iFCN: An Automated RTL-to-Device Framework for Molecular Field-Coupled Nanocomputing Circuits

Molecular Field-Coupled Nanocomputing (MolFCN) offers a promising post-CMOS alternative, characterized by ultra-low power consumption and high integration density. However, existing MolFCN design flows face critical challenges, including rigid clock-phase constraints, inefficient placement and routing, and the absence of accurate gate-to-device mapping. This paper introduces iFCN, an automated RTL-to-device-level design framework specifically optimized for MolFCN circuits. Building upon prior heuristic methods, iFCN incorporates inverter pruning at the RTL level to simplify circuit structure, utilizes Morton-coded linear quadtrees for efficient spatial indexing, and enhances A* routing to maximize path reuse for multi-fanout nets. To address limitations of fixed-phase clocking, we propose a hierarchical placement approach guided by Graph Convolutional Networks (GCN). The GCN learns connectivity-aware node embeddings that guide recursive partitioning and intra-layer ordering, significantly reducing wire crossings and improving placement quality. Subsequently, a lightweight adaptive method heuristically assigns clock phases to each layout layer, with careful consideration of timing and topological constraints. Additionally, we present an accurate gate-to-cell mapping algorithm to facilitate direct physical simulation and energy analyses. Benchmark results show a 30% reduction in runtime and a 10% improvement in layout area compared to our heuristic algorithm. Furthermore, the proposed method achieves comparable runtime performance to the state-of-the-art fiction tool, completing the layout of circuits with over 150 nodes in under one second. In addition, Comparative analyses against 12nm CMOS designs confirm that MolFCN circuits generated by iFCN exhibit superior area utilization and reduced power consumption, highlighting the practical potential of MolFCN for future ultra-low-power computing. All source code and data are available at https://github.com/li-yangshuai/iFCN