scispace - formally typeset

Journal ArticleDOI

A comparison of grain nucleation and grain growth during crystallization of HWCVD and PECVD a-Si:H films

15 Jan 2008-Thin Solid Films (Elsevier)-Vol. 516, Iss: 5, pp 529-532

AbstractFrom TEM, XRD and Raman measurements, we compare the crystallization kinetics when HWCVD and PECVD a-Si:H films, containing different initial film hydrogen contents (CH), are crystallized by annealing at 600 °C. For the HWCVD films, the nucleation rate increases, and the incubation time and the full width at half maximum (FWHM) of the XRD (111) peak decrease with decreasing film CH. However, the crystallization kinetics of HWCVD and PECVD films of similar initial film CH are quite different, suggesting that other factors beside the initial film hydrogen content affect the crystallization process. Even though the bonded hydrogen evolves very early from the film during annealing, we suggest that the initial spatial distribution of hydrogen plays a critical role in the crystallization kinetics, and we propose a preliminary model to describe this process.

Topics: Crystallization (56%), Nucleation (53%)

Summary (2 min read)

1. Introduction

  • The crystallization of as deposited a-Si:H thin films is becoming increasingly important because of its potential use to produce higher mobility polycrystalline materials for use in solar cells and high performance thin film transistors.
  • This process is believed, at relatively low anneal temperatures (b 1000 °C), to follow a classical model of nucleation and grain growth [1], where an amorphous incubation time, a steady state nucleation rate, grain growth of these nuclei, and a characteristic time of crystallization can be identified.
  • Further, recent results have shown that the crystallization time, for a given film CH, can depend upon the a-Si:H deposition method [4].

2. Experimental

  • H films were deposited by HWCVD and PECVD, using deposition conditions described previously [6,7], also known as A-Si.
  • Regarding the latter, the width of the XRD diffraction peaks may result from a combination of grain size, defect density, or strain effects.
  • This thickness was chosen because no sample thinning was required.
  • TEM analysis was performed on a CM200 Scanning TEM using a Phillips single-slit holder and a Gatan Model 652 double-slit heating holder for in-situ annealing.

3. Results and discussion

  • Fig. 1 shows TEM images of partially crystallized HWCVD and PECVD a-Si:H films annealed at 600 °C, with the annealing times indicated in the figures.
  • From this figure, clear differences are seen not only in the anneal time, but also in the grain density and grain morphology between the different films.
  • These differences become even more pronounced when a low CH HWCVD film is included in the comparison compared to the PECVD film seen in Fig. 1(b) [8,9]; in these comparisons, the PECVD film showed much lower grain densities and larger grains overall.
  • In Table 3 the authors first present a review of the NMR data, from which the densities of the isolated and clustered hydrogen distributions can be obtained.
  • H calculation, the authors use an averaged value of the clustered/isolated hydrogen ratio, obtained for ‘standard’ (low deposition rate) films deposited using 100% silane and a moderate (∼ 200–250 C) substrate temperature, also known as For the PECVD a-Si.

4. Summary and conclusions

  • The authors have presented the crystallization kinetics when HWCVD films of different film CH and ‘standard’ PECVD aSi:H films have been annealed at a temperature of 600 °C to induce film crystallization.
  • The authors find that the low CH HWCVD film nucleates first, and that the incubation time increases with increasing film CH.
  • Not surprisingly, the films which nucleate the fastest contain the smallest grains when crystallization is complete.
  • The increase in short range disorder upon film hydrogen evolution does not seem to play a primary role in the crystallization process.
  • A tentative model relating the crystallization kinetics to the initial hydrogen spatial distribution in the film is presented.

