Effect of hydrogen dilution on the structure of amorphous silicon alloys

Article abstract of DOI:10.1063/1.119928

We investigate why high levels of hydrogen dilution of the process gas lead to enhanced light soaking stability of amorphous silicon (a-Si) alloy solar cells by studying the microstructural properties of the material using high-resolution transmission electron microscopy (TEM) and Raman spectroscopy. The TEM results show that a-Si alloy (with or without hydrogen dilution) is a heterogeneous mixture of amorphous network and linear-like objects that show evidence of order along their length. The volume fraction of these ordered regions increases with increasing hydrogen dilution.

Full text of DOI:10.1063/1.119928

Effect of hydrogen dilution on the structure of amorphous silicon alloys
D. V. Tsu, B. S. Chao, S. R. Ovshinsky, S. Guha, and J. Yang
Citation: Applied Physics Letters 71, 1317 (1997); doi: 10.1063/1.119928
View online: http://dx.doi.org/10.1063/1.119928
View Table of Contents: http://scitation.aip.org/content/aip/journal/apl/71/10?ver=pdfcov
Published by the AIP Publishing
Articles you may be interested in
Effects of microwave power on the structural and emission properties of hydrogenated amorphous silicon carbide
deposited by electron cyclotron resonance chemical vapor deposition
J. Appl. Phys. 89, 2699 (2001); 10.1063/1.1344218
The effect of hydrogen on the network disorder in hydrogenated amorphous silicon
Appl. Phys. Lett. 75, 2803 (1999); 10.1063/1.125155
Structural, defect, and device behavior of hydrogenated amorphous Si near and above the onset of
microcrystallinity
Appl. Phys. Lett. 74, 1860 (1999); 10.1063/1.123693
Structural disorder induced in hydrogenated amorphous silicon by light soaking
Appl. Phys. Lett. 73, 3093 (1998); 10.1063/1.122683
Structural, optical, and electrical properties of hydrogenated amorphous silicon germanium alloys
J. Appl. Phys. 83, 4111 (1998); 10.1063/1.367229
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:
128.114.34.22 On: Thu, 27 Nov 2014 18:21:10

