CN116056553A - Ferroelectric memristor based on aluminum scandium nitride, preparation method and application thereof - Google Patents

Abstract

The invention provides a ferroelectric memristor based on aluminum scandium nitride, a preparation method and application thereof. The structural form of the memristor provided by the invention can be expressed as Pd/Al _0.77 Sc _0.23 N/TiN/Si. The preparation method of the memristor comprises the following steps: growing a TiN bottom electrode on the Si substrate, and growing Al on the TiN bottom electrode by a nitrogen reaction magnetron sputtering method _0.77 Sc _0.23 An N functional layer; at Al _0.77 Sc _0.23 And growing a palladium electrode layer on the N functional layer. The invention uses aluminum scandium nitride as the intermediate functional layer of the nonvolatile memory for the first time, has stronger self-rectification characteristic, can better solve the problems of cross talk and the like when being applied to a cross array, and simulates synapsesThe novel biological synapse function is developed and realized in a neuron system, the performance is good, and the novel biological synapse function is a nonvolatile memory with good storage performance, low starting voltage and wider application prospect, wherein the memory is used for simulating the biological synapse function.

Description

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

Technical Field

The invention relates to the technical field of memories, in particular to a ferroelectric memristor based on aluminum scandium nitride, a preparation method and application thereof.

Background

Memristors, collectively known as memristors, are circuit devices that represent a magnetic flux versus charge relationship. Memristors have the dimension of resistance, but unlike resistance, the resistance of a memristor is determined by the charge flowing through it. Therefore, by measuring the resistance of the memristor, the amount of charge flowing through the memristor can be known, and thus the memristor has the function of memory charge. The earliest person proposing the memristor concept was the scientist Cai Shaotang of chinese. Briefly, a memristor is a nonlinear resistor with a memory function. The resistance value can be changed by controlling the change of the current, and if the high resistance value is defined as '1', and the low resistance value is defined as '0', the resistance can realize the function of storing data. In fact, the non-linear resistor with the memory function. When the current is the water quantity passing through the common water pipe and the resistor is the thickness of the water pipe, the water pipe becomes thicker along with the water flow when the water flows from one direction, and the thickness of the water pipe is kept unchanged if the water flow is turned off; conversely, when water flows in the opposite direction, the water pipe becomes thinner and thinner. Such a component is referred to as a memristor because it "remembers" the amount of current before. Later in 2008, researchers at the hewlett-packard laboratory fabricated the first memrist device, after which the scientific community has turned on the research of memristors hot.

Memristors are a two-terminal device and therefore can be integrated at high density. In addition, it transitions very quickly between different resistance states, compatible with CMOS processes. Due to these characteristics of memristors, they have unique advantages in information storage and neurosynamics simulation.

Memristors are used as non-volatile impedance memories, and are one of the powerful candidates for next-generation high-density memories because of their advantages of simple structure, fast access speed, low power consumption, easy integration, etc., and have been widely studied. Resistive random access memories are typically sandwich structures based on top electrode-dielectric layer-bottom electrode. The data storage method mainly utilizes the reversible transition phenomenon between high and low resistance states of the intermediate dielectric layer under the action of different electric excitations to store data.

Memristors can also be used for logic computation, which was announced again by the hewlett-packard laboratory in 2010, and this discovery shakes the computer community. The national defense science and technology university researchers who have led to develop the 'Tianhe' series supercomputer consider that all the current digital logic circuits can be completely replaced by memristors theoretically after tracking and investigation.

Conventional memristors operate by voltage induced material displacement, but material displacement inside conventional memristors is extremely unstable. The ferroelectric material is used as a functional layer of the memristor, namely the ferroelectric memristor, the change of the resistance is determined by the polarization inversion of a ferroelectric film of the ferroelectric memristor, and the ferroelectric memristor has the characteristics of stable performance, low power consumption, flexible plasticity and the like, so that the defect of unstable displacement of substances in the traditional memristor is overcome, and in a tunnel junction with a ferroelectric potential barrier, the change of the resistance of a ferroelectric tunnel junction is caused by the polarization inversion of the ferroelectric film. However, most ferroelectric memristors are unipolar or bipolar, and when the integration density is gradually increased, obvious crosstalk phenomenon exists between devices.

