阅读说明：本技术 电阻式存储器元件累进电阻特性的控制方法 (Control method for progressive resistance characteristic of resistive memory element ) 是由 蒋光浩 林榆瑄 于 2019-02-21 设计创作，主要内容包括：一种电阻式存储器累进电阻特性的控制方法,包括：对一电阻式存储器元件施加一第一写入脉冲集合,以获取一参考累进电阻分布；根据参考累进电阻分布对电阻式存储器元件施加第二写入脉冲集合,使电阻式存储器元件具有一预设累进电阻分布。(A method of controlling a progressive resistance characteristic of a resistive memory, comprising: applying a first set of write pulses to a resistive memory element to obtain a reference progressive resistance distribution; applying a second set of write pulses to the resistive memory element according to the reference progressive resistance distribution to provide the resistive memory element with a predetermined progressive resistance distribution.)

1. A method of controlling a progressive resistance characteristic of a resistive memory, comprising:

applying a first write pulse set to a resistive memory element to obtain a reference progressive resistance distribution; and

applying a second set of write pulses to the resistive memory element according to the reference progressive resistance distribution to provide the resistive memory element with a predetermined progressive resistance distribution.

2. The method of claim 1, wherein the first set of write pulses comprises a plurality of first write pulses; the second set of write pulses comprises a plurality of second write pulses; at least one of the first write pulses is different from a corresponding one of the second write pulses.

3. The method as claimed in claim 1, wherein the predetermined progressive resistance distribution is a linear function; the reference progressive resistance distribution is an exponential function.

4. A method of controlling a progressive resistance characteristic of a resistive memory, comprising:

forming a resistive memory element, such that the resistive memory element comprises:

a first metal layer;

a second metal layer;

a first metal oxide layer disposed on the first metal layer

And the second metal layer; and

an oxygen content control layer disposed between the first metal oxide layer and one of the first metal layer and the second metal layer; and

an oxygen content distribution of the oxygen content control layer is adjusted such that the resistive memory element has a predetermined progressive resistance distribution when a predetermined set of write pulses is applied.

5. The method as claimed in claim 4, wherein the oxygen content distribution is gradually increased away from the first metal oxide layer toward one of the first metal layer and the second metal layer.

6. The method as claimed in claim 5, wherein the oxygen content distribution is gradually increased along the direction to reach a first local maximum; gradually decreasing and then gradually increasing to a second local maximum.

7. The method as claimed in claim 5, wherein the resistive memory element further comprises a second metal oxide layer disposed between the other of the first metal layer and the second metal layer and the oxygen content control layer.

8. The method of claim 5, wherein the oxygen content distribution is gradually increased at a decreasing rate, an increasing rate, or an average rate.

9. The method as claimed in claim 4, wherein the oxygen content control layer is formed by a deposition process.

10. The method of claim 4, wherein the step of adjusting the oxygen content distribution comprises providing a thermal stress to drive oxygen atoms in the metal oxide layer into the oxygen content control layer.

Technical Field

The present invention relates to a method for controlling a Resistive Random-Access Memory (ReRAM) device, and more particularly, to a method for controlling a progressive resistance characteristic of a Resistive Memory device.

Background

Resistive memory elements, such as resistive random access memory elements, are based on the interpretation of information storage states such as "0" and "1" by applying a pulse voltage to the metal oxide film of the memory element to generate a resistance difference.

A typical resistive memory device includes a vertically stacked lower Metal-Insulator-Metal (MIM) stack structure, which can be used to realize high-density storage of a three-dimensional crossbar array (crossbar) structure. The memory layer is generally a resistance switching layer (resistance switching layer) composed of Transition Metal Oxides (TMO), and the oxidation degree of the transition metal oxides is a main factor affecting the resistance switching characteristics (resistance switching characteristics) and the operation efficiency of the resistive memory element. Binary oxides, such as titanium oxide (TiOx), are currently used as transition metal oxides for the resistance transition layer of the resistive memory element.

In recent years, with the rise of Neural Network (NN) systems of artificial intelligence and internet of things, resistive memories, also called memristors, have the advantages of high device density (device sensitivity), high power consumption, fast programming/erasing speed, and three-dimensional space stacking, can meet the requirements of electronic synapse (electronic synapse) applications, and are considered to be one of the most potential candidate devices. However, with the use of the resistive memory element with the transition metal oxide wire resistance switching (film resistive switching), the writing and erasing still have great variability; when the three-dimensional array product is manufactured, a leakage current path is easy to generate, reading errors are caused, and the application of a neural network system is not facilitated.

