Research Article

Cite This: ACS Appl. Mater. Interfaces 2018, 10, 29741−29749 www.acsami.org

All-Inorganic Bismuth Halide Perovskite-Like Materials A3Bi2I9 and

A3Bi1.8Na0.2I8.6 (A = Rb and Cs) for Low-Voltage Switching Resistive
Memory
Can Cuhadar,† Seul-Gi Kim,† June-Mo Yang,† Ja-Young Seo,† Donghwa Lee,*,‡
and Nam-Gyu Park*,†
†
School of Chemical Engineering, Sungkyunkwan University, Suwon 16419, Korea
‡
Department of Materials Science and Engineering, Pohang University of Science and Technology (POSTECH), Pohang 37673,
Downloaded via INDIAN ASSOCIATION CULTIVATION SCI on November 15, 2022 at 14:00:48 (UTC).

Korea
See https://pubs.acs.org/sharingguidelines for options on how to legitimately share published articles.

*
S Supporting Information

ABSTRACT: As silicon-based metal oxide semiconductor field effect transistors get

closer to their scaling limit, the importance of resistive random-access memory devices
increases due to their low power consumption, high endurance and retention
performance, scalability, and fast switching speed. In the last couple of years,
organic−inorganic lead halide perovskites have been used for resistive switching
applications, where they outperformed conventional metal oxides in terms of large on/
off ratio and low power consumption. However, there were scarce reports on lead-free
perovskites for such applications. In this report, we prepared lead-free Au/A3Bi2I9/Pt/
Ti/SiO2/Si (A is either Cs+ or Rb+) devices and tested their resistive switching
characteristics. They showed a forming step prior to repeating switching, low operating
voltage (0.09 V for Rb3Bi2I9 and 0.1 V for Cs3Bi2I9), large on/off ratio (>107), relatively
high endurance (200 cycles for Rb3Bi2I9 and 400 cycles for Cs3Bi2I9 cycles), and high
retention (1000 s). Such low voltage could be explained by grain boundary-modulated
ion drift. Difference in endurance was speculated to be due to the difference in the
surface roughness of films because Cs3Bi2I9 films are smoother. To get rid of the forming step, 10% of the Bi3+ cations were
substituted with Na+ cations. However, this method only worked on Rb-based structures. This phenomenon was explained by
the defect formation energy, which can only be negative in a corner-sharing Rb3Bi2I9 structure compared to a face-sharing
octahedral Cs3Bi2I9 structure. As a result, the forming step was removed, and 100 cycles endurance and 1000 s retention
performance were obtained. Similarly, the lower endurance is suspected to be due to the poor surface quality of the film.
KEYWORDS: lead-free perovskite, resistive switching, low power consumption, large on/off ratio, multilevel switching, substitution,
forming-free

■ INTRODUCTION
With silicon-based metal oxide field effect transistors reaching
their scaling limits, research on new kinds of memory devices
has gained great importance,1,2 mainly, ferroelectric random-
access memory (FeRAM),3 magnetoresistive RAM (MRAM),4
phase-change RAM (PRAM), 5 and resistive RAM
(ReRAM).1,2,6 Among these alternative to conventional
silicon-based memory devices, ReRAM stands out due to its
scalability, low power consumption, fast switching speed,
durability, and long data retention characteristics.2 ReRAMs
are composed of metal/insulator/metal structures in which the
insulator is the active film. The basic operation principle of
ReRAMs relies on resistive switching phenomenon, which
allows devices’ resistance to be changed under an applied
electric field between high resistance state (HRS) and low
resistance state (LRS). This phenomenon can be divided into
two types with respect to switching mechanism: interface and
filamentary types. Filamentary-type switching can occur via
electrochemical metallization (ECM) or valence change
memory (VCM). In case of ECM, switching occurs by the
formation and dissolution of metallic filament, which is
composed of atoms from an electrochemically active electrode.
However, VCM occurs by migration and filament formation of
internal defects in the active material. For both cases, usually a
forming step is required for repeatable switching.2,6,7 Most
research on ReRAM were focused on binary metal oxides such
as TaOx,8 HfOx,99 and TiOx10 and perovskite oxides such as
BiFeO311 and SrTiO3.12
Organic−inorganic halide perovskites having an ABX3
structure, where A is a monovalent cation, B is a divalent
cation, and X is a halide, have been investigated greatly for
their tremendous solar cell performance since the first report of
high-efficiency solid-state devices.13 Since then, these materials
have been incorporated with other types of electronic devices
Received: May 1, 2018
Accepted: July 3, 2018
Published: July 3, 2018

© 2018 American Chemical Society 29741 DOI: 10.1021/acsami.8b07103

ACS Appl. Mater. Interfaces 2018, 10, 29741−29749
ACS Applied Materials & Interfaces Research Article

Figure 1. (a) Schematic device structure and cross-sectional SEM images of (b) Rb3Bi2I9 film and (c) Cs3Bi2I9 film. Scale bars are 1 μm. XRD
patterns of (d) Cs3Bi2I9 film and (e) Rb3Bi2I9 film coated onto Pt/Ti/SiO2/Si substrates with the calculated patterns.

