LETTERS

PUBLISHED ONLINE: 19 MAY 2013 | DOI: 10.1038/NMAT3649

Ferroelectric-field-effect-enhanced
electroresistance in metal/ferroelectric/
semiconductor tunnel junctions
Zheng Wen1,2,3 , Chen Li1,2,3 , Di Wu1,2,3 *, Aidong Li1,2,3 and Naiben Ming1,2,3
Ferroelectric tunnel junctions (FTJs), composed of two metal
electrodes separated by an ultrathin ferroelectric barrier, have
attracted much attention as promising candidates for nonvolatile resistive memories. Theoretical14 and experimental59
works have revealed that the tunnelling resistance switching in
FTJs originates mainly from a ferroelectric modulation on the
barrier height. However, in these devices, modulation on the
barrier width is very limited, although the tunnelling transmittance depends on it exponentially as well10 . Here we propose a
novel tunnelling heterostructure by replacing one of the metal
electrodes in a normal FTJ with a heavily doped semiconductor.
In these metal/ferroelectric/semiconductor FTJs, not only the
height but also the width of the barrier can be electrically
modulated as a result of a ferroelectric field effect11,12 , leading
to a greatly enhanced tunnelling electroresistance. This idea
is implemented in Pt/BaTiO3 /Nb:SrTiO3 heterostructures, in
which an ON/OFF conductance ratio above 104 , about one to
two orders greater than those reported in normal FTJs, can be
achieved at room temperature6,9,1318 . The giant tunnelling electroresistance, reliable switching reproducibility and long data
retention observed in these metal/ferroelectric/semiconductor
FTJs suggest their great potential in non-destructive readout
non-volatile memories.
In quantum mechanics, the wavefunction of an electron can
leak through a barrier as long as the barrier is sufficiently low and
sufficiently thin. The electron therefore has a finite probability of
being found on the opposite side of the barrier10 . Conventionally, a
tunnel junction consists of two metal electrodes and a nanometrethick insulating barrier layer sandwiched inside. The transmittance
of a tunnel junction depends exponentially on the height and the
width of the barrier10 . Superconductor Josephson junctions19 and
magnetic tunnel junctions20 are well-known examples that have
attracted much attention. Recently, great technological advances
have been achieved in the growth of perovskite oxide thin films.
Ferroelectricity is found to exist even in thin films of several unit
cells (u.c.) in thickness, which makes it possible to realize FTJs by
employing ultrathin ferroelectrics as barriers21,22 .
The concept of FTJs can be dated back to the early 1970s,
when Esaki et al. reported a possible FTJ using bismuth niobate as the barrier23 . However, the control mechanism of FTJs
was formulated only very recently24 . Unlike conventional tunnel
junctions, the barrier height seen by electrons in a FTJ can be
switched between a high and a low value by polarization re-

versal in the ferroelectric barrier24 . This results in an electrical

switching of tunnelling resistance, the so-called tunnelling electroresistance (TER). The electrically modulated barrier height has
been demonstrated by recent experiments59 . It has been verified
using scanning probe microscopy that resistance switching in ultrathin BaTiO3 (BTO)/SrRuO3 /SrTiO3 epitaxial heterostructures is
associated with barrier height switching in response to polarization
reversal7 . Giant TER with reliable switching has been reported in
FTJs using BTO and Pb(Zr,Ti)O3 as barriers6,9 . These achievements
suggest that the FTJ offers a promising alternative for non-volatile
resistive memories of pure electronic contribution6 . Moreover,
non-volatile switching between four tunnelling resistance states has
been realized in multiferroic tunnel junctions1318 . Most recently, it
has also been reported that a ferroelectric memristor behaviour can
be achieved in FTJs (refs 24,25).
As well as on the barrier height, the tunnelling transmittance
also depends exponentially on the barrier width according to
basic quantum mechanics10 . A greatly enhanced TER can be
expected if the barrier width can be significantly modulated
in parallel with the barrier height. The possibility of an FTJ
switching mechanism with an inherent ferroelectric modulation
of the barrier width has been proposed3 . However, such a
modulation in metal/ferroelectric/metal FTJs is limited by the short
screening length in metals. Here, we propose a novel FTJ scheme
employing a metal/ferroelectric/semiconductor heterostructure, as
shown in Fig. 1, using an n-type semiconductor as an example.
By polarization reversal in the ultrathin ferroelectric barrier, the
semiconductor surface can be switched between accumulation and
depletion of majority carriers as a result of a ferroelectric field
effect11,12 . Hence, along with the switching of barrier height in
response to the polarization reversal in the ferroelectric barrier,
there exists an additional tuning on the width of the barrier because
the tunnelling electrons have to experience an extra barrier over the
space charge region if the semiconductor surface is depleted.
As depicted schematically in Fig. 1a, if the ferroelectric polarization points to the semiconductor, positive bound charges
in the ferroelectric/semiconductor interface will drive the n-type
semiconductor surface into accumulation11,12 . The accumulated
semiconductor can be treated as a metal and the screening is then
similar to that in metal/ferroelectric/metal FTJs (refs 24). The
screening is usually incomplete and a depolarization field opposite
to the polarization develops in the ferroelectric barrier3 . This
depolarization field lowers the barrier height and generates a higher

