JP2008187048A - Magnetoresistive effect element - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To achieve both reduction in writing currents and improvement in MR ratio at reading. <P>SOLUTION: The magnetoresistive effect element comprises first and second ferromagnetic layers 11A and 11B of which magnetizing directions are fixed mutually opposite; a third ferromagnetic layer 12, which is arranged between the first and second ferromagnetic layers 11A and 11B with magnetizing direction changing; a first non magnetic layer 13A arranged between the first ferromagnetic layer 11A and the third ferromagnetic layer 12; and a second non magnetic layer 13B arranged between the second ferromagnetic layer 11B and the third ferromagnetic layer 12. When a reading voltage Ir is biased, MR ratio of a first unit U1 is in inverse state, while the MR ratio of a second unit U2 is in normal state. When writing voltages Iw"0" and Iw"1" are biased, the MR ratio of the first and second units U1 and U2 are both in the normal state. <P>COPYRIGHT: (C)2008,JPO&INPIT

Description

  The present invention relates to a magnetoresistive effect element having a dual pin structure.

  In recent years, a large number of memory elements that store data based on a new principle have been proposed. Among them, magneto-resistive elements (TMR: Tunneling Magneto Resistive) use magneto-resistive elements (MRAM: Magnetic Random Access Memory) and reconfigurable memory cells. Application to FETs (Field Effect Transistors) constituting a (reconfigurable) logic circuit is expected (see, for example, Patent Document 1 and Non-Patent Document 1).

  The basic structure of the magnetoresistive element is composed of a pinned layer in which the magnetization direction is fixed, a free layer in which the magnetization direction changes, and a tunneling barrier layer between them. Composed. Then, data is stored depending on the relative relationship between the magnetization directions of the pinned layer and the free layer, that is, whether the magnetization directions of the both are the same (parallel) or opposite (antiparallel).

  Here, for writing data to magnetoresistive effect elements, spin angular momentum transfer (SMT) writing method (hereinafter referred to as “Spin Momentum Transfer”) (hereinafter referred to as “Spin Momentum Transfer”) is possible. , Spin injection writing method) is effective.

  In the spin injection writing method, the write current Ic necessary for reversing the magnetization direction of the free layer is defined by the current density. The current density necessary for reversing the magnetization direction of the free layer is called a critical current density Jc, and a writing current Ic exceeding the density is required at the time of writing.

  The advantage of the spin injection writing method is that when the critical current density Jc is constant, the value of the write current Ic decreases as the size of the magnetoresistive element decreases. For this reason, in principle, it is excellent in scalability, and for example, it is not impossible to realize a large-capacity magnetic random access memory exceeding 256 megabits.

  However, there are many problems that must be solved for this purpose.

  One of them is reduction of critical current density Jc. This is because unless this is reduced, the write current Ic cannot be reduced when the size of the magnetoresistive element is constant.

  The technology devised there is a dual pin structure.

  The dual pin structure is a structure in which a pinned layer is provided on both sides of a free layer via a nonmagnetic layer (for example, a tunnel barrier layer). According to this structure, at the time of writing, spin-polarized electrons are injected from the two pinned layers into the free layer, so that the critical current density Jc is 1/2 or more compared to the basic structure in which the pinned layer is one. It can be:

However, the dual pin structure causes a new problem that the MR (magneto-resistive) ratio of the magnetoresistive element is lowered. That is, at the time of reading, a sufficient margin cannot be secured between the resistance value of “0” -data and the resistance value of “1” -data.
U.S. Patent 6,256,223 C. Slonczewski, “Current-driven ecitation of magnetic multilayers”, JORNAL OF MAGNETISM AND MAGNETIC MATERIALS, VOLUME 159, 1996, p.L1-L7

  In the example of the present invention, a technique is proposed that enables a reduction in write current and an improvement in MR ratio at the time of reading in a magnetoresistive element having a dual pin structure.

  The magnetoresistive effect element according to the example of the present invention is disposed between the first and second ferromagnetic layers whose magnetization directions are fixed opposite to each other, and the first and second ferromagnetic layers, and the magnetization direction changes. A third ferromagnetic layer, a first nonmagnetic layer disposed between the first and third ferromagnetic layers, and a second nonmagnetic layer disposed between the second and third ferromagnetic layers. When the read voltage is biased between the first and second ferromagnetic layers, the MR ratio of the first unit composed of the first ferromagnetic layer, the first nonmagnetic layer, and the third ferromagnetic layer is The MR ratio of the second unit that is in the inverse state and is composed of the second ferromagnetic layer, the second nonmagnetic layer, and the third ferromagnetic layer is in the normal state, and is between the first and second ferromagnetic layers. When the write voltage is biased, the MR ratios of the first and second units are both normal or inverse. A.

  The magnetoresistive effect element according to the example of the present invention is disposed between the first and second ferromagnetic layers whose magnetization directions are fixed in the same direction, and the first and second ferromagnetic layers, and the magnetization direction changes. A third ferromagnetic layer, a first nonmagnetic layer disposed between the first and third ferromagnetic layers, and a second nonmagnetic layer disposed between the second and third ferromagnetic layers. An MR ratio of a first unit comprising a first ferromagnetic layer, a first nonmagnetic layer, and a third ferromagnetic layer when a read voltage is biased between the first and second ferromagnetic layers, and The MR ratio of the second unit composed of the second ferromagnetic layer, the second nonmagnetic layer, and the third ferromagnetic layer is in a normal state or an inverse state, and is between the first and second ferromagnetic layers. When the write voltage is biased, the MR ratio of the first unit is in the inverse state, and the second unit R ratio is in a normal state.

  According to the example of the present invention, in a magnetoresistive effect element having a dual pin structure, it is possible to achieve both reduction in write current and improvement in MR ratio at the time of reading.

  The best mode for carrying out an example of the present invention will be described below in detail with reference to the drawings.

1. Overview
In the example of the present invention, it is composed of a first ferromagnetic layer (pinned layer) / first nonmagnetic layer / third ferromagnetic layer (free layer) / second nonmagnetic layer / second ferromagnetic layer (pinned layer). The dual pin structure magnetoresistive effect element.

  Here, / means that the layers existing on both sides thereof are stacked in contact with each other. same as below.

  Further, the first ferromagnetic layer / first nonmagnetic layer / third ferromagnetic layer is defined as a first unit, and the third ferromagnetic layer / second nonmagnetic layer / second ferromagnetic layer is defined as a second unit.

  When the magnetization directions of the first and second ferromagnetic layers are antiparallel, one of the MR ratios of the first and second units when the read voltage is biased to the magnetoresistive effect element is set to the inverse state. The other is set to the normal state, and both the MR ratios of the first and second units when the write voltage is biased to the magnetoresistive effect element are set to the normal state or the inverse state.

  Further, when the magnetization directions of the first and second ferromagnetic layers are in a parallel relationship, both the MR ratios of the first and second units when the read voltage is biased to the magnetoresistive effect element are in a normal state or an inverse. One of the MR ratios of the first and second units when the write voltage is biased to the magnetoresistive effect element is set to the inverse state, and the other is set to the normal state.

  Note that the magnetization direction is antiparallel means that the magnetization direction of the first and second ferromagnetic layers can be regarded as opposite or substantially opposite, and the angle θ between the magnetization directions is 90 ° < Including the case of θ ≦ 180 °.

  Further, the parallel magnetization directions mean that the magnetization directions of the first and second ferromagnetic layers can be regarded as the same direction or substantially the same direction, and the angle θ between the magnetization directions is 0 ° ≦ θ Also includes the case of <180 °.

  Thereby, in the dual-pin structure magnetoresistive effect element, it is possible to achieve both reduction of the write current and improvement of the MR ratio at the time of reading.

2. Dual pin structure
First, a dual pin structure magnetoresistive effect element which is an object of an example of the present invention will be described.

  1 and 2 show a magnetoresistive effect element having a dual pin structure.

  The basic structure is pinned layer 11A / spacer layer 13A / free layer 12 / spacer layer 13B / pinned layer 11B. Arrows indicate the direction of magnetization.

  The pinned layers 11A and 11B are made of a ferromagnetic material, and the magnetization direction is fixed by, for example, an antiferromagnetic material as a pin layer. The free layer 12 is made of a ferromagnetic material, and the magnetization direction is changed by a write current (spin injection current).

  The spacer layers 13A and 13B are made of a nonmagnetic material. If the nonmagnetic material is an insulator or a semiconductor, the spacer layers 13A and 13B become tunnel barrier layers, so that a tunnel magnetoresistance effect can be generated.

  At least one of the pinned layers 11A and 11B may have a magnetic layer / nonmagnetic layer / magnetic layer structure such as a SAF (Synthetic anti-ferromagnetic) structure as shown in FIG. 3, or as shown in FIG. The antiferromagnetic layer 14 may be added.

  The free layer 12 may have a laminated structure of a plurality of magnetic layers as shown in FIG. 5, or a magnetic layer / nonmagnetic layer / magnetic layer structure such as a SAF structure as shown in FIG.

  As for the spacer layers 13A and 13B, as long as at least one of them functions as a tunnel barrier layer, the rest may be made of a nonmagnetic material as a conductor.

  The difference between the structure of FIG. 1 and the structure of FIG. 2 is the magnetization direction of the pinned layers 11A and 11B. In the structure of FIG. 1, the magnetization directions of the pinned layers 11A and 11B are fixed in opposite directions, whereas in the structure of FIG. 2, the magnetization directions of the pinned layers 11A and 11B are fixed in the same direction.

  The magnetization direction of the residual magnetization (the easy axis direction) of the pinned layers 11A and 11B and the free layer 12 is the direction in which they are stacked (perpendicular magnetization).

  However, the present invention is not limited to this, and the magnetization direction (magnetization easy axis direction) of the residual magnetizations of the pinned layers 11A and 11B and the free layer 12 is perpendicular to the direction in which they are stacked (in-plane magnetization). It is good.

  The easy magnetization axis direction is a direction of spontaneous magnetization in which the internal energy of the macro ferromagnet is lowest in the absence of an external magnetic field. Further, the hard axis direction is the direction of spontaneous magnetization in which the internal energy of the macro ferromagnet becomes the highest in the absence of an external magnetic field.

  According to the dual pin structure, it is possible to reduce the write current (spin injection current) when the size of the magnetoresistive effect element is constant by reducing the critical current density Jc.

  Here, pinned layer 11A / spacer layer 13A / free layer 12 is defined as first unit U1, and free layer 12 / spacer layer 13B / pinned layer 11B is defined as second unit U2.

  In the structure of FIG. 1, the magnetization of the pinned layer 11A of the first unit U1 is upward on the paper surface, and the magnetization of the pinned layer 11B of the second unit U2 is downward on the paper surface. In this case, the effect of the dual pin structure, that is, reduction of the write current can be realized.

  However, when the magnetization of the free layer 12 is downward (solid line), the magnetization state of the first unit U1 is anti-parallel, whereas the magnetization state of the second unit U2 is parallel (parallel). ).

  Similarly, when the magnetization of the free layer 12 is upward (broken line), the magnetization state of the first unit U1 is parallel, while the magnetization state of the second unit U2 is antiparallel.

  For this reason, at the time of reading, the MR ratio by the first unit U1 and the MR ratio by the second unit U2 cancel each other.

  Therefore, in order to ensure the MR ratio as the magnetoresistive effect element, for example, the material and thickness of the spacer layers 13A and 13B must be varied to make a difference in the MR ratio of the first and second units U1 and U2. I must.

  However, since the MR ratio by the first unit U1 and the MR ratio by the second unit U2 still cancel each other, it is difficult to secure a sufficiently large MR ratio as a result.

  In the structure of FIG. 2, the magnetizations of the pinned layers 11A of the first and second units U1 and U2 are the same direction, and in this example, both are upward.

  In this case, when the magnetization of the free layer 12 faces downward (solid line), the magnetization states of the first and second units U1 and U2 are both antiparallel. When the magnetization of the free layer 12 is upward (dashed line), the magnetization states of the first and second units U1 and U2 are both parallel.

  Therefore, at the time of reading, the MR ratio by the first unit U1 and the MR ratio by the second unit U2 are not canceled out, and a sufficiently large MR ratio can be secured.

  However, since the magnetizations of the pinned layers 11A of the first and second units U1 and U2 are in the same direction, the reflection of spin-polarized electrons is caused when the magnetization states of the first and second units U1 and U2 are made parallel. Thus, the effect of improving the magnetization reversal efficiency cannot be obtained. In addition, it is difficult to make the magnetization states of the first and second units U1, U2 antiparallel.

  In the example of the present invention, all such problems are solved at once by utilizing the bias dependence of the TMR (tunnel magnetoresistance effect) ratio.

3. Normal state / inverse state
There are positive and negative MR ratios of magnetoresistive elements.

  A state in which the MR ratio is positive is defined as a normal state, and a tunnel magnetoresistance effect that causes such an MR ratio is defined as a normal TMR effect. Further, a state in which the MR ratio is negative is defined as an inverse state, and a tunnel magnetoresistance effect that causes such an MR ratio is defined as an inverse TMR effect.

  Specifically, it is as follows.

When the MR ratio is in the normal state, the resistance value when the magnetization state of the pinned layer and the free layer is parallel is Rp, and the resistance value when it is antiparallel is Rap.
Rap> Rp
It becomes. The MR ratio at this time is
MR ratio = [(Rap−Rp) / Rp] × 100> 0
It is represented by

When the MR ratio is an inverse state, the resistance value when the magnetization state of the pinned layer and the free layer is parallel is Rp, and the resistance value when the pinned layer is antiparallel is Rap.
Rap <Rp
It becomes. The MR ratio at this time is
MR ratio = [(Rap−Rp) / Rp] × 100 <0
It is represented by

  By the way, the MR ratio of the tunnel magnetoresistive element, that is, the TMR ratio has a bias dependence.

