COMMUNICATIONS
Memory Materials
Y. X. Yang, G. L. Yuan,* Z. B. Yan, Y. J. Wang, X. B. Lu,
J.-M. Liu*
...
1700425Flexible, Semitransparent, and Inorganic Resistive Memory based on BaTi0.95Co0.05O3 Film
Flexible mica/SrRuO3/BaTi0.95Co0.05O3/Au memory cells show bipolar resistive switching and the high/low resistance ratio is up to 50.
The resistive-switching properties show no obvious change after the 2.2 mm radius memory being writ-ten/erased for 360 000 cycles nor after the memory being bent to 3 mm radius for 10 000 times.
Memory Materials
www.advmat.de
Flexible, Semitransparent, and Inorganic Resistive Memory based on BaTi 0.95 Co 0.05 O 3 Film
Yuxi Yang, Guoliang Yuan,* Zhibo Yan, Yaojin Wang, Xubing Lu, and Jun-Ming Liu*
It is a big challenge for dynamic random access memory and flash memory to be miniaturized further due to the current lim-itations of semiconductor fabrication. On the contrary, resistive random access memory is promising for application in next-generation nonvolatile memory because of a variety of advan-tages, such as simple structure, high density, high speed, and low-power consumption.[1–6] At the same time, flexible resistive memory has attracted great attention with the rapid develop-ment of smart wearable devices and flexible display screens. Most studies of flexible resistive memory have been focused on organic substrates or films until now.[7–9] For example, Kim et al. successfully prepared a resistive switching memory with a quinoidal oligothiophene derivative and the memory with 0.5 mm bending radius was able to keep information over 6000 s or endure 100 writing/erasing cycles.[7] However, compared with inorganic memories, organic memories also
Dr. Y. X. Yang, Prof. G. L. Yuan, Prof. Y. J. Wang School of Materials Science and Engineering Nanjing University of Science and
Technology Nanjing 210094, P. R. China E-mail: yuanguoliang@njust.edu.cn Prof. Z. B. Yan, Prof. J.-M. Liu National Laboratory of Solid State Microstructures Nanjing University Nanjing 210093, P. R. China E-mail: liujm@nju.edu.cn Prof. X. B. Lu
Institute for Advanced Materials and Guangdong Provincial Laboratory of Quantum Engineering and Quantum Materials South China Normal University
Guangzhou 510006, China
DOI: 10.1002/adma.201700425
encounter some difficulties, such as poor stability at high temperature and easy oxi- dization under light illumination.[10,11]
Although many oxide resistive memo-ries have better endurance of writing/ erasing cycles, better temperature stability, and higher corrosion resistance than those of organic resistive memories, it is still a big challenge for these oxide memories to be flexible.[12–14] Recently, stable resistive switching characteristics, such as writing/
erasing less than 10/70 ns and over 105 cycles, have been demonstrated in Co-doped BaTiO3 film. Here, the migration of oxygen vacancies favors the formation of conductive filaments in the local regions under an
electric field, leading to an insu-
lator-to-metal transition.[2] Furthermore, the bandgap of 3.2 eV of BaTiO3 means that most visible light to pass can through its film.[15] Unfortunately, the used rigid substrates, such as Si and SrTiO3, limit the flexible performance of most oxide films, though freestanding oxide films are commonly thin enough to be flexible, in theory. Recently some epitaxial thin films (e.g.,
PbZr0.52Ti0.48O3 and GaAs) have been successfully peeled from their hard substrates to achieve flexible characteristics, so it is possible to prepare flexible resistive memories with oxide films.[16,17]
Here, SrRuO3/BaTi0.95Co0.05O3/Au (SRO/BTCO/Au) on mica is found to be a semitransparent, flexible, and cost-effective resistive memory device. The memory with 2.2 mm bending radius keeps the high/low resistance (RHRS/RLRS) ratio over 50 even after it was written/erased for 360 000 cycles or was annealed at 500 C.
