Photoacoustic Effect and The Physics of Waves

You might also like

Download as pdf
Download as pdf
You are on page 1of 7
Photoacoustic effect and the physics of waves F. Alan McDonald Department of Physics, Southern Methodist University, Dallas, Texas 75275 (Received 20 July 1978; accepted 9 July 1979) The photoacoustic effect is the basis for photoacoustic spectroscopy, a rapidly developing technique for studying optical absorption in many and liquid samples which are difficult to study by conventional means. Understanding of the effect requires knowledge of both acoustic and thermal waves, making this an interesting problem in the application of fundamental wave concepts and techniques. Both qualitative and quantitative discussions are presented for use as classroom examples at various levels. I. INTRODUCTION Almost 100 years ago Bell! discovered that an acoustic signal could be produced by illuminating a solid sample in an enclosed cell with periodically modulated (chopped) light. Subsequent work? showed that this “photoacoustic effect” also occurred with liquid and gas samples. Interest in the effect has been rekindled in recent years? because it Provides a means for studying optical absorption in various samples which are difficult to study by more conventional spectroscopic techniques. Photoacoustic spectroscopy of ‘Bases has been well understood for a number of years, but only recently®-9 has the photoacoustic effect with solids and liquids been given a satisfactory theoretical explanation in order to provide a basis for spectroscopic and other studies of such materials. In photoacoustic studies of solids or liquids the sample is placed in a closed cell containing a gas (such as air) and is illuminated with chopped monochromatic light (see Fig, 1). Absorption of the chopped light produces periodic heating of the sample, which causes mechanical motion of the sample (acoustic waves) as well asa temperature vari- ation in the sample and in a thin layer of the gas at the gas-sample boundary (thermal waves). The resulting pressure variation in the gas is detected by a microphone in the cell wall. This acoustic signal provides the basis for studying optical absorption in the sample, assuming a suitable theoretical model to relate absorption to signal, Photoacoustic techniques have been successfully used in studying various opaque materials, as well as samples in powder or gel forms, and thin films; studies which are dif= ficult or impossible with standard transmission or reflection techniques. The theoretical treatment of the photoacoustic effect is an interesting problem from a pedagogic viewpoint. It in- volves acoustic and thermal waves, as well as optical ra~ diation, providing an opportunity to compare and contrast the properties of these distinct wave types. Further, it shows the need for the physicist’s broad training, as one must synthesize knowledge from several seemingly distinct fields in order to understand the effect and use it for material analysis. The physical mechanisms underlying the effect may be treated qualitatively at an elementary level, or they may be considered quantitatively at a more advanced level (when students have had an introduction to wave equations and boundary value problems), The purpose of this article is to outline the physics involved in the photoacoustic effect, with the hope that it may serve as a useful classroom ex- ample of “applied physics.” 41 Am.J. Phys. 48(1), Jan. 1980 (0002.9505/80/010041-07500.50 Il, QUALITATIVE DESCRIPTION OF THE EFFECT. In this section we give a qualitative discussion of the photoacoustic signal produced when light is absorbed in a sample. After first indicating how thermal energy is pro- duced, we describe the nature of thermal waves and their ‘contribution to the signal, and then describe the contribu- tion that acoustic waves may make. The effects of thermal and acoustic waves are finally combined in a “composite- piston” model, which has been proved successful in repre- senting the photoacoustic signal. We focus on the important question of whether the signal is linear in the absorption coefficient, for then a pho- toacoustic spectrum can be directly interpreted as an ab- sorption spectrum. The answer to this question, for non- transparent samples, depends on the chopping frequency, so this frequency dependence will also be described. A. Optical absorption The absorbed optical energy which ultimately causes the photoacoustic signal is that fraction ofthe total absorbed ‘energy which is converted to heat (phonons) via nonradia~ tive deexcitation processes in the sample (photoacoustic study thereby complements fluorescence spectroscopy