1. When we started we promised a series of themes - well, like all the best ideas, this one has slipped a little. In Edition III we continue the subject raised last time - the atmosphere and the effect it has on our spectroscopy. As we previously discussed, the atmosphere (whether man-made or not) influences the background in infrared spectrometers and this can, in turn, affect our results. In this edition we expand a little on this subject and offer a few methods of solving "background problems". One of my ambitions for the Journal is to find really interesting instruments or applications and ask well qualified authors to tell us about them. Several people have already been invited and are hard at it, scribbling away. The first to arrive is a piece by Mike Pelletier. He works for Kaiser Optical, a company famous for their holographic filters. Recently the company announced a really novel Raman spectrometer incorporating several very unusual features. I am keen that IJVS does not publish commercials, so I asked Mike to write me a proper scientific account and to explain the more unusual features of the Kaiser machine. You can judge for me, but I hope you will agree that the article is interesting and definitely not an advertisement. Last autumn I had the honour of attending the 2nd Australian Conference on Vibrational Spectroscopy in Brisbane. The breadth of coverage and the quality impressed me enormously and I wandered round the poster sessions, asking the youngsters to produce pieces describing their work. Several are on their way and the first is included in this edition. The subject is specialised but the techniques used are not. The authors are telling us about an infrared band method of analysis on wool, or rather chlorinated wool. Their introduction describes the industrial processes involved and why the analysis is required. Even if you are not interested in wool (except for your socks!) and its chlorination, scan the paper - the approach may well be relevant to the work you do. In particular, they use ATR and then process the data quite extensively. Some readers may not be very comfortable with derivation, least squares and other mathematical procedures, others may not be familiar with ATR. Don't worry - we will be covering all of these matters in future editions of the Journal. Since writing last, I have been busy requesting articles and even whole editions for the future. A major piece on near infrared applications, another on optical materials, an edition devoted to chemometrics, another on inelastic neutron scattering - all are promised, together with many others. The number of responses from readers is increasing - please let this continue. We really have no other way of knowing what you think of the Journal. Please tell us - even if you think it is rubbish!! P. J. HENDRA Editor's Note Last time we began to explore the question of backgrounds in infrared spectroscopy. What is the effect of the atmosphere? How does water vapour absorption affect our results? In Edition II I made the point that since water vapour absorbs, it must have an effect on our spectra - if only to lower the available energy or produce interfering peaks. I then reported to readers on a survey we had made on good practice in infrared labs and showed how some labs will accept remarkably poor atmospheric atmospheres, others demand an environment inside their instruments that is essentially completely dry. Every time the latter practitioners open their instruments to introduce a sample, they allow the ingress of wet air and this can be hard to remove. Further, as it is removed the background continuously changes, opening up more potential problems. I therefore asked Bill Maddams to consider the "background problem". In the article which follows, he shows how important a stable background is, particularly if quantitative work is intended. Bill then goes on to discuss the effect the characteristics of the instrument has on quantitative results. Much more can be said on both these subjects and I plan further articles. But enough of my scribbling - here is Bill's piece.
Bill Maddams All the absorbers that lie in the infrared beam of a spectrometer will give rise to characteristic absorption peaks - and this means not only the sample but anything else that is in the beam - the atmosphere, coatings on mirrors, beam splitters or windows, anything! The 'instrumental' absorbers and those due to the atmosphere may well overlap those of the sample. The group of sharp infrared peaks from water vapour in the region 1400 - 1800cm-1 due to vibration rotation transitions, and the broader ones* from carbon dioxide at about 670 and 2340cm-1, were well known to spectroscopists four or five decades ago, when single beam dispersive spectrometers were the norm. The water vapour peaks posed greater problems than those due to CO2, as they cover the wavelength range where peaks characteristic for C=C and C=O lie, causing difficulties in analytical work, particularly in the estimation of relatively low concentrations of carbonyl-containing compounds. With the advent of double beam dispersion spectrometers in the late 50's,
such problems seemed to disappear. In these machines, and there were two types, the
so-called null balance machines and the more exotic ratio recording instruments, two beams
were created, one labelled sample and the other reference. The intensity of the radiation
passing through each beam at each infrared frequency was in effect compared. If the
atmosphere attenuated the total optical path the two beams, sample and reference, were
equally affected and no consequence was noted. If an infrared absorbing sample was present
in the sample beam then
*These too are closely spaced sharp multiplets visible as such at high resolution.
