Extended Data Fig. 1: Investigation on the possible effect of temperature rise on the observed memory effect.

a, Measured current of a VO₂ switch under triangular excitation. The MIT happens between points A and B with a short time separation of Δt_AB ≈ 100 ns. The inset presents the extracted resistance of the device in the IMT and MIT cycles showing that post excitation resistance is close (within 1%) to the pre-IMT resistance. This is another indication of the fast sub-microsecond cooling. b, Thermal microscopy of a two-terminal VO₂ switch triggered by 10-μs pulses (V_set = 2.1 V) with three different frequencies 0.1 Hz, 1 Hz and 10 Hz. The captured average temperature over the device does not show any notable difference between the three cases. This indicates that given the long time duration of our observed memory, this memory effect cannot be originated from thermal effects. c,d, Investigation of the memory effect for different excitation amplitudes V_set = 4.2 V and 2.1 V, respectively. If thermal effects induced the memory effect, then the 4.2 V excitation should result in a more pronounced change in the incubation time, because it leads to a higher temperature rise comparing to 2.1 V excitation. The measured incubation times, however, are identical, which disproves the role of thermal effects in our observed memory.