in the information it gives about a sample). Atomic or mo- lecular transitions or band structure (as examples) can be studied, provided only that the excited levels can undergo such nonradiative deexcitation; These deexcitation pro- cesses, and the subsequent thermalization of the energy, ‘occur on a time scale which is quite rapid compared to the light-chopping period (typically 0.1 sec), so one effectively has an instantaneous conversion of incident optical energy into a periodic thermal energy source distribution in the sample (one clearly can ignore the time for the light signal to traverse the sample). If the light intensity is assumed uniform across the sample, the amplitude of the thermal source depends only on |x|, the depth in the sample, de- ‘creasing exponentially as absorption decreases the available light intensity. The intensity at any point can be represented as I(x) = Toexp (Bx)(A)(1 + cos wt), a) where 1s the incident intensity, isthe optical absorption coefficient, and @ (=27/) is the chopping frequency; note that x is negative in the sample. Then the thermal energy produced per unit length is (© 1980 American Association of Physics Teachers 41 Tacident Backing As Semple] cos [window i © Tath Fig.l. Schematic diagram of typical photoacoustic cell, showing coor dinates used intext. ate y COs wi an Bloexp (8x)('2)1 + 1), Q) where we assume sinusoidal chopping and complete con- version of absorbed to thermal energy for simplicity. The periodic term ultimately produces the detected pho- toacoustic signal; the constant term leads only to a small static temperature variation in the cell Thermal waves The original explanation of the photoacoustic signal*® considered only the effect of thermal waves in the sample, and itis convenient to consider them first; in fact, in some situations acoustic waves in the sample do not contribute appreciably, so the thermal wave explanation is adequate. The most important fact about thermal waves is that they are rapidly attenuated. A plane thermal wave has the form exp (jat-ox), where o = (1 + )(«/2a)"? (see Sec. IIA) Iris thus damped by a factor e-' in a distance called the thermal diffusion length, 4. = (2c/«)'/?; in one thermal “wavelength” (\y = 2x) the amplitude is decreased by a factor exp (~2x) ~ 0.002. (At f = 100 Hz, typical diftu- sion lengths are 2 x 10-3, 3 X 10-2, and 7 X 10-2 em for water, air, and silver, respectively.) Any thermal wave generated at a sample depth greater than Ay will thus have negligible effect on the gas, and will not contribute to the photoacoustic signal. In addition the thermal waves which do reach the gas are confined to a thin layer near the sample, The transmitted thermal wave leads to expansion and contraction of the gas boundary layer which can be char~ acterized as a thermal “piston.” The thermal piston dis- placement is found? by recognizing that each infinitesimal layer of gas near x = 0 undergoes fractional, essentially isobaric expansion /contraction proportional to the ratio of periodic gas temperature ry to ambient temperature To, The net thermal displacement is the sum of these infini- tesimal displacements: tedx; 8) the upper limit need only be chosen such that 7 ~ 0. ‘The pressure variation in the gas is obtained by assuming an adiabatic response to the thermal piston. The gas length is typically a small fraction of the acoustic wavelength in the gas, so the gas may be treated as a bulk sample (reso: nance effects may occur at high frequencies, f ~ 10 kHz; see Sec. V). If PY? = const, where P and V are gas pressure and volume, and + is the ratio of specific heats, then Pe = YPo(BV/V) = (yPallg)Ox1. a) where py is the periodic gas pressure, Po is the ambient pressure, and /y is the gas length (Fig. 1). Thus the pho- 42 Am.J. Phys, Vol. 48, No.l, January 1980, toacoustic signal depends on physical and thermal param- eters of the sample and cel system, as well as on the optical absorption in the sample. This allows for some optimization of the signal through cell design. The key relation for photoacoustic analysis is that be- tween the sample's absorption coefficient 8 and the signal Pe. The most desirable relation is a linear one, for then Photoacoustic spectra are directly comparable to conven tional absorption spectra. Equations (3) and (4) show that B affects the signal through the dependence of the gas temperature r, on 8, while 7, is determined by the heat flux reaching the sample surface (at.x = 0). This heat flux can be affected only by eneray absorbed within a distance ), of the surface, as already explained. If BA, <« 1, Eq. (2) shows that the energy deposited near the boundary is proportional to 8, so that the signal will be linear in B as desired. How- ever, if the sample is so highly opaque that al! the incident energy is absorbed within the distance 2, (i.e..if BA, ~ 1), then the heat flux, and thus the signal, will be independent of 8. Nevertheless, since A, (or 2) is proportional to w!