Did you find this useful? Give us your feedback

...read more

Content maybe subject to copyright    Report

A comparison of grain nucleation and grain growth during
crystallization of HWCVD and PECVD a-Si:H films
A.H. Mahan
a,
, S.P. Ahrenkiel
a
, R.E.I. Schropp
b
,H.Li
b
, D.S. Ginley
a
a
National Renewable Energy Laboratory, Golden, CO 80401, United States
b
Utrecht University, Faculty of Science, 3508 TA Utrecht, The Netherlands
Available online 14 June 2007
Abstract
From TEM, XRD and Raman measurements, we compare the crystallization kinetics when HWCVD and PECVD a-Si:H films, containing
different initial film hydrogen contents (C
H
), are crystallized by annealing at 600 °C. For the HWCVD films, the nucleation rate increases, and the
incubation time and the full width at half maximum (FWHM) of the XRD (111) peak decrease with decreasing film C
H
. However, the
crystallization kinetics of HWCVD and PECVD films of similar initial film C
H
are quite different, suggesting that other factors beside the initial
film hydrogen content affect the crystallization process. Even though the bonded hydrogen evolves very early from the film during annealing, we
suggest that the initial spatial distribution of hydrogen plays a critical role in the crystallization kinetics, and we propose a preliminary model to
describe this process.
© 2007 Published by Elsevier B.V.
Keywords: Hydrogenated amorphous silicon; Annealing; Crystallization kinetics; Crystallite nucleation; Nuclear magnetic resonance
1. Introduction
The crystallization of as deposited a-Si:H thin films is
becoming increasingly important because of its potential use to
produce higher mobility polycrystalline materials for use in
solar cells and high performance thin film transistors. This
process is believed, at relatively low anneal temperatures
(b 1000 °C), to follow a classical model of nuclea tion and grain
growth [1], where an amorphous incubation time, a steady state
nucleation rate, grain growth of these nuclei, and a characteristic
time of crystallization can be identified. While the incubation
time can be examined by a varie ty of techniques, TEM
measurements during the early stages of nucleation are crucial
to determine the nucleation rate and grain growth rates, from
which a final grain size can be extrapolated. In limited previous
studies at an anneal temperature of 600 °C for PECVD a-Si:H,
all of these process steps were seen to depend on the film
substrate temperature (T
S
) [2,3]. While there was agreement in
these studies that lower T
S
films exhibited longer incubation
times and larger grain sizes, the latter due primarily to the
smaller nucleation rate, there was no general consensus as to
why this occurred. One approach linked the trends in grain size
with T
S
to differences in the Raman signatures of the as grown
films, which were then related to differences in short range
structural order [2,3] . However, these Raman analyses were
either not rigorous [2] or were done with an interpretation no
longer considered appropriate [3]. Further, recent results have
shown that the crystallization time, for a given film C
H
, can
depend upon the a-Si:H deposition method [4]. The present
work thus re-examines a-Si:H crystallization [5], with an
attempt to determine not only what factors limit grain size, but
also to understand how H influences nucleation.
2. Experimental
A-Si:H films were deposited by HWCVD and PECVD,
using deposition conditions described previously [6,7].Two
different film thicknesses were then analyzed. First, 1 μ m
thick a-Si:H films were deposited on 1737 Corning glass, and
Raman spectroscopy and XRD were used respectively to probe
the film short range order during the amorphous film incubation
(hydrogen evolution) period, and the crystallite defect structure
A
vailable online at www.sciencedirect.com
Thin Solid Films 516 (2008) 529 532
www.elsevier.com/locate/tsf
Corresponding author.
E-mail address: harv_mahan@nrel.gov (A.H. Mahan).
0040-6090/$ - see front matter © 2007 Published by Elsevier B.V.
doi:10.1016/j.tsf.2007.06.036

