A ferroelectric neural network unit and a preparation method thereof

Abstract

The application discloses a kind of all ferroelectric neural-paradigm units and preparation methods thereof.A kind of all ferroelectric neural-paradigm unit, the all ferroelectric neural-paradigm unit structure includes the upper electrode array, ferroelectric layer and lower electrode array sequentially arranged from top to bottom;The single electrode shape of the upper electrode array and the lower electrode array is rectangle, the width of single electrode is 1nm~10 μm respectively, the spacing between adjacent electrodes is 10nm~50 μm;The width ratio of the single electrode of the upper electrode array and the single electrode of the lower electrode array is 0.001~1000;The upper electrode array and the lower electrode array electrode staggered arrangement, the angle range of staggered angle is 0~180°.The application can efficiently simulate neuron firing, synaptic plasticity and other behaviors by fine changing electric field loading conditions to regulate polarization, shielding, space charge dynamics in ferroelectric functional layer, can also realize the simulation of neuron and synapse simultaneously, to build all ferroelectric brain network.

Description

Technical Field

[0001] This invention relates to the field of neuromorphic devices, specifically to an all-ferroelectric neuromorphic unit and its fabrication method. Background Technology

[0002] Silicon-based microelectronic devices with von Neumann architectures, due to their high energy consumption, are no longer able to meet the demands of artificial intelligence development. Neuromorphic devices capable of simulating the efficient computation of the biological brain are a cutting-edge research hotspot worldwide. The development of new productive forces urgently requires highly integrated, low-power neuromorphic units capable of simulating the behavior of neurons and synapses. In recent years, considerable progress has been made in neuromorphic devices that simulate plastic behavior based on properties such as memristors. Neural networks developed based on this technology can achieve complex functions such as handwriting recognition or supervised learning, and various materials with non-volatile conductivity modulation states have been explored.

[0003] Neuromorphic devices that simulate synaptic plasticity based on ferroelectric properties have attracted widespread attention. This is because the polarization state of ferroelectric materials can be directly manipulated by an electric field (rather than a current), offering inherent advantages in low power consumption, non-volatility, and fatigue resistance, which aligns with the requirements of neuromorphic computing. It is well known that the physical basis for realizing the functional characteristics of ferroelectric synaptic devices is the evolution of ferroelectric domains driven by an external field. However, domain evolution at the micro- and nano-scale is a complex, nonlinear process involving many-body interactions, making it difficult to improve the integration density of ferroelectric synaptic devices. More importantly, current ferroelectric domain-based synaptic devices typically require combination with other neuronal devices to achieve complete neuromorphic behavior and brain-like neural networks, which significantly hinders the development of highly integrated, low-power brain-like neural network devices.

[0004] Therefore, how to realize highly integrated, low-power neuromorphic units based on all-ferroelectric microstructures has become a key scientific problem that researchers in this field urgently need to solve. Summary of the Invention

[0005] To address the aforementioned technical problems, this invention provides a fully ferroelectric neuromorphic unit and its fabrication method. Based on the coexistence and co-evolutionary ecosystem of ferroelectric domains and domain walls, and the similarity to the connection and operational modes between neurons and synapses, this invention constructs a sandwich structure containing upper and lower electrode arrays. By utilizing pulse sequences to finely control the physical properties of the domains and domain walls in the ferroelectric functional layer, it can simultaneously simulate the plasticity behavior of neurons and synapses, thereby realizing a highly integrated, low-power fully ferroelectric neural network.

[0006] To achieve the above objectives, the technical solution adopted by the present invention is as follows:

[0007] The first aspect of the present invention provides a fully ferroelectric neuromorphic unit, the structure of which includes an upper electrode array, a ferroelectric layer and a lower electrode array arranged sequentially from top to bottom;

[0008] The upper electrode array and the lower electrode array each have a rectangular shape, with a width of 1 nm to 10 μm and a spacing of 10 nm to 50 μm between adjacent electrodes. The width ratio of a single electrode in the upper electrode array to that in the lower electrode array is 0.001 to 1000.

[0009] The upper electrode array and the lower electrode array are arranged in a staggered manner, with the staggered angle ranging from 0 to 180°.

[0010] Preferably, the ferroelectric layer is made of PbZr. x Ti 1-x One or more of O3, PbTiO3, BaTiO3, and BiFeO3; the PbZr x Ti 1-x In formula O3, 0 < x < 1; more preferably, the ferroelectric layer is made of PbZr. x Ti 1-x One of O3 and PbTiO3, where 0 < x < 1.

