1048 IR SPECTROMETERS

IR Spectrometers
RA Spragg, Perkin-Elmer Analytical Instruments, VIBRATIONAL, ROTATIONAL &
Beaconsfield, Bucks, UK
RAMAN SPECTROSCOPIES
Copyright 1999 Academic Press Methods & Instrumentation

Introduction
From the 1980s onwards Fourier-transform infrared
(FT-IR) spectrometers almost entirely replaced dis-
persive instruments. The most significant feature of
FT spectrometers is that radiation from all wave-
lengths is measured simultaneously. This is much
more efficient than the operation of the dispersive
Figure 1 The divisions of the IR region.
spectrometer, in which different wavelengths are
measured successively. This difference, known as the
Fellgett advantage, leads to much faster or more However, it is worth noting that interest in the near-
sensitive measurement. For example, a dispersive IR region has increased steadily in recent years.
spectrometer typically takes at least 3 min to scan
from 4000 to 400 cm1, while most FT instruments Dispersive IR spectrometers
can acquire a spectrum in 1 s or less. The first com-
mercial mid-IR Fourier transform spectrometer was The principle of operation of a dispersive spectrome-
introduced in 1969, but initially the cost of the neces- ter is very simple. A beam from a broad-band source
sary computing power confined the use of such of radiation is dispersed by a prism or grating. The
instruments to research applications. More recently, dispersed radiation falls on a slit through which a
FT-IR spectrometers have become less expensive in narrow range of wavelengths passes to reach a detec-
real terms than the dispersive instruments they have tor. Rotating the prism or grating brings successive
replaced. As this article will deal with instruments wavelengths onto the slit, so scanning the spectrum.
that are capable of recording complete spectra, it will For the mid IR, the source is a ceramic rod or wire
not consider laser-based measurements. heated to around 1300 K. In practice, instruments
use one or more gratings to cover the mid-IR region,
with filters to separate different orders of diffraction.
The usual detector is a thermocouple. Instruments
The IR region
operate in a double-beam mode (Figure 2) switching
The IR region covers wavelengths from about rapidly between two equivalent optical paths, one of
800 nm to 1 nm (Figure 1). However, spectro- which contains the sample. The transmission of the
scopists generally divide this into three, near IR sample is measured from the ratio between the detec-
(NIR) from the visible to 2.5 m, mid IR from 2.5 to tor signals from the two beams directly (ratio record-
25 m, and far IR (FIR) beyond 25 m. The main ing) or by moving an attenuator into the reference
significance of this division is that most fundamental beam until the detector signal matches that through
molecular vibrations occur in the mid IR, making the sample.
this region the richest in chemical information. The
near-IR region contains overtones and combinations
Fourier-transform spectrometers
of the fundamental vibrations, especially those in-
volving hydrogen atoms. Some electronic absorp- Most commercial Fourier-transform (FT) spectrome-
tions are also found in this region. The far-IR region ters are based on the Michelson interferometer
is mostly of interest for vibrations involving heavy (Figure 3). A collimated beam of radiation falls on a
atoms, lattice modes of solids and some rotational beam splitter that divides it into two beams with
absorptions of small molecules. approximately equal intensities. These two beams
This article will concentrate on the mid-IR region, strike mirrors that return them to the beam splitter
which covers the widest range of applications. where they recombine, creating interference, and
 IR SPECTROMETERS 1049

Figure 2 The optical arrangement of a double-beam dispersive IR spectrometer.

then proceed to a detector. Moving one of the Unlike dispersive spectrometers, FT instruments
mirrors creates a varying optical path difference operate in a single-beam mode. The sample is
(OPD) between the two beams. As the path differ- normally placed between the interferometer and the
ence changes, the two beams move in and out of detector, in contrast to the arrangement in dispersive
phase with each other, so that there is alternating spectrometers where the sample is placed between
constructive and destructive interference. the source and the monochromator. Two spectra are
Monochromatic radiation produces a sinusoidal recorded, first of the instrument alone and then with
signal, the frequency of which is directly propor- the sample in place. The ratio of these, calculated
tional to the wavenumber. The signal from a broad- automatically, is the transmission spectrum of the
band source consists of a combination of overlapping sample (Figures 6 and 7).
sinusoidal signals, called an interferogram. A typical
interferogram has a strong signal, called the centre
burst, at the point where the paths of the two beams Range and resolution in FT-IR
are equal, and decays as the path difference increases spectrometers
(Figure 4). The interferogram can be converted into
The detector in a FT spectrometer sees a range of fre-
the required spectrum of intensities at different wave-
quencies that is limited by the output of the source
numbers by a Fourier transformation (Figure 5).

