Various embodiments of the present invention are generally directed to a method and apparatus for writing data to a magnetic memory element, such as a spin-torque transfer random access memory (STRAM) memory cell.
In accordance with various embodiments, a multi-level cell (MLC) magnetic memory cell stack has first and second magnetic memory elements connected to a first control line and a switching device connected to a second control line. The first memory element is connected in parallel with the second memory element, and the first and second memory elements are connected in series with the switching device. The first and second memory elements are provisioned at different elevations within the memory cell stack.
Programming currents are passed between the first and second control lines to concurrently set the first and second magnetic memory elements to different programmed resistances. In some embodiments, a first write current concurrently programs the first and second elements, followed by application of a second write current in an opposing second direction to switch the first element to a different programmed resistance.
These and various other features and advantages which characterize the various embodiments of the present invention can be understood in view of the following detailed discussion and the accompanying drawings.
The present disclosure sets forth improvements in the manner in which data may be written to magnetic memory elements, such as but not limited to spin-torque transfer random access memory (STRAM) cells.
An array of solid-state magnetic memory cells can be used to provide non-volatile storage of data bits. Some magnetic memory cell configurations include a programmable resistive element, such as a magnetic tunneling junction (MTJ). An MTJ includes a pinned reference layer having a fixed magnetic orientation in a selected direction. A free layer is separated from the reference layer by a tunneling barrier, with the free layer having a selectively variable magnetic orientation. The orientation of the free layer relative to the fixed layer establishes an overall electrical resistance of the cell, which can be detected during a read sense operation.
While magnetic memory elements have been found to efficiently store data in a compact semiconductor array environment, one issue related to such elements is the general inability to write multiple bits to magnetic memory cells using multi-level cell (MLC) programming. That is, the free layer in many magnetic memory elements can be magnetically precessed between just two magnetic states (parallel and antiparallel). This allows each magnetic memory element to store only a single bit of data using single level cell, or SLC programming.
Accordingly, various embodiments of the present invention are generally directed to an apparatus and method for carrying out MLC programming to memory cells with magnetic memory elements. As explained below, each memory cell is provisioned with two (or more) magnetic memory elements coupled in parallel with each other within the cell. Different current densities can be applied to the cell to independently switch the respective memory elements to the desired resistive states.
The use of two programmable memory elements in each cell allows the storage of two bits of data (00, 01, 10 and 11, respectively) in each cell. It will be appreciated that any plural number of memory elements can be provided in each cell. For example, the use of three memory elements would allow the storage of up to three bits of data (from 000 to 111) in each cell, and so on.
The device 100 is shown to include a controller 102 and a memory module 104. The controller 102 provides top level control of the device including interface operations with the host (not separately shown). The controller functionality may be realized in hardware or via a programmable processor, or may be incorporated directly into the memory module 104. Other features may be incorporated into the device 100 as well including but not limited to an I/O buffer, ECC circuitry and local controller cache.
The memory module 104 includes a solid-state array of non-volatile memory cells 106 as illustrated in
In some embodiments, the memory cells 106 are characterized as spin-torque transfer random access memory (STRAM) cells. The memory elements 108 are characterized as magnetic tunneling junctions (MTJs), and the switching devices are characterized as nMOSFETs (n-channel metal oxide semiconductor field effect transistors). It will be appreciated that other cell configurations can readily be used including magnetic elements with giant magnetic resistance (GMR) constructions, current perpendicular to plane GMR (CPP and CCP) constructions, and other magnetic constructions that provide different resistances responsive to the application of suitable write currents thereto.
Access to the cells 106 is carried out through the use of various control lines, including bit lines (BL) 112, source lines (SL) 114 and word lines (WL) 116. All of the cells 106 along a selected word line 116 may form a page of memory that is currently accessed during read and write operations. The array may include any number of M×N memory cells arranged in rows and columns. A cross-point array can be used in which only two control lines are directly coupled to each cell.
