Efforts to emulate signaling produced at nerve cell synapses aren’t new. Many different approaches – CMOS circuits and ‘ionic-drift’ based memristor technology, for example – have been tried, all with various shortcomings. Last week, researchers from UMass, Loughborough University, Hewlett Packard Labs, and Brookhaven National Laboratory reported a new approach that closely mimics the Ca2+ diffusion dynamics that occur at synapses between human nerve cells.
Their work reported in Nature Materials (Memristors with diffusive dynamics as synaptic emulators for neuromorphic computing, Sept 26, 2016, online) could lead to new applications in neuromorphic computing. “In addition to providing a synapse emulator, the diﬀusive memristor can also serve as a selector with a large transient nonlinearity that is critical for the operation of a large crossbar array as a neural network. The results here provide an encouraging pathway toward synaptic emulation using diﬀusive memristors for neuromorphic computing,” report the researchers in the paper’s conclusion.
They note, “[CMOS] circuits have been employed to mimic synaptic Ca2+ dynamics, but three-terminal devices bear limited resemblance to bio-counterparts at the mechanism level and require significant numbers and complex circuits to simulate synaptic behavior. A substantial reduction in footprint, complexity and energy consumption can be achieved by building a two-terminal circuit element, such as a memristor directly incorporating Ca2+-like dynamics.”
Other efforts based on ionic drift have been used and, “Although qualitative synaptic functionality has been demonstrated, the fast switching and non-volatility of drift memristors optimized for memory applications do not faithfully replicate the nature of plasticity,” according to the paper.
Quoted in an article on nanotechweb.org (Memristor behaves like a synapse), lead author J. Joshua Yang of UMass said, “In our memristors, we looked at how metallic atoms, like silver (Ag) or copper (Cu), diffuse through dielectric oxide materials. The way these metals diffuse through a dielectric is very similar, physically, to the way Ca2+ diffuses through channels in biological synapses.” Perhaps most important, the memristors created by the researchers exhibited relaxation after being turned on.
As described in the article, researchers made their silver-in-oxide memristors with two Pt or Au inert electrodes sandwiching a switching layer of a dielectric film with embedded Ag nanoclusters. “The device is essentially a volatile memristor where Ag atoms diffuse under the influence of electrical bias,” said Yang. “This electrical stimulation turns on the devices to their low resistance state, forming a nanoscale conduction channel of Ag (around 4 nm in diameter). When the electrical bias is removed, the device spontaneously relaxes back to its high resistance state thanks to the silver channel reshaping into spherical silver clusters.”
In the concluding section of their paper the researchers write: “[W]e have constructed and demonstrated a new class of memristors as synaptic emulators that function primarily on the basis of diﬀusion (rather than drift) dynamics…The Ag dynamics of the diﬀusive memristors functionally resemble the synaptic Ca2+ behavior in chemical synapses and lead to a direct and natural emulation of multiple synaptic functions for both short-term and long-term plasticity…”
Link to the Nature Materials paper: http://www.nature.com/nmat/journal/vaop/ncurrent/full/nmat4756.html
Link to the article written by Belle Dumé on nanotechweb.org: http://nanotechweb.org/cws/article/tech/66462
Link to related HPCwire article, IBM Phase Change Device Shows Promise for Emerging AI Apps: https://www.hpcwire.com/2016/08/03/ibm-phase-change-device-shows-promise-emerging-ai-apps/