[0011] Preferably, the thickness of the ferroelectric layer is 0.001 to 1 μm; more preferably, the thickness of the ferroelectric layer is 0.01 to 0.3 μm.

[0012] Preferably, the materials of the upper electrode array and the lower electrode array are selected from one or more of Ag, Au, Pt, W, ITO, ZnO, and SrRuO3; more preferably, the materials of the upper electrode array and the lower electrode array are selected from one of Au, Pt, W, and SrRuO3.

[0013] Preferably, the thicknesses of the upper electrode array and the lower electrode array are 1 to 500 nm, respectively; more preferably, the thicknesses of the upper electrode array and the lower electrode array are 1 to 50 nm, respectively.

[0014] Preferably, the width of each individual electrode in the upper electrode array is 5–100 nm, and the spacing between adjacent electrodes is 0.1–5 μm; the width of each individual electrode in the lower electrode array is 200–2000 nm, and the spacing between adjacent electrodes is 1–5 μm. The individual electrodes in both the upper and lower electrode arrays have the same width.

[0015] Preferably, the lower electrode array and the upper electrode array have a projected overlap region; the width ratio of a single electrode in the upper electrode array to a single electrode in the lower electrode array is 0.01 to 100, excluding the value 1, meaning the upper and lower electrodes cannot be of equal width; this differentiated width of the upper and lower electrodes is an important prerequisite for generating and controlling neuromorphic behaviors such as domain wall tilting. This is because if the upper and lower electrodes are of equal width, they tend to form vertical domains, unlike the attached domains... Figure 2 The domain walls are generated in a sloping manner, as shown in the example of c. Therefore, the differential electrodes at the top and bottom can effectively simulate neuromimetic behavior.

[0016] Preferably, the upper electrode array and the lower electrode array are staggered, with the stagger angle ranging from 80 to 100°.

[0017] This invention employs a unique upper and lower electrode array arrangement, which has the following beneficial effects: ① By setting the electrode width to conform to the size range of the electric domains, and setting the width ratio of the upper and lower electrodes, it is possible to effectively simulate the generation of neurons, the plastic behavior of synapses (based on the controllable tilting of domain walls), and the neuromorphic behavior (based on the co-evolution of domains / domain walls) based on ferroelectric microstructures (domains / domain walls); ② It is possible to perform addressing operations through the numbering of the upper and lower electrodes, thereby constructing a fully ferroelectric brain-like network.

[0018] A second aspect of the present invention provides a method for preparing the aforementioned all-ferroelectric neuromorphic unit, comprising the following steps:

[0019] (1) A sacrificial transition layer and a ferroelectric layer are sequentially deposited on the upper surface of the substrate;

[0020] (2) Deposit an upper electrode array on the upper surface of the ferroelectric layer;

[0021] (3) Etch the sacrificial transition layer to obtain a free radical ferroelectric thin film containing an electrode array;

[0022] (4) A lower electrode array is deposited on the lower surface of the free radical ferroelectric thin film to obtain a fully ferroelectric neuromorphic unit.

[0023] Preferably, the substrate is (001) oriented SrTiO3.

[0024] Preferably, in step (1), the sacrificial transition layer is SrRuO3 with a thickness of 1 to 50 nm; more preferably, the thickness of the sacrificial transition layer is 5 to 30 nm.

[0025] Preferably, in step (1), the deposition method used is pulsed laser deposition, and the conditions are: the sample deposition temperature is 550–800℃, the deposition oxygen pressure is 50–300 mtorr, and the laser energy density is 100–1500 mJ·cm⁻¹. -2The laser pulse frequency is 5–10 Hz; more preferably, the pulsed laser deposition conditions are: a sample deposition temperature of 590–690 °C, a deposition oxygen pressure of 80–150 mtorr, and a laser energy density of 150–500 mJ·cm⁻¹. -2 The laser pulse frequency is 8-10 Hz.

[0026] In the method for preparing the ferroelectric neuromorphic unit of the present invention, the electrode arrays in steps (2) and (4) can be prepared by industry-standard methods such as electron beam direct writing, magnetron sputtering, and electron beam evaporation. The present invention does not limit this.

