UC Berkeley
UC Berkeley Previously Published Works
Title
Above-bandgap voltages from ferroelectric photovoltaic devices.
Permalink
https://escholarship.org/uc/item/6831m90k
Journal
Nature nanotechnology, 5(2)
ISSN
1748-3387
Authors
Yang, SY
Seidel, J
Byrnes, SJ
et al.
Publication Date
2010-02-01
DOI
10.1038/nnano.2009.451
Peer reviewed
eScholarship.org Powered by the California Digital Library
University of California
Above-bandgap voltages from ferroelectric
photovoltaic devices
S. Y. Yang
1
*
, J. Seidel
2,3
,S.J.Byrnes
2,3
,P.Shafer
1
, C.-H. Yang
3
,M.D.Rossell
4
,P.Yu
3
,Y.-H.Chu
5
,
J. F. Scott
6
,J.W.Ager,III
2
,L.W.Martin
2
and R. Ramesh
1,2,3
In conventional solid-state photovoltaics, electron–hole pairs are created by light absorption in a semiconductor and
separated by the electric field spaning a micrometre-thick depletion region. The maximum voltage these devices can
produce is equal to the semiconductor electronic bandgap. Here, we report the discovery of a fundamentally different
mechanism for photovoltaic charge separation, which operates over a distance of 1–2 nm and produces voltages that are
significantly higher than the bandgap. The separation happens at previously unobserved nanoscale steps of the
electrostatic potential that naturally occur at ferroelectric domain walls in the complex oxide BiFeO
3
. Electric-field control
over domain structure allows the photovoltaic effect to be reversed in polarity or turned off. This new degree of control,
and the high voltages produced, may find application in optoelectronic devices.
T
he conversion process of light energy to electrical energy in
photovoltaic devices relies on some form of built-in asymme-
try that leads to the separation of electrons and holes. The fun-
damental physics behind this effect (for example, in silicon-based
cells) is charge separation using the potential developed at a p–n
junction, or heterojunction
1–3
. This suggests the following ques-
tion—are there other pathways to accomplish charge separation in
materials to enable the next generation of photovoltaics? In the
past, anomalous photovoltaic effects in polar materials have been
found to arise from two mechanisms: (i) granularity
4,5
and (ii) the
inherent non-centrosymmetry in the bulk material, that is, the
absence of an inversion centre of symmetry
6–9
. The former mechan-
ism inevitably suffers from the granular interface being poorly
controlled, and the latter is typically seen in wide-bandgap semicon-
ductors (E
g
. 2.5 eV), which absorb very little of the visible spec-
trum. In this paper, we describe a new mechanism of charge
separation and photovoltage generation that occurs exclusively at
nanometre-scale ferroelectric domain walls in a model ferroelectric,
BiFeO
3
(BFO), under white-light illumination. In contrast to semi-
conductor-based photovoltaics
10
, the photovoltages observed here
are significantly higher than the electronic bandgap.
The rhombohedrally distorted perovskite structure of BFO leads
to eight ferroelectric polarization directions along the pseudocubic
111-directions, corresponding to four structural variants. The poss-
ible domain pattern formation in (001)-oriented epitaxial rhombo-
hedral perovskite ferroelectric films and their control has been well
established by earlier theoretical and experimental studies
11–13
.To
eliminate confusion, we use the notation set prescribed in ref. 11.
Domain walls in such materials are typically 1–2 nm wide
14
.
Additionally, recent studies have demonstrated that BFO has a
direct bandgap of 2.67 eV (465 nm)
15
and has been previously
shown to display a conventional photovoltaic effect (open-circuit
voltage V
OC
E
g
)
16
and photoconductivity
15
.