Figure 2. I−V curves of (a) forming step and (b) after forming step. Statistic data for (c) switching voltage and (d) forming voltage.

such as light-emitting diodes,14 lasers,15 thin-film transistors,16
X-ray imaging,17 and photodetectors18 due to their exceptional
electronic properties like long diffusion length,19 low binding
energy,20 and tunable band gap.21 Recently, lead halide
perovskite materials have been used in resistive switching
devices because they exhibit ionic migration character-
istics22−30 and demonstrate high on/off ratio, low operating
voltage, relatively fast switching, and relatively high endurance
29742
and data retention. For electronic applications, it is necessary
to develop lead-free resistive switching devices to avoid
underlying the toxicity of lead.31
Here, we report on resistive switching properties of
perovskite-related A3Bi2I9, where A is a cation that is either
Cs+ or Rb+. Difference in size of cation A in A3Bi2I9 can alter
the local geometry around the Bi ion between the zero-
dimensional (0D) face-sharing BiI63− octahedral (in case Cs+ is
DOI: 10.1021/acsami.8b07103
ACS Appl. Mater. Interfaces 2018, 10, 29741−29749
ACS Applied Materials & Interfaces
Figure 3. Endurance test performances of (a) Au/Rb3Bi2I9/Pt and (b) Au/Cs3Bi2I9/Pt devices. Set voltage was 0.5 V and reset voltage was −1 V
applied using dc voltage pulses having a pulse width of 50 ms; for both tests, the read voltage was kept at 0.2 V. Retention test performed for (c)
Au/Rb3Bi2I9/Pt device and (d) Au/Cs3Bi2I9/Pt device.
used) and the two-dimensional (2D) corner-sharing BiI63−
octahedral (in case Rb+ is used). The Rb3Bi2I9 structure was
first used for a resistive switching device. This study was
performed to get insights on the dependence of resistive
switching behavior on connections of transition metal
octahedral. In the end, both structures similarly showed an
initial forming step, very low switching voltage (lowest to date
with a gold top electrode), high on/off ratio, and relatively
high retention and endurance. Such similarity of switching
voltages could be explained by defect-modulated migrations of
vacancies. Endurance characteristics were higher for Cs3Bi2I9-
based devices, which were related to the difference in the
surface roughness of films. To remove the forming step, Bi3+
cations were partially substituted with Na+ cations, where 10%
substitution led to the best results. This modification only
worked on Rb3Bi2I9 structure, which could be explained by the
energetic stability of Na+ cations in a corner-sharing octahedral
structure compared to those in a face-sharing one. In the end,
the initial switching voltage dropped to about 0.3 V from 1.5 V,
and 100 cycles endurance and 1000 s retention performance
were obtained. Similarly, the lower endurance was suspected to
be due to the poor surface quality of the film.
■
RESULTS AND DISCUSSION
We fabricated vertically stacked devices in Au/A3Bi2I9/Pt/Ti/
SiO2/Si order, as shown in Figure 1a. Films were coated using
a simple antisolvent dripping method (details can be found in
29743
Research Article
the Experimental Section), which resulted in highly dense and
pinhole-free films having a thickness of about 180 nm for
Cs3Bi2I9 and 160 nm for Rb3Bi2I9 films (confirmed using
scanning electron microscopy (SEM) in Figure 1b,c). The
fabricated films also have a very smooth surface, which is very
important for electronic applications because rough films will
cause concern about differences in effective thickness and
localization of electric field.32 The surface quality of films was
checked using SEM and atomic force microscopy (AFM)
(Figure S1, Supporting Information). The Cs3Bi2I9 film and
the Rb3Bi2I9 film have a root mean square (RMS) roughness of
3.4 and 12.2 nm, respectively, which is in an acceptable range
when considering the thickness of these films. To confirm
phases of the films, X-ray diffraction (XRD) was used.
Comparing the calculated patterns based on previous
reports33,34 with the experimental ones proves that indeed
coated films have the desired phases (Figure 1d,e)0D face-
sharing BiI63− octahedral for Cs3Bi2I9 and 2D corner-sharing
octahedral for Rb3Bi2I9 films (Figure S2).
As seen in Figure 2a,b, devices switch in between resistance
states at different polarity voltages, which indicates that devices
show bipolar switching. A forming step is necessary for devices
to show repeatable switching characteristics, which is on
average 1.5 V for Au/Rb3Bi2I9/Pt and 1.3 V for Au/Cs3Bi2I9/
Pt devices (Figure 2d). After the forming step, I−V
characteristics of devices were obtained by applying direct
current (dc) voltage sweeps in 0 → 0.5 V → 0 → −0.5 V → 0
DOI: 10.1021/acsami.8b07103
ACS Appl. Mater. Interfaces 2018, 10, 29741−29749
ACS Applied Materials & Interfaces
Figure 4. log(I) vs log(V) plot of I−V characteristics of (a) Au/Rb3Bi2I9/Pt and (b) Au/Cs3Bi2I9/Pt. Curve fittings for HRS of (c) Au/Rb3Bi2I9/Pt
and (d) Au/Cs3Bi2I9/Pt, together with pristine states of (e) Au/Rb3Bi2I9/Pt and (f) Au/Cs3Bi2I9/P.
order to the top electrode while the bottom Pt electrode was
grounded. For both the forming step and other electrical
characterizations, current compliance was kept at 10−3 A to
prevent hard breakdown. Upon applying voltage from 0 to 0.5
V, switching from HRS (high resistance state) to LRS (low
resistance state) occurred at 0.09 V (5.6 × 103 V cm−1) for
Au/Rb3Bi2I9/Pt and at 0.1 V (5.5 × 103) for Au/Cs3Bi2I9/Pt
devices on average (Figure 2c). These devices showed the
lowest to date switching voltage with gold top electrodes
among perovskite-like halide materials. When voltage sweep
from 0 to −0.5 V was applied on devices until a certain voltage,
they retain their LRS; then, there is an abrupt drop of current
to about 10−10 A for Au/Rb3Bi2I9/Pt and about 10−11 A for
Au/Cs3Bi2I9/Pt, which refers to the reset process, switching
devices to HRS, during reset process no current compliance
was applied. Such abrupt changes and forming step suggest
that switching was modulated by filament formation. To show
reproducibility of results of the fabricated devices, voltage
29744
Research Article
sweeps were applied to 10 individual cells of each device, and
they showed a similar trend (Figure S3).
To show the reliability of devices, endurance tests were
conducted. During endurance tests, dc voltage pulses of 0.5 V
for set and −1 V for reset operation having a pulse width of 50
ms were applied to devices alternatingly, and the read voltage
was 0.02 V. Au/Rb3Bi2I9/Pt device showed switching for about
200 cycles with clear separation between LRS and HRS,
whereas Au/Cs3Bi2I9/Pt device showed about 400 cycles under
the same conditions (Figure 3a,b). On average, Au/Rb3Bi2I9/
Pt device had an on/off ratio of 2.9 × 107, whereas Au/
Cs3Bi2I9/Pt device had an average of 9.5 × 107, which is
consistent with their I−V curves because the HRS resistance of
Au/Cs3Bi2I9/Pt is higher than that of Au/Rb3Bi2I9/Pt. Such
difference in the endurance performance of devices can be
explained by differences in the surface roughness of films. As
previously reported for MAPbI3-based devices, increasing the
surface quality drastically increased the endurance performance
DOI: 10.1021/acsami.8b07103
ACS Appl. Mater. Interfaces 2018, 10, 29741−29749

