Large memcapacitance and memristance at Nb:SrTiO3/La0.5Sr0.5Mn0.5Co0.5O3-d topotactic redox interface

Large memcapacitance and memristance at Nb:SrTiO 3 /La 0.5 Sr 0.5 Mn 0.5 Co 0.5 O 3-d topotactic redox interface Cite as: Appl. Phys. Lett. 116, 063502 (2020); doi: 10.1063/1.5131854 Submitted: 18 October 2019 .Accepted: 27 January 2020 . Published Online: 10 February 2020 W. Rom an Acevedo, 1,2 C. A. M. van den Bosch, 3 M. H. Aguirre, 4,5,6 C. Acha, 2,7 A. Cavallaro, 3 C. Ferreyra, 1,2 M. J. S anchez, 8 L. Patrone, 9 A. Aguadero, 3,a) and D. Rubi 1,2,a) AFFILIATIONS 1 Comisi on Nacional de Energ ıa At omica and Instituto de Nanociencia y Nanotecnolog ıa, Centro At omico Constituyentes, 1650, Buenos Aires, Argentina 2 Consejo Nacional de Investigaciones Cient ıficas y T ecnicas, Godoy Cruz 2290 (1425), Buenos Aires, Argentina 3 Department of Materials, Imperial College London, London SW7 2AZ, United Kingdom 4 Departmento de F ısica de Materia Condensada, Universidad de Zaragoza, Pedro Cerbuna 12, 50009 Zaragoza, Spain 5 Laboratorio de Microscop ıas Avanzada (LMA), Instituto de Nanociencia de Arag on (INA)-Universidad de Zaragoza, C/Mariano Esquillor s/n. 50018 Zaragoza, Spain 6 Instituto de Ciencias de Materiales de Arag on (ICMA), Universidad de Zaragoza, 50009 Zaragoza, Spain 7 Depto. de F ısica, FCEyN, Universidad de Buenos Aires & IFIBA, UBA-CONICET, Pab I, Ciudad Universitaria, Buenos Aires (1428), Argentina 8 INN, Centro At omico Bariloche and Instituto Balseiro, 8400 San Carlos de Bariloche, Argentina 9 Instituto de Nanociencia y Nanotecnolog ıa (CONICET-CNEA), Centro At omico Bariloche and Instituto Balseiro, 8400 San Carlos de Bariloche, Argentina a) Authors to whom correspondence should be addressed: [email protected] and [email protected]m ABSTRACT The possibility to develop neuromorphic computing devices able to mimic the extraordinary data processing capabilities of biological systems spurs the research on memristive systems. Memristors with additional functionalities such as robust memcapacitance can outperform standard devices in key aspects such as power consumption or miniaturization possibilities. In this work, we demonstrate a large memcapacitive response of a perovskite memristive interface, using the topotactic redox ability of La 0.5 Sr 0.5 Mn 0.5 Co 0.5 O 3-d (LSMCO, 0 d0.62). We demonstrate that the multi-mem behavior originates at the switchable n-p diode formed at the Nb:SrTiO 3 /LSMCO interface. We found for our Nb:SrTiO 3 /LSMCO/Pt devices a memcapacitive effect C HIGH /C LOW 100 at 150 kHz. The proof-of-concept interface reported here opens a promising venue to use topotactic redox materials for disruptive nanoelectronics, with straightforward applications in neuromorphic computing technology. Published under license by AIP Publishing. https://doi.org/10.1063/1.5131854 Neuromorphic computing devices aim at mimicking biological systems and are expected to dramatically improve the performance and efficiency of electronic devices for advanced information technology. 1 Brain synapses can be emulated by memristors, 2,3 consisting of capacitor-like structures displaying a reversible and non-volatile electrical resistance change upon the application of electrical stimulus. 4,5 Other potential applications of memristors include nanoelectronic memories 2 and logic gates. 6 Memristive behavior is ubiquitously found in transition metal oxides, including perovskite manganites. 