The details of the growth of the BFO thin films used in our study
are given in the Methods. Piezoresponse force microscopy (PFM)
reveals that ordered arrays of 718 (Fig. 1a and schematically depicted
in Fig. 1b) and 109 8 domain walls with two in-plane variants (Fig. 1c
and schematically in Fig. 1d) have been created through such a
careful heteroepitaxial growth process. X-ray diffraction studies
(insets of Fig. 1a,c) confirm the presence of these two different
types of domain wall
17
. Additional X-ray diffraction reciprocal-
space-mapping studies (Supplementary Fig. S1) reveal the high
quality of these ordered stripe domains. In both cases, there is a
net polarization aligned in the plane of the film, that is, perpendicu-
lar to the projection of the domain wall plane on the (001) film
surface (Fig. 1b,d). Transmission electron microscopy (TEM)
images of the two different domain structures show that the 718
domain walls (Supplementary Fig. S2a) lie along 101-type planes,
whereas the 1098 domain walls (Supplementary Fig. S2b) lie along
100-type planes, consistent with theoretical predictions
11
. Detailed
analyses of the atomic structure at these domain walls reveal a
wall width of 1–2 nm, consistent with previous work
14,18
. Such
nanoscale domain-wall features are the focus of this work.
Photovoltaic device measurements
Test structures, based on symmetric platinum top electrodes with a
length of 500 mm and an inter-electrode distance of 200 mm, were
fabricated on top of 100-nm-thick films by photolithography in
two geometries: electrodes for electric transport measurements (i)
perpendicular (DW
?
) and (ii) parallel (DW
k
) to the domain walls
(DW, Fig. 2a,b, respectively). Current–voltage (I–V) characteristics
of samples in the two geometries, with ordered arrays of 718
domain walls, were measured under saturation illumination
(Supplementary Fig. S3) on the same film in both dark- and
white-light illumination (285 mW cm
22
) and reveal strikingly
different photovoltaic behaviours (Fig. 2c,d). In the DW
?
direction,
a large photo induced V
OC
of 16 V was measured, with in-plane
short-circuit current density J
sc
1.2 10
24
Acm
22
. In contrast,
dark and light I–V curves measured in the DW
k
direction exhibit
a significant photoconductivity, but no photo induced V
OC
.
1
Department of Materials Science and Engineering, University of California, Berkeley, Berkeley, California 94720, USA,
2
Materials Sciences Division,
Lawrence Berkeley National Laboratory, Berkeley, California 94720, USA,
3
Department of Physics, University of California, Berkeley, California 94720, USA,
4
National Center for Electron Microscopy, Lawrence Berkeley National Laboratory, Berkeley, California 94720, USA,
5
Department of Materials Science and
Engineering, National Chiao Tung University, HsinChu, Taiwan 30010,
6
Department of Earth Sciences, University of Cambridge, Cambridge CB2 3EQ, UK.
*
e-mail: syyang@berkeley.edu
ARTICLES
PUBLISHED ONLINE: 10 JANUARY 2010 | DOI: 10.1038/NNANO.2009.451
NATURE NANOTECHNOLOGY | VOL 5 | FEBRUARY 2010 | www.nature.com/naturenanotechnology 143
© 2010 Macmillan Publishers Limited. All rights reserved.
The photo induced voltages were found to increase linearly in
magnitude as the electrode spacing was increased (Fig. 3a). Most
importantly, a single domain sample (that is, with no domain
walls between the platinum contacts; black curve, Fig. 3a) show neg-
ligible levels of photovoltage, which rules out a ‘bulk’ photovoltaic
effect arising from non-centrosymmetry
6,7
. In turn, this strongly
suggests the prominent role of domain walls in creating the anom-
alous photovoltages. In fact, the magnitude of the overall potential
drop varies linearly with the total number of domain walls
between the electrodes (Fig. 3a). The thickness dependence of the
b
d
P
net
P
net
21.0 21.5 22.022.523.0
c
1 µm
a
[100]
pc
[100]
pc
[010]
pc
[010]
pc
109° DW
(001)
pc
71° DW
(110)
pc
1 µm
ω (deg)
ω (deg)
22.0 22.1 22.2 22.3 22.4
Intensity (a.u.)
Intensity (a.u.)
