0% found this document useful (0 votes)
44 views9 pages

1 s2.0 S0042207X23007650 Main

This document summarizes research on applying an artificial synapse based on aluminum-doped strontium titanate (STAO) thin film for neuromorphic computing. Key points: 1. STAO thin films were fabricated using a sol-gel method and doped with aluminum to improve electrical properties. 2. Electrical characterization found the STAO synapses could effectively emulate spike-time-dependent plasticity and other synaptic functions. 3. A convolutional neural network was constructed using the STAO synapses and achieved 96.2% accuracy recognizing handwritten digits, demonstrating applicability for neuromorphic computing. 4. Aluminum doping was found to improve stability, linearity and recognition

Uploaded by

jjhonhack
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
0% found this document useful (0 votes)
44 views9 pages

1 s2.0 S0042207X23007650 Main

This document summarizes research on applying an artificial synapse based on aluminum-doped strontium titanate (STAO) thin film for neuromorphic computing. Key points: 1. STAO thin films were fabricated using a sol-gel method and doped with aluminum to improve electrical properties. 2. Electrical characterization found the STAO synapses could effectively emulate spike-time-dependent plasticity and other synaptic functions. 3. A convolutional neural network was constructed using the STAO synapses and achieved 96.2% accuracy recognizing handwritten digits, demonstrating applicability for neuromorphic computing. 4. Aluminum doping was found to improve stability, linearity and recognition

Uploaded by

jjhonhack
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 9

Vacuum 217 (2023) 112568

Contents lists available at ScienceDirect

Vacuum
journal homepage: www.elsevier.com/locate/vacuum

Application of artificial synapse based on Al-doped SrTiO3 thin film in


neuromorphic computing
Zhi-Hao Shen , Wen-Hua Li *, Xin-Gui Tang , Hao Chen , Jia Hu , Kai-Yuan Wang , Ke Meng ,
Yan-Ping Jiang , Xiao-Bin Guo
School of Physics and Optoelectronic Engineering, Guangdong University of Technology, Guangzhou Higher Education Mega Center, Guangzhou, 510006, People’s
Republic of China

A R T I C L E I N F O A B S T R A C T

Handling Editor: Prof. L.G. Hultman To alleviate the von Neumann architecture bottleneck, scholars have proposed neuromorphic computing con­
sisting of artificial synapses and neural network algorithms, in which the ability to well simulate biological
Keywords: synaptic functions is considered a key step in building neuromorphic computing. In this work, the artificial
SrTiO3 synapse based on Al-doped SrTiO3 (STAO) thin film is fabricated by a low-cost sol-gel method. It effectively
Memristor
emulated a diverse array of significant synaptic functions, including spike-time-dependent plasticity and short/
Artificial synapse
long plasticity. We constructed a convolutional neural network (CNN) to perform the training/recognition task of
Neuromorphic computing
handwritten digit images and demonstrated the applicability of the memristor device in neuromorphic
computing. We compared STAO with SrTiO3 (STO) and found that the Al-doped devices had better stability,
higher linearity, and higher image recognition accuracy from 84.5% to 96.2%. We have studied the main
influencing factors for these improving performances, which may be due to the generation of more oxygen
vacancies by Al doping. The x-ray photoelectron spectroscopy (XPS) test showed an increase in the oxygen
vacancy content of the sample from 23.59% to 37.91%, which supports our hypothesis. This has positive ram­
ifications for building artificial neural networks capable of processing vast quantities of data.

1. Introduction transmission properties of neurosynapses, as well as possessing both


storage and processing properties, renders them ideal for the develop­
The current information processing is still based on the traditional ment of artificial synapses [6,7]. Utilizing a singular memristor is suf­
von Neumann architecture of computer systems, and its “storage and ficient to emulate the fundamental functions of a neurosynapse [8]. In
computation separation” feature severely restricts the efficiency of in­ particular, an array based on memristors [9] and synaptic transistors can
formation storage and transmission. In order to break the traditional von perform neuromorphic computing that simulates the functions of the
Neumann architecture, people are constantly looking for new possibil­ human brain, enabling high-performance hyper-parallel computing,
ities, and one of the most desirable solutions is to design a “storage and thus making it possible to achieve comparable to the biological brain in
computation integration” device and build an artificial neural network a single chip, which alleviates the von Neumann bottleneck to a certain
so that the speed of data processing and data storage can be greatly extent [10–12]. Therefore, finding suitable materials to develop artifi­
increased [1,2]. cial synapses, simulate biological synapses, and build neuromorphic
The human brain can efficiently process huge amounts of data [3] computing has been the concern of researchers [13–17].
through parallel neuron-synapse structures, which is mainly attributed As a typical perovskite material, strontium titanate (SrTiO3) has
to the unique “memory-computer” structure of synapses [4,5]. There­ become a model material for memristors due to its chemical defects,
fore, synapse-like devices (called artificial synapses) are ideal devices oxygen diffusion, and cation diffusion [18–20]. Its physical properties
that are pivotal for overcoming the von Neumann structural bottleneck. are susceptible to systematic changes with oxidation state and doping,
The advent of emerging non-linear electronic components, namely and thus may exhibit many excellent properties [21]. In this study,
memristors, which exhibit electrical characteristics akin to the based on SrTiO3 thin films, we improved the electrical properties of the

* Corresponding author.
E-mail address: [email protected] (W.-H. Li).

