Abstract
We present experiments to systematically study the time constants of spin-dependent recombination processes in semiconductors using pulsed electrically detected magnetic resonance (EDMR). The combination of time-programmed optical excitation and pulsed spin manipulation allows us to directly measure the recombination time constants of electrons via localized spin pairs and the time constant of spin-pair formation as a function of the optical excitation intensity. Using electron nuclear double resonance, we show that the time constant of spin-pair formation is determined by an electron capture process. Based on these time constants we devise a set of rate equations to calculate the current transient after a resonant microwave pulse and compare the results with experimental data. Finally, we critically discuss the effects of different boxcar integration time intervals typically used to analyze pulsed EDMR experiments on the determination of the time constants. The experiments are performed on phosphorus-doped silicon, where EDMR via spin pairs formed by phosphorus donors and Si/SiO${}_{2}$ interface dangling bond defects is detected.
Introduction
In semiconductor devices, electronic performance is often limited by the recombination of charge carriers at material defects, particularly at interfaces like the Si/SiO$_2$ boundary. This recombination process is inherently spin-dependent due to the Pauli Exclusion Principle: two electrons can only occupy the same defect state if their spins are antiparallel. Parallel spin pairs create a "bottleneck," remaining in a metastable state until a spin-flip occurs.
While Electrically Detected Magnetic Resonance (EDMR) has long been used to study these defects, traditional continuous-wave measurements operate in a steady state where formation and recombination processes are conflated. This research addresses the need to decouple these dynamics to independently measure the time constants for pair formation, antiparallel recombination, and parallel recombination.
Experimental Methodology
The researchers utilized a 22 nm-thick phosphorus-doped silicon layer on a Silicon-on-SOI substrate. The experimental core is a specialized Pulsed EDMR setup enhanced by two key innovations:
- Time-Programmed Optical Excitation: A red LED was modulated with microsecond precision to control carrier generation, allowing the system to be "cleared" and then "filled" at specific intervals to observe transient dynamics.
- Pulsed ENDOR (Electron Nuclear Double Resonance): By probing the nuclear spin of phosphorus atoms, the team directly measured the rate at which ionized donors capture conduction band electrons.
As shown in the figure above, the process involves (i) the initial dark state, (ii) carrier generation and capture under illumination, (iii) rapid recombination of antiparallel spin pairs ($\tau_{ap}$), and (iv) slow recombination of parallel pairs ($\tau_p$) limited by spin relaxation.
Results and Analysis
The study successfully decoupled the various time constants of the $^{31}$P-$P_{b0}$ spin-pair system. A critical finding was that the time constant for spin-pair formation is physically identical to the electron capture time constant ($\tau_{ec}$), which was measured at approximately 40 $\mu$s under 20 mW/cm² illumination.
The experimental data (shown above) reveals two distinct decay regimes in the charge transient $\Delta Q$ as a function of the waiting time $T$:
- Antiparallel Recombination ($\tau_{ap}$): Measured at $\approx 15.5$ $\mu$s.
- Parallel Recombination ($\tau_p$): Measured at $\approx 1.2$ ms.
The recombination of antiparallel pairs is roughly 100 times faster than that of parallel pairs. Furthermore, the decays followed a stretched exponential behavior (exponent $n \approx 0.5$), suggesting a distribution of donor-defect distances and varying tunneling rates.
Discussion: The "Boxcar Trap"
A significant methodological insight from this work concerns the use of boxcar integration in EDMR analysis. Typically, researchers integrate the current signal over a specific time window. This study demonstrates that if a system possesses a broad distribution of recombination rates, the choice of this integration window can bias the resulting time constant. This implies that many previously reported values in EDMR literature may be artifacts of measurement settings rather than intrinsic material properties.
Conclusion
By combining pulsed spin manipulation with time-resolved illumination, this research provides a rigorous framework for untangling the complex dynamics of spin-dependent recombination. While the study focuses on the $^{31}$P-$P_{b0}$ system in silicon, the techniques are broadly applicable to other semiconductors, including organic materials and different interface defects. Future work will aim to directly measure hole capture dynamics and map the spatial distribution of defects more precisely.
Summary updated