Jamming-controlled stochasticity in metal-insulator switching

Abstract

Understanding and controlling phase transitions is a fundamental part of physics and has been central to many technological revolutions, from steam engines to field-effect transistors. At present, there is strong interest in materials with strongly coupled structural and electronic phase transitions, which hold promise for energy-efficient technologies. Utilizing a structural phase transition and controlling its plasticity naturally leads to built-in memory, a key feature for emulating neurons and synapses in neuromorphic technologies. Here, $\textit{operando}$ Bragg X-ray photon correlation spectroscopy is used to study the evolution of the nano-domain distribution at the micron-scale in neuromorphic devices made from the archetypal Mott insulator vanadium dioxide. It is found that after electrical switching, slow nano-domain reconfiguration occurs on timescales of thousands of seconds and that the domains undergo a jamming transition, offering control over switching stochasticity at the micron scale. More precisely, repetitive above-threshold currents plastically drive the system into a jammed/glassy state where switching becomes deterministic, while sub-threshold currents erase the short-term memory contained in the nano-domain distribution, recovering stochastic switching, thus offering a path for in-device learning. The results illustrate the importance of studying the nanoscale physics associated with phase transitions in strongly correlated materials, even for macroscopic devices, and offer guidance for future device operation schemes.

Figures (5)

• Figure 1: X-ray diffraction and electrical setup.a and b, schematic representation of the X-ray photon correlation spectroscopy setup for a device in the Off and On states, respectively. The red area represents metallic domains within the insulating film (gray). Diffraction patterns are simulated based on the illustrated domain configurations. A current source connected to Ti/Au electrodes is used for electrical switching. c, experimental procedure for measuring switching reproducibility. The pulse sequences show the timing of electrical switching (top) and synchronized diffraction measurement (center). The bottom panel illustrates the transition from stochastic to deterministic switching observed in Fig. \ref{['fig_switch']}.
• Figure 2: Temperature dependence and activation energy. Two-times correlation at room temperature (a) and 340 K (b). The black arrow follows the $t_1 = t_2$ line and represents the aging time $t_\mathrm{age}$. The colored arrows ($t_1 + t_2 = t_\mathrm{age}$) are perpendicular to the black arrow, and correspond to the $g_2(t_\mathrm{age})$ data shown in c and d for various aging times at room temperature and 340 K, respectively. Fits were performed using equation \ref{['eq_exp_decay']} with a single $\tau$ value for all $t_\mathrm{age}$. The inset in d shows an Arrhenius plot, giving an activation energy $E_A = 2700 \pm 800$ K ($230 \pm 70$ meV).
• Figure 3: Short-term memory, erasure, and learning.a and b, two-times correlation measured as a sub-threshold current I = 0.26 mA and an above-threshold current I = 4 mA is set, respectively. The switching time is set to t = 0 s and the temperature at T = 313.15 K. Electrical switching sequences are illustrated in d, e, and f. In a the initial state remains unchanged for $\sim$1000 s after switching (memory). In b it is immediately erased. g and h, $t_\mathrm{age}$ dependent correlation decay $g_2(t)$ obtained from the colored arrows in a and b. Fits were performed using equation \ref{['eq_exp_decay']} with a single $\tau$ value for all $t_\mathrm{age}$. c, two-times correlation while switching on and off the device every three seconds. The top x-axis shows the cycle number. i, normalized correlation between subsequent switching events $\Delta G(n,n+1)$ as a function of cycle number, corresponding to the orange arrow in c. After 1000 cycles the device has learned a preferred nano-domain configuration resulting in deterministic switching (high correlation).
• Figure 4: Jamming transition. Stretching exponent $\beta$ obtained from fits using equation \ref{['eq_exp_decay']} as a function of $t_\mathrm{age}$ for three switching currents. $\beta$ = 1 signifies liquid-like dynamics, and $\beta$ = 1.5 jammed/glassy dynamics. The solid lines are sigmoid fits, used to extract the transition time $t_\mathrm{jam}$.
• Figure 5: Resistance vs temperature and current.a, thermal switching measured prior to the XPCS experiment for the device exposed to a 0.1 pC/$\mu$m$^{2}$ Ga$^+$ dose. b, electrical switching measured during the XPCS beamtime. $I_1$ and $I_2$ indicate the first and second switching currents.