Relationship between the optical gap and the optical-absorption tail breadth in amorphous GaAs J. H. Dias da Silva, R. R. Campomanes, D. M. G. Leite, Farida Orapunt, and Stephen K. O’Leary Citation: Journal of Applied Physics 96, 7052 (2004); doi: 10.1063/1.1797541 View online: http://dx.doi.org/10.1063/1.1797541 View Table of Contents: http://scitation.aip.org/content/aip/journal/jap/96/12?ver=pdfcov Published by the AIP Publishing [This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to ] IP: 186.217.234.225 On: Tue, 14 Jan 2014 13:05:34 http://scitation.aip.org/content/aip/journal/jap?ver=pdfcov http://oasc12039.247realmedia.com/RealMedia/ads/click_lx.ads/www.aip.org/pt/adcenter/pdfcover_test/L-37/1744363738/x01/AIP-PT/JAP_CoverPg_101613/aipToCAlerts_Large.png/5532386d4f314a53757a6b4144615953?x http://scitation.aip.org/search?value1=J.+H.+Dias+da+Silva&option1=author http://scitation.aip.org/search?value1=R.+R.+Campomanes&option1=author http://scitation.aip.org/search?value1=D.+M.+G.+Leite&option1=author http://scitation.aip.org/search?value1=Farida+Orapunt&option1=author http://scitation.aip.org/search?value1=Stephen+K.+O�Leary&option1=author http://scitation.aip.org/content/aip/journal/jap?ver=pdfcov http://dx.doi.org/10.1063/1.1797541 http://scitation.aip.org/content/aip/journal/jap/96/12?ver=pdfcov http://scitation.aip.org/content/aip?ver=pdfcov Relationship between the optical gap and the optical-absorption tail breadth in amorphous GaAs J. H. Dias da Silva, R. R. Campomanes, and D. M. G. Leite Departamento de Física, Faculdade de Ciências, Universidade Estadual Paulista, CEP 17033-360 Bauru, São Paulo, Brazil Farida Orapunt and Stephen K. O’Learya) Faculty of Engineering, University of Regina, Regina, Saskatchewan S4S 0A2, Canada (Received 7 July 2004; accepted 3 August 2004) We study the relationship between the optical gap and the optical-absorption tail breadth for the case of amorphous gallium arsenide(a-GaAs). In particular, we analyze the optical-absorption spectra corresponding to some recently prepared a-GaAs samples. The optical gap and the optical-absorption tail breadth corresponding to each sample is determined. Plotting the optical gap as a function of the corresponding optical-absorption tail breadth, we note that a trend, similar to that found for the cases of the hydrogenated amorphous silicon and hydrogenated amorphous germanium, is also found for the case of a-GaAs. The impact of alloying on the optical-absorption spectrum associated with a-GaAs is also briefly examined. ©2004 American Institute of Physics. [DOI: 10.1063/1.1797541] I. INTRODUCTION Amorphous gallium arsenide(a-GaAs) is a promising electronic material that is used in a plethora of device appli- cations. Thus far, this material has been used as an antiguide channel in vertical cavity surface-emitting lasers,1 as a buffer layer between silicon and gallium arsenide epitaxial layers,2 and as a host material for erbium doping.3 New applications for a-GaAs are being devised with each passing year. As with many other amorphous semiconductors, the optical proper- ties of this material have been the focus of considerable investigation.4–8 From this research, key insights into the electronic character of a-GaAs have been gleaned. The determination of the spectral dependence of the optical-absorption coefficient,as"vd, has been a principal focus of studies into the optical response of a-GaAs for many years. In a crystalline semiconductor, the disorder present leads to a narrow tail inas"vd, which encroaches into the otherwise empty gap region. In an amorphous semiconduc- tor, however, there is a considerably greater level of disorder present, and the tail exhibited byas"vd is much broader.9 It is now widely held that the tail in the optical-absorption spectrum of an amorphous semiconductor is attributable to optical transitions involving tail states, these states encroach- ing into the otherwise empty gap region of an amorphous semiconductor.10–13As the tail states arise as a result of the disorder inherent to the amorphous state, the breadth of the optical-absorption tail is often taken as a direct measure of the amount of disorder present in the material.14 In an amorphous semiconductor, such as a-GaAs, the disorder present has thermal and structural components. The thermal component reflects the thermal occupancy of the phonon states. The structural component, by a way of con- trast, owes its existence to the structural disorder inherent to the amorphous semiconductor under consideration. In an ef- fort to understand how these components individually con- tribute to the breadth of the optical-absorption tail associated with hydrogenated amorphous silicon(a-Si:H), Cody et al.14 performed an experimental investigation of their roles. In particular, structural disorder was deliberately intro- duced into a-Si:H films through the thermal evolution of hy- drogen. A variety of measurement temperatures were em- ployed in order to assess the role of thermal disorder. The data obtained from this study was consistent with the notion that the effect of the thermal and structural components of disorder are additive. The key result of this analysis, that the optical gap associated with a-Si:H decreases in response to the increases in the optical-absorption