JP2018055752A - Method of writing magnetic tunnel junction memory element and semiconductor memory - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide a method of writing a magnetic tunnel junction memory element writable in low power consumption without applying external magnetic field.SOLUTION: A method of writing a magnetic tunnel junction memory element includes each step of: writing the magnetic tunnel junction memory element in non-parallel state by applying a voltage to a perpendicular magnetization magnetic tunnel junction memory element, in which a magnetization fixed layer and a magnetization free layer are included, within only a first period so as to flow a current in a direction from the magnetization fixed layer to the magnetization free layer; and writing the magnetic tunnel junction memory element in parallel state by applying the voltage within only a second period shorter than the first period to the magnetic tunnel junction memory element so as to flow the current in the direction.SELECTED DRAWING: Figure 2

Description

  The present disclosure relates to a magnetic tunnel junction memory element writing method and a semiconductor memory device.

  Spin injection type magnetoresistive random access memory (hereinafter referred to as MRAM) has advantages in that the size of a magnetic tunnel junction (hereinafter referred to as MTJ) can be reduced and the switching current can be reduced. However, there is a problem that there is a trade-off between switching current reduction and thermal fluctuation resistance. Therefore, recently, the object of development has shifted from the MTJ of the in-plane magnetization method to the MTJ of the perpendicular magnetization method, which is advantageous for high density and has high thermal fluctuation resistance, and trial manufacture of a fine MTJ of 20 nm or less has been attempted. .

  It has been reported that when a voltage (electric field) is applied to a ferromagnetic layer via an insulator, the magnetic anisotropy of the ferromagnetic layer changes, and the voltage dependence of the magnetic anisotropy is reported. The voltage dependence of the magnetic anisotropy often changes linearly with respect to the voltage (electric field). In the case of a CoFeB / MgO / CoFeB structure often used in a spin injection type MRAM, when a positive voltage (upper is positive and lower is negative) is applied to CoFeB located under MgO, the holding force decreases, and the negative voltage When (the upper is negative and the lower is positive) is applied, the holding force increases. Further, in CoFeB positioned on MgO, holding force increases when a positive voltage (upward is positive and lower is negative) is applied, and holding force is reduced when a negative voltage (upward is negative and lower is positive) is applied.

  A non-volatile memory device has been proposed that uses the voltage dependence of magnetic anisotropy to change the coercive force of the magnetization free layer of the MTJ and realizes zero writing and one writing by applying a voltage in a single direction. (For example, see Patent Documents 1 and 2 and Non-Patent Documents 1 and 2). At this time, since the coercive force of the magnetization free layer of the MTJ is reduced by utilizing the voltage dependence of the magnetic anisotropy, it is expected to realize writing with low power consumption with a small current.

  Specifically, the magnetic anisotropy of the magnetization free layer is reduced by applying a certain voltage V1 to the MTJ, and magnetization reversal is performed by spin injection of a small current flowing in that state. As a result, “1” is written. Further, when the magnetic anisotropy of the magnetization free layer is further reduced by applying the voltage V2 (> V1) having the same polarity as the voltage V1, only the "0" state exists under a predetermined external magnetic field (bias magnetic field). Can create conditions that do not. As a result, “1” is written.

  In the above writing operation, it is necessary to apply an external magnetic field. However, in order to realize a high-density and low-cost MRAM, it is not preferable to add a configuration in which an external magnetic field is applied even though magnetic field inversion can be realized by spin injection.

  In addition, the MTJ used in Non-Patent Document 1 that has succeeded in the above writing operation has a relatively large size, and is not a size of, for example, 30 nm or less that is expected to be put to practical use. There is a report that the above write operation could not be reproduced with a vertical Top-pinned type element having a size of 65 nm (see Non-Patent Document 3, for example). Further, the MTJ that has succeeded in the writing operation used in Non-Patent Document 1 has a very high MgO layer thickness of 1.4 nm and high resistance. In the above-described “1” write operation, since the magnetization is reversed by spin injection by passing a current, it is not preferable to make the MTJ resistance value too high. In addition, since the element resistance increases in inverse proportion to the element area in the MTJ, the resistance value becomes a problem in the MTJ having a size expected to be put to practical use.

  Although the resistance value can be reduced by reducing the thickness of the MgO layer, a phenomenon in which the RH loop shifts due to the influence of current is observed when the thickness of the MgO layer is reduced. Moreover, this RH loop shift has a problem that it shifts in a direction that prevents the above-described "0" writing, that is, a direction that requires a larger external magnetic field.

US Patent Application Publication No. 2013/0015542 International Publication No. WO2012 / 159078 Pamphlet

Wel-Gang Wang and three others, "Electric-field-assisted switching in magnetic tunnel junctions," Nature Materials, Vol. 11, January 2012, pp 64-68 J. G. Alzate, 12 others, "Voltage-Induced Switching of Nanoscale Magnetic Tunnel Junctions," IEEE International Electron Devices Meeting (IEDM 2012), pp 29.5.1-29.5.4 Hao Meng, 3 others, "Electric field control of spin re-orientation in perpendicular magnetic tunnel junctions- CoVeB and MgO thickness dependence," Applied Physics Letters 105, 042410 (2014), pp 042410-1-042410-5

  In view of the above, it is desired to provide a writing method for a magnetic tunnel junction memory element that can be written with low power consumption without applying an external magnetic field.

  The magnetic tunnel junction storage element has a first writing method in which a current flows in the direction from the magnetization fixed layer to the magnetization free layer in the perpendicular magnetization type magnetic tunnel junction storage element including the magnetization fixed layer and the magnetization free layer. The voltage is applied only for a period of time to write the magnetic tunnel junction memory element in an antiparallel state, and the voltage is applied to the magnetic tunnel junction memory element for a second period shorter than the first period so that a current flows in the direction. To write the magnetic tunnel junction storage element in a parallel state.

  According to at least one embodiment, it is possible to provide a method for writing a magnetic tunnel junction memory element that can be written with low power consumption without applying an external magnetic field.

