Influence of annealing and Al
2
O
3
properties on the hydrogen-induced passivation of
the Si/SiO
2
interface
G. Dingemans, F. Einsele, W. Beyer, M. C. M. van de Sanden, and W. M. M. Kessels
Citation: Journal of Applied Physics 111, 093713 (2012);
View online: https://doi.org/10.1063/1.4709729
View Table of Contents: http://aip.scitation.org/toc/jap/111/9
Published by the American Institute of Physics
Articles you may be interested in
Status and prospects of Al
2
O
3
-based surface passivation schemes for silicon solar cells
Journal of Vacuum Science & Technology A: Vacuum, Surfaces, and Films 30, 040802 (2012);
10.1116/1.4728205
Hydrogen induced passivation of Si interfaces by films and stacks
Applied Physics Letters 97, 152106 (2010); 10.1063/1.3497014
On the surface passivation mechanism by the negative-charge-dielectric
Journal of Applied Physics 104, 113703 (2008); 10.1063/1.3021091
Controlling the fixed charge and passivation properties of Si(100)/Al
2
O
3
interfaces using ultrathin SiO
2
interlayers synthesized by atomic layer deposition
Journal of Applied Physics 110, 093715 (2011); 10.1063/1.3658246
Ultralow surface recombination of substrates passivated by plasma-assisted atomic layer deposited
Applied Physics Letters 89, 042112 (2006); 10.1063/1.2240736
Surface chemistry of atomic layer deposition: A case study for the trimethylaluminum/water process
Journal of Applied Physics 97, 121301 (2005); 10.1063/1.1940727
Influence of annealing and Al
2
O
3
properties on the hydrogen-induced
passivation of the Si/SiO
2
interface
G. Dingemans,
1
F. Einsele,
2,a)
W. Beyer,
2
M. C. M. van de Sanden,
1
and W. M. M. Kessels
1,b)
1
Department of Applied Physics, Eindhoven University of Technology, P.O. Box 513, 5600 MB Eindhoven,
The Netherlands
2
IEK-5, Forschungszentrum Juelich GmbH, 52425 Juelich, Germany
(Received 5 January 2012; accepted 29 March 2012; published online 7 May 2012)
Annealing at moderate temperatures is required to activate the silicon surface passivation by Al
2
O
3
thin films while also the thermal stability at higher temperatures is important when Al
2
O
3
is
implemented in solar cells with screenprinted metallization. In this paper, the relationship between
the microstructure of the Al
2
O
3
film, hydrogen diffusion, and defect passivation is explored in detail
for a wide range of annealing temperatures. The chemical passivation was studied using stacks of
thermally-grown SiO
2
and Al
2
O
3
synthesized by atomic layer deposition. Thermal effusion
measurements of hydrogen and implanted He and Ne atoms were used to elucidate the role of
hydrogen during annealing. We show that the passivation properties were strongly dependent on the
annealing temperature and time and were significantly influenced by the Al
2
O
3
microstructure. The
latter was tailored by variation of the deposition temperature (T
dep
¼ 50
C–400
C) with hydrogen
concentration [H] between 1 and 13 at.% and mass density q
mass
between 2.7 and 3.2 g/cm
3
.In
contrast to films with intermediate material properties, the passivation by low- and high density
films showed a reduced thermal stability at relatively high annealing temperatures (600
C). These
observations proved to be in good agreement with thermal effusion results of hydrogen and inert gas
atoms that were also strongly dependent on film microstructure. We demonstrate that the
temperature of maximum effusion decreased for films with progressively lower density (i.e., with
increasing [H]). Therefore, the reduced thermal stability of the passivation for low-density
hydrogen-rich ([H] >5 at. %) films can be attributed to a loss of hydrogen at relatively low
annealing temperatures. In contrast, the lower initial [H] for dense Al
2
O
3
films can likely explain
the lower thermal stability associated with these films. The effusion measurements also allowed us
to discuss the role of molecular- and atomic hydrogen during annealing.
