NONVOLATILE MEMORY

Abstract

According to one embodiment, a nonvolatile memory includes a conductive line including a first portion, a second portion and a third portion therebetween, a storage element including a first magnetic layer, a second magnetic layer and a nonmagnetic layer therebetween, and the first magnetic layer being connected to the third portion, and a circuit flowing a write current between the first and second portions, applying a first potential to the second magnetic layer, and blocking the write current flowing between the first and second portions after changing the second magnetic layer from the first potential to a second potential.

Description

CROSS-REFERENCE TO RELATED APPLICATIONS

This application is based upon and claims the benefit of priority from Japanese Patent Application No. 2016-155105, filed Aug. 5, 2016, the entire contents of which are incorporated herein by reference.

FIELD

Embodiments described herein relate generally to a nonvolatile memory.

BACKGROUND

Currently, nonvolatile memories such as static random access memory (SRSM) and dynamic random access memory (DRAM) are mainstream as a working memory used in various systems. However, these memories have a problem of high power consumption.

Thus, attempts to replace the working memory used in various system and further, storage memories with a magnetic memory that is faster and consumes less power have been examined. However, it is necessary to reduce a write error rate to apply magnetic memories to various systems.

BRIEF DESCRIPTION OF THE DRAWINGS

FIG. 1 shows a magnetic memory according to a first embodiment;

FIG. 2 is a diagram showing a relationship between V_assist and I_write;

FIG. 3 is a diagram showing a relationship between V_assist and I_write;

FIG. 4 is a diagram showing a relationship between V_assist and I_write;

FIG. 5 is a diagram showing the start of a write operation (0-write) in a first characteristic;

FIG. 6 is a diagram showing the end of the write operation (0-write) in the first characteristic;

FIG. 7 is a diagram showing the start of a write operation (1-write) in the first characteristic;

FIG. 8 is a diagram showing the end of the write operation (1-write) in the first characteristic;

FIG. 9 is a diagram showing a read operation in the first characteristic;

FIG. 10 is a diagram showing the start of the write operation (0-write) in a second characteristic;

FIG. 11 is a diagram showing the end of the write operation (0-write) in the second characteristic;

FIG. 12 is a diagram showing the start of the write operation (1-write) in the second characteristic;

FIG. 13 is a diagram showing the end of the write operation (1-write) in the second characteristic;

FIG. 14 is a diagram showing the read operation in the second characteristic;

FIG. 15 is a diagram showing a magnetic memory according to a second embodiment;

FIG. 16 is a diagram showing a magnetic memory according to a third embodiment;

FIG. 17 is a diagram showing an example of a device structure of a unit cell;

FIG. 18 is a diagram showing an example of the device structure of the unit cell;

FIG. 19 is a diagram showing an example of the device structure of the unit cell;

FIG. 20 is a diagram showing an example of the device structure of a memory cell;

FIG. 21 is a diagram showing an example of the device structure of the memory cell;

FIG. 22 is a diagram showing an example of the device structure of the memory cell;

FIG. 23 is a diagram showing an example of a read/write circuit;

FIG. 24 is a diagram showing an example of a write operation (first);

FIG. 25 is a diagram showing an example of a write operation (second);

FIG. 26 is a waveform chart showing changes of main signals in a write operation;

FIG. 27 is a diagram showing characteristics of a magnetic memory according to a fourth embodiment;

FIG. 28 is a diagram showing the start of the write operation (first);

FIG. 29 is a diagram showing the end of the write operation (first);

FIG. 30 is a diagram showing the start of the write operation (second) of a selected bit;

FIG. 31 is a diagram showing the end of the write operation (second) of the selected bit;

FIG. 32 is a diagram showing the start of the write operation (second) of a non-selected bit;

FIG. 33 is a diagram showing the end of the write operation (second) of the non-selected bit;

FIG. 34 is a diagram showing characteristics of a magnetic memory according to a fifth embodiment;

FIG. 35 is a diagram showing the start of the write operation (first);

FIG. 36 is a diagram showing the end of the write operation (first);

FIG. 37 is a diagram showing the start of the write operation (second) of the selected bit;

FIG. 38 is a diagram showing the end of the write operation (second) of the selected bit;

FIG. 39 is a diagram showing the start of the write operation (second) of the non-selected bit;

FIG. 40 is a diagram showing the end of the write operation (second) of the non-selected bit; and

FIGS. 41A and 41B are signal waveform diagrams shown for explaining an example in which a write error rate is reduced.

DETAILED DESCRIPTION

In general, according to one embodiment, a nonvolatile memory comprises: a conductive line including a first portion, a second portion and a third portion therebetween; a storage element including a first magnetic layer, a second magnetic layer and a nonmagnetic layer therebetween, and the first magnetic layer being connected to the third portion; and a circuit flowing a write current between the first and second portions, applying a first potential to the second magnetic layer, and blocking the write current flowing between the first and second portions after changing the second magnetic layer from the first potential to a second potential.

Hereinafter, the embodiments will be described with reference to the drawings.

First Embodiment

FIG. 1 shows a magnetic memory according to a first embodiment.

The magnetic memory is what is called a SOT (Spin-Orbit Torque) magnetic memory.

A conductive wire 11 has a first portion E₁, a second portion E₂, and a third portion E₃ therebetween. For example, the first and second portions E₁, E₂ correspond to two ends of the conductive wire 11 in a direction in which the conductive wire 11 extends and the third portion E₃ corresponds to a center portion of the conductive wire 11.

A storage element MTJ is a 2-terminal element having a first terminal and a second terminal.

For example, the storage element MTJ is a magnetoresistive effect element. In this case, the storage element MTJ includes a first magnetic layer (first terminal) FL having a variable magnetization direction, a second magnetic layer (second terminal) RL having an invariable magnetization direction, and a nonmagnetic layer (tunnel barrier layer) TN between the first and second magnetic layers FL, RL and the first magnetic layer FL is connected to the third portion E₃.

A first circuit 12 can generate one of a first current I_{w_ap} and a second current I_{w_p} opposite to each other between the first and second portions E₁, E₂.

For example, the first circuit 12 includes driver/sinkers D/S_A, D/S_B capable of generating one of the first current I_{w_ap} and the second current I_{w_p} between the first and second portions E₁, E₂ in accordance with write data (0 or 1) and a transfer gate TG.

In this case, when the write data is 1, for example, the driver/sinker D/S_A outputs V_{dd_W1} (positive potential) and the driver sinker D/S_B outputs a ground potential V_ss. When a control signal φ₃ becomes active (1), the transfer gate TG is turned on and a write pulse WP_A is generated. Thus, the first current I_write (=I_{w_ap}) flows from the first portion E₁ toward the second portion E₂.

Also, when the write data is 0, for example, the driver/sinker D/S_B outputs V_{dd_W1} (positive potential) and the driver sinker D/S_A outputs the ground potential V_ss. When the control signal 93 becomes active (1), the transfer gate TG is turned on and a write pulse WP_B is generated. Thus, the second current I_write (=I_{w_p}) flows from the second portion E₂ toward the first portion E₁.

In a write operation, a second circuit 13 can apply one of a first potential V₁ and a second potential V₂ that are difficult from each other to the second magnetic layer (second terminal) RL of the storage element MTJ. Also, in a read operation, the second circuit 13 can apply a read potential V_read to the second magnetic layer (second terminal) RL of the storage element MTJ.

For example, the second circuit 13 includes a selector 14, for example, a multiplexer MUX that outputs one of the first potential V₁, the second potential V₂, and the read potential V_read based on a control signal φ₁. The potential output from the selector 14 is applied to the second magnetic layer (second terminal) RL of the storage element MTJ.

In this case, the control signal φ₁ is active (01) or non-active (00) in a write operation. When the control signal φ₁ is active (01), for example, the selector 14 selects the first potential V₁. The first potential V₁ is, for example, a negative potential. The first potential V₁ is different from the potential of the third portion E₃ when the first or second current I_write (I_{w_ap} or I_{w_p}) flows between the first and second portions E₁, E₂.

That is, the first potential V₁ is an assist potential V_assist to generate a voltage that assists in reversing magnetization of the first magnetic layer FL between the second magnetic layer RL of the storage element MTJ and the third portion E₃ of the conductive wire 11 when the first or second current I_write (I_{w_ap} or I_{w_p)} flows between the first and second portions E₁, E₂.

When the control signal φ₁ is non-active (00), for example, the selector 14 selects the second potential V₂. The second potential V₂ is, for example, the ground potential V_ss. The second potential V₂ is a potential on standby, that is, when none of the write operation and read operation is performed.

Also, the control signal φ₁ is active (10) or non-active (00) in a read operation. When the control signal φ₁ is active (10), for example, the selector 14 selects the read potential V_read. The read potential V_read is, for example, a positive potential.

A controller 15 controls read operations and write operations.

In a write operation, for example, the controller 15 makes the control signal φ₁ active (01)/non-active (00) and applies the first potential V₁ or the second potential V₂ to the second magnetic layer RL of the storage element MTJ. Also, the controller 15 makes the control signal φ₃ active/non-active and generates the first or second current I_write (I_{w_ap} or I_{w_p}) between the first and second portions E₁, E₂.

In this case, the controller 15 controls the potential of the second magnetic layer RL of the storage element MTJ and the first or second current I_write (I_{w_ap} or I_{w_p}) in the order below.