Disclosure of Invention

The invention aims to provide a ferroelectric memristor based on aluminum scandium nitride, a preparation method and application thereof, so as to solve the problem that crosstalk phenomenon can be formed between devices when the integration density of the existing ferroelectric memristor is increased, and the ferroelectric memristor has a nerve synapse bionic function.

The invention is realized in the following way:

the ferroelectric memristor based on aluminum scandium nitride provided by the invention has the structure that a TiN bottom electrode layer and Al are sequentially formed on a Si substrate _0.77 Sc _0.23 An N functional layer and a Pd top electrode layer.

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

Preferably, the Al is _0.77 Sc _0.23 The thickness of the N functional layer was 30nm.

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

The preparation method of the ferroelectric memristor based on aluminum scandium nitride comprises the following steps:

(a) Sequentially cleaning Si substrate with ultrasonic wave in acetone, alcohol and deionized water, and placing Si substrate into hydrogenCleaning in hydrofluoric acid diluent (hydrofluoric acid: deionized water=1:3) for 30 seconds to remove SiO on Si substrate ₂ And (3) cleaning the layer in deionized water to remove residual HF solution, and finally taking out the layer and drying the substrate by a nitrogen gun.

(b) Fixing Si substrate on sample stage of magnetron sputtering equipment cavity, and evacuating back of cavity to 2×10 ^-4 Argon is introduced into the cavity below Pa and used as sputtering gas, an air inlet valve is adjusted to enable the pressure in the cavity to be maintained at 0.8Pa, a radio frequency source for controlling the ignition of TiN is opened, the power of the radio frequency source is adjusted to 15W, so that the target TiN is ignited, and sputtering is performed for 1-5 min; and finally, formally sputtering for 1h, and forming a TiN bottom electrode layer on the Si substrate.

(c) Fixing an aluminum target with purity of 99.99% on a radio frequency source, fixing a scandium target with purity of 99.99% on a direct current source, fixing a TiN substrate on a sample stage of a cavity of a magnetron sputtering device, and vacuumizing the cavity to 2×10 ^-4 Pa, introducing nitrogen into the cavity, taking the nitrogen as sputtering gas and reaction gas, opening a radio frequency source for controlling the starting of the aluminum target and a direct current source for controlling the starting of the scandium target, adjusting the power of the radio frequency source to 300W, adjusting the power of the direct current source to 150W, starting the aluminum target and the scandium target together, adjusting an air inlet valve to maintain the pressure in the cavity at 0.2Pa, raising the temperature of the tray to 350 ℃, pre-sputtering for 1-5 min, opening a baffle plate, performing formal sputtering for 2min, and forming Al on a TiN bottom electrode _0.77 Sc _0.23 And an N functional layer.

(d) Placing the mask plate on Al _0.77 Sc _0.23 Vacuum-pumping the cavity to 2×10 on the substrate with N functional layer ^-4 Pa, introducing argon into the cavity, adjusting an air inlet valve to maintain the pressure in the cavity at 1Pa, opening a direct current source for controlling the ignition of the palladium target, adjusting the power of the direct current source to be 10W, and pre-sputtering the palladium target for 1-5 min; then formally sputtering for 15min, at Al _0.77 Sc _0.23 And forming a Pd top electrode layer on the N functional layer.

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

In the above preparation method, al is formed in the step (c) _0.77 Sc _0.23 The thickness of the N functional layer was 30nm.

In the preparation method, circular holes with the diameter of 100 mu m are uniformly distributed on the mask plate in the step (d).