V
C
2012 American Institute
of Physics.[http://dx.doi.org/10.1063/1.4709729]
I. INTRODUCTION
Recombination of charge carriers at silicon surfaces can
be a major contributor to suboptimal energy conversion effi-
ciencies of silicon solar cells. In recent years, aluminum
oxide (Al
2
O
3
) thin films, mainly synthesized by atomic layer
deposition (ALD) and plasma-enhanced chemical vapour
deposition (PECVD), have been used to significantly
improve solar cell efficiencies by providing effective surface
passivation.
1–11
On both n- and p-type Si surfaces, Al
2
O
3
induces surface recombination velocities, S
eff
, typically well
below 5 cm/s. The effective passivation is related to a sign ifi-
cant reduction in defect density at the Si interface (D
it
10
11
eV
1
cm
2
), after annealing the Al
2
O
3
films at rela-
tively low temperatures (400
C).
12
Simultaneously, the
fixed negative charge density associated with the Al
2
O
3
films
increases up to Q
f
¼(3–10) 10
12
cm
2
leading to field-
effect passivation through a reduction of the electron density
at the Si surface.
5,12
Recently, Al
2
O
3
films have also been applied as capping
layers on SiO
2
.
13–16
It was found that the advantageous
effect of the capping layer was chiefly related to the chemi-
cal passivation at the remote interface. Most notably, for
stacks comprising SiO
2
synthesized at low-temperatures by
PECVD and ALD, very low S
eff
values < 3 cm/s and low
D
it
values < 10
11
cm
2
eV
1
were obtained,
13,15
which
remained stable up to high annealing temperatures.
13
The
low defect densities were explained by an effective hydro-
genation of defects present at the buried Si/SiO
2
interface
under influence of the Al
2
O
3
capping layer. Secondary-ion-
mass-spectroscopy (SIMS) demonstrated that a fraction of
the hydrogen incorporated in the Al
2
O
3
during deposition
penetrated into SiO
2
and diffused toward the Si/SiO
2
inter-
face during annealing.
14
At this interface, the Si-dangling
bond is the most prominent electrically active defect
(P
b
-type defect).
17–20
As the dangling bond is chemically
active, reactions with hydrogen can lead to the elimination
of these defects. On the other hand, the fixed charge density
responsible for field-effect passivation was found to be very
small for thick (e.g., 50–200 nm) thermally-grown SiO
2
layers capped by Al
2
O
3
.
14,16,21,22
These stacks can therefore
be regarded as a model system for studying the chemical
a)
Present address: Robert Bosch GmbH, Stuttgart, Germany.
b)
Author to whom correspondence should be addressed. Electronic mail:
w.m.m.kessels@tue.nl.
0021-8979/2012/111(9)/093713/9/$30.00
V
C
2012 American Institute of Physics111, 093713-1
JOURNAL OF APPLIED PHYSICS 111, 093713 (2012)
passivation induced by Al
2
O
3
.
14
Moreover, when Al
2
O
3
is
deposited on H-terminated Si, an interfacial SiO
x
is formed.
1
The corresponding electronic interface properties were found
to be comparable to other Si/SiO
2
interfaces.
23
Given the important role of hydrogen in the passivation
properties of Al
2
O
3
, it is vital to explore a number of open
questions related to the improvement in chemical passivation
during annealin g. For example, what is the effect of a varia-
tion in annealing temperature and time? What is the influ-
ence of the structural properties of Al
2
O
3
, i.e., mass density
and hydrogen concentration, on the diffusion of hydrogen?
In what form, atomic or molecular, is hydrogen transported
in the Al
2
O
3
films? In this paper, these questions will be
addressed by the combination of lifetime spectroscopy meas-
urements on SiO
2
/Al
2
O
3
stacks and thermal effusion
measurements.
Thermal effusion measurements have proven to be
powerful for studying the material microstructure and hydro-
gen diffusion in thin films by the detection of volatile species
that are released from a film during annealing (T ¼ 200
–1100
C).