during the crystallization process. Regarding the latter, the
widthoftheXRDdiffractionpeaksmayresultfroma
combination of grain size, defect density, or strain effects.
Second, thinner ( 0.1 μm) films were grown directly on C-
coated, 200 mesh Mo TEM grids which were mounted on the
glass substrates using colloidal graphite paste. This thickness
was chosen because no sample thinning was required. TEM
analysis was performed on a CM200 Scanning TEM using a
Phillips single-slit holder and a Gatan Model 652 double-slit
heating holder for in-situ annealing. TEM images were acquired
in the conical dark-field mode with a 12-bit digital camera from
Soft Imaging System, with the direct beam tilted by 0.70° and
dynamically pressed about the optic axis of the microscope. To
determine the crystalline volume fraction X
c
(= V
c
/V, where V
c
is the crystalline volume and V is the total volume) from TEM
images, we assume grain growth to be two dimensional and use
the magic wand tool in Photoshop for boundary recognition to
binarize the images, converting crystalline (amorphous) regions
to black (gray) respectively. Histograms of the one-bit images
were then compu ted in Digital Micrograph to determine X
c
.
Direct grain counting was used to determine the areal grain
number densities r
g
(= N
g
/A, where N
g
is the number of grains
in area A). Each image series was analyzed incrementally,
starting at the early stages of crystallization, and keeping a
running total while counting emerging nuclei in subsequent
images.
The XRD measurements were obtained using a Scintag X1
diffractometer. XRD spectra of all samples were measured for
2ϑ between 20 and 60°, and the correlation length were obtained
from the Sherrer formula by fitting the full width at half
maximum (FWHM) of the Si(111) peaks using Pearson-7
lineshapes. Raman scattering spectroscopy was performed with
a laser wavelength of 514.5 nm. The power density of the
incident laser beam was carefully adjusted to avoid the influence
from unintentional sample heating. A long integration time was
used when necessary. An optical polarizer was used to suppress
the scattered light from the substrate (Corning glass) and the
background signal. Data were first normalized in the range
between 150 cm
1
and 600 cm
1
. The lowest signal within this
range was treated as the base line. No additional background
fitting was used. FWHM values were obtained by fitting the
scattering spectra between 460 cm
1
and 560 cm
1
.
3. Results and discussion
Fig. 1 shows TEM images of partially crystallized HWCVD
and PE CVD a-Si:H films annealed at 600 °C, with the
annealing times indicated in the figures. Based upon infrared
measurements of the SiH wag mode, the film H contents
(C
H
's) are roughly equivalent ( 1113 at.%). In addition, the
X
c
's of these partially crystallized films are roughly similar.
These images clearly show that crystallization occurs via grain
nucleation, which occurs at random in the amorphous matrix,
and grain grow th into the surrounding amorphous material.
From this figure, clear differences are seen not only in the
anneal time, but also in the grain density and grain morphology
between the different films. These differences become even
more pronounced when a low C
H
HWCVD film is included in
the comparison compared to the PECVD film seen in Fig. 1(b)
[8,9]; in these compa risons, the PECVD film showed much
lower grain densities and larger grains overall. These differ-
ences are tabulated in Table 1, which gives the measured
crystallization parameters for the HWCVD and PECVD films
annealed at 600 °C [8]. The crystallite nucleation rate (r
n
) and
grain growth rate (s
g
) were determined by examining successive
TEM images at low X
c
, and the extrapolated final grain size (d
g
)
has been calcul ated using the methodology of Iverson and Reif
[1]. Also shown in Table 1 are the correlation lengths of the
three films, obtained from the full width at half maximum
Fig. 1. TEM images taken in the early crystalline regimes of high C
H
HWCVD
film (a) and (similar) high C
H
PECVD film (b). The crystalline volume fractions
are roughly similar. The 600 °C anneal times are included in the figures.
Table 1
Measured crystallization parameters for HWCVD and PECVD films annealed at
600 °C
Film type HWCVD (low H) HWCVD (high H) PECVD (high H)
r
n
(min/μm
3
)
1
2.3 0.16 0.027
s
g
(nm/min) 4.1 3.1 2.7
d
g
(x)(μm) 0.31 0.66 1.20
XRD correlation
length (μm)
0.08 0.055 0.045
530 A.H. Mahan et al. / Thin Solid Films 516 (2008) 529532

