Resonant states and nuclear dynamics in solid-state systems: the case of silicon-hydrogen bond dissociation

1. Problem Statement

The dissociation of silicon-hydrogen (Si-H) bonds by energetic carriers is a fundamental mechanism driving hot-carrier degradation in modern semiconductor devices (e.g., MOSFETs) and is central to phenomena like radiation damage and photocatalysis. While the macroscopic effects are well-documented, the microscopic mechanism has remained elusive.

Previous theoretical and experimental models relied on two primary hypotheses:

1. Direct Excitation Model: An electron is excited from the Si-H bonding state ($\sigma$) to the antibonding state ($\sigma^*$), creating a repulsive potential.
2. Inelastic Scattering Model: Multiple electrons sequentially excite vibrational modes via resonant scattering, eventually breaking the bond.

Key Challenges in Previous Approaches:

• Adiabatic Approximation Failure: Standard Density Functional Theory (DFT) uses the Born-Oppenheimer (adiabatic) approximation, which assumes electrons adjust instantaneously to nuclear motion. This leads to the hybridization of localized Si-H states with the bulk continuum, making it impossible to identify distinct bonding/antibonding states necessary for analyzing non-adiabatic dissociation.
• Discrepancy with Experiment: Previous models struggled to explain the "giant isotope effect" (H vs. D), the finite dissociation probability at low bias voltages (below the ~7 V threshold), and the temperature independence of the process.
• Lack of First-Principles Validation: No prior first-principles calculation could quantitatively reproduce experimental desorption yields or explain low-bias dissociation without invoking empirical parameters.

2. Methodology

The authors introduce a comprehensive non-adiabatic theoretical framework that explicitly treats the coupling between electronic and nuclear motions.

• State Partitioning (MLWF + PAO):
  • To overcome the delocalization issue in standard DFT, the authors employ a partitioning scheme. They use Maximally Localized Wannier Functions (MLWF) to transform extended Bloch states into localized orbitals.
  • For the antibonding states (which reside in the entangled conduction band continuum), they integrate the Pseudo-Atomic Orbital (PAO) method. This allows them to project and disentangle the specific Si-H antibonding orbital from the bulk continuum, effectively creating a localized "resonant state" description analogous to the Newns-Anderson Hamiltonian.
• Potential Energy Curves (PECs):
  • Using the partitioned states, they construct PECs for the ground state and excited states.
  • They compare three scenarios: (i) electron excitation ($\sigma \to \sigma^*$), (ii) electron injection into $\sigma^*$, and (iii) hole injection into $\sigma$.
• Nuclear Dynamics (MGR Model):
  • The nuclear motion of the hydrogen atom is treated quantum-mechanically using a wavepacket formulation.
  • They apply the Menzel-Gomer-Redhead (MGR) model. The process involves:
    1. A Franck-Condon transition to the excited (repulsive) potential.
    2. Propagation of the nuclear wavepacket on this repulsive surface for a residence time ($\tau_R$).
    3. Decay back to the ground state.
  • Dissociation occurs if the wavepacket gains sufficient kinetic energy to overcome the ground-state barrier during the residence time.
• Quantum Yield Calculation:
  • The dissociation probability is averaged over an exponential distribution of residence times to calculate the quantum yield, allowing direct comparison with experimental desorption yields.

3. Key Contributions

• Identification of Non-Adiabatic States: The paper successfully isolates the localized Si-H bonding and antibonding states within a bulk solid-state environment, resolving the hybridization problem that plagued previous DFT studies.
• Mechanism Clarification: The authors demonstrate that direct electron injection into the antibonding state (not excitation from the bonding state) is the primary driver of dissociation. The energy required for injection aligns with the experimental ~7 V threshold, whereas the excitation energy ($\sigma \to \sigma^*$) is calculated to be >12 eV, ruling out the traditional excitation model.
• Explanation of Low-Bias Dissociation: By accounting for the vibrational broadening of the resonant state (Franck-Condon factors), the model explains how dissociation can occur at bias voltages significantly lower than the antibonding state energy (down to 2 V), a phenomenon previously unexplained by single-electron models.
• Unified Framework: The approach unifies the explanation of STM surface experiments and oxide-stress device degradation experiments under a single physical mechanism.

4. Key Results

• Potential Energy Surfaces:
  • The ground-state PEC is anharmonic (Morse potential).
  • The excited-state PEC (for electron injection) is strongly repulsive and exponentially decaying.
  • The energy difference between the valence band maximum and the Si-H antibonding state is calculated at ~7.0 eV, matching the experimental threshold.
• Dissociation Dynamics:
  • Time Scale: Dissociation occurs on femtosecond timescales (residence times of ~6–8 fs for H).
  • Isotope Effect: The model reproduces the "giant isotope effect." Due to the mass difference, the deuterium wavepacket moves slower on the repulsive potential. For short residence times (typical of the process), the probability of deuterium gaining enough kinetic energy to dissociate is orders of magnitude lower than hydrogen. This explains the high $Q_H/Q_D$ ratio (50–200) observed in experiments.
  • Temperature Independence: Calculations show that up to room temperature, the quantum yield is dominated by the lowest vibrational state, explaining the observed temperature independence of the dissociation process.
• Quantitative Agreement:
  • STM: The calculated desorption yield spectrum matches experimental data, showing saturation above 7 V and an exponential drop below it.
  • Oxide-Stress: The model predicts trap generation probabilities in MOSFETs that align with DiMaria's experimental data, including the finite probability of trap creation at low biases (2 V).
  • Quantum Yield: The calculated quantum yield ($~10^{-5}$) and isotope ratios match experimental values when an electron lifetime of ~0.4–0.5 fs is assumed.

5. Significance

• Paradigm Shift: The paper refutes the long-held "multi-electron vibrational ladder" hypothesis for low-bias dissociation, proving that a single-electron injection mechanism is sufficient to explain all experimental observations, including low-bias effects.
• Device Reliability: By providing a first-principles basis for hot-carrier degradation, the framework offers a pathway to more accurate TCAD (Technology Computer-Aided Design) models for predicting device lifetime and reliability, moving away from empirical fitting parameters.
• General Applicability: The methodology (partitioning resonant states + non-adiabatic nuclear dynamics) is not limited to Si-H bonds. It is applicable to other carrier-induced bond-breaking phenomena, such as radiation damage, Auger-Meitner recombination effects in wide-bandgap semiconductors, and photocatalysis.

In conclusion, this work provides the first rigorous, first-principles explanation of Si-H bond dissociation driven by hot carriers, resolving decades of ambiguity regarding the mechanism, the role of isotopes, and the behavior at low bias voltages.