[Paper Review] Near-Zero-Field Spin-Dependent Recombination Current and Electrically Detected Magnetic Resonance from the Si/SiO$_2$ interface

Nicholas J. Harmon, James P. Ashton+2 more|arXiv (Cornell University)|Aug 18, 2020

Advancements in Semiconductor Devices and Circuit Design4 citations

TL;DR

This paper develops a semiclassical theory for spin-dependent recombination at the Si/SiO₂ interface, explaining near-zero-field magnetoresistance (NZFMR) and electrically detected magnetic resonance (EDMR) in MOSFETs. It shows that hyperfine interactions via nuclear spins enable spin mixing, and that a semiclassical model—unlike purely quantum models—successfully explains both NZFMR and EDMR responses simultaneously, resolving prior discrepancies in parameter fitting.

ABSTRACT

Dielectric interfaces critical for metal-oxide-semiconductor (MOS) electronic devices, such as the Si/SiO$_2$ MOS field effect transistor (MOSFET), possess trap states that can be visualized with electrically-detected spin resonance techniques, however the interpretation of such measurements has been hampered by the lack of a general theory of the phenomena. This article presents such a theory for two electrical spin-resonance techniques, electrically detected magnetic resonance (EDMR) and the recently observed near-zero field magnetoresistance (NZFMR), by generalizing Shockley Read Hall trap-assisted recombination current calculations via stochastic Liouville equations. Spin mixing at this dielectric interface occurs via the hyperfine interaction, which we show can be treated either quantum mechanically or semiclassically, yielding distinctive differences in the current across the interface. By analyzing the bias dependence of NZFMR and EDMR, we find that the recombination in a Si/SiO$_2$ MOSFET is well understood within a semiclassical approach.

Motivation & Objective

• To address the lack of a general theoretical framework for interpreting EDMR and NZFMR in Si/SiO₂ MOSFETs.
• To resolve inconsistencies between prior quantum models and experimental data for both NZFMR and EDMR.
• To explain why quantum models fail to simultaneously describe both EDMR and NZFMR line shapes.
• To establish a unified theoretical model for spin-dependent recombination at oxide-semiconductor interfaces.
• To demonstrate that a semiclassical approximation of hyperfine interactions better captures the experimental behavior than a fully quantum treatment.

Proposed method

• Generalizes Shockley-Read-Hall recombination theory using stochastic Liouville equations to model spin dynamics.
• Models spin mixing via hyperfine interactions between electron and nuclear spins at the Si/SiO₂ interface.
• Compares quantum and semiclassical treatments of hyperfine coupling, showing distinct differences in predicted current responses.
• Uses bias-dependent measurements of NZFMR and EDMR to constrain model parameters such as capture and dissociation rates (kS and kD).
• Performs nonlinear least squares fitting of experimental line shapes to extract hyperfine coupling constants and abundances.
• Analyzes the forward bias dependence of EDMR and NZFMR amplitudes to validate the model against dc-IV recombination current.

Experimental results

Research questions

• RQ1 Why do quantum models of hyperfine interaction fail to simultaneously explain both NZFMR and EDMR responses in Si/SiO₂ interfaces?
• RQ2 What is the correct theoretical framework for describing spin-dependent recombination currents under near-zero magnetic fields?
• RQ3 How do the rates of electron capture (kS) and dissociation (kD) influence the relative amplitudes of EDMR and NZFMR?
• RQ4 Can a semiclassical approximation of hyperfine coupling accurately describe both EDMR and NZFMR phenomena?
• RQ5 What explains the discrepancy between previous quantum fits to NZFMR and the negligible EDMR predicted by those same fits?

Key findings

• A semiclassical model of hyperfine interaction provides a consistent explanation for both NZFMR and EDMR, unlike purely quantum models.
• The model shows that smaller values of kS and kD—consistent with experimental trends—enhance EDMR relative to NZFMR, resolving prior inconsistencies.
• The ratio of experimental to theoretical EDMR amplitude increases with forward bias, indicating a strong correlation with the dc-IV recombination current.
• The hyperfine interaction at both the electron and defect spins is the most plausible explanation for the observed line shape features in both NZFMR and EDMR.
• The model attributes discrepancies in prior work to the assumption of constant kS and kD, which are instead better described as distributed across a range of values.
• The theory successfully explains the bias dependence of spin-dependent recombination, supporting its applicability to spin-dependent trap-assisted transport (SDTAT) in real devices.