Graphene Quantum Dots as an Oxygen Reservoir for Topotactic Phase Transition‐Based Memristive Devices

A novel class of transition metal oxides, capable of reversible topotactic phase transition between the oxygen‐deficient brownmillerite and oxygen‐rich perovskite, has emerged as a promising material for memristive and magnetoelectric devices. However, the absence of a local oxygen source in the device structure necessitates an oxygen exchange process between the surrounding atmosphere and the switching layer during operation, which can lead to unreliable device performance. In this study, graphene quantum dots (GQDs) are introduced into a SrFe0.5Co0.5Ox memristive device as an oxygen reservoir for the nanoscale topotactic redox process. The SrFe0.5Co0.5Ox memristive devices with GQDs exhibit reliable resistive switching performance compared to SrFe0.5Co0.5Ox devices without GQDs. To understand the effect of GQDs on the device structure, a pulse endurance test is carried out in a high vacuum. The devices with GQDs show rather good endurance behavior, while devices without GQDs exhibit endurance failure. These results provide a deeper understanding of the potential use of GQDs in enhancing the performance of SrFe0.5Co0.5Ox memristive devices, with implications for tuning nanoscale topotactic phase transition for multi‐functional properties.


Introduction
[3][4] However, the poor ionic migration in conventional metal oxides makes it challenging to achieve robust switching states. [5,6]To achieve more reliable switching, processes involving volumetric phase transitions induced by voltage-driven ion migration are necessary. [1,4,7]The brownmillerite (BM) structure enables reversible topotactic phase transition via a redox process by substantially varying the oxygen composition, leading to a range of physical and chemical properties, including electrical conductivity and magnetic states. [8,9]This characteristic is highly attractive for multiplex writing/reading of electric and magnetic signals, which is crucial for hybrid non-volatile memory applications. [6,10]ecently, we introduced BM-SrCoO 2.5 and BM-SrFeO 2.5 as novel material platforms for resistive switching devices by harnessing their exceptional oxygen ion transport properties. [11,12]he BM materials are well-known for their high oxygen ion conductivity at low temperatures due to the presence of ordered oxygen vacancy channels (OVCs). [13,14]Our previous work demonstrated a reliable resistive switching performance in BM-SrFeO 2.5 devices, with high endurance (>10 6 cycles), fast switching speed (10 ns), and high uniformity in key switching parameters. [15]In situ X-ray absorption spectromicroscopy and in situ TEM experiments on SrFeO x devices revealed a reversible topotactic phase transition between an insulating BM-SrFeO 2.5 and a conductive SrFeO 3 perovskite (PV) phase as the cause of the resistance change. [9,16]In contrast to other oxide-based memristive device materials, the conducting filament in SrFeO x devices is composed of a fully oxidized PV-SrFeO 3 phase, while the surrounding insulator matrix is an oxygen-deficient BM-SrFeO 2.5 phase. [9,17,18]he chemical species that mediate the conduction path formation in SrFeO x devices are oxygen ions, which was experimentally demonstrated through spectromicroscopy studies. [9,19]Additionally, we observed that during the forming process, the negative bias applied to the top electrode (TE) can incorporate oxygen species from the atmosphere (e.g., O 2 or H 2 O) into the BM-SrFeO 2.5 switching insulating layer, which then drifts toward the bottom electrode (BE) interface. [17]This oxygen exchange process between the surrounding atmosphere and the switching layer during operation is commonly also observed in other memristive devices. [20]It is as well-known that the oxygen exchange process leads to poor repeatability of resistive switching and their properties due to the uncontrolled nature of conducting filaments. [21,22]Therefore, a controlled oxygen supply in memristive devices is necessary for a reliable resistive switching process.
Graphene quantum dots (GQDs) are nanometer-sized fragments of graphene that are functionalized at the edges with a significant volume of oxygen functional groups. [23]These GQDs can be easily detached using electrical fields.The quantum confinement and edge effects in GQDs generate localized electrical fields, which makes them highly attractive for controlled conducting filament formation in resistive switching devices. [24]Similar to metal nanodots (Ru and Pt), GQDs may act as preferential sites for conducting filament formation. [25,26]In contrast to metal nanodots, which often require additional experimental steps and ex situ processes, GQDs can be readily spin-coated onto the top of the switching layer.Additionally, the unique combination of localized oxygen reservoirs and the electric field distribution generated by GQDs can result in superior and reliable resistive switching performance. [27,28]Therefore, GQDs represent a promising approach for inducing multiple conductivity and magnetic states in BM devices via nanoscale topotactic redox reactions.