  For example, when the bias voltage is zero and the TMR ratio is in the normal state, the TMR ratio decreases as the bias voltage increases. When the bias voltage reaches a predetermined value, the TMR ratio becomes zero. As the bias voltage is further increased, the TMR ratio (absolute value) becomes an inverse state and increases again.

  Further, when the bias voltage is zero and the TMR ratio is inverse, the TMR ratio (absolute value) decreases as the bias voltage increases. When the bias voltage reaches a predetermined value, the TMR ratio becomes zero. As the bias voltage is further increased, the TMR ratio becomes normal and increases again.

  The normal state / inverse state of the TMR ratio also affects the writing, that is, the effect of the spin injection torque on the free layer.

  For example, when the TMR ratio is in the normal state, when a write current flows from the free layer to the pinned layer, electrons spin-polarized in the same direction as the magnetization direction of the pinned layer pass through the pinned layer and enter the free layer. In order to give a spin injection torque, the magnetization direction of the free layer is the same (parallel) as the magnetization direction of the pinned layer.

  In addition, when a write current flows from the pinned layer to the free layer, electrons that are spin-polarized in the direction opposite to the magnetization direction of the pinned layer are reflected by the pinned layer and give a spin injection torque to the free layer. The magnetization direction of the layer is opposite (antiparallel) to the magnetization direction of the pinned layer.

  On the other hand, when the TMR ratio is in the inverse state, when a write current flows from the free layer to the pinned layer, electrons that are spin-polarized in the direction opposite to the magnetization direction of the pinned layer pass through the pinned layer and are free. Since the spin injection torque is applied to the layer, the magnetization direction of the free layer is opposite (antiparallel) to the magnetization direction of the pinned layer.

  Also, when a write current flows from the pinned layer to the free layer, among the electrons flowing from the free layer to the pinned layer, electrons that are spin-polarized in the same direction as the magnetization direction of the pinned layer are reflected by the pinned layer. In order to give spin injection torque to the free layer, the magnetization direction of the free layer is the same (parallel) as the magnetization direction of the pinned layer.

  As described above, the TMR ratio becomes a normal state or an inverse state according to the bias voltage, and whether the TMR ratio is the normal state or the inverse state has a significant influence on the read / write operation. .

  Accordingly, by controlling the bias dependence of the TMR ratio and setting the write voltage and the read voltage to appropriate values, it is possible to simultaneously reduce the write current and improve the MR ratio at the time of reading.

  In the example of the present invention, it is sufficient that at least one MR ratio of the first and second units U1 and U2 in FIGS. 1 and 2 has a bias dependency, so that at least one of the spacer layers 13A and 13B. One is a tunnel barrier layer.

4). Embodiment
(1) When the magnetization directions of the two pinned layers are opposite
7 to 10 show states at the time of writing / reading of the magnetoresistive effect element in which the magnetization directions of the two pinned layers are fixed in opposite directions.

First, the direction of electron spin is defined as follows.
A state in which the electron spin is directed upward in the drawing is referred to as upspin, and a state in which the electron spin is directed downward in the drawing is referred to as downspin.

  Arrows in the pinned layer and the free layer indicate the direction of magnetization.

  Since the structure of the magnetoresistive effect element has already been described with reference to FIG. 1, the description thereof is omitted here.

A. First example
In the example of FIGS. 7 and 8, when the write voltage is biased to the magnetoresistive effect element, the MR ratios of the first and second units U1 and U2 are both in the normal state, and when the read voltage is biased, The MR ratio of one unit U1 is in an inverse state, and the MR ratio of the second unit is in a normal state.

  First, as shown in FIG. 7, when the magnetization of the free layer 12 is downward on the paper surface, in order to reverse the magnetization upward on the paper surface, a write voltage (bias voltage) Vw “1” is applied to the magnetoresistive effect element. To do. The write voltage Vw “1” brings the pinned layer 11A to a low potential and the pinned layer 11B to a high potential.

  A bias for setting the pinned layer 11A to a low potential and setting the pinned layer 11B to a high potential is defined as a positive bias voltage.

  At this time, the write current (spin injection current) Iw “1” flows from the pinned layer 11B toward the pinned layer 11A.

  For this reason, the electrons spin-up polarized upward (upspin) pass through the pinned layer 11A and give a spin injection torque to the free layer 12. Further, the upspin is reflected by the pinned layer 11B and gives a spin injection torque to the free layer 12.

  Therefore, the magnetization of the free layer 12 changes from a downward direction on the paper surface to an upward surface on the paper surface.

  Thereafter, a read voltage (bias voltage) Vr is applied to the magnetoresistive effect element. The read voltage Vr is sufficiently smaller than the write voltage Vw “1”. The read voltage Vr brings the pinned layer 11A to a low potential and the pinned layer 11B to a high potential.

  At this time, the read current Ir flows from the pinned layer 11B toward the pinned layer 11A. In the state where the read voltage Vr is biased, the first unit U1 is in the inverse state and the second unit U2 is in the normal state.

  Therefore, in the first unit U1, since the magnetization state of the pinned layer 11A and the free layer 12 is parallel, the first unit U1 is in a high resistance state, and in the second unit U2, the magnetization state of the pinned layer 11B and the free layer 12 is opposite. Since it is parallel, it will be in a high resistance state.

  That is, since both are in a high resistance state, the MR ratio is not canceled between the first and second units U1 and U2, and both reduction in write current and improvement in MR ratio at the time of reading are achieved. Can do.

  Next, as shown in FIG. 8, when the magnetization of the free layer 12 is upward on the paper surface, in order to reverse the magnetization downward on the paper surface, a write voltage (bias voltage) Vw “0” is applied to the magnetoresistive effect element. Apply. The write voltage Vw “0” brings the pinned layer 11A to a high potential and the pinned layer 11B to a low potential.

  A bias for setting the pinned layer 11A to a high potential and setting the pinned layer 11B to a low potential is defined as a negative bias voltage.

  At this time, the write current (spin injection current) Iw “0” flows from the pinned layer 11A toward the pinned layer 11B.

  For this reason, the electrons (down spin) spin-polarized downward in the drawing pass through the pinned layer 11B and give a spin injection torque to the free layer 12. Further, the down spin is reflected by the pinned layer 11A and gives a spin injection torque to the free layer 12.

  Therefore, the magnetization of the free layer 12 changes from upward on the paper to downward on the paper.

  Thereafter, a read voltage (bias voltage) Vr is applied to the magnetoresistive effect element. The read voltage Vr is sufficiently smaller than the write voltage Vw “0”. The read voltage Vr brings the pinned layer 11A to a low potential and the pinned layer 11B to a high potential.

  At this time, the read current Ir flows from the pinned layer 11B toward the pinned layer 11A. In the state where the read voltage Vr is biased, the first unit U1 is in the inverse state and the second unit U2 is in the normal state.

  Accordingly, in the first unit U1, since the magnetization state of the pinned layer 11A and the free layer 12 is antiparallel, the resistance state is low, and in the second unit U2, the magnetization state of the pinned layer 11B and the free layer 12 is Since it is parallel, it will be in a low resistance state.

  That is, since both are in a low resistance state, the MR ratio is not canceled by the first and second units U1 and U2, and both reduction in write current and improvement in MR ratio at the time of reading are achieved. Can do.

  7 and 8, the first unit U1 may be in the normal state and the second unit U2 may be in the inverse state.

  In this case, in the state shown in FIG. 7, that is, when the magnetization of the free layer 12 is upward on the paper, both the first and second units U1 and U2 are in the low resistance state, and the state shown in FIG. When the magnetization of 12 is directed downward in the drawing, both the first and second units U1, U2 are in a high resistance state.

  As for the read voltage Vr, the pinned layer 11A is set to a low potential and the pinned layer 11B is set to a high potential, but instead, the pinned layer 11A may be set to a high potential and the pinned layer 11B may be set to a low potential.

B. Second example
In the example of FIGS. 9 and 10, when the write voltage is biased to the magnetoresistive effect element, the MR ratios of the first and second units U1 and U2 are both in an inverse state, and when the read voltage is biased, The MR ratio of one unit U1 is in an inverse state, and the MR ratio of the second unit is in a normal state.

  First, as shown in FIG. 9, when the magnetization of the free layer 12 is downward on the paper, in order to reverse the magnetization upward on the paper, the write voltage (negative bias voltage) Vw “1” is applied to the magnetoresistive effect element. Apply. The write voltage Vw “1” brings the pinned layer 11A to a high potential and the pinned layer 11B to a low potential.

  At this time, the write current (spin injection current) Iw “1” flows from the pinned layer 11A toward the pinned layer 11B.

  For this reason, the spin-polarized electrons (up spin) on the paper surface pass through the pinned layer 11B and give a spin injection torque to the free layer 12. The upspin is reflected by the pinned layer 11A and gives a spin injection torque to the free layer 12.

  Therefore, the magnetization of the free layer 12 changes from a downward direction on the paper surface to an upward surface on the paper surface.

  Thereafter, a read voltage (bias voltage) Vr is applied to the magnetoresistive effect element. The read voltage Vr is sufficiently smaller than the write voltage Vw “1”. The read voltage Vr brings the pinned layer 11A to a low potential and the pinned layer 11B to a high potential.

  At this time, the read current Ir flows from the pinned layer 11B toward the pinned layer 11A. In the state where the read voltage Vr is biased, the first unit U1 is in the inverse state and the second unit U2 is in the normal state.

  Therefore, in the first unit U1, since the magnetization state of the pinned layer 11A and the free layer 12 is parallel, the first unit U1 is in a high resistance state, and in the second unit U2, the magnetization state of the pinned layer 11B and the free layer 12 is opposite. Since it is parallel, it will be in a high resistance state.

  That is, since both are in a high resistance state, the MR ratio is not canceled between the first and second units U1 and U2, and both reduction in write current and improvement in MR ratio at the time of reading are achieved. Can do.

  Next, as shown in FIG. 10, when the magnetization of the free layer 12 is upward on the paper surface, in order to reverse the magnetization downward on the paper surface, the write voltage (positive bias voltage) Vw “0” is applied to the magnetoresistive effect element. "Is applied. The write voltage Vw “0” makes the pinned layer 11A have a low potential and the pinned layer 11B has a high potential.

  At this time, the write current (spin injection current) Iw “0” flows from the pinned layer 11B toward the pinned layer 11A.

  For this reason, the electrons (down spin) spin-polarized downward in the paper surface pass through the pinned layer 11A and give a spin injection torque to the free layer 12. Further, the down spin is reflected by the pinned layer 11B and gives a spin injection torque to the free layer 12.

  Therefore, the magnetization of the free layer 12 changes from upward on the paper to downward on the paper.

  Thereafter, a read voltage (bias voltage) Vr is applied to the magnetoresistive effect element. The read voltage Vr is sufficiently smaller than the write voltage Vw “0”. The read voltage Vr brings the pinned layer 11A to a low potential and the pinned layer 11B to a high potential.

  At this time, the read current Ir flows from the pinned layer 11B toward the pinned layer 11A. In the state where the read voltage Vr is biased, the first unit U1 is in the inverse state and the second unit U2 is in the normal state.

  Accordingly, in the first unit U1, since the magnetization state of the pinned layer 11A and the free layer 12 is antiparallel, the resistance state is low, and in the second unit U2, the magnetization state of the pinned layer 11B and the free layer 12 is Since it is parallel, it will be in a low resistance state.

  That is, since both are in a low resistance state, the MR ratio is not canceled by the first and second units U1 and U2, and both reduction in write current and improvement in MR ratio at the time of reading are achieved. Can do.

  As a state at the time of reading in FIGS. 9 and 10, the first unit U1 may be in a normal state and the second unit U2 may be in an inverse state.

  In this case, in the state shown in FIG. 9, that is, when the magnetization of the free layer 12 is upward on the paper surface, the first and second units U1 and U2 are both in the low resistance state, and the state shown in FIG. When the magnetization of 12 is directed downward in the drawing, both the first and second units U1, U2 are in a high resistance state.

  As for the read voltage Vr, the pinned layer 11A is set to a low potential and the pinned layer 11B is set to a high potential, but instead, the pinned layer 11A may be set to a high potential and the pinned layer 11B may be set to a low potential.

(2) When the magnetization directions of the two pinned layers are the same
11 to 14 show states at the time of writing / reading of the magnetoresistive effect element in which the magnetization directions of the two pinned layers are fixed in the same direction.

  First, regarding the direction of electron spin, as in the case where the magnetization directions of the two pinned layers are opposite to each other, a state in which the direction is upward is referred to as upspin, and a state in which the direction is downward is referred to as downspin. Called.

A. First example
In the example of FIGS. 11 and 12, when the write voltage is biased to the magnetoresistive effect element, the MR ratio of the first unit U1 is in the normal state, the MR ratio of the second unit is in the inverse state, and the read voltage is biased. Sometimes, the MR ratios of the first and second units U1, U2 are both in a normal state.

  First, as shown in FIG. 11, when the magnetization of the free layer 12 is downward on the paper surface, in order to reverse the magnetization upward on the paper surface, the write voltage (positive bias voltage) Vw “0” is applied to the magnetoresistive effect element. Is applied. The write voltage Vw “0” makes the pinned layer 11A have a low potential and the pinned layer 11B has a high potential.

