Analysis of Controlled Mixed-Phase „Amorphous+Microcrystalline… Silicon Thin ﬁlms by Real Time Spectroscopic Ellipsometry ͒ N

Home , Crystal growth

Analysis of controlled mixed-phase „amorphous+microcrystalline… silicon thin films by real time spectroscopic ellipsometry ͒ N. J. Podrazaa Department of Electrical Engineering, The Pennsylvania State University, University Park, Pennsylvania 16802 Jing Li Department of Materials Science and Engineering, The Pennsylvania State University, University Park, Pennsylvania 16802 C. R. Wronski Department of Electrical Engineering, The Pennsylvania State University, University Park, Pennsylvania 16802 E. C. Dickey Department of Materials Science and Engineering, The Pennsylvania State University, University Park, Pennsylvania 16802 R. W. Collins Center for Photovoltaics Innovation and Commercialization, University of Toledo, Toledo, Ohio 43606 ͑Received 1 June 2009; accepted 3 August 2009; published 14 September 2009͒ Engineered thin films consisting of periodic arrays of silicon microcrystallites in a hydrogenated amorphous silicon host matrix have been prepared by plasma-enhanced chemical vapor deposition where the hydrogen dilution of silane is modulated in multiple cycles. These types of films have been guided by a phase evolution diagram, depicting the deposition conditions and film thickness at which the material exhibits amorphous, microcrystalline, or mixed-phase ͑amorphous ͒ +microcrystalline characteristics, developed for intrinsic Si:H prepared with varying H2 dilution on unhydrogenated a-Si:H. Real time spectroscopic ellipsometry ͑RTSE͒ has been used in situ to noninvasively determine the phase evolution of the resulting hydrogenated mixed-phase ͑amorphous+microcrystalline͒ silicon thin films and corroborated with dark-field transmission electron microscopy. Such tailored microstructures are of growing interest as components of thin film photovoltaic devices, and RTSE is shown to be a key technique for structure verification. © 2009 American Vacuum Society. ͓DOI: 10.1116/1.3212893͔

I. INTRODUCTION Improvements in thin film Si:H solar cell efficiency have been achieved using these optimization principles facilitated ͑ ͒ Thin-film hydrogenated silicon Si:H based n-i-p and by phase evolution diagrams, depicting the deposition condi- p-i-n photovoltaic devices may incorporate amorphous sili- tions and film thickness at which the material exhibits amor- ͑ ͒ ͑␮ ͒ con a-Si:H , microcrystalline silicon c-Si:H , and even phous, microcrystalline, or mixed-phase ͑amorphous mixed-phase ͑amorphous+microcrystalline͒ silicon ͓͑a +microcrystalline͒ characteristics, as determined in real time +␮c͒-Si:H͔ intrinsic layer ͑i-layer͒ components. Optimum spectroscopic ellipsometry ͑RTSE͒ studies.2,5–7 Further im- single-step a-Si:H i-layers, as assessed by the stabilized ef- provements in the stability of a-Si:H based thin film solar ficiency of the resulting photovoltaic devices, are produced ͑ ␮ ͒ ͑ ͒ cells have been reported by incorporating a+ c -Si:H by plasma-enhanced chemical vapor deposition PECVD 8–10 from SiH at the maximum H dilution possible while avoid- i-layers into the device structure; however, optimization 4 2 ͑ ␮ ͒ ing the nucleation of crystallites from the amorphous phase and reproducible control of a+ c -Si:H deposition are a 1–3 challenge due to the rapid evolution of the microcrystallite throughout the i-layer thickness. The high H2 dilution ensures a well-ordered amorphous network that rapidly reaches content, the extreme substrate dependence, and thus the dif- its fully light-soaked state in less than 100 h at air mass 1.5. ficulty of controlling the volume fraction of crystallites at a In contrast, the corresponding optimum single-step ␮c-Si:H constant average value throughout the thickness. Although i-layers are produced by PECVD at the minimum H2 dilution continuous hydrogen dilution profiling has been widely while maintaining continuous crystal growth on a microcrys- adopted in order to optimize grain boundary passivation in 4,5 ␮ 11 talline surface. In this case, the low H2 dilution ensures c-Si:H i-layer materials throughout the thickness, the that grain boundaries are well passivated with an amorphous more challenging problem of modulating hydrogen dilution phase. to generate ͑a+␮c͒-Si:H with a controlled fraction of em- bedded microcrytallites has yet to be widely applied. The ͒ ͑ ␮ ͒ a Author to whom correspondence should be addressed; electronic mail present work demonstrates that a+ c -Si:H i-layer films [email protected] can be fabricated with periodic arrays of microcrystallites in

