Nature of doped a-Si:H/c-Si interface recombination
Stefaan De Wolf
1,a兲
and Michio Kondo
2
1
Ecole Polytechnique Fédérale de Lausanne (EPFL), Photovoltaics and Thin Film Electronics Laboratory,
Breguet 2, CH-2000 Neuchâtel, Switzerland
2
National Institute of Advanced Industrial Science and Technology (AIST), Central 2, 1-1-1 Umezono,
Tsukuba, Ibaraki 305-8568, Japan
共Received 10 December 2008; accepted 8 April 2009; published online 26 May 2009兲
Doped hydrogenated amorphous silicon 共a-Si:H兲 films of only a few nanometer thin find
application in a-Si: H/crystalline silicon heterojunction solar cells. Although such films may yield a
field effect at the interface, their electronic passivation properties are often found to be inferior,
compared to those of their intrinsic counterparts. In this article, based on H
2
effusion experiments,
the authors argue that this phenomenon is caused by Fermi energy dependent Si–H bond rupture in
the a-Si:H films, for either type of doping. This results in the creation of Si dangling bonds,
counteracting intentional doping of the a-Si:H matrix, and lowering the passivation quality. © 2009
American Institute of Physics. 关DOI: 10.1063/1.3129578兴
I. INTRODUCTION
Semiconducting heterostructures increasingly attract at-
tention for electronic junction formation in crystalline silicon
共c-Si兲 wafer based solar cells. A key point of such a device is
the displacement of highly recombination-active 共Ohmic兲
contacts from the silicon surface by insertion of a film with
wide bandgap. To reach the full device potential, the hetero-
interface state density should be minimal.
1
Practically, hy-
drogenated amorphous silicon 共a-Si: H兲 films are appealing
candidates for this: Their bandgap is wider than that of c-Si
and, when intrinsic, such films can reduce the c-Si surface
state density by hydrogenation.
2
In addition, these films can
be doped relatively easily, either n-orp-type, allowing for
the fabrication of electronically abrupt p-n and low-high het-
erojunctions 共HJ兲.
3
Doping of these films may be expected to
yield a built-in electrical field, repelling either electrons or
holes from the surface states. In principle, this could sup-
press the a-Si:H / c-Si interface recombination further, in a
similar way as, e.g., in back-surface-field homojunction solar
cells.
4
Experimentally, however, such layers have been found
to result sometimes in poorer electronic passivation of c-Si
surfaces than their intrinsic counterparts.
5–7
For this reason,
typically, a few nanometer thin intrinsic buffer layer is in-
serted between the c-Si surface and the doped a-Si:H films
for device fabrication.
8
For HJ solar cells featuring such
stacked film structures, impressive large area 共⬎100 cm
2
兲
energy conversion efficiencies 共⬎22%兲 have been
reported.
9,10
Despite this result, the fundamental origin of the
poor passivation of the doped a-Si:H/ c-Si interface is not
yet fully understood.
In this article, we propose that the dependency of the
electronic surface passivation on the a-Si:H film doping is
linked to Fermi energy 共E
F
兲 dependent Si–H bond rupture in
such films. The latter phenomenon is attributed to the posi-
tion of E
F
within the bandgap, influencing the generation of
共native兲 compensation defects in the semiconductor, counter-
acting intentional doping.
II. EXPERIMENTAL
For the experiments, 320
m thick 0.7 ⍀ cm
phosphorus-doped high quality float zone 共100兲共FZ兲-Si wa-
fers were used. Both substrate surfaces were mirror polished
to eliminate the influence of substrate surface roughness on
the passivation properties.