An equivalent problem may occur, and frequently does, in Fourier transform infrared instruments. These machines operate as double beam IN TIME rather than IN SPACE as was familiar in the dispersion machines. By this we mean that a background is measured - in effect, the spectrum of the source as attenuated or exaggerated by the instrument. One then measures the spectrum with the sample in place and compares the two. Absorption present in the second spectra but not the first must, of course, then be due to the sample. The procedure adopted is a point-by-point ratioing. If the atmosphere absorbs strongly at frequency v, the ratio determined is that of Intensityv Sample versus Intensityv Background and one is ratioing almost nothing against very little - a somewhat imprecise ratio! All is far from satisfactory but manageable if purging is used and hence atmospheric absorption minimised, but if laboratory practice is sloppy, further problems can be encountered. Problems may well arise in practice because when purging, dry air or nitrogen are employed to reduce the water vapour concentration in the spectrometer. With most types of instrument the insertion of the sample into the sample chamber will allow the ingress of air into the sample chamber at a relative humidity as high as 70-80%. This water level will then decrease after the sample chamber is closed, because of the continuous purging, to a level approaching equilibrium within the spectrometer. In those situations where interference from water vapour peaks poses problems, it may be necessary to delay the spectral measurement until the water vapour level has fallen appreciably, or is not changing rapidly with time. This may set a limit to the speed with which measurements may be made. An example from the Malaysian Rubber Producers Research Association was included in Vol.1, Edition II. The effect of carbon dioxide present in the spectrometer is seldom considered, primarily because the CO2 peak at 2350cm-1 is in a position which is clear of the characteristic frequencies for most functional groups. However, there may be situations where it is necessary to monitor or regulate the levels of carbon dioxide within the spectrometer, e.g. if low levels of CO2 in gas samples are being determined. Mismatch between the CO2 levels between background and sample runs is possible for the same reasons as those discussed above in connection with water vapour absorption, and a small negative peak at 2350cm-1 is sometimes seen. This indicates a higher CO2 concentration during the background spectral measurement than is present when the sample is scanned. Bearing in mind that water vapour will gradually displace CO2 from any dessicant in the spectrometer, it is sensible to monitor the carbon dioxide level from time to time by running a background spectrum. High levels should be removed by purging. There is a second factor which influences quantitative measurements on gas samples and also on liquid/solid samples having sharp peaks, and is probably not widely appreciated. This is the effect of spectrometer resolution and its effect, at least in qualitative terms, is easily demonstrated by reference to Figs. 1 to 4.
Fig. 1 shows the measurement of absorption at a peak position The situation shown in Fig. 1 is idealised. It assumes that at It is convenient to think about the real-life situation in terms of a dispersive
instrument, although the principles involved are exactly the same for interferometric
instruments. If the spectrometer is set at
This is illustrated in Fig. 2 which, for simplicity, is shown as a
triangular function, although in practice it is likely to be non-linear. At
What is the effect of this on the measured absorbance of a peak such as
the one shown in Fig. 3? This can be assessed qualitatively by
superimposing Fig. 2 and Fig. 3, as shown in Fig. 4. Here the absorption peak has a half-width (width at half the
maximum height) of 10cm-1 and the transmission function
has a half-width of 5cm-1. At the absorption maximum, at
Fig. 6 shows how the absorption peak of Fig. 3, together with a sharper transmission function, with a half-width
of 3cm-1, resulting from higher spectrometer resolution.
