Download I Physical characterization of a-Si thin films deposited by thermal I

J. Phys. D: A@. Phys. 24 (1991) 1015-1021. Printed in the UK
I
I
Physical characterization of a-Si thin
films deposited by thermal
decomposition of iodosilanes
G Tamizhmanit, Michael Cociverats, Richard T Oakleytg,
Charles FischerS and Minoru Fujimoto*
t Guelph-Waterloo Centre for Graduate Work in Chemistry. University of Guelph,
Guelph. Ontario. Canada N1G 2W1
Guelph-Waterloo Program for Graduate Work in Physics, University of Guelph,
Guelph, Ontario, Canada N1G 2W1
*
Received 30 April 1990, in final form 7 February 1991
I
Abstract. Amorphous silicon thin films have been deposited by a new chemical
vapour deposition process using iodosilane precursors at atmospheric pressure.
Fourier transform infrared, ultravioletlvisible. electron spin resonance and
Rutherford backscattering spectroscopies along with scanning electron microscopy
and resistivity measurements were used to determine some of the properties of
films that were prepared under various conditions and subjected to post-deposition
treatments. It was found that the dangling bond density could be decreased by
post-deposition heat treatment in an atmosphere of hydrogen or argon.
1. Introduction
Various methods have been used todeposit a-Si thin film
[l-31. Because chemical vapour deposition (CVD)
methods involve low-energy particles and need no
expensive equipment, they are considered to be among
the promising methods for producing large area samples.
The Fourier transform infrared (FTIR) spectrum of a-Si
films produced by conventional hetero chemical vapour
deposition (HECVD) systems using silane or higher molecular weight silanes exhibit no silicon-hydrogen
vibrations, indicating less than 0.5% hydrogen incorporation [l]. Consequently these films contain a large
number of recombination centres (deep states) due to
dangling bonds. On the other hand, homo chemical
vapour deposition (HOCVP) and plasma techniques provide films containing substantial quantities of hydrogen
[4]. In this paper, we report on the characterization of aSi films produced from di-iodosilanes and tri-iodosilanes
by a HECVD process in a flowing argon atmosphere.
Because of the non-explosive nature of the iodosilane/
air mixtures and the absence of gas phase nucleation of
chemical intermediates [ 5 ] ,iodosilane precursorsmay be
more advantageous precursors than silanes. The synthesisof these materials and the deposition technique are
discussed elsewhere [6]; hence, only a brief description
is presented in this paper. The properties of these films
prepared and treated under various conditions were
monitored using various techniques including
g To wham all correspondence should be addressed.
M)22-3727/91/061015
+ 07 $03.50 0 1991 IOP Publishing Ltd
ultraviolet/visible (UV/VIS), FTIR, electron spin resonance (ESR), Rutherford hack scattering (RBS) spectroscopies, scanning electron microscopy (SEM) and the
temperature dependence of the film resistivity.
2. Experimental procedure
Iodosilanes, which were synthesised as describe earlier
[6], were decomposed at the surface of substrates that
were heated in a specially designed double-wall quartz
reactor in a flowing argon atmosphere using inductive
heating [6]. Deposition on various substrates including
glass, graphite, KBr and a special ESR grade glass was
performed at 500, 550 and 600 "C using di-iodosilane
and 600°C using tri-iodosilanz. The film deposited on
glass was used for UV/VIS (Hewlett Packard, model
8451A) and resistivity (in vacuum) studies. Films
deposited on graphite, KBr and ESR grade glass were
used, respectively, for RBS, FTIR (Bomem, model MB
100) and ESR (Varian E-12) measurements. ESR
measurements were carried out at room temperature
with a microwave power of about 40mW and modulation frequency of 100 kHz. The g value of unpaired
spins in the film was determined by comparison to the
signal (g = 2.0037) of DPPH free radical (a,"-diphenylB-picryl hydroxyl). For RBS studies, a beam of 2 MeV
helium nuclei was used to analyse films deposited on
graphite substrates. This technique was used to determine the relative amounts of silicon, oxygen and iodine
in the film. The presence of hydrogen and oxygen was
1015
G Tamizhmani et a/
determined qualitatively by R I R . Analysis for iodine by
this technique was not possible because the vibrational
frequency range of the Si-I vibration falls beyond the
window limit of KBr. The surface morphology of these
films was observed with a Hitachi S-570 scanning electron microscope. These characterizations were carried
out at room temperature.
