Persistent conductive footprints of 109° domain walls in bismuth ferrite films

I. Stolichnov, M. Iwanowska, E. Colla, B. Ziegler, I. Gaponenko, P. Paruch, M. Huijben, G. Rijnders, and N.

Setter

Citation: Applied Physics Letters 104, 132902 (2014); doi: 10.1063/1.4869851

View online: http://dx.doi.org/10.1063/1.4869851
View Table of Contents: http://scitation.aip.org/content/aip/journal/apl/104/13?ver=pdfcov
Published by the AIP Publishing

Articles you may be interested in

Effects of coherent ferroelastic domain walls on the thermal conductivity and Kapitza conductance in bismuth
ferrite
Appl. Phys. Lett. 102, 121903 (2013); 10.1063/1.4798497

Three-dimensional ferroelectric domain imaging of epitaxial BiFeO 3 thin films using angle-resolved
piezoresponse force microscopy
Appl. Phys. Lett. 97, 112907 (2010); 10.1063/1.3487933

Ferroelectric domain wall pinning at a bicrystal grain boundary in bismuth ferrite

Appl. Phys. Lett. 93, 142901 (2008); 10.1063/1.2993327

Piezoelectric behavior of SrRuO 3 buffered lanthanum modified bismuth ferrite thin films grown by chemical
method
Appl. Phys. Lett. 93, 142902 (2008); 10.1063/1.2979692

Domain imaging and local piezoelectric properties of the (200)-predominant SrBi 2 Ta 2 O 9 thin film
Appl. Phys. Lett. 75, 1610 (1999); 10.1063/1.124770

This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to IP:
158.121.247.60 On: Thu, 06 Nov 2014 16:52:55
 APPLIED PHYSICS LETTERS 104, 132902 (2014)

Persistent conductive footprints of 109 domain walls in bismuth ferrite films

I. Stolichnov,1 M. Iwanowska,1 E. Colla,1 B. Ziegler,2 I. Gaponenko,2 P. Paruch,2
M. Huijben,3 G. Rijnders,3 and N. Setter1
1
Ceramics Laboratory, EPFL-Swiss Federal Institute of Technology, Lausanne 1015, Switzerland
2
DPMC-MaNEP, University of Geneva, 24 Quai Ernest Ansermet, 1211 Geneva 4, Switzerland
3
Faculty of Science and Technology and MESAþ Institute for Nanotechnology, University of Twente,
7500 AE Enschede, The Netherlands

(Received 6 February 2014; accepted 16 March 2014; published online 31 March 2014)
Using conductive and piezoforce microscopy, we reveal a complex picture of electronic transport at
weakly conductive 109 domain walls in bismuth ferrite films. Even once initial ferroelectric stripe
domains are changed/erased, persistent conductive paths signal the original domain wall position.
The conduction at such domain wall “footprints” is activated by domain movement and decays
rapidly with time, but can be re-activated by opposite polarity voltage. The observed phenomena
represent true leakage conduction rather than merely displacement currents. We propose a scenario
of hopping transport in combination with thermionic injection over interfacial barriers controlled by
C 2014 AIP Publishing LLC. [http://dx.doi.org/10.1063/1.4869851]
the ferroelectric polarization. V

