Resistive switching behavior of quasi-2D CsPbBr₃ memristors: Impact of ambient atmosphere on logic storage and computing integration with anti-crosstalk and reconfiguration features

In the realm of advanced electronics, the quest for efficient, scalable, and versatile memory and computing solutions has led to the exploration of novel materials and device architectures.^1–3 Memristive devices, characterized by their ability to retain a memory of past electrical states, have emerged as promising candidates for next-generation memory and computing applications.^3–5 Among these contenders, halide perovskites have risen to prominence as resistive switching (RS) materials, showcasing remarkable performance metrics such as an impressive ON/OFF ratio of 10⁹, a notably low set/reset voltage of 0.1 V, and an extended retention duration exceeding 10⁵ s.^6–10 Memory devices employing perovskites have demonstrated various types of switching, including unipolar, bipolar, and write-once-read-many times (WORM) switching.^11–13 In particular, many perovskite memory devices exhibit bipolar witching characteristics, wherein they can be activated under a forward bias and exclusively deactivated through a reverse bias.^14,15 However, when configured in a crossbar array, the attributes of bipolarresistive switches present a challenge in the form of crosstalk, as current may inadvertently flow through the unselected devices. Furthermore, these devices cannot execute diverse logic operations using simple resistance switching principles.^16,17 Several strategies, including 1T1R (one transistor, one resistor), 1S1R (one selector, one resistor), 1D1R (one diode, one resistor), CRS (complementary resistive switching), and antipolar resistive switches, have been proposed to address the prevalent crosstalk issue in ReRAM crossbar arrays.^18–22 However, these approaches require additional components such as transistors, selectors, or diodes, or involve the use of two resistive switching devices connected anti-serially (CRS). From the perspective of high-density integration of memory devices and 3D stacking, these structures are less suitable compared to passive devices.

In our previous work, it was observed that devices with an Ag/MAPbI₃(wire)/Ag setup exhibited antipolar resistive switching behavior akin to capacitive behavior under light exposure. The study revealed that the resistive switching behavior is correlated with the duration of exposure of the perovskite to air. This phenomenon was attributed to the formation of AgI_x at the interfaces of the two Ag/MAPbI₃ layers, thereby generating a tunnel barrier.^23–25 Additionally, we constructed resistive random access memories (ReRAMs) comprising ITO/(UVO-treated) amorphous ZnO/MAPbI₃/Ag, which effectively mitigated crosstalk current and enabled the execution of 12 distinct 2-input sequential logic functions with the characteristic of antipolar resistance switching behavior. Through comparative analysis involving crystalline ZnO, amorphous ZnO, and UV-ozone-treated amorphous ZnO, we discovered the capability of oxygen presence to modulate resistance switching behavior.²⁰ To the best of our knowledge, the achievement of suppressing sneak currents and achieving various reconfigurable logic memory functionalities within a single halide perovskite memristor is rarely reported, and its underlying mechanism remains elusive.

To further investigate the influence of air atmosphere on resistive switching behavior, we prepared quasi-2D CsPbBr₃ with good stability in air and fabricated resistive switching devices. Aging and testing in different environments revealed the crucial role of the combined effect of air and high voltage in forming negative differential resistance (NDR) effects and capacitive hysteresis loops. By comparing the resistive switching characteristics of devices with inert, low-active, and active electrodes in different test atmosphere conditions, we suggest that while the activity of the metal electrode modulates resistive switching behavior, it is not the decisive factor for this antipolar resistive switching feature; instead, the variation of surface defects and internally mobile ions in quasi-2D CsPbBr₃ is the determining factor under alternating current voltage. Exploiting the resistive characteristics of devices in different environments, a switching ratio of 10⁴ was achieved for (write-once-read-more) WORM in an Ar atmosphere, while operation in air yielded suppressed crosstalk and reconfigurable logic for integrated storage and computing. Gratifyingly, to the best of our knowledge, this unit displays the simplest structured non-volatile storage and computing integrated unit with the most diverse reconfigurable logic algorithms to date. This work not only demonstrates the significant role of air in resistive switching mechanisms but also provides reference and new insights for high-density, high-computing-power non-volatile storage, and computing integrated units.