If we again consider the data points It is clear from these results that the important factor is the half-width
of the absorption peak being measured relative to that of the transmission function. A
rough rule of thumb that may well prove useful in practice is that if the ratio is greater
than about five, the measured intensity reduction is quite small. The width of a spectral
band may, of course, be determined experimentally by measuring the spectrum of interest at
several resolution values, but this is rather time consuming. The pertinent question is,
what are the practical implications of the intensity reduction at the peak maximum? By
considering the matter in greater detail it may be shown that, for a given spectrometer
resolution and peak width, the reduction in absorbance increases at a rate greater than
linear with the absorbance of the peak being measured. Consequently, calibration curves of
absorbance vs. concentration are non-linear, and flatten out at higher concentrations and
absorbance values. This does not prohibit quantitative work, but the setting up of
non-linear calibration curves is more time consuming than for linear plots. In practice, it is sensible to work with a spectrometer resolution which is high enough to avoid markedly non-linear calibration curves. This, of course, will increase the required spectrometer scan time, for a given signal/noise ratio in the spectrum, and a compromise may be necessary. The problem will not arise to a marked degree in the case of measurements on liquid samples, although peaks from some types of vibrational modes, e.g. the out of plane deformation modes of the hydrogen atoms attached to aromatic rings, give sharp and analytically useful peaks which are narrow enough to give these non-linear calibration curves. Crystalline solids usually have peaks which are typically sharper than those of the materials when examined in the molten state or in solution. Gaseous samples, particularly of smaller molecules, e.g. water vapour and a range of atmospheric contaminants, show varying amounts of fine structure in their spectra and, ideally, should be examined with a resolution of 1cm-1 or better. So, to conclude this very preliminary opening essay on quantitative measurements - the atmospheric background can seriously affect the quality of the quantitative results we obtain (and can cause annoying interferences to qualitative results as well) and we must worry about resolution. In future editions we plan more detailed articles. REF: Int. J. Vib. Spect.,
[www.ijvs.com] 1, 3, 2 (1997) 3. Solving The Background Problem In The Mid-Infrared Patrick Hendra Several times throughout my career we have needed to make measurements in the 1600-1700 cm-1 area of the infrared or more rarely near 3000 cm-1 and have had serious problems with atmospheric absorption. In all cases we were using FT instruments flushed with dried air. The normally accepted practice is to run the background, introduce the sample into the chamber as quickly as possible and then wait 15 or 20 minutes before running a spectrum. It was assumed that during this period the wet air would be swept away but we found that the atmosphere in at least part of the instrument invariably contained a slightly different amount of water vapour than it did when the background was recorded. Further, try as we might, the instrument gradually became more contaminated as the day progressed and the situation less and less satisfactory. Let me tell you about one example of the work we were doing and you will
see the problem. When films of polyethylene are strained, in common with all
semi-crystalline polymers, massive irreversible stretching occurs to produce a highly
oriented sample with unique enhancements of their strength and stiffness. There is a major
research interest in working out what happens to the polymer molecules as the unoriented
material is deformed to an oriented one. Our approach was to deform and compare the i.r.
spectrum as deformation was increased. We were searching for C=O groups which we reasoned
would be produced when chains break under stress and the free radical ends oxidise. The
number of broken chains are expected to be very small, hence the concentration of C=O
groups must be minute. We wanted to detect these groups and to do so quantitatively. The
water vapour problem made the work impossible. To overcome the difficulty we first tried an air-trap approach. We bought some 'lay flat' tube of polythene film from our local DIY store. The 'lay flat' material was about 300mm wide. The sample area lid was removed and replaced by a simple metal sheet with a hole in it. The 'lay flat'* film was opened and taped around the hole and the sample then introduced through the tube. Once introduced, the end was rolled and clipped with a wide paper clip. The setup is shown below:
The arrangement is widely used and it works, but only in a limited sense. Every time a sample is introduced, a small amount of wet air inevitably gets in and therefore the background steadily deteriorates throughout the day. The deterioration is small and usually acceptable but the change in background is unacceptable if high sensitivity work is intended. Ratioing the background against the spectrum inevitably shows up water vapour bands and hence destroyed our results. We needed a system enabling us to change samples without the ingress of ANY air into the machine. An alternative is a method of coping with a background even though it is changing. Let me tell you about methods of tackling both. *Lay flat tubing is available from lab suppliers. Used with a strip welder, it is almost universally used in packaging. About ten years ago we developed a contraption at Southampton to allow us to introduce the sample and avoid contamination. The device, in effect, was a small elevator inside which was mounted a set of 2x3 inch 'industry standard' sample holding rails. The elevator fitted inside two concentric thin tubes, the inner of which was rigidly fixed and screwed to the sample area base. The outer tube came up through a lid and could be rotated. The device is illustrated below:
Winding the outer tube O moves the elevator E up and down and twists it at the same time. Hence, the holes cut through the tubes and the elevator to allow passage of the infrared beam can be allowed or obstructed.
