Systems and methods for dynamically reconfigurable artificial synapses and neurons with tunable activation functions
An ionic redox transistor comprises a solid channel, a solid reservoir layer, and a solid electrolyte layer disposed between the channel and the reservoir layer. The channel exhibits a substantially linear current-voltage relationship in a first range of voltages, and a nonlinear current-voltage relationship in a second range of voltages that is greater than the first range of voltages. One or both of the substantially linear current-voltage relationship or the nonlinear current-voltage relationship of the channel is varied by changing the concentration of ions such as oxygen vacancies in the channel. Ion or vacancy transport between the channel and the reservoir layer across the electrolyte layer occurs in response to applying a voltage between the channel and the reservoir layer. Subject to the first range of voltages, the channel can function as a synapse device. Subject to the second range of voltages, the channel can function as a neuron device.
1. A device for neuromorphic computing, comprising:
a first ionic redox transistor and a second ionic redox transistor, where each of the first ionic redox transistor and the second ionic redox transistor comprises:
a channel layer having a first end and a second end;
a first contact positioned at the first end of the channel layer;
a second contact positioned at the second end of the channel layer, wherein:
the first ionic redox transistor is configured to operate as a synapse in a neural network based upon a first voltage in a first range of voltages being applied between the first contact and the second contact of the first ionic redox transistor, where the channel layer of the first ionic redox transistor exhibits a substantially linear current-voltage relationship in response to the first voltage being applied between the first contact and the second contact of the first ionic redox transistor; and
the second ionic redox transistor is configured to operate as an activation function in the neural network based upon a second voltage in a second range of voltages being applied between the first contact and the second contact of the second ionic redox transistor, where the channel layer exhibits a nonlinear current-voltage relationship in response to the second voltage being applied between the first contact and the second contact of the second ionic redox transistor, the second range of voltages being higher than the first range of voltages.
2. The device of claim 1 , where each of the first ionic redox transistor and the second ionic redox transistor further comprises:
a reservoir layer; and
an electrolyte layer positioned between the reservoir layer and the channel layer, the reservoir layer and the channel layer comprising a mobile ion, wherein the electrolyte layer is configured to permit transport of the mobile ion across the electrolyte layer.
3. The device of claim 2 , where each of the first ionic redox transistor and the second ionic redox transistor further comprises a third contact positioned on the reservoir layer, wherein responsive to a voltage being applied between the third contact and one of the first or second contacts, the mobile ion migrates from the reservoir layer to the channel layer or from the channel layer to the reservoir layer, wherein migration of the mobile ion causes the substantially linear current-voltage relationship of the channel in the first ionic redox transistor and the nonlinear current-voltage relationship of the channel in the second ionic redox transistor to change.
4. The device of claim 3 , wherein migration of the mobile ion causes a slope of the substantially linear current-voltage relationship in the first ionic redox transistor to change.
5. The device of claim 3 , wherein migration of the mobile ion causes a peak current of the nonlinear current-voltage relationship in the second ionic redox transistor to change.
6. The device of claim 3 , wherein migration of the mobile ion causes an activation voltage of the nonlinear current-voltage relationship in the second ionic redox transistor to change.
7. The device of claim 2 , wherein the mobile ion is oxygen vacancies.
8. The device of claim 1 , wherein the channel layer comprises vanadium dioxide (VO 2 ).
9. The device of claim 8 , wherein the channel layer comprises VO 2 alloyed with an element selected from a list consisting of:
titanium;
aluminum;
gallium; and
indium.
10. The device of claim 1 , wherein the channel layer comprises niobium dioxide (NbO 2 ).
11. The device of claim 1 , wherein the channel layer comprises a rare earth perovskite material with the composition RNiO 3 where R is one of samarium (Sm), praseodymium (Pr) or neodymium (Nd).
12. The device of claim 1 , wherein the first range of voltages extends from 0V to 0.5V.
13. The device of claim 1 , wherein the first range of voltages extends from 0V to 0.25V.
14. The device of claim 1 , wherein the first range of voltages extends from 0V to 0.1V.
15. The device of claim 1 , wherein the second range of voltages is greater than 0.5 V.
16. The device of claim 1 , wherein the second range of voltages is greater than 2.5 V.
17. A method, comprising:
applying a first voltage between a first contact and a second contact, the first voltage being within a first range of voltages, the first contact and the second contact positioned at opposite ends of a channel layer, wherein responsive to the first range of voltages being applied between the first contact and the second contact, the channel layer exhibits a linear current-voltage relationship; and
applying a second voltage between the first contact and the second contact, the second voltage being within a second range of voltages, the second range of voltages non-overlapping with the first range of voltages, and wherein responsive to the second range of voltages being applied between the first contact and the second contact, the channel layer exhibits a nonlinear current-voltage relationship.
18. A device for neomorphic computing comprising:
a first ionic redox transistor comprising:
a first channel layer;
a first contact coupled to the first channel layer at a first side of the first channel layer; and
a second contact coupled to the first channel layer at a second side of the first channel layer, where the first ionic redox transistor is configured to generate a first output based upon a first voltage applied between the first contact and the second contact, where the first ionic redox transistor is configured to operate as a synapse in a neural network based upon the first voltage being within a first range of voltages;
a second ionic redox transistor comprising:
a second channel layer;
a third contact coupled to the second channel layer at a first side of the second channel layer;
a fourth contact coupled to the second channel layer at a second side of the second channel layer, where the second ionic redox transistor is configured to generate a second output based a second voltage applied between the third contact and the fourth contact, where the second ionic transistor is configured to act as an activation function in the neural network based upon the second voltage being within a second range of voltages that is higher than the first range of voltages, and further where the second voltage is based upon the first output.
19. The device of claim 18 , where the first channel layer and the second channel layer comprise vanadium dioxide (VO 2 ).
20. The device of claim 18 , wherein the first channel layer and the second channel layer comprise a rare earth perovskite material with the composition RNiO 3 where R is one of samarium (Sm), praseodymium (Pr) or neodymium (Nd).