A ferromagnetic thin-film based digital memory cell with a memory film of an anisotropic ferromagnetic material and with a source layer positioned on one side thereof so that a majority of conduction electrons passing therefrom have a selected spin orientation to be capable of reorienting the magnetization of the film. A disruption layer is positioned on the other side of the memory film so that conduction electrons spins passing therefrom are substantially random in orientation. The magnitude of currents needed to operate the cell can be reduced using coincident thermal pulses to raise the cell temperature.
|
1. A ferromagnetic thin-film based digital memory cell, said memory cell comprising:
a substrate; and
a bit structure formed of a plurality of layers provided in a selected sequence between a pair of electrodes in electrical contact therewith at opposite ends of said sequence and supported on said substrate comprising:
a memory film of an anisotropic ferromagnetic material capable of conducting an electrical current therethrough;
a source layer positioned on one side of said memory film capable of conducting an electrical current therethrough so that a majority of conduction electrons passing therefrom have a selected spin orientation; and
a disruption layer positioned on another side of said memory film capable of conducting an electrical current therethrough so that conduction electrons spins passing therefrom are substantially random in orientation.
26. A ferromagnetic thin-film based digital memory cell, said memory cell comprising:
a substrate; and
a bit structure formed of a plurality of layers provided in a selected sequence between a pair of electrodes in electrical contact therewith at opposite ends of said sequence and supported on said substrate comprising:
an electrically insulative intermediate layer;
a source layer on one side of said insulative intermediate layer capable of conducting an electrical current therethrough so that a majority of conduction electrons passing therefrom have a selected spin orientation;
a memory film of an anisotropic ferromagnetic material on another side of said insulative intermediate layer capable of conducting an electrical current therethrough; and
an antiparallel maintenance layer adjacent to a ferromagnetic material layer together on a side of at least one of said source layer and said memory film across from said intermediate layer.
44. A ferromagnetic thin-film based digital memory cell, said memory cell comprising:
a substrate; and
a bit structure formed of a plurality of layers provided in a selected sequence between a pair of electrodes in electrical contact therewith at opposite ends of said sequence and supported on said substrate comprising:
an electrically conductive intermediate layer;
a source layer on one side of said conductive intermediate layer capable of conducting an electrical current therethrough so that a majority of conduction electrons passing therefrom have a selected spin orientation;
a memory film of an anisotropic ferromagnetic material on another side of said conductive intermediate layer capable of conducting an electrical current therethrough, and
an antiparallel maintenance layer adjacent to a ferromagnetic material layer together on a side of at least one of said source layer and said memory film across from said intermediate layer.
2. The device of
3. The device of
4. The device of
5. The device of
6. The device of
7. The device of
8. The device of
9. The apparatus of
10. The apparatus of
11. The apparatus of
12. The apparatus of
13. The device of
14. The device of
15. The device of
16. The device of
17. The device of
18. The device of
19. The device of
20. The device of
21. The device of
22. The device of
23. The device of
24. The device of
25. The device of
27. The device of
28. The device of
29. The device of
30. The device of
31. The device of
32. The device of
33. The device of
34. The apparatus of
35. The device of
36. The device of
37. The device of
38. The apparatus of
39. The device of
40. The device of
41. The device of
42. The device of
43. The device of
45. The device of
46. The device of
47. The device of
48. The device of
49. The device of
50. The device of
51. The apparatus of
52. The device of
53. The device of
54. The apparatus of
55. The device of
56. The device of
57. The device of
58. The device of
59. The device of
0. 60. The device of claim 1, wherein when a write current flows through the bit structure between the pair of electrodes, the electrical current conducted through the memory film is capable of reorienting magnetization of the memory film by spin momentum transfer.
0. 61. The device of claim 60, wherein the source layer is part of a first synthetic antiferromagnet (SAF), wherein the memory film and the disruption layer are part of a second synthetic antiferromagnet (SAF), and wherein the source layer and the memory film are separated by a spacer layer.
0. 62. The device of claim 61, wherein the spacer layer comprises an electrically insulative layer.
0. 63. The device of claim 62, wherein the source layer has a fixed magnetization.
0. 64. The device of claim 63, and further comprising an antiferromagnetic layer in contact with the first SAF.
0. 65. The device of claim 60, wherein the source layer has a fixed magnetization.
0. 66. The device of claim 65, wherein the write current has a current density and polarization that are large enough to switch magnetization direction of the memory film.
0. 67. The device of claim 66, wherein the current density is at least about 107 amps/cm2.
0. 68. The device of claim 60, wherein the disruption layer is a ruthenium layer.
0. 69. The device of claim 1, wherein the bit structure is configured to allow magnetization of the memory film to change direction due to spin momentum transfer when a write current is passed through the bit structure.
0. 70. The device of claim 69, wherein the source layer is part of a first synthetic antiferromagnet (SAF), wherein the memory film and the disruption layer are part of a second synthetic antiferromagnet (SAF), and wherein the source layer and the memory film are separated by a spacer layer.
0. 71. The device of claim 69, wherein the spacer layer comprises an electrically insulative layer.
0. 72. The device of claim 71, wherein the source layer has a fixed magnetization.
0. 73. The device of claim 72, and further comprising an antiferromagnetic layer in contact with the first SAF.
0. 74. The device of claim 69, wherein the source layer has a fixed magnetization.
0. 75. The device of claim 69, wherein the write current has a current density and polarization that are large enough to switch magnetization direction of the memory film.
0. 76. The device of claim 69, wherein the current density is at least about 107 amps/cm2.
0. 77. The device of claim 69, wherein the disruption layer is a ruthenium layer.
0. 78. The device of claim 1, wherein the memory film is capable of switching magnetization direction in response to spin momentum transfer from electrons having the selected spin orientation that pass through the memory film.
0. 79. The device of claim 1, wherein magnetization direction of the memory film is changed by spin momentum transfer by electrical current from the source layer passing through the memory film.
0. 80. The device of claim 1, wherein the electrical contacts apply a write current through the bit structure that causes magnetization of the memory film to switch by spin momentum transfer.
0. 81. The device of claim 1, wherein the electrodes apply a write current through the bit structure to affect a direction of magnetization of the memory film.
0. 82. The device of claim 81, wherein the write current has a current density and polarization that are large enough to switch magnetization direction of the memory film.
0. 83. The device of claim 82, wherein the current density is at least about 107 amps/cm2.
0. 84. The device of claim 26, wherein when a write current flows through the bit structure between the pair of electrodes, the electrical current conducted through the memory film is capable of reorienting magnetization of the memory film by spin momentum transfer.
0. 85. The device of claim 84, wherein the source layer is part of a first synthetic antiferromagnet (SAF), and wherein the memory film and the disruption layer are part of a second synthetic antiferromagnet (SAF).
0. 86. The device of claim 85, wherein the source layer has a fixed magnetization.
0. 87. The device of claim 86, and further comprising an antiferromagnetic layer in contact with the first SAF.
0. 88. The device of claim 84, wherein the source layer has a fixed magnetization.
0. 89. The device of claim 88, wherein the write current has a current density and polarization that are large enough to switch magnetization direction of the memory film.
0. 90. The device of claim 89, wherein the current density is at least about 107 amps/cm2.
0. 91. The device of claim 84, wherein the disruption layer is a ruthenium layer.
