United States
Environmental Protection
Agency
Environmental Research
Laboratory
Athens GA 30605
EPA-600,4-79-064
October 1979
Research and Development
f/EPA
Automated
Measurements of
Infrared Spectra of
Chromatographically
Separated Fractions
-------
RESEARCH REPORTING SERIES
Research reports of the Office of Research and Development, U.S. Environmental
Protection Agency, have been grouped into nine series. These nine broad cate-
gories were established to facilitate further development and application of en-
vironmental technology. Elimination of traditional grouping was consciously
planned to foster technology transfer and a maximum interface in related fields.
The nine series are:
1. Environmental Health Effects Research
2. Environmental Protection Technology
3. Ecological Research
4. Environmental Monitoring
5. Socioeconomic Environmental Studies
6. Scientific and Technical Assessment Reports (STAR)
7. Interagency Energy-Environment Research and Development
8. "Special" Reports
9. Miscellaneous Reports
This report has been assigned to the ENVIRONMENTAL MONITORING series.
This series describes research conducted to develop new or improved methods
and instrumentation for the identification and quantification of environmental
pollutants at the lowest conceivably significant concentrations. It also includes
studies to determine the ambient concentrations of pollutants in the environment
and/or the variance of pollutants as a function of time or meteorological factors.
This document is available to the public through the National Technical Informa-
tion Service, Springfield, Virginia 22161
-------
EPA-600A-79-061*
October 1979
AUTOMATED MEASUREMENTS OF INFRARED
SPECTRA OF CHROMATOGRAPHICALLY SEPARATED
FRACTIONS
Peter R. Griffiths
Ohio University
Athens, Ohio 4 5701
Grant No. R80^333-01
Project Officer
Leo V. Azarraga
Analytical Chemistry Branch
Environmental Research Laboratory
Athens, Georgia 30605
ENVIRONMENTAL RESEARCH LABORATORY
OFFICE OF RESEARCH AND DEVELOPMENT
U.S. ENVIRONMENTAL PROTECTION AGENCY
ATHENS, GEORGIA 30605
-------
DISCLAIMER
This report has been reviewed by Environmental Research Laboratory,
U. S. Environmental Protection Agency, Athens, Georgia, and approved for
publication. Approval does not signify that the contents necessarily reflect
the views and policies of the U.S. Environmental Protection Agency, nor does
mention of trade names or commercial products constitute endorsement or
recommendation for use.
ii
-------
FOREWORD
Nearly every phase of environmental protection depends on a capability
to identify and measure specific pollutants in the environment. As part
of this Laboratory's research on the occurrence, movement, transformation,
impact and control of environmental contaminants, the Analytical Chemistry
Branch develops and assesses new techniques for identifying and measuring
chemical constituents of water and soil.
Gas and liquid chromatographic techniques are effective tools for
separating sample components to identify and measure trace organic pollutants
in water. The mass spectrometer, coupled to a gas chromatograh, has proved
highly successful as a detector for identifying and measuring the separated
components that are volatile enough to pass through a gas chromatograph.
Significant problems, however, are sometimes associated with identifying
certain compounds using mass spectrometry. This report examines the develop-
ment of an alternative method, Fourier transform infrared spectroscopy, that
shows promise for resolving some of these problems and for improving our
ability to identify less volatile water pollutants (using high pressure
liquid chromatography) at the milligram per liter or microgram per liter
level.
David W. Duttweiler
Director
Environmental Research Laboratory
Athens, Georgia
iii
-------
ABSTRACT
This research project was initiated with the overall objective of
improving the sensitivity of the interface of a Fourier transform infrared
(FT-IR) spectrometer with both gas and high performance liquid chromatographs
for the on-line measurement of the spectra of eluting species. These devices
have been applied to the identification of trace organic water pollutants.
To optimize the sensitivity of infrared measurements of gas chromato-
graphic effluents, the optimal dimensions for the light-pipe gas cells were
first calculated. The transmittance of light-pipes with these optimized
dimensions is so high that the signal-to-noise ratio of single-beam inter
ferograms measured using a mercury cadmium telluride photodetector is
limited by digitization noise. To get around the problem, the application
of a dual-beam FT-IR spectroscopy was tested and the sensitivity of measure-
ments was four times greater than the corresponding single-beam measurement.
It is probable that the sensitivity of dual-beam measurements is limited by
the poor reproducibility of the interferograms and the nonlinearity of
the detector response.
This system was applied to the identification of trace drganics in water
after extraction using neutral polystyrene resins, elution by diethyl ether
and separation by gas-liquid chromatography using 1/8-in. o.d. packed columns.
Detection limits of about 250 ng were achieved.
The system was then redesigned for increased sensitivity by reducing the
volume of peaks eluting from the chromatograph through the use of support-
coated open-tubular columns and shorter light-pipes. The system was not
quite completed by the end of the project period, but detection limits below
10 ng could be forecast.
The application of dual-beam FT-IR spectroscopy to the on-line identifi-
cation of peaks eluting from a high performance liquid chromatograph was
investigated. Typical detection limits in excess of 10 |ig were found, which
are too great for general analytical work. The feasibility of semi-continuous
measurements of liquid chromatographic effluents after eliminating the solvent
was studied. The lowest detection limits were found with a system based on
diffuse reflectance measurements of deposited solutes on KCl powder. Spectra
of submicrogram quantities were measured in preliminary Work.
The sensitivity of techniques for the _in situ identification of species
on specially prepared thin-layer chromatographic plates was improved by the
application of programmed multiple development. Detection limits of 100 ng
IV
-------
could be achieved in less than 5 seconds data acquisition time, and limits of
10 ng could be achieved after extended signal-averaging. The application of
diffuse reflectance spectroscopy to the identification of species on aluminum-
foil backed thin-layer plates was studied, and although submicrogram detection
limits were again found, compensation for strong absorption bands of the adsorbent
was found to be extremely difficult.
This report was submitted in fulfillment of Grant Ho. R8o4j33-01 by Ohio
University under the sponsorship of the U.S. Environmental Protection Agency.
This report covers the period April 1, 1976 to October Jl, 1978, and work was
completed as of November 30> 1978.
v
-------
CONTENTS
Foreword
Abstract ........ ............................................... iv
Figures [[[ .vili
List of Abbreviations and Symbols .............................. x
1. Introduction ........................................... 1
2. Conclusions ............................................ 3
3 . Recommendations ........................................ 5
k-. Theoretical Considerations in GC-FT-IR ................. 6
5. Dual-Beam Fourier Transform Infrared Spectrometry ...... 16
6. Gas Chromatography and FT-IR Spectrometry .............. 28
T. Liquid Chromatography and FT-IR Spectrometry ........... 37
8. Thin-Layer Chromatography and FT-IR Spectrometry ....... 57
-------
FIGURES
Number
1. Sample quantity and concentration in light-pipe as a
function of Vi/Vcell ........................................ 7
2. Peak spreading for Rg = 1.0 and Vi = Vceii .................... 8
3. Peak spreading for Rg = 0.8 and Vi = Vcell .................... 9
4. LTp vs L for 4-ml and 1-ml light-pipes ............... . ........ Li
5. Relative energies of Nernst glower and nichrome wire sources . . 13
6. Optical layout for dual -beam FT-IR spectrometry ............... ^
7. Optical layout for GC-IR using dual-beam FT-IR optics ......... 19
8. Calculation of a transmittance spectrum from dual-beam FT-IR
spectra [[[
9. Comparison of sensitivity of single-beam and dual-beam FT-IR
spectrometry .............................................. ,
10. Reproducibility of single-beam and dual -beam interferograms ... 2^
11. Reproducibility of dual-beam FT-IR spectra .................... 25
12. Linearity of the response of the MCT and TGS detectors at high
photon flux ................................................
13. On-line GC-IR spectra of strong infrared absorbers from a 2 ppb
aqueous solution ........ . ..................................
14. On-line GC-IR spectra of chlorinated pesticides from a 1 ppm
aqueous solution ............................ . .............. 32
15. Relative sensitivities of FT-IR spectrometers manufactured in
1973 and 1977 .............................................. 34
16. Ratio of the experimental to the optimal S/N as a function of
solvent transmittance ......................................
17- On-line LC-IR spectrum of 100 pg of anisole measured by dual-
beam FT-IR spectrometry ........ „ ........ . ..... . ............
18. On-line LC-IR spectrum of 150 jog of TDE measured by dual -beam
FT-IR spectrometry ......................................... 41
19A. Distribution of deposited solute in a light-pipe for different
light-pipe temperatures ....................................
19B. Distribution of deposited solute in a light-pipe for different
-------
Number Page
20. Spectra of solute deposited in light-pipe under different
measurement conditions ..................................... ^
21. Optical diagram for hemiellipsoidal diffuse reflectometer ..... 48
22. l6-cup carousel for LC-IR measurements by DRIFT spectrometry .. 50
25. Concentration and solution deposition stage for LC-IR measure-
ments by DRIFT spectrometry ................................ ^
2^. On-line LC-IR spectrum of 1 jog of Butter Yellow dye measured by
DRIFT spectrometry ...................... . .................. 53
25. On-line LC-IR spectrum of 1 pg of Indophenol Blue dye measured
by DRIFT spectrometry ...................................... ^
2.6. On-line LC-IR spectrum of 1 )jg of Sudan Red G dye measured by
DRIFT spectrometry ......................................... 55
27. TLC-IR spectra of Aldrin developed by PMD and by conventional
TLC [[[
28. TLC-IR spectra of Butter Yellow under different experimental
conditions .................................................
29- TLC-IR spectra of methylene blue as a function of sample size
and measurement time .......................................
-------
ABBREVIATIONS
LIST OF ABBREVIATIONS AHD SYMBOLS
a. c.
ADC
cfm
d.c.
DRIFT
FID
FT- IE
GO
GC-IR
GC-FT-IR
GC-VB
EFLC
i.d.
LC-IR
LC-FT-IR
MCT
o.d.
FMD
ppm
SCOT
S/N
TCD
TGS
TLC
TLC-IR
WCOT
M.-ATR
alternating current
analog-to-digital converter
cubic feet per minute
direct current
diffuse reflectance infrared Fourier transform
flame ionization detector
Fourier transform infrared
gas chromatography
interface between a gas chromatograph and an
infrared spectrometer
GC-IR performed using an FT-IR spectrometer
without trapping the sample
interface between a gas chromatograph and a
mass spectrometer
high performance liquid chromatography
internal diameter
interface between a liquid chromatograph and an
infrared spectrometer
LC-IR performed with an FT-IR spectrometer
mercury cadmium telluride
outside diameter
programmed multiple development
parts per billion ((ag/l)
parts per million (mg/l)
support-coated open tubular
signal-to-noise ratio
thermal conductivity detector
triglycine sulfate
th in-layer chromat ography
jLn situ infrared spectrometry of species on TLC plates
wall-coated open tubular
micro attenuated total reflectance
SYMBOLS
a
Ccell
C
max
D*
half-angle of beam
maximum value of the concentration of sample in the cell
maximum concentration of sample eluting from the
chromatograph
specific detectivity
-------
d,, beam diameter at a focus
d internal diameter of light-pipe
5 optical retardation in centimeters
!'(&) a.c. and d.c. components of the interferogram
I (&) a.c. component of interferogram
I (v) . source intensity as a function of wavenumber
L length of light-pipe or pathlength of cell
Q ,, maximum quantity of sample in a cell
Q. .. , total quantity of sample in a chromatographic peak
v wavenumber, cm'1
R reflectance of light-pipe coating
R~ relative retention distance in ILC
R reflectance of beamsplitter at wavenumber, v
RCT chromatographic resolution
s
R diffuse reflectance at infinite depth
t advance unit time in PMD
a
t development time for the n th cycle in PMD
T transmittance of beamsplitter at wavenumber, v
T (ix) transmittance of sample at wavenumber, v
T transmittance of light-pipe
0 phase lag in interferogram for wavenumber, v
Vi volume between the half-width points of a
8 chromatographic peak
V .,, volume of a cell
cell
xi
-------
SECTION 1
INTRODUCTION
The rapid identification of trace organic pollutants in water presents
one of the more severe problems for environmental analytical chemists today.
The usual procedure is to extract the organics from water by some means, and
to analyze the extract (often after further concentration) by one or more
chromatographic techniques., The most commonly used technique is gas
chromatography, although high performance liquid chromatography is being used
to an increasingly large extent for less volatile components. Thin-layer
chromatography is now used less frequently to identify individual pollutants,
but rather to determine the best conditions for their separation by high
performance liquid chromatography. Identification of components separated
by gas and liquid chromatography can be attempted by comparison of retention
times of the unknown peaks with the retention times of standards, but results
found in this way are equivocal at best. Spectroscopic identification of
chromatographically separated fractions, preferably without trapping each
sample, yields more certain identification of each peak.
The on-line identification of gas chromatographic peaks by mass spec-
trometry (GC-M3) is now a standard method for investigating organic water
pollutants, and detection limits in the low nanogram or high picogram
range may be obtained with state-of-the-art equipment. The primary dis-
advantages of GC-M3 are (a) that spectra of closely related compounds,
especially isomers, are not readily distinguished, (b) spectra may differ
from spectra in reference compilations because of different ionization
conditions in the spectrometer, and (c) reference spectra of the unknown
sample may not be available, in which case the spectrum must be interpreted
a priori. An alternative, or supplementary,general method for the on-line
identification of gas chromatographic peaks would be most beneficial for
studying water pollutants. Fourier transform infrared (FT-IR) spectroscopy
may be the most promising alternative, but until recently gas chromatographic
peaks have not been identified at submicrogram levels by FT-IR spectroscopy.
One objective of this study was to develop methods for increasing the
sensitivity of on-line FT-IR methods for identifying organic water pollutants
separated by gas chromatography (GC-FT-IR).
Components of mixtures separated by high performance liquid chromato-
graphy are-less easily identified by mass spectrometry because the samples
are generally less volatile than samples separated by gas chromatography and
because they are more difficult to separate from the mobile phase. FT-IR
spectroscopy therefore may be presented as a viable alternative for the
on-line identification of species separated by this technique, and a second
objective of this project was the development of FT-IR methods for identifying
-------
species separated by liquid chromatography with practical detection limits
of less than one microgram.