24–26
Only recently, effusion measurements have
been applied to study the properties of passivation
materials.
6,14,27–29
In the context of passivation, effusion
measurements are of primary interest as they help to eluci-
date the important role of hydrogen. It is important to
emphasize that the temperature range during the effusion
experiments encompasses the activation of the surface passi-
vation, which generally requires annealing at T ¼350
–450
C, and also the higher temperature range (>700
C)
typically used during the metallization processes of solar
cells. For example, during the manufacture of industrial
solar cells with screenprinted contacts, high temperatures
(T > 800
C) are typically required and the thermal stabil-
ity of the passivation film is crucial.
6,30,31
Effusion experi-
ments have, for instance, been applied to study the relation
between hydrogen effusion and film composition for
a-SiN
x
:H surface passivation films.
27
In addition, hydrogen
effusion from thermally grown SiO
2
exhibited maxima at
distinct annealing temperatures, which corresponded with a
deterioration in surface passivation quality.
14
For Al
2
O
3
, the
effusion of molecules such as H
2
O, H
2
, CO, and CO
2
from
the film bulk and/or interfaces has been observed in a wide
temperature range.
6
In this paper, both the effusion of intrin-
sic hydrogen species (i.e., originating from H incorporated
during deposition) and inert gas atoms which were implanted
in Al
2
O
3
will be studied.
In this paper, it will be shown that: (i) The hydrogena-
tion of the Si/SiO
2
interface by Al
2
O
3
exhibited an Arrhenius
behavior as the associated reaction rate increased with
increasing annealing temperature. The results suggested that
the passivation mechanism of the stacks was not diffusion-
limited at annealing temperatures of 400
C. (ii) The effu-
sion of hydrogen was strongly influenced by the Al
2
O
3
structural properties and the annealing temperature. In addi-
tion, the level of passivation achieved at high annealing
temperatures was reduced for low-density hydrogen-rich
films synthesized at low T
dep
(100
C) and for dense
films synthesized at high T
dep
(>300
C). The cause for
the lower thermal stability can be related to the hydrogen
transport toward the interface. For the relatively hydrogen-
rich films, rapid effusion takes place toward the ambient
already at low annealing T, while for dense Al
2
O
3
films, a
low initial hydrogen density may reduce the effectiveness of
interface hydrogenation. (iii) The measurements suggested
that hydrogen can diffuse in molecular as well as atomic
form, depending on annealing temperature and Al
2
O
3
struc-
tural properties.
After the experimental details in Sec. II, the annealing
kinetics of the SiO
2
/Al
2
O
3
stacks will be discussed in
Sec. III A. Subsequently, the effect of the Al
2
O
3
micro-
structure on the passivation properties will be reported in
Sec. III B. The effusion measurements are presented in
Sec. III C. The discussion in Sec. IV aims to combine the
experimental results to enable a better understanding of the
role of hydrogen in the passivation properties of Al
2
O
3
.
II. EXPERIMENTAL DETAILS
The Al
2
O
3
films were synthesized by thermal and
plasma ALD in an Oxford instruments OpAL reactor.
8
Al(CH
3
)
3
served as Al precursor, and either H
2
O (thermal
ALD) or a remote O
2
plasma (plasma ALD) was employed
during the oxidation step. The substrate temperature during
deposition T
dep
was varied between 50
C and 400
C
(plasma ALD), to obtain films with different hydrogen con-
tent and O/Al ratio.
32
Note that the actual wafer temperature
for the films deposited at 300 and 400
C may be somewhat
lower than the substrate temperature. The Al
2
O
3
atomic
composition was dete rmined by Rutherford backscattering
spectroscopy (RBS) and elastic-recoil detection (ERD).
