Method and device for reading data

Abstract

In the present invention, data represented in a spin state is read from a storage layer without destroying the data in a magnetic memory element of a type that transmits information based on domain wall motion. A method for reading data from the magnetic memory element comprises: a step (S1) for causing pulse current to flow through a layer structure 9 of a magnetic memory element 10 provided with a plurality of storage layers 1a-1d capable of switching a spin state and a boundary layer 2 disposed among the plurality of storage layers 1a-1d to constitute a magnetic domain wall; a step (S2) for reading a response signal from the layer structure 9 through which the pulse current has flown; and steps (S3-S6) for identifying a spin state for each of the plurality of storage layers 1a-1d on the basis of the response signal.

Description

Data reading method and device

[0001] The present invention relates to a method and apparatus for reading data from a magnetic memory element that transfers information based on domain wall motion.

[0002] With the dramatic increase in the amount of information, there is a need for memory devices that can record information at high density. Flash memory is currently widely used as such a memory device. However, due to its operating principle, flash memory has the drawbacks of limited number of write cycles due to deterioration of the oxide film, and the write speed slowing down as information is repeatedly written. For these reasons, various magnetic memories have been proposed in recent years as replacements for existing flash memory.

[0003] For example, Patent Document 1 discloses a racetrack memory, which is a three-dimensional magnetic memory. In the racetrack memory of Patent Document 1, ferromagnetic materials are separated by magnetic domains and arranged in stacks or lines. A bit is defined for each magnetic domain, and data is stored by the direction of magnetization in the magnetic domain. A current is passed through a thin wire (magnetic nanowire) of ferromagnetic material to move the domain wall. This causes the magnetization in the magnetic domains to move in one direction all at once, transmitting data.

[0004] The information transmission method based on domain wall motion as exemplified in Patent Document 1 still has problems such as a high drive current for domain wall motion and poor controllability of domain wall motion. In response to these problems, the present inventor has proposed magnetic memory elements as disclosed in Patent Documents 2 and 3, which are magnetic memory elements that transmit information based on domain wall motion similar to that of Patent Document 1.

[0005] US Patent No. 6,834,005 International Publication No. 2022 / 014529 Japanese Patent Application Laid-Open No. 2023-094193

[0006] The layer structures of the magnetic memory elements disclosed in Patent Documents 2 and 3 improve the drive current required for domain wall motion and the controllability of domain wall motion compared to conventional magnetic memory elements. Meanwhile, Patent Documents 2 and 3 also exemplify destructive readout as an example of a method for reading data from the storage layer. In destructive readout, data in the storage layer is destroyed when data is read from the storage layer. There is a demand for more refined data readout methods in magnetic memory elements that transmit information based on domain wall motion.

[0007] An object of the present invention is to read data from a storage layer without destroying the data in a magnetic memory element that transmits information based on domain wall motion.

[0008] The inventors of the present invention have conducted extensive research to solve the above-mentioned problems and have found that there is a correspondence relationship between a bit string represented by a plurality of memory cells and the waveform of a response signal corresponding to that bit string, as shown in Figures 5 and 6. Based on this correspondence relationship, it has been found that it is possible to identify a bit string corresponding to a response signal read from a layer structure including a plurality of memory cells. In this case, even if the response signal is read from the layer structure, the bits of the memory cells included in the layer structure are not destroyed.

[0009] That is, to achieve the above object, the present invention includes, for example, the following aspects: (Item 1) A method for reading data represented by spin states from a magnetic memory element that transmits information based on domain wall motion, the method including: passing a pulse current through a layer structure of the magnetic memory element including multiple memory layers whose spin states are switchable and boundary layers disposed between the multiple memory layers and constituting domain walls; reading a response signal from the layer structure through which the pulse current has passed, and identifying the spin state for each of the multiple memory layers based on the response signal; (Item 2) The method according to Item 1, in which the step of identifying the spin states identifies the combination of spin states corresponding to the read response signal based on a correspondence between the combination of spin states and multiple reference response signals; (Item 3) The method according to Item 2, in which the step of identifying the spin states includes: calculating an index relating to the signal similarity between the read response signal and each of the multiple reference response signals; and identifying the combination of spin states corresponding to the read response signal based on the calculated index. (Item 4) The method of item 3, wherein the step of identifying the spin states further includes a step of converting the read response signal from a time domain signal to a frequency domain signal, and the step of calculating the index includes a step of calculating a first index relating to a similarity between the response signal in the frequency domain and each of the plurality of reference response signals in the frequency domain, and a step of calculating a second index relating to a similarity between the response signal in the time domain and each of the plurality of reference response signals in the time domain, and the step of identifying the combination identifies the combination of spin states corresponding to the read response signal based on the calculated first index and second index. (Item 5) The method of item 4, wherein the first index and the second index are correlation coefficients. (Item 6) The method of item 4, wherein the first index is a correlation coefficient and the second index is a phase difference.(Item 7) An apparatus for reading data represented by spin states from a magnetic memory element that transmits information based on domain wall motion, comprising: a current source that applies a pulse current to a layer structure of the magnetic memory element, the layer structure including a plurality of memory layers whose spin states are switchable and boundary layers that are disposed between the plurality of memory layers and form domain walls; a sensor that reads a response signal from the layer structure through which the pulse current has been applied; and a spin state specifying unit that specifies the spin state for each of the plurality of memory layers based on the response signal. (Item 8) The apparatus according to Item 7, wherein the spin state specifying unit specifies the spin state combination corresponding to the read response signal based on a correspondence between the spin state combination and a plurality of reference response signals. (Item 9) The apparatus according to Item 8, wherein the spin state specifying unit includes: an index calculation unit that calculates an index related to signal similarity between the read response signal and each of the plurality of reference response signals; and a combination specification unit that specifies the spin state combination corresponding to the read response signal based on the calculated index. (Item 10) The device according to item 9, wherein the spin state identification unit further includes a signal conversion unit that converts the read response signal from a time domain signal to a frequency domain signal, and the index calculation unit includes a first index calculation unit that calculates a first index related to a similarity between the response signal in the frequency domain and each of the plurality of reference response signals in the frequency domain, and a second index calculation unit that calculates a second index related to a similarity between the response signal in the time domain and each of the plurality of reference response signals in the time domain, and the combination identification unit identifies the spin state combination corresponding to the read response signal based on the calculated first index and second index.