In the above preparation method, the Pd top electrode layer in step (d) includes a plurality of uniformly distributed Al _0.77 Sc _0.23 A circular electrode with a diameter of 100 μm on the N functional layer; the thickness thereof was 20nm.

The Al, sc, tiN, pd material belongs to commercial products.

The ferroelectric memristor based on aluminum scandium nitride provided by the invention has excellent device performance of wide bandgap semiconductor aluminum nitride scandium doped, and only aluminum nitride is used as the research of the memristor at present, and the ferroelectric memristor with aluminum nitride scandium doped as the self-rectifying characteristic is first proposed in the application. For the memristor prepared by the method, a series of electrical property tests are carried out, and voltage is applied to enable the polarization of the ferroelectric material of the functional layer to be inverted, so that the resistance of the device is changed. The ferroelectric memristor flexible adjustable capacity based on aluminum scandium nitride is shown through pulse adjustment tests of different parameters, the simulation of the nerve bionic function is realized, and the performance is good. The invention has wider application prospect in simulating the plasticity performance of biological nerve synapses.

Furthermore, the Al-based alloy prepared by the invention _0.77 Sc _0.23 The ferroelectric memristor with the N functional layer has ferroelectricity and self-rectification characteristic, can solve the unstable characteristic of the traditional memristor, can also solve the interference encountered by the large-scale integration of the common ferroelectric memristor, can simulate bionic researches such as the learning characteristic of nerve synapses and the like, and has great research value and application prospect.

Drawings

FIG. 1 is a schematic diagram of a ferroelectric memristor based on scandium aluminum nitride provided by the present invention.

FIG. 2 is a schematic diagram of a magnetron sputtering apparatus used in fabricating a ferroelectric memristor according to the present invention.

FIG. 3 is an I-V plot of ferroelectric memristors prepared in example 2, comparative example 1, and comparative example 2 of the present invention.

FIG. 4 is a diagram of Al in a ferroelectric memristor prepared in example 2 of the present invention _0.77 Sc _0.23 Morphology scan and PFM phase map of the N functional layer.

FIG. 5 is a 1000-turn I-V plot (corresponding to FIG. (a)) and a 1000-turn logarithmic I-V plot (corresponding to FIG. (b)) of the novel ferroelectric memristor prepared in example 2 of the present disclosure.

FIG. 6 is a pulse modulation curve of a novel ferroelectric memristor prepared in example 2 of the present disclosure.

Detailed Description

The following examples serve to further illustrate the invention in detail, but are not intended to limit the invention in any way. The reagents, methods and apparatus employed in the present invention are those conventional in the art and are not intended to limit the invention in any way unless otherwise specified.

Example 1

As shown in FIG. 1, the ferroelectric memristor based on aluminum scandium nitride provided by the invention comprises a Si substrate 1, a TiN bottom electrode layer 2 and Al which are sequentially arranged from bottom to top _0.77 Sc _0.23 An N functional layer 3 and a Pd top electrode layer 4.

Wherein the thickness of the TiN bottom electrode layer 2 is 50nm, al _0.77 Sc _0.23 The thickness of the N functional layer 3 is 30nm, and the thickness of the Pd top electrode layer 4 is 20nm. The preparation process of the Pd top electrode layer 4 specifically comprises the following steps: using several masks to put on Al _0.77 Sc _0.23 Sputtering Pd target material on the N functional layer to form a uniform Al distribution _0.77 Sc _0.23 A circular electrode with a diameter of 100 μm on the N functional layer.

Example 2

The preparation method of the ferroelectric memristor based on aluminum scandium nitride provided by the invention comprises the following steps:

(1) And sequentially removing residual grease from the Si substrate by using an acetone solution, removing residual acetone from the Si substrate by using absolute ethyl alcohol and removing an ethanol residual liquid from the Si substrate by using deionized water in an ultrasonic cleaning mode. Silicon oxidizes in air, so the oxide layer is removed before use, and the Si substrate is cleaned in a hydrofluoric acid diluent (hydrofluoric acid: deionized water=1:3)Washing for 30 seconds to remove SiO on the substrate ₂ And (3) cleaning the layer in deionized water to remove residual HF solution, and finally taking out the layer and drying the substrate by a high-purity nitrogen gun.

