You are on page 1of 5

1906

ANALYTICAL CHEMISTRY, VOL. 50, NO. 13, NOVEMBER 1978

Diffuse Reflectance Measurements by Infrared Fourier Transform Spectrometry


Michael P. Fuller and Peter R. Griffiths'
Department of Chemistry, Ohio University, Athens, Ohio 4570 1

An optical system for diffuse reflectance infrared spectrometry is described. The device uses an on-axis ellipsoidal collecting mirror and a commercial Fourier transform spectrometer. Almost any type of powdered sample can be studied and the feasibility of good quantitative analysis is discussed. Submicrogram detection limits are found for samples dispersed in KCI and for materials chromatographically separated directly on TLC plates. Results indicating the possibility of studying heterogeneous reactions of gases with powdered solids are shown.

T h e popularity of infrared spectrometry can in large part be attributed to its versatility, since samples may be readily studied as gases and liquids and, somewhat less readily, as solids. Two types of solid samples may be distinguished. Polymers may be prepared as free-standing or cast films and their spectra measured using conventional transmission or attenuated total reflection techniques. Powdered solids require preparation as solutions in a suitable solvent, as KBr disks, or as mineral oil mulls, and measurement of the spectra of powders without sample preparation is generally thought of as being rather difficult. Two possible techniques have been discussed in t h e past for measuring the infrared spectra of powdered samples without prior preparation, namely emission and diffuse reflection spectrometry. Infrared emission spectrometry is fraught with experimental hazards, in particular the effect of temperature gradients on the spectra and sample decomposition on heating. T o date no good quantitative measurements have been obtained by infrared emission spectrometry for powdered samples. Diffuse reflectance infrared spectrometry has been largely ignored in t h e past by analytical spectroscopists, since it has generally been considered that the intensity of radiation reflected from powders is too low to enable spectra to be measured at medium resolution (2-4 cm-') and high signal-to-noise ratio (SNR). Diffuse reflectance spectrometry has been widely used for the measurement of the ultraviolet-visible spectra of powders and turbid liquids, and a number of books on this subject have been written (1-5). The technique has been applied to clinical measurements ( 6 ) ,pharmaceutical quality control ( 7 ) ,heat transfer studies ( 8 ) , and in food science (9-11). T h e most commonly used device for collecting diffusely reflected UV-visible, and even near infrared, radiation from a sample is a n integrating sphere whose interior is coated with a diffusing, nonabsorbing powder, such as MgO or BaS04. T h e sample and detector are usually held a t t h e surface of the sphere, and the measured spectrum is, to a good approximation, independent of the spatial distribution of the reflected light and the relative position of the sample and the detector. T h e use of an integrating sphere for mid infrared diffuse reflectance spectrometry presents several problems. Firstly, the devices are intrinsically inefficient, and infrared detectors are several orders of magnitude less sensitive than photomultiplier tubes, so that the SNR of measured spectra would be expected to be low. Secondly, it is not easy to prepare
0003-2700/78/0350-1906$01 .OO/O

suitable coatings for the interior of spheres designed for mid infrared spectrometry with equivalent performance t o t h a t of MgO or BaS04 in the UV-visible region. As a result, reports of vibrational spectrometry using diffuse reflectance techniques are sparse, and most of the published work in this field is more concerned with the acquisition of accurate engineering data, and describes the optimization of sample illumination, scattered radiation collection, and focusing methods rather than chemical or analytical spectrometry. Previous devices which have been employed in the measurement of mid infrared diffuse reflectance spectra have included the Coblentz hemisphere ( 2 2 ) , the Gier-Dunkle cavity (13), hemispherical (14),hemiellipsoidal (15) and ellipsoidal (16-18) collecting mirrors, in addition to integrating spheres (19, 20). Of these devices, t h e hemiellipsoidal and ellipsoidal reflectometers are probably the most efficient in terms of collecting and transmitting diffusely reflected radiation to the detector. T h e feasibility of applying these mirrors in conjunction with prism monochromators t o chemical spectroscopy (15) and catalysis (16) has been demonstrated. Recently a special-purpose Fourier transform infrared (FT-IR) spectrometer has been interfaced to an integrating sphere for diffuse reflectance measurements (20), b u t the sphere is so inefficient that, in spite of t h e high performance of t h e spectrometer, spectra a t adequate resolution and SNR required rather long measurement times. I n this report we describe results obtained when a fairly efficient ellipsoidal collecting mirror is interfaced to a commercial rapid-scanning FT-IR spectrometer. We demonstrate that diffuse reflectance infrared Fourier transform (DRIFT) spectra of a variety of powdered samples can be measured rapidly a t medium resolution and high SNR. These spectra provide useful analytical information, and suggest that many types of previously intractable samples can be studied using this technique.