/2, one can imagine increasing the chopping frequency to achieve the desirable condition that 6X, < 1. (Note that for transparent samples 81, < 1, so the energy deposited in any thickness is essentially proportional to 8.) The possibility of obtaining favorable conditions for photoacoustic study of opaque samples by varying the chopping frequency leads one to look at the overall depen- dence of p, on w. The character of thermal waves is again the primary factor. The gas temperature will have the form "y exp (fiat ~ ox), so that the heat flux at x = Os, K_(~07/Ox) = keg, where Ky is the gas thermal con- ductivity. If BN, >> 1, stich that all the optical energy con- tributes to the heat flux at the boundary, the flux is inde- pendent of frequency. Since a « wV2, we must then infer that C, is inversely proportional to «!/2, In addition the exponential dependence of « shows that the region over which the integrand of Eq, (3) is appreciable shrinks as og increases, such that an additional factor of «o-"/? occurs in 6x,. For these conditions 5x, is thus inversely proportional tow. Consequently an attempt to achieve linearity in the relation of signal to absorption coefficient results in an ac- companying decrease in the signal magnitude (see Fig. 5). This reduction may be acceptable depending on such im- portant factors as signal-to-noise ratio, detection sensitivity ete. If the condition BA, < 1 holds, then the boundary heat flux will depend on co. This is because the region contrib- ting to the heat flux decreases as @="/2, so the flux (for given 8) will be proportional to w~"/?, When this factor is Combined with that discussed in the previous paragraph, the signal amplitude is seen to be proportional to w~¥/. The iransition in w dependence occurs near the frequency at which Bus ~ 1, with the sample diffusion length. C. Acoustic waves Recently? it has been shown that acoustic waves in the sample may contribute appreciably to the photoacoustic signal. These waves are caused by local expansion/con- traction of the sample due to periodic heating. They affect the signal by causing mechanical vibration of the sample surface, which would lead to pressure variation in the gas even in the absence of the thermal piston effec. It is shown in Sec. III that the amplitude of surface vibration is pro- F.Alan McDonald 42 portional to the total optical energy absorbed in the sample, and inversely proportional to «. The former property is reasonable because the acoustic waves are essentially un damped; waves generated throughout the sample are su- perimposed at the surface and, as the sample thickness is very much less than the acoustic wavelength, the superim- posed waves are virtually in phase, leading to constructive interference. The result is that the surface amplitude is linear in 8 only for transparent samples, because only then is the absorbed optical energy proportional to 8. D Composite piston model ‘The combined effect of thermal and acoustic’ waves in the sample is obtained by simply superimposing the thermal piston displacement on the mechanical displacement of the sample surface, giving a “composite piston” displacement This should seem physically reasonable, and we have shown® that this composite piston model (CPM) is in agreement with more elaborate calculations for typical experimental situations. The algebraic result is given in Eq (20), but some qualitative comments can be made. The mechanical displacement can become significant, compared to the thermal, when the attenuation inherent in thermal waves limits the optical energy contributing to the thermal piston displacement, compared to that contributing to the ‘mechanical. This will occur for “thermally thick” samples (Ig > Xx) when 8; < 1. Transparent samples certainly fit this condition, highly opaque samples (31, <« 1) do not, and others fall in between. The mechanical displacement is understandably proportional to the thermal expansion coefficient of the sample, so the magnitude of this factor enters into the comparison of mechanical and thermal displacements. ‘The dependence of the CPM signal on 8 and w can be somewhat complicated, since two terms having different dependence are being added. If one term or the other dominates the sum then itis easy to specify the dependences, In addition, transparent samples will always give signals linear in 8 because each term