Fig.1: The development of near-infrared spectroscopy in technology jumps.
Vibrational
Spectroscopy
A
Short Introduction
H. W. Siesler
Department of Physical Chemistry
University of Duisburg-Essen, Campus
Essen
D 45117 Essen, Germany
General Remarks
Although this homepage is dedicated to near-infrared (NIR) spectroscopy, its theory, instrumentation and applications can only be understood properly in the context of an overview of all the available vibrational spectroscopic techniques. Based on this knowledge, the potential of NIR spectroscopy can then be much better evaluated and appreciated in terms of its implementation as a quality- and process-control tool in an industrial environment.
At this point critics as well as enthusiastic users of this technique are addressed. Near-infrared spectroscopy - especially in combination with chemometric evaluation routines - is neither a black-box instrumentation nor should it be used for any analytical problem in an overenthusiastic fashion without prior detailed feasibility studies. Most disappointments with and opposition against this technique can be traced back to a negative experience based on a surficial, quick and dirty application trial which inevitably led to disastrous results. Furthermore, it should always be kept in mind, that in the majority of cases NIR spectroscopy requires reference techniques to build up calibration routines and to guarantee a proper maintenance of an established calibration with reference to outlier detection and troubleshooting. Finally, it should be emphasized, that near-infrared spectroscopy is not only a routine tool but has also a tremendous research potential, which can provide unique information not accessible by any other technique.
Historically, the discovery of near-infrared energy is ascribed to F. W. Herschel in 1800 /1/. In a recent meeting of the Royal Society of Chemistry the 200th anniversary of this scientific milestone was celebrated in London /2/. As far as the development of instrumentation and its break-through for industrial applications in the second half of the 20th century was concerned, it proceeded in technology jumps (Fig.l). In this respect, large credit has to be given to researchers in the field of agricultural science, foremost K. H. Norris /3,4/, who have recognized the potential of this technique already in the very early stage of development. At the same period, with few exceptions /5-7/, comparatively low priority has been given to NIR spectroscopy in chemical industry. This situation is also reflected by the fact, that the near-infrared spectral range was for a long time only offered as a low- or high-wavenumber add-on to ultraviolet-visible (UV-VIS) or mid-infrared (MIR) laboratory spectrometers. This situation has dramatically changed since about 1980 when stand-alone NIR instrumentation became widely available. Nevertheless, it needed another decade of intense conviction phase before near-infrared spectroscopy became a generally accepted technique. Since the early nineties the availability of efficient chemometric evaluation routines, light-fiber optics coupled with specific probes for a multitude of purposes and the fast progress in miniaturization - based on new monochromator/detection designs - has launched NIR spectroscopy into a new era for industrial quality- and process-control.
Fig.1: The
development of near-infrared spectroscopy in technology jumps.
The following short overview on mid-infrared, near-infrared and Raman spectroscopy
is intended to provide a minimum basis in order to put the available vibrational
spectroscopies in a perspective for their practical applications. Over the last
years these techniques have developed to indispensable tools for academic research
and industrial quality control in a wide field of applications ranging from
chemistry to agriculture and from life sciences to environmental analysis. In
what follows, the physical principles and some instrumental aspects of MIR,
NIR and Raman spectroscopy will be discussed and the advantages and disadvantages
of the individual techniques will be outlined with reference to their implementation
as industrial routine equipment. For more detailed information the interested
reader is referred to the pertinent literature /8-15/.
While mid-infrared spectroscopy has been a well-established research tool and structure elucidation technique in the academic as well as industrial environment over many decades, Raman spectroscopy was restricted for a long time primarily to academic research and has gained a broader industrial recognition only in combination with the Fourier-Transform (FT) technique and NIR-laser excitation in the late eighties. Just recently dispersive Raman spectroscopy with charge-coupled device (CCD) detection is undergoing a further renaissance with reference to process-control applications /9,11/. On the other hand, NIR spectroscopy has been used only occasionally since the early fifties for chemical analysis, however, more frequently in the field of agriculture, but a real break-through as quality- and process-control tool occurred only within the last decade since the introduction of efficient chemometric evaluation techniques and the development of light-fiber coupled probes. In actual fact, despite the lack of comparable specific spectral information, NIR spectroscopy is quickly overtaking Raman and primarily mid-infrared spectroscopy as a process-monitoring technique /10,11/. The main reason is the much easier sample presentation and the possibility to separate the sample measuring position and the spectrometer by light fibers over distances of several hundred meters. Although similar arguments hold for Raman spectroscopy, the restriction to only small sample volumes, comparatively low sensitivity, interference by fluorescence and safety arguments are still limiting its industrial application on a real broad scale.