Therefore, there is a need to provide a neural network system and a control method thereof suitable for product terms and operations to solve the problems faced by the prior art.

Disclosure of Invention

One embodiment of the present disclosure discloses a method for controlling a progressive resistance characteristic of a resistive memory, including: applying a first set of write pulses to a resistive memory element to obtain a reference progressive resistance distribution; applying a second set of write pulses to the resistive memory element according to the reference progressive resistance distribution to provide the resistive memory element with a predetermined progressive resistance distribution.

Another embodiment of the present disclosure discloses a method for controlling a progressive resistance characteristic of a resistive memory device, comprising: forming a resistive memory element such that the resistive memory element comprises: a first metal layer, a second metal layer, a first metal oxide layer and an oxygen content control layer. Wherein the first metal oxide layer is disposed between the first metal layer and the second metal layer; the oxygen content control layer is disposed between the first metal oxide layer and at least one of the first metal layer and the second metal layer. And selectively adjusting the oxygen content distribution in the metal oxide layer to make the resistive memory element have a preset progressive resistance distribution.

According to the above embodiments, a method for controlling the progressive resistance characteristic of a resistive memory device is provided, which applies a set of first write pulses to the resistive memory device to obtain a reference progressive resistance distribution. Then, a second write Pulse set with a current, voltage or Pulse length different from that of the first write Pulse set is selected according to the reference progressive resistance distribution to perform Incremental Step Pulse Programming (ISSP) to change the progressive resistance characteristic of the resistive memory element. Or in the process of manufacturing the resistive memory element, the oxygen content distribution of the oxygen content control layer in the resistive memory element is adjusted by different process steps to have a preset oxygen content distribution. When a default set of write pulses is applied to the resistive memory element, a default progressive resistance distribution is obtained. Therefore, the writing and erasing variability of the resistive memory element is reduced, so that the Multi-Level Cell (MLC) characteristic of the resistive memory element is improved, and the resistive memory element is more suitable for the application of a neural network system.

In order to make other aspects and advantages of the present invention more comprehensible, the following drawings and detailed description are given.

Drawings

Fig. 1A is a flowchart illustrating a method for controlling the progressive resistance of a resistive memory device according to an embodiment of the present disclosure.

Fig. 1B is a graph showing a relationship between a progressive resistance and a number of pulses for writing a resistive memory element with a quantity step pulse by using three different sets of write pulses according to an embodiment of the present disclosure.

Fig. 2A is a flow chart illustrating a method for controlling the progressive resistance of a resistive memory device according to another embodiment of the present disclosure.

Fig. 2B is a cross-sectional view of a resistive memory device according to an embodiment of the present disclosure.

FIG. 2C is a graph showing the oxygen content distribution in the oxygen content control layer of the resistive memory element according to FIG. 2B.

Fig. 3A is a cross-sectional view of a resistive memory device according to another embodiment of the present disclosure.

FIG. 3B is a graph showing the oxygen content distribution in the oxygen content control layer of the resistive memory element according to FIG. 3A.

FIG. 4 is a graph showing the relationship between the number of pulses and the number of progressive resistance after writing three resistive memory elements with different oxygen content distributions by using a default set of write pulses.

[ description of reference ]

101. 102, 103, 401, 402, 403: progressive resistance-pulse number relation curve

200. 300, and (2) 300: resistive memory element

201: a first metal layer

202: second metal layer

203. 303: metal oxide layer

204. 304: oxygen content control layer

206. 207, 208: oxygen content distribution curve

S11: applying a set of write pulses to the resistive memory elements to obtain a reference progressive resistance distribution

S12: providing a second set of write pulses according to the reference progressive resistance distribution, and applying the second set of write pulses to the resistive memory element to make the resistive memory element have a predetermined progressive resistance distribution

S21: forming a resistive memory element includes: first metal layer, second metal layer, metal oxide layer and oxygen content control layer

S22: the oxygen content distribution of the oxygen content control layer is regulated and controlled through the process steps to regulate and control the progressive resistance characteristic of the resistive memory element

Detailed Description

The present specification provides a method for controlling the progressive resistance characteristics of a resistive memory element, which can reduce the writing and erasing variability of the resistive memory element, making it more suitable for the application of a neural network system. In order to make the aforementioned and other objects, features and advantages of the present invention comprehensible, several preferred embodiments accompanied with figures are described in detail below.

