Schottky barrier height switching in thin metal oxide films studied in diode and solar
cell device configurations
Dahvyd Wing, Avner Rothschild, and Nir Tessler
Citation: Journal of Applied Physics 118, 054501 (2015); doi: 10.1063/1.4927839
View online: http://dx.doi.org/10.1063/1.4927839
View Table of Contents: http://scitation.aip.org/content/aip/journal/jap/118/5?ver=pdfcov
Published by the AIP Publishing
Articles you may be interested in
(Sr,Ba)(Si,Ge)2 for thin-film solar-cell applications: First-principles study
J. Appl. Phys. 115, 203718 (2014); 10.1063/1.4880662
Novel attributes in modeling and optimizing of the new graphene based InxGa1−xN Schottky barrier solar cells
J. Appl. Phys. 115, 194506 (2014); 10.1063/1.4878158
Transport mechanisms and effective Schottky barrier height of ZnO/a-Si:H and ZnO/μc-Si:H heterojunction solar
cells
J. Appl. Phys. 114, 184505 (2013); 10.1063/1.4831661
Epitaxial diodes of a half-metallic ferromagnet on an oxide semiconductor
J. Appl. Phys. 95, 7324 (2004); 10.1063/1.1669255
Measurements and modelling of the barrier heights and ideality factors in the metal/conducting polymer
composite Schottky device
J. Appl. Phys. 85, 3671 (1999); 10.1063/1.369732
[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:
132.68.239.10 On: Sun, 13 Sep 2015 08:06:03
JOURNAL OF APPLIED PHYSICS 118, 054501 (2015)
Schottky barrier height switching in thin metal oxide films studied
in diode and solar cell device configurations
Dahvyd Wing,1,2 Avner Rothschild,2 and Nir Tessler1,a)
1
Zisapel Nano-Electronic Center, Department of Electrical Engineering, Technion—Israel Institute
of Technology, Haifa 32000, Israel
2
Department of Materials Science and Engineering, Technion—Israel Institute of Technology,
Haifa 32000, Israel
(Received 20 April 2015; accepted 21 July 2015; published online 3 August 2015)
We investigate changes in the properties of 8 nm thin metal oxide (Nb-doped SrTiO3) films in
response to relatively high voltage or light soaking under three Sun excitation. We measure the
current–voltage behavior of metaljinsulatorjmetal type diodes and use a device model to relate
changes in device behavior to the metal oxide film properties. We find that the device’s resistive
switching is mainly associated with shifts (switching) of the metal oxide work function between
high and low injection barrier states. The method presented here can be used for in situ monitoring
of the contact work function and for quantifying the uniformity of this value across the device. We
also discuss the effect of non-uniform work function on the apparent diode’s ideality factor.
C 2015 AIP Publishing LLC. [http://dx.doi.org/10.1063/1.4927839]
V
I. INTRODUCTION
Thin metal oxide films are finding use in a range of
applications such as contacts for light emitting diodes, contacts for solar cells, and as memristor elements. Certain
metal oxides can display a several orders of magnitude
change in their resistance. This phenomenon is utilized in
memristors, and there is a vast effort to try to decipher the
mechanism causing this phenomenon; however, no clear
picture has yet evolved.1,2 When those same metal oxides
are used as device contacts, a change in their properties
during device operation would typically be considered as a
device instability that is best avoided. One salient example
of such an instability is the light soaking phenomenon in
organic photovoltaic solar cells, which use a transparent conductive oxide (TCO) as their cathode.3–6 However, numerous
reports show that these devices suffer from an s-bend kink in
their current-voltage behavior, strongly reducing the fill factor of these devices and their power conversion efficiency.
