Magnetic Density

Goal

To calculate the magnetic moments and optionally extract the spin density ($\rho_{\text{spin}}$) of magnetic materials using spin-polarized DFT calculations. This skill enables the characterization of magnetic ordering, local magnetic moments on individual atoms, and spatial distribution of spin density.

Instructions

1. Structure Preparation

Obtain or prepare the structure of the magnetic material you want to study. You can:

• Query from Materials Project using the base MCP tools (recommended - already DFT-optimized)
• Load from a local CIF/POSCAR file
• Use a previously relaxed structure

Note on relaxation: For magnetic moment calculations, relaxation is optional if using high-quality experimental or Materials Project structures. Magnetic moments are relatively insensitive to small structural variations. However, relaxation is recommended for:

• New/hypothetical structures
• Surfaces, interfaces, or defects
• Systems where you need accurate total energies (not just magnetic moments)
• Strongly correlated oxides with significant magnetic-structural coupling

2. Run Spin-Polarized DFT Calculation

Use the atomate2 MCP tool to run a spin-polarized static calculation. The mp preset (MPStaticSet) automatically enables spin polarization and applies appropriate settings for magnetic systems.

mcp_atomate2_run_atomate2_vasp_calculation(
    structures_path="structure.cif",  # Path to your structure file
    output_dir="magnetic_calc",       # Directory to save results
    preset_type="mp",  # MPStaticSet with PBE (includes spin polarization)
    calculation_type="static",         # Static calculation
)

Important - Functional Selection:

• For metallic ferromagnets (Fe, Co, Ni): Use mp preset (MPStaticSet with PBE). PBE provides excellent accuracy for magnetic moments (typically within ~2% of experimental values)[1].
• For strongly correlated oxides (NiO, CoO, FeO): Use mp preset (see Example 2 below). MPStaticSet automatically applies appropriate GGA+U corrections for transition metal oxides. Standard PBE fails to predict the correct insulating antiferromagnetic ground state and severely underestimates band gaps[3]. Note: +U corrections improve electronic structure but do not guarantee better magnetic moments (e.g., GGA+U may overestimate for NiO or underestimate for CoO due to missing orbital contributions)[4].
• Avoid r2SCAN for metallic ferromagnets: r2SCAN significantly overestimates magnetic moments in itinerant ferromagnets like Fe (by ~24% compared to experimental values)[2].

References: [1] PBE underestimates Fe magnetization by only 1.8%: Zhang et al., Phys. Rev. Materials 6, 013801 (2022). DOI: 10.1103/PhysRevMaterials.6.013801 [2] r2SCAN and SCAN overestimate Fe magnetization by 24% and 17% respectively: ibid. [3] PBE+U opens band gaps and corrects ground state in transition metal oxides: Kulik, J. Chem. Phys. 142, 240901 (2015). DOI: 10.1063/1.4922693 [4] CoO magnetic moments: PBE+U underestimates due to missing orbital contributions: Radi et al., Z. Naturforsch. A 70, 789 (2015). DOI: 10.1515/zna-2015-0216

See the "Functional Selection Guide" section below for detailed recommendations.

3. Monitor Job Status

After submitting the calculation, monitor its progress:

mcp_atomate2_get_atomate2_job_status(
    job_id="<job_id_from_step_2>"  # Job ID returned from step 2
)

4. Extract Magnetic Moments

Once the calculation is complete, retrieve the results to extract magnetic moments:

mcp_atomate2_get_atomate2_results_by_id(
    job_ids=["<job_id>"],  # List of job IDs
    save_to_file="magnetic_results.json"  # Save results to file
)

The results will include:

• Total magnetization: Net magnetic moment of the unit cell
• Site magnetic moments: Magnetic moment on each atom
• Energy: Total energy of the magnetic configuration

5. Parse and Analyze Results

Use the provided script to parse the magnetic moments from the results:

# Env: base-agent
python .agents/skills/mat-magnetic-density/scripts/parse_magnetic_moments.py magnetic_results.json --output magnetic_analysis.json

This script will:

• Extract site-resolved magnetic moments
• Calculate the total magnetization
• Identify the magnetic ordering pattern
• Generate a summary report

6. Extract Spin Density (Optional)

For detailed analysis of spin density distribution, use the spin density extraction script:

# Env: base-agent
python .agents/skills/mat-magnetic-density/scripts/extract_spin_density.py <output_dir> --output spin_density.json

This requires access to the CHGCAR file from the VASP calculation and will:

• Read the spin density from CHGCAR
• Calculate integrated spin moments
• Optionally generate 3D visualization data