Effect of hydrogen dilution on the structure of amorphous silicon alloys
D. V. Tsu, B. S. Chao, and S. R. Ovshinsky
Energy Conversion Devices, Inc., Troy, Michigan 48084
S. Guha and J. Yang
United Solar System Corp., Troy, Michigan 48084
͑Received 29 April 1997; accepted for publication 3 July 1997͒
We investigate why high levels of hydrogen dilution of the process gas lead to enhanced light
soaking stability of amorphous silicon ͑a-Si͒ alloy solar cells by studying the microstructural
properties of the material using high-resolution transmission electron microscopy ͑TEM͒ and
Raman spectroscopy. The TEM results show that a-Si alloy ͑with or without hydrogen dilution͒ is
a heterogeneous mixture of amorphous network and linear-like objects that show evidence of order
along their length. The volume fraction of these ordered regions increases with increasing hydrogen
dilution. © 1997 American Institute of Physics. ͓S0003-6951͑97͒02736-8͔
Light-induced degradation in amorphous silicon ͑a-Si͒
alloy has received a great deal of attention in recent years.1
Since the first observation2 of improvement in stability of
a-Si alloy films prepared from a dilute mixture of silane and
hydrogen, hydrogen dilution of the process gas mixtures has
been extensively used to improve the stability of a-Si alloy
materials and devices. Recent results indicate3 that signifi-
cant improvement in the stability of a-Si alloy cells takes
place when the intrinsic layers are prepared from a gas mix-
ture heavily diluted with hydrogen. While the hydrogen con-
tent of this material is similar to that in the alloy prepared
with low hydrogen dilution, hydrogen effusion from this film
takes place at a much lower temperature. Small-angle x-ray
scattering4 also showed that a more stable material was char-
acterized by an anisotropic structure with low void density,
and it was postulated5 that an improvement in the structure
of the material was contributing to the improved stability.
In this letter, we investigate films prepared with increas-
ing hydrogen dilution using Raman spectroscopy and high-
resolution transmission electron microscopy ͑TEM͒ and
demonstrate that a more ordered structure is obtained as the
hydrogen dilution increases.
medium, and high levels of H2 dilution, respectively. All of
the spectra in Fig. 1 were normalized with respect to the
baseline (ϳ560 cmϪ1) and the LO band (ϳ400 cmϪ1). The
sharper band in trace ͑d͒ at 516.5 cmϪ1 is due to microcrys-
talline inclusions as found in the TEM images ͑Fig. 2͒. This
film was made with high hydrogen dilution. Besides this fea-
ture, the most obvious trend with increasing H2 dilution is a
shift in the peak position of the principal Si TO band from
474.5 cmϪ1 ͑no H2͒ to 482.6 cmϪ1 ͑high H2͒, as shown by
the dashed line in Fig. 1. The two very sharp features at
529.64 and 560.80 cmϪ1 are laser plasma lines.
In order to qualitatively account for the trend in the Si
TO peak position, we perform spectral subtractions using the
Raman spectrum of the standard a-Si alloy film as reference
to reveal any underlying spectral component. We compared
the difference spectra of many different standard a-Si alloy
samples, and find that there is little net signal above the
noise. In contrast, when compared to any of the modified H2
Three a-Si alloy films were made from a gas mixture of
disilane and hydrogen under H2 dilution levels, from low to
medium to high, while one standard a-Si alloy film was
made from silane without any hydrogen dilution to serve as
reference. The hydrogen to disilane ratios were varied from
5:1 to 150:1. All of the samples were prepared in a radio-
frequency glow-discharge diode reactor on 7059 glass, and
the thickness of each sample was kept near 500 Å. Details of
deposition conditions are as follows3: Pressure—1 Torr,
temperature—200–300 °C, and power density—50–150
mW/cm2.
We begin with the Raman scattering measurements
where we first look for qualitative differences, then quantify
those differences by Gaussian deconvolution. The Raman
spectra of the Si TO region were collected in the near-
backscattering geometry using the 4880 Å line of an Arϩ
laser and a double monochrometer. Details are given
elsewhere.6 Trace ͑a͒ of Fig. 1 shows the spectrum of the
standard a-Si alloy sample, while traces ͑b͒ through ͑d͒ are
the spectra for the modified a-Si alloy samples using the low,
FIG. 1. Raman spectra vs hydrogen dilution of the process gas. 500 Å films
on 7059 Corning glass: ͑a͒ standard a-Si alloy, sample RF5349 ͑no dilu-
tion͒; and modified a-Si alloy with ͑b͒ low dilution, US8014; ͑c͒ medium
dilution, US8035; and ͑d͒ high dilution, US8013.
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:
Appl. Phys. Lett. 71 (10), 8 September 1997 0003-6951/97/71(10)/1317/3/$10.00 © 1997 American Institute of Physics 1317
128.114.34.22 On: Thu, 27 Nov 2014 18:21:10