7 The proposed memristive mechanisms for metal/manganite systems include the modulation of metal/insulator Schottky barriers due to oxygen vacancy (OV) electromigration, 8–10 or the interfacial redox reaction occurring when a reactive electrode (Ti or Al) is used. 11–13 In these cases, oxygen exchange with the environment is neglected. 9 Alternatively, reports on memristive perovskites claiming for oxygen exchange between the perovskite and the atmosphere 14 or the metallic electrode 15 can be found. For volumetric redox processes triggered by electrical stimuli, a robust memristive effect is expected for perovskites displaying topotactic redox ability, i.e., the capability of reversibly storing and releasing oxygen with slight structural changes that maintain Appl. Phys. Lett. 116, 063502 (2020); doi: 10.1063/1.5131854 116, 063502-1 Published under license by AIP Publishing Applied Physics Letters ARTICLE scitation.org/journal/apl the perovskite structure. 16,17 Here, the memristive effect relies on the electrical switch between oxidized and reduced phases with different electrical conductivities. A better stability for redox memristive behavior is envisaged for topotactic perovskites in comparison with standard ones, 18,19 as the structure changes—i.e., change in the perovskite space group 17 in the former case allows transitions between structures presenting well defined minimum energies. Memcapacitance—non-volatile change of a device capacitance C upon the application of electrical stress is an additional functionality of memristors that has scarcely been explored. 20–26 The proposed mechanisms include creation/annihilation of conducting nanofilaments, 20,21 modulation of Schottky barriers at interfaces, 21,22 oxidation/reduction of a TiO x active layer 23 or changes in the oxide permittivity upon OV electromigration. 24,25 While applications for memcapacitance, including neuromorphic computing devices, have been proposed, 27 the interest in this phenomenon has been hampered by the small reported figures to date (C HIGH /C LOW 10). 20–26 It was shown that associated capacitive networks, suitable for efficient pattern recognition, can be built from cells able to switch their capacitance between C HIGH and C LOW ,where the array size—linked to the device computing capability—scales with a C HIGH /C LOW ratio. 28 This evidences the high technological interest of memcapacitive systems with a large response. In this paper, we show that the interface between the topotactic redox perovskite manganite La 0.5 Sr 0.5 Mn 0.5 Co 0.5 O 3-d (LSMCO, 0 d 0.62, p-type) 17 and Nb:SrTiO 3 (NSTO, n-type) behaves as a switchable n-p diode with memristive and large memcapacitive behavior. We show that the observed multi-mem behavior is related to the electrical switch between LSMCO oxidized (d¼0) and reduced (d0.62) phases. Oxidized phase LSMCO thin films were grown by laser ablation on Nb:SrTiO 3 (0.5 wt. %, 001) substrates. The growth temperature, oxygen pressure, and laser fluence were fixed at 800 C, 0.04 mbar, and 0.5–1 J/cm 2 , respectively. Pristine LSMCO films were 20 nm thick and epitaxial, as shown in supplementary material Fig. S1. High resolution scanning transmission electron microscopy (STEM) was performed using a FEI Titan G2 microscope with a probe corrector. Microstructured top Pt electrodes, 20 nm thick, were fabricated by FIB or optical lithography. The platinum high work function (5.6eV) allows an Ohmic interface with p-type LSMCO. 