Figure 3 The basic optics of a Michelson interferometer. Figure 4 The interference of overlapping cosine waves.
1050 IR SPECTROMETERS

Figure 5 The interferogram from a broad-band source and the

corresponding spectrum.

and the transmission of the optics. The interfero-

gram is digitized prior to the Fourier transformation.
The range of frequencies that can be identified in the
Fourier analysis depends on the spacing of the data
points in the digitized signal. There must be at least
two data points for each cycle of the highest frequen-
cy in the analysis. This is known as the Nyquist crite-
rion. If there are fewer than two data points per Figure 7 Single-beam spectra and the resulting sample
cycle, the signal is indistinguishable from that of spectrum.
another, lower frequency, a phenomenon known as
aliasing. Optical or electrical filtering is therefore
needed to remove any frequencies that are above the
Nyquist limit (Figure 8).
The resolution in the spectrum depends on the
spacing of the frequencies provided by the Fourier Thus, the spectrum generated by the Fourier trans-
transform. This is determined by the length of the form contains wavenumbers separated by the inverse
interferogram. The frequencies used in a Fourier of the maximum path difference in centimetres.
transform are those such that an exact number of Instruments for routine analytical work typically
cycles fits the length of the interferogram from zero have highest resolution of 1 or 0.5 cm1, requiring
to the maximum OPD, (in cm). Two adjacent fre- a maximum path difference of 1 or 2 cm. This re-
quencies in the analysis, corresponding to wavenum- solution is constant across the spectrum. Research
bers 1 and 2 (cm1), have n and (n+1) cycles in this instruments have been built with resolutions as high
length. The wavenumbers are given by as 0.0009 cm1, which requires a path difference in
excess of 10 m.
In a practical instrument, the beam passing
through the interferometer cannot be perfectly paral-
lel. There are rays at different angles and these
From this it follows that the difference between the experience different changes in path length for a
wavenumbers is given displacement of the moving mirror. This

Figure 8 A signal sampled at less than two points per cycle

and a lower frequency with identical values at the sampling
Figure 6 The optical arrangement of a FT-IR spectrometer. points.
 IR SPECTROMETERS 1051

causes line broadening by an amount that is

proportional to the wavenumber. The range of
angles admitted to the interferometer has to be
restricted so that the required resolution is achieved.
Incorporating an aperture, called a Jacquinot stop, at
an appropriate point in the optical system controls
this. In some ways this aperture is analogous to the
slit in a dispersive spectrometer. For a given resolu-
tion, the optical throughput of a FT spectrometer,
limited by a circular aperture, can be higher than
that of a dispersive spectrometer with a rectangular
slit. This is referred to as the Jacquinot advantage Figure 9 A short length of an interferogram with the sampling
and contributes to the higher sensitivity of FT instru- points corresponding to zero-crossings of the laser signal.
ments. The size of this aperture depends on the reso-
lution and the wavenumber range. Lowering the in a single scan. The analogue-to-digital converter
resolution by a factor of 2 allows the energy reaching must be able to handle the very large centre burst
the detector to be doubled. However, in most spec- and the much smaller noise component. The preci-
trometers it is not possible to increase the energy sion needed for this is at least 16 bits.
above that permitted for 4 cm1 resolution because In an ideal interferometer, the complete spectral
of other limitations such as the range of the ana- information could be obtained by scanning the path
logue-to-digital converter. difference from zero to the value needed to achieve
the required resolution. However, data have to be
measured for a limited distance on both sides of the
Measuring the interferogram
nominal point of zero path difference in order to
To establish the wavelength scale of the spectro- carry out phase correction. For signal averaging, the
meter, the interferogram has to be measured at pre- data points corresponding to the same path differ-
cisely known intervals of the path difference. Gener- ence for successive scans have to be combined. The
ally this is achieved by using a heliumneon laser at laser signal measures changes in the path difference
0.6328 m (15 800 cm1) to provide a reference sig- but not its absolute value. In early FT instruments
nal. The laser radiation passes through the interfer- this was measured by including a broad-band white
ometer to a separate detector. This produces a light source, the interferogram from which is a very
sinusoidal signal with a period equal to the wave- sharp spike at zero path difference. The optics were
length of the laser (Figure 9). Typically the interfero- arranged so that zero path difference for the white
gram signal is measured at each point where the laser light was offset from that for the IR radiation. The
signal passes through zero. With this spacing of the white-light signal was used to trigger data collection
data points in the interferogram, all wavenumbers (Figure 10). The interferometer scanned in one
below 15 800 cm1 can be distinguished. The laser direction and collected a complete interferogram on
wavelength is known very accurately and is very sta-
ble. As a result, the wavelength scale of FT spectro-
meters is much more accurate than that of dispersive
instruments. This superiority is known as the Connes
advantage after the originator of the laser referenc-
ing technique. The absolute accuracy of the scale
depends on the spectrometer resolution, while the
scan-to-scan precision is of the order of 0.01 cm1.
The time taken for an interferometer to scan from
zero to the maximum path difference is typically of
the order of 1 s. For most routine measurements, the
interferogram signal is averaged over several scans to
reduce random noise that is largely generated by the
detector. By averaging N scans the noise level in the
final spectrum can be reduced by a factor of N rela-
tive to that from a single scan. For the signal averag-
ing to work properly, the interferogram must be Figure 10 The use of a white-light interferogram to trigger data
digitized with sufficient precision to record the noise collection.
1052 IR SPECTROMETERS