0. 92. The device of claim 26, wherein the bit structure is configured to allow magnetization of the memory film to change direction due to spin momentum transfer when a write current is passed through the bit structure.
0. 93. The device of claim 92, wherein the source layer is part of a first synthetic antiferromagnet (SAF), and wherein the memory film and the disruption layer are part of a second synthetic antiferromagnet (SAF).
0. 94. The device of claim 93, wherein the source layer has a fixed magnetization.
0. 95. The device of claim 94, and further comprising an antiferromagnetic layer in contact with the first SAF.
0. 96. The device of claim 92, wherein the source layer has a fixed magnetization.
0. 97. The device of claim 96, wherein the write current has a current density and polarization that are large enough to switch magnetization direction of the memory film.
0. 98. The device of claim 97, wherein the current density is at least about 107 amps/cm2.
0. 99. The device of claim 92, wherein the disruption layer is a ruthenium layer.
0. 100. The device of claim 26, wherein the memory film is capable of switching magnetization direction in response to spin momentum transfer from electrons having the selected spin orientation that pass through the memory film.
0. 101. The device of claim 26, wherein magnetization direction of the memory film is changed by spin momentum transfer by electrical current from the source layer passing through the memory film.
0. 102. The device of claim 26, wherein the electrical contacts apply a write current through the bit structure that causes magnetization of the memory film to switch by spin momentum transfer.
0. 103. The device of claim 26, wherein the electrodes apply a write current through the bit structure to affect a direction of magnetization of the memory film.
0. 104. The device of claim 26, wherein the write current has a current density and polarization that are large enough to switch magnetization direction of the memory film.
0. 105. The device of claim 26, wherein the current density is at least about 107 amps/cm2.
|
This application claims the benefit of Provisional Application No. 60/291,209 filed May 15, 2001 for “CURRENT SWITCHED MAGNETORESISTIVE MEMORY CELL”.
The present invention relates to ferromagnetic thin-film structures exhibiting relatively large response magnetoresistive characteristics and, more particularly, to such structures used for the storage and retrieval of digital data.
Many kinds of electronic systems make use of magnetic devices including both digital systems, such as memories, and analog systems such as magnetic field sensors. Digital data memories are used extensively in digital systems of many kinds including computers and computer systems components, and digital signal processing systems. Such memories can be advantageously based on the storage of digital symbols as alternative states of magnetization in magnetic materials provided in each memory storage cell, the result being memories which use less electrical power and do not lose information upon removals of such electrical power.
Such memory cells, and magnetic field sensors also, can often be advantageously fabricated using ferromagnetic thin-film materials, and are often based on magnetoresistive sensing of magnetic states, or magnetic conditions, therein. Such devices may be provided on a surface of a monolithic integrated circuit to provide convenient electrical interconnections between the device and the operating circuitry therefor.
Ferromagnetic thin-film memory cells, for instance, can be made very small and packed very closely together to achieve a significant density of information storage, particularly when so provided on the surface of a monolithic integrated circuit. In this situation, the magnetic environment can become quite complex with fields in any one memory cell affecting the film portions in neighboring memory cells. Also, small ferromagnetic film portions in a memory cell can lead to substantial demagnetization fields which can cause instabilities in the magnetization state desired in such a cell.
These magnetic effects between neighbors in an array of closely packed ferromagnetic thin-film memory cells can be ameliorated to a considerable extent by providing a memory cell based on an intermediate separation material having two major surfaces on each of which an anisotropic ferromagnetic memory thin-film is provided. Such an arrangement provides significant “flux closure,” i.e. a more closely confined magnetic flux path, to thereby confine the magnetic field arising in the cell to affecting primarily just that cell. This result is considerably enhanced by choosing the separating material in the ferromagnetic thin-film memory cells to each be sufficiently thin. Similar “sandwich” structures are also used in magnetic sensors.
In the recent past, reducing the thicknesses of the ferromagnetic thin-films and the intermediate layers in extended “sandwich” structures, and adding possibly alternating ones of such films and layers, i.e. superlattices, have been shown to lead to a “giant magnetoresistive effect” being present in some circumstances. This effect yields a magnetoresistive response which can be in the range of up to an order of magnitude or more greater than that due to the well known anisotropic magnetoresistive response.
In the ordinary anisotropic magnetoresistive response, varying the difference occurring between the direction of the magnetization vector in a ferromagnetic thin-film and the direction of sensing currents passed through that film leads to varying effective electrical resistance in the film in the direction of the current. The maximum electrical resistance occurs when the magnetization vector in the field and the current direction therein are parallel to one another, while the minimum resistance occurs when they are perpendicular to one another. The total electrical resistance in such a magnetoresistive ferromagnetic film can be shown to be given by a constant value, representing the minimum resistance, plus an additional value depending on the angle between the current direction in the film and the magnetization vector therein. This additional resistance has a magnitude characteristic that follows the square of the cosine of that angle.
Operating magnetic fields imposed externally can be used to vary the angle of the magnetization vector in such a film portion with respect to the easy axis of that film. Such an axis comes about in the film because of an anisotropy therein typically resulting from depositing the film during fabrication in the presence of an external magnetic field oriented in the plane of the film along the direction desired for the easy axis in the resulting film. During subsequent operation of the device having this resulting film, such operational magnetic fields imposed externally can be used to vary the angle to such an extent as to cause switching of the film magnetization vector between two stable states which occur for the magnetization being oriented in opposite directions along the film's easy axis. The state of the magnetization vector in such a film can be measured, or sensed, by the change in resistance encountered by current directed through this film portion. This arrangement has provided the basis for a ferromagnetic, magnetoresistive anisotropic thin-film to serve as a memory cell.
In contrast to this arrangement, the resistance in the plane of a ferromagnetic thin-film is isotropic for the giant magnetoresistive effect rather than depending on the direction of the sensing current therethrough as for the anisotropic magnetoresistive effect. The giant magnetoresistive effect involves a change in the electrical resistance of the structure thought to come about from the passage of conduction electrons between the ferromagnetic layers in the “sandwich” structure, or superlattice structure, through the separating nonmagnetic layers with the resulting scattering occurring at the layer interfaces, and in the ferromagnetic layers, being dependent on the electron spins. The magnetization dependant component of the resistance in connection with this effect varies as the sine of the absolute value of half the angle between the magnetization vectors in the ferromagnetic thin-films provided on either side of an intermediate nonmagnetic layer. The electrical resistance in the giant magnetoresistance effect through the “sandwich” or superlattice structure is lower if the magnetizations in the separated ferromagnetic thin-films are parallel and oriented in the same direction than it is if these magnetizations are antiparallel, i.e. oriented in opposing or partially opposing directions. Further, the anisotropic magnetoresistive effect in very thin films is considerably reduced from the bulk values therefor in thicker films due to surface scattering, whereas a significant giant magnetoresistive effect is obtained only in very thin films. Nevertheless, the anisotropic magnetoresistive effect remains present in the films used in giant magnetoresistive effect structures.
