Anal. Chem.

2005, 77, 5965-5972

Reduction of Detection Limits of the Direct

Deposition GC/FT-IR Interface by
Surface-Enhanced Infrared Absorption
David A. Heaps and Peter R. Griffiths*

Department of Chemistry, University of Idaho, Moscow, Idaho 83844-2343

Even though the interface of gas chromatography (GC) and

mass spectrometry (MS) is by far the most popular way
of identifying molecules eluting from a GC in real time,
the identification of compositional isomers by MS is
equivocal at best. Much better results would be found by
infrared spectrometry (IR) if the sensitivity of the GC/IR
interface could be improved. In this paper, we show how
the smallest quantity of molecules injected into a GC for
which an identifiable infrared spectrum can be measured
on-line has been reduced by a factor of 10 below the
detection limit of the most sensitive current technique. A
commercial direct deposition interface between a GC and
a Fourier transform infrared spectrometer was modified
by vapor-depositing an island film of silver on the surface
of the zinc selenide substrate. Band intensities in the
spectra of molecules located within ∼4 nm of the surface
of the silver islands were increased by at least 1 order of
magnitude through surface-enhanced infrared absorption
Figure 1. Reference mass spectra of the butylbenzene isomers
(SEIRA). The effectiveness of this approach was illustrated for the on-line NIST database. (A) isobutylbenzene, (B) n-butylben-
by comparing the limits of identification of butylbenzene zene, (C) sec-butylbenzene, and (D) tert-butylbenzene.
isomers measured with and without the silver film.
Comparison with the spectra of the same molecules
measured by mass spectrometry showed the increased that we discuss below, it may still be difficult to assign the correct
sensitivity and specificity of the GC/SEIRA interface. structure to a given GC peak.
It is well known that differences between the infrared spectra
Any spectroscopic technique that is to be considered for the of constitutional isomers are usually much greater than the
on-line identification of molecules eluting from a gas chromatog- corresponding mass spectra. As an example, let us compare the
raphy (GC) column must be fast, sensitive, and selective. Electron mass spectra and infrared spectra of the four butylbenzene
ionization (EI) mass spectrometry (MS) is certainly fast and isomers. The mass spectra of these molecules contained in the
sensitive and, for many compounds, gives a unique fragmentation NIST/EPA/NIH database3 are shown in Figure 1. It can be seen
pattern. By comparing measured mass spectra of GC peaks with that the spectra of isobutylbenzene and n-butylbenzene are very
the spectra in large spectral databases, many components of similar and quite different from the spectra of sec-butylbenzene
complex mixtures can be readily identified. The identification of and tert-butylbenzene.
GC peaks by mass spectral library searching has two drawbacks, The infrared spectra of these four isomers are shown in Figure
however. EI mass spectra can vary with the energy of the ionizing 2. The substitution pattern may be determined from the strong
electrons; for this reason, mass spectral libraries often contain bands absorbing between 900 and 650 cm-1. The strong bands at
multiple entries for a given compound. In addition, the mass ∼750 and ∼700 cm-1, assigned to the aromatic C-H out-of-plane
spectra of constitutional isomers can be very similar, especially deformation mode and a ring-puckering vibration, respectively,
for aromatic molecules. For example, it is well known that the EI immediately show that the aromatic ring is either monosubstituted
mass spectra of o-, m-, and p-xylene are essentially identical when (1) Wilkins, C. L.; Giss, G. N.; Brissey, G. M.; Steiner, S. Anal. Chem. 1981,
measured under identical conditions.1,2 Even when fragmentation 53, 113-117.
occurs on the substituent, as it does for the butylbenzene isomers (2) Crawford, R. W.; Hirschfeld, T.; Sanborn, R. H.; Wong, C. M. Anal. Chem.
1982, 54, 817-820.
* To whom correspondence should be addressed. E-mail: pgriff@ (3) NIST/EPA/NIH Mass Spectral Reference Database 1, National Institute of
uidaho.edu. Standards and Technology, Gaithersburg, MD, 2002.

10.1021/ac050585l CCC: $30.25 © 2005 American Chemical Society Analytical Chemistry, Vol. 77, No. 18, September 15, 2005 5965
Published on Web 08/11/2005