[0010] According to the present invention, in a magnetic memory element of the type that transfers information based on domain wall motion, data can be read from the storage layer without destroying the data.

[0011] FIG. 1 is a cross-sectional view schematically showing an example of the general configuration of a magnetic memory element from which data is read; FIG. 2 is a view for explaining the principle of a data reading method according to an embodiment of the present invention; FIG. 3 is a view for explaining the results of a simulation performed for the case where there are four memory cells; FIG. 4 is a view for explaining the results of a simulation performed for the case where there are four memory cells; FIG. 5 is a view showing the correspondence between all combinations of bit strings represented by four memory cells and the corresponding response signals; FIG. 6 is a view showing the correspondence between all combinations of bit strings represented by four memory cells and the corresponding response signals; FIG. 7 is a view for explaining a method for distinguishing and identifying bit strings represented by four memory cells when response signals in the frequency domain are similar or approximately identical; FIG. 8 is a view for explaining a method for distinguishing and identifying bit strings represented by four memory cells when response signals in the frequency domain are similar or approximately identical; FIG. 9 is a flowchart for explaining the procedure of a data reading method according to an embodiment of the present invention; FIG. 10 is a view schematically showing the general configuration of a data reading device according to an embodiment of the present invention; FIG. 11 is a view schematically showing the general configuration of a data reading device according to another embodiment of the present invention; FIG. 12 is a cross-sectional view schematically showing another example of the general configuration of a magnetic memory element from which data is read;

[0012] Hereinafter, embodiments of the present invention will be described in detail with reference to the accompanying drawings. In the following description and drawings, the same reference numerals will denote the same or similar components, and therefore, redundant descriptions of the same or similar components will be omitted.

[0013] In this specification, the term "layer" does not only mean a layer that is distinguished from other layers by the material used to form it or by physical and chemical properties such as magnetism and conductivity, but also means a film or region formed on the surface of or inside a material, such as an elemental metal or alloy, by a method such as sputtering. The term "layer" in the claims is to be interpreted as including the terms "film" and "region."

[0014] The magnetic memory element from which data is read in this embodiment is a magnetic memory element that transmits information based on domain wall motion. Hereinafter, an example of a method for non-destructively reading data from a memory layer of a magnetic memory element disclosed in Patent Document 2 will be described.

[0015] First, referring to Fig. 1, the configuration of the magnetic memory element exemplified in Patent Document 2 will be described as a magnetic memory element from which data is to be read. Next, referring to Figs. 2 to 9, the principles and methods for non-destructively reading data from the memory layer of this exemplified magnetic memory element will be described. [Configuration of Target Magnetic Memory Element]

[0016] FIG. 1 is a cross-sectional view showing a schematic example of the configuration of a magnetic memory element from which data is read.

[0017] A magnetic memory element 10 from which data is read in this embodiment includes a plurality of memory layers 1 (1a to 1d), a plurality of boundary layers 2 (2a to 2d), a first ferromagnetic layer 3, a first electrode 4, an insulating film 5, a second ferromagnetic layer 6, and a second electrode 7. The illustrated magnetic memory element 10 has a three-dimensional structure in which the first ferromagnetic layer 3, a layer structure 9 of the plurality of boundary layers 2 and the plurality of memory layers 1, the insulating film 5, and the second ferromagnetic layer 6 are stacked between the first electrode 4 and the second electrode 7 in this order from the bottom in the figure.

[0018] The multiple storage layers 1 (1a to 1d) are layers whose spin state can be switched. In the illustrated magnetic memory element 10, the storage layers 1 are ferromagnetic layers. In the illustrated embodiment, the spin state can have two states, for example, with the spin arrow pointing up or down, and one storage layer 1 functions as a memory cell that stores one bit of binary information. Illustratively, the storage layer 1 is made of a metal such as iron or cobalt, or an alloy of these metals, for example, Fe １－ｘ Ni ｘ , Fe １－ｘ Co ｘ , Co １－ｘ Pt ｘ, CoFeB, where x is the composition ratio of the alloy and takes a value in the range of 0<x<1.

[0019] The boundary layer 2 is disposed between the multiple storage layers 1 to form a domain wall. In the illustrated embodiment, the spin state of the boundary layer 2 can have three states, for example, with the spin arrow pointing upward, downward, or sideways. When a domain wall is formed in the boundary layer 2, the spin state of the boundary layer 2 is represented by a sideways arrow. For convenience of explanation, the sideways state of the spin arrow is only shown pointing rightward. In the illustrated magnetic memory element 10, the boundary layer 2 is formed using a non-magnetic material. For example, the non-magnetic material of the boundary layer 2 can be a non-ferromagnetic elemental metal such as copper or platinum, or an alloy of cobalt and platinum with a controlled composition, as described below. In the illustrated magnetic memory element 10, even though the boundary layer 2 is a non-magnetic material, its thickness is thinned, so that the influence of the ferromagnetic layer adjacent to the boundary layer 2 (the storage layer 1 or the first ferromagnetic layer 3) causes the boundary layer 2 to become a ferromagnetic material with a weak exchange stiffness constant due to the proximity effect.

[0020] In the layer structure 9, the description will focus on one boundary layer 2 and a pair of storage layers 1 (1a, 1b) sandwiching the boundary layer 2. In the layer structure 9 of the illustrated magnetic memory element 10, the boundary layer 2 generates a ferromagnetic interaction (magnetic stiffness) Aex between the multiple storage layers 1. More specifically, the boundary layer 2 has a thickness or composition that generates the ferromagnetic interaction Aex between the multiple storage layers 1. The ferromagnetic interaction Aex is an interaction that aligns the spin direction. The generation of the ferromagnetic interaction Aex between the multiple storage layers 1 improves the drive current required for domain wall motion and the controllability of domain wall motion in the magnetic memory element 10. Note that the ferromagnetic interaction Aex also occurs between the storage layer 1a and the first ferromagnetic layer 3, which sandwich the boundary layer 2.