ACS Applied Materials & Interfaces
of devices,25 which is related to electric field distributions
within the device; this in turn could cause multifilament
formation, hence lowering endurance.32 During endurance
tests, HRS resistance of both devices varied in about the 1000
Ω range (Figure 3a,b). This could be due to the formation of
multiple filaments because during the reset operation, some
remaining filaments might be present in the film, decreasing
the HRS resistance.29,32
Retention tests were conducted as well to confirm and
compare the nonvolatile memory characteristics of devices.
After dc voltage pulse was applied to switch devices, having the
same height and width as the pulses used in endurance tests,
from HRS to LRS, a constant voltage of 0.02 V was applied
while measuring the time. The same procedure was performed
to measure the HRS retention of devices; before the retention
test, again dc voltage pulse from the endurance test was used to
switch the device from LRS to HRS. Both devices showed
stable LRS and HRS for more than 1000 s (Figure 3c,d).
During HRS, although the resistance of devices fluctuated, yet
the on/off ratio higher than 107 for Au/Rb3Bi2I9/Pt and 106 for
Au/Cs3Bi2I9/Pt was preserved. Such behavior can be explained
by the trapping and detrapping of electrons in defect states.26
In addition, stability measurements of Rb3Bi2I9 devices were
performed as well because the stability of Cs3Bi2I9 devices was
reported before. After 30 days, devices stored under vacuum
showed the same switching characteristics as those in the initial
data (Figure S4). This indicates that the Rb3Bi2I9 devices are
stable for long-term use.
To find out the current conduction mechanisms of devices,
I−V curves on logarithmic scale were plotted (Figure 4a,b).
When linear fitting was done to devices at LRS, it was found
that the slope of the fitted line was very close to 1, which
indicates that current conduction for both devices is Ohmic
like at LRS.35 In case of HRS, the curves showed almost no
change in current as the voltage increases. Thus, we further
consider the hopping conduction mechanism in which current
conduction can occur by the tunneling of electrons between
trap states. In hopping conduction, current density has the
ÅÄÅ qaE É
following relationship with applied field35
E ÑÑ
J = qanν expÅÅÅÅ − a ÑÑÑÑ
ÅÅÇ kT kT ÑÑÖ
(1)
where q is electronic charge, a is hopping distance, n is electron
concentration in the conduction band, ν is the frequency of
thermal vibration of electrons at trap states, Ea (activation
energy) is the energy level from the trap states to the bottom
of the conduction band, E is the applied field, k is Boltzmann’ s
constant, and T is temperature. Hence, there should be a linear
relation between V and ln(I).36 To observe such a relationship
in our devices in the HRS region, ln(I) vs V curves were
plotted (Figure 4c,d); to plot this curve for Au/Cs3Bi2I9/Pt
device, the negative bias region was used because of small
switching voltages, fluctuated data, and capacitive effects.37
Linear curves prove that conduction at HRS occurs by the
tunneling of electrons between trap states. In the same manner,
the conduction mechanism at pristine state was investigated
because Schottky conduction was suspected. Schottky
conduction occurs by electrons obtaining enough thermal
energy to overcome the barrier height at the electrode
dielectric interface, which has such a relation of current
density and applied field35
29745
ÄÅ ÉÑ
Research Article
ÅÅ −q(ϕ − qE /(4πε ε ) ÑÑ
Å ÑÑ
J = A*T expÅÅÅÅ ÑÑ
ÅÅ ÑÑ
ÅÇ ÑÑÖ
2 B r 0
kT (2)
where A* is effective Richardson constant, qϕB is the Schottky
barrier height, ε0 is permittivity of vacuum, and εr is relative
permittivity. Therefore, between ln(I) and V1/2, there should
be a linear relationship for Schottky conduction. Indeed, ln(I)
vs V1/2 curves were linear for devices at pristine state (Figure
4e,f).
As mentioned before, devices require a forming step before
showing repeatable switching characteristics. During the
forming step for VCM, introduction of ionic defects, mainly
iodide vacancies (VI• in Kröger−Vink notation), occurs by
electrolytic reaction at the anode, which acts as a donor, thus
increasing conductivity. Because vacancies are mobile in the
film upon the application of an electric field, they migrate
toward the cathode and accumulate there. Then, these
vacancies form a conducting filament for the first time by
anisotropic accumulation due to their high conductivity.2,6,7,38
Such defect generation can be observed from a comparison of
conduction mechanisms between the pristine state and HRS
because new energy states appear in the band gap upon the
introduction of defects, which might cause change in the
conduction mechanism of devices from Schottky to hopping
conduction. Thus, considering these facts, we propose
switching mechanism as vacancy modulated filament formation
for both devices.
As seen from the statistical calculation of set voltages of Au/
Rb3Bi2I9/Pt and Au/Cs3Bi2I9/Pt devices (Figure 2c), there is a
very small difference between themabout 0.01 V, which is
very peculiar because the crystal structures of Rb3Bi2I9 and
Cs3Bi2I9 are completely different from each other. In previous