Large-scale and high-quality SRO/BTCO films were depos-ited on 10 m thick mica by pulsed laser deposition. Compared with the common methods for flexible oxide memories listed in Table 1, such as one that includes oxide film growth on a hard substrate and transferring oxide films to a flexible organic substrate, the preparation method used here is high efficient, reliable, and cost effective. Figure 1a shows the X-ray diffraction (XRD) patterns of mica/SRO/BTCO, where the (110) and (220) diffraction peaks of SRO/BTCO film just exist, except those from single-crystal mica substrate. This suggests the [110] preferred orientation growth of the SRO/BTCO films on mica. Besides, the free standing SRO/BTCO films can be peeled from the mica substrate by mechanical exfoliation without any damage at surface and interface according to the cross-sectional image observed by transmission electron microscopy (Figure S1, Sup-porting Information). The complete [110] preferred orientation
Adv. Mater. 2017, 1700425
1700425 (1 of 6)
© 2017 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim
Perovskite ceramics and single crystals are commonly hard and brittle due to their small maximum elastic strain. Here, large-scale BaTi0.95Co0.05O3 (BTCO) film with a SrRuO3 (SRO) buffered layer on a 10 m thick mica substrate
is flexible with a small bending radius of 1.4 mm and semitransparent for visible light at wavelengths of 500–800 nm. Mica/SRO/BTCO/Au cells show bipolar resistive switching and the high/low resistance ratio is up to 50. The resistive-switching properties show no obvious changes after the 2.2 mm radius memory being written/erased for 360 000 cycles nor after the memory being bent to 3 mm radius for 10 000 times. Most importantly, the memory works properly at 25–180 C or after being annealed at 500 C. The flexible and transparent oxide resistive memory has good prospects for application in smart wearable devices and flexible display screens.
www.advancedsciencenews.com
www.advmat.de
Table 1. Nine flexible resistive memories studied extensively.
Reference This work Kim et al.[7] Zheng et al.[8] Fang et al.[9] Kim et al.[25] Rani et al.[26] Hwang et al.[27] Kim et al.[28] Jeong et al.[29]
Material BTCO QQT(CN)4 GeOx HfO2 PVDF-TrFE GOa) CNTsb) NiO TiO2
Substrate Mica Organic Organic Organic Organic Organic Organic Organic Organic
Transferc) No No No No No No No Yes Yes
W/E Cyclesd) [N] 3.6 105 100 100 400 125 50 100 100 50
Retention [s] 8.6 104 6000 — 104 104 104 105 104 104
RHRS/RLRS ratio 50 103 20 10 10 105 100 30 10
Radiuse) [mm] 1.4 0.5 9 — 4 — 7 7.5 —
Bending Cycles 104 1000 105 — 1000 50 500 — —
Temp.f) [C] 180 25 25 80 — — — 25 25
Annealingg) [C] 500 180 — — — — — — —
a)Graphene oxide; b)Carbon nanotubes; c)Transferring film to flexible substrate; d)Writing/erasing cycles; e)Minimum bending radius; f)Highest working temperature; g)Highest anneal temperature.
of the BTCO film contributes an nanometer-scale smooth sur- wavelength of 500–800 nm, respectively, as shown in Figure 1c.
face like that of epitaxial BaTiO3 film on single-crystal SrTiO3 Because the bandgap of BTCO is 3.2 eV, the SRO buffered layer substrate.[18,19] An average surface roughness of 1.35 nm of the as the bottom electrode absorbs most incident visible light and BTCO film has been determined according to the surface mor- decreases the transmittance of mica/SRO/BTCO. Figure 1d,e phology observed by atomic force microscopy (AFM), as shown shows the optical images of the bending mica/SRO/BTCO in Figure 1b. The nanometer-scale smooth surface is favorable and the flat one on a paper with BTCO characteristics. The two for preparing high-density memories by using the tiny elec- images confirm the large-scale flexible and semitransparent trodes with a diameter of tens of nanometers, which are influ- mica/SRO/BTCO, which may be used in flexible screen in the
enced seriously by a rough surface. future.