1 National

Laboratory of Solid State Microstructures, Nanjing University, Nanjing 210093, China, 2 Department of Materials Science and Engineering,
College of Engineering and Applied Sciences, Nanjing University, Nanjing 210093, China, 3 National Center of Microstructures and Quantum Manipulation,
Nanjing University, Nanjing 210093, China. Present address: Ningxia Key Laboratory of Photovoltaic Materials, Ningxia University, Yinchuan 750021,
China. *e-mail: diwu@nju.edu.cn
NATURE MATERIALS | VOL 12 | JULY 2013 | www.nature.com/naturematerials

2013 Macmillan Publishers Limited. All rights reserved

617

NATURE MATERIALS DOI: 10.1038/NMAT3649

LETTERS
a

F
M

F
M

EF
Ec

EF
Ec

Ev

Ev

Figure 1 | Resistive switching mechanism. ad, Schematic drawings of the metal/ferroelectric/semiconductor structures and corresponding potential
energy profiles for the low (a,c) and the high (b,d) resistance states. In a,b, M, F, and S represent the metal, the ferroelectric and the semiconductor,
respectively. A heavily doped n-type semiconductor is taken as an example. In the ferroelectric barriers, the red arrows denote the polarization directions
and the large plus and minus symbols represent positive and negative ferroelectric bound charges, respectively. The plus and dot symbols in the metal
and n-type semiconductor electrodes represent holes and electrons, respectively. The circled plus symbols represent ionized donors. A rectangular
barrier, denoted by dashed lines in c,d, is assumed when the ferroelectric is unpolarized. As a simplification, we further assume that ferroelectric bound
charges at the metal/ferroelectric interface can be perfectly screened. The barrier height at the metal/ferroelectric interface is, therefore, fixed and does
not change with the polarization reversal.

tunnelling transmittance24 . The device is then set to a low resistance

ON state. However, as the polarization is reversed, pointing to the
metal electrode, as shown in Fig. 1b, the semiconductor surface is
depleted of electrons and the negative ferroelectric bound charges
have to be screened by the immobile ionized donors11,12 . In contrast
to the ON state, where majority carriers can be accumulated very
close to the ferroelectric/semiconductor interface, the immobile
screening charges in the depleted state spread over a space charge
region defined by the doping profile26 . On one hand, the incomplete
screening again produces a depolarization field, but this time
increases the barrier height24 . On the other hand, the tunnelling
electrons have to experience an extra barrier in the depleted space
charge region owing to band bending induced by ferroelectric
polarization in the barrier26 . The tunnelling transmittance can be
greatly decreased by this extra barrier. The device is then set to a
high resistance OFF state.
We calculated the conductance in the ON and the OFF states
for a model metal/ferroelectric/semiconductor FTJ (Supplementary
Information), following the method implemented by Zhuravlev
and co-workers3 . The TER value, as quantified by the calculated
ON/OFF conductance ratio, depends strongly on the OFF state
resistance, which increases abruptly when the polarization exceeds
a threshold corresponding to the critical polarization that could
bend the conduction band minimum above the Fermi level to
form the extra barrier. The TER value can reach 104 105 for
a moderate polarization of 1015 C cm2 , at least one to two
orders greater than that in metal/ferroelectric/metal FTJs (ref. 3;
Supplementary Fig. S2).
Single-crystalline 0.7wt% Nb-doped SrTiO3 (Nb:STO) is chosen
as the semiconductor substrate27 . A 7 u.c. thick BTO ultrathin film
was deposited epitaxially on (001)-oriented Nb:STO substrates by
pulsed laser deposition in a typical layer-by-layer growth mode
(Methods and Supplementary Fig. S3a). As shown in Fig. 2a, the
BTO/Nb:STO heterostructures exhibit atomically flat surfaces with
a root-mean-square roughness of approximately 0.17 nm over an
area of 55 m2 . Reciprocal space mapping around the (013) Bragg
reflection indicates coherent growth of BTO on Nb:STO substrates
(Supplementary Fig. S3b). The measured out-of-plane and in-plane
lattice constants of BTO are 0.42 and 0.39 nm, respectively. The
tetragonality (c/a ratio) of 1.08 implies a high Curie temperature
and a large ferroelectric polarization in the ultrathin BTO (refs 22,
28). Piezoresponse force microscopy (PFM) hysteresis loops shown
618