It is a figure which shows an example of a fundamental structure of a magnetic tunnel junction memory element. It is a figure which shows an example of the voltage waveform applied in a 1st period, and the voltage waveform applied in a 2nd period. It is a figure which shows an example of a structure of the 1st Example of an MTJ memory element. It is a figure which shows the relationship between the voltage which the magnetization free layer shown in FIG. 3 shows, and an anisotropic magnetic field. It is a figure which shows MH loop of the magnetization free layer shown in FIG. It is a figure which shows the waveform of the voltage pulse applied in a write-in operation | movement with respect to the MTJ memory element shown in FIG. FIG. 4 is a diagram showing a change in normalized TMR when a write operation is performed on the MTJ memory element shown in FIG. 3. FIG. 4 is a diagram showing in detail a change in normalized TMR when a voltage of a voltage value V2 is applied to the MTJ memory element shown in FIG. It is a figure which shows an example of the locus | trajectory of the magnetization direction at the time of the magnetization direction of the magnetization free layer shown in FIG. 3 changing. It is a figure which shows the change of normalized TMR when the voltage of the voltage value V2 is applied with respect to the MTJ memory element shown in FIG. It is a figure which shows the relationship between the voltage which the magnetization free layer of the MTJ memory element of a 2nd Example shows, and an anisotropic magnetic field. It is a figure which shows MH loop which the magnetization free layer in the MTJ memory element of a 2nd Example shows. It is a figure which shows the change of normalized TMR when write operation is performed with respect to the MTJ memory element of 2nd Example. It is a figure which shows the relationship between the voltage which the magnetization free layer of the MTJ memory element of a 3rd Example shows, and an anisotropic magnetic field. It is a figure which shows the change of normalization TMR when write-in operation | movement is performed with respect to the MTJ memory element of 3rd Example. It is a figure which shows an example of the relationship between resistance area product RA and a voltage magnetic anisotropy change. It is a figure which shows an example of a structure of a semiconductor memory device. FIG. 18 shows an example of the operation of the semiconductor memory device shown in FIG. 17.

  Hereinafter, embodiments of the present invention will be described in detail with reference to the accompanying drawings. In the following drawings, the same or corresponding components are referred to by the same or corresponding numerals, and the description thereof will be omitted as appropriate.

  FIG. 1 is a diagram illustrating an example of a basic configuration of a magnetic tunnel junction storage element. The MTJ memory element shown in FIG. 1 includes a lower electrode 10, a magnetization free layer 11, a tunnel insulating film 12, a magnetization fixed layer 13, and an upper electrode 14. The MTJ memory element shown in FIG. 1 is a perpendicular magnetization type MTJ, and the magnetization direction is oriented in a direction perpendicular to the surface of each layer (that is, the thickness direction of the layer), not in a direction parallel to the surface of each layer.

  In the example of FIG. 1, the magnetization direction of the magnetization fixed layer 13 is directed downward as indicated by an arrow in the layer. In this case, when the magnetization direction of the magnetization free layer 11 is directed downward, the magnetization direction of the magnetization free layer 11 and the magnetization direction of the magnetization fixed layer 13 are in a parallel state (a state in which they are directed in the same direction), and the MTJ memory is stored. The element is in a low resistance state showing a low resistance value. Further, when the magnetization direction of the magnetization free layer 11 is directed upward, the magnetization direction of the magnetization free layer 11 and the magnetization direction of the magnetization fixed layer 13 are in an antiparallel state (a state in which the magnetization direction is opposite), and the MTJ storage element Becomes a high resistance state showing a high resistance value. Information can be stored in the MTJ memory element by setting the MTJ memory element to a low resistance state or a high resistance state.

  In the conventional spin injection method, the magnetization of the magnetization free layer 11 can be set by using spin-polarized electron torque (STT). For example, a voltage is applied so that the lower electrode 10 on the magnetization free layer 11 side is connected to the positive electrode side and the upper electrode 14 on the magnetization fixed layer 13 side is connected to the negative electrode side. With this voltage application, electrons flow from the magnetization fixed layer 13 side to the magnetization free layer 11 side (that is, electrons flow downward in FIG. 1). While electrons having a spin opposite to the magnetization direction of the magnetization fixed layer 13 do not pass through the magnetization fixed layer 13, electrons having a spin in the same direction as the magnetization direction of the magnetization fixed layer 13 pass through the magnetization fixed layer 13. It reaches the magnetization free layer 11. Due to the influence of electrons having spins in the same direction as the magnetization direction of the magnetization fixed layer 13, the magnetization direction of the magnetization free layer 11 is set to a state having magnetization in the same direction as the magnetization direction of the magnetization fixed layer 13.

  Further, a voltage is applied so that the upper electrode 14 on the magnetization fixed layer 13 side is connected to the positive electrode side, and the lower electrode 10 on the magnetization free layer 11 side is connected to the negative electrode side. With this voltage application, electrons flow from the magnetization free layer 11 side to the magnetization fixed layer 13 side (that is, electrons flow upward in FIG. 1). Electrons having a spin in the same direction as the magnetization direction of the magnetization fixed layer 13 pass through the magnetization fixed layer 13, while electrons having a spin opposite to the magnetization direction of the magnetization fixed layer 13 are reflected by the magnetization fixed layer 13. The magnetization free layer 11 is affected. Due to the influence of electrons having a spin opposite to the magnetization direction of the magnetization fixed layer 13, the magnetization direction of the magnetization free layer 11 is set to a state having a magnetization opposite to the magnetization direction of the magnetization fixed layer 13.

  In the normal spin injection method described above, the direction of the voltage applied to the MTJ memory element differs depending on the state to be set, which is either the parallel state or the antiparallel state. Further, the voltage dependence of the magnetic anisotropy in the magnetization free layer 11 is not particularly considered.

  On the other hand, the writing method for the MTJ memory element disclosed in the present application enables writing with low power consumption by utilizing the voltage dependence of the magnetic anisotropy in the magnetization free layer 11. At this time, the direction of the voltage applied to the MTJ memory element during the writing operation is the same regardless of whether the writing is in the parallel state or the anti-parallel state. This voltage application direction is a direction in which the magnetic anisotropy of the magnetization free layer 11 decreases, that is, a direction in which the coercive force of the magnetization free layer 11 decreases. The magnetization fixed layer 13 side is positive and the magnetization free layer 11 side is negative. It is the direction that is. That is, a write operation is performed by applying a voltage in a direction in which a current flows from the magnetization fixed layer 13 toward the magnetization free layer 11.

  The MTJ memory element may be designed so that the voltage dependence of the magnetic anisotropy of the magnetization free layer 11 is sufficiently large. That is, the MTJ memory element may be designed such that the anisotropic magnetic field of the magnetization free layer 11 changes greatly according to the change in applied voltage.

  Applying a voltage only for a first period so that a current flows in a direction from the magnetization fixed layer 13 to the magnetization free layer 11 to the perpendicular magnetization type MTJ memory element including the magnetization fixed layer 13 and the magnetization free layer 11. Thus, the MTJ memory element is written in the antiparallel state. In addition, the MTJ memory element is written in a parallel state by applying a voltage to the MTJ memory element in a second period shorter than the first period so that a current flows in the same direction as the above direction. In the example of FIG. 1, the magnetization direction of the magnetization fixed layer 13 is downward, but the write operation to the anti-parallel state and the parallel state of the magnetization free layer 11 is downward even when the magnetization direction of the magnetization fixed layer 13 is upward. However, the same can be done.

