The preparation of insitu doped hydrogenated ...

The preparation of i n s i t u doped hydrogenated amorphous silicon by homogeneous chemical vapor deposition B. S. Meyerson, B. A. Scott, and D. J. Wolford Citation: Journal of Applied Physics 54, 1461 (1983); doi: 10.1063/1.332172 View online: http://dx.doi.org/10.1063/1.332172 View Table of Contents: http://scitation.aip.org/content/aip/journal/jap/54/3?ver=pdfcov Published by the AIP Publishing Articles you may be interested in Selective deposition of i n s i t u doped polycrystalline silicon by rapid thermal processing chemical vapor deposition Appl. Phys. Lett. 55, 2408 (1989); 10.1063/1.102029 Pressure dependence of i n s i t u borondoped silicon films prepared by lowpressure chemical vapor deposition. II. Resistivity J. Appl. Phys. 66, 4812 (1989); 10.1063/1.343795 Pressure dependence of i n s i t u borondoped silicon films prepared by lowpressure chemical vapor deposition. I. Microstructure J. Appl. Phys. 66, 4806 (1989); 10.1063/1.343794 Plasmaenhanced chemical vapor deposition of i n s i t u doped epitaxial silicon at low temperatures. II. Boron doping J. Appl. Phys. 65, 1067 (1989); 10.1063/1.343041 Plasmaenhanced chemical vapor deposition of i n s i t u doped epitaxial silicon at low temperatures. I. Arsenic doping J. Appl. Phys. 65, 1053 (1989); 10.1063/1.343040

[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: 129.186.1.55 On: Sun, 24 Aug 2014 19:14:44

The preparation of in situ doped hydrogenated amorphous silicon by homogeneous chemical vapor deposition B. S. Meyerson, B. A. Scott, and D. J. Wolford IBM T. J. Watson Research Center, Yorktown Heights, New York 10598

(Received 16 August 1982; accepted for publication 23 November 1982) Raman scattering, infrared absorption, conductivity measurements, electron microprobe, and secondary ion mass spectrometry (SIMS) were used to characterize boron and phosphorus doped hydrogenated amorphous silicon (a-Si:H) films prepared by Homogeneous Chemical Vapor Deposition (HOMOCVD). HOMOCVD is a thermal process which relies upon the gas phase pyrolysis of a source (silane containing up to 1.0% diborane or phosphine) to generate activated species for deposition upon a cooled substrate. Doped films prepared at 275°C by this process were found to contain - 12-at. % hydrogen as determined by infrared absorption. We examined dopant incorporation from the gas phase, obtaining values for a distribution coefficient CD (film dopant content/gas phase dopant concentration, atomic basis) ofO.33';;;CD ';;;0.63 for boron, while 0.4.;;; CD .;;; 10.75 in the limits 3.3 X 1O- 5 ';;;PH 3/SiH4';;;0.004. We interpret the data as indicative of the formation of an unstable phosphorus/silicon intermediate in the gas phase, leading to the observed enhancements in CD at high gas phase phosphine content. HOMOCVD films doped at least as efficiently as their prepared counterparts, but tended to achieve higher conductivities [a;;;.O.1 (n cm) - I for 4.0% incorporated phosphorus] in the limit of heavy doping. Raman spectra showed no evidence of crystallinity in the doped films. Film properties (conductivity, activation energy of of conduction) have not saturated at the doping levels investigated here, making the attainment of higher "active" dopant levels a possibility. We attribute the observation that HOMOCVD appears more amenable to high "active" dopant levels than plasma techniques to the low (-0.1 eV) thermal energy at which HOMOCVD proceeds, versus -1{}-100 eV for plasma techniques. Low substrate temperature (75°C) doped films were prepared with initial results showing these films to dope as readily as those prepared at high temperature (T - 275°C). PACS numbers: 73.60.Fw, n.20.Jv, 81.10.Bk

INTRODUCTION

Device (photovoltaic) quality hydrogenated amorphous silicon (a-Si:H) has been grown for several years via the glow discharge decomposition of silicon bearing gas species (SiH4' 1,2 Si2H 6,3,4 SiF/). Alternate a-Si:H preparation techniques such as sputtering6 and ion beam deposition 7 also appear in the literature, though with less frequency. To date, little is published regarding the in situ preparation of high quality (low spin density, photoconducting, hydrogen passivated) a-Si:H via the pyrolysis of silane (SiH 4 ), i.e., chemical vapor deposition (CVD). As CVD is an exceptionally clean and well controlled process, it is commonly used to prepare high quality polycrystalline 8 and epitaxial9 silicon thin films, yet to date it has been unsuitable for the preparation of aSi:H. The temperature required to pyrolyze SiH4 , 1';;;.500 °C, is the primary difficulty with CVD. In a hot gaslhot substrate CVD environment, the hydrogen required for defect passivation in the a-Si network is evolved from the growing film, resulting in a-Si displaying a high defect density. The Fermi level is thus pinned near midgap, precluding doping, and rendering the material unsuitable for most applications. Consequently, such CVD prepared films must be posthydrogenated,1O which reduces the defect density such that efficient doping may take place. It would obviously be preferable that as-prepared CVD material have sufficient hydrogen content for defect passivation, and two approaches towards resolving this have recently appeared. One method of en1461

