Mode selectivity in electron promoted vibrational relaxation of chemisorbed hydrogen on molybdenum and tungsten surfaces

Imagine a metal surface, like a sheet of Mothball (Molybdenum) or Tungsten, as a giant, trampoline-like dance floor. Now, imagine tiny hydrogen atoms landing on this floor. When they land, they don't just sit still; they start bouncing, wiggling, and vibrating, much like a dancer trying to find their rhythm.

This paper is about how quickly these dancing hydrogen atoms lose their energy and stop vibrating.

The Problem: The "Silent" Energy Leak

When a hydrogen atom lands on a metal surface, it has a lot of extra energy. It needs to get rid of this energy to settle down. Usually, it does this by shaking hands with the electrons in the metal. Think of the metal's electrons as a crowd of people in a dark room. When the hydrogen atom vibrates, it bumps into these electrons, transferring its energy to them. The electrons get excited (like a crowd cheering), and the hydrogen atom slows down.

The scientists wanted to know: How fast does this energy leak happen? And does the speed change depending on how crowded the dance floor is?

The Two Types of "Dance Moves" (Vibrations)

The hydrogen atoms can vibrate in different ways, and the paper found that these different moves behave very differently:

The "Fano" Moves (The Chaotic Dancers):
Some vibrations look messy and asymmetric in experiments. The scientists call these "Fano line shapes."
- The Analogy: Imagine a dancer spinning wildly and bumping into the crowd in a chaotic, unpredictable way.
- The Finding: For these moves, the computer simulations matched the real-world experiments perfectly. This confirms that the energy is leaking out almost entirely because the hydrogen is bumping into the metal's electrons. It's a direct, efficient energy transfer.
The "Lorentzian" Moves (The Smooth Dancers):
Other vibrations look smooth and symmetrical.
- The Analogy: Imagine a dancer gliding smoothly across the floor.
- The Finding: Here, the computer said the dancer should slow down slowly (because electron bumps are the only thing we calculated). But in the real world, the dancer stopped much faster.
- The Lesson: This means there are other things slowing them down that the computer didn't count! Maybe the dancers are bumping into each other (since there are many of them), or maybe the floor itself is shaking in a way the model didn't catch. The "electron crowd" isn't the only thing stealing their energy.

The "Crowded Dance Floor" Effect (Coverage)

This is the most surprising part of the paper. The scientists tested what happens when the dance floor is empty versus when it is packed with hydrogen atoms.

The Expectation: You might think that if you pack the floor with more dancers, they would bump into the electron crowd more often, making them lose energy faster.
The Reality: The opposite happened! When the floor was packed (high coverage), the hydrogen atoms held onto their energy longer. They vibrated for a longer time before stopping.
The Analogy: Imagine a crowded mosh pit. When it's empty, a single person can run and bump into the walls (electrons) easily. But when the pit is packed shoulder-to-shoulder, everyone is moving together in a synchronized wave. The individual "bumps" against the walls cancel each other out or become less effective. The crowd moves as one unit, making it harder for the energy to leak out into the metal.

Why This Matters

This discovery is a big deal for two reasons:

Fixing the Math: Many computer models used to simulate these surfaces assume the metal is always "pristine" (empty). This paper shows that if you have a crowded surface (like in a fusion reactor or a chemical factory), those old models are overestimating how fast energy is lost. They think the hydrogen cools down instantly, but in reality, it stays hot and energetic for longer.
Real-World Applications: This helps us understand how hydrogen behaves in fusion reactors (where hydrogen fuel is recycled) and in making fertilizers (the Haber-Bosch process). If we can predict exactly how long hydrogen stays "hot" on a surface, we can design better reactors and more efficient chemical plants.

The Bottom Line

The paper tells us that crowds change the rules. When hydrogen atoms are alone on a metal surface, they lose energy quickly by bumping into electrons. But when they are in a dense crowd, they stick together, move in sync, and hold onto their energy much longer. To understand how these systems work, we can't just look at the individual atoms; we have to look at the whole crowd.

Here is a detailed technical summary of the paper "Mode selectivity in electron promoted vibrational relaxation of chemisorbed hydrogen on molybdenum and tungsten surfaces."

1. Problem Statement

The adsorption of hydrogen on metal surfaces is a critical process in catalysis (e.g., Haber-Bosch, Fischer-Tropsch) and fusion reactor fuel recycling. A fundamental challenge in this field is understanding how adsorbed hydrogen atoms dissipate excess energy released during bond dissociation.

The Core Issue: While electron-hole pair (EHP) excitation is known to be the dominant energy dissipation mechanism on pristine metal surfaces, isolating its specific contribution to vibrational linewidths is difficult. Experimental spectra often contain contributions from inhomogeneous broadening, pure dephasing, and adsorbate-adsorbate interactions.
The Gap: Previous theoretical studies often relied on the Local Density Friction Approximation (LDFA), which assumes isotropic friction based on the pristine metal's electron density. There is a lack of first-principles, mode-resolved characterizations of vibrational linewidths for full monolayer coverages of hydrogen on Molybdenum (Mo) and Tungsten (W), particularly regarding the distinction between Fano-type (asymmetric) and Lorentzian (symmetric) line shapes.

2. Methodology

The authors employed first-order time-dependent perturbation theory (TDPT) within the framework of electronic friction, implemented using the FHI-aims software package.