29 The NSTO substrate wasgrounded(adropofsilverpaintwasusedtomakecontact)and the electrical stimulus was applied to the top Pt electrode. For the electrical characterization, we used a Keithley 2612 source-meter, an AutoLab PGSTAT302N impedance analyzer, and a standard LCRmeter. The virgin resistances of 37.5 10 3 lm 2 NSTO/LSMCO/Pt devices were 1MX, and the current–voltage (I–V) curve for low stimulus, shown in supplementary material Fig. S2, displays a rectifying behavior linked to the formation of an n-p diode at the NSTO/ LSMCO interface. A forming process is triggered when a 7 V pulse is applied, resulting in a sudden resistance drop to 50 X(supplementary material Fig. S2). After forming, dynamic I–V curves were obtained by applying a sequence of 1 ms voltage pulses of different amplitudes (0 !V MAX !V MIN !0), with the current measured during the application of the pulse. Additionally, after each voltage pulse a small reading voltage (100mV) was applied to determine the remnant resistance states, obtaining the hysteresis switching loops (HSLs). Figures 1(a) and 1(b) display typical I–V and HSL curves, both demonstrating the memristive properties of our devices. The device switches from a low resistance (R LOW ) to high resistive state (R HIGH ) (RESET process) upon the application of þ5.5 V, while the opposite behavior (SET process) is observed upon the application of –1.5 V. From the second cycle, R HIGH and R LOW stabilize to 2–6 kXand 100–200 X, respectively, giving an average ON/OFF ratio of 25. Figure 1(c) shows an endurance test with a stable behavior for 200 cycles, while Fig. 1(d) shows retention times of at least 10 4 s for both resistive states. Interestingly, the observed memresistance is concomitant with a large memcapacitive effect. Dynamic capacitance–voltage (C–V) curves were obtained by measuring the device capacitance upon the application of DC voltage pulses of increasing amplitude with a small superimposed AC signal (10 kHz, amplitude 200 mV). Figure 2(a) displays dynamic C–V curves measured on devices prepared in both R HIGH and R LOW states. An evident difference between both curves is observed, indicating a significant capacitance change between R HIGH and R LOW states, where the R HIGH (R LOW ) state corresponds to a low (high) capacitance C LOW (C HIGH ) one. The negative capacitance found for C LOW for positive voltages is attributed to the non-monotonic or positive-valued behavior of the time-derivative of the transient current in response to a small voltage step. 30 The memcapacitive effect was confirmed by remnant capacitance measurements, performed in both states by applying a pure AC signal at different frequencies. Figure 2(b) displays the evolution of both remnant C LOW and C HIGH states as a function of the frequency (f) of the excitation signal. C LOW 3.5 pF at 10 kHz and displays a subtle decrease with f, while C HIGH displays a stronger decrease, from 3nFat10kHzto0.13 nF at 300 kHz. The capacitance decrease is at a higher rate for f<100 kHz. The existence of leakage channels at the diode interface likely increases its effective capacitance at low f, 31 but other effects such as the presence of surface states might also contribute to the dependence of the capacitance with f. 32 In the low frequency range, there is a rapid drop in the C HIGH / FIG. 1. (a) Dynamic pulsed I–V curve recorded on a formed device, with arrows indicating the circulation direction of the curve. The inset displays the I–V curve corresponding to a non-formed device; (b) experimental hysteresis switching loop (HSL, symbols), recorded simultaneously with the I–V curve. The simulated HSL is displayed with a solid line; (c) retention experiments corresponding to a NSTO/ LSMCO/Pt device, for both R LOW and