Ferroelectric memristor based on scandium aluminum nitride, preparation method and application thereof

Abstract

This invention provides a ferroelectric memristor based on aluminum scandium nitride, its fabrication method, and its applications. The memristor structure provided by this invention can be represented as Pd / Al. 0.77 Sc 0.23 N / TiN / Si. The fabrication method of this memristor includes: growing a TiN bottom electrode on a Si substrate, and growing Al on the TiN bottom electrode by nitrogen reactive magnetron sputtering. 0.77 Sc 0.23 N functional layer; in Al 0.77 Sc 0.23 A palladium electrode layer is grown on the N-functional layer. This invention is the first to use aluminum scandium nitride as the intermediate functional layer of a non-volatile memory. It has strong self-rectification characteristics and can better solve problems such as crosstalk when applied to cross-arrays. It can develop and realize new biological synaptic functions in simulated synaptic neuron systems, and its performance is good. It is a non-volatile memory with good storage performance, low turn-on voltage, and broad application prospects that simulates biological synaptic behavior.

Description

Technical Field

[0001] This invention relates to the field of memory technology, specifically to a ferroelectric memristor based on aluminum scandium nitride, its preparation method, and its application. Background Technology

[0002] A memristor, short for memory resistor, is a circuit device that represents the relationship between magnetic flux and electric charge. While a memristor has the dimension of resistance, unlike a regular resistor, its resistance is determined by the charge flowing through it. Therefore, by measuring the resistance of a memristor, the amount of charge flowing through it can be determined, thus enabling it to "memorize" charge. The concept of a memristor was first proposed by the Chinese-American scientist, Shao-Tang Tsai. Simply put, a memristor is a non-linear resistor with a memory function. Its resistance can be changed by controlling the change in current. If a high resistance is defined as "1" and a low resistance as "0", then this resistor can store data. In essence, it is a non-linear resistor with a memory function. Using a common water pipe analogy, the current represents the amount of water flowing through it, and the resistance represents the pipe's diameter. When water flows in one direction, the pipe becomes wider with the flow rate. If the water flow is turned off, the pipe's diameter remains unchanged; conversely, when the water flows in the opposite direction, the pipe becomes narrower. Because such a component "remembers" the previous amount of current, it is called a memristor. Then in 2008, researchers at HP Labs created the first memristor device, sparking a surge of research into memristors in the scientific community.

[0003] Memristors are two-terminal devices, allowing for high-density integration. Furthermore, their transitions between different resistive states are rapid, making them compatible with CMOS processes. These characteristics give memristors unique advantages in information storage and neural synapse simulation.

[0004] The simplest application of memristors is as non-volatile impedance memory. Due to their simple structure, fast access speed, low power consumption, and ease of integration, they are a strong candidate for next-generation high-density memory and have been extensively studied. Resistive random access memory (RRAM) is a typical sandwich structure based on a top electrode-dielectric layer-bottom electrode. It primarily utilizes the reversible transition between high and low resistance states in the intermediate dielectric layer under different electrical stimuli to store data.

[0005] Memristors can also be used for logic calculations. In 2010, HP Labs announced that memristors could perform Boolean logic operations, a discovery that shook the computer science community. Researchers at the National University of Defense Technology, who led the development of the "Tianhe" series of supercomputers, believed after conducting follow-up research that, theoretically, memristors could completely replace all current digital logic circuits.

[0006] Traditional memristors operate by displacing matter within a material due to voltage, but this internal displacement is highly unstable. Ferroelectric memristors, which use ferroelectric materials as the functional layer, overcome this problem. The polarization reversal of the ferroelectric thin film in a ferroelectric memristor determines its resistance, resulting in stable performance, low power consumption, and flexible design. Furthermore, in tunnel junctions with ferroelectric barriers, the polarization reversal of the ferroelectric thin film leads to changes in the resistance of the ferroelectric tunnel junction. However, most ferroelectric memristors are unipolar or bipolar, and as integration density increases, significant crosstalk occurs between devices. Summary of the Invention

[0007] The purpose of this invention is to provide a ferroelectric memristor based on aluminum scandium nitride, its preparation method and application, so as to solve the problem of crosstalk between devices when the integration density of existing ferroelectric memristors increases, and the ferroelectric memristor has the function of neural synapse bionics.