https://doi.org/10.1016/j.vacuum.2023.112568
Received 29 May 2023; Received in revised form 28 August 2023; Accepted 30 August 2023
Available online 31 August 2023
0042-207X/© 2023 Elsevier Ltd. All rights reserved.
Z.-H. Shen et al. Vacuum 217 (2023) 112568

samples by Al doping and investigated their synaptic-related properties microstructure. Using X-ray Photoelectron Spectroscopy (XPS) (Escalab
such as long/short term plasticity. We constructed Au/STAO/LNO/Si 250Xi, Thermo Fisher, England) analyze the element’s chemical state
artificial synapses and realized the simulation of various synaptic [28,29]. The binding energy was calibrated by the work function
memory learning functions. In addition, to evaluate the possibility of method [30,31]. The spectra were shifted accordingly so that the C 1s
artificial synaptic devices in artificial neural networks, a convolutional peaks were aligned at (289.58 − ΦSA) eV, where ΦSA was obtained from
neural network (CNN) was constructed and used to recognize hand­ the literature for the work function of the sample with a value of 4.5 eV
written digit images, achieving an accuracy rate of 96.2%, demon­ [32]. The source gun type was monochromatic Al Kα. The electron
strating its feasibility for neuromorphic computation [2,22,23]. emission angle is 58◦ , and the sample analysis area size is 650 μm. The
instrument base pressure is 3 × 10− 7 mBar, and a charge neutralizer is
2. Experimental section used. Characterization of the sample cross-section morphology by
transmission electron microscopy (TEM) (Talos F200X G2, FEI, Amer­
The experiments were performed by the sol-gel method, which ica). Additionally, the electrical performances of the samples were
mainly involved the processes of precursor solution preparation, film evaluated using Keithley 2400 (Solon, OH, America) [33].
shaking by a homogenizer, baking by heating platform, and annealing
by annealing equipment [24–26]. 3. Results and discussion

2.1. Precursor solution preparation 3.1. Structure, morphology, and chemical states

LaNiO3 precursor solution: Nickel acetate Ni(OCOCH3)2⋅4H2O and Fig. 1 shows the surface morphology of the STAO and STO thin film.
lanthanum nitrate La(NO3)3⋅6H2O were adopted as raw materials, It can be seen from Fig. 1(a) that the STAO thin film is dense, smooth and
ethylene glycol methyl ether C3H8O2 and glacial acetic acid C2H4O2 as uniform with very high crystallinity and the grain size is about 70 nm.
solvents, and acetylacetone C5H8O2 as a stabilizer. Dissolve the desired The cross-sectional morphology (See Fig. S1 for details) of the thin film
raw materials in a 1:1 solvent. Stir for 1 h until the solute was completely is illustrated in Fig. 1(e), the interface of STAO/LNO heterostructure is
dissolved. clearly layered, the thickness of STAO is about 230 nm, and the thick­
SrTi0.9Al0.1O3-δ precursor solution: Tetrabutyl titanate ness of LNO is about 130 nm. By comparing Fig. 1(a) and (b), the grain
C16H36O4Ti, strontium acetate C2H3O2Sr, aluminum acetylacetonate size of the Al-doped STAO thin film is larger and more dense. The
C15H21AlO6 were adopted as raw materials, ethylene glycol methyl ether elemental composition of the facets of STAO was determined by energy
C3H8O2 and ice acetic acid C2H4O2 as solvents, and acetylacetonate dispersive spectroscopy (EDS), as shown in Fig. 1(c). It was found that
C5H8O2 as a stabilizer. Firstly, C2H3O2Sr was dissolved in C3H8O2 and the elements were uniformly distributed in the sample, which indirectly
C2H4O2 and stirred for 2 h at 60 ◦ C, C15H21AlO6 was dissolved in C2H4O2 indicated that the thin film samples were well grown. More details can
and stirred for 1 h at 50 ◦ C, C16H36O4Ti was dissolved in C5H8O2 and be found in Fig. S2. In addition, Fig. 1(d) shows the AFM image of the
stirred for 15 min at room temperature. Finally, the three solutions were STAO sample surface, which exhibits a smooth and uniform morphology
mixed together and stirred at room temperature for 35 min. with a root-mean-square roughness (Rq) of 3.98 nm. The AFM plot of the
SrTiO3 precursor solution: The preparation process is the same as STO sample is presented in Fig. 1(f), and its Rq is 2.73 nm. The result of
for SrTi0.9Al0.1O3-δ precursor solution but without the addition of Al. comparison shows that Al doping results in larger grains and higher Rq
All three precursor solutions mentioned above were at a concentra­ of the sample [34,35], which corresponds to SEM (Fig. 1(a) and (b)).
tion of 0.25 mol/L and left at room temperature for three days to ensure Fig. 1(g) depicts the 3D schematic of the Au/STAO/LNO/Si device.
no precipitation before being used to prepare the films [27]. To investigate the impact of Al doping on the chemical valence and
ratio of the sample elements, narrow spectra were scanned for Sr 3d, Ti
2.2. Thin film preparation 2p, O 1s and Al 2p of the samples. Fig. 2(a) illustrates the detection of
two discernible peaks of Sr 3d at 133.18 eV and 134.98 eV, signifying
Substrate cleaning: Experiments were performed using p-Si sub­ the presence of the Sr2+ state [36]. The Al spectrum is shown in Fig. 2
strates, which were cleaned with acetone, anhydrous ethanol, and iso­ (d), with a peak at 73.88 eV, indicating the Al3+ state [37]. As indicated
propanol in an ultrasonic bath for 35 min at room temperature. Each in Fig. 2(b) and (c), the sample has two main peaks near 463 eV and 458
cleaning in the ultrasonic bath was followed by several repeated rinses eV for Ti 2p3/2 and Ti 2p1/2, respectively, suggesting the dominance of
with deionized water to ensure a clean substrate. This is because the Ti4+ ions. In the STAO sample, additional shoulder-like features appear
cleanliness of the substrate is an important factor in the quality of the which are attributed to the Ti3+ state [38,39]. The decrease in Ti valence
thin film. state indicates the existence of oxygen vacancies. A brief defect reaction
Preparation of thin films: The substrate was placed on a Spin equation can be written to deepen the understanding of the Al ion
Coater, and an appropriate amount of precursor solution was added doping effect [40,41]:
dropwise onto the substrate by aspirating it with a syringe; the Spin
1
Coater speed was set to 1000 rpm for 15 s and 4000 rpm for 40 s. The wet VO2− ⟶VO.. + 2e′ + O2 (1)
2
film was placed on a baking platform at 100 ◦ C for 10 min and at 400 ◦ C
for 20 min in order to make the precursor solution in organic matter 2Ti4+ + 2e′⟶2Ti3+ (2)
decompose sufficiently to ensure the formation of nuclei within the film.
Repeat the above steps several times, depending on the required film Al3+ ′ 1
thickness. Finally, the films were annealed at 800 ◦ C for 20 min. SrTiO3 → Al SrTiO3 + VO.. + O2 (3)
2