Several studies have linked this s-bend kink to a combination
of the TCO having a larger work function than expected3,6,7
and having poor conductivity.8,9 Exposing these solar cells
to light for a period of time ranging from a few minutes up
to several hours, a process known as light soaking, causes
the s-bend kink in the device behavior to disappear due to a
photo-induced change in the TCO layer. Several explanations of the photo-induced changes in the oxide have been
proposed. One of the most cited being the release of oxygen
from the surface of the TCO by UV light, thereby increasing
the TCO’s electron carrier concentration.5,6 Another explanation is the filling of surface or bulk electronic traps in the
TCO by photoexcited charge carriers.3,7,10
While the instability of a metal oxide contact would
typically be undesirable, Yalon et al.11 were able to make
a)
Author to whom correspondence should be addressed. Electronic mail:
nir@ee.technion.ac.il
0021-8979/2015/118(5)/054501/7/$30.00
inferences about the switching phenomena in a metal oxide
by carefully studying the performance of bipolar transistors
that used the metal oxide as part of the contact stack. In their
work, the configuration of the bipolar transistors was [metalmetal oxidejsemiconductor]. Similarly, Kim et al.3 used a
[metaljmetal oxide] Schottky junction to study how resistive
switching affected the metal oxide interface. In this
paper, we suggest the use of a structure resembling the
[metaljinsulatorjmetal] (MIM) diode, i.e., [metaljmetal
oxidejorganic semiconductorjmetal]. We analyze the
current–voltage characteristics of a range of devices following either the application of switching-voltage or light soaking. We find that the change in the device resistance can
largely be attributed to reversible changes in the metal oxide
work function. Using a semiconductor device model, we
show that one can extract not only the effective contact barrier height but, to some extent, also the spatial nonuniformity
of the contact or the area fraction that switches. We believe
that this constitutes a method of characterizing the relative
work function of a deep buried contact interface in a
complete device that would be inaccessible to conventional
techniques such as the Kelvin probe and photoemission
spectroscopy.
II. EXPERIMENTAL
The Nb-doped SrTiO3 (SrTiO3:Nb) films used in this
study were made by pulsed laser deposition (PLD). First, a
target of the proper stoichiometry, SrTi.98Nb.02O3, was prepared using the conventional mixed-oxide solid state reaction route. SrCO3 (99.994% purity), TiO2 (99.995% purity),
and Nb2O5 (99.9985% purity) powders from Alfa Aesar
were mixed in the proper amounts to obtain the desired Nb
content of 2% of the B-site cations. After 24 h of ball milling, the powder was calcined at 1200 C for 16 h in air. The
calcined powder was then pressed into a pellet and then
118, 054501-1
C 2015 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:
132.68.239.10 On: Sun, 13 Sep 2015 08:06:03
054501-2
Wing, Rothschild, and Tessler
isostatically pressed for 4 min at 200 MPa. The pellet was
then sintered at 1450 C for 12 h.
SrTiO3:Nb films were deposited on commercial FTOcoated glass substrates (Solaronix, 15 X=ⵧ) via a PLD workstation from Surface Gmbh. The FTO substrates were heated
to 500 C in vacuum in which O2 gas was pumped in, creating
an oxygen pressure of 2 mTorr. The laser fluence was
0.78 J/cm2. Approximately 8 nm of SrTiO3:Nb films was deposited on the FTO-coated substrates, as determined by a
Dektak 150 surface profilometer. For such processes, ITO
coated glass is not adequate as the ITO is not stable at 500 C.
Electron only MIM type devices were made with the following geometry: FTO/SrTiO3:Nb/PCBM/Ca/Al/Au. A 25
mg/ml solution of [6,6]-phenyl-C61 butyric acid methyl ester
(PCBM) (purchased from Nano-C) in chloroform was spin
coated onto FTO/SrTiO3:Nb substrates at 1500 rpm for 80 s.
This created a 130 nm thick layer of PCBM, as determined by
surface profilometry. After the film dried, Ca, Al, and Au
layers were thermally evaporated onto the samples. 20 nm of
Ca was deposited at a rate of 0.3 nm/s. 100 nm of Al was deposited at a rate of 0.3 nm/s. Finally, 100 nm of Au was deposited at a rate of 0.36 nm/s. A shadow mask was used to pattern
the top electrode in order to define a device area of 1.5 mm2.
Hole only MIM type devices were made with the
following geometries: FTO/SrTiO3:Nb/P3HT/MoO3/Au and
FTO/P3HT/MoO3/Au. A 95 nm thick layer of poly(3-hexylthiophene) (P3HT) was created by spin coating a 25 mg/ml
solution of regio-regular P3HT (Rieke Metals) in (1,2)
dicholorobenzene onto the substrates at 1000 rpm for 100 s.
10 nm of MoO3 (99.95% purity from Testbourne) was thermally evaporated at a rate of 0.005 nm/s onto the samples.
240 nm of gold was then evaporated at a rate of 0.01 nm/s.
FTO/SrTiO3:Nb/P3HT:PCBM/MoO3/Au bulk heterojunction organic solar cells were fabricated by spin coating a
20 mg/ml solution of P3HT and PCBM in a 1:1 weight ratio
in (1,2) dichlorobenzene onto FTO/SrTiO3:Nb substrates
at 1000 rpm for 100 s. This created a 150 nm thick
P3HT:PCBM film. 10 nm of MoO3 and 150 nm of gold was
thermally evaporated using the same rates described for the
hole only MIM-type devices. The samples were then vacuum
annealed for 4 min at 130 C.