[0008] This invention is implemented as follows:

[0009] The ferroelectric memristor based on aluminum scandium nitride provided by this invention has a structure in which a TiN bottom electrode layer and an Al layer are sequentially formed on a Si substrate. 0.77 Sc 0.23 N functional layer and Pd top electrode layer.

[0010] Preferably, the thickness of the TiN bottom electrode layer is 50 nm.

[0011] Preferably, the above-mentioned Al 0.77 Sc 0.23 The thickness of the N-functional layer is 30 nm.

[0012] Preferably, the thickness of the Pd top electrode layer is 20 nm.

[0013] The above-mentioned method for fabricating ferroelectric memristors based on aluminum scandium nitride includes the following steps:

[0014] (a) The Si substrate was ultrasonically cleaned in acetone, alcohol and deionized water in sequence. Then the Si substrate was cleaned in hydrofluoric acid dilution solution (hydrofluoric acid: deionized water = 1:3) for 30 seconds to remove the SiO2 layer on the Si substrate. Then it was cleaned in deionized water to remove the residual HF solution. Finally, the substrate was taken out and dried with a nitrogen gun.

[0015] (b) Fix the Si substrate onto the sample stage of the magnetron sputtering equipment cavity and evacuate the cavity to a back-bottom vacuum of 2 × 10⁻⁶. -4Below Pa, argon gas is introduced into the cavity as sputtering gas. The inlet valve is adjusted to maintain the pressure in the cavity at 0.8 Pa. The radio frequency source controlling the TiN ignition is turned on and the power of the radio frequency source is adjusted to 15 W to ignite the TiN target material. Pre-sputtering is performed for 1 to 5 minutes. Then, formal sputtering is performed for 1 hour to form a TiN bottom electrode layer on the Si substrate.

[0016] (c) Fix a 99.99% pure aluminum target to the RF source, a 99.99% pure scandium target to the DC source, fix the TiN substrate to the sample stage of the magnetron sputtering equipment cavity, and evacuate the cavity to 2×10⁻⁶. -4 Nitrogen gas was introduced into the chamber at Pa, serving as both the sputtering gas and the reactant gas. The RF source controlling the ignition of the aluminum target and the DC source controlling the ignition of the scandium target were turned on, with the RF source power adjusted to 300W and the DC source power to 150W, allowing both the aluminum and scandium targets to ignite together. The inlet valve was adjusted to maintain the chamber pressure at 0.2 Pa. The tray temperature was raised to 350℃, and pre-sputtering was performed for 1–5 minutes. The baffle was then opened, followed by formal sputtering for 2 minutes, resulting in Al forming on the TiN bottom electrode. 0.77 Sc 0.23 N functional layer.

[0017] (d) Place the mask on Al 0.77 Sc 0.23 On the substrate of the N-functional layer, the cavity was evacuated to 2×10⁻⁶. -4 Argon gas is introduced into the chamber at 1 Pa, and the inlet valve is adjusted to maintain the pressure in the chamber at 1 Pa. The DC source controlling the ignition of the palladium target is turned on, and the power of the DC source is adjusted to 10 W to ignite the palladium target. Pre-sputtering is performed for 1–5 minutes; then formal sputtering is performed for 15 minutes in Al. 0.77 Sc 0.23 A Pd top electrode layer is formed on the N functional layer.

[0018] In the above preparation method, the thickness of the TiN bottom electrode layer formed in step (b) is 50 nm.

[0019] In the above preparation method, the Al formed in step (c) 0.77 Sc 0.23 The thickness of the N-functional layer is 30 nm.

[0020] In the above preparation method, the mask in step (d) is uniformly distributed with circular holes with a diameter of 100 μm.

