The emergence of photoelectric memristors has opened up new opportunities for the research community to realize the neuro-synaptic functionalities of photoelectric systems. Neuromorphic photoelectric memristors (NPMs) can directly respond to non-contact photonic signals while possessing the desirable features of high bandwidth, zero latency, and low crosstalk. With their capability to integrate the sensing, memory, and computing features, they can mimic the human vision system. Here, we propose a perovskite oxide (ABO3)-based NPM, where the active medium is comprised of oxygen rich and oxygen deficient layers of barium strontium titanate. Along with the analog-type resistive switching behavior, the device current modulation also enables the imitation of long term-potentiation/depression behaviors of the human brain. The designed convolutional neural network model achieves high accuracy even when tested with the damaged (noisy) face images of the Olivetti Research Laboratory dataset. The photo-excitation and photo-inhibition phenomena of NPM are observed under 405 and 633 nm illumination, respectively, and further utilized to realize the spike-intensity, spike-width, spike-rate, and spike-number dependent synaptic plasticity behaviors. These findings significantly inspire future research in the field of perovskite oxide based transparent photoelectric synaptic resistive switching memory devices.

A human brain with its complex neural network of 1011 neurons and 1015 synapses is considered as the most efficient and powerful processor in nature.1,2 In neural networks, the external information is transmitted from pre-synaptic neurons to post-synaptic neurons, and the synapses in between these neurons contribute to the computation by changing their transmitting efficiency (synaptic plasticity) and play the most vital role in information processing, learning, and memory by producing excitatory/inhibitory post-synaptic current (EPSC/IPSC).3–5 Artificial synaptic devices that can imitate the functions of the human brain with their parallel data processing and learning abilities would be an impressive solution to overcome the von Neumann bottleneck.6–9 To manifest the functionalities of neurons and synapses, analog memory devices, especially memristors, are widely acknowledged and considered as the most favorable electronic devices for the realization of brain-inspired computing10 due to CMOS process compatibility11 and integration density.12–14 

Although the human brain receives external information through the sensory cells, such as olfactory, tactile, gustatory, ocular, and auditory,15 among them, the “ocular sensory cell” plays a major role by receiving 80% of external information. In the human visual system, the eye, with its iris and lens, controls the incident light intensity and then passes the visual or optical information to the multilayered transparent membrane tissue referred to as the retina.16 This visual information is first detected by wavelength receptors and then distinguished into several colors and brightness via cones and rods to form images in the brain.17 Furthermore, it is converted into an electrical signal and transmitted to the visual cortex for storage and processing.13,18–21 A variety of traditional image sensors with the requirement of large memory space and high energy have been intensively studied.22 Therefore, it is necessary to develop a neuromorphic vision system23 that can perform as efficiently as the human vision system in an energy saving manner.17 Photoelectric synapses can operate similarly to human visual neurons,24 and they inspire the development of two-terminal photoelectric memristors.22,25 Those devices can momentarily memorize and process the relevant information extracted from optical signals in real time. Due to the integration of memory and synaptic functions with optical sensing,26 the photoelectric memristor, as an elementary unit, can provide an optimistic hardware foundation for the evolution of artificial ocular (or vision) systems.22 In this artificial intelligence related multidisciplinary research field, various photoelectric memristors have been developed and reported in recent years. Most of the reported photoelectric memristors27–30 work in a hybrid photoelectric mode (photonic potentiation and electrical depression) and suffer from the issues of control complexity and heat generation.31 Hence, the development of all-optically controlled synaptic memristors,20,32–35 especially with bidirectional photoelectric response (BPR),31,36,37 would be crucial to imitate the human vision system. The realization of BPR in a single memristive device constitutes a strict necessity for photo sensitive materials. In recent times, low-dimensional18 and halide perovskite38 materials have been extensively investigated to achieve BPR in photoelectric synapses. On the other hand, the available reports on “oxide-based all-optically controlled synaptic memristors with BPR” are very limited and summarized in Table I. In addition, the performance of perovskite oxide based memristors has been investigated and summarized in Table II.

TABLE I.

Comparison of the typical performance of previously reported oxide-based all-optically controlled memristors with our present device.

Memristor device structureAll-optically controlledBPRLight source (nm)STM–LTMPPF/PTPNeuromorphic application
Ga2O3/ZnO20  ✓ ⋯ 405/522 ⋯ ✓ ✓ 
ZnO/p+-Si31  ✓ ✓ 400/700 ✓ ✓ ⋯ 
NiO/TiO232  ✓ ⋯ 320/480 ✓ ✓ ✓ 
NiO/NbOx/TiO233  ✓ ⋯ 290–390 ✓ ✓ ✓ 
OD-/OR-IGZO34  ✓ ⋯ 420/800 ⋯ ⋯ ⋯ 
Zn2SnO4/ZnO35  ✓ ⋯ 405/633 ✓ ⋯ ✓ 
MoOx/ZnO36  ✓ ✓ 350/570 ✓ ✓ ⋯ 
WOx: Ag75  ✓ ⋯ 450–680 ✓ ⋯ ✓ 
Si: Ga2O3/ZnO76  ✓ ⋯ 255/370 ✓ ✓ ✓ 
TiO277  ✓ ⋯ 360/532 ✓ ⋯ ✓ 
Cu2O/WO378  ✓ ✓ 405/633 ✓ ⋯ ✓ 
OD-/OR-BST (this work) ✓ ✓ 405/633 ✓ ⋯ ✓ 
Memristor device structureAll-optically controlledBPRLight source (nm)STM–LTMPPF/PTPNeuromorphic application
Ga2O3/ZnO20  ✓ ⋯ 405/522 ⋯ ✓ ✓ 
ZnO/p+-Si31  ✓ ✓ 400/700 ✓ ✓ ⋯ 
NiO/TiO232  ✓ ⋯ 320/480 ✓ ✓ ✓ 
NiO/NbOx/TiO233  ✓ ⋯ 290–390 ✓ ✓ ✓ 
OD-/OR-IGZO34  ✓ ⋯ 420/800 ⋯ ⋯ ⋯ 
Zn2SnO4/ZnO35  ✓ ⋯ 405/633 ✓ ⋯ ✓ 
MoOx/ZnO36  ✓ ✓ 350/570 ✓ ✓ ⋯ 
WOx: Ag75  ✓ ⋯ 450–680 ✓ ⋯ ✓ 
Si: Ga2O3/ZnO76  ✓ ⋯ 255/370 ✓ ✓ ✓ 
TiO277  ✓ ⋯ 360/532 ✓ ⋯ ✓ 
Cu2O/WO378  ✓ ✓ 405/633 ✓ ⋯ ✓ 
OD-/OR-BST (this work) ✓ ✓ 405/633 ✓ ⋯ ✓ 
TABLE II.

Comparison of the typical performance of previously reported perovskite oxide-based memristors with our present device.

Functional layerEndurance cyclesHCS/LCS ratioArtificial synapseOptical sensingBPRNeuromorphic application
LaAlO379  2000 ∼100 ⋯ ⋯ ⋯ ⋯ 
SrCoOx80  100 >1000 ⋯ ⋯ ⋯ ⋯ 
AlFeO381  10 000 100 ⋯ ⋯ ⋯ ⋯ 
LaMnO382  50 ∼18 ⋯ ⋯ ⋯ ⋯ 
SrFeO2.583  1000 >10 ✓ ⋯ ⋯ ✓ 
SrTiO384  100 >10 ✓ ⋯ ⋯ ⋯ 
Nb: SrTiO385  ⋯ ⋯ ✓ ✓ ⋯ ✓ 
Nb: SrTiO386  ⋯ ⋯ ✓ ✓ ⋯ ✓ 
BST-Nd2O387  10 000 ∼10 ✓ ✓ ⋯ ✓ 
BST88  107 1000 ⋯ ⋯ ⋯ ⋯ 
BST58  1000 ∼97 ✓ ✓ ✓ ✓ 
OD-/OR-BST (this work) 10 800 ∼560 ✓ ✓ ✓ ✓ 
Functional layerEndurance cyclesHCS/LCS ratioArtificial synapseOptical sensingBPRNeuromorphic application
LaAlO379  2000 ∼100 ⋯ ⋯ ⋯ ⋯ 
SrCoOx80  100 >1000 ⋯ ⋯ ⋯ ⋯ 
AlFeO381  10 000 100 ⋯ ⋯ ⋯ ⋯ 
LaMnO382  50 ∼18 ⋯ ⋯ ⋯ ⋯ 
SrFeO2.583  1000 >10 ✓ ⋯ ⋯ ✓ 
SrTiO384  100 >10 ✓ ⋯ ⋯ ⋯ 
Nb: SrTiO385  ⋯ ⋯ ✓ ✓ ⋯ ✓ 
Nb: SrTiO386  ⋯ ⋯ ✓ ✓ ⋯ ✓ 
BST-Nd2O387  10 000 ∼10 ✓ ✓ ⋯ ✓ 
BST88  107 1000 ⋯ ⋯ ⋯ ⋯ 
BST58  1000 ∼97 ✓ ✓ ✓ ✓ 
OD-/OR-BST (this work) 10 800 ∼560 ✓ ✓ ✓ ✓ 