J. Appl. Phys. 118, 054501 (2015)
The samples were maintained in a N2 environment at all
times. Current–voltage measurements were taken with a
Keithley 2400 source/meter. The light source for the illuminated current-voltage measurements of the solar cells and the
light soaking treatment of the MIM-type devices was performed by an SS150W Class AAA solar simulator from
ScienceTech Inc. The solar simulator used a Xe arc lamp and
an AM1.5G air mass filter. It was calibrated by a broadband
thermopile detector by Gentec-EO so that it illuminated the
solar cells with the power equivalent of 1 Sun (100 W/cm2).
During the light soaking treatment, the devices were brought
closer to the light source so that they experienced the power
equivalent of 3 Suns.
III. RESULTS
The first device we report on is an electron only MIM
type device having the following layer structure: FTO/
SrTiO3:Nb/PCBM/Ca/Al/Au. With the workfunction of
SrTiO3:Nb and Ca positioned at 3.9 eV and 2.8 eV, the
position of the PCBM HOMO at 6 eV below the vacuum
level ensures that no holes can be injected in this device configuration, while PCBM’s LUMO at 3.5 eV allows electrons to be injected into the device; hence, the device is a
single carrier, electron only device. When the current–voltage measurements were restricted to the range of 62 V, the
curves were stable and repeatable, and no pronounced hysteresis was observed. When a voltage of 64 V was applied,
the diode resistance switched between low and high resistance states. Figure 1(a) shows the J-V curves where the bias
is such that the electrons are injected from the SrTiO3:Nb
contact. The numbers next to the lines depict the sequence of
measurements between which a switching voltage (64 V)
was applied. The plot is in a log-log scale, but the slope in
all the curves is close to 1, indicating resistor like behavior
where the resistance changes by about 2 orders of magnitude.
The origin of this resistance change is, however, not clear. In
Figure 1(b), we plot the J-V curves measured for the opposite
voltage bias such that electrons were injected from the Ca
contact. Examining this set of curves, we see diode behavior
with the leakage current changing between the different
curves.
FIG. 1. (a) Current density–voltage
measurements
for
a
SrTiO3:
Nb–PCBM-Ca electron only device
where SrTiO3:Nb is the injecting contact. The sequence of the measurements is depicted by the numbers on
the figure, and the switching between
high and low conductance was
achieved by applying a high switching
voltage in the 4–5 V range (either
positive or negative). (b) The current
density–voltage measurements for the
opposite voltage bias during the same
sequence of measurements are shown
in (a); here, Ca is the electron injecting
contact.
[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:
132.68.239.10 On: Sun, 13 Sep 2015 08:06:03
054501-3
Wing, Rothschild, and Tessler
We defer the explanation of the switching phenomena
and the apparent leakage currents to the analysis section, and
we continue to present other experimental data that can help
shed light on the topic at question.
The second type of device we constructed was a hole
only device, which had one of two different configurations.
The first one was FTO/SrTiO3:Nb/P3HT/MoO3/Au and the
second one omitted the SrTiO3:Nb layer so that the device
configuration was FTO/P3HT/MoO3/Au. In both cases, the
semiconductor (P3HT) had a thickness of about 100 nm. The
position of the P3HT LUMO at 3 eV below the vacuum
level ensures that no electrons can be injected in this device
configuration, while the HOMO at 4.8 eV enabled the
injection of holes so that the device was a hole only device.
In Figure 2(a), we plot both the forward and reverse characteristics of the two diodes. The dashed line is for injection
from the bottom contact (FTO or SrTiO3:Nb) and the solid
line is for injection from the MoO3 contact. AFM scans of
the FTO substrates confirmed that they contained no spikes
that could introduce leakage currents. The presence of an
asymmetry between forward and reverse currents for MIM
type devices has been attributed to the presence of a built in
field.12,13 This is to be expected for the SrTiO3:Nb–MoO3
electrode pair since their work functions are reported to be
3.9 eV (Refs. 14 and 15) and 6.9 eV,16 respectively. The lack
of an asymmetry in the FTO–MoO3 device and its presence
in the SrTiO3:Nb-MoO3 device strongly suggests that the
work function of the bottom contact is a dominant factor in
determining the J-V curves of these devices. The difference
between these two devices resembles the differences shown
in Figure 1 suggesting that the work function of the bottom
contact may have changed when switching voltages were
applied. In Figure 2(b), we plot the initial J-V curves of the
FTO/SrTiO3:Nb/P3HT/MoO3/Au device in the dark (1) and
the J-V curves of the device in the dark after prolonged light
soaking under simulated solar radiation (2). Light soaking
had a unidirectional effect resulting in lower currents that
would recover after prolonged periods in the dark. We note
that in this device configuration light soaking results in
reduced “leakage” currents or enhanced effective resistance.