[0021] In the above preparation method, the Pd top electrode layer in step (d) comprises several uniformly distributed layers in Al. 0.77 Sc 0.23 A circular electrode with a diameter of 100 μm on the N-functional layer; its thickness is 20 nm.

[0022] In this invention, Al, Sc, TiN, and Pd materials are commercially available products.

[0023] The ferroelectric memristor based on aluminum scandium nitride provided by this invention exhibits superior device performance due to the use of scandium-doped aluminum nitride, a wide-bandgap semiconductor. Currently, research on memristors has focused solely on aluminum nitride; this application marks the first time that scandium-doped aluminum nitride has been proposed as a self-rectifying ferroelectric memristor. A series of electrical performance tests were conducted on the memristor fabricated in this invention. Applying voltage caused the polarization of the ferroelectric material in the functional layer to reverse, altering the device's resistance. Pulse modulation tests with different parameters demonstrated the flexible tunability of the aluminum scandium nitride-based ferroelectric memristor, achieving the simulation of neuromorphic functions with excellent performance. This invention broadens the application prospects of memristors in simulating the plasticity of biological neural synapses.

[0024] Moreover, the Al-based material prepared in this invention 0.77 Sc 0.23 The N-functional-layer ferroelectric memristor possesses both ferroelectricity and self-rectification characteristics. It can solve the instability of traditional memristors and the interference encountered in the large-scale integration of ordinary ferroelectric memristors. Furthermore, the rectified ferroelectric memristor can be used for biomimetic research such as simulating the learning characteristics of neural synapses, and has great research value and application prospects. Attached Figure Description

[0025] Figure 1 This is a schematic diagram of the ferroelectric memristor based on scandium aluminum nitride provided by the present invention.

[0026] Figure 2 This is a schematic diagram of the magnetron sputtering equipment used in the fabrication of ferroelectric memristors according to the present invention.

[0027] Figure 3 These are the I-V curves of the ferroelectric memristors prepared in Embodiment 2, Comparative Example 1, and Comparative Example 2 of the present invention.

[0028] Figure 4 The Al in the ferroelectric memristor prepared in Example 2 of this invention 0.77 Sc 0.23 Topographic scan and PFM phase map of the N functional layer.

[0029] Figure 5 These are the 1000-turn I-V curve (corresponding to Figure (a)) and the 1000-turn logarithmic I-V curve (corresponding to Figure (b)) of the novel ferroelectric memristor prepared in Embodiment 2 of the present invention.

[0030] Figure 6 This is the pulse control curve of the novel ferroelectric memristor prepared in Example 2 of the present invention. Detailed Implementation

[0031] The following examples are provided to further illustrate the present invention, but they do not limit the invention in any way. Unless otherwise specified, the reagents, methods, and equipment used in this invention are conventional reagents, methods, and equipment in this technical field and do not limit the invention in any way.

[0032] Example 1

[0033] like Figure 1 As shown, the ferroelectric memristor based on aluminum scandium nitride provided by the present invention comprises, from bottom to top, a Si substrate 1, a TiN bottom electrode layer 2, and an Al... 0.77 Sc 0.23 N functional layer 3 and Pd top electrode layer 4.

[0034] The thickness of the TiN bottom electrode layer 2 is 50 nm, and the Al... 0.77 Sc 0.23 The thickness of the N-functional layer 3 is 30 nm, and the thickness of the Pd top electrode layer 4 is 20 nm. The specific fabrication process of the Pd top electrode layer 4 is as follows: several masks are placed on Al... 0.77 Sc 0.23 On the N-functional layer, Pd target sputtering is performed to form a uniform distribution on Al. 0.77 Sc 0.23 A circular electrode with a diameter of 100 μm on the N-functional layer.

[0035] Example 2

[0036] The method for fabricating a ferroelectric memristor based on aluminum scandium nitride provided by this invention includes the following steps:

[0037] (1) The Si substrate was ultrasonically cleaned, and then acetone solution was used to remove residual grease, anhydrous ethanol to remove residual acetone, and deionized water to remove residual ethanol. Silicon will oxidize in air, so the oxide layer needs to be removed before use. The Si substrate was placed in a diluted hydrofluoric acid solution (hydrofluoric acid: deionized water = 1:3) for 30 seconds to remove the SiO2 layer on the substrate. Then it was placed in deionized water to remove the residual HF solution. Finally, the substrate was taken out and dried with a high-purity nitrogen gun.