2.3. Characterization and measurement where VO2− represents lattice oxygen and VO.. represents oxygen
vacancies.
The X-ray Diffraction (XRD) (D8 ADVANCE, Bruker, Germany) was The lowering of the Ti valence state indicates that the Al element in
used to analyze the crystal structures of the samples with an incidence the sample successfully replaced the Ti element and caused oxygen va­
angle of 0.5◦ , a scan step of 0.02◦ , a scan time of 30 min, and a scan angle cancies at the same time [39,41,42]. It is also evidenced by the obser­
range of 20◦ –80◦ . Scanning Electron Microscopy (SEM) (SUB8220, vation of O 1s spectra. Fig. 2(e) and (f) depict the O 1s spectrum
Hitachi, Japan) and Atomic Force Microscope (AFM) (Dimension featuring a low binding energy (LBE) peak and a high binding energy
FastScan, Bruker, America) were performed to characterize the sample

2
Z.-H. Shen et al. Vacuum 217 (2023) 112568

Fig. 1. (a) and (b) SEM pictures of the sample microstructure. (c) The STAO sample’s EDS spectrum results. (d) and (f) show the two-dimensional and three-
dimensional AFM images. (e) Cross-section of the STAO sample. (g) Schematic illustration of the Au/STAO/LNO/Si device.

Fig. 2. XPS spectra of the STAO thin films: (a) Sr 3d, (b) Ti 2p, (d) Al 2p, (e) O 1s. XPS spectra of the STO thin films: (c) Ti 2p, (f) O 1s.

(HBE) peak [43]. The former is attributed to lattice oxygen, whereas the (111), (200), (220), (222), and (311) were seen in the XRD pattern
latter may be caused by oxygen vacancies in the thin films [41,43]. A (Fig. 3), in agreement with the data reported (JCPDF Card No. 35-0734),
comparison of the O 1s spectra of STO and STAO thin films revealed that showing the standard cubic phase. It is noteworthy that no significant
Al doping induced a boost in the amount of oxygen vacancies (from shift in the peak position of STAO was observed, which is likely due to
23.59% to 37.91%), which could enhance the synaptic performance of the low content of doped elements.
the devices [44].
The phase structure of the sample was analyzed by XRD test. Peaks

3
Z.-H. Shen et al. Vacuum 217 (2023) 112568

Fig. 3. XRD pattern of the thin films.

3.2. Device electrical properties

The synaptic structure in the human brain neural network and the
artificial synaptic device structure are depicted in Fig. 4. The device is
not only structurally similar to the biological synapse but also func­
tionally similar. Both are three-layer structures, and the device
conductance corresponds to the synaptic weight; biologically, the syn­
aptic weight is varied by internal Na + Ca2+ migration [45], while the
conductance of the device depends on the migration of oxygen vacancies
[46,47].
The electrical properties of the device are evaluated through the Fig. 5. (a) The process of formation and breakage of conductive filaments. (b)
application of a voltage or a designed pulsed sequence to the top elec­ Single I-V measurement curve of device. (c) I–V curve plotted on a double
trode. As shown in Fig. 5(b), the device is capable of showing successive logarithmic scale.
and regular changes under voltage sweeps, which means that the
conductance state of the device can be dynamically tuned. This feature Fig. 5(a). By treating the samples with a focused ion beam, possible
holds significant application potential for simulating biological regions of CF were found as depicted in Fig. S3.
synapses. The device’s double-logarithm fitting analysis, as shown in Fig. 5(c),
Changes in synaptic weight in biology are caused by changes in in­ was performed for studying more about the carrier transport mecha­
ternal ion concentrations, while changes in the conductivity of the de­ nism. When the voltage is scanned from 0 V to 6 V, the device is at a high
vice are caused by oxygen vacancies [36,48–50]. Vacancies migrate and resistance state (HRS) and the slope is 1.12, which is linear, indicating
progressively form conductive filaments (CF) [51,52] under the influ­ that the ohmic transport mechanism is followed at this time, and as the
ence of voltage. Changes in voltage lead to the formation and breaking of voltage increases, the slope rises to 4.63, which follows space charge
filaments [53–55], resulting in changes in conductivity, as shown in limited conduction (SCLC) [56,57]. Subsequently, the device is at a low
resistance state (LRS) as the voltage is scanned positively from 6 V to 0
V, with a slope close to 1, suggesting that the ohmic conduction mech­
anism is followed. In addition, the 100 I-V cycle test of the device is
illustrated in Fig. S4, and the STAO device has better stability and
durability.

3.3. Synaptic property simulation

Synapses are the sites where functional connections between neu­


rons occur and are key sites for information transfer, with unique
nonlinear transport properties. The intensity of synaptic links de­
termines the information transfer efficiency, and the strength of such
connections can be varied with the training of input signals. Therefore,
the primary requirement for simulating biological synapses is that the
conductance of artificial synaptic devices can be continuously regulated
in a simulated manner. As shown in Fig. 6(a) and (b), the conductance of
the device gradually increases or decreases when the device is subjected
to successive sweeps of positive and negative voltages. Interestingly, the
Fig. 4. Biological synapses and artificial synapses similarity. variation of the device conductivity can also be well controlled in a

4
Z.-H. Shen et al. Vacuum 217 (2023) 112568

Fig. 6. The I-V curves of positive and negative voltage sweeps are displayed in (a) and (b). Current modulation under (c) positive and (d) negative consecutive
triangular wave.