  At this time, the write current (spin injection current) Iw “0” flows from the pinned layer 11B toward the pinned layer 11A.

  For this reason, the electrons spin-up polarized upward (upspin) pass through the pinned layer 11A and give a spin injection torque to the free layer 12. Further, the upspin is reflected by the pinned layer 11B and gives a spin injection torque to the free layer 12.

  Therefore, the magnetization of the free layer 12 changes from a downward direction on the paper surface to an upward surface on the paper surface.

  Thereafter, a read voltage (bias voltage) Vr is applied to the magnetoresistive effect element. The read voltage Vr is sufficiently smaller than the write voltage Vw “0”. The read voltage Vr brings the pinned layer 11A to a low potential and the pinned layer 11B to a high potential.

  At this time, the read current Ir flows from the pinned layer 11B toward the pinned layer 11A. In addition, when the read voltage Vr is biased, both the first and second units U1 and U2 are in the normal state.

  Accordingly, in the first unit U1, since the magnetization state of the pinned layer 11A and the free layer 12 is parallel, the resistance state is brought about. In the second unit U2, the magnetization state of the pinned layer 11B and the free layer 12 is parallel. Therefore, a low resistance state is obtained.

  That is, since both are in a low resistance state, the MR ratio is not canceled by the first and second units U1 and U2, and both reduction in write current and improvement in MR ratio at the time of reading are achieved. Can do.

  Next, as shown in FIG. 12, when the magnetization of the free layer 12 is upward on the paper, in order to reverse the magnetization downward on the paper, the write voltage (negative bias voltage) Vw “1” is applied to the magnetoresistive effect element. "Is applied. The write voltage Vw “1” brings the pinned layer 11A to a high potential and the pinned layer 11B to a low potential.

  At this time, the write current (spin injection current) Iw “1” flows from the pinned layer 11A toward the pinned layer 11B.

  For this reason, the electrons (down spin) spin-polarized downward in the drawing pass through the pinned layer 11B and give a spin injection torque to the free layer 12. Further, the down spin is reflected by the pinned layer 11A and gives a spin injection torque to the free layer 12.

  Therefore, the magnetization of the free layer 12 changes from upward on the paper to downward on the paper.

  Thereafter, a read voltage (bias voltage) Vr is applied to the magnetoresistive effect element. The read voltage Vr is sufficiently smaller than the write voltage Vw “1”. The read voltage Vr brings the pinned layer 11A to a low potential and the pinned layer 11B to a high potential.

  At this time, the read current Ir flows from the pinned layer 11B toward the pinned layer 11A. In addition, when the read voltage Vr is biased, both the first and second units U1 and U2 are in the normal state.

  Accordingly, in the first unit U1, since the magnetization state of the pinned layer 11A and the free layer 12 is antiparallel, the high resistance state is obtained. In the second unit U2, the magnetization state of the pinned layer 11B and the free layer 12 is Since it is antiparallel, it will be in a high resistance state.

  That is, since both are in a high resistance state, the MR ratio is not canceled between the first and second units U1 and U2, and both reduction in write current and improvement in MR ratio at the time of reading are achieved. Can do.

  11 and 12, the first unit U1 may be in the inverse state and the second unit U2 may be in the normal state.

  In this case, when the bias condition as shown in FIG. 11 is given at the time of writing, the magnetization direction of the free layer 12 changes not upward on the paper but downward on the paper. Further, when a bias condition as shown in FIG. 12 is given at the time of writing, the magnetization direction of the free layer 12 changes not upward on the paper but upward on the paper.

  As for the read voltage Vr, the pinned layer 11A is set to a low potential and the pinned layer 11B is set to a high potential, but instead, the pinned layer 11A may be set to a high potential and the pinned layer 11B may be set to a low potential.

B. Second example
In the example of FIGS. 13 and 14, when the write voltage is biased to the magnetoresistive effect element, the MR ratio of the first unit U1 is in the normal state, the MR ratio of the second unit is in the inverse state, and the read voltage is biased. Sometimes, the MR ratio of the first and second units U1, U2 is both in an inverse state.

  First, as shown in FIG. 13, when the magnetization of the free layer 12 is downward on the paper surface, in order to reverse the magnetization upward on the paper surface, a write voltage (positive bias voltage) Vw “1” is applied to the magnetoresistive effect element. Apply. The write voltage Vw “1” brings the pinned layer 11A to a low potential and the pinned layer 11B to a high potential.

  At this time, the write current (spin injection current) Iw “1” flows from the pinned layer 11B toward the pinned layer 11A.

  For this reason, the electrons spin-up polarized upward (upspin) pass through the pinned layer 11A and give a spin injection torque to the free layer 12. Further, the upspin is reflected by the pinned layer 11B and gives a spin injection torque to the free layer 12.

  Therefore, the magnetization of the free layer 12 changes from a downward direction on the paper surface to an upward surface on the paper surface.

  Thereafter, a read voltage (bias voltage) Vr is applied to the magnetoresistive effect element. The read voltage Vr is sufficiently smaller than the write voltage Vw “1”. The read voltage Vr brings the pinned layer 11A to a low potential and the pinned layer 11B to a high potential.

  At this time, the read current Ir flows from the pinned layer 11B toward the pinned layer 11A. In addition, when the read voltage Vr is biased, both the first and second units U1 and U2 are in an inverse state.

  Therefore, in the first unit U1, since the magnetization state of the pinned layer 11A and the free layer 12 is parallel, the high resistance state is obtained. In the second unit U2, the magnetization state of the pinned layer 11B and the free layer 12 is parallel. Therefore, a high resistance state is obtained.

  That is, since both are in a high resistance state, the MR ratio is not canceled between the first and second units U1 and U2, and both reduction in write current and improvement in MR ratio at the time of reading are achieved. Can do.

  Next, as shown in FIG. 14, when the magnetization of the free layer 12 is upward on the paper surface, in order to reverse the magnetization downward on the paper surface, the write voltage (negative bias voltage) Vw “0” is applied to the magnetoresistive effect element. "Is applied. The write voltage Vw “0” brings the pinned layer 11A to a high potential and the pinned layer 11B to a low potential.

  At this time, the write current (spin injection current) Iw “0” flows from the pinned layer 11A toward the pinned layer 11B.

  For this reason, the electrons (down spin) spin-polarized downward in the drawing pass through the pinned layer 11B and give a spin injection torque to the free layer 12. Further, the down spin is reflected by the pinned layer 11A and gives a spin injection torque to the free layer 12.

  Therefore, the magnetization of the free layer 12 changes from upward on the paper to downward on the paper.

  Thereafter, a read voltage (bias voltage) Vr is applied to the magnetoresistive effect element. The read voltage Vr is sufficiently smaller than the write voltage Vw “0”. The read voltage Vr brings the pinned layer 11A to a low potential and the pinned layer 11B to a high potential.

  At this time, the read current Ir flows from the pinned layer 11B toward the pinned layer 11A. In addition, when the read voltage Vr is biased, both the first and second units U1 and U2 are in an inverse state.

  Accordingly, in the first unit U1, since the magnetization state of the pinned layer 11A and the free layer 12 is antiparallel, the resistance state is low, and in the second unit U2, the magnetization state of the pinned layer 11B and the free layer 12 is Since it is antiparallel, it will be in a low resistance state.

  That is, since both are in a low resistance state, the MR ratio is not canceled by the first and second units U1 and U2, and both reduction in write current and improvement in MR ratio at the time of reading are achieved. Can do.

  As a state at the time of writing in FIG. 13 and FIG. 14, the first unit U1 may be in an inverse state and the second unit U2 may be in a normal state.

  In this case, when a bias condition as shown in FIG. 13 is given at the time of writing, the magnetization direction of the free layer 12 changes not upward on the paper but downward on the paper. Further, when a bias condition as shown in FIG. 14 is given at the time of writing, the magnetization direction of the free layer 12 changes not upward on the paper but upward on the paper.

  As for the read voltage Vr, the pinned layer 11A is set to a low potential and the pinned layer 11B is set to a high potential, but instead, the pinned layer 11A may be set to a high potential and the pinned layer 11B may be set to a low potential.

(3) Control of bias dependence of MR ratio
In the example of the present invention, one of the MR ratios of the first and second units has a bias dependency, and one of the MR ratios of the first and second units has a bias dependency is the read voltage and the write voltage Between the normal state and the inverse state, the region changes from one to the other.

  Here, an example of a method for realizing the TMR characteristic with respect to such a bias voltage will be described.

  The following two types of tunnel magnetoresistive elements having the basic structure of magnetic free layer / tunnel barrier layer / magnetic pinned layer can be realized.

One is a magnetoresistive effect element having a normal TMR effect at a low bias and an inverse TMR effect at a high bias.
The other is a magnetoresistive element having an inverse TMR effect at a low bias and a normal TMR effect at a high bias.

  Whether the tunnel magnetoresistive element is in a normal state or an inverse state is influenced by the electrical conduction characteristics of the material constituting the tunnel barrier layer, and the tunnel barrier layer and the magnetic free layer. And the electron level at the interface between the tunnel barrier layer and the magnetic pinned layer.

  In the former case, the bias dependence of the MR ratio of the tunnel magnetoresistive element can be controlled by selecting the material constituting the tunnel barrier layer.

  In the latter case, the bias dependency of the MR ratio of the tunnel magnetoresistive element can be controlled by changing the combination of materials of the magnetic pinned layer, the tunnel barrier layer, and the magnetic free layer.

Typical examples of the tunnel barrier layer in which the normal TMR effect occurs at low bias include oxides such as Al ₂ O ₃ , TiO ₂ , MgO, CaO, SrO, and TiO.

Tunnel barrier layers composed of such oxides are, for example, Co _1-ab Fe _a Ni _b / oxide / Co _1-ab Fe _a Ni _b , Co _1-ab Fe _a Ni _b / Al ₂ O ₃ It is often used in the form of / La _1-x Sr _x MnO ₃ .

The basic unit that generates the inverse TMR effect at low bias is Co _1-ab Fe _a Ni _b / SrTiO ₃ / La _1-x Sr _x Mo ₃ , Co _1-ab Fe _a Ni _b / Zr _1-x O _x / Co _1-ab Fe _a Ni _b . These can generate a normal TMR effect at the time of high bias by devising a creation method.

  In addition, a material such as cosputtering can be used to form a tunnel barrier layer by mixing a material that generates an inverse TMR effect and a material that has a normal TMR effect.

  By the way, when the bias dependency of the MR ratio is controlled only by selecting such a material, the TMR characteristics between the normal state and the inverse state are reduced between the read voltage and the write voltage. There is a current situation where it is difficult to change from one to the other stably.

  As a countermeasure, it is conceivable to combine two or more units having different bias dependencies. In this case, it is possible to manufacture a tunnel magnetoresistive element in which switching of TMR characteristics (inverse / normal) is performed at an ideal point (bias voltage).

  For example, as shown in FIG. 15, a unit (a) having a small bias dependency and having a normal TMR effect at a low bias, and an inverse TMR effect at a low bias and having a large bias dependency. When combined with the unit having the same (FIG. 5B), the transition point Vc1 from the inverse state to the normal state is easily set between the write voltage Vw “0” (or Vw “1”) and the read voltage Vr. Yes ((c) in the figure).

  Further, as shown in FIG. 16, a unit (a) having a large bias dependency and having a normal TMR effect at a low bias and an inverse TMR effect at a low bias and having a small bias dependency. When combined with the unit having the same (FIG. 5B), the transition point Vc1 from the normal state to the inverse state can be easily changed between the write voltage Vw “1” (or Vw “0”) 1 and the read voltage Vr. It can be set ((c) in the figure).

  Here, the two write voltages Vw “0” and Vw “1” are present in the spin injection write method in the direction of the write current (spin injection current) according to the write data “0” and “1”. Due to changing.

  That is, when the first data “0” is written, a write voltage (for example, a positive bias voltage) Vw “0” is applied to the magnetoresistive element, and when the second data “1” is written, the write voltage ( For example, a negative bias voltage (Vw “1”) is applied to the magnetoresistive element.

  The values of the write voltages Vw “0” and Vw “1” are set to optimum values in consideration of the bias dependency of the MR ratio (change points Vc1 and Vc2). Further, the read voltage Vr is set to either the positive bias side or the negative bias side in consideration of a margin or the like.

  In order to generate the normal TMR effect and the inverse TMR effect according to the bias voltage in one basic unit, a nonmagnetic metal layer is inserted between the tunnel barrier layer and the free layer or between the tunnel barrier layer and the pinned layer. The vibration of the TMR ratio that appears in the case can also be used.

For example, in the case of the basic unit of Co _1-ab Fe _a Ni _b / Al ₂ O ₃ / Cu / Co _1-ab Fe _a Ni _b , the TMR ratio is vibrated between positive and negative by controlling the thickness of Cu. It is possible to make it.

The basic unit Co _1-ab Fe _a Ni _b / MgO / non-magnetic metal / Co _1-ab Fe _a Ni _b (however, non-magnetic metals are Au, Ru, Pt, Ir, Cu, Ag, Rh, In the case of Os), it is possible to vibrate the TMR ratio between positive and negative by controlling the thickness of the nonmagnetic metal.

  Even when nonmagnetic oxides such as CoO, NiO, MnO, CrO, FeO, VO, TiO are used instead of nonmagnetic metals, the normal TMR effect is generated at low bias and the inverse TMR effect is generated at high bias. Can do.