In this study, we have prepared epitaxial thin films of SrFe 0.5 Co 0.5 O x (SFCO), which is highly interesting for tuning multiple physical properties, including electrical and magnetic at temperatures above room temperature. [6]However, in the present investigation, we will be focusing on voltage-controlled electrical properties (i.e., memristive behavior) of SFCO thin film.To the best of our knowledge, there has been no investigation of memristive behavior in the SFCO compound prior to our study.As prepared SFCO memristive devices exhibited an eightwise resistive switching behavior.For improving switching properties, we have introduced GQDs as an oxygen reservoir for the nanoscale topotactic redox process in an SFCO device structure.Notably, the incorporation of GQDs into the SFCO memristive devices resulted in improved and consistent resistive switching performance compared to devices without GQDs.Moreover, a pulse endurance test conducted under a high vacuum revealed that devices with GQDs displayed favorable endurance behavior, while those without GQDs failed.To explain the resistive switching behavior and corresponding nanoscale topotactic phase transition, we developed a physical model based on joule heating and oxygen ions migration between the GQDs and the active switching layer.

Results and Discussion
Figure 1a shows the high-resolution X-ray diffraction (XRD) pattern of as-grown BM-SFCO/SrRuO 3 (SRO) heteroepitaxial thin film on the SrTiO 3 (001) substrate (STO).The diffraction pattern represents a single-phase nature of (001) oriented thin film.Two distinct peaks observed at the (002) peaks (enlarged peaks at the inset) belong to reflection from BM-SFCO and bottom SRO layers and the average out-of-plane lattice constant of the BM-SFCO and SRO is 3.972 and 3.951 Å, respectively.No half-ordered BM diffraction peaks appear, which suggests that the alternative stacking of the octahedral and tetrahedral layers of the BM structure is along the in-plane direction of the thin film. [13,29]The Xray rocking-curve analysis of the (002) peak of BM-SFCO shown in Figure 1b, revealed a full width at a half-maximum of 0.07°, which indicates that the film had excellent crystallinity.Reciprocal space mapping measurements were conducted around the (103) STO Bragg reflection to analyze the in-plane alignment and strain of BM-SFCO/SRO, which is shown in Figure 1c.The results revealed that SFCO, SRO, and STO reflections were aligned in the same pseudomorphic line, indicating that the heteroepitaxial growth of the material on the STO substrate was highly strained.Interestingly, the reflections from the SFCO and SRO are well separated along the out-of-plane axis as that of XRD in Figure 1a, signifying different out-of-plane lattice constants.The surface morphology of the film is shown in Figure 1d.The film shows a rather smooth surface over a relatively large area (10 × 10 μm) with root-mean-square roughness ≈0.7 nm.This observation could be understood by comparing with our previous BM-SrCoO 2.5 [30] and BM-SrFeO 2.5 [12] thin film surfaces; the BM-SrFeO 2.5 can be grown on an atomically smooth surface (RMS of ≈0.4 nm) because of the layer-by-layer growth mode, while unfortunately, BM-SrCoO 2.5 does not have layer-by layer growth behavior, resulting in a rather rough surface (RMS of ≈2 nm).Therefore, we believe that different competing (or combination) growth modes of BM-SFCO thin films resulted in a rather smooth surface.
Figure 2a,b depicts two types of memristive devices, namely i) without GQDs and ii) with GQDs at the TE interface in an Au/BM-SFCO/SRO/STO structure, respectively.Hereinafter, the devices without GQDs are referred to as SFCO devices, while those devices with GQDs are called G-SFCO devices.In Figure 2a, the blue polyhedrons represent the tetrahedral (FeO 4 ) layer, while the yellow polyhedrons represent the octahedral (FeO 6 ) layer in the BM-SFCO structure.The FeO 4 layers consist of vertical ordered OVCs along out-of-plane direction.On the other hand, in Figure 2b, GQDs are shown to be present at the TE interface, consisting of oxygen functional groups, such as hydroxyl (─OH), carboxyl (─COOH), and epoxide (─O─).The detailed chemical and microstructural properties of the BM-SFCO structure will be discussed in the following figures.