[0038] (2) Fabrication of the bottom electrode layer: using methods such as Figure 2 The magnetron sputtering equipment shown is, for example Figure 2As shown, open the magnetron sputtering equipment cavity, remove tray 7, and use sandpaper to polish the impurities on the first target stage 5 and tray 7. Clean the organic matter adhering to the surface of the pressing stage with acetone, and finally wipe it clean with alcohol. Apply silver paste to tray 7, place the cleaned Si substrate on tray 7, and place the Si substrate flat on the silver paste-coated area and fix it with a pressing device to ensure uniform film growth during sputtering. Place tray 7 on the sample stage 8 and rotate it to fix it. Place the titanium nitride target on the first target stage 5 and fix it, attach the target sleeve, close the cavity, and evacuate the cavity to a vacuum level of 2 × 10⁻⁶. -4 Argon gas is introduced into the chamber as the sputtering gas. The inlet valve 10 is adjusted to maintain the chamber pressure at 0.8 Pa. The RF source controlling the TiN ignition is turned on, and the RF source power is adjusted to 15 W to ignite the TiN target. Pre-sputtering is performed for 2 minutes. Pre-sputtering is to clean the target surface, so a baffle is needed to block the Si substrate during pre-sputtering to prevent the formation of unwanted films on the substrate. Afterwards, the baffle is opened (there is a baffle on tray 7). Figure 2 (No markings are provided) After 1 hour of sputtering, a TiN bottom electrode layer with a thickness of 50 nm was formed on the Si substrate.

[0039] (3) Preparation of the functional layer: Atmospheric atmosphere is introduced through the inlet valve 10, the magnetron sputtering equipment cavity is opened, the first target stage 5 and the second target stage 6 are taken out, impurities are sanded off with sandpaper, organic matter adhering to the surface is wiped off with acetone, and finally wiped clean with alcohol. Replace the target material, fix the 99.99% pure aluminum target material to the first target stage 5 and install the target sleeve, fix the 99.99% pure scandium target material to the second target stage 6 and install the target sleeve. After fixing, close the cavity and evacuate the cavity pressure to 2×10. -4 Pa.

[0040] Aluminum and scandium targets were used to grow aluminum scandium nitride thin films via nitrogen reactive magnetron sputtering. Nitrogen gas was introduced into the cavity through the gas filling valve 9. The radio frequency (RF) source and DC source were turned on. The RF source controlled the ignition of the aluminum target, and the DC source controlled the ignition of the scandium target. The power of the RF source was adjusted to 300W and the power of the DC source was adjusted to 150W to ignite the Al and Sc targets respectively. The gas inlet valve 10 was adjusted to maintain the pressure in the cavity at 0.5Pa. The temperature of tray 7 was raised to 350℃, and pre-sputtering was performed for 5 minutes. The baffle was opened, and formal sputtering was performed for 2 minutes, forming a 30nm thick aluminum scandium nitride functional layer on the TiN bottom electrode layer.

[0041] XPS analysis of the prepared aluminum scandium nitride functional layer revealed an Al:Sc ratio of 0.77:0.23. Therefore, the aluminum scandium nitride functional layer prepared in this embodiment is Al... 0.77 Sc 0.23 N functional layer.

[0042] (4) Preparation of the top electrode (Pd) layer: Atmosphere is introduced through the air inlet valve 10, the magnetron sputtering equipment cavity is opened, and the grown Al layer is removed. 0.77 Sc 0.23 For the N-functional layer, the mask was cleaned using ultrasonic cleaning, followed by acetone cleaning to remove organic matter adhering to the mask surface, and finally wiped clean with alcohol. Then, the grown Al... 0.77 Sc 0.23 A mask with a 100-micrometer diameter circular aperture is attached to the N-functional layer. After the electrode layer is sputtered, the area of ​​the circular aperture is the size of the effective working area of ​​this non-volatile memory.