[0027] A third aspect of the present invention provides a method for simulating neural behavior, employing the aforementioned all-ferroelectric neuromorphic unit, comprising the following methods:

[0028] (1) For the upper electrode array (X) a Input pulse electrical signal to monitor the lower electrode array (Y) b The electrical response of a neuron is used to simulate neuronal behavior, where the pulsed electrical signal has one or more of the following shapes: sine wave, square wave, step wave, triangular wave, and sawtooth wave, with an amplitude of 0.1–1.5V. c V c The coercive voltage of the ferroelectric layer has a frequency of 0.1–1 GHz and a pulse count of 1–10. 10 indivual;

[0029] (2) For the upper electrode array (X) a Input pulse electrical signal to monitor the lower electrode array (Y) b The electrical response of the pulsed electrical signal is used to simulate synaptic plasticity behavior, where the shape of the pulsed electrical signal is one or more of the following: sine wave, square wave, step wave, triangular wave, and sawtooth wave, with an amplitude of 0.9–3V. c V c The coercive voltage of the ferroelectric layer has a frequency of 0.1–1 GHz;

[0030] (3) For the upper electrode array (X) a Input pulse electrical signal to monitor the lower electrode array (Y) b The electrical response of the pulsed electrical signal is used to simultaneously simulate neuronal behavior and synaptic plasticity, wherein the shape of the pulsed electrical signal is one or more of the following: sine wave, square wave, step wave, triangular wave, and sawtooth wave, with an amplitude of 0.1–3V. c V c The coercive voltage of the ferroelectric layer has a frequency of 0.1–1 GHz and a pulse count of 1–10. 16 indivual;

[0031] (4) Simultaneously apply electrical signals to any two or more adjacent upper electrodes and monitor the electrical signal responses of any two or more adjacent lower electrodes to simultaneously simulate neuronal and synaptic plasticity and construct a brain-like network. The pulsed electrical signals are one or more of the following: sine wave, square wave, triangular wave, step wave, and sawtooth wave, with a frequency of 0.1–1 GHz and an amplitude of 0.1–3 V. c V c The coercive voltage of the ferroelectric layer, with a pulse count of 1 to 10. 16 indivual.

[0032] This invention proposes a neuromorphic approach based on an all-ferroelectric microstructure. By precisely altering the electric field loading conditions, it modulates the polarization, shielding, and space charge dynamics within the ferroelectric functional layers. This not only efficiently simulates neuronal ignition and synaptic plasticity but also simultaneously simulates neurons and synapses, thereby constructing an all-ferroelectric brain-like network. Compared to similar technologies, it boasts significant advantages in low power consumption and high integration, making it valuable in defense, industrial production, and information technology.

[0033] Compared with the prior art, the beneficial effects of the present invention are:

[0034] This invention provides a fully ferroelectric neuromimetic unit. Based on the coexistence and co-evolutionary ecosystem of ferroelectric domains and domain walls, and the similarity to the connection and working patterns between neurons and synapses, this invention proposes to utilize the co-evolutionary characteristics of ferroelectric domains / domain walls to simultaneously simulate the plasticity behavior of neurons and synapses, and constructs a fully ferroelectric neuromimetic unit. The fully ferroelectric neuromimetic unit disclosed in this invention can not only simulate the plasticity behavior of neurons and synapses by controlling any electrode region (X...) a Y b Applying specific signals can precisely simulate neuronal characteristics or synaptic plasticity, and can also simulate both simultaneously. Specifically, the all-ferroelectric neuromimetic unit provided by this invention can simulate neuromimetic behavior in the following ways: ① Applying specific signals to any electrode region (X... a Y b Applying a pulsed electrical signal sequence that is less than or close to the coercive voltage can simulate neuronal ignition characteristics through conductance changes caused by behaviors such as domain nucleation in the ferroelectric layer; ② For any electrode region (X a Y b Applying a pulsed electrical signal sequence that is close to or greater than the coercive voltage can simulate synaptic plasticity by mimicking conductance changes caused by processes such as domain wall tilting in the ferroelectric layer; ③ For any electrode region (X a Y bExcitation with a specific frequency and number of electrical pulse sequences can simultaneously simulate the plasticity of neurons and synapses through the co-evolutionary behavior of domains / domain walls, such as domain / domain wall connections and domain flipping modes (single-domain / multi-domain flipping); ④ By applying a specific electrical pulse sequence to any two or more adjacent electrode regions to change the domain / domain wall states of multiple different electrode regions corresponding to the ferroelectric layers, multiple neuron / synaptic units can be simulated, thereby realizing a fully ferroelectric neural network similar to a real biological brain network. It is evident that the fully ferroelectric neuromorphic unit provided by this invention offers a qualitative improvement in integration and power consumption compared to traditional ferroelectric brain-like devices, providing a new paradigm for the development of ferroelectric brain-like devices.