one side of zero path difference only. After each Phase correction can be avoided by scanning the full
scan, the interferometer would move back past the length of the interferogram on both sides of zero
start position to begin the next scan. Such a scheme path difference. This allows a magnitude spectrum to
is described as unidirectional single-sided scanning be calculated without phase correction. The Fourier
(Figure 11). It was necessary to trigger each scan transform itself uses the fast Fourier transform (FFT)
separately because the path difference could not be algorithm devised by Cooley and Tukey. This algo-
tracked through the change of direction at the end of rithm requires that the number of points to be trans-
the scan. Most systems now use two detectors with a formed is equal to a power of 2. However, the
phase shift between them to monitor the laser signal, number of data points collected by scanning to a
allowing forward and reverse motion to be distin- path difference of 1 cm, for example, does not corre-
guished. This quadrature detection arrangement can spond to a power of 2. The interferogram is there-
track the path difference continuously and so control fore extended by additional data points with the
the co-addition of successive scans without individ- value of zero to give the required number of points.
ual triggering. It also allows the interferometer to This process is called zero-filling. It is a common
collect data from scans in both directions. This bidi- practice to apply further zero-filling to double or
rectional scanning achieves more efficient data col- quadruple the number of points used. This improves
lection as there is no fly-back period between scans the appearance of the final spectrum by increasing
(Figure 11). the density of the data points, although it does not
improve the resolution. In some systems, interpola-
tion in the spectrum is used to give results that are
Processing the interferogram signal
equivalent to the additional zero-filling.
The signal processing that converts the interfero- Apodization is an operation that is necessary
gram into a spectrum usually involves phase correc- because of the finite length of the interferogram.
tion, zero-filling and apodization as well as the Monochromatic radiation produces a sinusoidal
Fourier transform. An ideal interferogram would be interferogram, but the length of this is limited by the
perfectly symmetrical about zero path difference. maximum path difference in the interferometer scan.
However, in practice there is some asymmetry The Fourier transform of such a signal is a line of
because of phase differences between the signals finite width with oscillations on either side. This line
from different wavenumbers. A phase correction has the form of a sinc function ((sin x/x))
routine is applied before the Fourier transform. (Figure 12). The oscillations (or side-lobes) that

Figure 11 Alternative scanning and data collection schemes.

 IR SPECTROMETERS 1053

Figure 13 Some common apodization functions and the re-

sulting line shapes.

Figure 12 The sinc function generated by Fourier transfor-

mation of a cosine wave of finite length.
The scanning mechanisms of interferometers have
accompany each line are obviously a problem to be designed to minimize any change in alignment
because they could obscure smaller features in the during the scan that would introduce noise or distor-
spectrum. They can be reduced at the expense of tion into the spectrum. In systems where a plane
some broadening of the lines by multiplying the mirror moves linearly, the problem is to avoid any
interferogram with a function that decreases with tilting of the moving mirror. In some spectrometers
increasing path difference. This process is called the alignment is monitored and corrected continu-
apodization. Various functions are used for this, rep- ously in a process called dynamic alignment. Multi-
resenting different compromises between line broad- ple detectors measure the laser reference signal from
ening and the reduction of the side-lobes. different points in the beam. Any mirror tilt intro-
Historically, a simple triangular function was often duces a phase shift between the signals from the
used, but other functions can reduce the side-lobes detectors because the path differences are not equal.
further with less broadening. By examining a very This is automatically corrected in real time by using
large number of alternative functions, Norton and piezoelectric transducers to tilt the non-scanning
Beer established empirical limits for the reduction in mirror. The problem can also be addressed by using
side-lobe intensity that could be achieved for a given a cube-corner reflector instead of a plane mirror.
increase in line width. Three functions identified by Tilting the cube corner does not alter the direction of
Norton and Beer are in common use, weak, medium the reflected radiation but lateral movement would
and strong (Figure 13), corresponding to increasing affect the alignment. There are also scanning
degrees of side-lobe reduction and line broadening. arrangements that completely avoid alignment errors
The use of no apodization in order to achieve the
narrowest possible lines is called boxcar truncation.