reports, it was observed that grain boundaries have a great
influence on the ion migration of halide perovskite
materials.39,40 Therefore, it is reasonable to say that vacancy
migration is modulated by grain boundaries. Such an
explanation can also interpret the very low switching voltage
of these devices. Grain boundaries are highly open structures
compared to bulk, so along grain boundaries, ions are more
mobile. As seen in Figure S1b,d, the grain size of films is very
small, making the grain boundary area very large. As a result,
highly mobile ions can easily drift along these paths.
Although both devices show similar switching behaviors,
there are some differences as well, such as pristine state and
HRS resistances. Au/Cs3Bi2I9/Pt devices have higher resist-
ance and hence higher on/off ratio, 1.45 × 107 for Au/
Rb3Bi2I9/Pt and 9.73 × 108 for Au/Cs3Bi2I9/Pt devices on
average, which was calculated from dc sweeps. To understand
the difference in resistivity and on/off ratio, we further perform
first-principles hybrid functional calculations. Our hybrid
calculations show that Cs3Bi2I9 films have a higher Schottky
barrier than Rb3Bi2I9 films. The estimated Schottky barrier at
the Au interface is 1.63 eV for Cs3Bi2I9 and 1.04 eV for
Rb3Bi2I9 (Figure S5). Thus, our calculation results successfully
show that the higher resistance of Cs3Bi2I9 films is a result of
the higher Schottky barrier in the pristine state. At HRS, I−V
characteristics show that the current conduction follows the
hopping mechanism, which depends on energy difference
between the trap state and the conduction band minimum
(CBM); see eq 1. To verify this, the electronic densities of
state (DOS) of both Cs3Bi2I9 and Rb3Bi2I9 were investigated.
Figure S6 shows the DOS profile of the perfect system (a and
DOI: 10.1021/acsami.8b07103
ACS Appl. Mater. Interfaces 2018, 10, 29741−29749
ACS Applied Materials & Interfaces
Figure 5. (a) Resistive switching performance of Au/Rb3Bi1.8Na0.2I8.6/Pt device I−V curves of the first and second sweeps. (b) Box plot of the
initial switching voltages. (c) Endurance test performance. (d) Retention test performance. All tests were conducted under the same conditions
with the previous devices.
b) and with VI• (c and d). Our hybrid calculations predict that
the energy level of VI• is 1.2 eV lower than CBM for Cs3Bi2I9,
whereas it is only 0.595 eV lower than CBM for Rb3Bi2I9. As a
result, the number of thermally excited electrons is less in
Cs3Bi2I9 and thus Cs3Bi2I9 shows a higher resistance in HRS.
Nonetheless, both devices have on/off ratio at least more than
107, which can be exploited for multilevel switching to be used
in high-density memory applications (Figure S7). Using
current compliance to adjust different resistance states, five
distinct states for Au/Rb3Bi2I9/Pt and four states for Au/
Cs3Bi2I9/Pt devices can be set. As seen in Figure S7, with
increase of current compliance, reset voltage increases as well.
This behavior is a universal characteristic for resistive switching
materials, which is due to either increase of filament radius or
decrease in tunneling gap so that more stable filaments are
formed, thus making it more difficult to rupture them.7
In terms of commercial device operation, forming step is not
desired. Thus, many studies have been done to remove this
step for oxide-based resistive memories; several methods have
been developed to this end, such as doping.7 However, because
this was the first time a halide perovskite was reported having a
forming step with the VCM mechanism, this will be the first
time such a thing is tried. In this report, we tried doping or
substituting Bi3+ cations with Na+. Na+ was used because it has
a similar ionic radius (102 pm) with Bi3+ (103 pm) so that
strain energy when two cations replaced will be zero.
Moreover, because Na+ cation has 2+ lower valency energy
for iodine, vacancy formation will be even lower.41 When Na+
was substituted, films having chemical formula A3Bi2−xNaxI9−2x
were fabricated. It was seen that 10% substitution gave the best
results, which corresponds to x = 0.2. The driving force behind
29746
Research Article
iodine vacancy creation is preserving charge neutrality, and the
creation of vacancies occurs as the following reaction:
NaI = Na″Bi + 2V •I + I×I (3)
To see whether substitution was successful or not, XRD
pattern and UV−vis spectra of films were taken (Figure S8). In
the XRD pattern of the Rb3Bi2Na0.2I8.6 film, there are some
new peaks which is due to change in the positions of ions
because Na+ has lower charge than Bi3+ so that bond distances
changes when Na+ substitutes Bi3+. As a result, structure factor
calculation changes and hence new peaks will appear. The
slope of the absorption curve in the UV−vis spectrum of the
Rb3Bi1.8Na0.2I8.6 film became less steep as well, which indicates
that it has higher Urbach energy, and it is calculated from the
slope of ln(α) vs photon energy curve, where α is the
absorption coefficient. Higher Urbach energy shows larger
disorder around the band edge (Figure S8).42 In the case of
Rb3Bi1.8Na0.2I8.6, this disorder is due to the substitution of Na+
cations. However, there are almost no differences in the XRD
pattern and the UV−vis spectrum of the Cs3Bi1.8Na0.2I8.6 film.