First, for example, the controller 15 writes first data (1) or second data (0) into the storage element MTJ by passing the first or second current I_write (I_{w_ap} or I_{w_p}) between the first and second portions E₁, E₂ and applying the first potential V₁ to the second magnetic layer (second terminal) RL of the storage element MTJ.

For example, the first data is written into the storage element MTJ when the first current I_{w_ap} is passed to between the first and second portions E₁, E₂ and the second data is written into the storage element MTJ when the second current I_{w_ap} is passed to between the first and second portions E₁, E₂.

Here timing t1 when the first or second current I_write (I_{w_ap} or I_{w_p}) is passed to between the first and second portions E₁, E₂ and timing t2 when the first potential V₁ is applied to the second magnetic layer RL of the storage element MTJ may be the same or different.

For example, as shown in FIG. 2, the timing t1 may be before the timing t2 or, as shown in FIG. 3, the timing t1 may be after the timing t2. Also, as shown in FIG. 4, the timing t1 and the timing t2 may be the same.

Next, after writing the first data into the storage element MTJ, the controller 15 changes the potential of the second magnetic layer (second terminal) RL of the storage element MTJ from the first potential V₁ to the second potential V₂. Then, the controller 15 shuts off the first or second current I_write (I_{w_ap} or I_{w_p}) between the first and second portions E₁, E₂.

That is, for example, as shown in FIGS. 2 to 4, timing t3 to change the potential of the second magnetic layer RL of the storage element MTJ from the first potential V₁ to the second potential V₂ is before timing t4 when the first or second current I_write (I_{w_ap} or I_{w_p}) between the first and second portions E₁, E₂ is shut off.

In a write operation, for example, the controller 15 makes the control signal φ₁ active (10)/non-active (00) and applies the read potential V_read to the second magnetic layer RL of the storage element MTJ. In a read operation, a read current flows between the second magnetic layer RL of the storage element MTJ and the third portion E₃ of the conductive wire 11.

That is, the path through which the read current flows as a write current is different from the path through which the first or second current I_write (I_{w_ap} or I_{w_p}) flows. Thus, even if the read current is set to be relatively large, a situation in which an erroneous write is caused by the read current can be inhibited.

To increase the effect still more, the second potential V₂ is desirably between the first potential V₁ and the read potential V_read. This will be described below.

In the magnetic memory in FIG. 1, the conductive wire 11 desirably has a material and a thickness capable of controlling the magnetization direction of the first magnetic layer FL of the storage element MTJ by the spin orbit coupling or Rashba effect. For example, the conductive wire 11 contains a metal such as tantalum (Ta), tungsten (W), or platinum (Pt) and has a thickness of 5 to 20 nm (for example, about 10 nm).

In this case, if the first or second current I_write (I_{w_ap} or I_{w_p}) is passed to the conductive wire 11, SOT (Spin-Orbit Torque) acts on the first magnetic layer (storage layer) FL of the storage element MTJ and thus, the magnetization direction of the first magnetic layer (storage layer) FL can be reversed. If, at this point, the above assist voltage is applied to the storage element MTJ, magnetic characteristics of the first magnetic layer FL are modulated by the field effect and the first or second current I_write (I_{w_ap} or I_{w_p}) needed to reverse the magnetization direction of the first magnetic layer FL can be made smaller.

Such a situation is shown in FIGS. 5 to 14.

That is, as shown in FIGS. 5 to 14, a first threshold line Th_p showing a boundary of whether the relation of magnetization directions of the first and second magnetic layers FL, RL is set to a parallel state and a second threshold line Th_ap showing a boundary of whether the relation of magnetization directions of the first and second magnetic layers FL, RL is set to an antiparallel state have fixed inclinations in a graph in which a current I_SO flowing between the first and second portions E₁, E₂ is as the x axis and the potential V_assist applied to the second magnetic layer RL of the storage element MTJ is set as the y axis.

When, for example, as shown in FIGS. 5 to 9, a negative potential is applied to the second magnetic layer RL of the storage element MTJ as V_assist, the current I_SO needed to reverse the magnetization direction of the first magnetic layer FL becomes smaller, that is, a first case (first characteristic) where the first and second threshold lines Th_p, Th_ap are open upward is created.

Also, when, as shown in FIGS. 10 to 14, a positive potential is applied to the second magnetic layer RL of the storage element MTJ as V_assist, the current I_SO needed to reverse the magnetization direction of the first magnetic layer FL becomes smaller, that is, a second case (second characteristic) where the first and second threshold lines Th_p, Th_ap are open downward is created.

However, in the first and second cases, a point X where the current I_SO flowing between the first and second portions E₁, E₂ is 0 and the potential V_assist applied to the second magnetic layer RL of the storage element MTJ is 0 is assumed to be the initial state. Also in the present example, only a voltage assist effect caused by V_assist is considered and the STT (Spin Transfer torque) effect accompanying V_assist is not considered. The STT effect accompanying V_assist will be described below.

Also, P indicates an area in which the relation of magnetization directions of the first and second magnetic layers FL, RL changes to a parallel state and AP indicates an area in which the relation of magnetization directions of the first and second magnetic layers FL, RL changes to an antiparallel state. P/AP indicates an area in which a parallel state is maintained when the relation of magnetization directions of the first and second magnetic layers FL, RL is the parallel state and an antiparallel state is maintained when the relation of magnetization directions of the first and second magnetic layers FL, RL is the antiparallel state.

The parallel state is a relation in which the magnetization directions of the first and second magnetic layers FL, RL are mutually the same direction and the antiparallel state is a relation in which the magnetization directions of the first and second magnetic layers FL, RL are mutually reverse directions.

Then, what can be known from the first case (FIGS. 5 to 9) is that the current I_SO needed to reverse the magnetization direction of the first magnetic layer FL can be made smaller by applying a negative potential to the second magnetic layer RL of the storage element MTJ as V_assist.

When, for example, as shown in FIG. 5, margins Δ_{w_p}, Δ_{w_ap} from the first and second threshold lines Th_p, Th_ap are secured in consideration of thermal disturbance in a write operation, write currents I_{w_p}, I_{w_ap} when V_assist is a negative potential are smaller than write currents I_{w_p}′, I_{w_ap}′ when V_assist is 0V. That is, write points W_p, W_ap can be set closer to 0 than write points W_p′, W_ap′.

In this case, for example, as shown in FIG. 9, the read potential V_read is desirably a potential of polarity that makes reversal of the magnetization direction of the first magnetic layer FL of the storage element MTJ difficult in a read operation. That is, in the first case (FIGS. 5 to 9), a distance Δr between a read point R and the first and second threshold lines Th_p, Th_ap increases in a direction in which V_assist is a positive potential and thus, the read potential V_read is desirably a positive potential.

Therefore, the second potential (for example, the ground potential V_ss) V₂ becomes a potential between the first potential (for example, a negative potential) V₁ and the read potential (for example, a positive potential) V_read.

However, the read potential V_read can be set to between the first potential V₁ and the second potential V₂.

Also, what can be known from the second case (FIGS. 10 to 14) is that the current I_SO needed to reverse the magnetization direction of the first magnetic layer FL can be made smaller by applying a positive potential to the second magnetic layer RL of the storage element MTJ as V_assist.

When, for example, as shown in FIG. 10, the margins Δ_{w_p}, Δ_{w_ap} from the first and second threshold lines Th_p, Th_ap are secured in consideration of thermal disturbance in a write operation, write currents I_{w_p}, I_{w_ap} when V_assist is a positive potential are smaller than write currents I_{w_p}′, I_{w_ap}′ when V_assist is 0V. That is, write points W_p, W_ap can be set closer to 0 than write points W_p′. W_ap′.

In this case, for example, as shown in FIG. 14, the read potential V_read is desirably a potential of polarity that makes reversal of the magnetization direction of the first magnetic layer FL of the storage element MTJ difficult in a read operation. That is, in the second case (FIGS. 10 to 14), the distance ar between the read point R and the first and second threshold lines Th_p, Th_ap increases in a direction in which V_assist is a negative potential and thus, the read potential V_read is desirably a negative potential.

Therefore, the second potential (for example, the ground potential V_ss) V₂ becomes a potential between the first potential (for example, a positive potential) V₁ and the read potential (for example, a negative potential) V_read.

However, the read potential V_read can be set to between the first potential V₁ and the second potential V₂.

In the first and second cases (FIGS. 5 to 14), the first or second current I_write (I_{w_ap} or I_{w_p}) flows between the first and second portions E₁, E₂ in a write operation. That is, while the first or second current I_write (I_{w_ap} or I_{w_p}) flows, the third portion E₃ has a predetermined potential (for example, a positive potential).

Therefore, in consideration of the predetermined potential generated in the third portion E₃ while the first or second current I_write (I_{w_ap} or I_{w_p}) flows, the first potential V₁ is set such that an appropriate assist voltage is applied to the storage element MTJ. That is, in the first case, the first potential V₁ may be, instead of a negative potential, 0V or a positive potential. In the second case, the first potential V₁ may be, instead of a positive potential, 0V or a negative potential.

A write operation (0-write) in the first case (FIGS. 5 to 9) is started by, as shown in FIG. 5, setting V_assist to the first potential V₁ and I_SO to the write current I_{w_p}. The order thereof may be that, as shown in FIG. 5, I_SO is set to the write current I_{w_p} after V_assist is set to the first potential V₁ (steps ST₀₁→ST₀₂) or V_assist is set to the first potential V₁ after I_SO is set to the write current I_{w_p} (steps ST₀₃→ST₀₄).