The final objective was the refinement of FT-IE methods for identifying
species separated by thin-layer chromatography, without removing the sample
from the plate. Wo general alternative method to infrared spectroscopy is
currently in use, although ultraviolet-visible spectrometry and fluorescence
spectrometry may both be used for certain types of compounds.
Each of these FT-IR methods could be of real importance in identifying
water pollutants at the parts-per-million or parts-per-billion level.
-------
SECTION 2
CONCLUSIONS
Calculations have shown that the optimum dimensions for "light-pipe"
gas-cells, through which the' effluent from a gas chromatograph is passed for
the FT-IE measurement, would cause the transmittance of the cell to be so
great that if interferograms were measured using a conventional rapid-scanning
interferometer and a mercury cadmium telluride photodetector, the signal-to-
noise ratio near the point of stationary phase would he so great that the
limiting source of noise would be the digitizing system rather than the
detector. To avoid this problem, dual-beam FT-IE spectroscopy can be used to
eliminate the large signal due to the source and leave only the small signal
caused by the absorbing sample. Even though the stability of our interfero-
meter was not good, and the response of the detector was nonlinear at the
high light levels of this experiment, the detection limits of dual-beam FT-
IE spectroscopy for the on-line identification of gas chromatographic peaks
were about four times lower than were found in the corresponding single-beam
measurement. When the system was applied to the identification of trace
organics in water after extraction by neutral polystyrene resins, detection
limits of less than one part-per-billion were observed. The system was
redesigned to be compatible with high resolution gas chromatography using
capillary columns, and preliminary results indicate that detection limits of
less than 10 nanograms of strongly absorbing samples will ultimately be
achieved.
The on-line identification of submicrogram amounts of samples separated
by high performance liquid chromatography without separating the solvent
proved to be impossible even using dual-beam FT-IR spectroscopy; detection
limits of about 10 |j.g for strong absorbers and 100 |_ig for weak absorbers were
found. Several semi-continuous methods of solvent removal followed by the
rapid measurement of the FT-IE spectra of each solute were studied. The most
successful method to date involves a preliminary concentration step, in which
the effluent from the chromatograph is sprayed into a heated tube, followed
by a deposition step, where the concentrated solution is sprayed onto a small cup
containing powdered potassium chloride. After the peak is eluted the cup
is automatically transferred to the focus of a diffuse reflectance attachment
of an FT-IE spectrometer. Submicrogram detection limits for nonvolatile
compounds are found with the technique.
Transmittance spectra of the sample spots on developed thin-layer
chromatography (TLC) plates can be measured provided that the adsorbent is
deposited on a silver chloride backing. The effect of scattering by the
adsorbent can be minimized by treating the plate with an oil whose refractive
index matches that of the adsorbent. Submicrogram detection limits can be
-------
achieved in very short measurement times if the spot size is minimized by
the use of programmed multiple development TLC, thereby demonstrating the
feasibility of an automated FT-IR scanner for TLC plates. The use of
diffuse reflectance FT-IR also appears to be a promising method for the
identification of sample spots on glass or aluminum-foil backed TLC plates;
again submicrogram detection limits have been observed.
-------
SECTION 3
RECOMMENDATIONS
The most important work to be finished in this study is the development
of the device for the semi-continuous identification of species eluting from
a high performance liquid chromatograph based on diffuse reflectance infrared
Fourier transform spectroscopy. Several parts of this device require
optimization. The propellant gas flow rate, temperature and dimensions of
the preconcentrator, the method of applying the concentrated solution to the
alkali halide powder, and the particle size, depth and temperature of the
powder all require further investigation to optimize the performance of this
device.
The true potential of dual-beam FT-IR spectroscopy should be studied
after stabilizing the interferometer with the fast retrace electronics. For
the study of gas chromatographic effluents, the increased stability should
lead to improvements in phase correction, reduction in the effect of atmospheric
interferences, and probably a small increase in sensitivity. For the study
of liquid chromatographic effluents, the increased interferometer stability
should lead to substantially improved solvent compensation and a proportion-
ately greater increase in sensitivity than for the gas chromatography work.
Further studies on improving the performance of the system designed to
measure the spectra of compounds eluting from gas chromatographs using
capillary columns are required. A dual-beam FT-IR system with 2mm diameter
light-pipes should be built, so that its performance can be tested against
the corresponding measurements made using 3mm diameter light-pipes. After
optimizing the performance of a. "breadboard" system, an all-glass system
should be constructed to allow all organic compounds to be measured without
decomposition, preferably using a nondestructive photoionization detector.
Ideally, a combined system for simultaneously identifying species eluting
from a gas chromatograph and a high performance liquid chromatograph using two
interferometers controlled by a single data system should be built. This
system should be applied to the identification of the organic pollutants in
actual water samples taken from a variety of sources.
Finally our preliminary investigation into the application of diffuse
reflectance infrared spectroscopy for the elucidation of a wide variety of
environmental problems should be expanded. Problems to be studied should
include not only the identification of species on TLC plates, but also the
investigation of organic compounds, especially pesticides, sorbed onto clays
and other minerals, and even the potential for qualitative and quantitative
analysis of organics sorbed onto charcoal and other adsorbents from industrial
atmospheres.
-------
SECTION k
THEORETICAL CONSIDERATIONS IN GC-FT-IR
OPTIMAL LIGHT- PIPE VOLUME
To optimize GC-FT-IR sensitivity, it is necessary to consider the per-
formance of the chromatograph, the spectrometer and the light-pipe gas cell
through which the effluent from the chromatograph is being passed while its
spectrum is being measured. The dimensions of the light-pipe are of critical
importance, and as the first part of this project >the optimum dimensions for
GC-IR light-pipes were calculated; these results have been published (l) and
are summarized below. Other factors which have been studied include the data
collection trigger, the performance of the spectrometer and detector, and the
chromatographic separation. The conclusions from these calculations are
summarized in the following subsections of this section of the report.
It is intuitively apparent that, if the volume of the light-pipe, V ,
is much larger than the volume between the half -width points of any chromato-
graphic peak, Vi, two peaks which are just resolved by the chromatograph could
be present simultaneously in the light-pipe. In this case the resolution of
the chromatograph could be seriously degraded. In addition the maximum value
of the average concentration of the sample in the cell, C , will be sub-
stantially less than the maximum concentration of the sample in the peak, C ,
max
as it elutes from the column. On the other hand, if V « Vi_, the maximum
C"_L J_ i^
quantity of sample in the cell, Q , will be much smaller than the total
quantity of sample in the peak, CL , , even though C approximates C
T3Ou3»J- CGJ __ L 1
The variation of C /C and Q -n/Q^ 4. T with V , -,/Vi was calculated
cell' max cell' ^total cell7 ^
for triangular GC peaks, and the result is shown in Figure 1. Calculations
which assume a Gaussian GC peak shape yield a very similar plot. The inter-
section of these two curves occurs at the optimum value of V ,/Yi, and it
o
can be seen from Figure 1 that this occurs when V ,-,-VJL. Under these con-
CGJ — L "o"
conditions the GC-IR peak (i.e. the variation of the absorbance of a sample
in the light- pipe as a function of time) should be wider than the GC peak by
approximately 50$,. We believe that this small degradation in resolution
is a small price to pay for optimizing the sensitivity of on- the- fly
GC-FT-IR measurements.
-------
To quantitate the effect of this peak broadening on spectral interferences
if V = Vi, the cross-contamination of neighboring GC peaks of equal
CGJ L p
concentration and half-width, separated with a chromatographic resolution,
Rg, was calculated. R is defined as the ratio of the difference in the
retention volumes of the peaks to the average of the half-widths (2). Figures
2 and 3 show the extent of interference by neighboring triangular peaks for
R = 1.0 and 0.8, respectively. It can be seen from these figures that, if
the data are only collected between the half-width points of the peaks (after
allowing a suitable delay), the amount of cross-contamination is negligibly
small, unless the interfering peak is a much stronger infrared absorber than
the peak of interest. Similar results may again be shown for Gaussian GC
peaks, the extent of cross-contamination being about 50$ greater than for
triangular peaks.
OPTIMAL DATA COLLECTION TRIGGER
Several methods of data collection for GC-FT-IR have been proposed. In
the simplest each interferogram is stored separately; however, such a scheme
1.00
0.75 -
0.50 -
0.25 -
0.8 1.2
1.8
V
2.0 2.4 2.8 3.2
CELL /Vl/2
Figure 1. Variation of Q__^/Q and
3.6 4.O
with V -i-i/Yi for triangular GC peaks.
v_S J—L o
-------
implies that 3600 interferegrams of 1 sec. scan time must be stored for a
chromatogram of 1 hour duration,. This in turn imples a very large data
system. For
cm
resolution interferograms, it is usual to collect 4K data
points in double-precis ion, so that about 2.6 x 10s words of storage are
needed. In addition the data would take an inordinately long time to compute,
even with a faster transform than is available on current commercial FT-IE
spectrometerSo It is possible to reduce the amount of data storage by
co-adding several successive interferograms, but this approach can degrade
the chromatographic resolution and still does not completely solve the
problem of an excessive data storage requirement.
A superior approach is to collect data only when a sample is present in
the light-pipe at a concentration which will yield spectra at a sufficiently
high signal-to-noise ratio (S/N) to permit several absorption bands to be
identifiable. The first GC-FT-IR systems used a sensor which permitted
interferograms to be signal-averaged during the period when the signal from
the GC detector exceeded a certain threshold, after allowing for a delay
time to take into account the time difference between the maximum signal at
the GC detector and the maximum sample concentration in the light-pipe. For
successful operation, this system relied on the operator to set the threshold
at a value which was large enough to allow small peaks (which would not yield
an identifiable spectrum) to pass through the light-pipe without being
measured, but small enough to permit a reasonably large number of interfer-
ograms to be averaged across any given GC peak. A further disadvantage of
this type of trigger is found for partially resolved GC peaks for which the
PEAK
PEAK B
i.o
4.0
2.0 3.0
V/VCELL
Figure 2. Peak spreading in a cell with V ,, = Vi for
triangular GC peaks separated by 2 V± (E = 1.0).
2 S
The triangular peaks represent the signal at the
GC detector, and the smooth curve gives the sum
of the values of 0 ,., for A and B. Data should-
cell
be collected between the half width points of the
GC peak after a time delay, t".
5.0
-------
dip between the peaks always exceeds the threshold; in this case, signal-
averaging continues across the unresolved peaks. In addition, a drift of
the "baseline of the chromatogram can cause data acquisition to be initiated
and continued until the operator resets the baseline below the threshold.
It can be shown (l) that a close approach to the optimum sampling
trigger for GC-FT-IR could be actuated by the inflexion points of the
Gaussian GC peaks. It is relatively simple to electronically sense these
points of inflexion using the zero-crossings of the second derivative of the
GC signal. We calculated that if signal-averaging were performed for this
period of time, the S/l would be greater than 90^ of the maximum possible
value. If the GC peaks are separated to the extent that there is even the
smallest dip between the peaks, interferograms for each such resolved peak
will be averaged separately. One of the disadvantages of a "second derivative
trigger" is that it is possible to initiate and terminate data collection by
noise, and one method of discriminating against the effect of noise is to
actuate the second derivative trigger only when the signal exceeds a certain
threshold value. A trigger of this type, however, would still be adversely
affected by baseline drift.
An even better system would be a microprocessor-based trigger which
would initiate data collection using a second-derivative trigger after
checking for a period of time double or triple the time constant of the GC
detector amplifier to check if a second zero crossing (presumably generated
Average % B during data collection - 1.96%
I Or
Figure 3. Peak spreading in a cell with V = VA for
t^G-L-L g
triangular GC peaks separated by 1.6 V± (R =0.8).
"a s
The upper broken line represents the ratio of the
quantity of peak A to the total sample quantity
in the light-pipe. If data acquisition occurs
between the half-width points of each peak, after
the optimum time delay, t", the average percentage
contamination of peak A. by peak B (as shown by
cross-hatching) is less than 2%.
-------
by noise) has occurred. We are currently building such a system based on
the KIM-1 microcomputer (MOS Technology, Norristown, PA). Because of the
difficulties involved with 60 Hz pick-up on the microvolt level signals-in
this device, we have not yet been able to construct a working trigger.
OPTIMAL LIGHT-PIPE DIMENSIONS
If an extremely low noise infrared spectrophotometer with a very wide
dynamic range was available for GC-IR measurements, it is apparent that the
light-pipe should be as long and narrow as possible, consistent with its
volume, V , being equal to VI as discussed above. However the baseline
noise level and dynamic range of rapid-scanning Fourier transform spectro-
meters are by no means ideal, and it is necessary to consider the trade-off
between the length, L, and diameter, d , of light-pipes on GC-IR sensitivity.
Mantz (3) has studied the problem fromp theoretical grounds and has derived
an equation relating the transmittance of a light-pipe, T , to the reflectivi-
ty of the coating, R, the half-angle of the incoming beam^ a, and the length
and diameter of the light-pipe:
T = R(-2L/d ) In cos a
P
If the reflectivity of the coating is known, the optimum dimensions of the
light-pipe could be calculated in this way, but coating imperfections and
the variation of R with polarization and incidence angle make this approach
less than useful in practice.
We developed a more practical basis for determining T . By using a
single-beam GC-FT-IR arrangement in which the diameter of the focused beam
at the entrance of the light-pipe, cU, was 0.3 cm, we were able to derive
an empirical expression for the transmittance of any light-pipe used on this
system with d < 0.3 cm:
Tp . 0.9 (A-) " 0.32 (o.ooa/ip)
In this expression the first term (0.9) is due to the reflection loss at the
windows, the second term is due to vignetting of the beam at the entrance
to the light-pipe, and the third term is due to reflection losses down the
light-pipe. If d > df, the second term is omitted.
The noise level of the spectral baseline (100$ -line) should be inversely
proportional to T , and for weakly absorbing bands (typical of the GC-FT-IR
experiment) the sample absorptance should be approximately proportional to
the absorbance which is proportional to L. Thus the S/N should be propor-
tional to the product of L and Tp. Plots of LT vs L for different values
of V revealed that the maximum usually occurred with d = df, 'and that
the transmittance of the cell at maximum efficiency is usually greater
than 20fo, see Figure i|-. This rather high transmittance has important
implications 'for FT-IR measurements in which a mercury cadmium telluride
(MCT) detector is used instead of the conventional triglycine sulfate (TGS)
detector, as discussed in the next -section.
