JP2010282724A - Thin film magnetic memory device - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To improve operation reliability and reduce magnetic noise by reducing data write current in an MRAM device including MTJ memory cells. <P>SOLUTION: In data writing, data write currents ±Iw to be supplied to a bit line pair BLP are supplied as reciprocating currents flowing in different directions in bit lines BL and /BL, respectively in a selected memory cell column. In the bit line pair BLP, the bit lines BL and /BL are formed using different metallic wiring layers so as to hold a magnetic tunnel junction part MTJ in a vertical direction. The bit lines BL and /BL are electrically coupled to each other at an end of a memory array 10 by an equalizer transistor 62, to supply data write currents. <P>COPYRIGHT: (C)2011,JPO&INPIT

Description

  The present invention relates to a thin film magnetic memory device, and more particularly to a random access memory including a memory cell having a magnetic tunnel junction (MTJ).

  An MRAM (Magnetic Random Access Memory) device has attracted attention as a storage device that can store nonvolatile data with low power consumption. An MRAM device is a storage device that performs non-volatile data storage using a plurality of thin film magnetic bodies formed in a semiconductor integrated circuit and allows random access to each of the thin film magnetic bodies.

  In particular, in recent years, it has been announced that the performance of an MRAM device is dramatically improved by using a thin film magnetic material using a magnetic tunnel junction (MTJ) as a memory cell. For MRAM devices with memory cells with magnetic tunnel junctions, see “A 10ns Read and Write Non-Volatile Memory Array Using a Magnetic Tunnel Junction and FET Switch in each Cell”, ISSCC Digest of Technical Papers, TA7.2, Feb 2000. and “Nonvolatile RAM based on Magnetic Tunnel Junction Elements”, ISSCC Digest of Technical Papers, TA7.3, Feb. 2000.

  FIG. 83 is a schematic diagram showing a configuration of a memory cell having a magnetic tunnel junction (hereinafter also simply referred to as “MTJ memory cell”).

  Referring to FIG. 83, the MTJ memory cell includes a magnetic tunnel junction MTJ whose resistance value changes according to the data level of stored data, and an access transistor ATR. Access transistor ATR is formed of a field effect transistor, and is coupled between magnetic tunnel junction MTJ and ground voltage Vss.

  For MTJ memory cells, write word line WWL for instructing data writing, read word line RWL for instructing data reading, and the level of stored data at the time of data reading and data writing A bit line BL which is a data line for transmitting the electrical signal is disposed.

FIG. 84 is a conceptual diagram illustrating a data read operation from the MTJ memory cell.
Referring to FIG. 84, magnetic tunnel junction MTJ includes magnetic layer (hereinafter also simply referred to as “fixed magnetic layer”) FL having a fixed magnetic field in a certain direction and magnetic layer (hereinafter simply referred to as “fixed magnetic layer”) having a free magnetic field. VL) (also referred to as “free magnetic layer”). A tunnel barrier TB formed of an insulator film is disposed between the fixed magnetic layer FL and the free magnetic layer VL. In the free magnetic layer VL, either one of the magnetic field in the same direction as that of the fixed magnetic layer FL and the magnetic field in a direction different from that of the fixed magnetic layer FL is nonvolatilely written in accordance with the level of stored data.

  At the time of data reading, access transistor ATR is turned on in response to activation of read word line RWL. As a result, a sense current Is supplied as a constant current from a control circuit (not shown) flows through a current path from the bit line BL to the magnetic tunnel junction MTJ to the access transistor ATR to the ground voltage Vss.

  The resistance value of the magnetic tunnel junction MTJ changes according to the relative relationship in the magnetic field direction between the fixed magnetic layer FL and the free magnetic layer VL. Specifically, when the magnetic field direction of the pinned magnetic layer FL and the magnetic field direction written in the free magnetic layer VL are the same, the resistance of the magnetic tunnel junction MTJ is greater than when both magnetic field directions are different. The value becomes smaller.

  Therefore, at the time of data reading, the voltage change caused at the magnetic tunnel junction MTJ by the sense current Is differs according to the magnetic field direction stored in the free magnetic layer VL. Thus, if the supply of the sense current Is is started after the bit line BL is once precharged to a high voltage, the level of the data stored in the MTJ memory cell is read by monitoring the voltage level change of the bit line BL. Can do.

FIG. 85 is a conceptual diagram illustrating a data write operation for the MTJ memory cell.
Referring to FIG. 85, at the time of data writing, read word line RWL is inactivated and access transistor ATR is turned off. In this state, a data write current for writing a magnetic field in free magnetic layer VL is supplied to write word line WWL and bit line BL. The magnetic field direction of free magnetic layer VL is determined by a combination of directions of data write currents flowing through write word line WWL and bit line BL, respectively.

  FIG. 86 is a conceptual diagram illustrating the relationship between the direction of data write current and the direction of magnetic field during data writing.

  Referring to FIG. 86, magnetic field Hx indicated by the horizontal axis indicates the direction of magnetic field H (WWL) generated by the data write current flowing through write word line WWL. On the other hand, the magnetic field Hy indicated on the vertical axis indicates the direction of the magnetic field H (BL) generated by the data write current flowing through the bit line BL.

  The magnetic field direction stored in the free magnetic layer VL is newly written only when the sum of the magnetic fields H (WWL) and H (BL) reaches the region outside the asteroid characteristic line shown in the figure. . That is, when a magnetic field corresponding to the region inside the asteroid characteristic line is applied, the magnetic field direction stored in the free magnetic layer VL is not updated.

  Therefore, in order to update the data stored in the magnetic tunnel junction MTJ by the write operation, it is necessary to pass a current through both the write word line WWL and the bit line BL. The magnetic field direction once stored in the magnetic tunnel junction MTJ, that is, the stored data is held in a nonvolatile manner until new data writing is executed.

  Even during the data read operation, sense current Is flows through bit line BL. However, since the sense current Is is generally set to be about 1 to 2 digits smaller than the data write current described above, the stored data in the MTJ memory cell is erroneously read at the time of data reading due to the influence of the sense current Is. The possibility of rewriting is small.

  The above-described technical literature discloses a technique for constructing an MRAM device that is a random access memory by integrating such MTJ memory cells on a semiconductor substrate.

FIG. 87 is a conceptual diagram showing MTJ memory cells integrated and arranged in a matrix.
Referring to FIG. 87, a highly integrated MRAM device can be realized by arranging MTJ memory cells in a matrix on a semiconductor substrate. FIG. 87 shows a case where MTJ memory cells are arranged in n rows × m columns (n, m: natural numbers).

  As already described, it is necessary to arrange the bit line BL, the write word line WWL, and the read word line RWL for each MTJ memory cell. Therefore, n write word lines WWL1 to WWLn and read word lines RWL1 to RWLn and m bit lines BL1 to BLm are arranged for n × m MTJ memory cells arranged in a matrix. There is a need.

  As described above, an MTJ memory cell is generally provided with an independent word line corresponding to each of the read operation and the write operation.

FIG. 88 is a structural diagram of an MTJ memory cell arranged on a semiconductor substrate.
Referring to FIG. 88, access transistor ATR is formed in p type region PAR on semiconductor main substrate SUB. Access transistor ATR has source / drain regions 110 and 120 which are n-type regions, and a gate 130. Source / drain region 110 is coupled to ground voltage Vss through a metal wiring formed in first metal wiring layer M1. For the write word line WWL, a metal wiring formed in the second metal wiring layer M2 is used. The bit line BL is provided in the third metal wiring layer M3.

  The magnetic tunnel junction MTJ is disposed between the second metal wiring layer M2 provided with the write word line WWL and the third metal wiring layer M3 provided with the bit line BL. Source / drain region 120 of access transistor ATR is connected to magnetic tunnel junction MTJ through metal film 150 formed in the contact hole, first and second metal wiring layers M1 and M2, and barrier metal 140. Electrically coupled. The barrier metal 140 is a cushioning material provided to electrically couple the magnetic tunnel junction MTJ and the metal wiring.

  As already described, in the MTJ memory cell, the read word line RWL is provided as a wiring independent of the write word line WWL. The write word line WWL and the bit line BL need to pass a data write current for generating a magnetic field having a magnitude greater than a predetermined value at the time of data writing. Therefore, the bit line BL and the write word line WWL are formed using metal wiring.

  On the other hand, the read word line RWL is provided to control the gate voltage of the access transistor ATR, and it is not necessary to actively flow a current. Therefore, from the viewpoint of increasing the degree of integration, the read word line RWL is formed using a polysilicon layer, a polycide structure, or the like in the same wiring layer as the gate 130 without newly providing an independent metal wiring layer.

Roy Scheuerlein and 6 others, “A 10ns Read and Write Non-Volatile Memory Array Using a Magnetic Tunnel Using FET Switches and Magnetic Tunnel Junctions in Each Cell Junction and FET Switch in each Cell), (USA), 2000 IEICE International Solid Circuit Conference and Technical Papers TA7.2 (2000 IEEE ISSCC Digest of Technical Papers, TA7.2), p. 128-129 D. Durlam and five others, “Nonvolatile RAM based on Magnetic Tunnel Junction Elements” (USA), 2000 International Solid State Circuit Conference / Technology of the Institute of Electrical and Electronics Engineers, 2000 Proceedings TA7.3 (2000 IEEE ISSCC Digest of Technical Papers, TA7.3), p. 130-131

  However, as described with reference to FIG. 84, data reading from the MTJ memory cell is executed based on a voltage change caused by passing a sense current (Is in FIG. 84) through the magnetic tunnel junction MTJ acting as a resistor. . Therefore, when the RC constant of the sense current path is large, this voltage change cannot be generated promptly, and it is difficult to increase the speed of the data read operation.

  Also, as shown in FIG. 86, data writing is executed in accordance with the magnitude of the magnetic field with respect to the asteroid characteristic line given as the threshold value, and therefore the variation in the asteroid characteristic line during the manufacture of the memory cell. There also arises a problem that it leads directly to variations in the write margin to the memory cell.

  FIG. 89 is a conceptual diagram for explaining the influence of manufacturing variation on the data write margin.

  Referring to FIG. 89, the design value of the asteroid characteristic line is indicated by reference sign ASd in the figure. Here, a case is considered in which the asteroid characteristic line of the memory cell is deviated from the design value as indicated by the reference sign Asa or ASb due to manufacturing variations of the MRAM device.

  For example, in an MTJ memory cell having an attodot characteristic line ASb, data writing cannot be performed even when a data write current is applied according to a design value and a data write magnetic field is applied.

  On the other hand, in the MTJ memory cell having the asteroid characteristic line Asa, data writing is executed even when a data writing magnetic field smaller than the design value is applied. As a result, the MTJ memory cell having such characteristics becomes very weak against magnetic noise.

  Such manufacturing variation of the asteroid characteristic line may be further increased in accordance with the downsizing of the memory cell with high integration. Therefore, not only the development of manufacturing technology that reduces manufacturing variations of asteroid characteristic lines, but also the technology that adjusts to ensure an appropriate data write margin in response to fluctuations in asteroid characteristic lines is the manufacturing yield. It will be necessary to secure.

  Further, as described with reference to FIGS. 85 and 86, at the time of data writing, it is necessary to pass a relatively large data write current through bit line BL and write word line WWL. When the data write current increases, the current density in the bit line BL and the write word line WWL increases, and there is a possibility that a phenomenon generally called electromigration occurs.

  As a result, when disconnection due to electromigration or a short circuit between the wirings occurs in these wirings, the operation reliability of the MRAM device may be impaired. Further, when the data write current becomes large, the influence of magnetic noise caused thereby may not be negligible. Therefore, it is desirable that the data write can be executed with a smaller data write current.

  As described with reference to FIGS. 87 and 88, since the number of wirings required to execute data writing and data reading with respect to the MTJ memory cell is large, a memory array in which MTJ memory cells are arranged in an integrated manner Therefore, it is difficult to reduce the chip area of the MRAM device.

  As a structure of an MTJ memory cell that can be further integrated as compared with the MTJ memory cell shown in FIG. 83, a structure using a PN junction diode as an access element instead of an access transistor is known.

FIG. 90 is a schematic diagram showing a configuration of an MTJ memory cell using a diode.
Referring to FIG. 90, an MTJ memory cell using a diode includes a magnetic tunnel junction MTJ and an access diode DM. Access diode DM is coupled between the two, with the direction from magnetic tunnel junction MTJ to word line WL as the forward direction. Bit line BL is provided in a direction crossing word line WL, and is coupled to magnetic tunnel junction MTJ.

  Data writing to the MTJ memory cell using the diode is performed by flowing a data write current through the word line WL and the bit line BL. The direction of the data write current is set according to the data level of the write data, as in the case of the memory cell using the access transistor.

  On the other hand, at the time of data reading, word line WL corresponding to the selected memory cell is set to a low voltage (for example, ground voltage Vss) state. At this time, by precharging the bit line BL to a high voltage (for example, power supply voltage Vcc) state, the access diode DM becomes conductive and the sense current Is can flow to the magnetic tunnel junction MTJ. On the other hand, since the word line WL corresponding to the non-selected memory cell is set to the high voltage state, the corresponding access diode DM maintains the off state, and the sense current Is does not flow.

  In this manner, data reading and data writing can be executed also in an MTJ memory cell using an access diode.

  FIG. 91 is a structural diagram when the MTJ memory cell shown in FIG. 90 is arranged on a semiconductor substrate.

  Referring to FIG. 91, an access diode DM is formed by N-type well NWL on semiconductor main substrate SUB and P-type region PRA provided on N-type well NWL.

  N-type well NWL corresponding to the cathode of access diode DM is coupled to word line WL arranged in metal interconnection layer M1. P-type region PRA corresponding to the anode of access diode DM is electrically coupled to magnetic tunnel junction MTJ through barrier metal 140 and metal film 150. Bit line BL is arranged in metal interconnection layer M2 and coupled to magnetic tunnel junction MTJ. In this way, by using an access diode instead of an access transistor, an MTJ memory cell advantageous for high integration can be configured.

  However, since data write current flows through word line WL and bit line BL during data writing, a voltage drop occurs in the data write current in these wirings. As a result of such a voltage drop, depending on the voltage distribution on the word line WL and the bit line BL, the PN junction of the access diode DM is turned on in a part of the MTJ memory cell not subjected to data writing. There is a risk that. As a result, an unexpected current may flow through the MTJ memory cell, which may cause erroneous data writing.

  As described above, the conventional MTJ memory cell using the access diode is advantageous for high integration, but has a problem that the data writing operation becomes unstable.

  The present invention has been made to solve these problems, and a first object of the present invention is to increase the speed of data reading in an MRAM device including MTJ memory cells.

  A second object of the present invention is to adjust the amount of data write current to ensure a predetermined data write margin by compensating for variations in magnetic characteristics due to manufacturing variations in an MRAM device having MTJ memory cells. It is to provide a configuration that can be easily executed.

  A third object of the present invention is to improve operation reliability and suppress magnetic noise by reducing a data write current in an MRAM device including MTJ memory cells.

  A fourth object of the present invention is to provide a configuration of an MTJ memory cell capable of high integration and having high operation reliability.

  A fifth object of the present invention is to reduce the chip area in an MRAM device having MTJ memory cells arranged in an array by improving the degree of freedom of layout and reducing the number of wires required for the entire memory array. It is to plan.

  The thin film magnetic memory device according to claim 1 includes a plurality of magnetic memory cells arranged in a matrix and each having one of a first resistance value and a second resistance value according to a level of stored data. A plurality of first bit lines each provided corresponding to a column of magnetic memory cells and a plurality of first bit lines provided corresponding to rows of magnetic memory cells and set to a first voltage; A plurality of magnetic memory cells corresponding to the address-selected row are electrically coupled between the first bit line and the second voltage, respectively, and a plurality of data read currents are passed through the magnetic memory cells. Read word line, first read data line for transmitting read data, and one voltage corresponding to one of the plurality of first bit lines corresponding to the address-selected column, Set the read data line voltage Comprising a read gate circuit because, depending on the voltage of the first read data lines, and a data read circuit for setting the data level of the read data.

  The thin film magnetic memory device according to claim 2 is the thin film magnetic memory device according to claim 1, wherein a pull-up for coupling the plurality of first bit lines to the first voltage at the time of data reading is performed. A circuit is further provided.

  The thin film magnetic memory device according to claim 3 is the thin film magnetic memory device according to claim 2, wherein the first bit line corresponding to the address-selected column and the pull-up circuit are provided at the time of data reading. A selection circuit for electrically coupling is further provided.

  The thin film magnetic memory device according to claim 4 is the thin film magnetic memory device according to claim 3, wherein the data write current supply supplies a data write current for writing stored data to the magnetic memory cell. A circuit, a write data line for transmitting a data write current, and a data write current supply circuit, a pull-up circuit, and a write data line in data writing and data reading, respectively. And a switch circuit for the purpose. The selection circuit includes a plurality of column selection gates respectively disposed between the write data line and the plurality of first bit lines, and one corresponding to the address-selected column of the plurality of column selection gates Is turned on both when data is written and when data is read.

  The thin film magnetic memory device according to claim 5 is the thin film magnetic memory device according to claim 1, wherein a precharge for precharging a plurality of first bit lines to a first voltage before data reading is performed. A charge circuit is further provided. A data read circuit amplifies a voltage difference between a voltage at an input node and a predetermined voltage and outputs the voltage, and a voltage on a first bit line corresponding to an address-selected column at an input node And a latch circuit for latching the output of the voltage amplifier circuit and generating read data at a predetermined timing.

  The thin film magnetic memory device according to claim 6 is the thin film magnetic memory device according to claim 1, wherein the thin film magnetic memory device is hierarchically provided with a plurality of first bit lines, and an address-selected column is provided at the time of data reading. Is further provided with a second read data line selectively coupled to the first bit line corresponding to. The read gate circuit has a current control circuit for forming a current path according to the voltage of the second read data line between the first read data line and the second voltage.

  The thin-film magnetic memory device according to claim 7 is the thin-film magnetic memory device according to claim 1, wherein the read gate circuit is provided corresponding to each column of the magnetic memory cells, and Between the read data line and the second voltage, there are a plurality of current control circuits for forming a current path corresponding to a corresponding one of the plurality of first bit lines.

  The thin film magnetic memory device according to claim 8 is the thin film magnetic memory device according to claim 1, wherein a plurality of second bit lines provided as complementary bit lines of the plurality of first bit lines are provided. And a second read data line provided as a complementary data line of the first read data line and a resistance value intermediate between the first and second resistance values, each of the first and second bit lines being It further includes a plurality of dummy memory cells coupled to any one of them and a plurality of dummy read word lines for selecting the plurality of dummy memory cells. The plurality of read word lines correspond to an address-selected row between one of the plurality of first and second bit lines set to the first voltage and the second voltage at the time of data reading. The magnetic memory cells are electrically coupled to each other, and the plurality of dummy read word lines are connected to the other one of the plurality of first and second bit lines set to the first voltage and the second voltage at the time of data reading. The dummy memory cells are electrically coupled to each other. The read gate circuit sets the voltage levels of the first and second read data lines according to the voltage level one by one corresponding to the address-selected column of the plurality of first and second bit lines. To do. The data read circuit sets the data level of the read data according to the voltage difference between the first and second read data lines.

  The thin film magnetic memory device according to claim 9 is a thin film magnetic memory device having a normal operation mode and a test mode, and includes a memory array having a plurality of magnetic memory cells arranged in a matrix. Each of the magnetic memory cells has a different resistance value depending on the level of stored data to be written when the data write magnetic field applied by the first and second data write currents is larger than a predetermined magnetic field. And a plurality of write word lines provided corresponding to the rows of the magnetic memory cells and selectively activated in accordance with a row selection result at the time of data writing, and a plurality of activated writes A write word line driver for supplying a first data write current having a current amount corresponding to the voltage level of the first control node to the word line, and a second control node at the time of data writing Voltage level A data write circuit for supplying a second data write current having a current amount according to the level and a column of the magnetic memory cell are provided corresponding to the column selection result at the time of data writing. A plurality of bit lines selectively connected to the data write circuit, and at least one of the write word line driver and the data write circuit corresponds to the first and second control nodes in the test mode. And an input terminal for setting one voltage level from the outside.

  The thin film magnetic memory device according to claim 10 is the thin film magnetic memory device according to claim 9, wherein the input terminal is electrically connected to one of the corresponding ones of the first and second control nodes in the test mode. And a reference voltage input terminal capable of inputting a predetermined voltage from the outside.

  The thin film magnetic memory device according to claim 11 is the thin film magnetic memory device according to claim 9, wherein at least one of the write word line and the data write circuit corresponds to the first and second control nodes. A reference voltage adjusting circuit for generating a reference voltage, and the reference voltage adjusting circuit includes a plurality of programs that change in a non-volatile manner from the first state to the second state in response to an external blow input. A voltage adjusting unit that sets the voltage level of the reference voltage according to the combination of the elements and the states of the program elements.

  The thin-film magnetic memory device according to claim 12 is the thin-film magnetic memory device according to claim 11, wherein the reference voltage adjustment circuit is provided corresponding to each of the plurality of program elements, and each of the test voltages is externally tested. A plurality of test gate circuits for forming an electrical connection state similar to the case where the corresponding program element transitions to the second state in response to the signal, and the input terminal corresponds to the plurality of test gate circuits. And a plurality of test terminals for inputting test signals respectively corresponding to the plurality of test gate circuits.

  14. A thin film magnetic memory device according to claim 13, comprising a memory array having a plurality of magnetic memory cells arranged in rows and columns, wherein each of the plurality of magnetic memory cells has first and second data writing. A magnetic memory including a magnetic storage unit having one of a first resistance value and a second resistance value according to a level of stored data written when a data write magnetic field applied by an electric current is larger than a predetermined magnetic field A plurality of write word lines, each provided corresponding to a row of cells and selectively activated in accordance with an address selection result for flowing a first data write current during data writing; And a plurality of bit line pairs each including a first bit line and a second bit line, each of which is provided corresponding to a column of magnetic memory cells. Each bit line The first and second bit lines are formed by using wirings formed on the first and second metal wiring layers arranged on the semiconductor substrate with the magnetic memory portion interposed therebetween. There is further provided a coupling circuit for electrically coupling the two, and the second data write current flows as a current that reciprocates between the first and second bit lines electrically coupled by the coupling circuit.

  The thin film magnetic memory device according to claim 14 is the thin film magnetic memory device according to claim 13, wherein each of the first bit lines has a wiring formed in the first metal wiring layer, Each of the second bit lines has a wiring formed in the second metal wiring layer, and the thin film magnetic memory device is selected according to the address selection result of the plurality of bit line pairs. Further includes a data write circuit for setting one end of each of the first and second bit lines included in each of the high potential state and the low potential state, and the coupling circuit corresponds to the plurality of bit line pairs. Each includes a plurality of bit line current control circuits for electrically coupling the other ends of the first and second bit lines during data writing.

  The thin film magnetic memory device according to claim 15 is the thin film magnetic memory device according to claim 13, wherein each of the first and second bit lines intersects each other in a predetermined region on the memory array. The first and second metal wiring layers are formed, and each magnetic memory cell is coupled to one of the first and second bit lines in the first metal wiring layer. The thin film magnetic memory device further includes a plurality of read word lines provided corresponding to the rows of the magnetic memory cells and selectively activated according to a row selection result at the time of data reading, A plurality of dummy memory cells having a resistance value intermediate between the second resistance values, each of which is coupled to one of the first and second bit lines, and a plurality of dummy for selecting the plurality of dummy memory cells A read word line and a data read circuit for setting a data level of read data in accordance with a voltage difference of each of the first and second bit lines corresponding to the selected column. . The plurality of read word lines and the plurality of dummy read word lines have a plurality of first and second bit lines set to a first voltage and a second voltage in order to pass a data read current according to a row selection result. The magnetic memory cell and the dummy memory cell are electrically coupled to the voltage.

  17. The thin film magnetic memory device according to claim 16, comprising a memory array having a plurality of magnetic memory cells arranged in rows and columns, wherein each of the plurality of magnetic memory cells has first and second data writing. Including magnetic storage units having different resistance values depending on the level of stored data to be written when a data write magnetic field applied by an electric current is larger than a predetermined magnetic field, and each provided corresponding to a column of magnetic memory cells Are provided corresponding to a plurality of bit lines and a row of magnetic memory cells, respectively, provided to selectively pass a first data write current during data writing. A plurality of write word lines that are selectively activated according to an address selection result to cause the second data write current to flow, and each write word line has a magnetic storage portion on the semiconductor substrate Up Thin film magnetic memory device corresponding to each write word line, including first and second sub-write word lines respectively formed in first and second metal wiring layers arranged in the direction And a coupling circuit for electrically coupling each of the first and second sub write word lines, wherein the second data write current is electrically coupled by the coupling circuit. It flows as a current that reciprocates between the first and second sub-write word lines.

  The thin film magnetic memory device according to claim 17 is the thin film magnetic memory device according to claim 16, and is provided corresponding to each of a plurality of write word lines, and a plurality of write magnetic memory devices are provided according to an address selection result. A plurality of write word drivers for setting one end of a first sub-write word line included in a corresponding one of the embedded word lines to a first voltage, and each second sub-write word One end of the line is coupled to the second voltage, and the coupling circuit includes a wiring for coupling between the other ends of the first and second sub-write word lines.

  The thin film magnetic memory device according to claim 18 is the thin film magnetic memory device according to claim 17, wherein the plurality of write word drivers are adjacent to the memory array in the row direction every predetermined number of rows. Are divided and arranged.

  The thin film magnetic memory device according to claim 19 is the thin film magnetic memory device according to claim 16, wherein one end of each of the first and second sub-write word lines is connected to the first and second voltages. The coupling circuits are respectively provided corresponding to the plurality of write word lines, and first and second included in corresponding ones of the plurality of write word lines according to the row selection result. A switch circuit for electrically coupling the other ends of the sub-write word lines is included.

  21. The thin film magnetic memory device according to claim 20, comprising a memory array including a plurality of magnetic memory cells arranged in a matrix, wherein each of the plurality of magnetic memory cells includes a first data write and a second data write. A magnetic memory unit having a resistance value that differs depending on the level of stored data to be written when a data write magnetic field applied by an electric current is larger than a predetermined magnetic field is provided corresponding to a row of magnetic memory cells. In data reading, a plurality of read word lines driven to a first voltage according to a row selection result are provided corresponding to the row, and a first data write current flows in data writing. Are provided in a direction intersecting with the plurality of write word lines corresponding to the columns of the magnetic memory cells, each of which is selectively activated according to the row selection result. Multiple combined with storage One of the plurality of bit lines selected according to the column selection result passes a data read current and a second data write current during data read and data write, respectively. Each magnetic memory cell further includes a rectifying element connected between the magnetic memory portion and the read word line.

  A thin film magnetic memory device according to a twenty-first aspect is the thin film magnetic memory device according to the twentieth aspect, wherein adjacent magnetic memory cells share one of a plurality of write word lines.

  The thin film magnetic memory device according to claim 22 is the thin film magnetic memory device according to claim 20 or 21, wherein each write word line has a larger cross-sectional area than each bit line.

  The thin-film magnetic memory device according to claim 23 is the thin-film magnetic memory device according to claim 20, wherein the plurality of write word lines are made of a material having better electromigration resistance than the plurality of bit lines. The

  25. The thin film magnetic memory device according to claim 24, comprising a memory array including a plurality of magnetic memory cells arranged in rows and columns, wherein each of the plurality of magnetic memory cells has first and second data writing. A magnetic memory unit having a resistance value that differs depending on the level of stored data to be written when a data write magnetic field applied by an electric current is larger than a predetermined magnetic field is provided corresponding to a row of magnetic memory cells. A plurality of word lines shared between magnetic memory cells adjacent to each other in the column direction, and a plurality of word lines for flowing a first data write current and a data read current in data writing and data reading, respectively. A word line driver for activating one of the word lines selected according to the row selection result and a direction crossing the plurality of word lines corresponding to the column of the magnetic memory cells. A plurality of bit lines coupled to the magnetic storage unit, and one of the plurality of bit lines selected according to the column selection result is a data read current during data read and data write. Each magnetic memory cell further includes a rectifying element connected between the magnetic memory portion and the word line.