in Fig. 2b indicate the ferroelectric nature of the ultrathin BTO. The

local coercive voltages are about +2.0 and 1.0 V, as indicated by
the minima of the amplitude loop. Figure 2c,d shows the out-ofplane PFM phase and amplitude images of ferroelectric domains
written on the BTO surface. The 180 phase contrast reveals that
the polarization is antiparallel in the two domains. Clear phase
contrast can be observed during repetitive measurements for hours
(Supplementary Fig. S4). Current mapping over the two antiparallel
ferroelectric domains is shown in Fig. 3e. A larger current is
observed over the central domain with a ferroelectric polarization
pointing to the semiconductor substrate, in agreement with the
proposed mechanism in Fig. 1. Similar current mappings have
been used as essential evidence to demonstrate the polarizationdependent TER effect in FTJs (refs 5,7,8).
Pt top electrodes were deposited on BTO/Nb:STO heterostructures to form FTJs (Supplementary Fig. S5). The non-volatile
resistance switching in Pt/BTO/Nb:STO is demonstrated by the
clear hysteretic variation of the tunnelling resistance, as shown in
Fig. 3a. Positive pulses set the device to the low resistance ON state
by driving the polarization to point to the Nb:STO semiconductor,
whereas negative pulses switch the device to the high resistance OFF
state by polarization reversal. The corresponding I V characteristics are highly nonlinear in both ON and OFF states (Supplementary
Fig. S6b). Hysteresis measurements starting from the ON and the
OFF states generate identical resistance loops, demonstrating good
reproducibility. A giant ON/OFF ratio above 104 is clearly observed.
We note that the switching from the OFF state to the ON state is relatively sharp, consistent with the observation that the ON state resistance is not sensitive to the value of the effective polarization as long
as the polarization is switched downward, pointing to the semiconductor (Supplementary Fig. S2a). In contrast, the gradual switching
from the ON to the OFF state is due to the development of a depleted
space charge region on the Nb:STO surface. By applying negative
pulses with increasing amplitude, the volume of upward domains
increases by consuming downward domains under the electrode
(Supplementary Figs S7 and S8). Correspondingly, the tunnelling
resistance increases in response to the increase of effective upward
polarization pointing to the Pt electrode (Supplementary Fig. S8),
which requires widening of the space charge region to screen the
increased ferroelectric bound charges.
Figure 3b shows the conductance ratiowrite amplitude relation
of Pt/BTO/Nb:STO. The volume of switched domains and the
NATURE MATERIALS | VOL 12 | JULY 2013 | www.nature.com/naturematerials

2013 Macmillan Publishers Limited. All rights reserved

NATURE MATERIALS DOI: 10.1038/NMAT3649

LETTERS
a

F
M

F
M

EF
Ec

EF
Ec

Ev

Ev

Figure 1 | Resistive switching mechanism. ad, Schematic drawings of the metal/ferroelectric/semiconductor structures and corresponding potential
energy profiles for the low (a,c) and the high (b,d) resistance states. In a,b, M, F, and S represent the metal, the ferroelectric and the semiconductor,
respectively. A heavily doped n-type semiconductor is taken as an example. In the ferroelectric barriers, the red arrows denote the polarization directions
and the large plus and minus symbols represent positive and negative ferroelectric bound charges, respectively. The plus and dot symbols in the metal
and n-type semiconductor electrodes represent holes and electrons, respectively. The circled plus symbols represent ionized donors. A rectangular
barrier, denoted by dashed lines in c,d, is assumed when the ferroelectric is unpolarized. As a simplification, we further assume that ferroelectric bound
charges at the metal/ferroelectric interface can be perfectly screened. The barrier height at the metal/ferroelectric interface is, therefore, fixed and does
not change with the polarization reversal.