10
-------
SPECTROMETER PERFORMANCE
_In.t erferomet er
Commercial FT-IR spectrometers usually have a 50 mm aperture and a
Ge:KBr beamsplitter. In theory there is little one should be able to do to
improve the performance with the exception of increasing the aperture (to
increase the optical throughput) or changing the scan speed of the moving
mirror. The former operation cannot be performed on any interferometer once
it is built, while the scan speed cannot be changed on the majority of
commercially available interferometers. However it should be recognized that
the nature of the detector determines the optimum scan speed. The TGS
detector works best at relatively low modulation frequencies, thereby
necessitating a slower scan speed than if the MCT detector is being used.
In a comparative study of the relative performance of the spectrometer
being used for this project at Ohio University, a Digilab Model FTS-1^
(Digilab, Inc., Cambridge, MA) which was constructed in 1973, and a recent
instrument with the same specifications, it was found that the modern
•3r
T • 0.070
10
30 40 SO 60 70
LENGTH OF LIGHT-PIPE (CM)
80
90
too
Figure 4.
Calculated variation of LTp as a function of light-pipe
length for k-mL (upper trace) and 1-ml (lower trace)
light-pipes. When Tp exceeds 0.07, the S/N of measure-
ments made with an MCT detector is limited by digitiza-
tion noise rather than detector noise.
11
-------
instrument gave substantially superior performance (k), especially at high
frequency. It is believed that the improvements to commercially available
spectrometers are due to improvements in beamsplitter design and in the
signal electronics. Comparison between the work described in this report and
published work performed on later instruments should be made with the
decreased performance of the early instruments in mind.
Source
The two commercially available sources for mid infrared FT-IR spectro-
meters are the nichrome wire and the globar. lichrome wire sources operate
at a lower temperature than globars and, unlike globars, do not require
water-cooling; however the emitted energy from a nichrome wire source is
somewhat less than that from a globar at all wavenumbers.
A compromise between the convenience and cheapness of nichrome wire
sources and the performance of a globar is found in the Wernst glower. The
Nernst glower operates at a higher temperature than the globar but does not
require water cooling. The primary disadvantage of the Nernst glower is its
low emissivity above about 2500 cm -1, but it is an excellent source for the
region from 2000 to 100 cm"1. Comparative spectra of the Nernst glower and
the nichrome wire source measured .identically in all other respects are
shown in Figure 5« The nichrome wire was an original component of our
Digilab FTS-1^ spectrometer and the Nernst glower was an x37 type (Perkin-
Elmer Corp., Norwalk, CT) powered wrfti the original d,c. supply of the
FTS-14.
Detector
The TGS pyroelectric bolometer is usually used on mid infrared FT-IR
spectrometers. The TGS detector is not as sensitive as conventional thermal
detectors such as the thermocouple, but it has the advantage of a very short
response time which is necessary because the scan speed of the moving mirror
of a rapid-scanning interferometer has to be kept high enough to keep the
S/N of the interferogram near the point of stationary phase below the
dynamic range of the analog-to-digital converter (ADC) being used to
digitize the signal. For an interferometer with a 50 mm aperture set up for
mid infrared spectrometry (hOOO-kOO cm x) using a TGS detector, the scan
speed to keep the S/N below the dynamic range of a 15-bit ADC is approxi-
mately 3 mm/seco
The average detectivity, D*, of the MCT detector is about twenty times
greater than that of the TGS detector and, if an MCT detector were directly
substituted for the TGS without attenuating the source in any way, the scan
speed of the interferometer mirror would have to be increased by a factor of
(20)2 to avoid digitization noise. Even though the MCT detector could still
respond to all frequencies in the interferogram, the sampling frequency
would have to be raised to about 1 MHz and the data acquisition rate would
far exceed the capability of a state-of-the-art 15-bit ADC. Even if a
light-pipe with Tp = 0.25 were interposed between the interferometer and the
light-pipe, a sampling frequency in excess of 100 kHz is required which still
exceeds the maximum rate for a 15-bit ADC.
12
-------
To optimize the sensitivity of GC-FT-IR measurements it is still
necessary to use the most sensitive detector possible, but under conditions
where the signal is limited by detector noise rather than digitization noise
and without changing the scan speed of the interferometer (so that the limits
of the band pass filter in the amplifier do not have to be changed from the
values currently available on commercial FT-IR spectrometers). To achieve
this goal it is necessary to eliminate the very high signal near the point
of stationary phase of the interferogram, which is due to the intense
continuous spectrum of the source, while retaining the small modulations due
to the sample. Some possible ways of achieving this end are as follows:
or
UJ
z
UJ
UJ
o
cr
o
z
o
tr
UJ
UJ
z
tr
UJ
z
4000
3000 2000
FREQUENCY (CM"1)
1000
Figure. 5-
Relative energies of a Nernst glower and a nichrome
wire infrared source as a function of wavenumber.
The effect of the low emissivity of the Nernst
glower around JOOO cm"1 is apparent.
13
-------
a. Reducing the spectral range of the measurement, either by
an optical filter or by reducing the temperature of the source.
Either of these two courses of action would reduce the
multiplex advantage of FT-IR spectrometry and is less than
desirable.
b. By "clipping" or "blanking" the interferogram (5).
technique severely degrades the photometric accuracy of
the measurement and creates a very poor baseline on which
it is extremely difficult to observe weak absorption bands.
c. By "chirping" the interferometer (6,?). This technique
increases transform time substantially and has never been
tested for Interferograms of high dynamic range since it is
believed that the S/N of spectra computed from chirped
interferograms is less than that of spectra computed from
unchirped or only slightly chirped interferograms „
d. Through the use of dual-beam, or optical subtraction,
techniques (8-ll). The apparent advantages of dual-beam FT-IR
appear to substantially outweigh the potential disadvantages
and this was the technique by which we decided to attempt to
study the use of optimized light- pipes for high sensitivity
GC- FT-IR. The theory of dual-beam FT-IR is described in
Section 5> together with descriptions of our optical system
and comparative results. The application to GC- FT-IR is
described in Section 6.
OPTIMAL CHROMATOGRAPHIC PERFORMANCE
A truly versatile GC- FT-IR system should be able to be used to identify
peaks separated on both packed and capillary columns. However one important
goal of GC-FT-IR research in general, and this project in particular, is to
reduce detection limits (in terms of nanograms of a given component injected
onto the column) to the- maximum possible extent. It can be shown (l) that
the small Vi_ of peaks separated on capillary columns allows one to use cells
of smaller 2 volume than is necessary for packed columns, and that smaller
detection limits should therefore be predicted. It should be noted that the
increased capacity of packed columns may sometimes offset the low detection
limits obtainable with capillary columns by allowing a greater volume of the
analyte to be injected onto the column initially. This effect is particularly
important for wall -coated open tubular (WCOT) capillary columns, onto which
it is customary to .inject about 10 nl of the analyte solution. However the
capacity of support- coated open tubular (SCOT) columns is usually sufficiently
large that microgram quantities of components can be injected without
degrading the performance of the column. The use of SCOT columns therefore
presents a good compromise between the resolution of WCOT columns and the
capacity of packed columns, and these columns should allow the lowest
detection limits of GC-FT-IR to be attained.
14
-------
In this project we first attempted to optimize GC-FT-IR performance
using 1/8-in, o.d. packed columns, since this is the type of column most
generally used in gas chromatography today. Towards the end of the project
we fitted the chromatograph with a SCOT column to determine the extent to
which GC-FT-IR detection limits would be reduced. The results are described
in Section 6.
15
-------
SECTION 5
DUAL-BEAM FOURIER TRANSFORM INFRARED SPECTROMETRY
THEORY
If the collimated input beam of a conventional Michelson interferometer
enters at the usual angle of +5° to the plane of the beamsplitter, one of
the emerging beams returns to the source and is not measured; this beam is
usually called the "reflected beam", and is denoted by the subscript B in
this report. The other beam (the "transmitted beam" which is denoted by the
subscript A in this report) is passed through the sample and is then measured
at the detector. If the beam enters at an angle other than ^5° to the
beamsplitter both the transmitted and the reflected beams could be detected.
The intensity of the interferogram of the transmitted beam, !'(§), is
given as a function of retardation, 5 cm, by the equation:
+00
l(v).dv + 2 I R T l(v) cos(2itv8 +0v).dv
where l(v) is the intensity of the source at wavenumber vcm"1,
R is the reflectance of the beamsplitter at vcm"1,
T is the transmittance of the beamsplitter of vcm"1,
and QV is a small wavenumber dependent phase angle.
The modulation of the reflected beam is 180° out-of-phase with the
transmitted beam, and it can be shown that:
+00 +00
I'(5)T3 = 1 (R2 + T2) I(v).dv - 2 \ R T I(v) cos(2rtv8 + 9 ).dv
r3 \ V V I V V V
o o
If both beams are somehow measured at the same detector, the signal is
given by: +
+00
+ I'(5L = 2 f R T l(v) dv + f (R2 + T2) l(v).dv
.D I V V I V V
o o
oo
= f (R + T )2 l(v).dv
\ V V
16
-------
For a non-absorbing beamsplitter:
R + T = i
v v
Therefore
4-co
B = ^ I(v>'dv
o
i.e. the total energy entering the interferometer equals the total energy
emerging and the sinusoidal components of I'(s)A and I'(o)B are nulled out.
In rapid-scanning mid infrared FT-IR spectrometers, the amplifier for the
detector is a.c. coupled so that only the sinusoidal (modulated) component
of the interferogram is measured. Thus if no sample is in either beam, no
a. c. signal would be measured. On the other hand, if a sample of trans-
mi ttance T(V) is placed in the reflected beam:
00 00
I'(oL = \ (R2 + T2)T(v)l(v).dv - 2 ( R T T(v)l(v) cos (2rtv6 + 0 ).dv
O \ V V \ V V V
o o
If the modulated components of the transmitted and reflected interferograms
are denoted by 1(5)fl and 1(5) , respectively, we have that:
s\ -D
+00
\ EvTv[l-T
l(o)A + l(s)B = 2 \ ET[l-T(v)] I(v) cos (2jtv&
Therefore in this case, the greater the transmittance of the sample,
the smaller is the magnitude of the a. c. interferogram. For very small
sample quantities, [l-T(v)] is usually very small, so that if an MCT detector
were directly substituted for a TGS detector without changing the scan speed
of the interferometer, the resultant interferogram should not be digitization
noise limited.
T(V) can be calculated by dividing the measured dual-beam spectrum by
the single-beam spectrum, l(v), calculated from the transform of l(s) , and
subtracting the result from unity.
OPTICAL DESIGN
The optical system (12) that we designed for dual-beam FT-IR is shown
in Figure 6. The source was the Rernst glower described in the previous
section. The beam from this source was collimated by a k^° off-axis segment
of a 75 mm focal- length paraboloidal mirror (Special Optics Corporation,
Little Falls, NJ), which provided a beam of the maximum allowed throughput
for wavenumbers below if-000 cm l for measurements made with a resolution,
Av> 2 cm"1. The interferometer was a Model 296 Michelson interferometer,
which is a component of the Digilab Model FTS-14 Fourier transform spectro-
photometer. The emerging beams were picked off by two front- surf ace, gold-
17
-------
coated plane mirrors (Edmund Scientific Corporation, Harrington, N.J.) and
focused using two off-axis paraboloidal mirrors identical to the collimating
mirror for the source; the foci for the sample and reference beams were
separated by 6 inches, which was the separation of the two light-pipes in
our preliminary GC-IR set-up, vide infra. The beams were then refocused
onto a liquid nitrogen cooled MCT detector of 2mm side (Texas Instruments,
Dallas, TX) using two off-axis sections of an ellipsoid of focal lengths
109 and 262 mm (Special Optics Corp.).
The system was designed to allow the detector and the two ellipsoids to
be moved back as a unit, so that dual parallel light-pipes could be placed
with their entrance apertures at the sample and reference foci, as shown in
Figure 7- Without the light-pipes present, this dual-beam FT-IR system
could be used for studying conventional solutions at low concentration. If
a flow-through cell was placed in the reflected beam and a reference cell of
equal pathlength were placed in the transmitted beam (or vice versa), this
system could be used for on-line measurements of samples eluting from a
high performance liquid chromatograph.
SOURCE
Figure 6. Optical layout for dual-beam FT-IR spectrometry.
18
-------
DIGILAB
MODEL 296
INTERFEROMETER
IR SOURC
OFF - AXIS
PARABOLOIDAL
MIRROR '
I
rMCT DETECTOR
OFF - AXIS
ELLIPSOID
Figure J. Optical layout for GC-IB using dual-beam FT-IR optics.
19
-------
HIACTICAL ffiRFORMMCE
Nulling Ratio and Baseline Compensation
With no sample present in either beam it was never possible to completely
null the interfere gram. At the minimum amplitude of the dual-beam signal,
the transform of this interferogram did not have the same profile as the
single-beam spectrum, which would be the case if the incomplete null was due
to optical misalignment. Apparently much of the spectral structure is due to
species present on the surface of the germanium beamsplitter next to the
compensator plate. Similar bands had been observed previously during earlier
work using a low resolution rapid-scanning interferometer (9,10). To
eliminate the effect of these bands and to create a flat baseline, a dual-
beam interferogram was measured with no sample in either beam, and both this
interferogram and its transform were stored. After the dual-beam inter-
ferogram of the sample was measured, baseline compensation was achieved by
either subtracting the background interferogram from the sample interferogram
or subtracting the background spectrum from the sample spectrum. For certain
measurements the former procedure gave the flattest baseline while for others
the latter procedure gave the superior results. We were unable to develop
any rationale to forecast which procedure gave better background compensation.
Transmittance Spectra
That spectra corresponding to conventionally measured transmittance
spectra can be calculated from dual-beam spectra is illustrated in Figure 8.
In this series of spectra, which were measured using a TGS detector, A shows
a dual-beam spectrum of a 50 jjm film of polyethylene held in beam A, B shows
a dual-beam background spectrum, and C shows the result of subtracting B
from A. D is the single-beam background spectrum and E is the result of
dividing spectrum C by -D and subtracting the result from unity. A
comparison transmittance spectrum of polyethylene measured in the conven-
tional fashion by ratioing a-single-beam sample spectrum by a single-beam
reference spectrum is shown in F. It can be seen that the sharp band at
720 cm •"• is distorted. This distortion is apparently caused by a problem
in phase correcting the dual-beam interferogram; this problem will be dis-
cussed later in this section. However spectra E and F are otherwise identical.