The transmittances of the pure mica substrate and the mica/ The conductive filament can form and rupture through the SRO/BTCO are over 80% and 50% for visible light with a migration of oxygen vacancies in BTCO; thus, SRO/BTCO/Au cells show good resistive switching character- istics. An optical image of the SRO/BTCO/Au memory device is shown in Figure 2a. Fur- ther current versus voltage (I–V) curves were measured to study resistive switching char- acteristics in Figure 2b. A voltage sweep was carried out from 0 to 17 V with a compliance current of 1 mA in order to set the device from the high-resistance status (HRS) to the low-resistance status (LRS), then another voltage sweep was performed from 0 to −17 V to reset the device back to the HRS, where the voltages at current mutations are defined as VSET and VRESET and the resistance of the HRS and the LRS at 2 V are defined as RHRS and RLRS, respectively. The VSET/VRESET can change from 17/−17 V to 8/−8 V when the BTCO thickness reduces by half (Figure S2, Supporting Information). Furthermore, six different chips were measured (Figure S3, Supporting Information) and the chip-to- chip variations were small according to the standard deviation (i.e., the error bars in Figure 2b). The RHRS/RLRS ratio is always over 50 at 2 V bias here. Both RHRS and RLRS
are able to maintain over 2 104 s without Figure 1. a,b) XRD pattern (a) and AFM surface morphology (b) of mica/SRO/BTCO. c)
Trans-
obvious decaying, and they also remain non- degradable after 105 writing/erasing (i.e., set/
parency of mica substrate and mica/SRO/BTCO sample. d,e) Optical images of bending mica/
SRO/BTCO (d) and the flat one upon a written paper (e). reset) cycles (Figure 2c). Similar to earlier
www.advancedsciencenews.com
www.advmat.de
Figure 2. a) Optical image of mica/SRO/BTCO/Au memory, where the BTCO film at bottom right corner was removed to expose bottom electrode. b) I–V curves of the flat sample, where the error bars present the standard deviation of chip-to-chip variations. c) Dependence of RHRS and RLRS on retention time (red curves) and writing/erasing cycles (black diamonds). d) Resistances change when a memory cell was set from the HRS to the LRS and then its top electrode was cut into two parts, where the sketch maps of insets show the cross section of cells with conductive filaments.
reports, the bipolar resistive switching of BTCO mainly comes from the formation and rupture of conductive filaments.[2] To verify the inhomogeneous conduction and the filament model of the memory cell, we set the memory cell to the LRS and then cut the top electrode into two parts using the probe tip, which shows the HRS and LRS, respectively (Figure 2d). The LRS part was cut into two smaller parts again, which show HRS and LRS, too (Figure S4, Supporting Information). It is argued that a conductive filament exists in the LRS part, but not in the HRS part. Although both Joule heat and migration of oxygen vacan-cies can form and rupture the conductive filament, the migra-tion of oxygen vacancies mainly introduces resistive switching in our SRO/BTCO/Au cell due to two factors.[20,21] First, both SRO and Au electrodes are too inactive to introduce a filament as the Ag electrode does, and the maximum current during fila-ment rupture is smaller than the compliance current of 1 mA. Second, the LRS parts of the I–V curves show semiconductor characteristics and fit well with I2 V, i.e., the relationship of space-charge-limited- conduction (Figure S5, Supporting Infor-mation). According to previous studies,[22] the form and rupture of filaments due to the migration of oxygen vacancies are more stable and durable than that as a result of Joule heat.