R HIGH states; (d) endurance test performed by applying single SET and RESET voltage pulses with opposite polarities. Applied Physics Letters ARTICLE scitation.org/journal/apl Appl. Phys. Lett. 116, 063502 (2020); doi: 10.1063/1.5131854 116, 063502-2 Published under license by AIP Publishing C LOW ratio, from 900 at 10 kHz to 130 at 100 kHz, likely due to leakage effects. For higher frequencies, leakage effects should not contribute 31 and the dependence of C HIGH /C LOW with f is milder. We obtained C HIGH /C LOW 100 at 150 kHz, which is around one order of magnitude larger than memcapacitive figures reported to date for other systems. 20–26 The ability of our devices to reversibly change their capacitance between two non-volatile states upon consecutive cycling is confirmed by Fig. 2(c),whileFig. 2(d) displays retention times for both C HIGH and C LOW higher than 10 3 s, respectively. Figure 3(a) shows a scanning electron microscopy top-view of the device after forming, where three distinct zones are identified. Zone 1 has a diameter of 5lm and corresponds to the contact position of the tip. Zone 2 is a ring of higher contrast than Zone 1 (diameter: 15 lm) where the material from the film and the Pt electrode has been expelled during the forming process probably due to the release of O 2 gas. 33 At Zone 1, STEM-HAADF analysis shows that, upon forming, the pristine LSMCO epitaxial nanostructure [shown in supplementary material Fig. S3(a)] re-crystallizes due to selfheating effects. This re-crystallization can comprise the complete LSMCO thickness, leading to an arrangement of nanograins as the one displayed in Figs. 3(b) and S3(c). The bottom LSMCO nanograins, in contact with the NSTO substrate, retain a (001) out-of-plane orientation, while the top grains, in contact with the Pt top electrode, are in general not coherent with the bottom grains and present a tilted (001) direction, as shown in the zoomed-in image of Fig. 3(c). LSMCO recrystallization at Zone 1 can also be partial, leading to top LSMCO non-coherent nanograins located on top of epitaxial LSCMO [see Fig. S3(c)]. The LSMCO re-crystallization process is likely driven by the presence of thermal self-accelerated effects 34 where both the Pt electrode [see supplementary material Figs. S3(b) and S3(c)] and (part of) LSMCO melt during electroforming, followed by a fast cooling after theendoftheformingvoltagethatquenchesmeltedLSMCOintoa non-coherent nanograins arrangement. TEM analysis also shows that part of LSMCO becomes reduced upon during forming. Figure 4 shows atomic images and Fast Fourier Transforms corresponding to two nanograins at Zone 1. The nanograin displayed in Figs. 4(a) and 4(b), located close to the LSMCO/NSTO interface, remains structurally very similar to pristine (oxidized) LSMCO, while the one displayed in Figs. 4(c) and 4(d), located close to the Pt top electrode, is reduced and presents an ordered structure of OV that double the LSMCO unit cell along the (001) direction, resembling previous reports in brownmillerite. 