For understanding GQDs dispersion after the spin coating, the surface morphology of G-SFCO the device was investigated using field emission scanning electron microscopy (FESEM), which is shown in Figure 3a.GQDs are well distributed (granular structures) over the BM-SFCO surface.However, the GQDs might have aggregated (average diameter of 50 nm) rather than monodispersed. [23]Figure 3b shows a cross-sectional transmission electron microscopy (TEM) image of G-SFCO devices.An overview TEM image of device cross-section is presented in Figure S1 (Supporting Information).The TEM image exhibited atomically flat interfaces and sharp boundaries between the SRO and STO layers and between the BM-SFCO and SRO layers.The observed thicknesses of the SRO and BM-SFCO layers were ≈70 and ≈100 nm, respectively.The spin-coated GQD can also be seen between Au TE and BM-SFCO with a diameter of ≈50 nm.Interestingly, the interface between the GQD and BM-SFCO is visible with a bright circular-like boundary.Figure 3c displays a highresolution STEM-HAADF cross-section corresponding to a BM-SFCO film, evidencing in-plane-ordered alternative octahedra (FeO 6 or CoO 6 ) and tetrahedra (FeO 4 or CoO 6 ) layers (dark strips) in the BM-SFCO layer.It is expected that the in-plane-ordered oxygen-deficient tetrahedral layers (FeO 6 or CoO 6 ) in the BM-SFCO facilitate the formation of vertical ordered OVCs.The observed microstructural properties of BM-SFCO structure are consistent with the XRD measurements in Figure 1a.Unfortunately, it can be noticed that an additional rather large dark stripe, identified by the green-dashed rectangle in Figure 3c, was observed.This stripe, which corresponds to a planar defect, formed vertically and was present even within the films (see Figure 3b).These unwanted stacking faults were also observed in our previous BM-SrFeO 2.5 thin films, where the planar defects were formed horizontally for stabilizing b-axis growth (i.e., out-of-plane) of the octahedral and tetrahedral layers. [31]It can be therefore inferred that the vertical stacking faults in the BM-SFCO structure might result in the stabilization of the in-plane growth of alternative octahedral and tetrahedral layers (i.e., a-axis growth).
Raman spectra of GQDs on the BM-SFCO surface exhibit a disordered (D) band at 1405 cm −1 and a crystalline (G) band at 1719 cm −1 (see Figure 3d).The intensity ratio of the D to the G band (I D /I G ) is 0.92, which suggests a high occurrence of structural defects in GQDs. [32]In Figure 3e, the X-ray photoelectron spectroscopy (XPS) analysis of GQDs/SFCO films shows an asymmetric O 1s peaks, which can be attributed to the different chemical states of oxygen present in the system.The fitting of the spectra using Gaussian functions reveals three distinct peaks with energies of 528.92, 530.72, and 531.40 eV.The peak at 528.92 eV is assigned to the O 2− ions in the BM-SFCO lattice, which are commonly referred to as lattice oxygen (O L ). [33,34] The peak at 530.72 eV is attributed to the presence of oxygen-deficient regions, such as oxygen vacancies (O V ). [33]The highest energy peak at 531.40 eV is assigned to carbonates and chemisorbed oxygens (O C ) on the film surface. [28,35]The maximum intensity of the peak at 531.40 eV may be attributed to the abundance of GQDs with oxygen functional groups present on the film surface. [23,28]imilarly, the high-resolution C 1s spectra (Figure 3f) reveal that GQDs consist mostly of C═O, C─C, C─O, and C─O, with corresponding energy levels of 284.1, 285.3, 287.5, and 289.8 eV, respectively.Notably, the intensity of C─C is much stronger than that of C─O, indicating the prevalence of oxygen functional groups in GQDs. [27]s-prepared SFCO memristive devices exhibit eightwise bipolar resistive switching behavior after an initial forming step, as shown in Figure 4a.When an initial bias was applied at the TE along the negative direction for the forming process, the devise resistance changed from the pristine state to high resistance state (HRS) without any sharp current jump.This behavior is consistent with our previous (111) oriented SrFeO x devices. [15]The resistive switching set process was then initiated with a positive voltage sweep to 3 V, the device was switched to the low resistance state (LRS) from the HRS.When the voltage sweep was back to zero, the device still held the LRS, indicating the non-volatile behavior of resistance change.The subsequent reset process (LRS to HRS) was achieved by sweeping to a negative voltage of -4 V at the TE.The endurance test was performed by repeatedly applying a write pulse of +5 V/0.1 ms and -6 V/0.1 ms in the ambient atmosphere, which is shown in Figure 4b.Note, the resistance state of the device was measured using a read pulse of 