In the development of oxide-based bilayer neuromorphic photoelectric memristors (NPMs), modulating the oxygen content of functional material is emerging as a new trend. At first, Hu et al.34 developed a non-transparent, all-optically controlled memristor based on OD-/OR-IGZO homo-junction and reported the unidirectional current response by exposing it to a wide range of light spectrum (420–1000 nm). Song et al.39 used the different wavelengths of UV-light (330–370 nm) to illuminate an ITO/OR-/OP-IGZO/Ti/Pt based device and achieved only an EPSC response. On the other hand, Wang and co-workers developed one non-transparent memristor, Ag/Ta2O5–x/Ta2O5/N–Si,40 and one transparent memristor, ITO/Ta2O5–3x/Ta2O5–x/ITO.41 For the former device, a significant improvement in current has been recorded for blue and red light illuminations, whereas for the latter device, UV light (365 nm) was used to emulate synaptic functions such as paired pulse facilitation (PPF), short- to long-term memory (STM–LTM) transition, and classical eye blink reflex with EPSC response. However, wavelength shorter than 380 nm and longer than 780 nm are invisible to the human eye,37,42,43 and none of the abovementioned works34,39–41 achieved the BPR by using visible light only. It indicates that the development of an artificial ocular system in a single memristor, which requires photo-excitation and photo-inhibition44 with biologically adaptive light pulses, is challenging. Some previously reported two-terminal synaptic devices with a single material based functional layer composed of two different oxygen concentrations are summarized in Table III.

TABLE III.

Comparison of the previously reported memristive synapses based on a single material as a functional layer, comprised of different oxygen concentrations, with our present device.

Material used as functional layer with different oxygen contentMemristive synapseOptical sensingWavelength (nm)BPRNeuromorphic application
TiO255  ✓ ⋯ ⋯ ⋯ ⋯ 
Ta2O541  ✓ ✓ 365 ⋯ ✓ 
Ta2O540  ✓ ✓ 360–620 ⋯ ✓ 
Zn2SnO489  ✓ ✓ 405–633 ⋯ ⋯ 
Ga2O390  ✓ ✓ 254, 365 ⋯ ✓ 
InGaZnO39  ✓ ✓ 330 ⋯ ✓ 
InGaZnO34  ✓ ✓ 420–1000 ⋯ ✓ 
BST (this work) ✓ ✓ 405/633 ✓ ✓ 
Material used as functional layer with different oxygen contentMemristive synapseOptical sensingWavelength (nm)BPRNeuromorphic application
TiO255  ✓ ⋯ ⋯ ⋯ ⋯ 
Ta2O541  ✓ ✓ 365 ⋯ ✓ 
Ta2O540  ✓ ✓ 360–620 ⋯ ✓ 
Zn2SnO489  ✓ ✓ 405–633 ⋯ ⋯ 
Ga2O390  ✓ ✓ 254, 365 ⋯ ✓ 
InGaZnO39  ✓ ✓ 330 ⋯ ✓ 
InGaZnO34  ✓ ✓ 420–1000 ⋯ ✓ 
BST (this work) ✓ ✓ 405/633 ✓ ✓ 

Motivated in this fashion, we develop an ABO3-based perovskite memristor45,46 with a bilayer stacking of oxygen rich (OR) and oxygen deficient (OD) thin films of barium strontium titanate (BST). In this work, an AZO/OD-BST/OR-BST/ITO based fully transparent NPM was exposed to 405 and 633 nm wavelengths of the visible light spectrum and produced the photo-excitation and photo-inhibition phenomena, respectively. The recorded EPSC and IPSC responses47 were utilized to imitate STM–LTM transition28 via spike intensity dependent plasticity (SIDP), spike width dependent plasticity (SWDP), spike rate dependent plasticity (SRDP), and spike number dependent plasticity (SNDP) behaviors. We also simulated the image sharpening9 image memory48 functions of the human vision system. The synaptic weight obtained from electrically induced long term-potentiation/depression (LT-P/D) characteristics was integrated into a convolutional neural network (CNN)20 to classify face images of the Olivetti Research Laboratory (ORL)49 dataset. These findings demonstrate that the negative and positive photo responses are highly pursued in the design of retina-inspired artificial ocular systems.

At first, isopropyl alcohol and then de-ionized water were used to clean the commercially available ITO-coated glass substrate via ultra-sonication. A 4 in. sputtering target of BST material was used to deposit ∼10 nm OR (Ar/O2:12/12 SCCM) and then ∼10 nm OD (Ar:24 SCCM) layers. Finally, a 70 nm thick layer of ZnO:Al2O3 (98:2 wt. %) was deposited as the top electrode (TE) by RF sputtering and patterned via a metallic shadow mask. 10 mTorr work pressure was maintained during the thin films’ deposition. This process resulted in a transparent AZO/OD-BST/OR-BST/ITO/glass-based memristor [Fig. 1(a)]. The optical photo was captured by a micromanipulator from the top view of the completely fabricated devices and is shown in Fig. 1(b). The device’s measurements were carried out by an Agilent B1500A attached to a probe station. Violet/red color LASER sources (405/633 nm) were used for photoelectric measurements, where the device current was read out at 0.2 V.

FIG. 1.

(a) Schematic representation of fabricated perovskite memristor, (b) optical photo of the fully fabricated perovskite memristors captured from the top view, (c) cross-sectional TEM image of AZO/OD-/OR-BST/ITO/glass substrate device, and (d) XRD spectra of OR- and OD-BST layers. XPS spectra of each element such as (e) barium, (f) strontium, (g) titanium, and (h) and (i) oxygen, presented in the BST layer.

FIG. 1.

(a) Schematic representation of fabricated perovskite memristor, (b) optical photo of the fully fabricated perovskite memristors captured from the top view, (c) cross-sectional TEM image of AZO/OD-/OR-BST/ITO/glass substrate device, and (d) XRD spectra of OR- and OD-BST layers. XPS spectra of each element such as (e) barium, (f) strontium, (g) titanium, and (h) and (i) oxygen, presented in the BST layer.

Close modal

The layer-by-layer sputtering deposition of each material, such as BST and AZO, on the ITO-coated glass substrate was confirmed by capturing a cross-sectional transmission electron microscopy (TEM) image at a magnification scale of 20 nm [Fig. 1(c)]. To confirm the phase structure of the functional layer, we deposited the OR- and OD-BST thin films separately on two glass substrates, and then a high resolution x-ray diffraction (HR-XRD) tool was used to scan their spectra from 10° to 80°. No discernible peaks belonging to crystalline-BST50 are observed in Fig. 1(d) and demonstrate the amorphous nature51 of thin films deposited through the sputtering method. Moreover, a broad peak, centered at around 22°, corresponds to the glass substrate.48,52 The core level spectra of Ba, Sr, Ti, and O elements presented in the BST thin film were analyzed by using x-ray photoelectron spectroscopy (XPS). Figure 1(e) indicates that the peak fitting of the Ba3d spectrum reveals two peaks, Ba3d5/2 at 778.9 eV and Ba3d3/2 at 794.2 eV. The binding energy separation of 15.3 eV between the two spin states of Ba confirms the +2 oxidation state.53, Figure 1(f) shows the Sr3d spectrum where the peaks at 132.6 and 134.3 eV confirm the binding energies of Sr3d5/2 and Sr3d3/2, respectively, indicating the presence of the Sr2+ state.54 Similarly, Fig. 1(g) shows the spectrum of Ti2p, where the peaks at 457.5 and 463.2 eV confirm the binding energies of Ti2p3/2 and Ti2p1/2, respectively, which suggests that the Ti element exists in the state of Ti4+.53  Figures 1(h) and 1(i) show the O1s spectra of OR- and OD-BST thin films, where the spectra can be divided into two peaks using Gaussian fitting and confirm the presence of lattice oxygen and oxygen vacancies. The amount of oxygen vacancies in OR-BST is smaller than that of OD-BST thin film, which is attributed to the oxygen incorporation during the deposition process of the OR layer.