[0035] This invention also provides a method for preparing the all-ferroelectric neuromorphic unit. This method is simple to operate. By changing process parameters such as annealing rate, oxygen pressure and laser energy, an epitaxial ferroelectric functional layer that meets the requirements of neuromorphism can be prepared. Combined with conventional electrode array preparation process, a high-performance all-ferroelectric neuromorphic unit can be obtained. Attached Figure Description

[0036] Figure 1 This is a schematic diagram of the structure of an all-ferroelectric neural unit according to the present invention;

[0037] Reference numerals: 1. Upper electrode array; 2. Ferroelectric layer; 3. Lower electrode array;

[0038] Figure 2 This is an embodiment of the present invention, which describes the independent realization and in-situ observation of the plasticity of neurons and synapses in the ferroelectric layer under electrical pulse modulation: (a, c) TEM dark-field images of the unit device at different loading times, represented by different dashed boxes, wherein the initial polarization direction is downward and the scale is 100nm; (b) the correlation curve between the conductance of the unit device and the applied voltage when a small electric field signal of a step wave is applied; (d) the correlation curve between the conductance of the unit device and the continuous tilt angle of the domain wall; (e) the different dashed boxes are the TEM dark-field images of the unit device at times t1, t2 and t3, respectively, wherein the polarization direction of the written domain is upward and the scale is 100nm; (f) the relationship between the domain area value and the number of pulses when a -7.1V, 1Hz square wave pulse sequence is applied.

[0039] Figure 3 This is an in-situ observation of the neuromorphic behavior of the unit device under external electric field control in Embodiment 1 of the present invention: (a) The relationship between the domain area in the unit device and the number of pulses (labeled as the i-th, ii-th, and iii-th pulses, respectively) when a +7.1V, 1Hz square wave pulse sequence is applied. The inset shows the electric pulse excitation signal and response current signal received by the thin film; (b) The unit device is respectively under t i t ii and tiii (c) The TEM dark field image corresponding to the time, with the initial polarization facing downwards and the scale bar at 100 nm; (d) The correlation curve between the conductivity of the unit device and the area of ​​the domain (dotted line graph). Detailed Implementation

[0040] The specific embodiments of the present invention will be further described below. It should be noted that these descriptions are for the purpose of aiding understanding the present invention, but do not constitute a limitation thereof. Furthermore, the technical features involved in the various embodiments of the present invention described below can be combined with each other as long as they do not conflict with each other.

[0041] Unless otherwise specified, the experimental methods used in the following embodiments are conventional methods, and the experimental materials used in the following embodiments are all available through conventional commercial channels.

[0042] Example 1:

[0043] like Figure 1 The diagram shows a fully ferroelectric neuromorphic unit, whose structure, from top to bottom, includes: an upper electrode array, a ferroelectric functional layer, and a lower electrode array.

[0044] In this embodiment, a thickness of 1 mm and a size of 1 cm are selected. 2 (001)-oriented SrTiO3 was used as the substrate; a (001)-SrRuO3 film was deposited on the substrate as an etching sacrificial layer by pulsed laser deposition, followed by the deposition of (001)-PbZr. 0.2 Ti 0.8 An O3 thin film was used as the ferroelectric layer, W was deposited as the upper electrode array, and SrRuO3 was deposited as the lower electrode array using sodium periodate as the etchant. The specific experimental steps are as follows:

[0045] (1) A (001)-SrRuO3 etching sacrificial layer was epitaxially grown on a (001)-SrTiO3 substrate using pulsed laser deposition; wherein the operating conditions of the pulsed laser deposition method included: a sample deposition temperature of 690℃, a deposition oxygen pressure of 80 mtorr, and a laser energy density of 0.633 J·cm. -2 The deposition time was 5 minutes and the laser pulse frequency was 10 Hz.

[0046] (2) Using pulsed laser deposition, (001)-PbZr is grown on the upper surface of the (001)-SrRuO3 etched sacrificial layer in step (1). 0.2 Ti 0.8 O3 ferroelectric layer; wherein the operating conditions of the pulsed laser deposition method include: sample deposition chamber temperature of 600℃, deposition oxygen pressure of 160 mtorr, and laser energy density of 0.82 J·cm⁻¹.-2 The deposition time was 100 min, and the laser pulse frequency was 10 Hz; then it was cooled to room temperature at a cooling rate of 15 °C / min.