  The thin film magnetic memory device according to claim 25 is the thin film magnetic memory device according to claim 24, wherein each word line has a larger cross-sectional area than each bit line.

  A thin film magnetic memory device according to a twenty-sixth aspect is the thin film magnetic memory device according to the twenty-fourth aspect, wherein the plurality of word lines are made of a material having higher electromigration resistance than the plurality of bit lines.

  28. A thin film magnetic memory device according to claim 27, comprising a memory array having a plurality of magnetic memory cells arranged in a matrix, wherein each of the plurality of magnetic memory cells includes a first data write and a second data write. A magnetic memory unit having a different resistance value depending on the level of stored data to be written when the data write magnetic field applied by the current is larger than a predetermined magnetic field, and the data read current passes through the magnetic memory unit during data reading And a plurality of read operations for operating the corresponding memory cell selection gate according to the address selection result at the time of data reading, each of which is provided corresponding to a row of magnetic memory cells. Address selection to provide the first data write current at the time of data writing, provided corresponding to each word line and column of magnetic memory cells A plurality of write word lines that are selectively driven to an active state according to the result and a plurality of writes for supplying a second data write current at the time of data writing, respectively, corresponding to the row Data lines and a plurality of read data lines provided corresponding to the columns and allowing a data read current to flow during data read, and adjacent magnetic memory cells include a plurality of write word lines, a plurality of write word lines, A corresponding one of at least one of the read word line, the plurality of write data lines, and the plurality of read data lines is shared.

  The thin film magnetic memory device according to claim 28 is the thin film magnetic memory device according to claim 27, wherein each of the plurality of read data lines includes each of a plurality of magnetic memory units belonging to a corresponding row, and Electrically coupled through the memory cell select gate.

  A thin-film magnetic memory device according to claim 29 is the thin-film magnetic memory device according to claim 27 or 28, wherein the adjacent magnetic memory cell is one of the corresponding write word line and write data line. One of the write word line and the write data line has a larger cross-sectional area than the other of the corresponding write word line and the write data line.

  The thin film magnetic memory device according to claim 30 is the thin film magnetic memory device according to claim 27 or 28, wherein one of the write word line and the write data line farther from the magnetic memory portion is a write It is formed of a material having higher electromigration resistance than the other of the embedded word line and the write data line.

  A thin film magnetic memory device according to a thirty-first aspect is the thin film magnetic memory device according to the twenty-seventh or twenty-eighth aspect, wherein two of the plurality of read data lines are read data line pairs during data reading. And a plurality of magnetic memory cells selected by the same read word line are respectively connected to one of the two read data lines constituting the read data line pair, and the data read current is determined by column selection. It is supplied to each of the two read data lines constituting the read data line pair corresponding to the result.

  A thin film magnetic memory device according to a thirty-second aspect is the thin film magnetic memory device according to the twenty-seventh or twenty-eighth aspect, wherein two of the plurality of write data lines are written at the time of data writing. A plurality of magnetic memory cells constituting a data line pair and selected by the same write word line are respectively connected to one of two write data lines constituting the write data line pair. The thin film magnetic memory device is a data write for setting two write data lines constituting a write data line pair selected according to an address selection result to one of a high potential state and a low potential state. The circuit further includes a control circuit and a short circuit for short-circuiting between two write data lines constituting each write data line pair at the time of data writing.

  A thin film magnetic memory device according to a thirty-third aspect is the thin film magnetic memory device according to the twenty-seventh or twenty-eighth aspect, wherein two of each of the plurality of read data lines are read data line pairs during data reading. And a plurality of magnetic memory cells selected by the same read word line are respectively connected to one of the two read data lines constituting the read data line pair, and the data read current is determined by column selection. Supplyed to each of the two read data lines constituting the read data line pair corresponding to the result, two of the plurality of write data lines are used as write data line pairs during data writing. A plurality of magnetic memory cells configured and selected by the same write word line at the time of data writing are respectively connected to one of two write data lines constituting the write data line pair. . The thin film magnetic memory device is a data write for setting two write data lines constituting a write data line pair selected according to an address selection result to one of a high potential state and a low potential state. The circuit further includes a control circuit and a short circuit for short-circuiting between two write data lines constituting each write data line pair at the time of data writing.

  The thin film magnetic memory device according to claim 34 is the thin film magnetic memory device according to claim 27 or 28, wherein the plurality of read data lines are set to the first voltage before the data read is executed, The plurality of read data lines are set to the first voltage at the time of data writing.

  36. The thin film magnetic memory device according to claim 35, comprising a memory array having a plurality of magnetic memory cells arranged in rows and columns, wherein each of the plurality of magnetic memory cells has first and second data writing. A magnetic storage unit having a different resistance value depending on the level of stored data to be written when the data write magnetic field applied by the current is larger than a predetermined magnetic field, and a data read current is passed through the storage unit during data reading A plurality of read words, each provided corresponding to a row of magnetic memory cells for operating the corresponding memory cell selection gate in accordance with an address selection result at the time of data reading Corresponding to a plurality of write data lines and columns for supplying a first data write current at the time of data writing. Each of the plurality of common lines, each of the plurality of common lines is selectively supplied with a data read current according to an address selection result during data reading, and each of the plurality of common lines is In data writing, according to the address selection result, the second data write current is selectively driven to flow the second data write current, and the second voltage different from the first voltage is shared with each other. A current control circuit that couples and disconnects the wiring in each of the data writing and data reading states; and the adjacent magnetic memory cell includes a plurality of write data lines, a plurality of read word lines, and a plurality of read data lines A corresponding one of at least one of the common wirings is shared.

  The thin-film magnetic memory device according to claim 36 is the thin-film magnetic memory device according to claim 35, wherein each of the plurality of common wirings is connected to each of the plurality of storage units belonging to the corresponding row and each memory cell selection. Electrically coupled through the gate.

  The thin-film magnetic memory device according to claim 37 is the thin-film magnetic memory device according to claim 35 or 36, wherein the adjacent magnetic memory cell is a magnetic field among the corresponding common wiring and write data line. One of the common wiring and the write data line is shared, and one of the common wiring and the write data line has a larger cross-sectional area than the other of the common wiring and the write data line.

  The thin film magnetic memory device according to claim 38 is the thin film magnetic memory device according to claim 35 or 36, wherein one of the common wiring and the write data line farther from the magnetic storage portion is a common wiring line. And, it is made of a material having higher electromigration resistance than the other of the write data lines.

  The thin-film magnetic memory device according to claim 39 is the thin-film magnetic memory device according to claim 35 or 36, wherein two of the plurality of common wirings each have a read data line pair at the time of data reading. The plurality of magnetic memory cells configured and selected by the same read word line are respectively connected to one of the two common lines constituting the read data line pair, and the data read current is determined by the column selection result. It is supplied to each of the two common wirings constituting the corresponding read data line pair.

  The thin film magnetic memory device according to claim 40 is the thin film magnetic memory device according to claim 35 or 36, wherein two of the plurality of write data lines are written at the time of data writing. A plurality of magnetic memory cells constituting a data line pair and selected by the same common wiring at the time of data writing are respectively connected to one of two write data lines constituting the write data line pair. Is done. The thin film magnetic memory device is a data write for setting two write data lines constituting a write data line pair selected according to an address selection result to one of a high potential state and a low potential state. The circuit further includes a control circuit and a short circuit for short-circuiting between two write data lines constituting each write data line pair at the time of data writing.

  The thin-film magnetic memory device according to claim 41 is the thin-film magnetic memory device according to claim 35 or 36, wherein two of the plurality of common wirings each have a read data line pair at the time of data reading. The plurality of magnetic memory cells configured and selected by the same read word line are respectively connected to one of the two common lines constituting the read data line pair, and the data read current is determined by the column selection result. Supplyed to each of the two common wires constituting the corresponding read data line pair, and two of the plurality of write data lines constitute a write data line pair at the time of data writing, A plurality of magnetic memory cells selected by the same common wiring at the time of data writing are respectively connected to one of two write data lines constituting a write data line pair. A thin film magnetic memory device writes data for setting two write data lines constituting a write data line pair selected according to a row selection result to one of a high potential state and a low potential state. The circuit further includes a control circuit and a short circuit for short-circuiting between two write data lines constituting each write data line pair at the time of data writing.

  The thin-film magnetic memory device according to claim 42 is the thin-film magnetic memory device according to claim 35 or 36, wherein the common wiring is precharged to the second voltage before the data reading is executed, At the time of loading, the common wiring that was not selected as a result of address selection is set to the second voltage.

  The thin-film magnetic memory device according to any one of claims 1, 2, 5 and 7 can execute data reading by reducing the RC constant of the data read current path without passing the data read current through the first read data line. . Therefore, the voltage change in the first bit line can be caused quickly, and the data reading can be speeded up.

  The thin-film magnetic memory device according to claim 3 can supply the data read current only to the selected first bit line according to the column selection result, thereby reducing current consumption during data read. can do.

  In the thin film magnetic memory device according to the fourth aspect, the data write system circuit can be shared and the pull-up of the first bit line at the time of data read can be executed, so that the area of the peripheral circuit is reduced. be able to.

  In the thin film magnetic memory device according to the sixth aspect, since the read gate circuit can be shared by the entire memory cell column, in addition to the effect exhibited by the thin film magnetic memory device according to the first aspect, the area of the peripheral circuit can be reduced. Can do.

  Since the thin film magnetic memory device according to claim 8 can execute data reading by the bit line pair having the folded configuration, in addition to the effect exhibited by the thin film magnetic memory device according to claim 1, an operation margin in data reading is achieved. Can be secured.

  In the thin film magnetic memory device according to claim 9 and 10, since at least one of the first and second data write currents can be set from the outside in the test mode, manufacturing variations in the magnetic characteristics of the MTJ memory cell can be reduced. It is possible to easily execute the adjustment test of the data write current amount to compensate and appropriately secure the data write margin.

  The thin film magnetic memory device according to the eleventh aspect can set the data write current in a nonvolatile manner by an external blow input in addition to the effect exhibited by the thin film magnetic memory device according to the ninth aspect.

  The thin film magnetic memory device according to claim 12 has the same state as that during the actual operation by blowing the program element in a pseudo manner during the test operation in addition to the effect exhibited by the thin film magnetic memory device according to the eleventh aspect. The data write current can be adjusted based on the above.

  15. The thin film magnetic memory device according to claim 13, wherein a data write magnetic field acting in the same direction is generated by a data write current flowing back and forth through the electrically coupled first and second bit lines. Since it can be generated in the magnetic storage unit, the data write current required for generating the data write magnetic field with the same strength can be reduced. As a result, it is possible to reduce the power consumption of the MRAM device, improve the operation reliability by reducing the current density of the bit line, and reduce the magnetic field noise during data writing.

  The thin film magnetic memory device according to claim 15, wherein the magnetic memory cells are equally coupled to the first and second bit lines constituting the bit line pair to balance the RC load between the two, and the folded type Data reading based on the bit line configuration can be executed. As a result, in addition to the effect exhibited by the thin film magnetic memory device according to the thirteenth aspect, the stability of the data read operation can be improved.

  18. The thin film magnetic memory device according to claim 16 or 17, wherein the data write acting in the same direction by the data write current flowing back and forth through the electrically coupled first and second sub-write word lines. Since the embedded magnetic field can be generated in the magnetic storage unit, the data write current required for generating the data write magnetic field having the same strength can be reduced. As a result, it is possible to reduce the power consumption of the MRAM device, improve the operation reliability by reducing the current density of the bit line, and reduce the magnetic field noise during data writing.

  In the thin film magnetic memory device according to claim 18, since the write word driver for supplying the data write current can be divided and arranged on both sides of the memory array, the effect obtained by the thin film magnetic memory device according to claim 17 is achieved. In addition, the pitch constraint in the column direction can be relaxed, and the area of the memory array and its peripheral circuits can be reduced.

  The thin film magnetic memory device according to claim 19 rapidly changes the voltage of the write word line when the thin film magnetic memory device according to claim 16 shifts to a non-selection state or the like in addition to the effect exhibited by the thin film magnetic memory device according to claim 16. be able to.

  The thin film magnetic memory device according to claim 20 can maintain the off state of the rectifying element reliably in the non-selected magnetic memory cell in the magnetic memory cell that is advantageous for high integration using the rectifying element. As a result, it is possible to achieve both high integration and ensuring operation reliability.

  In the thin film magnetic memory device according to the twenty-first aspect, the number of write word lines arranged in the entire memory array can be reduced, and magnetic memory cells can be arranged with high integration. As a result, in addition to the effect exhibited by the thin film magnetic memory device according to the twentieth aspect, the chip area of the MRAM device can be reduced.

  27. The thin film magnetic memory device according to claim 22, 23, 25, and 26 is a write word line that is arranged farther from the magnetic memory portion and needs to pass a large current among wirings for passing a data write current. The occurrence of electromigration can be suppressed and the operational reliability of the MRAM device can be improved.

  25. The thin film magnetic memory device according to claim 24, wherein magnetic memory cells that are advantageous for high integration using rectifying elements are arranged by further reducing the number of word lines arranged in the entire memory array. it can. As a result, the chip area of the MRAM device can be reduced.

  29. The thin film magnetic memory device according to claim 27, wherein the read word line and the write data line are arranged corresponding to the row and column of the magnetic memory cell, respectively, and the read word line and the write word line are selected. The layout freedom can be improved by disposing the circuit for driving automatically. Further, at least one of the write word line, the read word line, the write data line, and the read data line can be shared between adjacent memory cells, thereby reducing the wiring pitch in the memory array. As a result, the degree of integration of the MRAM device can be increased.

  In the thin film magnetic memory device according to claim 28, since only the memory portion of the magnetic memory cell to be read is coupled to the read data line, in addition to the effect exhibited by the thin film magnetic memory device according to claim 27, As a result, the capacity of the read data line can be reduced and the data read speed can be increased.

  30. The thin film magnetic memory device according to claim 29, wherein, among the two types of wirings through which the data write current flows, it is necessary to flow a larger data write current. A cross-sectional area can be secured. As a result, in addition to the effect exhibited by the thin film magnetic memory device according to claim 27 or 28, the electromigration resistance of the wiring through which the data write current flows can be improved, and the operation reliability can be improved.

  In the thin film magnetic memory device according to the thirty-third aspect, one of the two types of wirings through which the data write current flows needs to flow a larger data write current is formed of a material having high electromigration resistance. As a result, in addition to the effect exhibited by the thin film magnetic memory device according to claim 27 or 28, the reliability of operation can be improved.

  The thin film magnetic memory device according to claim 31 performs data reading using two read data lines forming a pair. In addition to the effect exhibited by the thin film magnetic memory device according to claim 27 or 28, An operation margin at the time of data reading can be secured.

  The thin film magnetic memory device according to the thirty-second aspect performs data writing by using two write data lines that form a pair, so that the thin film magnetic memory device according to the thirty-seventh or twenty-eighth effect is effective. In addition, it is possible to secure an operation margin and reduce magnetic field noise during data writing.

  The thin film magnetic memory device according to claim 33 performs data reading and data writing using two read data lines and a write data line forming a pair, respectively, so that the thin film according to claim 27 or 28 In addition to the effects exhibited by the magnetic memory device, it is possible to secure an operation margin during data reading and data writing and to reduce data writing noise.

  In the thin film magnetic memory device according to the thirty-fourth aspect, it is not necessary to start a new precharge operation for the read data line before data reading. Therefore, the precharge operation can be made efficient, and the data reading can be speeded up.

  The thin film magnetic memory device according to the thirty-fifth aspect can reduce the number of wirings by sharing the function of the read data line at the time of data reading and the function of the write word line at the time of data writing to the common wiring. In addition, a circuit for selectively driving each of the common wiring lines functioning as a read word line and a write word line at the time of data writing can be arranged independently to improve the flexibility of layout. it can. Furthermore, at least one of the read word line, the write data line, and the common data line is shared between adjacent memory cells, so that the wiring pitch in the memory array can be reduced. As a result, the degree of integration of the MRAM device can be increased.

  36. The thin film magnetic memory device according to claim 36, wherein only the memory portion of the magnetic memory cell that is the target of data reading is coupled to a common wiring that functions as a read data line. In addition to the effects produced by the above, it is possible to reduce the load capacity of the common wiring at the time of data reading and to speed up data reading.

  In the thin film magnetic memory device according to claim 37, among the two types of wirings through which the data write current flows, it is necessary to flow a larger data write current. A cross-sectional area can be secured. As a result, in addition to the effects exhibited by the thin film magnetic memory device according to the thirty-fifth or thirty-sixth aspect, the electromigration resistance of the wiring through which the data write current flows can be improved, and the operation reliability can be improved.

  In the thin film magnetic memory device according to the thirty-eighth aspect, one of the two kinds of wirings through which the data write current flows is made of a material having a high electromigration resistance. As a result, in addition to the effect exhibited by the thin film magnetic memory device according to the thirty-fifth or thirty-sixth aspect, operation reliability can be improved.

  In the thin film magnetic memory device according to claim 39, since data reading is performed using two common wires forming a pair, in addition to the effect exhibited by the thin film magnetic memory device according to claim 35 or 36, An operation margin at the time of data reading can be secured.

  The thin film magnetic memory device according to claim 40 performs data writing using two write data lines forming a pair, so that the thin film magnetic memory device according to claim 35 or 36 exhibits the effect In addition, it is possible to secure an operation margin and reduce magnetic field noise during data writing.

  The thin-film magnetic memory device according to claim 41 executes data reading and data writing using two common wires and a write data line forming a pair, respectively, so that the thin-film magnetic memory device according to claim 35 or 36 is used. In addition to the effects achieved by the body storage device, it is possible to secure an operation margin during data reading and data writing and to reduce data writing noise.

  In the thin film magnetic memory device according to the forty-second aspect, it is not necessary to start a new precharge operation for the common wiring before data reading. Therefore, the precharge operation can be made efficient, and the data reading can be speeded up.

It is a schematic block diagram which shows the whole structure of the MRAM device 1 according to Embodiment 1 of this invention. 2 is a diagram for illustrating a configuration according to a first embodiment of memory array 10 and its peripheral circuits. FIG. FIG. 3 is a circuit diagram showing a configuration of a data write circuit 51a shown in FIG. FIG. 3 is a circuit diagram showing a configuration of a data read circuit 55a shown in FIG. 5 is a timing chart for illustrating data read and data write operations in the MRAM device according to the first embodiment. FIG. 11 is a diagram for describing a configuration according to a first modification of the first embodiment of memory array 10 and its peripheral circuits. FIG. 7 is a circuit diagram showing a configuration of a data write circuit 51b shown in FIG. FIG. 7 is a circuit diagram showing a configuration of a data read circuit 55b shown in FIG. 10 is a timing chart for illustrating data reading and data writing operations in the MRAM device according to the first modification of the first embodiment. It is a figure for demonstrating the structure according to the modification 2 of Embodiment 1 of the memory array 10 and its peripheral circuit. It is a figure for demonstrating the structure according to the modification 3 of Embodiment 1 of the memory array 10 and its peripheral circuit. FIG. 11 is a circuit diagram showing a configuration of a data write circuit according to a second embodiment. FIG. 6 is a circuit diagram showing a configuration example of a word line driver according to a second embodiment. FIG. 11 is a circuit diagram showing a configuration of a data write current adjustment circuit 230 according to a modification of the second embodiment. FIG. 3 is a diagram for explaining a configuration of a memory array 10 and its peripheral circuits in an MRAM device that executes data reading without using a read gate. It is a block diagram illustrating the arrangement of bit lines according to the third embodiment of the present invention. FIG. 11 is a structural diagram showing a first arrangement example of bit lines according to the third embodiment. FIG. 11 is a structural diagram showing a second arrangement example of bit lines according to the third embodiment. FIG. 10 is a conceptual diagram illustrating the arrangement of bit lines according to a first modification of the third embodiment. FIG. 16 is a structural diagram illustrating an arrangement of write word lines WWL according to a second modification of the third embodiment. It is a conceptual diagram explaining the coupling | bonding between the sub word lines which form the same write word line. FIG. 25 is a diagram for explaining the arrangement of write word lines according to a third modification of the third embodiment. It is a figure explaining arrangement | positioning of the write word line according to the modification 4 of Embodiment 3. FIG. FIG. 22 is a diagram for explaining the arrangement of write word lines according to a fifth modification of the third embodiment. FIG. 10 shows a structure of an MTJ memory cell according to the fourth embodiment. FIG. 26 is a structural diagram when the MTJ memory cell shown in FIG. 25 is arranged on a semiconductor substrate. FIG. 26 is a timing chart illustrating a read operation and a write operation for the MTJ memory cell shown in FIG. 25. FIG. 26 is a conceptual diagram showing a configuration of a memory array in which MTJ memory cells shown in FIG. 25 are arranged in a matrix. FIG. 3 is a conceptual diagram showing a configuration of a memory array formed by MTJ memory cells arranged in a matrix while sharing a write word line WWL. FIG. 16 is a conceptual diagram showing an arrangement according to a modification of the fourth embodiment of the MTJ memory cell. FIG. 10 is a schematic block diagram showing an overall configuration of an MRAM device 2 according to a fifth embodiment. FIG. 10 is a circuit diagram showing a connection mode of MTJ memory cells according to a fifth embodiment. FIG. 16 is a timing chart for describing data writing and data reading with respect to an MTJ memory cell according to the fifth embodiment. FIG. 16 is a structural diagram illustrating an arrangement of MTJ memory cells according to a fifth embodiment. FIG. 10 is a diagram for illustrating a configuration according to a fifth embodiment of memory array 10 and its peripheral circuits. It is a circuit diagram showing a configuration of a data read circuit 55e. It is a figure for demonstrating the structure according to the modification 1 of Embodiment 5 of the memory array 10 and its peripheral circuit. It is a figure for demonstrating the structure according to the modification 2 of Embodiment 5 of the memory array 10 and its peripheral circuit. It is a figure for demonstrating the structure according to the modification 3 of Embodiment 5 of the memory array 10 and its peripheral circuit. It is a figure for demonstrating the structure according to the modification 4 of Embodiment 5 of the memory array 10 and its peripheral circuit. It is a figure for demonstrating the structure according to the modification 5 of Embodiment 5 of the memory array 10 and its peripheral circuit. FIG. 23 is a circuit diagram showing a connection mode of MTJ memory cells according to a sixth embodiment. FIG. 17 is a structural diagram illustrating an arrangement of MTJ memory cells according to a sixth embodiment. It is a figure for demonstrating the structure according to Embodiment 6 of the memory array 10 and its peripheral circuit. It is a figure for demonstrating the structure according to the modification 1 of Embodiment 6 of the memory array 10 and its peripheral circuit. It is a figure for demonstrating the structure according to the modification 2 of Embodiment 6 of the memory array 10 and its peripheral circuit. It is a figure for demonstrating the structure according to the modification 3 of Embodiment 6 of the memory array 10 and its peripheral circuit. It is a figure for demonstrating the structure according to the modification 4 of Embodiment 6 of the memory array 10 and its peripheral circuit. It is a figure for demonstrating the structure according to the modification 5 of Embodiment 6 of the memory array 10 and its peripheral circuit. FIG. 17 is a circuit diagram showing a connection mode of MTJ memory cells according to a seventh embodiment. FIG. 17 is a structural diagram showing an arrangement of MTJ memory cells according to a seventh embodiment. It is a figure for demonstrating the structure according to Embodiment 7 of the memory array 10 and its peripheral circuit. It is a figure for demonstrating the structure according to the modification 1 of Embodiment 7 of the memory array 10 and its peripheral circuit. It is a figure for demonstrating the structure according to the modification 2 of Embodiment 7 of the memory array 10 and its peripheral circuit. It is a figure for demonstrating the structure according to the modification 3 of Embodiment 7 of the memory array 10 and its peripheral circuit. It is a figure for demonstrating the structure according to the modification 4 of Embodiment 7 of the memory array 10 and its peripheral circuit. It is a figure for demonstrating the structure according to the modification 5 of Embodiment 7 of the memory array 10 and its peripheral circuit. FIG. 20 is a circuit diagram showing a connection mode of MTJ memory cells according to an eighth embodiment. FIG. 20 is a structural diagram showing an arrangement of MJT memory cells according to an eighth embodiment. It is a figure for demonstrating the structure according to Embodiment 8 of the memory array 10 and its peripheral circuit. It is a figure for demonstrating the structure according to the modification 1 of Embodiment 8 of the memory array 10 and its peripheral circuit. It is a figure for demonstrating the structure according to the modification 2 of Embodiment 8 of the memory array 10 and its peripheral circuit. It is a figure for demonstrating the structure according to the modification 3 of Embodiment 8 of the memory array 10 and its peripheral circuit. It is a figure for demonstrating the structure according to the modification 4 of Embodiment 8 of the memory array 10 and its peripheral circuit. It is a figure for demonstrating the structure according to the modification 5 of Embodiment 8 of the memory array 10 and its peripheral circuit. FIG. 30 is a circuit diagram showing a connection mode of MTJ memory cells according to the ninth embodiment. FIG. 29 is a timing chart for describing data writing and data reading with respect to an MTJ memory cell according to the ninth embodiment. FIG. 30 is a structural diagram illustrating an arrangement of MTJ memory cells according to a ninth embodiment. It is a figure for demonstrating the structure according to Embodiment 9 of the memory array 10 and its peripheral circuit. It is a figure for demonstrating the structure according to the modification 1 of Embodiment 9 of the memory array 10 and its peripheral circuit. It is a figure for demonstrating the structure according to the modification 2 of Embodiment 9 of the memory array 10 and its peripheral circuit. It is a figure for demonstrating the structure according to the modification 3 of Embodiment 9 of the memory array 10 and its peripheral circuit. It is a figure for demonstrating the structure according to the modification 4 of Embodiment 9 of the memory array 10 and its peripheral circuit. It is a figure for demonstrating the structure according to the modification 5 of Embodiment 9 of the memory array 10 and its peripheral circuit. FIG. 32 is a circuit diagram showing a connection mode of MTJ memory cells according to the tenth embodiment. FIG. 22 is a configuration diagram showing an arrangement of MTJ memory cells according to a tenth embodiment. It is a figure for demonstrating the structure according to Embodiment 10 of the memory array 10 and its peripheral circuit. It is a figure for demonstrating the structure according to the modification 1 of Embodiment 10 of the memory array 10 and its peripheral circuit. It is a figure for demonstrating the structure according to the modification 2 of Embodiment 10 of the memory array 10 and its peripheral circuit. It is a figure for demonstrating the structure according to the modification 3 of Embodiment 10 of the memory array 10 and its peripheral circuit. It is a figure for demonstrating the structure according to the modification 4 of Embodiment 10 of the memory array 10 and its peripheral circuit. It is a figure for demonstrating the structure according to the modification 5 of Embodiment 10 of the memory array and its peripheral circuit. It is the schematic which shows the structure of the memory cell which has a magnetic tunnel junction part. It is a conceptual diagram explaining the data read-out operation | movement from an MTJ memory cell. It is a conceptual diagram explaining the data write-in operation | movement with respect to an MTJ memory cell. It is a conceptual diagram explaining the relationship between the direction of a data write current at the time of data writing, and a magnetic field direction. It is a conceptual diagram which shows the MTJ memory cell integratedly arranged by the matrix form. 2 is a structural diagram of an MTJ memory cell disposed on a semiconductor substrate. FIG. It is a conceptual diagram for demonstrating the influence which manufacture variation has on a data write margin. It is the schematic which shows the structure of the MTJ memory cell using a diode. FIG. 93 is a structural diagram when the MTJ memory cell shown in FIG. 90 is arranged on a semiconductor substrate.

Hereinafter, embodiments of the present invention will be described in detail with reference to the drawings.
[Embodiment 1]
FIG. 1 is a schematic block diagram showing an overall configuration of an MRAM device 1 according to the first embodiment of the present invention.

  Referring to FIG. 1, MRAM device 1 performs random access in response to external control signal CMD and address signal ADD, and executes input of write data DIN and output of read data DOUT.

