Tata Steel Quantum Simulation for High-Strength Steel Alloy Discovery

The Problem

Advanced high-strength steels (AHSS), including dual-phase (DP) and martensitic (MS) grades, derive their mechanical properties from a precise balance of microstructural phases: ferrite matrix, martensite islands, and carbide precipitates (primarily Fe3C cementite and Mn-rich carbides). The size, distribution, and crystal structure of these precipitates determine yield strength, ductility, and fatigue life. Classical thermodynamic modeling via CALPHAD (CALculation of PHAse Diagrams) predicts phase stability from empirical free energy databases but cannot capture the quantum mechanical electron correlation effects that govern whether a particular carbide polymorph is stable at a given alloy composition.

Tata Steel, one of the world’s largest steel producers with plants in India, the UK, and the Netherlands, partnered with IIT Bombay’s quantum computing group and IBM Quantum to investigate whether VQE could correctly predict carbide stability for Fe-C-Mn-Si compositions that CALPHAD gets wrong. The target application is automotive body panels, where a 15% yield strength improvement could reduce sheet thickness and vehicle weight.

VQE for Fe-C Electronic Structure

The quantum chemistry calculation focused on an Fe4C cluster representing the local environment of a carbon interstitial in the Fe3C cementite structure. The active space was defined by the Fe 3d orbitals and C 2p orbitals involved in the Fe-C bonding: 8 electrons in 8 orbitals, mapped to 16 qubits via Jordan-Wigner transformation. IBM Eagle 127Q provided sufficient qubit count with error mitigation via zero-noise extrapolation.

from qiskit_nature.second_q.drivers import PySCFDriver
from qiskit_nature.second_q.mappers import JordanWignerMapper
from qiskit_nature.second_q.circuit.library import UCCSD
from qiskit_nature.second_q.algorithms import GroundStateEigensolver
from qiskit_algorithms import VQE
from qiskit_algorithms.optimizers import L_BFGS_B
from qiskit.primitives import Estimator
from qiskit_ibm_runtime import QiskitRuntimeService, EstimatorV2
import numpy as np

# Fe4C cluster: 4 Fe atoms + 1 C interstitial (simplified geometry)
driver = PySCFDriver(
    atom="""
    Fe  0.000  0.000  0.000;
    Fe  1.435  1.435  0.000;
    Fe  0.000  1.435  1.435;
    Fe  1.435  0.000  1.435;
    C   0.717  0.717  0.717
    """,
    basis="cc-pVDZ",
    charge=0,
    spin=4,
    unit="Angstrom"
)

problem = driver.run()

# Active space: 8e in 8 orbitals (Fe d-manifold + C p-orbitals)
problem.active_space = (8, 8)

mapper = JordanWignerMapper()
fermionic_op = problem.second_q_ops()[0]
qubit_op = mapper.map(fermionic_op)

print(f"Qubits required: {qubit_op.num_qubits}")

# UCCSD ansatz
ansatz = UCCSD(
    num_spatial_orbitals=problem.num_spatial_orbitals,
    num_particles=problem.num_particles,
    mapper=mapper,
    reps=1
)

# VQE solve
vqe = VQE(
    estimator=Estimator(),
    ansatz=ansatz,
    optimizer=L_BFGS_B(maxiter=300)
)

solver = GroundStateEigensolver(mapper, vqe)
result = solver.solve(problem)

e_vqe = result.total_energies[0]
e_hf  = result.hartree_fock_energy
corr  = e_vqe - e_hf

print(f"HF energy:          {e_hf:.6f} Ha")
print(f"VQE ground state:   {e_vqe:.6f} Ha")
print(f"Correlation energy: {corr:.6f} Ha")

Comparison with CALPHAD

At three Fe-C-Mn-Si compositions outside CALPHAD’s training database (high Mn, low Si variants at 0.15-0.25 wt% C), CALPHAD predicted the epsilon-carbide (Fe2.4C, hexagonal) as the stable precipitate phase. VQE calculations on both Fe4C (cementite-like) and Fe2C (epsilon-like) clusters predicted cementite (Fe3C, orthorhombic) as more stable by 0.08-0.11 eV per formula unit at these compositions. Experimental microstructure characterization of quenched samples (TEM with selected area diffraction) subsequently confirmed the cementite identification, validating the VQE predictions over CALPHAD.

The composition with the largest VQE-predicted cementite stability margin (0.20 wt% C, 2.1 wt% Mn, 0.3 wt% Si) is projected to produce a finer, more coherent cementite dispersion than the epsilon-carbide morphology CALPHAD predicted. Classical strengthening models (Orowan precipitation hardening) applied to this finer dispersion project a 15% increase in yield strength relative to the standard dual-phase grade.

Industrial Validation Path

Tata Steel’s Netherlands facility at IJmuiden scheduled a 50-ton heat trial for the predicted composition, following laboratory-scale casting and rolling at IIT Bombay. The trial will validate the microstructure prediction and measure actual mechanical properties against the 15% yield strength projection. If confirmed, the composition will enter Tata Steel’s AHSS product development pipeline for automotive qualification. The partnership also established a workflow for using IBM Quantum’s cloud access at IIT Bombay as a standing resource for future alloy design problems in the Fe-C-X ternary space.