The flexibility of the mica/SRO/BTCO/Au memory satisfies the ordinary demands of flexible wearable devices and display screens. As shown in Figure 3a, in situ resistive-switching measurements were performed with our flexible memory arrays rolled on a commercial refill of a ballpoint pen, then
a number of bends were automatically carried out using a mechanical apparatus before further measurements. The sur-face morphology of the BTCO film after bending to a 1.4 mm radius is the same as that of the original flat film (Figure 3b), indicating that the bending strain did not change the micro-structure of the mica/SRO/BTCO. The 50 mm 30 mm mica/SRO/BTCO is also homogeneous and flexible as shown in Video S1 (Supporting Information). Figure 3c shows three similar I–V curves, which were measured on the flat memory cell, the cell with 2.2 mm bending radius, and the cell after
being bent to 3 mm radius for 10 000 times. Here, the VSET and VRESET of these three I–V curves are similar. Furthermore, the memory cells were folded inward to minimize the bending radius and then the normal resistive-switching effect in the unfolded cells is observed again (Figure S6, Supporting Infor- mation). Figure 3d shows the RHRS and RLRS on retention time and writing/erasing cycles for the memory cell with 2.2 mm bending radius and the unfolded cell after being bent to a
3 mm radius 10 000 times, respectively. The RHRS and RLRS of the bending and unfolded cells do not show any obvious decay within 10 000 s. Most importantly, the RHRS/RLRS ratio is always over 50 when the bending cell endures 360 000 writing/erasing cycles or when the unfolded cell endures 5000 writing/erasing cycles. In a word, the bending memory cell shows excellent resistive-switching performance, as the flat cell does, and the repeated bending does not influence the resistive-switching performance.
Adv. Mater. 2017, 1700425
1700425 (3 of 6)
© 2017 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim
Figure 3. a) Sketch map on how to measure the bent mica/SRO/BTCO/Au cells in situ and how to bend cells repeatedly with mechanical apparatus. b) AFM surface morphologies of mica/SRO/BTCO before and after it was bent to 1.4 mm radius. c) I–V curves of original flat cell, the cell with 2.2 mm bending radius and the cell after being bent to a 3 mm radius 10 000 times. d) RHRS and RLRS on retention time and writing/erasing cycles for the cell with 2.2 mm bending radius and the cell after being bent to a 3 mm radius 10 000 times, respectively.
That is to say that the bending strain at a radius of 1.4–3 mm is too small to introduce enough defects and dislocations to influence the form and rupture of conductive filament. When an h1-thick oxide film on an h2-thick substrate is bent to a radius r, the maximum bending strain is expressed as Equation (1):
δmax ≈ h1 h2
/r (1) Here, the BTCO film is much thinner than mica in our cases, so the thinner the mica is, the more flexible the memory will be.The flexible memory cell under light illumination shows the same resistive-switching characteristics as in darkness. When the flat memory and that with 2.2 mm bending radius are illu-minated under visible light with wavelength of 350–700 nm and intensity of 50, 100, and 150 mW cm−2 from a xenon lamp, their I–V curves are the same as those in darkness (Figure 4a and Figure S7, Supporting Information). When the memory was illuminated by visible light with an intensity of 150 mW cm−2,
RHRSand RLRSdo not deteriorate obviously within 10 000 s or during 5000 writing/erasing cycles (Figure 4b). The memory is promising for application in flexible display screens due to its advantages of semitransparency, flexibility, and antiradia-tion. In fact, there has been a growing interest from numerous
consumer electronics manufacturers to apply the flexible dis-play technology in e-readers, mobile phones, and other con-sumer electronics recently.
One of the most important advantages of the inorganic mica/
SRO/BTCO/Au memory is its temperature stability. As shown in Figure 4c, our memory cell can operate properly at 25–180 C or after postannealing at 500 C. Both RLRS and RHRS written at 25
C decrease with increasing temperature, and the RHRS at
180 C is over 3.3 times that of the RLRS at 25 C (Figure 4d). The RHRS and RLRS written at 180 C increase with temperature decreasing to 25 C and the final values are very close to those written at 25 C. This suggests that our resistive memory can work in the temperature range of 25–180 C safely. Besides, the dependence of RLRS on temperature shows semiconductor char- acteristics, being consistent with a conductive filament com-posed of oxygen vacancies and other ion dopants. Furthermore, the mica/SRO/BTCO/Au memory was postannealed at 500 C and then measured at 25 C again. Although all the conductive filaments already broke during the postannealing, its resistive- switching parameters and flexible performance could recover to those of the original memory after the postannealed memory underwent electric forming at room temperature (Figure S8, Supporting Information). In a word, this temperature stability is better than those of the widely used flash memory and most organic memories under study.[23,24]
Adv. Mater. 2017, 1700425