  In the writing in the antiparallel state, the magnetization direction of the magnetization free layer 11 is changed by the spin injection effect acting in a state where the magnetic anisotropy of the magnetization free layer 11 is reduced (that is, the coercive force is reduced) by voltage application. Set in antiparallel direction. Accordingly, low power consumption name writing with a small current is possible. Since the spin injection effect takes some time after the voltage is applied and the current starts to flow, the first period in which the voltage is applied at the time of writing in the antiparallel state is set to a period in which the spin injection effect appears sufficiently. May be.

  More specifically, the magnetization direction precesses as the magnetic anisotropy of the magnetization free layer 11 decreases. The central direction of this precession is the same as that in the example of FIG. 1 in the initial period of voltage application. The direction is closer to the downward direction than the horizontal direction. When the center direction of the initial precession is directed downward, there is an influence of the magnetization direction of the magnetization fixed layer 13, and the MH loop (magnetic hysteresis curve) of the magnetization free layer 11 is on the positive magnetic field side. If it is shifted, the magnetization direction tends to face downward. Thereafter, as the effect of spin injection that works to turn the magnetization direction upward increases, the center direction of the precession moves in a direction closer to the upward direction than the horizontal direction in FIG. 1 and gradually approaches upward.

  Compared with the period of time required for the spin injection effect to become more dominant among the effect of lowering the magnetization direction due to the decrease in magnetic anisotropy and the effect of increasing the magnetization direction by spin injection, The first period in which the voltage is applied at the time of writing in the parallel state may be a longer period. When the voltage application is stopped after waiting for the spin injection effect to become dominant, the magnetization direction of the magnetization free layer 11 is set to an antiparallel direction according to the spin injection effect.

  In the parallel writing described above, the second period in which the voltage is applied during the parallel writing may be set to a period shorter than the period required for sufficient spin injection effect to appear. That is, compared with the period of time required for the spin injection effect to become more dominant among the effect of directing the magnetization direction due to the decrease in magnetic anisotropy and the effect of directing the magnetization direction by spin injection upward, The second period in which the voltage is applied at the time of writing in the parallel state may be a shorter period. When the voltage application is stopped before the spin injection effect becomes dominant, that is, while the influence of precession accompanying the decrease in magnetic anisotropy is stronger than the influence of the spin injection effect, the magnetization direction of the magnetization free layer 11 is parallel. Set to direction.

  In the above-described writing operation of the MTJ memory element, writing in the antiparallel state and writing in the parallel state are performed in the state where the external magnetic field is not applied. Since it is not necessary to provide a function of applying an external magnetic field, a semiconductor memory device with high density and low power consumption can be realized.

  FIG. 2 is a diagram illustrating an example of a voltage waveform applied in the first period and a voltage waveform applied in the second period. FIG. 2A shows a voltage waveform applied to write the MTJ memory element in the anti-parallel state, and FIG. 2B shows a voltage waveform applied to write the MTJ memory element in the parallel state.

  The pulse width T1 of the voltage pulse applied to write the MTJ memory element shown in FIG. 2A in the anti-parallel state is the voltage applied to write the MTJ memory element shown in FIG. 2B in the parallel state. It is longer than the pulse width T2 of the pulse. That is, the first period T1 in which the voltage is applied to write the MTJ memory element in the antiparallel state is longer than the second period T2 in which the voltage is applied to write the MTJ memory element in the parallel state.

  Further, in the above-described write operation of the MTJ memory element, the voltage V2 applied in the second period may have a higher voltage value than the voltage V1 applied in the first period. In the second period, the magnetization free layer 11 is set in the parallel direction by utilizing the effect of directing the magnetization direction due to the decrease in magnetic anisotropy due to voltage application in the parallel direction (downward in the case of FIG. 1). Since it is magnetized, it is preferable to greatly reduce the magnetic anisotropy. In the second period, it is possible to more easily set the magnetization free layer 11 in the parallel state by applying a higher voltage V2. In the first period, writing with lower power consumption is possible by applying a lower voltage V1.

  FIG. 3 is a diagram showing an example of the configuration of the first embodiment of the MTJ storage element. The MTJ memory element shown in FIG. 3 has a configuration in which a plurality of layers from the first layer 22 to the tenth layer 31 are sandwiched between the lower electrode 21 and the upper electrode 32. As for the shape of the MTJ memory element, the horizontal cross section (cross section along the in-plane direction of the layer) is circular, and the diameter is 20 nm.

  The first layer 22 is a leakage magnetic field suppression layer and is CoPt having a thickness of 5 nm. The second layer 23 is Ru having a thickness of 3 nm. The third layer 24 is Ta having a thickness of 1 nm. The fourth layer 25 is a magnetization free layer and is CoFeB having a thickness of 1 nm. The fifth layer 26 is a barrier layer (tunnel insulating film) and is MgO having a thickness of 1.1 nm.

  The sixth layer 27 to the tenth layer 31 collectively function as one magnetization fixed layer. The sixth layer 27 is CoFeB having a thickness of 1.7 nm. The seventh layer 28 is Ta having a thickness of 0.4 nm. The eighth layer 29 is CoPt having a thickness of 5 nm. The ninth layer 30 is Ru having a thickness of 1 nm. The tenth layer 31 is CoPt having a thickness of 7 nm. By using Ta, CoPt and CoFeB having different crystal structures can be ferromagnetically coupled as one magnetic material. Ru is inserted for the purpose of inducing an antiferromagnetic exchange interaction and bringing the two CoPt layers into an antiparallel magnetization state. When the two CoPt layers are antiparallel, the leakage magnetic field from the pinned layer can be reduced.

  In the MTJ memory element having the structure shown in FIG. 3, it is assumed that the magnetic anisotropy of the magnetization free layer 25 changes linearly with respect to the voltage. This assumption substantially matches the behavior of the actual CoFeB layer. At this time, the rate of change of magnetic anisotropy with respect to the voltage change, that is, the slope of the straight line indicating the relationship between the voltage and the anisotropic magnetic field is a main parameter for determining whether or not the above-described write operation is successful. A micromagnetic simulation was performed on the MTJ memory element having the above-described structure, and the smallest inclination necessary to succeed in the above-described write operation was obtained. Hereinafter, a micromagnetic simulation for the MTJ memory element having the structure shown in FIG. 3 will be described. In this simulation, the characteristics of the MTJ memory element in a state where no external magnetic field is applied are examined.