  The MRAM device 1 includes a control circuit 5 that controls the overall operation of the MRAM device 1 in response to a control signal CMD, and a memory array 10 having a plurality of MTJ memory cells arranged in a matrix of n rows × m columns. Prepare. Although the configuration of the memory array 10 will be described in detail later, a plurality of write word lines WWL and read word lines RWL are arranged corresponding to the rows of MTJ memory cells, respectively. In addition, a bit line pair configured in a folded type provided corresponding to each column of MTJ memory cells is arranged. The bit line pair is constituted by bit lines BL and / BL. In the following, a set of bit lines BL and / BL is also collectively referred to as a bit line pair BLP.

  The MRAM device 1 further includes a row decoder 20 that performs row selection in the memory array 10 according to the row address RA indicated by the address signal ADD, and a column in the memory array 10 according to the column address CA indicated by the address signal ADD. A column decoder 25 for performing selection, a word line driver 30 for selectively activating read word line RWL and write word line WWL based on a row selection result of row decoder 20, and a write word at the time of data writing A word line current control circuit 40 for supplying a data write current to the line WWL, and a read / write control circuit 50 for supplying a data write current ± Iw and a sense current Is during data reading and data writing, 60.

  FIG. 2 is a diagram for explaining in detail the configuration of memory array 10 and its peripheral circuits according to the first embodiment.

  Referring to FIG. 2, memory array 10 includes MTJ memory cells MC having the configuration shown in FIG. 83 arranged in n rows × m columns (n, m: natural numbers). Read word lines RWL1 to RWLn and write word lines WWL1 to WWLn are provided corresponding to MTJ memory cell rows (hereinafter also simply referred to as “memory cell rows”). Corresponding to a column of MTJ memory cells (hereinafter, also simply referred to as “memory cell column”), bit lines BL1, / BL1 to BLm, / BLm that respectively constitute bit line pairs BLP1 to BLPm are provided.

  The MTJ memory cell MC is connected to one of the bit lines BL and / BL for each row. For example, the MTJ memory cell belonging to the first memory cell column will be described. The MTJ memory cell in the first row is coupled to the bit line / BL1, and the MTJ memory cell in the second row is connected to the bit line BL1. Combined with. Similarly, each of the MTJ memory cells is connected to one of the bit line pairs / BL1 to / BLm in the odd-numbered row, and is connected to the other bit lines BL1 to BLm in the even-numbered row.

  Memory array 10 further includes a plurality of dummy memory cells DMC coupled to bit lines BL1, / BL1 to BLm, / BLm, respectively. Dummy memory cell DMC is coupled to one of dummy read word lines DRWL1 and DRWL2 and arranged in 2 rows × m columns. Dummy memory cells coupled to dummy read word line DRWL1 are coupled to bit lines BL1, BL2-BLm, respectively. On the other hand, the remaining dummy memory cells coupled to dummy read word line DRWL2 are coupled to bit lines / BL1, / BL2- / BLm, respectively.

  As already described, the resistance value of the MTJ memory cell MC varies depending on the level of stored data. Here, when the resistance value of the MTJ memory cell MC when storing the H level data is Rh and the resistance value when storing the L level data is Rl, the resistance value Rd of the dummy memory cell DMC is Rl and Rh. Is set to an intermediate value. In the embodiment of the present invention, it is assumed that Rl <Rh.

  In the following, when the write word line, the read word line, the dummy read word line, the bit line and the bit line pair are collectively expressed, the symbols WWL, RWL, DRWL, BL (/ BL) and BLP are used. In order to indicate specific write word lines, read word lines, bit lines and bit line pairs, subscripts are added to these symbols, and RWL1, WWL1, BL1 (/ BL1), BLP1 It shall be written as

  Write word lines WWL 1 to WWLn are coupled to ground voltage Vss by word line current control circuit 40. As a result, the data write current Ip is supplied to the write word line WWL activated by the word line driver 30 to the selected state (high voltage state: power supply voltage Vcc).

  Hereinafter, the high voltage state (power supply voltage Vcc) and the low voltage state (ground voltage Vss) of the signal line are also referred to as H level and L level, respectively.

  Corresponding to the memory cell columns, write column selection lines WCSL1 to WCSLm are arranged for executing column selection at the time of data writing. Similarly, corresponding to memory cell columns, read column selection lines RCSL1 to RCSLm are provided for executing column selection at the time of data reading.

  Column decoder 25 activates one of write column selection lines WCSL1 to WCSLm to a selected state (H level) at the time of data writing in accordance with the decoding result of column address CA, that is, the column selection result. At the time of data reading, column decoder 25 activates one of read column selection lines RCSL1 to RCSLm to a selected state (H level) according to the column selection result.

  Furthermore, a write data bus pair WDBP for transmitting write data and an RDBP for transmitting read data are arranged independently. Write data bus pair WDBP includes write data buses WDB and / WDB. Similarly, read data bus pair RDBP includes read data buses RDB and / RDB.

  Read / write control circuit 50 includes write column select gates WCSG1 to WCSGm, read column select gates RCSG1 to RCSGm, and a read column provided corresponding to data write circuit 51a, data read circuit 55a, and memory cell columns, respectively. Gates RG1 to RGm.

  One of write column selection gates WCSG1 to WCSGm is turned on according to the column selection result of column decoder 25, and write data buses WDB and / WDB constituting write data bus pair WDBP are connected to corresponding bit line BL. Each binds to / BL.

  For example, write column select gate WCSG1 is electrically coupled between an N-type MOS transistor coupled between write data bus WDB and bit line BL1, and electrically coupled between write data bus / WDB and bit line / BL1. And an N-type MOS transistor. These MOS transistors are turned on / off according to the voltage level of write column select line WCSL1. That is, when write column select line WCSL1 is activated to a selected state (H level), write column select gate WCSG1 electrically couples write data buses WDB and / WDB to bit lines BL1 and / BL1, respectively. To do. The write column selection gates WCSG2 to WCSGm provided corresponding to the other memory cell columns have the same configuration.

  Data write circuit 51a operates in response to control signal WE activated (to H level) during data writing and control signal RE activated (to H level) during data reading.

  In the following, the read column selection lines RCL1 to RCSLm, the write column selection lines WCSL1 to WCSLm, the read column selection gates RCSG1 to RCSGm, the write column selection gates WCSG1 to WCSGm, and the read gates RG1 to RGm are collectively expressed. In this case, the symbols RCSL, WCSL, RCSG, WCSG, and RG are used.

FIG. 3 is a circuit diagram showing a configuration of data write circuit 51a.
Referring to FIG. 3, data write circuit 51a has a data write current supply circuit 52 for supplying data write current ± Iw and a bit line BL, / BL for pulling up at the time of data reading. And a pull-up circuit 53.

  Data write current supply circuit 52 includes a P-type MOS transistor 151 for supplying a constant current to internal node Nw 0, a P-type MOS transistor 152 constituting a current mirror circuit for controlling the passing current of transistor 151, and a current Source 153.

  Data write current supply circuit 52 further includes inverters 154, 155, and 156 that operate by receiving an operation current supplied from internal node Nw0. Inverter 154 inverts the voltage level of write data DIN and transmits it to internal node Nw1. Inverter 155 inverts the voltage level of write data DIN and transmits it to the input node of inverter 156. Inverter 156 inverts the output of inverter 155 and transmits the result to internal node Nw2. Therefore, data write circuit 51a sets the voltages of internal nodes Nw1 and Nw2 to one of power supply voltage Vcc and ground voltage Vss according to the voltage level of write data DIN.

  Pull-up circuit 53 has P-type MOS transistors 157 and 158 electrically coupled between power supply voltage Vcc and nodes Np1 and Np2, respectively. / RE which is an inverted signal of the control signal RE is input to the gates of the transistors 157 and 158.

  Data write circuit 51a further selects switch SW1a for selectively coupling one of nodes Nw1 and Np1 with write data bus WDB, and selects one of nodes Nw2 and Np2 with write data bus / WDB. And a switch SW1b for coupling them. The switches SW1a and SW1b operate according to the control signal RWS.

  Switches SW1a and SW1b connect nodes Nw1 and Nw2 to write data buses WDB and / WDB, respectively, at the time of data writing. As a result, at the time of data writing, in order to flow data write current ± Iw, the voltages of write data buses WDB and / WDB are either one of power supply voltage Vcc or ground voltage Vss according to the data level of the write data. Set to

  On the other hand, at the time of data reading, switches SW1a and SW1b couple nodes Np1 and Np2 to write data buses WDB and / WDB, respectively. Thereby, at the time of data reading, each of write data buses WDB and / WDB is pulled up to power supply voltage Vcc by pull-up circuit 53.

  Referring to FIG. 2 again, each of read column select gates RCSG1 to RCSGm and read gates RG1 to RGm respectively arranged corresponding to the memory cell column has the same configuration, and therefore bit line BL1. , / BL1 corresponding to the configuration of read column selection gate RCSG1 and read gate RG1 will be described representatively.

  Read column select gate RCSG1 and read gate RG1 are coupled in series between read data buses RDB, / RDB and ground voltage Vss.

  Read column select gate RCSG1 includes an N-type MOS transistor coupled between read data bus RDB and node N1a, and an N-type MOS transistor electrically coupled between read data bus / RDB and node N1b. Have These MOS transistors are turned on / off according to the voltage of the read column selection line RCSL1. That is, when read column select line RCSL1 is activated to a selected state (H level), read column select gate RCSG1 electrically couples read data buses RDB and / RDB to nodes N1a and N1b, respectively.

  Read gate RG1 has N-type MOS transistors Q11 and Q12 electrically coupled between nodes N1a and N1b and ground voltage Vss, respectively. Transistors Q1 and Q2 have their gates coupled to bit lines / BL1 and BL1, respectively. Therefore, the voltages of nodes N1a and N1b change according to the voltages of bit lines / BL1 and BL1, respectively.

  Specifically, when the voltage of the bit line BL1 is higher than the voltage of the bit line / BL1, the node N1b is more strongly pulled to the ground voltage Vss by the transistor Q12, so that the voltage of the node N1a is the voltage of the node N1b. Higher than. On the other hand, when the voltage of the bit line BL1 is lower than the voltage of the bit line / BL1, the voltage of the node N1b becomes higher than the voltage of the node N1a.

  The voltage difference between nodes N1a and N1b generated in this way is transmitted to the voltage difference between read data buses RDB and / RDB via read column select gate RCSG1. Data read circuit 55a amplifies the voltage difference between read data buses RDB and / RDB constituting read data bus pair RDBP to generate read data DOUT.

FIG. 4 is a circuit diagram showing a configuration of data read circuit 55a.
Referring to FIG. 4, data read circuit 55 a has a differential amplifier 56. Differential amplifier 56 receives voltages on read data buses RDB and / RDB, amplifies the voltage difference between the two, and generates read data Dout.

  Referring to FIG. 2 again, read / write control circuit 60 has equalize transistors 62-1 to 62-m which are turned on / off in response to bit line equalize signal BLEQ. Equalize transistors 62-1 to 62-m are provided corresponding to the memory cell columns, respectively. For example, equalize transistor 62-1 is provided corresponding to the first memory cell column, and electrically connects bit lines BL1 and / BL1 in response to activation (H level) of bit line equalize signal BLEQ. To join.

  Similarly, equalize transistors 62-2 to 62-m respectively provided corresponding to the other memory cell columns respond to the activation of the bit line equalize signal BLEQ in the corresponding memory cell column. The bit lines BL and / BL are electrically coupled.

  Read / write control circuit 60 further includes precharge transistors 64-1a, 64-1b to 64-ma provided between bit lines BL1, / BL1 to bit lines BLm, / BLm and ground voltage Vss, respectively. 64-mb. The precharge transistors 64-1a, 64-1b to 64-ma, 64-mb are turned on in response to activation of the bit line precharge signal BLPR, whereby the bit lines BL1, / BL1 to bit lines BLm, / BLm is precharged to the ground voltage Vss.

  In the following, equalize transistors 62-1 to 62-m and precharge transistors 64-1a, 64-1b to 64-ma, 64-mb are collectively referred to as equalize transistor 62 and precharge transistor 64, respectively. .

  The bit line equalize signal BLEQ generated by the control circuit 5 constitutes each bit line pair BLP provided in a folded type except during the data read operation during the standby period of the MRAM device 1 and the active period of the MRAM device 1. In order to short-circuit bit lines BL and / BL, they are activated to H level.

  On the other hand, in the data read operation during the active period of the MRAM device, bit line equalize signal BLEQ is inactivated to L level. In response to this, in each memory cell column, the bit lines BL and / BL constituting each bit line pair BLP are blocked.

  Similarly, the bit line precharge signal BLPR is generated by the control circuit 5. Bit line precharge signal BLPR is activated to H level in an active period of MRAM device 1 at least in a predetermined period before data reading is executed. On the other hand, in the data read operation during the active period of MRAM device 1, bit line precharge signal BLPR is inactivated to L level and precharge transistor 64 is turned off.

Next, operations during data writing and data reading will be described.
FIG. 5 is a timing chart for explaining data read and data write operations in the MRAM device according to the first embodiment.

First, the operation at the time of data writing will be described with reference to FIG.
The write column selection line WCSL corresponding to the column selection result is activated to the selected state (H level), and the corresponding write column selection gate WCSG is turned on. In response, bit lines BL and / BL corresponding to the column selection result are coupled to write data buses WDB and / WDB, respectively.

  Further, at the time of data writing, equalize transistor 62 is turned on to short-circuit between bit lines BL and / BL.

  As already described, data write circuit 51a sets the voltages of write data buses WDB and / WDB to either one of power supply voltage Vcc and ground voltage Vss. For example, when the data level of write data DIN is L level, the voltages of nodes Nw2 and Nw1 shown in FIG. 3 are set to power supply voltage Vcc and ground voltage Vss, respectively. A data write current -Iw for writing L level data is applied. Data write current -Iw is supplied to bit line BL via write column select gate WCSG.

  The data write current −Iw passed through the bit line BL is turned back by the equalize transistor 62. As a result, data write current + Iw in the opposite direction flows through the other bit line / BL. Data write current + Iw flowing through bit line / BL is transmitted to write data bus / WDB via write column select gate WCSG.

  Furthermore, any one of the write word lines WWL is activated to the selected state (H level) according to the row selection result, and the data write current Ip flows. Therefore, in the memory cell column corresponding to the column selection result, data writing is performed on the MTJ memory cell corresponding to the selected write word line WWL. At this time, L level data is written into memory cell MC coupled to bit line BL, and H level data is written into memory cell MC coupled to bit line / BL.

  When the data level of write data DIN is H level, the voltage setting of nodes Nw1 and Nw2 is opposite to the above case, and data writing in the opposite direction to bit line BL and / BL is performed. Current is supplied and data writing is executed. In this manner, data write current ± Iw having a direction corresponding to the data level of write data DIN is supplied to bit lines BL and / BL.

  At the time of data writing, read word line RWL is maintained in a non-selected state (L level).

  Further, for example, by activating bit line precharge signal BLPR (to H level) even during data writing, the voltages of bit lines BL and / BL during data writing become the precharge voltage during data reading. The ground voltage Vss corresponding to the level is set.

  Similarly, read data buses RDB and / RDB are set to power supply voltage Vcc corresponding to the precharge voltage at the time of data reading. As described above, the data read operation is performed by matching the voltage at the time of data writing to the bit lines BL and / BL corresponding to the non-selected columns and the read data buses RDB and / RDB with the precharge voltage at the time of data reading. It becomes unnecessary to execute a new precharge operation before, and the data read operation can be speeded up.

Next, the operation at the time of data reading will be described.
Before data reading, read data buses RDB, / RDB and bit lines BL, / BL are precharged to power supply voltage Vcc and ground voltage Vss, respectively.

  At the time of data reading, each of write data buses WDB and / WDB is pulled up to power supply voltage Vcc by pull-up circuit 53. Furthermore, both the corresponding read column selection line RCSL and write column selection line WCSL are activated to the selected state (H level) according to the column selection result.

  Thus, write data buses WDB and / WDB are electrically coupled to bit lines BL and / BL corresponding to the selected column via write column selection gate WCSG. Therefore, at the time of data reading, each of bit lines BL and / BL corresponding to the selected memory cell column is pulled up to power supply voltage Vcc.

  Any one of read word lines RWL is activated to a selected state (H level) according to a row selection result, and corresponding memory cell MC is coupled to one of bit lines BL and / BL.

  Further, one of dummy read word lines DRWL1 and DRWL2 is activated, and the other of bit lines BL and / BL, which is not coupled to MTJ memory cell MC, is coupled to dummy memory cell DMC.

  When an odd row is selected according to the row selection result and bit line / BL and MTJ memory cell MC are coupled, dummy read word line DRWL1 is activated and bit line BL and dummy memory cell DMC are activated. And are combined. On the other hand, when the even-numbered row is selected according to the row selection result and the bit line BL and the MTJ memory cell MC are coupled, the dummy read word line DRWL2 is activated, and the bit line / BL and the dummy row are activated. Memory cell DMC is coupled.

  In the selected MTJ memory cell MC, when the access transistor ATR is turned on, a sense current Is flows between the pulled-up bit line BL or / BL, memory cell MC, and ground voltage Vss. Therefore, voltage change ΔV1 corresponding to the stored data level occurs on one of bit lines BL and / BL coupled to the MTJ memory cell. FIG. 5 shows, as an example, a voltage change when MTJ memory cell MC that is a target of data reading holds H level data, that is, when MTJ memory cell MC has resistance value Rh.

  As already described, the resistance value Rd of the dummy memory cell DMC is set to an intermediate value between the resistance values Rh and Rl of the MTJ memory cell MC. Therefore, voltage change ΔVm corresponding to intermediate resistance value Rd occurs on the other of bit lines BL and / BL coupled to dummy memory cell DMC.

  Therefore, the relative relationship between the voltages of bit lines BL and / BL constituting bit line pair BLP corresponding to the selected memory cell column changes according to the level of the read storage data. Due to the voltage difference between the bit lines BL and / BL, the read data buses RDB and / RDB are driven through the read gate.

  That is, when the voltage of the bit line BL is higher than the voltage of the bit line / BL, the read data bus / RDB is driven more strongly to the ground voltage Vss side than the read data bus RDB by the read gate RG. (Voltage change ΔVb1> ΔVbm in FIG. 5). The voltage difference between read data buses RDB and / RDB generated in this way can be amplified by data read circuit 55a, and H level read data Dout can be output.

  On the other hand, when the MTJ memory cell MC from which data is read holds L level data, that is, when the voltage of the bit line / BL is higher than the voltage of the bit line BL, the read data is read by the read gate RG. The bus RDB is driven to the ground voltage Vss side more strongly than the read data bus / RDB. The voltage difference between read data buses RDB and / RDB generated in this way can be amplified by data read circuit 52 to output read data Dout at L level.

  In this manner, by adopting a configuration in which read data buses RDB and / RDB are driven via read gate RG, data reading can be executed without causing a sense current to flow through read data buses RDB and / RDB. As a result, the RC load on the sense current path can be reduced, and a voltage change necessary for data reading can be quickly generated in bit lines BL and / BL. As a result, data can be read at high speed, and access to the MRAM device can be speeded up.

  In addition, since the pulled-up write data buses WDB and / WDB are coupled to the bit lines BL and / BL via the write column selection gate WCSG, the sense current Is is supplied. Sense current Is can be supplied only to bit lines BL and / BL corresponding to the memory cell column to be. Thereby, unnecessary power consumption at the time of data reading can be avoided.

  Further, since the data write current is turned back by the equalizing transistor by the folded bit line pair, the voltage at one end of each bit line BL and / BL is only controlled to one of the power supply voltage Vcc and the ground voltage Vss. , Data write currents in different directions can be supplied. In this way, voltages with different polarities (negative voltages) are not required, and the direction of the current is switched only by setting the voltages of the write data buses WDB and / WDB to either the power supply voltage or the ground voltage. Therefore, the configuration of the data writing circuit 51a can be simplified. In read / write control circuit 60, it is not particularly necessary to provide a configuration for sinking data write current ± Iw (current path to ground voltage Vss), and data write current ± Iw can be generated only by equalizing transistor 62. Can be controlled. As a result, the circuit configuration related to data write current ± Iw in read / write control circuits 50 and 60 can be reduced in size.

  In addition, since data reading is performed using dummy memory cells under a configuration in which folded bit line pairs are provided, a sufficient data reading margin can be ensured.

[Variation 1 of Embodiment 1]
FIG. 6 is a diagram for describing a configuration according to the first modification of the first embodiment of memory array 10 and its peripheral circuits.

  Referring to FIG. 6, in the first modification of the first embodiment, precharge transistors 64-1a, 64-1b to 64-ma, 64-mb connect bit lines BL1, / BL1 to BLm, / BLm. It differs from the first embodiment in that it is provided for precharging to the power supply voltage Vcc. A data write circuit 51b is arranged instead of the data write circuit 51a, and a data read circuit 55b is arranged instead of the data read circuit 55a. Since other configurations are the same as those in FIG. 2, detailed description thereof will not be repeated.

FIG. 7 is a circuit diagram showing a configuration of data write circuit 51b.
Referring to FIG. 7, data write circuit 51b has data write current supply circuit 52 shown in FIG. Data write circuit 51b directly couples output nodes Nw1 and Nw2 of data write current supply circuit 52 to write data bus pair WDB and / WDB, respectively. Data write circuit 51b does not include pull-up circuit 53 and switches SW1a and SW1b, and does not perform a pull-up operation during data reading.

FIG. 8 is a circuit diagram showing a configuration of data read circuit 55b.
Referring to FIG. 8, data read circuit 55b has transfer gates TGa and TGb provided between read data buses RDB and / RDB and the input node of differential amplifier 56, respectively. Transfer gates TGa and TGb couple read data buses RDB and / RDB to the input node of differential amplifier 56 in response to trigger pulse φr.

  Data read circuit 55b further includes a latch circuit 57 for latching the output of differential amplifier 56, and a transfer gate TGc provided between differential amplifier 56 and latch circuit 57. Similarly to transfer gates TGa and TGb, transfer gate TGc operates in response to trigger pulse φr. Latch circuit 57 outputs read data DOUT.

  Therefore, data read circuit 55b sets the data level of read data DOUT by amplifying the voltage difference between read data buses RDB and / RDB at the timing when trigger pulse φr is activated to H level. During the inactivation (L level) period of trigger pulse φr, the data level of read data DOUT is held by latch circuit 57.

  FIG. 9 is a timing chart for describing data read and data write operations in the MRAM device according to the first modification of the first embodiment.

  Referring to FIG. 9, the precharge voltages of bit lines BL and / BL before data writing are set to power supply voltage Vcc. In data writing, trigger pulse φr is maintained in an inactive state (L level). Since the operation at the time of data writing excluding these points is the same as that in the timing chart shown in FIG. 5, detailed description will not be repeated.

  Next, the operation at the time of data reading will be described. Before data reading, bit lines BL and / BL and read data buses RDB and / RDB are precharged to power supply voltage Vcc. On the other hand, at the time of data reading, write column selection line WCSL is maintained in an inactive state (L level). That is, in modification 1 of the first embodiment, unlike the case of the first embodiment, bit lines BL and / BL are not pulled up to power supply voltage Vcc during data reading.

  When the read word line RWL is selectively activated in accordance with the row selection result from the state where the bit lines BL and / BL are precharged to the power supply voltage Vcc, the MTJ memory cell MC from which data is to be read , The access transistor ATR is turned on to form a path for the sense current Is. As a result, the voltages of the bit lines BL and / BL start to decrease.

  In this case, the voltage drop rate of bit lines BL and / BL is determined according to the resistance value of memory cell MC or dummy memory cell DMC coupled to bit lines BL and / BL. That is, the voltage drop speed of the bit lines BL, / BL coupled to the memory cells MC storing the L level data is fast, and the bit lines BL, / BL coupled to the memory cells MC storing the H level data are fast. The voltage drop rate is slow. The voltage drop speed of the bit lines BL and / BL coupled to the dummy memory cell DMC is an intermediate value between them.

  In FIG. 9, as an example, the waveform of the bit line when the MTJ memory cell MC that is the target of data reading holds L level data is shown along with the waveform of the bit line coupled to the dummy memory cell DMC.

  The voltage drop of bit lines BL and / BL is transmitted to read data buses RDB and / RDB via read gate RG, as in the first embodiment. Therefore, by catching the timing when the voltages of the read data buses RDB and / RDB are in the process of decreasing, the trigger pulse φr is activated and the voltage difference between the read data buses RDB and / RDB is taken into the latch circuit 57. High-speed data reading similar to that in the first embodiment can be performed.

  In the configuration according to the first modification of the first embodiment, it is not necessary to supply the sense current Is at the time of data reading, so that it is possible to further reduce power consumption.

[Modification 2 of Embodiment 1]
FIG. 10 is a diagram for describing a configuration according to the second modification of the first embodiment of memory array 10 and its peripheral circuits.

  In the second modification of the first embodiment, the data read via the read gate RG described in the first embodiment and the first modification is applied to the open bit line configuration.

  Referring to FIG. 10, open-type bit lines BL1 to BLm are provided corresponding to the memory cell columns, respectively. Write column select gates WCSG1 to WCSGm are provided between write data bus WDB and bit lines BL1 to BLm, respectively. Write column select gates WCSG1 to WCSGm are turned on / off according to the voltages on write column select lines WCSL1 to WCSLm.

  Read / write control circuit 60 includes bit line current control transistors 63-1 to 63-m provided between write database / WDB and bit lines BL1 to BLm, respectively. Bit line current control transistors 63-1 to 63-m are turned on / off according to the voltages of write column selection lines WCSL1 to WCSLm, respectively, similarly to write column selection gates WCSG1 to WCSGm.

  Precharge transistors 64-1 to 64-m precharge bit lines BL1 to BLm to power supply voltage Vcc in response to bit line precharge signal BLPR.

  Data write current ± Iw is supplied to write data buses WDB and / WDB by data write circuit 51b, as in FIG. With such a configuration, the data write current can be supplied to the selected memory cell column in the same manner as in the first modification of the first embodiment.

  In each memory cell column, read column select gate RCSG and read gate RG are coupled in series between read data bus RDB and ground voltage Vss. For example, in the first memory cell column, a read column selection gate RCSG1 formed of an N-type MOS transistor that is turned on / off according to a read column selection line RCSL1 between the read data bus RDB and the ground voltage Vss. A read gate RG1 formed of an N-type MOS transistor having a gate coupled to the bit line BL1 is coupled in series.

  With such a configuration, in the selected memory cell column, the read data bus RDB can be driven according to the voltage of the corresponding bit line BL via the read gate RG. Therefore, when the read word line RWL is activated from the state in which the bit lines BL1 to BLm are precharged to the power supply voltage Vcc, the bit line BL (power supply voltage Vcc precharge) to MTJ memory is selected in the selected memory cell. A sense current path from the cell to the ground voltage Vss can be formed.

  As a result, a voltage drop at a speed corresponding to the storage data level of the selected MTJ memory cell MC occurs in the corresponding bit line BL. Therefore, as in the first modification of the first embodiment, the voltage level of the bit line is taken into the data read circuit 55c at an appropriate timing while the voltage of the read data bus RDB is decreasing. By performing voltage comparison with reference voltage Vm determined based on the voltage drop speed of dummy memory cell DMC in Modification 1, read data Dout can be output. That is, the configuration of data read circuit 55c is the same as that of data read circuit 55c shown in FIG. 8, and one of the input nodes of differential amplifier 56 is replaced with the voltage of read data bus / RDB. Vm can be realized.

  Data column selection gate WCSG and bit line current control transistor 62 are turned on / off in the same manner as in the first embodiment, and data including pull-up circuit 53 is substituted for data write circuit 51b. It is also possible to perform data reading similar to that in the first embodiment in a state where write circuit 51a is arranged and bit line BL is pulled up to power supply voltage Vcc.

  In this case, depending on the column selection result, write column selection gate WCSG is turned on both at the time of data reading and at the time of data writing, but bit line current control transistor 62 is turned on only at the time of data writing. What is necessary is just composition.