Figure 2. DD GC/FT-IR of 30 ng of the butylbenzene isomers
deposited on bare ZnSe at -95 °C measured in 1 s. (A) tert-
butylbenzene, (B) n-butylbenzene, (C) sec-butylbenzene, and (D)
isobutylbenzene.
or (with lower probability) 1,3-disubstituted.4,5 The spectra in the
aliphatic C-H stretching region (3000-2800 cm-1) readily dis-
tinguish between the number of CH2 and CH3 groups in the butyl
group. Thus tert-butylbenzene, with three CH3 groups and no CH2
groups is readily distinguished from n-butylbenzene, with its one
CH3 group and three CH2 groups. However, the spectra of
isobutylbenzene and sec-butylbenzene are very similar in this
region, because each has two CH3 groups and one CH2 group.
These two compounds may be distinguished by the fact that the
two CH3 groups in isobutylbenzene are attached to the same
carbon atom so that some of the vibrations of these CH3 groups
are coupled, whereas this is not the case for sec-butylbenzene. As
a result, the symmetric CH3 deformation band at ∼1380 cm-1 is
split, while this mode is a singlet in the spectrum of sec-
butylbenzene. The spectra of the four butylbenzene isomers in
this region are shown in Figure 3.
In the first, and still the most commonly used, approach to
GC/FT-IR, the sample is passed in the vapor phase through a
heated light-pipe gas cell.6-8 To identify each component, the
measured spectrum is compared to a library of vapor-phase
spectra, for which the largest database (Aldrich) contains fewer
than 10 000 reference spectra. Detection limits in infrared spec-
trometry for different molecules can vary by at least 1 order of
magnitude because the absorptivity of bands in IR spectra depends
on the change in dipole moment during the vibration. Thus, the
limit of detection (LOD) of nonpolar molecules (such as the
butylbenzene isomers on which we report in this paper) may be
(4) Lin-Vien, D.; Colthup, N. B.; Fateley, W. G.; Grasselli, J. G. The Handbook
of Infrared and Raman Characteristic Frequencies of Organic Molecules;
Academic Press: San Diego, CA, 1991.
(5) Mayo, D. W.; Miler, F. A.; Hannah, R. W. Course Notes on the Interpretation
of Infrared and Raman Spectra; Wiley-Interscience: New York, 2004.
(6) Azarraga, L. V. Appl. Spectrosc. 1980 34, 224-226.
(7) Herres, W. HRGC-FTIR: Capillary Gas Chromatography Fourier Transform
Infrared Spectroscopy, Theory and Applications; Huethig: Heidelberg, Ger-
many, 1987.
(8) Reedy, G. T.; Bourne, S.; Cunningham, P. T. Anal. Chem. 1979, 51, 1535-
1540.
5966 Analytical Chemistry, Vol. 77, No. 18, September 15, 2005
Figure 3. ATR spectra of the butylbenzene isomers in the spectral
range between 1420 and 1320 cm-1. (A) tert-butylbenzene, (B)
n-butylbenzene, (C) sec-butylbenzene, and (D) isobutylbenzene.
much greater than the LOD of polar molecules. With a light-pipe
GC/FT-IR interface, the LOD is typically between ∼50 (for
nonpolar compounds) and ∼10 ng (for polar compounds.)
The sensitivity of GC/FT-IR measurements was increased
beyond that obtained with a light-pipe by isolating each separated
component in a matrix of solid argon in the manner developed
by Reedy et al.8-10 Real-time measurements were not possible with
the matrix isolation (MI) GC/FT-IR interface, and the spectrum
of each analyte had to be measured after all components had
eluted. Because the samples are trapped in an area that is smaller
than the cross-sectional area of a light-pipe, detection limits with
the MI GC/FT-IR interface are at least a factor of 10 lower than
for the light-pipe interface. Again, however, a special-purpose
library, this time of MI spectra, was required; this database
contained significantly fewer entries than the library of vapor-phase
spectra. This factor, in combination with the mechanical complex-
ity of the MI GC/FT-IR interface, led researchers to consider other
approaches.
The third, and most sensitive, type of GC/FT-IR interface relies
on the direct deposition of the effluent from the GC column onto
a slowly moving, liquid nitrogen (LN2)-cooled zinc selenide
window so that each component condenses as a very small spot
on the window.10-13 Immediately after the deposition, each spot
passes through the beam of a simple FT-IR microscope operating
in the transmission mode, so that spectra are measured at intervals
of 1 s or less. The smaller the diameter of the spot, the thicker is
the sample and the greater the peak absorbance. Because the
diameter of each deposited component can be made as small as
∼100 µm, the amount of polar molecules required to give a
spectrum in real time (i.e., as the spot passes through the infrared
(9) Reedy, G. T.; Ettinger, D. G.; Schneider, J. F.; Bourne, S.; Cunningham,
Anal. Chem. 1985, 57, 1602-1609.
(10) Visser. T. Gas Chromatography/Fourier Transform Infrared Spectroscopy.
In Handbook of Vibrational Spectroscopy; Chalmers, J. M., Griffiths, P. R.,
Eds.; John Wiley and Sons: Chichester, U.K., 2002; Vol. 2, pp 1605-1626.
(11) Haefner, A. M.; Norton, K. L.; Griffiths, P. R.; Bourne, S.; Curbelo, R, Anal.
Chem. 1988, 60, 2441-2444.
(12) Bourne, S.; Haefner, A. M.; Norton, K. L.; Griffiths, P. R. Anal. Chem. 1990,
62, 2448-2452.
(13) Norton, K. L.; Griffiths, P. R. J. Chromatogr., A 1995, 703, 383-392.