tail breadth, regardless of the exact nature of the disorder, suggests that disorder, rather than hydrogen content, determines the optical gap of a-Si:H. Persanset al.15 performed a similar study for the case of the hydrogenated amorphous germanium(a-Ge:H). Rather than considering a-Ge:H as a homogeneous material, they instead envisioned it as a binary alloy, Ge1−xsGeHdx, and provided experimental evidence to suggest that disorder re- lated to alloying makes a significant contribution to the over- all structural disorder present within this material; Cody et al.14 did not find compelling evidence suggesting disorder related to alloying present within a-Si:H, probably as a result of the fact that the hydrogen contents considered in the study of Cody et al.14 were too small. In a compound amorphous semiconductor, such as a-GaAs, which is fundamentally an alloy, i.e., a-Ga1−xAsx, wherex denotes the alloy content, the disorder related to alloying is expected to make a potentially even more significant contribution to the overall structural disorder present. In this paper, we study the relationship between the op- tical gap and the optical-absorption tail breadth for the case of a-GaAs. In particular, we analyze the optical-absorptiona)Electronic mail: stephen.oleary@uregina.ca JOURNAL OF APPLIED PHYSICS VOLUME 96, NUMBER 12 15 DECEMBER 2004 0021-8979/2004/96(12)/7052/8/$22.00 © 2004 American Institute of Physics7052 [This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to ] IP: 186.217.234.225 On: Tue, 14 Jan 2014 13:05:34 http://dx.doi.org/10.1063/1.1797541 spectra corresponding to some recently prepared a-GaAs samples. The optical gap and the optical-absorption tail breadth corresponding to each sample is determined. Then, plotting the optical gap as a function of the corresponding optical-absorption tail breadth, we will determine whether or not a trend, similar to that observed by Codyet al.14 for the case of a-Si:H and Persanset al.15 for the case of a-Ge:H, is also found for the case of a-GaAs. In an effort to assess the generality of these results, a comparison with other a-GaAs data from the literature is performed. We complete this analysis by considering the optical-absorption spectrum cor- responding to a number of a-Ga1−xAsx samples of varying alloy contents, with a view to developing an appreciation as to how the alloy content influences the optical absorption in this material. In order to provide a proper theoretical basis for the interpretation of these experimental results, the semi- classical optical-absorption analysis of O’Learyet al.11 is employed. From this analysis, insights into the optical re- sponse of a-GaAs are obtained. This paper is organized in the following manner. In Sec. II, the experimental means, whereby the a-GaAs samples were prepared and their optical-absorption spectra were de- termined, is presented. Then, in Sec. III, we analyze the re- sultant optical-absorption spectra. A comparative analysis of a-GaAs and a-Si:H is also presented. The optical-absorption spectra corresponding to a number of a-Ga1−xAsx samples, with varying alloy contents,x, are then considered in Sec. IV, with a view to determining how the alloy content influences the optical absorption in this material. Finally, conclusions are drawn in Sec. V. II. FILM PREPARATION AND EXPERIMENT The a-GaAs films used to perform this study were de- posited using the flash-evaporation technique on Corning 7059 glass substrates. The powder used for the depositions was obtained from high-purity small crystalline gallium ars- enide(c-GaAs) pieces. The crucible temperature,Tcruc, was varied from sample to sample over the range from 1240 to 1900 °C. The substrate temperature was kept con- stant at 25 °C for all of the film depositions.16 The thickness of the resultant films, measured with a profilometer, was found to range from 400 to 1000 nm. The composition of the films, determined using energy-dispersive x-ray spectros- copy, suggested an excess of arsenic for all the samples pre- pared; it was found thatx=0.55±0.01 for all of the a -Ga1−xAsx samples considered.Ex situ x-ray photoelectron measurements detected surface oxygen on the samples at the same concentration as on a reference c-GaAs wafer, i.e., 5 at. %. We thus attribute the presence of this impurity to the surface oxidation of the air-exposed films. No other impuri- ties were found within the detectability limit of the charac- terization techniques used.17 Further details on the prepara- tion of these films, and on their characterization, are provided in the literature.7,8,18 Measurements of the transmittance and reflectance spec- tra of these a-GaAs films were obtained using a Lambda-9 Perkin-Elmer spectrophotometer in the visible and near- infrared ranges. The corresponding optical-absorption spec- tra were determined using standard thin-film optics techniques.19 For regions of the optical-absorption spectrum requiring a greater sensitivity, we employed photothermal deflection spectroscopy(PDS). The PDS results, which do not provide for an absolute determination of the optical- absorption coefficient, were then calibrated with the results obtained from the transmittance and reflectance spectra. From this analysis, the optical-absorption spectrum corre- sponding to each a-GaAs