  Although a detailed configuration is not shown, instead of data read circuit 55c, according to the comparison result between the voltage of write data bus WDB and the reference voltage set corresponding to resistance value Rd of dummy memory cell DMC. A differential amplifier circuit that generates read data DOUT may be disposed.

  As described above, data reading and data writing can be performed even under an open-type bit line configuration, as in the first embodiment and the first modification thereof.

[Modification 3 of Embodiment 1]
In the third modification of the first embodiment, the number of gate circuits related to column selection can be reduced.

  FIG. 11 is a diagram for describing a configuration according to the third modification of the first embodiment of memory array 10 and its peripheral circuits.

  Referring to FIG. 11, in the third modification of the first embodiment, a data input / output line pair DI / OP formed by data input / output lines IO and / IO is arranged.

  Column select gates CSG1 to CSGm are provided between data input / output line pair DI / OP and bit line pairs BLP1 to BLPm, respectively. Column selection gates CSG-CSGm are turned on / off according to the voltages of column selection lines CSL1-CSLm selectively activated to H level by column decoder 25 according to the selection result. That is, column selection gates CSG1 to CSGm are controlled to be turned on / off in common according to the column selection result in both data reading and data writing.

  Similarly, the column selection gates CSG1 to CSGm are represented by the symbol CSG when collectively expressed.

  A read gate for speeding up data reading is coupled between read data bus pair RDBP and data input / output line pair DI / OP as common read gate RCG. A write selection gate WCG is further provided between the data input / output line pair DI / OP and the write data bus pair.

  Since configurations of memory array 10 and read / write control circuit 60 are the same as those in FIG. 2, detailed description thereof will not be repeated. The configuration and operation of data write circuit 51a and data read circuit 55a are also as already described, and therefore detailed description will not be repeated.

  Read gate RCG has N-type MOS transistors Qc1 and Qc3 coupled in series between read data bus RDB and ground voltage Vss, and N coupled in series between read data bus / RDB and ground voltage Vss. Type MOS transistors Qc2 and Qc4. Control signal RE is input to the gates of transistors Qc1 and Qc2. Transistors Qc3 and Qc4 have their gates connected to data input / output lines / IO and IO, respectively.

  With such a configuration, at the time of data reading in which the control signal RE is activated to H level, the selected memory cell column is connected via the column selection gate CSG and the data input / output line pair DI / OP. Read data buses RDB and / RDB can be driven by corresponding bit lines BL and / BL.

  Therefore, the common read gate RCG can be shared by the memory cell columns in the memory array 10 to reduce the circuit area. Even with the common read gate RCG, high-speed data reading can be executed without passing the sense current Is through the read data buses RDB and / RDB.

  Write select gate WCG has an N-type MOS transistor Qc5 electrically coupled between write data bus WDB and data input / output line IO, and electrically connected between write data bus / WDB and data input / output line / IO. And an N-type MOS transistor Qc6 coupled to each other. Control signal SG is input to the gates of transistors Qc5 and Qc6. Control signal SG is activated in writing data according to control signal WE. Further, at the time of data reading, by activating control signal SG in accordance with control signal RE, transistors Qc5 and Qc6 are turned on, and the memory selected by pull-up circuit 53 in data writing circuit 51a is selected. The bit lines BL and / BL corresponding to the cell columns can be pulled up to supply the sense current Is.

  During data writing, transistors Qc1 and Qc2 in common read gate RCG are turned off, so that voltages on read data buses RDB and / RDB are independent of data input / output lines IO and / IO.

  On the other hand, in response to activation of control signal SG (H level), write data buses WDB and / WDB are electrically connected to data input / output lines IO and / IO by transistors Qc5 and Qc6 in write select gate WCG, respectively. Combined with Thus, data write current ± Iw can be supplied to bit lines BL and / BL corresponding to the selected memory cell column.

  Similarly to the case of FIG. 6, data write circuit 51b and data read circuit 55b are arranged in place of data write circuit 51a and data read circuit 55a, and bit lines BL1, / BL1 to BLm, / BLm By setting the precharge voltage to the power supply voltage Vcc, it is also possible to execute data reading according to the voltage drop rate in the bit line, as in the first modification of the first embodiment.

  Further, in this case, it is necessary to deactivate control signal SG to L level and turn off write selection gate WCG at the time of data reading. For example, the control signal WE may be directly used instead of the control signal SG and input to the gates of the transistors Qc5 and Qc6.

[Embodiment 2]
In the second embodiment, a configuration for adjusting the data write current to secure a data write margin corresponding to a change in magnetic characteristics of the memory cell due to manufacturing variations will be described.

FIG. 12 is a circuit diagram showing a configuration of the data write circuit according to the second embodiment.
Referring to FIG. 12, the data write circuit according to the second embodiment is different from the configuration of data write circuit 51a shown in FIG. 3 in that data write current adjusting circuit 200 is further provided.

  Data write current adjustment circuit 200 outputs a reference voltage Vrw for controlling the amount of current of current source 153 in data write current supply circuit 52. Data write current supply circuit 52 has an N-channel MOS transistor corresponding to current source 153 and receiving an input of reference voltage Vrw at its gate. Therefore, in data write current supply circuit 52, the amount of current supplied to node Nw0 by transistor 151 that forms a current mirror with transistor 152, that is, the amount of data write current ± Iw is equal to reference voltage Vrw. Can be adjusted accordingly.

  Data write current adjustment circuit 200 includes a reference voltage external input terminal 202 for inputting reference voltage Vre1 from the outside, and a test mode entry signal TE for switching generation of reference voltage Vrw in the test mode / normal mode. Includes a test input terminal 204 for inputting a reference voltage and an internal reference voltage generation circuit 206 for generating a reference voltage Vri1 internally.

  Data write current adjustment circuit 200 further includes a transfer gate TGf1 coupled between reference voltage external input terminal 202 and node Nf1, and a transfer gate disposed between internal reference voltage generation circuit 206 and node Nf1. TGf2. Transfer gates TGf1 and TGf2 are complementarily turned on in response to test mode entry signal TE. Node Nf1 is coupled to the gate of an N channel MOS transistor corresponding to current source 153.

  With such a configuration, transfer gate TGf2 and transfer gate TGf1 are turned on and off, respectively, during a normal operation in which test mode entry signal TE is inactivated to L level. Accordingly, the reference voltage Vri1 generated by the internal reference voltage generation circuit 206 is input to the gate of the transistor corresponding to the current source 153 as the reference voltage Vrw.

  On the other hand, in the test operation in which test mode entry signal TE is activated to H level, transfer gate TGf1 and transfer gate TGf2 are turned on and off, respectively. As a result, the reference voltage Vre1 externally applied to the reference voltage external input terminal 202 is input to the gate of the transistor corresponding to the current source 153.

  Therefore, in the test mode, by activating test mode entry signal TE, it is possible to input a reference voltage Vre1 of an arbitrary voltage level from the outside and execute a data write margin test. As a result, it is possible to execute a data write current amount adjustment test to compensate for manufacturing variations in the magnetic characteristics of the MTJ memory cell and to appropriately secure a data write margin. In this adjustment test, for example, the data write current ± Iw may be gradually lowered from the standard value to confirm whether or not a desired data write margin is secured in all MTJ memory cells.

  The level of the voltage Vri1 generated by the internal reference voltage generation circuit 206 may be set to an appropriate value of the reference voltage Vrw found by such an adjustment test.

  As a result, it is possible to compensate for fluctuations in the magnetic characteristics of the MTJ memory cell due to manufacturing variations, and to execute a data write operation during normal operation based on an appropriate amount of data write current.

FIG. 13 is a circuit diagram showing a configuration example of the word line driver according to the second embodiment.
Referring to FIG. 13, the word line driver according to the second embodiment has write word drivers WWD1 to WWDn provided corresponding to write word lines WWL1 to WWLn, respectively. Each of write word drivers WWD1 to WWDn is formed of an inverter, for example. In the following, when the write word drivers WWD1 to WWDn are collectively described, the symbol WWD is simply used.

  Row decoder 20 activates one of row decode signals RD1 to RDn corresponding to the selected row to L level according to row address RA. Row decode signals RD1 to RDn are transmitted to word line driver 30. In word line driver 30, write word drivers WWD1 to WWDn receive row decode signals RD1 to RDn, respectively, and select the corresponding write word line WWL when the corresponding row decode signal is activated to L level. Activated to (H level).

  Word drivers WWD1 to WWDn supply data write current Ip to write word line WWL corresponding to the selected row during data writing.

  Word line driver 30 further includes a data write current supply circuit 32 for supplying data write current Ip to word drivers WWD1 to WWDn, and a data write current for adjusting the amount of data write current Ip. And an adjustment circuit 210.

  Data write current supply circuit 32 is electrically connected between nodes Np0 and Np1 and power supply voltage Vcc, P channel MOS transistors 33a and 33b electrically connected, and node Np1 and ground voltage Vss, respectively. N-channel MOS transistor 34 coupled. Data write current Ip supplied to each write word driver WWD is transmitted to node Np0.

  Node Np1 is electrically coupled to the gates of transistors 33a and 33b. A reference voltage Vrp output from the data write current adjustment circuit is input to the gate of transistor 34. Thereby, the transistor 34 operates as a current source for supplying a current amount corresponding to the reference voltage Vrp. On the other hand, since transistors 33a, 33b and 34 constitute a current mirror circuit, the amount of current supplied to node Np0 by data write current supply circuit 32, that is, the amount of data write current Ip is set to the data write current. The adjustment can be made according to the reference voltage Vrp output from the adjustment circuit 210.

Data write current adjustment circuit 210 has the same configuration as data write current adjustment circuit 200 described with reference to FIG. 12. That is, data write current adjustment circuit 210 has a reference voltage for inputting reference voltage Vre2 from the outside. It includes an external input terminal 212, a test input terminal 214 for inputting a test mode entry signal TE, and an internal reference voltage generation circuit 216 for internally generating a reference voltage Vri2.

  Data write current adjustment circuit 210 further includes a transfer gate TGf3 coupled between reference voltage external input terminal 212 and node Nf2, and a transfer gate disposed between internal reference voltage generation circuit 216 and node Nf2. TGf4. Transfer gates TGf3 and TGf4 are complementarily turned on in response to test mode entry signal TE. Node Nf2 is coupled to the gate of transistor 34 operating as a current source.

  Therefore, in each of the normal operation and the test mode, the reference voltage Vri2 generated by the internal reference voltage generation circuit 216 and the reference voltage Vre2 applied from the outside to the reference voltage external input terminal 212 according to the test mode entry signal TE. Is input to the gate of the transistor 34.

  As a result, in the test mode, a reference voltage Vre2 of an arbitrary voltage level can be input from the outside to perform a data write margin test. As a result, it is possible to easily execute a data write current amount adjustment test for compensating for manufacturing variations in the magnetic characteristics of the MTJ memory cell and ensuring a data write margin appropriately. In this adjustment test, for example, the data write current Ip may be gradually lowered from the standard value to confirm whether or not a desired data write margin is secured in all MTJ memory cells.

  The level of the voltage Vri2 generated by the internal reference voltage generation circuit 216 may be set to an appropriate value of the reference voltage Vrw found by such an adjustment test.

  As a result, it is possible to compensate for fluctuations in the magnetic characteristics of the MTJ memory cell due to manufacturing variations, and to execute a data write operation during normal operation based on an appropriate amount of data write current.

[Modification of Embodiment 2]
FIG. 14 is a circuit diagram showing a configuration of data write current adjustment circuit 230 according to the modification of the second embodiment.

  Referring to FIG. 14, data write current adjustment circuit 230 outputs a reference voltage Vref for adjusting the amount of data write current. 14 adjusts data write current adjustment circuit 200 for adjusting the data write current ± Iw for the bit line and data write current Ip for the write word line. The data write current adjustment circuit 210 can be replaced and applied.

  Referring to FIG. 14, data write current adjustment circuit 230 includes a tuning input unit 231a and a voltage adjustment unit 231b that adjusts reference voltage Vref according to the setting for tuning input unit 231a.

  Voltage adjusting unit 231b obtains a voltage difference between a P-channel MOS transistor 232 electrically coupled between node Nt1 for generating reference voltage Vref and power supply voltage Vcc, and a voltage at node Nt2 and predetermined voltage Vref0. An operational amplifier 234 that amplifies and outputs the amplified signal to the gate of the transistor 232.

  Voltage adjusting unit 231b further includes a P channel transistor 240 electrically coupled between nodes Nt1 and Nt2, and P channel MOS transistors 241, 242, coupled in series between node Nt2 and ground voltage Vss. 243 and 244. Transistors 240-244 have their gates coupled to ground voltage Vss. Thereby, the transistors 240 to 244 function as resistance elements.

  By controlling the gate voltage of the transistor 232 by the operational amplifier 234, the voltage level of the reference voltage Vref is controlled so that the voltage of the node Nt2 becomes equal to the predetermined voltage Vref0. The predetermined voltage Vref0 is set in consideration of the reference voltage Vref.

  Here, the voltage Vα at the node Nt2 is obtained by dividing the reference voltage Vref by the transistors 240 to 244 acting as resistance elements. When this voltage division ratio is defined as α (α = Vref / Vα), the reference voltage Vref is expressed by Vref = α · Vref0 using a predetermined voltage Vref0 input to the operational amplifier 234.

  The voltage division ratio α is determined by a ratio between a resistance value between the node Nt1 and the ground voltage Vss and a resistance value between the node Nt2 and the ground voltage Vss set according to the input to the tuning input unit 231a. The

  Thus, by programming the voltage dividing ratio α related to the input voltage to the operational amplifier 234 without directly programming the reference voltage Vref, the responsiveness and noise resistance of the reference voltage Vref can be improved.

  Tuning input unit 231a has a set of a fuse element, which is a program element, and a transfer gate, which are provided in parallel with each of transistors 241 to 243. For example, a transfer gate TGt1 and a fuse element 251 are connected in series with the transistor 241 in parallel. For transistor 242, transfer gate TGt2 and fuse element 252 connected in series are arranged. Similarly, a transfer gate TGt3 and a fuse element 253 connected in series are arranged in parallel with the transistor 243.

  The fuse can be blown by directing laser light directly into the fuse elements 251 to 253 or inputting a high voltage signal from the outside via the blow input nodes 281 to 283.

  Tuning input unit 231a further includes an input terminal 270 that receives a control signal TT that is activated when a tuning test of the data write current is executed, and input terminals 271 to 273 for receiving tuning test signals TV1 to TV3, respectively. A logic gate 261 for controlling on / off of the transfer gate TGt1 according to the levels of the control signal TT and the tuning test signal TV1, and an on / off of the transfer gate TGt2 according to the levels of the control signal TT and the tuning test signal TV2 And a logic gate 263 for controlling on / off of the transfer gate TGt3 in accordance with the levels of the control signal TT and the tuning test signal TV3.

  During normal operation, control signal TT is inactivated to L level, so that the output signals of logic gates 261 to 263 are set to H level, respectively. In response to this, since the transfer gates TGt1 to Tgt3 are all turned on, the voltage dividing ratio α is determined according to whether or not the fuse elements 251 to 253 are blown.

  In the tuning input unit 231a, the output signals of the logic gates 262 to 264 are set to the L level by the input signals to the input terminals 270 to 273, and the corresponding transfer gates TGt1, TGt2, TGt3 are turned off, thereby making a pseudo- It is possible to create a state where the fuse is blown.

  For example, when the tuning test is executed by activating the control signal TT (to H level), the transfer gate TGt1 can be turned off by activating the tuning test signal TV1 to H level, and the fuse element 251 It is possible to create a state equivalent to blowing.

  Similarly, a pseudo blown state can be set for fuse elements 252 and 253 as well.

  Therefore, the reference voltage Vref for adjusting the data write current can be variably set by changing the voltage dividing ratio α by the control signal TT and the tuning test signals TV1 to TV3 input to the input terminals 270 to 273. it can.

  Therefore, during the tuning test, it is easy to adjust the data write current amount in order to ensure a proper data write margin by reversibly adjusting the voltage division ratio α without actually blowing the fuse. Can be executed.

  After the tuning test is completed, the reference voltage Vref for obtaining an appropriate data write current can be programmed in the tuning input unit 231a in a nonvolatile manner by actually blowing the fuse element based on the test result. As a result, the data write current adjustment circuit 230 generates an appropriate programmed reference voltage Vref during normal operation. Therefore, the data write current adjustment circuit 230 compensates for manufacturing variations in the magnetic characteristics of the MTJ memory cell, A write operation can be executed.

  In FIG. 14, the reference voltage external input terminals 202 (212) and 204 (214) for inputting the reference voltage from the outside, and the transfer gates TGf1 (TGf3) and TGf2 (TGf4) are combined. Although shown, the data write current tuning test can be executed by omitting these elements and inputting the reference voltage Vref directly to the gate of the transistor 153 (34).

  By adopting such a configuration, the tuning test can be efficiently executed only by inputting a digital signal, as compared with the configuration of the data write current adjustment circuits 200 and 210 shown in FIGS. . In addition, since it is not necessary to make an adjustment corresponding to the output voltage adjustment of internal reference voltage generation circuits 206 and 216 in data write current adjustment circuits 200 and 210, the adjustment load is reduced.

  Note that the number of transistors for setting the voltage division ratio α is not limited to the example shown in FIG. 13, and an arbitrary plurality of transistors can be provided. In this case, if a set of transfer gates and fuse elements controlled similarly and an input terminal for a control signal are provided in parallel with a plurality of transistors functioning as resistance elements, the reference voltage Vref The setting level can be further refined.

  Further, in the configuration of FIG. 14, the configuration using a fuse element that is cut off after blow input is exemplified as the program element. However, a so-called antifuse element that becomes conductive after blow input can also be used. In this case, the same effect can be obtained if each of the transfer gates (TGt1 to TGt3 in FIG. 14) for executing the tuning test is provided in parallel with the antifuse element.

  The adjustment of the data write current described in the second embodiment and its modification is not limited to the MRAM device that performs data reading via the read gate described in the first embodiment and its modification. The present invention can be applied to an MRAM device having a different configuration.

  FIG. 15 shows a configuration example of an MRAM device that executes data reading without using a read gate.

  15 is compared with FIG. 2, in the configuration shown in FIG. 15, column select gates CSG1 to CSGm are arranged corresponding to the memory cell columns, respectively. Each column selection gate couples between corresponding bit line pair BLP and data input / output line pair DI / OP according to the column selection result. For example, column select gate CSG1 uses data input / output lines IO and / IO constituting data input / output line pair DI / OP in accordance with the voltage on column select line CSL1, and bit lines constituting corresponding bit line pair BLP1. It binds to BL1 and / BL1, respectively.

  Supply of data write current ± Iw to data input / output line pair DI / OP can be executed by data write circuit 51b described with reference to FIG. In order to adjust the amount of current of current source 153 in data write current supply circuit 52 included in data write circuit 51b, data write current adjustment circuit 200 or 230 shown in FIGS. 12 and 14 is provided, respectively. Thus, the same adjustment of the data write current can be executed.

  The data write current Ip for the write word line WWL is executed by the word line driver 30. By applying the configuration described in FIG. 13 to the configuration of the word line driver 30, the same as in the second embodiment. The data write current can be adjusted.

  In the MRAM device having the configuration shown in FIG. 15, it is necessary to supply sense current Is during data reading by data reading circuit 55d.

  Data read circuit 55d receives power supply voltage Vcc and is electrically coupled between current sources 161 and 162 for supplying a constant current to internal nodes Ns1 and Ns2, respectively, and internal node Ns1 and node Nr1. Amplifies the voltage level difference between n-type MOS transistor 163, N-type MOS transistor 164 electrically coupled between internal node Ns2 and node Nr2, and internal nodes Ns1 and Ns2, and outputs read data DOUT And an amplifier 165.

  A reference voltage Vrr is applied to the gates of the transistors 163 and 164. The supply current amount of the current sources 161 and 162 and the reference voltage Vrr are set according to the current amount of the sense current Is. Resistors 166 and 167 are provided for pulling down internal nodes Ns1 and Ns2 to ground voltage Vss. Further, nodes Nr1 and Nr2 are coupled to data input / output lines IO and / IO, respectively.

  With such a configuration, data read circuit 55d supplies sense current Is to each of data input / output lines IO and / IO during data read. Further, read data DOUT is output in accordance with voltage changes occurring in data input / output lines IO and / IO, respectively, in accordance with the level of data stored in the MTJ memory cells connected via the column selection gate and the bit line pair. .

[Embodiment 3]
In the third embodiment, a configuration in which a bit line BL and a write word line WWL for flowing a data write current are formed over a plurality of wiring layers will be described.

FIG. 16 is a block diagram illustrating the arrangement of bit lines according to the third embodiment of the present invention.
Referring to FIG. 16, data reading and data writing with respect to memory array 10 is performed by setting data input / output line pair DI / OP by data writing circuit 51b and data reading circuit 55d based on the same configuration as in FIG. It shall be executed via

  Corresponding to each of the memory cell columns, bit lines BL1 to BLm, / BL1 to / BLm forming column line pairs BLP1 to BLPm, column selection gates CSG1 to CSGn, and column selection lines CSL1 to CSLm are provided.

  Bit lines BL1 to BLm and bit lines / BL1 to / BLm are formed in different wiring layers. For example, each of bit lines BL1 to BLm is formed in metal interconnection layer M3, and each of bit lines / BL1 to / BLm is formed in metal interconnection layer M4.

  Memory cell MC is coupled to one bit line BL forming each bit line pair. On the other hand, dummy memory cell DMC is coupled to the other bit line / BL forming each bit line pair.

  Read / write control circuit 60 includes equalize transistors 62-1 to 62-m provided corresponding to the memory cell columns, respectively. Equalize transistor 62 short-circuits between bit lines BL and / BL formed in different metal wiring layers in response to bit line equalize signal BLEQ. Bit line equalize signal BLEQ is activated / deactivated in the same manner as described in the first embodiment.

  Therefore, at the time of data writing, data write current ± Iw supplied to bit line pair BLP is supplied as a round-trip current flowing through bit lines BL and / BL in different directions in the selected memory cell column. . Therefore, similarly to the first embodiment, the configuration of data write circuit 51b including data write current supply circuit 52 can be applied.

  As a result, as in the first embodiment, the equalization transistor 62 can provide a return path for the data write current ± Iw, so that the read / write control circuit 60 side can sink the data write current. There is no need for special arrangement, and the layout of the peripheral circuits can be reduced.

FIG. 17 is a structural diagram showing a first arrangement example of bit lines according to the third embodiment.
Referring to FIG. 17, write word line WWL is formed in metal interconnection layer M2. The bit line pair BLP includes a bit line BL formed in the metal wiring layer M3 and a bit line / BL formed in the metal wiring layer M4. Thus, the bit lines BL and / BL are formed using different metal wiring layers so as to sandwich the magnetic tunnel junction MTJ in the vertical direction. Bit lines BL and / BL are electrically coupled by an equalize transistor 62 at the end of memory array 10 to pass a data write current.

  Therefore, data write current ± Iw at the time of data writing flows in different directions in each of bit lines BL and / BL. Therefore, in magnetic tunnel junction MTJ, the data write magnetic field generated by data write current ± Iw acts in the direction in which the magnetic field generated by bit line BL and the magnetic field generated by bit line / BL are intensified. Thereby, the data write current ± Iw at the time of data writing can be reduced. As a result, the current consumption of the MRAM device can be reduced, the reliability can be improved by lowering the bit line current density, and the generated magnetic field noise during data writing can be reduced.

  On the contrary, in the peripheral portion including other memory cells, the magnetic fields generated by the bit lines BL and / BLb act in the direction of canceling each other. As a result, magnetic field noise during data writing can be further suppressed.

FIG. 18 is a structural diagram showing a second arrangement example of the bit lines according to the third embodiment.
Referring to FIG. 18, write word line WWL is arranged in metal interconnection layer M3. Bit lines BL and / BL are arranged in different metal wiring layers M2 and M4 so as to sandwich magnetic tunnel junction MTJ in the vertical direction. Even with such a configuration, the direction of the magnetic field generated by the data write current ± Iw is the same as in FIG. Therefore, the same effect as when the structure shown in FIG. 17 is adopted can be obtained.

  Referring to FIG. 16 again, in the third embodiment, MRAM is applied to data write circuit 51b for supplying a data write current during data writing and word line driver 30 for activating write word line WWL. The external power supply voltage Ext. Vcc is supplied directly.

  The MRAM device 1 further includes an external power supply voltage Ext. Vcc is stepped down to reduce the internal power supply voltage Int. A voltage drop circuit (VDC: Voltage Down Converter) 7 for generating Vcc is provided.

  The internal power supply voltage Int. Vcc is supplied to internal circuits that perform data reading and address processing, such as data reading circuit 55d, column decoder 25, control circuit 5, and row decoder 20.

  With such a configuration, a data write circuit that supplies a relatively large data write current ± Iw and a word line driver that supplies a data write current Ip to the write word line WWL are externally provided during data writing. External power supply voltage Ext. These data write currents can be quickly supplied by being driven by Vcc.

  On the other hand, for the internal circuits other than the circuit for supplying the data write current, the internal power supply voltage Int. By driving with Vcc, it is possible to reduce power consumption in these internal circuits and to ensure reliability corresponding to device miniaturization for high integration.

[Modification 1 of Embodiment 3]
FIG. 19 is a conceptual diagram illustrating the arrangement of bit lines according to the first modification of the third embodiment.

  Referring to FIG. 19, bit lines BL and / BL constituting each bit line pair BLP are provided so as to intersect in region CRS in memory array 10 using metal wiring layers M3 and M4.

  That is, in the configuration shown in FIG. 19, in the left region of region CRS, bit lines BL and / BL are formed by wirings arranged in metal wiring layers M3 and M4, respectively. On the other hand, in the right region of region CRS, bit lines BL and / BL are formed by wirings arranged in metal wiring layers M4 and M3, respectively.

  Wirings corresponding to the bit lines BL formed in the metal wiring layers M3 and M4 are coupled in the region CRS. Similarly, wirings corresponding to bit lines / BL formed in metal wiring layers M3 and M4 are coupled in region CRS.

  Bit lines BL and / BL are coupled to memory cell MC in one of the metal wiring layers. In FIG. 18, bit lines BL and / BL are coupled to memory cell MC in lower metal wiring layer M3 which is structurally small in distance from magnetic tunnel junction MTJ.

  Thus, memory cells MC belonging to the same memory cell column are coupled to one of bit lines BL and / BL. Therefore, a dummy memory cell DMC coupled to bit line BL and a dummy memory cell DMC coupled to bit line / BL are arranged corresponding to each memory cell column. A dummy read word line DRWL1 is arranged in common for the dummy memory cells DMC coupled to the bit line BL. Similarly, dummy read word line DRWL2 is arranged for dummy memory cell DMC coupled to bit line / BL.

  Equalize transistors 62-1 to 62-m are provided corresponding to the memory cell columns, respectively, and couple between bit lines BL and / BL constituting a bit line pair in response to bit line equalize signal BLEQ.

  By adopting such a configuration, in the selected memory cell column, data based on the folded bit line configuration is caused by flowing a round-trip current folded by the equalizing transistor 62 to the bit lines BL and / BL. Write can be performed.

  As described above, in the arrangement of the bit lines shown in FIG. 19, the number of memory cells coupled to each of bit lines BL and / BL constituting the bit line pair can be equalized. The RC load imbalance between the bit lines BL and / BL forming the line can be corrected. Furthermore, since the data read operation based on the folded bit line configuration can be executed using the dummy read cell, the operation margin at the time of data read can be further improved.

  Since the structure of the other parts and the basic operation at the time of data reading and data writing are the same as those of FIG. 15, detailed description will not be repeated.

[Modification 2 of Embodiment 3]
In the following, a configuration when the write word line WWL is formed using a plurality of metal wiring layers will be described.

  FIG. 20 is a structural diagram illustrating the arrangement of write word lines WWL according to the second modification of the third embodiment.

  Referring to FIG. 20, write word line WWL includes WWLl formed in metal wiring layer M2 and WWLu formed in fourth metal wiring layer M4. The sub write word lines WWLu and WWLl are arranged so as to sandwich the magnetic tunnel junction MTJ in the vertical direction.