[0021] When a simple metal is used as the non-magnetic material of the boundary layer 2, the boundary layer 2 is formed using a simple metal with a thickness that causes ferromagnetic interaction Aex between the multiple storage layers 1. For example, when the boundary layer 2 is formed using copper, the thickness of the boundary layer 2 is preferably within the range of 1 to 3 copper atoms. More preferably, the thickness of the boundary layer 2 is within the range of 1 to 2 copper atoms. For example, when the boundary layer 2 is formed using platinum, the thickness of the boundary layer 2 is preferably within the range of 1 to 4 platinum atoms. More preferably, the thickness of the boundary layer 2 is within the range of 1 to 3 platinum atoms.

[0022] When an alloy is used as the nonmagnetic material of the boundary layer 2, the boundary layer 2 is formed using an alloy with a composition that generates ferromagnetic interaction Aex between the multiple storage layers 1. By controlling the composition ratio of the alloy used to form the boundary layer 2, the magnitude of the ferromagnetic interaction Aex generated between the multiple storage layers 1 can be controlled. Since the Curie temperature Tc is the transition temperature at which a ferromagnetic material changes into a paramagnetic material, by controlling the Curie temperature Tc of the alloy, it is possible to control whether the alloy exhibits the properties of a ferromagnetic material or a paramagnetic material (i.e., a nonmagnetic material). The Curie temperature Tc and the ferromagnetic interaction Aex are proportional. On the other hand, the Curie temperature Tc of the alloy can be controlled by controlling the composition ratio of the alloy. For example, SA Ahern, MJC Martin and Willie Sucksmith, “The spontaneous magnetization of nickel + copper alloys”, Proc. Math. Phys. Eng. Sci., United Kingdom, The Royal Society, 11 November 1958, Volume 248, Issue 1253, p.145-152, https: / / doi.org / 10.1098 / rspa.1958.0235, Fig. 3 shows the structure of Ni １－ｘ Cu ｘ The composition dependence of the Curie temperature Tc in alloys is described. １－ｘ Cu ｘ Not limited to alloys, Co １－ｘ Ptｘ Similar control can be applied to alloys. Therefore, by controlling the composition ratio of the alloy used to form the boundary layer 2, it is possible to control the Curie temperature Tc of the alloy and to control whether the alloy exhibits ferromagnetic or paramagnetic properties, thereby controlling the magnitude of the ferromagnetic interaction Aex.

[0023] The first ferromagnetic layer 3 is a ferromagnetic layer whose spin state can be switched. The first ferromagnetic layer 3 is disposed on the side of the memory layer 1a located at the bottom in the layer structure 9, with the boundary layer 2 sandwiched therebetween. The first ferromagnetic layer 3 functions as a layer for writing one bit of binary information to the memory layer 1a. For example, the first ferromagnetic layer 3 can be formed using an alloy of cobalt and platinum or an alloy of iron and nickel. The first ferromagnetic layer 3 can be made of various materials used for the magnetization fixed layer in a magnetoresistive random access memory (MRAM).

[0024] The first ferromagnetic layer 3 has a higher coercivity than the storage layer 1. For example, when at least one of the following three conditions is satisfied, the first ferromagnetic layer 3 has a higher coercivity than the storage layer 1. The first condition is that the first ferromagnetic layer 3 is formed to be thicker than the storage layer 1, even if the first ferromagnetic layer 3 and the storage layer 1 are formed using the same material. The second condition is that the first ferromagnetic layer 3 is formed using a material having a higher magnetic anisotropy than the storage layer 1, even if the thickness of the first ferromagnetic layer 3 is the same as that of the storage layer 1. The third condition is that the first ferromagnetic layer 3 is formed using a material having a sufficiently higher magnetic anisotropy than the storage layer 1, even if the thickness of the first ferromagnetic layer 3 is thinner than that of the storage layer 1, and as a result, the coercivity of the first ferromagnetic layer 3 is higher than that of the storage layer 1.

[0025] The first electrode 4 is disposed adjacent to the first ferromagnetic layer 3 and uses spin-orbit torque to switch the spin state of the first ferromagnetic layer 3. The first electrode 4 includes a spin-orbit torque (SOT) layer 41 and two bottom electrodes 42 (42 a, 42 b) electrically connected to the spin-orbit torque layer 41.

[0026] By passing a drive current between the terminals 21 and 22 of the first electrode 4 to switch the spin state of the first ferromagnetic layer 3, a write current Iw, indicated by a dashed line in the figure, flows through the spin-orbit torque layer 41, and the spin-orbit torque generated by spin-orbit interaction switches the spin state of the first ferromagnetic layer 3. The spin state of the first ferromagnetic layer 3 is determined according to the direction of the write current Iw. In the illustrated magnetic memory element 10, the write current Iw is pulsed. Exemplarily, the spin-orbit torque layer 41 can be formed using a heavy metal such as platinum. The bottom electrode 42 can be formed using various conductive metals such as gold and copper.

[0027] The insulating film 5 and the second ferromagnetic layer 6 function as a magnetic tunnel junction (MTJ) in combination with the memory layer 1d located at the upper side in the figure in the layer structure 9. The memory layer 1d functions as a free layer of the magnetic tunnel junction.