The write operation (0-write) in the first case is terminated by setting, as shown in FIG. 6, V_assist to the second potential V₂ and then I_SO to 0 (steps ST₀₅→ST₀₆). This is because, as shown in FIG. 6, by taking the route from step ST₀₅ to step ST₀₆, the minimum margin between the route and the second threshold line Th_ap becomes Δ.

When, for example, the route from step ST₀₇ to step ST₀₈ is taken, the minimum margin Δ becomes larger than a minimum margin Δ′ between the route and the second threshold line Th_ap. Therefore, when the write operation (0-write) is terminated, a 1-write is not erroneously generated due to thermal disturbance or the like so that a write error rate can be reduced.

It is assumed here that a 0-write means a write operation that puts the storage element MTJ into a parallel state (low-resistance state).

A write operation (1-write) in the first case (FIGS. 5 to 9) is started by, as shown in FIG. 7, setting V_assist to the first potential V₁ and I_SO to the write current I_{w_ap}. The order thereof may be that, as shown in FIG. 7, I_SO is set to the write current I_{w_ap} after V_assist is set to the first potential V₁ (steps ST₁₁→ST₁₂) or V_assist is set to the first potential V₁ after I_SO is set to the write current I_{w_ap} (steps ST₁₃→ST₁₄).

The write operation (1-write) in the first case is terminated by setting, as shown in FIG. 8, V_assist to the second potential V₂ and then I_SO to 0 (steps ST₁₅→ST₁₆). This is because, as shown in FIG. 8, by taking the route from step ST₁₅ to step ST₁₆, the minimum margin between the route and the first threshold line Th_p becomes Δ.

When, for example, the route from step ST₁₇ to step ST₁₈ is taken, the minimum margin Δ becomes larger than a minimum margin Δ′ between the route and the first threshold line Th_p. Therefore, when the write operation (1-write) is terminated, a 0-write is not erroneously generated due to thermal disturbance or the like so that a write error rate can be reduced.

It is assumed here that a 1-write means a write operation that puts the storage element MTJ into an antiparallel state (high-resistance state).

A read operation in the first case (FIGS. 5 to 9) is performed by, as shown in FIG. 9, setting V_assist to the read potential V_read. The write current I_SO is 0 in the read operation and thus, a 0-write or 1-write is not generated. In the read operation, however, in consideration of thermal disturbance or the like, it is desirable to make a margin Δr between the read point R and the first and second threshold lines Th_p, Th_ap as large as possible.

Therefore, the read point R is desirably set in an opening direction of the first and second threshold lines Th_p, Th_ap, that is, in a direction in which the width of the first and second threshold lines Th_p, Th_ap broadens. In the present example, the read point R is set in a direction in which the read potential V_read becomes a positive potential.

A write operation (0-write) in the second case (FIGS. 10 to 14) is started by, as shown in FIG. 10, setting V_assist to the first potential V₁ and I_SO to the write current I_{w_p}. The order thereof may be that, as shown in FIG. 10, I_SO is set to the write current I_{w_p} after V_assist is set to the first potential V₁ (steps ST₂₁→ST₂₂) or V_assist is set to the first potential V₁ after I_SO is set to the write current I_{w_p}(steps ST₂₃→ST₂₄).

The write operation (0-write) in the second case is terminated by setting, as shown in FIG. 11, V_assist to the second potential V₂ and then I_SO to 0 (steps ST₂₅→ST₂₆). This is because, as shown in FIG. 11, by taking the route from step ST₂₅ to step ST₂₆, the minimum margin between the route and the second threshold line Th_ap becomes Δ.

When, for example, the route from step ST₂₇ to step ST₂₈ is taken, the minimum margin Δ becomes larger than a minimum margin Δ′ between the route and the second threshold line Th_ap. Therefore, when the write operation (0-write) is terminated, a 1-write is not erroneously generated due to thermal disturbance or the like so that a write error rate can be reduced.

A write operation (1-write) in the second case (FIGS. 10 to 14) is started by, as shown in FIG. 12, setting V_assist to the first potential V₁ and I_SO to the write current I_{w_ap}. The order thereof may be that, as shown in FIG. 12, I_SO is set to the write current I_{w_ap} after V_assist is set to the first potential V₁ (steps ST₃₁→ST₃₂) or V_assist is set to the first potential V₁ after I_SO is set to the write current I_{w_ap} (steps ST₃₃→ST₃₄).

The write operation (1-write) in the second case is terminated by setting, as shown in FIG. 13, V_assist to the second potential V₂ and then I_SO to 0 (steps ST₃₅→ST₃₆). This is because, as shown in FIG. 13, by taking the route from step ST₃₅ to step ST₃₆, the minimum margin between the route and the first threshold line Th_p becomes Δ.

When, for example, the route from step ST₃₇ to step ST₃₈ is taken, the minimum margin Δ becomes larger than a minimum margin Δ′ between the route and the first threshold line Th_p. Therefore, when the write operation (1-write) is terminated, a 0-write is not erroneously generated due to thermal disturbance or the like so that a write error rate can be reduced.

A read operation in the second case (FIGS. 10 to 14) is performed by, as shown in FIG. 14, setting V_assist to the read potential V_read. The write current I_SO is 0 in the read operation and thus, a 0-write or 1-write is not generated. In the read operation, however, in consideration of thermal disturbance or the like, it is desirable to make a margin Δr between the read point R and the first and second threshold lines Th_p, Th_ap as large as possible.

Therefore, the read point R is desirably set in an opening direction of the first and second threshold lines Th_p, Th_ap, that is, in a direction in which the width of the first and second threshold lines Th_p, Th_ap broadens. In the present example, the read point R is set in a direction in which the read potential V_read becomes a negative potential.

To further reduce the write error rate, as shown in FIG. 41A, a third potential V₃ can be added during writing. At this time, the third potential V₃ is a potential of polarity that makes reversal of the magnetization direction of the first magnetic layer FL of the storage element MTJ difficult (for example, a negative potential). Thus, the write error rate can be reduced by increasing the resistance to the thermal agitation in comparison with the standby state and preventing back-hopping.

As shown in FIG. 41B, slope may be actively added to each of the voltages V₁ to V₃ and the write current I_write. In particular, when the temperature of the write line is increased by the write current I_write, the error rate can be decreased by writing data with slope.

This modification can be also applied to the following embodiments.

Second Embodiment

FIG. 15 shows a magnetic memory according to a second embodiment.

The magnetic memory is what is called a SOT magnetic memory.

A conductive wire 11 has a first portion E₁, a second portion E₂, and a third portion E₃ therebetween. For example, the first and second portions E₁, E₂ correspond to two ends of the conductive wire 11 in a direction in which the conductive wire 11 extends and the third portion E₃ corresponds to a center portion of the conductive wire 11.

Storage elements MTJ₁, MTJ₂ are 2-terminal elements having a first terminal and a second terminal.

For example, the storage elements MTJ₁, MTJ₂ are a magnetoresistive effect element. In this case, the storage elements MTJ₁, MTJ₂ include a first magnetic layer (first terminal) FL having a variable magnetization direction, a second magnetic layer (second terminal) RL having an invariable magnetization direction, and a nonmagnetic layer (tunnel barrier layer) TN between the first and second magnetic layers FL, RL and the first magnetic layer FL is connected to the third portion E₃.

A first circuit 12 can generate one of a first current I_{w_ap} and a second current I_{w_p} opposite to each other between the first and second portions E₁, E₂.

For example, the first circuit 12 includes driver/sinkers D/S_A, D/S_B capable of generating one of the first current I_{w_ap} and the second current I_{w_p} between the first and second portions E₁, E₂ in accordance with write data (0 or 1) and a transfer gate TG.

In this case, when the write data is 1, for example, the driver/sinker D/S_A outputs V_{dd_W1} (positive potential) and the driver sinker D/S_B outputs a ground potential V_ss. When a control signal φ₃ becomes active (1), the transfer gate TG is turned on and a write pulse WP_A is generated. Thus, the first current I_write (=I_{w_ap}) flows from the first portion E₁ toward the second portion E₂.

Also, when the write data is 0, for example, the driver/sinker D/S_B outputs V_{dd_W1} (positive potential) and the driver sinker D/S_A outputs the ground potential V_ss. When the control signal φ₃ becomes active (1), the transfer gate TG is turned on and a write pulse WP_B is generated. Thus, the second current I_write (=I_{w_p}) flows from the second portion E₂ toward the first portion E₁.

In a write operation, second circuits 13₁, 13₂ can apply one of a first potential V₁, a second potential V₂, and a third potential V₃ that are difficult from each other to the second magnetic layer (second terminal) RL of the storage elements MTJ₁, MTJ₂. Also, in a read operation, the second circuits 13₁, 13₂ can apply a read potential V_read to the second magnetic layer (second terminal) RL of the storage element MTJ.

For example, the second circuits 13₁, 13₂ include selectors 14₁, 14₂, for example, multiplexers MUX that output one of the first potential V₁, the second potential V₂, the third potential V₃, and the read potential V_read based on control signals φ₁₁, φ₁₂ respectively. The potential output from the selectors 14₁, 14₂ is applied to the second magnetic layer (second terminal) RL of the storage elements MTJ₁, MTJ₂.