is linear. Also, any sample having BA, > 1 must give a signal proportional to a=!, again because each term has that dependence. More ‘quantitative information is presented in Sec. IV (See Figs. 3-5), The presence of an appreciable mechanical term in the CPM may make lifea little more complicated for the pho- toacoustic spectroscopist. In any spectrum which includes an absorption coefficient satisfying 8X, ~ 1 there may be distortion in the photoacoustic signal, i¢., lack of linearity in B. Proper interpretation of the spectrum will require extraction of 8 from the data, rather than the direct com- Parison of photoacoustic spectrum with a conventional absorption spectrum. Il, QUANTITATIVE TREATMENT In this section we present a more quantitative treat- ‘ment of the photoacoustic effect. After considering the nature of the thermal and acoustic wave solutions, we give an approximate treatment of the boundary-value problem, leading to the composite piston model result, Eq. (20). The complete boundary value problem (in one dimension) has been presented previously.? 43 Am.J. Phys, Vol. 48, No.1, January 1980 ‘A. Mathematical basis The quantitative treatment of the photoacoustic effect in fluids is based on coupled linear equations for the time- dependent pressure p and temperature T (ignoring viscous effects) (3) 6) here po and To are ambient density and temperature, B is the isothermal bulk modulus, 87 is the (volume) coefficient of thermal expansion, «is the thermal conductivity, ais the thermal diffusivity (a= kpo/Cp, with Cp the heat capacity at constant pressure), and S represents the thermal energy source due to optical absorption. Ifthe gas is assumed ideal, then for it B and 8 are replaced by Po (the ambient pres- sure) and Ty!, respectively; the source term S is absent in the gas. In solids one would employ coupled equations for medium displacement and temperature; however, it can be shown!" that plane wave solutions of such equations for isotropic solids have a form identical to that in nonviscous liquids, so we will use Eqs. (5) and (6) as the basis for our mathematical development (factors of order 1, involving ic constants, occur in the solutions for solids). We consider one-dimensional solutions of Eqs. (5) and (6), assuming time dependence exp (jut) for all quantities, with real parts of the complex solutions representing the physical variables. The uncoupled equations (i. ignoring the terms on right-hand side of each equation) are the ‘acoustic wave equation and the thermal diffusion equation, respectively. The acoustic wave equation has the sotutions exp (jest + jkx), with KP = pow/B. Q) The thermal equation leads to terms having similar form, exp (jut + ox), but jaja, (8) so that (1+ f(w/2a)", (Oo) The imaginary part of isthe thermal wave number, while the real part leads to rapid attenuation of thermal waves. Mathematically, the attenuation results because the first time derivative of 7 appears in Eq. (6), rather than the second time-derivative. Physically, the attenuation relates to the phase relation between heat flux and heat gain loss by the material. A typical thermal waveform is shown Fig. 2. The presence of a periodic thermal source S {omit- ting the constant term in Eq, (2)] requires an additional term in r of form exp (jer + Bx). The complete solution of the one-dimensional problem ‘of the photoacoustic cell requires knowledge of proper so- lutions in each region, together with appropriate boundary ‘conditions. The coupling of p and 7 in Eqs. (5) and (6) re- ‘uires that each contain terms of both acoustic and thermal types, as well as the source-related term. Then boundary conditions on temperature, heat flux, particle velocity, and pressure are imposed in order to evaluate coefficients for all terms. This solution has been obtained, by numerical and algebraic techniques,° but the details are omitted here be- ‘cause a simpler approach gives essentially equivalent results F.Alan McDonald 43 e287 q o 3 30 75 To uy Fig. 2. Illustration ofa thermal wave, #= 4 exp (ist-ox); the real pat of 7 is shown on the graph. The waveform i shown at ¢ = O and at 1 = 0.257, where T= f-!, Note the distortion from sinusoidal form due to damping in most situations, with the advantage of accompanying physical insight into the problem. B. Composite piston model CPM extracts the essential physics from the complicat- ed mathematical procedure just outlined. We have already discussed the model qualitatively in Sec. 11 D, so here we concentrate on its mathematical basis. Implementation of this simplified treatment rests on two facts: the temperature distribution in the system is essentially unaffected by the