(FWHM) of the Si (111) XRD diffraction peak after crystal-
lization is complete. As can be seen, while the TEM final grain
size increases from low C
H
HWCVD to high C
H
HWCVD to
(high C
H
) PECVD films, the correlation length shows the
opposite trend, becoming smaller with the same film progres-
sion. As all three types of films have been shown to exhibit
roughly the same film stress, the increasingly larger discrepancy
between TEM grain size and XRD correlation length, in the
progression from low C
H
HWCVD to (high C
H
) PECVD films,
may suggest an increasing crystallite defect density. This
interpretation is supported by the TEM images of Fig. 1, which
show increasingly jagged grain surfaces for the PECVD fil m
(smallest correlation length) compared to that for the HWCVD
film (larger correlation length), which shows, on average,
smoother grain surfaces. Once again, this interpretation is even
clearer when the TEM image of the low C
H
HWCVD film is
included [8,9].
Since the (HWCVD, PECVD) films exhibiting the smallest
XRD correlation lengths contain the most initial bonded C
H
,we
explore whether there is any correlation between the a-Si:H
short range order, as measured by the half width at half
maximum (HWHM) of the Raman transverse optical mode
[10], and the amount of film C
H
evolved from the films during
their respective incubation periods. The idea behind this
examination comes from AFM measurements of high C
H
(N 10 at.%) a-Si:H films undergoing rapid thermal anneals at
temperatur es N 600 °C while still remaining within t he
amorphous incubation period. As these films are literally
blown apart by the rapid evolution of the bonded C
H
[5],itis
interesting to exami ne if the film short range order is reduced,
and if so, does this increased film damage play any role in
subsequent crystallite formation. These results are shown in
Table 2, where the time for the 600 °C anneal for each film has
been adjusted so that each film has been annealed for 3/4 of
its respective incubation period. Preliminary measurements for
the HWCVD films have been presented elsewhere [8]. As can
be seen, although the Raman HWHM's for all films do broaden
a little during their respective film incubation periods upon
600 °C anneal, indicating a reduction in film short range order,
the amount of this broadening is not significantly larger for
films containing more initial film C
H
. These results suggest that
the evolution of different amounts of film C
H
seems not to be
correlated with defect creation on the crystallite surfaces. This,
upon reflection, may not be surprising since the film C
H
evolves
very early in the incubation period and long before nucleation
commences.
It has also been suggested that short range order affects the
PECVD a-Si:H nucleation rate upon film anneal [3,4]. In these
works, the increase in disorder was larger for the lower substrate
temperature (higher C
H
) films, and was correlated with lower
nucleation rates. We do not observe this trend, as the short range
disorder for the three films examined, presen ted in Table 2,
increases in roughly a similar manner for all films, irrespective
of the film C
H
or film deposition type. Therefore, we suggest
that while film disorder may play some role in the nucleation
process, the relationship between film disorder and nucleation
rate may be only of secondary importance.
We advance the idea, on the other hand, that the hydrogen
spatial distributions in the films, measured in their as grown
states, play a defining role in the crystallization process. As seen
in Table 1, the low C
H
HWCVD film crystallizes most quickly,
while the (higher C
H
) PECVD film takes the longest to
crystallize. The bonded hydrogen distributions in these two
extreme cases have been rigorously explor ed by NMR [1114].
Of critical interest is the ratio of the isolated to clustered
hydrogen in the two comparative films, as well as the number of
hydrogen atoms in the clusters, as obtained by multiple
quantum NMR measurements [15].InTable 3 we first present
a review of the NMR data, from which the densities of the
isolated and clustered hydrogen distributions can be obtained.
For the PECVD a-Si:H calculation, we use an averaged value of
the clustered/isolated hydrogen ratio, obtained for standard
(low deposi tion rate) films deposited using 100% silane and a
moderate ( 200250 C) substrate temperature. As can be seen,
both densities are roughly an order of magnitude lower for the
low C
H
HWCVD film compared to those for the (higher C
H
)
PECVD film, and are due not only to the different film C
H
's but
also to the different clustered/isolated hydrogen ratios. In
particular, the low C
H
HWCVD film is seen to contain minimal
isolated hydrogen.
In the lower part of the table we present calculations of the
average distances between the hydrogen containing regions.
Assuming that both types of hydrogen, as probed by NMR, are
randomly dis tributed in th e films, we ca n, to a fi rst
approximation, calculate the sizes of the regions in the films
that contain no, or minimal hydrogen. If we further assume the
shapes of these hydrogen deficient regions to be spherical, we
can then calculate the number of Si atoms in these regions.
These numbers are presented in the last row of the table. It is
interesting to compare these numbers with the sizes of silicon
crystallites that are stable in a-Si:H. Based upon free energy
considerations, the stability of a crystallite is a sum of two
Table 2
Raman to HWHM for HWCVD and PECVD films as grown and annealed at
600 °C while still remaining in incubation period
Film type HWCVD (low H)
(cm
1
)
HWCVD (high H)
(cm
1
)
PECVD (high H)
(cm
1
)
As grown 28.3 28.6 28.0
600 °C anneal 29.3 31.9 29.6
Table 3
H distributions in as grown a-Si:H
HWCVD (low C
H
) PECVD (high C
H
)
Clustered/isolated 9/1 6/4
H/cluster N 15 6
Clusters/cm
3
6×10
19
6×10
20
Isolated/cm
3
1×10
20
2×10
21
Distance between
H clusters 25 Å 11 Å
Isolated H 21 Å
Spherical volume 7238 Å
3
380 Å
3
# Si/sphere 360 19
531A.H. Mahan et al. / Thin Solid Films 516 (2008) 529532