In this case, in a write operation, the selectors 14₁, 14₂ select the first potential V₁ or the third potential V₃ based on the control signals φ₁₁, φ₁₂ respectively. The first potential V₁ is an assist potential that enables a write operation, for example, a negative potential. The third potential V₃ is an inhibit potential that inhibits a write operation, for example, a positive potential. The first and third potentials V₁, V₃ are different from the potential of the third portion E₃ when a first or second current I_write (I_{w_ap} or I_{w_p}) flows between the first and second portions E₁, E₂.

In a read operation, the selector 14 selects the read potential V_read. The read potential V_read is, for example, a positive potential.

The second potential V₂ is a potential selected by the selectors 14₁, 14₂ on standby, that is, when none of the write operation and read operation is performed.

A controller 15 controls read operations and write operations.

For example, a case in which the storage element MTJ₁ is selected to be written into and the storage element MTJ₂ is not selected to be written into in a write operation will be considered.

In this case, a controller 15 transfers the control signal φ₁₁ to a second circuit 13₁. The selector 14₁ applies the second potential V₂ to the second magnetic layer RL of the storage element MTJ₁ based on the control signal φ₁₁. Also, the controller 15 transfers the control signal φ₁₂ to a second circuit 13₂. The selector 14₂ applies the third potential V₃ to the second magnetic layer RL of the storage element MTJ₂ based on the control signal φ₁₂.

Further, the controller 15 transfers the control signal φ₃ to the first circuit 12. The first circuit 12 generates the first or second current I_write (I_{w_ap} or I_{w_p}) between the first and second portions E₁, E₂ based on the control signal φ₃.

Then, the controller 15 controls the potential of the second magnetic layer RL of the storage elements MTJ₁, MTJ₂ and the first or second current I_write (I_{w_ap} or I_{w_p}) in the order below.

The controller 15 applies the third potential V₃ to the second magnetic layer RL of the storage element MTJ₂. Then, the controller 15 passes the first or second current I_write (I_{w_ap} or I_{w_p}) to between the first and second portions E₁, E₂. Due to this order, data is inhibited from being erroneously written into the storage element MTJ₂ not to be written into.

On the other hand, the controller 15 applies the first potential V₁ to the second magnetic layer RL of the storage element MTJ₁. The timing when the first potential V₁ is applied to the second magnetic layer RL of the storage element MTJ₁ may be, as described with reference to FIGS. 2 to 4, after the first or second current I_write (I_{w_ap} or I_{w_p}) is passed to between the first and second portions E₁, E₂ or before the first or second current I_write (I_{w_ap} or I_{w_p}) is passed to between the first and second portions E₁, E₂.

The timing when the first potential V₁ is applied to the second magnetic layer RL of the storage element MTJ₁ may be the same as the timing when the first or second current I_write (I_{w_ap} or I_{w_p}) is passed to between the first and second portions E₁, E₂.

Then, the first or second current I_write (I_{w_ap} or I_{w_p}) flows between the first and second portions E₁, E₂ and the first potential V₁ is applied to the second magnetic layer RL of the storage element MTJ₁ to write, for example, first data (1) or second data (0) into the storage element MTJ₁.

Next, after the first or second data is written into the storage element MTJ₁, the controller 15 changes the potential of the second magnetic layer RL of the storage element MTJ₁ from the first potential V₁ to the second potential V₂. Then, the controller 15 shuts off the first or second current I_write (I_{w_ap} or I_{w_p}) between the first and second portions E₁, E₂.

That is, the timing to change the potential of the second magnetic layer RL of the storage element MTJ₁ as a write target from the first potential V₁ to the second potential V₂ is before the timing when the first or second current I_write (I_{w_ap} or I_{w_p}) between the first and second portions E₁, E₂ is shut off. Due to this order, data opposite to write data is prevented from being erroneously stored in the storage element MTJ₁ when the write operation is terminated.

Also, after the first or second data is written into the storage element MTJ₁, the controller 15 shuts off the first or second current I_write (I_{w_ap} or I_{w_p}) between the first and second portions E₁, E₂. Then, the controller 15 changes the potential of the second magnetic layer RL of the storage element MTJ₂ from the third potential V₃ to the second potential V₂.

That is, the timing to change the potential of the second magnetic layer RL of the storage element MTJ₂ not to be written into from the third potential V₃ to the second potential V₂ is after the timing when the first or second current I_write (I_{w_ap} or I_{w_p}) between the first and second portions E₁, E₂ is shut off. Due to this order, write data is prevented from being erroneously written into the storage element MTJ₂ when the write operation is terminated.

The read operation is the same as in the first embodiment and thus, the description thereof here is omitted.

Third Embodiment

FIG. 16 shows a magnetic memory according to a third embodiment.

The magnetic memory is SOT-MRAM.

SOT-MRAM 31 includes an interface 32, an internal controller 33, a memory cell array 34, and a word line decoder/driver 35. The memory cell array 34 includes n blocks (memory cores) BK_1 to BK_n. Where n is a natural number equal to 2 or greater.

A command CMD is transferred to the internal controller 33 via the interface 32. The command CMD includes, for example, a read command, a write command and the like.

When the command CMD is received, the internal controller 33 outputs, for example, control signals WE₁ to WE_n, WE1/2, W_{sel_1} to W_{sel_n} to execute the command CMD. The meanings and roles of these control signals will be described below.

An address signal Addr is transferred to the internal controller 33 via the interface 32. The address signal Addr is divided into a row address A_row and column addresses A_{col_1} to A_{col_n} in the interface 32. The row address A_row is transferred to the word line decoder/driver 35. The column addresses A_{col_1} to A_{col_n} are transferred to the n blocks BK_1 to BK_n.

DA₁ to DA_n are read data or write data transmitted/received in a read operation or write operation.

Each block BK_k includes a sub-array A_{sub_k}, a read/write circuit 36, and a column selector 37.

The column selector 37 selects one of j columns (j is a natural number equal to 2 or greater) CoL₁ to CoL_j and electrically connects the selected one column CoLp (p is one of 1 to j) to the read/write circuit 36.

The sub-array A_{sub_k} includes, for example, a cell unit CU_ij. The cell unit CU_ij includes memory cells MC₁ to MC₈ and transistors Q_S, Q_W. The transistors Q_S, Q_W are, for example, N-channel FET (Field effect transistor).

FIGS. 17 to 22 show examples of the cell unit CU_ij in FIG. 16.

A conductive wire 11 extends in a first direction. The cell unit CU_ij corresponds to the conductive wire 11 and includes a plurality of memory cells MC₁ to MC₈. The plurality of memory cells MC₁ to MC₈ includes eight memory cells in the present example, but the present embodiment is not limited to such an example. For example, the plurality of memory cells MC₁ to MC₈ may include two memory cells or more.

The plurality of memory cells MC₁ to MC₈ includes storage elements MTJ₁ to MTJ₈ and transistors T₁ to T₈ respectively.

The storage elements MTJ₁ to MTJ₈ are each magnetoresistive effect elements. For example, each of the storage elements MTJ₁ to MTJ₈ includes a first magnetic layer (storage layer) 22 having a variable magnetization direction, a second magnetic layer (reference layer) 23 having an invariable magnetization direction, and a nonmagnetic layer (tunnel barrier layer) 24 between the first and second magnetic layers 22, 23 and the first magnetic layer 22 is connected to the conductive wire 11.

In this case, the conductive wire 11 desirably has a material and a thickness capable of controlling the magnetization direction of the first magnetic layer of the storage elements MTJ₁ to MTJ₈ by the spin orbit coupling or Rashba effect. For example, the conductive wire 11 contains a metal such as tantalum (Ta), tungsten (W), or platinum (Pt) and has a thickness of 5 to 20 nm (for example, about 10 nm).

The transistors T₁ to T₈ are each, for example, N-channel FET (Field effect transistor). The transistors T₁ to T₈ are desirably what is called vertical transistors arranged on a semiconductor substrate and in which a channel (current path) is in a longitudinal direction in which the channel intersects the surface of the semiconductor substrate.

The storage element MTJd (d is one of 1 to 8) has a first terminal (storage layer) and a second terminal (reference layer) and the first terminal is connected to the conductive wire 11. The transistor T_d has a third terminal (source/drain), a fourth terminal (source/drain), a channel (current path) between the third and fourth terminals, and a control electrode (gate) that controls the generation of a channel and the third terminal is connected to the second terminal.

The conductive wires WL₁ to WL₈ extend, for example, in the first direction and are connected to control electrodes of the transistors T₁ to T₈. Conductive wires LBL₁ to LBL₈ each extend, for example, in a second direction intersecting the first direction and are connected to the fourth terminal of the transistors T₁ to T₈.

The transistor Q_S has a channel (current path) connected between the first portion E₁ of the conductive wire 11 and a conductive wire SBL and a control terminal (gate) that controls the generation of a channel. The transistor Q_W has a channel (current path) connected between the second portion E₂ of the conductive wire 11 and a conductive wire WBL and a control terminal (gate) that controls the generation of a channel.

A conductive wire SWL extends, for example, in the first direction and is connected to control electrodes of the transistors Q_S, Q_W. The conductive wires SBL, WBL each extend, for example, in the second direction.

In the present example, the transistor Q_S is connected to the first portion E₁ of the conductive wire 11 and the transistor Q_W is connected to the second portion E₂ of the conductive wire 11, but one of these transistors may be omitted.