High-quality SiO
2
layers with a thickness of 190 nm were
grown by (wet-) thermal oxidation in an H
2
O atmosphere at
a temperature of 900
C. As substrates, floatzone n-type
Si(100) wafers were used, with a resistivity of 2or12 X
cm. After oxidation, no cleaning step was performed prior to
Al
2
O
3
deposition. To activate the passivation, rap id thermal
annealing with a typical ramp-up time of 30
C/s was per-
formed in N
2
atmosphere for the SiO
2
/Al
2
O
3
stacks or in
forming gas (10% H
2
in N
2
) for single layer SiO
2
reference
samples.
Thermal effusion measurements were performed on HF-
last Si samples coated on both sides with 100 nm Al
2
O
3
films. The experiments took place in an evacuated quartz
tube under high vacuum conditions (10
7
mbar) and a linear
heating rate of 20
C/min was used. A quadrupole mass spec-
trometer was used for the detection of volatile species
released from the Al
2
O
3
films. Therefore, different masses
are detected sequentially at slightly different times and tem-
peratures with a resolution of approximately 20
C. He and
Ne atoms were incorporated in the films by ion implantation.
Prior to the measurement, the setup was calibrated by meas-
uring known gas flows of H
2
, He, and Ne. More details on
the effusion- and implantation experiments can be found
elsewhere.
24,33
The passivation properties of the stacks were
evaluated by measuring the effective minority carrier life-
time s
eff
by the photoconductance decay method (Sinton
WCT 100). For each sample, the lifetime was measured for a
spot in the middle of the wafer. The upper limit of the
093713-2 Dingemans et al. J. Appl. Phys. 111, 093713 (2012)
surface recombination velocity was calculated from s
eff
and
is quoted at an injection level of 5 10
14
cm
3
by assuming
an infinite bulk lifetime, i.e., by employing the relation S
eff
< W/(2s
eff
), with W the wafer thickness of 200 lm.
III. RESULTS
A. Effect of annealing time and temperature on
passivation properties
The injection-level dependent effective lifetime of a
SiO
2
/Al
2
O
3
stack after annealing at a temperature of 400
C
is shown in Fig. 1. The SiO
2
/Al
2
O
3
stack resulted in a higher
effective lifetime of s
eff
¼ 2.8 ms at Dn ¼ 5 10
14
cm
3
(2 X cm n-type Si) than obtained for a SiO
2
reference sam-
ple annealed in forming gas with s
eff
¼ 1.4 ms. The effective
lifetime induced by the stack corresponds to a value of S
eff
< 4 cm/s. In addition, we verified (on different type of
wafers) that the passivation performanc e of the stack was
similar to that obtained after “alnealing” (annealing the SiO
2
with a sacrificial Al-layer), which is known to lead to an
excellent passivation quality.
18,34,35
It was, fu rthermore,
found that annealing in N
2
led to the same passivation per-
formance for the stacks as annealing in forming gas. The
observation that the shape of the effective lifetime curves is
the same for forming-gas annealed SiO
2
and SiO
2
/Al
2
O
3
stacks is in accordance with a similar prevailing mechanism
underlying the passivation properties, i.e., a high level of
chemical passivation.
13,16
The negative Q
f
associated with
the Al
2
O
3
films was found to decrease strongly with SiO
2
interlayer thickness.
15,16
For the thick SiO
2
interlayers of
200 nm used here, the field-effect passivation was found to
be insignificant.
14,21
Similar results have recently been
reported for SiO
2
/a-SiN
x
:H stacks.
36
It was verified that
plasma and thermal ALD Al
2
O
3
capping layers resulted in a
comparable high level of passivation after annealing (see
also Fig. 2). However, prior to annealing, the plasma ALD
process led to a significant degradation of the interface
quality of the as-grown thermal SiO
2
, resulting in effective
lifetimes in the microsecond range. This degradation has
been attributed to the vacuum UV (VUV) radiation present
in the O
2
plasma, which does not play a role during thermal
ALD.