[0047] (3) A PMMA solution was spin-coated onto the surface of the thin film obtained in step (2) at a speed of 4200 r / min for 60 s. 40 μL of the solution was transferred and baked at 120 °C for 15 min to form a PMMA photoresist layer of about 50 nm on the surface of the thin film. Using an electron beam direct writing system (Pioneer Two manufactured by Raith Corporation), a square array pattern with a width of about 10 nm, a length of about 100 μm, a square shape, and a spacing of 5 μm between each other was written into the photoresist layer. Standard development, fixing, and baking operations were performed to obtain a thin film containing the square array pattern. A tungsten metal layer was deposited on its surface using a magnetron sputtering instrument (deposition power of 10 W for 30 s). The photoresist was removed using acetone to obtain a ferroelectric thin film containing the upper electrode array.

[0048] (4) The SrRuO3 sacrificial layer was etched using sodium periodate solution (concentration of 1 mol / L) to obtain a ferroelectric free radical film containing an upper electrode array;

[0049] (5) Using pulsed laser deposition technology and a mask (etched with a square array pattern, wherein the width of each square is about 1 μm, the length is about 100 μm, the shape is square, and the spacing between them is 4.5 μm), a SrRuO3 lower electrode array metal layer is deposited on the back side of the free radical film. The angle between the two electrode arrays is about 90°. The specific process parameters are the same as in step (1). The final result is as follows: Figure 1 The fully ferroelectric neuromorphic unit device shown.

[0050] To intuitively verify the neuromorphic performance of the all-ferroelectric neuromorphic unit provided in Embodiment 1, a high spatiotemporal resolution characterization system capable of observing the microscopic physical images and electrical properties of electric domains under in-situ electric field loading is required. Therefore, preprocessing of the all-ferroelectric neuromorphic unit device is necessary to meet the testing requirements for high spatiotemporal resolution characterization. First, the all-ferroelectric neuromorphic unit is bonded to a SrTiO3 substrate. Then, an ion thinning instrument (Gatan) is used to thin the all-ferroelectric neuromorphic unit to electron beam transparency (~100 nm), and the sample is bonded to a metal ring to obtain an in-situ TEM cross-section sample.

[0051] The neuromorphic behavior of a unit device was tested using a high spatiotemporal resolution system: First, a test circuit was constructed using a tungsten nanoprobe with a diameter of approximately 10 nm as the equivalent upper electrode and SrRuO3 as the bottom electrode. A 10 kΩ resistor was connected in the test circuit to modulate the shielding charge dynamics time. Then, the high spatiotemporal resolution system was used to apply... Figure 2(b) shows a stepped square wave (composed of many small square waves superimposed, with a total amplitude of 0 to -10V) used to excite the sample. It can be seen that when an electrical signal below or close to the coercive voltage is applied, as shown... Figure 2 (a) Nucleation begins in the domains of the ferroelectric layer. However, due to insufficient driving force from the electric field, the domains initially exhibit a wedge shape and low conductivity. As the electrical signal gradually increases and approaches the coercive voltage, the domains grow rapidly longitudinally, exhibiting cumulative positive flipping behavior. This results in a rapid increase and subsequent decrease in domain conductivity, with a change of up to 8.085 times. If this independent domain structure is considered as a neuron, it can accumulate electrical signal excitation after receiving a step-change electrical pulse, thus achieving the ignition behavior of the neuron. This demonstrates a pathway to simulate neuronal behavior through the accumulation and nucleation of domains.

