Spin orbit torque based electronic neuron

Abstract

An electronic neuron device that includes a thresholding unit which utilizes current-induced spin-orbit torque (SOT). A two-step switching scheme is implemented with the device. In the first step, a charge current through heavy metal (HM) places the magnetization of a nano-magnet along the hard-axis (i.e. an unstable point for the magnet). In the second step, the device receives a current (from an electronic synapse) which moves the magnetization from the unstable point to one of the two stable states. The polarity of the net synaptic current determines the final orientation of the magnetization. A resistive crossbar array may also be provided which functions as the synapse generating a bipolar current that is a weighted sum of the inputs of the device.

1 . A thresholding device for an electronic neuron, comprising:
 a) a heavy metal layer having a high spin orbit coupling;   b) a perpendicular magnetic anisotropy free layer having a bottom surface in contact with a top surface of the heavy metal layer;   c) a perpendicular magnetic anisotropy pinned layer; and   d) an oxide tunnel barrier connected between the free layer and the pinned layer, wherein the pinned layer, the oxide tunnel barrier, and the free layer form a magnetic tunnel junction.   
     
     
         2 . The thresholding device of  claim 1 , further comprising a current source, the current source configured to:
 a) supply a first charge current through the heavy metal layer, from a first end of the heavy metal layer to a second end of the heavy metal layer in a first direction along a first axis of the heavy metal layer to generate a torque which aligns the free layer magnetization in a direction along a second axis transverse to the first axis; and then   b) supply a second charge current from the pinned layer, through the magnetic tunnel junction, to the second end of the heavy metal layer to exert a torque on the magnetization of the free layer to align the free layer to either one of two orientations along a third axis, the third axis transverse to the first and second axis, the two orientations anti-parallel to one another.   
     
     
         3 . The thresholding device of  claim 1 , wherein the second supply charge current is substantially zero to allow the magnetic orientation of the free layer to randomly select between the two orientations along the third axis. 
     
     
         4 . The thresholding device of  claim 1 , wherein the heavy metal comprises beta-Tantalum, Tungsten, or Platinum. 
     
     
         5 . The thresholding device of  claim 1 , wherein the pinned layer and the free layer comprise a ferromagnetic material. 
     
     
         6 . The thresholding device of  claim 5 , wherein the pinned layer and the free layer comprise CoFe or CoFeB. 
     
     
         7 . The thresholding device according to  claim 5 , wherein the oxide tunnel barrier comprises MgO. 
     
     
         8 . A random number generating device, comprising:
 a) a heavy metal layer having a high spin orbit coupling;   b) a perpendicular magnetic anisotropy free layer having a bottom surface in contact with a top surface of the heavy metal layer;   c) a perpendicular magnetic anisotropy pinned layer;   d) an oxide tunnel barrier connected between the free layer and the pinned layer, wherein the pinned layer, the oxide tunnel barrier, and the free layer form a magnetic tunnel junction; and   e) a current switching device; and   f) a current source, the current source configured to supply a first charge current through the heavy metal layer, from a first end of the heavy metal layer to a second end of the heavy metal layer in a first direction along a first axis of the heavy metal layer to generate a torque which aligns the free layer magnetization in a direction along a second axis transverse to the first axis, and then turn off to allow a random thermal field to tilt the magnetization of the free layer  104  closer to one of two stable orientations.   
     
     
         9 . An artificial neural network arrangement, comprising:
 a) a plurality of electrically conductive row crossbars, each row crossbar connect to a first terminal of a plurality of resistive memory elements;   b) a plurality of electrically conductive column crossbars, each column crossbar connected to a second terminal of a plurality of resistive memory elements;   c) a plurality of thresholding devices, each of the thresholding devices comprising:
 i) a heavy metal layer having a high spin orbit coupling; 
 ii) a perpendicular magnetic anisotropy free layer having a bottom surface in contact with a top surface of the heavy metal layer; 
 iii) a perpendicular magnetic anisotropy pinned layer; 
 iv) an oxide tunnel barrier connected between the free layer and the pinned layer, wherein the pinned layer, the oxide tunnel barrier, and the free layer form a magnetic tunnel junction, wherein the pinned layer is connected to one of the column crossbars. 
   
     
     
         10 . The neural network arrangement according to  claim 9 , further comprising:
 a) a plurality of electronic switching devices each connecting the row crossbars to a voltage source.   
     
     
         11 . The neural network arrangement according to  claim 10 , wherein each of the electronic switching devices comprises a control input, and wherein each control input is connected to an input signal. 
     
     
         12 . The neural network arrangement according to  claim 11 , wherein each input signal is connected to the control inputs of a pair of the electronic switching devices; and wherein one of said pair of electronic switching devices is connected to a positive voltage source and the other one of said pair of electronic switching devices is connected to a negative voltage source. 
     
     
         13 . The neural network arrangement according to  claim 12 , wherein the electronic switching devices comprise a transistor. 
     
     
         14 . The neural network arrangement according to  claim 9 , wherein resistive memory elements comprise GeSbTe memristors. 
     
     
         15 . The neural network arrangement according to  claim 9 , wherein the resistive memory elements comprise Ag—Se memristors.