Scanning mechanisms
To create interference, the beams returned from the
two mirrors in the interferometer must overlap exact-
ly as they meet at the beam splitter. Any change in the
alignment of the mirrors reduces the efficiency of the
interference by an amount that increases with the
square of the wavenumber. Small mechanical chang-
es, for example associated with thermal expansion,
can cause the alignment to change slowly. Changes
that occur between measurement of the background
and sample spectra result in a curved baseline Figure 14 A parabolic baseline caused by misalignment of the
(Figure 14). interferometer.
1054 IR SPECTROMETERS

Figure 16 The instrument profile for a mid-lR spectrometer

and the resulting variation of the noise level.
Figure 15 A scanning mechanism with a rotating mirror pair.

during scanning. One of the simplest generates the range. In consequence, the noise level in the final spec-
path difference by rotating a pair of parallel mirrors trum increases at each end.
in one arm of the interferometer. The mirror pair
does not alter the direction of radiation passing Detectors
through it, whatever its orientation, so that it cannot
affect the alignment as it rotates (Figure 15). For routine measurements the most common detector
Some commercial instruments use a scanning material is deuterated triglycine sulfate (DTGS). This
mechanism that does not involve moving mirrors. is a thermal detector, the signal being generated in
Changing the refractive index instead of the physical response to the change in temperature caused by
length can alter the optical path. By moving a wedge absorption of the IR radiation. It is generally operated
of a material such as potassium bromide (KBr) into at a temperature close to ambient. The response time
one arm of an interferometer and displacing air, the of these detectors is less than 1 ms, allowing them to
optical path is increased. Because the mechanical tol- follow changes occurring at rates up to several kilo-
erances are less severe than for simple moving-mirror hertz. In spectrometers with DTGS detectors, the
systems, these instruments are mostly used in non- OPD is typically scanned at about 0.3 cm s1 because
laboratory applications. One disadvantage is that the faster scanning would reduce the response of the
refractive index varies with wavelength so that the detector. The time taken to scan the 0.25 cm needed
wavenumber scale is not inherently linear as it is in for a spectrum at 4 cm1 resolution is therefore
conventional instruments. approximately 1s. These detectors have excellent lin-
The range covered by a spectrometer depends pri- earity over a wide range of signal levels. Lithium tan-
marily on the properties of the source and the beam talate is an alternative room-temperature detector
splitter, and is sometimes limited by the detector. The material with somewhat lower performance but also
sources used in mid-IR spectrometers typically lower cost.
operate at temperatures around 1300 K. Their output When higher speed or sensitivity is needed, liquid
has a fairly similar profile to a black body and so pro- nitrogen-cooled MCT (mercury cadmium telluride)
vides significant energy across the entire mid-IR detectors are used. These are photon detectors in
region. An ideal beam splitter would provide uniform which electrons are excited directly by the absorp-
reflectivity of 50% and no absorption across the tion of radiation. Cooling to liquid-nitrogen temper-
region of interest. In practice, a single beam splitter ature is needed to avoid excitation of electrons by
made from KBr with a multilayer coating can be used thermal motion. These detectors do not cover the
from the near IR to about 400 cm1, where the KBr complete IR range because there is a minimum
absorbs strongly. Most interferometer designs incor- energy needed to excite an electron, corresponding
porate a compensator plate so that both beams pass to a long-wavelength limit. Changing the composi-
through the same thickness of KBr. A FT spectro- tion can vary this limit, but extending the range
meter can cover the entire mid-IR region in a single reduces the sensitivity. The highest sensitivity is
measurement. As can be seen from Figure 16 the obtained with narrow-band detectors, which have a
single-beam energy falls off towards the ends of the limit of about 750 cm1, while wide-band detectors
 IR SPECTROMETERS 1055