In the example of FIG. 17, the conductive wire 11 is arranged in an upper portion of a semiconductor substrate 41 and the transistors Q_S, Q_W are arranged in a surface area of the semiconductor substrate 41 as what is called horizontal transistor (FET). Here, the horizontal transistor is a transistor in which a channel (current path) is in a direction along the surface of the semiconductor substrate 41.

The storage elements MTJ₁ to MTJ₈ are arranged on the conductive wire 11 and the transistors T₁ to T₈ are arranged on the storage elements MTJ₁ to MTJ₈. The transistors T₁ to T₈ are what is called vertical transistors. Also, the conductive wires LBL₁ to LBL₈, SBL_j, WBL_j are arranged on the transistors T₁ to T₈.

In the example of FIG. 18, the conductive wire 11 is arranged in the upper portion of the semiconductor substrate 41 and the transistors Q_S, Q_W and the storage elements MTJ₁ to MTJ₈ are arranged on the conductive wire 11. The transistors T₁ to T₈ are arranged on the storage elements MTJ₁ to MTJ₈. The transistors Q_S, Q_W and the transistors T₁ to T₈ are what is called vertical transistors.

The conductive wires LBL₁ to LBL₈ are arranged on the transistors T₁ to T₈ and the conductive wires SBL_j, WBL_j are arranged on the transistors Q_S, Q_W.

In the example of FIG. 19, the conductive wires LBL₁ to LBL₈, SBL_j, WBL_j are arranged in the upper portion of the semiconductor substrate 41. The transistors T₁ to T₈ are arranged on the conductive wires LBL₁ to LBL₈ and the transistors Q_S, Q_W are arranged on the conductive wires SBL_j, WBL_j. The storage elements MTJ₁ to MTJ₈ are arranged on the transistors T₁ to T₈.

Also, the conductive wire 11 is arranged on the transistors T₁ to T₈ and the transistors Q_S, Q_W. The transistors Q_S, Q_W and the transistors T₁ to T₈ are what is called vertical transistors.

In the examples of FIGS. 17 to 19, the first and second magnetic layers 22, 23 have an easy-axis of magnetization in an in-plane direction along the surface of the semiconductor substrate 41 and in the second direction intersecting the first direction in which the conductive wire 11 extends.

For example, FIG. 20 shows an example of the device structure of a memory cell MC₁ in FIGS. 17 and 19. In this example, the transistor T₁ includes a semiconductor pillar (for example, a silicon pillar) 25 extending in a third direction intersecting the first and second directions, that is, in a direction intersecting the surface of the semiconductor substrate 41, a gate insulating layer (for example, silicon oxide) 26 covering a side surface of the semiconductor pillar 25, and a conductive wire WL_i covering the semiconductor pillar 25 and the gate insulating layer 26.

In the example of FIG. 20, the easy-axis of magnetization of the first and second magnetic layers 22, 23 is the second direction, but may be, as shown in the example of FIG. 21, the first direction or, as shown in FIG. 22, the third direction. The storage element MTJ₁ in FIGS. 20 and 21 is called an in-plane magnetization magnetoresistive effect element and the storage element MTJ₁ in FIG. 22 is called a perpendicular magnetization magnetoresistive effect element.

Incidentally, the memory cell MC₁ in FIG. 19 is obtained by turning the device structure in FIGS. 20 to 22 upside down.

The memory cell MC₁ in FIGS. 20 to 22 is characterized in that, as described above, the current path of the read current I_read used for read operation and the current path of the write current I_write used for write operation are different.

For example, the read current I_read in a read operation flows from the conductive wire LBL₁ toward the conductive wire 11 or from the conductive wire 11 toward the conductive wire LBL₁. In a write operation, by contrast, the write current I_write flows from right to left or from left to right inside the conductive wire 11.

If the current path of the read current I_read used for read operation and the current path of the write current I_write used for write operation are the same, sufficient margins have to be secured for the read current I_read and the write current I_write in consideration thermal stability to prevent a write phenomenon in a read operation from occurring.

However, the read current I_read and the write current I_write have both become sufficiently small due to increasingly finer structures of memory cells and the like, making it difficult to secure sufficient margins for both.

According to SOT-MRAM in the present example, the current path of the read current I_read and the current path of the write current I_write are different and thus, sufficient margins can be secured for both in consideration thermal stability even if the read current I_read and the write current I_write are both small due to increasingly finer structures of memory cells and the like.

Also, when, as described in the first and second embodiments (FIGS. 1 to 15), a 0/1 write is terminated in a write operation, a write error rate can be reduced by bringing the assist potential V_assist from the first potential V₁ back to the second potential (initial state) V₂ and then shutting off the write current I_write.

FIG. 23 is a diagram showing an example of the read/write circuit in FIG. 16.

The read/write circuit 36 performs a read operation or write operation in a read operation or write operation based on an instruction from the internal controller 33 in FIG. 15.

The read/write circuit 36 includes a read circuit and a write circuit.

Here, however, only the write circuit of the read/write circuit 36 will be described to simplify the description. This is because, like in the first and second embodiments, the third embodiment is characterized by the write operation to reduce the write error rate.

The write circuit includes ROM 45, 47, selectors (multiplexers) 46, 49, 51₁ to 51₈, write drivers/sinkers D/S_A, D/S_B, a transfer gate TG, a data register 48, voltage assist drivers 50₁ to 50₈, a delay circuit D, and select transistors (for example, N channel FET) T_S, T_U. The write drivers/sinkers D/S_A, D/S_B have the function of generating one of a first current I_{w_ap} and a second current I_{w_p}, which are opposite to each other, in, for example, the conductive wire 11 in FIGS. 17 to 19.

Here, the first current I_{w_ap} is a current to write 1 to, for example, the storage elements MTJ₁ to MTJ₈ in FIGS. 17 to 19 by the spin orbit coupling or Rashba effect, that is, a current to set the relation of magnetization directions of the first and second magnetic layers 22, 23 of the storage elements MTJ₁ to MTJ₈ in FIGS. 17 to 19 to an antiparallel state.

The second current I_{w_p} is a current to write 0 to, for example, the storage elements MTJ₁ to MTJ₈ in FIGS. 17 to 19 by the spin orbit coupling or Rashba effect, that is, a current to set the relation of magnetization directions of the first and second magnetic layers 22, 23 of the storage elements MTJ₁ to MTJ₈ in FIGS. 17 to 19 to a parallel state.

The first current I_{w_ap} and the second current I_{w_p} here correspond to the first current I_{w_ap} and the second current I_{w_p} in the first and second embodiments (FIGS. 1 to 15).

The voltage assist drivers 50₁ to 50₈ have the function of permitting/inhibiting a write operation using the first current I_{w_ap} and the second current I_{w_p}.

When, for example, a write operation is permitted, the voltage assist drivers 50₁ to 50₈ selectively apply the first potential V₁ that makes it easier to perform a write operation as the assist potential V_assist to, for example, the conductive wires LBL₁ to LBL₈ in FIGS. 17 to 19. In this case, an assist voltage that destabilizes the magnetization direction of the first magnetic layer (storage layer) 22 in FIGS. 17 to 19 is generated in the storage elements MTJ₁ to MTJ₈, which makes it easier for the magnetization direction of the first magnetic layer 22 to reverse.

When a write operation is inhibited, the voltage assist drivers 50₁ to 50₈ selectively apply the third potential V₃ that makes it more difficult to perform a write operation as an inhibit potential V_inhibit to, for example, the conductive wires LBL₁ to LBL₈ in FIGS. 17 to 19. In this case, an assist voltage that destabilizes the magnetization direction of the first magnetic layer (storage layer) 22 in FIGS. 17 to 19 is not generated in the storage elements MTJ₁ to MTJ₈ or an inhibit voltage that stabilizes the magnetization direction of the first magnetic layer 22 is generated in the storage elements MTJ₁ to MTJ₈, which makes it more difficult for the magnetization direction of the first magnetic layer 22 to reverse.

When a write operation is inhibited, instead of applying the inhibit potential V_inhibit to the conductive wires LBL₁ to LBL₈, the voltage assist drivers 50₁ to 50₈ may put the conductive wires LBL₁ to LBL₈ into an electrically floating state.

Next, an example of the write operation will be described.

Write Operation

When, for example, a write command CMD is received, the internal controller 33 in FIG. 16 controls a write operation. The internal controller 33 performs a write operation by a first write operation and a second write operation.

The first write operation is an operation to write the same data (for example, 1) into multiple bits (for example, eight bits) as write targets.

First, conductive wires WL₁ to WL₈, SWL are activated by the word line decoder/driver 35 in FIG. 16.

Next, the internal controller 33 in FIG. 16 sets, for example, a control signal WE1/2 to 0. The control signal WE1/2 is a signal to select one of the first write operation and the second write operation and when, for example, the control signal WE1/2 is 0, the first write operation is selected.

In this case, the selector 46 in the read/write circuit 36 in FIG. 23 selects 1 from the ROM 45 and outputs 1 as ROM data (1). Therefore, the driver/sinker D/S_A outputs, for example, a drive potential V_{dd_W1} as a write pulse signal and the driver sinker D/S_B outputs, for example, the ground potential V_ss.

In a write operation, the control signal 93 is activated (high level) and so the transfer gate TG is ON.

Therefore, the write pulse signal is applied to the conductive wire SBL via the transfer gate TG and the ground potential V_ss is applied to the conductive wire WBL via the transfer gate TG. At this point, for example, as shown in FIG. 24, the write current (first write current) I_write flows from the conductive wire SBL_j toward the conductive wire WBL_j, that is, from left to right inside the conductive wire 11.