sample was determined. Further details of our analysis are presented in the literature.7,8 III. RESULTS AND ANALYSIS We analyzed a total of ten flash-evaporated a-GaAs samples for this analysis. In Fig. 1(a), we plot the optical- absorption spectra corresponding to four of these samples. We note that there are significant variations in the optical- absorption spectra. In particular, it is seen that the optical- absorption spectrum corresponding to the sample prepared with Tcruc=1240 °C is much broader, and encroaches into the gap region much deeper, than that prepared withTcruc =1900 °C. It is clear that the crucible temperature is playing FIG. 1. (a) Optical-absorption spectra corresponding to four representative a-GaAs samples. The experimental data is depicted with the symbols. The fits of Eq. (1) are shown with the light solid lines.(b) The Tauc plots corresponding to the experimental data depicted in Fig. 1(a). The experi- mental data is depicted with the symbols. The fits of Eq.(2), and their extrapolations, are depicted with the solid and dotted lines, respectively. The Tauc optical gaps, determined from these extrapolations, are indicated with the small arrows. J. Appl. Phys., Vol. 96, No. 12, 15 December 2004 Dias da Silva et al. 7053 [This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to ] IP: 186.217.234.225 On: Tue, 14 Jan 2014 13:05:34 an important role in determining the electronic character of the resultant a-GaAs. Rather than determining the exact na- ture of this role, for the purposes of this analysis, we instead use the crucible temperature as a means of modulating the amount of disorder present in the samples. Through a direct comparison with the other a-GaAs results, the generality of our results may be confirmed. We now quantify the relationship between the optical gap and the optical-absorption tail breadth. In order to deter- mine the breadth of the optical-absorption tail, we follow the usual procedure in this analysis and fit an exponential func- tional dependence, as"vd = ao expS"v Eo D , s1d to our optical-absorption experimental data, whereao is a pre-exponential constant andEo denotes the breadth of the optical-absorption tail. Contrary to the approach adopted by Cody et al.14 in their analysis of a-Si:H, we do not,a priori, seek an Urbach “focus”. Rather, we fit, in a least-squares sense, Eq.(1) to our optical-absorption experimental data. We find that Eq.(1) may be fitted to our a-GaAs experimen- tal results for 103 cm−1øas"vdø104 cm−1.20 The fits corre- sponding to the four representative a-GaAs optical- absorption spectra depicted in Fig. 1(a) are also shown in Fig. 1(a). In Table I, the fits of Eq.(1), corresponding to the ten a-GaAs samples considered in this analysis, are tabu- lated. For the determination of the optical gap, we employ the “Tauc” approach.21 That is, we fit, in a least-squares sense, Îas"vd"v = CTaucs"v − EgTauc d, s2d to the relevant experimental data, whereCTauc denotes the slope of the Tauc extrapolation andEgTauc represents the cor- responding Tauc optical gap. The quantitative evaluation of CTaucandEgTauc depend critically upon the range of data over which the fit of Eq.(2) is taken. For the purposes of this analysis, we perform our fit of Eq.(2) to the experimental data for which 23104 cm−1øas"vdø63104 cm−1. The Tauc fits, corresponding to the four representative a-GaAs optical-absorption spectra depicted in Fig. 1(a), are shown in Fig. 1(b). In Table I, the fits of Eq.(2), corresponding to the ten a-GaAs samples considered in this analysis, are tabu- lated. In order to demonstrate that our experimental results are consistent with the other a-GaAs results presented in the lit- erature, in Fig. 2(a), we contrast the optical-absorption spec- tra of the four a-GaAs samples considered in Figs. 1(a) and 1(b) with representative a-GaAs optical-absorption experi- mental data from Theyeet al.4 and Manssor and Davis;22 Theyeet al.4 prepared their samples through a flash evapo- ration, whereas Manssor and Davis22 prepared their samples through sputtering. Differences in the optical-absorption spectra are evident. In particular, we find that the optical- absorption spectra corresponding to two of our a-GaAs samples, that prepared withTcruc=1240 °C and that prepared with Tcruc=1270 °C, exhibit much broader optical- absorption tails than the samples of Theyeet al.4 and Mans- sor and Davis.22 This suggests that theTcruc=1240 °C and Tcruc=1270 °C a-GaAs samples have substantially greater amounts of disorder than the other a-GaAs samples. Never- theless, the basic trend for all of the a-GaAs optical- absorption spectra depicted in Fig. 2(a) appear to be similar. In Fig. 2(a), we also present the optical-absorption spec- trum corresponding to crystalline gallium arsenide (c-GaAs).23 It is interesting to note that this spectrum forms a loose lower bound to all of the presented a-GaAs optical- absorption experimental data. That is, for any given value of "v, the optical-absorption spectrum corresponding to a-GaAs exceeds that associated with c-GaAs. Moreover, it is seen that as the disorder is decreased, i.e., asEo decreases, that the optical-absorption spectrum corresponding to a-GaAs appears to approach that corresponding to c-GaAs. One could speculate that as the disorderless limit is ap- proached, i.e., asEo→0, that the optical-absorption spectrum associated with a-GaAs approaches that associated with c-GaAs. This would suggest a smooth and a continuous tran- sition in the optical-absorption spectrum of gallium arsenide, from disordered a-GaAs to ordered c-GaAs. The situation for the case of a-Si:H, however, is quite different. In Fig. 2(b), we plot the optical-absorption spectra corresponding to four representative a-Si:H data sets from Cody et al.14 The optical-absorption spectra, corresponding to the four representative a-GaAs samples depicted in Figs. TABLE I. The fitting parameters that arise from fitting Eqs.