S/N Advantage
The advantage, in terms of S/N, of the dual-beam system with an MCT
detector over a conventional single-beam system with a TGS detector is
illustrated in Figure 9 t»y spectra of a 0.01$ solution of anisole in CC14
contained in a cell of 93 (Jin pathlength. For the dual-beam measurement, a
variable pathlength cell containing pure CC14 was held in the reference beam,
and its pathlength was adjusted to minimize the signal at the detector. For
the single-beam measurement, the single-beam spectrum of the solution was
ratioed against a single-beam spectrum of pure CC14 held in the same cell.
Both sets of spectra were measured with 10 scans for sample and reference
spectra. The strongest band in this spectrum had an absorbance of 1. TxlO~4,
and the noise equivalent absorbance was less than 1x10 5. It is apparent
20
-------
that with less than 1000 scans, bands with an absorbance of less than 10"6
could be observed. The performance of the dual-beam system below 2000 cm"1
is at least an order of magnitude better than any spectrometer known to the
principal investigator for highly transmitting samples.
Figure 8. Calculation of a transmittance spectrum from spectra
measured using a dual-beam FT-IE spectrometer (see
text for details).
21
-------
Fhase Correction
The biggest problem encountered in this work was the generation of
phase errors of the type mentioned earlier with reference to the distortion
of the 720 cm"1 band in the polyethylene spectrum shown in Figure 8. Several
approaches were taken to improve the accuracy of phase correction on our
dual-beam FT-IE system, including the following:
a. extending the number of data points around the point of
stationary phase used in the Mertz method of phase correction (13);
SB
DB
I6OO
8OO
Figure 9- Spectra of 0.01$ anisole in carbon tetrachloride at a
pathlength of 96 (im measured using a single-beam (SB)
FTrIR spectrometer with a TGS detector and a dual-beam
(DB) FT-IE spectrometer with an MCT detector.
22
-------
ID. storing a phase spectrum calculated from a dual-beam
interferogram of a screen with a spectrally uniform
transmittance of about 10$, and using this array for
subsequent transformations;
c. slightly misaligning the dual-beam optics to increase the
amplitude of the interferogram near the point of stationary phase.
None of these approaches, however, gave significantly improved band symmetry.
The root of the problem was ultimately traced to an instability in our
interferometer causing large variations in the shape of the measured inter-
ferogram from scan to scan. Although the magnitude of these variations was
rather small for single-beam measurements (typically 1-2$ near the point of
stationary phase), the dual-beam interferograms could vary by as much as
50$ in the same region, see Figures 10 and 11.
The problem appears to be confined to interferometers of the particular
model being used for this project. The interferometer, which was purchased
in 19T3j shows no significant design changes from the first ones made by
Block Engineering, and later by Digilab. When the moving mirror of these
interferometers is retraced after each active scan, it is stopped by a foam
rubber pad. An audible "thump" can be heard on impact. Apparently this
sets the mirror into oscillation, and it has to be held stationary for a
period of time (usually a little over a second) to allow these oscillations
to die down before the next scan is initiated. Because the mirror has to
be held stationary for such a long period of time, the duty cycle efficiency
for signal-averaging is rather low - about 20$ for measurements made at
8 cm"1 resolution on our instrument. Perhaps even more importantly for this
project, even though the amplitude of the mirror oscillations does decrease
during the period the mirror is held stationary, the mirror never seems to
be completely stable, so that even when the duty cycle efficiency was
intentionally decreased still further, significant scan to scan variations
were still observed in the dual-beam interferograms.
The duty cycle efficiency of Digilab interferometers made after 1975
was improved by changing the way in which the moving mirror is retraced
after each scan. As the mirror approaches the end of its retrace, its
position is sensed photoelectrically, and a retardation pulse is applied to
the coil of the drive transducer of large enough magnitude to halt the
mirror and start it moving forward. After this point, the mirror motion is
controlled by the same triangular ramp voltage as in the earlier inter-
ferometers. In this way the duty cycle for 8 cm"1 measurements has been
increased from 20$ to 90$• Thus four times as many scans can be taken per
unit time, and the resultant spectra have a S/W which is twice as great as
spectra measured in the same time on earlier interferometers.
Because this new fast retrace mechanism does not involve halting the
mirror by a mechanical stop, there is less cause for it to be set into
oscillation when it retraces; we therefore believed that the dual-beam
stability should be much better if a new interferometer were used. We
23
-------
',/VA '
Figure 10. Reproducibility of interferograms.
A. Interferograms measured on a dual-beam FT-IR spectrometer.
B. Interferograms measured on a single-beam FT-IR spectrometer.
24
-------
Figure 11. Reproducibility of dual-beam FT-IR spectra of 0.1$ anisole
in carbon tetrachloride in a $6 p.m cell. Each spectrum
was computed from a single-scan interferogram. The effect
of phase correction errors is readily apparent.
25
-------
tested this hypothesis by driving our dual-beam optical bench to the Digilab
facility at Cambridge., MA, and found that indeed the scan-to-scan variation
in dual-beam interferograms was less than 2$. Were a new interferometer
used for dual-beam measurements, the phase correction problem will be much
alleviated. The improved photometric accuracy might even result in a further
improvement in S/N, so that it may be possible to detect bands with an
absorbance significantly less than 10"s in a reasonable measurement time.
Detector Linearity
The actual sensitivity improvement of dual-beam FT-IE using an MCT
detector compared with single-beam FT-IE using a TGS detector for measure-
ments using typical sources at full throughput is of the order of factor of
ten. Our initial calculations using manufacturer's quoted D* data
suggested the improvement should have been about twice as good as it was
found to be. However, when the intensity of the collimated input beam was
attenuated using screens of various transmittance values the improvement
factor was found to increase,, This is presumed to be due to saturation of
the detector at high photon flux. The response of the TGS detector
increases linearly with photon flux, but the response of the MCT detector
only increases with (approximately) the square root of photon flux, see
Figure 12.
In spite of this factor operating against the use of dual-beam FT-IE,
if the S/N of single-beam interferograms measured with a TGS detector is
close to the digitization noise limit (which, in practice, it usually is)
and one desires to increase the sensitivity through the use of an MCT
detector, the largest advantage of dual-beam FT-IE compared to single-beam
FT-IE is found at the highest photon flux. With no sample or accessory
attenuating the beam, dual-beam FT-IE with an MCT detector gives about an
order of magnitude improvement over the corresponding single-beam measure-
ment. For GC-FT-IE measurements, the light-pipe usually attenuates at least
60$> of the incident radiation, so that dual-beam FT-IE gives less than an
order of magnitude advantage compared with the corresponding single-beam
measurement with an MCT detector, but the advantage is still appreciable.
This topic is discussed in greater detail in the next section.
26
-------
} 10 20 30 40 5O BO 7O 8O SO IOO
PERCENT OF SOURCE RADIATION
Figure 12. Linearity of the response of the MCT and TGS detectors at
high photon flux. The scale for the TGS detector has been
expanded by a factor of ten relative to the scale for the
MCT detector.
27
-------
SECTION 6
GAS CHROMATOGEAfflY AND FT-IR SEECTROSCOPY
GC-FT-IR SYSTEMS FOR PACKED GC COLUMNS
The first measurements in this project were optimized for gas chromatog-
raphy performed using 1/8-in. o.d. packed columns. Typical peak half-widths for
compounds separated in this way are k ml or greater,, To study these peaks,
we purchased and interfaced equipment consisting of two parallel light-pipes,
each of which was JO cm long and 4 mm square in cross-section (4.8 ml volume),
held 6 inches apart in an oven assembly (Norcon Instruments, S. Norwalk, CT).
Our first optical arrangement using these light-pipes was strictly a single-
beam FT-IR system using only one of the light-pipes. The system was designed
primarily for use with a TGS detector since the transmittance of these light-
pipes was so large (approximately JOfO that interferograms measured using an
MCT detector would have been digitization noise limited. The performance
of this system was described in a previous E.PoA. report from this laboratory
(l4). Identifiable spectra of microgram quantities of components could be
measured without trapping the samples, and a few bands of submicrogram
quantities of strongly absorbing compounds could be observed. Later single-
beam and dual-beam work (15) was performed using the optical system shown in
Figure 7-
Comparative GC-FT-IR detection limits for spectra measured on-the-fly
using a TGS detector (single-beam) and an MCT detector (single-beam and
dual-beam) are summarized in Table 1. The single-beam measurements were
taken on the dual-beam optical system with the reflected beam blocked, so
that these data are directly comparable. As forecast in the previous section,
it was found that dual-beam GC-FT-IR spectra measured with an MCT detector
are slightly more than an order of magnitude more sensitive than measurements
taken on the single-beam system with a TGS detector, and about a factor of
four better than measurements taken on the single-beam system with an MCT
detector.
The detection limits obtained were controlled not only by the type of
optical system being used but also by the nature of the samples and the
efficiency of the chromatography. Strongly polar, volatile compounds, such
as anisole, could be detected at very low quantities (less than 100 ng),
whereas less polar, less volatile compounds, such as the chlorinated pesticides,
presented far more of a problem and often required an order of magnitude more
sample to be injected into the chromatograph to yield an identifiable spectrum.
28
-------
TABLE 1
Experimental detection limits for a group of organic compounds
in the dual-beam and single-beam configuration
Compound
Anisole
Chlorobenzene
Diethyl Malonate
Acetonitrile
Benzonitrile
Aldrin
Perthane
p,p»-DDT
Heptachlor
Detection limits (p.g)
Single-beam
(TGS)
0.8
1.
2.
5.
10.
5.
10.
6.
8.
(MCT)
0.20
0.25
0.45
1.5
2.5
1.5
2.5
1.5
2.0
Dual-beam
(MCT)
0,050
0.075
0.100
O.lj-00
0.750
o.koo
0.750
0.400
0.600
All the detection limits in Table 1 refer to the amount of sample
injected into the chromatograph. It should be recognized that after elution
from the column the sample was passed into a 9-1 effluent splitter with 10$
passing to a flame ionization detector (FID) and the remaining 90/° passing
to the light-pipe. In view of the width of the GC peak, not all of this
fraction is present in the light-pipe at all moments during the measurement.
Thus we believe the detection limits in Table 1 to be realistically achiev-
able in practical situations. It may be noted that it has been a practice
of several people working in the field of GC-FT-LR in the past not to use an
effluent splitter and then only to quote the maximum amount of sample _in the
light-pipe rather than the total sample injected into the chromatograph.
IDENTIFICATION OF WATER POLLUTANTS BY GC-FT-IR
The identification of parts-per-million (ppm) and parts-per-billion
(ppb) amounts of organic compounds in water is an important environmental
problem. We have studied the feasibility of applying GC-FT-IR to assist in
identifying organics at this level after sorbing them onto neutral macro-
ret icular polystyrene resins and eluting them off the column using polar
solvents (l6-l8). Table 2 shows the percent recovery from 20-50 mesh XAD-2
resin (Malinkrodt Chemical Co., St. Louis, Mo) for solutions of chlorinated
pesticides in distilled water at a level of 50 ppb. The water flow rate for
most of these measurements was 20 ml/minute and 25 liters of this solution
was passed through the resin. When all the solution had been passed through
the column, the resin was allowed to dry. 100 ml of the eluting solvent was
then passed through the column at a flow rate of 2 ml/minute. This solution
was then evaporated down to a volume of 1 ml, and 10 .|al aliquots were
injected into the chromatograph.
Of the three eluting solvents tested, diethyl ether gave the best
percentage recovery for each pesticide. However, with a water flow rate of
29
-------
TABI£ 2
Effect of varying the eluting solvent and the water flow
rate on the recovery of some chlorinated pesticides using
20-50 mesh XAD-2 resin.
Pesticide
Aldrin
Lindane
DDD
p,p'-DDT
Perthane
Dieldrin
Heptachlor
Hexachlorobenzene
Percent Recovery
Acetonitrile
(20 ml/min)
50
73
78
69
85
89
76
83
Hexane
(20 ml/min)
^3
69
74
72
71
65
59
65
Diethyl Ether
(20 ml/min) (100 ml/min)
9^ 83
100 88
93 79
93 80
96 85
95 83
87 68
89 78
20 ml/minute, it takes 20 hours to elute a 25 liter sample. The effect of
increasing the flow rate above 20 ml/minute to 100 ml/minute decreased the
percentage recovery slightly (typically to about 8Q/0), but reduces the time
required to elute 25 liters of water to about 4 hours.
Representative GC-FT-IR spectra of solutes recovered from aqueous solu-
tion in this way using the dual-beam GC-FT-IR configuration shown in Figure
7 are illustrated in Figures 13 and l4. Figure 13 shows the spectra of these
solutes recovered from 25 liters of a solution spiked with 2 ppb of each
compound. A maximum of 500 ng of each was injected into the chromatograph
(assuming lOOfo recovery) with 10/o of the effluent stream being split to the
FID. Each of these three compounds was a relatively strong infrared absorber,
so that these spectra represent the highest sensitivity which we have been
able to achieve for recovered trace organics from water. Figure l4 shows the
spectra of three far weaker infrared absorbers, chlorinated pesticides,
recovered from 2 liters of a 50 ppb solution of each. The infrared spectra
of the two hexachlorocyclohexane isomers are quite different, in contrast to
their mass spectra which are almost identical. The acquisition of informa-
tion in isomeric molecules is, of course, one of the main advantages of GC-IR
over GC-^B.
The sensitivity of these GC-FT-IR measurements is still lower than state-
of-the-art GC-M3 systems, but the technique should still prove to be of
considerable benefit for the study of trace organics in water. It should be
possible to reduce the detection limits of trace organics in water by GC-FT-IR
in three ways:
a. The volume of the extract from the column could certainly be
reduced below 1 ml; reduction of the volume to 100 ^1 should
reduce the detection limits by about an order of magnitude.
In addition several efficient sample concentrators are now
commercially available (at a price usually between £51000 and ^2000).
Budgetary limitations have so far prevented us from investigating
30
-------
8OO
On-line GC-IR spectra of chlorobenzene, butyl ether and
anisole recovered from 25 liters of a water solution
containing 2 ppb of each component.
