Electron paramagnetic resonance (EPR) can help identify point defects and dopants in β-Ga2O3 by measuring how unpaired electron spins respond to a magnetic field and interact with nearby nuclei. A computational study of EPR in β-Ga2O3 compares magnesium (Mg) and zinc (Zn) acceptors at the material’s two gallium sites: tetrahedral and octahedral. The calculated g-tensors and hyperfine interactions offer distinct signatures for the candidate sites, support the established assignment for Mg, and provide predictions for Zn.
Why EPR signatures matter for β-Ga2O3 defects
β-Ga2O3 is an ultra-wide-band-gap semiconductor under consideration for transistor and transparent-conducting applications. As with other semiconductors, understanding its defects and dopants is important for interpreting and controlling its material properties. EPR is particularly useful because a defect that has an unpaired electron can produce measurable magnetic signatures influenced by its local atomic environment.
Two properties are central to this analysis. The g-tensor describes the directional dependence of the electron’s magnetic response. The hyperfine tensor describes the interaction between the electron spin and the nuclear spins of nearby atoms. Because these properties reflect the defect’s surroundings and spin distribution, they can help distinguish otherwise plausible defect configurations.
How the calculations treat spin localization
The calculations use the Gauge Including Projector Augmented Wave (GIPAW) method, based on density functional perturbation theory, to determine the linear magnetic response and calculate g-tensors. Hyperfine interaction has an isotropic Fermi contact component, which depends on spin density at a nucleus, and an anisotropic dipole component.
Reliable hyperfine predictions depend on how well the calculation represents the localization of the defect’s unpaired spin. For acceptors in β-Ga2O3, the spin is associated mainly with oxygen p-like states. Semilocal density-functional methods can spread this spin over neighboring atoms, affecting calculated hyperfine values. The study therefore uses structures relaxed with DFT+U, which helps localize the spin on a nearby oxygen atom. The g-tensors, however, are calculated at the GGA level because the GIPAW implementation used does not support g-tensor calculations with DFT+U or hybrid-functional electronic structures.
Mg in β-Ga2O3: evidence for the octahedral site
Both tetrahedral and octahedral substitutional sites were evaluated for Mg. The calculated EPR properties differ between these configurations, but the octahedral-site model, MgGa2, agrees with the reported experimental center. This supports the assignment of Mg to an octahedral gallium site, consistent with earlier calculations indicating that this site has lower energy for Mg.
The distinction is not based on a single tensor value. In the octahedral model, the calculated g-tensor follows the same ordering from smaller to larger values as the experimental measurements. The hyperfine interactions also reflect the local arrangement of neighboring gallium atoms: the octahedral configuration has two inequivalent Ga neighbors, whereas the tetrahedral model has two equivalent octahedral Ga neighbors. This difference in hyperfine behavior provides a more telling site-specific signature than small differences in g-values alone.
The calculated hyperfine magnitudes are not an exact match to experiment. For the octahedral Mg model, the study reports an underestimate of about 23% for one Ga interaction and an overestimate of about 10% for the other. The predicted anisotropy is small, and neither the calculation nor the compared hybrid-functional result reproduces the experimental ordering of the hyperfine tensor components. The overall agreement supports the site assignment while also showing that individual tensor details remain challenging to predict.
Zn results point to a different preferred site
The calculations predict a different site preference for Zn. In the neutral, EPR-active state considered, the tetrahedral Zn configuration, ZnGa1, is lower in energy than the octahedral configuration by 0.16 eV. Its predicted signature includes hyperfine interaction with two equivalent Ga atoms and a g-tensor whose maximum value is along the crystal c direction.
These Zn results are predictions rather than a confirmation against an established experimental assignment. The calculated signals are similar in some respects to those for Mg, and the Zn tetrahedral g-tensor resembles one previously associated with a gallium-vacancy-related center. This similarity could complicate interpretation from EPR parameters alone. The study suggests looking for the predicted Zn-related center in Zn-doped material without irradiation, which was associated with producing vacancies in earlier gallium-vacancy investigations.
The calculations also provide hyperfine parameters for the magnetic isotopes 67Zn and 25Mg, both of which have nuclear spin I = 5/2. Their natural abundances are low, so the corresponding splittings may be difficult to detect. For Zn, the predicted isotope hyperfine values are comparable to those for Ga and more anisotropic.
Interpreting the results and their limits
The work illustrates why EPR defect identification benefits from considering a set of signatures rather than relying on a single measurement direction or tensor component. The principal axes of the calculated tensors are not always aligned with the crystallographic axes. In particular, hyperfine tensor axes tend to follow local bond directions. Comparing measurements with calculations therefore requires attention to tensor principal values and orientations, as well as the local atomic environment.
The study also cautions against treating calculated acceptor transition levels as settled. The reported values vary between GGA and hybrid-functional approaches and are affected by methodological choices. These levels have not been established experimentally in the work described. The EPR calculations focus on the neutral state, which is the state considered active for EPR.
Conclusion
For Mg-doped β-Ga2O3, the calculated EPR properties support an octahedral gallium-site assignment, with unequal neighboring Ga hyperfine interactions helping distinguish it from the tetrahedral alternative. For Zn, the calculations predict a preference for the tetrahedral site and identify a candidate EPR signature for future investigation. For researchers interpreting defect measurements, the practical lesson is to combine g-tensor and hyperfine information with tensor orientation, local bonding, and the limits of the computational method. These results provide a framework for testing site assignments, not a substitute for experimental confirmation—especially in the predicted Zn case.