[0028] The insulating film 5 functions as a tunnel layer of the magnetic tunnel junction. The insulating film 5 is disposed between the memory layer 1d, located at the upper side of the layer structure 9 in the figure, and the second electrode 7. The second ferromagnetic layer 6 is a layer in which the spin state is fixed (in the illustrated embodiment, the arrow points upward) and functions as a fixed layer of the magnetic tunnel junction. In the illustrated magnetic memory element 10, the spin state of the second ferromagnetic layer 6 is fixed with the spin arrow pointing upward. The second ferromagnetic layer 6 is disposed between the insulating film 5 and the second electrode 7. The insulating film 5 can be formed using an oxide film such as magnesium oxide (MgO). The second ferromagnetic layer 6 can be formed using, for example, CoFeB, an alloy of cobalt, iron, and boron. The second ferromagnetic layer 6 can be formed from various materials used in the so-called magnetization fixed layer of MRAM.

[0029] The second electrode 7 is disposed adjacent to the second ferromagnetic layer 6 on the side of the storage layer 1d located at the top in the figure. By passing a drive current for moving the domain wall between the terminal 23 of the second electrode 7 and either the terminal 21 or the terminal 22 of the first electrode 4, a domain wall drive current Id, indicated by a two-dot chain line in the figure, flows between the second electrode 7 and the first electrode 4. This allows the domain wall to move in the multiple boundary layers 2 (2a to 2d) located between the second electrode 7 and the first electrode 4, and the spin states of the multiple storage layers 1 (1a to 1d) to shift sequentially in a racetrack-like manner. In the illustrated magnetic memory element 10, the domain wall drive current Id is pulsed. Exemplarily, the second electrode 7 can be formed using various conductive metals such as gold and copper. [Principle of Data Readout]

[0030] FIG. 2 is a diagram for explaining the principle of a data reading method according to an embodiment of the present invention.

[0031] In the following example, the magnetic memory element 10 has four memory cells (cell No. 1 to cell No. 4) and stores a total of four bits of binary information. For ease of explanation, the downward spin arrow indicates a value of "0" and the upward spin arrow indicates a value of "1."

[0032] Furthermore, a bit string refers to a sequence of 1 bit of binary information represented by each of multiple memory cells, and refers to a combination (or sequence) of spin states (value "0" or value "1") represented by each of multiple memory layers 1 (1a to 1d) included in the layer structure 9 shown in Figure 1.

[0033] 2, the bit string represented by the four memory cells is

[0010] in the example shown in (A) and (B), and

[1010] in the example shown in (C). The bit string represented by the four memory cells is different between (A) and (B) and (C).

[0034] A data reading method according to one embodiment is a method for non-destructively reading data from a magnetic memory element 10. In the data reading method according to one embodiment, as shown in FIG. 2A, a pulse current is passed through the layer structure 9 of the magnetic memory element 10 in which the bit strings represented by the multiple memory cells are in a certain state (e.g., the

[0010] state shown in the figure). The pulse current flows between the second electrode 7 and the first electrode 4, as indicated by the two-dot chain line in the figure. The magnitude of the pulse current is weaker than the domain wall-driving current Id described with reference to FIG. 1 and is strong enough not to move the domain walls formed in each of the multiple boundary layers 2. Exemplarily, the waveform of the pulse current passed through the layer structure 9 is rectangular, and the pulse length is approximately 1 ns (nanosecond). When the pulse current is passed through the layer structure 9, magnetization dynamics are induced in the layer structure 9 due to the spin torque effect, as shown in FIG. 2B. The magnetization dynamics are induced such that the spins in some of the storage layers 1 and boundary layers 2 contained in the layer structure 9 undergo, for example, precession.

[0035] On the other hand, as shown in FIG. 2C , even if a pulse current is passed through the layer structure 9 of the magnetic memory element 10 in which the bit strings represented by the multiple memory cells are in a state different from the state

[0010] shown in FIG. 2A (e.g., the illustrated

[1010] ), magnetization dynamics are induced in the layer structure 9 by the spin torque effect, as in the example shown in FIG. 2B . Here, attention is focused on the induced magnetization dynamics. Since the bit strings represented by the multiple memory cells included in the layer structure 9 are different between those shown in (B) and (C), the magnetization dynamics induced in the layer structure 9 by the spin torque effect are also different between those shown in (B) and (C). The magnetization dynamics induced in the layer structure 9 changes and decays over time.

[0036] That is, when a pulse current is passed through the layer structure 9 and a response signal from the layer structure 9 is read, the read response signal reflects differences in magnetization dynamics resulting from differences in bit strings within the layer structure 9. This makes it possible to identify the layer structure 9, i.e., to identify the bit strings represented by the multiple memory cells included in the layer structure 9, based on the response signal from the layer structure 9.

[0037] Next, several specific examples of bit strings represented by the four memory cells will be given to explain how these different bit strings can be distinguished using response signals from the layer structure 9.

[0038] 3 and 4 are diagrams for explaining the results of a simulation performed for the case of four memory cells. (A) is a diagram schematically showing the layer structure of the magnetic memory element used in the simulation. (B) is the simulation result of the response signal in the time domain, and (C) is the simulation result of the response signal in the frequency domain.

[0039] As shown in Figure 3A and Figure 4A, the bit strings represented by the four memory cells are different between the example shown in Figure 3 and the example shown in Figure 4. The bit string represented by the four memory cells is

[0001] in the example shown in Figure 3 and

[0010] in the example shown in Figure 4.

[0040] The signal shown in (B) is a response signal from the layer structure 9 when a pulse current is passed through the layer structure 9. The response signal read from the layer structure 9 is a signal whose signal intensity changes over time, i.e., a time-domain response signal. The signal shown in (B) can be read using a magnetic tunnel junction provided in the magnetic memory element 10. In the example of the layer structure 9 shown in FIG. 1, the insulating film 5 and the second ferromagnetic layer 6 that constitute the magnetic tunnel junction are provided on the upper side of the layer structure 9 in the figure. As shown in (B) of FIG. 3 and (B) of FIG. 4, the time-domain response signal read from the layer structure 9 differs between the example shown in FIG. 3 and the example shown in FIG. 4 due to differences in the bit strings within the layer structure 9.