TABLE I. Gaussian deconvolution of the Raman spectra of Fig. 1.
Raman band area ͑%͒
Primary TO Secondary TO
␮c-Si
Sample
H2 dilution (475 cmϪ1
)
(490 cmϪ1
)
(510/517 cmϪ1
)
RF5349
US8014
US8035
US8013
US8013a
none
low
medium
high
95.5
90.0
90.4
76.1
84.6
4.5
10.0
9.6
13.8
15.4
•••
•••
•••
10.1
•••
high
aNormalized without the microcrystalline band.
mary TO Gaussian are then fixed to these values for all of the
samples. The difference spectra demonstrate that the excess
Raman activity for all of the H2 diluted samples is centered
at 490 cmϪ1. Therefore, this position is chosen for the sec-
ondary TO Gaussian for all of the samples, and its width is
40 cmϪ1. The band areas of each component are summarized
in Table I for each sample.
It is the growth of the 490 cmϪ1 band relative to the
475 cmϪ1 band that causes the peak of the overall TO band
to shift to higher frequencies as observed in Fig. 1. However,
the results of this analysis indicate that even the so-called
standard a-Si alloy, i.e., made with no H2 dilution, contains
some excess scattering in the 490 cmϪ1 region. One could
argue that this 4.5% residual scattering is just an indication
of some non-Gaussian behavior that can be simulated by the
addition of a weak second Gaussian. However, based upon
our TEM analysis, it is apparent that even standard a-Si alloy
is not completely homogeneous. We, therefore, conclude that
this residual 4.5% scattering is caused by a secondary phase.
The thin a-Si alloy films were lifted off their glass sub-
strates by etching with a weak HF solution, and the films
were supported by a copper TEM grid. The microstructure of
the films was analyzed by a Jeol 2010 STEM microscope.
Figure 2͑a͒ shows the bright-field micrograph of the modi-
fied high H2 dilution film in relatively low magnification.
Here we observe that the gross microstructure of the sample
consists of distinct microcrystalline inclusions imbedded in
the ‘‘featureless amorphous’’ matrix. The typical dimension
of each inclusion is ϳ50 nm consisting of spherical
microcrystallites ranging from 2.5 to 10 nm in diameter, as
shown in the corresponding dark-field micrograph in Fig.
2͑b͒. The presence and size of these smaller crystallites result
in the Raman band ͑Fig. 1͒ that is shifted down to ϳ517
from the 521 cmϪ1 position of bulk crystalline Si.7 The dif-
fraction pattern recorded from the same region shows two
distinct halos, which are normally assigned to the so-called
‘‘amorphous’’ silicon component. In addition, we observe6
small, bright diffraction spots which were identified earlier7
to be due to the microcrystalline inclusions which possess
long-range order. On the other hand, no such diffraction
spots are found in the standard material, nor in modified ͑low
or medium͒ H2 dilution material.
FIG. 2. ͑a͒ Low magnification bright-field TEM micrograph of modified
a-Si alloy, sample US8013 ͑high hydrogen dilution͒. Arrows indicate the
␮c-Si inclusions. ͑b͒ Dark-field TEM micrograph of same region.
dilution samples, the difference spectra always show a sig-
nificant signal centered at 490 cmϪ1 having a full width of
40 cmϪ1. Although all modified samples show the residual
490 cmϪ1 band in their difference spectra, the subtraction
method cannot reliably quantify its magnitude. We thus turn
to Gaussian deconvolution.
In our deconvolution of the Raman data, we use a mini-
mal number of Gaussian functions that is consistent with the
entire data set. For all of the spectra, one Gaussian is used to
represent the LO shoulder at ϳ400 cmϪ1. For this band, its
position, amplitude, and full width are all allowed to vary
during the fit. Except for sample US8013, all of the other
samples use two Gaussians to represent the principal TO
band. Based on detailed analysis of the spectra, we find that
two Gaussians represent the minimal number that is consis-
tent with the entire data set. They should represent, then, two
distinct environments within the material. For US8013, two
additional Gaussians are required ͑at 510.2 and 517.4 cmϪ1
͒
to accurately fit the ␮c-Si TO band.7 In all these cases, two
more Gaussians are used for the two laser plasma lines to
accurately position the peaks of the Raman bands during the
fit.
On a much finer scale in Fig. 3 of the high H2 dilution
sample, we clearly observe the long parallel lines separated
by 0.31 nm that correspond to the d spacing of Si͑111͒
planes of the microcrystallites. Aside from this obvious
Experimentally, we measure the peak position and width
of the TO band of the standard sample to be 475 and
70.5 cmϪ1, respectively. The position and width of the pri-
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:
1318 Appl. Phys. Lett., Vol. 71, No. 10, 8 September 1997 Tsu et al.
128.114.34.22 On: Thu, 27 Nov 2014 18:21:10