A memory cell based on the “giant magnetoresistive effect” can be provided by having one of the ferromagnetic layers in the “sandwich” construction being prevented from switching the magnetization direction therein from pointing along the easy axis therein in one to the opposite direction in the presence of suitable externally applied magnetic fields while permitting the remaining ferromagnetic layer to be free to do so in the same externally applied fields. In one such arrangement, a “spin-valve” structure is formed by providing an antiferromagnetic layer on the ferromagnetic layer that is to be prevented from switching in the externally applied fields to “pin” its magnetization direction in a selected direction. In an alternative arrangement often termed a “pseudo-spin valve” structure, the ferromagnetic layer that is to be prevented from switching in the externally applied fields is made sufficiently thicker than the free ferromagnetic layer so that it does not switch in those external fields provided to switch the free layer.
Thus, a digital data memory cell based on the use of structures exhibiting the giant magnetoresistive effect is attractive as compared to structures based on use of an anisotropic magnetoresistive effect because of the larger signals obtainable in information retrieval operations with respect to such cells. Such larger magnitude signals are easier to detect without error in the presence of noise thereby leading to less critical requirements on the retrieval operation circuitry.
An alternative digital data bit storage and retrieval memory cell suited for fabrication with submicron dimensions can be fabricated that provides rapid retrievals of bit data stored therein and a low power dissipation memory through use of a cell structure that has a spin dependent tunneling junction (SDTJ), or magnetoresistive tunnel junction (MTJ), device therein based on a pair of ferromagnetic thin film layers having an electrical insulator layer therebetween of sufficient thinness to allow tunneling currents therethrough. This memory cell can be fabricated using ferromagnetic thin-film materials of similar or different kinds in each of the magnetic memory films present in a “sandwich” structure on either side of an intermediate nonmagnetic layer where such ferromagnetic films may be composite films, but this intermediate nonmagnetic layer conducts electrical current therethrough based primarily on the quantum electrodynamic effect “tunneling” current mentioned above.
This “tunneling” current has a magnitude dependence on the angle between the magnetization vectors in each of the ferromagnetic layers on either side of the intermediate layer due to the transmission barrier provided by this intermediate layer depending on the degree of matching of the spin polarizations of the electrons tunneling therethrough with the spin polarizations of the conduction electrons in the ferromagnetic layers, the latter being set by their magnetization directions to provide a “magnetic valve effect”. Such an effect results in an effective resistance or conductance characterizing this intermediate layer with respect to the “tunneling” current therethrough. In addition, an antiferromagnetic layer against one of the ferromagnetic layers is used in such a cell to provide different magnetization switching thresholds between that ferromagnetic layer and the other by fixing, or “pinning”, the magnetization direction for the adjacent ferromagnetic layer while leaving the other free to respond to externally applied fields. Such devices may be provided on a surface of a monolithic integrated circuit to thereby allow providing convenient electrical connections between each such memory cell device and the operating circuitry therefor.
A “sandwich” structure for such a memory cell, based on having an intermediate thin layer of a nonmagnetic, dielectric separating material with two major surfaces on each of which a anisotropic ferromagnetic thin-film is positioned, exhibits the “magnetic valve effect” if the materials for the ferromagnetic thin-films and the intermediate layers are properly selected and have sufficiently small thicknesses. The resulting “magnetic valve effect” can yield a response which can be several times in magnitude greater than that due to the “giant magnetoresistive effect” in a similar sized cell structure.
An example of a two state magnetoresistive device structure that is generally common to both of these kinds of memory cells is the “pinned sandwich” structure shown in the layer diagram of
A substrate, 2, supports an interconnection structure, 3, as the bottom contact electrode to a magnetic material (ferromagnetic material) free layer, 4, (meaning its magnetization is relatively free to be rotated to an alternative orientation) that is separated by a nonmagnetic material spacer layer, 5, from a magnetic material (ferromagnetic material) relatively fixed layer, 6, (meaning its magnetization is much less free to be rotated to an alternative orientation, i.e. “pinned”). This “pinning” of layer 6 is provided by a further magnetic material layer, 7, the “pinning” layer, that is of an antiferromagnetic material which is magnetically coupled to pinned layer 6 and thereby serves to make this two layer pinned structure relatively resistant to rotation of its initial joint magnetization direction in the presence of moderate external applied magnetic fields. An aluminum cap layer, 8, serves as the device top contact electrode providing a conductive path to a further interconnection, 9.
If spacer layer 5 is an electrical conductor, such as Cu, then the structure will exhibit the giant magnetoresistive (GMR) effect and be termed a “spin valve”. If spacer layer 5 is an electrical insulator, such as Al2O3, that is sufficiently thin, then the device will exhibit the spin dependent tunneling effect and be termed a “magnetic tunnel junction”. In either situation, the electrical resistance of the device is typically higher when the magnetizations of the free and fixed layers on either side of the spacer layer are oriented antiparallel to one another, and is lower when these magnetizations are oriented parallel to one another. The electrical resistance versus external applied magnetic field response characteristic for a spin valve that is measured for sense current being established across the magnetic material layers with the conductive layer therebetween is greater in terms of fractional change than that characteristic measured for the sense current established parallel to these layers because the entire collection of spins in the sense current electrons is forced to interact with both magnetic material layers for the sense current being established across these layers but only a fraction of these electrons interact with both layers for sense currents established parallel thereto.
Plots of the high externally applied magnetic field range and the low externally applied magnetic field range response characteristics of a typical spin valve are shown in the graphs of
As stated above, operating magnetic fields imposed externally can be used to vary the angle of the magnetization vector with respect to the easy axis in the ferromagnetic films of these various kinds of memory cell devices, particularly the free layers. Such operational magnetic fields imposed externally can be used to vary the angle to such an extent as to cause switching of the layer magnetization vector between two stable states which occur for the magnetization being oriented in opposite directions along the easy axis of the layer, the state of the cell determining the value of the binary bit being stored therein. One of the difficulties in such memories is the need to provide memory cells therein that have extremely uniform switching thresholds and adequate resistance to unavoidable interjected magnetic field disturbances in the typical memory cell state selection scheme used. This externally applied operating fields scheme is based on selective externally imposed magnetic fields provided by selectively directing electrical currents over or through sequences of such cells thereby giving rise to such magnetic fields so that selection of a cell occurs through coincident presences of such fields at that cell. Such a coincident interjected magnetic fields memory cell state selection scheme is very desirable in that an individual switch, such as that provided by a transistor, is not needed for every memory cell, but the limitations this selection mode imposes on the uniformity of switching thresholds for each memory cell in a memory make the production of high yields difficult.
Such memory cells can be modified to use Curie point or Néel point data storage, or writing, techniques based on the thermal pulse accompanying a current pulse provided in or near the cell. If such storing currents are established that are sufficient to heat these storage layers to the Curie temperature thereof, then much less magnetic field strength would be needed to change the magnetic states of the storage layers and the values of the storing currents could be much reduced to effectively avoid significant magnetization rotation thresholds in the device magnetic material layers.
With or without such thermal pulse techniques, however, as the dimensions of magnetoresistive elements in memory cells shrink to less than 100 nm in size, the problems that must be overcome in order to make successful operating magnetoresistive memories become more difficult. These problems include increasing thermal instability when states thereof are not being switched because less thermal energy is needed to upset the state of smaller memory cells, the needing of larger current densities to overcome greater demagnetization fields to store information, and the greater stray field interactions between adjacent memory cells on a common substrate. Thus, there is a desire for alternative, or supplemental methods, for storing information in magnetoresistance based memory cells.