Samples were also annealed inside a stainless steel
reactor [7] at 350 "C for 40 h in a hydrogen atmosphere
(40 psi), and subsequently characterized by FTIR, ESR
and UV/VIS. These measurements were repeated after
a second annealing for 65 h in hydrogen. Before and
after these treatments, resistivity against temperature
measurements were carried out in vacuum as described
earlier [7]. Temperatare regulation was accomplished
by a flow of compressed air or nitrogen gas, which
passed through an ice or liquid nitrogen bath and a
heating stage before reaching the sample holder. A
chromal-alumel thermocouple attached to the sample
holder provided feedback to the Eurotherm 808 temperature controller. For highly resistive samples
(deposited at 500°C using HzSiIz and 600°C using
HSiI,), the resistance was measured by a two-probe
method using colloidal graphite contacts [SI, which
were extended to copper wires embedded in epoxy.
These wires were connected to the instrumentation.
The resistance of less resistive films (deposited at
550°C from H,SiI,) was measured using a four-probe
method. Ail measurements were carried out in vacuum
to avoid leakage through adsorbed moisture and to
avoid heat transfer from the surroundings. For each
reading, dwell times of 5 and 10 min were maintained
for four-probe and two-probe measurements, respectively.
Attempts to deposit silicon by CVD on indium tin
oxide coated (ITO) glass using the iodosilanes resulted
in the removal of the ITO films before Si was deposited.
It was found that etching of the ITO could be avoided
by pyrolyzing droplets of di-iodosilane on the substrate
at 550 "C in an inert atmosphere. This method was also
adapted for the deposition of films on KBr and ESR
grade glass.
3. Results and discussion
3.1. Ultraviolet/visible absorption
The absence of long-range order and the presence of
dangling bonds of impurities in amorphous semiconductors create tail and gap states, respectively [9].
In amorphous silicon, optical absorption can involve
gap state to band, tail state to band as well as band to
band transitions. Because the band to band transition
can occur at a higher photon energy and with a higher
probability than other transitions, absorption coefficient values higher than lo3 or lo4"'
are often
used to determine the band gap using Tauc's plot [9].
The absorption coefficient CY as a function of photon
energy for the films deposited using H,Si12 at different
1016
hv leVl
Figure 1. Optical absorption edge for the films of a-Si
prepared by CVO using di-iodosilane and substrate
temperatures of (A) 500, (B) 550 and (C)600 "C.
15
1.1
19
2.1
hv la1
23
2.5
Figure 2. Tauc's plots for the films of a-Si prepared by CVD
process using di-iodosilane at the substrate temperatures
of (A) 500, (B) 550 and (C) 600 "C. The optical gap values
of these films were obtained by considering the absorption
coefficient (e)values in the order of 104cm-' and higher.
substrate temperatures (500, 550 and 600°C) is shown
in figure 1. Tauc's piots (figure 2j, using the data in
figure 1, gave band gap values that are similar for the
three films: 1.77eV (550°C); 1.71eV (500°C); and
1.68 eV (600 "C). The curvature in these plots is similar
to that observed for films prepared by other methods.
It is interesting that the KIR spectra (discussed below)
of these films indicated incorporation of hydrogen only
for films prepared at 550 "C. Tauc's plot (figure 3) of
the film deposited from HSil, at 600"C showed a band
gap (1.68eV) similar to those found for the films
deposited from H,SiI,. Annealing at 350°C under a
pressurized hydrogen atmosphere did not have any
effect on the band gap of these materials. Although
ESR data (discussed below) indicate that this annealing
treatment reduced the density of unpaired spins, incorporation of hydrogen does not appear to be the reason.