J. Appl. Phys. 54 (3), March 1983

hancing hydrogen incorporation during growth is to reduce CVD operating temperatures by using a more easily pyrolyzed silicon source, i.e., disilane (Si2H 6).11 Another approach, studied further here, is the use of a novel CVD technique, homogeneous CVD (HOMOCVD). HOMOCVD is a hot gas/cold substrate CVD technique, first used to studyl2.13 the kinetics and mechanism of film growth from SiH2 intermediates. Intrinsic films thus prepared were found to contain 4-40% hydrogen and displayed electrical characteristics (activation energy of conduction, dark conductivity, and photoconductivity) quite similar to their plasma prepared counterparts. 13 In the following we report our studies of the doping of HOMOCVD a-Si:H. In light of the relatively simple thermal chemistry involved in HOMOCVD, which yields low defect density material 14 over a wide range of preparation conditions, such studies are of special interest for comparison with the results of plasma deposited material. EXPERIMENTAL

Samples were prepared in a modified LPCVD (Low Pressure Chemical Vapor Deposition) system configured as in Fig. 1. Source gas mixtures containing up to 1% diborane (B2H6) or phosphine (PH 3) in SiH4 were used. The detailed operation of such a hot gas/cold substrate HOMOCVD system is set forth in earlier work,13 but in summary this technique relies on the gas phase (homogeneous) decomposition

0021-8979/83/031461-05$02.40

® 1983 American Institute of Physics

1461


MASS SPECTROMETER SAMPLING PORT

U1!s:\M~:

_- -L~_-_. . SOU~~~TGAS

- - --L---_-_',;...I

==========~~~C~OOLANT

l

I

F

FIG. 1. Homogeneous CVD apparatus.

IN

I../'\

~IT~H~E~R~M~O~C~O~UP~L=E~~~~COOLANT

PUMP PACKAGE

L_________ J t

OUT

QUARTZ FURNACE TUBE

/ FURNACE

of the source molecule, rather than on surface decomposition (heterogeneous) as in standard hot wallihot substrate CVD techniques. Typical operating conditions were substrate temperatures 25EA >0.25 eV for 0.26% B content, versus 8 X 1O- lO ";;O'RT ";;0.1 (il cm)-I and 0.7>EA >0.15 eV for a-Si:H containing up to 4.0% phsoporus. High conductivity in a-Si:H has been attributed in some reports to microcrystalline domains within an amorphous network. IS.16 To test for this we have used raman spectroscopy, which, for micro- or polycrystalline material, yields unmistakable two component scattering (crystalline and amorphous), 17.18 or spectra intermediate 19 between crystal Si (520 cm -I) and fully amorphized Si (-480 cm -I). Since conductivity was highest among the phosphorus doped films, all were tested. To eliminate the possibility of subsurface crystallinity, a phenomenon observed by Magarine and Kaplan, 16 samples were probed from both top and bottom surfaces (films were 3000 A thick). Typical results in back scattering geometry using 50 m W of 514.5-nm excitation are shown in Fig. 5. Spectra consist of the broad component (50100 cm- I ) centered near 485 em-I, which characterizes amorphous Si. 20 Significantly, none of the films, through the most conductive (top, Fig. 5), show evidence of the sharp 520-cm - 1line or narrow down-shifted "crystalline" scattering coexisting with the amorphous band. We may therefore attribute the high conductivities observed to doping rather than morphological effects. Infrared spectroscopy was used to determine hydrogen content, by integration of the 640-cm - I SiHx wagging mode absorption.21 For Ts = 275 DC, -12.0-at. % hydrogen content was observed, in good agreement with results for intrinsic films. The validity of this measurement technique has been confirmed for our samples by both nuclear magnetic resonance l3 and nuclear reaction techniques, with agreement among the measurements to within ± 10% of the H content found. 1463

J. Appl. Phys .. Vol. 54, No.3, March 1983

800

600

400

200

o

RAMAN SHIFT (em-I)

FIG. 5. Raman spectra for homogeneous CVD samples in the limits of phosphorus doping.