  FIG. 21 is a conceptual diagram illustrating coupling between sub word lines forming the same write word line.

  Referring to FIGS. 21A and 21B, sub word lines WWLu and WWLl forming the same write word line WWL are electrically coupled at the end of memory array 10. Thereby, data write current Ip can be made to flow as a round trip current using sub word lines WWLu and WWLl.

  FIG. 21A shows a configuration in which sub-write word lines WWLu and WWLl are electrically coupled through metal wiring 145 disposed in through hole 144.

  Further, as shown in FIG. 21B, a write word line current control switch TSW formed of a MOS transistor electrically coupled between the sub write word lines WWLu and WWLl is short-circuited between the two. It is also possible to arrange for this.

  With such a configuration, the data write current Ip can be turned back and passed as currents in opposite directions to the sub word lines WWLu and WWLl forming the same word line WWL.

  Referring to FIG. 20 again, by supplying data write current Ip in the reverse direction to sub write word lines WWLl and WWLu, respectively, magnetic fields are generated by sub write word lines WWLu and WWLl in the same manner as in FIGS. The data write magnetic field generated at each tunnel junction MTJ acts in the same direction.

  In the peripheral portion including other memory cells, the magnetic fields generated by these sub-write word lines WWLu and WWLl act in the direction of canceling each other. As a result, a larger data write magnetic field can be applied to the magnetic tunnel junction MTJ even with the same current value. As a result, the data write current required to generate the desired data write magnetic field is reduced.

  As a result, reduction of current consumption of the MRAM device, improvement of operation reliability by reduction of the current density of the write word line WWL, and reduction of generated magnetic field noise during data writing can be similarly realized.

[Modification 3 of Embodiment 3]
FIG. 22 is a diagram for explaining the arrangement of write word lines according to the third modification of the third embodiment.

  Referring to FIG. 22, write word drivers WWD1 to WWDn included in row decoder 20 and word line driver 30 are provided at one end of memory array 10 along the row direction. Write word drivers WWD1 to WWDn are provided corresponding to write word lines WWL1 to WWLn, respectively, and activate corresponding write word lines WWL and supply data write current Ip according to the decoding result of row decoder 20 To do.

  Each write word line WWL is arranged in the structure shown in FIGS. 20 and 21 (a). That is, sub write word lines WWLu and WWLl forming the same write word line WWL are electrically coupled to each other at the other end of memory array 10 by metal interconnection 145 through a through hole.

  Write word drivers WWD1 to WWDn supply data write current Ip to one of sub write word lines WWLu among corresponding write word lines WWL. The other sub-write word line WWLl forming the same write word line WWL is coupled to the ground voltage Vss at one end of the memory array 10 (on the write word driver WWD side).

  With such a configuration, in data writing, data write current Ip is used as a round-trip current that is folded back using subwrite word lines WWLu and WWLl in word line WWL corresponding to the selected memory cell column. Can flow. Incidentally, the connection relationship between the write word driver WWD and the ground voltage Vss and the sub write word lines WWLu and WWLl is switched, the sub write word line WWLl is coupled with the write word driver WWD, and the sub write word line WWLu is connected to the ground voltage. It is also possible to adopt a configuration that couples with Vss.

[Modification 4 of Embodiment 3]
FIG. 23 illustrates the arrangement of write word lines according to the fourth modification of the third embodiment.

  Referring to FIG. 23, in the configuration according to the fourth modification of the third embodiment, write word drivers WWD provided corresponding to each write word line WWL are divided and arranged at both ends of memory array 10. Therefore, the row decoder is also divided and arranged into a row decoder 20a for activating a write word driver corresponding to an odd row and a row decoder 20b for controlling a write word driver corresponding to an even row.

  As already described, write word driver WWD requires a relatively large size because it includes a transistor for supplying data write current Ip. Therefore, by dividing the write word driver WWD on both sides of the memory array in this way, the write word driver WWD can be arranged by utilizing the layout pitch for two rows. Thereby, the arrangement of the write word lines WWL in the row direction can be further integrated, and the area of the memory array 10 can be efficiently reduced.

  Since the configuration and operation of the other parts are the same as in FIG. 22, detailed description will not be repeated.

[Modification 5 of Embodiment 3]
FIG. 24 is a diagram for explaining the arrangement of write word lines according to the fifth modification of the third embodiment.

  Referring to FIG. 24, in the configuration according to the fifth modification of the third embodiment, sub-write word lines WWLu and WWLl forming the same word line WWL are provided at one end of memory array 10 (on the side of row decoder 20). They are electrically coupled by write word line current control switches TSW provided corresponding to the memory cell rows.

  FIG. 24 representatively shows, as an example, write word line current control switches TSW1 and TSW2 provided corresponding to write word lines WWL1 and WWL2, respectively. The write word line current control switch TSW is turned on when the corresponding memory cell row is selected as controlled by the row decoder 20.

  Sub write word lines WWLu and WWLl forming the same write word line WWL are coupled to power supply voltage Vcc and ground voltage Vss at the other end of memory array 10, respectively. Accordingly, when the write word line current control switch TSW is turned on based on the row selection result, the reciprocal data write current Ip is caused to flow to the sub write word lines WWLu and WWLl constituting the corresponding write word line WWL. Can do. Thereby, the same effects as those of Modifications 3 and 4 of Embodiment 3 can be obtained.

  In a period in which the corresponding write word line current control switch TSW is turned off, sub write word lines WWLu and WWLl are set to power supply voltage Vcc and ground voltage Vss, respectively. Therefore, it is possible to speed up the operation of returning the voltage of the write word line WWL to the standby state or the non-selected state after the selection operation of the write word line WWL.

  FIG. 24 illustrates a configuration in which sub-write word lines WWLu and WWLl are coupled to power supply voltage Vcc and ground voltage Vss at the other end of memory array 10 respectively. Write word lines WWLu and WWLl may be coupled to ground voltage Vss and power supply voltage Vcc, respectively.

  That is, the write word line WWL is made long in order to flow a round-trip data write current Ip at the time of data writing, but the write word line WWL is divided into sub-write word lines WWLu and WWLl and sub-write words By configuring each line to return to a predetermined voltage level, it is possible to speed up the operation of returning to the standby state or the non-selected state while enjoying the effect of flowing the data write current as a round-trip current. It becomes possible.

  In the modified examples 3 to 5 of the third embodiment, the dummy write word having the same configuration as that corresponding to the memory cell MC is also applied to the dummy memory cell DMC that is not originally related to the data write operation. At least one of lines DWWL1, DWWL2, write word drivers DWWD1, DWWD2, and write word line current control switches DTSW1, DTSW2 is arranged.

  However, since it is not necessary to pass a data write current to dummy memory cell DMC, the inputs of write word drivers DWWD1 and DWWD2 corresponding to the dummy memory cell are fixed at power supply voltage Vcc. Therefore, dummy write word lines DWWL1 and DWWL2 are always maintained in an inactive state (ground voltage Vss), and no current flows. Further, the gate of the N-type MOS transistor constituting the corresponding write word line current control switch DTSW is fixed to the ground voltage Vss, and the turn-off state is maintained.

  If the configuration in which the write word line WWL is not arranged only in the region corresponding to the dummy memory cell DMC is adopted, the shape continuity is lost, and thus a shape defect occurs during the formation of the MRAM device. there is a possibility. Therefore, even for a dummy memory cell that does not require a data write operation, a write word line, a write word driver, and its peripheral circuit having the same configuration as that for a normal memory cell MC (write word line current in FIG. By disposing the control switch TSW), it is possible to avoid a shape defect when forming the MRAM device.

  It should be noted that the arrangement of the bit lines and write word lines according to the third embodiment and the modification thereof may be configured as each of the first and second embodiments or a combination thereof. In this case, the data write circuit and the data read circuit may be configured as described in the first and second embodiments and their modifications.

[Embodiment 4]
FIG. 25 shows a structure of an MTJ memory cell according to the fourth embodiment.

  Referring to FIG. 25, MTJ memory cell MCD according to the fourth embodiment includes magnetic tunnel junction MTJ and access diode DM, similarly to the configuration shown in FIG. The MTJ memory cell MCD is different from the configuration shown in FIG. 90 in that the read word line RWL and the write word line WWL are separately arranged. Bit line BL is arranged in a direction crossing write word line WWL and read word line RWL, and is electrically coupled to magnetic tunnel junction MTJ.

  Access diode DM is coupled between the two, with the direction from magnetic tunnel junction MTJ toward read word line RWL as the forward direction. The write word line WWL is provided close to the magnetic tunnel junction MTJ without being connected to other wiring.

FIG. 26 is a structural diagram when the MTJ memory cell MCD is arranged on a semiconductor substrate.
Referring to FIG. 26, N-type well NWL formed on semiconductor main substrate SUB corresponds to the cathode of access diode DM. In the case where the MTJ memory cells are arranged in a matrix on the semiconductor substrate, for example, the read word line RWL is particularly connected to the MTJ memory cells belonging to the same row by electrically coupling the N-type wells NWL to each other. Without providing, the coupling relationship between the access diode DM and the read word line RWL shown in FIG. 25 can be realized.

  P-type region PAR provided on N-type well NWL corresponds to the anode of access diode DM. P-type region PAR is electrically coupled to magnetic tunnel junction MTJ through barrier metal 140 and metal film 150.

  Write word line WWL and bit line BL are arranged in metal interconnection layer M1 and metal interconnection layer M2, respectively. Bit line BL is arranged to be coupled to magnetic tunnel junction MTJ.

  FIG. 27 is a timing chart illustrating a read operation and a write operation for the MTJ memory cell MCD.

  Referring to FIG. 27, at the time of data writing, the voltage of read word line RWL, that is, N-type well NWL is set to the H level (power supply voltage Vcc). In data reading, no current flows through read word line RWL.

  A power supply voltage Vcc is applied to write word line WWL corresponding to the selected memory cell, and data write current Ip flows. For bit line BL, data corresponding to the data level of the write data is set by setting one of both ends of bit line BL to power supply voltage Vcc and ground voltage Vss according to the data level of the write data. The write current ± Iw can be supplied to the bit line BL.

  Data writing to the MTJ memory cell is executed by the data write currents Ip and ± Iw that flow in this way. In this case, since read word line RWL is set to power supply voltage Vcc, access diode DM is surely turned off during data writing. Therefore, the data write operation can be stabilized as compared with the MTJ memory cell shown in FIG.

Next, the operation at the time of data reading will be described.
Before data reading, the bit line BL is precharged to the ground voltage Vss.

  Read word line RWL corresponding to memory cell MCD to be read is driven to an active state (L level: ground voltage Vss) at the time of data reading. Accordingly, access diode DM is forward-biased, so that data read is performed by passing sense current Is through the path from bit line BL to magnetic tunnel junction MTJ to access diode DM to RWL (ground voltage Vss). be able to.

  Specifically, the data stored in the magnetic tunnel junction MTJ can be read by amplifying the voltage change generated in the bit line BL by the sense current Is.

  As shown in FIG. 26, the distance between the bit line BL and the magnetic tunnel junction MTJ is smaller than the distance between the write word line WWL and the magnetic tunnel junction MTJ. Even in this case, the magnetic field generated by the data write current flowing through the bit line BL is larger than the magnetic field generated by the data write current flowing through the write word line WWL.

  Therefore, in order to apply a data write magnetic field having substantially the same strength to the magnetic tunnel junction MTJ, it is necessary to pass a data write current larger than that of the bit line BL to the word line WWL. Bit line BL and write word line WWL are formed in a metal wiring layer in order to reduce the electrical resistance value. However, if the current density flowing in the wiring becomes excessive, a disconnection or a short circuit between wirings due to the electromigration phenomenon may occur, which may hinder operation reliability. For this reason, it is desirable to suppress the current density of the wiring through which the data write current flows.

  Therefore, when the MTJ memory cell shown in FIG. 25 is arranged on a semiconductor substrate, a large data can be obtained by making the sectional area of the write word line WWL larger than that of the bit line BL closer to the magnetic tunnel junction MTJ. The reliability of the MRAM device can be improved by suppressing the current density of the write word line WWL in which the write current needs to flow.

  Further, the metal wiring (the write word line WWL in FIG. 26) that has a large distance to the magnetic tunnel junction MTJ and needs to pass a larger data write current may be formed of a material having high electromigration resistance. Effective in improving reliability. For example, when other metal wiring is formed of an aluminum alloy (Al alloy), the metal wiring that needs to be considered for electromigration resistance may be formed of copper (Cu).

  FIG. 28 is a conceptual diagram showing a configuration of a memory array in which MTJ memory cells MCD are arranged in a matrix.

  Referring to FIG. 28, a highly integrated MRAM device can be realized by arranging MTJ memory cells in a matrix on a semiconductor substrate. FIG. 28 shows a case where MTJ memory cells MCD are arranged in n rows × m columns.

  As already described, it is necessary to arrange the bit line BL, the write word line WWL, and the read word line RWL for each MTJ memory cell MCD. Therefore, n write word lines WWL1 to WWLn and read word lines RWL1 to RWLn and m bit lines BL1 to BLm are arranged for n × m MTJ memory cells arranged in a matrix. The

  FIG. 29 is a conceptual diagram showing a configuration of a memory array formed by MTJ memory cells arranged in a matrix by sharing the write word line WWL.

  Referring to FIG. 29, read word line RWL and write word line WWL corresponding to MTJ memory cell MCD having the configuration shown in FIG. 25 are arranged along the row direction. Shared between adjacent memory cells.

  For example, the MTJ memory cell coupled to the read word line RWL1 and the MTJ memory cell coupled to the read word line RWL2 share the write word line WWL1.

  Thus, by sharing the write word line WWL, the number of write word lines WWL arranged in the entire memory array can be reduced. Thereby, the arrangement of the MTJ memory cells in the memory array can be highly integrated, and the chip area can be reduced.

  In addition, by reducing the number of write word lines WWL arranged in this way, the wiring pitch of the write word lines WWL pitch can be secured in the metal wiring layer M1 shown in FIG. Thereby, the wiring width of the write word line WWL can be easily increased. This facilitates setting the cross-sectional area of the write word line WWL to be larger than that of the bit line BL closer to the magnetic tunnel junction MTJ. As a result, it is possible to easily improve the reliability of the MRAM device by suppressing the occurrence of electromigration.

[Modification of Embodiment 4]
Such wiring sharing can also be applied to the MTJ memory cell having the configuration shown in FIG. 90 described in the prior art.

FIG. 30 is a conceptual diagram showing an arrangement according to a modification of the fourth embodiment of the MTJ memory cell.
30 shows a memory array in which MTJ memory cells MCD ′ having the configuration shown in FIG. 90 are integrated and arranged.

  Referring to FIG. 30, in the modification of the fourth embodiment, in the MTJ memory cells arranged in a matrix, memory cells MCD ′ adjacent in the column direction share the same word line WL. For example, the memory cell MCD ′ belonging to the first memory cell row and the memory cell MCD ′ belonging to the second memory cell row share the same word line WL1.

  With such a configuration, the number of word lines WL in the entire memory array can be reduced, the MTJ memory cells can be highly integrated, and the chip area can be reduced.

  Referring to FIG. 91 again, also in the MTJ memory cell shown in FIG. 90, the distance between word line WL and magnetic tunnel junction MTJ is the distance between bit line BL and magnetic tunnel junction MTJ. Therefore, a large data write current needs to flow through the word line WL. Therefore, in such an MTJ memory cell, it is important for securing the operation reliability to reduce the current density of the word line WL.

  In the modification of the fourth embodiment, since the wiring pitch of the word lines WL that needs to pass a larger data write current can be easily secured, the current density of the word lines WL is suppressed and the reliability of the MRAM device is increased. Improvements can be made. Further, as described in the fourth embodiment, the operation reliability of the MARA device is further improved by selecting a wiring material that requires a larger data write current to have a high electromigration resistance. be able to.

[Embodiment 5]
In the fifth and subsequent embodiments, the high integration of the memory array by adopting a configuration in which the read word line RWL and the write word line WWL are arranged along different directions will be described.

  FIG. 31 is a schematic block diagram showing an overall configuration of MRAM device 2 according to the fifth embodiment of the present invention.

  Referring to FIG. 31, in MRAM device 2, read word line RWL and write word line WWL are arranged on memory array 10 along the row direction and the column direction, respectively.

  Correspondingly, the bit line is divided into a read bit line RBL and a write bit line WBL, and arranged on the memory array 10 along the column direction and the row direction, respectively.

  Therefore, the MRAM device 2 is different from the MRAM device 1 shown in FIG. 1 in that the word line driver 30 is divided into a read word line driver 30r and a write word line driver 30w.

  Further, read / write control circuits 50 and 60 are also divided into write control circuits 50w and 60w and read control circuit 50r arranged adjacent to memory array 10 in the row direction.

  Since the configuration and operation of other parts are the same as those of MRAM 1, detailed description will not be repeated.

FIG. 32 is a circuit diagram showing a connection mode of MTJ memory cells according to the fifth embodiment.
Referring to FIG. 32, read word line RWL, write word line WWL, write bit line WBL and read bit line RBL are provided for the MTJ memory cell having magnetic tunnel junction MTJ and access transistor ATR. As the access transistor ATR, a MOS transistor which is a field effect transistor formed on the semiconductor substrate SUB is typically applied.

  Access transistor ATR has its gate coupled to read word line RWL. Access transistor ATR is turned on when read word line RWL is activated to a selected state (H level: power supply voltage Vcc) to form a current path including magnetic tunnel junction MTJ. On the other hand, when read word line RWL is deactivated to a non-selected state (L level: ground voltage Vss), access transistor ATR is turned off, so that a current path including magnetic tunnel junction MTJ is not formed.

  The write word line WWL and the write bit line WBL are arranged in directions orthogonal to each other so as to be close to the magnetic tunnel junction MTJ. Thus, by arranging the read word line RWL and the write word line WWL in directions orthogonal to each other, the read word line driver 30r and the write word line driver 30w can be divided and arranged.

  Further, the write word line WWL can be arranged independently without being coupled to other parts of the MTJ memory cell, so that the improvement in magnetic coupling with the magnetic tunnel junction MTJ is prioritized. can do. Thereby, the data write current Ip flowing through the write word line WWL can be suppressed.

  Since read word line RWL and write word line WWL are activated independently at the time of data reading and data writing, these drivers can be designed as originally independent. Therefore, the write word line driver 30w and the read word line driver 30r can be divided and reduced in size, and can be arranged in different regions adjacent to the memory array 10, thereby improving the flexibility of layout and improving the layout area. That is, the chip area of the MRAM device can be reduced.

  Magnetic tunnel junction MTJ is electrically coupled between read bit line RBL and access transistor ATR. Therefore, the read bit line RBL to the magnetic tunnel junction MTJ are set in response to the turn-on of the access transistor ATR by setting the voltage level of the write bit line WBL that does not require a current to flow at the time of data reading to the ground voltage Vss. A current path of access transistor ATR to write bit line WBL (ground voltage Vss) is formed. By flowing the sense current Is through this current path, a voltage change corresponding to the level of the storage data of the magnetic tunnel junction MTJ is generated in the read bit line RBL, and the storage data can be read.

  At the time of data writing, data write currents are supplied to write word line WWL and write bit line WBL, respectively, and the sum of the magnetic fields generated by these data write currents is a constant magnetic field, that is, an asteroid shown in FIG. By reaching the region exceeding the characteristic line, the stored data is written in the magnetic tunnel junction MTJ.

  FIG. 33 is a timing chart for describing data writing and data reading with respect to the MTJ memory cell according to the fifth embodiment.

First, the operation at the time of data writing will be described.
The write word line driver 30w drives the voltage of the write word line WWL corresponding to the selected column to the selected state (H level) according to the column selection result of the column decoder 25. In the non-selected column, the voltage level of write word line WWL is maintained in the non-selected state (L level). Since each write word line WWL is coupled to the ground voltage Vss by the word line current control circuit 40, the data write current Ip flows through the write word line WWL in the selected column.

  Read word line RWL is maintained in a non-selected state (L level) during data writing. At the time of data writing, read control circuit 50r does not supply sense current Is and precharges read bit line RBL to a high voltage state (Vcc). Since access transistor ATR maintains a turn-off state, no current flows through read bit line RBL during data writing.

  Write control circuits 50w and 60w generate a data write current in a direction according to the data level of write data DIN by controlling the voltage of write bit line WBL at both ends of memory array 10.

  For example, when the stored data “1” is written, the bit line voltage on the write control circuit 60 w side is set to a high voltage state (power supply voltage Vcc), and the bit line on the opposite write control circuit 50 w side is set. The voltage is set to a low voltage state (ground voltage Vss). Thereby, the data write current + Iw flows through the write bit line WBL in the direction from the write control circuit 60w to 50w.

  On the other hand, when writing the stored data of “0”, the bit line voltages on the write control circuit 50w side and 60w side are set to the high voltage state and the low voltage state, respectively, and the write control circuit 50w goes to 60w. Data write current -Iw flows through write bit line WBL in the direction. At this time, the data write current ± Iw is selectively supplied to the write bit line WBL corresponding to the selected row in accordance with the row selection result of the row decoder 20.

  Thus, by setting the direction of data write currents Ip and ± Iw, data write in the reverse direction is performed according to the level “1” or “0” of the stored data to be written at the time of data write. By selecting one of currents + Iw and -Iw, data write current Ip of write word line WWL can be fixed in a fixed direction regardless of the data level. As a result, the direction of the data write current Ip flowing through the write word line WWL can always be made constant, so that the configuration of the word line current control circuit 40 can be simplified as described above.

Next, the data read operation will be described.
At the time of data reading, write word line WWL is maintained in a non-selected state (L level), and its voltage level is fixed to ground voltage Vss by word line current control circuit 40. At the time of data reading, write control circuits 50w and 60w stop supplying data write current to write bit line WBL and set write bit line WBL to ground voltage Vss.

  On the other hand, the read word line driver 30r drives the read word line RWL corresponding to the selected row to the selected state (H level) according to the row selection result of the row decoder 20. In a non-selected row, the voltage level of read word line RWL is maintained in a non-selected state (L level). Read control circuit 50r supplies a certain amount of sense current Is for executing data read to read bit line RBL of the selected column during data read. Since read bit line RBL is precharged to a high voltage state (Vcc) before data reading, the turn-on of access transistor ATR in response to activation of read word line RWL causes the current path of sense current Is to be MTJ memory. A voltage change (drop) corresponding to the stored data occurs in the read bit line RBL.

  In FIG. 33, when the data level stored as an example is “1” and the magnetic field directions in the fixed magnetic layer FL and the free magnetic layer VL are the same, the stored data is “1”. The voltage change ΔV1 of the read bit line RBL is small, and the voltage change ΔV2 of the read bit line RBL when the stored data is “0” is larger than ΔV1. By detecting the difference between these voltage drops ΔV1 and ΔV2, the data stored in the MTJ memory cell can be read.

  In read bit line RBL, the precharge voltage for data reading and the set voltage for data writing are set to the same power supply voltage Vcc, so that the precharge operation at the start of data reading is made efficient. Therefore, the data reading operation can be speeded up. Even when the precharge voltage of the read bit line RBL is set to the ground line voltage Vss, the set voltage at the time of data writing may be set to the ground voltage Vss.

  Similarly, for the write bit line WBL that needs to be set to the ground voltage Vss at the time of data read, the data read operation can be speeded up by making the set voltage after the completion of data writing equal to the ground voltage Vss.

FIG. 34 is a structural diagram illustrating the arrangement of MTJ memory cells according to the fifth embodiment.
Referring to FIG. 34, access transistor ATR is formed in p-type region PAR on semiconductor substrate SUB. Write bit line WBL is formed in first metal interconnection layer M1, and is electrically coupled to one of source / drain regions 110 of access transistor ATR. The other source / drain region 120 is electrically connected to the magnetic tunnel junction MTJ via the metal wiring provided in the first metal wiring layer M1, the barrier metal 140, and the metal film 150 formed in the contact hole. Combined.

  The read bit line RBL is provided in the third metal wiring layer M3 so as to be electrically coupled to the magnetic tunnel junction MTJ. The write word line WWL is arranged in the second metal wiring layer M2. The write word line WWL can be arranged independently without being coupled to other parts of the MTJ memory cell, so that the magnetic coupling with the magnetic tunnel junction MTJ can be increased. Can be arranged.

  With such a configuration, the read word line RWL and the write word line WWL are arranged in directions perpendicular to each other with respect to the MTJ memory cell, and the read word line RWL and the write word line WWL are respectively read. The word line driver 30r and the write word line driver 30w can be arranged independently to increase the flexibility of layout. It is possible to prevent the word line drive current from becoming excessive at the time of data reading, and to prevent unnecessary magnetic noise from occurring.

  FIG. 35 is a diagram for describing a configuration according to the fifth embodiment of memory array 10 and its peripheral circuits.

  Referring to FIG. 35, in memory array 10 according to the fifth embodiment, memory cells MC having the configuration shown in FIG. 32 are arranged in a matrix. Read word line RWL and write word line WWL are arranged along the row direction and the column direction, respectively, and read bit line RBL and write bit line WBL are arranged along the column direction and the row direction, respectively.

  Word line current control circuit 40 couples each write word line WWL to ground voltage Vss. Thereby, the voltage and current of write word line WWL at the time of data reading and data writing can be controlled as shown in FIG.

  Memory cells adjacent in the row direction share the read bit line RBL. Memory cells adjacent in the column direction share the write bit line WBL.

  For example, the memory cell groups belonging to the first and second memory cell columns share the same read bit line RBL1, and the memory cell groups belonging to the third and fourth memory cell columns are the same. The read bit line RBL2 is shared. Further, the write bit line WBL2 is shared by the memory cell groups belonging to the second and third memory cell rows. For the subsequent memory cell rows and memory cell columns, read bit line RBL and write bit line WBL are similarly arranged.

  Corresponding to the same read bit line RBL or write bit line WBL, data collision occurs when a plurality of memory cells MC are subjected to data reading or data writing, so that the memory cells MC are alternately arranged.

  With such a configuration, the wiring pitch of the read bit line RBL and the write bit line WBL in the memory array 10 can be relaxed. As a result, the memory cells MC can be efficiently arranged to highly integrate the memory array 10 and the chip area of the MRAM device can be reduced.

  Next, the configuration of a peripheral circuit for flowing sense current Is and data write current ± Iw will be described.

  Column selection related to data reading is performed by a read column selection line RCSL and a read column selection gate RCSG provided for each read bit line RBL. FIG. 35 representatively shows read column select lines RCSL1, RCSL2 and read column select gates RCSG1, RCSG2 provided corresponding to read bit lines RBL1 and RBL2.

  Column decoder 25 activates one of a plurality of read column selection lines RCSL to a selected state (H level) in accordance with a column selection result at the time of data reading.

  The read column selection gate RCSG connects the read data line RDL and the corresponding read bit line RBL according to the voltage of the corresponding read column selection line RCSL. A sense current Is is supplied to the read data line RDL by the data read circuit 55e.

FIG. 36 is a circuit diagram showing a configuration of data read circuit 55e.
Referring to FIG. 36, data read circuit 55e differs from data read circuit 55d shown in FIG. 15 in that sense current Is is supplied only to node Nr1. Correspondingly, the transistor 164 shown in FIG. 15 is omitted, and the reference voltage Vrr is input only to the gate of the transistor 163.

  Data read circuit 55e compares the voltage drop caused by sense current Is with reference voltage drop ΔVr to detect the data level of read data DOUT. ΔVr is set to be an intermediate value between ΔVh and ΔVl, where ΔVh is the voltage drop of the data line when H level data is read and ΔVl is the voltage drop of the data line when L level data is read. Is done.

  Therefore, in data read circuit 55e, the resistance value of resistor 167 is set so that the voltage level of node Ns2 becomes (Vcc−ΔVr).

  Referring to FIG. 35 again, sense current Is is selectively supplied to read bit line RBL corresponding to the column selection result via read column selection gate RCSG.

  In accordance with the row selection result, read word line driver 30r selectively activates read word line RWL. Thus, the sense current Is can be supplied to the MTJ memory cell corresponding to the selected memory cell row.

