Identifying candidate hosts for quantum defects via data mining

The study addresses the need for a rational, systematic identification of solid-state host materials suitable for atom-like quantum defects used in quantum information science (QIS). While defect qubits in materials like diamond and silicon carbide have enabled advances in quantum networks and sensing, these systems represent only a small subset of potentially viable hosts and were largely discovered indirectly. The authors propose a comprehensive screening of inorganic materials to find hosts that support long spin coherence and efficient optical transitions. Ideal host properties include intrinsic diamagnetism (to reduce magnetic noise), a sufficiently large band gap to accommodate defect ground and excited states separated by optical frequencies, nonpolar crystal symmetry to minimize permanent electric dipole effects on defect optical coherence, high purity, and practical considerations such as dopability, surface control, and epitaxial growth capability. To reduce nuclear spin noise, elements with high natural abundance of zero-spin isotopes are favored, and isotopic enrichment is considered viable where zero-spin isotopes are ≥50%. The work aims to transform serendipitous discovery into a rational, database-driven search for host materials.

The paper situates its contribution within prior QIS work demonstrating solid-state spin qubits and their applications in quantum networks and nanoscale sensing, especially NV and related vacancy centers in diamond and defects in silicon carbide. It notes recent ab initio predictions for specific host-defect systems (e.g., SiC vacancy centers) and highlights the complementary approach of screening host materials by idealized criteria. The authors leverage several materials databases and computational frameworks—Materials Project (VASP GGA/GGA+U), JARVIS-DFT (OptB88vdW with mBJ), AFLOW (GGA+U with fitted band gaps), and AFLOW-ML (machine learning predictions)—and compare their band gap estimates with experimental literature values to mitigate known DFT biases (e.g., typical underestimation, especially with strong spin–orbit coupling).

The authors performed a four-stage screening of 125,223 inorganic compounds in the Materials Project (MP) database, with manual verification at each step and cross-referencing to JARVIS-DFT, AFLOW, AFLOW-ML, and literature. - Stage 1 (Automated MP query): Admit only phases (i) crystallizing in nonpolar space groups, (ii) derived from an experimental ICSD entry, (iii) composed exclusively of elements with >50% natural abundance of zero nuclear-spin (I=0) isotopes, and (iv) calculated by MP to be nonmagnetic (net magnetic moment ≈ 0). This stage produced the largest reduction in candidates (~97.3%). - Stage 2 (Manual curation): Remove phases containing uranium, thorium, cadmium, and mercury (radioactive/toxic), all noble gas-containing phases (no stable solids at STP), and all rare-earth-element-containing phases (difficulty in obtaining magnetically pure, spin-zero materials). Remove the one noble-gas solid that passed earlier filters (XeO3) due to instability under standard conditions. - Stage 3 (Band gap screening): For remaining phases, compile computed and predicted band gaps from MP (GGA with limited +U), JARVIS-DFT (OPT functional with mBJ correction; bulk structures re-optimized with OPT), AFLOW (GGA-PBE with PAW, relaxed with dense k-meshes, +U where needed; extraction of standard gap Esp, and a literature-calibrated fitted gap Egfit), and AFLOW-ML (metal/insulator classification and band gap regression using fragment descriptors). Retain phases likely to have band gaps ≥1.1 eV (silicon as lower bound); immediately discard those with MP-calculated band gaps <0.5 eV; assess borderline cases using literature reports and trends (e.g., approving some families like Ba–Hf–S based on general trends). Note that standard DFT gaps are often underestimated relative to experiment; JARVIS OPT-mBJ partially corrects this; AFLOW ‘fitted’ gaps may overestimate on average. - Stage 4 (Final validation): Confirm intrinsic diamagnetism using electron-counting rules and literature; verify phase stability at STP by removing high-pressure/high-temperature polymorphs not quenchable to ambient. Exclude phases with MP energy above hull (E Above Hull) >0.2 eV/atom as likely too unstable. Where needed, compare calculated and experimental band gaps, accounting for systematic under/overestimation biases (standard DFT and ML underestimation ~40–50%; JARVIS OPT-mBJ ~18.4% underestimation average; AFLOW fitted ~1.8% overestimation on average, with some outliers). Additional notes: some known stable phases (e.g., BaGeO3, BaGe2O5, C70) were missing from ICSD/MP and thus not included; fcc-C60 included via ICSD though absent in some databases. Single-crystal growth reports were noted but not used as exclusion criteria; air/moisture instability was noted but not exclusionary.