ence of excess hydrogen in the gas mixture passivates the
dangling bonds on the growing surface, and the impinging
species have more time to find favorable sites. This allows
for formation of more ordered structure. Excess hydrogen
also etches the growing surface, and in the extreme case of
heavy hydrogen dilution, strong etching eliminates part of
the disordered configurations.8 Since the crystalline phase is
the lowest energy configuration, it is often the surviving
structure. High dilution, therefore, favors the growth of mi-
crocrystalline material. The presence of microcrystalline re-
gions in large quantities affects the performance of the ma-
terial because of problems associated with grain boundaries.
The best material is obtained under the conditions just below
the threshold of microcrystalline growth. The research chal-
lenge will be to obtain conditions that will inhibit microcrys-
talline growth but will still promote the growth of the or-
dered regions. Other precursor gases could be useful. In fact,
we have been able to grow materials with similar ordered
regions using deuterated gases.
FIG. 3. High magnification bright-field TEM micrograph of ‘‘amorphous’’
matrix, sample US8013. Arrows show ordered linear-like objects.
In addition to creating conditions for the growth of a
more ordered material, hydrogen can also etch away weak or
strained bonds which will enhance stability. Although we see
a threefold increase in the volume fraction of the ordered
region as we go from the ‘‘standard’’ to the best quality
material, the fraction of the ordered region is still small. It is
possible that hydrogen dilution is causing other changes in
the structure or the bonding that are not detected by the
structural tools we have used.
In conclusion, we have shown that the volume fraction
of an ordered microstructure in a-Si alloy films increases
with increasing hydrogen dilution in the process gas mixture.
We suggest that the enhanced stability of solar cells prepared
with hydrogen dilution is caused by the improved micro-
structure.
structure, we also observe that the ‘‘amorphous’’ matrix is
not entirely featureless. In particular, as indicated by the ar-
rows, we find linearlike objects ͑i.e., one dimensional, hav-
ing widths 2–3 nm, and lengths up to a few 10’s of nm͒ that
appear to have some degree of order along their length.
These ‘‘linear’’ objects appear to meander throughout the
matrix in no particular direction. Similar linear objects are
also found in the rest of the samples, even in the standard
material! The ‘‘amorphous’’ matrix then consists of a hetero-
geneous mixture of a truly amorphous component, as de-
scribed by the short-range order of a continuous random net-
work ͑CRN͒, and objects that are intermediate in order
between the CRN and crystalline phases, i.e., the somewhat
ordered linear-like objects. It is this newly identified struc-
ture that we believe produces the 490 cmϪ1 band, and if so,
the results of the Raman deconvolution indicate that the vol-
ume fraction occupied by them within the ‘‘amorphous’’ ma-
trix increases with increasing H2 dilution.
The authors thank X. Xu and S. Sugiyama for discus-
sions, Dr. John Bradley at MVA, Inc. for TEM examina-
tions, and V. Trudeau for manuscript preparation.
The presence of the ordered linear-like regions and its
impact on stability merits some discussion. As the volume
fraction of these ordered regions grows, the quality of the
material improves as evidenced by the increased stability.
The best stability is obtained for the high hydrogen dilution
case where the volume fraction of these ordered regions is
the highest. It is interesting to point out that this material
actually includes a small amount of microcrystalline inclu-
sion, and effusion of hydrogen occurs more readily from this
material. Further increase in hydrogen dilution increases the
microcrystalline volume fraction further, and the solar cell
shows lower open-circuit voltage and fill factor characteristic
of microcrystalline solar cells.
1 For a recent review, see H. Fritzsche, Amorphous and Microcrystalline
Silicon Technology–1997, Materials Research Society Symposium Pro-
ceedings, 1997 ͑to be published͒.
2 S. Guha, K. L. Narasimhan, and S. M. Pietruszko, J. Appl. Phys. 52, 859
͑1981͒.
3 X. Xu, J. Yang, and S. Guha, J. Non-Cryst. Solids 198–200, 60 ͑1996͒.
4 D. L. Williamson, in Amorphous Silicon Technology–1995, Materials Re-
search Society Symposium Proceedings Vol. 377, edited by M. Hack, E.
A. Schiff, A. Madan, M. Powell, and A. Matsuda ͑Materials Research
Society, Pittsburgh, 1995͒, pp. 251–262.
5 S. Guha, Proc. 25th IEEE PVSC, 1017 ͑1996͒.
6 D. V. Tsu, B. Chao, S. R. Ovshinsky, S. Guha, and J. Yang ͑unpublished͒.
7 R. Tsu, S. S. Chao, M. Izu, S. R. Ovshinsky, G. J. Jan, and F. H. Pollack,
J. Phys. ͑France͒ C4, 269 ͑1981͒.
8 C. C. Tsai, in Amorphous Silicon and Related Materials, edited by H.
Fritzsche ͑World Scientific, Singapore, 1988͒, Vol. 1, p. 123.
What is the role of hydrogen? We believe that the pres-
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:
Appl. Phys. Lett., Vol. 71, No. 10, 8 September 1997 Tsu et al. 1319
128.114.34.22 On: Thu, 27 Nov 2014 18:21:10