  FIG. 4 is a diagram showing the relationship between the voltage and the anisotropic magnetic field indicated by the magnetization free layer shown in FIG. The horizontal axis represents the electric field, and the left side of the vertical axis represents the anisotropic magnetic field. The electric field value on the horizontal axis indicates the voltage value applied per 1 nm of the tunnel insulating film. On the right side of the vertical axis, interfacial magnetic anisotropy energy corresponding to the left anisotropic magnetic field is shown. The relationship between the voltage indicated by the magnetization free layer 25 and the anisotropic magnetic field is shown as a straight line passing through the points 40 to 42. At a voltage of 0 V, the anisotropic magnetic field at the position indicated by the point 40 was set to 1.89 T (Tesla). At the voltage V1, the anisotropic magnetic field at the position indicated by the point 41 was set to 1.7T. At the voltage V2 (> V1), the anisotropic magnetic field at the position indicated by the point 42 was set to 1.47T. This anisotropic magnetic field does not include the influence of the demagnetizing field.

The saturation magnetization of the free layer was 1720 emu / cm ³ , the polarizability was 0.6, the exchange stiffness constant was 1 × 10 ⁻⁶ erg / cm ² , and the damping constant was 0.01.

  The anisotropic magnetic field of 1.89 T at the voltage of 0 V is a value that realizes a holding force necessary for the MTJ memory element to hold the memory state. 1.7T at the voltage V1 and 1.47T at the voltage V2 are determined as follows.

Whether writing in an antiparallel state or writing in a parallel state, the smaller the magnetic anisotropy, the easier it is to write. As described above, writing in the antiparallel state is realized by the spin injection effect acting in a state where the magnetic anisotropy is reduced (that is, the coercive force is reduced). The upper limit of the anisotropic magnetic field and the lower limit of the current value necessary for writing in the antiparallel state can be obtained by micromagnetic simulation. The anisotropic magnetic field thus obtained is 1.7T described above. Further, the current density thus determined was 2 × 10 ⁶ A / cm ² .

The writing in the parallel state is realized by the effect of directing the magnetization direction downward due to the decrease in magnetic anisotropy. The upper limit of the anisotropic magnetic field necessary for writing in the parallel state can be obtained by micromagnetic simulation. The anisotropic magnetic field thus obtained is 1.47T described above. The voltage V2 at this time is set to 2.0V. Since the relationship between the anisotropic magnetic field and the voltage is linear, when the voltage V2 when the anisotropic magnetic field is 1.47T is 2.0V, the voltage V1 when the anisotropic magnetic field is 1.7T is 0.9V. Since the current density at the voltage V1 (= 0.9 V) is 2 × 10 ⁶ A / cm ² , the resistance area product RA (resistance value per unit area) is 45 Ω · μm ² . The rate of change of magnetic anisotropy energy with respect to the voltage change at this time (straight line in FIG. 4) was 180 fJ / Vm.

  FIG. 5 is a diagram showing a case where no current flows in the MH loop shown by the magnetization free layer shown in FIG. The MH loop (magnetic hysteresis curve) of the magnetization free layer 25 shown in FIG. 5 is a characteristic obtained by micromagnetic simulation. Since the magnetic anisotropy of the magnetization free layer 25 changes according to the applied voltage, the MH loop (magnetic hysteresis curve) of the magnetization free layer 25 also changes according to the applied voltage. In FIG. 5, a hysteresis curve 51 is a characteristic when the applied voltage is 0 V and the anisotropic magnetic field is 1.89T. The hysteresis curve 52 is a characteristic when the anisotropic magnetic field is reduced to 1.7 T by voltage application. The hysteresis curve 53 is a characteristic when the anisotropic magnetic field is reduced to 1.6 T by voltage application. The hysteresis curve 54 is a characteristic when the anisotropic magnetic field is reduced to 1.3 T by voltage application. As shown in FIG. 5, it can be seen that as the applied voltage increases, the hysteresis curve becomes narrower and the inclination becomes smaller, and the coercive force of the magnetization free layer 25 becomes smaller.

  FIG. 6 is a diagram showing a waveform of a voltage pulse applied to the MTJ memory element shown in FIG. 3 in a write operation. The horizontal axis is time, and the vertical axis is voltage. In the micromagnetic simulation, the voltage value V1 of the voltage pulse applied at the time of writing in the antiparallel state is 0.9 V, and the width of the voltage pulse (that is, the length of the first period) is 50 ns. This voltage pulse is applied in the period from 0 ns to 50 ns in FIG. The voltage value V2 of the voltage pulse applied at the time of writing in the parallel state is 2.0 V, and the width of the voltage pulse (that is, the length of the second period) is 2 ns. This voltage pulse is applied in the period of 100 ns to 102 ns in FIG.

  FIG. 7 is a diagram showing a change in normalized TMR when a write operation is performed on the MTJ memory element shown in FIG. In the micromagnetic simulation, by applying the voltage pulse shown in FIG. 6 to the MTJ memory element shown in FIG. 3, the normalized TMR value changes as shown in FIG. By applying a voltage pulse having a voltage value V1 of 0.9 V and a pulse width of 50 ns, the MTJ storage element that has stored the 0 state (parallel state) is written in the 1 state (antiparallel state). In addition, by applying a voltage pulse having a voltage value V2 of 2.0 V and a pulse width of 2 ns, the MTJ storage element that has stored one state (anti-parallel state) is written to zero state (parallel state).

  When the power consumed in the above-described 1-state write and 0-state write is compared with the power consumed in the conventional write operation using only the spin injection effect, the average between the 1-state write and the 0-state write is about 1 Writing can be realized with power consumption of / 3. That is, it is possible to reduce the write power consumption by nearly 70%.

FIG. 8 is a diagram showing in detail the change in normalized TMR when the voltage V2 is applied to the MTJ memory element shown in FIG. When the application of a voltage of 2.0 V to the MTJ memory element after being written from the 0 state to the 1 state is started at time 100 ns, the normalized TMR rapidly decreases as shown by the waveform 61. At this time, the MTJ memory element changes from the high resistance state to the low resistance state, and the current density flowing through the element changes between 2 × 10 ⁶ A / cm ² and 4 × 10 ⁶ A / cm ² . If the voltage of 2.0 V is continuously applied as it is, the normalized TMR increases as shown by the waveform 62 and then continues to vibrate as shown by the waveform 63. This increase in normalized TMR is due to the spin injection effect caused by the current flowing in the MTJ memory element.

  If the voltage application is stopped at the time t1 of 2 ns after the voltage application is started instead of continuing to apply the voltage of 2.0 V, the normalized TMR converges to 0 as shown by the waveform 64. This is because when the voltage is not applied, the magnetic anisotropy becomes a sufficiently large value, and the magnetization free layer behaves so that the magnetization is oriented in the vertical direction.