(2) Preparing a bottom electrode layer: the magnetron sputtering equipment shown in fig. 2 is adopted, the cavity of the magnetron sputtering equipment is opened, the tray 7 is taken out, the first target table 5 and impurities on the tray 7 are polished by sand paper, the organic matters attached to the surface of the tabletting table are cleaned by acetone, and finally the organic matters are wiped clean by alcohol. Silver colloid is coated on a tray 7, the cleaned Si substrate is placed on the tray 7, the Si substrate is flatly placed at the place coated with the silver colloid and is fixed by a pressing sheet, and the substrate is flattened to ensure that a film grows uniformly during sputtering. Placing the tray 7 on the sample table 8, fixing by rotation, placing the titanium nitride target on the first target table 5, fixing with the target sleeve, closing the cavity, and vacuumizing the cavity to 2×10 ^-4 Pa, argon is introduced into the cavity as sputtering gas, the air inlet valve 10 is adjusted to maintain the pressure in the cavity at 0.8Pa, a radio frequency source for controlling the ignition of TiN is turned on, the power of the radio frequency source is adjusted to 15W, the target TiN is ignited, the sputtering is performed for 2min, and the surface of the target is cleaned, so that a Si substrate is blocked by a baffle plate during the pre-sputtering, and an unwanted film layer is not formed on the substrate. Thereafter, the shutter (the shutter on the tray 7 is not marked in fig. 2) was opened, and formally sputtered for 1h to form a TiN bottom electrode layer having a thickness of 50nm on the Si substrate.

(3) Preparation of the functional layer: and (3) introducing the air into the magnetron sputtering equipment through the air inlet valve 10, opening the cavity of the magnetron sputtering equipment, taking out the first target table 5 and the second target table 6, polishing impurities by sand paper, wiping organic matters attached to the surfaces by acetone, and finally wiping the organic matters by alcohol. Changing the target, fixing an aluminum target with the purity of 99.99% on a first target table 5, installing a target sleeve, fixing a scandium target with the purity of 99.99% on a second target table 6, installing the target sleeve, closing a cavity after fixing, and vacuumizing the cavity to 2X 10 ^-4 Pa。

And growing an aluminum scandium nitride film by using the aluminum target and the scandium target through nitrogen reaction magnetron sputtering. Nitrogen is introduced into the cavity through the inflation valve 9; the method comprises the steps of opening a radio frequency source and a direct current source, controlling an aluminum target to start by the radio frequency source, controlling a scandium target to start by the direct current source, adjusting the power of the radio frequency source to 300W, adjusting the power of the direct current source to 150W, enabling the Al target and the Sc target to start respectively, adjusting an air inlet valve 10 to enable the pressure in a cavity to be maintained at 0.5Pa, raising the temperature of a tray 7 to 350 ℃, pre-sputtering for 5 minutes, opening a baffle plate, formally sputtering for 2 minutes, and forming an aluminum scandium nitride functional layer with the thickness of 30nm on a TiN bottom electrode layer.

XPS test is carried out on the prepared aluminum scandium nitride functional layer to obtain the proportion of Al and Sc as 0.77:0.23, so the aluminum scandium nitride functional layer prepared by the embodiment is Al _0.77 Sc _0.23 And an N functional layer.

(4) Top electrode (Pd) layer preparation: the air is introduced through the air inlet valve 10, the cavity of the magnetron sputtering equipment is opened, and the grown Al is taken out _0.77 Sc _0.23 An N functional layer, using ultrasonic to clean the mask, using acetone to clean the organic matters attached to the surface of the mask, using alcohol to wipe the organic matters clean, and growing Al _0.77 Sc _0.23 And a mask plate with a round hole with the diameter of 100 microns is stuck on the N functional layer. The circular hole area is the size of the effective working area of the nonvolatile memory after the electrode layer is sputtered.

