Theoretical modelling and simulation form the core of WP3, where computational techniques are employed to predict and optimize the behaviour of ZnO TN–based devices, with emphasis on in-sensor preprocessors. A multi-scale approach will be used: DFT will analyse electronic structure, defect states, and functionalization effects at the atomic scale; mesoscopic percolation models and FEM will simulate charge transport, local fields, and nonlinear dynamics across larger network architectures. Together, these methods elucidate how ZnO TN structure, connectivity, and surface states govern responses to electrical bias and, where relevant, UV/chemo stimuli, and how these responses translate into neuromorphic behaviours such as multi-level states, fading memory, and adaptive temporal filtering within the sensing substrate. To link materials to function, simulations will be parameterized and validated with WP2 data (morphology from EM/AFM, crystallinity from XRD, I–V and stability metrics). The framework will explore device architectures, operating regimes, and connectivity patterns to identify configurations meeting project targets—latency, energy per update, accuracy parity with software baselines, and 24-h stability/drift—and to define practical bias/readout windows for prototypes. While reservoir computing will serve as one benchmark of computational capacity in time-varying signals, WP3 remains open to alternative paradigms (e.g., spiking neural networks and other physical learning models) to fully exploit ZnO TN potential for in-sensor computing. POT will lead charge-transport and nonlinear-dynamics modelling; IT will focus on device-level memory effects and readout strategies. Continuous iteration with experimental partners (including SMEs) will sustain a simulation–experiment feedback loop—simulations guide prototypes; measurements refine models—accelerating optimization toward TRL-3 demonstrators and ensuring practical implementation in functional neuromorphic in-sensor preprocessors based on ZnO TNs.