For example, in the case of the basic unit Co _1-ab Fe _a Ni _b / MgO / non-magnetic oxide / Co _1-ab Fe _a Ni _b , MgO has a NaCl structure and is preferentially oriented in the (100) plane. Therefore, it is desirable that the nonmagnetic oxide also has a NaCl structure and is preferentially oriented in the (100) plane.

  CoO, NiO, MnO, CrO, FeO, VO, and TiO described above have a NaCl structure. These materials have been confirmed to have antiferromagnetic magnetic properties or electronic states by an electronic state analysis or an analysis of magnetization arrangement by XPS or XMCD analysis.

  As the tunnel barrier layer having the NaCl structure, CaO, SrO, EuO, TiO, etc. are known in addition to MgO.

  The nonmagnetic oxide is formed after a nonmagnetic metal such as Co, Ni, Mn, Cr, Fe, V, Ti is formed between the tunnel barrier layer and the free layer or between the tunnel barrier layer and the pinned layer. It is formed by oxidizing these nonmagnetic metals.

  The nonmagnetic oxide can also be deposited directly on the free layer or pinned layer.

  At least one of the pinned layer and the free layer may have a structure of magnetic layer / intermediate layer / magnetic layer. In this case, the two magnetic layers have a magnetic coupling with each other via the intermediate layer.

For example, a SAF (Synthetic anti-ferromagnetic) structure in which two magnetic layers are antiferromagnetically coupled can be employed in at least one of the pinned layer and the free layer. The SAF structure mainly includes Co _1-ab Fe _a Ni _b / Ru / Co _1-ab Fe _a Ni _b , Co _1-ab Fe _a Ni _b / Os / Co _1-ab Fe _a Ni _b , Co _{1 -ab} Fe _a Ni _b / Ir / Co _1-ab Fe _a Ni _b is often used.

The pinned layer and the free layer have a multilayer structure, for example, [Co _1-ab Fe _a Ni _b / Cr] n, [Co _1-ab Fe _a Ni _b / Ru] n (where n is a structure in [] May be stacked in n stages.). If the thickness of the magnetic layer is optimized, a desired bias dependency can be obtained. In particular, the ratio between the thickness of the magnetic layer in contact with the tunnel barrier layer and the thickness of the magnetic layer not in contact with the tunnel barrier layer is important.

5. Example
Examples will be described below.
In each embodiment, arrows in the pinned layer and the free layer indicate the direction of magnetization.

(1) First embodiment
FIG. 17 shows the first embodiment.

  In the first embodiment, the magnetization directions of the two pinned layers 11A and 11B are set to be opposite to each other. The magnetization direction of the residual magnetization (the easy axis direction) of the pinned layers 11A and 11B and the free layer 12 is the direction in which they are stacked (perpendicular magnetization).

  The spacer layers (nonmagnetic layers) 13A and 13B are tunnel barrier layers, and are made of, for example, an insulator such as MgO or AlOx or a semiconductor such as Ga or Ge. In this case, both the first and second units U1, U2 have a bias dependency.

  The bias dependence of the MR ratio of the first unit U1 exhibits inverse TMR characteristics in the low bias state and normal TMR characteristics in the high bias state.

  The point Vc1 on the positive bias side where the characteristic is switched is between the write voltage Vw “1” and the read voltage Vr. Further, the negative bias side point Vc2 at which the characteristic is switched is in a region lower than the write voltage Vw “0”.

  The bias dependence of the MR ratio of the second unit U2 shows normal TMR characteristics in the low bias state and inverse TMR characteristics in the high bias state.

  In the second unit U2, the point at which the characteristics are switched is in a region sufficiently higher than the write voltages Vw “1” and Vw “0”.

  When a write voltage (positive bias voltage) Vw “1” is applied to such a magnetoresistive effect element, the magnetization state of the first unit U1 becomes parallel, the magnetization state of the second unit U2 becomes antiparallel, and data “1” "Is written.

  Further, when a write voltage (negative bias voltage) Vw “0” is applied to the magnetoresistive effect element, the magnetization state of the first unit U1 becomes antiparallel, the magnetization state of the second unit U2 becomes parallel, and data “0”. Is written.

  When the read voltage Vr is biased, the first unit U1 is in an inverse state and the second unit U2 is in a normal state, so that data “0” is in a low resistance state and data “1” is in a high resistance state.

(2) Second embodiment
FIG. 18 shows a second embodiment.

  In the second embodiment, the magnetization directions of the two pinned layers 11A and 11B are set to the same direction. The magnetization direction of the residual magnetization (the easy axis direction) of the pinned layers 11A and 11B and the free layer 12 is the direction in which they are stacked (perpendicular magnetization).

  The spacer layers (nonmagnetic layers) 13A and 13B are tunnel barrier layers, and are made of, for example, an insulator such as MgO or AlOx or a semiconductor such as Ga or Ge. In this case, both the first and second units U1, U2 have a bias dependency.

  The bias dependence of the MR ratio of the first and second units U1, U2 shows normal TMR characteristics in the low bias state and inverse TMR characteristics in the high bias state.

  However, in the first unit U1, the point at which the characteristics are switched is in a region sufficiently higher than the write voltages Vw “0” and Vw “1”.

  On the other hand, in the second unit U2, the point Vc1 on the positive bias side where the characteristic is switched is between the write voltage Vw “0” and the read voltage Vr. Further, the point Vc2 on the negative bias side where the characteristic is switched is in a region lower than the write voltage Vw “1”.

  When a write voltage (positive bias voltage) Vw “0” is applied to such a magnetoresistive element, the magnetization states of the first and second units U1 and U2 are both parallel, and data “0” is written. It is.

  When a write voltage (negative bias voltage) Vw “1” is applied to the magnetoresistive effect element, the magnetization states of the first and second units U1 and U2 are both antiparallel, and data “1” is written. It is.

  When the read voltage Vr is biased, the first and second units U1, U2 are both in the normal state, so that the data “0” is in the low resistance state and the data “1” is in the high resistance state.

(3) Third embodiment
FIG. 19 shows a third embodiment.

  In the third embodiment, the magnetization directions of the two pinned layers 11A and 11B are set to be opposite to each other. The magnetization direction of the residual magnetization (the easy axis direction) of the pinned layers 11A and 11B and the free layer 12 is the direction in which they are stacked (perpendicular magnetization).

  The spacer layer (nonmagnetic layer) 13A is a tunnel barrier layer and is made of, for example, an insulator such as MgO or AlOx or a semiconductor such as Ga or Ge. On the other hand, the spacer layer (nonmagnetic layer) 13B is a metal spacer layer and is made of a conductor such as Cu or Au. In this case, only the first unit U1 has a bias dependency.

  The bias dependence of the MR ratio of the first unit U1 exhibits normal TMR characteristics in the low bias state and inverse TMR characteristics in the high bias state.

  The point Vc1 on the positive bias side where the characteristic is switched is between the write voltage Vw “1” and the read voltage Vr. Further, the negative bias side point Vc2 at which the characteristic is switched is in a region lower than the write voltage Vw “0”.

  The MR ratio of the second unit U2 is constant regardless of the bias voltage, and always shows an inverse TMR characteristic.

  When a write voltage (positive bias voltage) Vw “1” is applied to such a magnetoresistive effect element, the magnetization state of the first unit U1 becomes antiparallel, the magnetization state of the second unit U2 becomes parallel, and data “1” "Is written.

  When a write voltage (negative bias voltage) Vw “0” is applied to the magnetoresistive effect element, the magnetization state of the first unit U1 becomes parallel, the magnetization state of the second unit U2 becomes antiparallel, and data “0”. Is written.

  When the read voltage Vr is biased, the first unit U1 is in the normal state and the second unit U2 is in the inverse state, so that the data “1” is in the high resistance state and the data “0” is in the low resistance state.

(4) Fourth embodiment
FIG. 20 shows a fourth embodiment.

  In the fourth embodiment, the magnetization directions of the two pinned layers 11A and 11B are set to the same direction. The magnetization direction of the residual magnetization (the easy axis direction) of the pinned layers 11A and 11B and the free layer 12 is the direction in which they are stacked (perpendicular magnetization).

  The spacer layer (nonmagnetic layer) 13A is a tunnel barrier layer and is made of, for example, an insulator such as MgO or AlOx or a semiconductor such as Ga or Ge. On the other hand, the spacer layer (nonmagnetic layer) 13B is a metal spacer layer and is made of a conductor such as Cu or Au. In this case, only the first unit U1 has a bias dependency.

  The bias dependence of the MR ratio of the first unit U1 exhibits normal TMR characteristics in the low bias state and inverse TMR characteristics in the high bias state.

  The point Vc1 on the positive bias side where the characteristic is switched is between the write voltage Vw “1” and the read voltage Vr. Further, the negative bias side point Vc2 at which the characteristic is switched is in a region lower than the write voltage Vw “0”.

  The MR ratio of the second unit U2 is constant regardless of the bias voltage, and always exhibits normal TMR characteristics.

  When a write voltage (positive bias voltage) Vw “1” is applied to such a magnetoresistive effect element, the magnetization states of the first and second units U1 and U2 are both antiparallel, and data “1” is stored. Written.

  When a write voltage (negative bias voltage) Vw “0” is applied to the magnetoresistive effect element, the magnetization states of the first and second units U1 and U2 are both parallel and data “0” is written. .

  When the read voltage Vr is biased, both the first and second units U1, U2 are in the normal state, so that the data “1” is in the high resistance state and the data “0” is in the low resistance state.

(5) Fifth embodiment
FIG. 21 shows a fifth embodiment.

  In the fifth embodiment, the magnetization directions of the two pinned layers 11A and 11B are set to be opposite to each other. The magnetization directions of the pinned layers 11A and 11B are fixed by the antiferromagnetic layer 14. In addition, the magnetization direction (easy magnetization axis direction) of the residual magnetization of the pinned layers 11A and 11B and the free layer 12 is a direction perpendicular to the direction in which they are stacked (in-plane magnetization).

  The pinned layer 11A has a three-layer structure of ferromagnetic layer / nonmagnetic layer / ferromagnetic layer, and the two ferromagnetic layers are magnetically coupled to each other via the nonmagnetic layer. In this example, the pinned layer 11A has SAF (synthetic anti-ferromagnetic) coupling in which the magnetization directions of the two ferromagnetic layers are opposite to each other.

  The pinned layer 11B has a five-layer structure of ferromagnetic layer / nonmagnetic layer / ferromagnetic layer / nonmagnetic layer / ferromagnetic layer, and the two ferromagnetic layers present on both sides of the nonmagnetic layer are magnetically coupled to each other. is doing. In this example, the pinned layer 11B has SAF coupling in which the magnetization directions of the two ferromagnetic layers are opposite to each other.

  If the pinned layers 11A and 11B have a ferromagnetic layer / nonmagnetic layer / ferromagnetic layer structure, the magnetization state is thermally stabilized, and fluctuations with respect to an external magnetic field are less likely to occur. Specifically, since the magnetization at the end of the ferromagnetic layer is close to a single domain structure, the apparent saturation magnetization from the outside can be made zero by magnetostatic coupling.

  The number of ferromagnetic layers / nonmagnetic layers / ferromagnetic layer structures in pinned layer 11A (one in this example) is the number of ferromagnetic layers / nonmagnetic layers / ferromagnetic layer structures in pinned layer 11B (in this example). Then, it is desirable to make it different from 2).

  As for the magnetization states of the first and second units U1 and U2, how the magnetization direction of the free layer 12 is relative to the magnetization direction of the ferromagnetic layer adjacent to the spacer layers (nonmagnetic layers) 13A and 13B. Or is determined by.

  That is, the magnetization state of the first unit U1 is determined by the magnetization directions of the two ferromagnetic layers adjacent to the spacer layer 13A, and the magnetization state of the second unit U2 is the magnetization of the two ferromagnetic layers adjacent to the spacer layer 13B. It depends on the direction.

  The ferromagnetic layer of the free layer 12 may have either a single layer structure or a laminated structure. Further, the free layer 12 may have a ferromagnetic layer / nonmagnetic layer / ferromagnetic layer structure, and the two ferromagnetic layers may have a magnetic interaction.

  The ferromagnetic layers in the pinned layers 11A and 11B are made of a ferromagnetic material such as fcc-Co, hcp-Co, bcc-Fe, fcc-CoFe alloy, bcc-CoFe alloy, and amorphous CoFeB alloy. The nonmagnetic layers in the pinned layers 11A and 11B are made of a conductor such as Ru, Ir, or Os. The antiferromagnetic layer is made of an antiferromagnetic material such as PtMn, IrMn, NiMn, FeMn, PtCrMn, RhMn, or FeRh.

  The spacer layers (nonmagnetic layers) 13A and 13B are tunnel barrier layers, and are made of, for example, an insulator such as MgO or AlOx or a semiconductor such as Ga or Ge. In this case, both the first and second units (TMR units) U1 and U2 have bias dependency.

  The bias dependence of the MR ratio of the first unit U1 exhibits normal TMR characteristics in the low bias state and inverse TMR characteristics in the high bias state.

  In the first unit U1, the point at which the characteristics are switched is in a region sufficiently higher than the write voltages Vw “0” and Vw “1”.

  The bias dependence of the MR ratio of the second unit U2 exhibits inverse TMR characteristics in the low bias state and normal TMR characteristics in the high bias state.

  The point Vc1 on the positive bias side where the characteristic is switched is between the write voltage Vw “0” and the read voltage Vr. Further, the point Vc2 on the negative bias side where the characteristic is switched is in a region lower than the write voltage Vw “1”.