It should be noted, however, that the specific embodiments and methods are not to be considered as limiting the invention. The invention may be embodied with other features, elements, methods, and parameters. The preferred embodiments are provided only for illustrating the technical features of the invention, and are not intended to limit the scope of the invention. Those skilled in the art will recognize that equivalent modifications and variations can be made in light of the following description without departing from the spirit of the invention. Like elements in different embodiments and drawings will be denoted by like reference numerals.

Referring to fig. 1A and 1B, fig. 1A is a flowchart illustrating a method for controlling a progressive resistance characteristic of a resistive memory device according to an embodiment of the present disclosure. Fig. 1B is a graph showing a relationship between a progressive resistance and a number of pulses for writing a resistive memory element with a quantity step pulse by using three different sets of write pulses according to an embodiment of the present disclosure.

The control method of the progressive resistance characteristic of the resistive memory comprises the following steps: first, a set of write pulses is applied to the resistive memory elements to obtain a reference progressive resistance distribution of the resistive memory elements (step S11). Curve 101 in fig. 1B represents the cumulative resistance-pulse number relationship obtained after the resistive memory element is written with incremental step pulses using the first set of write pulses. In the present embodiment, each write pulse in the first set of write pulses has a voltage of 2.6 volts (V), a current of 200 microamperes (μ a), and a pulse width of 500 nanoseconds (ns). The curve 101 may be a logarithmic function (logarithmic function), and the progressive resistance in the curve 101 exhibits a logarithmic increase (logarithmic growth) as the number of pulses increases.

After the erase operation, a second set of write pulses is provided according to the reference progressive resistance distribution (curve 101) and applied to the resistive memory element to make the resistive memory element have a predetermined progressive resistance distribution (step S12). For example, the curve 102 in fig. 1B represents the cumulative resistance-pulse number relationship obtained by writing the resistive memory element with the second set of write pulses by the quantity step pulse. In the present embodiment, the voltages 2.0, 2.2, 2.4, 2.6, and 2.8 volts (V) in the second set of write pulses are twenty pulses each having a pulse current of 200 microamperes (μ a) and a pulse width of 500 nanoseconds. The curve 102 may be a linear function (linear function), and the value of the progressive resistance in the curve 102 increases proportionally with the number of pulses.

It should be noted that the write pulse set described herein is not limited to the electromagnetic pulse applied to the resistive memory element by the voltage current operation, and any energy that can change the resistance of the resistive memory element can be applied to the resistive memory element as the write pulse. In addition, although each write pulse in the set of write pulses has a fixed operating current and pulse width in the present embodiment, in other embodiments, each write pulse in the set of write pulses may have a different operating current and a different pulse width.

In another embodiment, after another erase operation, a third set of write pulses subsequently applied to the resistive memory element can be adjusted according to the curves 101 and 102 to make the resistive memory element have another predetermined progressive resistance distribution. For example, the curve 103 in fig. 1B represents the relationship curve of the number of progressive resistance-pulse times obtained by writing the resistive memory element with the quantity step pulse using the third set of write pulses. In the present embodiment, in the third set of write pulses, twenty pulses of voltages 1.6, 1.8, 2.0, 2.2, and 2.8 volts (V) each have a pulse current of 200 microamperes (μ a) and a pulse width of 500 nanoseconds. The curve 103 may be an exponential function (exponential function), and the value of the progressive resistance in the curve 103 increases exponentially as the number of pulses increases.

By changing at least one of the resistance, voltage and pulse width of different write pulses applied to the same resistive memory element, the distribution of the progressive resistance of the resistive memory element (i.e., the function graph of the progressive resistance of the resistive memory element versus the number of pulses) can be changed. Therefore, the writing and erasing variability of the resistive memory element can be reduced, the switching of the set (set) state and the reset (reset) state of the resistive memory element can be effectively judged, the characteristics of the multilayer memory cells of the resistive memory element are improved, and the resistive memory element is more suitable for artificial intelligence and operation of a neural network system and the Internet of things of a circuit.

