The Basic Physics of Infrared Emission Spectroscopy

William Maddams University of Southampton Until recently Chief Vibrational Spectroscopist, BP Research, Sunbury on Thames, UK. The analytical potential of infrared emission spectroscopy as a characterisational technique rests upon the fundamental fact that if a body emits radiation of one or more wavelengths under given conditions it will also absorb those wavelengths under the same conditions. This tenet requires justification; fortunately, the required reasoning is quite simple. Consider a body in equilibrium with its surroundings in a uniform temperature enclosure, exchanging energy only by radiation. Under these conditions it must absorb exactly the same energy, in any time interval, as it emits, otherwise its temperature will rise or fall. Furthermore, it must absorb exactly the same wavelength as it emits, otherwise the character of the radiation within the enclosure would alter with time, and would depend on the contents of the enclosure and not simply on its temperature. More generally, it may be shown that in non-equilibrium conditions, which often occur in practice, the absorptions and emission processes, at one or more wavelengths will still occur, but not necessarily at the same temperature. Bunsen and Kirchhoff, some one hundred and fifty years ago, devised an elegant but simple experiment that demonstrates the correctness of what was deduced above by simple reasoning. Their equipment is shown schematically in Figure 1.

F is a sodium flame, such as is obtained by feeding sodium chloride into a Bunsen burner. When this is viewed through the prism spectroscopes A and B the well known bright yellow doublet, with lines and 5890 and 5896angstroms, is seen. When the strong white light source [WLS], shining through the flame and in line with A, is switched on, a continuous spectrum with two dark lines at 5890 and 5896A is observed through A. However, spectrometer B, at right angles to the line of WLS/F/A, still shows that yellow sodium doublet.

The explanation is that the flame has absorbed the wavelengths 5890 and 5896A from the white light source and re-radiated it in all directions, and the amount falling on the entrance slit of spectrometer A is so small that two dark lines appear against the otherwise continuous background. This, incidentally is the cause of the various Fraunhofer lines, characteristic for a number of elements, that appear in the solar spectrum. When the flame, with the white light illumination passing through it in the direction WLS/F/A, is viewed at right angles via spectrometer B., the two sodium yellow lines are still present at unchanged intensity. One clear consequence of the relation between emission and absorption is that a wholly transparent medium cannot emit. This may appear contrary to everyday experience, but it has been demonstrated by Professor R.W. Wood, who devised many beautiful and clever experiments in physical optics. He used a rod of very pure fused quartz, free from small particles of foreign matter. When this was heated in a Bunsen flame there was only a very small amount of light emission. However, when it was heated to a higher temperature in an oxy-coal gas flame it became vividly incandescent. The bluish luminosity faded when the rod was removed from the flame but, on cooling, it did not pass through the usual stages of yellow and red heat. The explanation for the bluish luminosity is that, at ambient temperatures, quartz has an absorption band well into the ultraviolet spectral regions. This moves to longer wavelengths with increasing temperature and, with an oxy-coal gas flame reaches the violet end of the visible spectrum. Hence, as there is absorption, emission is also possible. The absorption moves back into the ultraviolet region as the sample cools, so it cannot pass through the yellow and red heat stages.

Any absorption spectrum displays the peaks in terms of loss of intensity from the incident intensity, which is set at 100%, so the spectrum has percentage transmission as its intensity scale. The emission spectrum represents an increase in intensity from a baseline of zero intensity. It follows that the emission spectrum will be the mirror image of the absorption spectrum as is shown in Figure 2, where the intensity of emission scale has been adjusted, in arbitrary units so that the peak heights in the two spectra are equal.

Figure 2. Infrared absorption and emission spectra

The difference between the two spectra is that the absorption intensity has been measured in terms of a ratio whereas the emission intensity is a straight measurement and will depend upon several factors, most notably the temperature of the emitting

body. It is therefore highly desirable that the emission intensity should also be expressed as a suitable ratio. A convenient way for doing this emerges when the emission process is considered in terms of some simple mathematics. The first step is to quantify the term emission. A convenient unit is the total energy emitted per square centimetre per second, for a particular temperature. This is termed THE TOTAL EMISSIVE POWER, denoted by the symbole. Consider, once again, a body at equilibrium in a constant temperature enclosure. The energy emitted per square centimetre per second in all directions between wavelength and +d , at a given temperature, is e d , where e is the emissive power of the surface for that particular wavelength and temperature. Let the energy falling on unit area per second within the same wavelength limits be dQ. This depends only on the temperature of the enclosure. Also, let the fraction of the incident energy absorbed between the wavelengths and +d be a, an entity usually termed the ABSORPTIVE POWER of the surface. Hence, the total energy absorbed by the unit area per second is adQ. As the temperature of the body is constant, the rate of emission of energy must equal the rate of absorption. Thus, ed = adQ or

