Relaxation type memristor and preparation method and application thereof

Abstract

The invention belongs to the related technical field of semiconductor devices, and discloses a relaxation type memristor and a preparation method and application thereof.The relaxation type memristor comprises a bottom electrode layer, a resistive layer and a top electrode layer which are arranged from bottom to top, and the top electrode layer is made of an inactive metal elementary substance or alloy containing inactive metal; the resistive layer is an inorganic molecular crystal film. According to the memristor, an inactive metal is adopted as a material of a top electrode layer of the memristor, and a growth process of a conductive wire in an inorganic molecular crystal resistive layer is controlled, so that a stable and adjustable conductive state is obtained, accurate mapping of an analog electric signal is further realized, and higher precision is provided for neuromorphic calculation based on the memristor.

Description

Technical Field

[0001] The present invention belongs to the technical field related to semiconductor devices, and more specifically, relates to a relaxor-type memristor, a preparation method thereof, and an application thereof. Background Art

[0002] As a new type of electronic device with a resistance memory function, the resistance value of a memristor can be reversibly switched under the stimulation of an applied voltage or current, so as to store and process information in the form of resistance value; thanks to its inherent dynamic characteristics and non-linear resistive switching behavior, the memristor has become a powerful candidate for implementing artificial neural networks at the hardware level, showing great potential in simulating the plasticity of neural synapses and performing complex computing tasks.

[0003] Most of the reported memristors currently are based on conductive filaments of active metals (such as Cu, Ag, etc.) to achieve resistance switching. The conductance mutation phenomenon during the formation and breakage of the conductive filaments limits the number of adjustable conductance states and reduces the linearity of the system. The limited number of conductance states cannot fully reflect the original data information, which will lead to incorrect signal output during the execution of computing tasks and seriously affect the computing performance.

[0004] Analog coding based on multiple resistance states can more accurately retain the original features of data, which is crucial for improving the recognition accuracy of neural networks, so that neuromorphic computing is applicable to a wider range of application scenarios and more complex computing tasks. Therefore, the current focus of research and development work is to develop a memristor with adjustable conductance states and excellent short-term dynamic characteristics, and ensure that its internal storage state accurately maps the feature information of the input data. Summary of the Invention

[0005] Aiming at the above defects or improvement requirements of the prior art, the present invention provides a relaxor-type memristor, a preparation method thereof, and an application thereof, aiming to solve the problem of limited number of conductance states of existing memristors.

[0006] To achieve the above object, according to one aspect of the present invention, there is provided a relaxor-type memristor, which includes a bottom electrode layer, a resistive switching layer, and a top electrode layer arranged from bottom to top. The material of the top electrode layer is an inactive metal single substance or an alloy containing an inactive metal; the resistive switching layer is an inorganic molecular crystal thin film.

[0007] Further, the inactive metal single substance is any one of antimony, gold, ruthenium, and platinum.

[0008] Further, the alloy contains two or more of antimony, gold, ruthenium, and platinum, and the atoms of any doping component are greater than 5%.

[0009] Further, the thickness of the top electrode layer is 5 nm - 200 nm.

[0010] Further, the material of the inorganic molecular crystal thin film is Sb 2 O 3 or As 2 O 3 One of them, and the thickness of the resistive switching layer is 5 nm - 100 nm.

[0011] Further, the material of the bottom electrode layer is any one of Pt, Cr, TiN and graphite.

[0012] Further, the material of the bottom electrode layer is Pt, Cr or TiN, and its thickness is 5 nm - 50 nm.

[0013] Further, the material of the bottom electrode layer is graphite, and its thickness is 0.3 nm - 30 nm.

[0014] The present invention provides a preparation method of a relaxor memristor, and the preparation method is used to prepare the relaxor memristor as described above.

[0015] The present invention also provides an application of the relaxor memristor as described above in neuromorphic computing.

[0016] Generally speaking, compared with the prior art through the above technical solutions conceived by the present invention, the relaxor memristor, its preparation method and application provided by the present invention mainly have the following beneficial effects: 1. The present invention uses inactive metals such as antimony (Sb), gold (Au), ruthenium (Ru) or platinum (Pt) as the material of the top electrode layer of the memristor. Based on the electrochemical metallization resistive switching mechanism, the metal ions are driven by an electric field to migrate and redox react in the resistive switching layer, resulting in the formation or breakage of the conductive filament, thereby realizing the switching between high and low resistance states. Different from other active metal top electrodes (such as silver Ag, copper Cu, titanium Ti, etc.), the diffusion activation energy of inactive metals is relatively high, and the growth of the conductive filament is relatively slow under the action of an external electric field, and the device conductance gradually changes during the resistive switching process, so a large number of distinguishable conductance states are obtained, realizing precise control of the resistive state of the memristor.

[0017] 2. The application of the invention will break through the limitation that neuromorphic computing is difficult to directly process analog signals through hardware, greatly improve the accuracy of complex signal processing, and provide a feasible solution for the hardware implementation of brain-like neuromorphic computing. Description of the Drawings

[0018] Figure 1 is a schematic structural diagram of the memristor in Embodiment 1 of the present invention.

