ML-Based Error Mitigation Approach for Overcoming NISQ Hardware Constraints in Fermionic Nanoelectronic Simulations

M. Przygocki, R. Kotas, M. Zubert (Lodz Univ. of Techn., Poland)

Simulating nanoelectronic solid-state systems on quantum computers presents a promising alternative to classically intractable computations. However, executing these models on Noisy Intermediate-Scale Quantum (NISQ) devices introduces significant decoherence and gate errors. In this paper, a 3-site asymmetric quantum wire is mapped using the tight-binding model and Jordan-Wigner transformation. The Trotter-Noise trade-off and algorithmic circuit depth are analyzed to identify an optimal practical baseline for simulation. Consequently, this reveals the critical limitations of current quantum processors, demonstrating systemic signal degradation. To overcome this without physical error correction overhead, a machine learning (ML)-based Quantum Error Mitigation (QEM) strategy using a Radial Basis Function (RBF) Kernel Ridge Regression (KRR) applied as a classical post-processing filter is proposed. Trained on stochastically generated Hamiltonians, the developed ML model successfully recovers the non-linear physical dynamics on an unseen cross-backend architecture. The classical training overhead is strictly negligible (millisecond-scale), proving the viability of scalable, ML-based quantum Electronic Design Automation (EDA).