(1) and(2) to the optical-absorption spectra for the ten a-GaAs samples considered in this analysis. While we strove to fit Eq.(1) to the experimental data for 103 cm−1øas"vdø104 cm−1 for all cases, for some of the data sets, we were forced to limit the range over which these fits were taken. The lower and the upper fit limits are specified in the table. Tcruc s°Cd ao scm−1d Eo (meV) Lower limit scm−1d Upper limit scm−1d EgTauc (eV) CTauc scm−1/2 eV−1/2d 1240 1.543103 366 53103 104 0.630 405 1270 1.483102 220 33103 104 0.765 479 1300 2.503101 178 23103 104 0.809 411 1350 9.93310−1 119 103 104 0.923 522 1375 5.553100 144 103 104 0.866 472 1440 1.183101 162 23103 104 0.827 411 1500 9.143100 157 103 104 0.815 396 1600 9.123100 151 23103 104 0.850 465 1770 9.603100 154 103 104 0.859 461 1900 3.273100 144 103 104 1.042 566 7054 J. Appl. Phys., Vol. 96, No. 12, 15 December 2004 Dias da Silva et al. [This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to ] IP: 186.217.234.225 On: Tue, 14 Jan 2014 13:05:34 1(a), 1(b), and 2(a), are also shown, as are the optical- absorption spectra corresponding to c-GaAs and crystalline silicon (c-Si).23 We note that as the disorder is decreased, i.e., asEo is decreased, that the optical-absorption spectrum cor- responding to a-Si:H does not approach that corresponding to c-Si. This suggests that there is a fundamental discontinu- ity in the optical-absorption spectrum of silicon, from disor- dered a-Si:H to ordered c-Si. We opine that the discontinuity observed for the case of silicon in the disorderless limit, i.e., asEo→0, fundamentally arises as a consequence of the in- direct nature of the band structure associated with c-Si, the band structure associated with c-GaAs being instead of a direct nature. Arguments in support of this point of view are presented by O’Leary.24 In order to develop a quantitative understanding as to how the disorder influences the optical-absorption spectrum associated with a-GaAs, we employ the semiclassical optical-absorption analysis of O’Learyet al.11 to interpret our optical-absorption experimental results. Noting that the spectral dependence of the optical-absorption coefficient, as"vd, is, in large measure, determined by the spectral de- pendence of the joint density of states(JDOS) function, Js"vd, the semiclassical analysis of O’Learyet al.11 focuses upon an evaluation of the JDOS function,Js"vd. In this model, the overall form of the JDOS function is determined by averaging a local JDOS over a Gaussian distribution of local energy gaps. Two principal parameters characterize this model:(1) the mean energy gap,Ego , and(2) the energy-gap variance,s2, wheres provides a measure of the amount of FIG. 2. (a) Optical-absorption spectrum of a-GaAs. The experimental data corresponding to the four representative a-GaAs samples depicted in Figs. 1(a) and 1(b) are shown with the light solid lines. The experimental data of Theyeet al. (Ref. 4) and Manssor and Davis(Ref. 22) are depicted with the symbols. The optical-absorption spectrum corresponding to c-GaAs is depicted with the light dotted line(Ref. 23). (b) The optical-absorption spectra corresponding to a-GaAs and a-Si:H. The experimental data is depicted with the symbols: diamonds for the case of our a-GaAs experimental data and open circles for the a-Si:H experimental data of Codyet al. (Ref. 14). The fits of Eq.(1) are shown with the light solid lines. Extrapolations of these fits are shown with the light dotted lines. The optical-absorption spectra corresponding to c-GaAs and c-Si are clearly marked and indicated with the light dotted lines(Ref. 23). (c) The optical-absorption spectra corresponding to the semiclassical analysis of O’Learyet al. (Ref. 11) fit to the four representative a-GaAs samples depicted in Figs. 1(a) and 1(b). The experimental data is depicted with the symbols. The fits of the semiclassical analysis of O’Learyet al. (Ref. 11) to this experimental data are represented with the heavy solid lines. We find that we are able to fit the semiclassical analysis of O’Learyet al. (Ref. 11) to our a-GaAs optical-absorption data withs set to 310, 350, 420, and 480 meV, andEgo being set to 1.4 eV for all cases, for the cases ofTcruc set to 1900, 1770, 1270, and 1240 °C, respectively. The disorderless limit result, i.e.,s=0 meV andEgo =1.4 eV, is also shown. The experimental c-GaAs results are shown with the open diamonds (Ref. 23). (d) Theoretical a-GaAs and a-Si:H optical-absorption results obtained using the semiclassical optical-absorption analysis of O’Learyet al. (Ref. 11). The results of the semiclassical optical-absorption analysis of O’Learyet al. (Ref. 11), corresponding tos set to 150, 270, 310, and 480 meV,Ego being set to 1.4 eV for all cases, are shown with the light solid lines. The fits of Eq.