31
-------
HtPTACMlOft
IINDANE OK
1.2.3.4.S.6-HEXACHIOI10CYCIOHEXANE<«-ISOMEII)
2000
1,2,3.4.5.6-MEXACHlOHO-
CYCIOHEXANE (-i-ISOMtn |
CM
-1
8010
Figure lk. On-line GC-IR spectra of Heptachlor, and two isomers of
1,2,3,4,5,6-hexachlorocyclohexane recovered from 2 liters
of a water solution containing 1 ppm of each component.
32
-------
their application to this project.
b<> We have demonstrated that the performance of the FT-IR
spectrometer used for this project was not as good as that
of contemporary systems (4). Figure 15 shows a comparison
of the performance of our spectrometer to that of a new
•instrument made to the same nominal specifications by the same
manufacturer for spectra measured by signal-averaging the same
number of scans. Upgrading both the optics and the electronics
of our spectrometer will certainly improve its performance,
especially at high wavenumbero Installation of the fast retrace
mechanism, discussed in Section 5> will not only increase the
duty-cycle efficiency by approximately a factor of four for low
resolution measurements, but could also yield a substantial
improvement in the sensitivity of dual-beam GC-FT-IR measurements
by improving the reproducibility of the interferograms„ Again,
because of budgetary limitations we were not able to update our
spectrometer and consequently the detection limits that could be
attained ultimately can only be estimated. An improvement of
at least a factor of four would have to be achieved since merely
improving the duty cycle efficiency would double the B/N attainable
in a given time. The remaining factor of two improvement is very
conservative in light of the data in Figure 15.
c. It is also possible to improve the chromatography and decrease
the detection limits of GC-FT-IR from our early results
summarized in Table 1. We have modified our chromatograph for
capillary columns and the results are summarized below.
GC-FT-IR WITH CAPILLARY COLUMNS
We have shown (l) that decreased detection limits for GC-FT-IR should
be achieved by switching from packed columns to capillary columns. The most
popular type of capillary column, the wall-coated open-tubular (WCOT) column,
does not have a high enough sample capacity to be useful for GC-FT-IR
measurements. However support-coated open-tubular (SCOT) columns, while .not
achieving quite the resolution of WCOT columns, have a substantially greater
sample capacity, and we have developed a GC-FT-IR system which is specifically
designed for identifying samples separated on SCOT columns. The injection
port, splitter assembly and transfer lines, and light-pipes were all
modified for this work.
Injection Fort
The on-column injector previously installed on the Mo'del 3920 gas
chromatograph (Perkin-Elmer Corporation, Norwalk, CT) which was previously
used with the packed columns was replaced by a "splitless injector"
(Scientific Glass Engineering, Dallas, TX). This device works on the same
principles as the original device developed by Grob and Grob (l9-2l), and
differs radically from conventional injectors used for capillary column
chromatography. On conventional injectors, a splitter typically allows ifo
33
-------
of the injected sample to pass onto the column; for a 1 |il injection,
therefore, only 10 nl passes onto the column to avoid overloading by any
component (especially the solvent). With a splitless injector, the whole
sample (again typically 1 p.l) is injected onto the column and can be treated
in several ways:
a. If all the peaks are well separated from the solvent peak, elution
can occur without interruption.
b. If there are several peaks which elute close to the solvent peak,
the injected sample is subjected to the flow of carrier gas for
about 20 seconds, to ensure that all the sample is on the column.
A valve is then opened for about UO seconds, and the back pressure
of carrier gas in the column forces the more volatile components
(especially the solvent) back down the column while the less
volatile components are retained. After the solvent has been
largely eliminated in this way, but before the other components
have a chance to be lost, the valve is closed and the elution
proceeds. Not only is the solvent interference reduced in this
way, but the capacity of the column for samples of only slightly
greater volatility than the solvent is increased. In addition,
while the valve is open all the components tend to be reconcentrated
4000
3000
2000
-I
IOOO
FREQUENCY (CM ')
Figure 15-
Improvement in S/N of a new FT-IR spectrometer over the
instrument used in this study.
34
-------
near the head of the col-umn (19-21) and the resolution of the
chromatogram is increased slightly in a way which is analogous to
the first step of programmed multiple development thin-layer
chromatography, which is described in Section 8.
c. If there is appreciable bleeding from the septum, the valve need
not be fully closed and a small fraction of the carrier gas is
not passed through the column but rather is vented to the atmosphere.
This proceedure extends the useful life of SCOT columns considerably
and we use it all the time (after the process described in b.
above, if necessary).
Columns
Both stainless steel- and glass-walled SCOT columns were used in this
work; all columns were 50 m in length. The glass columns gave the most
symmetrical peaks but were far more difficult to interchange on our
chromat ograph.
Detector and Splitter Assembly
In our earlier work with packed columns, the effluent from the column
was first passed through the thermal conductivity detector (TCD) which was
originally installed on our chromatograph. The effluent was then split, with
approximately 10$> of the gas passing to a flame ionization detector (FH))
(Gow-Mac Instrument Company, Madison, N.J.) and tyCtfo passing to the light-pipe.
The GC resolution was almost certainly degraded by the rather large dead
volume of the TCD. In addition, the performance of the FID was degraded by
the rather low flow rate of carrier gas, which was only 10$ of the flow rate
through the column. The flow rate through SCOT columns is typically about 2ml,
and we were unable to achieve good response from the FID when SCOT columns
were directly substituted for packed columns.
The problems were overcome first by decreasing the dead volume intro-
duced by the TCD by completely removing the TCD from the chromatograph and
installing the effluent splitter close to the end of the column in a well
insulated part of the chromatograph. The FID was then mounted close to the
effluent splitter in the same part of the chromatograph. A T-junction
was installed between the splitter and the FID into which make-up helium
could be passed to improve the performance of the detector. Increasing the
flow rate of carrier gas through the detector to.20 ml/minute was beneficial
when packed columns were used for the chromatography and absolutely essential
for SCOT columns.
Transfer Line
The dead volume between the splitter and the light-pipe was decreased by
substituting the 1/8-in. o.d. tubing used with packed columns by l/l6-in. tubing.
-------
Light- Pipes
The Worcon light-pipes used in our earlier work were too large for use
with SCOT columns. We therefore prepared our own light-pipes (optimized
for SCOT columns) using the method of Azarraga (22). Both glass and glass-
lined stainless steel tubing (Scientific Glass Engineering, Dallas, TX) were
coated by first scrupulously cleaning them and then coating them with a
layer of "Liquid Bright Gold" paint (Engelhard Industries, East Newark, W.J.).
After the paint, which is a gold complex, was deposited as a uniform layer,
the complex was decomposed by heating the tube in an atmosphere of air,
leaving a uniform layer of gold on the interior of the tube. ¥e first used a
design for the light-pipe also sent to us by Azarraga (22) but found the flow
characteristics of this light-pipe caused excessive tailing of sample peaks.
We therefore developed a new design in which the gas from the chromatograph
was fed directly into the light-pipe by means of a l/l6-in. o.d. tube. The end
of this tube was manually filed down until the wall thickness was very small.
A groove was then cut into the ends of the light-pipe and the filed end of
the tubing was sealed in place. A NaCl window was sealed onto the end of the
light-pipe with the same adhesive. Only glass tubes could be prepared in
this way, since we could find no way of preventing the thin glass layer of
glass-lined stainless steel tubing from cracking.
Preliminary results with this system have, frankly, been disappointing.
The problems have largely been associated with the chromatograph rather than
with the -spectrometer. If a flame ionization detector is used.to measure the
chromatogram, an effluent splitter must be used to allow most of the effluent
to pass through the light-pipe rather than the destructive GC detector. With
the 10:1 effluent splitter we were using, the chromatograms measured from SCOT
columns were very unreproducible, and we were never certain how much, if any,
samples was passing through the light-pipe. We believe that if the FID is
replaced by a non-destructive photoionization detector, thereby eliminating
the need for an effluent splitter, more reproducible results will be obtained.
We are currently in the process of obtaining such a detector, and we confidently
predict that when all the operating parameters are optimized, detection limits
for GC-FT-IR using SCOT columns will be reduced below 10 ng.
36
-------
SECTION 7
LIQUID CHROMTOGRAHff AMD FT-IE SPECTROSCOPE
USE OF FLOW-THROUGH CELLS
The problems associated with the on-line measurement of the infrared
spectra of samples as they elute from a liquid chromatograph (LC-IR) are
even more severe than those associated with GC-IR. The principal difference
between GC-IR and LC-IR lies in the fact that for GC-IR the mobile phase is
infrared transparent, so that light-pipes of any pathlength can be used
without interference through absorption by the carrier gas, whereas for LC-IR
all solvents used as the mobile phase absorb strongly at certain wavenumbers
and less strongly at all others. In particular, for the polar solvents used
in reverse-phase high performance liquid chromatography (HPLC) absorption
by the mobile phase is sufficient to eliminate wide regions of the spectrum
for study, even for pathlengths of 100 p,m or less«
In single-beam GC-FT-IR it is possible to reduce the background signal
at all wavenumbers using a long, narrow light-pipe with substantial reflection
losses, and to maintain a high S/N by using an MCT detector. The loss in
signal caused by reflection losses is approximately uniform across the
spectrum, and the response profile of an MCT detector in a rapid-scanning
FT-IR spectrometer above 800 cm"1 is not significantly different from that of
a TGS detector. Thus by suitably designing the dimensions of the light-pipe
so that detector noise exceeds digitization noise, substantial gain can be
made in single-beam GC-FT-IR when an MCT detector is used.
On the other hand, in a single-beam LC-IR experiment, if the pathlength
of the cell is increased, the transmittance at regions where there are no
absorption bands remains very high, but in the regions of strong solvent bands
the transmittance may change drastically. -For example, if a band has a peak
transmittance of 10/o for a certain cell thickness, it will have a transmittance
of only ifo if the pathlength is doubled. If a weak solute band absorbs at
the same wavenumber, doubling the pathlength will double the absorbance, and
to a good approximation the absorptance, of the band so that the S/N of that
band would be reduced by a factor of five on doubling the pathlength of the
cell. It can be readily shown that the optimum value for the S/N of solute
bands is achieved when the solvent transmittance is 1/e or 36.8/0. Of course
no solvent has a uniform transmittance, and the decision has to be made as to
the best pathlength to permit a "reasonably large" proportion of the spectrum
to be observed with an acceptable S/N after solvent compensation.
-------
The range of acceptable solvent transmittance which will permit solute
spectra to be measured at an acceptable S/N may be estimated by calculating
the product of the solute absorbance and solvent transmittance and plotting
the result against solvent transmittance, see Figure l6. This graph shows
that the optimum sensitivity is found when the solvent transmittance is
36.8/0, as expected, and that spectra may be measured within 79/° of the
optimum sensitivity if the solvent transmittance is between 65/0 and l4fo. The
corresponding 50$> points are between 79^ an-d Tfo solvent transmittance„ Thus
the pathlength of the cell should be chosen so that the transmittance of the
solvent in the fingerprint region of the spectrum is between about 80$> and
lO/o. If the transmittance falls below about 2.5/o, i.e. in the region of strong
solvent bands, essentially all solute information is lost.
For many organic solvents (with the exception of those containing a
nitrile group), the transmittance between 2700 and 1800 cm"1 exceeds 90$ when
the transmittance in the fingerprint region is optimized, but since so few
solutes have interesting absorption bands in this region, this high trans-
mittance cannot be used to increase analytical sensitivity (again, with the
exception of nitriles). The large amount of transmitted energy in this region
has the disadvantage of increasing the dynamic range problem for single-beam
LC-FT-IR measurements made with an MCT detector. If the optimum pathlength
for flow-through cells is achieved for measurements made in the fingerprint
region of the spectrum, it is apparent that high sensitivity LC-FT-IR must
use dual-beam techniques to attain the lowest possible detection limits.
For many solvents used in HPLC, a pathlength of approximately 100 \j.m
appears to be optimum, with only a few spectral regions where the transmittance
is less than 10$. However even for bands of intermediate strength, 10$> < T <
90/o, the problem of accurate solvent compensation to leave a flat baseline
for the solute spectrum may be quite severe. It should also be noted that the
absorbance of solutes eluting from an HPLC is going to be very small if a
100 urn cell is used, e>g. the absorbance of the strongest bands in a 1$> solution
of the pesticide DDD in hexane at this pathlength is about 0.1. Even though
a peak concentration of 10 |j.g/ml of solutes eluting from an HPLC would be
considered quite typical by most chromatographers, the maximum absorbance of
DDD measured under these conditions would be 10~4, or T = 99-98f0- This is a
rather low concentration to give detectable bands by single-beam FT-IE and if
the bands absorb close to solvent bands absorbing more than 70/o detectable
bands could not be measured in less than one minute's data acquisition time.
The need for dual-beam FT-IR to increase sensitivity again becomes apparent.
The optical arrangement we used for dual-beam LC-FT-IR is the same as
that shown in Figure 6, with a flow-through cell in the transmitted beam and
a variable pathlength reference cell in the reflected beam. The flow-through
cell was k mm in diameter and 93 um in pathle-ngth, so that its volume was
1.2 |_il. Since the half-width of HPLC peaks separated using columns packed
with 10 urn diameter particles is typically 500 (j.1-1 ml, it can be shown that
less than l
-------
100
= 36.8%
100
75 50 25
% TRANSMITTANCE OF SOLVENT
Figure 16. Ratio of the experimental to the optimal S/K for
weak absorption bands as a function of the trans-
mittance of the solvent. The sensitivity at any
given wavenumber may be increased by varying the
pathlength of the cell to yield a transmittanee
of 36.8$ at that wavenumber.
39
-------
length of the cell would increase the efficiency in regions where the solvent
absorbs weakly, but would increase the number and extent of the regions which
would be "blacked out" by strong solvent bands and also seriously increase
the problems of solvent compensation.
As might be expected the detection limits of on-line LC-FT-IR measure-
ments made using these optics were not spectacularly low. The best results
were found with strongly polar, low molecular weight solutes such as anisole.
For such samples, 100 ng of sample injected onto the chromatograph would
yield a readily identifiable spectrum, see Fig_ure IT, and detection limits of
1800
CM
-1
800
Figure 17.
On-line LC-FT-IR spectrum of 100 |ag of anisole, a strong
absorber, injected into the chromatograph, eluted with
hexane, and measured by dual-beam FT-IR spectrometry.