Figure 4. Infrared spectra of (A) 30 ng of isobutylbenzene on ZnSe
at -95 °C measured on-line in 1 s; (B) liquid film of isobutylbenzene
between two NaCl plates at 20 °C.
beam) is usually less than 500 pg. Jackson et al.14 and Visser10
have compared the GC/MI-FT-IR and GC/direct-deposition (DD)-
FT-IR interfaces and showed that, for spectra measured in equal
times, the sensitivity of the DD interface is between a factor of 5
and 10 greater than that of the MI interface. It should be
recognized, however, that only the GC/DD-FT-IR interface is
capable of measuring the spectra of separated components in real
time.
GC/FT-IR spectra measured by the DD technique have an
additional advantage in that the spectra are often very similar to
the standard reference spectra of solid samples prepared as KBr
disks12,13 or even liquid samples run as capillary films. As an
example, the DD GC/FT-IR spectrum of isobutylbenzene at -95
°C is compared to the liquid-phase spectrum of the same molecule
at ambient temperature in Figure 4. Because, for DD GC/FT-IR,
standard reference spectra of condensed-phase samples may be
used, the number of compounds for which reference spectra are
available is greatly increased, since libraries of condensed-phase
spectra exceed 100 000 entries.
With all early GC/FT-IR interfaces, the amount of sample that
had to be injected onto a GC column in order for a given
component to be identified, which we will call the limit of
identification (LOI), was usually at least 100 times greater for GC/
FT-IR than for GC/MS. This large difference in sensitivity has
led most GC practitioners to rely on MS as the sole technique
used for the on-line identification of all peaks in their chromato-
grams. Indeed, many chromatographers have complete faith in
the results of a GC/MS library search to the point and assign the
identity of each peak to the first “hit” in the list of possible matches
without performing a visual comparison of the measured spectrum
and the best match. A reliable GC/IR interface with the same
detection limits as even a benchtop GC/MS interface would
greatly increase the reliability to which each component of a
mixture that has been separated by gas chromatography could
be identified.
The goal of this project is to decrease the detection limits of
DD GC/FT-IR measurements by taking advantage of the phe-
(14) Jackson, P.; Dent, G.; Carter, D.; Schofield, D. J.; Chalmers, J. M.; Visser,
T.; Vredenbregt, M. J. High Resolut. Chromatogr. 1993, 16, 515-521.
nomenon of surface-enhanced infrared absorption (SEIRA)15,16 in
an analogous way to the off-line measurements reported for liquid
chromatography by Sudo et al.17 Our goal was to reduce the LOI
for spectra measured on-line using the DD GC/SEIRA interface
to the point that it is equal or less than that of GC/MS. Even
though SEIRA has been recognized for 25 years,18 it has found
far fewer applications to date than its vibrational spectroscopic
counterpart, surface-enhanced Raman scattering (SERS). In
SEIRA, the absorbance of molecules within a few nanometers of
the surface of island films of silver and gold is enhanced by up to
∼2 orders of magnitude, which is ∼100 times less than for SERS
in the absence of resonance enhancement. Other metals also give
rise to surface enhancement of infrared spectra, but usually by
only a factor of ∼10.19-23 In this paper, we report how coating the
surface of the ZnSe substrate with a thin island film of silver can
reduce the detection limit of the GC/FT-IR interface to the point
that it becomes comparable to those of GC/MS.
SEIRA spectra must be essentially identical to standard DD
GC/FT-IR spectra if molecules are to be identified by spectral
searching. Several research groups have reported22-26 that the
shape of bands in the SEIRA spectra of species on the surface of
metal-island films can become asymmetric under certain circum-
stances. Avoiding this asymmetric band shape is important if a
DD GC/SEIRA spectrum is to be input to a standard database
for spectral library searching.
EXPERIMENTAL SECTION
A Bourne Scientific (Newton, MA) infrared chromatograph
(IRC) DD GC/FT-IR interface was used to obtain all GC/FT-IR
spectra shown in this paper.27 In this instrument, the effluent from
the GC is passed through a heated transfer line through a fused-
silica deposition tip onto a cooled ZnSe window the position of
which is controlled by an x-y stage. The beam of a rudimentary
FT-IR microscope (beam diameter, 100 µm) is located less than
1 mm from the deposition tip so that the spectrum of each
component is measured very shortly after it is deposited. To avoid
the possibility of sample condensation, the temperature of the
transfer line was held 10 °C above final temperature of the GC
program. The temperature of the deposition tip was set at 250
°C. The temperature of the ZnSe substrate was thermostated at
-95 °C, with the pressure in the vacuum chamber at ∼5 × 10-5
Torr.
Thin films of silver were deposited using a custom-built
physical vapor deposition apparatus. For the deposition of silver,
(15) Osawa, M. Bull. Chem. Soc. Jpn. 1997, 70, 2861-2880.
(16) Aroca, R. F.; Ross, D. J.; Domingo, C. Appl. Spectrosc. 2004, 58, 324A-
338A.
(17) Hartstein, A.; Kirtley, J. R.; Tsang, J. C. Phys. Rev. Lett. 1980, 45, 201-4.
(18) Sudo, E.; Esaki; Y.; Sugiura, M. Bunseki Kagaku 2001, 50, 703-707.
(19) Bjerke, A. E.; Griffiths, P. R. Appl. Spectrosc. 2002, 56, 1275-1280.
(20) Bjerke, A. E.; Griffiths, P. R; Theiss, W. Anal. Chem. 1999, 71, 1967-
1974.
(21) Chen, Y.-J.; Sun, S.-G.; Chen, S.-P.; Li, J.-T.; Gong, H Langmuir 2004, 20,
9920-9925.
(22) Wang, H.-C.; Sun, S.-G.; Yan, J.-W.; Yang, H.-Z.; Zhou, Z.-Y. J. Phys. Chem.
B 2005, 109, 4309-4316.
(23) Krauth, O.; Fahsold, G.; Magg, N.; Puuci, A. J. Chem. Phys. 2000, 113,
6330-6333.
(24) Ishida, K. P.; Griffiths, P. R. Anal. Chem. 1994, 66, 522-530.
(25) Merklin; G. T.; Griffiths, P. R. J. Phys. Chem. B 1997, 101, 5810-5813.
(26) Brown, C. W. University of Rhode Island. Personal communication to P.R.G.,
2003.
(27) Bourne, S. Am. Lab. 1998, 30 (16), 17F-17J.
Analytical Chemistry, Vol. 77, No. 18, September 15, 2005 5967
Table 1. Results of Searching GC/MS Spectra of the Butylbenzene Isomers Injected at Levels of 2.9 ng