(1), for 103 cm−1øas"vdø104 cm−1 for the cases ofs set to 310 and 480 meV, are shown with the heavy solid lines. The fits of Eq.(1), for 53102 cm−1øas"vdø53103 cm−1 for the cases ofs set to 150 and 270 meV, are shown with the heavy solid lines. Extrapolations of these fits, beyond the fitting ranges used, are depicted with the light dotted lines. J. Appl. Phys., Vol. 96, No. 12, 15 December 2004 Dias da Silva et al. 7055 [This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to ] IP: 186.217.234.225 On: Tue, 14 Jan 2014 13:05:34 disorder. There is also a prefactor in the analysis, which is assumed to be a material constant for the case of a-GaAs. In Fig. 2(c), we find that with selections ofs between 310 and 480 meV,Ego being set to 1.4 eV, and the prefactor being set to a constant for all cases, the semiclassical optical- absorption analysis of O’Learyet al.11 is able to satisfacto- rily capture our a-GaAs experimental results over a broad range of photon energies; deviations from this semiclassical optical-absorption analysis arise as a result of defect absorp- tion, the spectral dependence of the optical transition matrix element,25 and the deviations from square-root distributions of electronic states that occur in the band regions.26–28 The semiclassical optical-absorption analysis of O’Leary et al.11 provides us with an analytical tool, with which we can now quantitatively probe the disorderless limit. In par- ticular, with all the semiclassical optical-absorption analysis parameters set as before, we study what happens ass is diminished to zero, this corresponding to the disorderless limit, i.e., s=0 meV, Ego =1.4 eV, and the same a-GaAs prefactor. It is interesting to note that in this disorderless limit, the low optical-absorption semiclassical result essen- tially coincides with that of c-GaAs; deviations between the disorderless limit semiclassical result, i.e.,s=0 meV, Ego =1.4 eV, and the same a-GaAs prefactor, and the optical- absorption experimental c-GaAs result, which occur for val- ues of optical absorption beyond 104 cm−1, arise as a conse- quence of the stringent momentum conservation rules that shape the spectral dependence ofas"vd for the case of c-GaAs, these conservation rules being relaxed for the case of a-GaAs. A similar result, corresponding to a different set of a-GaAs optical-absorption experimental data, was found previously by O’Leary.24 Recalling Fig. 2(b), it is interesting to note that the ex- ponential region corresponding to the a-GaAs samples ex- tends to higher optical-absorption values than that corre- sponding to a-Si:H. In particular, we find that the fit of Eq. (1) to the a-Si:H experimental data of Codyet al.14 is appli- cable for 53102 cm−1øas"vdø53103 cm−1, i.e., this is a factor of 2 lower than that found for the case of a-GaAs. In order to understand this observation, we once again resort to the semiclassical optical-absorption analysis of O’Learyet al.11 In particular, in Fig. 2(d), we plot the optical-absorption results corresponding to a-GaAs, determined from the semi- classical optical-absorption analysis of O’Learyet al.,11 s being set to 310 and 480 meV,Ego being set to 1.4 eV, and the prefactor being set as before for all cases, this corre- sponding to the range ofs found in our a-GaAs analysis. We also plot the optical-absorption spectra corresponding to the semiclassical optical-absorption analysis of O’Learyet al.,11 s being set to 150 and 270 meV,Ego being set to 1.4 eV, and the prefactor being set as before for all cases, this corre- sponding to the range ofs found in a previous a-Si:H analysis;24 we have selected a commonEgo , 1.4 eV, and a common prefactor value in order to facilitate a more direct comparison. It is noted that there is a distinctive curvature in these semiclassical optical-absorption spectra, which arises as a consequence of the Gaussian distribution of the local energy gaps.11 The presence of a curvature in the optical- absorption spectrum makes the determination of the optical- absorption tail breadth,Eo, ambiguous, as the exact range of as"vd, over which the fit of Eq.(1) to the experimental data should be taken, remains debatable.29,30 We note that with a broader optical-absorption tail, i.e., a largers, a region of apparently linear-exponential dependence is observed at larger values ofas"vd. Specifically, while the least-squares fits of Eq. (1) to the a-GaAs theoretical results, i.e.,s set to 310 and 480 meV for 103 cm−1øas"vdø104 cm−1, lead to exponential fits that are essentially indistinguishable from the exact theoretical result over a wide range of"v, if the same kind of fit is performed to the a-Si:H semiclassical results, i.e., s set to 150 and 270 meV, the resultant fits depart quite significantly from the theoretical results. On the other hand, the least-squares fits of Eq.(1) to the a-Si:H theoretical re- sults, i.e., s set to 150 and 270 meV for 53102 cm−1 øas"vdø53103 cm−1, yield exponential fits that are es- sentially indistinguishable from the exact semiclassical result over a wide range of"v. Thus, the fact that the fit of Eq.