Domain walls (DWs) separating regions with different ori-
entations of spontaneous electric polarization in ferroelectrics
can exhibit electronic and structural properties distinctly differ-
ent from the host material. Because of their responsiveness to
electrical and mechanical stimuli they offer tremendous poten-
tial for functional electronics,1 explored in a number of recent
studies. Among these, the groundbreaking discovery of conduc-
tivity in 180 and 109 DWs in BiFeO3 (BFO)2 was followed
by a series of works focused on their transport properties.
Beyond the initial observations of conduction through artifi-
cially created DWs,2 it was also reported in intrinsic 109
DWs3 in La-doped BFO, and later in 71 DWs.4 Enhanced DW
conductance was moreover recently observed in other ferro-
electrics like Pb(Zr,Ti)O3 (PZT)5 and hexagonal manganates,6
implying a general character of this phenomenon.
From these studies, the spectrum of proposed scenarios
explaining DW conduction remains confusingly broad. In
order to describe DW charge transport, nearly all conduction
mechanisms known for dielectrics have been suggested
including Schottky emission,2–5 Poole-Frenkel hopping,5
Fowler-Nordheim tunneling,3 and space-charge-limited cur-
rent.4,7 The physical models proposed rely on a local band
gap decrease due to polarization discontinuity,2 or on an
accumulation of charged defects, in particular, oxygen
vacancies3 as the microscopic origin of strong DW conduc-
tion. The importance of oxygen vacancies has been invoked
both in analyses of charge transport across the domain wall3
and injection through the Schottky barrier at the ferroelectric
film interface.7 The charged defects may also contribute to
dynamic effects such as enhanced conductance associated
with the DW movement/distortion.8 The sensitivity of DW
conduction to the defect chemistry is the most probable rea-
son for the strong variation in the transport properties of
BFO samples prepared by different groups, for example, in
samples with 71 DWs in which both insulating2 and con-
ductive4 behaviors have been reported.
In this Letter, we focus on 109 DWs in BFO thin films
showing weak and time-dependent conduction, quite unlike
the robust steady-state currents reported earlier. Our study of
the dynamics, decay, and recovery of this DW conduction in
relation to the polarization domain dynamics elucidates the
underlying physics and provides insight into the general pic-
ture of DW conduction in BFO.
La-doped (10%) BFO films with a thickness of 95 nm
were grown by pulsed laser deposition on DyScO3 (110) sub-
strates on top of 5 nm SrRuO3 bottom electrodes, as
described in Refs. 9 and 10. The intrinsic domain pattern of
the as-grown BFO films was dominated by 109 stripe
domains confirmed by out-of-plane and in-plane piezores-
ponse force microscopy (PFM) performed with an Asylum
Cypher AFM (Atomic Force Microscope) system.
Conductive atomic force microscopy (c-AFM) was car-
ried out using Asylec Pt/Ir-coated probes. The conductive
properties measured at 109 DWs were drastically different
from those reported earlier for similarly processed films,3
with no DW conduction observed in c-AFM scans at applied
sample bias up to þ1.5 V. At higher voltages, current signal
appeared concurrently with DW displacement, but surpris-
ingly, was measured at the original location of the DWs. An
obvious manifestation of this effect is presented in Fig. 1
where the conductive traces of electrically erased domain
walls are clearly seen. The array of 109 stripe domains
imaged in a pristine area (Figs. 1(a)–1(c)) totally disappears
during c-AFM scanning (Fig. 1(d)) with þ1.7 V applied to
the sample, as confirmed by the subsequent PFM image
(Figs. 1(e) and 1(f)). Thus, the conduction traces in Fig. 1(d)
mark the original DW locations, although the domains are no
longer present at the moment of the conduction measure-
ment. Likewise, for the wide domains observed in Fig. 1,
although they did not completely disappear after the c-AFM
scan, their shape was significantly modified. It is thus clear
from Fig. 1 (red circles) that the conduction traces corre-
sponding to the wide domains again represent the footprints
of the original DWs observed before the c-AFM scan.
Remarkably, when multiple c-AFM scans were carried
out on the same area, such conductive DW traces were only
observed during the first scan and were no longer visible in
subsequent scans.