[0052] For the electric domains that have completed the longitudinal nucleation stage, further excitation is applied using an electrical signal slightly larger than the coercive voltage of the ferroelectric thin film (e.g., ...). Figure 2 When the electrical signal in d is shown in the schematic diagram, it can cause the domain walls to exhibit continuous tilting behavior, and the dark field image is like... Figure 2 As shown in c. At this point, the thin film conductivity increases with increasing domain wall tilt angle (dotted line graph, where the curve is a fitted line obtained from characteristic data points), with a conductivity change of more than 2.84 times. This is because the change in tilt angle alters the charge accumulation state of the domain wall itself. It is noteworthy that the domain wall tilting is a relatively continuous dynamic behavior, implying the existence of a large number of distinguishable conductivity states, suggesting that the unit device has good potential in simulating synaptic plasticity. Besides simulating synaptic plasticity by using a large step wave to continuously tilt the domain wall, other methods such as... Figure 2 The illustration shows a series of consecutive square wave pulses used to excite the thin film, which can also cause the domain walls in the film to gradually move with the accumulation of electrical pulses, such as... Figure 2 The bar chart in f and Figure 2 As shown in the dark-field images at times t1, t2, and t3 in e, the domain area changes by several times or more, and the domain wall movement distance exceeds 300 nm. These results from electrical pulse excitations of different shapes and real-time dark-field images demonstrate that, through precise electrical pulse excitation, the expansion / tilting behavior of domain walls can be utilized to achieve long-term synaptic enhancement characteristics, and it also possesses the potential to remotely link multiple domains. In summary, the high spatiotemporal resolution characterization results indicate that the all-ferroelectric neuromimetic unit device can precisely and independently realize neuronal mimicry and synaptic plasticity behavior through external electric field loading.

[0053] Building upon the independent and controllable simulation of neuronal and synaptic behavior, simultaneous simulation of neuronal and synaptic behavior based on fully ferroelectric microstructures is also possible. In the actual operation of ferroelectric devices, domains cannot maintain an idealized specific distance; collisions, merging, and other multi-body interactions inevitably occur. Domain walls also link multiple domains through complex correlation behaviors, altering the domain's own state. By gradually injecting charge into the thin film during domain movement, a locally differentiated spatial charge region is formed. Thus, it becomes possible to observe phenomena such as... Figure 3 Figures a and b illustrate the phenomenon where the area and number of domains gradually change with the increase of the number of pulses. For example... Figure 3 As shown in Figure c, the electrical response data at corresponding times also indicate that the film conductivity nearly doubles with the increase in the number and area of ​​domains written. The film conductivity values ​​are extracted from the real-time current response curves based on 90% of the voltage hold time, i.e., t = 3.27, 4.27, and 5.27 s. The domain movement evolves from a single-domain mode to a multi-domain mode. Furthermore, this process is well-controllable; the number of domains can be increased from one domain to two domains and then to three domains with the increase in the number of pulses. Simultaneously, an approximate connection between domain walls can be observed between two domains, while domain walls at greater distances remain independent due to the low space charge density. This process is analogous to two neurons striving to extend and connect via synapses. Figure 3 It can be observed that although the diameter of the upper tungsten electrode is only 10 nm, the expansion range of the electric domains can exceed 1 μm. This means that the domain walls can expand to a distance hundreds of times the electrode size. This indicates that, based on achieving neuromorphic function in a single electrode region, a fully ferroelectric brain-like neural network can be constructed through the synergy of any two or more electrodes. In summary, the high spatiotemporal resolution characterization results demonstrate that the fully ferroelectric neuromorphic unit can achieve neuromorphic behavior based on the cooperative evolution of ferroelectric microstructures.

[0054] Example 2:

[0055] In this embodiment, a thickness of 1 mm and a size of 1 cm are selected. 2 (001) oriented SrTiO3 was used as the substrate; a (001)-SrRuO3 film was deposited on the substrate as an etching sacrificial layer by pulsed laser deposition, a BaTiO3 film was deposited as a ferroelectric layer, Pt was evaporated as the upper electrode array, and SrRuO3 was deposited as the lower electrode array using sodium periodate as an etchant. The specific experimental steps are as follows:

[0056] (1) A (001)-SrRuO3 etching sacrificial layer was epitaxially grown on a (001)-SrTiO3 substrate using pulsed laser deposition; wherein the operating conditions of the pulsed laser deposition method included: a sample deposition temperature of 690℃, a deposition oxygen pressure of 80 mtorr, and a laser energy density of 0.633 J·cm. -2 The deposition time was 5 minutes and the laser pulse frequency was 10 Hz.

[0057] (2) A BaTiO3 ferroelectric layer is grown on the upper surface of the (001)-SrRuO3 etched sacrificial layer in step (1) using pulsed laser deposition; wherein the operating conditions of the pulsed laser deposition method include: sample deposition chamber temperature of 650℃, deposition oxygen pressure of 200 mtorr, and laser energy density of 0.85 J·cm. -2 The deposition time was 60 min, and the laser pulse frequency was 10 Hz; then it was cooled to room temperature at a cooling rate of 15 °C / min.