Instead, the effect may be due to an increase in degree
of crystallinity, which apparently was not sufficient to
affect the band gap.
Physical properties of thin film silicon
21
Figure 3.Tauc's plot for the film of aSi prepared by cvo
process using tri-iodosilane at the substrate temperature of
600°C.The optical gap value of this film was obtained by
considering the absorption coefficient (U)values in the
order of IO4cm-' and higher.
I
4000
3300
ZMIO
1900
1200
I
500
Wave number Icm-'l
Figure 4. Infrared transmission spectra of a-Si deposited at
the substrate temperatures of (A) 500, ( 8 ) 550 and (C)
600 "Cusing di-iodosilane, and of (D) 600 "C using triiodosilane. The spectra are characterized by the presence
of Si-H. Si-0 and Si-OH vibrational modes.
3.2.
IR
transmission
The FTIR spectra of films deposited at various substrate
temperatures are shown in figure 4. The absorption
bands for the films correspond to vibrational modes of
Si-H (2027cm-'), Si-OH (3435 and 1603cm-I) and
Si-0-Si (1069cm-I). Others have been able to estimate the amount of hydrogen incorporation in amorphous silicon [lo]. Unfortunately, such an estimation was
not possible for our films because our process roughened the surface of the KBr substrate and precluded
measurements of film thickness. The reason for the
variability in hydrogen content is obscure at the
present, and a much more detailed study will be
needed. In addition, annealing the films in a hydrogen
atmosphere at 350°C did not result in the growth of
the intensity of the Si-H signal for any of the films.
Consequently, there is no evidence that this procedure
can incorporate hydrogen after deposition. At any rate
the incorporation of hydrogen by our process using
di-iodosilane is in contrast to the case of conventional
HECVD systems using silane precursors, which fail to
provide hydrogen incorporation during deposition [l].
For those systems hydrogenation is accomplished in a
second step after film deposition using hydrogen
plasma [ l l ] or H' ion implantation [12]. The absence
of any detectable hydrogen incorporation in the films
deposited at 6 0 0 T using HSiI, is a clear indication
that the pyrolysis mechanism for this precursor differs
from the one for di-iodosilane. Some tentative mechanisms have been proposed 161.
The silicon/oxygen stretch occurred at a significantly lower frequency (3460 cm-l than that for free
silanol (Si-OH) (3690 cm-'). This lowering of frequency might be due to hydrogen bonding to neighbouring groups in the solid. The RBS data presented
below indicate that oxygen was present on the surface
of the film, and none was detected deep within the
film. Consequently it would appear that oxygen was
not incorporated in the film during deposition and that
the surface oxide resulted from exposure of the film to
air after deposition. The presence of iodine could not
be checked by FTIR because the silicon/iodine stretching frequency [I31 falls beyond the window limit of
KBr. Hence, the presence of iodine was confirmed
by the RBS studies presented below. Annealing in a
hydrogen atmosphere does not have any effect on IR
spectra, although it did cause a decrease in the ESR
signal intensity as discussed in the next section.
3.3. ESR
The amorphous nature of the film is confirmed by
x-ray data [61 as well as the ESR spectrum. Fuhs 191
showed that unpaired electrons due to neutral dangling
bonds in a-Si are responsible for the ESR spectrum at
a g value of 2.0055 with a peak-to-peak linewidth
(AHpp)of 7 G . Similarly the observed value of g (of
2.0057) from the ESR spectra shown in figure 5 for the
film deposited at 500 "C using HzSiIz and figure 6 for
the film deposited at 600°C using HSiI, suggests the
presence of a high density of dangling bonds. Furthermore, the linewidth values of 7 G and 4 G determined for thick and thin (<0.1 pm) films, respectively
(figures 5 and 6) were found to be consistent with
the observation that the linewidth decreases as film
thickness decreases [14]. The spin density is expected
to decrease on annealing the films in the temperature
range between 100 and 350 "C, for which the formation
of twofold coordinated Si atoms has been proposed to
be responsible [14,15]. The reduction of ESR signal
intensity (figures 5 and 6) on first (40h) and second
(65 h) annealings at 350°C in a hydrogen atmosphere
could be interpreted in terms of hydrogen incorporation of structural effects. The FTIR study gave no
evidence that this treatment caused an increase in
hydrogen content. Consequently annealing at this temperature presumably caused an increase in the degree
of crystallinity of a-Si or, alternatively, caused a
1017
G Tamizhmani et a/
I
I
Magnetic field IGI
Magnetic field I G )
4
Figure 5. ESR spectra of a-Si deposited at the substrate
temperature of 500°C by the decomposition of diiodosiiane. The spectra are characterized by the reduction
of strength of signal of as-deposited film (A) on the first
annealing, and (6)second annealing, (C) in hydrogen
atmosphere (40 psi) at 350 "C.