The present invention provides a ferromagnetic thin-film based digital memory cell formed on a substrate supporting a bit structure with a memory film of an anisotropic ferromagnetic material capable of conducting an electrical current therethrough with a source layer positioned on one side thereof also capable of conducting an electrical current therethrough so that a majority of conduction electrons passing therefrom have a selected spin orientation to be capable of reorienting the magnetization of the memory film. A disruption layer is positioned on another side of the memory film capable of conducting an electrical current therethrough so that conduction electrons spins passing therefrom are substantially random in orientation. The magnitude of currents needed to operate the cell can be reduced using coincident thermal pulses to raise the cell temperature.
In high-density magnetic memory cell arrays, a common problem is to make the memory cells magnetically stable (so they retain information indefinitely without requiring energy) while also being easily switchable between alternative memory states as the basis for storing binary bits of data. The switching of one cell must not disturb the magnetic state of another nearby cell in the array. Such highly controllable switching events are desired to be performed using as little time and energy as possible.
In the case of both spin valve and magnetic tunnel junction devices described above, the purpose of the spacer layer is to keep the magnetizations of the free and fixed magnetic material layers separated from coupling to one another while allowing passage of electrons between these two layers. In a properly constructed spin valve or magnetic tunnel junction device, the passage of electrons through the metallic Cu spacer layer 5, or the insulating Al2O3 spacer layer 5, does not significantly affect the magnetic orientation of a given electron spin. Rather, the spin orientation is largely preserved until the electron interacts with the magnetic layer that it is passing into. These needs and this latter behavior are the basis for the memory cells to be described below.
A device structure, whose utility for these purposes will be explained in more detail later, is shown in
In addition, magnetically free layer 4 in the device of
In these new spin-polarized current written memory devices, the data storage method involves passing a flow of spin-polarized electrons into the magnetic storage layer in order to switch its magnetic orientation. This flow has been found needed to be at a density approaching 1×107 Amps/cm2 and to have a large spin polarization. Here, spin polarization is defined as: σ=[σ1−σ11]/[σ1↑+σ1] where σ1 and σ are the current densities of the up and down spin channels. “Up” and “down” are defined with respect to the axis of magnetization. When all of the electrons are spin up, the polarization is 1, or 100%, when there are equal numbers of spin up and spin down electrons, the polarization is 0.
Spin channels at the Fermi energy is a concept used to describe the conduction electrons in a solid. Only a fraction of all electrons in a solid are actually available for conduction, which are those having energies near the material Fermi energy, and these electrons are thus at an energy level that is consistent with motion through the atomic lattice of that solid. The “spin channel” idea is based on dividing up all the conduction electrons into two groups: spin up electrons and spin down electrons. To the extent that these two groups have minimal interaction with one another, they can be thought of as an electrical current conduction system having two separate channels. This “spin channel” concept is often referred to as the “two current model”. In the past, the two current model and variations thereof have been used with considerable success to describe many phenomena related to spin dependent transport. These observable phenomena include anisotropic magnetoresistance (AMR), giant magnetoresistance (GMR), spin dependent tunneling, and the extraordinary Hall effect. Since these phenomena occur in similar materials systems as are being proposed for the spin momentum transfer devices, the use of the two current model to describe the spin current properties is appropriate.
In transition metals and alloys, including ferromagnetic materials, the electrons that carry charge from place to place in a lattice (the conduction electrons) are known to be the same ones that provide the basis for the resulting magnetic properties of such a material. In other words, their conduction spin polarization is large in such materials. This is in contrast to materials where the magnetic properties of the material are not due to conduction electrons.
Spin-polarized currents can affect the magnetization of a ferromagnetic material through transfer of angular momentum, the angular momentum of an electron being well known to be coupled to its spin property. When an electron interacts with a material in such a way as to flip the direction of spin of that electron, the angular momentum lost by the electron must be transferred to some other entity. This transfer of angular momentum, in the atomic scale view, exerts magnetic torque on the entity with which the interaction occurred. In the case where the entity is the material magnetization, the spin angular momentum is transferred to the material magnetization. This transfer of angular momentum, in the atomic scale view, exerts magnetic torque on the magnetic moment in the material where the interaction occurred.
Consider a single electron traveling in the vertical conduction structure shown in
The absolute value of the angular momentum of each electron is
In practice, there are a large number of these interactions taking place per unit time in proportion to the electrical current flowing from or to one magnetic layer to or from another adjacent layer. In other words, the number per unit time=I/e where I=current in amps and e is the electron charge. The net spin transfer per unit time is also proportional to the spin polarization as defined above. Thus
dL/dt∝σI.
The magnetic torque exerted on the magnetic moment of the material at time t is defined as
τ(t)=dL/dt,
and therefore τ(t)=∝σI
This disturbance of an electron's spin, sometimes referred to as “spin flip scattering”, happens all the time in a material including when a current passes therethrough. But in most cases the net angular momentum transfer is nil because the number of flips from up to down are equal to the flips down to up. Only in the special case where the spin polarization of the incoming charge carriers is different than the spin polarization of the outgoing charge carriers is there a net transfer of angular momentum. When the sum of all these angular momentum transfers is high enough, the magnetization direction of a magnetic material can be caused to rotate.
In order to quantitatively calculate the net momentum transfer due to a polarized electron current flowing from one magnetic layer into another, one must know the total current, I, the net spin polarization, σ, and the angle θ between the magnetizations of the two layers. Assume that all of the spins from source layer 2 interact within target layer 1 so that the net spin polarization is completely dissipated. A formulation for the torque magnetic torque exerted on the magnetic moment of the material for arbitrary relative orientations of the layer magnetizations has been found to be
Here, {right arrow over (M)}1 and {right arrow over (M)}2 are the magnetizations of the target and source magnetic layers, respectively, having corresponding unit vectors {circumflex over (m)}1 and {circumflex over (m)}2; θ is the angle between {right arrow over (M)}1 and {right arrow over (M)}2; σ is the spin polarization as defined above, j=I/(area of target layer impingement surface) is the current density (Amps/cm2), e is the electron charge, and t is the thickness of the target layer (in cm). Note that the area times thickness is the volume of the target layer.
This torque calculation assumes a number of things, such as 100% spin transfer, and that all of the spin interactions occur within the target magnetic layer (i.e. the mean free path is much longer than the spacer between two magnetic layers, and much less than the thickness of each of the two layers). It also assumes that the layer target layer magnetization is free to rotate.
These relationships are, by definition, the result of the average of many electron-electron interactions. The polarization, θ, refers to the average spin polarization of the entire ensemble of conduction electrons. Likewise, the current density, j, is the average current density passing through the major surfaces of the magnetic material layers of the device.
This equation tells about the torque at a given point in time based on the relative orientations of magnetizations. A full calculation of the switching process (making the magnetization flip from one orientation to another) requires incorporating this instantaneous torque equation into the dynamic switching relationships known as the Landau-Lifshitz-Gilbert equation, or
d{right arrow over (M)}/dt=−γ{right arrow over (M)}×{right arrow over (H)}+(α/|{right arrow over (M)}|){right arrow over (M)}×(d{right arrow over (M)}/dt).
Adapting this equation to the present situation yields
where {right arrow over (H)}eff is the total effective magnetic field, γ the gyromagnetic ratio, and α is the damping constant. {right arrow over (H)}eff={right arrow over (H)}anis cos(θ)+{right arrow over (H)}−{right arrow over (H)}ex is the sum of anisotropy, applied, exchange, and demagnetizing fields, θ represents the angle between the magnetizations of the two magnetic layers. The first term is the net torque due to all magnetic fields, the second term is damping, and the third term is the spin momentum transfer.