In the read/write circuit 36 in FIG. 23, the selector 49 selects data stored in the ROM 47 and outputs the data as ROM data (11111111).

Therefore, all of a plurality of the voltage assist drivers 50₁ to 50₈ output, for example, the assist voltage V₁ to a plurality of the conductive wires LBL₁ to LBL₈.

That is, for example, as shown in FIG. 24, the write current (first write current) I_write flows from the conductive wire SBL toward the conductive wire WBL in a state in which the assist potential V₁ is applied to all of the plurality of conductive wires LBL₁ to LBL₈.

As a result, in the first write operation, the same data is written into all of the multiple bits (for example, eight bits) as write targets. It is assumed here that 1 is written in the first write, that is, all of the plurality of storage elements MTJ₁ to MTJ₈ are put into an antiparallel state.

The second write operation is an operation to hold (for example, if the write data is 1) or change from 1 to 0 (for example, if the write data is 0) the same data (for example, 1) written into the multiple bits (for example, eight bits) as write targets in accordance with the writ data.

First, the conductive wires WL₁ to WL₈, SWL are held in an activated state by the word line decoder/driver 35 in FIG. 16.

Next, the internal controller 33 in FIG. 16 sets, for example, the control signal WE1/2 to 1. When, for example, the control signal WE1/2 is 1, the second write operation is selected.

In this case, the selector 46 in the read/write circuit 36 in FIG. 23 selects 0 from the ROM 45 and outputs 0 as ROM data (0). Therefore, the driver/sinker D/S_B outputs, for example, the drive potential V_{dd_W1} as a write pulse signal and the driver sinker D/S_A outputs, for example, the ground potential V_ss.

The write pulse signal is applied to the conductive wire WBL via the transfer gate TG and the ground potential V_ss is applied to the conductive wire SBL via the transfer gate TG. At this point, for example, as shown in FIG. 25, the write current (second write current) I_write flows from the conductive wire WBL toward the conductive wire SBL, that is, from right to left inside the conductive wire 11.

In the read/write circuit 36 in FIG. 23, the selector 49 selects write data (for example, 01011100) stored in the data register 48 and outputs an inverted signal (for example, 10100011) of the write data. The write data is stored in advance in the data register 48 before the second write operation is performed.

Therefore, each of the plurality of voltage assist drivers 50₁ to 50₈ outputs the first potential V₁ as the assist potential V_assist when, for example, the inverted signal of write data is 1 and outputs the third potential V₃ as the inhibit potential V_inhibit when the inverted signal of write data is 0.

That is, when, for example, as shown in FIG. 25, the inverted signal of write data is 10100011, the write current (second write current) I_write flows from the conductive wire WBL_j toward the conductive wire SBL_j in a state in which the first potential V₁ is applied to the conductive wires LBL₁, LBL₃, LBL₇, LBL₈ and the third potential V₃ is applied to the conductive wires LBL₂, LBL₄, LBL₅, LBL₆.

As a result, in the second write operation, data of the storage elements MTJ₁, MTJ₃, MTJ₇, MTJ₈ of multiple bits (for example, eight bits) as write targets is changed from 1 to 0, that is, 0 is written thereinto. Also, data of the storage elements MTJ₂, MTJ₄, MTJ₅, MTJ₆ of multiple bits (for example, eight bits) as write targets holds 1, that is, 1 is written thereinto.

It is assumed here that in the second write operation, 0 is selectively written into the plurality of storage elements MTJ₁ to MTJ₈, that is, the plurality of storage elements MTJ₁ to MTJ₈ is selectively changed from the antiparallel state to the parallel state.

FIG. 26 shows a waveform chart of main signals in the above write operation.

What can be known from the waveform chart is that 1 is written into all the multiple bits (eight bits) in the write operation (first). At this point, the voltage assist is shut off and then the write current I_write is shut off at the end of the write operation. This is intended, as described above, to prevent a 0-write from erroneously occurring in the 1-write.

Also, in the write operation (second), 0 is selectively written into the multiple bits (eight bits). At this point, the voltage assist is shut off and then the write current I_write is shut off from the selected bit intended for 0-write at the end of the write operation. Accordingly, a 1-write is prevented from erroneously occurring in the 0-write.

For non-selected bits not intended for the 0-write, the voltage assist is applied and then the write current I_write is generated at the start of the write operation. Also, the write current I_write is shut off and then the voltage assist is shut off at the end of the write operation. Accordingly, a 0-write does not erroneously occur in the 0-write and 1 written in the write operation (first) can be held unchanged.

The relation (order of the application/shutoff) of the voltage assist and the write current I_write at the start/end of a write operation in the 0-write of a non-selected bit will be described in detail in fourth and fifth embodiments including the reason therefor.

Therefore, the timing of the application/shutoff of the voltage assist and the write current I_write is different between a selected bit and a non-selected bit. The control signal φ₁, the delay circuit D, the transistors T_S, T_U, and the selectors 51₁ to 51₈ in FIG. 23 are elements to implement the timing in FIG. 26.

In FIG. 23, for example, in the write operation (first), the selectors 51₁ to 51₈ select S (select) through control signals φ₂₁ to φ₂₈. In this case, the point at which the assist potential V_assist changes from V₁ to V₂ is before the point at which the write current I_write is shut off.

In the write operation (second), the selectors 51₁ to 51₈ selectively select S (select) or U (unselect) through the control signals φ₂₁ to φ₂₈. For example, the selectors 51₁ to 51₈ corresponding to selected bits intended for 0-write select S (select). Also, the selectors 51₁ to 51₈ corresponding to non-selected bits not intended for 0-write select U (unselect).

In this case, for a selected bit, like in the write operation (first), the point at which the assist potential V_assist changes from V₁ to V₂ is before the point at which the write current I_write is shut off.

For a non-selected bit, the point at which the assist potential V_assist changes from V₂ to V₁ is before the point at which the write current I_write is applied at the start of a write operation. Also for a non-selected bit, the point at which the assist potential V_assist changes from V₁ to V₂ is after the point at which the write current I_write is shut off at the end of a write operation.

Fourth Embodiment

FIG. 27 shows characteristics of a magnetic memory according to a fourth embodiment.

For example, as shown in the upper diagram of FIG. 27, magnetization reversal characteristics of a magnetic memory in consideration of the SOT effect and voltage assist effect exhibit a state in which first and second threshold lines Th_p, Th_ap are open upward. Also, in general, the first and second threshold lines Th_p, Th_ap are bilaterally symmetric with respect to I_SO=0.

However, if an assist potential V_assist is applied to a second magnetic layer RL of a storage element, a flow of electrons in a vertical direction, that is, in a direction in which the first and second magnetic layers FL, RL are stacked arises, in the first storage layer FL, which causes the STT effect.

For example, if the assist potential V_assist becomes increasingly higher than the potential of a third portion E₃ of a conductive wire 11, that is, with an increasing assist potential V_assist, the STT effect by electrons flowing from the first storage layer FL toward the second storage layer RL becomes conspicuous. In this case, electrons having a spin in a direction opposite to the magnetization direction of the second magnetic layer RL generate spin torque in the first storage layer FL and thus, the magnetization directions of the first and second storage layers FL, RL are more likely to be in an antiparallel state.

Therefore, as shown in the middle diagram of FIG. 27, magnetization reversal characteristics of the magnetic memory in consideration of the STT effect shift the first and second threshold lines Th_p, Th_ap to the left, that is, an antiparallel state is more likely to be entered and a parallel state is less likely to be entered with an increasing potential V_STT applied to the second magnetic layer RL.

Similarly, if the assist potential V_assist becomes increasingly lower than the potential of the third portion E₃ of the conductive wire 11, that is, with a decreasing assist potential V_assist, the STT effect by electrons flowing from the second storage layer RL toward the first storage layer FL becomes conspicuous. In this case, electrons having a spin in the same direction as the magnetization direction of the second magnetic layer RL generate spin torque in the first storage layer FL and thus, the magnetization directions of the first and second storage layers FL, RL are more likely to be in a parallel state.

Therefore, as shown in the middle diagram of FIG. 27, magnetization reversal characteristics of the magnetic memory in consideration of the STT effect shift the first and second threshold lines Th_p, Th_ap to the right, that is, a parallel state is more likely to be entered and an antiparallel state is less likely to be entered with a decreasing potential V_STT applied to the second magnetic layer RL.

From the above, as shown in the lower diagram of FIG. 27, magnetization reversal characteristics of the magnetic memory in consideration of the SOT effect, the voltage assist effect, and the STT effect decrease the inclination of the first threshold line Th_p indicating whether to enter a parallel state in a graph of I_SO (x axis)-V_assist (y axis) and increases the inclination of the second threshold line Th_ap indicating whether to enter an antiparallel state.

This means that with increasing V_assist, the first and second magnetic layers FL, RL are less likely to enter a parallel state. That is, it becomes easier to switch selected bits and non-selected bits in a 0-write (write operation to put a storage element into a parallel state) using the inclination of the first threshold line Th_p. On the other hand, it becomes more difficult to switch selected bits and non-selected bits in a 1-write (write operation to put a storage element into an antiparallel state).

Therefore, for example, in the write operation described in the third embodiment, it is desirable to write 1 into all of multiple bits (eight bits) by setting the 1-write as the write operation (first). Also, by setting the 0-write as the write operation (second), 0 can be selectively written into multiple bits (eight bits) using the inclination of the first threshold line Th_p in FIG. 27.

