Properties of interfaces in
amorphous/crystalline silicon
heterojunctions
Sara Olibet
*
,
1
, Evelyne Vallat-Sauvain
1
, Luc Fesquet
1
, Christian Monachon
2
,Aı
¨
cha Hessler-Wyser
2
,
Je
´
ro
ˆ
me Damon-Lacoste
1
, Stefaan De Wolf
1
, and Christophe Ballif
1
1
Ecole Polytechnique Fe
´
de
´
rale de Lausanne (EPFL), IMT, Photovoltaics and Thin Film Electronics Laboratory, Breguet 2,
2000 Neucha
ˆ
tel, Switzerland (until 31 December 2008 part of the University of Neucha
ˆ
tel)
2
Ecole Polytechnique Fe
´
de
´
rale de Lausanne (EPFL), Interdisciplinary Centre for Electron Microscopy (CIME), 1015 Lausanne,
Switzerland
Received 1 August 2009, revised 9 November 2009, accepted 13 November 2009
Published online 19 January 2010
PACS 68.35.Ct, 68.37.Lp, 72.20.Jv, 73.20.At, 73.40.Lq, 73.61.Jc
*
Corresponding author: e-mail solibet@1366tech.com, Phone: þ1-617-5120379
To study recombination at the amorphous/crystalline Si (a-
Si:H/c-Si) heterointerface, the amphoteric nature of silicon (Si)
dangling bonds is taken into account. Modeling interface
recombination measured on various test structures provides
insight into the microscopic passivation mechanisms, yielding
an excellent interface defect density reduction by intrinsic a-
Si:H and tunable field-effect passivation by doped layers. The
potential of this model’s applicability to recombination at other
Si heterointerfaces is demonstrated. Solar cell properties of a-
Si:H/c-Si heterojunctions are in good accordance with the
microscopic interface properties revealed by modeling, that are,
e.g., slight asymmetries in the neutral capture cross-sections
and band offsets. The importance of atomically abrupt
interfaces and the difficulties to obtain them on pyramidally
textured c-Si is studied in combination with transmission
electron microscopy.
ß 2010 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim
1 Introduction The efficiency of standard monocrys-
talline Si (c-Si) solar cells featuring diffused emitters and
aluminum back surface fields (BSF) is limited to moderate
values by interface recombination. This contrasts with
deposition of intrinsic/doped amorphous Si (a-Si:H) layer
stacks on c-Si, which effectively passivate the c-Si surfaces
and simultaneously form the emitter and BSF, while
avoiding the highly recombinative direct contact of metal
to c-Si. Such Si heterojunction (HJ) solar cells are fabricated
by the company Sanyo [1], resulting with 23% sunlight
conversion efficiency in the highest-efficient large area c-Si
solar cells [2]. Despite these excellent achievements, the
physical understanding of interfaces in a-Si:H/c-Si HJs is
limited. In this study, the amphoteric nature of Si dangling
bonds (DBs) is considered for modeling a-Si:H/c-Si inter-
face recombination [3], revealing the microscopic nature of
this interface passivation scheme. The intuitive interpret-
ation of measured injection-level dependent lifetimes at
various a-Si:H/c-Si interfaces is facilitated by means of
trajectories on three-dimensional surface recombination rate
plots [4]. The use of this amphoteric interface recombination
formalism to model recombination at other Si heterointer-
faces featuring DBs is demonstrated. For Si HJ solar cell
formation, layer stacks with the required properties were
identified by modeling. The measured solar cell parameters
confirm our modeling results and add up with other
researchers findings to a more complete picture of Si HJs.
The importance of atomically abrupt interfaces for highest
passivation quality was confirmed in this study by combining
lifetime measurements with transmission electron micro-
scopy (TEM), quantifying the detrimental effect of epitaxial
interfaces, that could finally also be suppressed in textured Si
HJs.
2 Experimental Hydrogenated amorphous and
microcrystalline silicon (a-Si:H and mc-Si:H) layers were
grown by very high frequency plasma enhanced chemical
vapor deposition (VHF-PECVD) in a single chamber
deposition system. SiH
4
,H
2
,PH
3
, and trimethylboron
(TMB: B(CH
3
)
3
) were used as precursor gases to grow
intrinsic (i) a-Si:H and doped mc-Si:H layers on c-Si
wafers.