  Further, when the voltage application is stopped at the time t2 of 50 ns after the start of voltage application (position of 150 ns in FIG. 8), the normalized TMR converges to 1 as shown by the waveform 65. As described above, when the voltage application is stopped after the normalized TMR is sufficiently close to 1 by the spin injection effect, the magnetization direction of the MTJ memory element is directed to the 1 side instead of the 0 side.

  FIG. 9 is a diagram illustrating an example of the locus of the magnetization direction when the magnetization direction of the magnetization free layer illustrated in FIG. 3 changes. In the micromagnetic simulation, when a voltage of 2.0 V is continuously applied to the MTJ memory element shown in FIG. 3, the magnetization direction of the magnetization free layer shows a locus as shown in FIG.

  9A and 9B, the vertical axis direction of the illustrated sphere corresponds to the thickness direction of the magnetization free layer, and the horizontal cross section of the sphere corresponds to the in-plane direction of the magnetization free layer. The magnetization of the magnetization free layer is represented by a vector starting from the center of the sphere. The locus shown in FIG. 9 is a locus drawn by the end point of the vector indicating the magnetization of the magnetization free layer.

  A point 71 represents the magnetization of the magnetization free layer of the MTJ memory element written in one state. When application of a voltage of 2.0 V to the MTJ memory element is started, the magnetization direction starts to move toward the zero state (downward in the drawing) as indicated by a locus 72. Thereafter, when a voltage of 2.0 V is continuously applied, the magnetization direction precesses while facing a direction closer to the lower side than the upper side as indicated by a locus 73.

  FIG. 9A shows a trajectory from the start of voltage application to 5 ns. FIG. 9B shows a locus from 5 ns to 10 ns after the start of voltage application. At 5 ns after the start of voltage application, the magnetization direction facing downward as indicated by a point 74 is directed to a direction closer to upward than downward and thereafter precessing while facing upward as indicated by a locus 75. to continue.

  Thus, the magnetization direction precesses as the magnetic anisotropy of the magnetization free layer decreases. The center direction of this precession is as shown in FIG. 9A during the initial period of voltage application. The direction is closer to the downward direction than the horizontal direction. When the initial precession center direction is close to the downward direction, there is an influence of the magnetization direction of the magnetization fixed layer, and the MH loop (magnetic hysteresis curve) of the magnetization free layer shifts to the positive magnetic field side. If it is, it becomes easy to turn the magnetization direction downward. Thereafter, as the effect of spin injection that works to turn the magnetization direction upward increases, the center direction of the precession moves closer to the upward direction than the horizontal direction as shown in FIG. Approaching upwards.

  FIG. 10 is a diagram showing a change in normalized TMR when a voltage of voltage value V2 is applied to the MTJ memory element shown in FIG. In the above-described micromagnetic simulation, by applying a voltage pulse having a voltage value V1 of 0.9 V and a pulse width of 50 ns, the MTJ storage element that has stored the 0 state (parallel state) is in the 1 state (antiparallel state). Is written on. In addition, by applying a voltage pulse having a voltage value V2 of 2.0 V and a pulse width of 2 ns, the MTJ storage element that has stored one state (anti-parallel state) is written to zero state (parallel state). The voltage value of 0.9 V used for writing in the antiparallel state is a voltage value corresponding to the minimum current value among writable current values in consideration of power consumption reduction. However, V1 <V2 is not an essential requirement for writing in the antiparallel state. For example, even when V1 is set to a voltage value (2.0 V) equal to V2, writing in the antiparallel state is possible. is there.

  FIG. 10 shows that in the micromagnetic simulation, by applying a voltage pulse having a voltage value of 2.0 V and a pulse width of 20 ns, the MTJ storage element storing the 0 state (parallel state) is in the 1 state (antiparallel state). ) Is written. Note that, under the conditions shown in FIG. 10, the voltage application is stopped 20 ns after the voltage application is started. As can be seen by comparing the waveform between 0 ns and 50 ns shown in FIG. 8 with the waveform shown in FIG. 10, the case where 2.0 V is applied rather than the case where 0.9 V is applied (FIG. 8) (FIG. 8). In 10), the timing of rising toward the antiparallel state is earlier. Therefore, in the case of FIG. 10, the writing in the antiparallel state may be completed in a time earlier than the writing in the antiparallel state shown in FIG. 8, and the voltage value and the current value are compared with the case of FIG. There is a possibility that the power consumption does not increase so much even if the value of.

  In the first embodiment of the MTJ memory element described above, the center of the MH loop is shifted to the positive magnetic field side as shown in FIG. In the second embodiment of the MTJ memory element described below, the center of the MH loop is adjusted so as to be located near the magnetic field zero.

  The structure of the MTJ memory element of the second embodiment is the same as the structure shown in FIG. 3, and has a circular shape with a diameter of 20 nm as in the first embodiment. However, the film thickness is different from that of the MTJ memory element of the first embodiment. In the MTJ memory element of the second embodiment, the CoPt of the first layer 22 has a thickness of 6 nm, which is thicker than the 5 nm of the first embodiment. The CoPt of the eighth layer 29 has a thickness of 6 nm, which is thicker than 5 nm of the first example. The CoPt of the tenth layer 31 has a thickness of 9 nm, which is thicker than 7 nm of the first embodiment. By increasing the CoPt film thickness in this way, the shift of the MH loop can be reduced.

  FIG. 11 is a diagram illustrating the relationship between the voltage and the anisotropic magnetic field exhibited by the magnetization free layer of the MTJ memory element according to the second embodiment. The horizontal axis represents the electric field, and the left side of the vertical axis represents the anisotropic magnetic field. The electric field value on the horizontal axis indicates the voltage value applied per 1 nm of the tunnel insulating film. On the right side of the vertical axis, interfacial magnetic anisotropy energy corresponding to the left anisotropic magnetic field is shown. The relationship between the voltage indicated by the magnetization free layer 25 and the anisotropic magnetic field is shown as a straight line passing through the points 80 to 82. At a voltage of 0 V, the anisotropic magnetic field at the position indicated by the point 80 was set to 1.89 T (Tesla). At the voltage V1, the anisotropic magnetic field at the position indicated by the point 81 was set to 1.7T. At the voltage V2 (> V1), the anisotropic magnetic field at the position indicated by the point 82 was set to 1.2T.

