J. Phys. E: Soi. strum. Vol 16, 1983. Printed in Grea Britain Determination of the thickness and optical constants of amorphous silicon IR Swanepoel Deparment of Physies, Rand Afrikaans University Johannesburg, South Africa Received 27 May 1983 Abstract. The rigorous expression forthe transmission ofa thin absorbing film on a transparent substrate is manipulated to yield formulae in closed form forthe refractive index and absorption coeficient. A procedure is presented to calculate the thickness to accuracy of better than 1% with similar accuracies inthe \alues of A method to correct for errors due t slit width is also givea. Various formulae to calculate the absorption ccoeficient accurately over almost three orders of magnitude are ‘discussed. Only data from the transmission spectrum are used tnd the procedure is simple, fast and very accurate. Al formulae are in closed form and can be used on a programmable pocket calculator. 1. Introduction Amorphous silicon or hydrogenated a-Si:H is an important ‘terial for photovoltaic devices. These devices consist of films with thicknesses of about I um and itis important to know the refractive index and absorption coefficient as function of wavelength to predict the photoelectric behaviour of a device. Knowledge of these optical constants is also necessary to etermine the optical gap or 10 verify theoretical models of (Clark 1980) The thickness of films can be determined using a surface- profiling stylus or by various interferometric methods (Bennet and Bennet 1967). A method is proposed in this paper to determine the thickness from the interference fringes of the twansmission spectrum to an accuracy of better than 1%, provided the films are of good quality. The refractive index and absorption coefficient are usually determined by elaborate computer iteration procedures (Lyashenko e/ af 1964, Wales et al 1967, Szczyrbowski et al 1977) using both the transmission and feflection spectra. A simple straightforward process has also. been devised (Manifucier er af 1976) for caleulating m and 2 but it will be ppoinced out in this paper that this process contains some mistakes. Methods are proposed in this paper to determine m(a) and (i) by simple straightforward calculations using the ‘wansmission spectrum alone. The accuracy is also of the order ‘of 1% which is even better than the accuracy of the elaborate iteration methods. In the discussion only the spectrum in the optical region is ‘considered. All formulae are, of course, also valid in the neat- inftared region and for films with small optical gaps it may be ‘essential to do measurements inthis region 2. Theory ‘The practical situation for a thin flm on a transparent substrate is shown in figure 1, The film has thickness d and complex refractive index m="n—ik, where m isthe refractive index and & the extinction coefficient which can be expressed in terms of the absorption coefficient using equation (27). The transparent © 1983 The Insitute of Physics Substrate Figure 1. System of an absorbing thin film on a thick finite transparent substrate substrate has a thickness several orders of magnitude larger then d and has index of refraction s and absorption coeficent a, = 0. ‘The index ofthe surrounding air snp = 1, Rigorous analysis has to take into account all the multiple reflections at the three interfaces when calculating 7; If the thickness d is not uniform or is slightly tapered, all, interference effects are destroved and the transmission is a smooth curve as shown by the dotted curve T; in figure 2. The spectrum can roughly be divided into four regions. In the Uwansparent region 2=0 and the transmission is determined by fand $ through multiple reflections. In the region of weak absorption a is small but starts to reduce the transmission. In the region of medium absorption 2s large and the transmission decreases mainly due to the effect of 2 In the region of strong. absorption the transmission decreases drastically due almost exclusively to the influence of a. The smooth transmission curve Tr is often used (Freeman and Paul 1979) to determine a2) in the optical and infrared region using a formula given in the appendix (A3), Ifthe thickness d is uniform, interference effects give rise 0 f spectrum shown by the full curve in figure 2. Far from being a