40
-------
approximately 10 (ig were obtained. However for less polar, more weakly
absorbing materials, such as the chlorinated pesticides, detection limits
about an order of magnitude greater than this were found, see Figure 18. To
what extent the rather poor solvent compensation evident in this spectrum is
caused by the instability of the interferometer cannot be judged. If this
effect is the principal cause of baseline variation, it is quite possible
that detection limits could be reduced considerably.
REFERENCE
ON-LINE
1700
CM
700
Figure 18. On-line LC-FT-IR spectrum of 150 jag of TDE, a weak absorber,
eluted with hexane, and measured by dual-beam FT-IR spec-
trometry. A reference spectrum of TDE is shown above.
41
-------
It is, of course, possible to increase the amount of solute present in
the flow-through cell by increasing the volume of sample injected into the
chromatograph. However we found that increasing the quantity of what is
already enough solute to exceed the linear capacity of the column degraded
the chromatogra.phic resolution. Only for mixtures with very few components was
increasing the volume injected found to be beneficial.
A further disadvantage of on-line LC-FT-IR measurements of the type
described above is the fact that they can only be performed for isocratic
separations. If gradient elution is required to improve the quality of the
chromatogram, the problems' of solvent compensation are so great that this
method of LC-FT-IR becomes essentially useless.
All the above problems stem from the absorption of the incident radiation
by the mobile phase. We have therefore studied various ways for performing
continuous or semi-continuous LC-FT-IR measurements after removing the solvent
in some fashion.
SOLVENT ELIMINATION TECHNIQUES
The first technique we studied was to evaporate the solution onto a
heated moving metal ribbon so that the solvent evaporated and the solute was
deposited on the ribbon. After removal of the solvent and the section of the
ribbon on which the solute was deposited has moved into the infrared beam, the
reflection-absorption spectrum of the solute could be measured. The problem
with this idea is that, if the flow rate through the chromatograph is greater
than about 500 [il/min, the area over which the solute is deposited is much
greater than 1 cm , whereas the maximum sensitivity is found if the area is
at least ten times smaller than this. We found that the best way to obtain
the smallest deposition area was to spray the solution onto the ribbon so
that much of the evaporation could take place before the solution came in
contact with the ribbon. However even with this technique, the area over
which the solute was deposited was so large that the detection limits were
merely comparable to those found for the on-line experiments made with a
flow-through cell.
Another method of obtaining low detection limits for solutes is to
evaporate the solution on the surface of a micro attenuated total reflection
(ja-ATR) crystal. By using small (0.5 or 1 mm thick) crystals of less than
1 cm2 surface area, detection limits of less than 100 ng have been reported
for off-line experiments using grating spectrophotometers (23). The usual
procedure in measurements of this type is to take a rather large volume of
the solution and evaporate off all but a few microliters of solvent by blowing
air through the warm solution. The remaining concentrated solution is applied
to the surface of the (i-ATR plate using a microsyringe and the remaining
solvent is evaporated off before measuring the spectrum of the deposited
solute. This two step procedure cannot be applied for semicontinuous
LC-FT-IR, and the evaporation of solvent at rates around 1 ml/min onto a
surface area less than 1 cm2 again proved very difficult. Although spraying
again improved the efficiency of the deposition, we were never able to
achieve the desired rate of deposition of 1 ml/min, and this approach was
also ultimately abandoned.
42
-------
It is apparent that for semi continuous LC-FT-IR, the requirements are
quite difficult to achieve. To achieve the maximum FT-IE sensitivity the
beam diameter for the measurement should be small (typically 2-*J- mm) but the
rate of evaporation of solvent should be fast, which implies that the solute
will be deposited over a large surface area. In addition the spectrum should
be measurable a short time after the peak has eluted from the chromatograph,
preferably using multiple passes of the beam to enhance sensitivity. One
technique in which these criteria appeared to be obeyed involves spraying
the solution into a heated vertical light-pipe. For a 3 ^m diameter light-
pipe of length 28 cm the interior surface area on which the sample Is
deposited is 26.4 cm2, while the area of the aperture is only 0.07 cm2. We
proposed that by suitably controlling the temperature of the light-pipe it
would be possible to deposit all the solute on the interior of the light-pipe
while evaporating the solvent and flushing it out of the tube by the spray
propellant gas. A beam entering the light-pipe would be multiply reflected
as it passed down the tube that the reflection absorption spectrum of the
sample would be enhanced.
We believed that it should be possible to construct a device in which
four light-pipes are placed in a carousel holder. Into one of these the solute
from the first HELC peak would be deposited. After this peak has eluted and
been deposited, the carousel would be rotated through 90°, after which the
second HELC peak would be deposited while the interferogram of the first peak
is measured. After the second peak has been deposited the carousel would
again be rotated through 90°. The third HPLC peak would then be deposited,
the second peak would be measured and the first peak is washed from the tube.
This process would be continued, with the fourth position on the carousel
being used for drying and heating the tubes in preparation for the following
deposition step. The complete system would be controlled by monitoring the
signal from the HPLC detector using a second derivative trigger of the type
described in Section k-. The circuit we used for this purpose was quite func-
tional; however ultimately we planned on replacing it with a microcomputer
used in the same fashion as in our later work with diffuse reflectance
spectrometry.
The preliminary experiments performed to test the feasibility of this
device were performed using solutions of a single colored component at low
concentration (0.1-100 ppm) flowing into the light-pipe at rates around
1 ml/min; the flow was generated by a peristaltic pump. The first results
from this system were marred when it was found that the principal sample
which was deposited on the light-pipe was the phthalate ester plasticizer
from the tubing of the peristaltic pump, although it should be noted that
UV-visible, spectrophotometry showed that the sample is deposited with high
efficiency at the same time as the plasticizer.
After changing all tubing to Teflon, we were surprized to find that it
was not possible to obtain identifiable*spectra from sample quantities smaller
than 100 jig even though all the sample was shown spectrophotometrically to be
present in the light-pipe. The reason for this low sensitivity became apparent
at the end of a series of experiments designed to determine the distribution of
the deposited sample down the light-pipe. By depositing a colored compound,
43
-------
methylene blue, it was possible to determine the distribution as a function of
spray propellant gas flow rate and the light-pipe temperature (or, more
specifically, the voltage applied to the nichrome wire heater coil wrapped
around the light-pipe). The quantity of methylene blue deposited at intervals
of 1 cm down the light-pipe was determined by dipping the first centimeter of
light-pipe into a known volume of methanol to dissolve all the dye in this
interval, and determining the concentration spectrophotometrically. This
process was repeated at 1 cm intervals up the tube. The variation of solute
quantity down the light-pipe is illustrated in Figure 19» The optimum results
should be found when the sample is most uniformly distributed down the light-
pipe, but even with a flow rate of 4 cfm of nitrogen and a Variac setting of
k, it was still found not to be possible to measure identifiable spectra from
less than 100|_ig of deposited sample,,
Several different attempts were made to improve the sensitivity of this
technique. Initially J mm diameter light-pipes were used whose transmittance
was so high that a screen of 12$ transmittance was placed in the beam to avoid
digitization noise. When 200 |ag of methylene blue was deposited and measured
in this way, the maximum absorptance was only about 10$, see Figure 20.a. To
increase the number of reflections down the light-pipe, the screen was removed
and replaced by an "antiaperture" to remove the central 90$ of the area of the
collimated beam emerging from the interferometer and leave only the outer ring.
Surprisingly no increase in the absorption band intensity was noted, see Figure
20.bo
Finally the optics were changed so that light-pipes of smaller diameter
could be studied. The 3 inch focal-length off-axis paraboloid used to focus
the collimated beam onto the 3 nnn diameter light-pipes was replaced by a
1.5 inch focal-length off-axis paraboloidal mirror. The reduction in tiie size
of the focus enabled 1.5 mm i.d. light-pipes to be used without significant
vignetting losses. In view of the reduced diameter of the light-pipe and the
increased beam half-angle an increase in the absorbance of the bands should
have been noted, but again, in practice, no improvement over the earlier measure-
ments, was noted, see Figure 20.c.
In each of the spectra shown in Figure 20, it can be seen that several
of the bands appear to be tfbottoming out" even though the apparent maximum
absorptance is less than 10$. One possible reason for this effect is that
the solute is not deposited as a uniform layer over the inside of the tube,
but rather forms very narrow streaks of material at high concentration when
the solvent evaporates, presumably due to the surface tension of the solvent.
(For the solvent with the highest surface tension, the droplets could never
be totally evaporated in the light-pipe and no solute was deposited). If
only a very small fraction of the surface of the light-pipe has sample present,
most of the radiation reflected down the tube would not interact with the
sample. Confirmatory evidence for this hypothesis was obtained by using an
uncoated glass tube, rather than a gold-coated light-pipe, for the deposition
step. In this case very narrow streaks of sample could indeed be seen on the
interior surface after spraying. We could find no way of improving the
deposition characteristics of this method and we have reluctantly abandoned
what initially appeared to be a very promising technique.
44
-------
2
g
UJ
o
o
UJ
Si
UJ
o:
D
I 23456789
FRACTION COLLECTED
Figure 19 A. Distribution of deposited solute down light-pipe with
a nitrogen flow-rate of 2 cfm for the following voltages
across the heating coil: A, IJv; B, 11v; C, 9v; D, 6v.
(A,B,C and D corresponded to Variac settings of 12, 10,
8 and 6, respectively).
-------
o
!5
LJ
o
z
o
o
LJ
LJ
CC
B
1 23456789
FRACTION COLLECTED
Figure 19B. Distribution of deposited solute down light-pipe with a
voltage of llv across the heating coil for the following
nitrogen flow rates: A, 6 cfm; B, 5 cfm; C, k cfm;
D, 3 cfm.
46
-------
10% T
3000
Figure 20.
2000
1000
-I
FREQUENCY (CM ')
Spectra of 200 jig of methylene blue deposited on light-pipe,
measured under the following conditions:
A. 3 wra. light-pipe, 12$ T screen in beam;
B. 3 mm light-pipe, central $0% of beam blocked;
C. 1. 5 mm light-pipe, no further screening.
47
-------
Some recent developments in this laboratory have finally led to a method
which, at the point of writing this report, still appears to be a potentially
powerful method for semi-continuous LC-FT-IR with solvent elimination. The
method is based on diffuse reflectance infrared Fourier transform (DRIFT)
spectroscopy, a technique which was being developed in our laboratory for a
purpose completely unrelated to this project. Our system is a hemiellipsoidal
diffuse reflectometer, which functions because a beam of light emanating from
one focus of an ellipsoid is reflected from the surface of the ellipsoid
through the second focus. A beam of radiation is passed into the ellipsoid
through a hole at the surface on the major axis and focused onto a diffuse
reflector held at the nearest focus to the hole. The diffusely reflected
radiation is collected by the ellipsoidal mirror and focused at the second
focus of the ellipsoid. If the input beam is modulated by a rapid-scanning
interferometer and a detector is held at the second focus of the ellipsoid,
the diffuse reflectance spectrum of the sample held at the first focus could
be measured. An optical schematic of our DRIFT apparatus is shqwn in Figure
21. In practice the magnification of the beam at the second focus is quite
large and we use an off-axis segment of another paraboloid placed before this
focus to reduce the size of the image to 2mm, the same size as our TGS and
MCT detectors (24,25).
It has been found that finely powdered samples of alkali halides have a
diffuse reflectance of nearly 100$ (26). We found that if small quantities
of strongly infrared absorbing compounds were mixed with powdered alkali halides,
Collimoted
Beam from
Interferometer
Figure 21. Optical diagram for hemiellipsoidal diffuse reflectometer.
48
-------
and an MCT detector was used for the measurement, detection limits of a few
nanograms were attained. We surmized that if a small quantity of powdered
KC1 were held on a fritted sample cup held at the first focus of the diffuse
reflectance apparatus shown in Figure 21, it might be possible to evaporate
the solvent from the HPLC effluent and deposit the solute on the surface of
the KC1, This powder has a rather large total surface area, but can be held
in a small (3mm diameter) cup; hence the sampling requirement for LC-FT-IR
discussed above appears to be fulfilled. By holding several such cups in a
carousel arrangement, a semi-continuous and completely automated LC-FT-IR
system has been designed, and a"b readboard" system showing submicrogram
detection limits has been constructed.
At least 30 sample cups are held around the circumference of the carousel
at equal intervals. Each cup is about 3 mm deep and 5 mm in diameter, and contains
about 100 mg of finely powdered KC1. The cup is lined with a fine mesh, so that
when a low vacuum is applied to the bottom of the cup through a flexible hose,
air is drawn through the KC1 powder, causing any volatile solvent also present
in the cup to be removed by evaporation. Holes were drilled through the wheel
along the line joining the center of each cup to the center of the wheel, each
hole being equidistant from the center. A small light bulb is held above the
carousel over one of the holes and a photodiode is positioned immediately below
the same hole. The wheel may be rotated using a belt driven by a motor, and its
position at any time is sensed by counting the number of holes which pass the
photodiode after its initial position. Three cup positions are used for sample
deposition, final solvent elimination and spectral measurement, respectively,
as shown in Figure 22 for a sixteen-cup carousel, and the whole device is con-
trolled by a microcomputer.
The effluent from the liquid c"hromatograph is sprayed continuously into a
heated concentrator tube, as shown in Figure 2J. The concentrator tube is a
thin glass tube wrapped with nichrome heating wire in such a fashion that the
number of turns of the wire is greater around the top of the concentrator tube
than near the bottom, so that the heat input to the spray is greater near the
top of the tube where most of the solvent evaporates than near the bottom of
the tube. The temperature of the interior, of the tube is controlled by adjusting
the voltage across the nichrome wire using a Variac so that about 90% of the
solvent evaporates in the tube. If the chromatographic flow rate is 1 ml/min,
the flow rate of the droplets emerging from the bottom of the tube is only 100
|il/min. The typical volume of solvent between the half-width points of an
HPLC peak is about 30 seconds, so that if only this fraction of each peak is
dropped onto the KC1 sample cup, only 50 (il of s.olvent must be evaporated for
complete elimination.
