 90,
1984.
11. Schneider, J. F., G. R. Reedy and D. G. Ettinger. GC/Matrix Isolation/
FTIR Applications: Analysis of PCBs. J. Chromatogr. Sci., 23:49, 1985.
12. Reedy, G. T., D. G. Ettinger, J. F. Schneider and S. Bourne. High
Resolution Gas Chromatography/Matrix Isolation Infrared Spectrometry.
Anal. Chem., 57:1602, 1985.
72
-------
13. Grainger, J. and L. T. Gelbaum. Tetrachlorodibenzo-p-Dioxin Isomer
Differentiation by Capillary Gas Chromatography Fourier Transform
Infrared Spectroscopy. Appl. Spectrosc., 41:809, 1987.
14. Griffiths, P. R. and D. E. Henry. Coupled Gas Chromatography and Fourier
Transform Infrared Spectrometry. Progr. Analyt. Spectrosc., 9:455, 1986.
15. Warthen, J. D. and T. P. McGovern. GC/FTIR Analyses of Trimedlure
Isomers and Related Esters. J. Chromatogr. Set., 24:451, 1986.
16. Griffiths, P. R. A Unified Approach to the Chroraatography/FT-IR
Interface. Mikrochim. Acta (Wien), in press, 1988.
17. Griffiths, P. R. and J. A. de Haseth. Fourier Transform Infrared
Spectrometry. Wiley-Intersciecice, New York, 1986.
73
------- |