Basic Principles of Vibrational Spectroscopy
Although the three spectroscopic techniques are very different in several aspects, their basic physical origin is the same: absorption bands in the MIR, NIR and Raman spectra of chemical compounds can be observed as a consequence of molecular vibrations. Considering the simple diatomic oscillator of Fig.2, the vibrational frequency ν (based on a harmonic oscillator approximation) can be correlated with molecular parameters by:
 |
(Eq.l) |
where the force constant f reflects the strength of the bond between m and M and the reduced mass µ is given by:
 |
(Eq.2) |
Thus, the vibrational frequencies are very sensitive to the structure of the investigated compound and this is the basis for the wide-spread application of MIR (and less frequently) Raman spectroscopy for structure elucidation.
Fig.2: Specific features of Raman, mid-infrared
and near-infrared spectroscopy (see text).
Furthermore, due to the different excitation conditions of MIR, NIR and Raman spectroscopy, the relationships between the absorption intensities and the addressed functionalities of the molecules under examination vary significantly. These specific features lead to extremely different responses of the same molecular vibration to the applied technique. Figure 2 summarizes the potential energy (V) transitions, excitation conditions and the most important instrumental as well as qualitative and quantitative aspects of the three spectroscopies. The principal difference between the three techniques is, that Raman spectroscopy is a scattering technique whereas mid- and near-infrared spectroscopy are absorption techniques. Thus, whereas scanning MIR and NIR spectrometers operate with a polychromatic source from which the sample absorbs specific frequencies corresponding to its molecular vibrational transitions (mostly fundamental vibrations for the MIR and overtone or combination vibrations for the NIR), in Raman spectroscopy the sample is irradiated with monchromatic laser light whose frequency may vary from the VIS to the NIR region. This radiation excites the molecule to a virtual energy state which is far above the vibrational levels of this anharmonic oscillator for a VIS laser and in the range of high overtones for NIR laser excitation (see also Fig.4). From this excited energy level the molecule may return back to the ground state by elastic scattering thereby emitting the Rayleigh line which has the same frequency as the excitation line and does not contain information in terms of the molecular vibration (this case is not shown). If it returns only to the first excited vibrational level by inelastic scattering, the emitted Raman line (so-called Stokes line) has a lower frequency and the frequency difference to the excitation line corresponds to the vibrational energy of the MIR absorption. Also not shown is the case of the anti-Stokes line, where the starting level is the first excited vibrational state and the molecule returns to the ground state by inelastic scattering, thereby emitting a Raman line of higher frequency (here too, the frequency difference to the excitation line corresponds to the vibrational energy of the MIR absorption) but of lower intensity compared to the Stokes line, due to the lower population probability of the excited state (law of Boltzmann). Commonly, the Stokes lines are used for practical Raman spectroscopy and refer to the same vibrational transitions as MIR spectroscopy. One of the limiting factors for the application of the Raman technique, however, becomes evident by comparing the intensities of the source and the scattered radiation:
 |
(Eq.3) |
From these figures it can readily be derived, that a sensitive detection of the Raman line alongside an efficient elimination of the Rayleigh line are experimental prerequisites for the successful application of Raman spectroscopy. As shown in Fig.2, Raman and MIR spectroscopy cover approximately the same wavenumber region with the Raman technique extending somewhat further into the far-infrared (FIR) region (ca. 50 cm-1) due to instrumental reasons (primarily because of the MIR detector cut-off). In some cases, this additional frequency range is valuable, since it often contains absorptions of lattice modes of molecular crystals which may be very characteristic for a specific polymorph.