The second target table 6 is taken out, the impurities are polished by sand paper, the organic matters attached to the surface are wiped by acetone, and finally the organic matters are wiped by alcohol. Changing target material, fixing palladium target material on second target table 6, installing target sleeve, and vacuumizing cavity to 2×10 ^-4 Pa, introducing argon into the cavity, adjusting an air inlet valve to maintain the pressure in the cavity at 1Pa, opening a direct current source for controlling the ignition of the palladium target, adjusting the power of the direct current source to be 10W, and pre-sputtering the palladium target for 3min; then formally sputtering for 15min, at Al _0.77 Sc _0.23 And forming a Pd top electrode layer with the thickness of 20nm on the N functional layer. The Pd top electrode layer is a circular pattern which is not shielded by the mask.

Comparative example 1

Compared with the example 2, in the preparation of the aluminum scandium nitride functional layer, the power of the direct current source used in the sputtering of the scandium target material is adjusted to be 50W in the process of growing the aluminum scandium nitride thin film by magnetron sputtering. The rest of the procedure was the same as in example 2.

Comparative example 2

Compared with the example 2, in the preparation of the aluminum scandium nitride functional layer, the power of the direct current source used in the sputtering of the scandium target material is adjusted to be 250W in the process of growing the aluminum scandium nitride thin film by magnetron sputtering. The rest of the procedure was the same as in example 2.

Performance testing

The ferroelectric memristors prepared in example 2, comparative example 1 and comparative example 2 were subjected to I-V curve test, and the results are shown in fig. 3. As can be seen from FIG. 3, the Al-based alloy prepared in example 2 _0.77 Sc _0.23 The ferroelectric memristor with the N functional layer has good self-rectifying characteristic, but the sample in the comparative example 1 does not have the self-rectifying characteristic, and the window of the sample in the comparative example 2 is small, so that more resistance states cannot be stored. Since the power of the dc source used to sputter the scandium target was adjusted in comparative examples 1 and 2, the ratio of aluminum to scandium in the finally formed aluminum scandium nitride functional layer was different from that of example 2, which would directly affect the performance of the final sample.

Al prepared in example 2 _0.77 Sc _0.23 The N functional layer was subjected to a topography characterization test using an Atomic Force Microscope (AFM), and a topography of a sample scanned over a 1 μm by 1 μm region is shown in FIG. 4 (a). It can be observed that Al _0.77 Sc _0.23 The surface of the N film is relatively flat. The PFM phase diagram is shown in FIG. 4 (b), from which it can be seen that Al is present when a forward bias of +6V is applied to the device _0.77 Sc _0.23 The N film is polarized downward. After scanning with a negative bias of-6V, al _0.77 Sc _0.23 The N film is polarized upward.

The current-voltage characteristic curve was measured by applying a scanning voltage to the nonvolatile ferroelectric memory prepared in example 2, and the result is shown in fig. 5. In fig. 5, (b) is a graph obtained by taking the log of the ordinate of (a). As can be seen from fig. 5, when the device is continuously scanned with positive (0 to 8V) and negative (0 to-8V) voltages for the scan voltage, the device conductivity continuously increases and decreases (positive increase, negative decrease) with the voltage scan, and the negative direction has good self-rectifying characteristics. As can be seen from fig. 5, the memristor prepared based on the method has stable biological nerve simulation characteristics.

The results of testing the nerve synapse simulation function of the device prepared in example 2 are shown in fig. 6, which shows that the device prepared in the invention has the biological synapse characteristics with the change of synaptic weights, and shows good nerve bionic effect.