During the electrical investigation, the ITO bottom electrode was kept grounded, while the AZO top electrode (TE) was connected to the test voltage. Initially, a positive forming process was needed to switch the low conductance state (LCS) of the pristine device into the high conductance state (HCS) [Fig. 2(a)]. Under the positive bias, the migration of oxygen ions toward TE generates oxygen vacancies and forms the conductive filaments (CFs) [Fig. 2(b)] in the active medium. Meanwhile, the interface/boundary in between the OD and OR layers moves toward the TE with oxygen ions.55 Now, the voltage is swept 0 → −1.4 → 0 V, referred to as the reset process [Fig. 2(a)]. Subsequently, oxygen ions discharged from the TE react with oxygen vacancies, causing a gradual rupture of CFs [Fig. 2(b)]. The introduction of the OD-BST layer in between the AZO and OR-BST layers improves the oxygen concentration gradient and presents gradual resistive switching (RS) behavior, in which the importance of oxygen vacancies to forming/rupturing CFs is highlighted.55 The I–V curve presents an “analog-type”56 RS behavior, which is suitable for neuromorphic computing applications. The voltage sweep process (0 → −1.4 → 0 → +2 → 0 V), referring to reset/set, was consecutively repeated more than 10 000 times. Afterward, the device current was read at 0.2 V in both conductance states and depicted the endurance characteristics of the device with a wide and stable memory window (×560) [Fig. 2(c)]. Moreover, 11 non-consecutive I–V curves recorded at different number of voltage sweeps, such as 1000th, 2000th, etc., were plotted, and each of them is shown in Fig. 2(d). The set voltages (Vset) extracted from Fig. 2(d) indicate a very narrow distribution in between 0.96 and 1.02 V [Fig. 2(e)]. On the other hand, no fluctuation in reset voltage is observed because the device transitions back to LCS in a very gradual manner, and the current starts to reduce only after reaching the reset-stop voltage.57 A comparison of the OD-/OR-BST based bilayer device with our previously reported58 BST based single layer device is summarized in Table IV.

FIG. 2.

(a) I–V characteristic with bipolar RS behavior. (b) Schematic of RS mechanism involving the formation and rupture of oxygen vacancy based CFs in pristine state, set, and reset (left to right). (c) Endurance characteristics. (d) I–V curves recorded at different number of voltage sweeps. (e) Distribution of set voltages extracted from 11 non-consecutive voltage sweeps shown in (d). LT-P/D characteristics with (f) NL values, (g) max/min conductance states, and (h) SEs.

FIG. 2.

(a) I–V characteristic with bipolar RS behavior. (b) Schematic of RS mechanism involving the formation and rupture of oxygen vacancy based CFs in pristine state, set, and reset (left to right). (c) Endurance characteristics. (d) I–V curves recorded at different number of voltage sweeps. (e) Distribution of set voltages extracted from 11 non-consecutive voltage sweeps shown in (d). LT-P/D characteristics with (f) NL values, (g) max/min conductance states, and (h) SEs.

Close modal
TABLE IV.

Comparison of our previously reported BST based single layer device with this work.

ParametersBST-based single layer device58 OD-/OR-BST based bilayer device (this work)
Substrate Glass 
Bottom electrode ITO 
Top electrode AZO 
Total functional layer thickness (nm) ∼30 ∼21 
Endurance cycles (#) 103 >104 
Set voltage 1.27 0.98 
Memory window ∼97 ∼560 
Non-linearity in conductance 2/2.4 0.35/1.3 
Symmetric error in device conductance ∼0.5 ∼0.03 
Exposure on/off-time ratio for the same wavelength ∼1:10 ∼1:25 
ParametersBST-based single layer device58 OD-/OR-BST based bilayer device (this work)
Substrate Glass 
Bottom electrode ITO 
Top electrode AZO 
Total functional layer thickness (nm) ∼30 ∼21 
Endurance cycles (#) 103 >104 
Set voltage 1.27 0.98 
Memory window ∼97 ∼560 
Non-linearity in conductance 2/2.4 0.35/1.3 
Symmetric error in device conductance ∼0.5 ∼0.03 
Exposure on/off-time ratio for the same wavelength ∼1:10 ∼1:25 

LT-P/D characteristics are usually considered to describe the long-term synaptic plasticity and can be implemented by triggering the synaptic device with successive voltage pulses of opposite polarities. As shown in Fig. 2(f), 250 identical and successive positive/negative voltage pulses are applied, which potentiate/depress the conductance states similar to the learning/forgetting process of organisms. The gradual set and reset operations in memristors often exhibit some non-ideal characteristics such as non-linearity (NL), asymmetry, etc., in conductance change.59 Therefore, the non-linearity (NL) values of LT-P/D curves are calculated and exhibited in Fig. 2(f), which play a decisive role in the performance of the neuromorphic computing system. Figure 2(g) depicts a good stability of the maximum/minimum conductance state (Gmax/Gmin) throughout the switching for 125 cycles. Figure 2(h) shows that the calculated symmetric errors (SEs) of some randomly selected LT-P/D cycles are close to their ideal value (i.e., 0), which indicates the good symmetry in the distributed conductance states. A detailed explanation of the NL and SE calculations was reported in Refs. 20 and 48.

The asymmetric weight update in an artificial synapse prevents a hardware neural network from yielding the same high-level accuracies as those yielded by a software neural network.60 Therefore, to observe the feasibility of OD-/OR-BST based NPM as an artificial synapse in an artificial neural network, we performed a software based CNN simulation to classify the images of humans’ faces. As shown in Fig. 3(a), the neural network architecture has a total of eight layers, including three convolutional layers (32, 32, 64 neurons), three max-pooling layers, and two fully connected layers (256, 40 neurons). The ORL dataset with a total of 400 gray scale face images (40 persons × 10 images; pixel-size 28 × 28) is used to train (280 images) and test (120 images) the model. The simulation was performed on the PyTorch platform and based on the backpropagation algorithm to minimize the error function by enabling the gradient descent.59 Here, the device conductance states [shown in Fig. 2(f)] were associated with the weight layers of the CNN model. The results based on the weight update characteristics of the memristor yield the image classification accuracy of 97.5% after #1000 epochs [Fig. 3(b)]. Furthermore, some amount of uniform noise61 was added to damage the input images [Fig. 3(c)], which helps to adapt to the existence of disturbances. Under the noise interference, the same CNN model was tested and acquired 86% image classification accuracy for a 20% noise level. In Fig. 3(d), the radar chart shows the degradation in classification accuracy with the increase in uniform noise level.

FIG. 3.

(a) Architecture of CNN model, (b) accuracy vs epoch curve of memristive synapse with different noise levels, (c) a face image belonging to the ORL dataset without and with uniform-noise interference, and (d) accuracy vs noise level spectra. The ORL dataset is freely available for research and educational purposes,49 and its credit is given to AT&T Laboratories Cambridge (https://www.cl.cam.ac.uk/research/dtg/attarchive/facedatabase.html). There is no need to obtain explicit permission from any individual or organization to use it as described in its usage agreement.

FIG. 3.

(a) Architecture of CNN model, (b) accuracy vs epoch curve of memristive synapse with different noise levels, (c) a face image belonging to the ORL dataset without and with uniform-noise interference, and (d) accuracy vs noise level spectra. The ORL dataset is freely available for research and educational purposes,49 and its credit is given to AT&T Laboratories Cambridge (https://www.cl.cam.ac.uk/research/dtg/attarchive/facedatabase.html). There is no need to obtain explicit permission from any individual or organization to use it as described in its usage agreement.