edQ
a

Since d is a chosen, fixed interval and dQ depends only on the temperature of the enclosure, the right hand side of the equation must be constant so

econstant
a

This equation is generally applicable and, in particular, to a black body, which is defined as a perfectly absorbing, perfectly emitting body. If, for this black body, the emissive and absorptive powers are denoted by E and A respectively, then

eE
a

Then, because absorptive power is defined as a fractional quantity, A = 1, giving

eE

Hence, at any given temperature, the ratio of the emissive power of any body to its absorbing power for that wavelength at that temperature is constant, and is equal to the emissive power of a black body at that temperature. This is Kirchhoff's Law, and provides the basis for infrared emission spectroscopy. Since E is a clearly defined entity, it is convenient for practical purposes to express the results of experimental measurements as the ratio e/E. This is known as the EMITTANCE, and spectra are a plot of emittance as a function of wavelength. It is useful to note, at this juncture that, for experimental purposes, a roughened surface coated with soot provides a good approximation to a black body. As noted above, emissive power is temperature dependent, increasing with increasing temperature. However, emittance is independent of temperature, because e and E are always measured at the same temperature. If a black body at temperature T is surrounded by a perfectly black surface at temperature To, the latter being greater than T, the gain of energy of the black body is given by the equation E = s (To4 - T4 ). This is Stefan's Law, which he deduced experimentally, and it was later given a theoretical proof by Boltzmann, on thermodynamic grounds. The practical implications of this equation are that the emission from a body at a higher temperature than its surrounding increases rapidly as the temperature difference between them increases. This is important because the emitted energy level is relatively small but, on the other hand, the chemical stability of the sample being heated may determine the maximum temperature to which it may be heated to obtain an infrared emission spectrum. Additionally, it is evident that if the sample is at ambient temperature, but the radiation detector is cooled it may be possible to measure a spectrum, a fact which has practical implications. A black body, by definition, can only absorb and emit. With real life samples some energy transfer by reflection will always occur. The degree of specular reflection as the emitted radiation leaves the sample, is a function of the refractive index of the sample, and at normal incidence and a refractive index of 1.5, which is typical for many organic compounds, it amounts to about 4%. This is not significant but its effect is evident, as a secondary effect, in distorting the shapes of stronger emission peaks. This is because the refractive index of the sample changes in the vicinity of these peaks, falling on the high frequency side of the maximum and rising on the low frequency side, to as much as 2.5. The consequential effect is a small reduction in the reflectance on the high frequency edge of the peak and a rather larger increase on the low frequency side. This, in turn, produced a degree of distortion in the emission

spectrum, but the effect is unlikely to be large enough to lead to a misinterpretation of the measured emission spectrum. To conclude, a few brief comments on the measurement of infrared emission spectra will complement and supplement the background information given above. The T4dependence of the emission intensity shows that a doubling of the difference in the sample and detector temperatures should lead to a sixteen-fold increase in intensity. In practice, the maximum usuable temperature may well be determined by the sample stability. Fortunately, the sensitivity of modern Fourier Transform spectrometers is such that emission intensity is seldom the limiting factor. Furthermore, instruments with cooled detectors are available, so that it may be possible to operate with the omitting sample at ambient temperature, given that the spectrometer design is such that radiation emitted by surfaces in the instrument itself does not reach the detector. The usual practice is to remove the normal spectrometer radiation source and replace it by the heated sample. This can take a number of forms. Solid films are conveniently dealt with mounted inside a heated liquid cell, whose transparent windows have a loe emissivity. Alternatively, a solid sample may be placed in contact with a polished metal surface and this is effective given that the sample is not too thick and there is a significant temperature gradient from the back to the front of the film. Such samples may occur in practice, in the form of thin polymer films on aluminium cans. When the emission spectrum of the sample has been recorded and stored in the computer memory, the sample is replaced by a black body at the same temperature. Its emission spectrum is measured and the ratio of the emission curves of the sample and the black body gives the required emission spectrum in units of emittance. REF: W. Maddams. Internet J. Vib. Spec.[www.ijvs.com] 5, 2, 2 (2001)