The damping factor, α, in the LLG equation is an empirically measured term that takes into account all means of energy transfer from magnetic rotation to other forms. Several physical phenomena play a role in damping. Excitations such as photons, phonons, and magnons contribute to a larger and lesser degree, respectively. Thermal heating due to eddy currents and interactions with atomic nuclei also contribute.
Spin currents can switch a magnetic cell having approximately 400 Oe of effective anisotropy (shape and crystalline anisotropy) with a current density of 107 A/cm2. This has been accomplished by flowing electrons first through a highly spin polarizing material (cobalt), then through a thin layer (probably epitaxial) layer of copper, and into a thinner layer of cobalt having the magnetization direction to be switched.
An important related demonstration is that sufficient electrical current density in the opposite direction results in reversing the magnetic switching. As mentioned above, the calculation of magnetization rotation assumes that the target layer is magnetically free to rotate. In cases where the source layer is freer to rotate than the target layer, a large fraction of one spin channel can be reflected at the spacer-target interface. These reflected spins then interact with the source layer and cause its magnetization to rotate.
The foregoing discussion has addressed the nature of angular momentum transfer from spins to other spins and magnetizations, the relationship between this transfer and torque on the magnetization of materials, and the dynamic “equations of motion” for magnetization under the influence of fields and spin currents. We further the explanation here with a discussion of the relationship between spin polarized currents and choices of material structures in the device. Certain thicknesses and arrangements of magnetic layers are advantageous for the exploitation of the spin current switching effect. These design choices depend on the location at which the spin polarized current interacts with the magnetization.
The depth of penetration, closely related to the “spin flip length”, of an incident spin current is highly dependent on the spin orientation. “Spin up” electrons are defined as those spins orientated parallel to the material magnetization, and “spin down” electrons are defined as those spins oriented anti-parallel to the material magnetization. The flipping of spin down electrons is the fundamental mechanism for the transfer of angular momentum to the magnetization and the resulting torque on the magnetization. Flipping of spin up electrons only reinforces the existing magnetization orientation. The scattering probability of spin down electrons is much higher than that of spin up electrons. This means that the momentum transfer of an incident spin down current takes place much sooner (shallower) than an incident spin up current. The ratio of the spin flip lengths is on the order of 10, so that the spin flip length (sfl) of spin down electrons is about 1 nm while the sfl of spin up electrons is about 10 nm. These two factors indicate that the optimum thickness for a target magnetic layer is longer than the spin down sfl, and shorter than the spin up sfl. This means the target layer thickness could be up to a few 10s of nm. Added thickness of the target layer, however, increases the total required momentum transfer since the magnetic anisotropy energy is proportional to the volume.
The introduction of thin ruthenium layers between ferromagnetic layers (NiFe, for instance) can help to optimize the angular momentum transfer in a relatively thin target layer. Thin Ru layers are effective spin scattering layers which randomize electron spins passing through that layer, and they strongly antiparallelly couple such ferromagnetic layers provided on either side thereof. In such a SAF pair, one of the two magnetic layers acts as the “target” layer for an incoming spin current. The SAF has several desirable features for dense MRAM. Self-demagnetizing effects are greatly reduced by the presence of compensating magnetic free poles at the device edges, regardless of the magnetization orientation. Stray fields, which can cause undesirable magnetic interactions with nearby bits, are greatly reduced when a SAF structure is used compared to a single magnetic film structure. The effective anisotropy of the SAF is the same as for a single film, and the anisotropy energy is proportional to the total volume of the two magnetic layers. The effect of external fields is drastically lower because the magnetostatic energy of the two antiparallel layers cancel each other. In total, the SAF structure is more sensitive to internal spin polarized switching currents, less sensitive to external stray fields, and has reduced magnetostatic interactions with nearby magnetic elements.
In addition to magnetostatic considerations, the Ru layer provides an important spin randomization effect. This effect prevents the spin polarized current from exerting significant torque on the non-target film of the SAF. In absence of this spin randomization effect, the spin polarized current which does not transfer its angular momentum to the target layer will then pass through to the non-target layer and exert a torque in the opposite direction to that intended. Due to the spin scattering effect of the Ru layer, the spin up electrons in the current through the target ferromagnetic layer are randomized before they reach the non-target ferromagnetic layer so that relatively small net torque is exerted upon the magnetization of that non-target layer. The antiferromagnetic coupling between the two ferromagnetic layers due to the ruthenium layer therebetween thus allows the torque in the first layer to switch magnetization directions of the two ferromagnetic layers together.
Magnetic tunnel junctions can be fabricated with down to 1 Ω-μm2 barrier resistance area product with magnetoresistance change ratio of 40%. Such a thin barrier will be necessary to allow for the occurrence of sufficient current density magnitudes so as to be capable of switching the magnetic storage layers in memory cells.
Minimizing the lateral scattering of injection spin current conduction electrons to reduce the polarization losses during transit for better switching requires that such currents be established through the smallest amount of a material layer as possible, i.e. through the thickness of a thin-film layer. This is supported by having the lateral dimensions of such a layer as small as possible to minimize the lateral spreading out of the current over the layer because such spreading leads to an effectively lengthened current path thereby also adding path resistance that does not contribute to the magnetoresistance change based signal. As a result, the lateral dimensions of a layer in a stack of layers forming an operative device will approach, or even be less than, the total thickness of the stacked layers in that device. In an alternative to devices with a magnetic junction, as just mentioned, where instead the barrier layer is replaced by an electrical conductor such as copper, an added benefit of such small lateral dimensions is an increase in path electrical resistance because of the resulting small path cross sectional area.
Thus, a vertical cell having two tunnel junctions one on the other but in series with an inner contact structure therebetween consisting of two ferromagnetic layers (10 to 50 Å thick) sandwiching a ruthenium layer of approximately 7 to 12 Å thickness forms a strongly antiparallel coupled device. Two outside electrodes are pinned synthetic antiferromagnets with the pinning antiferromagnetic materials on the side thereof away from the tunnel junctions. These outside electrodes are permanently magnetized so that the magnetizations of the magnetic layers next to the junction are both oriented in the same direction.
Such a vertically stacked memory cell structure, 10, is representatively shown in the side view of
A 9 Å thick ruthenium layer, 18, is provided directly supporting ferromagnetic layer 16, and another 9 Å thick ruthenium layer, 19, is provided directly on ferromagnetic layer 17. A ferromagnetic layer, 20, is provided directly supporting ruthenium layer 18, and a further ferromagnetic layer, 21, is provided directly on ruthenium layer 19. A chrome platinum manganese antiferromagnetic layer, 22, directly supports ferromagnetic layer 20 to form a pinned synthetic antiferromagnetic to pin the magnetization direction of ferromagnetic layer 16 in a selected direction. Similarly, a chrome platinum antiferromagnetic layer, 23, is directly on ferromagnetic layer 21 to form a pinned synthetic antiferromagnetic to pin the magnetization direction of ferromagnetic layer 17 in the same direction as that of ferromagnetic layer 16 along the length of the cell.