All chemicals used in this work were commercially available. Lead bromide (PbBr2, 99.99%), cesium bromide (CsBr, 99.99%), and phenethylamine bromide (PEABr, 99.99%) were purchased from Xi'an Yuri Solar Co. Dimethyl sulfoxide (DMSO, 99.99%) was purchased from Shanghai Aladdin Company. (18-Crown-6, 99.99%) was obtained from Sigma Aldrich (Shanghai) Trading Co. All materials were used directly without further purification. Quasi-2D chalcogenide (CsPbBr₃) films were deposited directly onto ITO-coated glass by a one-step spin-coating method in an N₂-filled glovebox environment. Before device fabrication, commercial ITO glass substrates were ultrasonically cleansed with detergent, de-ionized water, acetone, and ethanol for 20 min, then dried with an N₂ gas stream. The substrates were treated with oxygen plasma for 90 s to render them more hydrophilic. Preparation of quasi-2D perovskite (CsPbBr3) thin films involved dissolving 0.064 g of CsBr and 0.110 g of PbBr₂ (1:1 molar ratio), along with 0.004 g of 18-crown-6 ether and 0.024 g of PEABr (final molar proportion CsBr: PbBr₂: PEABr = 1:1:0.4) in 1 ml of dimethyl sulfoxide (DMSO) solution, and stirring at 60 °C for 12 h. The solution was then filtered through a 0.22 μm organic phase syringe filter for further use. Subsequently, in an N₂-filled glovebox, 50 ml of the perovskite precursor solution was spin-coated onto ITO substrates. The spin coater was accelerated to 500 rpm for 5 s followed by 3500 rpm for 25 s. Upon completion of spin coating, the samples were placed on a hot plate and annealed at 100 °C for 10 min. The thin film color transformed from transparent to brownish-yellow, indicating successful fabrication of CsPbBr₃ thin films. The ITO/CsPbBr₃ stacks were loaded into the chamber of a high vacuum electron beam physical vapor deposition system (Beijing Kejie Tech. Co., TEMD500). After the vacuum reached 5.0 × 10⁻⁴ Pa, the deposition process commenced, depositing Cu, Ag or Au at a rate of 0.2 Å/s until the thickness reached 10 nm. The rate was then increased to 0.4 Å/s until the thickness reached 30 nm. Each metal electrode has a uniform area of 2 × 13 mm². The effective area of the devices, where the electrodes interact with the active material, is 2 × 2 mm². In ITO/CsPbBr₃/C devices, C electrodes were implemented by using a carbon electrode paste. The area of C electrode was about 2 × 3 mm².

Morphological analysis of the sample surface and cross section was performed using a field emission scanning electron microscope (SEM, FEI Quanta 200FEG). X-ray diffraction (XRD) measurements were carried out using a PANalytical Empyrean x-ray diffractometer with Cu (Kα) as the emission source. X-ray photoelectron spectroscopy (XPS) analysis of elemental composition and valence states was conducted using a Thermo Scientific K-Alpha photoelectron spectrometer with Al (Kα) as the emission source. Absorption spectra of materials were measured using a UV spectrophotometer (HITACHI, UV4150). The electrical characterization was performed using a Keithley 2636B Source Measure Unit (SMU). During the testing process, a Keithley 2636B SMU was used in conjunction with a probe station to measure the electrical properties of the devices. The positive electrode was connected to the ITO bottom electrode, while the negative electrode was connected to the top electrode (Ag/Cu/Au/C). For measurements inside the glovebox, the probe station was placed inside the glovebox, and the connecting wires were led out to contact with the Keithley 2636B SMU. For aging device in air, the temperature and humidity are about 20 °C and 65%, respectively.