Figure 1 | Model domain-wall architectures. a, Piezoresponse force microscopy image of ordered arrays of 718 domain walls. Inset: corresponding X -ray
rocking curves, along two orthogonal crystal axes, demonstrating the high quality of the films. b, Schematic of the 718 domain-wall arrays. The various arrows
map out the different components of polarization (both in-plane and out-of-plane) as well as the net polarization direction (large arrow) in the samples.
Samples are found to hav e net polarization in the plane of the film. c, Piezoresponse force microscopy image of ordered arrays of 1098 domain walls. Inset
shows the corresponding X -r ay rocking curves, along two orthogonal crystal axes. d, Schematic of the 1098 domain-wall arra ys.
−3 × 10
−4
−15 −10 −5 0 5 10 15
Voltage (V)
−15 −10 −5 0 5 10 15
Voltage (V)
−2
× 10
−4
−1 × 10
−4
1 × 10
−4
2 × 10
−4
3 × 10
−4
0
Dark current
Photo current
Dark current
Photo current
Current density (A cm
−2
)
−6 × 10
−4
−4 × 10
−4
−2 × 10
−4
2 × 10
−4
4 × 10
−4
6 × 10
−4
0
Current density (A cm
−2
)
a
c
b
d
V V
P
net
P
net
Figure 2 | Light and dark I –V measurements. a,b, Schema tics of the perpendicular (DW
?
)(a)andtheparallel(DW
k
)(b) device geometries.
c,d, Corresponding I–V measur ements of the DW
?
(c)andDW
k
(d) devices, respectiv ely.
ARTICLES
NATURE NANOTECHNOLOGY DOI: 10.1038/NNANO.2009.451
NATURE NANOTECHNOLOGY | VOL 5 | FEBRUARY 2010 | www.nature.com/naturenanotechnology144
© 2010 Macmillan Publishers Limited. All rights reserved.
photovoltage provides another route to verify this conclusion,
because the wall density scales inversely with film thickness
19,20
.
From PFM analysis we have calculated the average
domain spacing and used this to calculate the potential drop
for each domain wall to be 10 mV, irrespective of the domain
width (Fig. 3b). This value is quite close to the theoretically
predicted 20 mV potential drop across 718 domain walls in
BFO (ref. 21).
Model for the photovoltaic effect
Figure 4a–d shows our proposed model for the effect described
above. Figure 4a is a schematic of the model domain structure
showing a series of 718 domain walls. Figure 4b shows the corre-
sponding position of the valence (VB) and conduction (CB)
bands in dark conditions. Recent ab initio calculations suggest
that ferroelectric domain walls have built-in potential steps
14,21
arising from the component of the polarization perpendicular to
the domain wall. The associated charge density,
r
¼ 2r
.
P, forms
an electric dipole, leading to an electric field within the wall and a
potential step from one side to the other. In a strongly correlated,
polar system such as BFO, the photo generated exciton is expected
to be localized and tightly bound. Therefore, such an exciton in the
bulk of the BFO (Fig. 4b(i)), is expected to quickly recombine,
resulting in no net photo effect. If the light is incident at the
domain wall (Fig. 4b(ii)), the significantly higher local electric
field enables a more efficient separation of the excitons, creating a
net imbalance in charge carriers near the domain walls and resulting
in the band diagram shown in Fig. 4c. This effect (analogous to the
type-II band alignment that drives polymeric solar cells) means that,
under illumination, a net voltage is observed across the entire
sample, resulting from the combined effect of the domain walls
and the excess charge carriers created by illumination (Fig. 4d).
Photo excited electron–hole pairs are separated and drift to either
side of the domain wall, building up an excess of charge. A close
inspection of the effects at a given domain wall (Fig. 4) reveals a
similar picture to a classic p–n junction. The key difference is the
magnitude of the electric field that drives charge separation. In a
classic silicon-based system (V
OC
0.7 V; depletion layer thickness,
1 mm), an effective electric field of 7kVcm
21
is obtained (com-
pared with the BFO system, with a field of 50 kV cm
21
) for each
domain wall. In open-circuit illuminated conditions, the electric
field across the domain walls should decrease relative to its
thermal-equilibrium value, creating a drift-diffusion current equal
and opposite to the photocurrent described above. The domains
themselves maintain the same electric field as in thermal equili-
brium, because this is already the correct field for zero net
current. Therefore, a net electric field would build up across the
sample (Fig. 4c).