  When a write voltage (positive bias voltage) Vw “0” is applied to such a magnetoresistive effect element, the magnetization state of the first unit U1 becomes parallel, the magnetization state of the second unit U2 becomes antiparallel, and the data “0”. "Is written.

  When a write voltage (negative bias voltage) Vw “1” is applied to the magnetoresistive effect element, the magnetization state of the first unit U1 becomes antiparallel, the magnetization state of the second unit U2 becomes parallel, and data “1”. Is written.

  When the read voltage Vr is biased, the first unit U1 is in the normal state and the second unit U2 is in the inverse state, so that the data “0” is in the low resistance state and the data “1” is in the high resistance state.

(6) Sixth embodiment
FIG. 22 shows a sixth embodiment.

The sixth embodiment is a modification of the fifth embodiment.
The sixth embodiment is different from the fifth embodiment in that the free layer 12 has a SAF structure.

  The free layer 12 has a five-layer structure of ferromagnetic layer / nonmagnetic layer / ferromagnetic layer / nonmagnetic layer / ferromagnetic layer, and the two ferromagnetic layers existing on both sides of the nonmagnetic layer are antiferromagnetic to each other. Magnetically coupled.

  When the free layer 12 has the SAF structure, the magnetization state is thermally stabilized, and fluctuations with respect to the external magnetic field are less likely to occur. Specifically, since the magnetization at the end of the ferromagnetic layer is close to a single domain structure, the apparent saturation magnetization from the outside can be made zero by magnetostatic coupling.

  The magnetization state of the first unit U1 is determined by the magnetization direction of the lowermost ferromagnetic layer of the free layer 12 with respect to the magnetization direction of the uppermost ferromagnetic layer of the pinned layer 11A. Is done. That is, the magnetization state of the first unit U1 is determined by the magnetization directions of the two ferromagnetic layers adjacent to the spacer layer 13A.

  The magnetization state of the second unit U2 is determined by the magnetization direction of the uppermost ferromagnetic layer of the free layer 12 with respect to the magnetization direction of the lowermost ferromagnetic layer of the pinned layer 11B. Is done. That is, the magnetization state of the second unit U2 is determined by the magnetization directions of the two ferromagnetic layers adjacent to the spacer layer 13B.

  The bias dependence of the MR ratio of the first unit U1 exhibits normal TMR characteristics in the low bias state and inverse TMR characteristics in the high bias state.

  In the first unit U1, the point at which the characteristics are switched is in a region sufficiently higher than the write voltages Vw “0” and Vw “1”.

  The bias dependence of the MR ratio of the second unit U2 exhibits inverse TMR characteristics in the low bias state and normal TMR characteristics in the high bias state.

  The point Vc1 on the positive bias side where the characteristic is switched is between the write voltage Vw “0” and the read voltage Vr. Further, the point Vc2 on the negative bias side where the characteristic is switched is in a region lower than the write voltage Vw “1”.

  When a write voltage (positive bias voltage) Vw “0” is applied to such a magnetoresistive effect element, the magnetization state of the first unit U1 becomes parallel, the magnetization state of the second unit U2 becomes antiparallel, and the data “0”. "Is written.

  When a write voltage (negative bias voltage) Vw “1” is applied to the magnetoresistive effect element, the magnetization state of the first unit U1 becomes antiparallel, the magnetization state of the second unit U2 becomes parallel, and data “1”. Is written.

  When the read voltage Vr is biased, the first unit U1 is in a normal state and the second unit U2 is in an inverse state, so that data “0” is in a low resistance state and data “1” is in a high resistance state.

(7) Seventh embodiment
FIG. 23 shows a seventh embodiment.

  In the seventh embodiment, the magnetization directions of the two pinned layers 11A and 11B are set to the same direction. The magnetization directions of the pinned layers 11A and 11B are fixed by the antiferromagnetic layer 14. In addition, the magnetization direction (easy magnetization axis direction) of the residual magnetization of the pinned layers 11A and 11B and the free layer 12 is a direction perpendicular to the direction in which they are stacked (in-plane magnetization).

  Each of the pinned layers 11A and 11B has a three-layer structure of ferromagnetic layer / nonmagnetic layer / ferromagnetic layer, and the two ferromagnetic layers are magnetically coupled to each other via the nonmagnetic layer. In this example, the pinned layers 11A and 11B have SAF coupling in which the magnetization directions of the two ferromagnetic layers are opposite to each other.

  The free layer 12 has a three-layer structure of ferromagnetic layer / nonmagnetic layer / ferromagnetic layer, and the two ferromagnetic layers are magnetically coupled to each other via the nonmagnetic layer. In this example, the free layer 12 has SAF coupling in which the magnetization directions of the two ferromagnetic layers are opposite to each other.

  If the pinned layers 11A and 11B and the free layer 12 have a ferromagnetic layer / nonmagnetic layer / ferromagnetic layer structure, the magnetization state is thermally stabilized, and fluctuations with respect to an external magnetic field are less likely to occur.

  The magnetization state of the first unit U1 is determined by the magnetization direction of the lowermost ferromagnetic layer of the free layer 12 with respect to the magnetization direction of the uppermost ferromagnetic layer of the pinned layer 11A. Is done.

  The magnetization state of the second unit U2 is determined by the magnetization direction of the uppermost ferromagnetic layer of the free layer 12 with respect to the magnetization direction of the lowermost ferromagnetic layer of the pinned layer 11B. Is done.

  The ferromagnetic layers in the pinned layers 11A and 11B and the free layer 12 are made of a ferromagnetic material such as fcc-Co, hcp-Co, bcc-Fe, fcc-CoFe alloy, bcc-CoFe alloy, and amorphous CoFeB alloy. . The nonmagnetic layers in the pinned layers 11A and 11B and the free layer 12 are made of a conductor such as Ru, Ir, or Os. The antiferromagnetic layer is made of an antiferromagnetic material such as PtMn, IrMn, NiMn, FeMn, PtCrMn, RhMn, or FeRh.

  The spacer layers (nonmagnetic layers) 13A and 13B are tunnel barrier layers, and are made of, for example, an insulator such as MgO or AlOx or a semiconductor such as Ga or Ge. In this case, both the first and second units (TMR units) U1 and U2 have bias dependency.

  The bias dependence of the MR ratio of the first unit U1 exhibits normal TMR characteristics in the low bias state and inverse TMR characteristics in the high bias state.

  In the first unit U1, the point at which the characteristics are switched is in a region sufficiently higher than the write voltages Vw “0” and Vw “1”.

  The bias dependence of the MR ratio of the second unit U2 exhibits inverse TMR characteristics in the low bias state and normal TMR characteristics in the high bias state.

  The point Vc1 on the positive bias side where the characteristic is switched is between the write voltage Vw “0” and the read voltage Vr. Further, the point Vc2 on the negative bias side where the characteristic is switched is in a region lower than the write voltage Vw “1”.

  When a write voltage (positive bias voltage) Vw “0” is applied to such a magnetoresistive effect element, the magnetization state of the first unit U1 becomes parallel, the magnetization state of the second unit U2 becomes antiparallel, and the data “0”. "Is written.

  When a write voltage (negative bias voltage) Vw “1” is applied to the magnetoresistive effect element, the magnetization state of the first unit U1 becomes antiparallel, the magnetization state of the second unit U2 becomes parallel, and data “1”. Is written.

  When the read voltage Vr is biased, the first unit U1 is in a normal state and the second unit U2 is in an inverse state, so that data “0” is in a low resistance state and data “1” is in a high resistance state.

(8) Eighth Example
FIG. 24 shows an eighth embodiment.

  In the eighth embodiment, the magnetization directions of the two pinned layers 11A and 11B are set to be opposite to each other. The magnetization directions of the pinned layers 11A and 11B are fixed by the antiferromagnetic layer 14. In addition, the magnetization direction (easy magnetization axis direction) of the residual magnetization of the pinned layers 11A and 11B and the free layer 12 is a direction perpendicular to the direction in which they are stacked (in-plane magnetization).

  The pinned layer 11A has a three-layer structure of ferromagnetic layer / nonmagnetic layer / ferromagnetic layer, and the two ferromagnetic layers are magnetically coupled to each other via the nonmagnetic layer. In this example, the pinned layer 11A has SAF coupling in which the magnetization directions of the two ferromagnetic layers are opposite to each other.

  The pinned layer 11B has a five-layer structure of ferromagnetic layer / nonmagnetic layer / ferromagnetic layer / nonmagnetic layer / ferromagnetic layer, and the two ferromagnetic layers present on both sides of the nonmagnetic layer are magnetically coupled to each other. is doing. In this example, the pinned layer 11B has SAF coupling in which the magnetization directions of the two ferromagnetic layers are opposite to each other.

  The free layer 12 has a three-layer structure of ferromagnetic layer / nonmagnetic layer / ferromagnetic layer, and the two ferromagnetic layers are magnetically coupled to each other via the nonmagnetic layer. In this example, the free layer 12 has SAF coupling in which the magnetization directions of the two ferromagnetic layers are opposite to each other.

  If the pinned layers 11A and 11B and the free layer 12 have a ferromagnetic layer / nonmagnetic layer / ferromagnetic layer structure, the magnetization state is thermally stabilized, and fluctuations with respect to an external magnetic field are less likely to occur.

  The magnetization state of the first unit U1 is determined by the magnetization direction of the lowermost ferromagnetic layer of the free layer 12 with respect to the magnetization direction of the uppermost ferromagnetic layer of the pinned layer 11A. Is done.

  The magnetization state of the second unit U2 is determined by the magnetization direction of the uppermost ferromagnetic layer of the free layer 12 with respect to the magnetization direction of the lowermost ferromagnetic layer of the pinned layer 11B. Is done.

  The ferromagnetic layers in the pinned layers 11A and 11B and the free layer 12 are made of a ferromagnetic material such as fcc-Co, hcp-Co, bcc-Fe, fcc-CoFe alloy, bcc-CoFe alloy, and amorphous CoFeB alloy. . The nonmagnetic layers in the pinned layers 11A and 11B and the free layer 12 are made of a conductor such as Ru, Ir, or Os. The antiferromagnetic layer is made of an antiferromagnetic material such as PtMn, IrMn, NiMn, FeMn, PtCrMn, RhMn, or FeRh.

  The spacer layers (nonmagnetic layers) 13A and 13B are tunnel barrier layers, and are made of, for example, an insulator such as MgO or AlOx or a semiconductor such as Ga or Ge. In this case, both the first and second units (TMR units) U1 and U2 have bias dependency.

  The bias dependence of the MR ratio of the first and second units U1 and U2 shows normal TMR characteristics in the low bias state and inverse TMR characteristics in the high bias state, respectively.

  However, in the first unit U1, the point at which the characteristics are switched is in a region sufficiently higher than the write voltages Vw “0” and Vw “1”.

  On the other hand, in the second unit U2, the point Vc1 on the positive bias side where the characteristic is switched is between the write voltage Vw “0” and the read voltage Vr. Further, the point Vc2 on the negative bias side where the characteristic is switched is in a region lower than the write voltage Vw “1”.

  When a write voltage (positive bias voltage) Vw “0” is applied to such a magnetoresistive element, the magnetization states of the first and second units U1 and U2 are both parallel and data “0” is written. It is.

  When a write voltage (negative bias voltage) Vw “1” is applied to the magnetoresistive effect element, the magnetization states of the first and second units U1 and U2 are both antiparallel, and data “1” is written. It is.

  When the read voltage Vr is biased, the first and second units U1, U2 are both in the normal state, so that the data “0” is in the low resistance state and the data “1” is in the high resistance state.

(9) Ninth embodiment
FIG. 25 shows a ninth embodiment.

  In the ninth embodiment, the magnetization directions of the two pinned layers 11A and 11B are set to the same direction. The magnetization directions of the pinned layers 11A and 11B are fixed by the antiferromagnetic layer 14. Further, the magnetization direction (easy magnetization axis direction) of the residual magnetization of the pinned layers 11A and 11B and the free layer 12 is a direction perpendicular to the stacking direction (in-plane magnetization).

  Each of the pinned layers 11A and 11B has a three-layer structure of ferromagnetic layer / nonmagnetic layer / ferromagnetic layer, and the two ferromagnetic layers are magnetically coupled to each other via the nonmagnetic layer. In this example, the pinned layers 11A and 11B have SAF coupling in which the magnetization directions of the two ferromagnetic layers are opposite to each other.

  The free layer 12 has a five-layer structure of ferromagnetic layer / nonmagnetic layer / ferromagnetic layer / nonmagnetic layer / ferromagnetic layer, and two ferromagnetic layers existing on both sides of the nonmagnetic layer are magnetically coupled to each other. is doing. In this example, the free layer 12 has SAF coupling in which the magnetization directions of the two ferromagnetic layers are opposite to each other.

  If the pinned layers 11A and 11B and the free layer 12 have a ferromagnetic layer / nonmagnetic layer / ferromagnetic layer structure, the magnetization state is thermally stabilized, and fluctuations with respect to an external magnetic field are less likely to occur.

  The magnetization state of the first unit U1 is determined by the magnetization direction of the lowermost ferromagnetic layer of the free layer 12 with respect to the magnetization direction of the uppermost ferromagnetic layer of the pinned layer 11A. Is done.

  The magnetization state of the second unit U2 is determined by the magnetization direction of the uppermost ferromagnetic layer of the free layer 12 with respect to the magnetization direction of the lowermost ferromagnetic layer of the pinned layer 11B. Is done.