[0058] (3) A PMMA solution was spin-coated onto the surface of the thin film obtained in step (2) at a speed of 4200 r / min for 60 s. 40 μL of the solution was transferred and baked at 120 °C for 15 min to form a PMMA photoresist layer of about 50 nm on the surface of the thin film. Using an electron beam direct writing system (Pioneer Two manufactured by Raith Corporation), a square array pattern with a width of about 50 nm, a length of about 100 μm, and a spacing of 10 μm was written onto the photoresist layer. Standard development, fixing, and baking operations were performed to obtain a thin film containing the square array pattern. A platinum metal layer was deposited on its surface using a magnetron sputtering instrument (deposition power of 5 W for 30 s). The photoresist was removed using acetone to obtain a ferroelectric thin film containing the upper electrode array.

[0059] (4) The SrRuO3 sacrificial layer was etched using sodium periodate solution (concentration of 1 mol / L) to obtain a ferroelectric free radical film containing an upper electrode array;

[0060] (5) Using pulsed laser deposition technology and a mask (etched with a square array pattern, wherein the width of each square is about 5 μm, the length is about 100 μm, the shape is square, and the spacing between them is 10 μm), a SrRuO3 lower electrode array metal layer is deposited on the back of the free radical film. The angle between the two electrode arrays is about 90°. The specific process parameters are the same as in step (1), and finally the all-ferroelectric neuromorphic unit device is obtained.

[0061] According to the test method of Example 1, the results show that the effect of Example 2 is roughly the same as that of Example 1.

[0062] Example 3:

[0063] In this embodiment, a thickness of 1 mm and a size of 1 cm are selected. 2 (001) oriented SrTiO3 was used as the substrate; a (001)-SrRuO3 film was deposited on the substrate as an etching sacrificial layer by pulsed laser deposition, a BiFeO3 film was deposited as a ferroelectric layer, Au was evaporated as the upper electrode array, and SrRuO3 was deposited as the lower electrode array using sodium periodate as an etchant. The specific experimental steps are as follows:

[0064] (1) A (001)-SrRuO3 etching sacrificial layer was epitaxially grown on a (001)-SrTiO3 substrate using pulsed laser deposition; wherein the operating conditions of the pulsed laser deposition method included: a sample deposition temperature of 690℃, a deposition oxygen pressure of 80 mtorr, and a laser energy density of 0.633 J·cm. -2 The deposition time was 5 minutes and the laser pulse frequency was 10 Hz.

[0065] (2) A BiFeO3 ferroelectric layer is grown on the upper surface of the (001)-SrRuO3 etched sacrificial layer in step (1) using pulsed laser deposition; wherein the operating conditions of the pulsed laser deposition method include: sample deposition chamber temperature of 690℃, deposition oxygen pressure of 200 mtorr, and laser energy density of 0.65 J·cm. -2 The deposition time was 120 min, and the laser pulse frequency was 10 Hz; then it was cooled to room temperature at a cooling rate of 15 °C / min.

[0066] (3) A PMMA solution was spin-coated onto the surface of the thin film obtained in step (2) at a speed of 4200 r / min for 60 s. 40 μL of the solution was transferred and baked at 120 °C for 15 min to form a PMMA photoresist layer of about 50 nm on the thin film surface. Using an electron beam direct writing system (Pioneer Two manufactured by Raith Corporation), a square array pattern with a width of about 100 nm, a length of about 100 μm, a square shape, and a spacing of 5 μm between each other was written onto the photoresist layer. Standard development, fixing, and baking operations were performed to obtain a thin film containing the square array pattern. An Au metal layer was deposited on its surface using a magnetron sputtering instrument (deposition power of 10 W for 30 s). The photoresist was removed using acetone to obtain a ferroelectric thin film containing the upper electrode array.

[0067] (4) The SrRuO3 sacrificial layer was etched using sodium periodate solution (concentration of 1 mol / L) to obtain a ferroelectric free radical film containing an upper electrode array;

[0068] (5) Using pulsed laser deposition technology and a mask (etched with a square array pattern, wherein the width of each square is about 10 μm, the length is about 100 μm, the shape is square, and the spacing between them is 8 μm), a SrRuO3 lower electrode array metal layer is deposited on the back side of the free radical film. The angle between the two electrode arrays is about 90°. The specific process parameters are the same as in step (1), and finally the all-ferroelectric neuromorphic unit device is obtained.

[0069] According to the test method of Example 1, the results show that the effect of Example 3 is roughly similar to that of Example 1.

