Thermal tuning of dynamic response in Ag-based nanowire networks

Abstract

Self-assembled networks of metallic nanowires (NWs) are being intensively explored as test benches for neuromorphic proposals. In this work, we study the electric transport properties of dense self-assembled networks of Ag-based NWs (AgNWNs) coated with a thin insulating layer, using DC and AC stimuli. The building blocks of this network are the metallic NWs and the NW-NW junctions, either metallic or memristive. In the pristine state, frequency independence of the impedance reveals an over-percolated purely resistive network. A combination of low-temperature annealing and AC stimulus is shown to drastically affect the resistivity of the sample (interpreted as a depopulation of purely metallic junctions), unveiling a rich dynamic response. This procedure triggers the achievement of a capacitive response, which is successfully rationalized using a previously introduced 'two junction model'. Thermal treatment appears to be an indirect strategy to effectively modify the humidity content at the NW-NW intersections and, consequently, enable multiple switching schemes suitable for brain-like processing alternatives.

Figures (11)

• Figure 1: Pristine state of dense Ag NWNs. (a) Impedance’s absolute value, Z, and (b) phase measured as a function of frequency, f (from 20 Hz to 2 MHz) with an AC amplitude of 20 mV, a bias voltage set to 0 V, and T = 20°C (RT). (c) Z as a function of temperature, recorded at different f. Multiple f sweeps from 20 Hz to 2 MHz are successively recorded. Z readings for specific f (1 kHz, 10 kHz, 100 kHz, and 1 MHz) were selected and averaged over the four runs. The mean value and standard deviation for each selected f are plotted as a function of T as solid circles and error bars, respectively. The inset depicts the time-domain representation of the data, where Z exhibits stochastic, non-monotonic excursions but a gradual overall rise, with no correlation to the excitation frequency.
• Figure 2: Before and after low-T annealing impedance responses recorded at RT. (a) Ten frequency, f, sweeps between 20 Hz and 2 MHz (AC amplitude = 20 mV and V$_{\mathrm{DC}}$ = 0 V) were applied to record the complex impedance response. The dependencies of (b) Z (logarithmic scale) and (c) phase (linear scale) as a function of f illustrate the contrast between the ohmic response in the pristine condition and the capacitive dependence obtained before and after the conducted annealing, respectively. In addition, Z exhibits intermittent instabilities, abruptly switching between capacitive and purely resistive behavior, consistent with sudden transitions from a high-resistance state to intermediate resistance levels. The black dashed line in Z indicates the instrumental limit (1 G$\Omega$lcr-meter).
• Figure 3: Successive runs of complex impedance measured upon different DC bias voltage conditions. (a) f and (b) V$_{\mathrm{DC}}$ programmed for the sequential runs. (c) Z and (d) phase recorded as a function of time. All the measurements were conducted using an AC amplitude of 20 mV.
• Figure 4: Bias voltage sequences. Successive runs of complex impedance measured as a function of the f (from 20 Hz to 2 MHz) recorded at different bias voltages. Four cycles of V$_{\mathrm{DC}} = \{0;0.5;0;1.2\}$ V are implemented to observe the progression of Z and phase.
• Figure 5: Temporal evolution of the impedance depending on the external temperature. (a) Frequency, f (b) impedance's absolute value, Z, and (c) phase as a function of time, for different external temperatures.