  The ferromagnetic layers in the pinned layers 11A and 11B and the free layer 12 are made of a ferromagnetic material such as fcc-Co, hcp-Co, bcc-Fe, fcc-CoFe alloy, bcc-CoFe alloy, and amorphous CoFeB alloy. . The nonmagnetic layers in the pinned layers 11A and 11B and the free layer 12 are made of a conductor such as Ru, Ir, or Os. The antiferromagnetic layer is made of an antiferromagnetic material such as PtMn, IrMn, NiMn, FeMn, PtCrMn, RhMn, or FeRh.

  The spacer layers (nonmagnetic layers) 13A and 13B are tunnel barrier layers, and are made of, for example, an insulator such as MgO or AlOx or a semiconductor such as Ga or Ge. In this case, both the first and second units (TMR units) U1 and U2 have bias dependency.

  The bias dependence of the MR ratio of the first and second units U1 and U2 shows normal TMR characteristics in the low bias state and inverse TMR characteristics in the high bias state, respectively.

  However, in the first unit U1, the point at which the characteristics are switched is in a region sufficiently higher than the write voltages Vw “0” and Vw “1”.

  On the other hand, in the second unit U2, the point Vc1 on the positive bias side where the characteristic is switched is between the write voltage Vw “0” and the read voltage Vr. Further, the point Vc2 on the negative bias side where the characteristic is switched is in a region lower than the write voltage Vw “1”.

  When a write voltage (positive bias voltage) Vw “0” is applied to such a magnetoresistive element, the magnetization states of the first and second units U1 and U2 are both parallel and data “0” is written. It is.

  When a write voltage (negative bias voltage) Vw “1” is applied to the magnetoresistive effect element, the magnetization states of the first and second units U1 and U2 are both antiparallel, and data “1” is written. It is.

  When the read voltage Vr is biased, the first and second units U1, U2 are both in the normal state, so that the data “0” is in the low resistance state and the data “1” is in the high resistance state.

(10) Tenth embodiment
FIG. 26 shows a tenth embodiment.

The tenth embodiment is a modification of the sixth embodiment.
The tenth embodiment is different from the sixth embodiment in that the spacer layer 13B of the second unit U2 is not a tunnel barrier layer (insulator or semiconductor) but a metal spacer layer (conductor). .

  In this case, the second unit U2 functions as a GMR unit, and the MR ratio has no bias dependency. In this example, the MR ratio of the second unit U2 has a constant value regardless of the bias voltage, and is always in the normal state.

  On the other hand, the first unit U1 functions as a TMR unit and exhibits inverse TMR characteristics in a low bias state and normal TMR characteristics in a high bias state.

  In the first unit U1, the point Vc1 on the positive bias side where the characteristic is switched is between the write voltage Vw “1” and the read voltage Vr. Further, the negative bias side point Vc2 at which the characteristic is switched is in a region lower than the write voltage Vw “0”.

  When a write voltage (positive bias voltage) Vw “1” is applied to such a magnetoresistive effect element, the magnetization state of the first unit U1 becomes parallel, the magnetization state of the second unit U2 becomes antiparallel, and data “1” "Is written.

  Further, when a write voltage (negative bias voltage) Vw “0” is applied to the magnetoresistive effect element, the magnetization state of the first unit U1 becomes antiparallel, the magnetization state of the second unit U2 becomes parallel, and data “0”. Is written.

  When the read voltage Vr is biased, the first unit U1 is in an inverse state and the second unit U2 is in a normal state, so that data “0” is in a low resistance state and data “1” is in a high resistance state.

(11) Eleventh embodiment
FIG. 27 shows an eleventh embodiment.

The eleventh embodiment is a modification of the seventh embodiment.
The difference between the eleventh embodiment and the seventh embodiment is that the spacer layer 13B of the second unit U2 is not a tunnel barrier layer (insulator or semiconductor) but a metal spacer layer (conductor). .

  In this case, the second unit U2 functions as a GMR unit, and the MR ratio has no bias dependency. In this example, the MR ratio of the second unit U2 has a constant value regardless of the bias voltage, and is always in the normal state.

  On the other hand, the first unit U1 functions as a TMR unit and exhibits inverse TMR characteristics in a low bias state and normal TMR characteristics in a high bias state.

  In the first unit U1, the point Vc1 on the positive bias side where the characteristic is switched is between the write voltage Vw “1” and the read voltage Vr. Further, the negative bias side point Vc2 at which the characteristic is switched is in a region lower than the write voltage Vw “0”.

  When a write voltage (positive bias voltage) Vw “1” is applied to such a magnetoresistive effect element, the magnetization state of the first unit U1 becomes parallel, the magnetization state of the second unit U2 becomes antiparallel, and data “1” "Is written.

  Further, when a write voltage (negative bias voltage) Vw “0” is applied to the magnetoresistive effect element, the magnetization state of the first unit U1 becomes antiparallel, the magnetization state of the second unit U2 becomes parallel, and data “0”. Is written.

  When the read voltage Vr is biased, the first unit U1 is in an inverse state and the second unit U2 is in a normal state, so that data “0” is in a low resistance state and data “1” is in a high resistance state.

(12) 12th embodiment
FIG. 28 shows a twelfth embodiment.

The twelfth embodiment is a modification of the eighth embodiment.
The twelfth embodiment is different from the eighth embodiment in that the spacer layer 13B of the second unit U2 is not a tunnel barrier layer (insulator or semiconductor) but a metal spacer layer (conductor). .

  In this case, the second unit U2 functions as a GMR unit, and the MR ratio has no bias dependency. In this example, the MR ratio of the second unit U2 has a constant value regardless of the bias voltage, and is always in the normal state.

  On the other hand, the first unit U1 functions as a TMR unit and exhibits normal TMR characteristics in a low bias state and inverse TMR characteristics in a high bias state.

  In the first unit U1, the point Vc1 on the positive bias side where the characteristic is switched is between the write voltage Vw “1” and the read voltage Vr. Further, the negative bias side point Vc2 at which the characteristic is switched is in a region lower than the write voltage Vw “0”.

  When a write voltage (positive bias voltage) Vw “1” is applied to such a magnetoresistive element, the magnetization states of the first and second units U1 and U2 are antiparallel, and data “1” is stored. Written.

  When a write voltage (negative bias voltage) Vw “0” is applied to the magnetoresistive effect element, the magnetization states of the first and second units U1 and U2 become parallel and data “0” is written. .

  When the read voltage Vr is biased, the first and second units U1, U2 are both in the normal state, so that the data “0” is in the low resistance state and the data “1” is in the high resistance state.

(13) Thirteenth embodiment
FIG. 29 shows a thirteenth embodiment.

The thirteenth embodiment is a modification of the ninth embodiment.
The thirteenth embodiment is different from the ninth embodiment in that the spacer layer 13B of the second unit U2 is not a tunnel barrier layer (insulator or semiconductor) but a metal spacer layer (conductor). .

  In this case, the second unit U2 functions as a GMR unit, and the MR ratio has no bias dependency. In this example, the MR ratio of the second unit U2 has a constant value regardless of the bias voltage, and is always in the normal state.

  On the other hand, the first unit U1 functions as a TMR unit and exhibits normal TMR characteristics in a low bias state and inverse TMR characteristics in a high bias state.

  In the first unit U1, the point Vc1 on the positive bias side where the characteristic is switched is between the write voltage Vw “1” and the read voltage Vr. Further, the negative bias side point Vc2 at which the characteristic is switched is in a region lower than the write voltage Vw “0”.

  When a write voltage (positive bias voltage) Vw “1” is applied to such a magnetoresistive element, the magnetization states of the first and second units U1 and U2 are antiparallel, and data “1” is stored. Written.

  When a write voltage (negative bias voltage) Vw “0” is applied to the magnetoresistive effect element, the magnetization states of the first and second units U1 and U2 become parallel and data “0” is written. .

  When the read voltage Vr is biased, the first and second units U1, U2 are both in the normal state, so that the data “0” is in the low resistance state and the data “1” is in the high resistance state.

(14) Other
In the first to thirteenth embodiments, when the first unit U1 is on the lower side and the second unit U2 is on the upper side, the first unit U1 is formed on a base layer made of a conductor. A cap layer made of a conductor is formed on the second unit.

  The read voltage Vr is a positive bias voltage (high potential on the second unit U2 side) in this example, but may instead be a negative bias voltage (high potential on the first unit U1 side).

6). Application examples
The example of the present invention is intended to achieve both the reduction of the write current and the improvement of the MR ratio at the time of reading in the magnetoresistive effect element having the dual pin structure. In the device, it is possible to realize a new technique in which binary data “0” and “1” can be written only by changing the magnitude of the write current (spin injection current) without changing the direction.

  According to this technique, the write current only needs to flow in one direction regardless of the value of the write data. Therefore, the driver / sinker as a peripheral circuit is simplified, and the chip size can be reduced.

  In this application example, a magnetoresistive effect element having a single pin structure is targeted, and the bias voltage is defined as follows.

  A bias that sets the free layer 12 at a high potential, a pinned layer 11 at a low potential is a positive bias voltage, a pinned layer 11 is at a high potential, and a bias that sets the free layer 12 at a low potential is a negative bias voltage.

(1) First application example
FIG. 30 shows a first application example.

  The magnetoresistive element U includes a pinned layer 11, a free layer 12, and a spacer layer 13 disposed therebetween. The spacer layer 13 is a tunnel barrier layer made of an insulator or a semiconductor.

  The pinned layer 11 and the free layer 12 are composed of a single layer or a plurality of layers of ferromagnetic materials. Further, the pinned layer 11 and the free layer 12 may have a ferromagnetic layer / nonmagnetic layer / ferromagnetic layer structure represented by the SAF structure.

  The bias dependency of the MR ratio of the magnetoresistive element U is that it exhibits normal TMR characteristics in a low bias state and inverse TMR characteristics in a high bias state.

  The point Vc1 at which the characteristic is switched is between the two write voltages Vw “0” and Vw “1”.

  The magnitude relationship of the voltages including the read voltage Vr is 0 <Vr <Vw “0” <Vc1 <Vw “1”. As a matter of course, Vw “1” is a value smaller than the breakdown voltage Vb of the magnetoresistive effect element U.

  In such a magnetoresistive effect element U, as shown in FIG. 31, when a write voltage (positive bias voltage) Vw “0” is applied, the magnetoresistive effect element U enters a normal state, and the magnetization of the pinned layer 11 Electrons that are spin-polarized in the same direction as the direction pass through the pinned layer 11 and give a spin injection torque to the free layer 12.

  As a result, the magnetization state of the magnetoresistive effect element U becomes parallel and data “0” is written.

  Further, as shown in FIG. 32, when a write voltage (positive bias voltage) Vw “1” is applied, the magnetoresistive effect element U is in an inverse state and spin-polarized in the direction opposite to the magnetization direction of the pinned layer 11. The passed electrons pass through the pinned layer 11 and give a spin injection torque to the free layer 12.

  As a result, the magnetization state of the magnetoresistive effect element U becomes antiparallel, and data “1” is written.

  Further, as shown in FIGS. 31 and 32, when the read voltage (positive bias voltage) Vr is biased, the magnetoresistive element U is in the normal state, so that the data “0” is the low resistance state, the data “ 1 ″ is in a high resistance state.

  Note that the read voltage Vr may be a negative bias voltage.

(2) Second application example
FIG. 33 shows a second application example.

The second application example is a modification of the first application example.
The second application example differs from the first application example in that the read voltage Vr and the write voltages Vw “0” and Vw “1” are all negative bias voltages.

  The point Vc2 at which the characteristic is switched is between the two write voltages Vw “1” and Vw “0”.

  The magnitude relationship between voltages including the read voltage Vr is 0 <| Vr | <| Vw "1" | <| Vc2 | <| Vw "0" |. Naturally, | Vw “0” | is a value smaller than the breakdown voltage | Vb | of the magnetoresistive element U.

  In such a magnetoresistive effect element U, as shown in FIG. 34, when a write voltage (negative bias voltage) Vw “1” is applied, the magnetoresistive effect element U enters a normal state, and the magnetization of the pinned layer 11 Electrons spin-polarized in the direction opposite to the direction are reflected by the pinned layer 11 and give a spin injection torque to the free layer 12.

  As a result, the magnetization state of the magnetoresistive effect element U becomes antiparallel, and data “1” is written.

  Further, as shown in FIG. 35, when a write voltage (negative bias voltage) Vw “0” is applied, the magnetoresistive effect element U is in an inverse state and spin polarized in the same direction as the magnetization direction of the pinned layer 11. The reflected electrons are reflected by the pinned layer 11 and give a spin injection torque to the free layer 12.

  As a result, the magnetization state of the magnetoresistive effect element U becomes parallel and data “0” is written.

  Further, as shown in FIGS. 34 and 35, when the read voltage (negative bias voltage) Vr is biased, the magnetoresistive effect element U is in the normal state, so that the data “0” is the low resistance state, the data “ 1 ″ is in a high resistance state.

  Note that the read voltage Vr may be a positive bias voltage.

(3) Third application example
FIG. 36 shows a third application example.

  The magnetoresistive element U includes a pinned layer 11, a free layer 12, and a spacer layer 13 disposed therebetween. The spacer layer 13 is a tunnel barrier layer made of an insulator or a semiconductor.

  The pinned layer 11 and the free layer 12 are composed of a single layer or a plurality of layers of ferromagnetic materials. Further, the pinned layer 11 and the free layer 12 may have a ferromagnetic layer / nonmagnetic layer / ferromagnetic layer structure represented by the SAF structure.

  The bias dependence of the MR ratio of the magnetoresistive effect element U indicates that it exhibits inverse TMR characteristics in a low bias state and normal TMR characteristics in a high bias state.