The morphology of the quasi-2D CsPbBr3 film and the thickness of each layer were detected utilizing SEM, as shown in Fig. 1(a). The quasi-2D perovskite sample shows complete coverage with particle sizes around 20 nm and no significant pinholes. The UV–visible absorption spectra of 3D CsPbBr₃ and quasi-2D CsPbBr₃ films were collected, as shown in Fig. 1(b), to confirm the existence of lower-dimensional phases. Compared to the absorption spectra of 3D perovskite, the quasi-2D perovskite film has weak exciton absorption peaks at 408, 438, and 463 nm, corresponding to n = 1, n = 2, and n = 3 phase perovskite, respectively.²⁶ XRD of fresh and aged (in air) quasi-2D CsPbBr₃ films was conducted, as shown in Fig. 1(c), confirming the cubic phase of CsPbBr3 (ICSD-29073). Even after one week of air exposure, no significant decomposition phases appeared. Figures 1(d)–1(h) illustrate the evolution of the current–voltage (I–V) characteristics of devices over aging time, where quasi-2D CsPbBr₃ serves as an intermediate layer between Ag and ITO electrodes, prepared in an N₂ atmosphere. The voltages for the driving device vary in magnitude, with scans conducted sequentially from 0 V to positive voltage, back to 0 V, then to negative voltage, and again to 0 V. Each I–V characteristic undergoes five consecutive cycles to demonstrate repeatability, with the compliance current set to 0.1 A. Figures 1(d) and 1(e) illustrate the evolution of the prepared device over aging time under an N₂ atmosphere. Initially, the device displayed excellent conductivity and maintained a low-resistance state (LRS) without hysteresis loops. Even after 8 days, the devices only showed minimal hysteresis within the range of 0 to −5 V. Subsequently, the device was transferred from the glovebox to ambient air for I–V testing, as depicted in Figs. 1(f)–1(h). At low voltages (below 5 V), the device exhibited high conductivity without any resistance variation [Fig. 1(f)]. Upon increasing the voltage to 5 V, symmetric negative differential resistance (NDR) characteristics emerged in the I–V curve, indicating a resistance switch from low to high resistance state (HRS) in both positive and negative voltage regions, with a HRS/LRS ratio of approximately 30 [Figs. 1(g) and 1(h)]. Comparative analysis conducted under different testing environments revealed that air exposure notably induced symmetric NDR switching characteristics in ITO/quasi-2D CsPbBr₃/Ag devices.

To explore how aging and testing in air affect the I-V characteristics of quasi-2D CsPbBr₃ devices, we first prepared ITO/quasi-2D CsPbBr₃ stacks in a glovebox. Subsequently, Ag electrodes were added following one week of air exposure to prevent prolonged contact between the Ag electrode and the quasi-2D CsPbBr₃ layer during aging. Figures 2(a)–2(d) showcase the I–V behavior of the devices tested at various voltages in N₂. At testing voltages below 5 V, the devices display higher conductivity, as evident in Figs. 2(a) and 2(b). With increasing testing voltage, the devices transition from LRS to HRS, failing to return to the former state in subsequent tests [Figs. 2(c) and 2(d)]. This behavior differs from that of quasi-2D CsPbBr₃ device aged in N₂, which maintains a high-conductivity under large voltage [Fig. 1(e)]. These results imply that aging in air prompts the devices to switch from LRS to HRS at a relative large voltage. Figures 2(e)–2(h) illustrate the I–V characteristics of the devices measured at different voltages in air. At a testing voltage of 1 V, the devices exhibit hysteresis, NDR characteristics at 4 V, and pronounced symmetric NDR behavior at 8 and 15 V, with an optimal HRS/LRS ratio of approximately 10². Compared to devices aged in an N₂ atmosphere [Figs. 1(g) and 1(h)], those aged in air demonstrate significant resistance switching phenomena and a larger switching ratio. The I–V plots of the device over 100 consecutive scanning cycles were displayed in Fig. 2(i), demonstrating excellent reproducibility. Unlike the behavior observed during N₂ testing [Figs. 2(c) and 2(d)], devices tested in air exhibit continuous I–V hysteresis cycles. Thus, the environment to which quasi-2D CsPbBr₃ devices are exposed, encompassing both aging and testing processes, dictates their capability to transition to HRS under high voltages, while the testing atmosphere determines their ability to exhibit cycle-repeatable resistance switching. The retention and durability of the four resistive states, measured at Vread = 0.5 V (for HRS+ and LRS+) and −0.5 V (for HRS- and LRS-), remain stable for up to 1 × 10⁴ s and 2000 cycles, respectively, as illustrated in Figs. 2(j) and 2(k). The noise in the I–V curves may originate from environmental factors, equipment, and device fabrication processes. Despite the presence of low-level noise, it does not have a significant impact on our main discussion. In the future, we will adopt advanced filtering and denoising techniques and optimize processes to reduce the noise.