35 These grains are likely the source of the released O 2 during forming.Zone3displaysasimilarcontrasttoZone1andpresentsan epitaxial structure with the presence of extended defects, as analyzed in supplementary material Fig. S4. All remnant resistive and capacitive statesareindependentofthe(virgin)devicearea(supplementary material Fig. S5), indicating that the forming process electrically decouples Zone 1 from the rest of the device. Thus, the multi-mem behavior is confined to Zone 1, presenting an effective area of 80 lm 2 . We propose that the LSMCO multi-mem behavior is related to its topotactic redox ability where oxidation (reduction) of LSMCO nanograins in Zone 1 is associated with the SET (RESET) process. This scenario is supported by memresistance experiments performed in vacuum (<110 2 mbar) where it is found that the SET event is not achieved (supplementary material Fig. S6), indicating that environmental O 2 is critical for the SET process through LSMCO oxidation. Further evidence about the link between memresistance and LSMCO redox was obtained by simulating the experimental HSL [Fig. 1(b)] with the voltage enhance OV drift model, 8,9,13 adapted to the present system. The simulation assumes a 1D chain of LSMCO nanodomains, able to accommodate different oxygen content which controls their resistivity, in contact with an oxygen reservoir (see the supplementary material for further details). The model simulates the oxygen dynamics related to the electrically induced LSMCO oxidation FIG. 2. (a) Dynamic capacitance–voltage (C–V) curves recorded both for oxidized (R LOW and C HIGH ) and reduced (R HIGH and C LOW ) LSMCO states; (b) evolution of remnant C HIGH and C LOW states with the frequency of the external AC signal; (c) reversible switch between C HIGH and C LOW (measured at 10 kHz) after the application of SET and RESET single pulses. C HIGH and C LOW values were in the ranges 3–20 nF and 3–9 pF, respectively; (d) retention experiments for C HIGH and C LOW states, measured at 10 kHz. FIG. 3. (a) Scanning electron microscopy image (top-view) of a formed device. Three zones are identified and described in the text; (b) STEM-HAADF cross section (Zone 1) corresponding to a formed device; (c) higher magnification STEMHAADF cross section evidencing the non-coherent nature of LSMCO nanograins at Zone 1 after forming. Atomic planes are indicated with green lines and the grain boundary is marked with white dotted line; (d) sketch of the forming, RESET, and SET processes in the NSTO/LSMCO/Pt devices. After forming, epitaxial LSMCO is re-crystallized and nanograins are formed. The transition between R LOW (C HIGH ) and R HIGH (C LOW ) is related to LSMCO oxidation and reduction. Applied Physics Letters ARTICLE scitation.org/journal/apl Appl. Phys. Lett. 116, 063502 (2020); doi: 10.1063/1.5131854 116, 063502-3 Published under license by AIP Publishing and reduction. The simulated HSL shows a very good agreement with the experimental one, as shown in Fig. 1(b). Further insight into the memristive mechanism was obtained from the analysis of the I–V curves corresponding to the two resistive states, by plotting the power exponent c¼d(ln(I))/dln(V) vs V 1/2 . 36 This method is useful for identifying the presence of multiple conduction mechanisms, 37–39 which often occurs at metal/complex oxide interfaces. 40–42 Figure 5(a) shows the complex evolution of cvs V 1/2 for both R HIGH (C LOW ) and R LOW (C HIGH ), indicating the presence of several circuit elements with relative weights that change between these states. The equivalent circuit that describes the cvs V 1/2 behavior is shown in Fig. 4(e) and includes the series combination of: (i) an n-p diode in parallel with a leakage channel R 1 , corresponding to the NSTO/LSMCO interface, (ii) the series resistor R 2 associated with the Ohmic conduction of non-interfacial LSMCO plus the LSMCO/Pt interface, and (iii) a Schottky diode linked to the external Ag/NSTO contact. The experimental I–V curves were fitted by numerically solving the implicit I–V equations of the circuit (see the