term coupling it to the pressure; the surface motion of the sample is virtually independent of the periodic pressure variation in the gas. One is then left with two independent problems to solve: determination of the temperature dis- tribution, leading to an expression for the coefficient Cy in Tg, and ultimately giving the thermal piston displacement ‘x, (see Sec. IIB); determination of the amplitude of sur- face vibration of the sample due to the presence of the thermal source S. The temperature distribution problem is further sim- plified by the attenuated nature of thermal waves. The gas is usually “thermally thick” (ly >> ig, where Hy is the gas thermal diffusion length), so that only thermal waves propagating in the +x sense need be included, and no thermal wave exists in the cell window. The backing is as- sumed thermally thick also, so that only a thermal wave in the —x sense occurs there. Then the temperature solutions in gas, sample and backing become = Ceexp (—0,x), (10) T= Cyexp ( —o4x) + De exp (a,x) + Taexp (Bx), ay. a (2) If temperature and heat flux are set equal at x = O and x =1,, one obtains four algebraic equations for Cy, C,, Ds. and Dp. If the solution for Cy is inserted in rg and the in- tegral in Eq. (3) is evaluated, one obtains an expression for x. Here we will assume further that the sample is ther- maily thick (true of most samples except in thin-film form) so that further simplification can be made to give 4x; = (~jlo/2gTo)(B/opoCp(g + Ir + 1, (3) Deexp (003). 44° Am.J. Phys, Vol 48, No.1, rary 1980 where g = Ketg/Ke0s,7= Blas. The acoustic wave problem is most simply treated by realizing that, for the acoustic mode, Eq. (6) reduces to kor wor the first term in Eq. (6) is negligible because it represents heat conduction, which occurs at a speed much smaller than the acoustic wave speed, so that essentially all the energy is available to produce the time-varying temperature . Then Eq. (5) becomes ap jg Br Blo att a °C, 2 ay where the time-dependent part of Eq. (2) has been used for the source term. The solution of this equation is = Acexp ( —jkex) + Byexp (iksx) + Agexp(Bx), (16) where Ag = (~joBBylo/2Cp (8? + x,2)-!, The velocity may be related to the pressure by i 2p @p0 2x" such that appropriate boundary conditions may be imposed: 0, = Qatx = =p, = Oat x = 0. These conditions give 4, and B;, which then allow the velocity at the sample surface to be written as 6 = Brlal2poCp)(1 + K3/B)"1 = exp(-Ble)], (18) Where we have used the condition ksls. << 1. Making the reasonable assumption that ks < 8, true for all but very transparent materials, and using the relation between dis- placement and velocity (x = —je/w) we obtain the me- chanical (surface) displacement as: (—/Brlo/2wpoCp){1 — exp(—Bl,)}. (19) ‘The mechanical displacement is proportional to the ex- pansion coefficient, Br, and it is also proportional to the ‘energy absorbed in the sample, (1/2) loll — exp(—Bl)] [The displacement 5x», may also be obtained by directly considering thermal expansion, providing ksly < 1. The expansion of a sample layer Ax will be B77 Ax, with Eq, (14) giving + = $/JwpoC,). Integrating over the sample depth gives the result of Eq. (19). The photoacoustic signal based on the CPM is found by adding the thermal and mechanical displacements, Eqs. (13) and (19), and then using Eq. (4) with the net dis- placement rather than dx, alone. The algebraic result is 2S; «aay ra exp (8x), a7) Xp @ Ip 2p0Cp 8 *\e, Tole + D+) = + Br(1—exp(—8t,)}. (20) ‘The reader is reminded that certain limiting, yet reasonable, assumptions have been made to obtain this specific ex- pression: the gas and sample are thermally thick, but acoustically thin. IV. DISCUSSION OF RESULTS The dependence of the photoacoustic signal on absorp- tion coefficient is illustrated in Figs. 3 and 4. The sample F.Alan McDonald 44 Fig. 3.. Dependence of pho- toacoustic signal (PAS) on biorpion coefficient 8 for transparent samples, Solid and short-dashed cures show values including the effect of acoustic waves inthe sample, with x = ly and OS, re- spectively. | Long-dashed curves show values omitting the effet of acoustic waves in ‘the sample. Material param- cles are given in Table . In addition, for water we use Br em) is assumed to be a dye-water solution, that is, the thermal and material parameters are those of water, the concen- tration of dye determining the absorption coefficient. In Fig 4, for opaque samples, the signal is almost entirely due to the thermal piston term, with some deviation oceurring at lower 8 values. The signal is (roughly) linear in 8 until “saturation” occurs, where 8, ~ 1; for higher 8 values the signal is independent