Phys. Status Solidi A, 1–6 (2010) / DOI 10.1002/pssa.200982845
pss
applications and materials science
a
status
solidi
www.pss-a.com
physica
ß 2010 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim
The properties of a-Si:H/c-Si heterointerfaces were
studied by means of the photoconductance technique
measuring the effective charge carrier lifetime t
eff
(s) as a
function of the excess carrier density Dn (cm
3
) [5], see
Fig. 1. Using the highest quality of c-Si wafers, hetero-
structures are dominated by their interface properties
quantified by the effective surface recombination velocity
S
eff
(cm/s), accessible from t
eff
by
S
eff
¼
ðt
1
eff
t
1
b;intr
ÞW
2
, (1)
where symmetrical surface passivation is assumed, W is the
wafer thickness and intrinsic bulk c-Si recombination
(t
b;intr
) only dominates at high Dn [6], see again Fig. 1. The
absolute lifetime values as well as their injection-level
dependencies are given by the specific interface passivation
mechanisms. From such t
eff
(Dn) plots one can additionally
extract the very valuable information of the 1-sun
illumination-level open-circuit voltage (V
oc
) that a solar
cell with these interface recombination properties, and
based on this wafer would have, called the implied V
oc
(implV
oc
), see again Fig. 1 [5].
3 Modeling The surface recombination rate U (cm
2
/s)
is related to the experimentally accessible S
eff
by
U ¼ S
eff
Dn, (2)
where Dn ¼ Dp is the bulk excess carrier density generated,
e.g., by illumination. Recombination through defect levels
in semiconductors is usually described by the Shockley–
Read–Hall (SRH) theory, where the single trap level surface
recombination rate U
SRH
is given by
U
SRH
¼
n
s
p
s
n
s
=s
p
þ p
s
=s
n
v
th
N
s
: (3)
Without additional surface charges, the surface carrier
densities n
s
, p
s
equal the bulk carrier densities n
b
¼ n
0
þ Dn
and p
b
¼ p
0
þ Dn, where n
0
, p
0
are the thermal equilibrium
charge carrier densities. s
n
and s
p
(cm
2
) are the capture
cross-sections of electrons and holes, v
th
(cm/s) is the
thermal velocity of the charge carriers, and N
s
(cm
2
) is the
surface defect density. Figure 2a visualizes U
SRH
ðn
s
; p
s
Þ
(Eq. 3) for the capture cross-section ratio generally assumed
to model SiO
2
/c-Si interface recombination, i.e.,
s
n
=s
p
¼ 100 [7]. In our experiments, the carrier density
Dn, thus n
s
and p
s
vary and U varies accordingly, as shown
in Fig. 2a by means of the trajectories for 1 V cm n- and
p-type c-Si. Figure 2b shows the corresponding S
eff;SRH
ðDnÞ
plots (Eq. 2). With the capture cross-section asymmetry of
s
n
> s
p
, surface recombination at the SiO
2
/p c-Si interface
is thus higher than at the SiO
2
/n c-Si interface.
A surface charge density Q
s
(cm
2
) results in a surface
band bending c
s
(V) and thus the surface carrier densities
differ from the ones in the bulk: n
s
¼ n
b
exp ðþqC
s
=kTÞ and
p
s
¼ p
b
exp ðqC
s
=kTÞ. Figure 3 shows the effect of a
positive surface charge, as e.g., reported at the SiN
x
/c-Si
interface, on the trajectories on the U
SRH
ðn
s
; p
s
Þ plot for
equal capture cross-sections and 1 V cm n- and p-type c-Si.
While repelling the minority holes from the n-type c-Si
surface reduces interface recombination, attracting the
2 S. Olibet et al.: Properties of interfaces in a-Si:H/c-Si HJs
physica
ssp
status
solidi
a
Figure 1 Effective charge carrier lifetime t
eff
as a function of the
excess carrier density Dn measured by the photoconductance tech-
nique.
Figure 2 Impact of the wafer doping for a given capture cross-
section ratio s
n
=s
p
on (a) the surface recombination rate
U
SRH
ðn
s
; p
s
Þ and (b) the surface recombination velocity
S
eff;SRH
ðDnÞ.