1255 J. Vac. Sci. Technol. A 27„6…, Nov/Dec 2009 0734-2101/2009/27„6…/1255/5/$25.00 ©2009 American Vacuum Society 1255 1256 Podraza et al.: Analysis of controlled mixed-phase „amorphous+microcrystalline… silicon thin films 1256 an amorphous matrix, whereby the average crystallite size as well as the in-plane and out-of-plane crystallite spacings can be separately modified solely by deposition conditions estab- lished and verified by RTSE.

II. EXPERIMENTAL DETAILS The ͑a+␮c͒-Si:H i-layer films studied here were deposited on single-phase a-Si:H layers, which were prepared, in turn, on native oxide coated crystalline silicon ͑c-Si͒ substrates using a single-chamber rf ͑13.56 MHz͒ PECVD sys- tem. The deposition parameters included a modulated hydro- ͓ ͔/͓ ͔ gen dilution ratio, R= H2 SiH4 , with all other parameters ͑ ͒ fixed as follows: a low substrate temperature Ts =200 °C , the minimum rf power possible for a stable plasma ͑ / 2͒ 0.08 W cm , a low partial pressure of the SiH4 source gas ͑ϳ0.06 Torr͒, and a low total pressure ͑Ͻ1.0 Torr͒. The initial a-Si:H layer was prepared under the same conditions but with fixed R=0. This layer serves to erase the memory of the c-Si wafer substrate and also mimics the underlying p-or FIG. 1. Surface roughness layer thickness ͑top͒ and microcrystallite volume n-layer for i-layer deposition in p-i-n or n-i-p solar cell con- fraction in the near-surface of the bulk layer ͑bottom͒ plotted vs bulk layer ͓ ͔/͓ ͔ figurations, respectively. An initial series of Si:H films was thickness for a high H2-dilution R= H2 SiH4 =40 film deposited on a crystalline silicon ͑c-Si͒ wafer substrate coated with undiluted R=0 amor- prepared on these structures at hydrogen dilution ratios of phous silicon ͑a-Si:H͒. The evolution of the microcrystallite fraction as 0ՅRՅ170 for phase evolution diagram development, while predicted by a cone growth geometrical model is also shown ͑solid line͒;the multilayer mixed-phase ͑a+␮c͒-Si:H films were prepared inset shows parametrizations of the imaginary parts of the dielectric function ␧ ͑ ͒ ͑ by modulating R during the deposition. The Si:H deposition spectra, 2, for the fully amorphous dotted line and microcrystalline solid line͒ R=40 hydrogenated silicon ͑Si:H͒ material. process was investigated in situ and in real time using a rotating-compensator multichannel ellipsometer.12 Films prepared at low H2 dilution, for example, R=0, parameters that include the volume fractions of Si:H crystal- remain in the amorphous growth regime throughout the lites ͑f␮ ͒ and voids ͑f ͒ in the outer layer, the outer-layer thickness. For such films, the time evolution of the bulk and c void growth rate r, and the surface roughness layer thickness ds. surface roughness layer thicknesses, as well as the bulk layer After this information is obtained, additional structural pa- ͑␧ ␧ ͒ dielectric functions 1 , 2 , were extracted from the RTSE rameters such as the areal density of microcrystallites in the data using a global ͚␴ minimization procedure, where ␴ is ͑ ͒ ͑␪͒ 13 amorphous matrix Nd and the half-angle of the inverted the unweighted error function. Films prepared with high H2 cones can be estimated using a geometric model.15 dilution tend to develop in the amorphous ͑protocrystalline͒ phase for an initial period of growth but subsequently