Initially a three-way solenoid valve was used to direct the effluent from
the HPLC either to the concentrator tube or to waste, depending on whether or
not a peak .was eluting from the chromatograph. However with this arrangement
the temperature of the concentrator tube would increase too much in the regions
between eluting peaks, and a steady state condition was preferable. To achieve
this end, the effluent was continuously sprayed into the concentrator tube, and
a small capillary tube which is attached to an aspirator is placed at its tip
so that all the liquid emerging is drawn down the capillary. A normally-open
solenoid valve is held between the aspirator and the capillary, and when this
49
-------
is in the closed position the emerging liquid is allowed to drop onto the KC1
powder in the cup below. The solenoid is actuated by a trigger actuated from
the HPLC detector after a preset delay to allow the peak to travel from the
detector to the tip of the concentrator tube.
When the first peak eluting from the chromatograph is sensed, it is assigned
a label by the microcomputer as are all peaks which elute subsequently. The
status of each peak is listed in a table and continuously monitored. After the
first peak has eluted from the concentrator tube, the solenoid valve is reopened
to prevent any more solvent depositing on the KC1, and the wheel is rotated so
that the cup containing this peak (cup A) reaches the solvent elimination position.
At this point a stream of air is drawn through the cup so that all the solvent is
evaporated. When the second peak reaches the tip of the concentrator tube, the
solenoid is closed so that this peak is deposited onto the cup which is now at
that position (cup B). At this time, we are allowing air to be drawn through
cup A from the moment it reaches the solvent elimination position to the time
FINAL
ELIMINATION
OF SOLVENT
~ ~ ~ , SAMPLE
IR MEASUREMENT > ° ° ( DEPOSITION
POSITION COUNTER
ABOVE: LIGHT
BELOW: PHOTODIODE
Figure 22. l6-cup carousel for LC-IR measurements by DRIFT spectrometry.
50
-------
the next peak has been deposited into cup B. However we believe that it will
be possible to reduce the time during which air is being passed through the cup
in the solvent elimination stage in order to reduce solute loss by evaporation
while still totally eliminating the solvent. Further work to optimize this
stage will be performed. After the second peak has eluted, the solenoid valve
is reopened to direct the flow of liquid emerging from the concentrator tube
back to the aspirator and the carousel is again rotated. Cup A is then in the
spectral measurement position, cup B is in the solvent elimination position, and
the third peak is deposited in cup C.
The diffuse reflectance spectrum of the sample in cup A is measured from
the time the carousel stops (if necessary after a short delay to allow all
vibrations to die out) until the third peak is deposited in cup C. The micro-
FROM
AIR HPLC
6
VARIAC
O
ASPIRATOR
MICROCOMPUTER-
CONTROLLED
SOLENOID VALVE
KCI Powder
FINE MESH
Figure 27). Concentration and solution deposition stage for LC-IR
measurements by DRIFT spectrometry.
51
-------
computer then waits for the scan currently in progress to finish and again
actuates rotation of the carousel. This process is repeated each time a new
peak is sensed at the HPLC detector, and is completely automatic. Two alterna-
tive methods are- available for rotating the carousel to allow the cups containing
the last two peaks to reach the spectral measurement position. The total number
of peaks known to be present in the chromatogram can be preprogrammed into the
microcomputer and the spectra of the samples in the last two cups can each be
measured for a certain time, say two minutes. Alternatively a character can be
entered into the keyboard of the microcomputer indicating that the last HPLC
peak has eluted; in this case the microcomputer will control the carousel
rotation and data acquisition in an exactly analogous way to the first method.
The first method has the advantage of requiring absolutely no interaction but
requires a chromatogram of that sample to have been run previously so that the
operator knows the number of peaks. The second method does not require prior
knowledge of the chromatogramj however the operator must be able to use some
intuition to estimate when the last significant peak in the chromatogram has
eluted.
The first "breadboard" system to be constructed certainly demonstrated the
feasibility of achieving submicrogram detection limits. The sensitivity of
this system was demonstrated using a 1 jil injection of Stahl1 s Test Dye Solution
(Malinckrodt Chemical Co., St. Louis, MO), which is a solution of 0.1$ each of
three dyes, Butter Yellow, Indophenol Blue and Sudan Red G. The chromatograph
used for this work was made by Tracer Instruments, Inc. (Dallas, TX) , and was
equipped with a 254 nm ultraviolet detector and a 25 cm long x 4 mm diameter
column packed with 10 (am particle size Partisil, which is a silica adsorbent.
The separation involved a normal-phase mechanism with a 99:1 hexane-propenol
mixture as the mobile phase. Each component of the injected dye mixture was
present at a total quantity of 1 jig, and the diffuse reflectance spectra of the
three dyes, which are shown in Figures 24-26, show a S/N of better than 100.
These spectra were measured using a completely unpurged spectrometer, and the
effects of uncompensated atmospheric water vapor can be seen in these spectra
between 1900 and 1700 cm -1. The "derivative-like" shape of these water lines
is caused by a slight difference in the alignment of the cups containing the
samples and the cup containing the reference KC1. Misalignment of the beam in
FT- IE spectrometry produces a wavelength shift and it became apparent that the
tolerances to which our first carousel was made were not good enough for this
work. A second carousel, made to far more exacting tolerances, is now being
constructed. A cover to allow water vapor to be purged from the optical train
is also being constructed. When this system is completed, we expect that
detection limits of less than 10 ng of sample injected into the chromatograph
will be attainable for nonvolatile solutes.
An interesting property of diffuse reflectance spectroscopy is beneficial
for infrared microsampling in general, and for LC-IR in particular. The S/N
is given by the ratio of (1-R ) for any absorption band to the noise at the
same wavenumber, where R is the reflectance of the sample at "infinite depth".
(¥e have shown (24) that°°the reflectance changes very little on increasing the
depth beyond 3 mm). According to the Kubelka-Munk equation (27) linking R to
the concentration, c, of -sample in a nonabsorbing matrix such as KC1: °°
52
-------
800
On-line LC-IR spectrum of 1 ^g of Butter Yellow dye,
measured after automatic solvent elimination.
53
-------
3200
Figure 25.
800
On-line LC-IR spectrum of 1 |j,g of Indophenol Blue dye,
measured after automatic solvent elimination.
54
-------
800
On-line LC-IR spectrum of 1 |j.g of Sudan Red G dye, measured
after automatic solvent elimination.
55
-------
where k is the absorption index, which is directly proportional to the absorptivity.
When c is small, R^ becomes approximately unity, so that the denominator of this
expression becomes constant, and the signal, (l-R^,) , becomes proportional to the
square root of the concentration. In almost every other spectroscopic technique,
the S/N becomes linearly proportional to the sample concentration near the detec-
tion limit, and this property should give DRIFT a real advantage in infrared
ultramicrosampling compared with other more conventional sampling techniques.
Although this LC-FT-IR method is obviously at a very early stage in its
development it shows real promise, and even at this stage shows detection limits
two orders of magnitude below methods in which the solvent is not eliminated and
the HPLC effluent is passed directly through a flow-through cell. This sensitivity
improvement is measured in regions where the solvent is transparent, and substan-
tially greater improvements are found in the regions of strong solvent absorption.
The method is of course, applicable even for gradient elution HPLC, thereby
obviating the need for the extremely difficult solvent compensation routines
needed when flow-through cells are used in this case. Another advantage of the
technique is that each peak is captured, and if the S/N is insufficient to allow
unequivocal identification, the cup may be readily returned to the spectral mea-
surement position and the spectrum may be measured using more extensive signal-
averaging than is possible in this automated measurement. As yet we have been
unable to totally eliminate aqueous solvents, largely because of the high surface
tension and latent heat of vaporization of water. ¥e believe that there are
several other ways of approaching this problem which have not yet been tested,
and we are not rejecting the idea that aqueous mobile phases cannot be separated
from the solute. However this type of chromatography is equally difficult to
investigate by conventional LC-FT-IR techniques because of the extremely high
absorptivity of water all through the spectrum. _ Overall the LC-FT-IR method
which we have developed appears to be a promising, and perhaps even a very
powerful, method for the on-line measurement of the infrared spectra of peaks
eluting from high performance liquid chromatographs.
56
-------
SECTION 8
THIN-LAYER CHROMATOGRA Hff AND FT- IE SJECTROSCOFT
TRANSMITTANCE MEASUREMENTS
We have previously developed a method by which the infrared spectra of
species separated by thin- layer chromatography (TLC) can be measured without
removing the sample from the adsorbent (28,29). In this technique the
adsorbent layer (silica gel or alumina) is deposited on an insoluble, chemically
inert, infrared transmitting plate (AgCl) so that after the chromatogram has
been developed, the absorption spectrum of each separated sample "spot" can
be measured by direct transmission. The plate may be treated with a small
quantity of Fluorolube (whose refractive index closely matches that of the
silica or alumina) prior to the infrared measurement to reduce the effect of
scattering at high wavenumber (29). The sensitivity of this technique, which
we abbreviate as TLC-IR, is determined in part by the spectrometer and in
part by the chromatography.
The sample absorbance of TLC-IR spectra depends on the number of analyte
molecules per unit area of surface. Until the spot size becomes smaller than
the area of the infrared beam at its focus (typically Jrm) , the smaller the
diameter of the spot can be made, the higher is the sample concentration and
the higher is the infrared absorbance. One of the limitations of conventional
TLC is that spots usually spread to a greater diameter than 3 mm during
development, thereby reducing the potential sensitivity of TLC-IR measurements.
The use of programmed multiple development (PMD) in TLC decreases the spot size
and increases chromatographic resolution (30-J2).
HO is a form of TLC in which the chromatography is performed using
several developments, each one allowing the solvent front to advance further
than the preceding one. A controlled solvent removal step, either using
radiant heat or by sweeping the plate with an inert gas, follows each solvent
advance returning the solvent to the spot origin. Each cycle consists of a
solvent advance (development) and a solvent removal (drying) step, with the
development time during the n th cycle, t^, depending on n and the advance unit
time, t . With our equipment (Regis Chemical Company, Morton Grove, IL),
any of the following modes may be selected:
Mode 1: t = n t
n a
n
Mode 2: t =
-------
Mode 3: t = n2 t
n a
where ta ranges from 10 to 100 seconds. Mode 1 yields the most cycles for a
given overall program time and best resolves spots of low R^ value. Mode 3
most closely resembles conventional TLC, yields the fewest cycles for a given
overall program time, and best resolves spots of high Rf value. Mode 2 is
intermediate between Modes 1 and 3<>
PMD has three major advantages over conventional TLC for TLC-IR measure-
ments. The first is the increased chromatographic resolution as mentioned
above. In addition, the original solution may be applied to the plate in
greater volume and over a larger area than for conventional TLC before the
spot size starts to increase. Finally the adsorbent layer may be loaded with
a much greater quantity of sample than conventional TLC before the chromato-
graphic resolution becomes degraded. Since the thickness of the layer for
TLC-IR experiments is ideally between 50 and 100 (jon (to avoid excessively
strong absorption bands due to adsorbent) compared to the usual range of 100
to 250 |im for many commercially available TLC plates, the probability of
exceeding the adsorbent capacity for conventional TLC is increased, and the
use of HO should be beneficial when TLC-IR measurements are desired.
TLC plates for this work were prepared by depositing a layer of alumina
(mean grain size, 10 to 40 (j,m) on a silver chloride plate, as described
previously (28,29). After development by conventional TLC or PMD, the plate
was treated with Fluorolube and held at the 3 nm sample focus in the trans-
mitted beam of the dual-beam FT-IR system shown in Figure 6 with the reflected
beam blocked. This measurement is equivalent to holding the sample at the
focus of a ij-x beam condenser in the sample compartment of an unmodified
Digilab FTS-14 spectrometer, but allows larger plates to be studied without
vigenetting the beam.
Two types- of compound, dyes and chlorinated pesticides, were investigated.
The dyes were methylene blue and Stahl's solution (an equimolar solution of
butter yellow, Sudan red G, and indophenol blue). Visualization of the spots
of separated pesticides was achieved by exposure of the dried plate to iodine
vapor. All spectra were measured at 8 cm"1 resolution in double-precision,
using either the TGS or MCT detector.
To compare the sensitivity of TLC-IR experiments performed by conventional
TLC and PMD, two identical chromatopiates were prepared, spotted with 500 ng
of the pesticide, Aldrin, and developed with n-heptane. For the PMD separation,
Mode 1 was used with n = 15 and ta = 10s, giving a total time of 30 minutes,
compared with 15 minutes for conventional TLC. The area of the spot on the
PMD chromatogram was approximately five times less than that for the conven-
tional TLC plate. The measured spectra are shown in Figure 27, and the
improvement in sensitivity for the PMD run is apparent.
The advantage of improving the chromatography by PMD and increasing the
spectrometric sensitivity by substituting the TGS detector by an MCT detector
is illustrated in Figure 28, for butter yellow dye separated from Stahl's
solution. The increase in concentration gained through the use of PMD is
58
-------
B
1700
CM"1
1300
Figure 27. TLC-IR spectra of 500 ng of Aldrin measured with an MCT
detector using 10 scans:
A. by PMD, Mode 1, n=15, t =10 s, total development
time = y> min;
B. by conventional TLC, development time = 15 min.
59
-------
32OO
28OO
32OO
28OO
B
195O
Figure 28.
155O
195O
1550
CM
-1
Portions of TLC-IE spectra of Butter Yellow after separation
from the other component's of Stahl' s solution, using benzene
as eluent:
A. 100 ng of sample separated using PMD, Mode 3, n=6, t =10 s,
total development time = 45 min, measured with an a
MCT detector using 10 scans;
500 ng of sample separated by conventional' TLC, development
time = 15 minj measured with a TGS detector using 400 scans.
B.
60
-------
shown by the increased absorbance of the bands, while the advantage of using
the MCT detector is reflected in the reduced measurement time required to
achieve a given noise level.
It -is noteworthy that when separation is effected by FMD and an MCT
detector is used for the infrared measurement the total (active) scan time
required to attain submicrogram sensitivity is less than 5 seconds, even for
such weakly absorbing samples as the chlorinated pesticides. These results
suggest that it may be feasible to construct a device to measure TLC-IR
spectra automatically across an entire chromatoplate developed by PMD.
Spectra could be measured after moving the plate by intervals as small as 1 mm
in a total measurement time less than the time taken to develop the plate.