[0070] The embodiments of the present invention have been described in detail above, but the present invention is not limited to the described embodiments. For those skilled in the art, various changes, modifications, substitutions, and variations can be made to these embodiments without departing from the principles and spirit of the present invention, and these variations still fall within the protection scope of the present invention.

Claims

1. A method of simulating neural behavior with a fully ferroelectric neural- like unit, characterized by, The full ferroelectric neural state unit structure comprises an upper electrode array, a ferroelectric layer and a lower electrode array arranged in sequence from top to bottom. The single electrode of the upper electrode array and the lower electrode array is rectangular in shape, the width of the single electrode is 1 nm-10 μm, and the spacing between adjacent electrodes is 10 nm-50 μm; the width ratio of the single electrode of the upper electrode array to the single electrode of the lower electrode array is 0.01-100. The upper electrode array and the lower electrode array are misaligned, and the angle of the misalignment angle ranges from 0 to 180 degrees. The thickness of the upper electrode array and the lower electrode array is 1-500 nm. The lower electrode array and the lower electrode array have a projection overlap region. The thickness of the ferroelectric layer is 0.001-1 μm. The method for simulating neural behavior Comprises the following modes: (1) input pulse electrical signal to the upper electrode array, monitor the electrical response of the lower electrode array to simulate the behavior of neurons, wherein the shape of the pulse electrical signal is one or more of sine wave, square wave, step wave, triangular wave and sawtooth wave, the amplitude is 0.1-1.5 V c , V c is the coercive voltage of the ferroelectric layer, the frequency is 0.1-1 GHz, and the pulse number is 1-10 10 ​ (2) input pulse electrical signals to the upper electrode array, monitor the electrical response of the lower electrode array to simulate synaptic plasticity behavior, wherein the shape of the pulse electrical signal is one or more of a sine wave, a square wave, a step wave, a triangular wave, and a sawtooth wave, and the amplitude is 0.9-3 V c , V c is the coercive voltage of the ferroelectric layer, and the frequency is 0.1-1 GHz; (3) inputting a pulse electric signal to the upper electrode array, monitoring the electric response of the lower electrode array to simultaneously simulate the neuron behavior and the synaptic plasticity behavior, wherein the pulse electric signal is one or more of a sine wave, a square wave, a ladder wave, a triangle wave and a sawtooth wave, and the amplitude is 0.1-3 V c , V c is the coercive voltage of the ferroelectric layer, the frequency is 0.1-1 GHz, and the pulse number is 1-10 16 . (4) simultaneously loading electrical signals on any two or more adjacent upper electrodes, monitoring the electrical signal response of any two or more adjacent lower electrodes, thereby simultaneously simulating neuron and synaptic plasticity behavior, and constructing a brain-like network, wherein the shape of the pulse electrical signal is one or more of a sine wave, a square wave, a triangular wave, a step wave, and a sawtooth wave, the frequency is 0.1-1 GHz, and the amplitude is 0.1-3 V c , V c is the coercive voltage of the ferroelectric layer, and the number of pulses is 1-10 16 .

2. The method of simulating neural behavior of claim 1, wherein, The material of the ferroelectric layer is PbZr x Ti 1-x one or more of O3, PbTiO3, BaTiO3, BiFeO3; the PbZr x Ti 1-x O3 formula, 0 < x < 1.

3. The method of simulating neural behavior of claim 1, wherein, The material of the upper electrode array and the lower electrode array is selected from one or more of Ag, Au, Pt, W, ITO, ZnO and SrRuO3.

4. The method of simulating neural behavior of claim 1, wherein, The preparation method of the full ferroelectric neural state unit comprises the following steps: (1) Depositing a sacrificial transition layer and a ferroelectric layer on the upper surface of the substrate in sequence; (2) Depositing an upper electrode array on the upper surface of the ferroelectric layer; (3) Etching the sacrificial transition layer to obtain a free radical ferroelectric thin film containing an electrode array; (4) Depositing a lower electrode array on the lower surface of the free radical ferroelectric thin film to obtain a full ferroelectric neural state unit.

5. The method of simulating neural behavior of claim 4, wherein, In step (1), the sacrificial transition layer is SrRuO3, and the thickness is 1-50 nm.

6. The method of simulating neural behavior of claim 4, wherein, In step (1), the deposition method used is pulsed laser deposition, and the conditions are: the sample deposition temperature is 550-800 ℃, the oxygen deposition pressure is 50-300mtorr, the laser energy density is 100-1500 mJ·cm -2 , and the laser pulse frequency is 5-10 Hz.

Images