hit direct match reverse search probability

compound rank molecule in MS “hit list” factor factor factor
isobutylbenzene 1 isobutylbenzene 598 641R 26.3P
2 n-butylbenzene 582 622R 15.2P
3 isobutylbenzene 573 616R 26.3P
4 acetophenone, 4-(benzyloxymethyl)- 570 607R 10.1P
5 isobutylbenzene 568 611R 26.3P
6 n-butylbenzene 568 610R 15.2P
7 bicyclo[2.2.2]oct-7-en-2-one, 5-methylene- 561 595R 7.33P
8 isobutylbenzene 561 622R 26.3P
9 n-butylbenzene 560 590R 15.2P
10 n-butylbenzene 559 599R 15.2P
n-butylbenzene 1 n-butylbenzene 720 741R 38.4P
2 n-Butylbenzene 717 743R 38.4P
3 n-butylbenzene 715 736R 38.4P
4 n-butylbenzene 709 744R 38.4P
5 n-butylbenzene 708 734R 38.4P
6 isobutylbenzene 706 732R 24.1P
7 isobutylbenzene 699 724R 24.1P
8 isobutylbenzene 689 714R 24.1P
9 1,2,3,4,4a,8a- hexahydronaphthalene 681 704R 7.34P
10 1-hexen-4-ol, 3-methyl-6-phenyl- 675 680R 5.77P
sec-butylbenzene 1 benzene, diethyl- 638 794R 41.7P
2 benzene, 1-methyl-3-propyl- 612 719R 12.5P
3 sec-butylbenzene 605 708R 9.57P
4 benzeneethanol, 3-methyl- 598 752R 7.33P
5 n-butylbenzene 582 747R 4.22P
6 hydrazine, (2-phenylethyl)- 580 585R 3.89P
7 benzene, 1-methyl-2-propyl- 574 678R 3.06P
8 benzene, 1-methyl-3-propyl- 571 608R 12.5P
9 oxirane, 2-methyl-3-phenyl- 570 694R 2.58P
10 benzeneethanol, β-methyl-, (S)- 569 574R 2.48P
tert-butylbenzene 1 benzene, 1-methyl-4-(1-methylethyl)- 619 728R 10.2P
2 benzene, 1,2,4,5-tetramethyl- 619 709R 10.2P
3 benzene, 1,2,3,5-tetramethyl- 618 752R 9.75P
4 tert-butylbenzene 616 731R 9.00P
5 benzene, 1-methyl-4-(1-methylethyl)- 615 724R 10.2P
6 benzene, 1-methyl-2-(1-methylethyl)- 611 716R 7.25P
7 tert-butylbenzene 609 717R 9.00P
8 tert-butylbenzene 608 736R 9.00P
9 benzene, 1-methyl-3-(1-methylethyl)- 606 701R 5.84P
10 oxazole, 2,5-dihydro-5-(4-methylphenyl)-4-phenyl- 603 716R 5.16P

