Platinum Adatom Diffusion on Pt(100) Surface

Introduction

Building on my Cu adatom diffusion tutorial, this post explores platinum adatom movement on Pt(100) surfaces. While the simulation setup is similar, platinum’s higher atomic mass and stronger metallic bonding create notably different diffusion behavior—valuable insights for both materials science and machine learning applications.

What you’ll learn:

• How atomic mass affects surface diffusion dynamics
• Setting up LAMMPS for precious metal simulations
• Comparing diffusion behavior across different elements
• Generating element-specific data for ML model training

New to LAMMPS surface simulations? Start with my Cu adatom diffusion tutorial for detailed background on simulation setup and analysis techniques.

Why Platinum Matters

Platinum plays a crucial role in modern technology, particularly in catalysis. Understanding how individual platinum atoms move on surfaces is essential for:

• Catalyst design: Most platinum catalysts work through surface reactions
• Fuel cells: Platinum electrodes rely on surface atom mobility
• Materials engineering: Platinum’s stability makes it ideal for harsh environments
• Machine learning: Element-specific behavior helps train more accurate models

From a simulation perspective, platinum offers interesting contrasts to copper:

• Heavier atoms: ~3x copper’s mass affects diffusion kinetics
• Stronger bonding: Higher melting point indicates stronger surface interactions
• Different lattice: Slightly larger lattice constant (3.96 Å vs 3.614 Å)
• Catalytic relevance: Direct applications in real-world chemical processes

Simulation Overview

This simulation models a single platinum adatom diffusing across a Pt(100) surface using molecular dynamics. The approach builds directly on the copper simulation framework but uses platinum-specific parameters.

Key Differences from Copper

Simulation Parameters

The simulation uses identical conditions to the copper case for direct comparison:

• System size: 8×8×6 unit cells
• Temperature: 600 K surface equilibrium, 850 K bulk heating
• Duration: 500 ps simulation time
• Timestep: 5 fs for numerical stability
• Potential: EAM potential from Zhou et al. (2004)

This parameter consistency enables meaningful comparisons between elements—exactly what’s needed for training element-aware ML models.

LAMMPS Input Script

The platinum simulation uses the same framework as the copper case, with element-specific modifications. Here’s the complete LAMMPS input script:

### Original Created by Eric N. Hahn  ###
### [email protected] ###

### Modifications by Hunter Heidenreich, CSE lab (Harvard, 2023)
### [email protected]
### 2023-09-01

### Simulating adatoms ###
### Version 0.2 ###


units metal
dimension 3
boundary p p s
atom_style atomic

lattice fcc 3.96
variable cubel equal 4
variable fixer1 equal "v_cubel+2"
variable fixer2 equal "v_cubel+1.49"
region  box block -${cubel} ${cubel} -${cubel} ${cubel} -${fixer1} 1 units lattice
region cbox block -${cubel} ${cubel} -${cubel} ${cubel} -${fixer1} 0 units lattice
create_box 1 box
create_atoms 1 region cbox
create_atoms 1 single -0.5 0 0.5 units lattice
region hold block INF INF INF INF -${fixer1} -${fixer2} units lattice
region temp block INF INF INF INF -${fixer2} -${cubel} units lattice
group hold region hold
group temp region temp

pair_style eam/alloy
pair_coeff * * Pt_Zhou04.eam.alloy Pt

timestep        0.005
compute         new all temp
velocity        temp create 600 12345
fix heater temp temp/rescale 1 850 850 5 1
fix nve all nve
fix freeze hold setforce 0 0 0

variable e     equal pe
variable k     equal ke
variable t     equal etotal
variable T     equal temp
fix energy all ave/time 1 50 50 v_k v_e v_t v_T file energy_avg.txt

minimize 1.0e-4 1.0e-6 1000 10000

dump eve all custom 5 dump.lammpstrj id type xu yu zu   # fx fy fz  # uncomment for forces
dump_modify eve sort id

thermo 50
run 100000  # 100_000 * 5 fs = 500 ps

Key Differences from Copper Script

The main changes from the copper simulation are:

1. Lattice constant: lattice fcc 3.96 (vs 3.614 for Cu)
2. Potential file: Pt_Zhou04.eam.alloy (vs Cu01.eam.alloy)
3. Element specification: Pt in the pair_coeff line

These simple changes capture the essential physics differences between elements while maintaining the same simulation protocol—ideal for generating comparative datasets for ML training.