As a result, the write error rate is further reduced in a write operation of the third embodiment. To make the STT effect more conspicuous and improve bit selectivity (make the inclination of the first threshold line Th_p still smaller), for example, techniques of reducing a resistance-area product (RA) of the storage element MTJ or increasing the spin polarization rate of the first and second magnetic layers FL, RL may be combined.

Regarding the STT effect, in contrast to the above description, a parallel state may also be likely to be entered when electrons flow from the first magnetic layer FL toward the second magnetic layer RL and an antiparallel state may also be likely to be entered when electrons flow from the second magnetic layer RL toward the first magnetic layer FL.

In such a case, the middle diagram of the FIG. 27 changes to characteristics in which a parallel state is more likely to be entered with increasing V_STT and an antiparallel state is more likely to be entered with decreasing V_STT. As a result, the lower diagram of FIG. 27 changes to characteristics in which the inclination of the first threshold line Th_p increases and the inclination of the second threshold line Th_ap decreases.

Therefore, in such a case, the 0-write may be set as the write operation (first) and the 1-write may be set as the write operation (second).

Which trend of the above two cases the STT effect exhibits depends on, for example, band filling of the magnetic material used for the first magnetic layer FL to the 3d orbit.

Next, examples of the write operation will be described.

[First Write Operation (all Bits: 1-Write)]

In the first write operation, a 1-write is performed for multiple bits (all bits).

A write operation (1-write) is started by, as shown in FIG. 28, setting V_assist to the first potential V₁ and I_SO to the write current I_{w_ap}. The order thereof may be that, as shown in FIG. 28, I_SO is set to the write current I_{w_ap} after V_assist is set to the first potential V₁ (steps ST₁₁→ST₁₂) or V_assist is set to the first potential V₁ after I_SO is set to the write current I_{w_ap} (steps ST₁₃→ST₁₄).

The write operation (1-write) is terminated by setting, as shown in FIG. 29, V_assist to the second potential V₂ and then I_SO to 0 (steps ST₁₅→ST₁₆). This is because, as shown in FIG. 8, by taking the route from step ST₁₅ to step ST₁₆, the minimum margin between the route and the first threshold line Th_p becomes Δ.

When, for example, the route from step ST₁₇ to step ST₁₈ is taken, the minimum margin Δ becomes larger than a minimum margin Δ′ between the route and the first threshold line Th_p. Therefore, when the write operation (1-write) is terminated, a 0-write is not erroneously generated due to thermal disturbance or the like so that a write error rate can be reduced.

[Second Write Operation: Selected Bit]

In the second write operation, a 0-write is performed for a selected bit for which the 0-write should be performed in the order below:

A write operation (0-write) is started by, as shown in FIG. 30, setting V_assist to the first potential V₁ and I_SO to the write current I_{w_p}. The order thereof may be that, as shown in FIG. 30, I_SO is set to the write current I_{w_p} after V_assist is set to the first potential V₁ (steps ST₀₁→ST₀₂) or V_assist is set to the first potential V₁ after I_SO is set to the write current I_{w_p} (steps ST₀₃→ST₀₄).

The write operation (0-write) is terminated by setting, as shown in FIG. 31, V_assist to the second potential V₂ and then setting I_SO to 0 (steps ST₀₅→ST₀₆). This is because, as shown in FIG. 6, by taking the route from step ST₀₅ to step ST₀₆, the minimum margin between the route and the second threshold line Th_ap becomes Δ.

When, for example, the route from step ST₀₇ to step ST₀₈ is taken, the minimum margin Δ becomes larger than a minimum margin Δ′ between the route and the second threshold line Th_ap. Therefore, when the write operation (0-write) is terminated, a 1-write is not erroneously generated due to thermal disturbance or the like so that a write error rate can be reduced.

[Second Write Operation: Non-Selected Bit]

In the second write operation, a 0-write is performed for a non-selected bit for which the 0-write should not be performed in the order below:

A write operation (0-write) is started by, as shown in FIG. 32, setting V_assist to the third potential V₃ and I_SO to the write current I_{w_p}. The order thereof is that, as shown in FIG. 32, V_assist is set to the third potential V₃ and then I_SO is set to the write current I_{w_p} (steps ST₂₁→ST₂₂).

A point Y is positioned inside an area P/AP and thus, a 0-write will not be performed.

However, if I_SO is set to the write current I_{w_p} and then V_assist is set to the third potential V₃, the point is close to the first threshold line Th_p or through a point Z beyond Th_p in a process of moving from the point X to the point Y so that a 0-write may erroneously occur (steps ST₂₃→ST₂₄).

Therefore, when a write operation (0-write) is started, it is desirable to set V_assist to the third potential V₃ and then I_SO to the write current I_{w_p} to reliably inhibit the occurrence of a 0-write of the non-selected bit.

The write operation (0-write) is terminated by setting, as shown in FIG. 33, I_SO to 0 and then V_assist to the second potential V₂ (steps ST₂₅→ST₂₆). This is because, as shown in FIG. 33, by taking the route from step ST₂₅ to step ST₂₆, the route does not cross the first threshold line Th_p.

In contrast, if, for example, V_assist is set to the second potential V₂ and then I_SO is set to 0, the point is close to the first threshold line Th_p or through the point Z beyond Th_p in a process of moving from the point Y to the point X so that a 0-write erroneously occurs (steps ST₂₇→ST₂₈).

Therefore, when the write operation (0-write) is terminated, it is desirable to shut off the write current I_{w_p} and then set V_assist to the second potential V₂ to reliably inhibit the occurrence of a 0-write of the non-selected bit.

From the above, the write error rate can be reduced without the occurrence of an erroneous 0-write for a non-selected bit.

Fifth Embodiment

FIG. 34 shows characteristics of a magnetic memory according to a fifth embodiment.

For example, as shown in the upper diagram of FIG. 34, magnetization reversal characteristics of a magnetic memory in consideration of the SOT effect and voltage assist effect exhibit a state in which the first and second threshold lines Th_p, Th_ap are open downward. Also, in general, the first and second threshold lines Th_p, Th_ap are bilaterally symmetric with respect to I_SO=0.

However, if an assist potential V_assist is applied to a second magnetic layer RL of a storage element, a flow of electrons in a vertical direction, that is, in a direction in which the first and second magnetic layers FL, RL are stacked arises, in the first storage layer FL, which causes the STT effect.

For example, if the assist potential V_assist becomes increasingly higher than the potential of a third portion E₃ of a conductive wire 11, that is, with an increasing assist potential V_assist, the STT effect by electrons flowing from the first storage layer FL toward the second storage layer RL becomes conspicuous. In this case, electrons having a spin in a direction opposite to the magnetization direction of the second magnetic layer RL generate spin torque in the first storage layer FL and thus, the magnetization directions of the first and second storage layers FL, RL are more likely to be in an antiparallel state.

Therefore, as shown in the middle diagram of FIG. 34, magnetization reversal characteristics of the magnetic memory in consideration of the STT effect shift the first and second threshold lines Th_p, Th_ap to the left, that is, an antiparallel state is more likely to be entered and a parallel state is less likely to be entered with an increasing potential V_STT applied to the second magnetic layer RL.

Similarly, if the assist potential V_assist becomes increasingly lower than the potential of the third portion E₃ of the conductive wire 11, that is, with a decreasing assist potential V_assist, the STT effect by electrons flowing from the second storage layer RL toward the first storage layer FL becomes conspicuous. In this case, electrons having a spin in the same direction as the magnetization direction of the second magnetic layer RL generate spin torque in the first storage layer FL and thus, the magnetization directions of the first and second storage layers FL, RL are more likely to be in a parallel state.

Therefore, as shown in the middle diagram of FIG. 34, magnetization reversal characteristics of the magnetic memory in consideration of the STT effect shift the first and second threshold lines Th_p, Th_ap to the right, that is, a parallel state is more likely to be entered and an antiparallel state is less likely to be entered with a decreasing potential V_STT applied to the second magnetic layer RL.

From the above, as shown in the lower diagram of FIG. 34, magnetization reversal characteristics of the magnetic memory in consideration of the SOT effect, the voltage assist effect, and the STT effect increase the inclination of the first threshold line Th_p indicating whether to enter a parallel state in a graph of I_SO (x axis)-V_assist (y axis) and decreases the inclination of the second threshold line Th_ap indicating whether to enter an antiparallel state.

This means that with decreasing V_assist, the first and second magnetic layers FL, RL are less likely to enter an antiparallel state. That is, it becomes easier to switch selected bits and non-selected bits in a 1-write (write operation to put a storage element into an antiparallel state) using the inclination of the second threshold line Th_ap. On the other hand, it becomes more difficult to switch selected bits and non-selected bits in a 0-write (write operation to put a storage element into a parallel state).

Therefore, for example, in the write operation described in the third embodiment, it is desirable to write 0 into all of multiple bits (eight bits) by setting the 0-write as the write operation (first). Also, by setting the 1-write as the write operation (second), 1 can be selectively written into multiple bits (eight bits) using the inclination of the second threshold line Th_ap in FIG. 35.

As a result, the write error rate is further reduced in a write operation of the third embodiment. To make the STT effect more conspicuous and improve bit selectivity (make the inclination of the second threshold line Th_ap still smaller), for example, techniques of reducing a resistance-area product (RA) of the storage element MTJ or increasing the spin polarization rate of the first and second magnetic layers FL, RL may be combined.