supplementary material), and all circuit parameters were extracted and are listed in supplementary material Table S1. Figures 5(a) and 5(b) show the excellent fits of the experimental and cvs V 1/2 and I–V curves for both R HIGH (C LOW )andR LOW (C HIGH ) states. The fittings indicate that the transition between R HIGH (C LOW )andR LOW (C HIGH )isdominatedby the metallization of the LSMCO/NSTO interface, reflected in the increase in the p-n diode inverse saturation I satpn from 2.5 to 5.6 lA and the decrease in the leakage resistance R 1 from 270 to 110 X. Impedance spectroscopy (IS) was performed to further investigate the AC response of the different multi-mem states. Figures 5(c) and 5(d) display the Cole–Cole plots for R HIGH (C LOW )andR LOW (C HIGH ), respectively. The equivalent circuit that allows a good fit of the R LOW (C HIGH ) state is shown in Fig. 5(f),andwasalsousedtosimulatethe R HIGH (C LOW ) spectrum. The fitted values for the circuit elements are shown in supplementary material Table S2. Again, it is found that the multi-mem effect is mainly localized at the NSTO/LCMO interface, characterized by the parallel combination of R  1 and C  1 [see Fig. 5(f)]. R  1 changes from 10 Xto 500 Xand C  1 from 310 8 F to 110 12 F between R LOW (C HIGH ) and R HIGH (C LOW ), reinforcing the idea of the presence of a switchable diode formed at the NSTO/LSMCO interface. Thus, an oxidized NSTO/LSMCO interface leads to a R LOW (C HIGH ) state, and a reduced interface results in a R HIGH (C LOW ) state. The key factor for this behavior is the topotactic redox ability of LSMCO, which tolerates large changes in its oxygen content between oxidized and reduced phases. The physical origin of the large memcapacitance is intriguing and can be attributed to different effects. A possible one is related to large variations in the donor/acceptor balance at the NSTO/LSMCO interface upon LSMCO redox. The p-character of oxidized LSMCO is determined by the 0.5 holes/f.u. introduced in the lattice when Sr 2þ ions replace La 3þ . In the reduced phase, each OV leaves two electrons behind, and part of these electrons will recombine with existing holes, reducing the number of uncompensated LSMCO acceptors. This could strongly change the (NSTO) donor/(LSMCO) FIG. 4. (a) and (b) Atomic image and FFT, respectively, corresponding to a postforming LSMCO nanograin (Zone 1) in contact with the NSTO substrate. An oxidized LSMCO perovskite structure is observed; (c) and (e) Same analysis for a LSMCO nanograin in contact with the Pt top electrode. The grain structure is perovskite-type, but with ordered OV that double the unit cell along the (001) direction (see additional diffraction spots, circled in red), indicating the presence of reduced LSMCO. The zone axis is [1,10] for both FFTs. These images correspond to a different lamella from that in images displayed in Fig. 3. FIG. 5. (a) cvs V 1/2 representation and (b) corresponding I–V curves for both the experimental (open symbols) and the calculated (solid line) R HIGH (C LOW )and R LOW (C HIGH ) states. The fits are performed by considering the (DC) equivalent circuit presented in (e). C 1 and C 2 were included for completeness but have no effect in the low frequency range used for the I–V experiments; (c) and (d) impedance spectroscopy spectra recorded for R HIGH (C LOW )and R LOW (C HIGH ) states. The experimental points are shown with open symbols, while the fittings are shown with solid lines; (f) equivalent (AC) circuit proposed to model the experimental impedance spectra. R  3 //C  3 corresponds