Close modal

When our NPM is exposed to the violet light (405 nm), then an initial abrupt reduction in the post-synaptic current (PSC) occurs, which further follows the EPSC behavior of the device [Figs. 4(a) and 4(b)]. The plausible reason behind that abrupt reduction in PSC is the charged-impurity-scattering effect,62,63 which originates from the ionization of shallowly located metal–metal bonding defects in the energy gap of BST.64 It causes a strong reduction in the mobility of electrons, and consequently, a sudden drop in PSC is recorded. The higher photon energy of violet light can ionize the deeply located oxygen vacancies65 and immediately revoke that scattering effect.58,63 Significantly generated photo electrons result in an increment in the PSC (i.e., EPSC). After turning-off the violet light source, the EPSC slightly increases first (instead of reducing), but just for a very short period of time [Fig. 4(c)]. It indicates the de-ionization of ionized metal–metal bonding defects first under a dark environment.58 This process is immediately influenced by the de-ionization of ionized oxygen vacancies and results in a persistent photo conductivity (PPC) effect. Under the influence of the PPC effect,66 the EPSC starts to recover itself very gradually and is referred to as the natural recovery/forgetting process. On the other hand, under the red light illumination, the device produces only an IPSC response [Fig. 4(d)], and the plausible reason is that the insufficient photon energy (1.95 eV) of red light could not ionize the deeply located oxygen vacancies. Therefore, the non-abolishment of the scattering effect leads to a continuous reduction in the PSC of the device. After turning-off the red light source, the de-ionization of ionized metal–metal bonding defects leads to the recovery of IPSC.58,63 These photo-excitation and photo-inhibition phenomena indicate that our OD-/OR-BST based NPM belongs to the category of all-optically controlled memristors with BPR.

FIG. 4.

(a) EPSC under the illumination of 405 nm, magnified-view of variation in PSC immediately after (b) turn-on, (c) turn-off the 405 nm light, and (d) IPSC under the illumination of 633 nm.

FIG. 4.

(a) EPSC under the illumination of 405 nm, magnified-view of variation in PSC immediately after (b) turn-on, (c) turn-off the 405 nm light, and (d) IPSC under the illumination of 633 nm.

Close modal

In recent times, so many reports on photoelectric synaptic devices67–69 have been published where the STM–LTM transition was achieved by simulating the SIDP, SWDP, SRDP, and SNDP behaviors of biological synapses. Among all these spike dependent plasticity behaviors, SIDP is emulated by modulating the input optical power (usually in μW and mW units), which subsequently affects the optical power density. In our experiment, the optical power density was modulated by varying the beam-angle70,71 of the violet LASER and the beam-distance48,72 of the red LASER without making any changes in input optical power. The LASER beam spot area on the device reduces (increases) with increasing beam angle (beam distance) and inversely affects the optical power density. Figure 5(a) shows that the beam angle of the violet LASER (optical power: 10 mW) increases from 40° to 80° and leads to an increase in EPSC, which is consistent with Ref. 70. Moreover, an enhanced memory effect (STM–LTM transition) was observed, which indicates that our OD-/OR-BST based NPM can determine the direction of the LASER beam source in 3D space under the premise of fixed wavelength and input power. As we know, among all wavelengths of visible light, red color is universally used for a “stop” signal because of its least scattering through air particles. Here, we used the red LASER to illuminate the device from different distances. Figure 5(b) shows that the beam distance of the red LASER (optical power: 10 mW) increases from 5 to 65 cm and causes a significant reduction in the absolute value of IPSC. In the field of photoelectric artificial synapses, things look bright at short distances and dim at long distances, which is close to the behavior of the human vision system. It is worth noting that the capability to determine the orientation and distance of an optical source is a vital indicator for simulating an artificial ocular system.

FIG. 5.

SIDP behavior by modulating the (a) beam angle and (b) beam distance. SWDP behavior under (c) violet and (d) red light illumination. Analysis of forgetting curves by fitting the dark current values with (e) Wickelgren’s power and (g) Kohlrausch functions, respectively. Radar charts showing the dependency of (f) forgetting rate parameter and (h) relaxation time on spike widths.

FIG. 5.

SIDP behavior by modulating the (a) beam angle and (b) beam distance. SWDP behavior under (c) violet and (d) red light illumination. Analysis of forgetting curves by fitting the dark current values with (e) Wickelgren’s power and (g) Kohlrausch functions, respectively. Radar charts showing the dependency of (f) forgetting rate parameter and (h) relaxation time on spike widths.

Close modal

To further gain insight into the STM–LTM transition, the width of violet (red) light spikes was manipulated from 10 to 50 s to demonstrate the excitatory (inhibitory)-induced SWDP behavior [Figs. 5(c) and 5(d)]. In both cases, the change in PSC increases with increasing width of the applied optical spike. Moreover, the natural forgetting (or recovery) process of PSCs was analyzed by fitting the dark current values recorded after closing the LASER sources with Wickelgren’s power law function73,74 (for EPSC) and Kohlrausch exponential function35,48 (for IPSC). Wickelgren’s power law function is mentioned in the inset of Fig. 5(e), where “ψ” is the forgetting parameter. The radar chart shown in Fig. 5(f) depicts the values of spike width dependent “ψ,” where “ψ” reduces with the increasing width of the spike. It can be noted that the rigorous learning can setup a more stable state, which also reduces the forgetting ability (small “ψ”). On the other hand, the Kohlrausch exponential function is mentioned in the inset of Fig. 5(g), where “τ” is the relaxation time of photo ionized carriers. The radar chart shown in Fig. 5(h) depicts the values of spike width dependent “τ,” where “τ” increases with the increasing width of the spike and results in a slower recovery of IPSC. All the results shown in Figs. 5(c)5(h) can relate to the Atkinson–Shiffrin model,68 a classic learning and memory model, which illustrates that the human brain can remember information for a longer time if there are longer impressions at the learning time.

Two other important synaptic functions of the human brain are SRDP and SNDP behaviors. In SRDP, the rate or frequency of spikes is controlled by adjusting the time interval. Figure 6(a) depicts that EPSC continuously increases with a rate of 10 consecutive violet light spikes, which results in distinct learning levels. Consequently, an obvious increment in EPSCgain (E10/E1) was observed and calculated by taking the ratio of EPSC values induced by the tenth spike to the first spike of the same stimulation set. Figure 6(b) shows the relationship in between EPSCgain and spike rate, which is fitted well by a sigmoidal shaped function [inset of Fig. 6(b)], similar to the high pass filtering (HPF) feature witnessed in biological synapses.9,72 The fitting result shows a cutoff frequency (fc) of 147.13 MHz and indicates that the device can evaluate the passing capacity of visual information and mimic the feature extraction of the human vision system. This HPF feature is usually utilized to strengthen the image’s sharpness. One gray scale image of a flower is processed by HPF (with fc = 147.13 MHz), in which the edge enhancement effect in flower petals and leaves is evident [Figs. 6(c) and 6(d)]. For emulating SNDP behavior, the spike number (n) of red light is increased, resulting in a larger change in IPSC [Fig. 6(e)]. An obvious improvement in the learning as well as memory level is observed, which can be further implemented as an image memory function48 of the human vision system. The 5 × 5-pixel images of four alphabets, “N,” “Y,” “C,” and “U,” are associated with 5, 10, 15, and 20 spikes, respectively [Fig. 6(f)]. We can observe that increasing “n” produces a clearer image, which is attributed to the larger change in synaptic weight in between neurons. As the memory level absolutely reduces over forgetting (or recovery) time, the images start to get blurred and almost disappear for a smaller number of spikes.

FIG. 6.

(a) SRDP behavior under violet light. (b) Plot of EPSC gain against the spike rate. Flower images (c) before and (d) after processing by HPF. (e) SNDP behavior under red light. (f) Image memory application.

FIG. 6.

(a) SRDP behavior under violet light. (b) Plot of EPSC gain against the spike rate. Flower images (c) before and (d) after processing by HPF. (e) SNDP behavior under red light. (f) Image memory application.

Close modal

The OD-/OR-BST based transparent NPM was successfully fabricated. Its reversibly tuned conductance states changed the connection strength between neurons and also influenced the learning ability of the CNN model. A high classification accuracy of 97.5% was achieved for the human face images of the ORL dataset. A successful transition from STM to LTM was achieved by varying the strength of optical stimuli via increasing (reducing) beam-angle (beam-distance) spike width, spike rate, and spike number. It exhibited the suitability of our NPM as an artificial ocular system. These findings highlight an effective and promising strategy for the development of oxide-based all-optically controlled neuromorphic memristors with BPR.

This work was supported by the National Science and Technology Council (NSTC), Taiwan (Project No. 111-2221-E-A49-160-MY3).

The authors have no conflicts to disclose.