This suggests that the substitution of Bi3+ cations with Na+
cations was unsuccessful. We further performed defect
formation energy (DFE) calculations to verify our exper-
imental observations. Our calculation results clearly show that
the substituted Na+ defect at Bi3+ site (NaBi″) in Cs3Bi2I9 has
positive DFE in the middle of the band gap; see the red line in
Figure S9. Thus, Na+ ion is not energetically favorable to
substitute Bi3+ ion in Cs3Bi2I9. On the contrary, NaBi″ in
Rb3Bi2I9 has negative DFE with the preferred charge state of
−2; see the blue line in Figure S9. This clearly shows that the
Na+ ion preferentially substitutes Bi3+ site in Rb3Bi2I9 and leads
to the formation of two VI• to maintain the charge balance.
DOI: 10.1021/acsami.8b07103
ACS Appl. Mater. Interfaces 2018, 10, 29741−29749
ACS Applied Materials & Interfaces
Thus, our DFE calculations successfully explain the preferential
formation of NaBi″ in a corner-shared Rb3Bi2I9 structure.
Measurements of Na+-substituted devices were made under
the same measurement condition as described previously. After
substituting Na+, there is no drastic difference between the first
and second switching voltages, which shows that these devices
are forming-free (Figure 5a). However, not all devices were
completely forming-free; some of them showed initial
switching up to 0.5 V, and when statistical calculations were
made, it was found that the average initial switching voltage
was about 0.3 V, which is far lower than that prior to
substitution (Figure 5b). Endurance and retention tests were
also carried out. Retention performance of these devices was
the same as with the Rb3Bi2I9-based ones (over 1000 s), but
endurance performance was lower (about 100 switching
cycles) (Figure 5c,d). Figure S10 visualizes both structural
and switching behaviors to explain different switching
mechanisms between Rb3Bi2I9 and Rb3Bi1.8Na0.2I8.6 films. To
understand the difference in endurance, RMS roughness of the
film was measured using AFM (Figure S11), and it was found
to be 17.3 nm, which is higher than the RMS roughness of the
Rb3Bi2I9 film (12.2 nm). This can explain the lower endurance
performance based on the reasons discussed previously.
■
CONCLUSIONS
We prepared Au/Rb3Bi2I9/Pt and Au/Cs3Bi2I9/Pt devices and
studied their resistive switching characteristics. It was found
that both devices show very low switching voltages, high on/off
ratio (>107), retention performance up to 1000 s, and
endurance more than 200 cycles for Au/Rb3Bi2I9/Pt and 400
cycles for Au/Cs3Bi2I9/Pt devices. Despite the difference in
vacancy connectivity due to the difference in octahedra
connection by face sharing (Cs case) and corner sharing (Rb
case), similarity of devices in terms of switching voltage was
explained by grain boundary-modulated drift of ions, whereas
the difference between endurance performance was explained
by the better surface roughness of Au/Cs3Bi2I9/Pt devices.
VCM was proposed due to the forming step and current
conduction mechanism change after forming. Slight difference
between HRS resistance was found due to the different band
gaps of films. It was also shown that the high on/off ratio of
devices can be exploited to be used for multilevel switching.
Forming-free devices were fabricated by substituting Na+
cation into Bi3+ cation sites. However, this procedure only
worked on Rb-based films. This is due to the fact that NaBi″
can only have negative DFE in a corner-shared Rb3Bi2I9
structure. The substituted devices had an initial switching
voltage of 0.3 V, and they showed 100 cycle endurance and
1000 s retention performance.
■ EXPERIMENTAL SECTION
Device Fabrication. First, 100 nm-thick Pt-coated silicon wafers
having 14 mm × 14 mm dimensions were cleaned using sonication
bath in deionized water, ethanol, acetone, and then again ethanol for
15 min each. Just before spin coating, substrates were treated by UV−
ozone for 20 min. Rb3Bi2I9 solution was prepared by mixing 0.1526
mmol RbI (Alpha Aesar, 99.8%) and 0.1017 mmol BiI3 (Sigma-
Aldrich, 99%) in 0.181 mL dimethylformamide (DMF) (Sigma-
Aldrich, 99.8%). Cs3Bi2I9 solution was prepared by mixing 0.127
mmol CsI (Sigma-Aldrich, 99.999%) and 0.0848 mmol BiI3 in 0.204
mL DMF. For Rb3Bi1.8Na0.2I8.6, 0.1526 mmol RbI, 0.0916 mmol BiI3,
and 0.0100 mmol NaI were mixed in 0.172 mL of DMF. For
Cs3Bi1.8Na0.2I8.6, 0.127 mmol CsI, 0.0763 mmol BiI3, and 0.0087
mmol NaI were mixed in 0.195 mL of DMF, where x = 0.2 for both
29747
Research Article
solutions, and they have the same concentration with solutions having
no NaI in them. The solutions were spin coated onto Pt-coated silicon
wafers at 4000 rpm for 20 s. (acceleration = 1200 rpm s−1); for
Rb3Bi2I9 films, after 10 s 0.4 mL of ether (Sigma-Aldrich, 99%) was
dripped, and for Cs3Bi2I9 films, 0.4 mL of chlorobenzene (Sigma-
Aldrich, 99.8%) was dripped after 8 s. Then, the films were annealed
at 100 °C for 15 min. Gold top electrodes were deposited using
thermal evaporation through shadow mask under 10−6 torr.
Characterization. Cross-sectional and surface images of films
were taken using a field emission scanning electron microscope
(JEOL, JSM7000F). All electrical characterizations were made using a