0003-6951/2014/104(13)/132902/4/$30.00 104, 132902-1 C 2014 AIP Publishing LLC

V

This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to IP:
158.121.247.60 On: Thu, 06 Nov 2014 16:52:55
 132902-2 Stolichnov et al.
FIG. 1. Conduction traces of electrically erased DWs. (a)–(c) PFM phase
(a), amplitude (b), and topography (c) images taken on the pristine area of
the BFO film. (d) c-AFM map measured on the same area under dc sample
bias of þ1.7 V. (e) and (f) Second measurement of phase (e) and amplitude
(f) PFM maps.
Single spot measurements with a stationary AFM probe
contacting the surface at the DW position show, moreover,
that the onset of the conduction state is not instantaneous.
After applying a þ1.5–2.2 V sample bias, no current is ini-
tially detected. However, after a millisecond to minute delay,
depending on the bias, an abrupt current increase was
observed. This conduction onset was always associated with
distortion of the domain pattern. The observed current gradu-
ally decayed over some hundreds of seconds.
This unusual current dynamics is well illustrated in an in-
dependent measurement using self-patterned gold nano-
electrodes (Fig. 2), formed by evaporating a 100 Å thick Au
layer and subsequent annealing at 385  C. After annealing the
continuous Au layer transforms into 200 6 100 nm islands,
with a thickness of 50 6 20 nm (Fig. 2(a)) (for details see sup-
plementary material10). The obvious advantage of these elec-
trodes compared to the AFM probe is their uniform field
distribution, combined with contact stability and accuracy of
current measurements. After current measurements the elec-
trodes were removed by AFM scanning with a high load
force, allowing access to the BFO surface for PFM imaging.
On such an Au nano-electrode, we initially detected no
current response under a dc bias of 2 V for a period of 75 s,
followed by an abrupt increase, and then gradual decrease of
current (Fig. 2(b)). Subsequent PFM images did not show
any switching outside the electrode (Fig. 2(c)). However, af-
ter electrode removal, we observed a modified domain pat-
tern directly underneath (Fig. 1(d)). Even though the exact
moment of this change in the polarization cannot be deter-
mined, it is natural to assume that it occurred concurrently
with the conduction onset.
Although this conduction behavior mimics the transient
response of displacement currents, a closer examination
This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to IP:
158.121.247.60 On: Thu, 06 Nov 2014 16:52:55
Appl. Phys. Lett. 104, 132902 (2014)
FIG. 2. DW conduction measurements using self-patterned gold nanoelectr-
odes. (a) Electrode used for the measurements (topography). (b) I-t curve
measured on this electrode with dc sample bias of þ2 V. (c) Amplitude PFM
image of the area including the measured electrode. (d) Amplitude PFM
images of the same area after removing the electrode.
shows that its nature is different. Indeed, in the case of
switching currents, the total amount of transferred charge is
limited to that associated with the reversed polarization. For
the measured capacitor with an area 5  1010 cm2 and
reversed polarization not exceeding 100 lC/cm2, the upper
estimate of this charge is less than 5  1014 C. On the other
hand, the charge obtained by integrating the I-t curve in
Fig. 2(b) for the initial period of 25 s is >5  1010 C, i.e., at
least 4 orders of magnitudes larger than the maximal charge
associated with polarization reversal. Thus, we conclude that
the measured current represents charge transport across the
film rather than a displacement current.
The essential observations from the experiments
described above can be summarized as follows: (i) DW con-
duction occurred concurrently with switching, and thus
appears to be triggered by DW motion; (ii) conduction traces
are observed at the initial positions of DWs rather than their
actual position; (iii) DW conduction is volatile and decreases
significantly within a time scale of 1–10 min. Feature (ii)
underlines the importance of charged defects accumulated in
the original DW locations, whereas feature (i) points to the
role of spontaneous polarization, suggesting that the conduc-
tion is limited by some interface-controlled mechanism. The