  On the other hand, column selection related to data writing is executed by selective activation of the write word line WWL by the write word line driver 30w according to the column selection result. Each write word line WWL is coupled to ground voltage Vss in word line current control circuit 40.

  Write bit line WBL is provided corresponding to the memory cell row in a direction orthogonal to write word line WWL. Therefore, row selection relating to data writing is executed by the write row selection line and write row selection gate provided for each write bit line WBL.

  FIG. 35 representatively shows write row selection lines WRSL1, WRSL2 and write row selection gates WRSG1, WRSG2 provided corresponding to write bit lines WBL1 and WBL2. In the following, when the read row selection line and the write row selection gate are collectively described, the symbols WRSL and WRSG are used, respectively.

  Write row select gate WRSG is electrically coupled between corresponding write bit line WBL and write data line WDL, and is turned on / off according to the voltage of corresponding write row select line WRSL.

  Read / write control circuit 60 includes bit line current control transistors arranged corresponding to write bit line WBL. FIG. 35 representatively shows bit line current control transistors 63-1 and 63-2 provided corresponding to write bit lines WBL1 and WBL2, respectively. In the following, when these bit line current control transistors are collectively referred to, reference numeral 63 is used.

  Bit line current control transistor 63 is electrically coupled between corresponding write bit line WBL and write data line / WDL, and is turned on / off according to the voltage of corresponding write row selection line WRSL.

  Data write current ± Iw is supplied to write data lines WDL and / WDL by data write current 51b shown in FIG. Therefore, data write current ± Iw can be supplied to write bit line WBL corresponding to the selected memory cell row in accordance with the row selection result in row decoder 20.

  Read / write control circuit 60 further includes a precharge transistor arranged corresponding to read bit line RBL and a write bit line voltage control transistor arranged corresponding to write bit line WBL.

  In FIG. 35, precharge transistors 64-1 and 64-2 provided corresponding to read bit lines RBL1 and RBL2, and write bit line voltage control transistor 65- provided respectively corresponding to write bit lines WBL1 and WBL2. 1,65-2 is representatively shown. In the following, the reference numeral 65 is used to collectively refer to the plurality of write bit line voltage control transistors.

  Each of write bit line voltage control transistors 65 is turned on during data reading, and couples corresponding write bit line WBL to ground voltage Vss in order to secure a current path for sense current Is. Other than during data reading, each write bit line voltage control transistor 65 is turned off, and each write bit line WBL is disconnected from the ground voltage Vss. The operation of precharge transistor 64 is similar to that described with reference to FIG. 2, and therefore description thereof will not be repeated.

  With such a configuration, at the time of data writing, write data line WDL to write row selection gate WRSG to write bit line WBL to bit line are applied to write bit line WBL corresponding to the selected memory cell row. Data write current ± Iw can be supplied to the path from current control transistor 63 to write data line / WDL. Note that the direction of the data write current ± Iw can be controlled by setting the voltages of the write data lines WDL and / WDL in the same manner as the write data buses WDB and / WDB in the first embodiment. Therefore, as in the first embodiment, the configuration of peripheral circuits related to data writing, that is, data write circuit 50w and read / write control circuit 60 can be simplified.

  As described above, even in a configuration in which the read word line RWL and the write word line WWL are arranged orthogonally and the write bit line WBL and the read bit line RBL are shared between adjacent memory cells, the data writing as shown in FIG. And data reading can be performed.

  With such a configuration, the wiring pitch of the write bit line WBL and the read bit line RBL in the memory array 10 can be relaxed. As a result, the memory cells MC can be efficiently arranged to highly integrate the memory array 10 and the chip area of the MRAM device can be reduced.

  Further, by reducing the wiring pitch of the write bit line WBL, the wiring width of the write bit line WBL can be secured wider. Thereby, the effect described below further occurs.

  As already described, at the time of data writing, it is necessary to pass a data write current through both write bit line WBL and write word line WWL.

As shown in FIG. 34, in the configuration of the MTJ memory cell according to the fifth embodiment, the distance between write bit line WBL and magnetic tunnel junction MTJ in the height direction is equal to write word line WWL and magnetic tunnel junction. Greater than distance to MTJ. Therefore, at the time of data writing, it is necessary to pass a larger current to write bit line WBL having a large distance from magnetic tunnel junction MTJ.

However, since the write bit line WBL is shared between adjacent memory cell columns, the write bit line WBL can be arranged using an arrangement space for two memory cell rows. Therefore, it is possible to suppress the current density by increasing the wiring width of each write bit line WBL to ensure at least a wider wiring width than the write word line WWL, that is, a large cross-sectional area.

  As described above, by configuring one wiring having a larger distance from the magnetic tunnel junction MTJ among the wirings through which the data write current flows, the memory cells of the MRAM device are shared. Reliability can be improved.

  In addition, it is effective in improving reliability to form a metal wiring (write bit line WBL in FIG. 34) having a large distance from the magnetic tunnel junction MTJ with a material having high electromigration resistance. For example, when other metal wiring is formed of an aluminum alloy (Al alloy), the metal wiring that needs to be considered for electromigration resistance may be formed of copper (Cu).

[Modification 1 of Embodiment 5]
FIG. 37 is a diagram for describing a configuration according to the first modification of the fifth embodiment of memory array 10 and its peripheral circuits.

  Referring to FIG. 37, in memory array 10 according to the first modification of the fifth embodiment, adjacent memory cells share the same write word line WWL. For example, the memory cell groups belonging to the first and second memory cell columns share one write word line WWL1. The write word lines WWL are similarly arranged for the subsequent memory cell columns.

  Here, in order to perform data writing normally, it is necessary that a plurality of memory cells MC arranged at the intersection of the same write word line WWL and the same write bit line WBL do not exist. Therefore, the memory cells MC are arranged alternately.

  Since the configuration of peripheral circuits related to data writing and data reading for read bit line RBL and write bit line WBL and the operation of each memory cell at the time of data reading and data writing are the same as in the fifth embodiment, Detailed description will not be repeated.

  With such a configuration, the wiring pitch of the write word lines WWL in the memory array 10 can be relaxed. As a result, the memory cells MC can be efficiently arranged to highly integrate the memory array 10 and the chip area of the MRAM device can be reduced.

[Modification 2 of Embodiment 5]
FIG. 38 is a diagram for describing a configuration according to the second modification of the fifth embodiment of memory array 10 and its peripheral circuits.

  Referring to FIG. 38, in memory array 10 according to the second modification of the fifth embodiment, compared with the configuration according to the first modification of the fifth embodiment, the same read word is used by the memory cells adjacent in the column direction. Line RWL is further shared. For example, the memory cell groups belonging to the first and second memory cell rows share the same read word line RWL1. The read word line RWL is similarly arranged for the subsequent memory cell rows.

  Here, in order to perform data reading and data writing normally, a plurality of memory cells MC selected by one read word line RWL or write word line WWL are connected to the same read bit line RBL or write bit. It is necessary not to be simultaneously coupled to the line WBL. Therefore, read bit line RBL and write bit line WBL are arranged for each memory cell column and each memory cell row, respectively, and memory cells MC are alternately arranged.

  Since the structure of other parts and the operation of each memory cell at the time of data reading and data writing are the same as those in the fifth embodiment, detailed description will not be repeated.

  With such a configuration, the wiring pitch of the read word line RWL and the write word line WWL in the memory array 10 can be relaxed. As a result, the memory cells MC can be efficiently arranged to highly integrate the memory array 10 and the chip area of the MRAM device can be reduced.

[Modification 3 of Embodiment 5]
FIG. 39 is a block diagram showing structures of memory array 10 and peripheral circuits according to the third modification of the fifth embodiment.

  Referring to FIG. 39, two memory cells corresponding to each group of memory cell columns formed by two adjacent memory cell columns are arranged for the memory cells according to the fifth embodiment arranged in a matrix. A folded bit line configuration is realized using the read bit line RBL. For example, a read bit line pair can be formed by read bit lines RBL1 and RBL2 corresponding to the first and second memory cell columns, respectively. In this case, the read bit line RBL2 is provided in a complementary manner with the read bit line RBL1, and is also referred to as a read bit line / RBL1.

  In the following, among the read bit lines constituting each read bit line pair, one corresponding to the odd-numbered memory cell column and the other corresponding to the even-numbered memory cell column are referred to as read bit line RBL and Also collectively referred to as / RBL.

  The read column selection line is provided for each read bit line pair, that is, for each set of memory cell columns. Therefore, the two read column selection gates RCSG corresponding to the same set are turned on / off in response to the common read column selection line RCSL.

  For example, the read column selection gates RCSG1 and RCSG2 corresponding to the first and second memory cell columns operate according to the common read column selection line RCSL1. Read column selection gates RCSG1, RCSG3,... Provided corresponding to odd-numbered read bit lines RBL are electrically coupled between corresponding read bit lines RBL and read data lines RDL. On the other hand, read column selection gates RCSG2, RCSG4,... Provided corresponding to even-numbered read bit lines / RBL are electrically coupled between corresponding read bit lines / RBL and read data lines / RDL. .

  In response to the read column selection line RCSL activated according to the column selection result, the corresponding two read column selection gates RCSG are turned on. As a result, read bit lines RBL and / RBL constituting the read bit line pair corresponding to the selected memory cell column are electrically coupled to read data lines RDL and / RDL constituting the read data line pair.

  Further, a precharge transistor 64 similar to that described with reference to FIG. 35 is arranged corresponding to each of read bit lines RBL and / RBL. As already described, precharge transistor 64 is turned off during data reading.

  As a result, sense current Is supplied by data read circuit 55d flows through each of read bit lines RBL and / RBL corresponding to the selected memory cell column. Since the configuration of data read circuit 55d has already been shown in FIG. 15, detailed description will not be repeated.

  Therefore, data reading is performed using dummy memory cell DMC similar to that of the first embodiment, which can be selectively coupled to one of read bit lines RBL and / RBL. Thus, a data read margin can be ensured based on a so-called folded bit line configuration.

  Similarly, a folded bit line configuration is realized by using two corresponding write bit lines WBL for each set of memory cell rows formed by two adjacent memory cell rows. For example, a write bit line pair can be configured by the write bit lines WBL1 and WBL2 corresponding to the first and second memory cell rows, respectively. In this case, the write bit line WBL2 is provided in a complementary manner to the write bit line WBL1, and is also referred to as a write bit line / WBL1.

  Similarly for the subsequent memory cell columns, each read bit line RBL and write bit line WBL are arranged so as to form a write bit line pair and a read bit line pair for each pair of memory cell columns and rows. .

  Of the write bit lines constituting each write bit line pair, one each corresponding to an odd-numbered memory cell row and the other corresponding to an even-numbered memory cell column are also collectively referred to as write bit lines WBL and / WBL. To do. Thus, data writing can be executed based on a so-called folded bit line configuration.

  The write row selection line WRSL is provided for each write bit line pair, that is, for each set of memory cell rows. Therefore, the two write row selection gates WRSG corresponding to the same set are turned on / off in response to the common write row selection line WRSL.

  For example, the write row selection gates WRSG1 and WRSG2 corresponding to the first and second memory cell rows operate according to the common write row selection line WRSL1.

  Write row selection gates WRSG1, WRSG3,... Provided corresponding to odd-numbered write bit lines WBL are electrically coupled between corresponding write bit lines WBL and write data lines WDL. On the other hand, write row selection gates WRSG2, WRSG4,... Provided corresponding to even-numbered write bit lines / WBL are electrically coupled between corresponding write bit lines / WBL and write data lines / WDL.

  In response to the write row selection line WRSL activated according to the row selection result, the corresponding two write row selection gates WRSG are turned on. As a result, write bit lines WBL and / WBL constituting the write bit line pair corresponding to the selected memory cell row are electrically coupled to write data lines WDL and / WDL constituting the write data line pair, respectively. The

  Further, in each write bit line pair, an equalize transistor 62 for connecting write bit lines WBL and / WBL is arranged in place of bit line current control transistor 63 shown in FIG. Equalize transistor 62 operates in response to control signal WE, for example, and short-circuits between two bit lines constituting the same write bit line pair at the time of data writing. A write bit line voltage control transistor 65 similar to that described with reference to FIG. 35 is arranged corresponding to each of write bit lines WBL and / WBL.

  Data write current ± Iw is supplied from data write circuit 51b to write data lines WDL and / WDL constituting the write data line pair, similarly to write data buses WDB and / WDB in the first embodiment. The Since data write circuit 51b has already been shown in FIG. 7, detailed description will not be repeated.

  As a result, as in the first embodiment, data writing can be executed by the reciprocating current turned back by the equalizer transistor 62 in the write bit line pair corresponding to the row selection result.

  By adopting such a configuration, the selected read bit line pair performs data reading by flowing a sense current in the same manner as the bit line pair of the first embodiment at the time of data reading. Similarly, the selected write bit line pair performs data writing by passing a data write current through corresponding equalizer transistor 62 in the same manner as the bit line pair of the first embodiment at the time of data writing. .

  Therefore, when memory cells according to the fifth embodiment capable of reducing the chip area are arranged in a matrix, an operation margin for data reading and data writing can be secured by using the folded bit line configuration. it can.

[Modification 4 of Embodiment 5]
In the fourth modification of the fifth embodiment, in addition to the folded bit line configuration shown in the third modification of the fifth embodiment, the write bit line WBL is shared between adjacent memory cells.

  FIG. 40 is a diagram for describing a configuration according to a fourth modification of the fifth embodiment of memory array 10 and its peripheral circuits.

  Referring to FIG. 40, in memory array 10 according to the fourth modification of the fifth embodiment, the memory cells adjacent in the column direction share the same write bit line WBL.

  At the time of data reading in which the read word line RWL is activated, every other memory cell column is connected to each read bit line RBL, so that the memory cell formed by two adjacent memory cell columns A read bit line pair is formed for each set of columns, and data reading similar to that of the third modification of the fifth embodiment based on the folded bit line configuration can be executed.

  On the other hand, at the time of data writing, in order to share write bit line WBL, data writing based on the folded bit line configuration cannot be performed. Therefore, in the fourth modification of the fifth embodiment, peripheral circuits related to selection of write bit line WBL are arranged in the same manner as shown in FIG. Thereby, as in the case of the fifth embodiment, data writing can be executed using the data writing circuit 51b having a simple circuit configuration.

  In addition, although data writing based on the folded bit line configuration cannot be executed, the wiring pitch of the write bit lines WBL in the memory array 10 can be relaxed. As a result, the chip area of the MRAM device can be further reduced due to the high integration of the memory array 10. Further, the reliability of the MRAM device can be improved by improving the electromigration resistance of the write bit line WBL.

  In the configuration of FIG. 40, the configuration in which the write bit line WBL is shared among adjacent memory cells among the data write system signal wirings is shown, but the write word line WWL is shared instead of the write bit line WBL. A configuration is also possible. In this case, however, the write bit line WBL cannot be shared and must be arranged for each memory cell row. Which wiring is shared and the wiring pitch is relaxed may be determined in consideration of structural conditions such as the distance from the magnetic tunnel junction MTJ, design convenience, and the like.

[Modification 5 of Embodiment 5]
In the fifth modification of the fifth embodiment, in addition to the folded bit line configuration shown in the third modification of the fifth embodiment, the read word line RWL is shared between adjacent memory cells.

  FIG. 41 is a diagram for describing a configuration according to the fifth modification of the fifth embodiment of the memory array 10 and its peripheral circuits.

  Referring to FIG. 41, in memory array 10 according to the fifth modification of the fifth embodiment, memory cells adjacent in the column direction share the same read word line RWL.

  Read / write control circuit 60 includes an equalize transistor 62, a precharge transistor 64, and a write bit line voltage control transistor 65 arranged in the same manner as in the third modification of the fifth embodiment.

  At the time of data writing in which the write word line WWL is activated, every other memory cell column is connected to each write bit line WBL, so a memory formed by two adjacent memory cell rows. A write bit line pair can be formed for each set of cell rows. As a result, data writing similar to that of the third modification of the fifth embodiment based on the folded bit line configuration is executed, and the same effect can be obtained.

  On the other hand, at the time of data reading in which read word line RWL shared between a plurality of memory cell rows is activated, data reading based on the folded bit line configuration cannot be performed. Therefore, in the fifth modification of the fifth embodiment, peripheral circuits related to selection of read bit line RBL are arranged in the same manner as shown in FIG.

  With such a configuration, an operation margin cannot be ensured by the folded bit line configuration, but data read can be normally executed after the wiring pitch of the read word lines RWL in the memory array 10 is relaxed. As a result, the chip area of the MRAM device can be reduced due to the high integration of the memory array 10.

  Therefore, using the memory cell according to the fifth embodiment, the operation margin is secured, the peripheral circuit is simplified and the data write noise is reduced, and the read word line RWL is shared by data writing based on the folded bit line configuration. The high integration of the memory array 10 based on the above can be realized at the same time.

  In the configuration of FIG. 41, the configuration in which the read word line RWL is shared between adjacent memory cells among the signal wirings in the data read system is shown, but the configuration in which the read bit line RBL is shared instead of the read word line RWL. It is also possible. However, in this case, the read word line RWL cannot be shared and must be arranged for each memory cell row. Which wiring is shared and the wiring pitch is relaxed may be appropriately determined in consideration of structural conditions and design convenience.

[Embodiment 6]
FIG. 42 is a circuit diagram showing a connection mode of MTJ memory cells according to the sixth embodiment.

  Referring to FIG. 42, in the MTJ memory cell according to the sixth embodiment, the connection relationship between read bit line RBL and write bit line WBL is different compared to the MTJ memory cell shown in FIG. That is, the read bit line RBL is not directly coupled to the magnetic tunnel junction MTJ, but is coupled to the magnetic tunnel junction MTJ according to the turn-on of the access transistor ATR. Further, write bit line WBL is coupled to magnetic tunnel junction MTJ and is included in the sense current path during data reading.

  Since other components including the arrangement direction of each signal wiring are the same as those in FIG. 32, detailed description will not be repeated. Further, the voltage and current waveforms of each wiring in data writing and data reading are the same as those in FIG. 33, and therefore detailed description will not be repeated.

  Therefore, the write word line WWL is provided close to the magnetic tunnel junction MTJ in the direction orthogonal to the write bit line WBL. As a result, the read word line driver 30r and the write word line driver 30w are arranged independently, and the same effect as in the fifth embodiment can be obtained.

  Further, the write word line WWL can be arranged with priority on improving the magnetic coupling with the magnetic tunnel junction MTJ without being coupled to other parts of the MTJ memory cell.

  In addition, since the read bit line RBL is joined to the magnetic tunnel junction MTJ via the access transistor ATR, the number of magnetic tunnel junctions MTJ coupled to the read bit line RBL is reduced, and the read bit line RBL Capacitance can be reduced and data read can be performed at high speed.

FIG. 43 is a structural diagram illustrating the arrangement of MTJ memory cells according to the sixth embodiment.
Referring to FIG. 43, read bit line RBL is provided in first metal interconnection layer M1 so as to be electrically coupled to source / drain region 110 of access transistor ATR. Read word line RWL is arranged in the same layer as gate 130 of access transistor ATR. The source / drain region 120 of the access transistor ATR is connected to the magnetic tunnel junction via the metal wiring provided in the first and second metal wiring layers M1 and M2, the barrier metal 140, and the metal film 150 provided in the contact hole. Combined with the part MTJ.

  The magnetic tunnel junction MTJ is disposed between the second metal wiring layer M2 and the third metal wiring layer M3. The write bit line WBL is electrically coupled to the magnetic tunnel junction MTJ and disposed in the third metal wiring layer M3. The write word line WWL is provided in the second metal wiring layer M2. At this time, the write word line WWL is arranged so that magnetic coupling with the magnetic tunnel junction MTJ can be enhanced.

  In the MTJ memory cell according to the sixth embodiment, the distance between write bit line WBL and magnetic tunnel junction MTJ can be reduced as compared with the MTJ memory cell according to the fifth embodiment shown in FIG. . Therefore, the amount of data write current flowing through write bit line WBL can be reduced.

  The distance between the magnetic tunnel junction MTJ and the write word line WWL is greater than that of the write bit line WBL. Therefore, in the MTJ memory cell according to the sixth embodiment, the distance relative to the write word line WWL is relatively large. It is necessary to pass a large data write current.

  FIG. 44 is a diagram for describing a configuration according to the sixth embodiment of memory array 10 and its peripheral circuits.

  Referring to FIG. 44, in memory array 10 according to the sixth embodiment, memory cells MC having the configuration shown in FIG. 42 are arranged in a matrix. Read word line RWL and write word line WWL are arranged along the row direction and the column direction, respectively, and read bit line RBL and write bit line WBL are arranged along the column direction and the row direction, respectively.

  Memory cells adjacent in the row direction share the read bit line RBL. Memory cells adjacent in the column direction share the write bit line WBL.

  For example, the memory cell groups belonging to the first and second memory cell columns share the same read bit line RBL1, and the memory cell groups belonging to the third and fourth memory cell columns are the same. The read bit line RBL2 is shared. Further, the write bit line WBL2 is shared by the memory cell groups belonging to the second and third memory cell rows. For the subsequent memory cell rows and memory cell columns, read bit line RBL and write bit line WBL are similarly arranged.

  Corresponding to the same read bit line RBL or write bit line WBL, data collision occurs when a plurality of memory cells MC are subjected to data reading or data writing, so that the memory cells MC are alternately arranged.

  By adopting such a configuration, the wiring pitch of the read bit line RBL and the write bit line WBL in the memory array 10 can be relaxed as in the fifth embodiment. As a result, the memory cells MC can be efficiently arranged to highly integrate the memory array 10 and the chip area of the MRAM device can be reduced.

  Since the configuration of the peripheral circuit for selectively supplying the data write current and the sense current to read bit line RBL and write bit line WBL is the same as that of FIG. 35, detailed description will not be repeated.

[Modification 1 of Embodiment 6]
FIG. 45 is a diagram for describing a configuration according to the first modification of the sixth embodiment of memory array 10 and its peripheral circuits.

  Referring to FIG. 45, in memory array 10 according to the first modification of the sixth embodiment, adjacent memory cells share the same write word line WWL. For example, the memory cell groups belonging to the second and third memory cell columns share one write word line WWL2. The write word lines WWL are similarly arranged for the subsequent memory cell columns.

  Here, in order to perform data writing normally, it is necessary that a plurality of memory cells MC arranged at the intersection of the same write word line WWL and the same write bit line WBL do not exist. Therefore, the memory cells MC are arranged alternately.

  Further, similarly to the sixth embodiment, memory cells adjacent in the row direction share read bit line RBL.

  Since the configuration of peripheral circuits related to data writing and data reading for read bit line RBL and write bit line WBL and the operation of each memory cell at the time of data reading and data writing are the same as in the sixth embodiment, Detailed description will not be repeated.

  As already described, in the MTJ memory cell according to the sixth embodiment, it is necessary to pass a relatively large data write current to write word line WWL. Therefore, by sharing the write word line WWL between adjacent memory cells and securing the wiring pitch, the wiring width, that is, the cross-sectional area of the write word line WWL can be secured and the current density can be suppressed. As a result, the reliability of the MRAM device can be improved. Furthermore, as described above, selecting the material of these wirings in consideration of electromigration resistance is also effective in improving the operation reliability.

[Modification 2 of Embodiment 6]
FIG. 46 is a diagram for describing a configuration according to the second modification of the sixth embodiment of memory array 10 and its peripheral circuits.

  Referring to FIG. 46, in memory array 10 according to the second modification of the sixth embodiment, the same read word line RWL is further shared by memory cells adjacent in the column direction. For example, the memory cell groups belonging to the second and third memory cell rows share the same read word line RWL2. The read word line RWL is similarly arranged for the subsequent memory cell rows.

  Here, in order to perform data reading normally, it is necessary that a plurality of memory cells MC selected by the same read word line RWL are not simultaneously coupled to the same read bit line RBL. Therefore, the read bit line RBL is arranged for each memory cell column, and the memory cells MC are alternately arranged.

  With such a configuration, the wiring pitch of the write word lines WWL in the memory array 10 can be relaxed. As a result, the memory cells MC can be efficiently arranged to highly integrate the memory array 10 and the chip area of the MRAM device can be reduced.

[Modification 3 of Embodiment 6]
FIG. 47 is a diagram for describing a configuration according to the third modification of the sixth embodiment of the memory array 10 and its peripheral circuits.

  Referring to FIG. 47, for the memory cells having the structure according to the sixth embodiment arranged in a matrix, two corresponding memory cells are formed for each pair of memory cell columns formed by two adjacent memory cell columns. A folded bit line configuration is realized using the read bit line RBL. For example, a read bit line pair can be configured by read bit lines RBL1 and RBL2 (/ RBL1) corresponding to the first and second memory cell columns, respectively.

  Similarly, a folded bit line configuration is realized by using two corresponding write bit lines WBL for each set of memory cell rows formed by two adjacent memory cell rows. For example, a write bit line pair can be configured by write bit lines WBL1 and WBL2 (/ WBL1) corresponding to the first and second memory cell rows, respectively.

  Supply of row selection and data write current ± Iw to write bit lines WBL and / WBL constituting the write bit line pair, column selection and sense current Is for read bit lines RBL and / RBL constituting the read bit line pair Since the configuration of the peripheral circuit for supplying is the same as that of FIG. 39, detailed description will not be repeated.

  Therefore, even when the memory cells according to the sixth embodiment are arranged in a matrix, an operation margin for data reading and data writing can be secured using the folded bit line configuration.

[Modification 4 of Embodiment 6]
In the fourth modification of the sixth embodiment, in addition to the folded bit line configuration shown in the third modification of the sixth embodiment, the write bit line WBL is shared between adjacent memory cells.

  FIG. 48 is a diagram for describing a configuration according to the fourth modification of the sixth embodiment of the memory array 10 and its peripheral circuits.

  Referring to FIG. 48, in memory array 10 according to the fourth modification of the sixth embodiment, the memory cells adjacent in the column direction share the same write bit line WBL.

  At the time of data reading in which the read word line RWL is activated, every other memory cell column is connected to each read bit line RBL, so that the memory cell formed by two adjacent memory cell columns A read bit line pair is formed for each set of columns, and data reading similar to that of the third modification of the sixth embodiment based on the folded bit line configuration can be executed.

  On the other hand, at the time of data writing, in order to share write bit line WBL, data writing based on the folded bit line configuration cannot be performed. Therefore, in the fourth modification of the sixth embodiment, peripheral circuits related to selection of write bit line WBL are arranged in the same manner as shown in FIG. Thereby, as in the case of the sixth embodiment, data writing can be executed using the data writing circuit 51b having a simple circuit configuration.

  Further, although data writing based on the folded bit line configuration cannot be executed, the wiring pitch of the write bits WBL in the memory array 10 can be relaxed. As a result, the chip area of the MRAM device can be further reduced due to the high integration of the memory array 10.

  In the configuration of FIG. 48, the configuration in which the write bit line WBL is shared between adjacent memory cells among the data write system signal wirings is shown, but the configuration in which the write word line WWL is shared instead of the write bit line. It is also possible. In this case, however, the write bit line WBL cannot be shared and must be arranged for each memory cell row. Which wiring is shared and the wiring pitch is relaxed may be determined in consideration of the distance from the magnetic tunnel junction MTJ.

[Modification 5 of Embodiment 6]
In the fifth modification of the sixth embodiment, in addition to the folded bit line configuration shown in the third modification of the sixth embodiment, the read word line RWL is shared between adjacent memory cells.

  FIG. 49 is a diagram for describing a configuration according to the fifth modification of the sixth embodiment of the memory array 10 and its peripheral circuits.

  Referring to FIG. 49, in memory array 10 according to the fifth modification of the sixth embodiment, the memory cells adjacent in the column direction share the same read word line RWL.

  Read / write control circuit 60 includes an equalize transistor 62, a precharge transistor 64, and a write bit line voltage control transistor 65 arranged in the same manner as in the third modification of the sixth embodiment.

  At the time of data writing in which the write word line WWL is activated, every other memory cell column is connected to each write bit line WBL, so a memory formed by two adjacent memory cell rows. A write bit line pair can be formed for each set of cell rows. As a result, data writing similar to that of the third modification of the fifth embodiment based on the folded bit line configuration is executed, and the same effect can be obtained.