For example, in the present embodiment, the resistive memory element is applied to a neural network system, and the first write pulse set, the second write pulse set, and the third write pulse set are respectively adopted to perform incremental step pulse writing on the resistive memory element, so as to perform face recognition calculation. As a result, it can be found that the operation accuracy using the first set of write pulses (causing the resistive memory element to produce the progressive resistance distribution as shown by the curve 101) is less than 30%; the computational accuracy using the second set of write pulses (to cause the resistive memory elements to produce a progressive resistance distribution as shown by curve 102) is about 70%; the accuracy of the operation using the third set of write pulses that produce a progressive resistance distribution for the resistive memory elements as shown by curve 103 is approximately 97. It can be seen that the progressive resistance distribution (as shown by the curve 103) generated by applying the third set of write pulses to the resistive memory element is more favorable for the neural network system to perform the face recognition calculation. In other words, by changing the current, voltage, or pulse length of the set of write pulses, the progressive resistance distribution of the resistive memory element can be changed, which is beneficial for the application of the resistive memory element in the neural network-like system.

It should be noted that, although the first set of write pulses, the second set of write pulses and the third set of write pulses in the foregoing embodiments have the same current and pulse width, only the voltages are different. However, the adjustment manner of the write pulse sets is not limited to this, and at least one of the current, the voltage and the pulse width of each write pulse set may be different. And the current, voltage and pulse width of each write pulse in each set of write pulses may also be the same or different from each other.

In addition, the progressive resistance characteristic of the resistive memory element can be adjusted by changing the physical structure of the resistive memory element through the modulation process step. For example, referring to fig. 2A, fig. 2A is a flow chart of a method for controlling the progressive resistance characteristic of the resistive memory device 200 according to another embodiment of the present disclosure. A method for controlling a progressive resistance characteristic of a resistive memory element, comprising the steps of: first, a resistive memory element 200 is formed, such that the resistive memory element 200 includes: a first metal layer 201, a second metal layer 202, a metal oxide layer 203, and an oxygen content control layer 204 (step S21).

Referring to fig. 2B, fig. 2B is a cross-sectional view illustrating a resistive memory device 200 according to an embodiment of the present disclosure. The first metal layer 201 faces the second metal layer 202, and the first metal layer 201 and the second metal layer 202 may be made of the same material or different materials. For example, in some embodiments of the present description, the first metal layer 201 and the second metal layer 202 may be composed of the same material. And the material constituting the first metal layer 101 and the second metal layer 102 may be tungsten (W), titanium (Ti), titanium nitride (TiN), aluminum (Al), nickel (Ni), copper (Cu), zirconium (Zr), niobium (Nb), tantalum (Ta), ytterbium (Yb), terbium (Tb), yttrium (Y), rhodium (La), (Sc), hafnium (Hf), chromium (Cr), vanadium (V), zinc (Zn), molybdenum (Mo), rhenium (Re), ruthenium (Ru), cobalt (C0), rhodium (Rh), cadmium (Pd), platinum (Pt), or an alloy of any combination thereof.

The metal oxide layer 203 is disposed between the first metal layer 202 and the second metal layer 202. In thatIn some embodiments of the present disclosure, the metal oxide layer 203, which may be an oxygen-rich barrier layer, may be formed of the chemical formula AO_xThe metal oxide compound represented by (1), wherein A is a metal selected from tungsten, titanium nitride, aluminum, nickel, copper, zirconium, niobium, tantalum, or any combination of these metals.

The oxygen content control layer 204 is disposed between at least one of the first metal layer 201 and the second metal layer 202 and the metal oxide layer 203. For example, in the present embodiment, the oxygen content control layer 204 is disposed between the first metal layer 201 and the metal oxide layer 203. Wherein the oxygen content controlling layer 204 may be formed of the chemical formula AO_xB_yThe metal oxide compound represented by (1), wherein A is a metal selected from tungsten, titanium nitride, aluminum, nickel, copper, zirconium, niobium, tantalum, or any combination of these metals. B is selected from the group consisting of nitrogen (N), silicon (Si), germanium (Ge), arsenic (As), gallium (Ga), indium (In), and phosphorus (P), and any combination thereof.

The oxygen content distribution of the oxygen content control layer 204 may be controlled by various steps of the process (i.e., adjusting the chemical formula AO in different portions of the oxygen content control layer 204)_xB_yX-y ratio) to achieve the purpose of modulating the progressive resistance characteristic of the resistive memory device 200 (step S22). For example, in some embodiments of the present description, the oxygen content control layer 204 may be formed by a Deposition method, such as a Chemical Vapor Deposition (CVD) or Physical Vapor Deposition (PVD) process. The content of oxygen atoms in different parts of the oxygen content control layer 204 can be modulated by adjusting the oxygen concentration of the reaction gas in the deposition process.

