Impact of TID on the Analog Conductance and Training Accuracy of CBRAM-Based Neural Accelerator

• Published in: IEEE transactions on nuclear science Vol. 70; no. 12; pp. 2572 - 2577
• Main Authors: Apsangi, P., Chamele, N., Barnaby, H. J., Gonzalez-Velo, Y., Holbert, K. E., Kozicki, M. N.
• Format: Journal Article
• Language: English
• Published: New York IEEE 01.12.2023; The Institute of Electrical and Electronics Engineers, Inc. (IEEE)
• Subjects: Analog synapse; Artificial neural networks; Conductance; conductive bridge random access memory (CBRAM); Conductivity measurement; Deep learning; Depression; Gamma rays; Ionization; Irradiation; Neural networks; neuromorphic accelerator; neuromorphic computing; Neuromorphic engineering; Performance evaluation; Programming; Radiation; Random access memory; Resistance; resistive random access memory (RRAM); Scanning electron microscopy; Switches; Synapses; total ionizing dose (TID); Training
• ISSN: 0018-9499; 1558-1578
• DOI: 10.1109/TNS.2023.3318511

The changes caused by total ionizing dose (TID) in the conductance of the analog response of Ag-Ge30Se70 conductive bridge random access memory (CBRAM) based synapses are studied. The conductance was seen to be severely affected by 60Co gamma ray exposure. The devices were tested up to a TID of 1 Mrad(Ge30Se70). Scanning electron microscopy imaging supports the conclusion that the device degradation is due to loss of filament contact after irradiation. The effect of irradiation was further analyzed by simulating the devices in an artificial neural network. The training accuracy was seen to degrade from 85% to 15% with increasing TID levels.