Conductance-dependent Photoresponse in a Dynamic SrTiO3 Memristor for Biorealistic Computing

Modern computers perform pre-defined operations using static memory components, whereas biological systems learn through inherently dynamic, time-dependent processes in synapses and neurons. The biological learning process also relies on global signals - neuromodulators - who influence many synapses at once depending on their dynamic, internal state. In this study, using optical radiation as a global neuromodulatory signal, we investigate nanoscale SrTiO3 (STO) memristors that can act as solid-state synapses. Via diverse sets of measurements, we demonstrate that the memristor’s photoresponse depends on the electrical conductance state, following a well-defined square root relation. Additionally, we show that the conductance decays after photoexcitation with time constants in the range of 1 - 10 s and that this effect can be reliably controlled using an electrical bias. These properties in combination with our device’s low power operation (< 1pJ per optical pulse) and small measurement variability may pave the way for space- and energy-efficient implementations of complex biological learning processes in electro-optical hardware.

💡 Research Summary

This paper presents a comprehensive study of a dynamic Pt/Cr–SrTiO₃–Ti/Pt memristor that exhibits a conductance‑dependent photoresponse, aiming to emulate the three‑factor learning rules observed in biological synapses. The device consists of two electrodes separated by a ~40 nm gap on a single‑crystal SrTiO₃ substrate, encapsulated by a SiN barrier to prevent oxygen and moisture ingress. Electrical characterization in the dark shows the characteristic “eight‑wise” hysteresis typical of SrTiO₃ memristors, which is attributed to modulation of the Schottky barrier by oxygen‑vacancy (V_O) creation and migration near the Pt/Cr–STO interface.

When illuminated with 365 nm UV light (photon energy 3.4 eV, above the SrTiO₃ bandgap), the current increases sharply. Importantly, after the optical pulse the conductance does not revert instantly; instead it decays exponentially over a timescale of 1–10 s. The decay exhibits an activation energy of ~0.33 eV, suggesting an ionic process rather than pure electronic trapping. By applying a read voltage of 0.6 V and varying the optical pulse width (up to 100 ms), the authors demonstrate that longer pulses produce larger steady‑state conductance increments, while the long‑term decay remains similar across pulse durations.

A key finding is that the magnitude of the photo‑induced conductance change depends on the pre‑existing conductance state. Experiments performed at two distinct initial conductances (low‑G and high‑G) reveal that the high‑G state yields a significantly larger immediate conductance jump under identical optical excitation. Quantitatively, the photo‑response follows a square‑root relationship: ΔG ∝ √P_opt, where P_opt is the optical power density. This state‑dependent behavior mirrors the biological concept that a neuromodulator’s effect is gated by the synapse’s current strength.

The authors further show that the decay dynamics can be precisely tuned by applying a DC bias during or after illumination. Positive bias accelerates the return to the baseline (shorter τ), while negative bias prolongs the elevated conductance (longer τ). This controllability stems from the electric field‑driven migration of V_O: a forward field drives vacancies toward the Pt electrode, facilitating recombination and rapid recovery; a reverse field hinders vacancy annihilation, preserving the photo‑induced state.

Density functional theory (DFT) calculations support the experimental interpretation. Simulations of the Pt‑STO interface indicate that UV photons generate electron‑hole pairs; holes become trapped near the interface, lowering the local Schottky barrier, while the resulting band bending promotes V_O formation. The calculated formation energy reduction near the high‑work‑function Pt electrode aligns with the observed low activation energy for the decay process.

Energy efficiency is highlighted: each optical pulse consumes less than 1 pJ (derived from a 65 mW cm⁻², 100 ms pulse over the device area). This is orders of magnitude lower than previously reported optically‑controlled memristors, making the approach attractive for large‑scale neuromorphic systems where power budgets are critical.

By integrating three elements—(i) a local electrical spike that creates a short‑term conductance change, (ii) a global optical neuromodulator whose effect scales with the current conductance, and (iii) an electrical bias that sets the decay time constant—the device naturally implements a three‑factor learning rule. Such a rule is essential for reinforcement learning, sequential learning, and adaptation in non‑stationary environments, offering a hardware substrate that can process both local and global signals with biologically realistic temporal dynamics.

In conclusion, the study demonstrates that SrTiO₃‑based dynamic memristors can provide conductance‑dependent, long‑lasting, and electrically‑tunable photoresponses at sub‑picojoule energy costs. These properties bridge the gap between conventional solid‑state memory and the plastic, state‑dependent modulation seen in biological synapses, paving the way for energy‑efficient, space‑saving electro‑optical neuromorphic circuits. Future work should explore scaling to dense cross‑bar arrays, wavelength‑dependent modulation, and integration with CMOS control logic to realize full‑system neuromorphic processors.