tunnelling transmittance24 . The device is then set to a low resistance

ON state. However, as the polarization is reversed, pointing to the
metal electrode, as shown in Fig. 1b, the semiconductor surface is
depleted of electrons and the negative ferroelectric bound charges
have to be screened by the immobile ionized donors11,12 . In contrast
to the ON state, where majority carriers can be accumulated very
close to the ferroelectric/semiconductor interface, the immobile
screening charges in the depleted state spread over a space charge
region defined by the doping profile26 . On one hand, the incomplete
screening again produces a depolarization field, but this time
increases the barrier height24 . On the other hand, the tunnelling
electrons have to experience an extra barrier in the depleted space
charge region owing to band bending induced by ferroelectric
polarization in the barrier26 . The tunnelling transmittance can be
greatly decreased by this extra barrier. The device is then set to a
high resistance OFF state.
We calculated the conductance in the ON and the OFF states
for a model metal/ferroelectric/semiconductor FTJ (Supplementary
Information), following the method implemented by Zhuravlev
and co-workers3 . The TER value, as quantified by the calculated
ON/OFF conductance ratio, depends strongly on the OFF state
resistance, which increases abruptly when the polarization exceeds
a threshold corresponding to the critical polarization that could
bend the conduction band minimum above the Fermi level to
form the extra barrier. The TER value can reach 104 105 for
a moderate polarization of 1015 C cm2 , at least one to two
orders greater than that in metal/ferroelectric/metal FTJs (ref. 3;
Supplementary Fig. S2).
Single-crystalline 0.7wt% Nb-doped SrTiO3 (Nb:STO) is chosen
as the semiconductor substrate27 . A 7 u.c. thick BTO ultrathin film
was deposited epitaxially on (001)-oriented Nb:STO substrates by
pulsed laser deposition in a typical layer-by-layer growth mode
(Methods and Supplementary Fig. S3a). As shown in Fig. 2a, the
BTO/Nb:STO heterostructures exhibit atomically flat surfaces with
a root-mean-square roughness of approximately 0.17 nm over an
area of 55 m2 . Reciprocal space mapping around the (013) Bragg
reflection indicates coherent growth of BTO on Nb:STO substrates
(Supplementary Fig. S3b). The measured out-of-plane and in-plane
lattice constants of BTO are 0.42 and 0.39 nm, respectively. The
tetragonality (c/a ratio) of 1.08 implies a high Curie temperature
and a large ferroelectric polarization in the ultrathin BTO (refs 22,
28). Piezoresponse force microscopy (PFM) hysteresis loops shown
618

in Fig. 2b indicate the ferroelectric nature of the ultrathin BTO. The

local coercive voltages are about +2.0 and 1.0 V, as indicated by
the minima of the amplitude loop. Figure 2c,d shows the out-ofplane PFM phase and amplitude images of ferroelectric domains
written on the BTO surface. The 180 phase contrast reveals that
the polarization is antiparallel in the two domains. Clear phase
contrast can be observed during repetitive measurements for hours
(Supplementary Fig. S4). Current mapping over the two antiparallel
ferroelectric domains is shown in Fig. 3e. A larger current is
observed over the central domain with a ferroelectric polarization
pointing to the semiconductor substrate, in agreement with the
proposed mechanism in Fig. 1. Similar current mappings have
been used as essential evidence to demonstrate the polarizationdependent TER effect in FTJs (refs 5,7,8).
Pt top electrodes were deposited on BTO/Nb:STO heterostructures to form FTJs (Supplementary Fig. S5). The non-volatile
resistance switching in Pt/BTO/Nb:STO is demonstrated by the
clear hysteretic variation of the tunnelling resistance, as shown in
Fig. 3a. Positive pulses set the device to the low resistance ON state
by driving the polarization to point to the Nb:STO semiconductor,
whereas negative pulses switch the device to the high resistance OFF
state by polarization reversal. The corresponding I V characteristics are highly nonlinear in both ON and OFF states (Supplementary
Fig. S6b). Hysteresis measurements starting from the ON and the
OFF states generate identical resistance loops, demonstrating good
reproducibility. A giant ON/OFF ratio above 104 is clearly observed.
We note that the switching from the OFF state to the ON state is relatively sharp, consistent with the observation that the ON state resistance is not sensitive to the value of the effective polarization as long
as the polarization is switched downward, pointing to the semiconductor (Supplementary Fig. S2a). In contrast, the gradual switching
from the ON to the OFF state is due to the development of a depleted
space charge region on the Nb:STO surface. By applying negative
pulses with increasing amplitude, the volume of upward domains
increases by consuming downward domains under the electrode
(Supplementary Figs S7 and S8). Correspondingly, the tunnelling
resistance increases in response to the increase of effective upward
polarization pointing to the Pt electrode (Supplementary Fig. S8),
which requires widening of the space charge region to screen the
increased ferroelectric bound charges.
Figure 3b shows the conductance ratiowrite amplitude relation
of Pt/BTO/Nb:STO. The volume of switched domains and the
NATURE MATERIALS | VOL 12 | JULY 2013 | www.nature.com/naturematerials