The chemical reactions between Ag electrodes and the decomposition products of perovskite result in the formation of halides such as AgI_x and AgBr.^27,28 In contrast, copper (Cu) is well-known for its outstanding corrosion resistance, especially in non-oxidizing acidic environments such as hydrogen iodide (HI), hydrogen bromide (HBr), and halogens.²⁹ To investigate whether the reversible resistive switching effect illustrated in Fig. 1(g) arises from interactions between Ag electrodes and quasi-2D CsPbBr₃, devices with Cu electrodes were fabricated in air and glovebox, respectively. The I–V characteristics of these devices are illustrated in Fig. 3. Under the same atmosphere, devices with Cu electrodes exhibit similar I–V characteristics to those with Ag electrodes. Figures 3(a) and 3(b) depict the I–V curves of devices prepared and tested in N₂ within a glovebox, showcasing LRS even at high voltages. Conversely, devices prepared in air and tested in N₂ exhibit HRS under high voltages, as depicted in Figs. 3(c) and 3(d). Devices tested in air exhibit a repeatable NDR resistive switching effect in both N₂ and air fabricated environments [Figs. 3(e)–3(h)], with an HRS/LRS ratio of approximately 5 × 10². To gain a deeper understanding of the ion transport mechanisms, we conducted measurements of the capacitance–frequency response of the RRAM device in both its HRS and low-resistance state LRS. Figure 3(i) reveals distinct behaviors in the high-frequency and low-frequency regions, which can be attributed to bulk transport and interfacial transport mechanisms, respectively. Notably, at high frequencies, the capacitance-vs-logarithmic frequency (C−log ω) curves for both HRS and LRS align, indicating that the bulk resistance of the perovskite material undergoes minimal change during the resistive switching process between LRS and HRS. Conversely, at low frequencies, the capacitance of the HRS is markedly higher than that of the LRS. This discrepancy leads us to conclude that the transition from LRS to HRS is primarily influenced by interfacial effects, rather than bulk effects.

Some reports suggest that copper easily oxidizes in the presence of oxygen and moisture, forming potential oxidation products such as Cu(OH)₂ or CuO_x, which could react with degradation by-products of perovskite such as HI, thereby leading to perovskite decomposition.^29–31 To investigate the chemical state of the CsPbBr₃/Cu interface, a layer of 10 nm thick Cu was deposited onto CsPbBr₃. XPS was conducted after subjecting the sample to cyclic voltages ranging from −10 to 10 V, as depicted in Figs. 3(j)–3(l). In Fig. 3(j), the XPS spectra of Br 3d exhibit two peaks centered at 68.94 and 67.52 eV, corresponding to Br 3d_3/2 and Br 3d_5/2 of CsPbBr₃, respectively.^32–34, Figure 3(k) illustrates the O 1s XPS spectra, revealing oxygen incorporation within the lattice, with discernible peaks at 530.75 eV, indicative of oxygen ions in either CuO or Cu₂O. This inference finds support in the high-resolution Cu 2p XPS core level spectrum. As shown in Fig. 3(l), the observed peaks at 954.10 and 934.32 eV correspond to the Cu2p_1/2 and 2p_3/2 orbitals of CuO, respectively.³⁵ Peaks positioned at 944.10 and 941.5 eV correspond to the Cu2p_1/2 sat. of CuO. Additionally, peaks at 951.53 and 931.73 eV are aligned with the Cu2p_1/2 and 2p_3/2 orbitals of Cu₂O.³⁶ It is noteworthy that the binding energies of Cu(II) oxides and bromides, as well as Cu(I) oxides and bromides, exhibit proximity.^37,38 It can be deduced that the peaks observed in Fig. 3(l) represent copper oxides and bromides. Moreover, we constructed devices using inert Au and C electrodes in air. The I-V characteristic curves, depicted in Figs. 4(a) and 4(b), demonstrate a consistent and reproducible NDR resistive switching. Given the stability of Au and C in both air and halogen environments, we argue that the oxidation of metal electrodes (whether in air or halogen) does not significantly contribute to the NDR resistive switching phenomenon. It may be attributed to the simultaneous influence of ion migration or surface potential barriers under the conditions of air exposure and high voltage. Thus, we conclude that although the electrode undergoes oxidation, this oxidation is not the pivotal factor contributing to the generation of the symmetric NDR effect.