[0041] The signal shown in (C) is a frequency-domain response signal obtained by converting the time-domain response signal shown in (B) into the frequency domain by, for example, Fourier transform. As shown in (C) of Figure 3 and (C) of Figure 4, the frequency-domain response signal also differs between the example shown in Figure 3 and the example shown in Figure 4 due to differences in the bit sequence within the layer structure 9, just like the time-domain response signal.

[0042] As described above, when the time-domain response signals of the bit string

[0001] and the bit string

[0010] are compared, these time-domain response signals exhibit completely different time dependencies. Therefore, it is possible to distinguish between different bit strings contained in the layer structure 9 based on the difference in the time-domain response signals. Similarly, when the frequency-domain response signals of the bit string

[0001] and the bit string

[0010] are compared, these frequency-domain response signals also exhibit completely different frequency dependencies. Therefore, it is possible to distinguish between different bit strings contained in the layer structure 9 based on the difference in the frequency-domain response signals. Note that, as shown by comparing (B) and (C) of Figures 3 with (B) and (C) of Figures 4, it is possible to distinguish between bit strings more accurately using the frequency-domain response signals than using the time-domain response signals.

[0043] Next, it will be explained that for all combinations of bit strings represented by the four memory cells, it is possible to distinguish between these different bit strings using the response signals from the layer structure 9.

[0044] 5 and 6 are diagrams showing the correspondence between all combinations of bit strings represented by four memory cells and the corresponding response signals. ４ 5 and 6 show the correspondence between the bit strings and the waveforms of the response signals corresponding to the bit strings for each of 16 possible combinations of bit strings. In order to facilitate visual comparison and understanding of the invention, the correspondence between the bit strings represented by the memory cells and the response signals is shown as an example in which the response signals are frequency domain response signals.

[0045] 5 shows eight correspondence relationships between bit strings and response signals, and the waveforms of the eight response signals for these eight correspondence relationships are different from one another. Therefore, if the waveforms of the eight response signals shown in FIG. 5 are used as reference response signals and correspondence relationships between these reference response signals and bit strings are prepared in advance, it is possible to identify the bit string corresponding to the response signal read from the layer structure 9 based on such correspondence relationships between the reference response signals and bit strings. Matching between the response signal read from the layer structure 9 and the reference response signal can be performed, for example, by calculating the signal similarity. Examples of indices related to the signal similarity include a cross-correlation coefficient and the magnitude of the signal phase difference.

[0046] As with the example shown in Fig. 5, the waveforms of the eight response signals for the eight correspondence relationships shown in Fig. 6 are different from one another. Therefore, it is possible to identify the bit string corresponding to the response signal read from the layer structure 9 based on the correspondence relationships between the reference response signals and the bit strings that have been prepared in advance.

[0047] In this way, correspondence relationships between reference response signals and bit strings (16 in the illustrated example) are prepared in advance, and the signal similarity is calculated between the prepared reference response signals and the response signals read from the layer structure 9. Then, based on the calculated similarity, one reference response signal corresponding to the response signal read from the layer structure 9 is identified, and the bit string corresponding to the identified reference response signal is identified. This makes it possible to identify the bit string corresponding to the response signal from the layer structure 9.

[0048] On the other hand, when the correspondence relationship shown in Figure 5 and the correspondence relationship shown in Figure 6 are compared, there are cases where the waveforms of the response signals in the frequency domain are similar or roughly the same, even though the bit sequences are different. For example, when the bit sequence

[0001] shown in Figure 5 and the bit sequence

[1110] shown in Figure 6 are compared, the waveforms of the response signals in the frequency domain are roughly the same, even though the bit sequences are different. In such cases, simply converting the time-domain response signal read from the layer structure 9 into the frequency domain and comparing the converted frequency-domain response signal with a previously prepared reference frequency-domain response signal is not enough to uniquely identify the bit sequence corresponding to the read response signal.

[0049] In such a case, as will be explained below, by using not only the frequency domain response signal but also the time domain response signal to identify the bit string, it is possible to uniquely identify the bit string corresponding to the response signal read from the layer structure 9.

[0050] 7 and 8 are diagrams illustrating a method for distinguishing and identifying bit strings represented by four memory cells when the response signals in the frequency domain are similar or substantially identical. (A) is a diagram schematically illustrating the layer structure of the magnetic memory element used in the simulation. (B) is a simulation result of the response signal in the time domain, and (C) is a simulation result of the response signal in the frequency domain. (D) is a simulation result of the resistance value of the memory cell (memory layer 1d of cell No. 4) that constitutes the magnetic tunnel junction.

[0051] The bit strings represented by the four memory cells are

[0001] in the example shown in Fig. 7 and

[1110] in the example shown in Fig. 8. Although the bit strings represented by the four memory cells are different between the example shown in Fig. 7 and the example shown in Fig. 8, the waveforms of the response signals in the frequency domain are similar or roughly the same, as shown in Fig. 7C and Fig. 8C. In such a case, simply using the response signal in the frequency domain to identify the bit string cannot uniquely identify the bit string corresponding to the response signal read from the layer structure 9.

[0052] In such a case, in a data reading method according to an embodiment, not only the response signal in the frequency domain but also the response signal in the time domain is used to identify the bit string, thereby making it possible to uniquely identify the bit string corresponding to the response signal read from the layer structure 9.

[0053] As described above, the waveforms of the response signals in the frequency domain are similar or substantially the same between the examples shown in FIGS. 7 and 8 . However, as shown in FIGS. 7B and 8B , there are differences in the waveforms of the response signals in the time domain. Specifically, a phase difference occurs in the time domain response signals between the examples shown in FIGS. 7 and 8 . More specifically, the phase of the response signals in the time domain is inverted between the examples shown in FIGS. 7 and 8 . The reason for the phase inversion is that the difference in the spin state in the memory cell (memory layer 1d of cell No. 4) located at the upper side of the layer structure 9 constituting the magnetic tunnel junction is reflected in the time domain response signal read from the layer structure 9. This can be understood from the fact that the simulation result of the resistance value of the memory cell constituting the magnetic tunnel junction shown in FIG. 7D corresponds to the simulation result of the time domain response signal shown in FIG. 7B . As with FIG. 7 , the simulation result shown in FIG. 8D corresponds to the simulation result shown in FIG. 8B . [Data Readout Procedure]

[0054] FIG. 9 is a flowchart illustrating the procedure of a data reading method according to one embodiment of the present invention.