2013 Macmillan Publishers Limited. All rights reserved

NATURE MATERIALS DOI: 10.1038/NMAT3649

a

LETTERS
b

200
1.00

Phase ()

nm

0
0.10

Amplitude (a.u.)

100

100
1 m
200

0.01

Voltage (V)

500 nm

180

()

a.u.

30
pA

Figure 2 | Morphology and ferroelectricity of ultrathin BTO and correlation between polarization reversal and resistance switching. a, Morphology of a
7 u.c. BTO surface on Nb:STO semiconductor substrates. b, Local PFM hysteresis loops: top, phase signal; bottom, amplitude signal. c,d, PFM out-of-plane
phase (c) and amplitude (d) images recorded after writing an area of 3 3 m2 with 5 V and then the central 1.5 1.5 m2 square with +5 V using a
biased conductive tip. e, Current mapping over the same area shown in c,d, demonstrating the polarization direction-dependent resistance in the
BTO/Nb:STO heterostructure.

effective polarization in the barrier increases with increasing write

amplitude. Therefore, Fig. 3b can be regarded as an experimental
relation between the conductance ratio and the effective polarization, which is qualitatively in agreement with the calculated relation
from the resistance switching scenario proposed (Supplementary
Fig. S2b). At low write amplitudes, the conductance ratio is close to
unity, because neither the positive nor the negative pulse is strong
enough to polarize the ferroelectric barrier. At about 1.5 V, close
to the coercive voltage, the barrier starts to reorient in response
to write pulses. The ON state is established by the positive write
pulse and an abrupt increase of the conductance ratio is observed.
However, as noticed in Fig. 3a, the ON state resistance saturates
with increasing positive write amplitude, whereas the OFF state
resistance increases exponentially with increasing negative write
amplitude. Therefore, the conductance ratio increases continuously
with further increasing write amplitude.
It is also observed that the TER of Pt/BTO/Nb:STO increases
with decreasing Nb doping concentration in the substrate, as shown
in Fig. 3c (see also Supplementary Fig. S6). This is expected because
a wider space charge region (hence, a wider extra tunnelling barrier)
should be formed in semiconductor substrates with a lower doping
concentration to screen the ferroelectric bound charges in the
OFF state. The resistance switching characteristics of Pt/SrTiO3 (7
u.c.)/Nb:STO and Pt/Nb:STO heterostructures, where there is
no ferroelectric barrier, have also been measured. The observed
resistance change is less than five times (Supplementary Fig. S9),

in contrast to over 104 times observed in Pt/BTO/Nb:STO. This

clearly demonstrates the indispensable function of the ferroelectric
barrier on the observed TER. It is the creation and elimination
of the extra tunnelling barrier due to the switching between
depletion and accumulation of the semiconductor in response to
spontaneous polarization reversal that result in the observed great
enhancement of TER.
The existence of the extra barrier in the OFF state is demonstrated by capacitance measurements (Supplementary Fig. S10).
The Pt/BTO/Nb:STO heterostructure can be regarded as the capacitor of the BTO barrier connected in series with the capacitor of
the space charge region in the Nb:STO semiconductor. When the
semiconductor surface is accumulated, the ON state capacitance
of the space charge region is negligible. The measured capacitance
of the OFF state should be smaller than that of the ON state
owing to the existence of the space charge region capacitance.
This reduction of capacitance in the OFF state is indeed observed,
indicating the ferroelectric modulation on the width of the space
charge region in Nb:STO.
The ON and OFF resistance values, read at +0.1 V, of 20
different Pt/BTO/Nb:STO devices are shown statistically in Fig. 3d.
The ON/OFF ratio fluctuates around 104 , with an average value
of 12,800. These suggest good reproducibility and uniformity of
the Pt/BTO/Nb:STO FTJs. Data retention and switching properties
are the most important reliability issues associated with ferroelectric memory devices. Figure 3e shows the retention property of