to the LSMCO/Pt interface, R  1 //C  1 to the NSTO/LSMCO interface, and the series resistance R  2 includes the non-interfacial LSMCO plus the contribution of the Ag/NSTO interface. We note that R 2 in (e) includes both R  2 and R  3 of the AC equivalent circuit of (f) and that the n-p diode AC contribution is included in R 1  . Applied Physics Letters ARTICLE scitation.org/journal/apl Appl. Phys. Lett. 116, 063502 (2020); doi: 10.1063/1.5131854 116, 063502-4 Published under license by AIP Publishing acceptor balance at the interface, affecting the diode depletion layer and its capacitance. Other possible origin for the large memcapacitance is the Maxwell-Wagner effect, related to the creation of metallic zones embedded in the dielectric oxide, acting as the parallel plates of nanocapacitors, 43 that could produce large effective dielectric constants. 44 The metallization of the NSTO/LSMCO interface in the R LOW (C HIGH ) state could take place inhomogeneously, generating an interfacial capacitance significantly larger than that expected for a homogeneous interface. Further studies are necessary to confirm and get deeper insight into these mechanisms, together with the development of strategies to circumvent the electroforming process. See the supplementary material for additional structural and electrical characterization of our devices, together with details about the simulation and fittings of the electrical response. D.R. acknowledges financial support from ANPCyT, Project Nos. PICTs 0867 and 1836. Helpful discussions with S. Menzel, P. Stoliar, M. Rozenberg V. Ferrari, and P. Levy are acknowledged. We thank F. Golmar, from INTI, for the access to the FIB facility. A.A., C.vdB., and A.C. acknowledge the support of the Engineering and Physical Sciences Research Council (EPSRC), Grant Nos. EP/ M014142/1, EP/P026478/1, and EP/L504786/1. A.A. and A.C. also acknowledge FETPROACT-2018-2020 “HARVESTORE” 824072 project. M.A. acknowledges financial support of H2020-MSCARISE-2016 SPICOLOST Grant No. 734187 to perform TEM studies at LMA-INA, University of Zaragoza. REFERENCES 1 S. Yu, Neuro-Inspiring Computing Using Resistive Synaptic Devices (Springer International Publishing, 2017). 2 D. Ielmini and R. Waser, Resistive Switching from Fundamentals of Nanoionic Redox Processes to Memristive Device Applications (Wiley-VCH, 2016). 3 Z. Wang, S. Joshi, S. E. Savel’ev, H. Jiang, R. Midya, P. Lin, M. Hu, N. Ge, J. P. Strachan, Z. Li, Q. Wu, M. Barnell, G. L. Li, H. L. Xin, R. S. Williams, Q. Xia, and J. J. Yang, Nat. Mater. 16, 101 (2017). 4 A. Sawa, Mater. Today 11, 28 (2008). 5 R. Waser, R. Dittmann, G. Staikov, and K. Szot, Adv. Mater. 21, 2632 (2009). 6 J. Borghetti, G. S. Snider, P. J. Kuekes, J. J. Yang, D. R. Stewart, and R. S. Williams, Nature 464, 873 (2010). 7 S. Q. Liu, N. J. Wu, and A. Ignatiev, Appl. Phys. Lett. 76, 2749 (2000). 8 M. J. Rozenberg, I. H. Inoue, and M. J. S anchez, Phys. Rev. Lett. 92, 178302 (2004). 9 M. J. Rozenberg, M. J. S anchez, R. Weht, C. Acha, F. Gomez-Marlasca, and P. Levy, Phys. Rev. B 81, 115101 (2010). 10 D. Rubi, F. Tesler, I. Alposta, A. Kalstein, N. Ghenzi, F. Gomez-Marlasca, M. Rozenberg, and P. Levy, Appl. Phys. Lett. 103, 163506 (2013). 11 A. Herpers, C. Lenser, C. Park, F. Offi, F. Borgatti, G. Panaccione, S. Menzel, R. Waser, and R. Dittmann, Adv. Mater. 26, 2730 (2014). 12 S. Asanuma, H. Akoh, H. Yamada, and A. Sawa, Phys. Rev. B 80, 235113 (2009). 13 W. Rom an Acevedo, C. Ferreyra, M. J. S anchez, C. Acha, R. Gay, and D. Rubi, J. Phys. D: Appl. Phys. 51, 125304 (2018). 