Saransh Shrivastava: Conceptualization (lead); Data curation (equal); Formal analysis (equal); Methodology (equal); Validation (lead); Writing – original draft (lead); Writing – review & editing (lead). Stephen Ekaputra Limantoro: Formal analysis (equal); Software (lead). Hans Juliano: Data curation (equal); Methodology (equal). Tseung-Yuen Tseng: Funding acquisition (lead); Project administration (lead); Resources (lead); Supervision (lead); Visualization (lead); Writing – review & editing (equal).

The data that support the findings of this study are available from the corresponding authors upon reasonable request.

1.
C.
Yoo
,
T.-J.
Ko
,
M. G.
Kaium
,
R.
Martinez
,
M. M.
Islam
,
H.
Li
,
J. H.
Kim
,
J.
Cao
,
M.
Acharya
,
T.
Roy
, and
Y.
Jung
, “
A minireview on 2D materials-enabled optoelectronic artificial synaptic devices
,”
APL Mater.
10
(
7
),
070702
(
2022
).
2.
J.
Zhu
,
T.
Zhang
,
Y.
Yang
, and
R.
Huang
, “
A comprehensive review on emerging artificial neuromorphic devices
,”
Appl. Phys. Rev.
7
(
1
),
011312
(
2020
).
3.
D.
Kuzum
,
S.
Yu
, and
H.-S.
Philip Wong
, “
Synaptic electronics: Materials, devices and applications
,”
Nanotechnology
24
(
38
),
382001
(
2013
).
4.
R.
Yang
,
H.
Huang
, and
X.
Guo
, “
Memristive synapses and neurons for bioinspired computing
,”
Adv. Electron. Mater.
5
(
9
),
1900287
(
2019
).
5.
Y.
Sun
,
H.
Wang
, and
D.
Xie
, “
Recent advance in synaptic plasticity modulation techniques for neuromorphic applications
,”
Nano-Micro Lett.
16
(
1
),
211
(
2024
).
6.
Y.
Zhang
,
Z.
Wang
,
J.
Zhu
,
Y.
Yang
,
M.
Rao
,
W.
Song
,
Y.
Zhuo
,
X.
Zhang
,
M.
Cui
,
L.
Shen
,
R.
Huang
, and
J.
Joshua Yang
, “
Brain-inspired computing with memristors: Challenges in devices, circuits, and systems
,”
Appl. Phys. Rev.
7
(
1
),
011308
(
2020
).
7.
Y.
He
,
Y.
Ting
,
H.
Hu
,
T.
Diemant
,
Y.
Dai
,
J.
Lin
,
S.
Schweidler
,
G. C.
Marques
,
H.
Hahn
,
Y.
Ma
,
T.
Brezesinski
,
P. M.
Kowalski
,
B.
Breitung
, and
J.
Aghassi‐Hagmann
, “
Printed high‐entropy prussian blue analogs for advanced non‐volatile memristive devices
,”
Adv. Mater.
37
(
8
),
2410060
(
2025
).
8.
R.
Yang
,
D.
Hu
,
Q.
Chen
,
Z.
Wang
,
B.
Lu
,
Z.
Ye
,
X.
Li
, and
J.
Lu
, “
SnS‐facilitated ZnAlSnO‐based fully optically modulated artificial synaptic device for image processing
,”
Adv. Funct. Mater.
35
,
2414210
(
2024
).
9.
M.
Ma
,
C.
Huang
,
M.
Yang
,
D.
He
,
Y.
Pei
,
Y.
Kang
,
W.
Li
,
C.
Lei
, and
X.
Xiao
, “
Ultra‐low power consumption artificial photoelectric synapses based on Lewis acid doped WSe2 for neuromorphic computing
,”
Small
20
(
51
),
e2406402
(
2024
).
10.
X.
Duan
,
Z.
Cao
,
K.
Gao
,
W.
Yan
,
S.
Sun
,
G.
Zhou
,
Z.
Wu
,
F.
Ren
, and
B.
Sun
, “
Memristor‐based neuromorphic chips
,”
Adv. Mater.
36
(
14
),
2310704
(
2024
).
11.
D.
Kumar
,
A.
Saleem
,
L. B.
Keong
,
A.
Singh
,
Y. H.
Wang
, and
T.-Y.
Tseng
, “
ZnSnOy/ZnSnOx bilayer transparent memristive synaptic device for neuromorphic computing
,”
IEEE Electron Device Lett.
43
(
8
),
1211
1214
(
2022
).
12.
H.
Seok
,
D.
Lee
,
S.
Son
,
H.
Choi
,
G.
Kim
, and
T.
Kim
, “
Beyond von Neumann architecture: Brain‐inspired artificial neuromorphic devices and integrated computing
,”
Adv. Electron. Mater.
10
(
8
),
2300839
(
2024
).
13.
Y.
Lin
,
W.
Wang
,
R.
Li
,
J.
Kim
,
C.
Zhang
,
H.
Kan
, and
Y.
Li
, “
Multifunctional optoelectronic memristor based on CeO2/MoS2 heterojunction for advanced artificial synapses and bionic visual system with nociceptive sensing
,”
Nano Energy
121
,
109267
(
2024
).
14.
R.
Li
,
W.
Wang
,
Y.
Li
,
S.
Gao
,
W.
Yue
, and
G.
Shen
, “
Multi-modulated optoelectronic memristor based on Ga2O3/MoS2 heterojunction for bionic synapses and artificial visual system
,”
Nano Energy
111
,
108398
(
2023
).
15.
H.
Jang
,
J.
Lee
,
C.
Beak
,
S.
Biswas
,
S.
Lee
, and
H.
Kim
, “
Flexible neuromorphic electronics for wearable near‐sensor and in‐sensor computing systems
,”
Adv. Mater.
37
,
2416073
(
2025
).
16.
H.
Li
,
H.
Yu
,
D.
Wu
,
X.
Sun
, and
L.
Pan
, “
Recent advances in bioinspired vision sensor arrays based on advanced optoelectronic materials
,”
APL Mater.
11
(
8
),
080601
(
2023
).
17.
L.
Zhu
,
S.
Li
,
K.
Shu
,
S.
Ke
,
X.
Wan
,
H.
Sun
,
S.
Yan
,
Y.
Xu
,
C. L.
Tan
,
G.
He
,
Z.
Yu
, and
C.
Wan
, “
Visible-light responsive CdS-QDs modified InGaZnO synapse for biologically plausible color-to-gray conversion
,”
Appl. Phys. Lett.
125
(
3
),
033502
(
2024
).
18.
P.
Xie
,
D.
Li
,
S.
Yip
, and
J. C.
Ho
, “
Emerging optoelectronic artificial synapses and memristors based on low-dimensional nanomaterials
,”
Appl. Phys. Rev.
11
(
1
),
011303
(
2024
).
19.
Y.
Zhang
,
H.
Chen
,
W.
Sun
,
Y.
Hou
,
Y.
Cai
, and
H.
Huang
, “
All‐photonic synapses for biomimetic ocular system
,”
Adv. Funct. Mater.
34
,
2409419
(
2024
).
20.
S.
Shrivastava
,
H.-N.
Chi
,
S. E.
Limantoro
,
H.
Juliano
, and
T.-Y.
Tseng
, “
Fully visible light controlled neuro-synaptic ReRAM device based on metal oxide heterojunction
,”
Appl. Phys. Lett.
124
(
13
),
133503
(
2024
).
21.
J. H.
Jeong
,
J. H.
Ma
,
M. H.
Park
,
H. J.
Ha
,
S. J.
Kang
,
J. M.
Yun
,
Y. b.
Kim
,
E.
Kim
, and
S. J.
Kang
, “
Oxide based pentachromatic‐vision inspired optoelectronic synaptic transistor with large conduction states over 512
,”
Adv. Funct. Mater.
34
(
38
),
2402222
(
2024
).
22.
C.
Yang
,
B.
Sun
,
G.
Zhou
,
T.
Guo
,
C.
Ke
,
Y.
Chen
,
J.
Shao
,
Y.
Zhao
, and
H.
Wang
, “
Photoelectric memristor-based machine vision for artificial intelligence applications
,”
ACS Mater. Lett.
5
(
2
),
504
526
(
2023
).
23.
S. W.
Cho
,