NATURE MATERIALS | VOL 12 | JULY 2013 | www.nature.com/naturematerials

2013 Macmillan Publishers Limited. All rights reserved

619

NATURE MATERIALS DOI: 10.1038/NMAT3649

LETTERS
b

Resistance ()

106

104

105
104

OFF
ON

103

103

102

101

100
2

1
0
1
Voltage (V)
ON
OFF

105
103

105
104

105
103

104
103

2 4 6 8 10 12 14 16 18 20
Sample number

0.5 1.0 1.5 2.0 2.5 3.0 3.5

Write amplitude (V)

103
0.0 0.2 0.4 0.6 0.8 1.0
Nb concentration (wt%)

107
106

101

ON/OFF ratio

10 years

OFF
As-written
After 24 h

101

OFF
105

Resistance ()

107

Current (A)

Resistance ()

109

Resistance ()

104

w
5

105

wr

OFF

ON

wr

ON/OFF ratio

107

Conductance ratio

ON
104
OFF
107
0.6 0.3
0
0.3 0.6
Voltage (V)
ON
101

103

105
Time (s)

104

103

107

109

ON
100

101

102
Cycles

103

Figure 3 | TER of Pt/BTO/Nb:STO FTJs at room temperature. a, Resistance hysteresis loops. The loop composed of dark squares (green circles) is
measured using the pulse train shown schematically in the bottom-left (top-right) inset. The device is preset to the ON (OFF) state by a positive (negative)
3.6 V pulse. The corresponding domain evolution is shown schematically in the bottom-right, top-left and middle-left insets for the ON, OFF and
intermediate (ON OFF) states, respectively. The testing pulse train is composed of write pulses (w) following a triangular profile between +3.6 and
3.6 V and read pulses (r) of +0.1 V following each write pulse. b, Conductance ratio as a function of write amplitude. The measurement starts from an
unpolarized state. At a certain write amplitude, a pair of write pulses with opposite polarity is applied, as illustrated in the inset. The data points are the
ratio between tunnelling currents collected by +0.1 V read (r) pulses after the positive and negative write (w) pulses. c, ON/OFF ratio read at +0.1 V as a
function of Nb doping concentration in Nb:STO. The data points are averaged over 20 devices for each Nb concentration. d, ON and OFF state resistances
(upper panel) and the corresponding ON/OFF ratios (bottom panel) of 20 different devices. e, Retention properties. The inset shows currentvoltage
curves collected before and after 24 h retention measurement. f, Bipolar resistance switching properties.

Pt/BTO/Nb:STO FTJs. There is no significant reduction in the

resistance contrast within 24 h. The inset shows identical I V
curves in both ON and OFF states, before and after a 24-h retention
time. An ON/OFF ratio of over three orders of magnitude can
still be retained by extrapolating the retention data to over ten
years. These suggest the excellent resistance stability of the proposed
metal/ferroelectric/semiconductor FTJs during repetitive readouts
and long-time retention. Recent advances have shown that data retention in ultrathin ferroelectrics can be further improved by interface engineering29 . The switching characteristic of Pt/BTO/Nb:STO
FTJs was tested by switching the ferroelectric barrier repeatedly
with 2.5 V pulses. As shown in Fig. 3f, typical bipolar resistance
switching with good reproducibility is achieved. The ON/OFF ratio
of 100 is maintained over 3,000 write/read cycles. Cycling with
3.5 V bipolar pulses could switch the devices with an ON/OFF ratio above 10,000. However, it leads to peeling of the Pt top electrodes
after several tens of cycles. However, if 2.2 V pulses are applied, an
ON/OFF ratio of approximately 40 can be maintained to more than
10,000 cycles (Supplementary Fig. S11). These factors indicate that
there is still tremendous scope to improve the device performance.
In summary, we have proposed a metal/ferroelectric/
semiconductor heterostructure that may greatly enhance the
non-volatile memory effect by employing tunnelling through a
ferroelectric barrier. The greatly enhanced TER is ascribed to the
creation/elimination of an extra barrier on the semiconductor
620

surface in response to the polarization reversal in the ferroelectric barrier. The enhanced TER, long retention and good
switching reproducibility make the proposed scheme a promising
candidate for non-volatile resistive memories. If a ferromagnetic
semiconductor and/or a multiferroic barrier are used, the idea
presented here can be readily extended to multiferroic tunnelling
devices, where four or more non-volatile states with high resistance contrast may be realized in a single memory node1318 .
Moreover, the continuously tunable OFF state resistance observed
here actually functions as a memristor and suggests potential
applications of metal/ferroelectric/semiconductor heterostructures
in memristive devices24,25 .