The last set of experimental data we present is that of a
strong light soaking improvement in bulk heterojunction organic photovoltaic cells made from a photo-absorbing layer
J. Appl. Phys. 118, 054501 (2015)
of P3HT blended with PCBM sandwiched between a TCO
cathode of SrTiO3:Nb and an anode of MoO3 (see Figure
3(a)). The solar cells had an inverted configuration where the
bottom contact was SrTiO3:Nb and the top contact was
MoO3 (for the full device structure see the inset of Figure
3(b)). Figure 3(a) shows that before light soaking (0 min),
the J-V curve exhibits an S shape, which is commonly attributed to a high effective contact resistance. As light soaking
is applied, the S shape gradually disappears and both the
short-circuit current and the fill-factor improve. Figure 3(b)
shows the open circuit voltage as a function of light soaking,
and we see that in the first 10 min, it rises by about 0.2 V and
then saturates at a constant value of about 0.57 V. While the
open circuit voltage may depend on several factors, we may
still, carefully, deduce that it seems that both resistance and
built in potential are changing in response to light soaking.
IV. ANALYSIS
Before describing our analysis, we discuss relevant theories that have been developed in the context of MIM devices.
MIM devices with an inorganic crystal as the insulating layer
were an active area of research in the 1940s–1980s.17–19
These studies usually used the same metal for both the anode
and cathode of the device. This created a device with no
built-in voltage between the anode and cathode and ensured
that only one carrier type was being injected, transported,
and extracted from the device (see Figure 4(a)). For these
devices, analytical models were developed based on the
drift-diffusion equation. These models were often reduced to
simple expressions for two limiting cases. The first was
when the charge injected into the device was high enough
that space charge limited current would form (Eq. (1)) and
the second was when the injected charge was so low that it
limited the current giving rise to an Ohmic like relation (Eq.
(2)),17
9 V2
J ¼ el 3 ;
8 L
V
J ¼ qn0 l :
L
(1)
(2)
Here, J is the current density, q is the charge of an electron,
l is the electron mobility of the insulator, e is the insulator’s
FIG. 2. (a) Current–voltage measurements for FTO/P3HT/MoO3/Au and
FTO/SrTiO3:Nb/P3HT/MoO3/Au, hole
only devices (b) The characteristics of
the FTO/SrTiO3:Nb/P3HT/MoO3/Au
device before (1) and after (2) light
soaking.
[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:
132.68.239.10 On: Sun, 13 Sep 2015 08:06:03
054501-4
Wing, Rothschild, and Tessler
J. Appl. Phys. 118, 054501 (2015)
FIG. 3. (a) Sequential current–voltage
measurements of a FTO/SrTiO3:Nb/
P3HT:PCBM/MoO3/Au solar cell illuminated at 1 Sun for longer and longer
periods of time (light soaking). (b)
Open circuit voltage as a function of
light exposure time. The inset shows a
cross section view of the solar cell’s
configuration.
permittivity, V is the applied voltage, L is the thickness of
the insulating layer, and n0 is the charge density next to the
injecting contact.
This simple model was extended to the case where two
different metals with different Fermi-energy levels, i.e.,
workfunctions, were used so that the device had a built-in
voltage.13,20,21 One of the results of this model is that at large
voltage biases, there is an asymmetry in forward and reverse
bias conduction that can be several orders of magnitude in
size. This stems from the interplay between diffusion and
drift. At zero bias, the built-in field is opposed by an equally
strong diffusion gradient (see Figure 4(b)). For small forward
biases, a diffusion current is responsible for current flow.
The current is low because diffusion is opposed by a small
drift force in the opposite direction due to the built-in field.