FIG. 6 is a forward square wave applied to the device for 50 consecutive turns, and FIG. 6 (a) shows the change between the current passed by the device and the number of pulses, with a pulse amplitude of 6V and a pulse width and pulse spacing fixed at 0.5 μs; (b) The change of the current and the pulse amplitude of the device is shown, the number of the pulses is 50, and the pulse width and the pulse interval are fixed to be 0.5 mu s; (c) The change between the current passing through the device and the pulse width is shown, the number of the pulses is 50, and the pulse amplitude and the pulse interval are fixed to be 6V and 0.5 mu s; (d) The variation between the current passed by the device and the pulse interval is shown, with 50 pulses, and the pulse amplitude and pulse width fixed at 6V and 0.5 mus. As can be seen from fig. 6, the current passing through the device can be effectively controlled by changing the parameters of the pulse number, the pulse amplitude, the pulse width, the pulse interval and the like.

The structural form of the novel high-performance resistive random access memory prepared by the invention can be expressed as Pd/Al _0.77 Sc _0.23 N/TiN/Si, exhibits typical bipolar nonvolatile memory performance.

Claims (7)

1. A ferroelectric memristor based on aluminum scandium nitride is characterized by comprising a Si substrate, a TiN bottom electrode layer and Al which are sequentially arranged from bottom to top _0.77 Sc _0.23 An N functional layer and a Pd top electrode layer.

2. The aluminum scandium nitride based ferroelectric memristor according to claim 1, wherein the TiN bottom electrode layer has a thickness of 50nm.

3. The aluminum scandium nitride based ferroelectric memristor according to claim 1, wherein the AlScN functional layer has a thickness of 30nm.

4. The aluminum scandium nitride based ferroelectric memristor according to claim 1, wherein the Pd top electrode layer has a thickness of 20nm.

5. A preparation method of a ferroelectric memristor based on aluminum scandium nitride is characterized by comprising the following steps:

(a) Preprocessing a Si substrate;

(b) Fixing a Si substrate on a substrate table of a cavity of a magnetron sputtering device, vacuumizing, introducing argon into the cavity as sputtering gas, opening a radio frequency source for controlling the starting of a TiN target, adjusting the power of the radio frequency source to ensure that the TiN target starts, pre-sputtering for 1-5 min, and performing formal sputtering to form a TiN bottom electrode layer on the Si substrate;

(c) Fixing an aluminum target on a radio frequency source, fixing a scandium target on a direct current source, fixing a substrate with a TiN bottom electrode layer on a substrate table of a cavity of a magnetron sputtering device, vacuumizing, introducing nitrogen into the cavity, opening the radio frequency source for controlling the starting of the aluminum target and the direct current source for controlling the starting of the scandium target, adjusting the power of the radio frequency source to 300W, adjusting the power of the direct current source to 150W, starting both the aluminum target and the scandium target, pre-sputtering for 1-5 min, opening a baffle plate, performing formal sputtering, and forming Al on the TiN bottom electrode layer _0.77 Sc _0.23 An N functional layer;

(d) Placing the mask plate on the surface where Al is formed _0.77 Sc _0.23 Vacuumizing the cavity on the substrate with the N functional layer, introducing argon into the cavity, opening a direct current source for controlling the ignition of the palladium target, adjusting the direct current source to ignite the palladium target, pre-sputtering for 1-5 min, performing formal sputtering, and performing Al-sputtering on the substrate _0.77 Sc _0.23 And forming a Pd top electrode layer on the N functional layer.

6. The method for fabricating a scandium aluminum nitride based ferroelectric memristor according to claim 5, wherein step (a) specifically comprises: sequentially cleaning the Si substrate in acetone, alcohol and deionized water by ultrasonic waves, and then cleaning the Si substrate in hydrofluoric acid diluent to remove SiO on the Si substrate ₂ The layer is then put into deionized water to be cleaned to remove residual HF solution, and finally taken out and dried by a nitrogen gun.

7. Use of the aluminum scandium nitride based ferroelectric memristor according to any one of claims 1 to 4 in neurosynamics.

Images