To validate this model, we first rule out the bulk photovoltaic
effect previously observed in other ferroelectric crystals such as
LiNbO
3
(LNO). It is useful to make comparisons with known
results on periodically poled LNO, because BFO and LNO have
the same symmetry and LNO is an extensively studied photovoltaic
ferroelectric material
22
. There have been no reports of large photo-
voltages being generated in undoped LNO and, because LNO and
BFO both have a bulk symmetry R3c, this implies that such high-
voltage output in the latter is very unlikely to be a bulk property.
Additionally, despite possessing the same bulk symmetry, the
domain structures in LNO and BFO are very different. LNO has a
rhombohedral–rhombohedral crystal class-preserving ferroelectric
phase transition. As a result, it cannot be ferroelastic, and only
1808 domain walls can exist
23
. These apparently play no part in
any large photovoltage output. In contrast, BFO has a rhombo-
hedral–orthorhombic transition at its Curie temperature. This is a
ferroelastic phase transition with 718,1098 and 1808 domain walls.
Thus, quantitative differences in photovoltaic response suggest the
role of either 718 or 1098 domain walls.
Finally, we note that the bulk photovoltaic tensor is generally
third-rank and non-diagonal in R3c materials such as LNO
(ref. 24). Thus, application of an optical field is, in general, affected
not only by the r
33
photovoltaic coefficient
22
, but also by the r
15
coef-
ficient
25
. In a typical experiment on LNO (refs 26,27), this off-diag-
onal term produces a field of 40 kV cm
21
perpendicular to the
threefold polar axis for 500 mW of 514.5 nm laser light weakly
focused to a 50-mm spot diameter. This number may be compared
with those in the present study and suggests that a fully quantitative
analysis must involve the full off-diagonal photovoltaic tensor. This
was suggested earlier for puzzling results regarding the photovoltaic
response in YBaCuO high-T
C
materials
28
. We also note that the
photovoltaic response perpendicular to the polar threefold axis can
be compensated or enhanced by a strong thermal gradient
29–31
.
Because certain domain walls conduct electricity in BFO, this could
involve local heating
21
. Thus, comparison of the present data with
those for LNO supports the argument that the new effects reported
here cannot be bulk in nature.
Final evidence of a completely new photovoltaic mechanism
comes from the fact that the direction of the measured J
SC
in our
BFO films is parallel to the net in-plane polarization. This current
direction is opposite to what has been observed for granular ferro-
electric materials
5
. In turn, we have observed that there is a drop
in the potential in the direction of the net in-plane polarization in
these epitaxial BFO films. The expected magnitude of J
SC
can be
0 50 100
100 150 200 250
150 200
0
5
10
15
20
0
5
10
15
20
Single domain
100
nm
200
nm
500
nm
V
OC
(V)
Electrode distance (µm)
Potential drop (mV)
Domain width (nm)
a
b
Figure 3 | Role of domain-walls in the photov oltaic response. a, Study of the evolution of V
OC
as a function of electrode spacing for four different samples:
718 domain-walls samples with thicknesses of 100 nm (red), 200 nm (blue) and 500 nm (green) as well as a monodomain BFO film having no domain walls
(black). A clear correla tion between the number of domain walls and the magnitude of V
OC
is observed. b,Thepotentialdropinrelationtodomain-wall
width is essentially constant regardless of the spacing betw een domain walls. The dashed line repre sents the theoretically expected potential drop per
domain wall from ref. 21.
NATURE NANOTECHNOLOGY DOI: 10.1038/NNANO.2009.451
ARTICLES
NATURE NANOTECHNOLOGY | VOL 5 | FEBRUARY 2010 | www.nature.com/naturenanotechnology 145
© 2010 Macmillan Publishers Limited. All rights reserved.
predicted, and is consistent with measurements (details of the calcu-
lation are given in the Supplementary Information).