[0043] Remove the second target stage 6 and sand away any impurities. Wipe away any organic matter adhering to the surface with acetone, and finally wipe it clean with alcohol. Replace the target material, fix the palladium target material in the second target stage 6, install the target sleeve, and evacuate the cavity to 2×10⁻⁶. -4 Argon gas was introduced into the chamber at 1 Pa, and the inlet valve was adjusted to maintain the pressure in the chamber at 1 Pa. The DC source controlling the palladium target ignition was turned on, and the power of the DC source was adjusted to 10 W to ignite the palladium target. Pre-sputtering was performed for 3 minutes; then formal sputtering was performed for 15 minutes in Al. 0.77 Sc 0.23 A 20nm thick Pd top electrode layer is formed on the N functional layer. The Pd top electrode layer is the circular pattern that is not masked on the photomask.

[0044] Comparative Example 1

[0045] Compared to Example 2, in this comparative example, during the preparation of the aluminum scandium nitride functional layer, the power of the DC source used for sputtering the scandium target was adjusted to 50W during the magnetron sputtering growth of the aluminum scandium nitride thin film. All other steps were the same as in Example 2.

[0046] Comparative Example 2

[0047] Compared to Example 2, in this comparative example, during the preparation of the aluminum scandium nitride functional layer, the power of the DC source used for sputtering the scandium target was adjusted to 250W during the magnetron sputtering growth of the aluminum scandium nitride thin film. All other steps were the same as in Example 2.

[0048] Performance testing

[0049] The ferroelectric memristors prepared in Example 2, Comparative Example 1, and Comparative Example 2 were subjected to IV curve testing, and the results are as follows: Figure 3 As shown. By Figure 3 It can be seen that the Al-based material prepared in Example 2 0.77 Sc 0.23The ferroelectric memristor with the N-functional layer exhibits excellent self-rectification characteristics, while the sample in Comparative Example 1 lacks these characteristics. The sample in Comparative Example 2 has a small window, preventing it from storing more resistive states. In Comparative Examples 1 and 2, the power of the DC source used for sputtering the scandium target was adjusted, resulting in a difference in the aluminum to scandium ratio in the final aluminum nitride scandium functional layer compared to Example 2. This directly affects the performance of the final sample.

[0050] The Al prepared in Example 2 0.77 Sc 0.23 The N-functional layer was characterized using atomic force microscopy (AFM). The morphology of the sample scanned in a 1 μm × 1 μm region is shown in the image below. Figure 4 As shown in (a), Al can be observed. 0.77 Sc 0.23 The surface of the N film is relatively flat. The PFM phase diagram is shown below. Figure 4 As shown in (b), the PFM image reveals that when a forward bias of +6V is applied to the device, Al 0.77 Sc 0.23 The N-film is downward polarized. When a negative bias of -6V is applied during scanning, Al... 0.77 Sc 0.23 N thin films are upward polarized.

[0051] The current-voltage characteristic curves were measured by applying a scanning voltage to the non-volatile ferroelectric memory prepared in Example 2. The results are shown in [Figure 1]. Figure 5 . Figure 5 In the graph, (b) is the plot of the logarithm of the ordinate of (a). From... Figure 5 It can be seen that when the device is continuously scanned with positive (0 to 8V) and negative (0 to -8V) voltages, the device conductivity continuously increases and decreases with voltage scanning (increasing in the positive direction and decreasing in the negative direction), and exhibits good self-rectification characteristics in the negative direction. From Figure 5 As can be seen, the memristor prepared based on this method has stable bioneural simulation characteristics.

[0052] The neural synapse simulation function of the device prepared in Example 2 was tested, and the results are shown in [Figure 2]. Figure 6 This demonstrates that the device prepared by the present invention has biological synaptic characteristics with synaptic weight changes, exhibiting good neuro-biomimetic effects.