the vacuum chamber was first evacuated to 0.1 Torr and a plasma
glow discharge was used to clean the ZnSe surface prior to the
plating of the silver. The glow discharge was formed at 3 mA and
3 kV for 5 min at a pressure of 0.1-0.05 Torr. After the glow
discharge was terminated, the pressure was reduced to 1 × 10-6
Torr with an oil diffusion pump. The amount of silver deposited
was measured with a quartz crystal microbalance; the thickness
and deposition rates given in this paper are calculated by assuming
that the deposited layer is uniform. All results shown in this paper
were obtained with a 5-nm-thick silver-island film deposited at a
rate of 1.2 × 10-2 nm s-1.
Separations were performed on a Hewlett-Packard 5890 II GC
equipped with a 30:1 inlet splitter. The column was 15 m long
with a 0.25-mm internal diameter (i.d.); the stationary phase was
a 0.25-µm film of 5% methyl, 95% phenyl polysiloxane (ZB-5,
Phenomenex). Most separations were performed by holding the
column at 150 °C for 10 min, followed by a temperature ramp of
5 °C/min to a final temperature of 225 °C. The column head
pressure was 10 psi. Solutions of the butylbenzene isomers (99%
pure, Aldrich) in methanol at concentrations of 1500, 86, and 8.6
ppm solutions were prepared. For all separations reported in this
paper, 1 µL of these solutions was injected. Since the GC was
equipped with a 30:1 inlet splitter, 50 ng, 2.9 ng, and 290 pg of
the butylbenzenes were deposited on the bare or silver-coated
ZnSe window.
GC/MS analyses were performed by injecting 1.0-µl aliquots
of each solution into a Hewlett-Packard Model 6890 chromato-
graph equipped with a 30:1 inlet splitter and a 30-m, 0.25-mm-i.d.
column with a 0.25-µm film of DB-5. The column was held at a
temperature of 50 °C for 2 min and then increased at the rate of
15 °C/minute until all components had eluted. The output of the
column was passed into a JMS-AX505HA mass spectrometer
(JEOL) that was scanned from 30 to 300 m/z every 1.0 s, with
unit resolution. The mass spectrometer was operated with an
ionization voltage of 70 eV, source temperature of 180 °C, and
detector voltage of 2 kV.
RESULTS
Mass spectra are usually identified by a combination of a
forward search and a reverse search.28 In the forward search, a
similarity value is calculated that is a measure of how closely the
intensities of a certain number (often 5 or 10) of lines in the sample
(28) Stein, S. E.; Scott, D. R. J. Am. Soc. Mass Spectrom. 1994, 5, 859-866.

5968 Analytical Chemistry, Vol. 77, No. 18, September 15, 2005
spectrum match the intensities of the corresponding lines for all Table 2. Results of Searching DD GC/FT-IR Spectra of
library spectra. In reverse searches, the intensities of the lines in the Butylbenzene Isomers Injected at Levels of 2.9 ng
the library spectra are compared to the intensity of the corre-
hit
sponding lines in the sample spectra. Forward searching generally compound rank molecule in IR “hit list” match
works best when the GC peak contains only one component,
n-butylbenzene 1 n-butylbenzene 89.28
whereas a reverse search is more useful if a component with a 2 amylbenezene 86.84
similar retention time coelutes with the GC peak of interest. Some 3 propylbenzene 81.47
4 tetraphenylethylene 77.33
library searching programs include an algorithm that gives a 5 benzhydrol 76.85
probability factor that a given peak has been correctly identified. 6 stilbene oxide, trans- 76.70
If the library contains several entries for a given molecule and 7 triphenylmethane 75.73
8 m-terphenyl 73.34
more than one of these entries are found to be in the top 10 9 triphenylphosphine 71.31
matches, the probability of a correct identification is increased. 10 benzyl sulfide 70.75
If, on the other hand, only one reference spectrum of a given sec-butylbenzene 1 sec-butylbenzene 78.18
molecule is contained in the library, the probability factor of a 2 cumene 77.51
3 poly(styrene-acrylonitrile), 74.85
correct match is reduced. 25% acrylonitrile
To illustrate some of the limitations of contemporary mass 4 3-phenyl-1-butanol 72.01
spectral library searching algorithms, GC/MS spectra of 2.9 ng 5 poly(styrene-vinylidene 70.84
chloride)
each of the four butylbenzenes were measured. There are five 6 tert-butylbenzene 68.22
reference spectra each of n-butylbenzene and sec-butylbenzene 7 benzyl mercaptan 68.09
8 ABS plastic (ATR Ge crystal) 67.01
and four spectra each of isobutylbenzene and tert-butylbenzene 9 ABS plastic (ATR corrected) 66.91
in the MS database. The results of library searching are shown 10 (+)-2-phenyl-1-propanol 66.28
in Table 1. It can be seen that the first five hits for n-butylbenzene tert-butylbenzene 1 tert-butylbenzene 82.51
were each of the authentic compound, despite the great similarity 2 cumene 75.26
3 sec-butylbenzene 68.25
of the spectra of n-butylbenzene and isobutylbenzene seen in 4 3-phenyl-1-butanol 63.46
Figure 1. Not unexpectedly, the next three hits were also for 5 poly(styrene-vinylidene 62.83
isobutylbenzene. In light of the similarity of the reference spectra chloride)
6 poly(styrene-acrylonitrile), 60.63
of n-butylbenzene and isobutylbenzene, this result could certainly 25% acrylonitrile
be classified as a success. The results for isobutylbenzene were 7 (+)-2-phenyl-1-propanol 59.27
8 (()-1-phenyl-1-propanol 56.33
more equivocal, with isobutylbenzene appearing as hit ranks 1, 9 benzyl chloride 56.25
3, 5, and 8 and n-butylbenzene as hit ranks 2, 6, 9 and 10. Despite 10 ABS plastic (ATR Ge crystal) 55.93
the spectral similarity to n-butylbenzene, isobutylbenzene was
successfully identified.
We expected the results for sec-butylbenzene and tert-butyl- of the DD GC/FT-IR spectra shown in Figure 2 was of 29 ng of
benzene to be even better, since the mass spectra of these each butylbenzene isomer injected on the column. On-line DD
compounds are quite different from each other and from the GC/FT-IR spectra of equivalent quality of sample quantities as
spectra of the n and iso isomers (see Figure 1). However, this low as 300 pg of 3-ethylphenol12 and 75 pg of barbiturates13 have
was not found to be the case. sec-Butylbenzene was identified as been reported previously. However, these molecules are far more
1-methyl,3-propylbenzene, with diethylbenzene as the second polar than the butylbenzene isomers examined in this investiga-
match and sec-butylbenzene third. Note that even though all three tion, and consequently, the absorptivities of the stronger bands
compounds are isomeric with monosubstituted butylbenzenes, the are significantly higher. Furthermore, the spectra in refs 12 and
first two hits are disubstituted molecules; the probability factor 13 were measured using the Bio-Rad Tracer, for which the
that the compound was in fact sec-butylbenzene was less than 10%. baseline noise is a factor of almost 10 lower than noise measured
Even worse results were found with tert-butylbenzene, which was in an equivalent time with the Bourne Scientific IRC.
ranked as having the fourth highest probability, with the more The result of searching these spectra against several libraries
highly ranked compounds being either di- or tetrasubstituted. Only of condensed-phase infrared spectra (Aldrich Condensed Phase
the fact that there are four reference spectra of tert-butylbenzene Sample Library, Georgia State Crime Lab Sample Library, Hummel
in the database caused the probability of this compound being Polymer Sample Library, Nicolet Condensed Phase Academic
identified correctly to be as high as it was. Despite the uncertain Sampler, and Sigma Biological Sample Library) is shown in Table
identification of the butylbenzene isomers by GC/MS and the 2. The reference spectrum of isobutylbenzene was not in these
expectation that infrared spectrometry would allow these isomers databases and so the results are not shown for this molecule. The
to be distinguished with much more certainty than MS, the low corresponding result obtained when the concentration of the
sensitivity of IR spectrometry leads the vast majority of gas analytes was reduced by 1 order of magnitude is shown in Table
chromatographers to use MS as their primary peak identification 3. To increase the match factor for Table 3 (290-pg sample),
tool. spectra measured over a 5-s interval were averaged together. Even
As noted above, infrared spectra allow many constitutional for the data shown in Table 3, for which the signal-to-noise ratio
isomers to be identified unequivocally but the detection limits of (SNR) of the input spectra was relatively poor, the authentic
GC/FT-IR interfaces are usually far worse than those on GC/ compound was the first hit, demonstrating the success of infrared
MS, especially when the analytes are nonpolar. For example, each spectrometry in identifying the correct isomer. Not surprisingly
Analytical Chemistry, Vol. 77, No. 18, September 15, 2005 5969
Table 3. Results of Searching DD GC/FT-IR Spectra of
the Butylbenzene Isomers Injected at Levels of 290 pg