[0055] In this embodiment, the magnetic memory element from which data is read is a magnetic memory element 10 having the configuration described with reference to Fig. 1. The magnetic memory element 10 includes a plurality of storage layers 1 whose spin states are switchable, and boundary layers 2 disposed between the plurality of storage layers 1 and constituting domain walls.

[0056] In a data read method according to an embodiment, a bit string represented by a plurality of memory cells is identified using a response signal in the frequency domain and a response signal in the time domain. The correspondence between the reference response signal and the bit string described with reference to FIGS. 5 and 6 is prepared in advance before the data read method according to the present embodiment is performed.

[0057] In this embodiment, as described above, the bit strings represented by the plurality of memory cells are identified using the response signal in the frequency domain and the response signal in the time domain, so that the correspondence relationships prepared in advance for the reference response signal are also prepared as a total of two sets: a first set relating to the correspondence relationship between the response signal in the frequency domain and the bit string, and a second set relating to the correspondence relationship between the response signal in the time domain and the bit string. For example, when the magnetic memory element 10 stores a total of four bits of binary information, the number of memory cells required is four as shown in the figure, and in this case, the number of first sets prepared in advance is two. ４ = 16, and the number of the second set prepared in advance is 2 ４ = 16 pieces.

[0058] In step S1, a pulse current is passed through the layer structure of a magnetic memory element 10 from which data is to be read. When the pulse current is passed through the layer structure 9 of the magnetic memory element 10, magnetization dynamics based on the spin torque effect is induced in the layer structure 9.

[0059] In step S2, a response signal is read from the layer structure through which the pulse current has been passed. In this embodiment, a time-domain response signal is read from the layer structure 9. The response signal reflects differences in magnetization dynamics resulting from differences in bit sequences within the layer structure 9.

[0060] In step S3, the response signal read in S2 is converted from a time domain signal to a frequency domain signal. In this embodiment, the time domain response signal read from the layer structure 9 is converted into a frequency domain response signal by Fourier transform.

[0061] In step S4, a first index of signal similarity is calculated between the frequency-domain response signal converted in step S3 and each of a plurality of frequency-domain reference response signals prepared in advance. In this embodiment, the first index is a cross-correlation coefficient. Since methods for performing cross-correlation analysis on two different signals are well known, further detailed description thereof will be omitted in this specification.

[0062] In step S5, a second index of signal similarity is calculated between the time-domain response signal read in S2 and each of a plurality of previously prepared time-domain reference response signals. In this embodiment, the second index is a phase difference. There are various methods for calculating the phase difference between two different signals, and all of these are well known, so further detailed description will be omitted in this specification.

[0063] In step S6, a bit string corresponding to the time-domain response signal read in step S2 is identified based on the first and second indices related to the signal similarity calculated in steps S4 and S5. First, as described with reference to FIGS. 3 to 8, one reference response signal that most certainly corresponds to the time-domain response signal read in step S2 is identified based on the indices related to the signal similarity. Next, a bit string corresponding to the identified reference response signal is identified based on a correspondence relationship between the reference response signal and the bit string that has been prepared in advance. This uniquely identifies a bit string corresponding to the response signal read from the layer structure 9. The identified bit string is a bit string represented by multiple memory cells included in the layer structure 9 and is a combination of the spin states represented by these multiple memory cells.

[0064] As described above, according to the data reading method according to one embodiment of the present invention, data can be read from the memory layer (memory cell) without destroying the data in the memory layer.

[0065] FIG. 10 is a diagram showing a schematic configuration of a data reading device according to an embodiment of the present invention.

[0066] The data reading device 20 according to one embodiment is a device that reads data represented by spin states from the magnetic memory element 10, and includes a current source 24, a sensor 25, and a spin state specifying unit 29. The spin state specifying unit 29 can be implemented using, for example, various semiconductor integrated circuits and electric circuits.

[0067] The current source 24 passes a pulse current for data reading from the second electrode 7 to the first electrode 4 of the magnetic memory element 10. The sensor 25 measures the current value of the current flowing through the layer structure 9 of the magnetic memory element 10. The current source 24, the sensor 25, and the spin state specifying unit 29 are connected to a memory controller 26. The memory controller 26 controls the operation of the current source 24, the sensor 25, and the spin state specifying unit 29, thereby controlling the data read operation from the magnetic memory element 10, as described with reference to FIGS. 2 to 9 . Data read via the spin state specifying unit 29 is transmitted and received via a data bus 27. The connection from the current source 24 to the first electrode 4 and the second electrode 7 of the magnetic memory element 10 is switched using, for example, switches 28 a and 28 b. The operation of the switches 28 a and 28 b is controlled, for example, by the memory controller 26. A plurality of magnetic memory elements 10 can be arranged in an array to form a memory array.

[0068] The current source 24 applies a pulse current for data reading to the layer structure 9 of the magnetic memory element 10 from which data is to be read, thereby inducing magnetization dynamics based on the spin torque effect in the layer structure 9. The sensor 25 reads a response signal from the layer structure 9 through which the pulse current is applied.

[0069] The spin state determination unit 29 determines bit strings represented by the plurality of memory cells based on the response signals read by the sensor 25. In this embodiment, the spin state determination unit 29 (29a) includes a signal conversion unit 291, a first index calculation unit 292, a band-pass filter 293, a second index calculation unit 294, a combination determination unit 295, and a reference data holding unit 299. The first index calculation unit 292 and the second index calculation unit 294 may be integrated to form an index calculation unit. As will be described later with reference to FIG. 11 , in other embodiments, the spin state determination unit 29 (29b) may be configured without the band-pass filter 293 and the second index calculation unit 294.