ß 2010 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim www.pss-a.com
minority electrons to the p-type c-Si surface increases the
surface recombination rate, as compared to the flatband case.
The standard SRH surface recombination model is based
on interface traps having two possible charge states. Because
the bare c-Si surface as well as a-Si:H feature DBs, we
propose to model heterostructure interface recombination by
amphoteric defects, i.e., DBs having three possible charge
states. For this, we extended a model previously established
for bulk a-Si:H recombination assuming one single recom-
bination level with three charge conditions [8] to the c-Si
surface, resulting in the surface DB recombination rate U
DB
U
DB
¼
n
s
s
0
n
þ p
s
s
0
p
ðp
s
=n
s
Þðs
0
p
=s
þ
n
Þþ1 þðn
s
=p
s
Þðs
0
n
=s
p
Þ
v
th
N
s
,
(4)
where s
0
n
, s
0
p
are the capture cross-sections of the neutral
states and s
þ
n
, s
p
are the capture cross-sections of the
charged states. Note that to find Eq. (4) the illumination
level must be high enough to neglect emission from DB
states [8]. Best fits of our experimental data are obtained
with s
0
n
=s
0
p
¼ 1=20 and s
þ
n
=s
0
n
¼ s
p
=s
0
p
¼ 500. The same
capture cross-section hierarchy of s
0
n
< s
0
p
< s
þ
n
< s
p
is
found in amorphous semiconductors by Street [9], although
much less pronounced. The surface plot of U
DB
ðn
s
; p
s
Þ,
shown in Fig. 4, has a local minimum whose position is
given mainly by the neutral capture cross-section ratio.
Within its width determined by the charged to neutral
capture cross-section ratios, recombination is dominated by
majority carriers and thus opposite to the common SRH
recombination. Comparing the Dn-dependent trajectories on
the U
DB
ðn
s
; p
s
Þ plot in Fig. 4 to the ones similar to SiO
2
- and
SiN
x
-passivated c-Si in Figs. 2 and 3, shows that a-Si:H
passivation is more symmetrical as far as surface passiva-
tion of both wafer doping types is concerned. Fits
to measured t
eff
ðDnÞ plots are obtained by combining
Eq. (1), (2), and (4)
t
1
eff
¼ t
1
b
þ
2
W
1
Dn
U
DB
ðDn; n
0
; p
0
; Q
s
; N
s
; s
0
p
; s
0
n
=s
0
p
; s
þ;
=s
0
Þ:
(5)
4 Results
4.1 i a-Si:H passivation of various c-Si Figure 5
shows by symbols the measured injection-level dependent
lifetimes of i a-Si:H passivating variously doped flat c-Si,
implying excellent open-circuit voltages implV
oc
over
700 mV throughout, see legend. Thus, the simple low
temperature a-Si:H passivation scheme compares favorably
to best performing SiO
2
and SiN
x
layers [6, 10]. Best a-Si:H/
c-Si interface passivation with surface recombination
Phys. Status Solidi A (2010) 3
Original
Paper
Figure 4 Surface recombination rate U
DB
ðn
s
; p
s
Þ based on the
amphoteric nature of c-Si surface dangling bonds, including exam-
ples of Dn-dependent trajectories.
Figure 3 Impact of the surface charge density Q
s
and the wafer
doping on the trajectories on the surface recombination rate plot
U
SRH
ðn
s
; p
s
Þ.
Figure 5 Lifetimes t
eff
of i a-Si:Hpassivating differently doped flat
c-Si. Symbols are measurements and lines are fits with our ampho-
teric interface recombination model using the parameters listed in
Table 1.
www.pss-a.com ß 2010 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim
velocities S
eff
down to 1 cm/s were measured on (111) crystal
oriented c-Si, likely related to the possibility of perfect
monohydride hydrogen termination of such (111) surfaces
[11] despite their higher DB density. This is opposite to SiO
2
and SiN
x
where the passivation of (111) oriented c-Si is
inferior to the one of (100) oriented c-Si [12]. Concerning the
application of passivation in the industrially relevant
textured monocrystalline Si solar cells featuring (111)
oriented pyramidal facets, this is an important finding.