nucleate isolated microcrystallites from within the amorphous ma- III. RESULTS trix at a well-defined thickness. These nuclei grow in the Figure 1 shows the evolution of the surface roughness form of inverted cones, and eventually coalesce to single- thickness and microcrystallite volume fraction in the outer phase ␮c-Si:H. Consequently, the final films exhibit an layer ͑top 3 Å͒ for an R=40 Si:H film prepared on a native amorphous phase near the substrate interface and a microc- oxide/c-Si substrate coated with R=0 a-Si:H. Microcrystal- rystalline phase at the top of the film, each with distinct lites first nucleate from the amorphous phase at a bulk layer dielectric functions. An intervening structurally graded ͑a thickness of db =160 Å, which is denoted as the amorphous +␮c͒-Si:H phase exists, wherein the microcrystallite frac- to mixed-phase ͑amorphous+microcrystalline͓͒a→͑a ␮ ͔͒ tion increases with bulk layer thickness from zero ͑initial + c transition. At db =500 Å, the microcrystallites coa- nucleation͒ to near unity ͑full coalescence͒. lesce to form single-phase ␮c-Si:H, which is referred to as RTSE data collected during the growth of such structur- the mixed-phase to single-phase microcrystalline ͓͑a+␮c͒ ally graded ͑a+␮c͒-Si:H are conveniently interpreted using →␮c͔ transition. The microcrystallite nucleation density is ͑ ͒ 14,15 ϳ ϫ 10 −2 virtual interface analysis VIA . This analysis utilizes a estimated at Nd 6.0 10 cm , which corresponds to an four medium optical model consisting of ͑i͒ the ambient, ͑ii͒ initial in-plane microcrystallite spacing of ϳ400 Å, and the a surface roughness layer, ͑iii͒ an outer layer that contains cone half-angle is estimated at ␪ϳ35°. the most recently deposited material, and a virtual interface The inset in Fig. 1 shows the spectra in the imaginary part ͑ ͒ ␧ to iv the pseudosubstrate which contains the past history of of the dielectric function, 2, for the near-substrate amor- the deposition. The surface roughness layer is modeled using phous and near-surface microcrystalline R=40 Si:H materi- the Bruggeman effective medium theory as a 0.5/0.5 volume als. The dielectric function for the amorphous phase was fraction mixture of the outer-layer material and void.16 The determined by exact inversion assuming bulk and surface VIA approach is based on least-squares regression with free roughness thicknesses obtained by ͚␴ minimization using

J. Vac. Sci. Technol. A, Vol. 27, No. 6, Nov/Dec 2009 1257 Podraza et al.: Analysis of controlled mixed-phase „amorphous+microcrystalline… silicon thin ﬁlms 1257

FIG. 2. Deposition phase diagram for hydrogenated silicon ͑Si:H͒ ﬁlms deposited by rf PECVD ͑f =13.56 MHz͒ onto an undiluted amorphous silicon ͑ ͓ ͔/͓ ͔ ͒ R= H2 SiH4 =0, a-Si:H substrate. The ﬁxed deposition parameters include a relatively low substrate temperature, Ts =200 °C, the minimum power for a stable plasma, P=0.08 W/cm2, and a low total pressure, p Ͻ1 Torr. The variable parameter, forming the abscissa of the diagram, is ͓ ͔/͓ ͔ the H2-dilution ratio R= H2 SiH4 , which is varied from R=0 to 170. The ordinate axis represents the bulk layer thickness at which the amorphous to mixed-phase ͑amorphous+microcrystalline͓͒a→͑a+␮c͔͒ ͑solid squares͒ and mixed-phase to single-phase microcrystalline ͓͑a+␮c͒→␮c͔͑open squares͒ transitions occur.