different way, through a triangular wave. As illustrated in Fig. 6(c) and


(d), the device conductance gradually increases at positive pulses (+4 V)
and decreases at negative pulses (− 4 V). Fig. 6 reveals that the device
conductance can be modulated by changing the input signal, and this
nonlinear electrical property closely resembles the nonlinear transport
property of the neurosynapse.
Long-term potentiation (LTP) and long-term depression (LTD) are
crucial synaptic functions necessary to facilitate synaptic weight updates
for neuromorphic computing [58]. To evaluate the LTP and LTD be­
haviors, 55 pulse stimuli (±4 V) were consecutively applied to the de­
vice at a pulse width of 100 ms. As demonstrated in Fig. 7, we observe
the continuous increase (decrease) of conductance under the continuous
positive (negative) pulse stimulation, which effectively simulates the
LTP (LTD) behavior. In addition, the processes can be reproduced
cyclically, as shown in Fig. 7(b), and good potentiation and depression
were observed even after many repetitions, indicating the device’s
excellent stability. The device exhibits good symmetry and linearity as
shown in Fig. 7, indicating that it satisfies the demands of neural
network image recognition applications. Meanwhile, the LTP and LTD
characteristics of STO are shown in Fig. S5, and the comparison shows
that the LTP and LTD characteristics of STAO have better symmetry and
linearity [59], which will be beneficial to enhance the accuracy of the
device in neural network image recognition applications.
The human brain has a highly complex process for the formation and
storage of memory, which can be broadly classified into two categories:
short-term and long-term memory. Short-term memory is mainly
dominated by short-term plasticity [60]. Paired pulse facilitation (PPF),
a typical manifestation of short-term neural activity, is a kind of
short-term synaptic plasticity. In biological synapses, PPF diminishes Fig. 7. (a) LTP and LTD properties of the device. (b) Stability of the device.
with increasing stimulus interval time, which symbolizes that memory
diminishes with increasing learning interval. The ability of a device to device exhibits the PPF behavior described. A series of pulses (amplitude
exhibit short-term memory potential is predicated on the ability to 4 V, width 100 ms) are applied and the current is adjusted with time
exhibit this short-term plasticity. After several experiments, the results interval as shown in Fig. 8(a). By continuously adjusting the pulse in­
showed the amplitude of excitatory postsynaptic current (EPSC) could terval, ranging from 100 ms to 1200 ms, the PPF function was obtained.
be adjusted by pulses, and the greater the pulse interval, the lower the The currents after stimulation of the two pulses before and after are
amplitude of EPSC enhancement [45]. Fig. 8(b) demonstrates that the denoted as I1 and I2, respectively. The PPF gain can be expressed by the

5
Z.-H. Shen et al. Vacuum 217 (2023) 112568

Fig. 8. (a) Schematic illustration of the input and output signals of PPF. (b) Variation of PPF index with time interval. (c) The input pulse schematic of STDP. (d)
Synaptic weights fitted with time.

following formula: Where A0 and A1 are the conductance values before and after applying a
I2 − I1 pulse to the device, respectively. Δw represents the change in synaptic
PPF = × 100% (4)
I1 weight. The fitted curve of STDP is represented by the dotted line in
Fig. 8(d) and can be described by the following equations:
The PPF data obtained from the experiment and the corresponding
{
fitted PPF index curve are presented in Fig. 8(b). The equation used for C1 exp(− t/τ1 ), Δt > 0
fitting the PPF to the pulse interval is presented in equation [60]: STDP = (7)
C2 exp(− t/τ2 ), Δt < 0
PPF = A1 ∗ exp(− t/τ1 ) + A2 ∗ exp(− t/τ2 ) (5)
The results suggest that the Δw of the device decreases with
The t is time interval. τ1 and τ2 are 82 ms and 786 ms, respectively, increasing Δt, meaning the shorter the stimulus interval, the more sen­
which closely match the relaxation time of human brain synapses [61]. sitive the synaptic response. This is in accordance with the Hebbian
As shown in Fig. 8(b), the PPF growth rate increases from 101% to 127% synaptic learning rule, and the device exhibits STDP.
as the time interval decreases. This demonstrates the enhanced
short-term synaptic weight modulation when the signal interval of 3.4. Application: Identifying handwritten digits
synaptic transmission is short enough, a behavior that is very similar to
the role of neurotransmitter release in synapses. To test the aptitude of artificial synaptic devices for employment in
In brain neural activity, electrical stimulation can not only neuromorphic computing, a convolutional neural network (CNN) was
strengthen or weaken synapses, but also generate a range of functions constructed and used to recognize images in the MNIST dataset [64,65].
associated with space and time. One of the typical properties when As shown in Fig. 9(a), our CNN has three convolutional layers for feature
implementing synaptic learning functions is their ability to conform to map extraction, three pooling layers for redundant information reduc­
one of the important synaptic adaptation rules in Hebbian theory, Spike- tion, and one fully connected layer for image classification [66]. The
timing-dependent plasticity (STDP), and also necessary to implement network consists of 784 inputs neurons, 256 hidden neurons and 10
when simulating brain functions in neuromorphic computation [62,63]. outputs neurons for distinguishing numbers between 0 and 9 [67–69].
Hence, accurate simulation of the STDP learning rule is an important The handwritten digital images in the MNIST dataset are shown in Fig. 9
requirement for the development of brain-like synaptic devices. The (b), which includes 60,000 images, 48,000 for training, and 12,000 for
design of the pre/post-synaptic membrane pulses is in Fig. 8(c). testing. We mapped the measured device conductance to the numerical
Fig. 8(d) depicts the simulated results of STDP, where a pair of pre/ weights of the neural network and obtained the recognition rate after
post-synaptic spikes with amplitudes of +/- 4 V, respectively, were 100 epochs. Fig. 9(c) shows the variation of the image recognition ac­
applied to the device. The change in synaptic weight (Δw) was measured curacy of the simulated network with training time. We compared the
as a function of the relative time (Δt) between the two spikes. The recognition rates of STAO and STO, and it is clear that the recognition
definition of Δw, given by accuracy obtained based on the synaptic data of Al-doped samples is
higher, indicating that Al doping facilitates the artificial synaptic de­
A1 − A0
Δw = × 100% (6) vices to exhibit better performance, which is very meaningful.
A0

6
Z.-H. Shen et al. Vacuum 217 (2023) 112568

Fig. 9. (a) Convolutional neural network framework diagram. (b) Handwritten digital image of the MNIST dataset. (c) Recognition accuracy based on STAO and
STO samples.