7. Visualize Magnetic Ordering (Optional)

Visualize the magnetic ordering by creating a structure with magnetic moment vectors:

# Env: base-agent
python .agents/skills/mat-magnetic-density/scripts/visualize_magnetic_structure.py structure.cif magnetic_analysis.json --output magnetic_structure.png

Examples

Example 1: Calculate magnetic moments for bulk Fe

# Step 1: Get Fe structure from Materials Project
mcp_base_search_materials_project_by_formula(
    formula="Fe",
    save_to_file="Fe_mp.cif"
)

# Step 2: Run spin-polarized static calculation
mcp_atomate2_run_atomate2_vasp_calculation(
    structures_path="Fe_mp.cif",
    output_dir="Fe_magnetic",
    preset_type="mp",  # MPStaticSet with PBE
    calculation_type="static",
    execution_mode="remote"
)

# Step 3: After completion, retrieve results
mcp_atomate2_get_atomate2_results_by_id(
    job_ids=["<job_id>"],
    save_to_file="Fe_magnetic_results.json"
)

# Step 4: Parse magnetic moments
# Env: base-agent
python .agents/skills/mat-magnetic-density/scripts/parse_magnetic_moments.py Fe_magnetic_results.json --output Fe_moments.json

Example 2: NiO with automatic GGA+U

# For transition metal oxides like NiO, use 'mp' preset
# MPStaticSet automatically applies appropriate U parameters (e.g., U=6.2 eV for Ni in oxides)
# This ensures correct insulating antiferromagnetic ground state

mcp_atomate2_run_atomate2_vasp_calculation(
    structures_path="NiO.cif",
    output_dir="NiO_magnetic",
    preset_type="mp",  # MPStaticSet automatically handles +U for transition metal oxides
    calculation_type="static",
    execution_mode="remote"
)

Functional Selection Guide

Choosing the right exchange-correlation functional is critical for accurate magnetic property calculations:

For Metallic Ferromagnets (Fe, Co, Ni, etc.)

Recommended: PBE (use preset_type="mp" for MPStaticSet)

• PBE underestimates magnetic moments by only ~1.8% for Fe
• Provides excellent agreement with experimental spin magnetization
• Computational efficiency superior to meta-GGA functionals
• Do NOT use: r2SCAN overestimates by ~24% for Fe

Example: Bulk Fe experimental value = 2.2 μB/atom

• PBE prediction: ~2.15 μB/atom (2% error) ✓
• r2SCAN prediction: ~2.73 μB/atom (24% error) ✗

For Transition Metal Oxides and Strongly Correlated Systems

Recommended: PBE with automatic +U (use preset_type="mp" - MPStaticSet handles this automatically)

When to use PBE+U:

1. Localized d-electrons: Systems where d-electrons exhibit strong correlation (NiO, CoO, Fe₂O₃)
2. Band gap issues: Standard PBE underestimates band gaps significantly
3. Magnetic ground states: PBE gives incorrect magnetic ordering or underestimates moments
4. Redox chemistry: Systems involving oxidation state changes

U parameter selection:

• U values are material-dependent and typically calibrated against experimental band gaps or magnetic moments
• Common values: U = 3-5 eV for 3d transition metals in oxides
• For NiO: U = 5-6 eV is typical
• Consult literature for your specific system

Why standard GGA fails for oxides:

• Self-interaction error artificially delocalizes d-electrons
• Underestimates band gaps and magnetic exchange interactions
• May predict incorrect magnetic ground states

Constraints

• Spin Polarization: The mp preset (MPStaticSet) automatically enables spin polarization (ISPIN=2) for magnetic systems. MAGMOM values are also automatically initialized (default 0.6 per magnetic atom).
• Initial Magnetic Moments: For complex magnetic ordering (e.g., antiferromagnetic), you should provide initial MAGMOM values through the config parameter to help convergence.
• Environment: The parsing scripts require the base-agent conda environment with pymatgen installed.
• Remote Execution: DFT calculations are computationally expensive and should typically be run on remote clusters using execution_mode="remote".
• CHGCAR Access: Extracting spin density requires access to the CHGCAR file, which may not be automatically retrieved. You may need to manually download it from the remote execution directory.
• Convergence: Magnetic systems can be challenging to converge. If calculations fail to converge, try:
  • Adjusting MAGMOM initial values
  • Increasing NELM (max electronic steps)
  • Using a denser k-point mesh
  • Enabling LDAU for strongly correlated systems

References

• VASP Manual: Spin-polarized calculations
• Pymatgen Documentation: Magnetic Structure Analysis

---

---

Author: Bowen Deng Contact: GitHub @learningmatter-mit