The various bit, source and control lines 112, 114 and 116 represented in
The free layer 128 has a selectively programmable magnetic orientation that is established responsive to the application of write current to the element 108. The programmed magnetic orientation of the free layer 128 may be in the same direction as the orientation of the reference layer 124 (parallel), or may be in the opposing direction as the orientation of the reference layer 126 (antiparallel). Parallel orientation provides a lower resistance RL through the memory cell, and antiparallel orientation provides a higher resistance RH through the cell.
It is contemplated that the magnetization direction of the reference and free layers 124, 128 will be perpendicular to the axial direction through the cell as shown, but this is not necessarily required. For reference, the parallel orientation of the free layer provides a magnetization along an easy axis of the layer, and the antiparallel orientation of the free layer provides a magnetization along a hard axis of the layer.
While not shown in
The magnetic memory element 108 can take a variety of forms. An exemplary construction is a cylinder shape as generally depicted by an exploded representation in
An electrically conductive structure 142 extends from the doped region 138 to support a bridging electrode 144. Two side-by-side, in-plane magnetic memory elements 108, denoted as MTJ1 and MTJ2, are supported on the electrode 144. A second extension electrode 146 extends from the MTJ1 and MTJ2 elements to contactingly engage the associated bit line 112. A third electrode structure 148 interconnects the doped region 136 with a longitudinally extending source line 114. Although not separately denoted, it will be appreciated that insulating layers of material such as silicon dioxide extend between the various elements shown in
The current density required to switch the first magnetic memory element MTJ1 is selected to be different from the current density required to switch the second magnetic memory element MTJ2. The current density through the memory cell 106 is regulated by the conductivity of the MOSFET 110 which in turn is established by the voltage potential supplied to the word line 116. The magnitude and direction of current through the cell is established through the application of appropriate potentials to the bit and source lines 112, 114.
It is noted that the respective MTJ1 and MTJ2 elements lie along the same plane and can be formed concurrently during semiconductor fabrication. The different switching densities can be established by providing different sizes and/or shapes to the respective elements. For example, MTJ2 is shown to be larger in cross-sectional area than MTJ1 in
Providing the MTJ1 and MTJ2 memory elements 108 in different planes as shown can advantageously improve writing operations and can affect electrical resistance response because in-plane magnetic field effects from one element to the other are substantially avoided. For example, magnetic fields generated during a programming operation upon the free layer in MTJ1 may not affect the free layer in MTJ2, and vice versa, because of the differences in vertical elevation of these respective layers.
Moreover, providing the MTJ1 and MTJ2 memory elements in different, non-overlapping planes as in
The amount of vertical axial separation between the respective out-of-plane MTJ1 and MTJ2 elements can vary as desired, including an out-of-plane separation that includes some amount of overlap, a lowermost portion of MTJ2 can be in the same plane as an uppermost portion of MTJ1 while the elements remain out-of-plane. In some embodiments, it may be beneficial to align the bottom electrode (BE) of MTJ2 with the top electrode (TE) of MTJ1 so that these respective elements are formed using the same metallization process. Nevertheless, these elements would remain out-of-plane for purposes herein.
An extension electrode 150 supports MTJ2 as shown to raise the MTJ2 element to the second, higher elevation. This electrode can be formed during the formation of MTJ1, and then MTJ2 can be formed subsequently on top of this electrode. As before, MTJ2 is provisioned with a different size and/or shape as compared to MTJ1, including different thicknesses of the respective internal layers, to provide different switching characteristics between the respective elements.
The above exemplary constructions allow the memory cell 106 to take four different resistive states. The first state occurs when both of the memory elements 108 are at a lower resistance RL (0,0). The second state occurs when MTJ1 is at a higher resistance RH and MTJ2 remains at RL (1,0). The third state occurs when MTJ1 is at the lower resistance RL and MTJ2 is at RH (0,1). The fourth state is when both elements are at RH (1,1).
Because of the different current switching densities of the respective memory elements 108 in each cell 106, the low resistance of MTJ1, denoted as RH, will be different from the low resistance of MTJ2, denoted as RL2. Similarly, the high resistance of MTJ1, denoted as RH1, will be different from the high resistance of MTJ2, denoted as RH2. Generally, it is contemplated that RL2 will be greater than RL1 (RL2>RL1) and RH2 will be greater than RH1 (RH2>RH1).