11
Prior to deposition, the samples
were immersed in a piranha solution 共H
2
SO
4
:H
2
O
2
兲共4:1兲 for
10 min to grow a chemical oxide, followed by rinsing in
de-ionized water. The oxide was then stripped off in a dilute
HF solution 共5%兲 for 30 s. To avoid cross contamination, a
clustered multichamber parallel plate plasma enhanced
chemical vapor deposition 共PECVD兲 system, consisting
of separate chambers for a-Si: H共i兲, a-Si:H共n
+
兲, and
a-Si: H共p
+
兲 layer deposition was used. After transfer of the
samples to the relevant deposition chambers and mounting at
the top electrodes, the wafer surfaces were exposed to a 200
SCCM 共SCCM denotes cubic centimeter per minute at STP兲
H
2
flow for 20 min at a pressure of 0.5 Torr for temperature
stabilization of the samples. The spacing between electrode
and sample was 20 mm when depositing intrinsic films,
whereas for doped layer depositions this was 22 mm. During
film deposition, all chambers were operated at radio fre-
quency 共13.56 MHz兲 power. For soft film deposition, the
used power was consistently the minimum required to main-
tain a stable plasma. The B
2
H
6
and PH
3
concentrations were
4660 and 1550 ppm in H
2
, respectively. No additional H
2
dilution was used during deposition. To avoid epitaxial
growth during deposition, the intrinsic and p
+
-films were de-
posited at 155 °C. Due to the relatively higher H
2
dilution
factor for the n
+
-films, their deposition temperature T
depo
was
reduced to 120 °C, for the same reason. All deposition con-
ditions are summarized in Table I, unless otherwise stated.
To evaluate the surface passivation quality of the respec-
tive films, identical structures were deposited on both wafer
a兲
Electronic mail: stefaan.dewolf@epfl.ch.
JOURNAL OF APPLIED PHYSICS 105, 103707 共2009兲
0021-8979/2009/105共10兲/103707/6/$25.00 © 2009 American Institute of Physics105, 103707-1
Author complimentary copy. Redistribution subject to AIP license or copyright, see http://jap.aip.org/jap/copyright.jsp
surfaces. For the stacked doped structures, first the intrinsic
buffer layer was deposited on both sides of the samples prior
to doped layer deposition, to avoid cross contamination of
the interfaces. Before each intrinsic deposition, the samples
were immersed in a dilute 共5%兲 HF solution for 30 s. Post-
deposition annealing offers in a straightforward way a single
parameter to vary both electronic and material properties of
the samples under study. For this, following deposition, the
samples were consecutively subjected to stepwise isochronal
annealing treatments in a vacuum furnace 共20 ° C increment
per step of 30 min, with annealing temperatures 共T
ann
step
兲 rang-
ing from 120 ° C to 260 ° C兲. In between these annealing
steps, the value for effective carrier lifetime 共
eff
兲 of the
samples was measured with a Sinton Consulting WCT-100
quasi-steady-state photoconductance 共QSSPC兲 system,
12
op-
erated in the so-called generalized mode. Unless otherwise
stated, all reported values for
eff
were evaluated at a con-
stant carrier injection density, ⌬n = ⌬p= 1.0⫻10
15
cm
−3
.
Since high quality 共FZ兲-Si wafers were used, the values for
eff
can be considered as a measure for the surface passiva-
tion quality. The deposited film thickness 共d
bulk
兲 as well as
other film properties were determined by measuring ellip-
sometry spectra 共
,⌬兲 using a variable angle Woollam
M-2000 rotating-compensator instrument. These data were
then fitted to a two-layer 共Tauc–Lorentz兲 model, taking the
50% void surface roughness thickness 共d
rough
兲 of the depos-
ited material into account.
13
For bulk characterization of the
films, thermal desorption spectroscopy 共TDS兲 measurements
were taken of 1 cm
2
samples. For this an ESCO EMD-
WA1000S system operated at ultrahigh vacuum 共⬍1.0
⫻10
−9
Torr兲 was used in which the samples are lamp-heated
up to 1000 ° C, with a linear temperature ramp of
20 K min
−1
. During the annealing, a Balzers AG QMG 421
quadrupole mass spectrometer was used to determine the H
2
effusion rate from the a-Si:H films.
III. RESULTS
Figure 1共a兲 shows the change in surface passivation
quality, expressed by
eff
, as a function of the described step-
wise annealing treatment for, respectively, a few nanometer
thin intrinsic, n
+
- and p
+
-doped single film a-Si:H/ c-Si het-
erostructures. Whereas annealing up to 260 °C has a benefi-
cial effect on the surface passivation properties of a-Si: H共i兲
films, the same treatment is seen to result immediately in
losses for the a-Si:H共p
+
兲 case. For a-Si: H共n
+
兲 films, initially
an improvement in passivation quality can be seen. Never-
theless, annealing above 220 °C results also here in losses.
Figure 1共b兲 shows how the optical bandgap 共E
G
opt
兲 of these
films, extracted from spectroscopic ellipsometry 共SE兲 mea-
surements, changes during the annealing treatment. Interest-
ingly, in both cases where the passivation degrades, it coin-
cides with a collapse of E
G
opt
. The latter phenomenon may
point at defect formation in the amorphous host.