The anisotropic magnetic field of 1.89 T at the voltage of 0 V is a value that realizes a holding force necessary for the MTJ memory element to hold the memory state. 1.7T at the voltage V1 and 1.2T at the voltage V2 are determined in the same manner as in the first embodiment. That is, the upper limit of the anisotropic magnetic field and the lower limit of the current value necessary for writing in the antiparallel state can be obtained by micromagnetic simulation. The anisotropic magnetic field thus obtained is 1.7T described above. Further, the current density thus determined was 2 × 10 ⁶ A / cm ² .

In addition, the upper limit of the anisotropic magnetic field necessary for writing in the parallel state can be obtained by micromagnetic simulation. The anisotropic magnetic field thus obtained is 1.2T described above. The voltage V2 at this time is set to 2.0V. Since the relationship between the anisotropic magnetic field and the voltage is linear, when the voltage V2 when the anisotropic magnetic field is 1.2T is 2.0V, the voltage V1 when the anisotropic magnetic field is 1.7T is It will be about 0.55V. Since the current density at the voltage V1 (= about 0.55 V) is 2 × 10 ⁶ A / cm ² , the resistance area product RA (resistance value per unit area) is 28 Ω · μm ² . Further, the rate of change of magnetic anisotropy energy with respect to the voltage change at this time (straight line in FIG. 11) was 297 fJ / Vm.

  FIG. 12 is a diagram showing a case where no current flows in the MH loop indicated by the magnetization free layer in the MTJ memory element of the second embodiment. The MH loop (magnetic hysteresis curve) of the magnetization free layer 25 shown in FIG. 12 is a characteristic obtained by micromagnetic simulation. In FIG. 12, a hysteresis curve 91 is a characteristic when the applied voltage is 0 V and the anisotropic magnetic field is 1.89T. The hysteresis curve 92 is a characteristic when the anisotropic magnetic field is reduced to 1.7 T by voltage application. The hysteresis curve 93 is a characteristic when the anisotropic magnetic field is reduced to 1.45 T by voltage application. The hysteresis curve 94 is a characteristic when the anisotropic magnetic field is reduced to 1.3 T by voltage application. As shown in FIG. 12, it can be seen that as the applied voltage increases, the hysteresis curve becomes narrower and the inclination becomes smaller, and the coercive force of the magnetization free layer 25 becomes smaller. It can also be seen that the shift of the hysteresis curve is small and the loop center of the hysteresis curve is adjusted so as to be located in the vicinity of substantially zero magnetic field.

  FIG. 13 is a diagram illustrating a change in normalized TMR when a write operation is performed on the MTJ memory element according to the second embodiment. The voltage pulse condition applied to the MTJ memory element of the second embodiment is the same as that of the first embodiment except that the voltage value of the voltage pulse for writing to the 1 state is 0.55V instead of 0.9V. This is the same as the voltage pulse condition shown in FIG. 6 applied to the MTJ memory element. In the micromagnetic simulation, the MTJ memory element that has stored the 0 state (parallel state) is written to the 1 state (anti-parallel state) by applying a voltage pulse having a voltage value V1 of 0.55 V and a pulse width of 50 ns. It is.

  When application of a voltage of 2.0 V is started at time 100 ns to the MTJ memory element after being written from the 0 state to the 1 state, the normalized TMR rapidly decreases as shown by the waveform 131. If the voltage of 2.0 V is continuously applied as it is, the normalized TMR increases as shown by the waveform 132 and then continues to vibrate as shown by the waveform 133. This increase in normalized TMR is due to the spin injection effect caused by the current flowing in the MTJ memory element.

  If the voltage application is stopped at the time t21 of 2 ns after the voltage application is started instead of continuing to apply the voltage of 2.0 V, the normalized TMR converges to 0 as shown by the waveform 134. This is because when the voltage is not applied, the magnetic anisotropy becomes a sufficiently large value, and the magnetization free layer behaves so that the magnetization is oriented in the vertical direction. Thus, by applying a voltage pulse having a voltage value V2 of 2.0 V and a pulse width of 2 ns, the MTJ memory element that has stored one state (anti-parallel state) is written to the zero state (parallel state). Is possible.

  When the voltage application is stopped at time t22 (at the position of 120 ns in FIG. 13) 20 ns after the start of voltage application, the normalized TMR converges to 1 as shown by the waveform 135. As described above, when the voltage application is stopped after the normalized TMR is sufficiently close to 1 by the spin injection effect, the magnetization direction of the MTJ memory element is directed to the 1 side instead of the 0 side.

  In the case of the MTJ memory element of the first embodiment in which the MH loop is shifted to the positive magnetic field side, the anisotropic magnetic field required for writing in the 0 state is 1.47 T, and the magnetic field against the voltage change at this time The change rate of the anisotropic energy was 180 fJ / Vm. On the other hand, in the case of the MTJ memory element of the second embodiment in which the shift of the MH loop is small, the anisotropic magnetic field required for writing in the 0 state is 1.2 T. The change rate of the isotropic energy was 297 fJ / Vm. Thus, it can be seen that the larger the MH loop shift amount, the smaller the voltage dependency of the anisotropic magnetic field necessary for realizing the rewrite operation.

  In the first and second embodiments of the MTJ memory element described above, the element diameter was 20 nm. In the third example of the MTJ memory element described below, the diameter of the element was set to 30 nm, and the characteristics were examined by micromagnetic simulation.

The structure of the MTJ memory element of the third embodiment is the same as the structure shown in FIG. 3, but has a circular shape with a diameter of 30 nm, unlike the MTJ memory element of the first and second embodiments. The thickness of the MTJ memory element of the third embodiment is the same as that of the MTJ memory element of the first embodiment.
FIG. 14 is a diagram illustrating the relationship between the voltage and the anisotropic magnetic field exhibited by the magnetization free layer of the MTJ memory element according to the third embodiment. The horizontal axis represents the electric field, and the left side of the vertical axis represents the anisotropic magnetic field. The electric field value on the horizontal axis indicates the voltage value applied per 1 nm of the tunnel insulating film. On the right side of the vertical axis, interfacial magnetic anisotropy energy corresponding to the left anisotropic magnetic field is shown. The relationship between the voltage indicated by the magnetization free layer 25 and the anisotropic magnetic field is shown as a straight line passing through the points 140 to 142. At a voltage of 0 V, the anisotropic magnetic field at the position indicated by the point 140 was set to 1.89 T (Tesla). At the voltage V1, the anisotropic magnetic field at the position indicated by the point 141 was set to 1.75T. At the voltage V2 (> V1), the anisotropic magnetic field at the position indicated by the point 142 was set to 1.55T.

The anisotropic magnetic field of 1.89 T at the voltage of 0 V is a value that realizes a holding force necessary for the MTJ memory element to hold the memory state. 1.75T at voltage V1 and 1.55T at voltage V2 were determined in the same manner as in the first embodiment. The current density was 0.8 × 10 ⁶ A / cm ² at the lower limit of the current value required for writing in the antiparallel state.