nuisance, these fringes can be used to calculate the optical ‘constants of the film as wil be shown in this paper. ‘Considering the thick substrate alone in the absence of a film, the imerference-free transmission is given by the well known expression elt? TR where Relis—Dis+ DF 0 and @ ‘The basic equation for interference iingesis tnd= mi °Determination of thickness and optical constants of a-Si:H absorption Strong —t4— Netlar +4 eos cs Figure 2. Simulated transmission spectrum (ull curve) for @ Lum film of 2-Si:H on a finite glass substrate with transmission 7, Curves Ty, Ta. T, and T= according to the text where mm is an integer for maxima and half integer for minima. Equation (3) contains information on the product of» and d and there is n0 way of obtaining information on either 2 or d separately using this equation only ‘The transmission T for the case of figure 2 is a complex function (Keradec 1973, Mini 1982) and is given in the appendix an TM. It's is known it is convenient to write the above equation in terms of m(2) and the absorbance x2), where x is defined in equation ($ 1) od 2 T=TMn9 ‘The expression (A1) becomes much simpler if we put k=0, an approximation that is indeed valid over most of the region of the spectrum in figure 2.(A1) then becomes ax ‘Ba Crcoso= Det o where A= l6n's (Sa) (n= ines) (3b) 2(n? = in? 5) (Se) Da(n—1n-s) (Sd) (se) we The extremes ofthe interference fringes ean be writen a Ty= Ae o T= o B+ Cee De For further analyses Ty and T= are now considered to be continuous functions of 4 and thus of n(2) and x42) (Manifacier ef al 1976), as is shown by the envelopes in figure 2. For any i, Ty has a corresponding value Ty. In figure 2 for example Tir has the corresponding value Ty and Tay has the corresponding value Ti 2.1. The transparent region In the transparent region a=0 or «= 1 in equations (6) and (7) Substituting equations (5) into (6) yields @ Equation (8) is identical to equation (1) and the maxima of the interference fringes area function of s only and coincide with 7. When the maxima depart from T, it denotes the onset of absorption. Equation (8) can be used to calculate s in the ‘tanspareat region using the form of equation (2). ‘Substitetng equation ($) in equation (7) for yields 4ats weniss Das (MP —9)!9)" o where TT is thus a function of both m and 5, and m can be calculated from T using equation (9). 2.2. The region of weak and medium absorption In this region 20 and x< 1. Subtracting the reciprocal of equation (6) from the reciprocal of equation (7) yields an rasR Swanepoel ‘expression that is independent of x es ao Tea Subsisng equation (5) ino (10) and sling for niles n= [No (Nt sy) ay tee Tu-Ts S+1 Me ET. Equation (11) can be used to caleulate (2) from Ty and Ta. It is identical to the formula derived by Manifacier er al (1976) Using the theory for an infinite substrate, Once n(i) is known, all the constants in equation ($) are ‘known and x can be calculated in various ways. Both (6) and (3) are quadratic equations in x that can be solved for x and the results simplified using equation (S). Solving equs at 2 as) oe were Bn's oa Ew Th Xn? —s?), ‘Solving equation (7) gives: Fa-[Eh * 13 as) oo were Bnés 2 52) Fao — 1Xn? ="), Adding the ecprocals of equation (6) and (7) yields 2Iu To aa) Ta Tn Solving fr. thie gives F-[F?=(n? = 1)? -s4)? wefan) as where 8nis aT and 2TWTs 16 TutTs ‘ From equation (14) and (4) it can be seen tht 7; represents a curve passing through the inflection poins of the fringes as shown in figure 2. ‘The imererence free transmission can be calculated fom the interference fringes by integrating equation (4) berween a maximum and an adjacent minimum Ler aw "= 5 |, Fa Creosee De ‘Assuming « narrow integration region where all parameters are ‘constant, the integral yields ax BGs DVB Ce DAT an 1216 ‘Substitution of equations (6) and (7) nto (17) yields T1= VT To us) Tr is thus just the geometric mean of Ty and T and equation (18)is a very useful relation. Solving equation (17) for x gives 2 Gle pie as) oH + Dor Equation (19) is equivalent to a well known equation often used in optical and infrared studies, (A3). It can also be used 10 etermine m from T in the tansparent region where 2=0, Putting x= 1 in equation (19) and solving form gives nas) 20) where a fel Gn? Manifacier etal (1976) also derived an equation for x using the theory for an infinite substrate. Tie formula is @ en Unilke the case with the refractive index, equation (21) is nor ‘equivalent to equations (12), (13. (15) and (19) 2.3. The region of sirong absorptton {In the region of strong absorption the interference fringes disappear. There is no way to calculate and x independently in this region from the transmission spectrum alone. Values of ‘can be estimated by extrapolating the values calculated in the fother parts of the spectrum. The values of x can then be calculated using any of the four formulae presented in the previous section with their appropriate curves. For very large 2 the four curves Ty, Ts, Ti and Tm converge to a single curve Ty. I imerference effects are ignored, equation (4) can be written for elas Dar Tes en 24. Numerical simulation ‘To test the accuracy of the theory presented here, a film with the following properties is postulated. Substrate refractive index (constant) Film thickness d= 1000am Film refractive index 3x10" Film absorption coeficient 15x10" pan 5 8 (ainam™Determination of thickness and optical constants of a-Si:H ‘The above values of n(2) and a(2) represent typical values for a-Si:H (Freeman and Paul 1979). The spectrum in figure 2is ‘a plot of 714) as calculated by (A1) using the above postulated properties of the film. The true values of n(2), a(2) and d can thus be used to determine the accuracy of ealeulated values. The ‘smooth envelopes Ty and Ty are constructed on this simulated spectrum and the values from these ‘experimental’ envelopes are used in the calculation inthe following sections. The width ofthe interference pattern (Ty ~T) inereases with increase in (n—5) ‘and statements about accuracy in the Following sections refer to ‘typical spectrum of @ SiH asin gure 2. 25. Infate substrate approximation ‘Another expression for T, assuming an infinite substrate and thus ignoring the contribution of muliple refections from the back of the substrate, has been used by many workers (Marifacier eral 1976, Lyashenko 1977, Hadley 1947). One equivalent form of this expression is given in the appendix (A2). ‘A plot of T calculated from equation (A2) using the fm properties given in the previous section is shown by the full ‘curve in Bgure 3. The broken curve shows again the values of T ealeuiated by equation (AI). Tt can be seen that equation (A2) yields higher values for both Thx and T than equation (A). A disturbing aspect is that ‘equation (A2) predicts a transmission inthe transparent region that is about 4% larger than the wansmission of the substrate alone. Ths is impossible as can be verified experimentally. Itis the author's opinion that caleulated transmission curves using equation (A2) or equivalent forms should be regarded with suspicion; the equation does nor represent the experimental ‘ales of 7 fr real fms on transparent substrates. Equation (A2) can also be written in the form of equations (4) and (8) forthe case of k=O. The only siference is in the expressions for B and D in (S6) and (Sd), ie. Bin + n+ 9" and Dan(n— 143) Since the expression for 1 equation (1D, is independent of B and D it is valid for both (A) and (Ad). The expressions for x do contain B and D and will thus be different for the two cases. 3, Determination ofthe refractive index ‘The refractive index of the substrate can be determined by measuring the transmission spectrum of the clean substrate alone and using equation (2) to calculate s. If significant dispersion is present the data can be fited to a simple linear function to give s2). In this work sis assumed to be constant at a value $= 1.51, yielding a transmission of 0.921, es shown by 7, in igure 2. For the calculation of m in the region of weak and medium absorption the values of Ty and Tat different i must be ‘obtained. The accuracy to which & can be measured depends on the scale used and for the case of figure 2 the maximum accuracy is about +1nm or about 0.1%. The maximum absolute accuracy of Ty and T is also about 0.001 or 0.1%. These two values set the limits forthe accuracy of the calculated values of m and a. Table I shows the values at the extremes ofthe spectrum of 4, Ty. and T obtained from figure 2. The transmission values should be read om the curves of Ty and Ts at each