Patent Grant
11917924
The present invention relates to a logical multiplier. In particular, the invention relates to an Interlayer Exchange Coupling (IEC) multiplier formed using ferromagnetic layers separated by a spacer layer of thickness sufficient to generate anti-ferromagnetic coupling of magnetic flux across the spacer layer.
Various device configurations have been composed using the Giant Magneto Resistive (GMR) effect, where a first ferromagnetic layer is separated from a second ferromagnetic layer by an extremely thin non-magnetic material (typically on the order of several atomic thickness of a conductor such as copper). The GMR effect results in a magnetic field in the plane of one of the ferromagnetic layers generating an anti-ferromagnetic field in the other layer, the anti-ferromagnetic field being co-planar and of opposite sense.
A multiplier device in the prior art uses semiconductor logic devices, which rely on semiconductor switching characteristics such as transconductance via field effect at an input to form the many logic elements required to perform multiplication. Semiconductor logic devices must remained powered on to retain the multiplication result, which consumes power.
A truth table for the inputs and outputs of a multiplier are shown in the table below:
A prior art device for ferromagnetic logic operations utilizes dipole magnetic field edge coupling of regions of ferromagnetic deposition onto a passive substrate. A dimensional limitation of nanoscale construction of these devices is that when the dipole coupled nanomagnets are scaled down to sub 50 nm, the reduction in associated energy barriers required to maintain magnetic state results in increased susceptibility towards soft errors whereby thermal noise causing perturbations of the ferromagnetic material can randomly flip the state of the nanomagnet and corrupt the information stored in it. Therefore, in order to successfully transfer information from one ferromagnetic dot to the other, the coupling energy between the nanomagnets should also exceed thermal noise. Although these devices work at low temperatures approaching 0° K, thermal noise interfering with the magnetic field orientation until the loss of magnetic properties associated with the material-specific Curie temperature is reached, and the device is no longer able to function. Additionally, when the dipole coupled nanomagnets are scaled down to sub 50 nm, the reduced size of the nanomagnets result in reduction of their energy barriers, which leads to increased susceptibility towards soft errors, where temperature fluctuations randomly flip the state of the nanomagnet and erase the stored information in it. Thus, in order to successfully transfer information from one dot to the other the coupling energy between the nanomagnets should be larger than thermal noise.
It is desired to provide an architecture and structures to perform arithmetic operations such as a two bit multiplication using ferromagnetic films, which operate reliably at room temperatures (25° C.) and above, and which retain the multiplication result in the absence of applied energy.
A first object of the invention is multiplier comprising an electrically conductive layer having a continuous surface of ferromagnetic material on one surface and a plurality of ferromagnetic regions on an opposite surface, the electrically conductive layer sufficiently thin to generate a ferromagnetic response, the plurality of ferromagnetic regions being arranged in a sparce rectangular array comprising at least five columns and 10 rows, the ferromagnetic regions further comprising an a0 input, an a1 input, a b0 input, and a b1 input, the ferromagnetic regions also including an m0 output, a plurality of m1 outputs, a plurality of m2 outputs, and a plurality of m3 outputs located on the second, third, and fourth columns of the sparse rectangular array.
An Interlayer Exchange Coupling (IEC) logic cell in a first example of the invention comprises a continuous bottom layer of ferromagnetic material, a layer of an extremely thin non-magnetic electrically conductive layer placed on the continuous layer of ferromagnetic material, and a plurality of regions of ferromagnetic material positioned on the extremely thin non-magnetic electrically conductive layer, the plurality of regions of ferromagnetic material formed on a top layer. The extremely thin non-magnetic electrically layer is sufficiently thin so as to generates the GMR effect and associated anti-ferromagnetic effect between top ferromagnetic layer regions and associated bottom ferromagnetic layer regions which form superposition magnetic fields in areas where not driven anti-ferromagnetic by the upper layer region.
A multiplier for [a1:a0] by [b1:b0] to generate an output [m3 m2 m1 m0] is formed with regions of ferromagnetic material on a first surface of a thin conductor such as copper, with a continuous layer of ferromagnetic material on a second surface opposite the first surface of the thin conductor. The regions of ferromagnetic material are arranged as a sparse rectangular array of ferromagnetic material regions in a first column, second column, third column, fourth column, and fifth column, the sparse rectangular array having a first row, second row, third row, fourth row, fifth row, sixth row, seventh row, eighth row, ninth row, and tenth row.
The first row has an a0 input ferromagnetic region at the second column.
The second row has a b0 input ferromagnetic region at the first column and an m0 output ferromagnetic region at the second column.
The third row has a zero ferromagnetic region in the second and fourth columns and a one ferromagnetic region in the third column.
The fourth row has a b0 input ferromagnetic region at the first column, m1 output ferromagnetic regions at the second column, third column, and fourth column, and a b1 input ferromagnetic region at the fifth column.
The fifth row has an a1 input ferromagnetic region at the second column, a zero ferromagnetic region at the third column, and an a0 input ferromagnetic region at the fourth column.