14 J. C. Gonzalez-Rosillo, R. Ortega Hernandez, B. Arndt, M. Coll, R. Dittmann, X. Obradors, A. Palau, J. Su~ ne, and T. Puig, Adv. Electron. Mater. 5, 1800629 (2019). 15 D. Cooper, C. Baeumer, N. Bernier, A. Marchewka, C. La Torre, R. E. DuninBorkowski, S. Menzel, R. Waser, and R. Dittmann, Adv. Mater. 29, 1700212 (2017). 16 A. Nemudry, E. Goldberg, M. Aguirre, and M. Alario-Franco, Solid State Sci. 4, 677–690 (2002). 17 A. Aguadero, H. Falcon, J. M. Campos-Martin, S. M. Al-Zahrani, J. L. G. Fierro, and J. A. Alonso, Angew. Chem. Int. Ed. 50, 6557 (2011). 18 S. K. Acharya, N. V. Raveendra, O. Togibasa, B. W. Lee, C. Liu, C. U. Jung, B. H. Park, J.-Y. Park, Y. Cho, D.-W. Kim, J. Jo, D.-H. Kwon, M. Kim, C. S. Hwang, and S. C. Chae, ACS Appl. Mater. Interfaces 8, 7902 (2016). 19 S. K. Acharya, J. Janghyun, N. Venkata Raveendra, U. Dash, M. Kim, H. Baik, S. Lee, B. H. Park, J. S. Lee, S. C. Chae, C. S. Hwang, and C. U. Junga, Nanoscale 9, 10502 (2017). 20 I. Salaoru, A. Khiat, Q. Li, R. Berdan, and T. Prodromakis, Appl. Phys. Lett. 103, 233513 (2013). 21 S. X. Wu, H. Y. Peng, and T. Wu, Appl. Phys. Lett. 98, 093503 (2011). 22 A. A. Bessonov, M. N. Kirikova, D. I. Petukhov, M. Allen, T. Ryh€ anen, and M. J. Bailey, Nat. Mater. 14, 199 (2015). 23 I. Salaoru, Q. Li, A. Khiat, and T. Prodromakis, Nanoscale Res. Lett. 9, 552 (2014). 24 P. Yang, H. Jun Kim, H. Zheng, G. Won Beom, J. S. Park, C. Jung Kang, and T. S. Yoon, Nanotechnolgy 28, 225201 (2017). 25 D. Park, P. Yang, H. J. Kim, K. Beom, H. H. Lee, C. J. Kang, and T. S. Yoon, Appl. Phys. Lett. 113, 162102 (2018). 26 R. Liu, R. Dong, X. Yan, S. Yuan, D. Zhang, B. Yang, and X. Xiao, Appl. Phys. Express 11, 114103 (2018). 27 D. Tran and C. Teuscher, preprint arXiv:1704.05921 (2017). 28 L. Nielen, A. Siemon, S. Tappertzhofen, R. Waser, S. Menzel, and E. Linn, IEEE J. Emerg. Sel. Top. Circuits Syst. 5, 153 (2015). 29 A. Herpers, “Electrical characterization of manganite and titanate structures,” Ph.D. thesis (Forschungszentrum J€ulich, 2014). 30 M. Ershov, H. C. Liu, L. Li, M. Buchanan, Z. R. Wasilewski, and A. K. Jonscher, IEEE Trans. Electron Devices 45, 2196 (1998). 31 N. Novkoski and E. Atannasova, Adv. Mater. Sci. Eng. 2017, 9746934. 32 A. E. Sapega, Phys. Status Solidi A 42, 585 (1977). 33 J. Joshua Yang, F. Miao, M. D. Pickett, D. A. A. Ohlberg, D. R. Stewart, C. N. Lau, and R. Stanley Williams, Nanotechnology 20, 215201 (2009). 34 U. Russo, D. Ielmini, C. Cagli, and A. L. Lacaita, IEEE Trans. Electron Devices 56, 193 (2009). 35 L. Yao, S. Inkinen, and S. van Dijken, Nat. Commun. 8, 14544 (2017). 36 C. Acha, J. Appl. Phys. 121, 134502 (2017). 37 C. Acha, A. Schulman, M. Boudard, K. Daoudi, and T. Tsuchiya, Appl. Phys. Lett. 109, 011603 (2016). 38 W. R. Acevedo, C. Acha, M. J. S anchez, P. Levy, and D. Rubi, Appl. Phys. Lett. 110, 053501 (2017). 39 N. Ghenzi, M. Barella, D. Rubi, and C. Acha, J. Phys. D: Appl. Phys. 52, 125401 (2019). 40 C. Acha, J. Phys. D: Appl. Phys. 44, 345301 (2011). 41 B. Ghosh, K. Das, and A. K. Raychaudhuri, J. Appl. Phys. 109, 083934 (2011). 42 F. Gomez-Marlasca, N. Ghenzi, A. G. Leyva, C. Albornoz, D. Rubi, P. Stoliar, and P. Levy, J. Appl. Phys. 113, 144510 (2013). 43 A. R. Von Hippel, Dielectrics and Waves (John Wiley and Sons, 1962), pp. 228–234. 44 P. Lunkenheimer, S. Krohns, S. Riegg, S. G. Ebbinghaus, A. Reller, and A. Loidl, Eur. Phys. J. Spec. Top. 180, 61 (2009). Applied Physics Letters ARTICLE scitation.org/journal/apl Appl. Phys. Lett. 116, 063502 (2020); doi: 10.1063/1.5131854 116, 063502-5 Published under license by AIP Publishing