Keithley 4200 semiconductor parametric analyzer under vacuum
about 10−2 torr; all voltages were applied to gold top electrodes while
the Pt bottom electrode was grounded. I−V curves were obtained
using direct current voltage sweeping, and cycling measurements were
done using direct voltage pulses having width 50 ms. X-ray diffraction
patterns were obtained using a Rigaku SmartLab with Cu Kα
radiation. Data were recorded under ambient conditions with a step
size of 0.02°. AFM measurements were done using SPA-400 Seiko
Instruments Inc. under ambient conditions.
Theoretical Calculations. First-principle density functional
theory (DFT) calculations were performed with the projector
augmented wave method43 and the revised Perdew, Burke, and
Ernzerhof (PBEsol) generalized gradient approximation44,45 for the
exchange−correlation potential as implemented in the Vienna ab
initio simulation package (VASP) code.46 The spin−orbital coupling
(SOC) effect is considered to correctly predict the localized electronic
state of the Bi ion. Because the DFT calculations underestimate the
band gap, the hybrid functional (PBE0),47 which mixes Hartree−Fock
(HF) exchange into DFT, is performed to improve the accuracy of
our DFT calculation. In this study, 15% of the HF exchange is mixed
with the PBEsol exchange and correlation potential. Our hybrid
functional calculations increase the band gap from 1.71 eV (DFT +
SOC) to 2.42 eV (PBE0 + SOC) for Cs3Bi2I9 and from 1.40 eV (DFT
+ SOC) to 2.10 eV (PBE0 + SOC) for Rb3Bi2I9. The supercell with
112 atoms (800 electrons) is chosen for both Cs3Bi2I9 and Rb3Bi2I9.
The optimized lattice parameters a and c are 8.57 and 21.73 Å for
Cs3Bi2I9, respectively, and the lattice parameters a, b, and c are 14.4,
7.98, and 20.63 Å for Rb3Bi2I9, respectively. Periodic boundary
condition and Monkhorst−Pack k-point sampling48 with a Γ-centered
k-point grid of up to 4 × 4 × 4 were used for Brillouin zone
integration. An energy cutoff of 400 eV was used for the plane-wave
representation of the wavefunctions. Atomic structures were relaxed
until all Hellman−Feynman forces were below 0.01 eV/Å. Defect
formation energy is calculated by subtracting the energy of the
nondefective system and defective species from the total energy of the
system, including defect. Rhombohedral crystal structure is used to
calculate the energy of Bi vacancy, and the FCC crystal structure is
used to calculate the energy of Na substitutional defect.
■
*
ASSOCIATED CONTENT
S Supporting Information
The Supporting Information is available free of charge on the
ACS Publications website at DOI: 10.1021/acsami.8b07103.
SEM images, AFM images, crystal structures, I−V
curves, band diagrams, electronic density of state
diagrams, XRD patterns, UV−vis spectra, Urbach energy
calculations, defect energy formation calculations,
schematic diagram (PDF)
■
AUTHOR INFORMATION
Corresponding Authors
*E-mail: donghwa96@postech.ac.kr (D.L.).
*E-mail: npark@skku.edu (N.-G.P.).
ORCID
Donghwa Lee: 0000-0002-8956-3648
Nam-Gyu Park: 0000-0003-2368-6300
DOI: 10.1021/acsami.8b07103
ACS Appl. Mater. Interfaces 2018, 10, 29741−29749
ACS Applied Materials & Interfaces
Notes
The authors declare no competing financial interest.
■
ACKNOWLEDGMENTS
This work was supported by the National Research
Foundation of Korea (NRF) grants funded by the Ministry
of Science and ICT (MSIT) of Korea under contracts NRF-
2016M3D1A1027663 and NRF-2016M3D1A1027664 (Future
Materials Discovery Program) and in part NRF-
2012M3A6A7054861 and NRF-2014M3A6A7060583 (Global
Frontier R&D Program on Center for Multiscale Energy
System) and NRF-2017H1A2A1046990 (NRF-2017-Fostering
Core Leaders of the Future Basic Science Program/Global
Ph.D. Fellowship Program).
■
REFERENCES
(1) Lee, J. S.; Lee, S.; Won Noh, T. Resistive Switching Phenomena:
A Review of Statistical Physics Approaches. Appl. Phys. Rev. 2015, 2,
No. 031303, DOI: 10.1063/1.4929512.
(2) Pan, F.; Gao, S.; Chen, C.; Song, C.; Zeng, F. Recent Progress in
Resistive Random Access Memories: Materials, Switching Mecha-
nisms, and Performance. Mater. Sci. Eng., R 2014, 83, 1−59.
(3) Ishiwara, H. Ferroelectric Random Access Memories. J. Nanosci.
Nanotechnol. 2012, 12, 7619−7627.
(4) Johan, Å. Magnetoresistive Random Access Memory. Handb.
Spin Transp. Magn. 2011, 104, 685−697.
(5) Wong, H.-S. P.; Raoux, S.; Kim, S.; Liang, J.; Reifenberg, J. P.;
Rajendran, B.; Asheghi, M.; Goodson, K. E. Phase Change Memory.
Proc. IEEE 2010, 98, 2201−2227.
(6) Waser, R.; Dittmann, R.; Staikov, C.; Szot, K. Redox-Based
Resistive Switching Memories Nanoionic Mechanisms, Prospects, and
Challenges. Adv. Mater. 2009, 21, 2632−2663.
(7) Ielmini, D.; Waser, R. Resistive Switching: From Fundamentals of
Nanoionic Redox Processes to Memristive Device Applications; Wiley,
2015.
(8) Lee, M. J.; Lee, C. B.; Lee, D.; Lee, S. R.; Chang, M.; Hur, J. H.;
Kim, Y. B.; Kim, C. J.; Seo, D. H.; Seo, S.; Chung, U. I.; Yoo, I. K.;
Kim, K. A Fast, High-Endurance and Scalable Non-Volatile Memory
Device Made from Asymmetric Ta2O5-xx/TaO2-Xbilayer Structures.
Nat. Mater. 2011, 10, 625−630.
(9) Wedig, A.; Luebben, M.; Cho, D. Y.; Moors, M.; Skaja, K.; Rana,
V.; Hasegawa, T.; Adepalli, K. K.; Yildiz, B.; Waser, R.; Valov, I.