12,37,38
1. Annealing kinetics
To investigate the kinetics associated with the passiva-
tion of the SiO
2
/Al
2
O
3
stacks during annealing, the annealing
time and temperature T were varied. Al
2
O
3
capping layers
synthesized by plasma and thermal ALD are compared in
Figure 2. An important observation was that the improve-
ment in the effective lifetime occurred faster for progres-
sively higher annealing temperatures (i.e., Arrhenius
behavior). In addition, the level at which the passivation
reached a maximum was controlled by the annealing temper-
ature. The highest level of passivation was found for
T ¼400
C, independent of the ALD method applied.
However, the annealing time required to reach saturation
was longer for Al
2
O
3
capping layers synthesized by thermal
ALD than by plasma ALD. This difference may be related to
a slight difference in structural properties resulting from the
two ALD methods.
32
In addition, it is likely that the incorpo-
ration of hydrogen into SiO
2
during plasma ALD, as was
corroborated by secondary ion mass spectrometry measure-
ments,
14
can also play a role in the observed differences. The
comparatively short annealing times associated with plasma
ALD capping layers may therefore be partially attributed to
the larger hydrogen reservoir in the SiO
2
layer.
Although the passivation performance after annealing at
450
C was slightly below that obtained at 400
C, a signifi-
cant deterioration of the passivation properties occurr ed only
at higher temperatures of 600–700
C,
14
as will also be dis-
cussed later. The degradation depended strongly on the
duration of the annealing step. The decrease in surface
FIG. 1. Injection-level-dependent effective lifetime for a SiO
2
/Al
2
O
3
stack
after annealing at 400
C (10 min, N
2
). As references, data for as-grown
SiO
2
and after subsequent forming gas annealing (400
C, N
2
/H
2
, 10 min)
are shown. The film thickness was 190 and 30 nm for SiO
2
and Al
2
O
3
,
respectively. Al
2
O
3
was deposited at T
dep
¼ 200
C by thermal ALD. As
substrates, 2 X cm n-type c-Si wafers were used.
FIG. 2. Effective lifetime (Dn ¼ 5 10
14
cm
3
) as a function of annealing
time for SiO
2
/Al
2
O
3
stacks. The Al
2
O
3
capping films were synthesized by
(a) plasma ALD and (b) thermal ALD, using T
dep
¼ 200
C. The film thick-
ness was 190 and 30 nm for SiO
2
and Al
2
O
3
, respectively. As substrates,
12 X cm n-type c-Si wafers were used. Annealing took place in N
2
. Note the
different scale of the horizontal axis in (a) and (b). The lines are exponential
fits to the data (see Eq. (1)). For every annealing temperature, a separate
sample was used, which was annealed and measured in consecutive steps.
093713-3 Dingemans et al. J. Appl. Phys. 111, 093713 (2012)
passivation quality at elevated T is indicative of an increase
in the interface defect density.
14,20,39
2. Activation energy
To obtain an estimate of the activation energy associated
with the passivation kinetics of the stacks, the trends in
Fig. 2 were fitted following the approach in Ref. 40 by the
expression:
s
eff
¼½b þ a expðt
anneal
=s
pass
Þ
1
; (1)
with t
anneal
the annealing time. Subsequently, the characteris-
tic time constant s
pass
was plo tted against 1/k
B
T to obtain an
estimate of the activatio n energy E
A
using the equation
s
1
pass
¼ c expðE
A
=k
B
TÞ (2)
From the Arrhenius plot in Fig. 3, a value of
E
A
¼ 0.9 6 0.2 eV was determined for plasma ALD capping
layers. For thermal ALD, the fitting was less accurate and
resulted in E
A
¼ 1.2 6 0.5 eV. E
A
can be regarded as the
effective activation energy representative for the processes
that lead to the hydrogenation of the interface during
annealing.
In the literature, a wide variety of activation energies are
reported for the diffusion of hydrogen in SiO
2
, but the gen-
eral understanding is that atomic H and H
2
can rapidly
migrate through SiO
2
during annealing.