C.
Jo
,
Y.-H.
Kim
, and
S. K.
Park
, “
Progress of materials and devices for neuromorphic vision sensors
,”
Nano-Micro Lett.
14
(
1
),
203
(
2022
).
24.
L.
Zhao
,
Y.
Gao
,
X.
Fu
,
Y.
Chen
,
B.
Zhang
, and
F.
Xuan
, “
Photodual‐responsive anthracene‐based 2D covalent organic framework for optoelectronic synaptic devices
,”
Small Methods
9
,
2401341
(
2024
).
25.
J.
Wang
,
N.
Ilyas
,
Y.
Ren
,
Y.
Ji
,
S.
Li
,
C.
Li
,
F.
Liu
,
D.
Gu
, and
K.
Ang
, “
Technology and integration roadmap for optoelectronic memristor
,”
Adv. Mater.
36
(
9
),
2307393
(
2024
).
26.
S.
B
and
P.
Dwivedi
, “
Scalable integrated sensing and computing memory devices for logic circuit realization: V2O5/WO3 heterostructures
,”
ACS Appl. Electron. Mater.
7
(
8
),
3191
3200
(
2025
).
27.
D.
Kumar
,
H.
Li
,
U. K.
Das
,
A. M.
Syed
, and
N.
El‐Atab
, “
Flexible solution‐processable black‐phosphorus‐based optoelectronic memristive synapses for neuromorphic computing and artificial visual perception applications
,”
Adv. Mater.
35
(
28
),
2300446
(
2023
).
28.
D.
Cui
,
M.
Pei
,
Z.
Lin
,
H.
Zhang
,
M.
Kang
,
Y.
Wang
,
X.
Gao
,
J.
Su
,
J.
Miao
,
Y.
Li
,
J.
Zhang
,
Y.
Hao
, and
J.
Chang
, “
Versatile optoelectronic memristor based on wide-bandgap Ga2O3 for artificial synapses and neuromorphic computing
,”
Light: Sci. Appl.
14
(
1
),
161
(
2025
).
29.
S.
Shrivastava
,
Y.-T.
Lin
,
B.
Pattanayak
,
S.
Pratik
,
C.-C.
Hsu
,
D.
Kumar
,
A. S.
Lin
, and
T.-Y.
Tseng
, “
Zn2SnO4 thin film based nonvolatile positive optoelectronic memory for neuromorphic computing
,”
ACS Appl. Electron. Mater.
4
(
4
),
1784
1793
(
2022
).
30.
Y.
Liang
,
F.
Zhang
,
S.
Qi
,
D.
Dong
,
X.
Ma
,
Y.
Yang
,
W.
Tang
,
H.
Yang
, and
Z.
Wu
, “
Deep ultraviolet optoelectronic memristors based on gallium oxide for biomimetic visual processing and neuromorphic memory applications
,”
Appl. Phys. Lett.
126
(
14
),
141102
(
2025
).
31.
X.
Li
,
Z.
Fang
,
X.
Guo
,
R.
Wang
,
Y.
Zhao
,
W.
Zhu
,
L.
Wang
, and
L.
Zhang
, “
Light-induced conductance potentiation and depression in an all-optically controlled memristor
,”
ACS Appl. Mater. Interfaces
16
(
21
),
27866
27874
(
2024
).
32.
C.
Lu
,
J.
Meng
,
J.
Song
,
T.
Wang
,
H.
Zhu
,
Q.-Q.
Sun
,
D. W.
Zhang
, and
L.
Chen
, “
Self-rectifying all-optical modulated optoelectronic multistates memristor crossbar array for neuromorphic computing
,”
Nano Lett.
24
(
5
),
1667
1672
(
2024
).
33.
C.
Lu
,
J.
Meng
,
J.
Song
,
K.
Xu
,
T.
Wang
,
H.
Zhu
,
Q.-q.
Sun
,
D. W.
Zhang
, and
L.
Chen
, “
Reconfigurable selector-free all-optical controlled neuromorphic memristor for in-memory sensing and reservoir computing
,”
ACS Nano
18
(
43
),
29715
29723
(
2024
).
34.
L.
Hu
,
J.
Yang
,
J.
Wang
,
P.
Cheng
,
L. O.
Chua
, and
F.
Zhuge
, “
All‐optically controlled memristor for optoelectronic neuromorphic computing
,”
Adv. Funct. Mater.
31
(
4
),
2005582
(
2021
).
35.
S.
Shrivastava
,
L. B.
Keong
,
S.
Pratik
,
A. S.
Lin
, and
T.
Tseng
, “
Fully photon controlled synaptic memristor for neuro‐inspired computing
,”
Adv. Electron. Mater.
9
(
3
),
2201093
(
2023
).
36.
J.
Zheng
,
Y.
Du
,
Y.
Dong
,
X.
Shan
,
Y.
Tao
,
Y.
Lin
,
X.
Zhao
,
Z.
Wang
,
H.
Xu
, and
Y.
Liu
, “
Fully light-modulated memristor based on ZnO/MoOx heterojunction for neuromorphic computing
,”
Appl. Phys. Lett.
124
(
13
),
133502
(
2024
).
37.
M.
Zhang
,
Y.
Jiang
,
Y.
Zhang
,
L.
Yu
,
S.
Zhang
,
X.
Li
,
X.
Song
,
K.
Liu
,
D.
Wang
,
J.
Wang
, and
C.
Xia
, “
Avian eye-inspired tetrachromatic-bidirectional synaptic transistor for multi-spectral recognition
,”
Appl. Phys. Lett.
126
(
13
),
133501
(
2025
).
38.
Z.
Feng
,
J.
Wang
,
F.
Chen
,
B.
Dong
,
X.
Ma
,
T.
Mei
,
N.
Yang
,
X.
Guan
,
L.
Hu
,
C.
Lin
,
Z.
Li
,
T.
Wu
, and
D.
Chu
, “
Coupling light into memristors: Advances in halide perovskite resistive switching and neuromorphic computing
,”
Small Methods
(published online) (
2025
).
39.
J.
Song
,
J.
Meng
,
C.
Lu
,
T.
Wang
,
H.
Zhu
,
Q.
Sun
,
D. W.
Zhang
, and
L.
Chen
, “
Photoelectric synaptic device based on bilayerd OR/OP-InGaZnO for neuromorphic computing
,”
IEEE Electron Device Lett.
45
(
1
),
120
123
(
2024
).
40.
W.
Wang
,
F.
Yin
,
H.
Niu
,
Y.
Li
,
E. S.
Kim
, and
N. Y.
Kim
, “
Tantalum pentoxide (Ta2O5 and Ta2O5–x)-based memristor for photonic in-memory computing application
,”
Nano Energy
106
,
108072
(
2023
).
41.
W.
Wang
,
N.-Y.
Kim
,
D.
Lee
,
F.
Yin
,
H.
Niu
,
E.
Ganbold
,
J.-W.
Park
,
Y.-K.
Shin
,
Y.
Li
, and
E.-S.
Kim
, “
Operant conditioning reflex implementation in a transparent Ta2O5–3x/Ta2O5–x homo-structured optoelectronic memristor for neuromorphic computing application
,”
Nano Energy
119
,
109102
(
2024
).
42.
H.
Zhang
,
X.
Ju
,
D.
Chi
,
L.
Feng
,
Z.
Liu
,
K.
Yew
,
M.
Zhu
,
T.
Li
,
R.
Wei
,
S.
Wang
,
L.
Sun
,
Z.
Wang
, and
Y.
Wu
, “
A neuromorphic bionic eye with broadband vision and biocompatibility using TIPS-pentacene-based phototransistor array retina
,”
Appl. Mater. Today
33
,
101885
(
2023
).
43.
Y.
Yang
,
Y.
Li
,
D.
Chen
, and
G.
Shen
, “
Multicolor vision perception of flexible optoelectronic synapse with high sensitivity for skin sunburn warning
,”
Mater. Horiz.
11
(
8
),
1934
1943
(
2024
).
44.
C.
Jin
,
J.
Wang
,
S.
Yang
,
Y.
Ding
,
J.
Chang
,
W.
Liu
,
Y.
Xu
,
X.
Shi
,
P.
Xie
,
J. C.
Ho
,
C.
Wan
,
Z.
Zheng
,
J.
Sun
,
L.
Liao
, and
J.
Yang
, “
Bidirectional photovoltage‐driven oxide transistors for neuromorphic visual sensors
,”
Adv. Mater.
37
(
1
),
2410398
(
2025
).
45.
M.
Liu
,
Z.
Cao
,
X.
Wang
,
S.
Mao
,
J.
Qin
,
Y.
Yang
,
Z.
Rao
,
Y.
Zhao
, and
B.
Sun
, “