Operation is as follows:
Thus - no delay is involved - the sample can be introduced and removed repeatedly and quickly. How does it work? One of my postgrads at the time in question was Jackie Ackavan. She claimed the device worked well enough that the background did not effectively change throughout a whole day of continuous use. To be honest (and supervisors have been known to be such) the favourable outcome may well have been influenced by the fact that the FTIR used an air bearing (an ancient Nicolet MX-1) hence the flow of dry air was copious. If a relatively slow purge is used, contamination might have occurred after many sample changes. Anyway - several of these devices were built and used by users of several different Nicolet instruments. Perhaps the sample changer needs re-invention! An alternative approach is to return to the old double beam technique - or rather the "ratio recording" rather than the "null balance" variants thereof. Some years ago, one of the commercial FTIRs was provided with a "shuttle". I am not sure whether it was a great commercial success but the idea was simple. After each scan of the interferometer, a solenoid operated and switched the "sample" for a "reference". Let us say the "reference" was nothing - just air - then the reference interferogram is that of the background. The "sample" interferogram (which must always be of lower intensity) is that of the background attenuated by the sample. The software puts each into a memory scan after scan, Fourier Transforms, ratios and comes up with a SAMPLE SPECTRUM. Now, this is exactly what we always do in our infrareds, the difference being that if the background is VARYING, the system allows for it. The old monochromator machines did this as a matter of routine, but the FTs can do the same if called upon. I have no idea whether all or only some of the manufacturers offered this
facility, but P-E
did on the 1700 series of instruments. Try asking your manufacturer what they can do. If
the answer is "Nothing", please FAX/e-mail the Editor and we will
see what can be done. 4. Integrated Raman Analyzer Using Volume Holographic Optical
Elements M. J. Pelletier Introduction Recent developments in optical, laser, and detector technology have made possible a new generation of Raman instruments. These Raman instruments meet the demands of industrial process analysis and control, and still provide the performance of state-of-the-art laboratory instruments. This report will describe one such instrument, the Kaiser Optical Systems HoloProbe.
A photograph of the integrated Raman system is shown in Figure 1. It consists of a computer, a base unit having a 23 inch by 18 inch footprint, and a probehead. The computer is connected to the base unit by a single high speed serial cable. The probehead is connected to the base unit by a single cable that contains two optical fibers. This fiber optic cable may be any length from 1.9 meters to greater than 100 meters. Raman spectra are collected by pointing the probehead at the sample, and clicking the acquire button.