The relative values of RL1, RL2, RH1 and RH2 will vary depending on the construction of the respective memory elements 108.
A current density J1 is defined as the current density required to achieve RL1 for MTJ1 and RL2 for MTJ2. Current densities J2, J3 and J4 achieve the remaining three states as shown. It will be noted that the polarity (direction of current flow) of current densities J1 and J3 are opposite that of current densities J2 and J4. In some embodiments, the (0,0) state may be achieved using a bulk refresh operation on a section of the memory array 104 by establishing a potential difference between the bit line 112 and the well of the transistor 110 to which the drain is connected.
It may be necessary in some embodiments to apply multiple successive write currents through the cell to set the MTJ1 and MTJ2 elements to the desired states. For example, to write the state (0,1) from an indeterminate initial state, it may be necessary to first apply a current with density J4 to set the state of the cell to (1,1), followed by application of a lower current of opposite polarity with density J3 to switch MTJ1 back to low resistance, e.g., (0,1). On the other hand, other states may be written using a single application of write current in a single direction, such as states (0,0) or (1,1). These and other operational requirements will depend on the configuration of a given cell, but can be readily incorporated by the skilled artisan in view of the present disclosure.
The MTJ1 and MTJ2 memory elements 108 operate as resistances in parallel, so that each of the above resistive states will provide a memory element resistance R(x,y) across the memory elements as follows:
Ignoring parasitic effects from the conductors and other effects, the total resistance RT of the memory cell can thus be approximated as:
R
T
=R(x,y)+RS (5)
where R(x,y) is the respective memory element resistance from equations (1)-(4) and RS is the drain-source resistance of the transistor 110. The programmed state of the memory cell can be sensed in any suitable manner, such as by applying a small read current through the memory cell 106 from bit line 112 to source line 114 and sensing the total voltage drop across the cell using a sense amplifier (not shown).
At step 202, data to be written to the array are initially received from a host device or some other source (such as internal metadata). During this processing, one or more cells in the array 104 will be selected to receive the writeback data. The controller 102 or other control circuitry identifies the desired resistive state for each selected cell in turn, as shown by step 204. Using the above example of two-bits per cell, desired resistive state will be (0,0), (1,0), (0,1) or (1,1).
To write the selected cell to the state (0,0), the flow passes to step 206 where a relatively large current I1 is applied through the cell having current density and polarity J1 (see
To write the selected cell to the state (1,0), the flow passes to step 208 where the relatively large current I1 is applied having current polarity and density J1 to set both MTJ1 and MTJ2 to state (0,0). This is followed by step 210 where a relatively smaller current I2 having current polarity and density J2 is applied to switch MTJ1 to the antiparallel state. This second smaller current I2 is insufficient to switch MTJ2 to the antiparallel state, so that the final state is (1,0).
To write the selected cell to the state (0,1), the flow passes to step 212 where a relatively large current I4 is applied having current polarity and density J4 to set both MTJ1 and MTJ2 to state (1,1). This is followed at step 214 with the application of a relatively smaller current I3 with polarity and density J3 to switch MTJ1 to the parallel state. This relatively smaller current I3 will be insufficient to switch MTJ2 to the parallel state, resulting in the final state of (0,1).
Finally, to write the selected cell to the state (1,1), the flow passes to step 216 where the relatively large current I4 is applied to set both MTJ1 and MTJ2 to the antiparallel state (1,1).
As desired, a read verify operation can be carried out at step 218 to validate the correct memory state was achieved, after which the process ends. It will be appreciated that the above steps 204-218 are carried out for each MLC cell to be written in turn.
The various embodiments disclosed herein are suitable for use in a write-once or write many memory. While STRAM memory cells have been used as an illustrative embodiment, the present disclosure is not so limited, as any number of different types of magnetic element constructions can incorporate the above techniques.
It is to be understood that even though numerous characteristics and advantages of various embodiments of the present invention have been set forth in the foregoing description, together with details of the structure and function of various embodiments of the invention, this detailed description is illustrative only, and changes may be made in detail, especially in matters of structure and arrangements of parts within the principles of the present invention to the full extent indicated by the broad general meaning of the terms in which the appended claims are expressed.