0.5 V/1 ms after every write pulse.The resistance values of LRS (R LRS ) and HRS  (R HRS ) exhibited stable behavior during cycling, an on/off ratio of 10 was maintained for up to 5000 cycles.To understand the switching performance of the SFCO device in a vacuum, the endurance measurements were carried out in different devices on the same chip after the initial forming process in a high vacuum of 2 × 10 −6 mbar (see Figure 4c).The initial forming and switching process in a vacuum is discussed in of Figure S2a (Supporting Information).In contrast to endurance characteristics in the ambient atmosphere, the R LRS and R HRS values exhibited unstable behavior during cycling in a vacuum.Moreover, no considerable on/off ratio was observed after 250 cycles, indicating switching failure.It can also notice that R LRS and R HRS values are gradually increased with the number of cycles for up to 250 cycles; this behavior might be due to depletion of oxygen in the device structure during cycling in a vacuum.This trend also was observed in other devices in the same chip, which is showed Figure S2a,b (Supporting Information).
To understand the effect of GQDs on the resistive switching properties of the devices, we investigated the resistive behavior of G-SFCO devices in both ambient and vacuum conditions, as shown in Figure 4c-e.Upon applying an initial bias along the negative direction for the forming process, the device resistance sharply increased from the pristine state to a HRS with a current jump at 4 V (see Figure 4c).In contrast, the forming process of SFCO devices without GQDs showed a gradual resistance change from the pristine state to HRS.After the forming process, the G-SFCO memristive devices exhibited eightwise bipolar resistive switching behavior similar to that of SFCO devices.Interestingly, the set and reset processes also showed sharp current jumps at +1 and −2 V, respectively, unlike the switching process of SFCO devices.Figure 4e shows the endurance test of G-SFCO devices in ambient conditions, which involved applying a write pulse of +5 V/0.1 ms and −6 V/0.1 ms repeatedly.The resistance values of LRS (R LRS )and HRS (R HRS ) remained stable during cycling, and an on/off ratio of ≈10 was maintained for up to 5000 cycles.To investigate the switching performance of G-SFCO devices in a vacuum, we performed endurance measurements in different devices on the same chip after the initial forming process in a high vacuum of 2 × 10 −6 mbar (see Figure 4f).The initial forming and switching process in a vacuum is discussed in Figure S2b (Supporting Information).Interestingly, unlike SFCO devices, the R LRS and R HRS values of G-SFCO devices exhibited a clear on/off ratio up to 5000 cycles.However, we observed a gradual decrease in the R LRS and on/off ratio values up to 4000 cycles then showed stable switching behavior with constant on/off ratio, which might be due to a large amount of oxygen ion migration from GQDs during the initial cycling process.This trend was also observed in other devices on the same chip, as shown in Figure S3 (Supporting Information).
To investigate the impact of GQDs on the switching reliability of SFCO devices, we analyzed the statistical distribution of key switching parameters, such as resistance values at R HRS and R LRS , set voltage (V S ), and reset voltage (V R ).Typical hysteresis I-V loops for 100 consecutive cycles in the SFCO and G-SFCO devices are shown in Figure 5a,b, respectively.The SFCO device displayed poor resistive switching behavior, whereas the G-SFCO device showed good switching behavior and could be switched for more than 100 cycles.In Figure 5c, we compared the V S and V R values between the G-SFCO and SFCO devices.We found that the set and reset voltages of G-SFCO (1.1 and −2.8 V, respectively) were lower than those of SFCO devices (2.0 and −3.8 V), indicating an easier and localized access to oxygen ions from GQDs.Both SFCO and G-SFCO devices exhibited low relative fluctuations (/μ) in V S and V R , with values of 0.25 and 0.14 for SFCO and 0.15 and 0.17 for G-SFCO, respectively.Interestingly, the relative fluctuation of V S decreased significantly in G-SFCO compared to SFCO devices.Figure 5d illustrates the cumulative probability distribution of R HRS and R LRS values measured at 0.20 V. Notably, the relative fluctuation of R HRS for G-SFCO was reduced by almost 45%, with a value of 0.28, while R LRS exhibited a similar relative fluctuation of ≈20% in both devices.Note, the switching key switching parameters of G-SFCO devices showed rather improved behavior compare to SFCO.This might be due resistance a large area device (50 × 50 μm) with multiple GQD dots, leading to the formation of multiple filaments during the forming and switching processes.