[0070] The signal conversion unit 291 converts the response signal read by the sensor 25 from a time domain signal to a frequency domain signal. In this embodiment, the time domain response signal read from the layer structure 9 is Fourier transformed. In this case, the signal conversion unit 291 can be called a Fourier transform unit 291.

[0071] The first index calculation unit 292 calculates a first index relating to the degree of similarity between the frequency-domain response signal converted by the signal conversion unit 291 and each of a plurality of frequency-domain reference response signals prepared in advance. In this embodiment, the first index is a cross-correlation coefficient. The calculated first index is input to the combination identification unit 295, which is a subsequent circuit.

[0072] In addition, in this embodiment, the response signal read by the sensor 25 is input to a second index calculation unit 294 via a band-pass filter 293 .

[0073] The second index calculation unit 294 calculates a second index relating to the degree of similarity between the response signal read by the sensor 25 and each of a plurality of reference response signals in the time domain prepared in advance. In this embodiment, the second index is a phase difference. The calculated second index is input to the combination identification unit 295, which is a subsequent circuit.

[0074] The combination identification unit 295 identifies a bit string corresponding to the time-domain response signal read in S2 based on the first index and second index related to the signal similarity calculated by the first index calculation unit 292 and the second index calculation unit 294.

[0075] The reference data storage unit 299 stores data relating to the correspondence between the reference response signal and the bit strings represented by the multiple memory cells as data used as a reference when identifying the spin state. In this embodiment, the reference data storage unit 299 stores a first data set relating to the correspondence between the frequency domain response signal and the bit string and a second data set relating to the correspondence between the time domain response signal and the bit string, which are prepared in advance. The reference data storage unit 299 can be implemented using a non-volatile storage circuit such as an EEPROM.

[0076] As described above, the data reading device according to one embodiment of the present invention can read data from the memory layer (memory cell) without destroying the data in the memory layer.

[0077] Although the present invention has been described above with reference to specific embodiments, the present invention is not limited to the above-described embodiments.

[0078] The number of memory cells included in the illustrated magnetic memory element 10 is not limited to four, and the magnetic memory element 10 can include a greater number of memory cells.

[0079] The structure of the illustrated magnetic memory element 10 is a three-dimensional structure in which various layers are stacked vertically, but the structure of the magnetic memory element can also be a structure in which regions corresponding to the various layers are arranged horizontally and mounted on a plane. The layer structure described in the claims does not only mean a three-dimensional structure in which various layers are stacked vertically, but also means a structure in which various regions are arranged horizontally in this way and mounted on a plane.

[0080] In the above embodiment, a Fourier transform is used to convert a time domain signal into a frequency domain signal, but the calculation method used to convert the signal is not limited to a Fourier transform. For example, a fast Fourier transform or a discrete Fourier transform may be used instead of a Fourier transform to convert a time domain signal into a frequency domain signal.

[0081] In the above-described embodiment, in order to facilitate visual comparison and understanding of the invention, the correspondence relationship between the bit strings represented by the memory cells and the response signals is illustrated as an example in which the response signals are frequency domain response signals, but the illustrated correspondence relationship between the bit strings and the response signals is not limited to the case in which the response signals are frequency domain response signals. Similarly to Figures 5 and 6, the correspondence relationship between the bit strings represented by the memory cells and the response signals can also be illustrated in the case in which the response signals are time domain response signals.

[0082] In the above-described embodiment, in step S5, a phase difference is calculated as a second index of signal similarity between the time-domain response signal and each of a plurality of prepared time-domain reference response signals. However, the second index calculated in step S5 is not limited to a phase difference. The index calculated in step S5 may be any index related to signal similarity that reflects a phase difference between the signals, for example, a cross-correlation coefficient. In this case, in step S4, a cross-correlation coefficient is calculated as a first index between the frequency-domain response signal and each of a plurality of prepared frequency-domain reference response signals, and in step S5, a cross-correlation coefficient is calculated as a second index between the time-domain response signal and each of a plurality of prepared time-domain reference response signals.

[0083] FIG. 11 is a diagram illustrating a schematic configuration of a data reading device according to another embodiment of the present invention. In the above-described embodiment, bit strings represented by multiple memory cells are identified using frequency-domain response signals and time-domain response signals. However, the response signals used to identify the bit strings do not necessarily need to include both frequency-domain and time-domain response signals. In the above-described embodiment, as shown by comparing FIGS. 3B and 3C with FIGS. 4B and 4C, using frequency-domain response signals allows for more accurate discrimination of bit strings than using time-domain response signals. However, it is also possible to identify bit strings represented by multiple memory cells using, for example, only time-domain response signals. In this case, for example, a cross-correlation coefficient may be calculated as an index of signal similarity between the time-domain response signal and each of multiple reference time-domain response signals prepared in advance. In this case, the spin state identification unit 29 (29b) may be configured without the signal conversion unit 291 and the second index calculation unit 294, as illustrated in FIG. 11.

[0084] In the above-described embodiment, the magnetic memory element 10 illustrated in Fig. 1 is the target of data read, but the magnetic memory element from which data is read is not limited to the magnetic memory element 10 illustrated in Fig. 1. The magnetic memory element from which data is read in the present invention may be any magnetic memory element that transmits information based on domain wall motion, and may be any magnetic memory element in which magnetization dynamics based on the spin torque effect are induced in the layer structure. Another example of a magnetic memory element from which data is read in the present invention is shown in Fig. 12.

[0085] 12 is a cross-sectional view showing another example of the general configuration of a magnetic memory element from which data is read. The magnetic memory element shown in FIG. 12 is the magnetic memory element exemplified in Patent Document 3. Unless otherwise specified, the configurations of the other examples of magnetic memory elements described below are the same as the configuration of the magnetic memory element 10 exemplified as the target of data read in the above-described embodiment, and therefore, redundant explanations will be omitted.