  On the other hand, at the time of data reading in which read word line RWL shared between a plurality of memory cell rows is activated, data reading based on the folded bit line configuration cannot be performed. Therefore, in the fifth modification of the sixth embodiment, peripheral circuits related to selection of read bit line RBL are arranged in the same manner as shown in FIG.

  With such a configuration, an operation margin cannot be ensured by the folded bit line configuration, but data read can be normally executed after the wiring pitch of the read word lines RWL in the memory array 10 is relaxed. As a result, the chip area of the MRAM device can be reduced due to the high integration of the memory array 10.

  Therefore, using the memory cell according to the sixth embodiment, the operation margin is secured, the peripheral circuit is simplified and the data write noise is reduced, and the read word line RWL is shared by data writing based on the folded bit line configuration. The high integration of the memory array 10 based on the above can be realized at the same time.

  In the configuration of FIG. 49, the read word line RWL is shared between adjacent memory cells among the signal wirings in the data read system, but the read bit line RBL is shared instead of the read word line RWL. It is also possible. In this case, however, the read word line RWL cannot be shared and must be arranged for each memory cell row. Which wiring is shared and the wiring pitch is relaxed may be appropriately determined in consideration of structural conditions and design convenience.

[Embodiment 7]
FIG. 50 is a circuit diagram showing a connection mode of MTJ memory cells according to the seventh embodiment.

  Referring to FIG. 50, read bit line RBL is coupled to magnetic tunnel junction MTJ through access transistor ATR. Magnetic tunnel junction MTJ is coupled between write word line WWL and access transistor ATR. Read word line RWL is coupled to the gate of access transistor ATR. Also in the configuration of FIG. 50, the read word line RWL and the write word line WWL are arranged in directions orthogonal to each other.

FIG. 51 is a structural diagram showing an arrangement of MTJ memory cells according to the seventh embodiment.
Referring to FIG. 51, read bit line RBL is arranged in metal interconnection layer M1. Read word line RWL is formed in the same layer as gate 130 of access transistor ATR. Read bit line RBL is coupled to source / drain region 110 of access transistor ATR. The source / drain region 120 is coupled to the magnetic tunnel junction MTJ through the metal wiring provided in the first and second metal wiring layers M1 and M2, the barrier metal 140, and the metal film 150 provided in the contact hole. The

  The write bit line WBL is provided in the second metal wiring layer M2 in the vicinity of the magnetic tunnel junction MTJ. The write word line WWL is electrically coupled to the magnetic tunnel junction MTJ and disposed in the third metal wiring layer M3.

  With such a configuration, read bit line RBL is coupled to magnetic tunnel junction MTJ through access transistor ATR. As a result, read bit line RBL is electrically coupled only to MTJ memory cells MC that belong to a data read target, that is, belong to a memory cell row in which corresponding read word line RWL is activated to a selected state (H level). Is done. As a result, the capacity of the read bit line RBL can be suppressed and the data read operation can be speeded up.

  In the MTJ memory cell according to the seventh embodiment, the voltage and current waveform of each wiring at the time of data writing and data reading are the same as those in FIG. 33, and therefore detailed description will not be repeated.

  Also in the MTJ memory cell according to the seventh embodiment, the distance between write bit line WBL and magnetic tunnel junction MTJ can be reduced as compared with the MTJ memory cell according to the fifth embodiment shown in FIG. . Therefore, the amount of data write current flowing through write bit line WBL can be reduced.

  In addition, since the distance from the magnetic tunnel junction MTJ is larger in the write bit line WBL than in the write word line WWL, in the MTJ memory cell according to the seventh embodiment, the write bit line WBL has a larger distance. It is necessary to pass a relatively large data write current.

  FIG. 52 is a diagram for describing a configuration according to the seventh embodiment of memory array 10 and its peripheral circuits.

  Referring to FIG. 52, in memory array 10 according to the seventh embodiment, memory cells MC having the configuration shown in FIG. 50 are arranged in a matrix. Read word line RWL and write word line WWL are arranged along the row direction and the column direction, respectively, and read bit line RBL and write bit line WBL are arranged along the column direction and the column direction, respectively.

  Memory cells adjacent in the row direction share the read bit line RBL. Memory cells adjacent in the column direction share the write bit line WBL.

  For example, the memory cell groups belonging to the first and second memory cell columns share the same read bit line RBL1, and the memory cell groups belonging to the third and fourth memory cell columns are the same. The read bit line RBL2 is shared. Further, the write bit line WBL2 is shared by the memory cell groups belonging to the second and third memory cell rows. For the subsequent memory cell rows and memory cell columns, read bit line RBL and write bit line WBL are similarly arranged.

  Corresponding to the same read bit line RBL or write bit line WBL, data collision occurs when a plurality of memory cells MC are subjected to data reading or data writing, so that the memory cells MC are alternately arranged.

  With such a configuration, the wiring pitch of the read bit line RBL and the write bit line WBL in the memory array 10 can be relaxed. As a result, the memory cells MC can be efficiently arranged to highly integrate the memory array 10 and the chip area of the MRAM device can be reduced.

  Since the configuration of the peripheral circuit for selectively supplying the data write current and the sense current to read bit line RBL and write bit line WBL is the same as that of FIG. 35, detailed description will not be repeated.

  As already described, in the MTJ memory cell according to the seventh embodiment, it is necessary to pass a relatively large data write current to write bit line WBL. Therefore, by sharing the write bit line WBL between adjacent memory cells and securing the wiring pitch, the wiring width, that is, the cross-sectional area of the write bit line WBL can be secured and the current density can be suppressed. As a result, the reliability of the MRAM device can be improved. Furthermore, as described above, selecting the material of these wirings in consideration of electromigration resistance is also effective in improving the operation reliability.

[Modification 1 of Embodiment 7]
FIG. 53 is a diagram for describing a configuration according to the first modification of the seventh embodiment of the memory array 10 and its peripheral circuits.

  Referring to FIG. 53, in memory array 10 according to the first modification of the seventh embodiment, adjacent memory cells share the same write word line WWL. For example, the memory cell groups belonging to the second and third memory cell columns share one write word line WWL2. The write word lines WWL are similarly arranged for the subsequent memory cell columns.

  Here, in order to perform data writing normally, it is necessary that a plurality of memory cells MC arranged at the intersection of the same write word line WWL and the same write bit line WBL do not exist. Therefore, the memory cells MC are arranged alternately.

  Further, similarly to the seventh embodiment, the memory cells adjacent in the row direction share the read bit line RBL.

  Since the configuration of peripheral circuits related to data writing and data reading for read bit line RBL and write bit line WBL and the operation of each memory cell at the time of data reading and data writing are the same as in the seventh embodiment, Detailed description will not be repeated.

  With such a configuration, the wiring pitch of the read bit line RBL and the write word line WWL in the memory array 10 can be relaxed. As a result, the memory cells MC can be efficiently arranged to highly integrate the memory array 10 and the chip area of the MRAM device can be reduced.

[Modification 2 of Embodiment 7]
FIG. 54 is a diagram for describing a configuration according to the second modification of the seventh embodiment of the memory array 10 and its peripheral circuits.

  Referring to FIG. 54, in memory array 10 according to the second modification of the seventh embodiment, the same read word line RWL is shared by memory cells adjacent in the column direction. For example, the memory cell groups belonging to the second and third memory cell rows share the same read word line RWL2. The read word line RWL is similarly arranged for the subsequent memory cell rows.

  Furthermore, the same write word line WWL is shared by memory cells adjacent in the row direction. For example, the memory cell groups belonging to the second and third memory cell columns share the same write word line WWL2. The write word line RWL is similarly arranged for the subsequent memory cell columns.

  Here, in order to perform data reading and data writing normally, a plurality of memory cells MC selected by the same read word line RWL or write word line WWL are connected to the same read bit line RBL or write bit line WBL. Must not be simultaneously bonded. Therefore, read bit line RBL and write bit line WBL are arranged for each memory cell column and each memory cell row, respectively, and memory cells MC are alternately arranged.

Since the configuration of other parts is the same as that of the seventh embodiment, detailed description will not be repeated.
With such a configuration, the wiring pitch of the write word line WWL and the read word line RWL in the memory array 10 can be relaxed. As a result, the memory cells MC can be efficiently arranged to highly integrate the memory array 10 and the chip area of the MRAM device can be reduced.

[Modification 3 of Embodiment 7]
FIG. 55 is a diagram for describing a configuration according to a third modification of the seventh embodiment of memory array 10 and its peripheral circuits.

  Referring to FIG. 55, two memory cells corresponding to each set of memory cell columns formed by two adjacent memory cell columns are arranged for the memory cells according to the seventh embodiment arranged in a matrix. A folded bit line configuration is realized using the read bit line RBL. For example, a read bit line pair can be configured by read bit lines RBL1 and RBL2 (/ RBL1) corresponding to the first and second memory cell columns, respectively.

  Similarly, a folded bit line configuration is realized by using two corresponding write bit lines WBL for each set of memory cell rows formed by two adjacent memory cell rows. For example, a write bit line pair can be configured by write bit lines WBL1 and WBL2 (/ WBL1) corresponding to the first and second memory cell rows, respectively.

  Supply of row selection and data write current ± Iw to write bit lines WBL and / WBL constituting the write bit line pair, column selection and sense current Is for read bit lines RBL and / RBL constituting the read bit line pair Since the configuration of the peripheral circuit for supplying is the same as that of FIG. 39, detailed description will not be repeated.

  Therefore, even when memory cells according to the seventh embodiment are arranged in a matrix, an operation margin for data reading and data writing can be secured using the folded bit line configuration.

[Modification 4 of Embodiment 7]
In the fourth modification of the seventh embodiment, in addition to the folded bit line configuration shown in the third modification of the seventh embodiment, the write word line WWL is shared between adjacent memory cells.

  FIG. 56 is a diagram for describing a configuration according to the fourth modification of the seventh embodiment of the memory array 10 and its peripheral circuits.

  Referring to FIG. 56, in memory array 10 according to the fourth modification of the seventh embodiment, the memory cells adjacent in the row direction share the same write word line WWL.

  At the time of data reading in which the read word line RWL is activated, every other memory cell column is connected to each read bit line RBL, so that the memory cell formed by two adjacent memory cell columns A read bit line pair is formed for each set of columns, and data reading similar to that of the third modification of the seventh embodiment based on the folded bit line configuration can be executed.

  On the other hand, at the time of data writing, since the write word line WWL is shared, data writing based on the folded bit line configuration cannot be performed. Therefore, in the fourth modification of the seventh embodiment, peripheral circuits related to selection of write bit line WBL are arranged in the same manner as shown in FIG. Thereby, as in the case of the seventh embodiment, data writing can be executed using the data writing circuit 51b having a simple circuit configuration.

  Further, although data writing based on the folded bit line configuration cannot be executed, the wiring pitch of the write word lines WWL in the memory array 10 can be relaxed. As a result, the chip area of the MRAM device can be further reduced due to the high integration of the memory array 10.

  In the configuration of FIG. 56, the configuration in which the write word line WWL is shared between adjacent memory cells among the signal wirings in the data write system is shown, but the configuration in which the write bit line WBL is shared instead of the write word line. It is also possible. However, in this case, the write word line WWL cannot be shared and must be arranged for each memory cell column. Which wiring is shared and the wiring pitch is relaxed may be determined in consideration of the distance from the magnetic tunnel junction MTJ.

[Modification 5 of Embodiment 7]
In the fifth modification of the seventh embodiment, in addition to the folded bit line configuration shown in the third modification of the seventh embodiment, the read word line RWL is shared between adjacent memory cells.

  FIG. 57 is a diagram for describing a configuration according to the fifth modification of the seventh embodiment of the memory array 10 and its peripheral circuits.

  Referring to FIG. 57, in memory array 10 according to the fifth modification of the seventh embodiment, the memory cells adjacent in the column direction share the same read word line RWL.

  Read / write control circuit 60 includes an equalize transistor 62, a precharge transistor 64, and a write bit line voltage control transistor 65 arranged in the same manner as in the third modification of the seventh embodiment.

  At the time of data writing in which the write word line WWL is activated, every other memory cell column is connected to each write bit line WBL, so a memory formed by two adjacent memory cell rows. A write bit line pair can be formed for each set of cell rows. As a result, data writing similar to that of the third modification of the fifth embodiment based on the folded bit line configuration is executed, and the same effect can be obtained.

  On the other hand, at the time of data reading in which read word line RWL shared between a plurality of memory cell rows is activated, data reading based on the folded bit line configuration cannot be performed. Therefore, in the fifth modification of the seventh embodiment, peripheral circuits related to column selection of read bit line RBL are arranged in the same manner as shown in FIG.

  With such a configuration, the operation margin cannot be ensured by the folded bit line configuration, but the data read can be normally executed after the wiring pitch of the read word lines RWL in the memory array 10 is relaxed. As a result, the chip area of the MRAM device can be reduced due to the high integration of the memory array 10.

  Therefore, using the memory cell according to the seventh embodiment, the operation margin is secured, the peripheral circuit is simplified and the data write noise is reduced, and the read word line RWL is shared by data writing based on the folded bit line configuration. The high integration of the memory array 10 based on the above can be realized at the same time.

  In the configuration of FIG. 57, the configuration in which the read word line RWL is shared between adjacent memory cells among the signal wirings in the data read system is shown, but the configuration in which the read bit line RBL is shared instead of the read word line RWL. It is also possible. In this case, however, the read word line RWL cannot be shared and must be arranged for each memory cell row. Which wiring is shared and the wiring pitch is relaxed may be appropriately determined in consideration of structural conditions and design convenience.

[Embodiment 8]
FIG. 58 is a circuit diagram showing a connection mode of MTJ memory cells according to the eighth embodiment.

  Referring to FIG. 58, the MTJ memory cell according to the eighth embodiment replaces the arrangement of read bit line RBL and write word line WWL as compared with the MTJ memory cell according to the seventh embodiment shown in FIG. It has a configuration. Since the arrangement of the other wirings is the same as that in FIG. Even in such a configuration, the read word line RWL and the write word line WWL can be arranged in directions orthogonal to each other.

FIG. 59 is a structural diagram showing an arrangement of MJT memory cells according to the eighth embodiment.
Referring to FIG. 59, in the MTJ memory cell according to the eighth embodiment, write word line WWL and read bit line RBL are arranged as compared with the structure of MTJ memory cell according to the seventh embodiment shown in FIG. The position is changed. That is, the write word line WWL is provided in the first metal wiring layer M1 and coupled to the source / drain region 110 of the access transistor ATR. On the other hand, the read bit line RBL is provided in the third metal wiring layer M3 so as to be electrically coupled to the magnetic tunnel junction MTJ.

  As described above, in the configuration according to the eighth embodiment, since read bit line RBL is directly coupled to magnetic tunnel junction MTJ, the speed of the data read operation as shown in the seventh embodiment cannot be increased. . However, in the configuration according to the eighth embodiment, the read word line driver 30r and the write word line driver 30w can be arranged independently to obtain the same effect as in the seventh embodiment.

  In the MTJ memory cell according to the eighth embodiment, the voltage and current waveform of each wiring at the time of data writing and data reading are the same as those in FIG. 33, and therefore detailed description will not be repeated.

  In the MTJ memory cell according to the eighth embodiment, the distance to the magnetic tunnel junction MTJ is larger in the write word line WWL than in the write bit line WBL. It is necessary to pass a relatively large data write current.

  FIG. 60 is a diagram for describing a configuration according to the eighth embodiment of memory array 10 and its peripheral circuits.

  Referring to FIG. 60, in the memory array according to the eighth embodiment, memory cells MC having the configuration shown in FIG. 58 are arranged in a matrix. Read word line RWL and write word line WWL are arranged along the row direction and the column direction, respectively, and read bit line RBL and write bit line WBL are arranged along the column direction and the row direction, respectively.

Memory cells adjacent in the row direction share the write word line WWL.
For example, the memory cell groups belonging to the first and second memory cell columns share the same write word line WWL1, and the memory cell groups belonging to the third and fourth memory cell columns are the same. The write word line WWL2 is shared. The write word lines WWL are similarly arranged for the subsequent memory cell columns.

  Corresponding to the same write bit line WBL, data collision occurs when a plurality of memory cells MC become data write targets. Therefore, the memory cells MC are alternately arranged.

  With such a configuration, the wiring pitch of the write word lines WWL in the memory array 10 can be relaxed. As a result, the memory cells MC can be efficiently arranged to highly integrate the memory array 10 and the chip area of the MRAM device can be reduced.

  Since the configuration of the peripheral circuit for selectively supplying the data write current and the sense current to read bit line RBL and write bit line WBL is the same as in FIG. 35, detailed description will not be repeated.

  As already described, in the MTJ memory cell according to the eighth embodiment, it is necessary to flow a large data write current to write word line WWL as a whole. Therefore, by sharing the write word line WWL between adjacent memory cells and securing the wiring pitch, the wiring width, that is, the cross-sectional area of the write word line WWL can be secured and the current density can be suppressed. As a result, the reliability of the MRAM device can be improved. Furthermore, as described above, selecting the material of these wirings in consideration of electromigration resistance is also effective in improving the operation reliability.

[Modification 1 of Embodiment 8]
FIG. 61 is a diagram for describing a configuration according to the first modification of the eighth embodiment of the memory array 10 and its peripheral circuits.

  Referring to FIG. 61, in memory array 10 according to the first modification of the eighth embodiment, adjacent memory cells share the same read bit line RBL. For example, the memory cell groups belonging to the second and third memory cell columns share the same read bit line RBL2. The read bit line RBL is similarly arranged for the subsequent memory cell columns.

  In order to perform data reading normally, it is necessary that a plurality of memory cells MC arranged at the intersection of the same read word line RWL and the same read bit line RBL do not exist. Therefore, the memory cells MC are arranged alternately.

  Furthermore, the same write bit line WBL is shared by adjacent memory cells. For example, the memory cell groups belonging to the first and second memory cell rows share the same write bit line WBL1. The write bit line WBL is similarly arranged for the subsequent memory cell rows.

  Here, in order to perform data writing normally, it is necessary that a plurality of memory cells MC arranged at the intersection of the same write word line WWL and the same write bit line WBL do not exist.

  Since the configuration of peripheral circuits related to data writing and data reading for read bit line RBL and write bit line WBL and the operation of each memory cell at the time of data reading and data writing are the same as in the eighth embodiment, Detailed description will not be repeated.

  With such a configuration, the wiring pitch of the read bit line RBL and the write bit line WBL in the memory array 10 can be relaxed. As a result, the memory cells MC can be efficiently arranged to highly integrate the memory array 10 and the chip area of the MRAM device can be reduced.

[Modification 2 of Embodiment 8]
FIG. 62 is a diagram for describing a configuration according to the second modification of the eighth embodiment of the memory array 10 and its peripheral circuits.

  Referring to FIG. 62, in memory array 10 according to the second modification of the eighth embodiment, the same read word line RWL is shared by memory cells adjacent in the column direction. For example, the memory cell groups belonging to the second and third memory cell rows share the same read word line RWL2. The read word line RWL is similarly arranged for the subsequent memory cell rows.

  Further, the same write bit line WBL is shared by memory cells adjacent in the column direction. For example, the memory cell groups belonging to the first and second memory cell rows share the same write bit line WBL1. The write bit line WBL is similarly arranged for the subsequent memory cell rows.

  Here, in order to perform data reading normally, it is necessary that a plurality of memory cells MC selected by the same read word line RWL are not simultaneously coupled to the same read bit line RBL. Therefore, the read bit line RBL is arranged for each memory cell row, and the memory cells MC are alternately arranged.

Since the configuration of other parts is the same as that of the eighth embodiment, detailed description will not be repeated.
With such a configuration, the wiring pitch of the read word lines RWL in the memory array 10 can be relaxed. As a result, the memory cells MC can be efficiently arranged to highly integrate the memory array 10 and the chip area of the MRAM device can be reduced.

[Modification 3 of Embodiment 8]
FIG. 63 is a diagram for describing a configuration according to the third modification of the eighth embodiment of the memory array 10 and its peripheral circuits.

  Referring to FIG. 63, two memory cells corresponding to each set of memory cell columns formed by two adjacent memory cell columns are arranged for the memory cells according to the eighth embodiment arranged in a matrix. A folded bit line configuration is realized using the read bit line RBL. For example, a read bit line pair can be configured by read bit lines RBL1 and RBL2 (/ RBL1) corresponding to the first and second memory cell columns, respectively.

  Similarly, a folded bit line configuration is realized by using two corresponding write bit lines WBL for each set of memory cell rows formed by two adjacent memory cell rows. For example, a write bit line pair can be configured by write bit lines WBL1 and WBL2 (/ WBL1) corresponding to the first and second memory cell rows, respectively.

  Supply of row selection and data write current ± Iw to write bit lines WBL and / WBL constituting the write bit line pair, column selection and sense current Is for read bit lines RBL and / RBL constituting the read bit line pair Since the configuration of the peripheral circuit for supplying is the same as that of FIG. 39, detailed description will not be repeated.

  Therefore, even when memory cells according to the eighth embodiment are arranged in a matrix, it is possible to secure an operation margin for data reading and data writing using the folded bit line configuration.

[Modification 4 of Embodiment 8]
In the fourth modification of the eighth embodiment, in addition to the folded bit line configuration shown in the third modification of the eighth embodiment, the write word line WWL is shared between adjacent memory cells.

  FIG. 64 is a diagram for describing a configuration according to a fourth modification of the eighth embodiment of the memory array 10 and its peripheral circuits.

  Referring to FIG. 64, in memory array 10 according to the fourth modification of the eighth embodiment, memory cells adjacent in the row direction share the same write word line WWL.

  At the time of data reading in which the read word line RWL is activated, every other memory cell column is connected to each read bit line RBL, so that the memory cell formed by two adjacent memory cell columns A read bit line pair is formed for each set of columns, and data reading similar to that of the third modification of the eighth embodiment based on the folded bit line configuration can be executed.

  On the other hand, at the time of data writing, since the write word line WWL is shared, data writing based on the folded bit line configuration cannot be performed. Therefore, in the fourth modification of the eighth embodiment, peripheral circuits related to selection of write bit line WBL are arranged in the same manner as shown in FIG. Thereby, as in the case of the eighth embodiment, data writing can be executed using the data writing circuit 51b having a simple circuit configuration.

  Further, although data writing based on the folded bit line configuration cannot be executed, the wiring pitch of the write word lines WWL in the memory array 10 can be relaxed. As a result, the chip area of the MRAM device can be further reduced due to the high integration of the memory array 10.

  In the configuration of FIG. 64, the configuration in which the write word line WWL is shared between adjacent memory cells among the data write system signal wirings is shown, but the configuration in which the write bit line WBL is shared instead of the write word line. It is also possible. However, in this case, the write word line WWL cannot be shared and must be arranged for each memory cell column. Which wiring is shared and the wiring pitch is relaxed may be determined in consideration of the distance from the magnetic tunnel junction MTJ.

[Modification 5 of Embodiment 8]
In the fifth modification of the eighth embodiment, in addition to the folded bit line configuration shown in the third modification of the eighth embodiment, the read word line RWL is shared between adjacent memory cells.

  FIG. 65 is a diagram for describing a configuration according to the fifth modification of the eighth embodiment of the memory array 10 and its peripheral circuits.

  Referring to FIG. 65, in memory array 10 according to the fifth modification of the eighth embodiment, memory cells adjacent in the column direction share the same read word line RWL.

  Read / write control circuit 60 includes an equalize transistor 62, a precharge transistor 64, and a write bit line voltage control transistor 65 arranged in the same manner as in the third modification of the eighth embodiment.

  At the time of data writing in which the write word line WWL is activated, every other memory cell column is connected to each write bit line WBL, so a memory formed by two adjacent memory cell rows. A write bit line pair can be formed for each set of cell rows. As a result, data writing similar to that of the third modification of the eighth embodiment based on the folded bit line configuration is executed, and the same effect can be obtained.

  On the other hand, at the time of data reading in which read word line RWL shared between a plurality of memory cell rows is activated, data reading based on the folded bit line configuration cannot be performed. Therefore, in Modification 5 of Embodiment 8, the peripheral circuits related to selection of read bit line RBL are arranged in the same manner as shown in FIG.

  With such a configuration, an operation margin cannot be ensured by the folded bit line configuration, but data read can be normally executed after the wiring pitch of the read word lines RWL in the memory array 10 is relaxed. As a result, the chip area of the MRAM device can be reduced due to the high integration of the memory array 10.

  Therefore, using the memory cell according to the eighth embodiment, it is possible to secure an operation margin, simplify peripheral circuits, reduce data write noise, and share the read word line RWL by data writing based on the folded bit line configuration. It is possible to achieve both high integration of the memory array 10 based on the above.

  In the configuration of FIG. 65, the read word line RWL is shared between adjacent memory cells among the signal wirings of the data read system, but the read bit line RBL is shared instead of the read word line RWL. It is also possible. In this case, however, the read word line RWL cannot be shared and must be arranged for each memory cell row. Which wiring is shared and the wiring pitch is relaxed may be appropriately determined in consideration of structural conditions and design convenience.

[Embodiment 9]
FIG. 66 is a circuit diagram showing a connection manner of MTJ memory cells according to the ninth embodiment.

  Referring to FIG. 66, access transistor ATR is electrically coupled between magnetic tunnel junction MTJ and write bit line WBL. Magnetic tunnel junction MTJ is coupled between access transistor ATR and common line CML. Access transistor ATR has its gate coupled to read word line RWL. Also in the configuration of FIG. 66, the common wiring CML functioning as the write word line WWL and the read word line RWL are arranged in directions orthogonal to each other. The degree can be improved.

  FIG. 67 is a timing chart for describing data writing and data reading for the MTJ memory cell according to the ninth embodiment.

  Referring to FIG. 67, at the time of data writing, data write current ± Iw is supplied to write bit line WBL. Further, when a current control transistor described later is turned on, a data write current Ip flows through the common line CML corresponding to the selected column according to the column selection result. Thus, the voltage and current of common wiring CML at the time of data writing are set similarly to write word line WWL shown in FIG.

  As a result, a magnetic field corresponding to the data level of the write data DIN can be written to the magnetic tunnel junction MTJ. Also, as shown in FIG. 33, the read bit line RBL is not particularly required at the time of data writing, so that both functions can be integrated into the common wiring CML.

  The current control transistor described above is turned off except during data writing, and the common line CML is precharged to the ground voltage Vss before data reading.

  At the time of data reading, the voltage level of write bit line WBL is set to ground voltage level Vss. Further, a sense current Is for data reading is supplied to the common line CML. Therefore, at the time of data reading, by activating read word line RWL to a selected state (H level), access transistor ATR is turned on, and common line CML-magnetic tunnel junction MTJ-access transistor ATR-write bit is turned on. A sense current Is can flow through the path of the line WBL.

  When the current path of the sense current Is is formed in the MTJ memory cell, an electric voltage change (rise) corresponding to the stored data occurs in the common line CML.

  In FIG. 67, when the data level stored as an example is “1” and the magnetic field directions in the fixed magnetic layer FL and the free magnetic layer VL are the same, the stored data is “1”. The voltage change ΔV1 of the common line CML is small, and the voltage change ΔV2 of the common line CML when the stored data is “0” is larger than ΔV1. By detecting the difference between the voltage changes ΔV1 and ΔV2 generated in the common wiring CML, the storage data of the MTJ memory cell can be read.

  Further, as shown in FIG. 33, write word line WWL is not particularly required at the time of data reading, so that write word line WWL and read bit line RBL can be integrated into common wiring CML.

  As described above, the same data writing and data reading can be executed for the MTJ memory cell in which the number of wirings is reduced using the common wiring CML in which the functions of the write word line WWL and the read bit line RBL are integrated. .

  Further, in the common wiring CML functioning as the read bit line RBL, the precharge voltage for data reading and the set voltage at the time of data writing are set to the same ground Vss, so that the precharge at the start of data reading is performed. The operation can be made efficient and the data read operation can be speeded up.