To describe the interfacial issues, we have fabricated a device configuration of ITO/quasi-2D CsPbBr₃/AgI_x/Cu, where the I ions in AgI_x are utilized to replace the aggregated Br⁻ ions at the interface location. The I–V curves are shown in Figs. 4(c) and 4(d). As can be observed, its initial state exhibits a slight hysteresis loop and manifests as a distinct Schottky barrier contact, contrasting the initial state of the ITO/quasi-2D CsPbBr₃/Cu structured device, which displays an ohmic contact characterized by excellent conductivity. This discrepancy arises from the presence of an interfacial barrier introduced by the AgI_x layer at the Cu interface initially, resulting in the behavior depicted in Fig. 4(c) under low voltage conditions. Nevertheless, under high voltage, a symmetrical NRD resistive switching effect emerges. This validates the interface barrier regulation mechanism assumed in this work.

In order to better understand the formation process of resistive switching, we investigated the process of resistive switching in devices from their initial states. This process is illustrated in Figs. 4(e)–4(h). All quasi-2D CsPbBr₃ layers of these devices in Fig. 4 were prepared in air. It can be observed that devices with Ag electrode and Cu electrode [Figs. 4(e) and 4(g)] exhibit non-volatile resistive switching during testing. A transition from LRS to HRS occurred during the second cycle and remained in the HRS throughout subsequent cyclic I–V measurement. This aligns with the typical characteristics of write-once-read-many (WORM) technology, allowing data to be written to the storage medium only once and then read many times. WORM is commonly used in applications where data need secure storage and multiple accesses but once written cannot be modified, ensuring data integrity and preventing accidental or malicious alterations to stored information. The HRS/LRS ratios for WORM with Ag electrode and Cu electrode tested in an N₂ atmosphere were 10³ and 10⁴, respectively. Figure 5(a) depicts the WORM testing under direct current, demonstrating a stable read process after a write operation.

Figures 4(f) and 4(h) illustrate the I–V measurements conducted on devices featuring Ag and Cu electrodes in air ambient. The voltage was swept from 0 to −5 V, then returned to 0 V, progressed to 5 V, and finally reverted to 0 V, cycling thereafter. Notably, during the voltage sweep from 0 V to higher values, both positive and negative voltage regimes displayed the NDR phenomenon. As the scan voltage increased, the resistance shifted from LRS to HRS, persisting in the HRS during voltage return until a reverse polarity scan voltage was applied, prompting the resistance to revert to LRS. This resistive switching behavior repeated with each testing cycle. The HRS/LRS ratio for both Ag and Cu electrode devices was approximately 10². To ascertain whether the characteristics depicted in Figs. 4(d) and 4(f) signify unipolar resistive switching (i.e., switching solely dependent on voltage magnitude regardless of polarity), the measurements were exclusively cycled within positive or negative voltage ranges, as shown in Figs. 5(b) and 5(d) for validation.

Figures 5(b)–5(d) illustrate the I–V curves of ITO/quasi-2D CsPbBr₃/Ag across various scanning voltage ranges. Specifically, Fig. 5(c) showcases characteristics under cyclic scanning from −10V to 10 V. In Fig. 5(b) and 5(d), the device's I–V characteristics are shown under only negative and only positive cyclic voltages ranging from 0 to −10 V and 0 to 10 V, respectively, with voltage polarity unchanged. The cyclic I–V curve in Fig. 5(c) demonstrates excellent repeatability of the resistive switching effect. Contrarily, in Figs. 5(b) and 5(d), only the first cycle exhibits a switch from LRS to HRS, with subsequent cycles maintaining the device in HRS. This observation suggests that this type of resistive switching does not belong to unipolar resistive switching. The transition from HRS to LRS depends on voltage polarity, while the reverse transition relies on voltage magnitude. This behavior diverges from conventional unipolar or bipolar switches, which are solely dependent on voltage amplitude or polarity.^39,40 The hysteresis loop of the device exhibits characteristics more akin to antipolar resistive switching.^39,41 This resistive switching, responsive to both voltage polarity and magnitude, presents a promising approach for designing memory systems capable of effectively mitigating sneak pathway effects and enabling reconfigurable logic operations.²⁰ 