[0086] As shown in FIG. 12 , in another example of a magnetic memory element 10 exemplified as a target for data reading, the storage layer 1 is an antiferromagnetic layer. The thickness of the boundary layer 2 is such that a domain wall can be formed therein and the layer can function as a domain wall layer. This antiferromagnetic layer 1, which functions as the storage layer 1, does not have magnetization as a whole, but has a magnetic order at the micro level, and one bit of binary information, "0" or "1," can be assigned to this state. In other words, the antiferromagnetic layer 1 (storage layer 1) having a switchable spin state can be expressed as a layer to which a spin state corresponding to one bit of binary information is assigned.

[0087] The antiferromagnetic layer 1 (storage layer 1) includes a plurality of stacked ferromagnetic layers 11 and 12 and a non-magnetic layer (not shown) disposed between the plurality of ferromagnetic layers 11 and 12. The non-magnetic layer in the antiferromagnetic layer 1 may be any configuration, and in another example, the non-magnetic layer may be omitted. In the antiferromagnetic layer 1, an interaction Jex that maintains the spin directions in opposite directions occurs between the stacked ferromagnetic layers 11 and 12. As a result, within one antiferromagnetic layer 1, the spin direction in one ferromagnetic layer 11 and the spin direction in the other ferromagnetic layer 12 are maintained in opposite directions. For example, in the antiferromagnetic layer 1, binary information "0" can be assigned when the spin arrow in the ferromagnetic layer 11 points upward and the spin arrow in the ferromagnetic layer 12 points downward. Binary information "1" can be assigned when the spin arrow in the ferromagnetic layer 11 points downward and the spin arrow in the ferromagnetic layer 12 points upward. The allocation of binary information in the antiferromagnetic layer 1 is not limited to this illustrated mode, and the binary information to be allocated may be logic opposite to that illustrated.

[0088] For example, the antiferromagnetic layer 1 may have a Co / Pt / Gd laminated structure in which cobalt, platinum, and gadolinium are laminated, or a Co / Pt structure in which iridium is laminated between two layers of cobalt and platinum. ６ / Ir / (Co / Pt) ６ The symbol " / " indicates the lamination of the layer structure. These lamination structures can be formed by, for example, sputtering. In the Co / Pt / Gd lamination structure that constitutes the antiferromagnetic layer 1, the Co layer and the Gd layer correspond to the ferromagnetic layers 11 and 12, and the Pt layer corresponds to the non-magnetic layer. Illustratively, the thickness of the Co layer is about 1 nm, the thickness of the Pt layer is about 0.1 nm, and the thickness of the Gd layer is about 1 nm. (Co / Pt) ６ / Ir / (Co / Pt) ６ In the layer structure, the two (Co / Pt) layers correspond to the ferromagnetic layers 11 and 12, and the Ir layer corresponds to the non-magnetic layer. ６ The thickness of the layer is about 2.4 nm, and the thickness of the Ir layer is about 0.5 nm.

[0089] 12, the boundary layer 2 is formed using an oxide film such as magnesium oxide (MgO). Illustratively, the thickness of the MgO layer used for the boundary layer 2 is approximately 1 nm. The boundary layer 2 can be formed by, for example, a sputtering method.

[0090] 1 (1a to 1d) Memory layer (ferromagnetic layer or antiferromagnetic layer) 2 (2a to 2d) Boundary layer 3 First ferromagnetic layer 4 First electrode 5 Insulating film 6 Second ferromagnetic layer 7 Second electrode 9 Layer structure 10 Magnetic memory element 11 Ferromagnetic layer 12 Ferromagnetic layer 20 Data reading device 21 to 23 Terminal 24 Current source 25 Sensor 26 Memory controller 27 Data bus 28 (28a, 28b) Switch 29 (29a, 29b) Spin state determination unit 291 Signal conversion unit 292 First index calculation unit 293 Bandpass filter 294 Second index calculation unit 295 Combination determination unit 299 Reference data holding unit 41 Spin-orbit torque layer 42 (42a, 42b) Bottom electrode Aex Ferromagnetic interaction Id Domain wall driving current Iw Write current

Claims

1. A method for reading data represented by a spin state from a magnetic memory element that transmits information based on domain wall motion, comprising: a step of passing a pulse current through a layer structure of the magnetic memory element having a plurality of memory layers whose spin state can be switched, and boundary layers disposed between the plurality of memory layers and constituting domain walls; and a step of reading a response signal from the layer structure through which the pulse current has been passed, and identifying the spin state for each of the plurality of memory layers based on the response signal.

2. The method according to claim 1, wherein the step of identifying the spin states includes identifying the combination of spin states corresponding to the read response signal based on a correspondence between the combination of spin states and a plurality of reference response signals.

3. The method according to claim 2, wherein the step of identifying the spin state includes the steps of: calculating an index relating to the signal similarity between the read response signal and each of the plurality of reference response signals; and identifying the combination of spin states corresponding to the read response signal based on the calculated index.

4. The method of claim 3, wherein the step of identifying the spin states further includes a step of converting the read response signal from a time domain signal to a frequency domain signal, the step of calculating the index includes a step of calculating a first index relating to a similarity between the response signal in the frequency domain and each of the plurality of reference response signals in the frequency domain, and a step of calculating a second index relating to a similarity between the response signal in the time domain and each of the plurality of reference response signals in the time domain, and the step of identifying the combination identifies the combination of spin states corresponding to the read response signal based on the calculated first index and second index.

5. The method of claim 4, wherein the first index and the second index are correlation coefficients.

6. The method of claim 4, wherein the first index is a correlation coefficient and the second index is a phase difference.

7. An apparatus for reading data represented by a spin state from a magnetic memory element that transmits information based on domain wall motion, comprising: a current source that passes a pulse current through a layer structure of a magnetic memory element having a plurality of memory layers whose spin states can be switched and boundary layers that are disposed between the plurality of memory layers and form domain walls; a sensor that reads a response signal from the layer structure through which the pulse current has been passed; and a spin state determination unit that determines the spin state for each of the plurality of memory layers based on the response signal.

Images