DISCUSSION

When one calculates22 the expected dependence of CD upon the phosphine/silane ratio, assuming independent decomposition pathways for each gas, it is straightforward to demonstrate that CD would be independent of phosphine partial pressure. This is inconsistent with the results we have obtained, and we interpret our findings as indicating the presence of a phosphorus/silicon gas phase chemistry, yielding the observed enhancements in CD' Several observations lend credence to this interpretation. First, the interaction must proceed in the gas phase as our substrate temperatures are well below those required to dissociate either source gas in a heterogeneous (surface) process. Additionally, we examined phosphorus incorporation over the range of substrate temperatures investigated here, and found it to be temperature independent, again making heterogeneous chemistry unlikely. Finally, we could induce the homogeneous nucleation ofpartic1es down-stream of the reaction upon the introduction of small concentrations of phosphine into otherwise stable hot (- 700 DC) silane gas within the reactor, indicative ofa homogeneous process. Thus, to account for the enhancement observed in CD' we postulated the presence of a silicon/phosphorus intermediate in the gas phase, and have calculated22 the dependence of CD upon the gas phase phosphine/silane ratio. Assuming the gas phase generation of a phosphorus/silicon intermediate, monosilyiphosphine (SiPHs) being a reasonable first approximation, we have obtained an expression in good agreement with our results. The presence of this intermediate in our system has been verified Meyerson, Scott, and Wolford

1463


by mass spectrometry, and such a reaction pathway has been observed by other authors. Blazejowski and Lampe 23 found that SiPH5 was produced via silylene (SiH 2) insertions into a phosphorus-hydrogen bonds during the photodecomposition of silane in the presence of phosphine. Our earlier work 13 established SiH 2 as the primary precursor of films grown by HOMOCVD, and thus it is quite reasonable to expect that silylene insertions into phosphine molecules occur in our SiH4/PH 3 gas mixtures. The considerable enhancement of bulk phosphorus content with increasing gas phase PH 3 concentration would therefore result from the presence of SiPH5 (or other unstable Si-P compounds) so generated, as by their generation they deplete silylene radicals otherwise available for film formation, inducing a second-order effect. The detailed reaction pathway leading to this result will be published shortly. The great contrast in the behavior of CD between HOMOCVD and plasma techniques, a consequence of this gas phase chemistry, results in order of magnitude corrections to the doping curves (Figs. 3 and 4) for solid versus gas phase dopant content, and emphasizes the need to used solid phase dopant levels in presenting data for doping efficiencies when comparing thermal and plasma preparation methods. Electrical measurement data plotted (Figs. 3 and 4) against the corrected bulk dopant content reveal a trend in which HOMOCVD films achieve both higher conductivities as well as lower activation energies in the dopant limit investigated here. Note that saturation of either (TRT or E A has not yet occurred for HOMOCVD material, suggesting that higher values of (TRT and lower E A may be attainable in more heavily doped films. The finding that one may more heavily dope HOMOCVD compared to plasma prepared a-Si:H can be ascribed to either an intrinsically lower defect density in the films, and/or a greater proportion of "active" dopant centers relative to similarly doped plasma films. ESR 14 has shown spin densities of _10 15 cm- 3 in undoped HOMOCVD films prepared at 275 ·C, a concentration comparable though not superior to that found in plasma prepared aSi:H.24 We note however that ESR cannot detect spinless defects such as gap states induced by the presence of weak or strained bonds, and thus is not an absolute indicator of the equivalence of defect densities in the two materials discussed here. More efficient incorporation of "active" dopant centers appears the more likely explanation of these results, an interpretation consistent with our finding that electrical properties are not yet saturated at relatively high doping levels. Efficient "active" dopant incorporation could be a consequence of the low energy thermal regime in which HOMOCVD operates, E-O.1 eV, versus plasma methods where ionic species occur which impinge the substrate at -10-100 eV energies. The greatly reduced energy available at the surface during HOMOCVD could result in the reduced surface mobility of adsorbed dopant atoms, slowing their migration to potentially inactive sites at defects or at dopant clusters. Clustering of dopant atoms effectively limits the "solubility" of dopants in the a-Si:H network, and the onset of clustering is the point at which additional dopant incorporation ceases to significantly alter electrical proper1464

J. Appl. Phys., Vol. 54, No.3, March 1983

ties. 25 Thus, the reduced dopant surface mobility expected during HOMOCVD should make possible the achievement of greater concentrations of "active" dopant centers than with plasma preparation techniques. Dopant atoms may also be rendered inert by coordination with one or more hydrogen atoms in the a-Si:H network (it is as yet unresolved as the exact coordination necessary to render a dopant atom inactive), as has recently been observed by Greenbaum et al.,26 and this would seem more likely to occur during a plasma process where there are greater concentrations of free H available to bind to the dopant atoms. The behavior of HOMOCVD a-Si:H departs radically from that of plasma prepared material for films prepared with Ts