As the bias increases, the electric field inside the device
decreases until it reaches a transition point and flips
FIG. 4. (a) Energy band diagrams for forward and reverse biased MIM device with no built-in field across the device (i.e., the contacts have the same
work function). Conduction is identical for reverse and forward bias. The
arrows represent the strength and direction of the drift electron flow. (b)
Energy band digrams for a MIM device with a built-in field. This built-in
field causes a competition between drift and diffusion forces for negative
biases and for small positive biases. The current is greatly reduced because
they oppose each other. At positive biases larger than the built-in voltage,
drift and diffusion point in the same direction and SCL current behavior is
obtained.13
directions, resulting in a large increase in current flow and
charge carrier density. In the reverse direction, a strong electric field is opposed by a strong diffusion gradient. There is
no transition point, and the contact that is injecting current is
a poor injector resulting in a current that is much lower than
the forward current at the same bias. This does not exist in
devices with symmetric contacts, for them Eqs. (1) and (2)
describe both forward and reverse biases. This asymmetry
depends on the strength of the built-in voltage and thus provides information on the device’s electrodes’ work functions.
Unfortunately, closed form solutions for the currents in MIM
devices with asymmetric contacts are extremely difficult to
attain, although rigorous solutions have been shown under
specific approximations.13 Because of this, we use numerical
simulations to analyze the device behavior.
Figure 5(a) shows current–voltage curves calculated
using a semiconductor device model. The simulation uses a
drift-diffusion model, which incorporates the contacts in a
self-consistent manner as described in Ref. 22. In these calculations, we assumed a 100 nm thick intrinsic semiconductor having a band gap of 2 eV. One electrode (defined as
Ohmic) had an injection barrier of 0.1 eV and the other one
had an injection barrier that was varied between 0 and 0.9 eV
(see values in the figure). The model assumed no electronic
disorder and that the semiconductor is non degenerate. The
charge mobility value was taken to be 103 cm2 v1 s1.
The thick solid lines, denoted as forward, were calculated for
the Ohmic contact being the electron injecting contact and
the thin lines are for the opposite bias where the other contact is injecting the electrons. For contact barriers between
zero and 0.3 eV, there is no difference between forward and
reverse bias indicating the range where the contact is still
Ohmic. At a barrier height of 0.4 eV, a difference starts to
evolve and reaches up to 9 orders of magnitude for an injection barrier of 0.9 eV.
Comparing Figure 5(a) to Figure 1 or Figure 2, we note
that it captures the change in the measured J-V curves
following either switching voltage or light soaking. Using
the simulated results, we note that in the reverse direction,
where the high barrier contact is injecting, the curves have
an almost linear slope and that the device resistance is
changing as a function of the contact barrier height. This can
be understood with the aid of Eq. (2), which tells us that for
a high barrier contact the current is proportional to the
[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:
132.68.239.10 On: Sun, 13 Sep 2015 08:06:03
054501-5
Wing, Rothschild, and Tessler
J. Appl. Phys. 118, 054501 (2015)
FIG. 5. (a) Simulated current–voltage
characteristics of a single carrier device with different injection barrier
heights at the bottom electrode and a
fixed injection barrier of 0.1 eV
(Eg ¼ 2 eV) at the top electrode. (b)
The measured data of Figure 2(b)
(symbols) compared to the best fit
simulated results (lines). The simulated
curves used injection barrier heights of
0.55 V (red lines and circles) and
0.6 eV (black lines and plus signs).
charge density at the contact (n0), which is determined
by
the contact barrier (DE) following n0 ¼ NDOS exp DE
kT
with NDOS being the semiconductor density of states. The
theory behind Eq. (2) shows that in the case where the current is contact limited, the charge density is almost uniform
across the film and equals n0. Namely, the resistance that is
changing is the bulk resistance that is directly related to the
charge density. This “resistor” also plays a role when the
Ohmic contact is the injecting contact. In that case, it appears
as the leakage path that masks the diode-like shape of the
J-V curve at low bias levels. The space charge limited regime appears only at applied voltages above the built in
potential, and indeed the exponential rise in the current is
followed by a power law with an exponent of 2.
To test the model in a more quantitative way, we fitted
the calculated curves to the measurements of Figure 2(b)
(see Figure 5(b)). As the mobility used for the simulation is
not necessarily the same as the mobility in the measured
device, we focused on fitting the shape of the measured
curves rather than their absolute values. Namely, instead
of finding the simulated curve, JðDEi Þ, that best fit the
measured data, we used C JðDEi Þ for the fit, with C being a
fitting parameter. This fitting procedure resulted in the
SrTiO3:Nb having an injection barrier of 0.55 eV relative to
P3HT’s HOMO, which then increased to 0.6 eV upon light
soaking (i.e., the SrTiO3:Nb workfunction went up towards
the vacuum level).