4. Conclusions Appendix A. Supplementary data

In conclusion, the Au/STAO/LNO/Si memristor prepared by the sol- Supplementary data to this article can be found online at https://doi.
gel method is capable of achieving functions of biological synapses like org/10.1016/j.vacuum.2023.112568.
STP, LTP, PPF, and STDP. We constructed a CNN to train the MNIST digit
recognition task, and the recognition accuracy reached 96.2% after 100 References
epochs. We compared STAO with STO and found that the Al-doped
device has better stability, higher linearity, and higher accuracy in ap­ [1] M. Di Ventra, Y.V. Pershin, The parallel approach, Nat. Phys. 9 (2013) 200–202,
https://doi.org/10.1038/nphys2566.
plications in image recognition. These results indicate that our devices [2] S.H. Jo, T. Chang, I. Ebong, B.B. Bhadviya, P. Mazumder, W. Lu, Nanoscale
are beneficial for the future realization of neuromorphic computing. memristor device as synapse in neuromorphic systems, Nano Lett. 10 (2010)
1297–1301, https://doi.org/10.1021/nl904092h.
[3] C. Eliasmith, T.C. Stewart, X. Choo, T. Bekolay, T. DeWolf, Y. Tang, D. Rasmussen,
CRediT authorship contribution statement A large-scale model of the functioning brain, Science 338 (2012) 1202–1205,
https://doi.org/10.1126/science.1225266.
Zhi-Hao Shen: Writing – original draft, Methodology, Data curation, [4] Y. van de Burgt, E. Lubberman, E.J. Fuller, S.T. Keene, G.C. Faria, S. Agarwal, M.
J. Marinella, A. Alec Talin, A. Salleo, A non-volatile organic electrochemical device
Conceptualization. Wen-Hua Li: Supervision. Xin-Gui Tang: Valida­ as a low-voltage artificial synapse for neuromorphic computing, Nat. Mater. 16
tion, Supervision. Hao Chen: Investigation. Jia Hu: Investigation. Kai- (2017) 414–418, https://doi.org/10.1038/nmat4856.
Yuan Wang: Writing – review & editing. Ke Meng: Software. Yan-Ping [5] R. Rahimian, M.M. Waldrop, Computer chips inspired by human neurons can do
more with less power, Nature 503 (2013) 22–24, https://doi.org/10.1038/
Jiang: Supervision, Investigation. Xiao-Bin Guo: Supervision,
503022a.
Conceptualization. [6] D.B. Strukov, G.S. Snider, D.R. Stewart, R.S. Williams, The missing memristor
found, Nature 453 (2008) 80–83, https://doi.org/10.1038/nature06932.
Declaration of competing interest [7] J. Borghetti, G.S. Snider, P.J. Kuekes, J.J. Yang, D.R. Stewart, R.S. Williams,
‘Memristive’ switches enable ‘stateful’ logic operations via material implication,
Nature 464 (2010) 873–876, https://doi.org/10.1038/nature08940.
The authors declare that they have no known competing financial [8] C. Diorio, P. Hasler, A. Minch, C.A. Mead, A single-transistor silicon synapse, IEEE
interests or personal relationships that could have appeared to influence Trans. Electron. Dev. 43 (1996) 1972–1980, https://doi.org/10.1109/16.543035.
[9] S. Thakoor, A. Moopenn, T. Daud, A.P. Thakoor, Solid-state thin-film memistor for
the work reported in this paper. electronic neural networks, J. Appl. Phys. 67 (1990) 3132–3135, https://doi.org/
10.1063/1.345390.
Data availability [10] Q. Xia, J.J. Yang, Memristive crossbar arrays for brain-inspired computing, Nat.
Mater. 18 (2019) 309–323, https://doi.org/10.1038/s41563-019-0291-x.
[11] W. Wang, G. Pedretti, V. Milo, R. Carboni, A. Calderoni, N. Ramaswamy, A.
Data will be made available on request. S. Spinelli, D. Ielmini, Learning of spatiotemporal patterns in a spiking neural
network with resistive switching synapses, Sci. Adv. 4 (2018), eaat4752, https://
doi.org/10.1126/sciadv.aat4752.
Acknowledgments [12] B. Sun, T. Guo, G. Zhou, S. Ranjan, Y. Jiao, L. Wei, Y.N. Zhou, Y.A. Wu, Synaptic
devices based neuromorphic computing applications in artificial intelligence,
This work was supported by the National Natural Science Foundation Materials Today Physics 18 (2021), 100393, https://doi.org/10.1016/j.
mtphys.2021.100393.
of China (Grant Nos. 12172093), the Guangdong Basic and Applied
[13] Z. Wang, H. Wu, G.W. Burr, C.S. Hwang, K.L. Wang, Q. Xia, J.J. Yang, Resistive
Basic Research Foundation (Grant Nos. 2021A1515012607), the Open switching materials for information processing, Nat. Rev. Mater. 5 (2020)
Project Program of Guangdong Provincial Key Laboratory of Electronic 173–195, https://doi.org/10.1038/s41578-019-0159-3.
Functional Materials and Devices, Huizhou University, China (Grant No. [14] L. Zhang, Z. Tang, J. Fang, X. Jiang, Y.-P. Jiang, Q.-J. Sun, J.-M. Fan, X.-G. Tang,
G. Zhong, Synaptic and resistive switching behaviors in NiO/Cu2O heterojunction
EFMDN2021003Z). We would like to thank Mr. Kai Yan, Mr. Yanxue Wu memristor for bioinspired neuromorphic computing, Appl. Surf. Sci. 606 (2022),
at Analysis and Test Center of Guangdong University of technology for 154718, https://doi.org/10.1016/j.apsusc.2022.154718.
their assistance with XPS and SEM analysis. We thank Figdraw (www.fi [15] J.J. Yang, D.B. Strukov, D.R. Stewart, Memristive devices for computing, Nat.
Nanotechnol. 8 (2013) 13–24, https://doi.org/10.1038/nnano.2012.240.
gdraw.com) for expert assistance in the pattern drawing. [16] B. Cui, Z. Fan, W. Li, Y. Chen, S. Dong, Z. Tan, S. Cheng, B. Tian, R. Tao, G. Tian,
D. Chen, Z. Hou, M. Qin, M. Zeng, X. Lu, G. Zhou, X. Gao, J.-M. Liu, Ferroelectric

