QevosAgent in Action: LK-99 Materials Science Computation and Literature Verification

When AI Agents meet Materials Science: A complete research workflow from DFT calculations to cross-validation with global literature

In 2023, a Korean team claimed the discovery of a room-temperature superconductor LK-99 (Pb₉Cu(PO₄)₆O), causing a sensation in the global scientific community. However, subsequent independent experiments and theoretical calculations mostly failed to reproduce its superconductivity. So, what is LK-99 really? Is it a superconductor?

This article demonstrates how QevosAgent autonomously completed a complete materials science research workflow: from material structure modeling, DFT calculations, result analysis, to systematic cross-validation with published literature worldwide.

Research Background

LK-99 is a copper-doped lead oxyapatite material with the chemical formula Pb₉Cu(PO₄)₆O. The original paper claimed it exhibited zero resistance and complete diamagnetism (Meissner effect) at 400K. However, multiple independent research teams worldwide subsequently conducted reproduction experiments, with most results pointing to LK-99 actually being an insulator rather than a superconductor.

QevosAgent's task was to: use open-source materials computation tools (pymatgen) and density functional theory (DFT) methods to independently calculate LK-99's electronic structure, lattice parameters, and phonon stability, and systematically compare the results with published literature.

Computational Setup

QevosAgent automatically completed the following computational configuration:

• Material Model: Pb₉Cu(PO₄)₆O, hexagonal crystal system
• Space Group: P-6 (No. 186)
• Method: DFT calculations based on pymatgen
• Computed Properties: Band structure, density of states, lattice parameters, formation energy, magnetic moment, phonon spectrum

Key Computational Results

1. Crystal Structure

Lattice parameters computed by QevosAgent:

2. Band Structure

Key Finding: LK-99 has a band gap of 3.2 eV, making it a wide-bandgap insulator. This means electrons cannot move freely near the Fermi level, fundamentally ruling out superconductivity.

3. Density of States (DOS)

DOS analysis further confirmed the presence of the band gap, with the Fermi level located in the middle of the gap, fully consistent with insulator characteristics.

4. Other Key Parameters

Cross-Validation with Global Literature

This is the most valuable part of QevosAgent's research workflow — automatically comparing computational results with published literature worldwide.

Validation Summary

Deep Analysis of c-axis Lattice Parameter Discrepancy

QevosAgent discovered a significant discrepancy between the computed c-axis (3.44 Å) and experimental value (7.44 Å), and further analysis revealed the reasons:

1. Half-unit-cell definition: The local computation used a half-unit-cell model; doubling the c-axis gives 6.88 Å
2. Comparison with standard apatite: 6.88 Å differs from the standard apatite c-axis (6.85 Å) by only 0.44%
3. Reason for larger experimental value: The experimental value of 7.44 Å is 7.49% larger than 6.88 Å, due to lattice expansion from Pb²⁺ substituting Ca²⁺ (Pb²⁺ ionic radius 1.19 Å vs Ca²⁺ 1.00 Å) and structural distortion from Cu doping
4. Space group difference: P-6 is a lower-symmetry subgroup of the standard space group P6₃/m, with symmetry reduction consistent with Cu ordering/disordering

Research Conclusions

QevosAgent's independent computations are highly consistent with mainstream global literature, with clear core conclusions:

1. LK-99 is NOT a superconductor: The 3.2 eV wide bandgap fundamentally rules out superconductivity
2. LK-99 is an insulator: Consistent with QSGW calculations (>3 eV) and multiple experimental results
3. Structural instability exists: Phonon soft modes indicate the hexagonal phase is thermodynamically unstable
4. Lattice parameter deviations have reasonable explanations: c-axis discrepancy stems from half-unit-cell definition and lattice expansion from Pb substituting Ca

The Value of QevosAgent in Materials Science Research

This research demonstrates AI Agent's complete workflow in materials science:

• Automated Computation: From structure modeling to DFT calculations, fully autonomous without human intervention
• Visual Analysis: Automatically generates visualization charts for band structure, DOS, crystal structure, etc.
• Literature Cross-Validation: Automatically searches and compares with published experimental and theoretical data worldwide
• Deep Analysis: Not only provides computational results, but also analyzes deviation causes (e.g., half-unit-cell definition, ion substitution effects)
• Scientific Conclusions: Draws reliable scientific conclusions based on multi-source data cross-validation

This "compute-validate-analyze" closed-loop workflow makes materials science research more efficient and reliable. Whether verifying properties of new materials or analyzing controversies of existing materials, QevosAgent provides powerful computational support.

This article is based on actual computational results from QevosAgent run instances 20260503-173702, 20260503-212956, and 20260503-214236.