Control of the photovoltaic response by electric fields
To demonstrate an additional level of control of the photovoltaic
effect in these films, we have studied the evolution of photovoltaic
properties as a function of domain switching in planar device struc-
tures. I–V characterization of an as-grown device structure in the
DW
||
geometry is shown in Fig. 5a. Consistent with data in Fig. 2,
there is no observable photovoltaic response in this geometry.
Using a device spacing of 10 mm, we can then apply voltage
pulses of 200 V between the two in-plane electrodes to induce
ferroelectric domain switching. Following application of such a
field (E 200 kV cm
21
) for a pulse of 100 ms, a corresponding
rotation of the ferroelectric domain structure was observed,
thereby creating a system with the DW
?
geometry. Subsequent
light I–V measurements reveal the formation of an anomalous
photovoltaic effect in this film (blue curve, Fig. 5a). The correspond-
ing PFM image following the þ200 V pulse reveals that the domain
structure is effectively rotated by 908 from the original configuration
(Fig. 5b, top and middle panels). It is clear that this domain
configuration is essential to create the potential drop necessary for
the anomalous photovoltaic effect. Furthermore, upon application
of a –200 V/100 ms pulse, the polarity of the photo-induced
voltage and current can be flipped (red curve, Fig. 5a). This is
explained by a change in the direction of the net, in-plane polariz-
ation of the BFO film (Fig. 5b, bottom panel).
Our earlier theoretical work showed that the magnitude of the
potential step is higher in the case of 1098 domain walls (150 mV,
compared to 20 mV for 718 domain walls). We were constrained
in terms of a macroscopic measurement of the 1098 domain
samples due to the presence of a random distribution of the two
in-plane variants. We therefore carried out microscopic measure-
ments (details of measurements are given in Supplementary
Fig. S4), which revealed an 4 larger potential drop per domain
wall compared to the 718 walls.
Conclusion
In summary, we have shown that a photovoltaic effect in BFO thin
films arises from a unique, new mechanism—namely, structurally
driven steps of the electrostatic potential at nanometre-scale
domain walls. These potential steps, which had been hypothesized,
Wall
CB
(i)
a
b
c
d
2 nm
e
−
e
−
h
+
h
+
~150 nm
(ii)
VB
Wall
Wall
Figure 4 | Band structure in dark conditions and under illumination. a, Schematic of four domains (three domain walls) in an order array of 718 domain
walls. b, Corresponding band diagram showing the valence band (VB) and conduction band (CB) across these domains and domain walls in the dark. Note
that there is no net voltage across the sample in the dark. Section (i) illustrates a photon hitting in the bulk of a domain and section (ii) a photon hitting at a
domain wall. c, Evolution of band structure upon illumination of the domain wall array. d, Detailed picture of the build-up of photo excited charges at a
domain wall.
BFO
DSO
−200 V pulse
As-grown
−1.0 −0.5 0.0 0.5 1.0
Voltage (V)
+200
V pulse
V
Pt
−3 × 10
−4
−2 × 10
−4
−1 × 10
−4
1 × 10
−4
2 × 10
−4
3 × 10
−4
0
a
b
Current density (A cm
−2
)
−200
V
+200
V
As-grown
Figure 5 | Domain-wall switching effect. a,LightI–V measurement in the DW
k
geometry sho ws no observable photo v oltaic effect, as gr o wn. On rotation of
the domain structure to the DW
?
configuration after application of +200 V voltage pulses to the in-plane device structure, a photovoltaic effect is observed.
b, Corresponding PFM images of the as-grown (top panel), 200 V poled (middle panel), and 2200 V poled (bottom panel) device structures. The arrows
indicate the in-plane projection of the polarization and the net polarization direction for the entire device structure.
ARTICLES
NATURE NANOTECHNOLOGY DOI: 10.1038/NNANO.2009.451
NATURE NANOTECHNOLOGY | VOL 5 | FEBRUARY 2010 | www.nature.com/naturenanotechnology146
© 2010 Macmillan Publishers Limited. All rights reserved.