7
Z.-H. Shen et al. Vacuum 217 (2023) 112568

photosensor network: an advanced hardware solution to real-time machine vision, on its electronic and magnetic properties, J. Appl. Phys. 122 (2017), 085310,
Nat. Commun. 13 (2022) 1707, https://doi.org/10.1038/s41467-022-29364-8. https://doi.org/10.1063/1.5000133.
[17] W. Li, Z. Fan, Q. Huang, J. Rao, B. Cui, Z. Chen, Z. Lin, X. Yan, G. Tian, R. Tao, [39] Y. Zhang, J. Xu, Y. Yu, W. Zheng, Z. Xu, L. Lu, Z. Wang, C. Lu, Z. Wen, Changeable
D. Chen, M. Qin, M. Zeng, X. Lu, G. Zhou, X. Gao, J.-M. Liu, Polarization- electroresistance in Pt/Pb(Zr,Ti)O3/(La,Sr)MnO3 tunnel junctions and memristive
dominated internal timing mechanism in a ferroelectric second-order memristor, properties for synaptic plasticity emulation, Appl. Phys. Lett. 120 (2022), 203501,
Phys. Rev. Appl. 19 (2023), 014054, https://doi.org/10.1103/ https://doi.org/10.1063/5.0093112.
PhysRevApplied.19.014054. [40] J. Li, X. Tang, Q. Liu, Y. Jiang, Z. Tang, Resistive switching and optical properties
[18] R.A. De Souza, V. Metlenko, D. Park, T.E. Weirich, Behavior of oxygen vacancies in of strontium ferrate titanate thin film prepared via chemical solution deposition,
single-crystal SrTiO3: equilibrium distribution and diffusion kinetics, Phys. Rev. B J Adv Ceram 10 (2021) 1001–1010, https://doi.org/10.1007/s40145-021-0483-0.
85 (2012), 174109, https://doi.org/10.1103/PhysRevB.85.174109. [41] X. Shi, J. Yang, X. Wen, F. Tian, C. Li, Oxygen vacancy enhanced biomimetic
[19] Y.-Y. Pai, A. Tylan-Tyler, P. Irvin, J. Levy, Physics of SrTiO3-based heterostructures superoxide dismutase activity of CeO2-Gd nanozymes, J. Rare Earths 39 (2021)
and nanostructures: a review, Rep. Prog. Phys. 81 (2018), 036503, https://doi.org/ 1108–1116, https://doi.org/10.1016/j.jre.2020.06.019.
10.1088/1361-6633/aa892d. [42] H.-C. Zhou, Y.-P. Jiang, X.-G. Tang, Q.-X. Liu, W.-H. Li, Z.-H. Tang, Excellent
[20] L. Gao, Q. Ren, J. Sun, S.-T. Han, Y. Zhou, Memristor modeling: challenges in bipolar resistive switching characteristics of Bi4Ti3O12 thin films prepared via sol-
theories, simulations, and device variability, J. Mater. Chem. C 9 (2021) gel process, Nanomaterials 11 (2021) 2705, https://doi.org/10.3390/
16859–16884, https://doi.org/10.1039/D1TC04201G. nano11102705.
[21] M. Tang, L. Dai, M. Cheng, Y. Zhang, Y. Wang, X. Zhong, J. Wang, F. An, M. Ma, [43] L.-Q. Wang, W.-H. Li, X.-G. Tang, X.-B. Guo, Q.-X. Liu, Y.-P. Jiang, Z.-H. Tang,
M. Huang, C. Li, J. Li, G. Zhong, High-throughput screening thickness-dependent Bipolar resistive switching characteristics of PbZrO3/LaNiO3 heterostructure thin
resistive switching in SrTiO3 thin films for robust electronic synapse, Adv. Funct. films prepared by a sol–gel process, Ceram. Int. 47 (2021) 5617–5623, https://doi.
Mater. 33 (2023), 2213874, https://doi.org/10.1002/adfm.202213874. org/10.1016/j.ceramint.2020.10.146.
[22] R. Ai, W. Luo, X. Liu, Rectifier characteristics and analog bipolar resistive [44] T.R. Silva, V.D. Silva, L.S. Ferreira, A.J.M. Araújo, M.A. Morales, T.A. Simões, C.
switching behavior in CoO-based memristor for artificial synapse application, A. Paskocimas, D.A. Macedo, Role of oxygen vacancies on the energy storage
Vacuum 188 (2021), 110190, https://doi.org/10.1016/j.vacuum.2021.110190. performance of battery-type NiO electrodes, Ceram. Int. 46 (2020) 9233–9239,
[23] B. Patel, R.K. Pati, I. Mukhopadhyay, A. Ray, Effect of vacuum and sulphur https://doi.org/10.1016/j.ceramint.2019.12.176.
annealing on the structural properties of spray deposited Cu2SnS3 thin films, [45] Z. Wang, S. Joshi, S.E. Savel’ev, H. Jiang, R. Midya, P. Lin, M. Hu, N. Ge, J.
Vacuum 158 (2018) 263–270, https://doi.org/10.1016/j.vacuum.2018.10.015. P. Strachan, Z. Li, Q. Wu, M. Barnell, G.-L. Li, H.L. Xin, R.S. Williams, Q. Xia, J.
[24] Z.-H. Shen, W.-H. Li, X.-G. Tang, J. Hu, K.-Y. Wang, Y.-P. Jiang, X.-B. Guo, An J. Yang, Memristors with diffusive dynamics as synaptic emulators for
artificial synapse based on Sr(Ti, Co)O3 films, Mater. Today Commun. 33 (2022), neuromorphic computing, Nat. Mater. 16 (2017) 101–108, https://doi.org/
104754, https://doi.org/10.1016/j.mtcomm.2022.104754. 10.1038/nmat4756.
[25] M. Surmeneva, K. Prosolov, I. Glukhov, I. Grubova, V. Botvin, R. Surmenev, [46] Z.Q. Wang, H.Y. Xu, X.H. Li, H. Yu, Y.C. Liu, X.J. Zhu, Synaptic learning and
Y. Sharkeev, Effects of annealing in vacuum on the microstructure of silicon- memory functions achieved using oxygen ion migration/diffusion in an amorphous
containing calcium phosphate films deposited on a ZrNb alloy by radio-frequency InGaZnO memristor, Adv. Funct. Mater. 22 (2012) 2759–2765, https://doi.org/
magnetron sputtering, Vacuum 212 (2023), 112028, https://doi.org/10.1016/j. 10.1002/adfm.201103148.
vacuum.2023.112028. [47] Y. She, F. Wang, X. Zhao, Z. Zhang, C. Li, H. Pan, K. Hu, Z. Song, K. Zhang, Oxygen
[26] W. Sailuam, T. Saisopa, I. Fongkaew, L. Ngamwongwan, T. Eknapakul, vacancy-dependent synaptic dynamic behavior of TiOx-based transparent
K. Seawsakul, M. Horprathum, P. Amonpattaratkit, N. Chanlek, memristor, IEEE Trans. Electron. Dev. 68 (2021) 1950–1955, https://doi.org/
P. Songsiriritthigul, S. Limpijumnong, R. Yimnirun, A. Jiamprasertboon, 10.1109/TED.2021.56333.
A. Bootchanont, Correlation of conductivity enhancement and Al-site defects in [48] X. Pan, Y. Shuai, C. Wu, W. Luo, X. Sun, Y. Yuan, S. Zhou, X. Ou, W. Zhang,
nanocolumnar ZnO films under vacuum annealing by experimental and Resistive switching behavior in single crystal SrTiO3 annealed by laser, Appl. Surf.
calculations, Appl. Surf. Sci. 613 (2023), 155985, https://doi.org/10.1016/j. Sci. 389 (2016) 1104–1107, https://doi.org/10.1016/j.apsusc.2016.08.013.
apsusc.2022.155985. [49] R. Su, R. Xiao, C. Shen, D. Song, J. Chen, B. Zhou, W. Cheng, Y. Li, X. Wang,
[27] W.-M. Zhong, Q.-X. Liu, Y.-P. Jiang, M.-L. Deng, W.-P. Li, X.-G. Tang, Ultra-high X. Miao, Oxygen ion migration induced polarity switchable SrFeOx memristor for