terms, a (negative) volume term proportional to the total number
of Si atoms in the crystallite, and a (positive) surface term
proportional to the number of atoms on the crystallite surface.
The volume term is equal to the free energy difference between
the amorphous and crystalline phases, while the surface term is
related to the surface tension at the amorphous/crystalline
interface [16]. The number of Si atoms in this critical crystallite
size, above which the crystallite grows and below which it tends
to shrink, has been calculated theoretically, and has been found
to range between 40 [16] and 110 [17] Si atoms respectively.
It is interesting that both of these predictions of stable
crystallite size lie between the values of the hydrogen deficient
regions for the low C
H
HWCVD and (higher C
H
) PECVD
films. Based upon this argument we suggest that, if the
hydrogen deficient regions become nucleation centers, they are
already larger than the critical size for the low C
H
HWCVD film
when they crystallize, and can grow larger immediately upon
further annealing, resulting in a comparatively short incubation
time. Conversely, the much smaller hydrogen deficient regions
for the (higher C
H
) PECVD film will have considerable trouble
growing upon film anneal when they nucleate, and this results in
much longer incubation times.
Of course, this argument presupposes we associate these
hydrogen deficient regions with nucleation centers, which then
grow according to the manner described above. While extensive
positron annihi lation measurements have shown that no
crystallites exist in as grown HWCVD a-Si:H [18], XRD
measurements which probe the film medium range order have
been perfor med [19], and have shown that the low C
H
HWCVD
films are better ordered than are the PECVD films which
contain higher film C
H
[20]. Since the low C
H
HWCVD films
contain so little C
H
, the XRD measurements on this fil m
primarily probe the hydrogen deficient regions. The nature of a
nucleation center has to date not been conclusively formulated;
models proposed range from highly disordered regions which
gain the most energy when they crystallize [21] to more ordered
regions which take less energy to crystallize [22]. The present
results favor the latter model, and measurements are in progress
to link the amorphous incubation time to the XRD medium
range order parameter 2ϑ.
4. Summary and conclusions
We have presented t he crystallization kinetics when
HWCVD films of different film C
H
and stand ard PECVD a-
Si:H films have been annealed at a temperature of 600 °C to
induce film crystallization. We find that the low C
H
HWCVD
film nucleates first, and that the incubation time increases with