Regarding the STT effect, in contrast to the above description, a parallel state may also be likely to be entered when electrons flow from the first magnetic layer FL toward the second magnetic layer RL and an antiparallel state may also be likely to be entered when electrons flow from the second magnetic layer RL toward the first magnetic layer FL.

In such a case, the middle diagram of the FIG. 34 changes to characteristics in which a parallel state is more likely to be entered with increasing V_STT and an antiparallel state is more likely to be entered with decreasing V_STT. As a result, the lower diagram of FIG. 34 changes to characteristics in which the inclination of the first threshold line Th_p decreases and the inclination of the second threshold line Th_ap increases.

Therefore, in such a case, the 1-write may be set as the write operation (first) and the 0-write may be set as the write operation (second).

Which trend of the above two cases the STT effect exhibits depends on, for example, band filling of the magnetic material used for the first magnetic layer FL to the 3d orbit.

Next, examples of the write operation will be described.

[First Write Operation (all Bits: 0-Write)]

In the first write operation, a 0-write is performed for multiple bits (all bits).

A write operation (0-write) is started by, as shown in FIG. 35, setting V_assist to the first potential V₁ and I_SO to the write current I_{w_p}. The order thereof may be that, as shown in FIG. 35, I_SO is set to the write current I_{w_p} after V_assist is set to the first potential V₁ (steps ST₀₁→ST₀₂) or V_assist is set to the first potential V₁ after I_SO is set to the write current I_{w_p} (steps ST₀₃→ST₀₄).

The write operation (0-write) is terminated by setting, as shown in FIG. 36, V_assist to the second potential V₂ and then I_SO to 0 (steps ST₀₅→ST₀₆). This is because, as shown in FIG. 6, by taking the route from step ST₀₅ to step ST₀₆, the minimum margin between the route and the second threshold line Th_ap becomes Δ.

When, for example, the route from step ST₀₇ to step ST₀₈ is taken, the minimum margin Δ becomes larger than a minimum margin Δ′ between the route and the second threshold line Th_ap. Therefore, when the write operation (0-write) is terminated, a 1-write is not erroneously generated due to thermal disturbance or the like so that a write error rate can be reduced.

[Second Write Operation: Selected Bit]

In the second write operation, a 1-write is performed for a selected bit for which the 1-write should be performed in the order below:

A write operation (1-write) is started by, as shown in FIG. 37, setting V_assist to the first potential V₁ and I_SO to the write current I_{w_ap}. The order thereof may be that, as shown in FIG. 37, I_SO is set to the write current I_{w_ap} after V_assist is set to the first potential V₁ (steps ST₁₁→ST₁₂) or V_assist is set to the first potential V₁ after I_SO is set to the write current I_{w_ap} (steps ST₁₃→ST₁₄).

The write operation (1-write) is terminated by, as shown in FIG. 38, setting V_assist to the second potential V₂ and then I_SO to 0 (steps ST₁₅→ST₁₆). This is because, as shown in FIG. 38, by taking the route from step ST₁₅ to step ST₁₆, the minimum margin between the route and the first threshold line Th_p becomes Δ.

When, for example, the route from step ST₁₇ to step ST₁₈ is taken, the minimum margin Δ becomes larger than a minimum margin Δ′ between the route and the first threshold line Th_p. Therefore, when the write operation (1-write) is terminated, a 0-write is not erroneously generated due to thermal disturbance or the like so that a write error rate can be reduced.

[Second Write Operation: Non-Selected Bit]

In the second write operation, a 1-write is performed for a non-selected bit for which the 1-write should not be performed in the order below:

A write operation (1-write) is started by, as shown in FIG. 39, setting V_assist to the third potential V₃ and I_SO to the write current I_{w_ap}. The order thereof is that, as shown in FIG. 39, V_assist is set to the third potential V₃ and then I_SO is set to the write current I_{w_ap} (steps ST₃₁→ST₃₂).

The point Y is positioned inside the area P/AP and thus, a 1-write will not be performed.

However, if I_SO is set to the write current I_{w_ap} and then V_assist is set to the third potential V₃, the point is close to the second threshold line Th_ap or through the point Z beyond Th_ap in a process of moving from the point X to the point Y so that a 1-write may erroneously occur (steps ST₃₃→ST₃₄).

Therefore, when a write operation (1-write) is started, it is desirable to set V_assist to the third potential V₃ and then I_SO to the write current I_{w_ap} to reliably inhibit the occurrence of a 1-write of the non-selected bit.

The write operation (1-write) is terminated by, as shown in FIG. 40, setting I_SO to 0 and then V_assist to the second potential V₂ (steps ST₃₅→ST₃₆). This is because, as shown in FIG. 40, by taking the route from step ST₃₅ to step ST₃₆, the route does not cross the second threshold line Th_ap.

In contrast, if, for example, V_assist is set to the second potential V₂ and then I_SO is set to 0, the point is close to the second threshold line Th_ap or through the point Z beyond the second threshold line Th_ap in a process of moving from the point Y to the point X so that a 1-write erroneously occurs (steps ST₃₇→ST₃₈).

Therefore, when the write operation (1-write) is terminated, it is desirable to shut off the write current I_{w_ap} and then set V_assist to the second potential V₂ to reliably inhibit the occurrence of a 1-write of the non-selected bit.

From the above, the write error rate can be reduced without the occurrence of an erroneous 1-write for a non-selected bit.

SUMMARY

According to an embodiment, as described above, the write error rate of a magnetic memory can be reduced.

While certain embodiments have been described, these embodiments have been presented by way of example only, and are not intended to limit the scope of the inventions. Indeed, the novel embodiments described herein may be embodied in a variety of other forms; furthermore, various omissions, substitutions and changes in the form of the embodiments described herein may be made without departing from the spirit of the inventions. The accompanying claims and their equivalents are intended to cover such forms or modifications as would fall within the scope and spirit of the inventions.

Claims

1-20. (canceled)

21: A nonvolatile memory comprising: a conductive line including a first portion, a second portion, and a third portion between the first portion and the second portion;a storage element including a first magnetic layer, a second magnetic layer, and a nonmagnetic layer between the first magnetic layer and the second magnetic layer, the first magnetic layer being connected to the third portion;a current source connected to the first portion and the second portion, and configured to output a first write current and a second write current, a value of the first write current being different from a value of the second write current, anda potential source connected to the second magnetic layer, and configured to output a first potential and a second potential,wherein the first write current is outputted, subsequently the first potential is outputted, subsequently the second potential is outputted, and subsequently the second write current is outputted.

22: The memory of claim 21, wherein the first potential is higher than the second potential.

23: The memory of claim 21, wherein the first potential is lower than the second potential.

24: The memory of claim 21, wherein: the first potential is higher than the second potential, andthe second potential is outputted, subsequently a potential with a positive slope is outputted, subsequently the first potential is outputted, subsequently a potential with a negative slope is outputted, and subsequently the second potential is outputted.

25: The memory of claim 21, wherein: the value of the first write current is higher than the value of the second write current, andthe second write current is outputted, subsequently a write current with a positive slope is outputted, subsequently the first write current is outputted, subsequently a write current with a negative slope is outputted, and subsequently the second write current is outputted.

26: A nonvolatile memory comprising: a conductive line including a first portion, a second portion, and a third portion between the first portion and the second portion;a storage element including a first magnetic layer, a second magnetic layer, and a nonmagnetic layer between the first magnetic layer and the second magnetic layer, and the first magnetic layer being connected to the third portion;a current source connected to the first portion and the second portion, and configured to output a first write current and a second write current, a value of the first write current being different from a value of the second write current; anda potential source connected to the second magnetic layer, and configured to output a first potential and a second potential,wherein the first potential is outputted, subsequently the first write current is outputted, subsequently the second potential is outputted, and subsequently the second write current is outputted.

27: The memory of claim 26, wherein the first potential is higher than the second potential.

28: The memory of claim 26, wherein the first potential is lower than the second potential.

29: The memory of claim 26, wherein: the first potential is higher than the second potential, andthe second potential is outputted, subsequently a potential with a positive slope is outputted, subsequently the first potential is outputted, subsequently a potential with a negative slope is outputted, and subsequently the second potential is outputted.

30: The memory of claim 26, wherein: the value of the first write current is higher than the value of the second write current, andthe second write current is outputted, subsequently a write current with a positive slope is outputted, subsequently the first write current is outputted, subsequently a write current with a negative slope is outputted, and subsequently the second write current is outputted.

31: A nonvolatile memory comprising: a conductive line including a first portion, a second portion, a third portion, and a fourth portion, the third portion being provided between the first portion and the second portion, and the fourth portion being provided between the second portion and the third portion;a first storage element including a first terminal connected to the third portion, and a second terminal;a current source connected to the first portion and the second portion, and configured to output a first write current and a second write current, a value of the first write current being different from a value of the second write current; anda potential source connected to the second terminal, and configured to output a first potential and a second potential,wherein the first write current is outputted, subsequently the first potential is outputted, subsequently the second potential is outputted, and subsequently the second write current is outputted.

32: The memory of claim 31, wherein a third potential is outputted.

33: The memory of claim 31, wherein a third potential or a fourth potential is outputted.

34: The memory of claim 31, wherein the first potential is different from a potential of the third portion.

35: The memory of claim 31, wherein a polarity of the second potential is different from a polarity of the first potential.