Silver nanowires coated in a polyvinylpyrrolidone (PVP) shell are memristive devices: each junction between two wires behaves as an electrochemical metallization cell, where the formation and dissolution of a silver filament tunes the junction conductance based on the history of the applied electrical stimulus. When thousands of these junctions self-assemble into a network, the collective evolution of the interacting memristive cells gives rise to complex, bio-inspired spatiotemporal dynamics. Within the physical reservoir computing framework, computation relies on the ability of such a physical system to transform input signals into a high-dimensional dynamical state space that is then analyzed by a trained readout layer. The network dynamics itself is central to information processing.

Yet how these dynamics depend on material-level parameters set during self-assembly has remained insufficiently understood. Areal mass density (AMD), the mass of nanowires deposited per unit area, directly affects the number of possible junctions and conductive pathways, and therefore the overall connectivity of the network. Conventional two-terminal measurements provide only a global description of the electrical response, whereas the internal processes relevant for computation are inherently distributed across the network. To resolve this internal organization, the researchers used a multiterminal approach with a 4 × 4 array of 16 gold-plated spring-loaded pogo pins pressed onto the sample surface, so that the mechanical force of the pins locally disrupts the PVP shell and establishes electrical contact with the silver nanowire core. The electrodes were connected to a characterization board equipped with 16 source-measurement units, enabling simultaneous electrical stimulation and readout from all terminals.

The study was conducted on 45 samples in total, with three samples analyzed for each density value in the investigated range between 11.9 and 39 mg/m². Networks were fabricated on 1 × 1 cm quartz substrates by drop-casting a commercially available suspension of silver nanowires in isopropyl alcohol, with the AMD controlled by diluting the initial suspension before deposition. During operation, one electrode located at a corner of the array was used as the bias input, while the opposite corner electrode was used as ground, in order to maximize the effective signal propagation across the network. The remaining 14 electrodes were used as floating readout nodes, and the recorded floating voltages served as reservoir states for the computational benchmarks, which were carried out in Python using the RCbench package developed for benchmarking physical reservoir systems.

Before analyzing the multiterminal spatiotemporal response, the pristine two-terminal conductance measured between opposite corner electrodes was evaluated as a function of nanowire density. The conductance follows a roughly sigmoid-like dependence on density, with a low-conductance branch at low densities, a clear activation jump around 30 mg/m², and a high-conductance branch above the jump. This trend arises from a percolation-driven transition in the network, where increasing nanowire density progressively enhances interwire connectivity and junction density, producing low conductance below the percolation threshold, rapid conductance activation near the critical density, and saturation at high densities. The general behavior shows first signs of global activation towards 25 mg/m² and conductance saturation above 32 mg/m². Notably, samples with similar AMD may still exhibit different initial conductance states, and the existence of a structural backbone connecting stimulating electrodes alone does not guarantee a conductive pathway, since memristive cells along the pathway must allow electronic conduction.

The multiterminal analysis then revealed a clear density dependence in how the network responds in a coordinated versus differentiated way. Pairwise zero-lag correlation, the Pearson coefficient between voltage signals recorded at two electrodes simultaneously, was computed for every electrode pair. The resulting correlation matrices show that the most heterogeneous patterns are observed below the activation jump, at densities up to about 30 mg/m², where highly correlated and weakly correlated electrode pairs coexist within the same sample. This indicates a strongly differentiated response, consistent with uneven signal propagation across a sparse conductive backbone in which only a subset of pathways carries most of the current. At higher densities, the matrices become progressively more uniform and dominated by high correlation values, reflecting a transition toward a more homogeneous and synchronized collective behavior.

Temporal dynamics told a complementary story. Fading-memory profiles showed that lower-density networks retain higher autocorrelation values over longer lags, with a smoother and slower decay, corresponding to a longer persistence of past input information at low density, whereas denser networks discard prior inputs more rapidly. Spectral centroid analysis, which reports the amplitude-weighted mean frequency of the signal’s power spectral density, showed that the spectral centroid increases monotonically with density, indicating that denser networks respond on shorter timescales. This behavior is expected from network literature, where studies demonstrate how a more interconnected network has faster information flow. Taken together, decreasing density shifts the network toward slower temporal dynamics, while increasing density produces faster responses.

The spatial distribution of computationally relevant activity was probed by evaluating the contribution of each electrode to the memory capacity (MC) benchmark task as a function of delay. At low density, the MC maps show that relevant contributions are unevenly distributed across the array, with a pronounced electrode-to-electrode contrast indicative of a spatially heterogeneous representation of temporal information. At intermediate density, the overall contribution decreases and the electrode-to-electrode contrast narrows, with contributions becoming more uniform across the array. At the highest density, memory-related contributions are strongly reduced and become nearly absent across the array, resulting in a uniformly low response. Voltage maps confirmed that the electrical response remains spatially nonuniform over the full density range, but the MC analysis was needed to reveal how these changes affect the spatial distribution of processing, since voltage maps alone do not directly identify which electrodes provide the most relevant contribution to information processing.

The implications are striking: the most useful computational regime is not the most synchronized one. The results indicate that nanowire AMD does not simply tune the overall conductance of the network, but also controls the balance between connectivity, temporal persistence, and synchronization. In the sparse-connectivity regime, the coexistence of highly correlated and weakly correlated responses indicates a richer internal dynamical organization, in which different regions of the network evolve with different degrees of coupling and contribute unequally to the overall response. Such spatial differentiation is consistent with the spatially heterogeneous distribution of memory-related contributions and with the longer temporal persistence emerging from the fading-memory analysis. By contrast, increasing density progressively drives the network toward a more homogeneous response, reducing the diversity of accessible internal states and, consequently, the temporal richness available for computation. This points to a density-dependent trade-off between persistence and speed of response: lower densities favor longer temporal traces of the input, while higher densities favor faster but less memory-rich dynamics.

For the emerging field of physical reservoir computing, where self-organizing memristive nanowire networks are promising substrates because their computational properties emerge from complex spatiotemporal dynamics distributed across the network, AMD therefore emerges as a key experimental parameter for designing multiterminal nanowire networks with targeted dynamical regimes. The most evident condition for useful computation in the present system is one in which sufficient connectivity coexists with substantial spatiotemporal heterogeneity, most evident in the sparse-connectivity regime below the activation jump, where the network shows the strongest spatial differentiation while still maintaining distributed signal propagation. As neuromorphic and reservoir-computing applications move from proof-of-concept toward practical devices, the ability to dial in memory, speed, and the spatial distribution of processing through a simple self-assembly parameter could prove decisive.

Subject of Research: Nanotechnology

Article Title: The role of areal mass density on spatiotemporal dynamics of multiterminal memristive nanowire networks

Article References: Pilati, D., Michieletti, F., Milano, G., & Ricciardi, C. (2026). The role of areal mass density on spatiotemporal dynamics of multiterminal memristive nanowire networks. Frontiers in Nanotechnology, 8, Article 1900378. https://doi.org/10.3389/fnano.2026.1900378

Image Credits: AI Generated

DOI: 10.3389/fnano.2026.1900378

Keywords: memristive networks, nanowire networks, neuromorphic computing, reservoir computing, spatiotemporal dynamics, unconventional computing, role, areal, mass, density, spatiotemporal, dynamics