Identifying candidate hosts for quantum defects via data mining

Quantum information science (QIS) aims to develop systems for storing, processing, and transmitting information using quantum states. Atom-like defects in solid-state hosts are promising platforms for QIS, particularly in quantum networks. Several systems like vacancy centers in diamond and silicon carbide have been studied extensively for applications in quantum networks, magnetometry, and nanoscale sensing. Transition-metal and rare-earth ion impurities are another class of defects, but exploring single defects as qubits is a more recent development. To achieve long spin coherence and high-efficiency optical transitions needed for QIS, host substrates must be highly pure, intrinsically diamagnetic, and possess a large band gap. The host should also be free of paramagnetic impurities and have nuclei with zero magnetic moments. Materials with controllable surfaces and known methods of epitaxial thin-film production are preferred for device integration. Most known quantum information materials have been discovered serendipitously, hence the need for a rational search of inorganic materials to find candidates with superior properties. This study uses a computational search to narrow the field of potential candidates by postulating ideal properties for host materials and filtering materials based on those properties.

The introduction extensively reviews existing research on atom-like defects in solid-state hosts for quantum information systems. It highlights successful examples such as vacancy centers in diamond and silicon carbide, and transition-metal and rare-earth ion impurities. The limitations of current approaches, mainly relying on serendipitous discoveries, are emphasized, leading to the motivation for a systematic search method.

The researchers employed a four-stage screening process to identify potential host materials for quantum defects. Stage 1 used the Materials Project's API to filter the database, selecting phases crystallizing in nonpolar space groups, experimentally derived from the ICSD, containing elements with >50% natural abundance of zero nuclear spin isotopes, and calculated to be nonmagnetic. Stage 2 involved manual removal of phases containing uranium, thorium, cadmium, mercury, noble gases, and rare-earth elements due to their radioactivity, toxicity, or difficulty in obtaining pure starting materials. Stage 3 utilized calculated band gaps from the Materials Project, JARVIS-DFT, and AFLOW databases, along with literature values. Phases with a Materials Project band gap of ≥2.0 eV were considered suitable; those with <0.5 eV were removed. Literature values were incorporated to account for computational limitations in band gap prediction. Stage 4 confirmed the intrinsic diamagnetic character and phase stability using electron-counting rules and literature reports. Phases with a Materials Project-calculated E Above Hull value >0.2 eV/atom were excluded. The reliability of calculated band gap values from different databases was assessed and discussed, noting systematic underestimation by standard DFT methods. Additional considerations included the presence of experimental crystal structures in the ICSD database and single-crystal growth feasibility. The study used different DFT functionals (PBE-GGA, OptB88vdW, mBJ) and machine learning methods (AFLOW-ML) to obtain band gap values, comparing and contrasting their predictive ability in relation to experimental values.

From 125,223 inorganic compounds in the Materials Project database, the study identified 541 (0.43%) potentially viable host substrates for quantum defects. The most significant reduction in candidates occurred in Stage 1 (97.3%). Of the 541 phases, 16 are unary, 74 binary, 322 ternary, and 129 quaternary or higher order. Many are oxides and chalcogenides, suggesting ease of fabrication and study. The band gap of 521 candidates could be confirmed or assumed to be ≥1.1 eV (similar to Silicon). Nickel and Iron were not found in any of the identified phases. A significant portion of Os-containing phases were transition-metal carbonyl cluster compounds. The distribution of E Above Hull values (a measure of thermodynamic stability) was analyzed, and those greater than 0.2 eV/atom were excluded. The accuracy of the band gap calculations from various sources was examined and showed that DFT calculations generally underestimated band gaps, with different methods exhibiting varying degrees of accuracy, highlighting the importance of considering multiple calculation approaches.

The study presents a comprehensive catalog of potentially viable host materials for atom-like defect-based quantum information systems. The combination of automated data mining and manual screening ensures the inclusion of necessary electronic, magnetic, and optical properties and stability. The large number of candidates (541) remains challenging for extensive experimental studies, but the identification of 90 relatively basic (unary and binary) systems simplifies the search for practical materials. Future computational studies to refine band structure calculations and predict application-specific properties, such as surface stability and excited-state lifetimes, will further refine the list of promising candidates. This systematic screening process can be applied to other materials systems and applications.

This research provides a valuable resource for researchers seeking suitable host materials for quantum defects. The systematic data-mining approach significantly narrowed the vast space of inorganic materials, identifying 541 promising candidates. Future work should focus on experimental validation of the identified materials and further refinement of the selection criteria based on specific applications and more advanced computational modeling.

The study acknowledges that the number of potentially viable phases may be significantly larger than the identified 541, as the analysis is limited to currently known materials. Some materials may have been excluded due to a lack of experimental crystal structures in the ICSD, which is a limitation of the databases utilized. Also, the accuracy of band gap calculations is subject to limitations inherent in DFT methodologies. Finally, while single crystal synthesis was considered, it was not a primary exclusionary criterion.