To optimize memristors for high-intensity data storage applications, they must be integrated into high-density crossbar arrays, where managing sneak current is crucial. This current issue intensifies with the presence of low-resistance paths, determined by the state of memristor cells.⁴² Various solutions, including 1T1R, 1S1R, 1D1R, CRS, and 1R structures, have been proposed to tackle crosstalk in ReRAM crossbar arrays.^7,18,21,22 Among these, the 1R structure, which has a simple device structure without additional components offers advantages, as a passive device with a minimum unit area of 4F², supports high-density memory device integration and facilitates straightforward 3D stacking. Figure 5(e) demonstrates the device's ability to effectively suppress sneak pathways. Initially, the device is in the HRS. Upon applying a V_set of −10 V, it transitions to the LRS. Subsequently, voltages of +0.5 and −0.5 V are utilized to readout. During the +0.5 V read, the device exhibits LRS, while during the −0.5 V read, it remains in HRS. This indicates that when in LRS, the device presents high resistance under reverse current, effectively inhibiting reverse current through unselected devices.

The nonvolatile memory characteristics of the device were assessed, and the results are presented in Figs. 5(f) and 5(g). In Fig. 5(f), the write–erase process is depicted, with the write, erase, and read voltages set to −10, 10, and −0.5 V, respectively. Upon applying the −10 V set voltage, the resistance state switches from the HRS to the LRS, effectively preserving the written electrical information. Subsequently, when 10 V is applied for reset, the resistance state switches back to HRS, leading to the erasure of stored information. Generally, resistive switching devices display LRS after the set operation and HRS after the reset operation. However, the resistive state illustrated in Fig. 5(f) exhibits the opposite behavior. This discrepancy stems from the symmetrical NDR feature in the resistive switching device discussed here. Referring to Fig. 5(c), it is evident that the resistance transitions from LRS to HRS after the application of −10 V and returns to LRS after a forward bias voltage is applied. The multiple write–erase cycles are demonstrated in Fig. 5(g), showcasing the excellent repetitive editing capabilities of the perovskite devices.

Above, we present I–V testing for resistive switching using various electrodes and infer that the formation of this resistive switching is not primarily governed by electrode oxidation. Although electrode oxidation with Br or O can influence the resistive behavior, it is not the determining factor for the observed NDR here. Additionally, quasi-2D CsPbBr₃ devices prepared in N₂ exhibit a constant LRS under testing condition of high voltage coupling N₂ atmosphere, with no resistance switching observed. This suggests that quasi-2D CsPbBr₃ devices in N₂ do not exhibit NDR effects even when in direct contact with electrodes under high voltage. We further investigate devices under high voltage and N₂ testing conditions, where quasi-2D CsPbBr₃, exposed to air for one week (or prepared in air) and subsequently electrode-evaporated, reveal WORM behavior that switching from LRS to HRS [Figs. 2(d), 3(c), 3(d), 4(c), and 4(e)]. This indicates that exposure of quasi-2D CsPbBr₃ to air leads to NDR characteristics in the I–V curve. As halide perovskites are exposed to ambient air, they develop various types of defects due to environmental factors such as oxygen and moisture. These defects include surface defects and ionic vacancies. Surface defects result from reactions between oxygen and moisture with the perovskite surface, leading to the formation of vacancies, dangling bonds, or surface states.^43–45 Ionic vacancies arise from oxygen-induced formation of vacancies within the halide perovskite lattice, resulting in charged defects.^46–48 Under the influence of high voltage, Br⁻ undergoes significant migration. In an N₂ atmosphere, alternating high voltage prompts the migration and aggregation of Br⁻ onto defect-rich regions on both upper and lower surfaces of quasi-2D CsPbBr₃. This process forms a surface energy level pinning effect, leading to the symmetric NDR phenomenon illustrated in Figs. 4(c) and 4(e). As Br⁻ accumulates on the surface, it depletes mobile Br⁻ within quasi-2D CsPbBr₃, resulting in the absence of hysteresis loops in the I–V curve. Conversely, when large alternating voltages are applied in air, surface defects are first combined with O, while a substantial amount of Br⁻ remains within quasi-2D CsPbBr₃. This causes quasi-2D CsPbBr₃ to exhibit a capacitive-like behavior under alternating current, thereby inducing an antipolar resistive switching effect.