3. Recording emission Spectra

The Editor Research grade FTIR is usually have an emission port (although this may be occupied by an FT Raman accessory if you are lucky enough to have one). Removing some sort of plug or cover and usually swinging away a mirror inside can access the port . However, there is a snag - you need a collimated beam as you will see in Figure 1. Bill suggested replacing the source by the emitting sample but this is usually very difficult to do. Illuminating the interferometer through the emission port is optically identical to replacing the source. Where do you go from here?

Figure 1. If you are wealthy enough to have an emission accessory or can beg, steal of borrow one to try the technique all is well, but most of you just want to try the method to see if there is anything in it for you. So just as most of us paint our houses because we cannot afford to have someone do it for us. I suggest a little bit of "do-it-yourself". To make a cheap, functional emission accessory - see Figure 2 - you need two mirrors, ideally M, they should be an off-axis paraboloid but a front surface aluminised shaving mirror quality spherical will do just fine. M2 is plane and must again be front surface aluminised. Window glass quality is fine.

Figure 2.
Cheap optics are not difficult to buy but if you have trouble contact Comar Instruments on +44 (0)1223 245470, ask for Mr Marsh. Either of the two gentlemen of this name will sort you out with advice and mirrors and not charge an arm and a leg.

To support the mirror and sample, I recommend you find a piece of steel (3mm thick is fine but thicker is no problem). Most workshops can help with this. Find a lab jack and for safety screw the bottom to a wooden board and the top to the steel sheet. If you find the latter is beyond you use contact adhesive. Now - engineering suppliers sell magnetic bases for "dial test indicators" and "marking out" accessories. I draw one in Figure 3. Buy two of them and mount the mirrors as I indicate. The whole setup shouldn't cost more than $100.

Figure 3. If your instrument has a laser spot on axis lining up the accessory is easy but if not you will have to be a little cleverer. To do so - place M1 by eye, similarly M2. Cover the instrument detector, switch on the source and put a strip of "magic tape" or tracing paper at the normal transmission sample point. Operate the instrument as a transmission machine and mark the patch of light on the tape or paper with a pencil. Using a small torch bulb find a position where the light from the bulb goes through your home built accessory and the interferometer and illuminates centre of the mark on the white screen. Now adjust all the mirrors to get a good optical path and note the position of the bulb filament. This will be your sample point. Uncover the detector, put a beaker containing hot water at the sample point and carefully tweak the mirrors to give you maximum energy. You are now ready to proceed except for one little problem - moist air. The beam splitter in your instrument will suffer if it is exposed to the atmosphere . However, a few minutes exposure will cause no damage. Find some thin polythene sheet and some masking tape and cover your accessory carefully so that it is as sealed as you can make it. Now purge your instrument and the accessory will be purged as well.
In the tropics be VERY careful. IF the atmosphere is hot and humid and you do not have air conditioning DO NOT open the interferometer at all.

What to do if you have no emission part or the one you have is occupied? There is an alternative method which will fit into the transmission sample area. As you will see in Figure 4, the beam passing through an interferometer goes forwards and backwards i.e. 50% passes through the device and 50% returns to the source. Although not as efficient as replacing the IR source by your sample, you can exploit this property of interferometers.

Figure 4. In Figure 5 I show you how. The emitted radiation passes into the interferometer and the returned radiation goes to the detector.

Figure 5. Obviously, half the exit J-stop is used to illuminate and half to view the interferometer. There are two problems with the method and both are experimental and simple to solve. a) Mirror M2 must not cast a shadow due to its thickness so it must have a bevelled edge and b) The mirror has to be very accurately positioned and angled. The problem is that interferometer have a real entrance J-stop (piece of metal with a hole in it) but a virtual exit J-stop i.e. the exit J-stop is the reverse projected image of the detector - not a great help when you are trying to set up a do-it-yourself accessory. Proceed as follows. Set the sample stand at the focus of the instrument. Put a strip of magic tape or tracing paper over the cell holder as suggested above and carefully mark with a pen, the position of the illuminated patch. Using a piece of steel as a base as above and magnetic bases, move the mirror M2 until it cuts of the patch. Illuminate M1 with the light from the interferometer , adjust M1and locate the emission sample point. Remember - before you make a measurement, turn off the normal IR source. You will find the measurements are not difficult to make especially if a cooled detector is used and the results can be very interesting indeed. Good Luck

