Grain Boundary Modeling

# Grain Boundary Modeling

## Introduction & Motivation

Predicting grain boundary properties and segregation enables design of materials with optimized microstructure and mechanical properties. ML models learn grain boundary energies and properties from computational and experimental data.

Motivation: Predict grain boundary properties and segregation.

Applications: Grain boundary energy, segregation prediction, microstructure design, mechanical property optimization.

---

## Core Concepts & Theory

### Grain Boundaries

Interface between grains.

### Misorientation

Crystallographic relationship.

### Boundary Energy

Interfacial tension.

### Segregation

Impurity accumulation.

---

## Mathematical Formulation

Grain Boundary Energy:
$$\gamma_{GB} = \frac{E_{ ext{GB}} - E_{ ext{bulk}}}{A}$$

Coincident Site Lattice:
$$\Sigma = \frac{V_{ ext{unit cell}}}{V_{ ext{CSL}}}$$

Segregation Energy:
$$E_{ ext{seg}} = E_{ ext{solute at GB}} - E_{ ext{solute in bulk}}$$

---

## Advanced Theory & Extensions

### Dislocation Content

Structural elements.

### Boundary Kinetics

Migration rates.

### Complexion Phases

Interfacial phases.

---

## Computational Considerations

Boundary Construction: O(N_atoms·D) complexity.

Energy Model: O(D²) network.

Prediction: O(D) per boundary.

---

## Practical Implementation Strategies

### Misorientation Encoding

Euler angles representation.

### Interface Descriptors

Local atomic environment.

### Segregation Features

Solute-matrix interaction.

---

## Benchmark Datasets & Evaluation

Computational Database: GBDB simulations.

Literature Data: Published values.

Experimental Measurements: Property data.

---

## Key Challenges & Limitations

### Boundary Complexity

Large structural space.

### Kinetic Barriers

Migration barriers.

### Temperature Dependence

Thermal effects.

---

## Hyperparameter Tuning

Hidden units: 128-256 neurons.

Dropout: 0.2-0.4 regularization.

Learning rate: 1e-4 to 1e-2 schedule.

---

## Real-World Applications & Case Studies

Steels: Grain boundary strengthening.

Aluminum: Mechanical properties.

Ceramics: Fracture behavior.

---

## Integration with Other Methods

GB ML + DFT; + experiments; + microstructure simulation.

---

## Summary & Key Takeaways

ML models grain boundary properties efficiently.

Principles:
1. Misorientation: Crystallographic encoding.
2. Structure: Boundary configuration.
3. Energy: Thermodynamic prediction.
4. Segregation: Impurity behavior.
5. Properties: Mechanical effects.

---

## Appendix: Practical Labs

### Lab 1: Misorientation Encoding

import numpy as np

def encode_misorientation(euler_angles):
 """Encode grain boundary misorientation"""
 # Euler angles to rotation descriptor
 descriptor = np.array([
 np.cos(euler_angles[0]),
 np.sin(euler_angles[0]),
 np.cos(euler_angles[1]),
 np.sin(euler_angles[1])
 ])
 assert len(descriptor) == 4, "Descriptor dimension error"
 return descriptor

angles = np.array([np.pi/4, np.pi/6, np.pi/3])
descriptor = encode_misorientation(angles)

assert descriptor.shape == (4,), "Encoding failed"
print(f"✓ Misorientation descriptor: {descriptor}")

### Lab 2: Grain Boundary Energy

import numpy as np

class GBEnergyPredictor:
 def __init__(self, descriptor_dim=10):
 self.weights = np.random.randn(descriptor_dim) * 0.01
 self.bias = 1.0
 
 def predict_gb_energy(self, descriptor):
 """Predict grain boundary energy"""
 energy = descriptor @ self.weights + self.bias
 return energy

descriptor = np.random.randn(10)
predictor = GBEnergyPredictor()
energy = predictor.predict_gb_energy(descriptor)

assert isinstance(energy, (float, np.ndarray)), "Prediction failed"
print(f"✓ GB energy: {energy:.3f} J/m²")

### Lab 3: Segregation Prediction

import numpy as np

def predict_segregation_energy(solute, matrix, gb_type):
 """Predict segregation energy"""
 # Interaction energy model
 interaction = np.sin(solute) * np.cos(matrix)
 gb_factor = 1.5 if gb_type == 'high_angle' else 1.0
 
 seg_energy = -2.0 * interaction * gb_factor
 return seg_energy

seg_e = predict_segregation_energy(1.0, 0.5, 'high_angle')

assert isinstance(seg_e, (float, np.ndarray)), "Prediction failed"
print(f"✓ Segregation energy: {seg_e:.3f} eV")

### Lab 4: Boundary Property Optimization

import numpy as np

class GBOptimizer:
 def __init__(self, target_energy=1.0):
 self.target = target_energy
 
 def optimize_misorientation(self, n_iterations=20):
 """Optimize misorientation for target GB energy"""
 best_angles = np.random.randn(3)
 best_error = float('inf')
 
 for _ in range(n_iterations):
 angles = best_angles + np.random.randn(3) * 0.1
 
 energy = 1.5 + 0.5 * np.sum(np.cos(angles))
 error = abs(energy - self.target)
 
 if error < best_error:
 best_error = error
 best_angles = angles
 
 return best_angles

opt = GBOptimizer(target_energy=1.2)
optimal = opt.optimize_misorientation()

assert optimal.shape == (3,), "Optimization failed"
print(f"✓ Optimized misorientation angles: {optimal}")

---