If the voltage V2 during the parallel writing operation is set to 2.0V, the voltage V1 when the anisotropic magnetic field is 1.75T is about 0.82V. Since the current density at the voltage V1 (= about 0.82 V) is 0.8 × 10 ⁶ A / cm ² , the resistance area product RA (resistance value per unit area) is about 100 Ω · μm ² . Further, the rate of change of magnetic anisotropy energy with respect to the voltage change at this time (straight line in FIG. 11) was 146 fJ / Vm.

  FIG. 15 is a diagram showing a change in normalized TMR when a write operation is performed on the MTJ memory element of the third embodiment. The voltage value of the voltage pulse for writing to one state is 0.82 V, and the pulse width of this voltage pulse is 50 ns. The voltage value of the voltage pulse for writing to the 0 state is 2.0 V, and the pulse width of this voltage pulse is 5 ns. In the micromagnetic simulation, the MTJ memory element that has stored the 0 state (parallel state) is written to the 1 state (anti-parallel state) by applying a voltage pulse having a voltage value V1 of 0.82 V and a pulse width of 50 ns. It is.

  When an application of a voltage of 2.0 V is started at the time 70 ns to the MTJ memory element after being written from the 0 state to the 1 state, the normalized TMR rapidly decreases as shown by the waveform 151. If the voltage of 2.0 V is continuously applied as it is, the normalized TMR increases as shown by the waveform 152. This increase in normalized TMR is due to the spin injection effect caused by the current flowing in the MTJ memory element.

  If the voltage application is stopped at the time t31 of 5.0 ns after the voltage application is started instead of continuing to apply the voltage of 2.0 V, the normalized TMR converges to 0 as shown by the waveform 153. This is because when the voltage is not applied, the magnetic anisotropy becomes a sufficiently large value, and the magnetization free layer behaves so that the magnetization is oriented in the vertical direction. In this way, by applying a voltage pulse having a voltage value V2 of 2.0 V and a pulse width of 5.0 ns, the MTJ memory element that has stored one state (anti-parallel state) is changed to zero state (parallel state). It becomes possible to write.

  When the voltage application is stopped at time t32 (position of 90 ns in FIG. 13) 20 ns after the start of voltage application, the normalized TMR converges to 1 as shown by the waveform 154. As described above, when the voltage application is stopped after the normalized TMR is sufficiently close to 1 by the spin injection effect, the magnetization direction of the MTJ memory element is directed to the 1 side instead of the 0 side.

  In the case of the MTJ memory element of the first embodiment having a diameter of 20 nm, the anisotropic magnetic field required for writing in the zero state is 1.47 T, and the change in magnetic anisotropy energy with respect to the voltage change at this time The rate was 180 fJ / Vm. On the other hand, in the case of the MTJ memory element of the third embodiment having a diameter of 30 nm, the anisotropic magnetic field required for writing in the 0 state is 1.55 T, and the magnetic anisotropy with respect to the voltage change at this time The change rate of sexual energy was 146 fJ / Vm. Thus, it can be seen that the larger the size (diameter) of the MTJ memory element, the smaller the voltage dependency of the anisotropic magnetic field necessary for realizing the rewrite operation.

  FIG. 16 is a diagram illustrating an example of the relationship between the resistance area product RA and the voltage magnetic anisotropy change. The horizontal axis represents the resistance area product RA, and the vertical axis represents the rate of change of magnetic anisotropy energy with respect to voltage change (the amount of change in magnetic anisotropy energy per unit voltage change). A curve 161 is calculated based on the data of the MTJ memory element of the first embodiment having a diameter of 20 nm, and the magnetic anisotropy change (magnetic anisotropy per unit voltage change) capable of executing the above-described write operation. The lower limit of the amount of change in sexual energy). That is, if there is a change in voltage magnetic anisotropy that exceeds the value indicated by the curve 161 (the region above the curve 161 in the graph), the write operation disclosed in this application for the MTJ memory element of the first embodiment is successful. be able to.

  A curve 162 is calculated based on the data of the MTJ memory element of the second embodiment having a diameter of 30 nm, and indicates the lower limit of the change in voltage magnetic anisotropy at which the above write operation can be performed. That is, if there is a change in voltage magnetic anisotropy greater than the value indicated by this curve 162 (the region above the curve 162 in the graph), the write operation disclosed in this application for the MTJ memory element of the third embodiment is successful. be able to.

  The current density is a parameter that controls the success or failure of writing in one state. However, when a current having the same current density is passed, the voltage decreases in proportion to the decrease rate when the resistance area product RA decreases. In order to maintain an anisotropic magnetic field having the same value in spite of this decrease in voltage, it is necessary to increase the voltage time anisotropy in inverse proportion to the decrease in current. Therefore, the curves 161 and 162 draw curves that are inversely proportional to the resistance area product RA.

  By appropriately designing the structure, material, film thickness, etc. of the MTJ memory element so as to satisfy the conditions of voltage magnetic anisotropy change shown in FIG. 16, the same polarity when no external magnetism is applied The write operation described above can be performed by the voltage pulse. In the case of an MTJ memory element having a diameter of 20 nm or 30 nm as shown in FIG. 16, for example, if the MTJ memory element is designed to have a voltage magnetic anisotropy change of about 200 fJ / Vm, the write operation disclosed in this application can be performed. It becomes possible.

  FIG. 17 is a diagram illustrating an example of the configuration of the semiconductor memory device. In FIG. 17, the boundary between each circuit or functional block indicated by each circuit symbol or box and another circuit or functional block basically indicates a functional boundary. It does not necessarily correspond to separation of electrical signals, separation of control logic, and the like. Each circuit or functional block may be one hardware module physically separated to some extent from another block, or one function in a hardware module physically integrated with another block May be shown.

  The semiconductor memory device shown in FIG. 17 includes a controller 170, a word line driver 171, a bit line driver 172, an MTJ memory element 173, a MOS transistor 174, a reference element 175, and a sense amplifier 176. The plurality of MTJ storage elements 173 are arranged in a matrix in a plurality of rows and a plurality of columns. One end of one MTJ storage element 173 is connected to a corresponding one of the plurality of bit lines BL1, BL2,..., And the other end is connected to one end of a channel of the MOS transistor 174. The other end of the channel of the MOS transistor 174 is connected to the ground, and the gate end is connected to a corresponding one of the n wart lines WL1 to WLn.