A tungsten plug, 24, extending through an electrical insulating layer directly supports and electrically contacts antiferromagnetic layer 22, and an aluminum cap, 25, is directly supported on chrome platinum antiferromagnetic layer 23. A lower electrical interconnection, 26, is in electrical contact with plug 24, and a further electrical interconnection, 27, is in electrical contact with cap 25.
Consider a current being passed through the two barrier junctions such that the electrons pass from top to bottom in
Retrieving stored information is accomplished by injecting smaller currents (less in magnitude than needed for ferromagnetic layer magnetization switching) into each of the outside contacts, and taking both currents out through the center electrode. Observing the voltage difference between the outside electrodes with respect to the inner electrode, which is essentially the voltage drops across the two barrier junctions, will provide the output signal to indicate the storage state as one barrier junction will be in its higher magnetoresistance magnetic state and the other in its lower magnetoresistance magnetic state. A “one” magnetic state and a “zero” magnetic state will provide corresponding output signals which are of opposite algebraic signs.
A small shape anisotropy for the cell of 40 Oe would require that the current density for switching to be 106 A/cm2. This would require 1.0 mA into a junction of 0.1 square micron area (10−9 cm2). At 100 mV across each junction, the resistance of the junction would have to be 100 Ohm, which is equivalent to 1.0 Ω-μm2 barrier resistance area product.
Thus, this memory cell can have the magnetization directions of the data storage layers 11 and 12 switched by a fairly low spin injected current of 1 mA, and the output signal, using 0.1 mA to read, would be about 8 mV across the cell for a 40% magnetoresistance. The cell would have a small area although making a contact to the inner contact at the device center is a fabrication process challenge.
A structural variation whereby the bottom tunneling layer is replaced with a thin copper layer simplifies making electrical contact for measuring the voltage drop due to the barrier junction tunneling current. A second variation is shown in
In the first variation, a thin layer of copper, 14′, (thin enough to scatter spins significantly) is used in place of bottom tunnel junction barrier layer 14 of memory cell 10 as shown for a corresponding variation memory cell, 10′, in the layer diagram of shown in the fragmentary side view of
A conductive riser to electrically contact interconnection 26 (fabrication process second metal layer) is formed of a tungsten plug, 30, a further (first metal) aluminum layer, 31, and a final tungsten plug, 32, extends up from a n+-type conductivity terminating region, 33, in a p+-type conductivity tab, 34, of a metal-oxide-semiconductor-field-effect transistor (MOSFET), 35, formed in an n-type conductivity silicon substrate, 36. MOSFET 35 has another n+-type conductivity terminating region, 37, in tub 34 separated from terminating region 33 by a channel region in tub 34 over which an n+-type conductivity polysilicon gate, 38, is provided separated from tub 34 by a gate oxide layer, 39. Another tungsten plug, 32′, rises from terminating region 37 to another portion of first layer metal, 31′. Interconnection 26 here serves as a lead to a voltage measurement circuit, and interconnection 31′ is adapted for connection to an electrical energy source with the connection thereto being controlled by the gate signal on MOSFET 35 serving as an electrical switch.
As indicated above, random access magnetoresistive memory cell arrays providing two combined operating magnetic fields from coincident currents to switch the magnetic state of a selected memory cell in storing therein a data bit require cell selection electrical currents to be provided along two separate current paths in or adjacent to the array. In some arrangements, the previously described memory cells based on the use of transistor controlled spin injection currents can avoid this sort of cell selection. However, in other arrangements, this sort of selection will still be relied upon, and so have the attendant difficulties to be described in more detail below.
In such a coincident current selection arrangement, only that cell in the vicinity of the crossing location, or intersection, of these two paths experience sufficient magnetic field intensities due to the summing of the fields due to these two currents to cause such a magnetic state change therein. Cells in the array that are located far away from both of these two current paths are not significantly affected by the magnetic fields generated by such currents in the paths because such fields diminish in intensity with distance from the source thereof. Cells, however, located in relatively close proximity to one, but not two, of these two paths do experience more significant magnetic fields thereabout, and those immediately in or adjacent to one such path experience sufficient field intensities to be considered as being “half-selected” by the presence of current in that path intended to participate in fully selecting a different cell along that path at the intersection with the other path on which a selection current is present. Half-selection means that a bit is affected by magnetic fields from the current through one path but not another.
The stability of magnetic memory bits can be considered in terms of an “energy well”. When the magnetostatic energy of a data storage magnetic material layer (defined below) of a cell is plotted vs. the angle between the magnetization and the easy axis of that layer, there is an energy minimum at the angular value of zero or, with this angle designated as θ, at θ=0 as shown in the graph of
E=½ sin2 θ|{right arrow over (M)}|HkV
where {right arrow over (M)} is the magnetization, Hk is the anisotropy, V is the volume, and θ is the angle of {right arrow over (M)} from the easy axis. The magnetization orientation will tend to orient to minimize the magnetostatic energy; i.e. θ will tend toward zero degrees.
The graph of
The energy expression above, when modified to include the effects of {right arrow over (H)}w and {right arrow over (H)}s, becomes
E=½ sin2 θ|
Here we assume that {right arrow over (H)}w is parallel to the effective easy axis while {right arrow over (H)}s is perpendicular to this axis. The easy axis is parallel to Hk.
If a half-select word field is applied (i.e. |{right arrow over (H)}w|=½ Hk and |{right arrow over (H)}s|=0), the energy expression becomes:
E=½ sin2 θ|{right arrow over (M)}|Hk+|{right arrow over (M)}∥{right arrow over (H)}w|V cos θ,
where the second term is the energy due to the applied word field. If a half-select sense field is applied (i.e. |{right arrow over (H)}s|=½ Hk and |{right arrow over (H)}w|=0), the energy expression becomes:
E=½ sin2 θ|{right arrow over (M)}|HkV−|{right arrow over (M)}∥{right arrow over (H)}s|V sin 0,
where the second term is the energy due to the applied sense field.
These two equations are plotted in the graphs of
Consider, for instance, a 256 megabit data storage capacity memory cell array provided as part of a monolithic integrated circuit chip organized so as to have 16 byte data blocks (8×17 binary bits) and having implemented therein the well known Hamming single bit error correction code which adds 8 additional bits. A reasonable sub-array to operate would be organized so that the word lines would each be immediately adjacent to 1088 cells to provide magnetic fields thereabout and the sense lines would each connect in series 128 cells for storing and retrieving data binary bits. The data is to be accessed in sequence in two byte groups.
A worst case can be taken to occur if one block had the data therein retrieved continuously for a year. In this case only one block would experience error correction while the remaining 2 million blocks would not. A total of 78 blocks (64−1+16−1) would be continuously half selected. The unselected blocks with 4 times the well depth can be ignored because their thermally induced failure rate would be negligible.
First, consider the situation without any error correction. For randomly occurring failures in a non-redundant system, reliability theory shows that the total failure rate of the 78 half selected blocks is the sum of the individual failure rates. Also each byte that is present in a block is accessed only ⅛ of the time. Thus to achieve a desired 10−5 yearly failure rate for the memory, the failure rate for each cell per ⅛ year is given by
10−5/(78 blocks*128 cells/block)=1.0016×10−9 failure/(⅛ a year).
Noting that there are 3.942*106 seconds in ⅛ year, that the relaxation time for a magnetic element of the type considered is about 0.5×10−9 seconds, and that if a cell is thermally excited above the barrier, there is a 50% chance it will end in the wrong state, one can compute the required barrier height (Eb) as
1.0016×10−9=0.5*e−(E
Eb=56.6 kT
(Note: the energy state is randomized with respect to each relaxation time (tr) so that there are 1/tr chances to fail per second.)