INSTRUMENTAL The optical configuration of the diffuse reflectometer used in this work is shown in Figure 1. The beam from a modified Nernst glower source (Perkin-Elmer Corp., Nonvalk, Conn.) is collimated using a 7.5-cm focal length, 45' off-axis paraboloidal mirror (Special Optics Corp., Little Falls, N.J.) and passed through a Model 296 Michelson interferometer (Digilab, Inc., Cambridge, Mass.). The collimated exit beam is then reflected by two plane mirrors to a 3.7-ern focal length, 90' off-axis paraboloidal mirror (Special Optics Corp.) which focuses the beam onto the powdered sample held in a small cup at one focus of an ellipsoidal mirror. This mirror (Special Optics Corp.) is the ellipsoid of revolution about the x axis of the ellipse:
X2

(992)*

+--Y 2

(386)'

-1

(1)

A small hole centered on the major axis is drilled to allow the beam

from the paraboloid to pass onto the sample. Because of this geometry, any specularly reflected radiation from the sample returns through the hole, but the diffuse component is collected by the ellipsoid and focused at the other principal focus of the ellipsoid. In principle a detector could be held a t this focus, but in practice the magnification is so great that we use a second
C 1978 American Chemical Society

ANALYTICAL CHEMISTRY, VOL 50, NO. 13, NOVEMBER


~

1978

1907

~_____~____---~~

Collimated Beam from

Figure 1. Optical diagram of the diffuse reflectometer. Radiation from the interferometer is focused on the sample, S, by the off-axis paraboloid, P, and the diffusely reflected radiation is collected by t h e ellipsoid, E, and then focused onto t h e detector, D. (Note that the detector is not coplanar with the other optics)

3.7-cm focal length, 90 off-axis paraboloid which focuses the beam to a diameter of approximately 2 mm. Both a triglycine sulfate (TGS) and a mercury cadmium telluride (MCT) detector were used for the diffuse reflectance measurements. We favored the use of the MCT detector for most measurements since the time required to achieve an acceptable SNR (less than 0.170 noise) could be reduced below 5 min. The detector is a short range (Arnm = 12 Fm), high sensitivity device (Texas Instruments, Dallas, Texas) and generally necessitated that the output beam from the interferometer had to be attenuated by a wire mesh screen to avoid digitization noise. If a state-of-the-art interferometer was used with a new TGS detector, we believe that a TGS detector could be used for all the measurements described in this paper. Particle size classification was achieved using a Sonic Sifter (ATM Corp, Milwaukee, Wis.) with appropriately sized sieves, after grinding the samples manually or in a Wig-L-Bug (Crescent Dental Manufacturing Co., Chicago, Ill.). Powdered samples were transferred to a 4-mm diameter cup without compression, and were levelled using a spatula.

~~

4303
F R E Q U E N C Y I CU- I

t 1 0

Figure 2. Reflectance spectra of powdered KCI for different particle sizes, d , vs reflectance of KCI sized between 75 and 90 p m . (A) d < 10 pm, (6)10 < d < 75 p m , (C) 75 d < 90 pm, and (D) a >

90 pm

. :

QUANTITATIVE MEASUREMENTS A general theory for diffuse reflectance a t scattering layers within powdered samples has been developed by Kubelka and Munk (21,22). This theory relates sample concentration and scattered radiation intensity much in the same manner as the Bouger-Beer law of transmittance spectrometry. For an infinitely thick layer, the Kubelka--Munk equation may be written as:

in reflectance around 1000 cm-l. This effect presumably occurs when the wavelength of the radiation is approximately equal t o t h e diameter of t h e particles. For brevity we refer to all subsequent spectra as R , or f(R,) in common with previous investigators, whereas strictly they should be referred t o as R, and f ( R , ), respectively. T h e Kubelka-Munk theory predicts a linear relationship between t h e molar absorption coefficient, h , and the peak value of f(R,) for each band, provided s remains constant. Since s is dependent on particle size and range, these parameters should be made as consistent as possible if quantitative data are needed. It has been shown (23)that for dilute samples in low absorbing or nonabsorbing matrices:

k = 2.303~
where R , is t h e absolute reflectance of the layer, s is a scattering coefficient, and k is the molar absorption coefficient. In practice a perfect diffuse reflection standard has not been found and R , in Equation 2 may be replaced by R , , where

(4)

where t is t h e molar absorptivity and c is the molar concentration. Therefore:

R, =

R, (sample) Rm( s t a n d a r d )
(3)
where k = s/2.303t. Thus diffuse reflectance spectra of samples dispersed in finely powdered KBr or KC1 might be expected to be quite similar to the absorbance spectra of the same samples prepared as a KBr disk. T h a t this is the case is shown in Figures 3 and 4 for cholic acid and carbazole, respectively. I t is well known t h a t t h e spectra of samples prepared as KBr disks can vary depending on the way in which the sample is prepared. Figure 3B was obtained from a disk made by adding about 5 mg of cholic acid of less than 10-pm particle size to about 300 mg of KBr and pressing the disk directly. T h e absorbance spectrum of this sample is almost identical to the Kubelka-Munk spectrum of neat cholic acid, and both spectra show band splittings due to the crystallinity of t h e sample. The spectrum shown in Figure 3C was obtained from about 1 mg of cholic acid which was ground with about 300

R, (sample) represents the single-beam reflectance spectrum of the sample and R, (standard) is the single-beam reflectance spectrum of a selected nonabsorbing standard exhibiting high
diffuse reflectance throughout t h e wavelength regions being studied. We investigated the diffuse reflectance of powdered gold, germanium, and several alkali halides and found t h a t finely ground KCl (Harshaw Chemical Company, Crystal and Electronic Products Dept., Solon, Ohio) has t h e fewest interferences and the highest overall reflectance. The reflectance of KC1 was found to vary with particle size, as shown in Figure 2 , with t h e highest reflectance a t high wavenumber being found for the smallest average particle size. We selected KCl with a particle size less than 10 pm as the reflectance standard for this work. However it is interesting to note the small drop

1908

ANALYTICAL CHEMISTRY, VOL. 50, NO. 13, NOVEMBER 1978

,
/

29

3-

50

55

h E , G i- p E R C C \ T

a 2 3 B E h L ELI E

3200

CM-

800

Figure 5.

Kubelka-Munk spectrum of neat cholic acid vs. a KCI reference. (B) Linear absorbance spectrum of a KBr disk of cholic acid prepared by adding 5 mg of cholic acid (particle size less than 10 p m ) to 100 mg of KBr, and pressing the disk directly. (C) Linear absorbance spectrum o a KBr disk prepared by grinding 1 mg of cholic f acid with 100 mg of KBr in a Wig-L-Bug, with subsequent pressing
Figure 3. (A)

Variation of f(R,) with weight percent azobenzene in KCI for the 1960 cm band of azobenzene
I

I33

2300

5c.c

OC

b7EOLth Y

-r.

2303

1500
FPEOLIENCI ICM-

,033

Effect of particle size on the diffuse reflectance spectrum, plotted as f(R,), of neat azobenzene vs a KCI reference (A) d > 90 p m , (B) 75 < d < 90 pm, (C) 10 < d < 75 pm, and (D) d < 10
Figure 6.

w-

spectrum of a KBr disk of carbazole. (B) Kubelka-Munk diffuse reflectance spectrum of 5 % carbazole in KCI vs. a KCI reference mg of KBr, after which the disk was pressed. No band splittings are observed in this spectrum, and the relative intensities of the bands are somewhat different from those of the other two spectra. The carbazole spectra shown in Figure 4 are more typical of results found for most organic compounds, and few, if any, crystal splittings are observed. Minor differences between the absorbance spectra of the KBr disks of most organics and their diffuse reflectance spectra may be attributed to the effect of particle size and deviations from the Kubelka-Munk equation, which is only expected to hold for moderately absorbing species at controlled particle size. A typical Kubelka-Munk law plot is shown in Figure 5 , and deviations from linearity a t high sample concentration are indeed observed. It is also noteworthy that Christiansen filter effects (24) are completely absent from diffuse reflectance spectra. The effect of particle size on the diffuse reflectance spectra of pure compounds is illustrated in Figure 6. I t can be seen t h a t the bandwidths and relative intensities change con-

Figure 4. (A) Linear absorbance

siderably with particle size. The bandwidths decrease substantially as the particle diameter is reduced. It is interesting to note that the intensity of certain bands appears to change more dramatically with particle size than others. This effect is particularly noticeable for the azobenzene bands a t 1540 and 1405 cm shown in Figure 6. In view of the tremendous importance of scaled absorbance subtraction procedures in the study of mixtures by computerized infrared spectrophotometers (25-27), we were curious as to the feasibility of the analogous procedure, Kubelka Munk subtractions, for diffuse reflectance spectra. In general, we found that spectral subtraction is possible provided that f(Rm) both components is low, but that accurate subtraction for over the complete spectrum for mixtures which were not diluted with KC1 was not possible. In spite of this observation, we believe that diffuse reflectance difference measurements may be extremely important for detecting and quantitating small differences between powdered mixtures. As an example, the result of subtracting the f(R,) spectrum of a 2.9% mixture of magnesium stearate in lactose from a 5.1% mixture is shown in Figure 7 . All the positive-going bands are due to magnesium stearate while the negative-going bands may be as-