10 G
0
L
-
It is known that oxygen contamination reduces the
g value below 2.0055 (101. The observed g value of
our films along with the FLIR results suggests oxygen
contamination within the film is low. The brown or
grey colour of these films also suggests low oxygen
content in the samples [16]. This conclusion is supported by RBS results discussed below.
3.4.
B
Magnetic field (GI +
Figure 6. ESR spectra of a-Si deposited at the substrate
temperature of 600 "C by the decomposition of triiodosilane. The spectra are characterized by the reduction
of strength of signal of as-deposited film (A) on first
annealing, (B) in hydrogen atmosphere (40psi) at 350°C.
decrease in the dangling bond density in the manner
proposed earlier [15]. This conclusion is supported by
the decrease of spin density on first (40 h) and second
(65 h) annealing at 350°C under an argon atmosphere
(figure 7).
1018
-
Figure 7. ESR spectra of a-Si deposited at the substrate
temperature of 500 "C by the decomposition of diiodosilane. The spectra are characterized by the reduction
of strength of signal of as-deposited film (A) on first
annealing, and (6) second annealing, (C) in argon
atmosphere (40 psi) at 350 "C.
RES
analysis
The 2 MeV, 180" Rutherford backscattering spectrum
and the simulated ji7j KBS spectrum (smooth curve)
are shown in figure 8. The channel [energy) is indicated
along with the surface scattering energies for Si, 1 and
0 on the film surface. The experimental data are consistent with a film in which the most abundant element
is Si. Iodine is distributed throughout the Si film at a
concentration not exceeding 1.7%. It is clearly evident
from the RBS spectrum that iodine incorporation is not
uniform. The simulation, which is a refinement of that
reported earlier [6], indicates that the iodine concentration decreases as the film depth increases. Oxygen was also observed, but only in a surface layer. For
the simulation, it was assumed that this oxide layer
consisted of S i 0 2 plus iodine, and the areal thickness
molecules cm-'. The thinwas found to be 200 x
ness of this layer in relation to the overall film thickness
and the fact that it is confined to the surface indicates
that it most likely resulted from oxidation of the Surface
of the film upon exposure to air.
Physical properties of thin film silicon
10
Energy
12
IMeVl
14
16
18
J
,
a240 a.zi4
~~
'
a248 '
o.k.2
T4'
Channel
Figure 8. RES spectrum of a-Si deposited on graphite disc
by decomposition of di-iodosilane in cv0. The spectrum is
characterized by the presence of abundant silicon, a small
amount of oxygen at the surface and of iodine throughout
the bulk.
28
32
36
40
1WO/T IK-' 1
4L
48
I
Figure 9. Plots of Ino against reciprocal temperature for
the films deposited at (A) 500°C and (8)550°C using diiodosilane, and at (C) 600-Cusing tri-iodosilane. The data
show the hopping nature of conductivity.