triggering of DW conduction by domain motion and its sub-
sequent decay suggest that only freshly reversed polarization
lowers the Schottky barrier at the ferroelectric interface.
Within this rationale, gradual time-dependent screening of
this polarization should decrease DW conduction, in agree-
ment with the observation of its volatility (feature (iii)).
To further clarify the role of such polarization screening,
we explored the phenomenon of electrical recovery of DW
conduction. In this experiment, a series of c-AFM maps were
sequentially collected on the same area (Fig. 3), alternating
the sample bias every second image between þ2 V and
2 V. In order to monitor any possible change of domain
 132902-3 Stolichnov et al.
FIG. 3. DW conduction recovery. (a) Phase PFM map taken on the pristine
BFO area. (b) c-AFM map measured under dc sample bias of þ2 V. A frag-
ment of the PFM phase image superimposed on the conduction map shows
that the conductive traces correspond to the domain boundaries (inset). (c)
Second PFM phase image performed after the c-AFM scan. (d) Second
c-AFM scan performed under the same conditions as (b) after “recovery” scan.
(e) Third PFM phase image performed after the second c-AFM scan. (f) Third
c-AFM scan performed under the same conditions as (b) after “recovery” scan.
pattern, PFM images were taken between the c-AFM scans.
Fig. 3(a) shows the original domain pattern before any
c-AFM measurements. The first c-AFM map (Fig. 3(b)) with
þ2 V sample bias shows clear traces of DW conduction. A
fragment of the PFM phase image superimposed on the con-
duction map shows that these conduction traces correspond
to the original domain boundaries (Fig. 3(b), inset).
The map in Fig. 3(b) includes two bands where a sample
bias was applied, separated by a section where the bias was
zero. This technique enables visualization of DW distortions
in relation to the conduction. The PFM phase image taken
after the initial c-AFM map (Fig. 3(c)) shows that the domains
with an upward-oriented vertical polarization component are
broadened in the regions where DW conduction was detected.
In the experiments discussed in the first part of this
Letter, DW conduction was observed only in the first c-AFM
measurement, while subsequent scans on the same area
yielded no current. However, the situation drastically changes
if the same area is re-scanned with opposite 2 V sample
bias. Although during this scan no current is detected, it ena-
bles a recovery of the DW conduction traces for subsequent
c-AFM imaging with positive þ2 V sample bias, allowing the
conductive traces of the original DWs to be reproduced
(Fig. 3(d)). To confirm this effect, another set of measure-
ments was performed according to the same procedure that
consists of PFM mapping (Fig. 3(e)) followed by two c-AFM
This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to IP:
158.121.247.60 On: Thu, 06 Nov 2014 16:52:55
Appl. Phys. Lett. 104, 132902 (2014)
scans with 2 V and then þ2 V. Fig. 3(f) shows the last
c-AFM image (with þ2 V applied to the sample) where DW
conduction traces can be clearly seen. The comparison of the
three sets of PFM images shows that the domains with
upward-oriented polarization tend to grow, curving outwards
as their DWs move. However, the conduction traces remain
straight lines situated at the original DW locations. Their sta-
bility indicates that the measured conduction is a true leakage
rather than dielectric displacement current.
Our qualitative interpretation of the observed transport
properties of DWs is based on defect-mediated transport in
combination with polarization-controlled Schottky injection
previously analysed in Ref. 11. In agreement with previous
BFO DW studies,3,8 we assume electron conductivity, which
implies injection through the top interface under positive sam-
ple bias. In the pristine sample, the polarization is nearly fully
compensated by mobile charges trapped in the interface-
adjacent layer, and therefore influences the Schottky barrier
only weakly. For sample bias up to 2 V, this barrier keeps the
current below the noise level of our measurement. For higher
bias, the conduction increase observed concurrently with
switching (DW displacement) indicates charge injection