ANALYTICAL CHEMISTRY. VOL. 50, NO. 13, NOVEMBER 1978

1909

075

1I
2 5 h

0 7 T 050-

0501

3200

c CM
800

5%

3200

CM'

800

'

CM

150

Figure 8. Microsampling by diffuse reflectance: spectra of (A) 12 pg, (B) 1.2 p g , and (C) 180 ng of carbazole dispersed in KCI vs. a KCI reference

c
3200
C M.'

800

3200

Chi.'

800
~

Figure 7. (A) Diffuse reflectance spectrum of pure lactose. (B) Diffuse reflectance spectrum of pure magnesium stearate. (C) Spectrum of 5.1 % magnesium stearate and 94.9% lactose. (D) Difference spectrum of the f(R,) spectra of 5.1 'YO magnesium stearate and 2.9% magnesium stearate in lactose. Positive-going bands due to magnesium stearate are marked MS

signed to lactose. These two materials happen to be commonly used excipients for pharmaceutical tablets, and we believe that diffuse reflectance difference spectrometry may prove to be a very useful technique for quality assurance and content uniformity determinations of pharmaceutical products, presumably after the careful construction of working curves. Further investigations into this area are now being performed in this laboratory and will be reported a t a later date.

2000

F?EPUENCY ICM 1

3c0

QUALITATIVE ANALYSIS
The first and most obvious application of qualitative analysis by DRIFT spectrometry is the rapid identification of powders without sample preparation. As mentioned above, not only is sample preparation time very short but the amount of sample required to produce a strong spectrum is also quite small; 100 pg of most samples is all that is required to give a strong spectrum, and bands from less than 200 ng of moderately absorbing compounds can usually be observed, as shown in Figure 8. For selected strong absorbers we have been able to observe bands from as little as 2 ng of sample dispersed in KCl. The depth of the KCl matrix required to yield samples of "infinite thickness", Le., where further increasing the thickness causes no change in the spectrum, appears to be iess than 5 mm. However, the difference in band intensity between samples of, for example, 2 mm and 5 mm depth is not very great. The depth of the samples whose spectra are shown in Figure 8 was approximately 0.5 mm, since the greatest contribution to f ( R p ) originates in the first few layers of sample. I t is surprizing to find such low detection limits using a technique which has been so little used in the past because of its supposedly poor SNR. T h e nature of the matrix affects detection limits for microsampling by D R I F T spectrometry, but strong bands of adsorbed compounds can be seen even for strongly absorbing matrices such as silica gel. This observation suggested the possibility of measuring the spectra of species separated by

Figure 9. In situ spectrum of 1.2 pg of methylene blue on an aluminum-backed silica gel TLC plate. Bands marked A are due to the uncompensated adsorbent alone, bands marked B are due to both the adsorbent and sample, and all other bands are due to the sample

thin-layer chromatography (TLC-IR) without sample preparation. The most common method currently used for TLC-IR involves extracting the spots using a polar solvent, evaporating down the extract, and measuring the spectrum by microtransmittance or ATR techniques. An in situ method for TLC-IR has been suggested (28-30), but requires the preparation of the TLC plate on an infrared transmitting plate, such as silver chloride. We have measured TLC-IR spectra of samples separated on commercially available aluminum-foil-backed silica gel TLC plates after punching out the spots with a standard 6-mm diameter file punch and holding each spot a t the sample focus of the D R I F T spectrometer. Although this system is by no means optimized for TLC-IR measurements and accurate compensation for the strong SiOa bands has not yet been ,achieved, the spectrum of 1.2 pg of methylene blue dye on a silica gel TLC plate shown in Figure 9 indicates the feasibility of TLC-IR by diffuse reflectance, even in the region of strong absorption bands of the adsorbent. These results suggest that DRIFT measurements may be very useful for following reactions of adsorbed species on catalyst surfaces without having to follow the currently accepted practice of compressing the catalyst into a thin self-supporting disk. With such a system, the kinetics of

1910

ANALYTICAL CHEMISTRY, VOL. 50, NO. 13, NOVEMBER 1 9 7 8

The number of potential applications for diffuse reflectance spectrometry appears to be enormous. We believe that this technique will provide analytical chemists and applied spectroscopists with an important tool for studying previously intractable samples.