Another important relation for the comparison of VIS- versus NIR-Raman spectroscopy, is the dependence of the scattered Raman intensity IRaman on the fourth power of the excitation frequency νexc:
 |
(Eq.4) |
The consequences of
this relationship with reference to the application of VIS- or NIR-Raman spectroscopy
for an individual problem will be outlined below.
NIR spectroscopy covers
the wavenumber range adjacent to the MIR and extends up to the VIS region. NIR
absorptions are based on overtone and combination vibrations of the investigated
molecule and due to their lower transition probabilities, the intensities usually
decrease by a factor of 10-100 for each step from the fundamental to the next
overtone /10/ Thus, the intensities of absorption bands successively
decrease from the MIR to the visible region, thereby allowing an adjustment
of the sample thickness (from millimeters up to centimeters), depending on the
rank of the overtone. This is a characteristic difference to MIR and Raman spectra,
where the intensities of the fundamentals vary irregularly over the whole frequency
range and depend exclusively on the excitation conditions of the individual
molecular vibrations. Thus, for a Raman band to occur, a change of the polarizability
α (the ease to shift electrons) has to take place during
the variation of the vibrational coordinate q. Alternatively, the occurrence
of an MIR band requires a change of the dipole moment µ during the vibration
under consideration. These different excitation conditions lead to the complementarity
of the Raman and MIR technique as structural elucidation tools, because Raman
spectroscopy predominantly focuses on vibrations of homonuclear functionalities
(e.g. C=C, C-C, S-S) whereas the most intense MIR absorptions can be traced
back to polar groups (e.g. C-F, Si-O, C=O and C-O). In a simplified form, NIR
spectroscopy requires - additionally to the dipole moment change - a large mechanical
anharmonicity of the vibrating atoms (see Fig.2) /16/. This
becomes evident from the analysis of the NIR spectra of a large variety of compounds,
where the overtone and combination bands of CH, OH and NH functionalities dominate
the spectrum whereas the corresponding overtones of the most intense MIR fundamental
absorptions are rarely represented. One reason for this phenomenon is certainly
the fact, that most of the X-H fundamentals absorb at wavenumbers > 2000
cm-1 so that their first overtones appear already in the NIR frequency
range. The polar groups leading to the most intense fundamental absorptions
in the MIR on the other hand, absorb usually at wavenumbers < 2000 cm-1,
so that their first (and sometimes higher) overtones still occur in the MIR
region. Due to the intensity loss for each step from the fundamental to the
next overtone, the absorption intensities of these functionalities have become
negligible by the time they occur in the NIR range.
The superposition of
many different overtone and combination bands in the NIR region causes a very
low structural selectivity for NIR spectra compared to the Raman and MIR analoga
where many fundamentals can usually be observed in isolated positions. Nevertheless,
also NIR spectra should be assigned as detailed as possible with reference to
their molecular origin /17/ thereby allowing a more effective
application for research purposes and combination with chemometric evaluation
procedures. For the assignment of overtones and combination bands in the NIR
to their corresponding fundamentals in the MIR it is recommendable to use the
wavenumber notation instead of the wide-spread wavelength (nm or µm) scale.
It should be mentioned, however, that the wavenumber positions of the overtones
deviate with increasing multiplicity from the exact multiples of their fundamentals
due to the anharmonicity of the vibrations
/10,15-17/.
As far as the quantitative
evaluation of vibrational spectra is concerned, MIR and NIR spectroscopy follow
Beer's law whereas the Raman intensity is directly proportional to the concentration
of the compound to be determined (Fig.2). To avoid compensation problems, in
most cases quantitative Raman spectroscopy is performed with an internal reference
band in the vicinity of the analytical absorption.
An important issue
for the implementation of a technique as an industrial routine tool is the sample
preparation required for this technique. In this respect it can be seen from
Fig.2, that Raman and NIR spectroscopy have considerable advantages over MIR
spectroscopy, which usually requires individual sample preparation steps before
data acquisition. Only the technique of attenuated total reflection (ATR) /18/
circumvents time-consuming sampling procedures for MIR spectroscopy.