Figure 5 includes by lines fits to the experimental curves
obtained with our amphoteric interface recombination
model. The extracted values of the interface DB density N
s
and the charge density Q
s
for the different wafers are listed in
Table 1. Q
s
is the DB charge within the passivating a-Si:H
layer inducing the image charge Q
Si
in the c-Si surface [4]. In
the present case of intrinsic a-Si:H passivating variously
doped c-Si, the interface DB charge is determined by the
wafer doping type and level, by band offset asymmetries and
the lightly n-type doped character of nominally intrinsic a-
Si:H. For example, best fits obtained with negative charge in i
a-Si:H at the interface to clearly n-type doped c-Si and
positive charge in i a-Si:H to p-type doped c-Si confirm the
amphoteric nature of a-Si:H/c-Si interface defects.
4.2 a-Si:H/c-Si field-effect passivation Unlike
SiN
x
(and less pronounced also SiO
2
) featuring a positive
interface charge when grown on c-Si, in the case of a-Si:H,
field-effect passivation of both charge types can be
controlled by varying the average state of charge on the
interface DBs, e.g., by an overlaying doped mc-Si:H layer.
The i a-Si:H buffer layer in such intrinsic/doped layer stacks
ensures a low interface defect density N
s
. Figure 6 shows
experimental t
eff
ðDnÞ plots of intrinsic/doped layer stacks
such as used for silicon HJ solar cell formation on 2.5 V cm
n-type (a) and p-type (b) c-Si together with their fits. The
subsequent representation in terms of trajectories over
surface recombination rate plots in Fig. 6c allows for an
easier, more intuitive interpretation of these t
eff
ðDnÞ plots.
The U
DB
ðn
s
; p
s
Þ plot in Fig. 6c thus illustrates that the
measured lower lifetimes at low injection levels on p-type
(open symbols) than on n-type (full symbols) c-Si result from
the slight neutral capture cross-section asymmetry, leading
in general to higher recombination when p
s
> n
s
.
4.3 Modeling of SiO
2
/c-Si interface rec om-
bination by amphoteric states The identification of
SiO
2
/c-Si interface defects as DBs [13] allows for a broader
application of our amphoteric interface recombination for-
malism. Yablonovitch et al. [14] measured SiO
2
/c-Si interface
recombination as a function of n
s
/p
s
, see symbols in Fig. 7, and
4 S. Olibet et al.: Properties of interfaces in a-Si:H/c-Si HJs
physica
ssp
status
solidi
a
Figure 6 Additional field-effect passivation by internally polarizing i a-Si:H layers with an overlaying doped m c-Si:H layer on 2.5 V cm n-
type (a) and p-type (b) c-Si. Symbols are measured t
eff
ðDnÞ plots and lines are their fits, also represented in terms of trajectories over surface
recombination rate plots to facilitate their intuitive interpretation (c).
Table 1 Model parameter couples interface dangling bond
density N
s
/charge density Q
s
, giving best accordance between
the measured and calculated t
eff
ðDnÞ plots in Fig. 5.
s
0
p
¼ 10
16
cm
2
is assumed.
i a-Si:H on N
s
(10
9
cm
2
) Q
s
(10
10
cm
2
)
n 2.5 V cm 1.0 2.2
n (111) 30 V cm 0.45 0.5
n60V cm 1.6 þ1.1
>15 kV cm (n) 1.6 þ0.1
p 2.5 V cm 1.4 þ1.8
p 130 V cm 3 þ0.5
Figure 7 SiO
2
/c-Si interface recombination as a function of n
s
/p
s
measured by Yablonovitch et al. [14]. Fits with two standard SRH
interfacestates ofdifferentcapture cross-sectionratios (dashedlines,
adding up) and fits with amphoteric interface states (solid line)
coincide.
ß 2010 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim www.pss-a.com
fitted their data by adding up the SRH recombination due to
two individual interface states of different capture cross-
section ratios. The dashed lines in Fig. 7 show such a fit with a
dominant defect of s
n
=s
p
¼ 60 and a lesser one of
s
n
=s
p
¼ 1=25. Our modeling of the same data with
amphoteric interface DBs represented by the bold line in
Fig. 7 yields exactly the same fit, but with a physical meaning
attributed to the fit parameters, that is larger charged than
neutral capture cross-sections.