Figure 2 shows for the a-Si : H共p
+
兲/ c-Si case the inverse
TABLE I. Film deposition conditions, unless otherwise stated
Parameter a-Si:H共i兲 a-Si:H共p
+
兲 a-Si: H共n
+
兲
Power density 共mW cm
−2
兲 12 36 36
Electrode distance 共mm兲 20 22 22
关SiH
4
兴共SCCM兲 20 10 10
关B
2
H
6
兴共SCCM兲 ¯ 30 ¯
关PH
3
兴共SCCM兲 ¯¯ 100
Pressure 共Torr兲 0.5 0.5 0.5
T
depo
共°C兲 155 155 120
t
depo
共s兲 108 60 150
FIG. 1. 共Color online兲共a兲 Influence of stepwise annealing on the interface
passivation quality of a few nanometer thin single film a-Si:H / c-Si struc-
ture, expressed by
eff
共at ⌬n =⌬p =1.0⫻10
15
cm
−3
兲 and extracted from
QSSPC measurements. No intrinsic buffer layers are present underneath the
doped films. Symmetric structures have been deposited on both wafer sur-
faces. 共b兲 Influence of stepwise annealing treatment on the optical bandgap
E
G
opt
of the same films as in 共a兲, extracted from SE measurements. Symbols
represent experimental data; the lines are guides for the eye.
FIG. 2. 共Color online兲 Inverse effective carrier lifetime 共corrected for Auger
recombination兲 as function of carrier injection density for an n-type FZ-Si
wafer bifacially passivated by a few nanometer thin a-Si:H共p
+
兲 films. The
respective curves show data following thermal annealing at given tempera-
tures. The dashed lines are asymptotic fits of the data to expression 共1兲.The
graph in the inset shows the extracted J
0e
values as function of the inverse
annealing temperature.
103707-2 S. De Wolf and M. Kondo J. Appl. Phys. 105, 103707 共2009兲
Author complimentary copy. Redistribution subject to AIP license or copyright, see http://jap.aip.org/jap/copyright.jsp
effective carrier-lifetime corrected for Auger-recombination,
eff
−1
−
Auger
−1
, as a function of the carrier-injection density,
throughout the annealing experiment. All data can be fitted
asymptotically to the expression of Kane and Swanson,
14
eff
−1
−
Auger
−1
=
SRH
−1
+2J
0e
共N
D
+ ⌬n兲
qn
i
2
W
, 共1兲
where
SRH
is the carrier lifetime limited by Shockley–Read–
Hall 共SRH兲 recombination in the wafer,
15,16
J
0e
is the emitter
saturation current, N
D
and n
i
are, respectively, the back-
ground doping and intrinsic carrier concentrations of the sub-
strate and q is the elementary charge. W is the wafer thick-
ness. This equation is valid for wafers featuring identical
junctions at both surfaces. Here, it suggests that the surface
passivation by the doped films is provided by a field effect
rather than by chemical surface state passivation. Since high
quality substrates were used, the SRH term may be ne-
glected. Consequently, the fitting lines cut the abscissa at
about −N
D
. The figure shows that the slope of the curves
共which is proportional to the value of J
0e
兲 increases signifi-
cantly with annealing 共here with an activation energy, E
A
,of
about 0.7 eV兲. This may point at a reducing field effect,
likely due to defect generation, counteracting the doping.
It is worth noting that at low injection the carrier lifetime
increases remarkably, irrespective of the annealing condi-
tions. This could point at carrier trapping,
17
which is known
to occur in multicrystalline silicon,
18,19
but perhaps also at
c-Si surfaces. This phenomenon is however absent in the
a-Si: H共n
+
兲/ c-Si共n兲 high-low junction case 共not shown兲.
Hence, more likely, the steady-state photoconductance of the
a-Si: H共p
+
兲/ c-Si共n兲 HJ is dominated by depletion-region
modulation effects, at low injection.