hit
compound rank molecule in IR “hit list” match
n-butylbenzene 1 n-butylbenzene 43.37
2 propylbenzene 41.43
3 amylbenzene 40.22
4 ethylbenzene 38.87
5 diphenylmercury 36.34
6 biphenyl 35.73
7 triphenylphosphine 34.69
8 triphenylarsine 34.63
9 sec-butylbenzene 34.33
10 benzphetamine in KBr 34.13
sec-butylbenzene 1 4-phenyltoluene 39.13
2 1-phenyl-1-propyne 36.79
Figure 5. Spectra of a self-assembled monolayer of p-nitrothiophe-
3 cumene 36.77
4 sec-butylbenzene 35.77 nol adsorbed on silver films deposited on a ZnSe substrate at the
5 poly(styrene) 31.81 same rate: (A) 5-nm-thick film; (B) 6-nm-thick film.
6 poly(styrene) atactic 31.34
7 tert-butylbenzene 31.21
8 poly(styrene-acrylonitrile), 28.95
25% acrylonitrile
9 benzyl mercaptan 28.15
10 poly(styrene-vinylidene 28.04
chloride)
tert-butylbenzene 1 benzonitrile 29.95
2 benzonitrile 29.95
3 phenylacetylene 26.42
4 1-phenyl-1-propyne 25.97
5 tert-butylbenzene 25.40
6 ethyl benzoate 23.61
7 ethyl benzoate 23.60
8 cumene 23.51
9 p-chlorobiphenyl 23.05
10 (E)-stilbene 22.84