increasing film C
H
. We also find a dependence upon film
deposition type, as PECVD films containing a C
H
similar to an
HWCVD film take a much longer time to nucleate. Not
surprisingly, the films which nucleate the fast est contain the
smallest grains when crystallization is complete. The increase in
short range disorder upon film hydrogen evolution does not
seem to play a primary role in the crystallization process. A
tentative model relating the crystallization kinetics to the initial
hydrogen spatial distribution in the film is presented.
Acknowledgements
The authors thank D.L. Williamson for XRD measurements.
This work was funded by the United States DOE under
subcontract number DE-AC36-99-GO10337.
References
[1] R.B. Iverson, R. Reif, J. Appl. Phys. 62 (1987) 1675.
[2] J.N. Lee, B.J. Lee, D.G. Moon, B.T. Ahn, Jpn. J. Appl. Phys. 36 (1997) 6862.
[3] K. Nakazawa, K. Tanaka, J. Appl. Phys. 68 (1990) 1029.
[4] D.L. Young, P. Stradins, E. Iwaniczko, B. To, B. Reedy, Y. Yan, H.M.
Branz, J. Lohr, M. Alvarz, J. Booske, A. Marconnet, Q. Wang, MRS
Symp. Proc. 862 (2005) 233.
[5] A.H. Mahan, B. Roy, R.C. Reedy Jr., D.W. Readey, D.S. Ginley, J. Appl.
Phys. 99 (2006) 023507.
[6] A.H. Mahan, J. Carapella, B.P. Nelson, R.S. Crandall, I. Balberg, J. Appl.
Phys. 69 (1991) 6728.
[7] The substrate temperature for the PECVD deposition was 350 °C, resulting
in C
H
12 at.%.
[8] A.H. Mahan, S.P. Ahrenkiel, B.Roy, R.E.I. Schropp, H. Li, D.S. Ginley,
Proc. WCPEC4 (Hawaii, May 2006), in press.
[9] S.P. Ahrenkiel, A.H. Mahan, B.Roy, D.S. Ginley, Proc. 2006 MRS Spring
Conference (Symposium A) (San Francisco, May 2006), in press.
[10] D. Beeman, R. Tsu, M.F. Thorpe, Phys. Rev. B 32 (1985) 874.
[11] Y. Wu, J.T. Stephen, D.X. Han, J.M. Rutland, R.S. Crandall, A.H. Mahan,
Phys. Rev. Lett. 77 (1996) 2049.
[12] J.A. Reimer, R.W. Vaughan, J.C. Knights, Phys. Rev. B 24 (1981) 3360.
[13] K.K. Gleason, M.A. Petrich, J.A. Reimer, Phys. Rev. B 36 (1987) 3259.
[14] W.E. Carlos, P.C. Taylor, Phys. Rev. B 26 (1982) 3605.
[15] J. Baum, K.K. Gleason, A. Pines, A.N. Garroway, J.A. Reimer, Phys. Rev.
Lett. 56 (1986) 1377.
[16] C. Spinella, S. Lombardo, F. Priolo, J. Appl. Phys. 84 (1998) 5383.
[17] K.N. Tu, Appl. Phys. A 53 (1991) 32.
[18] D.T. Britton, A. Hempel, M. Harting, G. Kogel, P. Sperr, W. Triftshauser,
C. Arendse, D. Knoesen, Phys. Rev. B 64 (2001) 075403.
[19] D.L. Williamson, MRS Symp. Proc. 557 (1999) 251.
[20] A.H. Mahan, L.M. Gedvilas, J.D. Webb, J. Appl. Phys. 87 (2000) 1650.
[21] R.W. Readey, private communication.
[22] A.T. Voutsas, M.K. Hatalis, J. Electrochem. Soc. 140 (1993) 871.
532 A.H. Mahan et al. / Thin Solid Films 516 (2008) 529532
Citations
More filters