dielectric tuning performance and double-set resistive switching effect achieved on high-precision handwriting recognition, Appl. Surf. Sci. 617 (2023), 156620,
the Bi2NiMnO6 thin film prepared by sol–gel method, J. Colloid Interface Sci. 606 https://doi.org/10.1016/j.apsusc.2023.156620.
(2022) 913–919, https://doi.org/10.1016/j.jcis.2021.08.109. [50] J. Joshua Yang, F. Miao, M.D. Pickett, D.A.A. Ohlberg, D.R. Stewart, C.N. Lau, R.
[28] G. Greczynski, L. Hultman, A step-by-step guide to perform x-ray photoelectron S. Williams, The mechanism of electroforming of metal oxide memristive switches,
spectroscopy, J. Appl. Phys. 132 (2022), 011101, https://doi.org/10.1063/ Nanotechnology 20 (2009), 215201, https://doi.org/10.1088/0957-4484/20/21/
5.0086359. 215201.
[29] G. Greczynski, L. Hultman, C 1s peak of adventitious carbon aligns to the vacuum [51] W. Sun, B. Gao, M. Chi, Q. Xia, J.J. Yang, H. Qian, H. Wu, Understanding
level: dire consequences for material’s bonding assignment by photoelectron memristive switching via in situ characterization and device modeling, Nat.
spectroscopy, ChemPhysChem 18 (2017) 1507–1512, https://doi.org/10.1002/ Commun. 10 (2019) 3453, https://doi.org/10.1038/s41467-019-11411-6.
cphc.201700126. [52] P.S. Subin, A. Antony, K.J. Saji, M.K.E. Jayaraj, Observation of magnéli phase
[30] G. Greczynski, L. Hultman, Referencing to adventitious carbon in X-ray filament formation in MoOx artificial synapse, Adv Elect Materials 8 (2022),
photoelectron spectroscopy: can differential charging explain C 1s peak shifts? 2200729, https://doi.org/10.1002/aelm.202200729.
Appl. Surf. Sci. 606 (2022), 154855 https://doi.org/10.1016/j. [53] H. Sun, Q. Liu, C. Li, S. Long, H. Lv, C. Bi, Z. Huo, L. Li, M. Liu, Direct observation
apsusc.2022.154855. of conversion between threshold switching and memory switching induced by
[31] G. Greczynski, L. Hultman, Reliable determination of chemical state in x-ray conductive filament morphology, Adv. Funct. Mater. 24 (2014) 5679–5686,
photoelectron spectroscopy based on sample-work-function referencing to https://doi.org/10.1002/adfm.2014014.
adventitious carbon: resolving the myth of apparent constant binding energy of the [54] J. Xu, H. Wang, Y. Zhu, Y. Liu, Z. Zou, G. Li, R. Xiong, Tunable digital-to-analog
C 1s peak, Appl. Surf. Sci. 451 (2018) 99–103, https://doi.org/10.1016/j. switching in Nb2O5-based resistance switching devices by oxygen vacancy
apsusc.2018.04.226. engineering, Appl. Surf. Sci. 579 (2022), 152114, https://doi.org/10.1016/j.
[32] H. Zhang, D. Zhang, X. Lu, C. Liu, G. Zhou, X. Ma, L. Wang, P. Jiang, Q.-K. Xue, apsusc.2021.152114.
X. Bao, Origin of charge transfer and enhanced electron–phonon coupling in single [55] P.S. Subin, P.S. Midhun, A. Antony, K.J. Saji, M.K. Jayaraj, Optoelectronic synaptic
unit-cell FeSe films on SrTiO3, Nat. Commun. 8 (2017) 214, https://doi.org/ plasticity mimicked in ZnO-based artificial synapse for neuromorphic image
10.1038/s41467-017-00281-5. sensing application, Mater. Today Commun. 33 (2022), 104232, https://doi.org/
[33] J. Hu, W. Li, X. Tang, Z. Shen, K. Wang, Y. Zhang, S. Zhang, Y. Jiang, X. Guo, 10.1016/j.mtcomm.2022.104232.
Enhancement of energy storage density and efficiency of PbHfO3 doped with La [56] S.-R. Zhang, L. Zhou, J.-Y. Mao, Y. Ren, J.-Q. Yang, G.-H. Yang, X. Zhu, S.-T. Han,
antiferroelectric thin films, ACS Appl. Energy Mater. 6 (2023) 120–126, https:// V.A.L. Roy, Y. Zhou, Artificial synapse emulated by charge trapping-based resistive
doi.org/10.1021/acsaem.2c02584. switching device, Adv. Mater. Technol. 4 (2019), 1800342, https://doi.org/
[34] H. Xu, H. Hao, Y. Xie, X. Wang, J. Geng, M. Cao, Z. Yao, H. Liu, Microcrystalline 10.1002/admt.201800342.
structure modulation and energy storage properties of BaZr0.25Ti0.75O3 thin films, [57] Y. Yang, B. Sun, G. Zhou, C. Ke, J. Zhang, Y. Zhou, S. Mao, J. Qin, Y. Zhao,
J. Alloys Compd. 907 (2022), 164236, https://doi.org/10.1016/j. Improved resistive switching performance and in-depth mechanism analysis in Mn-
jallcom.2022.164236. doped SrTiO3-based RRAM, Mater. Today Commun. 35 (2023), 105512, https://
[35] J. Liu, Y. Wang, X. Zhai, Y. Xue, L. Hao, H. Zhu, C. Liu, H. Cheng, J. Ouyang, doi.org/10.1016/j.mtcomm.2023.105512.
Energy storage properties of sol–gel-processed SrTiO3 films, Materials 16 (2022) [58] T. Ohno, T. Hasegawa, T. Tsuruoka, K. Terabe, J.K. Gimzewski, M. Aono, Short-
31, https://doi.org/10.3390/ma16010031. term plasticity and long-term potentiation mimicked in single inorganic synapses,
[36] K.-H. Son, H.-S. Lee, Grain boundary effect on the resistive switching Nat. Mater. 10 (2011) 591–595, https://doi.org/10.1038/nmat3054.
characteristics of SrTi1-xFexO3 directly patterned via photochemical organic-metal [59] G. Zhong, M. Zi, C. Ren, Q. Xiao, M. Tang, L. Wei, F. An, S. Xie, J. Wang, X. Zhong,
deposition, Appl. Surf. Sci. 575 (2022), 151754, https://doi.org/10.1016/j. M. Huang, J. Li, Flexible electronic synapse enabled by ferroelectric field effect
apsusc.2021.151754. transistor for robust neuromorphic computing, Appl. Phys. Lett. 117 (2020),
[37] X.-L. Lu, X.-B. Guo, F.-C. Su, W.-H. Qiu, Z. Su, J. Li, W.-H. Li, Y.-P. Jiang, Z.- 092903, https://doi.org/10.1063/5.0013638.
H. Tang, X.-G. Tang, Photoelectric characteristics of Al-doped ZnO/p-Si diode [60] R.S. Zucker, W.G. Regehr, Short-term synaptic plasticity, Annu. Rev. Physiol. 64
prepared by radio frequency magnetron sputtering, J. Phys. D Appl. Phys. 55 (2002) 355–405, https://doi.org/10.1146/annurev.physiol.64.092501.114547.
(2022), 485101, https://doi.org/10.1088/1361-6463/ac95a1. [61] X. Yan, Q. Zhao, A.P. Chen, J. Zhao, Z. Zhou, J. Wang, H. Wang, L. Zhang, X. Li,
[38] G. Panchal, D.K. Shukla, R.J. Choudhary, V.R. Reddy, D.M. Phase, The effect of Z. Xiao, K. Wang, C. Qin, G. Wang, Y. Pei, H. Li, D. Ren, J. Chen, Q. Liu, Vacancy-
oxygen stoichiometry at the interface of epitaxial BaTiO3/La0.7Sr0.3MnO3 bilayers induced synaptic behavior in 2D WS2 nanosheet–based memristor for low-power