The experimental results presented in this paper reveal a clear improvement in resistive switching performance, particularly in terms of endurance (in vacuum) and uniformity of key switching parameters (i.e., V S and R HRS ) upon the incorporation of GQDs into SFCO devices.To gain a deeper understanding of the underlying switching mechanism and the role of GQDs in this process, we developed a comprehensive switching model, as depicted in Figure 6.During the initial forming process, the application of a high voltage results in the generation of a large amount of heat due to joule heating within the device structure. [36,37]he presence of contact resistance at the Au/GQDs TE results in the dissipation of most of the heat at the Au/GQDs/SFCO interface, generating the highest temperatures in this region.This elevated temperature promotes the detachment of oxygen functional groups from GQDs owing to the applied electrical field and joule heating.These detached oxygen groups then migrate into the switching layer and move toward the BE interface, i.e., SFCO/SRO, under the influence of a strong electric field.This results in the accumulation of an oxygen-rich PV-SFCO conducting filament, as illustrated in Figure 6a.However, due to the strong electric field, an oxygen-deficient phase develops near the top interface, which leads to an intermediate device resistance overall.In the subsequent positive bias set process, the device resistance is switched to a low-resistance state (ON state), and highly Ohmic I-V behavior is observed for the reverse sweep direction, as shown in Figure 6b.This set process involves the internal redistribution of O 2-ions from the BE interface to the TE interface (Au/GQD/SFCO), facilitated by the applied electric field.This leads to the formation of an oxygen-rich conducting filament throughout the device structure, as shown in the schematic in Figure 6b.The resulting highly conductive filament is responsible for the observed LRS.
During the reset process in Figure 6c, a negative bias is applied to the TE, causing oxygen ions from the oxygen-rich PV-SFCO filament to migrate back to the SRO BE interface.The localized electrical field at the Au/GQDs interface may serve as a driving force for pushing the oxygen ions from the TE interface filament region to the BE interface. [25,30]This results in the local reduction of the PV-SFCO filament back to the original BM-SFCO phase at the TE interface, resulting in an OFF state due to the high resistive region at the TE interface.Thus, the incorporation of GQDs in SFCO devices can facilitate local oxygen sources for the formation and reduction of an oxygen-rich PV-SFCO filament during the set and reset processes, respectively, resulting in a local topotactic phase transition.The presence of functional groups on GQDs allows for oxygen absorption and storage, while the unique nanostructure and redox activity facilitate oxygen exchange with the SFCO layer during the resistive switching operation.Additionally, the GQDs can create preferential sites for localizing the electrical fields, leading to a confined filament along the same path at each voltage stress.This may contribute to the stable re-sistive switching behavior observed in G-SFCO devices compared to SFCO devices.Unlike many oxide-based memristive devices, such as SrTiO 3−x , TiO 2−x , HfO 2−x , and Ta2O 5−x , which predominantly rely on oxygen vacancy-rich filaments, the resistive switching mechanism in SFCO involves an oxygen-rich conductive PV-SFCO filament.Introducing GQDs as an oxygen reservoir at the local level serves a dual purpose: it acts as a source of oxygen and creates a favorable environment for filament growth.This unique approach not only provides the necessary oxygen for the switching process, but also enhances the overall switching performance by promoting filament formation and stability.

Conclusion
In summary, our study successfully demonstrated the deposition of a multi-functional BM-SFCO epitaxial thin film on a SRO/STO (001) substrate, resulting in an atomically smooth surface suitable for the memristive application.By introducing GQDs into the SFCO memristive device, we were able to enhance its performance by providing a nano-oxygen reservoir for the nanoscale topotactic redox process.The devices with GQDs exhibited reliable resistive switching behavior and good endurance properties, while devices without GQDs failed to show such behavior.Our study also proposed a physical model based on joule heating and oxygen ions from GQDs to the active switching layer to explain the resistive switching behavior and corresponding nanoscale topotactic phase transition.The findings of our study provide valuable insights into the potential of tuning nanoscale topotactic phase transitions for developing high-performance, multi-functional devices.