(1) occurs over a higher range ofas"vd for the case of a-GaAs when contrasted with the case of a-Si:H simply arises as a consequence of the fact that the disorder in our a-GaAs samples is greater than that associated with the a-Si:H samples considered by Codyet al.14 The fits of Eq. (1), presented in Fig. 2(b), were per- formed without any assumption regarding the presence of an Urbach focus. In order to determine the applicability of the Urbach focus concept for the case of a-GaAs, in Fig. 2(b), we plot the extrapolations of the fits of Eq.(1) that were found. We note that these extrapolations do not all converge at a common point. In particular, it is seen that these fits intercept over a broad range of photon energies, between 1.2 and 1.6 eV, the corresponding optical-absorption values be- ing between 104 and 106 cm−1. We thus conclude that the Urbach focus concept does not, strictly speaking, apply for the case of a-GaAs.31,32 In Fig. 3(a), we plot the optical gap as a function of the corresponding optical-absorption tail breadth for the a-GaAs samples we have considered in this analysis. While there are anomalies, in particular, the data point corresponding to the sample prepared withTcruc=1900 °C, the overall trend is clear, i.e., the optical gap diminishes in response to the in- creases in the optical-absorption tail breadth. This result demonstrates that the trend observed by Codyet al.14 for the case of a-Si:H and Persanset al.15 for the case of a-Ge:H also occurs for the case of a-GaAs. A linear least-squares fit of this experimental data, excluding theTcruc=1900 °C anomalous point, yields EgTauc = 1.02 − 1.09Eo, s3d where EgTauc and Eo are in units of eV, the error inEgTauc being of the order of ±20 meV in all cases. We note that, with the exception of the anomalous point, our experimental data lies very close to Eq.(3). In order to demonstrate that the observed trend is not simply unique to our particular experimental data, in Fig. 3(a), we also include a-GaAs data obtained from other sources. In particular, results obtained from the experimental data of Theyeet al.4 and Manssor and Davis22 are included in Fig. 3(a). It is worth noting that while the data of Theyeet 7056 J. Appl. Phys., Vol. 96, No. 12, 15 December 2004 Dias da Silva et al. [This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to ] IP: 186.217.234.225 On: Tue, 14 Jan 2014 13:05:34 al.4 and our data correspond to unhydrogenated a-GaAs samples prepared through flash evaporation, the samples of Manssor and Davis22 correspond to intentionally hydrogen- ated a-GaAs samples prepared through radio-frequency sput- tering. The data corresponding to two recently prepared flash-evaporated a-GaAs samples, with the substrate tem- perature held at 100 °C, are also included; our other samples were prepared with the substrate temperature held at 25 °C. In order to allow for a fair comparison, we analyzed the other experimental data in exactly the same manner as was em- ployed for the analysis of our own experimental data. We see that Eq.(3) forms a lower bound to this experimental data. We opine that the variations in the sample-preparation method are responsible for the observed deviations from Eq. (3), as suggested by the deviations from Eq.(3) in evidence for our samples prepared using the higher substrate tempera- ture [see the open-triangle data points depicted in Fig. 3(a)]. We now perform a comparative analysis of a-GaAs and a-Si:H. For the case of a-Si:H, Fig. 3 of Codyet al.14 sug- gests that EgTauc = 2.10 − 6.07Eo, s4d whereEgTauc and Eo are in units of eV. Eq.(4) has a form similar to that of Eq.(3). We do note, however, that the a-Si:H experimental data of Codyet al.14 exhibits consider- ably smaller values ofEo than our a-GaAs experimental data. In order to allow for a fair comparison, we contrast all of the a-GaAs experimental data depicted in Fig. 3(a) with the a-Si:H experimental data of Codyet al.14 and Viturro and Weiser33 in Fig. 3(b), the optical-absorption tails exhibited by the a-Si:H experimental data of Viturro and Weiser33 be- ing considerably broader than those of Codyet al.14 It is seen that the rate of decrease in the optical gap with the increased Eo starts to decline for the higher values ofEo. This effect may well account for the differences in the slopes of Eqs.(3) and(4), i.e., the relationship betweenEgTauc andEo is funda- mentally nonlinear, Eqs.(3) and (4) being applicable over different ranges ofEo. Focusing solely on the a-Si:H experi- mental data presented in Fig. 3(b) with Eoù100 meV, a lin- ear least-squares fit analysis demonstrates that EgTauc = 1.60 − 1.02Eo, s5d whereEgTauc andEo are in units of eV, the error being similar to that found for the case of Eq.