This particular type of resistive switching, influenced by both voltage polarity and magnitude, is notably capable of executing 12 distinct 2-input sequential logic functions.²⁰  Table I succinctly illustrates the operational characteristics of these 12 sequential logic functions, including TRUE, FALSE, p, NOT p, q, NOT q, AND, RIMP, NIMP, NOR, XOR, and XNOR. The left section of the table (highlighted in blue) delineates the rules governing logic operations, while the right side (highlighted in orange) elaborates on the strategies for configuring the logical variables p and q. Here, p represents the first input, and q represents the second input, with “0” and “1” denoting low and high voltage states, respectively. Prior to initiating operations, devices are initialized in the HRS. The “Step Count” indicates the essential number of operational steps required to accomplish the specified logic function. The resistive switching behavior observed in this passive 1R structure (ITO/quasi-2D CsPbBr₃/electrode) can realize an integrated storage and computing unit capable of performing 12 reconfigurable logic operations and supporting non-volatile memory. As far as we know, this unit is the non-volatile storage and computing integrated unit with the simplest structure and the most diverse reconfigurable logic algorithms to date.

In conclusion, our study highlights the significant impact of air on the functionality of memristors, specifically those designed to suppress sneakpath current and operate reconfigurable logic functions. That is, upon contact with air, quasi-2D CsPbBr₃ devices exhibit NDR effects in their I–V characteristics. Moreover, testing the devices in air results in antipolar resistive switching behavior, while testing in a N₂ atmosphere only shows unidirectional switching from LRS to HRS. Through an analysis of the impact of environmental conditions and electrodes on the resistive switching behavior of devices, we propose that defects generated in quasi-2D CsPbBr₃ when exposed to air induce ion aggregation and energy level pinning effects on the surfaces under high voltage conditions, therefore leading to the NDR effect. In N₂, the application of high alternating voltage drives the migration of Br⁻ ions, localizing them on the surface, resulting in an irreversible switch from LRS to HRS. Conversely, in air, O ions interact with these defects under high alternating voltage, leading to the retention of a significant quantity of Br⁻ ions within the quasi-2D CsPbBr₃, thereby exhibiting capacitive-like behavior. The unique resistive switching behaviors observed under different atmospheric conditions broaden the potential applications of resistive switching devices utilizing quasi-2D CsPbBr₃. Specifically, devices operated in a N₂ atmosphere demonstrate a switch ratio of 10⁴, enabling WORM data storage, ensuring data integrity, and safeguarding against unintended alterations. Devices operated in air can suppress crosstalk currents in a crossbar array setup. Furthermore, they can realize an integrated non-volatile storage and computing unit capable of executing 12 reconfigurable logic operations, within such a passive 1R structure (ITO/quasi-2D CsPbBr₃/metal electrode). To the best of our knowledge, this unit stands as the simplest structured non-volatile storage and computing integrated unit to date, boasting a wide array of reconfigurable logic algorithms. This study not only underscores the pivotal role of air in the resistive switching mechanism but also furnishes valuable insights for the development of next-generation memories tailored for high-density integrated circuits and storage-computing integration.

This work was partially supported by the National Natural Science Foundation of China and Natural Science Foundation of Jiangxi Province (Nos. 62464011, 20242BAB25226, 51571107, and 20202ACB202002).

The authors have no conflicts to disclose.

Mufan Zhu: Conceptualization (equal); Formal analysis (equal); Investigation (equal); Methodology (equal); Validation (equal). ChuTing Yao: Conceptualization (equal); Formal analysis (equal); Investigation (equal); Software (equal). Xiaofei Zhang: Conceptualization (equal); Formal analysis (equal); Investigation (equal); Software (equal). Song He: Conceptualization (equal); Formal analysis (equal); Investigation (equal); Software (equal). Baochang Cheng: Conceptualization (equal); Funding acquisition (equal); Supervision (equal); Validation (equal); Writing – review & editing (equal). Jie Zhao: Conceptualization (equal); Formal analysis (equal); Funding acquisition (equal); Supervision (equal); Writing – original draft (lead); Writing – review & editing (equal).

The data are available from the corresponding author upon reasonable request.