Experimental Section
SrFe 1−x Co x O 3− and SRO polycrystalline targets were prepared by the solid-state reaction method.Initially, to prepare the device, the SRO BE (≈70 nm thickness) was deposited on top of STO (001) substrate us-ing the pulsed laser deposition method, which consists of a KrF excimer laser of wavelength 248 nm and an ultrahigh vacuum chamber (base pressure can reach ≈10 −6 Torr).During SRO deposition, the laser fluence and repetition rate were kept at ≈2.5 J cm −2 , 4 Hz, respectively, while 750 °C of substrate temperature was set during deposition.During the SRO deposition, oxygen partial pressure was maintained ≈175 mTorr to achieve sufficient oxygen stoichiometry for obtaining a fully oxidized SRO BE with improved conductivity.Soon after the SRO deposition, SFCO deposition was performed at a set substrate temperature of 700 °C with a repetition rate of 2 Hz, while keeping laser fluence 2.1 j.cm −2 .To obtain the oxygen-deficient BM-SFCO phase, the oxygen partial pressure was reduced to 1 mTorr during the SFCO deposition.After the deposition, the SFCO/SRO/STO (001) heterostructure was cooled down to room temperature at the same oxygen partial pressure at a cooling rate of 20 °C min −1 .Memristive devices with GQDs were fabricated by spincoating solutions of GQDs in isopropyl alcohol with a concentration of 1 mg mL −1 onto the top of SFCO thin films.The GQDs used in this study were commercially sourced from Sigma-Aldrich.The structural characteristics of the formed SFCO/SRO/STO (001) heterostructure were characterized by high-resolution XRD (HR-XRD, Bruker D8).The thin films were imaged at an atomic resolution using a TEM with double Cs correctors, which was operated at 300 kV (Titan Themis 60-300 cubed, FEI).The electrical characteristics were measured under ambient conditions and high vacuum conditions (10 −6 torr) at room temperature using a semiconductor device parameter analyzer (Agilent B1540), triaxial cable, and a pulse unit (16440A).

Figure 1 .
Figure 1.a) X-ray diffraction (XRD) −2 patterns of as-deposited BM-SFCO/SRO/STO (001) thin films.The asterisks indicate peaks of the STO substrate.b) Full width half maximum of BM-SFCO (002) peak.(c) Reciprocal space maps near the (103) STO Bragg reflection, showing the epitaxial growth of the BM-SFCO thin film.d) Atomic force microscopy (AFM) image showing the surface morphology of the BM-SFCO layer.

Figure 2 .
Figure 2. a) Schematic drawing of an Au/BM-SFCO/SRO/STO (001) device with vertical ordered oxygen vacancy channels along the tetrahedral layers (blue polyhedral) of the BM-SFCO layer.b) Schematic of an Au/BM-SFCO/SRO/STO (001) device with graphene quantum dots between the Au top electrode and insulator BM-SFCO layer.

Figure 4 .
Figure 4. a) Representative I-V switching curves of SFCO devices without GQDs, with arrow directions indicating voltage sweeps.b,c) Pulsed-based endurance tests in ambient atmospheric and high vacuum conditions, respectively.d) Representative I-V switching curves of SFCO devices with GQDs, with arrow directions indicating voltage sweeps.e,f) Pulsed-based endurance tests in ambient atmospheric and high vacuum conditions, respectively.

Figure 5 .
Figure 5. a,b) Representative set and reset I-V curves for 100 consecutive cycles of SFCO devices with and without GQDs, respectively.c) Cumulative plots of set and reset voltages for both devices.d) Cumulative probability graphs of the high resistance state (HRS) and low resistance state (LRS) values for positive forward and reverse biases for both devices.

Figure 6 .
Figure 6.Schematic illustration of the electric-field-induced localized phase transition mechanism with graphene quantum dots (GQDs) serving as nano oxygen reservoirs within the filament structure in G-SFCO device for a) forming process, b) SET, and c) RESET operations.The yellow and blue spheres represent oxygen in the octahedral and tetrahedral configurations, respectively, while the red spheres indicate the electric-field-induced PV-SFCO phase change.