While there is a rather good resemblence between the
shape of the simulated curves and the measured ones, there
are clear differences, especially when considering the forward curves. The part that exhibits the exponential rise is
more shallow in the measured curves. Spatial disorder in
either the film23,24 or the contacts25–27 could be considered
as a source of such phenomena. Detailed studies of
SrTiO3:Nb do indicate the presence of spatial nonuniformities,28 and studies of Pt/SrTiO3 junctions have indicated
workfunction inhomogeneities.2,29 Also, the results of Figure
5(a) show the high impact of the barrier height, hence we
would follow the route of suggesting that the spatial disorder
in the contact barrier would create a series of parallel connected diodes each having a unique barrier height. Namely,
we fit the neasured current density (JD) to a linear combination of the currents calculated for different barrier heights
(DEi), as expressed in the following equation:
JD ¼
X
ai JðDEi Þ:
(3)
i
In Figures 6(a) and 6(c), we present the results of the fitting
procedure using Eq. (3). We first performed the fit using the
finite set of DE 2 ½0:35 eV; 0:40 eV; :::; 0:9 eV. Next, we
chose the 3 most significant values and repeated the fit using
them only. In Figures 6(b) and 6(d), we present the coefficients (ai) as a function of the barrier height (DEi). As the
reverse currents are featureless, we use the forward currents
to evaluate the coefficients and then the same coefficients
are used to calculate the reverse currents. The symbols in
Figure 6(a) represent the measured data of the hole only
device before and after light soaking (Figure 2(b)). The lines
show the best fit curves. One can see that the fit for the forward current is excellent and the reverse current fits are not
so bad. The most meaningful coeficients are of relatively
high barrier heights as would be expected for the injection of
holes from SrTiO3:Nb to P3HT. The resulting coefficients in
Figure 6(b) can be taken to represent the relative contact
area for each barrier height.25–27 Namely, Figure 6(b)
implies that the change seen in the J-V curves is attributed to
a relatively small fraction of the device area. Since the sum
of the coeficients for barrier heights below 0.9 eV is less than
10% of the total sum, we can deduce that less than 10% of
the sample was involved in the light soaking phenomena. In
Figure 6(c), we show the resulting fitting procedure when
applied to the switching voltage measurements of electron
only devices presented in Figure 1(b) (curves 4 and 5). The
obtained coefficients are shown in Figure 6(d). For the low
current case, the close to Ohmic contact comprises less than
10% of the contact area. In the high current case, the Ohmic
contact comprises about 90% of the contact area. The low
barrier heights are in agreement with SrTiO3:Nb being a relatively good electron injecting contact to PCBM. Since the
high current case, which shows no indication of built in field,
is featureless, the only conclusion we can make is that the
area covered by the Ohmic contact is at least 90% of the total
contact area and hence that most of the device area took part
in the switching phenomena. It is also interesting to compare
Figures 6(b) and 6(d). Figure 6(b) indicates that most of the
contact area has a workfunction that is about 0.9 eV above
the HOMO of P3HT (4.8 eV) suggesting that it is at about
3.9 eV. Figure 6(d) indicates that large parts of the contact
area are 0.3–0.4 eV below the LUMO of PCBM (3.5 eV)
[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:
132.68.239.10 On: Sun, 13 Sep 2015 08:06:03
054501-6
Wing, Rothschild, and Tessler
J. Appl. Phys. 118, 054501 (2015)
FIG. 6. (a) Experimental data of
Figure 2(b) (symbols) and the best fit
using a linear combination of diode
responses (lines). (b) The coefficient
(weight) of each diode curve used in
(a) as a function of the respective barrier height. (c) Experimental data of
curves 4 and 5 in Figure 1(b) and the
best fit using a linear combination of
diode responses (lines). (d) The coefficient (weight) of each diode curve
used in (c).
suggesting that it is at 3.8–3.9 eV, in good agreement with
Figure 6(b) and with reported values for n-type SrTiO3.14,15
V. SUMMARY
In summary, current–voltage measurements of single
carrier injection into MIM type devices can provide quantitative data on the built-in voltage inside the devices. This can
be used to measure changes in the Fermi energy level of the
contacts in situ following various treatments that modify the
contacts’ work functions. In the context of memristors, we
found, for the first time, that the resistance switching associated with the metal oxide films studied here is accompanied
by a large change in the work function of the film. In the
devices studied here, this was the dominant effect. The above
conclusions were made possible largely due to the device
configuration. First, the other contact had a workfunction
outside the semiconductor bandgap, so that it was pinned to
the respective energy level, HOMO or LUMO. Second, the
semiconductor used was a low mobility semiconductor
ensuring that the resistance of the 8 nm SrTiO3:Nb film was
negligible in comparison and thus changes in the SrTiO3:Nb
film resistance would not have affected the device performance and could be neglected. The results shown in Figure 6
are in qualitative agreement with those reported in Ref. 29,
and the actual differences may be attributed to the differences in prepraration protocols of the SrTiO3:Nb films.