3.5. Resistivity measurement
Adsorbed gas can cause band bending at the semiconductor surface (101. To avoid this complication, resistivity measurements were made after de-gassing at
90°C in vacuum. In the plots of In(u) against 1/T
(figure 9) (191, the deviation from linearity suggests a
hopping conduction [lo, 181. Variable range hopping
(VRH)at the Fermi level via localized states proposed
by Mott [lo, 191 is described by:
U
'
InQ)
0.'256 '
'
o
h
Figure 10. Plots of Ino against T-'I4 for the films
deposited at (A) 500 "C and ( 8 ) 550 "C using di-iodosilane,
and (C)at 600°C using tri-iodosilane. These results are
consistent with variable range hopping at the Fermi level.
In(o) against T11
(figure
4 10) is consistent with VRH at
the Fermi level. The density of states at the Fermi level
g(EF)deduced from the slope To of these plots is listed
in table 1 (using an assumed value of l/aw= 0.25 nm
[lo]). Upon annealing in hydrogen, the film deposited
at 500°C exhibited the same hopping conduction
(figure 11) with slightly higher resistivity at room temperature, which is consistent with the ESR data. On the
other hand, samples deposited at 550°C and 600°C
showed lower resistivity with extended-state conduction (figure 11) after annealing at 350°C in hydrogen atmosphere. The activation energies for the
extended-state conduction of these films are given in
table 1. This variability in response to preparation conditions indicates that more than one conduction mechanism is possible. A much more detailed study is
needed to ascertain the relation between preparation
conditions and conduction processes.
3.6.
SEM
Figure 12 shows the SEM micrographs of the films prepared at different substrate temperatures. The film prepared at 500 "C showed island-like structures with
microscopic voids. However, for the films prepared
at 550 and 600"C, no such resolvable structure was
observed and clearly indicates a different growth pattern compared to lower temperatures. Annealing in
hydrogen atmosphere at 350 "C did not have any distinguishable effect on surface morphology. As it is
known that the presence of the island structure would
result in electronically active defects [ZO], film deposition at 550 "C or above is probably preferred.
= uHe x p [ - ( ~ ~ / r ) l ' ~ ]
with
4. Summary
To = 2.law/[kg(E,)l
where l/awis the radius of the localized wavefunction;
k is Boltzmann's constant; and g(EF) is the density of
states a t the Fermi level. The linearity of the plots of
Thin-film amorphous silicon has been prepared by a
CVD process involving new precursors-di-iodosilanes
and tri-iodosilanes. The amorphous nature of the films
was confirmed by several techniques including: uv/vis
1019
G Tamizhmani et al
Table 1. Experimental values acquired from conductivitv studies,
As-deposited
PRT (Q cm)
pRr
To
dEF)
EA
Precursor
Decomposition
temperature
("C)
(Q cm)
(xIO'K)
(lO"eV-' c m 3 )
(ev)
H2SiI2
H,Si12
HSiI,
500
550
600
2.9x I O 6
371
1.4x I O 6
4.4x I O 6
3.8
6.2x IO5
2.5
2.2
4.8
6.3
7.1
3.3
-
2.8
3.2
36
Annealed
L.0
LL
48
lOOO/T IK-'i
Figure 11. Plots of Ino against reciprocal temperature for
annealed (350"C in hydrogen atmosphere) films, which
550 "Cusing diwere deposited at (A) 500 "Cand (6)
iodosilane, and at (C)600 "C using tri-iodosilane. The data
S~!QW +!e aciivaied naiure of conducii.ity !or ile iiims
deposited at 550 and 600 "C,and hopping nature of
conductivity for the films deposited at 500 "C.
Annealed
0.06
0.36
intensity after this treatment. However, a similar
decrease occurred when the films were annealed in Ar
under the same conditions. Consequently, these results
indicated that the decrease in dangling bond density is
probably not due to Si-H bond formation. Perhaps the
heat treatment facilitated Si-Si bond formation or the
formation of neutral divalent Si having paired spins.