FIG. 68 is a structural diagram illustrating the arrangement of MTJ memory cells according to the ninth embodiment.
Referring to FIG. 68, write bit line WBL is arranged in first metal interconnection layer M1, and read word line RWL is arranged in the same layer as gate 130 of access transistor ATR. Write bit line WBL is electrically coupled to source / drain region 110 of access transistor ATR. The other source / drain region 120 is coupled to the magnetic tunnel junction MTJ through the metal wiring provided in the first metal wiring layer M1, the barrier metal 140, and the metal film 150 provided in the contact hole.

  The common wiring CML is provided in the second metal wiring layer M2 so as to be electrically coupled to the magnetic tunnel junction MTJ. Thus, in addition to the effect of the MTJ memory cell according to the sixth embodiment by having both the read bit line RBL and the write word line WWL functions in the common wiring CML, the number of wirings and the number of metal wiring layers This can reduce the manufacturing cost.

  In the MTJ memory cell according to the ninth embodiment, the distance from the magnetic tunnel junction MTJ is larger in the write bit line WBL than in the common wiring CML functioning as the write word line WWL. As a result, in the MTJ memory cell according to the ninth embodiment, it is necessary to flow a relatively large data write current to write bit line WBL.

  FIG. 69 is a diagram for describing a configuration according to the ninth embodiment of memory array 10 and its peripheral circuits.

  Referring to FIG. 69, in memory array 10 according to the ninth embodiment, memory cells MC having the structure shown in FIG. 66 are arranged in a matrix. Read word line RWL and write bit line WBL are arranged along the row direction. The common wiring CML is arranged along the column direction.

Memory cells adjacent in the row direction share a common wiring CML.
For example, the memory cell groups belonging to the first and second memory cell columns share the same common wiring CML1, and the memory cell groups belonging to the third and fourth memory cell columns are the same in common. The wiring CML2 is shared. The common wiring CML is similarly arranged for the subsequent memory cell columns.

  Corresponding to the same common wiring CML, data collision occurs when a plurality of memory cells MC are subjected to data writing and data reading. Therefore, the memory cells MC are alternately arranged.

  With such a configuration, the wiring pitch of the common wiring CML in the memory array 10 can be relaxed. As a result, the memory cells MC can be efficiently arranged to highly integrate the memory array 10 and the chip area of the MRAM device can be reduced.

  For common wiring CML, a peripheral circuit for selectively supplying a sense current provided for read bit line RBL in FIG. 35 is arranged.

  Further, a current control transistor is arranged corresponding to each common line CML. FIG. 69 representatively shows current control transistors 41-1 and 41-2 corresponding to common lines CML1 and CML2, respectively. In the following, when the current control transistors are collectively described, reference numeral 41 is simply used.

  The current control transistor 41 is disposed between the corresponding common line CML and the ground voltage Vss. The current control transistor 41 is turned on in response to activation of the control signal WE during data writing in which the common line CML functions as the write word line WWL. As a result, the data write current Ip can be supplied to the common line CML activated to the selected state (power supply voltage Vcc) by the write word line driver 30w.

  As described with reference to FIG. 67, the precharge voltage before data reading of the common line CML is set to the ground voltage Vss, so that the current control transistor 41 is further operated in response to the bit line precharge signal BLPR. Thus, the arrangement of the precharge transistor 44 can be omitted.

  On the other hand, the configuration of the peripheral circuit for selectively supplying the data write current to write bit line WBL is the same as that of FIG. 35, and therefore detailed description will not be repeated.

[Variation 1 of Embodiment 9]
FIG. 70 is a diagram for describing a configuration according to the first modification of the ninth embodiment of the memory array 10 and its peripheral circuits.

  Referring to FIG. 70, in memory array 10 according to the first modification of the ninth embodiment, adjacent memory cells share the same write bit line WBL. For example, the memory cell groups belonging to the second and third memory cell rows share the same write bit line WBL2. The write bit line WBL is similarly arranged for the subsequent memory cell columns.

  In order to perform data writing normally, it is necessary that there are not a plurality of memory cells MC arranged at the intersections of the same common line CML and the same write bit line WBL. Therefore, the common wiring CML is arranged for each column, and the memory cells MC are alternately arranged.

  Since the configuration of peripheral circuits related to data writing and data reading with respect to common wiring CML and write bit line WBL and the operation of each memory cell at the time of data reading and data writing are the same as those in the ninth embodiment, details are described. The explanation is not repeated.

  With such a configuration, the wiring pitch of the write bit lines WBL in the memory array 10 can be relaxed. As a result, the memory cells MC can be efficiently arranged to highly integrate the memory array 10 and the chip area of the MRAM device can be reduced.

  As already described, in the MTJ memory cell according to the ninth embodiment, it is necessary to flow a large data write current to write bit line WBL as a whole. Therefore, by sharing the write bit line WBL between adjacent memory cells and securing the wiring pitch, the wiring width, that is, the cross-sectional area of the write bit line WBL can be secured and the current density can be suppressed. As a result, the reliability of the MRAM device can be improved. Furthermore, as described above, selecting the material of these wirings in consideration of electromigration resistance is also effective in improving the operation reliability.

[Modification 2 of Embodiment 9]
FIG. 71 is a diagram for describing a configuration according to the second modification of the ninth embodiment of the memory array 10 and its peripheral circuits.

  Referring to FIG. 71, in memory array 10 according to the second modification of the ninth embodiment, the same read word line RWL is shared by memory cells adjacent in the column direction. For example, the memory cell groups belonging to the first and second memory cell rows share the same read word line RWL1. The read word line RWL is similarly arranged for the subsequent memory cell rows.

  Further, the same write bit line WBL is shared by memory cells adjacent in the column direction. For example, the memory cell groups belonging to the second and third memory cell rows share the same write bit line WBL2. The write bit line WBL is similarly arranged for the subsequent memory cell rows.

  Here, in order to perform data reading normally, it is necessary that the plurality of memory cells MC selected by the same read word line RWL are not simultaneously coupled to the same common line CML. Therefore, the common wiring CML is arranged for each memory cell row, and the memory cells MC are alternately arranged.

Since the configuration of other parts is the same as that of the ninth embodiment, detailed description will not be repeated.
With such a configuration, the wiring pitch of the read word line RWL and the write bit line WBL in the memory array 10 can be relaxed. As a result, the memory cells MC can be efficiently arranged to highly integrate the memory array 10 and the chip area of the MRAM device can be reduced.

[Modification 3 of Embodiment 9]
FIG. 72 is a diagram for describing a configuration according to the third modification of the ninth embodiment of the memory array 10 and its peripheral circuits.

  Referring to FIG. 72, with respect to the memory cells having the structure according to the ninth embodiment arranged in a matrix, two corresponding memory cells are formed for each pair of memory cell columns formed by two adjacent memory cell columns. The common bit line CML is used to realize a folded bit line configuration. For example, a common line CML1 and CML2 (/ CML1) respectively corresponding to the first and second memory cell columns can form a data line pair corresponding to the read bit line pair.

  Similarly, a folded bit line configuration is realized by using two corresponding write bit lines WBL for each set of memory cell rows formed by two adjacent memory cell rows. For example, a write bit line pair can be configured by write bit lines WBL1 and WBL2 (/ WBL1) corresponding to the first and second memory cell rows, respectively.

  Since the configuration of the peripheral circuit for performing row selection and supply of data write current ± Iw to write bit lines WBL and / WBL constituting the write bit line pair is the same as that of FIG. 39, detailed description will not be repeated.

  In addition, when one and the other of the common lines constituting the data line pair at the time of data reading are collectively referred to by reference numerals CML and / CML, column selection and sense for the read bit lines RBL and / RBL in the structure of FIG. Peripheral circuits for supplying current Is are arranged corresponding to common lines CML and / CML, respectively.

  Therefore, even when memory cells according to the ninth embodiment are arranged in a matrix, an operation margin for data reading and data writing can be secured using the folded bit line configuration.

[Modification 4 of Embodiment 9]
In the fourth modification of the ninth embodiment, in addition to the folded bit line configuration shown in the third modification of the ninth embodiment, the write bit line WBL is shared between adjacent memory cells.

  FIG. 73 is a diagram for explaining a configuration according to the fourth modification of the ninth embodiment of the memory array 10 and its peripheral circuits.

  Referring to FIG. 73, in memory array 10 according to the fourth modification of the ninth embodiment, write bit line WBL is shared by memory cells adjacent in the column direction.

  On the other hand, at the time of data reading when the read word line RWL is activated, every other memory cell column is connected to each common line CML functioning as the read bit line RBL. Data line pairs can be formed for each set of memory cell columns formed of cell columns, and data reading similar to that of the third modification of the ninth embodiment based on the folded bit line configuration can be executed.

  On the other hand, at the time of data writing, in order to share write bit line WBL, data writing based on the folded bit line configuration cannot be performed. Therefore, in the fourth modification of the ninth embodiment, peripheral circuits related to selection of write bit line WBL are arranged in the same manner as shown in FIG. Thereby, as in the case of the ninth embodiment, data writing can be executed using the data writing circuit 51b having a simple circuit configuration.

  Further, although data writing based on the folded bit line configuration cannot be executed, the wiring pitch of the write word lines WWL in the memory array 10 can be relaxed. As a result, the chip area of the MRAM device can be further reduced due to the high integration of the memory array 10.

[Variation 5 of Embodiment 9]
In the ninth modification 5, in addition to the folded bit line configuration shown in the ninth modification 3, the read word line RWL is shared between adjacent memory cells.

  FIG. 74 is a diagram for describing a configuration according to the fifth modification of the ninth embodiment of the memory array 10 and its peripheral circuits.

  Referring to FIG. 74, in memory array 10 according to the fifth modification of the ninth embodiment, memory cells adjacent in the column direction share the same read word line RWL.

  Read / write control circuit 60 includes an equalize transistor 62 and a write bit line voltage control transistor 65 arranged in the same manner as in the third modification of the ninth embodiment.

  At the time of data writing, every other memory cell column is connected to each write bit line WBL, so that a write bit line is set for each pair of memory cell rows formed by two adjacent memory cell rows. Pairs can be formed. As a result, data writing similar to that of the third modification of the ninth embodiment based on the folded bit line configuration is executed, and the same effect can be obtained.

  On the other hand, at the time of data reading in which read word line RWL shared between a plurality of memory cell rows is activated, data reading based on the folded bit line configuration cannot be performed. Therefore, in the fifth modification of the ninth embodiment, peripheral circuits related to the selection of the common wiring CML functioning as the read bit line RBL are arranged in the same manner as shown in FIG.

  With such a configuration, an operation margin cannot be ensured by the folded bit line configuration, but data read can be normally executed after the wiring pitch of the read word lines RWL in the memory array 10 is relaxed. As a result, the chip area of the MRAM device can be reduced due to the high integration of the memory array 10.

  Therefore, using the memory cell according to the ninth embodiment, it is possible to secure an operation margin, simplify peripheral circuits, reduce data write noise, and share the read word line RWL by data writing based on the folded bit line configuration. It is possible to achieve both high integration of the memory array 10 based on the above.

[Embodiment 10]
FIG. 75 is a circuit diagram showing a connection mode of MTJ memory cells according to the tenth embodiment.

  Referring to FIG. 75, access transistor ATR is coupled between common line CML and magnetic tunnel junction MTJ. Read word line RWL is coupled to the gate of access transistor ATR. Write bit line WBL is arranged in the same direction as read word line RWL, and is electrically coupled to magnetic tunnel junction MTJ.

  The common line CML is selectively activated by the write word line driver 30w in the same manner as the write word line WWL during data writing. On the other hand, at the time of data reading, the sense current Is is supplied to the common line CML.

  At the time of data writing, the common line CML activated to the selected state (H level) by turning on the current control transistors 41-1 to 41-m has the data write current Ip as in the write word line WWL. Flowing. On the other hand, at the time of data reading, current control transistors 41-1 to 41-m are turned off and flow through the path of common line CML, magnetic tunnel junction MTJ, access transistor ATR, and write bit line WBL (ground voltage Vss). As described with reference to FIG. 67, the sense current Is causes a voltage change corresponding to the data stored in the magnetic tunnel junction MTJ in the common wiring CML.

  Therefore, as in the ninth embodiment, the number of wirings can be reduced by combining the function of the write word line WWL at the time of data writing and the function of the read bit line RBL at the time of data reading in the common wiring CML. .

  In addition, since the read word line RWL and the common wiring CML that functions as a write word line at the time of data writing are arranged in directions orthogonal to each other, the read word line driver 30r and the write word line driver 30w are arranged independently, The same effect as in the sixth embodiment can be obtained.

FIG. 76 is a configuration diagram showing the arrangement of MTJ memory cells according to the tenth embodiment.
Referring to FIG. 76, common wiring CML is arranged in first metal wiring layer M1, and is electrically coupled to source / drain region 110 of access transistor ATR. Read word line RWL is formed in the same layer as gate 130 of access transistor ATR.

  The source / drain region 120 is coupled to the magnetic tunnel junction MTJ through the metal wiring formed in the first metal wiring layer M1, the barrier metal 140, and the metal film 150 formed in the contact hole. Write bit line WBL is arranged in second metal interconnection layer M2 so as to be electrically coupled to magnetic tunnel junction MTJ.

  Thus, the common line CML and the magnetic tunnel junction MTJ are coupled via the access transistor ATR, so that the common line CML is coupled to the magnetic tunnel junction MTJ only when the access transistor ATR is turned on. . As a result, the capacity of the common wiring CML functioning as the read bit line RBL at the time of data reading can be suppressed, and the data reading operation can be further speeded up.

  In the MTJ memory cell according to the tenth embodiment, the voltage and current waveform of each wiring at the time of data writing and data reading are the same as those in the ninth embodiment, and therefore detailed description will not be repeated.

  In the MTJ memory cell according to the tenth embodiment, the distance from the magnetic tunnel junction MTJ is larger in the common wiring CML functioning as the write word line WWL than in the write bit line WBL. As a result, in the MTJ memory cell according to the tenth embodiment, it is necessary to flow a relatively large data write current through common wiring CML.

  FIG. 77 is a diagram for describing a configuration according to the tenth embodiment of memory array 10 and its peripheral circuits.

  Referring to FIG. 77, in memory array 10 according to the tenth embodiment, memory cells MC having the configuration shown in FIG. 75 are arranged in a matrix.

  Read word line RWL and write bit line WBL are arranged along the row direction. The common wiring CML is arranged along the column direction.

Memory cells adjacent in the row direction share a common wiring CML.
For example, the memory cell groups belonging to the first and second memory cell columns share the same common wiring CML1, and the memory cell groups belonging to the third and fourth memory cell columns are the same in common. The wiring CML2 is shared. The common wiring CML is similarly arranged for the subsequent memory cell columns.

  Corresponding to the same common wiring CML, data collision occurs when a plurality of memory cells MC are subjected to data writing and data reading. Therefore, the memory cells MC are alternately arranged.

  With such a configuration, the wiring pitch of the common wiring CML in the memory array 10 can be relaxed. As a result, the memory cells MC can be efficiently arranged to highly integrate the memory array 10 and the chip area of the MRAM device can be reduced.

  Since the configuration of the peripheral circuit for selectively supplying the data write current to common wiring CML and write bit line WBL is the same as that of FIG. 69, detailed description will not be repeated.

  As already described, in the MTJ memory cell according to the tenth embodiment, it is necessary to flow a large data write current to the common wiring CML as a whole. Therefore, by sharing the common wiring CML between the adjacent memory cells and securing the wiring pitch, it is possible to secure the wiring width, that is, the cross-sectional area of the common wiring CML and suppress the current density. As a result, the reliability of the MRAM device can be improved. Furthermore, as described above, selecting the material of these wirings in consideration of electromigration resistance is also effective in improving the operation reliability.

[Modification 1 of Embodiment 10]
FIG. 78 is a diagram for describing a configuration according to the first modification of the tenth embodiment of the memory array 10 and its peripheral circuits.

  Referring to FIG. 78, in memory array 10 according to the first modification of the tenth embodiment, adjacent memory cells share the same write bit line WBL. For example, the memory cell groups belonging to the second and third memory cell rows share the same write bit line WBL2. The write bit line WBL is similarly arranged for the subsequent memory cell rows.

  In order to perform data writing normally, it is necessary that there are not a plurality of memory cells MC arranged at the intersections of the same common line CML and the same write bit line WBL. Therefore, the common wiring CML is arranged for each row, and the memory cells MC are alternately arranged.

  Since the configuration of peripheral circuits related to data writing and data reading for common wiring CML and write bit line WBL and the operation of each memory cell at the time of data reading and data writing are the same as those in the tenth embodiment, details are described. The explanation is not repeated.

  With such a configuration, the wiring pitch of the write bit lines WBL in the memory array 10 can be relaxed. As a result, the memory cells MC can be efficiently arranged to highly integrate the memory array 10 and the chip area of the MRAM device can be reduced.

[Modification 2 of Embodiment 10]
FIG. 79 is a diagram for describing a configuration according to the second modification of the tenth embodiment of the memory array 10 and its peripheral circuits.

  Referring to FIG. 79, in memory array 10 according to the second modification of the tenth embodiment, the same read word line RWL is shared by memory cells adjacent in the column direction. For example, the memory cell groups belonging to the first and second memory cell rows share the same read word line RWL1. The read word line RWL is similarly arranged for the subsequent memory cell rows.

  Further, the same write bit line WBL is shared by memory cells adjacent in the column direction. For example, the memory cell groups belonging to the second and third memory cell rows share the same write bit line WBL2. The write bit line WBL is similarly arranged for the subsequent memory cell rows.

  Here, in order to perform data reading normally, it is necessary that the plurality of memory cells MC selected by the same read word line RWL are not simultaneously coupled to the same common line CML. Therefore, the common wiring CML is arranged for each memory cell row, and the memory cells MC are alternately arranged.

  Since the configuration of other parts is the same as that of the tenth embodiment, detailed description will not be repeated.

  With such a configuration, the wiring pitch of the read word line RWL and the write bit line WBL in the memory array 10 can be relaxed. As a result, the memory cells MC can be efficiently arranged to highly integrate the memory array 10 and the chip area of the MRAM device can be reduced.

[Modification 3 of Embodiment 10]
FIG. 80 is a diagram for describing a configuration according to the third modification of the tenth embodiment of memory array 10 and its peripheral circuits.

  Referring to FIG. 80, two memory cells corresponding to each set of memory cell columns formed by two adjacent memory cell columns are arranged for the memory cells according to the tenth embodiment arranged in a matrix. The common bit line CML is used to realize a folded bit line configuration. For example, a common line CML1 and CML2 (/ CML1) respectively corresponding to the first and second memory cell columns can form a data line pair corresponding to the read bit line pair.

  Similarly, a folded bit line configuration is realized by using two corresponding write bit lines WBL for each set of memory cell rows formed by two adjacent memory cell rows. For example, a write bit line pair can be configured by write bit lines WBL1 and WBL2 (/ WBL1) corresponding to the first and second memory cell rows, respectively.

  Since the configuration of the peripheral circuit for performing row selection and supply of data write current ± Iw to write bit lines WBL and / WBL constituting the write bit line pair is the same as that of FIG. 72, detailed description will not be repeated.

  Similarly, the configuration of the peripheral circuit for performing column selection and supply of sense current Is to common lines CML and / CML constituting the data line pair at the time of data reading is similar to that of FIG. Does not repeat.

  Therefore, even when memory cells according to the tenth embodiment are arranged in a matrix, an operation margin for data reading and data writing can be secured using the folded bit line configuration.

[Modification 4 of Embodiment 10]
In the fourth modification of the tenth embodiment, in addition to the folded bit line configuration shown in the third modification of the tenth embodiment, the write bit line WBL is shared between adjacent memory cells.

  FIG. 81 is a diagram for describing a configuration according to the fourth modification of the tenth embodiment of the memory array 10 and its peripheral circuits.

  Referring to FIG. 81, in memory array 10 according to the fourth modification of the tenth embodiment, write bit line WBL is shared by memory cells adjacent in the column direction.

  At the time of data reading in which the read word line RWL is activated, every other memory cell column is connected to each common line CML functioning as the read bit line RBL, so that two adjacent memory cell columns Data line pairs can be formed for each set of memory cell columns formed in the above, and data reading similar to that of the third modification of the tenth embodiment based on the folded bit line configuration can be executed.

  On the other hand, at the time of data writing, in order to share write bit line WBL, data writing based on the folded bit line configuration cannot be performed. Therefore, in the fourth modification of the tenth embodiment, the peripheral circuits related to selection of write bit line WBL are arranged in the same manner as shown in FIG. Thereby, as in the case of the tenth embodiment, data writing can be executed using the data writing circuit 51b having a simple circuit configuration.

  Further, although data writing based on the folded bit line configuration cannot be executed, the wiring pitch of the write word lines WWL in the memory array 10 can be relaxed. As a result, the chip area of the MRAM device can be further reduced due to the high integration of the memory array 10.

[Variation 5 of Embodiment 10]
In the fifth modification of the tenth embodiment, in addition to the folded bit line configuration shown in the third modification of the tenth embodiment, the read word line RWL is shared between adjacent memory cells.

  FIG. 82 illustrates a configuration according to the fifth modification of the tenth embodiment of memory array 10 and its peripheral circuits.

  82, in memory array 10 according to the fifth modification of the tenth embodiment, memory cells adjacent in the column direction share the same read word line RWL. Read / write control circuit 60 includes an equalize transistor 62 and a write bit line voltage control transistor 65 arranged in the same manner as in the third modification of the tenth embodiment.

  At the time of data writing, since every other memory cell column is connected to each write bit line WBL, a write bit line is set for each pair of memory cell rows formed by two adjacent memory cell rows. Pairs can be formed. As a result, data writing similar to that of the third modification of the tenth embodiment based on the folded bit line configuration is executed, and the same effect can be obtained.

  On the other hand, at the time of data reading in which read word line RWL shared between a plurality of memory cell rows is activated, data reading based on the folded bit line configuration cannot be performed. Therefore, in the fifth modification of the tenth embodiment, the peripheral circuits related to the selection of the common wiring CML functioning as the read bit line RBL are arranged in the same manner as shown in FIG.

  With such a configuration, an operation margin cannot be ensured by the folded bit line configuration, but data read can be normally executed after the wiring pitch of the read word lines RWL in the memory array 10 is relaxed. As a result, the chip area of the MRAM device can be reduced due to the high integration of the memory array 10.

  Therefore, using the memory cell according to the tenth embodiment, the operation margin is secured, the peripheral circuit is simplified and the data write noise is reduced, and the read word line RWL is shared by data writing based on the folded bit line configuration. The high integration of the memory array 10 based on the above can be realized at the same time.

  The embodiment disclosed this time should be considered as illustrative in all points and not restrictive. The scope of the present invention is defined by the terms of the claims, rather than the description above, and is intended to include any modifications within the scope and meaning equivalent to the terms of the claims.

  10 memory array, 20 row decoder, 25 column decoder, 30 word line driver, 30r read word line driver, 30w write word line driver, 40 word line current control circuit, 50, 60 read / write control circuit, 51a, 51b data Write circuit, 52 data write current supply circuit, 55a, 55b, 55c, 55d, 55e data read circuit, 62 equalize transistor, 63 bit line current control transistor, 64 precharge transistor, 65 write bit line voltage control transistor, 200 , 210, 230 Data write current adjustment circuit, ATR access transistor, BL, / BL bit line, CML common wiring, CSG column selection gate, DM access diode, DMC dummy memory cell, FL pinned magnetic layer, MC , MCD memory cell, MTJ magnetic tunnel junction, RBL read bit line, RG read gate, RCG common read gate, RCSG read column select gate, RWL read word line, TB tunnel barrier, VL free magnetic layer, WCSG write column select gate , WRSG write row selection gate, WBL, / WBL write bit line, WWL write word line.

Claims (7)

A memory array having a plurality of magnetic memory cells arranged in a matrix;
Each of the plurality of magnetic memory cells has first and second data depending on the level of stored data written when a data write magnetic field applied by the first and second data write currents is larger than a predetermined magnetic field. Including a magnetic memory unit having one of the second resistance values;
A plurality of write words provided corresponding to the rows of the magnetic memory cells and selectively activated in accordance with an address selection result for flowing the first data write current during data writing Lines and,
A plurality of bit line pairs including first and second bit lines, each provided corresponding to a column of the magnetic memory cells for flowing the second data write current;
Each of the first and second bit lines is configured by using wirings formed on first and second metal wiring layers disposed on the semiconductor substrate with the magnetic memory portion interposed therebetween,
A coupling circuit for electrically coupling between each of the first bit lines and each second bit line;
The thin film magnetic memory device, wherein the second data write current flows as a current that reciprocates between the first and second bit lines electrically coupled by the coupling circuit. Each of the first bit lines has a wiring formed in the first metal wiring layer,
Each of the second bit lines has a wiring formed in the second metal wiring layer,
The thin film magnetic memory device includes:
For setting one end of each of the first and second bit lines included in one selected according to an address selection result of the plurality of bit line pairs to one of a high potential state and a low potential state. A data writing circuit;
The coupling circuit is:
A plurality of bit line currents provided corresponding to the plurality of bit line pairs, respectively, for electrically coupling the other ends of the first and second bit lines during the data writing. The thin film magnetic memory device according to claim 1, comprising a control circuit. Each of the first and second bit lines is formed using first and second metal wiring layers so as to cross each other in a predetermined region on the memory array,
Each magnetic memory cell is coupled to one of the first and second bit lines in the first metal wiring layer,
The thin film magnetic memory device further includes:
A plurality of read word lines provided corresponding to the rows of the magnetic memory cells and selectively activated according to a row selection result at the time of data reading;
A plurality of dummy memory cells each having a resistance value that is intermediate between the first and second resistance values, each coupled to one of the first and second bit lines;
A plurality of dummy read word lines for selecting the plurality of dummy memory cells;
A data read circuit that sets a data level of read data according to a voltage difference of each of the plurality of first and second bit lines corresponding to the selected column;
The plurality of read word lines and the plurality of dummy read word lines have the plurality of first and second bit lines set to a first voltage to flow a data read current according to the row selection result. 2. The thin film magnetic memory device according to claim 1, wherein the magnetic memory cell and the dummy memory cell are electrically coupled to each other between the first voltage and the second voltage. A memory array having a plurality of magnetic memory cells arranged in a matrix;
Each of the plurality of magnetic memory cells has different resistance values depending on the level of stored data written when the data write magnetic field applied by the first and second data write currents is larger than a predetermined magnetic field Including a magnetic storage unit having
A plurality of bit lines provided corresponding to the columns of the magnetic memory cells, respectively, for selectively passing the first data write current during data writing;
A plurality of writes provided corresponding to the rows of the magnetic memory cells and selectively activated in accordance with an address selection result for flowing the second data write current during the data write With word lines,
Each of the write word lines is formed in first and second sub-write layers, which are respectively formed on the semiconductor substrate with the magnetic memory portion sandwiched in the vertical direction. Including word lines,
A coupling circuit provided corresponding to each of the write word lines and electrically coupling between the first and second sub-write word lines;
The thin film magnetic memory device, wherein the second data write current flows as a current that reciprocates between the first and second sub-write word lines electrically coupled by the coupling circuit. The thin film magnetic memory device further includes:
The first sub-write word lines included in the corresponding one of the plurality of write word lines are provided corresponding to the plurality of write word lines, respectively, according to the address selection result. A plurality of write word drivers for setting one end to a first voltage;
One end of each second sub-write word line is coupled to a second voltage,
The coupling circuit is:
5. The thin film magnetic memory device according to claim 4, further comprising a wiring for coupling between the other ends of each of the first and second sub-write word lines.   6. The thin film magnetic memory device according to claim 5, wherein the plurality of write word drivers are divided and arranged in a region adjacent to the memory array in the row direction for each predetermined number of rows. One end of each of the first and second sub-write word lines is coupled to first and second voltages, respectively
The coupling circuit is:
The first and second sub-write words provided corresponding to the plurality of write word lines and included in the corresponding one of the plurality of write word lines according to a row selection result 5. The thin film magnetic memory device according to claim 4, further comprising a switch circuit for electrically coupling the other ends of the wires.

Images