However, we believe that at least some of the differences are
attributed to the different analysis method used here. It has
been suggested that switching in SrTiO3 would involve both
barrier height modifications and filamentary conduction.30
The use of thin SrTiO3:Nb film on top of a low mobility
semiconductor ensures that the resistance change of the
SrTiO3:Nb film does not affect the analysis of the barrier
height distribution. Otherwise, the analysis would over
emphasize the regions exhibiting lower SrTiO3:Nb film resistance. All measurements were performed at the same temperature thus minimizing the potential effect of changes to
the bulk resistance in the resulting analysis. Finally, as we
performed numerical fitting, we did not assume the inhomogeneity distribution to be Gaussian in nature.
In the context of MIM diodes or solar cells, applying
this method to devices using SrTiO3:Nb contacts revealed
that light soaking increased SrTiO3:Nb’s Fermi energy
level bringing it closer to the vacuum level. Through the
analysis of the J-V curves, we conclude that the Ohmic like
current, which in the context of diodes is the leakage current, is attributed to the finite charge density at the collecting electrode. To strengthen this point, we plot in Figure 7
the calculated charge density distribution at a small forward
bias (V ¼ 10 mV) for two barriers heights at the collecting
electrode. These were 0.5 eV (solid line) and 0.45 eV
(dashed line). It is clear from this plot that the collecting
electrode induces a background charge density which is independent of the charge injected at x ¼ 0, thus creating a
background conduction or apparent leakage current (parallel resistance).
The other point we could deduce from our analysis is
that the serial resistance found in the reverse bias is induced
by the contact and hence it can be called a contact resistance.
However, the resistance is that of the semiconductor due to
the very low charge density imposed by the contact.
[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:
132.68.239.10 On: Sun, 13 Sep 2015 08:06:03
054501-7
Wing, Rothschild, and Tessler
J. Appl. Phys. 118, 054501 (2015)
an ideality factor n close to 1 (n ¼ 1.1). The filled symbols
are the data points of Figure 2(b) curve 1 and the solid line is
the best fit to an exponential resulting in an ideality factor of
n ¼ 2.9.
ACKNOWLEDGMENTS
This research was partially supported by the Israeli
Nanotechnology Focal Technology Area on Nanophotonics
for Detection, the Grand Technion Energy Program (GTEP),
and the Russel Berrie Nanotechnology Institute and
comprises part of The Leona M. and Harry B. Helmsley
Charitable Trust reports on Alternative Energy series of the
Technion, Israel Institute of Technology, and the Weizmann
Institute of Science.
FIG. 7. Calculated charge density distribution for a device having injection
barrier of 0.1 eV at x ¼ 0 and injection barrier of 0.5 eV (solid line) or
0.45 eV (dashed line) at x ¼ 0.1 lm. The applied forward bias is 10 mV.
Finally, it is well accepted that thin metal oxide films
operating as memristor elements are non uniform to the point
they may support filamentary currents.1 We found that when
the thin metal oxide film acts as part of the electrode, the
non-uniformity affects the device performance and through
analysis we could quantify such non-uniformities (Figures
6(b) and 6(d)). In Ref. 25, the effect of the inhomogeneous
barrier height was considered within the framework of
thermionic emission, and it was shown that such nonuniformity or inhomogeneity would lead to ideality factor
larger than 1. In low mobility semiconductors, the back diffusion is significant, and hence the contact is better treated
assuming equilibrium or continuity of the Fermi level at the
contact interface.22,31 Nevertheless, the concept that a combination of different barrier heights could lead to ideality
factor significantly above 1 holds true here too. In Figure 8,
the symbols represent data points and the lines represent the
best fit to an exponentially rising current. The empty symbols are from the simulation data presented
in Figure 5(a)
qV
and the dashed lines are the best fits to J / enkT resulting in
FIG. 8. Current density as a function of applied bias. Empty symbols are
simulation data from Figure 5(a) and the filled symbols are the data of curve
1 in Figure 2(b). The lines are exponential fits performed to extract the ideality factor n.