Methods
BTO thin films were epitaxially grown on (001) single-crystalline Nb:STO (0.7wt%
Nb) substrates by pulsed laser deposition using a KrF ( = 248 nm) excimer laser
(Coherent COMPexPro 205). The films were deposited with 2.5 J cm2 laser energy
density at 2 Hz repetition, keeping the substrate temperature at 750 C and the
O2 pressure at 5 103 mbar. The thickness of BTO was controlled at 7 u.c. by
counting the intensity oscillations of the specular spot in the reflective high-energy
electron diffraction pattern.
The crystal structure of the ultrathin BTO was characterized using
high-resolution X-ray diffraction at the BL14B1-XRD beam line of Shanghai
Synchrotron Radiation Facility. The surface morphology, local ferroelectric
properties and current mapping were measured using an Asylum Research
Cypher scanning probe microscope. Olympus AC240TM Pt/Ti-coated silicon
cantilevers were adopted in the PFM measurements. Hysteresis loops were
NATURE MATERIALS | VOL 12 | JULY 2013 | www.nature.com/naturematerials

2013 Macmillan Publishers Limited. All rights reserved

NATURE MATERIALS DOI: 10.1038/NMAT3649

collected in the DART (dual a.c. resonance tracking) mode with a triangular
pulse of 5.0 V in amplitude applied to the tip. Phase and amplitude images
were recorded in single-frequency PFM mode. Current mapping was recorded
using conductive-diamond-coated silicon cantilevers (CDT-NCHR, NanoWorld)
biased at +0.5 V.
Pt top electrodes of 30 m in diameter and 300 nm in thickness were
sputter-deposited with a shadow mask. A Keithley 2400 SourceMeter was used
to measure the I V characteristics and resistance switching properties of the
FTJs. The electrical measurements were performed on a LakeShore CRX-4K
probe station by positioning BeCu probes on the electrodes. The testing pulses
were applied to the Pt electrodes and the Nb:STO substrate was always grounded
through an indium ohmic contact pad. Both the write and the read pulses
were 150 s in width.

Received 30 August 2012; accepted 5 April 2013; published online

19 May 2013

References
1. Tsymbal, E. Y. & Kohlstedt, H. Tunneling across a ferroelectric. Science 313,
181183 (2006).
2. Kohlstedt, H., Pertsev, N. A., Contreras, J. R. & Waser, R. Theoretical
currentvoltage characteristics of ferroelectric tunnel junctions. Phys. Rev. B
72, 125341 (2005).
3. Zhuravlev, M. Y., Sabirianov, R. F., Jaswal, S. S. & Tsymbal, E. Y. Giant
electroresistance in ferroelectric tunnel junctions. Phys. Rev. Lett. 94,
246802 (2005).
4. Pantel, D. & Alexe, M. Electroresistance effects in ferroelectric tunnel barriers.
Phys. Rev. B 82, 134105 (2010).
5. Garcia, V. et al. Giant tunnel electroresistance for non-destructive readout of
ferroelectric states. Nature 460, 8184 (2009).
6. Chanthbouala, A. et al. Solid-state memories based on ferroelectric tunnel
junctions. Nature Nanotech. 7, 101104 (2012).
7. Gruverman, A. et al. Tunneling electroresistance effect in ferroelectric tunnel
junctions at the nanoscale. Nano Lett. 9, 35393543 (2009).
8. Crassous, A. et al. Giant tunnel electroresistance with PbTiO3 ferroelectric
tunnel barriers. Appl. Phys. Lett. 96, 042901 (2010).
9. Pantel, D. et al. Tunnel electroresistance in junctions with ultrathin ferroelectric
Pb(Zr0.2 Ti0.8 )O3 barriers. Appl. Phys. Lett. 100, 232902 (2012).
10. Griffiths, D. Introduction to Quantum Mechanics (Pearson Prentice Hall, 2005).
11. Mathews, S., Ramesh, R., Venkatesan, T. & Benedetto, J. Ferroelectric field
effect transistor based on epitaxial perovskite heterostructures. Science 276,
238240 (1997).
12. Miller, S. L. & McWhorter, P. J. Physics of the ferroelectric nonvolatile memory
field effect transistor. J. Appl. Phys. 72, 59996010 (1992).
13. Garcia, V. et al. Ferroelectric control of spin polarization. Science 327,
11061110 (2010).
14. Gajek, M. et al. Tunnel junctions with multiferroic barriers. Nature Mater. 6,
296302 (2007).
15. Pantel, D., Goetze, S., Hesse, D. & Alexe, M. Reversible electrical switching
of spin polarization in multiferroic tunnel junctions. Nature Mater. 11,
289293 (2012).
16. Hambe, M. et al. Crossing an interface: Ferroelectric control of tunnel
currents in magnetic complex oxide heterostructures. Adv. Fun. Mater. 20,
24362441 (2010).