The sixth row has a b1 input ferromagnetic region at the first column, an m3 output ferromagnetic region at the second column, third column, fourth column, and a b0 input ferromagnetic region at the fifth column.
The seventh row has a zero ferromagnetic region at the second and fourth columns, and an m3 output region at the third column.
The eighth row has an a1 input ferromagnetic region at the second column and an m3 output ferromagnetic region at the third column.
The ninth row has a b1 input ferromagnetic region at the first column, an m2 output ferromagnetic region at the second and third columns, and a one ferromagnetic region at the fourth column.
The tenth row has a zero ferromagnetic region at the second column.
The a0, a1, b0, and b1 inputs are provided with a “1” magnetic field input with a magnetic field applied in a first direction parallel to the columns, and a “0” magnetic field with a magnetic field applied in a second direction opposite from the first direction. The one ferromagnetic regions have a magnetic field applied in the first direction and the zero ferromagnetic regions have a magnetic field applied in the second direction.
The m3, m2, m1, and m0 outputs are magnetic fields which are sensed as “1” in the first direction and “0” in the second direction.
As described in the background of the invention, it is desired to provide a nanomagnetic two input multiplier device which operates at higher temperatures than the prior art. The Interlayer Exchange Coupled (IEC) topology of the present invention has been shown to offer stronger interaction between thin ferromagnetic layers, resulting in greater scalability and better data retention at the deep sub-micron level, hence allowing magnetic interaction to be manipulated both in the vertical and lateral directions at the same time. According to known scientific approximations, a nominal value indicates as “on the order of” a nominal value is understood to be on the order of magnitude of the nominal value, or in the range one-tenth of the nominal value to ten times the nominal value. A value which is “approximately” a nominal value is understood to be +/−50% of the nominal value.
In the present disclosure, by convention, the example logic term a1, a0, b1, and b0 are binary magnetization values. The binary magnetization values of ‘1’ and ‘0’ will have magnetization values in generally opposing directions taken as a whole over the associated ferromagnetic regions. For simplicity in understanding the invention, a “1” value is understood as a magnetic vector where the majority of magnetic flux is oriented in an opposite direction from a “0” value, in the present examples, a “1” value has magnetic flux directed upwards on the sheet (along the long axis of the multiplier structure), and a “0” value is understood as a magnetic vector where the majority of magnetic flux is oriented in the opposite direction of the “1” value.
The ferromagnetic regions 204 and the top layer regions of first through fifth column and first through 10th row such as 210, 216, etc may be formed from an alloy such as 80% Nickel (Ni) and 20% Iron (Fe), and the conductive non-ferromagnetic layer 104 may be formed of Ruthenium (Ru), Copper (Cu), or other electrically conductive material. The ferromagnetic regions may be formed in a thickness on the order of 5 nm, and of dimensions on the order of 100 nm×100 nm. In one example of the invention, the conductive non-ferromagnetic layer may be in the range 1 nm to 10 nm thickness, or alternatively, on the order of magnitude of 1 nm to 10 nm thickness.
In the present series of figures, it is understood that the particular convention of logic “1” and “0” is arbitrary so long as uniformly applied, but for clarity in understanding the present examples, a “1” will represent a magnetic field pointing generally upwards toward a top edge of the drawings (or a majority of the field direction is upwards) and in the plane of the drawings, and a “0” will represent a magnetic field pointing downwards (or a majority of the field direction is downwards) and in the plane of the drawings. Other conventions are possible where the “1” and “0” magnetic sense is opposite each other, the convention in the present description is used only for clarity in understanding the invention.
The devices operate with two magnetization inputs for each multiplicand [a1 a0] and [b1 b0] and generate an output [m3 m2 m1 m0]. The multiplier inputs have identical response times, and since the magnetization is stored in the ferromagnetic layer until it is changed, the device does not dissipate any power after the input magnetic field is applied to ferromagnetic inputs a1, a0, b1, and b0. In general, the magnetic field polarizations of the multiplier inputs in the top layer establish anti-ferromagnetic responses in the bottom layer with superposition of the magnetic field surrounding the driven regions to provide a desired anti-ferromagnetic polarity in a top ferromagnetic region to form the four outputs [m3:m0]. The input ferromagnetic layers thereby generate an anti-ferromagnetic response in the bottom ferromagnetic layer, and these anti-ferromagnetic responses generate a superposition of magnetic fields in a region of the bottom layer which is separated from and anti-ferromagnetically coupled to a top layer output ferromagnetic region to form outputs at the various output regions.
The
The present examples are provided for illustrative purposes only, and are not intended to limit the invention to only the embodiments shown.
| Number | Name | Date | Kind |
|---|---|---|---|
| 20070223150 | Fukuzawa | Sep 2007 | A1 |
| 20170092843 | Torok | Mar 2017 | A1 |
| 20180176078 | Nigro | Jun 2018 | A1 |
| 20180228410 | Yeh | Aug 2018 | A1 |
| 20220385565 | Wiser | Dec 2022 | A1 |
| Number | Date | Country | |
|---|---|---|---|
| 20220336729 A1 | Oct 2022 | US |