20,21
Figure 3 shows how the presence of a few nanometer
thin intrinsic buffer layer underneath the doped films may
affect the surface passivation quality throughout the anneal-
ing experiment. Note that for all films of similar dopant type,
the deposition-times were exactly the same 共see Table I兲.Itis
hence assumed that the superposed thickness of intrinsic and
doped single-layers is equal to that of the corresponding
doped/intrinsic stacked structure. The graph shows how for
the a-Si:H共p
+
兲 case the presence of an intrinsic buffer layer
initially results in an improving passivation quality, although
at about 220 °C degradation sets in. Underneath a
a-Si: H共n
+
兲 layer, the presence of a similar buffer layer film
is even more benign. Here, the passivation quality improves
throughout the full annealing cycle.
Figure 4 shows H
2
effusion data for the samples dis-
played in Fig. 3. Panel 共a兲 and 共b兲, respectively, give data for
n
+
- and p
+
-type a-Si:H structures. In addition, also the su-
perposed signals of the single layers are displayed 共label ⌺兲.
Doping of such thin a-Si:H films profoundly influences the
H
2
effusion rate. Figure 4共b兲 shows that, compared to thin
intrinsic a-Si:H films, for p
+
films, H
2
effusion occurs at
significantly lower temperatures. Moreover, relatively more
H
2
effuses at lower temperatures from a
a-Si: H共p
+
兲/ a-Si:H共i兲 stack than for the superposed films
measured separately. Figure 4共a兲 gives TDS data for the
a-Si: H共n
+
兲 case. Also here, n
+
-type doping results in H
2
ef-
fusing at lower temperatures, although the effect is not as
drastic as for the p
+
-type case. The data for the superposed
intrinsic and n
+
-type films 共measured separately兲 seem now
to match well to that of the a -Si : H共n
+
兲/ a-Si:H共i兲 stack,
within experimental error.
IV. DISCUSSION
A. Intrinsic a-Si:H film passivation
For an atomically sharp a-Si:H共i兲 / c-Si interface, low-
temperature annealing improves the electronic interface
passivation.
22
This is confirmed in Fig. 1 for films of only a
few nanometer thick. Isothermal annealing of a-Si:H 共i兲/ c-Si
structures has been observed to yield stretched-exponential
recombination decay.
23
For bulk a-Si: H material, defect re-
duction following similar functionals was explained in the
FIG. 3. 共Color online兲 Influence of stepwise annealing treatment on the c-Si
surface passivation quality, expressed by
eff
共evaluated at ⌬n =⌬p =1.0
⫻10
15
cm
−3
兲, for doped a-Si: H stacks as shown in the inset sketches. Open
symbols represent doped single films, closed symbols represent stacks, fea-
turing an intrinsic buffer layer. Symmetric structures have been deposited on
both wafer surfaces. Results for as deposited material are indicated in the
abscissa by the label a.d. The data for the a-Si: H共p
+
兲 films are taken from
Ref. 34. Symbols represent experimental data; the lines are guides for the
eye.
FIG. 4. 共Color online兲 H
2
effusion data of the structures shown in Fig. 3.In
each panel, the total thickness of the stacked layers is equal to the sum of the
respective intrinsic and doped films. Panel 共a兲 represents data for n
+
-type,
whereas in 共b兲 the data are for p
+
-type a-Si :H films. The latter data have
been taken from Ref. 34.
103707-3 S. De Wolf and M. Kondo J. Appl. Phys. 105, 103707 共2009兲
Author complimentary copy. Redistribution subject to AIP license or copyright, see http://jap.aip.org/jap/copyright.jsp
past as arising either from dispersive 共i.e., time dependent兲
hydrogen diffusion
24
or from retrapping included hydrogen
motion.
25
For the a-Si:H共i兲 / c -Si interface, based on the lat-
ter interpretation, the passivation improvement has been at-
tributed to a transfer of hydrogen from a higher hydride state
in the a-Si:H film 共known to be dominant close to the
interface兲,
26
to a monohydride c-Si surface state.
23,27
Conse-
quently, the a-Si: H共 i兲 / c-Si interface passivation is likely due
to chemical surface state passivation, rather than due to a
field effect.
23
It is worth noting that for a few nanometer thin
a-Si: H共i兲 films, typically higher dangling bond densities are
measured, compared to their thicker counterparts.
28
This may
explain why also here the a-Si:H共i兲 / c-Si passivation quality
is slightly inferior compared to previously obtained results
with thicker 共50 nm兲 films, deposited under similar
conditions.