across the interface. Because the switching occurs via an
expansion of upward-polarized domains, the positive charge
of freshly reversed polarization lowers the barrier for electron
injection from the top interface, thus triggering conduction.
For further analysis of the interplay between polariza-
tion and conduction, we use a “passive layer” model of a
ferroelectric interface. This passive layer accounts for the
gradual decrease of ferroelectric polarization near the inter-
face and/or a nonzero polarization screening length inside
the electrode.12 It also separates the bound charge of sponta-
neous polarization from the mobile or trapped screening
charge inside the film, resulting in incomplete compensation.
This model explains a number of interface-related phenom-
ena in ferroelectrics including Schottky barrier lowering,
depolarization field in the ferroelectric film, and opposite
electric field in the interface-adjacent “passive” layer.
The effect of barrier lowering by switched polarization
is strongest immediately after a switching event, then gradu-
ally decreases due to polarization screening, leading to a
decreasing current. The evolution of DW conduction
observed in this study is depicted schematically in Fig. 4.
Initially (Fig. 4(a)) the polarization is almost fully screened,
and thus has no impact on the Schottky barrier height. In this
simplified model, the electron electrostatic energy is consid-
ered constant within the ferroelectric layer (neglecting the
built-in charge), as shown in Fig. 4(a), right panel. This state
corresponds to the initial phase of the conduction measure-
ments in which no current was detected. Once a microscopic
switching event occurs, changing the domain pattern, the
freshly switched non-screened polarization lowers the barrier
and prompts electron injection (Fig. 4(b)). At this stage, the
ferroelectric layer is exposed to the depolarization field,
hence the electrostatic energy of electrons changes across
the ferroelectric layer, and the two interfacial barriers are dif-
ferent (Fig. 4(b), right panel). Finally, the trapped charge
again screens the polarization in this area, reducing the
Schottky-barrier-lowering effect, and eventually inhibiting
the conduction. This description also clarifies the observed
 132902-4 Stolichnov et al.
FIG. 4. Effect of spontaneous polarization on the Schottky barrier at the fer-
roelectric film interface. (a) Initially, polarization is fully screened and has
no impact on Schottky barrier. Right panel: electrostatic energy of electrons
vs. depth, the built-in charge in ferroelectric layer is neglected. (b) Freshly
switched non-screened polarization lowers the Schottky barrier and prompts
the injection.
conduction recovery under opposite polarity bias, which
changes the charge balance in the interface-adjacent layer by
at least partially removing the trapped charge. Thus, the
bound charge of spontaneous polarization is “re-activated”
and lowers the Schottky barrier.
We also point out that the Schottky barrier change
occurring at the areas where the polarization was switched
does not make these areas entirely conductive: the observed
conduction traces were located only at the original DW posi-
tions. This can be understood if in addition to injection the
transport trough the film bulk is considered. A transport
mechanism mediated by defects accumulated at the original
DW locations is a plausible explanation of the observed con-
duction traces. Because of the relatively low mobility of the
defects at room temperature they remain virtually immobile
within the time scale of the experiment after the domain pat-
tern changes. Hence, they continue to provide a conductive
path at the same location, even though the actual DW has
meanwhile moved away. The nature of these defects cannot
be identified within this study. The oxygen vacancies gener-
ally considered in DW conduction models presented so far
are unlikely candidates due to their high mobility.13
However, their ordering at high enough density in perov-
skites could lead to significantly lower mobilities.14 Another
scenario involves charged dislocations accumulated at DWs,
however the issue requires further analysis.
Thus, to rationalize the entire set of observations from
