Algorithmic Discovery of Room-Temperature Superconductor Candidates via Topological Stress Filtering in Causal-AI Networks

The precise geometric and topological mechanisms governing high-temperature and potential room-temperature superconductivity (RTS) remain obscured by the mathematical rigidity of continuous differential equations. In this paper, we operationalize the theoretical premise that superconductivity is not merely a quantum particle pairing compulsion, but a topological tension-release mechanism of the spatial network. Utilizing a Causal-AI physics engine that extracts deterministic Taylor sensitivity operators ($j_1, j_2, j_3$) from discrete spatial grids, we executed a forward-sensitivity scan across 57,426 known inorganic crystal structures sourced from the Materials Project database. By filtering for purely metallic conduction (band gap of 0.0 eV), the engine identified two distinct classes of anomalies: 1. Stable Topological Anomalies: We present 5 highly stable intermetallic and oxide phases (e.g., $\text{Be}_{17}\text{Os}_3$, $\text{Ta}_{11}(\text{Co}_{52}\text{B}_{15})_2$) offering immediate, testable targets for cryogenic geometric resonance. 2. Extreme Metastable Candidates (RTS Targets): We present the Top 10 global RTS candidates (e.g., $\text{Eu}_3\text{Si}_2\text{BO}_{10}$, $\text{LiReN}_2$) which exhibit extreme structural anisotropy and metastable geometric strain (energy above hull 0.05–0.20 eV), perfectly matching the dimensional frustration required for flat-band electron correlation. This work bridges abstract topological network theory with actionable materials science, providing experimental physicists with a definitively narrowed, computation-backed search space for the next generation of superconductors.