Potential Details

The simulation uses the Zhou et al. EAM potential for platinum, available from the NIST repository. This potential accurately reproduces:

• Bulk properties (lattice constant, elastic constants)
• Surface energies and reconstruction patterns
• Defect formation energies
• Thermal expansion behavior

The EAM formalism captures metallic bonding better than simple pair potentials, making it essential for realistic surface dynamics.

Expected Behavior vs. Copper

Based on the physical differences, we expect platinum to show:

Slower Diffusion

Heavier atoms move more slowly at the same temperature. Platinum’s ~3x greater mass should reduce diffusion rates compared to copper.

Higher Energy Barriers

Stronger metallic bonding in platinum creates deeper potential wells, requiring more thermal energy for diffusion hops.

Different Pathways

The larger lattice constant changes the energy landscape, potentially favoring different diffusion mechanisms.

Temperature Sensitivity

The higher melting point suggests diffusion will be more sensitive to temperature changes—valuable for training temperature-dependent ML models.

Machine Learning Applications

This simulation generates platinum-specific training data for several ML applications:

Element-Aware Models

Comparing Cu and Pt trajectories enables training models that account for:

• Atomic mass effects on dynamics
• Element-specific binding strengths
• Lattice parameter influences on pathways
• Temperature scaling across different metals

Transfer Learning

Models trained on copper data can be fine-tuned with platinum trajectories, testing how well learned representations transfer between elements.

Multi-Element Datasets

Combined with copper simulations, this creates datasets spanning different regions of the periodic table—essential for developing generalizable materials models.

Catalyst Design

Platinum’s catalytic importance makes these simulations directly relevant for training models that predict catalyst performance and optimization.

Analysis and Visualization

Use the same analysis tools from the copper simulation to examine platinum diffusion behavior. The identical output format enables direct comparison between elements.

Trajectory Analysis

Track the platinum adatom position over time using:

# Same analysis script as copper simulation
python analyze_trajectory.py --input dump.lammpstrj --output pt_trajectory.png --id 1665

Expected differences from copper:

• Slower motion: Longer time between diffusion hops
• Different patterns: Heavier atoms may favor different pathways
• Temperature effects: More pronounced temperature dependence

Energy Analysis

Monitor system energetics with:

python analyze_energy.py --input energy_avg.txt --output pt_energy.png --skip 30

Platinum should show:

• Higher potential energies: Stronger metallic bonding
• Different fluctuation patterns: Mass effects on thermal motion
• Stability differences: Deeper energy wells for surface sites

Comparative Studies

The real power comes from comparing platinum and copper simulations side-by-side:

1. Diffusion rates: Quantify how atomic mass affects mobility
2. Energy landscapes: Compare activation barriers between elements
3. Temperature scaling: Test how diffusion rates scale with temperature
4. Pathway preferences: Identify element-specific diffusion mechanisms

Summary and Applications

This platinum simulation extends the adatom diffusion framework to precious metals, generating valuable comparative data for machine learning applications. Key insights include:

Physics Understanding

• Heavier atoms diffuse more slowly under identical conditions
• Stronger metallic bonding creates higher energy barriers
• Element-specific effects are crucial for accurate materials modeling

ML Applications

• Element-aware training datasets for neural network potentials
• Transfer learning between similar metals (Cu → Pt)
• Benchmarking models across different atomic species
• Catalyst design through surface dynamics prediction

Future Directions

This simulation framework enables systematic studies across the periodic table:

1. Other precious metals: Au, Pd, Ag diffusion comparisons
2. Alloy surfaces: Mixed Cu-Pt systems for complex behavior
3. Temperature studies: Activation energy determination across elements
4. ML model training: Large-scale datasets for robust model development

The combination of copper and platinum simulations provides an excellent foundation for developing element-aware machine learning models for materials science applications.

Interested in extending these simulations to other elements or applying the data to machine learning models? Feel free to reach out—I’m always excited to discuss computational materials science and its applications.