(3). It is noted that the slopes of Eqs. (3) and (5) are, within the range of experimental error, essentially identical. We thus assert that the relation- ship between the optical gap and the optical-absorption tail breadth is similar in a-GaAs and a-Si:H. Further insight into these results may be obtained through the use of the semiclassical optical-absorption analy- sis of O’Learyet al.11 In particular, using an approach simi- lar to that outlined in Ref. 34 to determine the Tauc optical gap and the breadth of the optical-absorption tail, we plot the functional dependence of the Tauc optical gap on the optical- absorption tail breadth. In order to allow for a fair compari- son with the experimental results, we performed the determi- nation of the Tauc optical gaps and the optical-absorption tail breadths in exactly the same manner as that employed for our experimental analysis. It is seen that the resultant depen- dence, depicted in Figure 3(b) with the heavy solid line, exhibits a functional form that captures the essence of the a-GaAs experimental results. We also note that as the disor- derless limit is approached, i.e., asEo→0, that the Tauc op- tical gap approaches the disorderless energy gap,Ego . This further bolsters our previously mentioned disorderless limit assertion. IV. THE ROLE OF ALLOYING We now study how the alloy content,x, plays a role in influencing the optical response of a-Ga1−xAsx. Five a-Ga1−xAsx samples were prepared for the purposes of this investigation, the alloy content,x, being varied from sample to sample. These samples were deposited using the same flash-evaporation technique described previously. To produce variations in the alloy content, two different types of input FIG. 3. (a) Optical gap of a-GaAs as a function of the optical-absorption tail breadth. All the a-GaAs experimental data is depicted with the symbols. The linear least-squares fit of our experimental data, i.e., Eq.(3), is depicted with the light dotted line. The additional data from our laboratory corresponds to flash-evaporated a-GaAs samples prepared with the substrate temperature held at 100 °C.(b) The optical gap of a-GaAs and a-Si:H as a function of the optical-absorption tail breadth. The experimental a-GaAs data from us, Theyeet al. (Ref. 4), and Manssor and Davis(Ref. 22), are included. The experimental a-Si:H data from Codyet al. (Ref. 14) and Viturro and Weiser (Ref. 33) are included. The a-GaAs experimental data is depicted with the solid points. The a-Si:H experimental data is depicted with the open points. The linear least-squares fits, i.e., Eqs.(3)–(5), are shown with the light dotted lines. The theoretical result, obtained using the semiclassical optical- absorption analysis of O’Learyet al. (Ref. 11), is shown with the heavy solid line. J. Appl. Phys., Vol. 96, No. 12, 15 December 2004 Dias da Silva et al. 7057 [This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to ] IP: 186.217.234.225 On: Tue, 14 Jan 2014 13:05:34 material powders, namely, 6N purity As powder and pc- GaAs powder, were fed into the resistively heated tungsten crucible. For the flash-evaporation parameters used for these depositions, i.e., a crucible temperature of 1380 °C and a substrate temperature of 20 °C, the alloy content of the re- sultant films were found to be primarily determined by the input powder proportions used. The alloy contents found for the resultant a-Ga1−xAsx films were 0.47, 0.55, 0.67, 0.84, and 1.00. In Fig. 4(a), we plot the spectral dependence of the optical-absorption spectrum corresponding to the five a -Ga1−xAsx samples considered in this analysis, these spectra being determined in exactly the same manner as specified previously, i.e., through measurements of the transmittance and the reflectance spectra and through PDS measurements. In order to provide for a basis of comparison, we also plot the spectral dependence of the optical-absorption coefficient corresponding to ourTcruc=1240 °C andTcruc=1900 °C a-GaAs results, these results already being depicted in Figs. 1(a), 1(b), and 2(a)–2(c). We note that the spectral depen- dence of the optical-absorption coefficient associated with a -Ga1−xAsx is strongly influenced by the alloy content,x. In particular, it is seen that the optical gap and the breadth of the optical-absorption tail vary greatly with the alloy content, x. Plotting Eo as a function of the alloy content,x, in Fig. 4(b), whereEo is as determined previously, we find thatEo varies considerably withx, exhibiting a peak in the neighbor- hood of x.0.5. Unfortunately, there are not enough data points, and the error associated with each such data point is too great, to draw definitive conclusions. Clearly, further re- search will be required in order to understand the exact role that the disorder related to alloying plays in contributing to the overall structural disorder present in a-GaAs. V. CONCLUSIONS In conclusion, we have studied the relationship between the optical gap and the optical-absorption tail breadth for the case of a-GaAs. By plotting the optical gap as a function of the corresponding optical-absorption tail breadth, we found that a trend, similar to that found for the cases of a-Si:H and a-Ge:H, is also found for the case of a-GaAs. We conclude that the critical role played by the disorder in linking the optical gap with the breadth of the optical-absorption tail, observed in a-Si:H and a-Ge:H, also occurs for the case of a-GaAs, despite the disparate nature of these materials. Fi- nally, the impact of alloying on the optical-absorption spec- trum of a-Ga1−xAsx was briefly examined. ACKNOWLEDGMENTS The authors gratefully acknowledge financial support from FAPESP(Grant No. 97/06278-6). Two of the authors (F.O. and S.K.O.) acknowledge financial support from the Natural Sciences and Engineering Research Council of Canada. The use of equipment granted from the Canadian Foundation for Innovation, and equipment loaned from the Canadian Microelectronics Corporation, is acknowledged. Finally, the authors would like to thank the Photovoltaic Conversion Group at the University of Campinas for the use of their PDS facility. 