Computational Setup:
- Method: Density Functional Theory (DFT) with the PBE exchange-correlation functional and scalar-relativistic corrections (ZORA).
- Models: Periodic slabs (6 layers) representing Mo(100), Mo(110), W(100), and W(110) surfaces.
- Coverage: Simulations covered full monolayer ( $\Theta = 1$ ML) and lower coverages ( $\Theta = 1/4$ to $1/36$ ML) using supercells to investigate coverage dependence.
- Isotopes: Calculations were performed for both Hydrogen (H) and Deuterium (D) to analyze kinetic isotope effects.
Theoretical Framework:
- Electron-Phonon Coupling: The frequency-dependent relaxation rate tensor $\Lambda(\hbar\omega)$ was calculated using TDPT, considering interband electronic transitions induced by atomic displacement.
- Linewidth Calculation: Vibrational linewidths ( $\gamma$ ) were derived by projecting the friction tensor onto mass-weighted normal mode displacement vectors.
- Mode Analysis: The study distinguished between:
  - Coherent modes ( $\gamma_0$ ): All adsorbates oscillating in phase ( $q=0$ ).
  - Q-averaged modes ( $\bar{\gamma}$ ): An ensemble average representing instantaneous dephasing, calculated by summing linewidths over folded $q$ -points in supercells. This serves as an upper bound for the linewidth.

3. Key Contributions

First-Principles Fano Characterization: The study provides the first ab initio characterization of experimental IR and EELS linewidths for full monolayer H/Mo and H/W surfaces, explicitly linking Fano line shapes to EHP-mediated relaxation.
Mode Selectivity: It demonstrates that the relaxation mechanism is highly mode-dependent. Modes exhibiting Fano line shapes are well-described by EHP excitation, whereas Lorentzian modes suggest significant contributions from other mechanisms (e.g., adsorbate-adsorbate interactions).
Coverage Dependence: The paper quantifies how electronic friction and vibrational linewidths change drastically with hydrogen coverage, challenging the validity of constant-friction models (like LDFA) at high coverages.
Anisotropy: It reveals that the electronic friction tensor is highly anisotropic at high coverages, particularly for Mo(110), contradicting the isotropic assumptions often used in molecular dynamics simulations.

4. Key Results

A. Linewidths and Line Shapes

Fano Modes (Asymmetric): For modes exhibiting Fano line shapes (e.g., asymmetric stretch on Mo(100)/W(100), rocking on Mo(110)), the calculated q-averaged linewidths ( $\bar{\gamma}$ ) show excellent agreement with experimental Full Width at Half Maximum (FWHM) values. This confirms that EHP excitation is the dominant relaxation channel for these specific modes.
- Example: The calculated $\bar{\gamma}$ for the asymmetric stretch on H/W(100) is 15 cm $^{-1}$ , matching the experimental value of 22 cm $^{-1}$ within a factor of 1.5.
Lorentzian Modes (Symmetric): For modes with symmetric Lorentzian profiles (e.g., symmetric stretch on Mo(100) and W(100)), the calculated linewidths are significantly smaller than experimental values (e.g., predicted ~29 cm $^{-1}$ vs. experimental 65 cm $^{-1}$ for Mo(100) symmetric stretch). This indicates that EHP excitation alone cannot explain the broadening; other mechanisms like adsorbate-adsorbate coupling or phonon-phonon interactions are dominant.

B. Isotope Effect

The study calculated the Kinetic Isotope Effect (KIE = $\gamma_H / \gamma_D$ ).
For most modes, the KIE is close to the theoretical expectation of $\approx 2$ (inverse mass scaling), confirming weak coupling limits.
Exception: The asymmetric stretch on W(100) showed a negligible isotope effect experimentally (KIE $\approx$ 0.9), which the authors attribute to strong non-adiabatic effects that reduce mass dependence, a nuance their standard TDPT approach struggles to fully capture without strong coupling corrections.

C. Coverage Dependence

Inverse Relationship: As hydrogen coverage decreases, the vibrational linewidths (and thus electronic friction) increase significantly (up to a factor of 8).
Mechanism: At high coverage ( $\Theta = 1$ ML), the adsorbate layer modifies the surface electronic structure, reducing the density of available electronic states for EHP excitation.
Anisotropy:
- Mo(110): Becomes strongly anisotropic at full coverage (friction along the surface normal is distinct from parallel directions).
- W(110): Remains relatively isotropic even at high coverage, though friction magnitude decreases.
Friction Tensor: The off-diagonal elements of the friction tensor (e.g., $\Lambda_{yz}$ ) become significant at low coverages, suggesting that simple isotropic friction models are insufficient for modeling diffusion or scattering on these surfaces.

D. Comparison with LDFA

The study compares their TDPT results with the Local Density Friction Approximation (LDFA).
At low coverage, TDPT and LDFA yield similar results.
At high coverage, LDFA overestimates the electronic friction by a factor of ~20 for W(110) compared to TDPT. This suggests that previous molecular dynamics simulations using LDFA on dense hydrogen layers may have overestimated non-adiabatic energy dissipation.

5. Significance and Implications

Refining Dissipation Models: The findings necessitate a shift from isotropic, coverage-independent friction models (LDFA) to coverage-dependent, anisotropic friction tensors for accurate modeling of hydrogen dynamics on metal surfaces.
Fusion and Catalysis: Accurate modeling of hydrogen sticking coefficients and energy dissipation is vital for fusion reactor fuel recycling and heterogeneous catalysis. The paper suggests that at high coverages, adsorbate-adsorbate interactions may play a larger role in energy dissipation than previously assumed.
Spectroscopic Interpretation: The work provides a robust theoretical basis for interpreting complex vibrational spectra (Fano vs. Lorentzian), allowing researchers to distinguish between electronic and non-electronic relaxation pathways in surface science experiments.
Methodological Validation: It validates the use of TDPT with supercell folding to access finite-q phonon modes, offering a practical route to study dephasing and linewidths without explicit q-dependent calculations.