Instrumentation
Figure 3 summarizes the present state of the most frequently used monochromator/detection principles for the different scanning spectroscopies.
Fig.3: The current monochromator/detection principles of scanning MIR, NIR and Raman spectrometers.
As mentioned above, in Raman spectroscopy two techniques are presently in current use:
Fig.4: Fluorescence and scattering
efficiency in NIR- and VIS-Raman spectroscopy /22/.
Both alternatives involve
compromises and the choice of the applied technique depends on the individual
problem. In Fig.4 the trends of the main limiting factors - fluorescence and
low scattering efficiency - have been outlined with reference to the two excitation
mechanisms. If a molecule is irradiated with visible radiation, it may be excited
to an energy level of the next-higher electronic state. Return to the ground
state or an excited vibrational level of the original electronic state can easily
proceed via fluorescence as shown in Fig.4. Thus, for a large proportion of
samples, irradiation with visible light causes strong fluorescence by additives
or impurities (or by the sample itself) which will superimpose and in many cases
,,inundate" the Raman spectrum of the sample. As an example, the effect
of fluorescence in the Raman spectrum of thioindigo measured with Ar+-ion
(488nm) laser excitation compared to the fluorescence-free Raman spectrum obtained
by NIR laser (Nd:YAG, 1064nm) excitation is demonstrated in Fig.5. A number
of time-consuming approaches have been proposed to circumvent the problem /14/,
including prior purification or ,,burning out" the fluorescence. These
approaches have at best been only partially successful. Additionally, highly
coloured samples may absorb the Raman photons thereby preventing them from reaching
the detector and leading to thermal degradation of the investigated material.
The use of an NIR laser excitation confers a number of advantages on a Raman
system. Both, fluorescence and self-absorption of the Raman signal are very
much reduced and due to the lower energy of the excitation radiation thermal
degradation is also less of a problem. However, these advantages are partly
neutralized by the disadvantages of using a low-frequency laser as source. The
NIR-Raman technique is obviously less sensitive due to the ν4-dependence
of the scattering efficieny (see also Fig.4). Thus, a shift of the excitation
line from the VIS region (e.g. Ar+-ion laser, 488 nm/20492 cm-1)
to the NIR region (1064 nm/9398 cm-1) reduces the scattering intensity
within the relative Raman region from 0 to 4000 cm-1 for factors
between 23 and 87, respectively /19,20/.
As shown in Fig.3, however, NIR-Raman spectroscopy is performed on FT-spectrometers
and the sensitivity loss can be quite efficiently compensated by accumulation
of multiple scans. As a valuable compromise to suppress fluorescence and at
the same time retain an acceptable scattering efficiency, excitation with a
diode laser at 785 nm (12739 cm-1) is increasingly used /9/.
Fig.5: Raman spectra of thioindigo (top: Ar+-ion laser excitation, bottom: NIR laser excitation) /20/.
As far as MIR spectroscopy
is concerned, today almost exclusively FT-based instruments are in routine use
(Fig.3). Besides the optical principle of the Michelson interferometer also
polarization interferometers based on a stationary and a moving MIR-transparent
(KBr) wedge have been integrated in scanning spectrometers /9/.
Contrary to Raman and MIR spectroscopy, scanning NIR spectroscopy offers
the largest multiplicity of monochromator principles. Thus, apart from different
designs with moving parts, such as grating instruments and FT-spectrometers
with Michelson and polarization interferometers (with NIR-transparent quartz
wedges) two fast-scanning approaches with no moving parts are available: diode-array
systems and acousto-optic tunable filters (AOTF) /9,10/.
In this context it should be mentioned that miniaturization has certainly progressed
most significantly for NIR spectroscopy. A similar trend is also observable
for VIS-Raman CCD spectrometers.
Process-Monitoring
In the last two rows
of Fig. 2 the most important aspects for the implementation of the individual
spectroscopies as process-monitoring tools are addressed. The very small representative
sample volume or thickness in Raman and MIR/ATR spectroscopy may certainly lead
to problems if special care is not taken to avoid the formation of a stationary
layer on the reactor window or on the ATR crystal. In this respect, NIR spectroscopy
is the method of choice in view of the comparatively large sample volume/thickness
involved in these measurements. The ability, to separate the spectrometer from
the point of sampling is certainly a great advantage for Raman and NIR spectroscopy.