REF: P.J. Hendra. Internet J. Vib. Spec.[www.ijvs.com] 5, 2, 3 (2001)

4. Some thoughts on Catalysis and the use of IR & Raman in its study
The Editor
There is absolutely no doubt that vibrational spectroscopy has been a prime analytical tool in heterogeneous catalysis for half a century but dare I suggest that in a way this is a little surprising. In heterogeneous catalysis reactants are passed over a catalytically active surface usually but not always heated well above ambient and as if by magic, the desired products are produced. [I will always remember that my organic tutor at University back in the 1950's - a man very classical in his views, always used to mutter -" any fool can shove a gas down a hot tube and get a load of old rubbish out the end"]. Modern catalytic systems when operated at precise temperatures under carefully purified reactants at controlled pressures can produce desired reactions with precision and at very high throughput. These processes are the heart of the large-scale chemical industry if only because they are flow systems and not batch procedures. In investigating these reactions and in devising new systems the nature of the reactants, the products, temperature pressures and flow rates are all easy to control and monitor. What is really hard is to actually see the reaction(s) itself. The reactants sorb to the surface. At reactive sites rearrangements, decomposition or addition occurs and the products usually vacate the surface very rapidly. As a consequence, exactly as is the case in non-catalytic processes where the activated complex is completely transitory, the reacting species at the surface are there in only minute concentration i.e. they are hard or impossible to study. There is a large range of methods available for the detailed analysis of surfaces and some are listed in the table below.
Pressure Range Several Atmospheres Method IR Raman Information gained Molecular Structure Molecular Structure but at

10-

10

atmos

As above

Scanning Tunneling Microscopy Not too far from 1 atmos NMR unless surface has chemisorbed species on it Very low pressures over Variety of electron clean surfaces spectroscopies and Molecular Spectroscopy

very low pressures need SERS Shape of molecules on well defined surfaces Very detailed molecular structure Usually elemented analysis and/or valency state. Rarely molecular structure.

Table 1. Some of the methods available to study surfaces. If our quest is to analyse the molecular species involved it seems to be absolutely obvious that the analytical procedure must operate at sensible pressures (from say 20 atmospheres) and representative temperatures (RT 100K) because only then will the reaction under study be typical of what it is that we are hoping to achieve. Under this set of constraints, infrared and Raman spectroscopies appear to be ideal because you can heat and pressurise and actually 'look' at the surfaces whilst they are reacting. However, the view is nave - the concentration of active sites and even worse the concentration of reacting species on the surface is swamped by the unreacted reactants and generated products. So where do we go from here? The examination of clean highly characterised surfaces such as single crystals to which are sorbed sub-monolayers of reactive gases or vapours must obviously be far more likely to yield results of significance on catalytic processes. The problem is that by definition there is almost nothing to examine! To give you an idea: assuming our technique examines 1 sqcm of surface, a monolayer of benzene lying flat on the surface would weigh only 8 micrograms and the thickness of the layer would be ~ 1angstroms. If we wish to carry out a simple transflectance measurement the sample thickness would be ~2 angstroms i.e. ~10-5 of the conventional infrared pathlength. If one planned to try Raman the sample size is absurdly small. So the outlook looks bleak but not disastrously so. Once the performance of infrared spectrometers had improved sufficiently it was demonstrated that it was possible to observe incredibly weak bands. If multi-reflection techniques were devised and/or glancing reflection resorted to, the effective path length in these cases could be raised. Infrared experimenters as early as the 1970's were able to study CO over metal crystal surfaces and the field has expanded from there very rapidly.

Raman is obviously not attractive - its lack of sensitivity and the minute sample volume inevitable with such a thin specimen make the method impracticable. However, the SERS effect can enhance the spectrum by 106or much more in favourable cases and so SERS is a technique with immense potential* success has already been achieved - see reference 1. * Readers should consult the Special Edition on SERS published last year. Many catalysts are highly porous. Compounds such as silica aluminas or zeolites have huge areas to which molecules can sorb. In microporous materials the holes may be of molecular size and limit access to sites nearer the surface, but the surface area of these materials can be huge indeed - values of 100 1000m2g-1 are not unusual. Sorbing a monolayer to these materials does provide a decent amount of material and so vibrational spectroscopy started with this type of catalyst system. Pioneers such as Norman Sheppard adsorbed methane to silicas and using infrared were able to show how the molecules interacted. The selection rules dictate that the bands expected for I are not the same as for II or III.