1B.-S. Yoo, H. Y. Chu, H.-H. Park, H. G. Lee, and J. Lee, IEEE J. Quantum Electron. 33, 1794(1997). 2K. I. Cho, W. K. Choo, J. Y. Lee, S. C. Park, and T. Nishinaga, J. Appl. Phys. 69, 237 (1991). 3A. R. Zanatta, Appl. Phys. Lett.75, 3279(1999). 4M. L. Theye, A. Gheorghiu, K. Driss-Khodja, and C. Boccara, J. Non- Cryst. Solids77 & 78, 1293(1985). 5R. Murri, L. Schiavulli, N. Pinto, and T. Ligonzo, J. Non-Cryst. Solids 139, 60 (1992). 6S. H. Baker, S. C. Bayliss, S. J. Gurman, N. Elgun, J. S. Bates, and E. A. Davis, J. Phys.: Condens. Matter5, 519 (1993). 7J. H. Dias da Silva and R. R. Campomanes, J. Non-Cryst. Solids299–302, 328 (2002). 8R. R. Campomanes, J. Ugucione, and J. H. Dias da Silva, J. Non-Cryst. Solids 304, 259 (2002). 9G. D. Cody, in Hydrogenated Amorphous Silicon, Semiconductors and Semimetals Vol. 21B, edited by J. I. Pankove(Academic, New York, 1984), p. 11. 10S. K. O’Leary, S. Zukotynski, and J. M. Perz, Phys. Rev. B51, 4143 (1995). 11S. K. O’Leary, S. Zukotynski, and J. M. Perz, Phys. Rev. B52, 7795 (1995). 12S. K. O’Leary, S. R. Johnson, and P. K. Lim, J. Appl. Phys.82, 3334 (1997). 13S. M. Malik and S. K. O’Leary, Appl. Phys. Lett.80, 790 (2002). 14G. D. Cody, T. Tiedje, B. Abeles, B. Brooks, and Y. Goldstein, Phys. Rev. Lett. 47, 1480(1981). FIG. 4. (a) Optical-absorption spectra corresponding to five a-Ga1−xAsx samples of varying alloy contents. This experimental data is depicted with the symbols. The optical-absorption spectra corresponding to the Tcruc=1240 °C and theTcruc=1900 °C a-GaAs results previously de- picted are plotted with the light solid lines.(b) The dependence of the breadth of the optical-absorption tail,Eo, on the alloy content,x, for four of the a-Ga1−xAsx samples considered in Fig. 4(a). Error bars are shown. 7058 J. Appl. Phys., Vol. 96, No. 12, 15 December 2004 Dias da Silva et al. [This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to ] IP: 186.217.234.225 On: Tue, 14 Jan 2014 13:05:34 15P. D. Persans, A. F. Ruppert, G. D. Cody, B. G. Brooks, and W. Lanford, AIP Conf. Proc. 120, 349 (1984). 16The substrate holder temperature was monitored during all the depositions using an externally fitted thermocouple. The thermocouple, situated at a distance of 3 mm from the substrate surface, was placed in a hole in the substrate holder from the external part of the chamber. Thermal contact of the thermocouple with the hole end was facilitated through the application of thermal grease. The thermocouple was encapsulated by a ceramic insu- lating tube so that only its tip reached the measurement point. We noted that the temperature of this point only changed by a few degrees during the depositionss3 °C–8 °Cd. The fact that the depositions only lasted for 5 to 6 minutes coupled with the fact that the stainless-steel substrate holder is massive(approximately 2 kg), and was connected to the chamber top flange (stainless steel) through a thermally conducting short neck, ac- counts for this observation. 17We can set an upper limit for the presence of hydrogen bonded to gallium or arsenic in the samples considered here as 0.1 at. %. This limit was established by comparing the spectra of our samples with that of samples of similar thicknesses prepared through rf sputtering in the presence of hydrogen. In the hydrogenated samples, concentrations of less than 1 at. % hydrogen correspond to the easily detectable spectral features, while in our samples the spectra were found to be completely featureless. 18R. R. Campomanes, J. H. Dias da Silva, J. Vilcarromero, and L. P. Car- doso, J. Non-Cryst. Solids299–302, 788 (2002). 19R. Swanepoel, J. Phys. E16, 1214(1983). 20The a-GaAs experimental data does not span across this fitting range for all of the samples considered in this analysis. For these samples, the fits of Eq. (1) are performed with the available experimental data. 21J. Tauc, R. Grigorovici, and A. Vancu, Phys. Status Solidi15, 627(1966). 22M. I. Manssor and E. A. Davis, J. Phys.: Condens. Matter2, 8063(1990). 23S. M. Sze,Physics of Semiconductor Devices, 2nd ed.(Wiley, New York, 1981). 24S. K. O’Leary, Appl. Phys. Lett.72, 1332(1998). 25S. K. O’Leary, Appl. Phys. Lett.82, 2784(2003). 26W. B. Jackson, S. M. Kelso, C. C. Tsai, J. W. Allen, and S.-J. Oh, Phys. Rev. B 31, 5187(1985). 27S. M. Malik and S. K. O’Leary, J. Non-Cryst. Solids336, 64 (2004). 28S. M. Malik and S. K. O’Leary, submitted for publication. 29Such an ambiguity was commented upon by Roxloet al.30 for the specific case of a-Si:H. 30C. B. Roxlo, B. Abeles, C. R. Wronski, G. D. Cody, and T. Tiedje, Solid State Commun.47, 985 (1983). 31A recent Urbach focus analysis for the case of a-Si:H has been presented by Orapunt and O’Leary(Ref. 32). 32F. Orapunt and S. K. O’Leary, Appl. Phys. Lett.84, 523 (2004). 33R. E. Viturro and K. Weiser, Philos. Mag. B53, 93 (1986). 34S. K. O’Leary, S. Zukotynski, and J. M. Perz, J. Non-Cryst. Solids210, 249 (1997). J. Appl. Phys., Vol. 96, No. 12, 15 December 2004 Dias da Silva et al. 7059 [This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to ] IP: 186.217.234.225 On: Tue, 14 Jan 2014 13:05:34