A schematic diagram of the instrument is shown in Figure 2. Light from the laser is coupled into a probehead through an optical fiber. The probehead illuminates the sample and collects Raman scattered light from the sample. The Raman scattered light is sent back to the base unit through a second optical fiber. In the base unit, residual laser light is removed and the Raman light is spectrally dispersed onto a CCD detector. The signal generated from the image on the CCD detector is sent to the computer for conversion into a spectrum. HoloGRAMS software is used for instrument configuration and control. HoloGRAMS also works with other software packages such as GRAMS, Excel, or Spectacle through the Windows DDE interface. The HoloProbe design represents a significant departure from traditional Raman instrument design. Traditional laboratory Raman instruments based on triple spectrographs are extremely flexible instruments. They are designed to operate with a wide range of excitation wavelengths, spectral resolutions, and instrumental configurations. The price for this flexibility is compromised performance and increased complexity. The HoloProbe design gives up much of this flexibility in exchange for improved performance, simpler, more stable alignment, and easier coupling to the sample. The critical components and assemblies that make up the HoloProbe are described in greater detail in the following sections of this paper. CCD detector The CCD detector [1] is actually a two
dimensional array of nearly ideal detectors called pixels. Each pixel is a silicon
detector. Photons absorbed by the pixel silicon release electrons that accumulate in that
pixel. When the CCD detector is "read out", the total accumulated charge in each
pixel is determined. Values proportional to the charge in each pixel are stored in the
computer for further analysis. More detailed description of CCD detector design,
operation, and performance have been published [2-6]. In nearly all practical cases the noise from the CCD detector is much lower than shot noise, the statistical noise due to the quantum nature of light itself. The signal-to-noise ratio of Raman spectra collected with a CCD detector is therefore proportional to the square root of the number of detected photons. Raman photons are usually in short supply, so increasing the number of detected Raman photons will usually increase the Raman instrument sensitivity and quantitative accuracy. The standard CCD detector used in the HoloProbe is a rectangular array of 1024 by 256 pixels. Since each pixel is an independent detector, more than 250,000 independent measurements can be made simultaneously. The HoloProbe uses the massively parallel detection capability of the CCD detector to measure all of the wavelengths in the Raman spectrum simultaneously and continuously during the data acquisition time. In fact, a four channel adapter allows the HoloProbe to simultaneously measure the complete Raman spectra from 4 independent points. Simultaneous measurement of all wavelengths eliminates sampling artifacts that occur in single-channel or multiplex detection systems when the spectrum changes during a single scan. Simultaneous measurement therefore gives the HoloProbe immunity to artifacts generated by laser noise, bubbles, turbulence, or moving samples. An important limitation of CCD detectors is the small size of their pixels. The standard CCD detector used in the HoloProbe has square pixels 27 microns on a side. The HoloProbe combines the signal of several pixels into a single "super pixel" in order to make the effective detector size larger, but even so, super pixels are still very small. As a result, not only does the Raman instrument need to deliver as many Raman photons as possible to the CCD detector, the instrument must also concentrate those photons into a very small area. These two goals define the major design objectives for the HoloProbe spectrograph. Spectrograph The spectrograph section in the base unit disperses light from the filter stage into a spectrum and images the spectrum onto a CCD detector. The spectrograph is an axial transmissive design [2], rather than the more traditional Czerny-Turner design. These two designs are compared in Figure 3. The axial transmissive spectrograph design uses lenses instead of mirrors in order to get good imaging quality with an aperture ratio (focal length divided by focusing optic diameter) of f/1.8. The light gathering ability at the detector decreases with the square of aperture ratio, so lens-based spectrographs with small focal ratios give higher optical throughput to CCD detectors [3].
Traditional Raman spectrographs operate in the f/5 to f/8 range or higher in order to achieve acceptable imaging quality and linear dispersion with simple spherical or toroidal mirrors. Image blurring at the detector increases rapidly as the focal ratio is reduced. Even at f/4 the imaging quality of a Czerny-Turner spectrograph is much worse than that of an axial transmissive design. This blurring degrades the spectral resolution of the Raman instrument, limits the number of independent measurements channels, and decreases Raman peak heights. The axial transmissive spectrograph also uses a unique volume holographic transmission grating [9] discussed later in this paper, that stacks two spectral regions on top of each other at the detector, effectively doubling the spectral coverage that can be obtained with a given CCD detector. Since the entire spectrum is imaged on the CCD detector simultaneously, there is no need to ever move the grating. The elimination of grating movement and adjustment hardware simplifies the spectrograph and removes a potential source of misalignment and failure. Unlike traditionally used reflection gratings, the transmission grating produces only a small spectral shift at the detector when the grating angle is changed. This property makes the axial transmissive spectrograph wavelength calibration more stable than a typical spectrograph using reflection gratings. The diffraction grating used by the HoloProbe needs to have a very large angular dispersion due to the short focal length (85 mm) of the spectrograph. Angular dispersion refers to the different directions light of different wavelengths takes as it leaves the diffraction grating. High angular dispersion often causes some diffracted light to get lost because it spreads out too quickly to be fully captured by the spectrograph focusing lens (or mirror). This kind of light loss is called vignetting. The axial transmissive spectrograph design minimizes vignetting by placing the focusing lens very close to the grating. Optical Fibers and the Fiber optic Probehead Raman light is delivered to the base unit by an optical fiber having a core diameter of 100 micrometers and a numerical aperture of 0.29. It is this relatively large diameter and large numerical aperture that makes the low focal ratio of the spectrograph so important. A spectrograph having a larger focal ratio cannot efficiently deliver the light from the optical fiber to a target as small as the CCD pixels. The amount of light that can be coupled from the sample into the optical fiber increases with core diameter and numerical aperture, though. Probehead alignment tolerances also increase with fiber core diameter, so it is usually advantageous to use the largest collection fiber that the spectrograph can efficiently handle.