Table 1

Range
Region (cm1) Source Beam splitter Detector
Mid IR 7 000400 Ceramic, KBr, ZnSe DTGS,
wire,Sic LiTa,
MCT,
InSb
Near IR 15 0004 000 Tungsten CaF2, quartz DTGS,
halogen PbS,
InGaAs,
InSb
Far IR 40020 Ceramic, Mylar, wire grid, DTGS, Si
wire, Csl (>200 cm1)
mercury
arc
Figure 17 The instrument profile with a cooled MCT detector
showing linear and nonlinear behaviour.

reach 450 cm1, close to the limit for KBr optics. to the fixed mirror. In practice, a constant path dif-
Scan rates are not limited by the detector response, ference can be maintained either by moving the two
and indeed the detector performance improves as mirrors synchronously or by holding both station-
the scan rate is increased. This allows data to be ary. Step-scanning instruments can address a much
acquired at rates up to 100 scans per second in wider range of dynamic measurements than continu-
commercial instruments. The response of MCT ously scanning systems because the signal modula-
detectors is inherently nonlinear (see Figure 17). tion frequencies are not restricted by the velocity of
This restricts their use to low signal levels, typically the moving mirror.
less than 10% of the energy available. However this The principal applications have been to very rapid
is not a problem in applications where the signal kinetics, photoacoustic spectroscopy and spectro-
level is restricted, such as microscopy. Because of the scopic imaging with array detectors. The kinetic
nonlinearity the centre burst signal is compressed rel- applications involve processes that are repeated for
ative to the rest of the interferogram. After Fourier each step of the path difference. The signal can be
transformation, the primary consequence is an offset measured at time intervals limited only by the detec-
in transmission, so that totally absorbing bands do tor response and electronics. At each path difference
not appear at zero. The presence of nonlinearity can measurements are made at the same times after initi-
be detected by examining the single-beam signal in ation of the process. The data are then reshuffled to
the region beyond the long-wavelength limit of the create a series of interferograms in which each point
detector. Any nonlinearity causes this signal to be was measured at the same time into the process. This
nonzero. Various methods, both in electronics and in approach has allowed FT-IR measurements to reach
software, have been employed to minimize the the nanosecond timescale. In photoacoustic spectro-
effects of nonlinearity. None of these is entirely suc- scopy there are advantages because the modulation
cessful. For this reason, room-temperature detectors frequency is constant across the spectrum, making it
are preferred for quantitative measurements. feasible to measure the spectra of subsurface layers.
The combinations of components found in use for In the imaging application, the sample is focused
the three regions are summarized in Table 1. onto an array detector. The signal from each pixel in

Step-scan instruments
In most spectrometers the path difference is scanned
continuously, but so-called step-scan instruments
have become important for some research applica-
tions (see Figure 18). In these instruments the path
difference is held constant at one point while the
interferogram signal is measured before being moved
to the next point. The IR signal has to be modulated
to match the frequency response of the detector. This Figure 18 The difference between continuous and step
is achieved by applying a small-amplitude oscillation scanning.
1056 IR SPECTROMETERS