  The point Vc2 at which the characteristic is switched is between the two write voltages Vw “1” and Vw “0”.

  The magnitude relationship between voltages including the read voltage Vr is 0 <| Vr | <| Vw "1" | <| Vc2 | <| Vw "0" |. Naturally, | Vw “0” | is a value smaller than the breakdown voltage | Vb | of the magnetoresistive element U.

  In such a magnetoresistive effect element U, as shown in FIG. 37, when a write voltage (negative bias voltage) Vw “1” is applied, the magnetoresistive effect element U enters an inverse state, and the magnetization of the pinned layer 11 Electrons that are spin-polarized in the same direction as the direction are reflected by the pinned layer 11 and give a spin injection torque to the free layer 12.

  As a result, the magnetization state of the magnetoresistive effect element U becomes parallel and data “1” is written.

  As shown in FIG. 38, when a write voltage (negative bias voltage) Vw “0” is applied, the magnetoresistive effect element U is in a normal state and spin-polarized in the direction opposite to the magnetization direction of the pinned layer 11. The reflected electrons are reflected by the pinned layer 11 and give a spin injection torque to the free layer 12.

  As a result, the magnetization state of the magnetoresistive effect element U becomes antiparallel, and data “0” is written.

  Further, as shown in FIGS. 37 and 38, when the read voltage (negative bias voltage) Vr is biased, the magnetoresistive element U is in an inverse state, so that data “0” is in a low resistance state, data “ 1 ″ is in a high resistance state.

  Note that the read voltage Vr may be a positive bias voltage.

(4) Fourth application example
FIG. 39 shows a fourth application example.

The fourth application example is a modification of the third application example.
The fourth application example is different from the third application example in that the read voltage Vr and the write voltages Vw “0” and Vw “1” are all positive bias voltages.

  The point Vc1 at which the characteristic is switched is between the two write voltages Vw “0” and Vw “1”.

  The magnitude relationship of the voltages including the read voltage Vr is 0 <Vr <Vw “0” <Vc1 <Vw “1”. As a matter of course, Vw “1” is a value smaller than the breakdown voltage Vb of the magnetoresistive effect element U.

  In such a magnetoresistive effect element U, as shown in FIG. 40, when a write voltage (positive bias voltage) Vw “0” is applied, the magnetoresistive effect element U enters an inverse state, and the magnetization of the pinned layer 11 Electrons that are spin-polarized in the direction opposite to the direction pass through the pinned layer 11 and give a spin injection torque to the free layer 12.

  As a result, the magnetization state of the magnetoresistive effect element U becomes antiparallel, and data “0” is written.

  As shown in FIG. 41, when a write voltage (positive bias voltage) Vw “1” is applied, the magnetoresistive effect element U is in a normal state and spin-polarized in the same direction as the magnetization direction of the pinned layer 11. The passed electrons pass through the pinned layer 11 and give a spin injection torque to the free layer 12.

  As a result, the magnetization state of the magnetoresistive effect element U becomes parallel and data “1” is written.

  Further, as shown in FIGS. 40 and 41, when the read voltage (positive bias voltage) Vr is biased, the magnetoresistive effect element U is in the inverse state, so that the data “0” is the low resistance state, the data “ 1 ″ is in a high resistance state.

  Note that the read voltage Vr may be a negative bias voltage.

(5) Summary
Thus, according to the magnetoresistive effect element having a single pin structure as an application example, binary data can be written to the magnetoresistive effect element without changing the direction of the write current.

7). Application examples
According to the dual pin type magnetoresistive effect element relating to the example of the present invention, it is possible to simultaneously realize the reduction of the write current and the improvement of the MR ratio at the time of reading. Also, according to the single pin type magnetoresistive effect element to which the example of the present invention is applied, it is possible to realize binary data writing without changing the direction of the write current.

  Therefore, by applying these magnetoresistance effect elements to a new device such as a magnetic random access memory or a spin FET, it is possible to contribute to the practical use of the new device.

(1) 1Tr-1MTJ type magnetic random access memory
A. First example
FIG. 42 shows a first example of the 1Tr-1MTJ type magnetic random access memory.

  The memory cell array 21 is composed of a plurality of memory cells. Each memory cell includes a magnetoresistive element MTJ and a select transistor (MISFET) ST connected in series with each other.

  As the magnetoresistive effect element MTJ, a dual pin type magnetoresistive effect element according to an example of the present invention is used.

  The word lines WL1, WL2,... Extend in the row direction, and one end thereof is connected to the row decoder / word line driver 22. Further, the word lines WL1, WL2,... Are connected to the gate electrode of the selection transistor ST.

  Lower bit lines BLd1, BLd2, BLd3,... And upper bit lines BLu1, BLu2, BLu3,... Extend in the column direction intersecting the row direction.

  One end of each of the lower bit lines BLd1, BLd2, BLd3,... Is connected to a column decoder / bit line driver / sinker 23A. Further, the lower bit lines BLd1, BLd2, BLd3,... Are connected to one of the two source / drain diffusion layers of the selection transistor ST.

  One end of each of the upper bit lines BLu1, BLu2, BLu3,... Is connected to a column decoder / bit line driver / sinker sense amplifier 24A. The upper bit lines BLu1, BLu2, BLu3,... Are connected to one end of the magnetoresistive element MTJ. The other end of the magnetoresistive element MTJ is connected to the other of the two source / drain diffusion layers of the select transistor ST.

  In such a magnetic random access memory, when data is written to the memory cell MC1, the row decoder / word line driver 22 activates the word line WL1, for example, “H”, and turns on the selection transistor ST in the memory cell MC1. To.

  Further, the lower bit line BLd1 is selected by the column decoder 23A, and the upper bit line BLu1 is selected by the column decoder 24A.

  When writing the first data (for example, “0”), the write current Iw is supplied from the bit line driver / sinker 23A toward the bit line driver / sinker 24A.

  When writing the second data (for example, “1”), the write current Iw is supplied from the bit line driver / sinker 24A to the bit line driver / sinker 23A.

  When reading data from the memory cell MC1, the row decoder / word line driver 22 activates the word line WL1, for example, “H”, and turns on the selection transistor ST in the memory cell MC1.

  Further, the lower bit line BLd1 is selected by the column decoder 23A, and the upper bit line BLu1 is selected by the column decoder 24A.

  Then, a read current Ir flows from the sense amplifier 24A toward the bit line driver / sinker 23A.

  Note that the read current Ir is made smaller than the write current Iw.

  Here, since the dual pin type magnetoresistive effect element according to the example of the present invention is used for the magnetoresistive effect element MTJ, it is possible to simultaneously reduce the write current and improve the MR ratio at the time of read.

B. Second example
FIG. 43 shows a second example of the 1Tr-1MTJ type magnetic random access memory.

  The second example is different from the first example in that a single pin type magnetoresistive effect element according to an application example of the present invention is used as the magnetoresistive effect element MTJ.

  As a result, since binary data can be written without changing the direction of the write current Iw, a column decoder / bit line sinker 23B is provided at one end of the lower bit lines BLd1, BLd2, BLd3,. A column decoder / bit line driver / sense amplifier 24B is connected to one end of the upper bit lines BLu1, BLu2, BLu3,.

  In such a magnetic random access memory, when data is written in the memory cell MC1, the row decoder / word line driver 22 activates the word line WL1, for example, “H”, and turns on the selection transistor ST in the memory cell MC1. To.

  Further, the lower bit line BLd1 is selected by the column decoder 23B, and the upper bit line BLu1 is selected by the column decoder 24B.

  When writing the first data (for example, “0”), a write current Iw having a first value is supplied from the bit line driver 24B toward the bit line sinker 23B.

  When writing the second data (for example, “1”), a write current Iw having a second value different from the first value is supplied from the bit line driver 24B toward the bit line sinker 23B.

  When reading data from the memory cell MC1, the row decoder / word line driver 22 activates the word line WL1, for example, “H”, and turns on the selection transistor ST in the memory cell MC1.

  Further, the lower bit line BLd1 is selected by the column decoder 23B, and the upper bit line BLu1 is selected by the column decoder 24B.

  Then, a read current Ir flows from the sense amplifier 24B toward the bit line sinker 23B.

  Note that the read current Ir is made smaller than the write current (first value and second value) Iw.

C. Device structure
FIG. 44 shows an example of the device structure of the memory cell MC1 shown in FIGS.

  An element isolation insulating layer 32 having an STI (shallow trench isolation) structure is formed in a semiconductor substrate 31 such as silicon. In the element region surrounded by the element isolation insulating layer 32, a selection transistor ST including a source / drain diffusion layer 33, a gate insulating layer 34, and a gate electrode (word line WL1) 35 is formed.

  One of the source / drain diffusion layers 33 is connected to the lower bit line BLd1 through the contact plug 36. The other one of the source / drain diffusion layers 33 is connected to the lower electrode 40 through a contact plug 37, an intermediate conductive layer 38 and a via plug 39.

  A magnetoresistive element MTJ is formed on the lower electrode 40. A cap layer (hard mask) 41 made of a conductor is formed on the magnetoresistive element MTJ. On the cap layer 41, the upper bit line BLu1 is formed.

  The lower bit line BLd1, the upper bit line BLu1, the contact plugs 36, 37, the intermediate conductive layer 38, the via plug 39, the lower electrode 40, and the cap layer 41 are made of a conductor such as W, Al, Cu, AlCu.

  When using Cu as these materials, it is desirable to employ a damascene process or a dual damascene process.

  FIG. 45 shows another example of the device structure of the memory cell MC1 shown in FIGS.

  The feature of this example is that the magnetoresistive effect element MTJ is directly disposed on the via plug 39. That is, when the lower electrode 40 in the device structure of FIG. 44 is removed, the device structure of FIG. 45 is obtained.

  In the case of this structure, the size of the magnetoresistive element MTJ is preferably smaller than the size of the via plug 39. Specifically, the layout is such that the magnetoresistive element MTJ is accommodated in the via plug 39 when viewed from above the semiconductor substrate 31 (in plan view).

  However, the present invention is not limited to this, and the magnetoresistive element MTJ may slightly protrude from the via plug 39 in plan view. Further, the size of the magnetoresistive effect element MTJ may be larger than the size of the via plug 39.

Here, if the minimum processing dimension determined by lithography or etching technique is defined as F (Minimum Feature Size), the minimum cell size when the structure of FIG. 44 is used is 8F ² . On the other hand, the minimum cell size when the structure of FIG. 45 is used is 4F ² .

  In the write method in which the magnetization reversal of the free layer is performed by a magnetic field, the value of the write current Iw increases as the cell size is reduced. Therefore, it is desirable to increase the size of the magnetoresistive element MTJ as much as possible. Therefore, the device structure is necessarily as shown in FIG.

  On the other hand, when the spin injection writing method is employed, the value of the write current Iw decreases as the cell size decreases, and therefore it is desirable to reduce the size of the magnetoresistive element MTJ as much as possible. Therefore, it is easy to adopt the structure of FIG. 45 as the device structure.

D. Summary
According to the 1Tr-1MTJ magnetic random access memory to which the magnetoresistive effect element according to the example of the present invention is applied, it is possible to obtain an accompanying effect of improving the writing speed as the magnetization reversal efficiency is improved. Specifically, the writing speed can realize a value within a range from several nanoseconds to several microseconds.

  It is desirable that the read current Ir supplied to the magnetoresistive element MTJ during reading has a shorter pulse width (supply time) than the write current Iw supplied to the magnetoresistive element MTJ during writing.

  In this case, disturb (erroneous writing) during reading can be prevented. This is because when the pulse width of the write current Iw is shortened, the absolute value of the value of the write current becomes too large.

  That is, the value of the write current Iw is reduced to reduce the current consumption, and the pulse width of the read current Ir is made shorter than the pulse width of the write current Iw in order to prevent disturbance during reading.

(2) Cross-point magnetic random access memory
A. First example
FIG. 46 shows a first example of a cross-point type magnetic random access memory.

  The memory cell array 51 is composed of a plurality of memory cells. Each memory cell includes a magnetoresistive element MTJ. As the magnetoresistive effect element MTJ, a dual pin type magnetoresistive effect element according to an example of the present invention is used.

  The word lines WL1, WL2, WL3,... Extend in the row direction, and one end thereof is connected to the row decoder / word line driver / sinker 52A. The word lines WL1, WL2, WL3,... Are connected to one end of the magnetoresistive element MTJ.

  Bit lines BL1, BL2,... Extend in the column direction intersecting the row direction. One end of each of the bit lines BL1, BL2,... Is connected to a column decoder / bit line driver / sinker 53A. The bit lines BL1, BL2,... Are connected to the other end of the magnetoresistive element MTJ.

  The other ends of the bit lines BL1, BL2,... Are connected to a sense amplifier (S / A) 54 via a column selection switch (MISFET) CSW. The gate electrode of the column selection switch CSW is connected to the column decoder 55.

  In such a magnetic random access memory, when data is written to the memory cell MC1, the word line WL1 is selected by the row decoder 52A. Further, the bit line BL1 is selected by the column decoder 53A.

  When writing the first data (for example, “0”), the write current Iw is supplied from the word line driver / sinker 52A to the bit line driver / sinker 53A.

  When writing the second data (for example, “1”), the write current Iw is supplied from the bit line driver / sinker 53A toward the word line driver / sinker 52A.

  When reading data from the memory cell MC1, the word line WL1 is selected by the row decoder 52A, and the bit line BL1 is selected by the column decoder 55. Then, a read current Ir flows from the sense amplifier 54 toward the word line driver / sinker 52A.