wavelength and not on the actual spectrum. This procedure partly compensates for the approximations made in deriving equation (8) from equation (AL). If the intermediate values is to be calculated by computer, a parabolic interpolation between three nearest points should be done, since linear interpolation is not accurate enough. Attempts should not be made to fit Ty, oF T= to Some mathematical fUnction since this leads to unacceptable errors and relevant physical information may be lst. ‘The values of refractive index m, as calculated from equation (11) is shown in table 1. There is a fair agreement with the true values my. The accuracy can be improved after calculating d, as willbe described in the next section. A relative error in sof 1% leads to a relative error of about 0.5% in m over the whole spectrum. An absolute error of 1% in a s 3 7 vovelength fon 3 0 Figure 3. Simulated transmission according to the theory for an infinite substrate full eurve) in comparison with that of a finite substrate (broken curve). 1217R Swanepoel Table 1, Values of, Ty and Tq forthe spectrum of figure 2. Calculation of» and d. a Tw Te om mo dsm te 889 0.919 0478 3.018 7 73007 3.006 8I¢ 0919 0470 3.082 78 1000 3.083 3.053, 775 0.916 0.460 3.098 1035 «81002 3.100 3.099, 7400913 O.4d8 3.149 988 «RS 999 BLS 3.1K 710 0.908 0437 3198 979-9 999-3195 Rtgs 683 0.896 0426 1023 9S 1001-3244 3.243 68908820813 1013 10 1000-3398 3.291 636-0847 0.398 102 10.$ 1001-3339 3.341 617 80S 0378-3400 9781 998 3304 3.389 $98 OTIS 0.384 3420 1049 11S 1005 -3.439 3.439 $82 0.616 0319 3.496 1006 12 999 3492 38d S64 0.432 0286 3570 8912S 3582 SSS 0328 0210 3668 $3813 3.573 8, = 1008; 9, = 25:2; = 1000; 0; =2 Ty leads to a relative of about 0.5% in m in the region of weak absorption but increases to about 1% in the region of medium absorption. An absolute error of 19% in Ty yields a relative error of about 1.5% in m in the region of weak absorption and this increases to about 3% in the region of medium absorption. Equation (11) is thus more sensitive to errors in Ty than in Ti. To obtain 1% accuracy in m, Ty and T,, should be measured to about 0.2% absolute accuracy. ‘The values of my in table | are calculated atthe extremes of, the spectrum only for illustrative purposes. It can of course be calculated at any value of Z using the smooth envelopes Ty. and Te 4. Determination of the thickness d Im, and m; are the refractive indices at wo adjacent maxima (or minima) at 1, and Ay, it fellows from equation (3) that the thickness is given by asks i ey 20m = Asm) Equation (23) is very sensitive to errors in m and is not ve accurate. The values of d calculated from equation (23) are shown as dy in table 1. There is some dispersion in the values of 4d; but the last ewo values deviate considerably from the other ‘values. This deviation is an indication that m, as calculated from ‘equation (11) and 4 is not accurate enough and these values rust be rejected. As a general rule the last two extremes of the spectrum should not be used in equation (11). The average value of d:, ignoring the last two values, is dy 1008 = 25 nm. This value of dj can now be used with 1, to determine the order numbers forthe extremes from equation (3). ‘A big increase in accuracy now results in taking the exact integer or halfinteger values of m for each 3 and calculating the thickness d; from equation (3) using the values of m, again. If the values of m cannot be determined with certainty by inspection using equation (3), few sets of values should be tried and d; calculated. The set of values of m that gives the smallest ispersion in d; should be taken. The average value of d; in table | is d= 1000 +2 nm. This is an accuracy better than 19%. Even if the values of & are rounded off to even values and the values of Ty and Ty rounded off to 1% precision the procedure stil yields an answer of d;=1000+3nm, illustrating. the accuracy ofthe procedure to determine d. 28 Using the accurate values of m and dj, can again be 0.25. Equation (19) can however not be used for values of 7'< 0.01 since due to its structure, round-of errors in caleulations lead to errors. The values for a as calculated by equation (A3) again agree excellently with the other values. ‘A relative error of 1% in s leads to an absolute error of Table 3, Valves