the array can be read and averaged at each path dif- to 5 cm1. Although there have been relatively few
ference. This is not feasible with continuous scan- reported applications, some spectrometers are also
ning spectrometers because the array cannot be read capable of measuring throughout the visible region.
out sufficiently fast. With a step-scan spectrometer a The range of scan speeds in research instruments
64 64-pixel image containing 4096 spectra can be is greater than in routine instruments. The higher
generated in a few minutes. speeds provide better performance with cooled
detectors and for kinetics, and the lower speeds are
used with photoacoustic detectors. Digital signal
Commercial instruments processing is increasingly being used to facilitate
Commercial spectrometers can be grouped into experiments involving external modulation in which
those intended for routine measurements and two channels of information have to be acquired
research instruments. In general, the routine instru- simultaneously. Examples include polarization mod-
ments are designed to simplify operation while ulation in reflectionabsorption of species on metal
research instruments aim to offer versatility. In virtu- surfaces and in measuring the dynamic dichroism
ally all cases the instrument is controlled by a per- associated with polymer stretching. Research instru-
sonal computer that handles an increasing amount of ments generally have both continuous-scanning and
the digital signal processing in the system. step-scan modes as discussed earlier.
A typical routine mid-IR spectrometer has KBr Since the introduction of the first commercial FT
optics, best resolution of around 1 cm1, and a room- instrument in the late 1960s, there have been great
temperature DTGS detector. Noise levels below improvements in speed and sensitivity. Routine
0.1%T peak-to-peak can be achieved in a few sec- measurements using the common sampling tech-
onds. To reduce the problem of fluctuating levels of niques now take only seconds, so that further
atmospheric water vapour and carbon dioxide, improvements in performance will bring little bene-
much of the optical path is sealed and desiccated. fit. Future advances are more likely to be seen in the
The sample compartment will accommodate a development of smaller and more robust spectrome-
variety of sampling accessories such as those for ters. There are areas where improved performance is
ATR (attenuated total reflection) and diffuse reflec- desirable, such as the sensitivity of chromatographic
tion. In most cases the output from the interferom- interfaces and the spatial resolution of IR micro-
eter can be directed through an external port, scopes. However, there has been little recent
principally for coupling to an IR microscope. Other progress and there is little prospect of significant
external accessories provide coupling to a gas advances in these areas. The advent of step-scan
chromatograph or thermogravimetric analyser. spectrometers has led to great advances, but these
These instruments often have versions for near-IR have been in rather specialized applications, apart
operation in which the source, beam splitter and from imaging. Thus FT-IR instrumentation can be
detector are different. classified as mature. There is little prospect of any
Research spectrometers are generally larger, per- alternative methods of measuring IR spectra replac-
mitting higher optical throughput and hence better ing FT instruments for most purposes.
signal-to-noise performance, and higher resolution.
There may be multiple external ports for the interfer- List of symbols
ometer output so that several experiments can be
installed simultaneously. The major components n = number of cycles in ; N = number of scans that
source, beam splitter and detector are generally are averaged; %T = percent transmission; =
exchangeable, so that a wide spectral range can be maximum optical path difference; = frequency.
covered. A common arrangement is to have alterna-
tive sources and detectors permanently installed and See also: ATR and Reflectance IR Spectroscopy,
selectable by switching mirrors. Different ranges can Applications; Chromatography-IR Applications;
then be selected simply by changing the beam split- Chromatography-IR, Methods and Instrumentation;
ter. For far-IR operation, the beam splitters are Far IR Spectroscopy, Applications; Fourier Transfor-
generally Mylar films, which have a narrow range, mation and Sampling Theory; High Resolution IR
or metal grids on a polymer substrate. Cooled silicon Spectroscopy (Gas Phase), Applications; High Reso-
bolometers are the detectors of choice for high reso- lution IR Spectroscopy (Gas Phase), Instrumentation;
lution and sensitivity. Some instruments have a IR Spectral Group Frequencies of Organic Com-
vacuum version that removes the very strong water pounds; IR and Raman Spectroscopy of Inorganic,
absorption in this region, allowing measurement out Coordination and Organometallic Compounds; IR
 IR SPECTROMETERS 1057

Spectroscopy, Theory; Near-IR Spectrometers; Sur- Haseth, this describes the optical layouts of both dis-
face Studies by IR Spectroscopy. persive and FT commercial spectrometers.)
Griffiths PR and de Haseth JA (2000) Fourier Transform
Infrared Spectroscopy. 2nd edn. New York: Wiley.
(This is the standard book on the subject with detailed
Further reading discussion of the principles of operation and design of
FT-IR spectrometers.)
Analytical Chemistry (1995) 67: 381A385A, (1998) 70: Mackenzie MW (ed) (1989) Advances in Applied Fourier
273A276A. (Reviews of currently available commer- Transform Spectroscopy. Chichester: Wiley.
cial laboratory spectrometers appear every few years in Persky MJ (1995) Review of Scientific Instruments 66:
this journal. These are the most recent.) 47634797. (A review of spaceborne Fourier transform
Chalmers JM and Dent G (1997) Industrial Analysis with spectrometers for remote sensing. This is an excellent
Vibrational Spectroscopy, RSC Analytical Spectroscopy review of non-laboratory instrumentation with a very
Monographs. Cambridge: Royal Society of Chemistry. clear summary of the difficulties. Numerous designs are
(More practically oriented than Griffiths and de described.)

IR Spectroscopic Studies of Hydrogen Bonding

See Hydrogen Bonding and other Physicochemical Interactions Studied By IR and Raman Spectroscopy.

IR Spectroscopy in Forensic Science

See Forensic Science, Applications of IR Spectroscopy.

IR Spectroscopy in Medicine
See Medical Science Applications of IR.

IR Spectroscopy of Surfaces
See Surface Studies By IR Spectroscopy.