22
B. Doped a-Si:H film passivation
For thin film p-i-na-Si:H solar cells, the p-layer often
has been argued to limit device performance. On the one
hand, exposure of surfaces simultaneously to B
2
H
6
and SiH
4
共before or after PECVD兲 may give rise to CVD growth of
highly defective a-SiB
x
:H layers, even at very low tempera-
tures, resulting in poor p-i interfaces.
29,30
On the other hand,
recombination in the a-Si:H共p
+
兲 bulk itself was already for
the first p-i- n devices recognized to hamper device
performance.
31
Likely, this originates from doping induced
localized states in the film.
32
Here, as E
G
opt
is mainly affected by the bonded H content
共independent from the doping of the film兲,
33
its degradation
under annealing 关see Fig. 1共b兲兴 suggests doping dependent
Si–H rupture. The created defects may counteract the film
doping and reduce thus the field effect 共and interface passi-
vation, see Fig. 2兲. The link between doping dependent Si–H
rupture in the film and a-Si:H共p
+
兲/ c-Si interface recombina-
tion was recently established based on H
2
effusion experi-
ments as well.
34
The effect of n
+
-type doping appears to be less detrimen-
tal on the passivation properties, compared to that of p
+
-type
doping. Nevertheless, the study of equally thin phosphorous-
doped a-Si:H共n
+
兲 films by near-UV photoelectron constant
final state yield spectroscopy 共CFSYS兲
35,36
has led to even
more direct proof that also for a few nanometer thin films
increased doping leads to increasing defect densities.
7
For
n
+
-type doping, these defects hinder the displacement of E
F
toward the conduction band minimum 共CBM兲, where also
here such increased defect densities have been linked to en-
hanced recombination at the a-Si: H / c-Si interface.
7
H
2
effusing at lower temperatures from doped 共com-
pared to undoped or compensated兲 a-Si:H films has been
seen in the past as a proof that Si–H bond rupture 共and thus
defect generation兲 in such films depends on the position of
E
F
rather than on the physical nature of the present
dopants.
37,38
This argument was originally used to explain
doping dependent hydrogen diffusion phenomena in a-Si: H
films.
39
The hydrogen diffusion energy E
D
ⴱ
, defined by D
H
=D
0
ⴱ
exp共−E
D
ⴱ
/ kT兲,ina-Si: H and microcrystalline silicon
共
c-Si兲 is displayed as function of E
F
in Fig. 5. The data are
taken from Ref. 40. The diffusion coefficient, D
H
, describes
the motion of hydrogen in the silicon matrix, where D
0
ⴱ
=10
−3
cm
2
s
−1
is the theoretical diffusion coefficient unaf-
fected by traps,
41
k is the Boltzmann constant, and T is the
temperature. The diffusion activation energy E
D
ⴱ
equals
E
S
-
H
, where E
S
is the saddle point for interstitial H migra-
tion and
H
is the chemical potential of H atoms.
42,43
The
shown data imply that when E
F
is closer to the valence band
maximum 共VBM兲共i.e., p-type doping兲 this rapidly results in
decreasing values for E
D
ⴱ
共see crosshatched region with label
①兲. For n-type doping, E
F
should be brought relatively
closer to the CBM to yield a similar drop 共crosshatched re-
gion with label ③兲.
The data in Fig. 4 confirm H
2
effusion occurring at
higher temperatures for n- than for p-type a-Si:H films. This
phenomenon is reflected in the passivation results as well
共see Fig. 1兲. It is only at higher annealing temperatures that
losses start to occur for n-type films, compared to their
p-type counterparts. These trends suggest that a-Si: H / c -Si
interface recombination may be related to E
F
dependent
Si–H rupture in the film, irrespective whether this film is n-
or p-type.
Fundamentally, as the bandgap of a semiconductor in-
creases, it often becomes increasingly difficult to dope it in a
symmetric way 共both n- and p-type兲. On the one hand, un-
intentional conductivity may originate from the presence of
extrinsic impurities, such as, e.g., is the case for H in ZnO.
