High-Throughput Screening of Bulk Boride Superconductors with Honeycomb-Ruby Frameworks

Metal borides are promising phonon-mediated superconductors, yet discovering new high superconducting transition temperature (T_c) bulk phases beyond MgB₂-type motifs remains challenging. Here we propose a bulk layered boride prototype built from a honeycomb-ruby-honeycomb trilayer boron framework with three inequivalent metal sites for chemical tuning. Using a universal machine learning interatomic potential-accelerated high-throughput workflow combined with first-principles calculations, we screen over 120,000 candidates and identify 38 dynamically stable, near-hull compounds with T_c > 5 K, reaching T_c = 28.9 K for LiTiRuB₁₈. Statistical analysis reveals a strong positive correlation between T_c and the B-p density of states at Fermi surfaces. Representative electronic and electron-phonon coupling analyses show multiband Fermi surfaces formed jointly by B-σ and B-π states and demonstrate that, beyond high-frequency B-B modes, metal-regulated soft B-M hybrid modes dominate the variation of electron-phonon coupling. These results expand the structural diversity of layered borides and establish an efficient data-driven route to bulk boride superconductors at ambient pressure.