For example, referring to fig. 2C, fig. 2C is a graph illustrating oxygen content distribution curves 206, 207, and 208 in the oxygen content control layer 204 of the resistive memory element 200 according to fig. 2B. The oxygen content in the oxygen content control layer 204 gradually increases along the direction R away from the metal oxide layer 203 with increasing depth in a gradient manner until the first metal layer 201 reaches a maximum. In one embodiment of the present description, the oxygen content is increased at an increasing rate (as shown by oxygen content profile 206). In another embodiment of the present description, the oxygen content increases at a decreasing rate (as shown by oxygen content profile 208). In yet another embodiment of the present description, the oxygen content is increased at an average rate of gradual increase (as shown by oxygen content profile 207).

However, the process steps for changing the oxygen content distribution in the oxygen content control layer 204 of the resistive memory device 200 are not limited thereto. In some embodiments of the present disclosure, after the first metal layer 201, the second metal layer 202, the metal oxide layer 203, and the oxygen content control layer 204 are formed, thermal stress may be provided to the metal oxide layer 203 and the oxygen content control layer 204, for example, a thermal annealing process may be performed to the metal oxide layer 203 and the oxygen content control layer 204 to drive a plurality of oxygen atoms in the metal oxide layer 203 into the oxygen content control layer 204.

In addition, the structure of the resistance variable memory element is not limited thereto. For example, referring to fig. 3A and 3B, fig. 3A is a cross-sectional view illustrating a resistive memory element 300 according to another embodiment of the present disclosure. FIG. 3B is a graph showing an oxygen content distribution curve 301 in the oxygen content control layer 303 of the resistive memory element 300 according to FIG. 3A. The structure of the resistive memory element 300 is substantially similar to the structure of the resistive memory element 200, except that the resistive memory element 300 may further include another metal oxide layer 303 between the first metal layer 201 and the oxygen content control layer 304. In this embodiment, the metal oxide layers 203 and 303 may be made of the same material.

The oxygen content distribution in the oxygen content control layer 304 can be changed by modulating the thermal stress applied to the metal oxide layers 203 and 303 and the oxygen content control layer 304 in the resistive memory element 300. For example, in the present embodiment (as shown in fig. 3B), the oxygen content in the oxygen content control layer 304 gradually increases along the direction R away from the metal oxide layer 203 to reach the first local maximum 302; and then gradually decreases and then increases again to reach a second local maximum 304, and then gradually decreases again to the metal oxide layer 303.

Referring to fig. 4, fig. 4 is a graph showing a relationship between a progressive resistance and a number of pulses obtained after writing the three resistive memory elements with different oxygen content distributions (shown in the oxygen content distribution curves 206, 207 and 208 respectively in fig. 2C) by using the same set of write pulses. Curves 401, 402 and 403 represent the curves of the progressive resistance-pulse number relationship of the three resistive memory elements with the oxygen content distribution curves 206, 207 and 208 as shown in fig. 2C, respectively. Fig. 4 shows that by selectively adjusting the oxygen content distribution of the oxygen content control layer 204 through the process steps, the written progressive resistance characteristic of the resistive memory element can be effectively changed, which is beneficial to the application of the resistive memory element in the neural network-like system.

According to the above embodiments, a method for controlling the progressive resistance characteristic of a resistive memory device is provided, in which a set of first write pulses is applied to the resistive memory device to obtain a reference progressive resistance distribution. And selecting a second write pulse set with current, voltage or pulse length different from that of the first write pulse set according to the reference progressive resistance distribution to perform incremental step pulse writing so as to change the progressive resistance characteristic of the resistance-type memory element. Or in the process of manufacturing the resistive memory element, the oxygen content distribution of the oxygen content control layer in the resistive memory element is adjusted by different process steps to have a preset oxygen content distribution. When a default set of write pulses is applied to the resistive memory element, a default progressive resistance distribution is obtained. Therefore, the writing and erasing variability of the resistive memory element is reduced, so that the characteristics of the multilayer memory unit of the resistive memory element are improved, and the resistive memory element is suitable for application of a neural network system.

The above-mentioned embodiments are intended to illustrate the objects, technical solutions and advantages of the present invention in further detail, and it should be understood that the above-mentioned embodiments are only exemplary embodiments of the present invention and are not intended to limit the present invention, and any modifications, equivalents, improvements and the like made within the spirit and principle of the present invention should be included in the protection scope of the present invention.