----- = surface

Whilst the sorbtion process will alter the strength and hence the 'rigidity' of the CH bands lowering their vibrational frequencies. This approach is half a century old but still vitally important. The experimental methods available have improved but the analysis is based on the same logic. Perhaps the main limitation with this method is that the predominant species must always be the support and hence the spectrum of the adsorbate must overlie that of the catalyst itself. Silica based materials absorb strongly in the 1000-1500cm-1region so as a result much of the 'fingerprint' region is obliterated. Non crystalline and partially crystalline ceramics often have very weak Raman spectra so it is obvious that monolayer coverages of good Raman scatters over porous surfaces might well be accessible.

They are and pioneering data appeared as early as 1970 but the field was dogged by fluorescence. The mere act of baking the catalyst under vacuum to drive off water and contaminants prior to sorbing the molecule of interest could produce astronomic levels of fluorescence. The problem was eventually traced to oil and grease in the vacuum system and was tamed but as a field of study, Raman has not been nearly as significant as infrared. So - if the study of molecules sorbed directly to surfaces is of considerable fundamental interest, but tricky to do, what other measurements are of value? Catalysts are frequently highly heterogeneous and their preparation may well be complex and chemically rather obscure e.g. highly significant catalysts such as those used in vehicle exhaust systems involve tiny clusters of metal atoms supported on a porous ceramic material. The methods by which these materials are produced are of very considerable commercial importance. The best method of production can well result in market domination. In general, the ceramic is treated with a solution of the metal(s) as salts or as organometallics. Once dried and often roasted, the catalyst is then frequently reduced using H2 or cracked ammonia generating the desired clusters of metal atoms. Vibrational spectroscopy has been remarkably successful in following the reactions involved. Most of the competing surface analytical techniques are relatively unattractive on these incredibly complex heterogeneous mixtures. Zeolite and silica alumina catalysts have a variety of acidic sites on their surfaces. They may contain surface OH groups (I) or Lewis acid acceptor sites (II)

D = Donor Molecule

or even Bronsted acid sites allowing proton abstraction III.

The performance of the catalyst is frequently dominated by the relative and absolute density of these sites on the surface. This is another area where vibrational spectroscopy is unequally valuable. The trick is to use a tell-tale molecule. Ammonia has always been popular but pyridene can also be used. Let us take pyridene as an example. Liquid pyridene shows its ring breathing mode at 990cm-1. Pyridene/water mixtures show a band near 1000cm-1. The pyridene cation vibrates still faster whilst coordination of pyridene to a metal pushes the frequency into the 1025cm-1area. Quite obviously, if one sorbs the pyridene to the catalyst and runs the spectrum, the nature of the species and their relative distribution can quickly be measured. The example I give is based on Raman bands but infrared is just as useful, but relies on bands at much higher frequencies. This type of surface assay is routinely used to characterise and/or monitor catalysts. You can be even more clever - the actual frequency of the tell-tale molecule vibration is closely related to the strength of the adsorbate adsorbent interaction. So a careful analysis of the band shape can yield data on the distribution of activities of a specific type of site e.g. the Lewis acid activity on a cracking catalyst. See ref 2. All surface scientists concern themselves with the adsorption isotherm. The subject is covered in all classical Physical Chemistry textbooks, but let me over simplify. In many examples of porous catalysts, the plot of the amount of material absorbed vs. the pressure over the system is shown below.

Monolayer coverage coincides with the sharp downward bend in the curve. Above this level, sorption produced multi-layers and eventually liquid trapped in the pores until at Y free liquid wets the catalyst. Surface scientists measure these isotherms and can make much of their shape. Since vibrational spectroscopy is applied to samples in

enclosed cells at controlled temperatures, spectra can be recorded at precisely known points on the isotherm. This is particularly and almost unequally valuable in both fundamental and applied work. So - vibrational spectroscopy is valuable because of its enormous versatility. Haven't I heard that before somewhere?