in view of the very strong characteristic bands at ∼750 and 700

cm-1, every single molecule listed in Tables 2 and 3 contains a
monosubstituted aromatic ring. For n-butylbenzene, the second Figure 6. DD GC/FT-IR spectra of 2.9 ng of tert-butylbenzene on
and third “hits” were n-amyl and n-propylbenzene, which illustrates (A) 5-nm silver on ZnSe and (B) bare ZnSe.
that infrared spectroscopy has more difficulty in distinguishing
between members of a homologous series than between individual
yield real-time measurements. In this paper, we report the
substitutional isomers. It may also be noted that no n-alkylbenzene
feasibility of increasing the SNR of on-line DD GC/FT-IR spectra
was ever listed among the top hits for the branched isomers. Last,
by coating the ZnSe substrate with a thin island film of silver so
and not surprisingly, by comparing the results shown in Tables 2
that the absorbance of the analyte molecules that are located
and 3, it is very clear that the match factor for the first hit, and
within ∼5 nm of the surface is significantly enhanced by SEIRA.
hence the confidence that can be placed in the result, was far
Both the thickness and the morphology of the silver film must
worse when the SNR was poor, varying from between 89 and 78
be optimized if the highest SNR is to be attained for SEIRA spectra
for the high-concentration sample to between 43 and 30 for the
of thin films of organic molecules. We have found that the
low-concentration sample. Even though the first compound listed
thickness of the metal film and the rate at which it is deposited
in Table 3 was correct, a match factor value of less than 50 usually
have an important effect on both the factor by which the band
indicates either that the authentic reference spectrum is not in
intensities of thin films of organic molecules are enhanced and
the library or that the SNR was too low to have any confidence in
the shape of the absorption bands.29 In general, as the nominal
the result. It is also noteworthy that the reference spectra of the
thickness of the metal film is increased from ∼1 to 10 nm, the
butylbenzene isomers were measured with the sample in the liquid
enhancement factor increases by more than 1 order of magnitude.
phase at ambient temperature, while for the DD GC/FT-IR
However, when the density of the metal islands reaches the point
spectra, the samples were solid at -95 °C.
in which they start to coalesce, or percolate, the absorption bands
The results shown in Tables 2 and 3 suggest that even higher
of any molecule near the surface of the islands become asym-
match factors could be achieved if the SNR of the spectra could
metrical, severely reducing the success of spectral library search-
be increased above the level of the spectra shown in Figure 2.
ing. The rate at which the silver is vapor deposited also has an
Since each separated component is condensed on the LN2-cooled
effect on the thickness sat which percolation starts. At a deposition
substrate, one way to decrease the noise level of any spectrum is
by postrun signal averaging. However, this approach would not (29) Heaps, D. A.; Griffiths, P. R., unpublished results, 2003.

5970 Analytical Chemistry, Vol. 77, No. 18, September 15, 2005
Table 4. Results of Searching DD GC/SEIRA Spectra of
the Butylbenzene Isomers Injected at Levels of 2.9 ng