LITERATURE CITED
(1) A. C. Hardy, "Handbook of Colorimetry", M.1T. Press, Cambridge, Mass., 1936. (2) D. 8. Judd and G. Wyszecki, "Coior in Business, Science and Industry", 2nd ed., John Wiley and Sons, New York, 1963. (3) F. W. Billmeyer, Jr., and M. Saltzman, "Principles of Color Technology", John Wiley and Sons. New York, 1966. (4) G. Koctum, 'Reflectance Spectroscopy", Springer-Verbg, New Yo&, 1969. ( 5 ) R. W. Frei and J. D. MacNeil, "Diffuse Reflectance Spectroscopy in Environmen!al Problem-Soiving", CRC Press, Cleveland, Ohio, 1973. (6) F. A. Rodrigo, A m . HeartJ., 45, 809 (1953). (7) C. W McKeehan and J. E. Christian, A m . Heart J . , 46, 631 (1957). (8) E. V. Ashburn and R. G. Waldon. J . Pharm. Sci.. 46, 442 (1956). (9) N. B. Guerrant. J . Agnc. Food. Chem.. 5 , 207 (1957). (10) D. W Kent-Jones and W. Martin, Analyst (London), 75, 127 (1950). (1 1) K. Yamaguchi, S. Fujii, T. Tobatata, and S. J. Kato, J . Pharm. SOC. Jpn., 74, 1322 (1954). (12) W. W. Coblentz. Natl. Bur. Stand. ( U . S . ) .Bull, 9, 283 (1913). (13) J. T. Gier, R. V. Dunkle, and J. T. Bevans, J . Opt. SOC. A m . , 44, 558 (1954) (14) J. V. White, J . Opt. SOC.A m . , 54, 1332 (1964). (15) B. E. Wood, J. G. Pipes, A. M. Smith, and J. A. Roux, Appl. Opt.. 15, 940 (1976). (16) G. Kortum and H. Delfs, Specbochim. Acta, 20, 405 (1964). (17) W. R. Blevin and W. J. Brown, J , Sci. Instrum., 42, 385 (1965). (18) S. T. Dunn, J. C. Richmond, and J. A. Wiebelt, J , Res. Natl. Bur. Stand., Sect. C, 70, 75 (1966). (19) S. T. Dum, "Flux Averaging Devices for the Infrared", Natl. Bur. Stand. ( U . S . ) Tech. Note. No. 279 (19651. (20) k. R . Willey, Appl. Spectrosc., 30. 593 (1976) (21) P. Kubelka and F. Munk, Z . Tech. Phys., 12, 593 (1931). (22) P. KuDelka, J . Opt. SOC.A m . , 38, 448 (1948). (23) P.Kortum, W. Braun, and G. Herzog, Angew. Chem., Int. Ed. Engl., 2, 333 (1963). (24) G. Duyckaerts, Analyst(London), 84, 201 (1959). (25) J. L. Koenig, Appl. Spectrosc., 29, 293 (1975). (26) M. K. Antoon, J. H. Koenig, and J. L. Koenig, Appi. Spectrosc., 31, 518 (1977). (27) R. W. Hannah, S. C. Pattacini, J. G. Grasselli, and S. E. Mocadlo, Appl. Spectrosc., 32, 69 (1978). (28) C. J. Percival and P. R. Griffiths, Anal. Chem., 47, 154 (1975). (29) M. M. Gornez-Taylor, D. Kuehl, and P. R. Griffiths, Appl. Spectrosc., 3 0 , 447 (1976). (30) M. M: Gomez-Taylor and P. R. Griffiths, Appl. Spectrosc., 31, 528 (1977).

I \

p ,)

I
4000
3000

2003 FREOUENCY \ C Y ' 1

003

Figure 10. Diffuse reflectance spectra of a sample of powdered poly(dimethy1fulvene)as a function of time; (A) 15 min, (B) 7 0 min, (C) 255 min, and (D) 1285 min after sample preparation

heterogeneous reactions may be more realistically studied than with more conventional techniques. The use of DRIFT spectrometry for studying the kinetics of reactions is illustrated in Figure 10 for a powdered sample of a rather unstable polymer, poly(dimethylfu1vene). This compound reacts with air and, at least in the latter stages of the reaction, the oxidation rate appears to be diffusion controlled. Although the time interval between the spectra shown in Figure 10 is relatively long, useful spectra could readily be measured at intervals of approximately 1 min.

RECEIVED review May 22, 1978. Accepted August 24, 1978. for
Work supported by Merrell-Kational Laboratories, Division of Richardson-Merrell, Inc., Exxon Research and Engineering Corporation, and the Environmental Protection Agency.

You might also like