If there is single bit error correction, then any single bit error in any half-selected block can be corrected and a failure occurs at the end of the year only if two or more errors occur in a block. For small failure probabilities, the failure rate for two or more elements failing in a block is only very slightly more than two elements failing in a block. For convenience, calculations will be based on two bits failing in a block.
Let “f” be the cell failure probability for an element for a year. Using the fact that the number of ways 2 elements can be selected out of 138 is given by 138×137/2, the required value of f can be calculated as follows:
10−5/(78*69*137)=f2; f=0.3683*10−6
The required well depth can then be computed directly as follows:
0.3683*10−6=0.5*e−(E
Eb=50.7 kT 4*Eb=203 kT
Incidentally, if the full memory was read every 10 hours and corrections made, the required well depth would be reduced by a small amount to 47.3 kT.
Taking the more conservative value of Eb=228 kT for unselected memory cells, and assuming the temperature of a half selected cell rises to 400 K during the writing procedure of another cell, and that the saturation magnetization of the storage layer is about 10,000 emu/cm3, and the effective anisotropy of the storage layer is 20 Oe, the minimum volume required to maintain thermal stability is given by the difference between the maximum and minimum energy values, which occur at θ=0 and θ=π/2, through the following relationship:
½ sin2 θ|{right arrow over (M)}|{right arrow over (H)}kV=228 kT;
V=2*228(1.381×10−23 Joules/Kelvin)(400 Kelvin)/(10,000 emu/cm3)(20 Oe)=12.6*10−24 m3=12,600 nm3.
As memory array cell densities increases force reaching such cell dimensions, other ways of making cell selections in the memory array must be used that will allow thermal stability. One such method is the exploitation of the temperature properties of magnetic materials in the cell structures. Specifically, cell structures can be made that have high anisotropy (and a correspondingly deep energy well) at the standard data retrieving temperature, but whose magnetic properties diminish or vanish at a higher data storing temperature. Thus, the storing of data bits is accomplished by coincident applications of heat pulses and externally applied magnetic fields. As described above, the external applications of magnetic fields can be supplemented or replaced by the use of polarized spin currents for switching the cells magnetic states in the storing of data bits therein.
In such arrangements where memory cell structures have data stored therein by selectively heating the structures, the desired rapidity of storage is achieved by using thin isolating layers such as the dielectrics. In thin dielectrics, thermal equilibrium is reached in a very short time. This can be directly calculated by use of the Fourier heat conduction equation and the appropriate material parameters. The temperature rise is given in general by that equation as
where T is the temperature, t is the time, kth is the thermal conductivity, P is the density, Cp is the specific heat at constant pressure, and x-y-z are the coordinates. The thermal diffusivity constant is given by
In one dimension which is a reasonable approximation with a large heat sink directly below the heating sources, the temperature rise is given by
Consider a one micron wide line made of 100 Å thick tungsten over 500 Å of silicon dioxide in a substrate having a silicon integrated circuit therebelow with one of the memory cells provided in the silicon dioxide 200 Å below the word line. The thermal conductivity of silicon dioxide is 0.014 Watts/° C. cm, for tungsten 1.87 watts/° C. cm, and for silicon 1.5 watts/° C. cm. Because the silicon has more than 100 times the thermal conductivity of the silicon dioxide, the silicon in the remaining portions of the substrate can be treated as an infinite heat sink. The volume specific heat for silicon dioxide is 2.27 joules/° C. cm3 and for tungsten is 3.66 joules/° C. cm3 The tungsten line has a resistivity of about 8 ohms per square.
For this example, a 10 mA current was stepped abruptly into the word line and the temperature transient was determined using the above parameters in a numerical solution of the above on dimension diffusion equation.
In operation, using coincident pulses in the word current, Iw, and in the sense current, Is, to store and retrieve information from these kinds of memory cells, there will be typical waveforms of the kinds illustrated in
A similar full magnitude current pulse during that first time period is applied through the cell as a sense current as shown in
This storage is accomplished by letting the cell cool below the pertinent one of the Curie or blocking temperatures (represented by the long dashed line in
The sense current in the second time period shown in
Retrieving information from the same cell as to the direction of magnetization of the ferromagnetic material associated with the magnetically harder side of the nonmagnetic intermediate layer is shown in
During retrieval of data from this memory cell, field magnitudes only relatively near zero are applied to these characteristics so that the operating point moves from one to the other of the two resistance lobes shown in
A cross section view of a spin-current based data storage memory array provided as part of a monolithic integrated circuit also using thermal pulses to store data is shown in
The thermally operated memory cells in the array must function independently of one another, and so the thermal conduction paths in and around each cell must allow one cell to be brought above its storing temperature based on the Curie or blocking temperatures of the magnetic material layers therein while nearby cells remain below their thermally stable temperature. This balance is achieved by limiting thermal conductivity between the adjacent cells through suitable choices of a bottom electrode separating dielectric material layer, 41, a device separating dielectric material layer, 42, and a word line and device separating dielectric material layer, 43, both as to kind of material and the separating thicknesses thereof. Similarly, the material and dimensions of bottom electrode 2, and the material and dimensions of common bit line 9, must be chosen suitably to limit there contributions sufficiently to the thermal conductivity between adjacent devices.
The memory cell devices in
Also, separately from the foregoing, or together with it, a Néel point cell could be made by substituting an anti ferromagnet material for ruthenium layer 13¢ between storage layers 11′ and 12′. If the storing currents heat the antiferromagnet between storage layers 11′ and 12′ above the Néel point, then the antiferromagnetic/ferromagnetic exchange coupling between these two storage layers would be broken, anrd the magnetic field associated with the storing current would cause the corresponding magnetization state to occur in these storage layers to thereby store the data bit desired to be stored in that memory cell. As the cell cools, the cell will have a deep storage energy well by virtue of the very large exchange fields that the antiferromagnetic provided between the storage layers provides.
Note that in this case, the antiparallel coupling between the two storage layers can no longer be provided by the omitted ruthenium layer, and that instead, magnetostatic coupling would have to supply the antiparallel coupling. In even very thin films, this coupling becomes very large as cell dimensions approach and become less than 1000 Å. Even a slight ellipticity in the cell shape will give a very substantial anisotropy. For example, a 30 Å circular film 100 nm in diameter with an Ms of 10K Oe would have a self-demagnetization field of about 240 Oe. Ellipticity would make the magnetization lie along the long axis due to shape anisotropy, and the stray fields in one storage layer would magnetize the other storage layer in an antiparallel state.
A simpler structure is described below but this structure has shortcomings which will make the advantages of the previously described devices clearer. The layer diagram for this simpler structure memory cell, 10″, is shown in
A copper intermediate layer in the giant magnetoresistive devices means that the output signal will be taken from a vertical GMR device with only two interfaces yielding a relatively small output signal value. Also, because of symmetry, there is no signal unless storage layer 54 is in the output interconnection. The output signal in the memory cells of
The device of
Stray magnetic fields between memory cells of the kind shown in
Stray fields can be reduced even further by using a “doughnut” shape for the memory cells rather than the long and narrow cells of
Memories as described above use a transistor per cell to select the cell for data storage and retrieving. The low current levels needed to operate the cells allow the transistor to be small in size. Thermal pulse data storage techniques can also be used to enhance stability.