The RES studies confirmed the FTIR results, which indicated the presence of oxygen, and demonstrated that
the oxygen occurred in a thin layer at the surface of
the film. This layer probably developed when the film
was exposed to air after deposition. The resistivity
against temperature studies of the as-deposited films
suggested the variable range hopping model for conduction although more detailed studies are needed to
understand the effect of preparation conditions on the
conduction process. Annealing in hydrogen atmosphere at 350°C changed the conduction process from
hopping to activated for the films deposited at 550 and
600°C. SEM micrographs suggest 3 dense microstructure for the films deposited at 550 and 600°C and
an island-structure for the film deposited at 500°C.
C
I
i
I
I
.I
,
L
.
i
I
i
SEM micrographs of a-Si films deposited using di-iodosilane at (a) 500°C.(b) 550°C and lriiodosilane at (c) 600°C.Film thickness is approximately 1 pm. These micrographs reveal an island structure for
the film deposited at 500 "Cand a dense microstructure for the films deposited at 550 and 600 "C.
Figure 12.
spectroscopy, which indicated a band gap around
1.7 eV; ESR spectroscopy, which indicated the presence
or dangling bonds; and x-ray diffraction, which indicated the absence of crystallinity o r crystallites that
were very small. The FTIK indicated Si-H, Si-0-Si, SiOH vibrational modes. The Si-H hand intensity was
not affected by annealing the film at 350°C in a hydrogen atmosphere. On the other hand the ESR signal,
which was assigned to dangling bonds decreased in
1020
Acknowledgment
This work was supported in part by a grant to MC and
RTO from thc Strategic Program of the Natural
Science and Engineering Research Council of Canada.
References
[ I ] Hirose M 1984 Chemical vepor deposition Semicond.
Semimern1.s 21 A 109: 1Y84 Hydrogeiiulcd
Physical properties of thin film silicon
Amorohous Silicon ed J 1 Pankove (Orlando. FL:
Acadimic)
121
. . Madan A. Ovshinskv S R and Benn E 1979 Phil. Max.
40 259
A K and Austin A E 1981 J . Elecfrochem.
.131. Aerawal
Soc. 128 2292
[4] Scott B A 1984 Homogeneous chemical vapor
deposition Semicond. Semimetals 21 A 123; 1984
Hydroyenared Amorphous Silicon ed J I Pankove
(Orlando, FL: Academic)
[5] Sharp K G 1988 Patent Eur. Pat. Appl. EP 262, 980
[6] Tamizhmani G, Cocivera M, Oakley R T and Belluz P
D P 1990 Ckem. Mafer. 2 473
[7] von Windheim J A and Cocivera M 1990 J . Phys. D . :
Appl. Pkys. 23 581
[SI Goodman N B 1982 Phil. Mag. B 45 407
[9] Fuhs W 1986 Electronic properties of plasma-deposited
semiconductor films Plasma Deposited Thin Films
ed J Mort and F Jansen (Florida: CRC) p 45
1101 Strauven H. Stesmans A . Winters J . SDinnewiin J and
Verbeke 0 B 1987 J . Appl. Pkys. 62 28361
.111. KaDlan
. D. Sol N , Velasco G and Thomas P A 1978
Appl. Phys. Lett. 33 440
[I21 Suzuki T, Hirose M and Osaka Y 1980 Japan. J . Appl.
Phys. Suppl. 19 (2) 91
[I31 Smith A L 1963 Spectrockim. Acta 19 849
[14] Thomas P A , Brodsky M H, Kaplan D and Lepine D
1978 Phys. Re". B IS 3059
1151 Voeet-Grote U. Kummerle W. Fisher R and Stuke J
1980 Phil. Mag. 5 41 127
1161
. . Kubler L. Jaeele A and Koulmann J J 1979 Phys
Status.' S o l k b 95 307
(171 Doolittle L R 1986 Nuclear Instrum. Merhods B 15 227
[I81 Mott N F 1969 Phil. Mag. 19 835
[19] Mott N F and Davis D A 1979 Non-crystalline
semiconductors Electronic Processes in NonCrystalline Materials (Oxford, Clarendon) p 199
[20] Knights J C and Lujan R A 1979 Appl. Phys. Lett. 35
244
.
.
I
1
I
102t