Although light-fibers based on ZrF4 and AgCl are also available for
MIR spectroscopy, it should be mentioned that their cost, attenuation properties
and mechanical/chemical stability are still inferior compared to the well-established
quartz fibers /10/. Specific probes are available for all
three techniques. Especially NIR spectroscopy offers a wide range of in-line
and at-line transmission and diffuse-reflection probes designed for the measurements
of liquids and solids. Large differences can also be identified with respect
to the ability of measuring aqueous solutions. Water is an extremely strong
absorber in the MIR and also a strong NIR absorber thereby limiting the available
wavenumber regions in both techniques. In contrast, it is a weak Raman scatterer
and it is recommended to consider Raman spectroscopy as an analytical tool for
aqueous solutions. Care has to be taken, however, with the NIR-Raman FT-technique
(1064nm), because due to the absorption of the water overtone vibration the
Raman spectrum may be modified relative to the VIS-laser excited Raman spectrum
/21/.
In conclusion, over
the last years MIR, NIR and Raman spectroscopy have been further developed to
a point where each technique can be considered a candidate for industrial quality-control
and process-monitoring applications. However, adding up the specific advantages
and disadvantages of the individual techniques, NIR spectroscopy is certainly
the most flexible alternative.
References
/1/
W. Herschel, Phil.
Trans. Roy. Soc.(London), 284 (1800)
/2/ A. M. C. Davies, NIR
news, 11(2), 3 (2000)
/3/ A. W. Brant, A. W. Otte and
K. H. Norris, Food Tech., 5(9), 356 (1951)
/4/ G. S. Birth and
K. H. Norris, Food Tech., 12(11), 592 (1958)
/5/
W. Kaye, Spectrochim. Acta, 6, 257 (1954)
/6/ R. G. J. Miller
and H. A. Willis, J. Appl. Chem., 6, 385 (1956)
/7/ R. F. Goddu and
D. A. Delker, Anal. Chem., 32, 140 (1960)
/8/ B. Schrader (ed.), Infrared
and Raman Spectroscopy, Wiley-VCH, Weinheim (1995)
/9/ J. Coates, Appl.
Spectrosc. Revs., 33(4), 267 (1998)
/10/ H. W. Siesler, Y.
Ozaki, S. Kawata, H. M. Heise (eds.), Near-Infrared Spectroscopy,
Wiley-VCH, Weinheim (2002)
/11/ J. J. Workman
Jr., Appl. Spectrosc. Revs., 34(1&2), 1-89 (1999)
/12/
P. R. Griffiths and J. A. de Haseth, Fourier Transform Infrared Spectrometry,
Wiley-Interscience, New York (1986)
/13/ H. Günzler,
H. M. Heise, IR-Spektroskopie, VCH, Weinheim (1995)
/14/ P. Hendra, C. Jones and G. Warnes, Fourier
Transform Raman Spectroscopy, Ellis Horwood Ltd., Chichester (1991)
/15/ H. W. Siesler and K. Holland-Moritz, Infrared
and Raman Spectroscopy of Polymers, Marcel Dekker, New York (1980)
/16/ W. Groh, Makromol. Chem., 189, 2861 (1988)
/17/ P. Wu and H. W. Siesler, J. Near Infrared
Spectrosc., 7, 65 (1999)
/18/ F. M. Mirabella Jr. (ed.), Internal Reflection
Spectroscopy, Marcel Dekker, New York (1993)
/19/ H. W. Siesler,
Revue de l´Institut Francais du Petrole, 48 (3), 223 (1993)
/20/ A. Hoffmann, PhD-Thesis, University
of Essen (1991)
/21/ W.-D. Hergeth, Chemie Ingenieur Technik,
70(7), 894 (1998)
/22/
B. Schrader, Chemie in unserer Zeit, 31(5), 229 (1997)