FIG. 3. Bulk layer thickness evolution of the surface roughness ͑top͒, the data acquired in the first ϳ150 Å of R=40 a-Si:H film microcrystalline volume fraction in the near-surface of the bulk layer growth on top of the R=0 layer. The inverted dielectric func- ͑middle͒, and the void volume fraction ͑bottom͒ for a modulated film struc- ͑ ͓ ͔/͓ ͔ ͒ ͑ ͒ 17 ture prepared by alternating high R= H2 SiH4 =40 and low R=0 tion was then fit to a Cody–Lorentz oscillator function. The ͑ ͒ H2-dilution hydrogenated silicon Si:H , as determined by VIA applied to dielectric function for the microcrystalline phase was deter- RTSE data. mined using the VIA technique, and the final inverted dielectric function was fit to two Tauc–Lorentz oscillators ͑simu- ͒ lating the E1 and E2 transitions of crystalline Si , sharing a the plane. The particular film described here consists of three 18,19 common band gap, Eg. pairs of R=0/R=40 layers deposited onto a native oxide Figure 2 shows a phase evolution diagram produced for a covered c-Si substrate. Figure 3 shows the surface roughness series of Si:H films prepared with H2 dilution ranging from layer thickness and the microcrystalline and void volume R=0 to R=170 on undiluted R=0 coated native oxide/c-Si fractions in the outer layer ͑3–20 Å thick for R=40to0, substrates. The ͓a→͑a+␮c͔͒ and ͓͑a+␮c͒→␮c͔ transitions respectively͒ as functions of the bulk layer thickness, which are observed and depicted as functions of R. Films prepared is obtained as a time integral of the instantaneous growth rate Ͻ ͑ ͒ ͑␧ ␧ ͒ at R 15 do not exhibit these transitions but remain amor- r t determined from the VIA. The 1 , 2 spectra used in phous throughout ϳ4000 Å of growth. Films prepared at the analysis of Fig. 3 have been deduced from the previous 15ഛRഛ150 initially nucleate in the amorphous phase but R=0 and R=40 depositions. Figure 4 shows a cross-sectional undergo the ͓a→͑a+␮c͔͒ transition at a bulk layer thick- transmission electron microscopy ͑TEM͒ dark-field image of ness, dependent on the H2 dilution R. The bulk layer thick- the structure in Fig. 3. ness at which this transition occurs decreases with increasing R throughout the series. As with the ͓a→͑a+␮c͔͒ transition, IV. DISCUSSION the ͓͑a+␮c͒→␮c͔ transition occurs at a bulk layer thickness, dependent on the H2 dilution R, and this thickness de- The microstructural evolution of Si:H materials, as shown creases with increasing R. It should be noted that for R in Fig. 1, has been used to produce a phase evolution dia- Ͼ170, the film initially nucleates as ␮c-Si:H on the amor- gram as a function of the hydrogen dilution ratio R, as shown phous R=0 substrate. in Fig. 2. This phase diagram evolution diagram, coupled A modulated Si:H structure was then deposited while with previous microcrystallite nucleation density studies,15 measuring by RTSE with the intention of fabricating a film has been used to guide the modulation of R during the course with a volume average microcrystalline fraction. The modu- of a deposition to produce a ͑a+␮c͒-Si:H film shown in lated film was designed to consist of alternating high-dilution Figs. 3 and 4 with a controllable microcrystallite growth as a ͑R=40͒ crystallite nucleation regions and low-dilution ͑R function of depth. =0͒ crystallite suppression regions in order to produce a A comparison of Figs. 3 and 4 corroborates the thick- multilayer mixed-phase film with microcrystallites spaced nesses of the amorphous and mixed-phase regions of the ͑a approximately 400 Å apart both in the film plane and out of +␮c͒-Si:H film as determined by the VIA. Both VIA and