Nanoscale Cation Motion in TaOx, HfOx and TiOx Memristive
Systems. Nat. Nanotechnol. 2016, 11, 67−74.
(10) Yang, J. J.; Pickett, M. D.; Li, X.; Ohlberg, D. A. A.; Stewart, D.
R.; Williams, R. S. Memristive Switching Mechanism for Metal/oxide/
metal Nanodevices. Nat. Nanotechnol. 2008, 3, 429−433.
(11) Jiang, A. Q.; Wang, C.; Jin, K. J.; Liu, X. B.; Scott, J. F.; Hwang,
C. S.; Tang, T. A.; Lu, H. B.; Yang, G. Z. A Resistive Memory in
Semiconducting BiFeO3 Thin-Film Capacitors. Adv. Mater. 2011, 23,
1277−1281.
(12) Menzel, S.; Waters, M.; Marchewka, A.; Böttger, U.; Dittmann,
R.; Waser, R. Origin of the Ultra-Nonlinear Switching Kinetics in
Oxide-Based Resistive Switches. Adv. Funct. Mater. 2011, 21, 4487−
4492.
(13) Kim, H. S.; Lee, C. R.; Im, J. H.; Lee, K. B.; Moehl, T.;
Marchioro, A.; Moon, S. J.; Humphry-Baker, R.; Yum, J. H.; Moser, J.
E.; Michael, G.; Park, N. G. Lead Iodide Perovskite Sensitized All-
Solid-State Submicron Thin Film Mesoscopic Solar Cell with
Efficiency Exceeding 9%. Sci. Rep. 2012, 2, No. 591.
(14) Lee, J. W.; Choi, Y. J.; Yang, J. M.; Ham, S.; Jeon, S. K.; Lee, J.
Y.; Song, Y. H.; Ji, E. K.; Yoon, D. H.; Seo, S.; Shin, H.; Han, G. S.;
Jung, H. S.; Kim, D.; Park, N. G. In-Situ Formed Type I
Nanocrystalline Perovskite Film for Highly Efficient Light-Emitting
Diode. ACS Nano 2017, 11, 3311−3319.
29748
Research Article
(15) Veldhuis, S. A.; Boix, P. P.; Yantara, N.; Li, M.; Sum, T. C.;
Mathews, N.; Mhaisalkar, S. G. Perovskite Materials for Light-
Emitting Diodes and Lasers. Adv. Mater. 2016, 28, 6804−6834.
(16) Wu, Y.; Li, J.; Xu, J.; Du, Y.; Huang, L.; Ni, J.; Cai, H.; Zhang, J.
Organic−inorganic Hybrid CH 3 NH 3 PbI 3 Perovskite Materials as
Channels in Thin-Film Field-Effect Transistors. RSC Adv. 2016, 6,
16243−16249.
(17) Kim, Y. C.; Kim, K. H.; Son, D. Y.; Jeong, D. N.; Seo, J. Y.;
Choi, Y. S.; Han, I. T.; Lee, S. Y.; Park, N. G. Printable
Organometallic Perovskite Enables Large-Area, Low-Dose X-Ray
Imaging. Nature 2017, 550, 87−91.
(18) Kwon, K. C.; Hong, K.; Van Le, Q.; Lee, S. Y.; Choi, J.; Kim, K.
B.; Kim, S. Y.; Jang, H. W. Inhibition of Ion Migration for Reliable
Operation of Organolead Halide Perovskite-Based Metal/Semi-
conductor/Metal Broadband Photodetectors. Adv. Funct. Mater.
2016, 26, 4213−4222.
(19) Xing, G.; Mathews, N.; Sun, S.; Lim, S. S.; Lam, Y. M.; Gratzel,
M.; Mhaisalkar, S.; Sum, T. C. Long-Range Balanced Electron- and
Hole-Transport Lengths in Organic-Inorganic CH3NH3PbI3. Science
2013, 342, 344−347.
(20) Miyata, A.; Mitioglu, A.; Plochocka, P.; Portugall, O.; Wang, J.
T. W.; Stranks, S. D.; Snaith, H. J.; Nicholas, R. J. Direct
Measurement of the Exciton Binding Energy and Effective Masses
for Charge Carriers in Organic-Inorganic Tri-Halide Perovskites. Nat.
Phys. 2015, 11, 582−587.
(21) Park, N. G. Perovskite Solar Cells: An Emerging Photovoltaic
Technology. Mater. Today 2015, 18, 65−72.
(22) Choi, J.; Park, S.; Lee, J.; Hong, K.; Kim, D.-H. H.; Moon, C.
W.; Park, G. D.; Suh, J.; Hwang, J.; Kim, S. Y.; Jung, H. S.; Park, N.
G.; Han, S.; Nam, K. T.; Jang, H. W. Organolead Halide Perovskites
for Low Operating Voltage Multilevel Resistive Switching. Adv. Mater.
2016, 28, 6562−6567.
(23) Muthu, C.; Agarwal, S.; Vijayan, A.; Hazra, P.; Jinesh, K. B.;
Nair, V. C. Hybrid Perovskite Nanoparticles for High-Performance
Resistive Random Access Memory Devices: Control of Operational
Parameters through Chloride Doping. Adv. Mater. Interfaces 2016, 3,
No. 1600092.
(24) Gu, C.; Lee, J. S. Flexible Hybrid Organic-Inorganic Perovskite
Memory. ACS Nano 2016, 10, 5413−5418.
(25) Choi, J.; Le, Q. V.; Hong, K.; Moon, C. W.; Han, J. S.; Kwon,
K. C.; Cha, P. R.; Kwon, Y.; Kim, S. Y.; Jang, H. W. Enhanced
Endurance Organolead Halide Perovskite Resistive Switching
Memories Operable under an Extremely Low Bending Radius. ACS
Appl. Mater. Interfaces 2017, 9, 30764−30771.
(26) Hwang, B.; Gu, C.; Lee, D.; Lee, J. S. Effect of Halide-Mixing
on the Switching Behaviors of Organic-Inorganic Hybrid Perovskite
Memory. Sci. Rep. 2017, 7, No. 43794.
(27) Hu, Y.; Zhang, S.; Miao, X.; Su, L.; Bai, F.; Qiu, T.; Liu, J.;
Yuan, G. Ultrathin Cs3Bi2I9Nanosheets as an Electronic Memory
Material for Flexible Memristors. Adv. Mater. Interfaces 2017, 4,
No. 1700131.
(28) Han, J. S.; Le, Q. V.; Choi, J.; Hong, K.; Moon, C. W.; Kim, T.
L.; Kim, H.; Kim, S. Y.; Jang, H. W. Air-Stable Cesium Lead Iodide
Perovskite for Ultra-Low Operating Voltage Resistive Switching. Adv.
Funct. Mater. 2017, No. 1705783.
(29) Seo, J.-Y.; Choi, J.; Kim, H.-S.; Kim, J.; Yang, J.-M.; Cuhadar,
C.; Han, J. S.; Kim, S.-J.; Lee, D.; Jang, H. W.; Park, N. G. Wafer-Scale
Reliable Switching Memory Based on 2-Dimensional Layered
Organic-Inorganic Halide Perovskite. Nanoscale 2017, 9, 15278−
15285.
(30) Sun, Y.; Tai, M.; Song, C.; Wang, Z.; Yin, J.; Li, F.; Wu, H.;
Zeng, F.; Lin, H.; Pan, F. Competition between Metallic and Vacancy
Defect Conductive Filaments in a CH 3 NH 3 PbI 3 - Based Memory
Device. J. Phys. Chem. C 2018, 122, 6431−6436.
(31) Babayigit, A.; Ethirajan, A.; Muller, M.; Conings, B. Toxicity of
Organometal Halide Perovskite Solar Cells. Nat. Mater. 2016, 15,
247−251.
DOI: 10.1021/acsami.8b07103
ACS Appl. Mater. Interfaces 2018, 10, 29741−29749
ACS Applied Materials & Interfaces Research Article