[0019] Figure 2 is the I-V characteristic diagram of the memristor in Embodiment 1.

[0020] Figure 3 It is the relaxation characteristic diagram of the memristor in Embodiment 1.

[0021] Figure 4 It is the pulse response characteristic diagram of the memristor in Embodiment 1.

[0022] Figure 5 It is the 5-bit pulse coding diagram of the memristor in Embodiment 1. Detailed implementation manners

[0023] In order to make the objectives, technical solutions and advantages of the present invention clearer and more understandable, the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not used to limit the present invention. In addition, 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.

[0024] The present invention provides a relaxation-type memristor. The memristor uses an inactive metal as the material of the top electrode layer of the memristor and controls the growth process of the conductive filaments in the inorganic molecular crystal resistive switching layer, so as to obtain a stable and adjustable conductance state, and further achieve an accurate mapping of analog electrical signals, providing higher precision for memristor-based neuromorphic computing.

[0025] The memristor includes a bottom electrode layer, a resistive switching layer and a top electrode layer arranged from bottom to top. The material of the top electrode layer is an inactive metal element or an alloy containing an inactive metal; the resistive switching layer is an inorganic molecular crystal thin film.

[0026] The inactive metal element is any one of antimony (Sb), gold (Au), ruthenium (Ru), and platinum (Pt); the alloy contains two or more of antimony (Sb), gold (Au), ruthenium (Ru), and platinum (Pt), and the atoms of any doping component are greater than 5%. The thickness of the top electrode layer is 5 nm - 200 nm.

[0027] The material of the inorganic molecular crystal thin film is Sb 2 O 3 or As 2 O 3 One of them, and the thickness of the resistive switching layer is 5 nm - 100 nm.

[0028] The material of the bottom electrode layer includes Au, Pt, Cr, TiN and graphite; when the material of the bottom electrode layer is Pt, Cr or TiN, its thickness is 5 nm - 50 nm; when the material of the bottom electrode layer is graphite, its thickness is 0.3 nm - 30 nm.

[0029] The present invention also provides a method for preparing the relaxation-type memristor as described above, and the preparation method includes the following steps: (1) Depositing a bottom electrode layer on a substrate, or forming a bottom electrode layer by mechanically exfoliating few-layer graphene.

[0030] (2) Preparing an inorganic molecular crystal thin film on the bottom electrode to form a resistive switching layer.

[0031] (3) Depositing an inactive metal element or an alloy containing an inactive metal on the resistive switching layer to form a top electrode layer.

[0032] The method for depositing an inactive metal element on the resistive switching layer includes: forming a thin film on a substrate by thermal evaporation deposition, electron beam evaporation deposition or magnetron sputtering deposition; the method for depositing an alloy containing an inactive metal on the resistive switching layer includes: forming an alloy thin film on a substrate by multi-source evaporation plating, evaporating an alloy metal source or alternately depositing different metal thin films and annealing.

[0033] In one embodiment, the resistive switching layer is Sb 2 O 3 or As 2 O 3 thin film grown directly on the bottom electrode layer by thermal evaporation deposition.

[0034] The bottom electrode layer is one of Au, Pt, Cr, TiN deposited by thermal evaporation deposition, electron beam evaporation deposition or magnetron sputtering deposition, or graphene exfoliated mechanically.

[0035] The present invention also provides an application of the relaxation-type memristor as described above in neuromorphic computing, including pulse signal processing, timing data prediction and image classification, The following uses specific embodiments to further elaborate on the present invention in detail.

[0036] Example 1 This Example 1 provides a memristive device, the structure of which is as Figure 1 shown, wherein the top electrode layer is a metal Sb elemental thin film.

[0037] The preparation method of the memristive device in this Example 1 includes the following steps: (1) Using a silicon wafer as a substrate, performing oxygen plasma treatment on the substrate, and the plasma treatment conditions are: the oxygen flow rate is 20 sccm, the power is 80 W, and the cleaning time is 10 min. Subsequently, stick the tape with exfoliated graphene on the substrate surface, place the substrate on a heating plate, and heat it at 100 °C for 2 min. After heating, remove the tape, and the graphene is transferred to the substrate surface. Select the graphene with a thinner thickness under a light microscope as the bottom electrode.

[0038] (2) Use electron beam lithography to fabricate lead patterns on graphene. The photoresist used is PMMA. After spin-coating, the photoresist is baked at 150 °C for 5 minutes. After exposure, the sample is immersed in the developer for 5 s - 10 s, then immersed in the fixer for 10 s and dried with a nitrogen gun.

[0039] (3) Use thermal evaporation deposition to deposit Cr / Au films on the lead patterns prepared in step (2) for pin testing. The thicknesses of the Cr and Au films are 10 nm and 50 nm respectively. The process conditions are: using Cr or Au particles as the evaporation source, nitrogen as the evaporation atmosphere, an evaporation rate of 0.2 Å / s, and a chamber pressure less than 8×10 -4 Pa. After coating, the sample with the bottom electrode is placed in acetone, and the acetone is heated to 60 °C - 70 °C and soaked for 20 min to complete degluing.