- From 125,223 MP inorganic compounds, 541 phases (0.43%) were identified as potentially viable host materials for quantum defects, a 99.57% reduction relative to all known phases (largest cull in Stage 1 at ~97.3%). - Composition breakdown of viable hosts: 16 unary, 74 binary, 322 ternary, and 129 quaternary or higher-order compounds; many are oxides and chalcogenides. - Band gap viability: At least 521 of the 541 candidates could be confirmed or reasonably assumed to have band gaps ≥1.1 eV (Si), based on computed and/or reported experimental values; the remaining 20 candidates had gaps slightly below or only tentatively extrapolated to ~Si values. - Elemental presence: Of the admitted elements, only Ni and Fe did not appear in any listed viable phases (Ni/Fe-containing phases considered were intrinsically paramagnetic). Many Os-containing and most Ru-containing entries in the tabulations corresponded to transition-metal carbonyl clusters. - Stability filtering (Stage 4): Excluding phases with MP E Above Hull >0.2 eV/atom removed 39 (largely ternary/quaternary) entries from otherwise viable sets; distributions of E Above Hull were analyzed by composition class. - Band gap benchmarking: Relative to experimental gaps, standard DFT and ML predictions (MP, JARVIS OPT-only, AFLOW unfit, AFLOW-ML) underestimated by ~40–50% on average; JARVIS OPT-mBJ underestimation averaged ~18.4%; AFLOW fitted gaps overestimated by ~1.8% on average, with notable outliers. - Practical implication: Although 541 candidates are still numerous, 90 are unary/binary and thus are generally simpler to synthesize and study; further application-specific criteria and higher-fidelity computations can narrow the set further.

The work provides a comprehensive, cross-validated catalog of host materials likely to support atom-like quantum defects for QIS applications. By integrating automated database queries with manual curation and literature verification, the authors ensure that identified hosts satisfy essential criteria: intrinsic diamagnetism, nonpolar symmetry, sufficient band gaps, and ambient stability. The findings address the research goal by shifting from serendipitous discovery to a systematic, scalable approach that can be extended to new materials and applications. The identified candidates, especially the simpler unary and binary systems, can prioritize experimental exploration for quantum networks and quantum sensors. The cross-database validation of band gaps and stability mitigates known computational biases, particularly DFT underestimation and heavy-element SOC effects. The approach and dataset enable targeted follow-up studies, including device-specific filtering (e.g., surface stability, defect formation energies, optical transition lifetimes), thereby accelerating the development pipeline for defect-based quantum technologies.

This study establishes a rational, four-stage screening framework that reduces over 125,000 inorganic phases to 541 viable candidates for hosting quantum defects, encompassing a broad chemical space with many oxides and chalcogenides and a significant subset of simpler unary/binary systems. The methodology, which blends database mining with manual verification and literature benchmarking, is generalizable to emerging phases and other materials discovery challenges. Future work should incorporate higher-accuracy electronic structure methods, defect-specific modeling (defect formation energies, charge states, zero-phonon lines, excited-state lifetimes), surface stability assessments, and application-driven filters to further refine candidates. Experimental validation—particularly crystal growth, isotopic enrichment, and characterization of defect coherence and optical properties—will be essential to realize practical quantum devices.

- Dependence on database coverage: Some known stable phases lack experimental ICSD entries and were thus absent from MP, leading to potential omissions (e.g., BaGeO3, BaGe2O5, C70). - Computational biases: Standard DFT calculations systematically underestimate band gaps, especially for heavy elements with strong SOC; even corrected/fitted approaches have residual errors and outliers. - Strict filters may exclude viable hosts: Excluding rare-earth-containing phases and elements without >50% zero-spin isotopes reduces nuclear-spin noise but may eliminate promising systems under certain conditions (e.g., isotopic purification or low concentration impurities). - Stability criterion approximation: The E Above Hull threshold (>0.2 eV/atom) is a practical proxy for ambient stability but may exclude metastable yet synthesizable phases or include phases sensitive to processing conditions. - Experimental practicality: Many candidates may be challenging to synthesize as large, defect-free, optically clear single crystals; air/moisture instability and contamination (e.g., oxygen defects) can affect magnetic character assessments and device performance. - Remaining breadth: Even after screening, 541 candidates remain too numerous for exhaustive experimental study, necessitating further application-specific narrowing.