of in units of 10" formula inthe text. cm ealoulated from the diferent Ty TT. a ay Gy) (a8) 859) 0.013 0.033 0042 0.070 0.047 si4 0.013 0.011 0.010 0.006 0.026 ms 0.036 0.026 0.02 0.008 0.043 740 0.088 0.062 0064 0.070 0.078 710 0.095 0.101 105 OS 0.119 683 0.183 0.177 0.174 G64 0.198 689 0.287 0291 0294 0301 0.310 636 0885 0542 0534 0514 0.61 617 0.890 0.912 0924 0.957 0.912 598 1686 1.652 L634 1.581 1.669 582 2712 2732-2741 2770 2.748. $64 277 $329 S381 3.329 585 7428 7318 7.720 1220Determination of thickness and optical constants of a-SI:H Table 4. Change in values of a (em™') due to changes i the parameters of the formulae. ‘Change in Tu(12) 7,09) Ts) T3(13) parameters z=10 S000 10° $000 10 © $00010—=—«$000 50015 +30 -10 10-20 30 30 80 —70 n—001n 0 =40 +60 +80 +100 +100 +200 ~180 Ty-001 +80 +180 +50 +110 +40 +9 0 0 T~—001 0 0 ~100 180 -140 240 ~280 —s00 Tw-001,Tq +001 ~80 +180 $0 -70 ~100 ~140 ~280 —so0 Table $. Values of 2in units of 10° em™' for the region of strong absorption using the ifferent formulas in the text. Tj tT TT. Te Ta To a2) a0) a) 0223 lost 909-893 6909.00 9.10 0.179 1248121110949 Ltn 038, 14.96 13.95 1389 1269 1392 13.96 0.095 17401737 1652 17.38 17.40 0.060 2193 2192 2139 2192 21.93 0.034 anss 2784 2724-7788 2788 0016 M9S 3502 3489 3802 35.03 ou! 3866 3875 386638753875 0.007 4439 43.28 43:19 4325 43.28 0.008 4846 4659 46.55 0.003, S168 $1.65 0.002 §571 $5.70 about 20cm™' in a and a relative error of 1% in n leads to an absolute error of about 100 em™' in over the whole region of, strong absorption. From equation (22) it follows that an absolute error of AT in Tp leads to an absolute error of, (i/dyln(1 + AT T5") in a. The accuracy of a thus decreases for T<0.01. In conclusion the author recommends that Ty with equation (12) be used to calculate a over the whole range of the spectrum. Ty is also the easiest curve to construct experimentally In cases where Ty, cannot be constructed accurately, eg. for very thick samples where slt errors may be present or very thin samples with a few fringes, 7; can be constructed and 2 be celculated from equation (15). 2 ean be ealeulated with about 1% accuracy in the range of about 100.cm~' to $x 10em~" Dut the accuracy decreases outside this region. Once a{i) is known, A(2) can be calculated from the equation kmai/as which completes the calculation ofthe optial constants an 7. Conclusion Formulae and procedures have been presented to calculate 2), (4) and d for a-Si films to an accuracy of the order of 1% using data from the transmission spectrum alone, ll formulae are in closed form and can easily be used on a programmable pocket calculator. The procedure has been used on a large umber of a-Si:H films using a Varin DMS-90 spectrophotometer and the results showed that the accuracy claims are justified. Appendix ‘The rigorous expression for the transmission T for a system as shown in figure lis ax FOE (an) 6stn? +3) (n= DP + YlOn+ Dore) +2] H)— 24S + D]2c08 0 14K os HR SF (5 I? = KEY Y= 88) +) +K)2sin © Di=ln-1F comand), a=dzk/i. -xp(-ad), In the ease of an infinite substrate the transmission T's given by Ate Foe (a2) where 6 son? +) (n= DP +A)? +] (nF = Xn KH) 4ks]2c080 K2(n 58 8) + 29001 +42 2sin © (n= DF AAI —3)* +) The absorbance x for a system as shown in figure 1 is given in 1221R Swanepoel terms of the interference-free transmission T,, by NeIPs [PA 2QTAU-RRIVQ (AD where = 27a(RyR + RRy~2RiRIRs) Pe(R,~14R; 141) d-ma+me Ryalin—sitn=sP Ry=[s—1Ms+ DF. References Bennet H E and Bennet J M 1967 Physics of Thin Films ed G Hass and R E Thun (New York: Academic) vol. 4 pp 1-96 Clark AH 1980 Polverystalline and Amorphous Thin Film Devices ed. L L Kazmerski (New York: Academic) chap. 4 pp 135-182 Freeman E C and Paul W 1979 Phys, Rev, B20 716-7 Hadley LN 1947 J. Opt. Soe. Am. 37.451 Keradec J 1973 Thesis L’Université Scientifique et Médicale de Grenoble Lyashenko $ P and Miloslavski V K 1964 Opt. Spectrase. 16 80-1 Manifacier 1 C, Gasiot J and Fillard 1 P 1976 A simple method forthe determination ofthe optical constants n.k and the thickness of a weakly absorbing flm J. Phys. E: Sl. Instrum. 9 1002—4 Mini A 1982 Thesis L'Universive Scientifique et Médicale de Grenoble Szezyrbowski J and Czapla A 1977 Thin Soltd Films 46 127-37 Wales J. Lovitt GJ and Hill R A 1967 Thin Solid Films 1137-80 1222