19,41
For instance,
values of E
A
¼ 0.2–0.3 eV have been reported by Tuttle et al.
and Burte et al. for atomic H diffusion,
43,44
while Reed et al.
obtained a higher value of 0.75 eV.
18
For the diffusion of H
2
,
a value of 0.5 eV was reported by Fink et al.
41
It is gener-
ally assumed that the hydrogenation of electronically active
defects at the Si/SiO
2
interface is a reaction-limited process
rather than diffusion-limited.
18,20,42
The activation energies
that we have obtained are higher than the reported activation
energies for hydrogen diffusion and, therefore, also suggest
that the passivation kinetics of the stacks were not diffusion-
limited. Furthermore, the values of E
A
fall in the range of
earlier reported activation energies during alnealing
(1.2 eV) (Ref. 18) and forming gas annealing (1.5 eV).
20
B. Effect of Al
2
O
3
structural properties on the
passivation
To study the influence of the Al
2
O
3
structural properties
on the hydrogenation of the Si/SiO
2
interface, Al
2
O
3
films
were synthesized by plasma ALD at various substrate tem-
peratures (T
dep
¼ 50–400
C).
32
Figure 4 shows that the
hydrogen content in Al
2
O
3
decreased with increasing T
dep
from approximately 13 at. % (50
C) to <1 at. %
(400
C). Infrared absorption measurements indicated that
hydrogen is mostly incorporated as OH groups.
45
For films
deposited at T
dep
200
C, the O/Al ratio approached the
value of 1.5 representative for stoichiometric Al
2
O
3
.
32
For
lower T
dep
, the O/Al ratio was observed to increase, which
can be partially ascribed to the incorporation of more OH
groups. The mass density exhibited an opposite trend to [H]
and increased for films deposited at higher T
dep
. Although
less pronounced, an increase was also observed for the total
number density of atoms. The observed correlation between
the mass density and the hydrogen content shows that both
variables cannot be controlled independently by varying the
substrate temperature.
The effect of T
dep
on the passivation quality and the
thermal stability of the SiO
2
/Al
2
O
3
stacks is shown in Figure
5. The level of passivation in the as-deposited state improved
with increasing T
dep
. This effect, albeit less pronounced, has
also been reported for as-deposited single-layer Al
2
O
3
.
32
We
attribute the improvement to “in situ annealing” (i.e., anneal-
ing during deposition) which becomes more effective at
higher T
dep
. During the subsequent post-deposition annealing
at 400
C, the passivation performance of all the stacks
improved to values of S
eff
< 10 cm/s, but the impact of
annealing was especially significant for T
dep
< 300
C. The
optimal passivation was reached for T
dep
in the range of
100–200
C, leading to S
eff
values < 3 cm/s. However, the
optimum in T
dep
shifted to intermediate values of
200–300
C, when the annealing temperature was increased
up to 600
C. The passivation performance of the stacks
comprising Al
2
O
3
deposited at low T
dep
of 50
C and 100
C
degraded significantly after annealing at 600
C. These films
exhibited a high hydrogen concentration >8 at. % and
FIG. 3. Arrhenius plot corresponding to Fig. 2(a). s
pass
is the time constant
corresponding to the increase in the effective lifetime during annealing at
temperatures T. An activation energy of E
A
¼ 0.9 6 0.2 eV was deduced for
the passivation kinetics during annealing (T ¼ 320-450
C).
FIG. 4. Mass density q
mass
and hydrogen concentration [H] as a function of
T
dep
for plasma ALD Al
2
O
3
films. Data were extracted from RBS and ERD
experiments and thickness information from spectroscopic ellipsometry
093713-4 Dingemans et al. J. Appl. Phys. 111, 093713 (2012)
![FIG. 4. Mass density qmass and hydrogen concentration [H] as a function of Tdep for plasma ALD Al2O3 films. Data were extracted from RBS and ERD experiments and thickness information from spectroscopic ellipsometry](/figures/fig-4-mass-density-qmass-and-hydrogen-concentration-h-as-a-1oenfokb.webp)