JVST A - Vacuum, Surfaces, and Films 1258 Podraza et al.: Analysis of controlled mixed-phase „amorphous+microcrystalline… silicon thin films 1258 ͑ R=0 a-Si:H layers and in the H2-dilution of the high R a +␮c͒-Si:H layers can be used to produce films with periodic arrays of microcrystallites, adjusting for substrate depen- dences so as to achieve controllable separations in the plane of the film and out of the plane.

V. SUMMARY In conclusion, RTSE with virtual interface analysis has been applied to study the growth of mixed-phase ͑a +␮c͒-Si:H films fabricated so as to maintain a constant volume average microcrystalline fraction with thickness. RTSE studies have been used to determine the thickness and sub- FIG. 4. Dark-field cross-sectional TEM image of the modulated hydrogen- strate dependence of the structural evolution, and thereby ͑ ͒ ated silicon Si:H film structure in Fig. 3 prepared by alternating Si:H with identify the required H dilution and bulk layer thicknesses high ͑R=͓H ͔/͓SiH ͔=40͒ and low ͑R=0͒ H dilution. 2 2 4 2 of components in engineered Si:H films that can result in specific isotropic or anisotropic distributions of microcrystallites. It is found that microcrystallite nucleation can be TEM results demonstrate that it is possible for the low- stopped and restarted with well-defined conical crystallite dilution R=0 deposition to suppress further microcrystallite decay and growth behaviors via the use of alternating low growth on the R=40 mixed-phase material and thereby pro- and high H2-dilution layers; however, the protocrystalline duce alternating amorphous and mixed-phase regions, the layers decrease in thickness with increasing cycle number latter with a defined size and average spacing for the crys- due to the increase in ordering of the substrate with thick- talline inclusions. Subsequent R=0/40 cycles, however, ness. As a result, to obtain periodic arrays of crystallites, the have shown variations in the microcrystallite evolution de- hydrogen dilution R of the high-dilution layers must be de- pending on the underlying structure which provide additional creased with increasing cycle number. insights into the well-known substrate dependence of phase evolution.6 The most obvious example of such variations is the additional voids that appear with microcrystallite nucle- ACKNOWLEDGMENTS ation for the second and third R=40 layers. For these R =40 layers, the surface roughness thicknesses at the a→͑a This research was sponsored by the U.S. Army Research ␮ ͒ + c transition have increased from ds =10 Å for the first Office and U.S. Army Research Laboratory and was accom- plished under Cooperative Agreement No. W911NF-0-2- R=40 layer to ds =30 Å and ds =60 Å, respectively. As the surface roughness thickness increases, voids first appear in 0026. ͑ ͒ the mixed-phase layer at ds =30 Å and then stabilize in ͑ 1D. V. Tsu, B. S. Chao, S. R. Ovshinsky, S. Guha, and J. Yang, Appl. Phys. volume fraction throughout the mixed-phase regime at ds ͑ ͒ ͒ Lett. 71, 1317 1997 . =60 Å . This behavior can be attributed to the inability of 2J. Koh, Y. Lee, H. Fujiwara, C. R. Wronski, and R. W. Collins, Appl. the crystalline phase to conformally cover rough surfaces and Phys. Lett. 73, 1526 ͑1998͒. the associated shadowing and failure of the amorphous phase 3J. Yang, K. Lord, and S. Guha, Mater. Res. Soc. Symp. Proc. 609, A15.4 ͑ ͒ the fill the interstices. A second example of the effect of 2000 . 