Although the present invention has been described with reference to preferred embodiments, workers skilled in the art will recognize that changes may be made in form and detail without departing from the spirit and scope of the invention.
Daughton, James M., Tondra, Mark C., Pohm, Arthur V.
Patent | Priority | Assignee | Title |
Patent | Priority | Assignee | Title |
4476454, | Jun 30 1983 | International Business Machines Corporation | New magnetoresistive materials |
4731757, | Jun 27 1986 | Micron Technology, Inc | Magnetoresistive memory including thin film storage cells having tapered ends |
4780848, | Jun 03 1986 | Micron Technology, Inc | Magnetoresistive memory with multi-layer storage cells having layers of limited thickness |
5565695, | Apr 21 1995 | Seagate Technology International | Magnetic spin transistor hybrid circuit element |
5587943, | Feb 13 1995 | Integrated Magnetoelectronics | Nonvolatile magnetoresistive memory with fully closed flux operation |
5629549, | Apr 21 1995 | Seagate Technology International | Magnetic spin transistor device, logic gate & method of operation |
5640343, | Mar 18 1996 | GLOBALFOUNDRIES Inc | Magnetic memory array using magnetic tunnel junction devices in the memory cells |
5650958, | Mar 18 1996 | GLOBALFOUNDRIES Inc | Magnetic tunnel junctions with controlled magnetic response |
5654566, | Apr 21 1995 | Seagate Technology International | Magnetic spin injected field effect transistor and method of operation |
5659499, | Nov 24 1995 | Everspin Technologies, Inc | Magnetic memory and method therefor |
5691936, | Aug 21 1995 | Matsushita Electric Industrial Co., Ltd. | Magnetoresistive element and memory element |
5695864, | Sep 28 1995 | International Business Machines Corporation | Electronic device using magnetic components |
5715121, | Dec 19 1995 | Western Digital Technologies, INC | Magnetoresistance element, magnetoresistive head and magnetoresistive memory |
5734605, | Sep 10 1996 | Everspin Technologies, Inc | Multi-layer magnetic tunneling junction memory cells |
5764567, | Nov 27 1996 | International Business Machines Corporation | Magnetic tunnel junction device with nonferromagnetic interface layer for improved magnetic field response |
5768181, | Apr 07 1997 | Everspin Technologies, Inc | Magnetic device having multi-layer with insulating and conductive layers |
5953248, | Jul 20 1998 | Everspin Technologies, Inc | Low switching field magnetic tunneling junction for high density arrays |
5966323, | Dec 18 1997 | Everspin Technologies, Inc | Low switching field magnetoresistive tunneling junction for high density arrays |
6055179, | May 19 1998 | Canon Kabushiki Kaisha | Memory device utilizing giant magnetoresistance effect |
6069820, | Feb 20 1998 | Kabushiki Kaisha Toshiba | Spin dependent conduction device |
6124711, | Jan 19 1996 | Fujitsu Limited | Magnetic sensor using tunnel resistance to detect an external magnetic field |
6130814, | Jul 28 1998 | International Business Machines Corporation | Current-induced magnetic switching device and memory including the same |
6140838, | Apr 21 1995 | Seagate Technology International | High density and high speed magneto-electronic logic family |
6166948, | Sep 03 1999 | International Business Machines Corporation | Magnetic memory array with magnetic tunnel junction memory cells having flux-closed free layers |
6172902, | Aug 12 1998 | ECOLE POLYTECHNIQUE FEDERALE DE LAUSANNE (EPFL) | Non-volatile magnetic random access memory |
6178111, | Dec 07 1999 | U S BANK NATIONAL ASSOCIATION, AS COLLATERAL AGENT | Method and apparatus for writing data states to non-volatile storage devices |
6178112, | May 13 1998 | Sony Corporation | Element exploiting magnetic material and addressing method therefor |
6185143, | Feb 04 2000 | SAMSUNG ELECTRONICS CO , LTD | Magnetic random access memory (MRAM) device including differential sense amplifiers |
6256223, | Jul 28 1998 | International Business Machines Corporation | Current-induced magnetic switching device and memory including the same |
6269018, | Apr 13 2000 | International Business Machines Corporation | Magnetic random access memory using current through MTJ write mechanism |
6272036, | Dec 20 1999 | U Chicago Argonne LLC | Control of magnetic direction in multi-layer ferromagnetic devices by bias voltage |
6292389, | Jul 19 1999 | Everspin Technologies, Inc | Magnetic element with improved field response and fabricating method thereof |
6376260, | Jul 19 1999 | Everspin Technologies, Inc | Magnetic element with improved field response and fabricating method thereof |
6385082, | Nov 08 2000 | International Business Machines Corp.; International Business Machines Corporation | Thermally-assisted magnetic random access memory (MRAM) |
6396734, | Aug 26 1999 | Hitachi Global Storage Technologies Japan, Ltd | Giant magnetoresistive sensor, thin-film read/write head and magnetic recording apparatus using the sensor |
6436526, | Jun 17 1999 | MATSUSHITA ELECTRIC INDUSTRIAL CO , LTD | Magneto-resistance effect element, magneto-resistance effect memory cell, MRAM and method for performing information write to or read from the magneto-resistance effect memory cell |
6438026, | Mar 09 2000 | Koninklijke Philips Electronics N V | Magnetic field element having a biasing magnetic layer structure |
6469926, | Mar 22 2000 | Everspin Technologies, Inc | Magnetic element with an improved magnetoresistance ratio and fabricating method thereof |
6535416, | Jun 18 1999 | NVE Corporation | Magnetic memory coincident thermal pulse data storage |
6538919, | Nov 08 2000 | International Business Machines Corporation | Magnetic tunnel junctions using ferrimagnetic materials |
6611405, | Sep 16 1999 | TOSHIBA MEMORY CORPORATION | Magnetoresistive element and magnetic memory device |
6657888, | May 11 2001 | J A WOOLLAM CO INC | Application of high spin polarization materials in two terminal non-volatile bistable memory devices |
EP910092, | |||
EP913830, | |||
WO9820496, | |||
WO9914758, |
Executed on | Assignor | Assignee | Conveyance | Frame | Reel | Doc |
Sep 14 2012 | NVE Corporation | (assignment on the face of the patent) | / |
Date | Maintenance Fee Events |
Nov 24 2015 | M2553: Payment of Maintenance Fee, 12th Yr, Small Entity. |
Date | Maintenance Schedule |
May 06 2017 | 4 years fee payment window open |
Nov 06 2017 | 6 months grace period start (w surcharge) |
May 06 2018 | patent expiry (for year 4) |
May 06 2020 | 2 years to revive unintentionally abandoned end. (for year 4) |
May 06 2021 | 8 years fee payment window open |
Nov 06 2021 | 6 months grace period start (w surcharge) |
May 06 2022 | patent expiry (for year 8) |
May 06 2024 | 2 years to revive unintentionally abandoned end. (for year 8) |
May 06 2025 | 12 years fee payment window open |
Nov 06 2025 | 6 months grace period start (w surcharge) |
May 06 2026 | patent expiry (for year 12) |
May 06 2028 | 2 years to revive unintentionally abandoned end. (for year 12) |