Perovskite material-based memristors for applications in information processing and artificial intelligence
,”
J. Mater. Chem. C
11
(
39
),
13167
13188
(
2023
).
46.
B.
Sun
,
G.
Zhou
,
L.
Sun
,
H.
Zhao
,
Y.
Chen
,
F.
Yang
,
Y.
Zhao
, and
Q.
Song
, “
ABO3 multiferroic perovskite materials for memristive memory and neuromorphic computing
,”
Nanoscale Horiz.
6
(
12
),
939
970
(
2021
).
47.
D.
Kim
and
J.
Lee
, “
Neurotransmitter‐induced excitatory and inhibitory functions in artificial synapses
,”
Adv. Funct. Mater.
32
(
21
),
2200497
(
2022
).
48.
S.
Shrivastava
,
W.
Dai
,
S. E.
Limantoro
,
H.
Juliano
, and
T.
Tseng
, “
A violet‐light‐responsive ReRAM based on Zn2SnO4/Ga2O3 heterojunction as an artificial synapse for visual sensory and in‐memory computing
,”
Adv. Electron. Mater.
11
(
3
),
2400527
(
2025
).
49.
S.
Lawrence
,
C. L.
Giles
,
A.
Chung Tsoi
, and
A. D.
Back
, “
Face recognition: A convolutional neural-network approach
,”
IEEE Trans. Neural Networks
8
(
1
),
98
113
(
1997
).
50.
S.
Liu
,
B.
Zhang
,
L.
E
,
D.
Zhao
, and
Z.
Liu
, “
High efficient PEC performance via mechanical stress field and temperature field coupling assisted Ba0.7Sr0.3TiO3 under visible light irradiation
,”
Int. J. Hydrogen Energy
48
(
70
),
27269
27279
(
2023
).
51.
Z.
Xu
,
Y.
Tanushi
,
M.
Suzuki
,
K.
Wakushima
, and
S.
Yokoyama
, “
Optical properties of amorphous Ba0.7Sr0.3TiO3 thin films obtained by metal organic decomposition technique
,”
Thin Solid Films
515
(
4
),
2326
2331
(
2006
).
52.
R.
Yang
,
Y.
Tian
,
L.
Hu
,
S.
Li
,
F.
Wang
,
D.
Hu
,
Q.
Chen
,
X.
Pi
,
J.
Lu
,
F.
Zhuge
, and
Z.
Ye
, “
Dual-input optoelectronic synaptic transistor based on amorphous ZnAlSnO for multi-target neuromorphic simulation
,”
Mater. Today Nano
26
,
100480
(
2024
).
53.
D. K.
Bhat
,
H.
Bantawal
, and
U. S.
Shenoy
, “
Rhodium doping augments photocatalytic activity of barium titanate: Effect of electronic structure engineering
,”
Nanoscale Adv.
2
(
12
),
5688
5698
(
2020
).
54.
P. K.
Gopi
,
B.
Muthukutty
,
S.-M.
Chen
,
T.-W.
Chen
,
X.
Liu
,
A. A.
Alothman
,
M. A.
Ali
, and
S. M.
Wabaidur
, “
Platelet-structured strontium titanate perovskite decorated on graphene oxide as a nanocatalyst for electrochemical determination of neurotransmitter dopamine
,”
New J. Chem.
44
(
42
),
18431
18441
(
2020
).
55.
Y.
She
,
F.
Wang
,
X.
Zhao
,
Z.
Zhang
,
C.
Li
,
H.
Pan
,
K.
Hu
,
Z.
Song
, and
K.
Zhang
, “
Oxygen vacancy-dependent synaptic dynamic behavior of TiOx-based transparent memristor
,”
IEEE Trans. Electron Devices
68
(
4
),
1950
1955
(
2021
).
56.
S.
Shrivastava
,
S.
Pratik
,
A. S.
Lin
, and
T.-Y.
Tseng
, “
Emulating synaptic and nociceptive behavior via negative photoconductivity of a memristor
,”
IEEE Trans. Electron Devices
70
(
7
),
3530
3535
(
2023
).
57.
Y.
Qin
,
Z.
Wang
,
Y.
Ling
,
Y.
Cai
, and
R.
Huang
, “
A TaOx-based RRAM with improved uniformity and excellent analog characteristics by local dopant engineering
,”
Electronics
10
(
20
),
2451
(
2021
).
58.
S.
Shrivastava
,
S. E.
Limantoro
,
H.
Juliano
, and
T.-Y.
Tseng
, “
Bidirectional photoelectric response of perovskite oxide-based synaptic ReRAM device
,”
IEEE Electron Device Lett.
45
(
3
),
384
387
(
2024
).
59.
D.
Ielmini
and
G.
Pedretti
, “
Resistive switching random-access memory (RRAM): Applications and requirements for memory and computing
,”
Chem. Rev.
(published online) (
2025
).
60.
S.
Seo
,
B.-S.
Kang
,
J.-J.
Lee
,
H.-J.
Ryu
,
S.
Kim
,
H.
Kim
,
S.
Oh
,
J.
Shim
,
K.
Heo
,
S.
Oh
, and
J.-H.
Park
, “
Artificial van der Waals hybrid synapse and its application to acoustic pattern recognition
,”
Nat. Commun.
11
(
1
),
3936
(
2020
).
61.
Y.
Wang
,
M.
Cao
,
J.
Bian
,
Q.
Li
, and
J.
Su
, “
Flexible ZnO nanosheet‐based artificial synapses prepared by low‐temperature process for high recognition accuracy neuromorphic computing
,”
Adv. Funct. Mater.
32
(
52
),
2209907
(
2022
).
62.
A.
Reed
,
C.
Stone
,
K.
Roh
,
H. W.
Song
,
X.
Wang
,
M.
Liu
,
D.-K.
Ko
,
K.
No
, and
S.
Lee
, “
The role of third cation doping on phase stability, carrier transport and carrier suppression in amorphous oxide semiconductors
,”
J. Mater. Chem. C
8
(
39
),
13798
13810
(
2020
).
63.
Y.
Mi
,
C.
Yang
,
L.
Shih
, and
J.
Chen
, “
All‐optical‐controlled excitatory and inhibitory synaptic signaling through bipolar photoresponse of an oxide‐based phototransistor
,”
Adv. Opt. Mater.
11
(
14
),
2300089
(
2023
).
64.
A.
Ghalem
,
L.
Huitema
,
A.
Crunteanu
,
M.
Rammal
,
L.
Trupina
,
L.
Nedelcu
,
M. G.
Banciu
,
P.
Dutheil
,
C.
Constantinescu
,
P.
Marchet
,
F.
Dumas-Bouchiat
, and
C.
Champeaux
, “
Electrical transport properties and modelling of electrostrictive resonance phenomena in Ba2/3Sr1/3TiO3 thin films
,”
J. Appl. Phys.
120
(
18
),
184101
(
2016
).
65.
C.-C.
Hsu
,
S.
Shrivastava
,
S.
Pratik
,
S.
Chandrasekaran
, and
T.-Y.
Tseng
, “
ZTO/MgO-based optoelectronic synaptic memristor for neuromorphic computing
,”
IEEE Trans. Electron Devices
70
(
3
),
1048
1054
(
2023
).
66.
S.-L.
Gao
,
L.-P.
Qiu
,
J.
Zhang
,
W.-P.
Han
,
S.
Ramakrishna
, and
Y.-Z.
Long
, “
Persistent photoconductivity of metal oxide semiconductors
,”
ACS Appl. Electron. Mater.
6
(
3
),
1542
1561
(
2024
).
67.
L.
Dong
,
S.
Yuan
,
G.
Wei
,
P.
Zhu
,
S.
Ma
,
B.
Xu
, and
Y.
Yang
, “
Artificial optoelectronic synapse based on violet phosphorus microfiber arrays
,”
Small
20
(
13
),
2306998
(
2024
).
68.
M.
Cui
,
Z.
Wei
,
P.
Zhao
,
C.
Jiang
,
C.
Luo
,
H.
Lin
,
J.
Zhao
,
X.
Tang
, and
H.
Peng
, “