(32) Valov, I. Interfacial Interactions and Their Impact on Redox-

Based Resistive Switching Memories (ReRAMs). Semicond. Sci.
Technol. 2017, 32, No. 093006.
(33) Chabot, B.; Parthé, E. Cs3Sb2I9 and Cs3Bi2I9 with the
Hexagonal Cs3Cr2Cl9 Structure Type. Acta Crystallogr., Sect. B:
Struct. Crystallogr. Cryst. Chem. 1978, 34, 645−648.
(34) Lehner, A. J.; Fabini, D. H.; Evans, H. A.; Hébert, C. A.; Smock,
S. R.; Hu, J.; Wang, H.; Zwanziger, J. W.; Chabinyc, M. L.; Seshadri,
R. Crystal and Electronic Structures of Complex Bismuth Iodides
A3Bi2I9(A = K, Rb, Cs) Related to Perovskite: Aiding the Rational
Design of Photovoltaics. Chem. Mater. 2015, 27, 7137−7148.
(35) Chiu, F. C. A Review on Conduction Mechanisms in Dielectric
Films. Adv. Mater. Sci. Eng. 2014, 2014, 1−18.
(36) Chiu, F. C.; Shih, W. C.; Feng, J. J. Conduction Mechanism of
Resistive Switching Films in MgO Memory Devices. J. Appl. Phys.
2012, 111, No. 094104.
(37) Oh, S.-I.; Rani, J. R.; Hong, S.-M.; Jang, J.-H. Self-Rectifying
Bipolar Resistive Switching Memory Based on an Iron Oxide and
Graphene Oxide Hybrid. Nanoscale 2017, 9, 15314−15322.
(38) Kim, K. M.; Jeong, D. S.; Hwang, C. S. Nanofilamentary
Resistive Switching in Binary Oxide System; a Review on the Present
Status and Outlook. Nanotechnology 2011, 22, No. 254002.
(39) Yuan, Y.; Huang, J. Ion Migration in Organometal Trihalide
Perovskite and Its Impact on Photovoltaic Efficiency and Stability.
Acc. Chem. Res. 2016, 49, 286−293.
(40) Shao, Y.; Fang, Y.; Li, T.; Wang, Q.; Dong, Q.; Deng, Y.; Yuan,
Y.; Wei, H.; Wang, M.; Gruverman, A.; Shield, J.; Huang, J. Grain
Boundary Dominated Ion Migration in Polycrystalline Organic−
inorganic Halide Perovskite Films. Energy Environ. Sci. 2016, 9, 1752−
1759.
(41) Akbari, M.; Lee, J.-S. Control of Resistive Switching Behaviors
of Solution-Processed HfO X -Based Resistive Switching Memory
Devices by N-Type Doping. RSC Adv. 2016, 6, 21917−21921.
(42) Masuda, A.; Itoh, K.; Matsuda, K.; Yonezawa, Y.; Kumeda, M.;
Shimizu, T. Nitrogen-Doping Effects on Electrical, Optical, and
Structural Properties in Hydrogenated Amorphous Silicon. J. Appl.
Phys. 1997, 81, 6729.
(43) Blöchl, P. E. Projector Augmented-Wave Method. Phys. Rev. B
1994, 50, 17953−17979.
(44) Perdew, J. P.; Burke, K.; Ernzerhof, M. Generalized Gradient
Approximation Made Simple. Phys. Rev. Lett. 1996, 77, 3865−3868.
(45) Perdew, J. P.; Ruzsinszky, A.; Csonka, G. I.; Vydrov, O. A.;
Scuseria, G. E.; Constantin, L. A.; Zhou, X.; Burke, K. Restoring the
Density-Gradient Expansion for Exchange in Solids and Surfaces.
Phys. Rev. Lett. 2008, 100, 136406.
(46) Kresse, G.; Furthmüller, J. Efficiency of Ab-Initio Total Energy
Calculations for Metals and Semiconductors Using a Plane-Wave
Basis Set. Comput. Mater. Sci. 1996, 6, 15−50.
(47) Becke, A. D. Density-Functional Thermochemistry. III. The
Role of Exact Exchange. J. Chem. Phys. 1993, 98, 5648−5652.
(48) Pack, J. D.; Monkhorst, H. J. “Special Points for Brillouin-Zone
Integrations”-a Reply. Phys. Rev. B 1977, 16, 1748−1749.

29749 DOI: 10.1021/acsami.8b07103

ACS Appl. Mater. Interfaces 2018, 10, 29741−29749