of 8, as already discussed (Sec. II B). In Eq. (20) this saturation results because B(r + 1)~! > «. In Fig. 3, for transparent samples, the mechanical term becomes the dominant one, atleast at the higher chopping frequencies. At lower 8 values the signal is again linear in 8, because both mechanical and thermal terms are linear. In Eq. (20) the linearity of the mechanical term arises be- cause [1 ~ exp(Bl)] > Bl, if Ble —> 1. The frequency dependence of the signal is shown in Fig. 5. For lower 8 values the mechanical term becomes im- Portant as the frequency increases, because the thermal term behaves as w~2/2, but the mechanical term decreases only as "1, In Eq. (20) the extra factor of w™1/? arises because ag = w~"? [at lower frequencies where oy is small enough that r >> 1, a, (r+ 1) —> Bog/o,, which is inde- pendent of w]. For higher 8 values the thermal term dom- inates the signal. ors 0 0F 10 tem) Fig. 4. Dependence of photoacoustic signal (PAS) on absorption coefficient for opaque samples. Notation is that of Fig. 3 45° Am.4. Phys. Vol. 4, No.1, January 1980 Table 1. Material properties used in theoretical ealeulations of pho- toacousie signal Material alm?/s) x(W/m°K)—__plke/m) Air 215% 10-8 0.0258, 12 Water 1.46 10-7 061 1000.0 Brass 29 x 10-5 963 Plexighs 1.13 10-7 016 Glass S110 038, The results of the CPM calculation agree with those of the full coupled-equation treatment outlined in Sec. III A, over the usual frequency range (10-1000 Hz). It has also been demonstrated? that this model gives results consistent with experimental data, so that the CPM, particularly Eq. (20), provides a handy basis for photoacoustic studies. One application in which this is important is the determi of absolute absorption coefficients (rather than relative spectra). It has been shown'? that the frequency dependence of the photoacoustic signal may be used to determine 8, essentially because the character of a curve such as one of those of Fig. 5 is uniquely related to a single B, provided thermal and material parameters are known. . FURTHER WAVE EFFECTS Resonance may occur in the gas at greater-than-nor- ‘mal chopping frequencies, because the gas acoustic wave- length is no longer much larger than the gas length. Such effects may be treated by a simple extension of the CPM of Sec. III. One assumes the piston displacement dx = 6x, + dxm at x = 0, as well as vg = Oat x= Jp, using these two boundary conditions to determine 4, and By in Eq. (11). ignoring the thermal mode terms. The result for the pressure amplitude is simply py of Eq. (20) times a resonance factor R(x) = Kgl cOs(Kgly — Kex)/sin( Kyle). (21) For f < 1000 Hz and /, = 1.74.em, R(x) ~ 1, but for f > 1000 Hz the spatial dependence given by R(x) becomes important, such that the signal depends on microphone placement. Illustration of the resonance effect is included in Fig. 5 for x = [p/2 and x = lp. Again, this extension of the CPM gives results which agree with the full coupled- equations treatment. For many types of waves the determination of trans- = Fig. 5, Dependence of pho- 2 toacoustic signal (PAS) on 2 chopping frequency fo three values of absorption coefi= cient 8. rr a F.Alan McDonald 45 mission and reflection coefficients at a boundary is an im- portant problem. For the thermal waves considered here, ‘transmission at the gas-sample boundary affects the mag- nitude of the photoacoustic signal; for thermally thin samples reflection at the sample-backing boundary leads to superimposed waves at the gas-sample boundary, and this also affects the ultimate signal. Consider a thermal wave incident on the plane x = 0 from the sample side: T. = Ceexp (— osx) + Deexp (ass), (22) 14 = Ceenp (“a4 23) ‘The boundary conditions of continuous temperature and thermal flux allow Ce and D, to be determined in terms of the incident amplitude C,; itis C, that is important in Eq, (3). The temperature transmission coefficient 'r is Cy/C;, the reflection coefficient I'g is Dy/C,. Carrying through the algebra one obtains Te=(1- gl te) (24) Tr = 2/1 +8). 