8
Z.-H. Shen et al. Vacuum 217 (2023) 112568

neuromorphic computing, Small 15 (2019), 1901423, https://doi.org/10.1002/ [66] N. Duan, Y. Li, H.-C. Chiang, J. Chen, W.-Q. Pan, Y.-X. Zhou, Y.-C. Chien, Y.-H. He,
smll.201901423. K.-H. Xue, G. Liu, T.-C. Chang, X.-S. Miao, An electro-photo-sensitive synaptic
[62] G. Bi, M. Poo, Synaptic modifications in cultured hippocampal neurons: transistor for edge neuromorphic visual systems, Nanoscale 11 (2019)
dependence on spike timing, synaptic strength, and postsynaptic cell type, 17590–17599, https://doi.org/10.1039/C9NR04195H.
J. Neurosci. 18 (1998) 10464–10472, https://doi.org/10.1523/JNEUROSCI.18-24- [67] Y. Chang, H. Cong, R. Zhou, W. Zhang, Y. Qin, X. Liu, F. Wang, Enhanced artificial
10464.1998. synaptic properties enabled by arrays of electrolyte-gated electrospun InZnO
[63] Y. Li, Y. Zhong, J. Zhang, L. Xu, Q. Wang, H. Sun, H. Tong, X. Cheng, X. Miao, nanowires, ACS Appl. Electron. Mater. 4 (2022) 2570–2579, https://doi.org/
Activity-dependent synaptic plasticity of a chalcogenide electronic synapse for 10.1021/acsaelm.2c00326.
neuromorphic systems, Sci. Rep. 4 (2015) 4906, https://doi.org/10.1038/ [68] J.-H. Ryu, S. Kim, Artificial synaptic characteristics of TiO2/HfO2 memristor with
srep04906. self-rectifying switching for brain-inspired computing, Chaos, Solitons & Fractals
[64] H. Bian, Y.Y. Goh, Y. Liu, H. Ling, L. Xie, X. Liu, Stimuli-responsive memristive 140 (2020), 110236, https://doi.org/10.1016/j.chaos.2020.110236.
materials for artificial synapses and neuromorphic computing, Adv. Mater. 33 [69] K.C. Kwon, Y. Zhang, L. Wang, W. Yu, X. Wang, I.-H. Park, H.S. Choi, T. Ma, Z. Zhu,
(2021), 2006469, https://doi.org/10.1002/adma.202006469. B. Tian, C. Su, K.P. Loh, In-plane ferroelectric tin monosulfide and its application in
[65] M.E. Pam, S. Li, T. Su, Y. Chien, Y. Li, Y.S. Ang, K. Ang, Interface-modulated a ferroelectric analog synaptic device, ACS Nano 14 (2020) 7628–7638, https://
resistive switching in Mo-irradiated ReS2 for neuromorphic computing, Adv. doi.org/10.1021/acsnano.0c03869.
Mater. 34 (2022), 2202722, https://doi.org/10.1002/adma.202202722.

You might also like