The only unfavorable factor for such a device would be the difference between
the shape of the infrared beam at its focus (circular) and the shape of spots
separated by PMD, which are elongated ellipses, typically approximately 1 by
6 mm. There are two ways to improve this situation. It might be possible to
use a toroidal mirror rotated 90° through its optical axis to deform the beam
to the desired shape as suggested by Low (33). It has also been shown that
it is possible to modify the PMD experiment to obtain circular spots Cjl).
If FMD is used for TLC-IR measurements, detection limits as low as 10 ,ng
can be achieved with extended signal-averaging, as shown in Figure 29. It
may be noted that since the diameter of the infrared beam is larger than the
minor diameter of the elliptical spot on the chromatoplate, a little over
half the beam misses the sample completely. Thus if better matching could be
achieved by either of the techniques discussed above, the detection limits
should be further reduced by about a factor of two»
DIFFUSE REFLECTANCE MEASUREMENTS
The only major disadvantage of the TLC-IR measurements described above
is that chromatopiates on AgCl substrates are not commercially available.
Commercial TLC plates are made on either glass, plastic or metal foil backings.
None of these materials is sufficiently infrared transmitting to be usable for
TLC-IR measurements by direct transmission. We therefore decided to test the
feasibility of diffuse reflectance spectroscopy for TLC-IR measurements. One
advantage of DRIFT for this application is that it is not necessary to treat
the plate with Nujol or Fluorolube after development to eliminate the effects
of scattering at high wavenumbers, since the technique only works for
scattering samples. On our particular system for DRIFT measurements (Figure
21), we can only measure samples of 4 mm diameter, so that TLC-IR spectra can-
not be obtained without removing the sample from the plate in some fashion.
We are able to punch out small samples from TLC plates on plastic or metal
foil backings and place the spots at the sample focus to investigate the
feasibility of using DRIFT for TLC-IR measurements.
For silica TLC plates on polyethylene terephthalate backings (Eastman
Chemical Company, Rochester, NY) we found that the penetration depth was
greater than the thickness of the silica layer (100 |im) so that the polymer
bands were very evident in the measured spectra. We therefore investigated
the feasibility of using aluminum foil backed silica TLC plates (Brinkmann
Instruments, Hauppage, NY). Several problems were encountered in this work.
61
-------
2000
CM"1
1200
Figure 29. TLC-IR spectra of methylene blue developed by PMD using
methanol as eluent, Mode 1, n=10, "ta=10 s, measured
using an MCT detector;
A. 100 ng of sample, 10 scans;
B. 10 ng of sample, 1000 scans.
62
-------
Firstly it was found that the diffuse reflectance from silica, and other
materials with bands of unusually high absorptivity, is strongly dependent on
the particle size (25). When the particle size is between about 100 jjm and 10
jjm, the spectrum appears to be a combination of the specular reflectance spectrum
and the diffuse reflectance spectrum, with the proportion of the diffuse reflectance
spectrum increasing as the particle size is decreased. For particles of approxi-
mately kO |jm diameter, the bands which would normally show as pure absorption
bands in diffuse reflectance spectra appear to insert near the band center because
of the specular reflectance component. The specular component of the beam does
not penetrate the sample at all so that even though there appears to be enough
energy across the entire spectrum that absorption bands of sorbed species should
be measurable at all wavenumbers, these bands are in fact masked near the maxima
of the adsorbent bands. Only by using an adsorbent with an average particle
size considerably less than 10 jjn could one expect to make the contribution of
the specular reflectance component negligibly small.
The second problem concerned the feasibility of adequately compensating the
bands due to water associated with the silica gel. The intensity of these bands
would change before and after chromatography, so that it was necessary to use a
region of the chromatoplate close to that of the spot being studied to obtain a
good matching of the intensity of the water bands and hence to obtain a good
baseline in the compensated spectrum. ¥e were never able to obtain a perfect
baseline in any TLC-IR measurement by DRIFT, although several spectra with less
than % deviations around the 100$ reflectance line were obtained.
Finally if an alcohol, such as methanol, was used to develop the chromatogram,
reaction between the silanol group on the silica gel and the methanol was shown
to occur:
= SiOH + CH3OH -> = SiO CH3 + H20
The extent of this reaction was dependent on the Rf value of the spot, so that
once again great care had to be taken in the compensation of the background
absorption.
In spite of all these problems, we were able to measure spectra of materials
sorbed on silica-gel TLC plates after elution with a variety of solvents with
submicrogram detection limits, see Figure JO. The difficulties involved with
background compensation make it appear that the development of an on-line scanner
for TLC plates based on diffuse reflectance spectrometry would be of marginal use
to analytical chemists involved in trace organic analysis.
63
-------
2000
Figure JO.
CM
-I
800
TLC-IR spectra of 1.2 |j.g of methylene blue
measured by DRIFT spectrometry- About one
half of the observable bands, on this spectrum
are caused by the adsorbate, and "the rest are
caused by the silica gel adsorbent.
64
-------
REFERENCES
1. Griffiths, P.R. Optimized Sampling in, the Gas Chromatography-Infrared
Spectroscopy Interface, Appl. Spectres., 31:284-288, 1977.
2. Dean, J.A. Chemical Separation Methods. Van Nostrand Reinhold
Company, New York, 1969. 398 pp.
3. Mantz, A.W. Sensitivity in Complex Analysis. Industrial Research,
19(2): 90-94, 1977-
4. Griffiths, P.R., Sloane, H.J., and Hannah, R.W. Interferometers vs.
Monochromators:Separating the Optical and Digital Advantages. Appl.
Spectrosc.,31:485-495, 1977.
5. Mark, H. , and Low, M. J. D. Background Elimination in Spectra Generated
by Fourier Transform Spectrometers. Appl. Spectrosc.,25:605-608, 1971.
6. Hohnstreiter, G.F., Sheahen, T. P. , and Howell, W. Michelson Inter-
ferometer for On-Board Space Vehicle Measurements. In: Aspen
International Conference on Fourier Spectroscopy, 1970, G.A. Vanasse,
A.T. Stair, and D. J. Baker, eds. AFCRL-71-0019, pp. 243-254.
7. Sheahen, T. P. Use of Chirping to Distinguish Good from Bad Inter-
ferograms and Spectra. Appl. Spectrosc., 28:283-285, 1974.
8. Bar-Lev, H. A Dual-Beam Interferometer Spectrometer. Infrared Phys.,
7:93-98, 1967.
9- Low, M.J.D., and Mark, H. Infrared Fourier Transform Spectroscopy in
the Coatings Industry. II. -Optical Subtraction. J. Paint Technol.,
^3(553): 31-M, 1971.
10. Griffiths, P.R. Chemical Infrared Fourier Transform Spectroscopy.
Wiley-Interscience, New York, 1975- 3^ pp.
11. Chrandrasekhar, H.R., Genzel, L. , and Kuhl, J. Double-Beam Fourier
Spectroscopy with Interferometric Background Compensation. Optics
Commun., 17( l) : 106-110, 1976.
12. Kuehl, D., and Griffiths, P.R. A Dual-Beam Fourier Transform Infrared
Spectrometer. Anal. Chem., 50:4l8-422, 1978.
13. Mertz, L. Auxilliary Computation for Fourier Spectrometer. Infrared
Phys., 7:17-23, 1967.
65
-------
14. Griffiths, P.R. On-Line Measurement of the Infrared Spectra of Gas
Chromatographic Effluents. ERL-026, U.S. Environmental Protection
Agency, Athens, Georgia, 1976. 19 pp.
15. Gomez-Taylor, M.M.,, and Griffiths, P.R. On-Line Identification of Gas
Chromatographic Effluents by Dual-Beam Fourier Transform Infrared
Spectrometry. Anal. Chem., 50:422-425, 1978.
16. Burnham, A. K., Calder, G.V., Fritz, J.S. , Junk, G.A., Svec, H.J.,
and Willis, R. Identification and Estimation of Neutral Organic
Contaminants in Potable Water. Anal. Chem., 44: 139-142, 1972.
17. Glaze, ¥.H., Hendersen, J.E., Bell, J.E., and Wheeler, V.A. Analysis
of Organic Materials in Wastewater Effluents after Chlorination.
J. Chromatogr. Sci., 11:580-584, 1973.
18. Junk, G.A., Richard, J.J., Grieser, M.D., Witiak, J.L., Arguello, M.D.,
Vick, R., Svec, H. J. , Fritz, J.S., and Calder, G.V. Use of
Macroreticular Resins in the Analysis of Water for Trace Organic
Components. J. Chromatogr., 99:745-762, 1974.
19. Grob, K., and Grob, G. Splitless Injection on Capillary Columns,
Part I. The Basic Technique; Steriod Analysis as an Example. J.
Chromatogr. Sci., 7:584-586, 1969.
20. Grob, K., and Grob, G. Splitless Injection on Capillary Columns,
Part II. Conditions and Limits, Practical Realization. J. Chromatogr.
Sci., 7:587-591, 1969.
21. Grob, K., and Grob, K., Jr. Isothermal Analysis on Capillary Columns
without Stream Splitting. The Role of the Solvent. J. Chromatogr.,
9^:53-64, 1974.
22. Azarraga, L.V. E.P.A. Environmental Research Laboratory, Athens,
Georgia. Personal Communication, 1976.
•
23. Hannah, R.W., Pattacini, S.C., Grasselli, J.G., and Mocadlo, S.E.
Trace Analysis by Infrared Spectroscopy using Preconcentration,
Separation and Computer Techniques. Appl. Spectres., 32:69-79, 1978.
24. Fuller, M.P., and Griffiths, P.R. Diffuse Reflectance Measurements
by Infrared Fourier Transform Spectrometry. Anal. Chem., 50:1906-1910,
1978.
25. Fuller, M.P., and Griffiths, P.R. Infrared Analysis by Diffuse Reflectance
Spectrometry. Amer. Lab., 10(10)169-80, 1978.
26. Wood, B.E., Pipes, J. G., Smith, A.M., and Roux, J.A. Hemi-Ellipsoidal
Mirror Infrared Refectometer:Development and Operation. Applied
Optics, 15:9^-950, 1976.
66
-------
27- Kubelka, P. Hew Contributions to the Optics of Intensely Light Scattering
Materials, Part I. J. Opt. Soc. Am., 38:448-460, 1948, and references
contained therein.
28. .Percival, C. J., and Griffiths, P.R. Direct Measurement of the Infrared
Spectra of Compounds Separated by Thin-Layer Chromatography. Anal.
Chem. , 47:154-156, 1975.
29. Gomez-Taylor, M.M., Kuehl, D., and Griffiths, P.R. Vibrational
Spectrometry of Pesticides and Related Materials on Thin-Layer
Chromatography Adsorbents. Appl. Spectrosc., 30:447-452, 1976.
30. Perry, J.A., Haag, K.W., and Glunz, L.J. Programmed Multiple
Development in Thin-Layer Chromatography. J. Chromatogr. Sci.,
11:447-453, 1973-
31. Perry, J.A. Programmed Multiple Development Lateral Spot
Reconcentration. J. Chromatogr., 110:27-35, 1975.
32. Perry, J.A., Jupille, T.H., and Glunz, L.J. TLC:Programmed Multiple
Development. Anal. Chem., 47:65A-74A, 1975-
33- Low, M.J.D. Infrared Fourier Transform Spectroscopy in Flavor
Analysis. IV. Spectra of Gas Chromatography Fractions. J. Agr.
Food Chem., 19:1124-1127, 1971.
67
-------
TECHNICAL REPORT DATA
(Please read Instructions on the reverse before completing)
1. REPORT NO.
EPA-600/4-79-064
3. RECIPIENT'S ACCESSION>NO.
4. TITLE AND SUBTITLE
Automated Measurements of Infrared Spectra
of Chromatographically Separated Fractions
5. REPORT DATE
October 1979 issuing date
6. PERFORMING ORGANIZATION CODE
'. AUTHOR(S)
Peter R. Griffiths
8. PERFORMING ORGANIZATION REPORT NO,
9. PERFORMING ORGANIZATION NAME AND ADDRESS
Department of Chemistry
Ohio University
Athens OH 45701
10. PROGRAM ELEMENT NO.
1BD713
11. CONTRACT/GRANT NO.
R804333-01
12. SPONSORING AGENCY NAME AND ADDRESS
Environmental Research Laboratory—Athens, GA
Office of Research and Development
U.S. Environmental Protection Agency
Athens GA 30605
13. TYPE OF REPORT AND PERIOD COVERED
Final, 4/76-10/78
14. SPONSORING AGENCY CODE
EPA/600/01
15 SUPPLEMENTARY NOTES
16ABSTRACT
To optimize the sensitivity of infrared measurements of gas chromatographic
effluents, the optimal dimensions for the light-pipe gas cells were first calculated.
The transmittance of light-pipes with these optimized dimensions is so high that the
signal-to-noise ratio of the single-beam interferograms measured using a mercury cadmi-
um telluride photodetector is limited by digitization noise. To get around this proble
the application of dual-beam Fourier transform infrared spectroscopy was tested and the
sensitivity of measurements was four times greater than the single beam measurement.
Detection limits of less than 1 ppb were observed when this system was applied to trace
organics in water. SCOT columns and shorter light pipes were expected to produce de-
tection limits below 10 ng. The application of dual-beam FT-IR spectroscopy to the on-
line identification of peaks eluting from a high performance liquid chromatograph was
investigated. Typical detection limits in excess of 10 yg were found, which are too
great for general analytical work. Spectra of submicrogram quantities were measured in
preliminary work with a system based on diffuse reflectance measurements of deposited
solutes on KC1 powder. The sensitivity of techniques for the in situ identification of
species on specially prepared thin-layer chromatographic plates was improved by the ap-
plication of programmed multiple development. Detection limits of 100 ng could be
achieved in less than 5 seconds data acquisition time, and limits of 10 ng could be
achieved after extended signal-averaging.
17.
KEY WORDS AND DOCUMENT ANALYSIS
DESCRIPTORS
b.lDENTIFIERS/OPEN ENDED TERMS
c. COSATI Field/Group
Infrared Spectroscopy
Organic Chemistry
Gas Chromatography
Chemical Analysis
Liquid Chromatography
07C
68D
3. DISTRIBUTION STATEMENT
RELEASE TO PUBLIC
19.
21. NO. OF PAGES
80
20. SECURITY CLASS (Thispage)
UNCLASSIFIED
22. PRICE
EPA Form 2220-1 (9-73)
68
» U.S. GOVERNMENT PRINTING OFFICE: 1979657.146/5480
------- |