5) For a water-air boundary one finds g < 1, so that this sys- tem is virtually optimized to give maximum temperature coefficient Cy." Differences in reflection of thermal waves at the sam- ple-backing interface can lead to substantial variation in the photoacoustic signal for thermally thin samples. [For ‘these samples one can neglect the acoustic mode, but the thermal mode must be treated as outlined following Eq (12), without the simplification leading to Eq. (13): the algebraic result is given in Ref. 6,] An interesting com- parison is shown in Fig. 6, where the signals for plexiglas and brass backing are shown. The reflection coefficient is like that in Eq. (25), but with g replaced by b = Kpr/&s0 For plexiglas backing (water sample) b ~ 0.3.50 l'g = 0. for brass b = 11,80 T'g = —0.8. When the sample thic ness is small (J, < 4), direct and reflected waves reach the plane x = 0 with little attenuation, so that the reflection coefficient determines the nature of the interference. A plexiglas backing gives a reflected wave in phase with the direct wave, whereas a brass backing causes a 180° phase change on reflection (I’ <0). The constructive interfer- ence in the case of plexiglas leads to a much greater signal 2 Pasinsme) 10 Fig. 6, Dependence of photoacoustic signal on sample thickness for thermally thin samples (/'= 100 He). Xi the rato of sample thickness to thermal diffasion length, In this Figure the slid curves show values for Pleniglas backing, dashed curves for brass backing. Cell parameters are ‘those in Fig. 3. The optical absorption length is wg = 8 46° Amd. Phys., Vol. 48, No. January 1980 than that using brass. As , > j,, however, the reflected wave is attenuated so that it contributes less to the signal, and the signal is thereby reduced. As J, becomes greater than ws the phase change due to the propagation through the sample results in a slight destructive interference effect, which in turn disappears as J; increases such that attenua- tion eliminates any effect of reflected waves. With brass backing the increased attenuation as /, increases causes less destructive interference, such that the signal increases. For 1, > py the signal with brass actually becomes slightly greater than that with plexiglas, because the phase change ‘due to propagation combines with that due to reflection to cause constructive interference, albeit with a greatly at- tenuated reflected wave amplitude. In the lower curve for plexiglas the direct wave is still increasing as ly —> as, more than compensating for the loss of reflected wave. Finally, if l, > is the reflected wave does not contribute to the sig- nal, so the signal is independent of backing material. VI. CONCLUSION Photoacoustic spectroscopy, and other studies using the photoacoustic effect, add to the arsenal of techniques available for the study of various aspects of matter. Inter- pretation of data requires an understanding of the pho- toacoustic effect, particularly the dependence of signal on absorption coefficient and chopping frequency. It is sug- gested that the wave approach described herein makes understanding of the effect accessible even at an elementary level ‘Variations on the photoacoustic technique described here have been suggested recently. Direct detection of the acoustic wave has been achieved in liquids" and solids.' In another case'® the temperature wave (“second sound”) has been detected directly by immersing a solid sample in liquid helium at low temperature. It appears we may look forward to expanded use of techniques of this general sort in the near future ACKNOWLEDGMENT The author wishes to express appreciation to Professor Grover C. Wetsel for suggesting the photoacoustic effect as an interesting problem and for continuing useful dis- cussions of the effect and its application, "AG. Bell, Am. J. Sci 20, 308 (1880) 24, Tyndall, Proc. R. Soc. London 31,307 (1881); W.C. Rontgen, Philos Mag. 11, 308 (1881); A. G, Bell, bid. 1,510 (1881), 5A, Rosencwag, Phys. Today 28, 23 (1975), and references therein, *D. W. Hill and T. Powell, Non-Dispersice Infrared Gas Analysis in Sclence Medicine, and Industry (Plenum, NY. 1968) 51.6. Parker, Appl Opt. 12,2974 (1973). 6A. Rosencwaig and A. Gersho, J. Appl. Phys. 47,64 (1976) TEA. MeDonald and G. C. Wetsel, Jr. J. Acoust. Soe. Am. 60, $52 (1976): G. C, Weisel, Je-and F_A. McDonald, ibid. 60,882 (1976). 5L.C. Aamodt, J.C. Murphy, and JG, Parker, App, Phys. 48, 927 (1979). °F. A. MeDonald and G. C. Wetsel, dry J. Appl. Phys. 49, 2313, (1978). "0p. M. Morse and K. U. Ingard, Theoretical Acoustics (MeGraw-Hil, NY, 1968), UR.N. Thurston, Physical Acoustics (Academic, NY, 1964) Vol, Part. A F.Alan McDonald 46 "9G. C, Wels Je and F. A. McDonald, App. Phys. Lett 30, 252 (1977). ">The signal, from Eg. (3, also depends on 0. orc: it has been pointe out in Ref. 8 that a helium el givesa signal almost thre times that from an air el al ther things being equal, which is due tothe 47 Am.J. Phys. Vol 48, No.1, January 1980 thermal diffusty (a) of helium being about eight times that of ar. "*Y, Kohanzadeh, J. R, Whinnery, and M. M, Carol. J, Acoust So. Am. 87,67 (1995). 154, Hordvik and H. Schlossberg, Appl. Opt 16,101 (1977) ‘6B. Smith and G. A. Laguna, Phys. Lett. A $6, 223 (1976), F, Alan MeDonald a

You might also like