[0053] Figure 6 It involves continuously applying a positive square wave for 50 cycles to the device. Figure 6(a) shows the variation between the current flowing through the device and the number of pulses, with a pulse amplitude of 6V and a fixed pulse width and pulse interval of 0.5μs; (b) shows the variation between the current flowing through the device and the pulse amplitude, with 50 pulses and a fixed pulse width and pulse interval of 0.5μs; (c) shows the variation between the current flowing through the device and the pulse width, with 50 pulses and a fixed pulse amplitude and pulse interval of 6V and 0.5μs; (d) shows the variation between the current flowing through the device and the pulse interval, with 50 pulses and a fixed pulse amplitude and pulse width of 6V and 0.5μs. Figure 6 As can be seen, the current passing through the device can be effectively controlled by changing parameters such as the number of pulses, pulse amplitude, pulse width, and pulse interval.

[0054] The novel high-performance resistive switching memory fabricated in this invention can be represented by the structure of Pd / Al. 0.77 Sc 0.23 N / TiN / Si exhibits typical bipolar non-volatile memory performance.

Claims

1. A ferroelectric memristor based on aluminum scandium nitride, characterized in that its... The structure includes, from bottom to top, a Si substrate, a TiN bottom electrode layer, and an Al layer. 0.77 Sc 0.23 N functional layer and Pd top electrode layer.

2. The ferroelectric memristor based on aluminum scandium nitride according to claim 1, characterized in that, The thickness of the TiN bottom electrode layer is 50 nm.

3. The ferroelectric memristor based on aluminum scandium nitride according to claim 1, characterized in that, The Al 0.77 Sc 0.23 The thickness of the N functional layer is 30 nm.

4. The ferroelectric memristor based on aluminum scandium nitride according to claim 1, characterized in that, The thickness of the Pd top electrode layer is 20 nm.

5. A method for fabricating a ferroelectric memristor based on aluminum scandium nitride, characterized in that, Includes the following steps: (a) Pretreatment of the Si substrate; (b) Fix the Si substrate onto the substrate stage of the magnetron sputtering equipment cavity, evacuate the cavity, introduce argon gas into the cavity as the sputtering gas, turn on the radio frequency source that controls the TiN target to ignite, adjust the power of the radio frequency source to ignite the TiN target, pre-sputter for 1~5 minutes, and then perform formal sputtering to form a TiN bottom electrode layer on the Si substrate. (c) Fix the aluminum target to the RF source and the scandium target to the DC source. Fix the substrate with the TiN bottom electrode layer to the substrate stage of the magnetron sputtering equipment cavity. Evacuate the cavity and introduce nitrogen gas. Turn on the RF source controlling the ignition of the aluminum target and the DC source controlling the ignition of the scandium target. Adjust the RF source power to 300W and the DC source power to 150W to make both the aluminum and scandium targets ignite. Pre-sputter for 1-5 minutes, open the baffle, and then perform formal sputtering. Al is formed on the TiN bottom electrode layer. 0.77 Sc 0.23 N functional layer; (d) Place the mask on the surface where Al is formed. 0.77 Sc 0.23 On the N-functional layer substrate, the cavity was evacuated, argon gas was introduced into the cavity, the DC source controlling the palladium target ignition was turned on, the DC source was adjusted to ignite the palladium target, pre-sputtering was performed for 1-5 minutes, followed by formal sputtering on Al. 0.77 Sc 0.23 A Pd top electrode layer is formed on the N functional layer.

6. The method for fabricating a ferroelectric memristor based on aluminum scandium nitride according to claim 5, characterized in that, Step (a) specifically involves: cleaning the Si substrate sequentially with acetone, alcohol, and deionized water using ultrasonic cleaning; then cleaning the Si substrate in a diluted hydrofluoric acid solution to remove the SiO2 layer on the Si substrate; then cleaning it in deionized water to remove any residual HF solution; and finally removing it and drying it with a nitrogen gun.

7. The application of the aluminum scandium nitride-based ferroelectric memristor according to any one of claims 1 to 4 in neural synapse bionics.

Images