LETTERS
17. Yin, Y. W. et al. Coexistence of tunneling magnetoresistance and
electroresistance at room temperature in La0.7 Sr0.3 MnO3 /(Ba,Sr)TiO3 /
La0.7 Sr0.3 MnO3 multiferroic tunnel junctions. J. Appl. Phys. 109,
07D915 (2011).
18. Yin, Y. W. et al. Enhanced tunneling electroresistance effect due to a
ferroelectrically induced phase transition at a magnetic complex oxide
interface. Nature Mater. 12, 397402 (2013).
19. Giaever, I. Energy gap in superconductors measured by electron tunneling.
Phys. Rev. Lett. 5, 147148 (1960).
20. Tedrow, P. M. & Meservey, R. Spin-dependent tunneling into ferromagnetic
nickel. Phys. Rev. Lett. 26, 192195 (1971).
21. Fong, D. D. et al. Ferroelectricity in ultrathin perovskite films. Science 304,
16501653 (2004).
22. Kim, Y. S. et al. Critical thickness of ultrathin ferroelectric BaTiO3 films.
Appl. Phys. Lett. 86, 102907 (2005).
23. Esaki, L., Laibowitz, R. B. & Stiles, P. J. Polar switch. IBM Tech. Discl. Bull. 13,
21612162 (1971).
24. Chanthbouala, A. et al. A ferroelectric memristor. Nature Mater. 11,
860864 (2012).
25. Kim, D. J. et al. Ferroelectric tunnel memristor. Nano Lett. 12,
56975702 (2012).
26. Sze, S. M. & Ng, K. K. Physics of Semiconductor Devices 3rd edn, Ch. 4
(Wiley, 2007).
27. Takizawa, M. et al. Angle-resolved photoemission study of Nb-doped SrTiO3 .
Phys. Rev. B 79, 113103 (2009).
28. Choi, K. J. et al. Enhancement of ferroelectricity in strained BaTiO3 thin films.
Science 306, 10051009 (2004).
29. Lu, H. et al. Enhancement of ferroelectric polarization stability by interface
engineering. Adv. Mater. 24, 12091216 (2012).

Acknowledgements
This work was jointly sponsored by State Key Program for Basic Research of China
(2009CB929503), Natural Science Foundation of China (51222206 and 91022001), and
Natural Science Foundation of Jiangsu Province (BK2012016). Shanghai Synchrotron
Radiation Facility is greatly acknowledged for providing the beam time and technical
assistance. The authors would like to thank J. Yao and D. Beck from Asylum Research for
valuable advice on PFM measurements.

Author contributions
D.W. conceived this work and conducted the calculations. D.W. and Z.W. designed
the experiment. Z.W. fabricated the devices and performed the measurements.
Z.W., C.L., D.W., A.L. and N.M. analysed the data. Z.W. and D.W. wrote the
manuscript. D.W. and N.M. directed the project. All authors discussed the data and
contributed to the manuscript.

Additional information
Supplementary information is available in the online version of the paper. Reprints and
permissions information is available online at www.nature.com/reprints. Correspondence
and requests for materials should be addressed to D.W.

Competing financial interests

The authors declare no competing financial interests.

NATURE MATERIALS | VOL 12 | JULY 2013 | www.nature.com/naturematerials

2013 Macmillan Publishers Limited. All rights reserved

621