44
Native defects, on the other hand, can act as compensating
centers that counteract intentional doping. Their formation
depends on the position of E
F
, according to the relation,
45
⌬H共D
q
兲 = qE
F
+ n
D
共
D
−
SH
兲 + ⌬E
b
. 共2兲
This relation describes the formation enthalpy of dopant D
q
of charge state q 共where q =−1, 0, or +1兲 in the semiconduct-
ing host. Here,
D
and
SH
are the chemical potentials of the
dopants and host, n
D
is the number of dopants, ⌬E
b
=E共host+ defect兲 –E共host兲, and E is the total energy. The first
term accounts for the fact that D
+
donates an electron, and
FIG. 5. 共Color online兲 H diffusion energy E
D
ⴱ
in a-Si:H 共stars兲,
c-Si:H
共open circles兲 or c-Si: H 共closed circles兲 as a function of E
F
,atT
ann
=350 ° C. All data have been taken from Ref. 40.Thead hoc superposed
straight lines represent the dependence of the formation enthalpy ⌬H of
defect D
q
in the respective charge states q =+,0,and⫺,onE
F
共adapted after
Ref. 45兲.
103707-4 S. De Wolf and M. Kondo J. Appl. Phys. 105, 103707 共2009兲
Author complimentary copy. Redistribution subject to AIP license or copyright, see http://jap.aip.org/jap/copyright.jsp
D
−
accepts one. The transition level 共q +1/ q 兲 between
charge states q +1 and q is defined as the position of E
F
for
which the formation energies of these charge states are
equal.
46
Deliberate p-type 共n-type兲 doping of the material by
acceptors 共donors兲 will shift E
F
toward the VBM 共CBM兲 and
decrease the formation energy of native donors 共acceptors兲 to
a point where they are created spontaneously. Often, E
F
can-
not be brought beyond a certain point, the so-called the n-or
p-type pinning energy E
F
共n or p兲
, due to the occurring elec-
tronic compensation.
45
In a-Si: H, it may be speculated that the described 共asy-
metric兲 E
F
dependent defect generation is related to its rela-
tively wide bandgap as well. Relation 共2兲 can be sketched in
Fig. 5 by the ad hoc superposed straight lines, approximately
following the same trends as the experimental data. Here, the
most likely formed defect by either type of doping is the
amphoteric Si dangling bond D
3
, backbonded to three Si
atoms, via Si–H rupture.
47
At equilibrium, according to the
position of E
F
, this defect is either neutral 共D
3
0
兲, positively
共D
3
+
兲, or negatively 共D
3
−
兲 charged, accommodating, respec-
tively, 1, 0, and 2 electrons, forming the foundations of the
a-Si: H defect-pool model.
48,49
From the +/ 0 and 0 / − tran-
sition levels, a correlation energy U =共0 / −兲− 共+ / 0兲⬇
+0.38 eV is found, agreeing with values known for D
3
in
a-Si: H.
50
This energy originates from Coulomb repulsion,
possibly lowered by lattice relaxation at the defect-site.
51
Note that it is thanks to D
3
’s positive-U that a-Si: H can be
intentionally doped relatively well; negative-U defects pin
E
F
.
51
Nevertheless, when E
F
is located either in the 关VBM,
共+ / 0兲兴 or 关共0/ −兲, CBM兴 area of the bandgap 共both of
which are crosshatched in Fig. 5, respectively, with labels ➀
and ➂兲, D
3
will behave as a compensating center, counter-
acting the intentional doping. In these cases, the formation
energy of a dangling bond, compared to its neutral state, is
reduced by an amount, respectively, equal to ⌬E
+
共E
F
兲= 关共
+/ 0兲 − E
F
兴 and ⌬E
−
共E
F
兲= 关E
F
−共0 / −兲兴.
52,53
The diffusion
activation energy E
D
ⴱ
is reduced by similar amounts. These
arguments lead to conclude that the doping asymmetry of
a-Si: H films 共and thus, as argued, their difference in c-Si
surface passivation quality兲 originates from the asymmetrical
location of the D
3
+
and D
3
−
states in the a-Si: H bandgap.
To increase the thermal stability 共and thus doping effi-
ciency兲 of p-type 共n-type兲 material, it has been argued that
the VBM 共CBM兲 should be closer to 共further away from兲 the
vacuum level.
45
Fullfilling both conditions yields a reduced
bandgap material. For a-Si : H films, this can be accom-
plished by optimizing deposition conditions toward a low-
ered bonded hydrogen content of the films,
33
corresponding
to denser material.
54,55
This argument may explain why the
surface passivation properties of
c-Si共p
+
兲 films 关deposited
on a-Si:H共i兲 buffer layers兴 appear to be superior, compared
to their wider bandgap a-Si: H共p
+
兲 counterparts.