4O. Vetterl, F. Finger, R. Carius, P. Hapke, L. Houben, O. Kluth, A. Lam- roughness is a loss in the regularity of the microcrystallite bertz, A. Muck, B. Rech, and H. Wagner, Sol. Energy Mater. Sol. Cells profile, which occurs for the third cycle in Figs. 3 and 4. 62,97͑2000͒. 5 It is also important to note in Fig. 3 that the ͓a→͑a J. A. Stoke, N. J. Podraza, J. Li, X. Cao, X. Deng, and R. W. Collins, J. ␮ ͔͒ Non-Cryst. Solids 354, 2435 ͑2008͒. + c transition shifts to lower bulk layer with increasing 6R. W. Collins et al., Sol. Energy Mater. Sol. Cells 78, 143 ͑2003͒. cycle number, as indicated by the bulk thickness required to 7X. Cao et al., J. Non-Cryst. Solids 354, 2397 ͑2008͒. 8 reach a given f␮ value, e.g., if f␮ =0.25, then d T. Kamei, P. Stradins, and A. Matsuda, Appl. Phys. Lett. 74,1707͑1999͒. c c b 9 =320, 220, 140 Å for cycles 1, 2, and 3, respectively. This M. Ito, N. Myojin, M. Kondo, A. Matsuda, M. Shiratani, and Y. Wa- tanabe, Proceedings of the Third World Conference on Photovoltaic En- shift can be attributed to the improved ordering of the under- ergy Conversion, 2003 ͑unpublished͒, p. 1592. lying R=0 materials with increasing cycle number due to the 10S. Y. Myong, S. W. Kwon, K. S. Lim, and M. Konagai, Sol. Energy influence of the underlying ͑a+␮c͒-Si:H. In order to im- Mater. Sol. Cells 85, 133 ͑2005͒. 11 prove the periodicity of these structures, thicker R=0 layers B. Yan, G. Yue, J. Yang, S. Guha, D. L. Williamson, D. Han, and C.-S. Jiang, Appl. Phys. Lett. 85, 1955 ͑2004͒. may be necessary with increasing cycle number. Alterna- 12I. An, J. A. Zapien, C. Chen, A. S. Ferlauto, A. S. Lawrence, and R. W. tively, the H2 dilution of the high R layers could be reduced Collins, Thin Solid Films 455–456, 132 ͑2004͒. slightly with increasing cycle number. This behavior would 13Y. Cong, I. An, K. Vedam, and R. W. Collins, Appl. Opt. 30,2692 ͑ ͒ shift the ͓a→͑a+␮c͔͒ transition to higher bulk layer thick- 1991 . 14H. Fujiwara, J. Koh, and R. W. Collins, Thin Solid Films 313–314,474 ness and reduce the crystallite nucleation density as well. ͑1998͒. Thus, it should be clear that variations in the thickness of the 15A. S. Ferlauto, G. M. Ferreira, R. J. Koval, J. M. Pearce, C. R. Wronski,

J. Vac. Sci. Technol. A, Vol. 27, No. 6, Nov/Dec 2009 1259 Podraza et al.: Analysis of controlled mixed-phase „amorphous+microcrystalline… silicon thin ﬁlms 1259

R. W. Collins, M. M. Al-Jassim, and K. M. Jones, Thin Solid Films 17A. S. Ferlauto, G. M. Ferreira, J. M. Pearce, C. R. Wronski, R. W. Collins, 455–456, 665 ͑2004͒. X. Deng, and G. Ganguly, J. Appl. Phys. 92,2424͑2002͒. 16H. Fujiwara, J. Koh, P. I. Rovira, and R. W. Collins, Phys. Rev. B 61, 18G. E. Jellison, Jr. and F. A. Modine, Appl. Phys. Lett. 69,371͑1996͒. 10832 ͑2000͒. 19G. E. Jellison, Jr. and F. A. Modine, Appl. Phys. Lett. 69,2137͑1996͒.

JVST A - Vacuum, Surfaces, and Films