hit
compound rank molecule in IR “hit list” match
n-butylbenzene 1 n-butylbenzene 83.62
2 amylbenzene 80.59
3 propylbenzene 78.09
4 benzhydrol 68.10
5 tetraphenylethylene 68.09
6 stilbene oxide, trans- 67.97
7 m-terphenyl 67.03
8 triphenylmethane 65.99
9 triphenylphosphine 65.35
10 sec-butylbenzene 63.80
sec-butylbenzene 1 sec-butylbenzene 83.22
2 cumene 78.78
3 3-phenyl-1-butanol 70.73
4 tert-butylbenzene 69.83
5 ABS plastic (ATR Ge crystal) 65.94
6 ABS plastic (ATR corrected) 63.16
7 (()-1-phenyl-1-propanol 61.87 Figure 7. DD GC/FT-IR spectra of 290 pg of tert-butylbenzene on
8 (+)-2-phenyl-1-propanol 61.48 (A) 5-nm silver on ZnSe and (B) bare ZnSe.
9 poly(styrene-vinylidene 60.88
chloride)
10 poly(styrene-acrylonitrile), 59.97 Table 5. Results of Searching DD GC/SEIRA Spectra of
25% acrylonitrile the Butylbenzene Isomers Injected at Levels of 290 pg
tert-butylbenzene 1 tert-butylbenzene 81.67 hit
2 cumene 71.86 compound rank molecule in IR “hit list” match
3 sec-butylbenzene 63.49
4 3-phenyl-1-butanol 59.40 n-butylbenzene 1 n-butylbenzene 76.41
5 (+)-2-phenyl-1-propanol 54.26 2 amylbenzene 75.82
6 tert-butylbenzene 50.03 3 tetraphenylethylene 70.43
7 poly(styrene- vinylidene 49.36 4 octane 69.46
chloride) 5 stilbene oxide, trans- 68.55
8 (()-1-phenyl-1-propanol 48.85 6 benzhydrol 67.84
9 ABS plastic (ATR Ge crystal) 48.65 7 triphenylmethane 67.58
10 2,2,4-trimethylpentane 47.83 8 amylamine 67.10
9 tetrabutylammonium bromide 66.20
10 propylbenzene 65.69
sec-butylbenzene 1 sec-butylbenzene 76.47
rate of ∼0.01 nm s-1, percolation of the silver islands starts when 2 cumene 71.67
the thickness reaches ∼6 nm. As the deposition rate is increased, 3 tert-butylbenzene 63.25
4 3-phenyl-1-butanol 55.56
the onset of percolation starts at lower film thicknesses. For the 5 tert-butylbenzene 54.67
measurements described here, silver was deposited at 1.2 × 10-2 6 (()-1-phenyl-1-propanol 53.43
nm s-1. At this deposition rate, a thickness of 5 nm of silver gave 7 2,2,4-trimethylpentane 53.30
8 o-diethylbenzene 52.96
the maximum enhancement without percolation occurring. The 9 (1R,2S,5R)-(-)-menthol 52.72
spectra of a self-assembled monolayer (SAM) of p-nitrothiophenol 10 ethylbenzene 52.72
(PNTP) prepared in the way described by Carron and Kennedy30 tert-butylbenzene 1 tert-butylbenzene 73.72
on 5- and 6-nm-thick silver films are shown in Figure 5. It can be 2 cumene 67.90
3 sec-butylbenzene 61.52
seen that bands in the spectra of PNTP adsorbed on a 6-nm silver 4 tert-butylbenzene 59.03
film are asymmetric while those on the 5-nm silver film are not, 5 2,2,4-trimethylpentane 54.55
indicating the onset of percolation at a nominal thickness of 6 6 3-phenyl-1-butanol 48.56
7 cyclopentane 48.28
nm. 8 3,3-dimethylpentane 46.66
A 5-nm island film of silver deposited on the ZnSe substrate 9 2,4,4-trimethyl-2-pentene 44.56
of a DD GC/FT-IR interface reduced the LOI to below the level 10 o-diethylbenzene 43.95
achievable with an uncoated ZnSe window. The DD-GC/SEIRA
spectra of 2.9 ng of tert-butylbenzene measured with and without
the silver film are shown in Figure 6. Note that the absorbance
Osawa showed that the absorption of only those molecules
of corresponding bands in the spectra measured with the silver
that are within 4 nm of the surface is enhanced.15 For 2.9-ng
film has increased significantly because of surface enhancement.
injections of n-, sec-, and tert-butylbenzene, the thickness of the
With 5-nm films, the baseline noise was increased by less than a
sample is much greater than 4 nm, so that the effect of surface
factor of 2 because the transmittance of the silver film is greater
enhancement was minimal. The average value of the match factors
than 50%. The results of searching the spectra of the three
obtained with and without the silver film was ∼83 in both cases.
butylbenzene isomers whose spectra were available in the
Thus, it is likely that for the thicker deposits, most of the spectral
database that we used are shown in Table 4.
intensity originates from molecules that are more than 4 nm from
(30) Carron, K. T.; Kennedy, B. J. Anal. Chem. 1995, 67, 3353-3356. the metal surface.
Analytical Chemistry, Vol. 77, No. 18, September 15, 2005 5971
 For the 290-pg injections, the average value of the match factors
obtained on bare ZnSe was ∼37, while the corresponding values
when the ZnSe substrate had been prepared with a 5-nm silver
film was ∼76 (Figure 7). The significantly improved match factor
of the latter spectra can be largely attributed to the increased SNR
caused by SEIRA, but these high match factors could also only
be achieved if the amount of spectral distortion introduced by the
5-nm silver film was negligible (Table 5). We also note that both
the DD GC/FT-IR and GC/SEIRA spectra are very similar to the
liquid-phase spectra in the database, allowing for library searching
against existing condensed-phase libraries even when the sensitiv-
ity has been enhanced through the SEIRA effect.
The intensity of bands in the reflection-absorption spectra of
very thin layers of molecules on any metal surface is governed
by the surface selection rule; i.e., only those vibrational modes
that have a component of the dipole moment derivative that is
perpendicular to the surface of the metal are infrared active. The
same rule applies to SEIRA spectra, as can be seen by comparing
the SEIRA spectrum of a SAM of PNTP to a spectrum of a KBr
disk of the same molecule. In the spectrum of PNTP prepared as
a KBr disk, the intensities of the symmetric NO2 stretching mode
at 1340 cm-1 and the antisymmetric NO2 stretching mode at 1515
cm-1 are approximately equal. When PNTP is adsorbed onto a
flat metal surface, the dipole moment derivative of the symmetric
NO2 stretch is perpendicular to the surface of the silver islands,
while for the antisymmetric NO2 stretch, the dipole moment
derivative is approximately parallel to the silver surface. The
5972 Analytical Chemistry, Vol. 77, No. 18, September 15, 2005
intensity of symmetric stretching band is, therefore, considerably
more intense than that of the antisymmetric stretching band. An
exactly analogous result is seen in the SEIRA spectrum. If the
analyte molecules were oriented on deposition in DD GC/SEIRA
measurements, one would expect to observe significant difference
between the SEIRA spectrum and the standard reference spec-
trum. Since we do not observe such differences, we must conclude
that the molecules are randomly oriented.
CONCLUSION
Coating the ZnSe substrate that is used for DD GC/FT-IR
measurements leads to a reduction in the limit at which molecules
eluting from a gas chromatograph can be identified in comparison
to any other GC/FT-IR technique. Provided that the silver film is
sufficiently thin, distortions in the measured spectra are precluded
and the analyte can be identified by library searching against
standard databases of condensed-phase infrared spectra at levels
that are equal to or lower than standard GC/MS interfaces and
with better specificity for the identification of compositional
isomers.
ACKNOWLEDGMENT
We thank Kathryn Kalasinsky of the Armed Forces Institute
of Pathology for the generous loan of the Bourne Scientific IRC.
Received for review April 6, 2005. Accepted July 10, 2005.
AC050585L