The HoloProbe fiber optic probehead [10,11] is
described in detail, along with other fiber optic probe designs, by Lewis et al. [12,13], so it will be described only briefly here. A schematic
diagram of the probehead is shown in Figure 4. The probehead uses
the same detachable optic to focus laser light on the sample, and to collect Raman
emission from the sample. Permanent alignment inside the probehead ensures that the
excitation optical path and the collection optical path are always properly aligned,
regardless of the collection optic used. Numerous collection optics are available to
provide the user with a wide range of working distances and interfaces to process
equipment. For example, one class of collection optics consists of a lens and window in a
hollow tube. This optic can be immersed in transparent or opaque liquids. It is
automatically optimally focused on the sample. No optical alignment is required from the
user. Laser light is delivered to the probehead by a single optical fiber. This laser light is contaminated by fluorescence and Raman emission from the optical fiber itself. The probehead therefore contains a miniature volume holographic monochromator to filter this contamination from the laser light. Light scattered by the sample is coupled into the second optical fiber, but only after it is filtered by a volume holographic notch filter to reject laser light scattered by the sample. If this laser light were coupled into the second optical fiber, it would produce fluorescence and Raman emission from the second optical fiber that would interfere with the Raman signal from the sample. Holographic Optical Elements The HoloProbe uses several high performance volume holographic optical elements (HOEs). These relatively new devices consist of a volume hologram in a thin film of dichromated gelatin that is sealed between two plates of glass [14]. The hologram is an optical interference pattern recorded as refractive index variations in the gelatin. Unlike artistic holograms, which produce 3-dimensional images, the hologram(s) in HOEs simulate optical devices or combinations of optical devices. This technology has been used in military fighter aircraft for many years, but only recently became available to the spectroscopic community. Volume holographic notch filters are used in both the base unit and in the probehead to reject laser light while efficiently transmitting Raman light. They offer a simple, small, less expensive, higher throughput, and more mechanically robust alternative to the traditional subtractive double spectrograph. Volume holographic notch filters do not distort Raman spectra the way that dielectric filters or subtractive double spectrographs do because the volume holographic notch filter transmission curves are very smooth and flat throughout nearly all of the Raman spectral region. Volume holographic notch filters transmit Raman light as close as 50 cm-1 from the laser line while reducing the intensity at the laser wavelength by more than a factor of one million. A volume holographic beam splitter is used in the fiber optic probehead to combine the excitation and collection paths. This beam splitter is really just another volume holographic notch filter. Volume holographic diffraction gratings are used in the miniature monochromator inside the fiber optic probehead, and in the 785 nm laser filter. In addition to their high diffraction efficiency, these gratings have much lower optical absorption than metal reflection gratings, and therefore a much higher laser damage threshold. These gratings are exposed to the full laser power, so high damage threshold can be as important as high throughput. Finally, the most complex volume holographic optical element in the HoloProbe is the diffraction grating in the spectrograph section described earlier. Lasers The HoloProbe is configured with one of four different types of laser: a
532 nm frequency-doubled Nd:YAG laser, a 785 nm external cavity stabilized diode laser, a
1064 nm Nd:YAG laser, or a 633 nm HeNe laser. Each of these lasers has no need for
alignment or maintenance for the entire lifetime of the laser, which tends to be one year
or longer of continuous operation. A volume holographic laser bandpass filter is used with
the diode laser to reject the intense superradiant emission. Without this filter, the
superradiant emission can add an overwhelming continuum background to the low wavenumber
region of the Raman spectrum. Lasers external to the HoloProbe base unit can also be used