Journal ArticleDOI
Abstract: Utilizing the concepts of a critical crystallite size and local film inhomogeneity, it is shown that nucleation in thermally annealed hydrogenated amorphous silicon occurs in the more well ordered spatial regions in the network, which are defined by the initial inhomogeneous H distributions in the as-grown films. Although the film H evolves very early during annealing, the local film order is largely retained in the still amorphous films even after the vast majority of the H is evolved, and the more well ordered regions which are the nucleation center sites for crystallization are those spatial regions which do not initially contain clustered H, as probed by H NMR spectroscopy. The sizes of these better ordered regions relative to a critical crystallite size determine the film incubation times (the time before the onset of crystallization). Changes in film short range order upon H evolution, and the presence of microvoid type structures in the as grown films play no role in the crystallization process. While the creation of dangling bonds upon H evolution may play a role in the actual phase transformation itself, the film defect densities measured just prior to the onset of crystallization exhibit no trends which can be correlated with the film incubation times.

23 citations


Journal ArticleDOI
Abstract: Amorphous Si (a-Si) films with lower hydrogen contents show better adhesion to glass during flash lamp annealing (FLA). The 2.0 µm-thick a-Si films deposited by plasma-enhanced chemical vapor deposition (PECVD), containing 10% hydrogen, start to peel off even at a lamp irradiance lower than that required for crystallization, whereas a-Si films deposited by catalytic CVD (Cat-CVD) partially adhere even after crystallization. Dehydrogenated Cat-CVD a-Si films show much better adhesion to glass, and are converted to polycrystalline Si (poly-Si) without serious peeling, but are accompanied by the generation of crack-like structures. These facts demonstrate the superiority of as-deposited Cat-CVD a-Si films as a precursor material for micrometer-thick poly-Si formed by FLA.

21 citations


Journal ArticleDOI
Abstract: In this paper the effect of the microstructure of remote plasma-deposited amorphous silicon films on the grain size development in polycrystalline silicon upon solid-phase crystallization is reported. The hydrogenated amorphous silicon films are deposited at different microstructure parameter values R* (which represents the distribution of SiHx bonds in amorphous silicon), at constant hydrogen content. Amorphous silicon films undergo a phase transformation during solid-phase crystallization and the process results in fully (poly-)crystallized films. An increase in amorphous film structural disorder (i.e., an increase in R*), leads to the development of larger grain sizes (in the range of 700-1100 nm). When the microstructure parameter is reduced, the grain size ranges between 100 and 450 nm. These results point to the microstructure parameter having a key role in controlling the grain size of the polycrystalline silicon films and thus the performance of polycrystalline silicon solar cells.

21 citations


Journal ArticleDOI
Abstract: The Johnson–Mehl–Avrami–Kolmogorov (JMAK) model is widely used to quantify the isothermal crystallization kinetics. The present work reports an analytical solution for the crystallization kinetics in the special case of plate-shaped samples with a finite thickness. As a result, we obtained an adapted JMAK model revealing the thickness range which influences the crystallization kinetics mode significantly. The analytical solution also provides theoretical bounds for the film thickness, where the assumption of 2D or 3D kinetics is accurate. Finally, the conclusions related to amorphous silicon and amorphous nickel-titanium thin films are reported.

14 citations


Book ChapterDOI
19 Sep 2012
Abstract: Polycrystalline silicon thin films have attracted the attention of semiconductor industries in the past few decades due to their wide applications in thin film transistors, solar cells, display units and sensors (Schropp & Zeman, 1998; Choi et al., 2005; Mahan et al., 2008). Polycrystalline Si thin films are generally fabricated by crystallizing amorphous Si (a-Si) thin films, because these can render larger grains compared to the conventional poly Si film deposition. As a consequence, a variety of methods for lowering the crystallization temperature of a-Si have been developed. Excimer laser annealing is one of the promising ways to achieve large grain size poly Si films at lower substrate temperatures. Its high costs and nonuniform grain size, however, are significant obstacles that prevent its wide use (Parr et al., 2002). The other promising technique is the solid phase crystallization method. But this technique is essentially a high-temperature process and many substrates, including most forms of glass, cannot withstand the thermal processing. In order to achieve lower costs and have a wider range of application, inexpensive materials such as glass and special polymers have to substitute quartz or PyrexTM substrates. In the case of glass substrates, all of the processing steps need to be limited to temperatures below 550 °C. The other known technique is rapid thermal annealing (RTA). In RTA infrared radiation is used as a heating source, and has the advantage of the high heating speed (up to 60 oC/s) that reduces the crystallization time. In RTA radiation is applied in pulses to heat the sample without heating the glass substrate (which is transparent to the infrared radiation). However, the grain size obtained in the crystallization of a-Si is also in the range of a few micrometers.

13 citations


References
More filters

Journal ArticleDOI
Abstract: Device‐quality hydrogenated amorphous silicon containing as little as 1/10 the bonded H observed in device‐quality glow discharge films have been deposited by thermal decomposition of silane on a heated filament. These low H content films show an Urbach edge width of 50 mV and a spin density of ∼1/100 as large as that of glow discharge films containing comparable amounts of H. High substrate temperatures, deposition in a high flux of atomic H, and lack of energetic particle bombardment are suggested as reasons for this behavior.

