Abstract: The electron paramagnetic resonance (EPR) linewidths of B acceptors in Si are found to reduce dramatically in isotopically purified $^{28}\text{Si}$ single crystals. Moreover, extremely narrow substructures in the EPR spectra are visible corresponding to either an enhancement or a reduction in the absorbed microwave on resonance. The origin of the substructures is attributed to a combination of simultaneous double excitation and spin relaxation in the four level spin system of the acceptors. A spin population model is developed which qualitatively describes the experimental results.
Introduction
Boron is the primary shallow acceptor in silicon, yet its physical properties are remarkably complex due to the four-fold degeneracy of its ground state. This degeneracy makes the boron acceptor highly sensitive to local internal strains. In natural silicon ($^{nat}\text{Si}$), the random distribution of isotopes ($^{28}\text{Si}$, $^{29}\text{Si}$, and $^{30}\text{Si}$) creates "isotope noise"—local fluctuations in the crystal potential that inhomogeneously broaden the energy levels. This broadening has historically obscured fine details in EPR spectra, including mysterious sharp "substructures" observed at the center of resonance lines. This study utilizes isotopically purified $^{28}\text{Si}$ to eliminate isotope-induced broadening and elucidate the origin of these substructures.
Methods
The researchers compared p-type silicon crystals of two isotopic compositions: natural silicon ($^{nat}\text{Si:B}$) and isotopically purified silicon ($^{28}\text{Si:B}$), enriched to 99.98% $^{28}\text{Si}$. EPR measurements were conducted at cryogenic temperatures (1.8 K to 3.2 K) using X-band microwave frequencies (~9.4 GHz) with the magnetic field $H$ parallel to the [100] crystal axis.
To interpret the experimental findings, an incoherent spin population model was developed. This model employs rate equations to track the populations of the four energy levels ($|1\rangle$ to $|4\rangle$) under the competing influences of microwave-induced transitions and thermal spin relaxation processes.
Isotopic Purification and Linewidth Narrowing
The transition from $^{nat}\text{Si}$ to $^{28}\text{Si}$ resulted in a dramatic reduction of EPR linewidths. In natural silicon, the random isotopic environment smears the resonance, while in the purified crystal, the "isotope noise" is virtually eliminated. This reveals sharp features at the resonance centers, which correspond to boron sites experiencing minimal local strain.
Substructure Dynamics and Temperature Effects
The high resolution afforded by $^{28}\text{Si}$ allows for a detailed study of the substructures. These features arise from "double excitations" in the four-level system. At zero-strain sites, multiple transitions (e.g., $|1\rangle \leftrightarrow |2\rangle$ and $|3\rangle \leftrightarrow |4\rangle$) occur at the same frequency, leading to interference in the absorption signal.
Whether a substructure appears as a "dip" (reduced absorption) or a "peak" (enhanced absorption) depends on the balance between the microwave excitation rate ($w$) and the spin-lattice relaxation rates ($\gamma$). The experimental results show that the $\Delta m=1$ transition consistently exhibits a negative substructure (a dip). In contrast, the $\Delta m=2$ transition is highly temperature-sensitive, flipping from a positive peak at 3.2 K to a negative dip at 1.8 K as relaxation rates change.
Discussion and Conclusion
The developed spin population model qualitatively reproduces the observed sign changes in the substructures, confirming that they are intrinsic to the dynamics of the four-level spin system rather than an artifact of isotopic composition. While the model captures the essential physics, it remains qualitative and does not perfectly match the observed microwave power saturation characteristics.
Future work will involve more rigorous quantitative modeling using density matrix formalisms or Redfield relaxation theory. Understanding these acceptor states is crucial for potential applications in quantum information processing, where the long coherence times of isotopically purified silicon offer a promising platform for spin-based qubits.
Summary updated