by coupling them directly into the laser delivery fiber that normally attaches to the
HoloProbe base unit. Base unit filter stage Light from the collection optical fiber entering the base unit first passes through the filter stage. The filter stage uses a volume holographic notch filter to remove residual laser light. It also contains a shutter to control the data acquisition time, and a lens to focus the Raman light onto the entrance slit of the spectrograph. The fiber optic connector that couples the optical fiber to the base unit is mounted on a plate that is kinematically mounted to the filter stage. This plate snaps in and out of the filter stage. It can be replaced with a 4 channel adapter that accepts four fiber optic cables, and directs their outputs through the filter stage to the proper locations on the spectrograph entrance slit. Raman Applications Raman spectroscopy was traditionally confined to the laboratory environment. The instrument described in this paper, however, is being used in many applications outside the laboratory environment. In most cases, the primary reason for choosing Raman spectroscopy for these applications over any other analytical technique was ease of sampling. Raman spectroscopy provides detailed analytical results without collecting or diverting a sample from the process. Raman spectroscopy provides non-invasive, non-contact, non-destructive analysis through common window materials such as glass, sapphire, or quartz. It can often be made insensitive to films that build up on windows by simply focusing at a point further from the window. Fiber optic probes allow the Raman instrument to be hundreds of meters away from the actual point of measurement. A detailed discussion of specific Raman applications is beyond the scope of this paper. A few brief descriptions of representative (and non-proprietary) applications, however, should illustrate the types of analytical problems that are well suited to Raman spectroscopy. Raman spectroscopy is being used to continuously monitor the contents of the reaction vessel used to manufacture PCl3 [15]. The Raman probe is held inside the reaction vessel. The previous analytical method required pumping the contents of the reaction vessel through an external flow cell and back into the reactor. While this sounds like an easy thing to do, the hazardous nature of the reactor contents made the external flow cell approach very expensive and undesirable. The in-situ Raman measurement provides the required analytical data in a safe, cost-effective, and low-maintenance manner. Raman spectroscopy is being used to replace gas chromatography in a distillation monitoring application [16]. In this case, the lower maintenance costs and supplies costs made the estimated total cost-of-ownership lower for Raman spectroscopy than for gas chromatography. Faster analytical results were an added benefit that provided unexpected insight into the true operation of the distillation column. Raman spectroscopy was used to non-destructively determine the integrity of micromechanical silicon accelerometer chip seals [17]. The analytical challenge was to quantitatively analyze the gas in a 70-micron-deep cavity between a silicon cap and a Pyrex window. One final example is the use of Raman spectroscopy to monitor the
composition of semiconductor cleaning baths [18]. In this case
the analysis had to be non-contact in order to avoid any possibility of contaminating the
ultra-pure cleaning reagents. The cleaning solution generated a continuous stream of
bubbles, so the measured Raman signal intensity varied randomly over a wide range. The
quantitation accuracy was better than 0.1%, however, because all of the Raman wavelengths
were simultaneously and continuously measured. References
|
![[upsilon]](upsilon.gif){kind=small}
0 . In the case of a double beam dispersive instrument, the
intensity of the reference beam is I0 and it also has
this value for an interferometric instrument in the absence of the sample. The intensity
value is I in the sample beam of the dispersive spectrometer after the radiation has
traversed the sample, and likewise in the interferometric instrument. The detector/output
systems of both types of instrument measure the ratio I0/I,
or its inverse, and display the result in terms of percentage transmission or absorbance,
i.e. log10 I0/I. ![[Delta]](delta-c.gif){kind=small}
v
where 
O. This is
not achievable in practice.