Optically stimulated artificial synapse with ferroelectric assistance for a neuromorphic visual system
,”
ACS Appl. Electron. Mater.
7
(
1
),
193
201
(
2025
).
69.
Y.
Wu
,
A.
Chang
,
W.
Chen
,
E.
Ercan
,
W.
Chen
,
B.
Lin
,
C.
Liu
,
Y.
Lin
, and
W.
Chen
, “
High‐performance synaptic phototransistor using A photoactive self‐assembled layer toward ultralow energy consumption
,”
Adv. Opt. Mater.
12
(
6
),
2302040
(
2024
).
70.
C.
Li
,
J.
Wang
,
D.
Li
,
N.
Ilyas
,
Z.
Yang
,
K.
Chen
,
P.
Gu
,
X.
Jiang
,
D.
Gu
,
F.
Liu
,
Y.
Jiang
, and
W.
Li
, “
An oxide-based heterojunction optoelectronic synaptic device with wideband and rapid response performance
,”
J. Mater. Sci. Technol.
123
,
159
167
(
2022
).
71.
C.
Li
,
N.
Ilyas
,
J.
Wang
,
Y.
Li
,
H.
Luo
,
D.
Li
,
D.
Gu
,
F.
Liu
,
Y.
Jiang
, and
W.
Li
, “
Nanostructured CuAlO2@ZnO optoelectronic device for artificial synaptic applications
,”
Appl. Surf. Sci.
611
,
155682
(
2023
).
72.
Y.
Chen
,
Y.
Huang
,
J.
Zeng
,
Y.
Kang
,
Y.
Tan
,
X.
Xie
,
B.
Wei
,
C.
Li
,
L.
Fang
, and
T.
Jiang
, “
Energy-efficient ReS2-based optoelectronic synapse for 3D object reconstruction and recognition
,”
ACS Appl. Mater. Interfaces
15
(
50
),
58631
58642
(
2023
).
73.
D.
Rao
,
A. I. K.
Pillai
,
M.
Garbrecht
, and
B.
Saha
, “
Scandium nitride as a gateway III‐nitride semiconductor for both excitatory and inhibitory optoelectronic artificial synaptic devices
,”
Adv. Electron. Mater.
9
(
3
),
2200975
(
2023
).
74.
M.
Huang
,
W.
Ali
,
L.
Yang
,
J.
Huang
,
C.
Yao
,
Y.
Xie
,
R.
Sun
,
C.
Zhu
,
Y.
Tan
,
X.
Liu
,
S.
Li
,
Z.
Li
, and
A.
Pan
, “
Multifunctional optoelectronic synapses based on arrayed MoS2 monolayers emulating human association memory
,”
Adv. Sci.
10
(
16
),
2300120
(
2023
).
75.
Y.
Cheng
,
Z.
Li
,
Y.
Lin
,
Z.
Wang
,
X.
Shan
,
Y.
Tao
,
X.
Zhao
,
H.
Xu
, and
Y.
Liu
, “
Color recognition achieved in multiwavelength controlled plasmonic optoelectronic memristor for neuromorphic visual system
,”
Adv. Funct. Mater.
35
(
5
),
2414404
(
2025
).
76.
S.
Sun
,
T.
Zhang
,
S.
Jin
,
X.
Pan
,
J.
Lu
,
Z.
Ye
, and
B.
Lu
, “
Fully UV modulated artificial synapses with integrated sensing, storage and computation
,”
Adv. Funct. Mater.
34
(
36
),
2401403
(
2024
).
77.
X.
Shan
,
C.
Zhao
,
X.
Wang
,
Z.
Wang
,
S.
Fu
,
Y.
Lin
,
T.
Zeng
,
X.
Zhao
,
H.
Xu
,
X.
Zhang
, and
Y.
Liu
, “
Plasmonic optoelectronic memristor enabling fully light‐modulated synaptic plasticity for neuromorphic vision
,”
Adv. Sci.
9
(
6
),
2104632
(
2022
).
78.
F.
Wu
and
T.-Y.
Tseng
, “
All-optically regulated ITO/Cu2O/WO3/ITO memristor for optoelectronic neuromorphic computing
,”
ACS Appl. Electron. Mater.
6
(
7
),
5212
5221
(
2024
).
79.
S.
Wu
,
L.
Ren
,
J.
Qing
,
F.
Yu
,
K.
Yang
,
M.
Yang
,
Y.
Wang
,
M.
Meng
,
W.
Zhou
,
X.
Zhou
, and
S.
Li
, “
Bipolar resistance switching in transparent ITO/LaAlO3/SrTiO3 memristors
,”
ACS Appl. Mater. Interfaces
6
(
11
),
8575
8579
(
2014
).
80.
H.-Y.
Lo
,
C.-Y.
Yang
,
G.-M.
Huang
,
C.-Y.
Huang
,
J.-Y.
Chen
,
C.-W.
Huang
,
Y.-H.
Chu
, and
W.-W.
Wu
, “
Observing topotactic phase transformation and resistive switching behaviors in low power SrCoOx memristor
,”
Nano Energy
72
(
1001
),
104683
(
2020
).
81.
B. N.
Rao
,
S.
Yasui
,
Y.
Han
,
Y.
Hamasaki
,
T.
Katayama
,
T.
Shiraishi
,
T.
Kiguchi
, and
M.
Itoh
, “
Redox-based multilevel resistive switching in AlFeO3 thin-film heterostructures
,”
ACS Appl. Electron. Mater.
2
(
4
),
1065
1073
(
2020
).
82.
Z.
Xu
,
K.
Jin
,
L.
Gu
,
Y.
Jin
,
C.
Ge
,
C.
Wang
,
H.
Guo
,
H.
Lu
,
R.
Zhao
, and
G.
Yang
, “
Evidence for a crucial role played by oxygen vacancies in LaMnO3 resistive switching memories
,”
Small
8
(
8
),
1279
1284
(
2012
).
83.
V. R.
Nallagatla
,
J.
Kim
,
K.
Lee
,
S. C.
Chae
,
C. S.
Hwang
, and
C. U.
Jung
, “
Complementary resistive switching and synaptic-like memory behavior in an epitaxial SrFeO2.5 thin film through oriented oxygen-vacancy channels
,”
ACS Appl. Mater. Interfaces
12
(
37
),
41740
41748
(
2020
).
84.
H.-X.
Fan
,
Z.-H.
Shen
,
W.-H.
Li
,
X.-G.
Tang
,
R.-K.
Zhao
,
J.
Hu
, and
K.-Y.
Wang
, “
Memristive synapses based on SrTiO3 thin film
,”
Mater. Sci. Eng.: B
309
,
117623
(
2024
).
85.
S.
Gao
,
G.
Liu
,
H.
Yang
,
C.
Hu
,
Q.
Chen
,
G.
Gong
,
W.
Xue
,
X.
Yi
,
J.
Shang
, and
R.-W.
Li
, “
An oxide Schottky junction artificial optoelectronic synapse
,”
ACS Nano
13
(
2
),
2634
2642
(
2019
).
86.
L.
Wang
,
Y.
Zhang
,
Z.
Guo
,
X.
Meng
,
Q.
Li
,
M.
Xu
,
R.
Gao
,
X.
Zhu
, and
P.
Wang
, “
High-precision attention mechanism for machine vision enabled by an artificial optoelectronic memristor synapse
,”
Nano Lett.
25
(
7
),
2716
2724
(
2025
).
87.
H.
He
,
Y.
Pei
,
L.
Chao
, and
X.
Yan
, “
Artificial vision system design and implementation based on BaSrTiO3 and Nd2O3 composite memristors for efficient pattern recognition
,”
J. Mater. Chem. C
13
,
10168
10177
(
2025
).
88.
Y.-C.
Wang
,
K.-H.
Chen
,
M.-C.
Kao
,
H.-C.
Chen
,
C.-M.
Cheng
,
H.-X.
Huang
, and
K.-C.
Huang
, “
Bipolar switching properties and reaction decay effect of BST ferroelectric thin films for applications in resistance random access memory devices
,”
Nanomaterials
15
(
8
),
602
(
2025
).
89.
D.
Kumar
,
A.
Saleem
,
L. B.
Keong
,
Y. H.
Wang
, and
T.-Y.
Tseng
, “
Light induced RESET phenomenon in invisible memristor for photo sensing
,”
IEEE Electron Device Lett.
43
(
7
),
1069
1072
(
2022
).
90.
R.
Li
,
Y.
Lin
,
Y.
Li
,
S.
Gao
,
W.
Yue
,
H.
Kan
,
C.
Zhang
, and
G.
Shen
, “
Amorphous gallium oxide homojunction-based optoelectronic synapse for multi-functional signal processing
,”
J. Semicond.
44
(
7
),
074101
(
2023
).