  The controller 170 operates in response to an external command input, clock input, address input, and the like, and controls the operation of the semiconductor memory device in synchronization with an externally input clock signal. The controller 170 performs a write operation and a read operation on the MTJ storage element 173 by decoding a command input from the outside and controlling each circuit of the semiconductor storage device according to the decoding result. Specifically, the controller 170 decodes an address input from the outside, and controls the word line driver 171 and the bit line driver 172 according to the decoding result, thereby selecting the word line selected according to the decoding result and The bit line is driven. The word line driver 171 activates the selected word line, and the bit line driver 172 activates the selected word line. A current flows through the MTJ memory element 173 and the MOS transistor 174 connected to the activated word line and the activated bit line, and a write operation or a read operation for the MTJ memory element 173 is executed.

  In the read operation, a voltage corresponding to the resistance value of the MTJ storage element 173 appears on the selected bit line. The sense amplifier 176 compares the voltage of the selected bit line applied to one input with the voltage corresponding to the reference element 175 applied to the other input, and outputs a data read result corresponding to the magnitude of both voltages. To do. The data read result output from the sense amplifier 176 is output to the outside of the semiconductor memory device as read data.

  In the write operation, the word line driver 171 and the bit line driver 172 apply a write voltage pulse to the MTJ storage element 173 by appropriately activating the selected word line and the selected bit line. By controlling the pulse width of the write voltage pulse and, if necessary, the voltage value, writing in the 0 state or writing in the 1 state to the MTJ storage element 173 is executed. The MTJ memory element 173 may be the MTJ memory element of any of the first to third embodiments described above, and the above-described write operation is executed by the voltage pulse illustrated in FIGS. 2A and 2B. It's okay.

  In the semiconductor memory device shown in FIG. 17, a mechanism for applying an external magnetic field to the MTJ memory element 173 is not provided. Then, a write operation is performed on the MTJ storage element 173. Accordingly, it is possible to realize a semiconductor memory device with low power consumption and high density.

  FIG. 18 shows an example of the operation of the semiconductor memory device shown in FIG. FIG. 18 shows a case where the bit line BL1 is a selected bit line and the word line WL1 is a selected word line WL1.

  FIG. 18A shows the voltage waveform of the selected bit line BL1. FIG. 18B shows the voltage waveform of the selected word line WL1. FIG. 18C shows the voltage waveform of the sense amplifier activation signal SA applied to the sense amplifier 176. FIG. 18D shows the output voltage waveform of the sense amplifier 176. FIG. 18 shows signal waveforms when a write operation in one state is performed, the written one data is read, a write operation in the zero state is performed thereafter, and the read operation of the written zero data is sequentially executed. Has been.

  In the one-state write operation, the selected word line WL1 is activated to make the MTJ storage element 173 to be written into a voltage-applicable state, and then the selected bit line BL1 is activated to store the one-state write voltage pulse in the MTJ storage. Applied to the element 173. The voltage waveform of the selected bit line BL1 may be, for example, the waveform of the write voltage pulse shown in FIG.

  In the subsequent read operation, the selected word line WL1 is activated to set the MTJ memory element 173 to be read to a voltage application enabled state, and then the selected bit line BL1 is activated to transfer the read voltage pulse to the MTJ memory element 173. Apply. At that time, the sense amplifier activation signal SA is applied from the controller 170 to the sense amplifier 176 to activate the sense amplifier 176. As a result, the sense amplifier 176 performs a read data determination operation, and the output signal of the sense amplifier 176 becomes a voltage value indicating data 1.

  In the write operation in the 0 state, the selected word line WL1 is activated to make the MTJ storage element 173 to be written into a voltage-applicable state, and then the selected bit line BL1 is activated to store the 0 state write voltage pulse in the MTJ storage. Applied to the element 173. The voltage waveform of the selected bit line BL1 may be, for example, the waveform of the write voltage pulse shown in FIG.

  In the subsequent read operation, the selected word line WL1 is activated to set the MTJ memory element 173 to be read to a voltage application enabled state, and then the selected bit line BL1 is activated to apply a read voltage pulse to the MTJ memory element 173. To do. At that time, the sense amplifier activation signal SA is applied from the controller 170 to the sense amplifier 176 to activate the sense amplifier 176. As a result, the sense amplifier 176 performs a read data determination operation, and the output signal of the sense amplifier 176 has a voltage value indicating data 0.

  Note that writing may be simultaneously performed on the plurality of MTJ storage elements 173 connected to the selected word line WL1, and different voltage pulses may be applied to each bit line depending on writing in the 0 state or 1 state. With this write operation, a bit value of 1 state or 0 state corresponding to the bit pattern may be written to external write data in each of the plurality of MTJ storage elements 173.

  As mentioned above, although this invention was demonstrated based on the Example, this invention is not limited to the said Example, A various deformation | transformation is possible within the range as described in a claim.

DESCRIPTION OF SYMBOLS 10 Lower electrode 11 Magnetization free layer 12 Tunnel insulating film 13 Magnetization fixed layer 14 Upper electrode 170 Controller 171 Word line driver 172 Bit line driver 173 MTJ memory element 174 MOS transistor 175 Reference element 176 Sense amplifier

Claims (5)

A voltage is applied to the perpendicular magnetization type magnetic tunnel junction memory element including the magnetization fixed layer and the magnetization free layer for a first period so that a current flows in a direction from the magnetization fixed layer to the magnetization free layer. Write the tunnel junction memory element in an antiparallel state,
A magnetic tunnel including each step of writing a voltage in a parallel state by applying a voltage for a second period shorter than a first period so that a current flows in the direction to the magnetic tunnel junction memory element. A method of writing a junction memory element.   The magnetic tunnel junction memory element writing method according to claim 1, wherein writing to the antiparallel state and writing to the parallel state are executed in a state where no external magnetic field is applied.   3. The magnetic tunnel junction memory element writing method according to claim 1, wherein the voltage applied in the second period has a higher voltage value than the voltage applied in the first period.   The first period is longer than the period required for the spin injection effect on the magnetization free layer to be dominant, and the second period is because the spin injection effect on the magnetization free layer is dominant. 4. The method for writing into a magnetic tunnel junction memory element according to claim 1, wherein the time period is shorter than a period required for the magnetic tunnel junction memory element. A perpendicular magnetization type magnetic tunnel junction memory element including a magnetization fixed layer and a magnetization free layer;
A voltage is applied to the magnetic tunnel junction memory element in a first period so that a current flows in a direction from the magnetization fixed layer to the magnetization free layer, and the magnetic tunnel junction memory element is written in an antiparallel state, A semiconductor memory device including a control circuit for applying a voltage for a second period shorter than the first period so as to cause a current to flow in the direction to the tunnel junction memory element and writing the magnetic tunnel junction memory element in a parallel state .

Images