  Note that the read current Ir is made smaller than the write current Iw.

  Here, since the dual pin type magnetoresistive effect element according to the example of the present invention is used for the magnetoresistive effect element MTJ, it is possible to simultaneously reduce the write current and improve the MR ratio at the time of read.

B. Second example
FIG. 47 shows a second example of the cross-point type magnetic random access memory.

  The second example is different from the first example in that a single pin type magnetoresistive effect element according to an application example of the present invention is used as the magnetoresistive effect element MTJ.

  As a result, since binary data can be written without changing the direction of the write current Iw, a row decoder / word line sinker 52B is connected to one end of the word lines WL1, WL2, WL3,. The column decoder / bit line driver 53B is connected to one end of the bit lines BL1, BL2,.

  Further, since the direction of the write current Iw for “0” / “1” write is the same, a complete cross-point type spin injection magnetic random access memory using the rectifier diode D can be realized.

  Here, the term “complete” is derived from the fact that it has been impossible in the past to combine the spin injection writing method with the cross-point type.

  The reason for this is that, in the case of the cross-point type, means for adding a rectifier diode to the magnetoresistive element MTJ is employed in order to prevent a sneak current at the time of reading. This is because it becomes impossible to change the direction of the write current Iw according to the write data.

  According to the application example of the present invention, since the direction of the write current Iw is constant regardless of the write data, a combination of the spin injection write method and the cross point type becomes possible.

  Further, in the case of the cross-point type, it is not necessary to connect a selection transistor to each magnetoresistive element MTJ, so that the problem that the magnitude of the write current Iw is limited to the size of the selection transistor does not occur.

  As a result, in some cases, a write current of 1 mA or more can be supplied to the magnetoresistive element MTJ.

As for the cell size, since there is no selection transistor, 4F ² which is the minimum size when the width of each of the word line and the bit line is F can be easily realized.

  In such a magnetic random access memory, when data is written to the memory cell MC1, the word line WL1 is selected by the row decoder 52B. Further, the bit line BL1 is selected by the column decoder 53B.

  When writing the first data (for example, “0”), a write current Iw having a first value is supplied from the bit line driver 53B toward the word line sinker 52B.

  When writing the second data (for example, “1”), a write current Iw having a second value different from the first value is supplied from the bit line driver 53B toward the word line sinker 52B.

  When reading data from the memory cell MC1, the word line WL1 is selected by the row decoder 52B, and the bit line BL1 is selected by the column decoder 55. Then, a read current Ir flows from the sense amplifier 54 toward the word line sinker 52B.

  Note that the read current Ir is made smaller than the write current (first value and second value) Iw.

C. Device structure
FIG. 48 shows an example of the device structure of the memory cell MC1 of FIG.

  A word line WL1 is formed on an upper portion of a semiconductor substrate 61 such as silicon. A magnetoresistive element MTJ is formed on the word line WL1. A cap layer (hard mask) 62 made of a conductor is formed on the magnetoresistive element MTJ. On the cap layer 62, the bit line BL1 is formed.

  The word line WL1, the bit line BL1, and the cap layer 62 are made of a conductor such as W, Al, Cu, and AlCu.

  When using Cu as these materials, it is desirable to employ a damascene process or a dual damascene process.

  FIG. 49 shows an example of the device structure of the memory cell MC1 of FIG.

  A feature of this example is that a rectifier diode D is formed between the word line WL1 and the magnetoresistive element MTJ. That is, when the rectifier diode D is added to the device structure of FIG. 48, the device structure of FIG. 49 is obtained.

  Note that the rectifier diode D may be formed between the magnetoresistive element MTJ and the cap layer 62.

D. Summary
According to the cross-point type magnetic random access memory to which the magnetoresistive effect element according to the example of the present invention is applied, an accompanying effect of improving the writing speed can be obtained as the magnetization reversal efficiency is improved. Specifically, the writing speed can realize a value within a range from several nanoseconds to several microseconds.

  It is desirable that the read current Ir supplied to the magnetoresistive element MTJ during reading has a shorter pulse width (supply time) than the write current Iw supplied to the magnetoresistive element MTJ during writing.

  In this case, disturb (erroneous writing) during reading can be prevented. This is because when the pulse width of the write current Iw is shortened, the absolute value of the value of the write current becomes too large.

  That is, the value of the write current Iw is reduced to reduce the current consumption, and the pulse width of the read current Ir is made shorter than the pulse width of the write current Iw in order to prevent disturbance during reading.

8). Conclusion
According to the example of the present invention, in a magnetoresistive effect element having a dual pin structure, it is possible to achieve both reduction in write current and improvement in MR ratio at the time of reading.

  The example of the present invention is not limited to the above-described embodiment, and can be embodied by modifying each component without departing from the scope of the invention. Various inventions can be configured by appropriately combining a plurality of constituent elements disclosed in the above-described embodiments. For example, some constituent elements may be deleted from all the constituent elements disclosed in the above-described embodiments, or constituent elements of different embodiments may be appropriately combined.

Sectional drawing which shows the example of a dual pin structure. Sectional drawing which shows the example of a dual pin structure. Sectional drawing which shows the example of a pinned layer. Sectional drawing which shows the example of a pinned layer. Sectional drawing which shows the example of a free layer. Sectional drawing which shows the example of a free layer. The figure which shows the 1st example of the magnetoresistive effect element with which the magnetization direction of two pinned layers is reverse. The figure which shows the 1st example of the magnetoresistive effect element with which the magnetization direction of two pinned layers is reverse. The figure which shows the 2nd example of the magnetoresistive effect element in which the magnetization direction of two pinned layers is reverse. The figure which shows the 2nd example of the magnetoresistive effect element in which the magnetization direction of two pinned layers is reverse. The figure which shows the 1st example of the magnetoresistive effect element in which the magnetization direction of two pinned layers is the same. The figure which shows the 1st example of the magnetoresistive effect element in which the magnetization direction of two pinned layers is the same. The figure which shows the 2nd example of the magnetoresistive effect element in which the magnetization direction of two pinned layers is the same. The figure which shows the 2nd example of the magnetoresistive effect element in which the magnetization direction of two pinned layers is the same. The figure which shows the example of the control method of the bias dependence of MR ratio. The figure which shows the example of the control method of the bias dependence of MR ratio. The figure which shows 1st Example. The figure which shows 2nd Example. The figure which shows 3rd Example. The figure which shows 4th Example. The figure which shows 5th Example. The figure which shows 6th Example. The figure which shows 7th Example. The figure which shows 8th Example. The figure which shows 9th Example. The figure which shows 10th Example. The figure which shows 11th Example. The figure which shows 12th Example. The figure which shows 13th Example. The figure which shows a 1st application example. The figure which shows a 1st application example. The figure which shows a 1st application example. The figure which shows a 2nd application example. The figure which shows a 2nd application example. The figure which shows a 2nd application example. The figure which shows the 3rd application example. The figure which shows the 3rd application example. The figure which shows the 3rd application example. The figure which shows a 4th application example. The figure which shows a 4th application example. The figure which shows a 4th application example. The figure which shows the 1st example of 1Tr-1MTJ type | mold magnetic random access memory. The figure which shows the 2nd example of 1Tr-1MTJ type | mold magnetic random access memory. Sectional drawing which shows the structural example of a memory cell. Sectional drawing which shows the structural example of a memory cell. The figure which shows the 1st example of a crosspoint type | mold magnetic random access memory. The figure which shows the 2nd example of a crosspoint type | mold magnetic random access memory. Sectional drawing which shows the structural example of a memory cell. Sectional drawing which shows the structural example of a memory cell.

Explanation of symbols

  11A, 11B: Pinned layer (ferromagnetic layer), 12: Free layer (ferromagnetic layer), 13A, 13B: Spacer layer (nonmagnetic layer), 14: Antiferromagnetic layer, 21, 51: Memory cell array, 22: Row decoder / word line driver, 23A: Column decoder / bit line driver / sinker, 23B: Column decoder / bit line sinker, 24A: Column decoder / bit line driver / sinker sense amplifier, 24B: Column decoder / bit line driver / Sense amplifiers 31, 61: Semiconductor substrate, 32: Element isolation insulating layer, 33: Source / drain diffusion layer, 34: Gate insulating layer, 35: Gate electrode, 36, 37: Contact plug, 38: Intermediate conductive layer, 39 : Via plug, 40: Lower electrode, 41, 62: Cap layer, 52A: Row decoder / word line driver / sinker, 52B: Row decoder / word line sinker, 53A: Column decoder / bit line driver / sinker, 53B: Column decoder / bit line driver, 54: Sense amplifier, 55: Column decoder, MTJ: magnetoresistive effect element, ST: selection transistor, D: rectifier diode.

Claims (20)

The first and second ferromagnetic layers whose magnetization directions are fixed opposite to each other; the third ferromagnetic layer disposed between the first and second ferromagnetic layers, the magnetization direction of which varies; and the first And a first nonmagnetic layer disposed between the third ferromagnetic layer and a second nonmagnetic layer disposed between the second and third ferromagnetic layers,
The MR ratio of the first unit comprising the first ferromagnetic layer, the first nonmagnetic layer, and the third ferromagnetic layer when a read voltage is biased between the first and second ferromagnetic layers. Is in an inverse state, and the MR ratio of the second unit composed of the second ferromagnetic layer, the second nonmagnetic layer, and the third ferromagnetic layer is in a normal state,
The MR ratio of the first and second units is in the normal state or the inverse state when a write voltage is biased between the first and second ferromagnetic layers. Effect element. The first and second ferromagnetic layers whose magnetization directions are fixed in the same direction, the third ferromagnetic layer disposed between the first and second ferromagnetic layers and changing the magnetization direction, and the first And a first nonmagnetic layer disposed between the third ferromagnetic layer and a second nonmagnetic layer disposed between the second and third ferromagnetic layers,
The MR ratio of the first unit comprising the first ferromagnetic layer, the first nonmagnetic layer, and the third ferromagnetic layer when a read voltage is biased between the first and second ferromagnetic layers. And the MR ratio of the second unit composed of the second ferromagnetic layer, the second nonmagnetic layer, and the third ferromagnetic layer are both in a normal state or an inverse state,
When a write voltage is biased between the first and second ferromagnetic layers, the MR ratio of the first unit is in the inverse state, and the MR ratio of the second unit is in the normal state. A magnetoresistive effect element.   The MR ratio of the first and second units has a bias dependency, and one of the MR ratios of the first and second units is a region between the read voltage and the write voltage. The magnetoresistive effect element according to claim 1, wherein the magnetoresistive effect element changes from one of the normal state and the inverse state to the other.   4. The magnetoresistive element according to claim 3, wherein the first and second nonmagnetic layers are made of an insulator or a semiconductor and generate a tunnel magnetoresistive effect.   One of the MR ratios of the first and second units has a bias dependency, and one of the MR ratios of the first and second units having the bias dependency is the read voltage and the write voltage. 3. The magnetoresistive effect element according to claim 1, wherein the magnetoresistive effect element changes from one of the normal state and the inverse state to the other in a region between the first and second states.   6. The magnetoresistive effect according to claim 5, wherein when the first unit has the bias dependence, the first nonmagnetic layer is made of an insulator or a semiconductor and generates a tunnel magnetoresistive effect. element.   6. The magnetoresistive effect according to claim 5, wherein when the second unit has the bias dependency, the second nonmagnetic layer is made of an insulator or a semiconductor and generates a tunnel magnetoresistive effect. element.   The magnetoresistive effect element according to any one of claims 1 to 7, wherein the magnetization directions of the first, second, and third ferromagnetic layers are directions in which they are stacked.   8. The magnetism according to claim 1, wherein the magnetization directions of the first, second and third ferromagnetic layers are perpendicular to a direction in which they are stacked. Resistive effect element.   The magnetoresistive effect element according to claim 1, wherein the direction of the write voltage changes according to a value of write data.   11. A magnetic random access memory comprising the magnetoresistive effect element according to claim 1 as a memory cell. A first ferromagnetic layer whose magnetization direction is fixed, a second ferromagnetic layer whose magnetization direction changes, and a nonmagnetic layer disposed between the first and second ferromagnetic layers,
The MR ratio when the first write voltage is biased between the first and second ferromagnetic layers to write the first data is in a normal state;
A magnetoresistive effect element, wherein an MR ratio when a second write voltage is biased between the first and second ferromagnetic layers to write second data is in an inverse state.   The magnetoresistive effect element according to claim 12, wherein directions of the first and second write voltages are the same.   14. The magnetoresistive element according to claim 12, wherein the magnitudes of the first and second write voltages are different.   15. The MR ratio when a read voltage is biased between the first and second ferromagnetic layers is in the normal state or the inverse state. 15. Magnetoresistive effect element.   13. The MR ratio has a bias dependency and changes from one of the normal state and the inverse state to the other in a region between the first and second write voltages. The magnetoresistive effect element of any one of thru | or 15.   The magnetoresistive effect element according to claim 16, wherein the nonmagnetic layer is made of an insulator or a semiconductor and generates a tunnel magnetoresistive effect.   The magnetoresistive effect element according to any one of claims 12 to 17, wherein the magnetization directions of the first and second ferromagnetic layers are directions in which they are stacked.   18. The magnetoresistive element according to claim 12, wherein the magnetization directions of the first and second ferromagnetic layers are perpendicular to a direction in which the first and second ferromagnetic layers are stacked. .   20. A magnetic random access memory comprising the magnetoresistive effect element according to claim 12 as a memory cell.

Images