1
R. Waser, R. Dittmann, G. Staikov, and K. Szot, Adv. Mater. 21, 2632
(2009).
2
K. Heesub, P. Chanwoo, L. Sungjoo, and K. Dong-Wook, J. Phys. D:
Appl. Phys. 42, 055306 (2009).
3
J. Kim, G. Kim, Y. Choi, J. Lee, S. Park, and K. Lee, J. Appl. Phys. 111,
114511 (2012).
4
J. Wang, C. Lu, J. Hsu, M. Lee, Y. Hong, T. Perng, S. Horng, and H.
Meng, J. Mater. Chem. 21, 5723 (2011).
5
M. Lilliedal, A. Medford, M. Madsen, K. Norrman, and F. C. Krebs, Sol.
Energy Mater. Sol. Cells 94, 2018 (2010).
6
H. Schmidt, K. Zilberberg, S. Schmale, H. Flugge, T. Riedl, and W.
Kowalsky, Appl. Phys. Lett. 96, 243305 (2010).
7
G. Kim, J. Kong, J. Kim, H. Kang, H. Back, H. Kim, and L. Kwanghee,
Adv. Energy Mater. 5, 1401298 (2014).
8
B. Ecker, H. Egelhaaf, R. Steim, J. Parisi, and E. Von Hauff, J. Phys.
Chem. C 116, 16333 (2012).
9
A. Wagenpfahl, D. Rauth, M. Binder, C. Deibel, and V. Dyakonov, Phys.
Rev. B 82, 115306 (2010).
10
C. Kim, S. Lee, E. Gomez, J. Kim, and Y. Y. Loo, Appl. Phys. Lett. 94,
113302 (2009).
11
E. Yalon, A. Gavrilov, S. Cohen, D. Mistele, B. Meyler, J. Salzman, and
D. Ritter, IEEE Trans. Electron Device Lett. 33, 11 (2012).
12
J. Hwang, A. Wan, and A. Kahn, Mater. Sci. Eng., R 64, 1 (2009).
13
G. T. Wright, Solid-State Electron. 2, 165 (1961).
14
G. Campet, C. Geoffroy, J. P. Manaud, J. Portier, Z. W. Sun, J.
Salardenne, and P. Keou, Mater. Sci. Eng., B 8, 45 (1991).
15
M. Minohara, I. Ohkubo, H. Kumigashira, and M. Oshima, Appl. Phys.
Lett. 90, 132123 (2007).
16
J. Meyer, S. Hamwi, M. Kr€
oger, W. Kowalsky, T. Riedl, and A. Kahn,
Adv. Mater. 24, 5408 (2012).
17
N. F. Mott and R. W. Gurney, Electronic Processes in Ionic Crystals
(Oxford University Press, London, 1940).
18
M. A. Lampert and P. Mark, Current Injection in Solids (Academic Press,
New York, 1970).
19
K. C. Kao and W. Hwang, Electrical Transport in Solids (Pergamon press,
New York, 1981), Vol. 14.
20
D. J. Page, Solid-State Electron. 9, 255 (1966).
21
J. S. Bonham and D. H. Jarvis, Aust. J. Chem. 31, 2103 (1978).
22
N. Tessler, J. Polym. Sci., Part B: Polym. Phys. 52, 1119 (2014).
23
N. Rappaport, O. Solomesch, and N. Tessler, J. Appl. Phys. 99, 064507
(2006).
24
N. Rappaport, Y. Preezant, and N. Tessler, Phys. Rev. B 76, 235323
(2007).
25
J. H. Werner and H. H. Guttler, J. Appl. Phys. 69, 1522 (1991).
26
K. Maeda, Surf. Sci. 493, 644 (2001).
27
M. V. Schneider, A. Y. Cho, E. Kollberg, and H. Zirath, Appl. Phys. Lett.
43, 558 (1983).
28
M. Markovich, J. Roqueta, J. Santiso, E. Lakin, E. Zolotoyabko, and A.
Rothschild, Appl. Surf. Sci. 258, 9496 (2012).
29
E. Lee, M. Gwon, D.-W. Kim, and H. Kim, Appl. Phys. Lett. 98, 132905
(2011).
30
S. Xianwen, L. Guoqiang, Z. Xin’an, D. Linghong, and Z. Weifeng,
J. Phys. D: Appl. Phys. 44, 125404 (2011).
31
Y. Preezant and N. Tessler, J. Appl. Phys. 93, 2059 (2003).
[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:
132.68.239.10 On: Sun, 13 Sep 2015 08:06:03