56,57
More-
over, the use of
c-Si共p
+
兲 films may also resolve possible
contact problems between transparent conductive oxide and
the p-type film, during device fabrication.
57,58
Care has to be
taken to deposit such denser material for the intrinsic buffer
layer, however, as it will easily result in undesired epitaxial
growth.
C. Stacked film passivation: Intrinsic buffer layer
For structures that feature a thin intrinsic buffer layer,
initially the passivation improves for both the
a-Si: H共p
+
兲/ a-Si:H共i兲 / c-Si and the
a-Si: H共n
+
兲/ a-Si:H共i兲 / c-Si case 共Fig. 3兲. Likely, this origi-
nates again from hydrogenation of c-Si surface states at the
a-Si: H共i兲 / c-Si interface, as described before. Nevertheless,
for the a-Si:H共p
+
兲/ a-Si:H共i兲 / c-Si case, from annealing at
about 220 ° C onwards, degradation sets in. For c-Si共n兲 HJ
solar cells featuring such a a-Si: H共p
+
兲 emitter, postdeposi-
tion annealing at moderate temperatures has been observed
to result in a similar local optimum for the value of the open
circuit voltage, V
oc
.
59
It has been argued that the loss in pas-
sivation is caused by Si–H rupture in the intrinsic buffer
layer, occurring at lower temperatures for the stacked
a-Si: H共p
+
兲/ a-Si:H共i兲 / c-Si structure compared to the
a-Si: H共i兲 / c-Si case 共lacking the a-Si:H共p
+
兲 overlayer兲.
34
This is evidenced in Fig. 4, where for the stacked
a-Si: H共p
+
兲/ a-Si:H共i兲 / c-Si structure more H
2
effuses out at
low temperatures, compared to the superposed
a-Si: H共p
+
兲/ c-Si and a-Si: H共 i兲 / c-Si structures 共label ⌺ in
the figure兲. This phenomenon was also observed by Einsele
et al. 共Ref. 60兲, and can be explained by Fig. 5: Due to the
presence of the p
+
-type overlayer, an electrical field will now
be present in the intrinsic buffer layer, pulling also here E
F
closer to the VBM, likely into the 关VBM, 共+ / 0兲兴 area 共with
label ➀兲, and thus lowering the defect formation energy.
For the a-Si:H共n
+
兲/ a-Si:H共i兲 / c-Si case, the passivation
trend has the same tendency as the same structure lacking the
a-Si: H共n
+
兲 overlayer: No degradation is observed. This re-
sult agrees well with the H
2
effusion data presented in panel
a of Fig. 4. Here, the signals for the stacked
a-Si: H共n
+
兲/ a-Si:H共i兲 / c-Si case are 共within experimental er-
ror兲 equal to the superposed a-Si:H共i兲 / c-Si and
a-Si: H共n
+
兲/ c-Si cases 共label ⌺ in the figure兲. Consequently,
in contrast to the p
+
-type case, an n
+
-type overlayer does not
lower the Si–H bond rupture energy in the underlying intrin-
sic film. This is consistent again with Fig. 5. Also here it
must be assumed that E
F
is shifted closer to the CBM in the
buffer layer. Most likely, however, this shift is not suffi-
ciently large to bring E
F
into the 关共0 / −兲, CBM兴 region 共la-
bel ➂兲. The fact that the passivation quality of the
a-Si: H共n
+
兲/ a-Si:H共i兲 / c-Si structure is slightly superior to
that of the a-Si:H共i兲/ c-Si case could now perhaps be ex-
plained by an additional field effect, repelling holes from the
c-Si surface states, but unsufficiently strong to reduce the
chemical a-Si: H / c-Si interface passivation.
V. CONCLUSIONS
In this article, it has been argued that doped a-Si:H / c-Si
interface recombination may result from Fermi energy de-
pendent defect generation in the passivating layer, counter-
acting the intentional doping. For a-Si : H films, for both
types of doping, this defect is likely the amphoteric Si dan-
gling bond, created by Si–H rupture, as evidenced from H
2
effusion and spectroscopic ellipsometry data. Interface pas-
sivation losses that may occur in case an intrinsic buffer
layer is present 共underneath the doped layers兲 could be ex-
103707-5 S. De Wolf and M. Kondo J. Appl. Phys. 105, 103707 共2009兲
Author complimentary copy. Redistribution subject to AIP license or copyright, see http://jap.aip.org/jap/copyright.jsp