our experiments, a model incorporating both defect-
mediated (Poole-Frenkel) conduction through the bulk, and
Richardson-Schottky-Simmons injection through the barrier
tuned by spontaneous polarization is required. Obviously,
this picture does not address all conductive phenomena
occurring in ferroelectric DWs. In particular, in the artificial
DWs written by the AFM tip the conduction mechanism can
be different.2 On the other hand, the present analysis brings
insight into the physics of DW transport and can be applied
to intrinsic DWs regardless of the details of conduction
behavior. Specifically, this model can be interesting for
This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to IP:
158.121.247.60 On: Thu, 06 Nov 2014 16:52:55
Appl. Phys. Lett. 104, 132902 (2014)
La-doped BFO with 109 DW, where much higher levels of
conductivity have been reported, but where a dramatic con-
duction increase was likewise observed upon DW displace-
ment.8 We note that dynamic DW conduction and its
hysteretic behavior have been previously attributed to micro-
scopic reversible distortions of DWs and the effect of sponta-
neous polarization. Our results agree with this concept and
extend it to the case of strong irreversible changes of domain
patterns at much higher length scales.
In summary, we have explored electronic transport in
weakly conductive 109 DWs in La-doped BFO thin films.
The use of removable nano-electrodes and a special proce-
dure of c-AFM data acquisition revealed unexpected features
of the DW conduction, which could be linked to local polar-
ization reversal. A complex model involving defect-
mediated hopping and polarization-controlled injection
through the top interface is shown to be consistent with the
observed conduction behavior. These results provide insight
into the physical picture of DW conduction, which can be
useful for analysis of transport properties in ferroelectrics.
From the perspective of functional electronics, the conduc-
tion traces at the original DW positions indicate a possible
way towards the formation of arrays of tunable conductive
channels, which can remain stable regardless of the evolution
of polarization domains.
Co-authors from EPFL acknowledge the support from
the Swiss National Science Foundation: Project No.
200020_132724. EU is acknowledged for the support
through the projects ERC-268058 MOBILE-W and
FP7/2007-2013 ERC Grant Agreement No. 268058 and the
STSM COST-0904 ACTION.
UniGE coauthors acknowledge the support of the Swiss
National Science Foundation through the NCCR MaNEP.
1
G. Catalan, J. Seidel, R. Ramesh, and J. F. Scott, Rev. Mod. Phys. 84(1),
119 (2012).
2
J. Seidel, L. W. Martin, Q. He, Q. Zhan, Y. H. Chu, A. Rother, M. E.
Hawkridge, P. Maksymovych, P. Yu, M. Gajek, N. Balke, S. V. Kalinin,
S. Gemming, F. Wang, G. Catalan, J. F. Scott, N. A. Spaldin, J. Orenstein,
and R. Ramesh, Nature Mater. 8(3), 229 (2009).
3
J. Seidel, P. Maksymovych, Y. Batra, A. Katan, S. Y. Yang, Q. He, A. P.
Baddorf, S. V. Kalinin, C. H. Yang, J. C. Yang, Y. H. Chu, E. K. H. Salje,
H. Wormeester, M. Salmeron, and R. Ramesh, Phys. Rev. Lett. 105(19),
197603 (2010).
4
S. Farokhipoor and B. Noheda, Phys. Rev. Lett. 107(12), 127601 (2011).
5
J. Guyonnet, I. Gaponenko, S. Gariglio, and P. Paruch, Adv. Mater.
23(45), 5377 (2011).
6
D. Meier, J. Seidel, A. Cano, K. Delaney, Y. Kumagai, M. Mostovoy,
N. A. Spaldin, R. Ramesh, and M. Fiebig, Nature Mater. 11(4), 284
7
(2012).
S. Farokhipoor and B. Noheda, J. Appl. Phys. 112(5), 052003 (2012).
8
P. Maksymovych, J. Seidel, Y. H. Chu, P. P. Wu, A. P. Baddorf, L. Q.
Chen, S. V. Kalinin, and R. Ramesh, Nano Lett. 11(5), 1906 (2011).
9
Y. H. Chu, Q. He, C. H. Yang, P. Yu, L. W. Martin, P. Shafer, and R.
Ramesh, Nano Lett. 9(4), 1726 (2009).
10
See supplementary material at http://dx.doi.org/10.1063/1.4869851 for fur-
ther experimental details.
11
P. W. Blom, R. M. Wolf, J. F. M. Cillessen, and M. P. C. M. Krijn, Phys.
Rev. Lett. 73(15), 2107 (1994).
12
A. K. Tagantsev, G. Gerra, and N. Setter, Phys. Rev. B 77(17), 174111
(2008).
13
S. Gottschalk, H. Hahn, S. Flege, and A. G. Balogh, J. Appl. Phys.
104(11), 114106 (2008).
14
J. F. Scott and M. Dawber, Appl. Phys. Lett. 76(25), 3801 (2000).