[0040] (4) Use thermal evaporation deposition to deposit a layer of Sb 2 O 3 as the resistive switching layer. The thickness of the Sb 2 O 3 film is 10 nm. The process conditions are: using Sb 2 O 3 powder as the evaporation source, nitrogen as the evaporation atmosphere, an evaporation rate of 0.1 Å / s, and a chamber pressure less than 8×10 -4 Pa.

[0041] (5) Use electron beam lithography to fabricate top electrode patterns on the Sb 2 O 3 resistive switching layer. The process conditions are the same as those in step (2).

[0042] (6) Use thermal evaporation deposition to deposit a layer of Sb film as the top electrode layer on the top electrode patterns prepared in step (5). The thickness of the Sb film is 40 nm. The process conditions are: using Sb particles as the evaporation source, nitrogen as the evaporation atmosphere, an evaporation rate of 0.1 Å / s, and a chamber pressure less than 8×10 -4 Pa.

[0043] (7) Use thermal evaporation deposition to deposit a layer of Au film as the protective layer on the top electrode prepared in step (6). The thickness of the Au film is 20 nm. The process conditions are: using Au particles as the evaporation source, nitrogen as the evaporation atmosphere, an evaporation rate of 0.2 Å / s, and a chamber pressure less than 8×10 -4 Pa. After coating, the sample with the top electrode is placed in acetone, and the acetone is heated to 60 °C - 70 °C and soaked for 20 min to complete degluing, obtaining the memristor.

[0044] Result Analysis In this Example 1, the I-V characteristic curve of the memristor based on the Sb single-element top electrode layer is as follows Figure 2 shown. At a relatively low current limit (100 nA), the device exhibits typical non-threshold volatile resistive switching characteristics, and the conductance gradually changes during the set and reset processes; the set voltage of the device is 2.5 V, and the switching ratio is about 10 4 , with distinct high and low resistance states.

[0045] In this Example 1, the relaxation curve of the memristor based on the Sb single-element top electrode layer is as follows Figure 3 shown. A 2 V stimulating pulse is applied to the device, and then the device current is continuously read with a 0.1 V voltage. It is observed that the current gradually decays with time, and the relaxation time is about 5 μs; this indicates that the device has typical time-sequence dependent characteristics and can be used for the processing of time-sequence signals.

[0046] In this Example 1, the pulse response characteristics of the memristor based on the Sb single-element top electrode layer are as follows Figure 4 shown. Five identical pulses (pulse amplitude of 3.5 V, pulse width of 10 μs) with an interval of 2 μs are continuously applied and the dynamic response of the device is detected. The current gradually increases as the number of pulses accumulates. Further, a 5-bit pulse sequence is used to modulate the current of the memristor, and 32 distinguishable conductance states are obtained. The results are as follows Figure 5 shown. The above experimental results show that the conductance of this device can be precisely adjusted by electrical pulses.

[0047] Example 2 In this Example 2, the memristive device is fabricated using the same steps as in Example 1. The difference from Example 1 is that the top electrode is an alloy film deposited by co-sputtering, where the matrix metal is Au and the doped non-active metal is Sb, and the doping atomic ratio of Sb is 10%.

[0048] It is easy for those skilled in the art to understand that the above are only the preferred embodiments of the present invention and are not intended to limit the present invention. Any modifications, equivalent replacements, and improvements made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A relaxation memristor, characterized in that: The memristor comprises a bottom electrode layer, a resistive switching layer and a top electrode layer arranged from bottom to top. The material of the top electrode layer is an inactive metal element or an alloy containing an inactive metal; the resistive switching layer is an inorganic molecular crystal film.

2. The relaxation memristor according to claim 1, characterized in that: The inactive metal element is any one of antimony, gold, ruthenium and platinum.

3. The relaxation memristor according to claim 1, characterized in that: The alloy contains two or more of antimony, gold, ruthenium and platinum, and the atomic content of any doping component is greater than 5%.

4. The relaxation memristor according to claim 1, characterized in that: The thickness of the top electrode layer is 5 nm-200 nm.

5. The relaxation memristor according to claim 1, characterized in that: The material of the inorganic molecular crystal film is one of Sb2O3 and As2O3, and the thickness of the resistive layer is 5 nm-100 nm.

6. The relaxation memristor according to claim 1, characterized in that: The material of the bottom electrode layer is any one of Pt, Cr, TiN and graphite.

7. The relaxation memristor according to claim 1, characterized in that: The material of the bottom electrode layer is Pt, Cr or TiN, and its thickness is 5 nm-50 nm.

8. The relaxation memristor according to claim 1, characterized in that: The material of the bottom electrode layer is graphite, and the thickness thereof is 0.3 nm-30 nm.

9. A method for preparing a relaxation type memristor, characterized in that: The preparation method is used to prepare the relaxation memristor according to any one of claims 1 to 8.

10. Application of the relaxation memristor according to any one of claims 1 to 8 in neuromorphic computing.

Images