462 citations


Journal ArticleDOI
TL;DR: It is shown that the width of the ``optic peak'' increases roughly linearly with the rms bond-angle distortion of the network, consistent with model-building experience which shows that it is impossible to construct fully bonded amorphous networks with \ensuremath{\Delta}${\ensureMath{\theta}}_{b}$.
Abstract: The Raman scattering from various model structures for amorphous silicon is computed. It is shown that the width of the ``optic peak'' increases roughly linearly with the rms bond-angle distortion \ensuremath{\Delta}${\ensuremath{\theta}}_{b}$ of the network. The experimentally observed linewidths lead to 7.7\ifmmode^\circ\else\textdegree\fi{}\ensuremath{\le}\ensuremath{\Delta}${\ensuremath{\theta}}_{b}$\ensuremath{\le}10.5\ifmmode^\circ\else\textdegree\fi{}. The smaller linewidths (and hence angles) correspond to networks that have been annealed at higher temperatures. These results are consistent with model-building experience which shows that it is impossible to construct fully bonded amorphous networks with \ensuremath{\Delta}${\ensuremath{\theta}}_{b}$\ensuremath{\le}6.6\ifmmode^\circ\else\textdegree\fi{}.

379 citations


Journal ArticleDOI
Abstract: The solid phase crystallization of chemical vapor deposited amorphous silicon films onto oxidized silicon wafers, induced either by thermal annealing or by ion beam irradiation at high substrate temperatures, has been extensively developed and it is reviewed here. We report and discuss a large variety of processing conditions. The structural and thermodynamical properties of the starting phase are emphasized. The morphological evolution of the amorphous towards the polycrystalline phase is investigated by transmission electron microscopy and it is interpreted in terms of a physical model containing few free parameters related to the thermodynamical properties of amorphous silicon and to the kinetical mechanisms of crystal grain growth. A direct extension of this model explains also the data concerning the ion-assisted crystal grain nucleation.

327 citations


Journal ArticleDOI
Abstract: This paper presents a theoretical and experimental study of the recrystallization behavior of polycrystalline silicon films amorphized by self‐implantation. The crystallization behavior was found to be similar to the crystallization behavior of films deposited in the amorphous state, as reported in the literature; however, a transient time was observed, during which negligible crystallization occurs. The films were prepared by low‐pressure chemical vapor deposition onto thermally oxidized silicon wafers and amorphized by implantation of silicon ions. The transient time, nucleation rate, and characteristic crystallization time were determined from the crystalline fraction and density of grains in partially recrystallized samples for anneal temperatures from 580 to 640 °C. The growth velocity was calculated from the nucleation rate and crystallization time and is lower than values in the literature for films deposited in the amorphous state. The final grain size, as calculated from the crystallization param...

326 citations


Journal ArticleDOI
TL;DR: Using the fact that multiple-quantum excitation is limited by the size of the dipolar-coupled spin system, it is shown that the predominant bonding environment for hydrogen is a cluster of four to seven atoms.
Abstract: Multiple-quantum nuclear-magnetic-resonance techniques are used to study the distribution of hydrogen in hydrogenated amorphous silicon. Using the fact that multiple-quantum excitation is limited by the size of the dipolar-coupled spin system, we show that the predominant bonding environment for hydrogen is a cluster of four to seven atoms. For device quality films, the concentration of these cluster defects increases with increasing hydrogen content. At very high hydrogen content, the clusters are replaced with a continuous network of silicon-hydrogen bonds.

169 citations


Frequently Asked Questions (1)
Q1. What contributions have the authors mentioned in the paper "A comparison of grain nucleation and grain growth during crystallization of hwcvd and pecvd a-si:h films" ?

Even though the bonded hydrogen evolves very early from the film during annealing, the authors suggest that the initial spatial distribution of hydrogen plays a critical role in the crystallization kinetics, and they propose a preliminary model to describe this process.