4.4 a-Si:H/c-Si heterojunctions a-Si:H/c-Si HJ
solar cells must feature simultaneously lowest interface
defect densities for highest open-circuit voltages (V
oc
) and
highest doped layers for highest fill factors (FF), with the
thinnest layers possible, to allow for current extraction. Thus,
the modeling assisted interpretation of measured lifetime
curves permits the development of suitable layer stacks for Si
HJ solar cell formation, which were found to consist of very
thin i a-Si:H layers providing excellent interface passivation
to c-Si, followed by under high H
2
-dilution grown highest
doped mc-Si:H layers ensuring best carrier extraction from
the a-Si:H/c-Si HJ. A 19.1% efficient (4.5 mm
2
) surfaced Si
HJ solar cell was reached on flat n-type c-Si with [V
oc
(mV),
FF (%), J
sc
(mA/cm
2
), h (%)] ¼ [680, 82, 34, 19.1], where J
sc
is the short-circuit current density and h the efficiency. With
a slightly thicker i-layer, high V
oc
Si HJs were achieved with
n-wafer ¼ [705, 78, 32, 17.6] and p-wafer ¼ [690, 74, 32,
16.3]. The slight neutral capture cross-section asymmetry
and the band offset asymmetry, see Fig. 6 and Table 1
including related comments, give an explanation for the
poorer performance of Si HJs based on p-wafers. The best
V
oc
s of 730 mV were obtained on (111) oriented c-Si, as
predicted by i a-Si:H passivation, see Fig. 5. We relate the
poor FFs of typically 50% of such cells partially to the lower
doping level of this wafer (30 V cm) making the required
tunneling transport across the ITO(n)/p mc-Si:H/i a-Si:H/n c-
Si interface even more challenging [15].
4.5 Atomic a-Si:H/c-Si interfaces From the crystal-
lographic point of view, abrupt interfaces are the key to low
interface recombination. As first observed by Wang et al.
[16] interface recombination increases when epitaxized
interfaces occur. For example, under the same process
conditions grown highly H
2
diluted i a-Si:H layer leads to
best interface passivation on flat (111) oriented c-Si, but to
epitaxized (100) oriented c-Si interfaces as observed by high
resolution transmission electron microscopy (HR-TEM).
Presumably the (111) c-Si surface benefits from an increased
monohydride hydrogen passivation [11] while the (100) c-Si
surface suffers from an epitaxial interface of low quality and
an increased a-Si:H/c-Si interface area [4]. The resulting
increase in interface recombination by a factor of 40
corresponds, e.g., to an implV
oc
decrease from 710 down to
600 mV. On pyramidally textured c-Si featuring (111) facets,
local epitaxial growth in the pyramid grooves, presumably
resulting from stress, increases interface recombination by a
factor of 4 when featuring large pyramids, reducing implV
oc
,
e.g., from 720 to 675 mV. However, if epitaxial growth is
suppressed, implV
oc
s even higher than on flat (100) c-Si were
reached despite the increased textured interface area,
because of the superior passivation of the (111) oriented
pyramid facets over the flat (100) oriented c-Si.
On the Si HJ solar cell level, HR-TEM micrographs
confirm the abrupt nature of the interfaces in our best
performing flat Si HJ solar cells consisting of c-Si/i a-Si:H/
doped mc-Si:H layers, see Fig. 8a. On pyramidally textured
c-Si, abrupt facet interfaces but epitaxized texture grooves,
see Fig. 8b, related to the growth of only thin i a-Si:H layers
followed by the deposition of an overlaying doped mc-Si:H
layer grown under very high H
2
dilution, lead to lowered V
oc
s
of 660 mV on n-type doped c-Si. Reducing not only the
groove density per projected surface area by using large,
regular pyramidally textured c-Si, but blunting additionally
the sharp pyramidal grooves, could reduce the amount of
epitaxized a-Si:H/c-Si interfaces. Thus, textured Si HJ solar
Phys. Status Solidi A (2010) 5
Original
Paper
Figure 8 HR-TEM micrographs showing (a) abrupt flat crystallographic c-Si/a-Si:H/mc-Si:H interfaces of Si HJ solar cells, (b) an
epitaxially connected i a-Si:H interface passivation layer in the pyramidal groove of a textured Si HJ solar cell, and (c) a rough a-Si:H/c-Si
interface, also of a textured Si HJ solar cell.
www.pss-a.com ß 2010 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim
