VQE for First-Wall Electronic Structure

The variational quantum eigensolver estimates ground-state energies of first-wall materials by minimizing a classically driven cost over a quantum ansatz.

The variational principle on hardware

The variational quantum eigensolver (VQE) is the leading NISQ approach to electronic structure. It prepares a parameterized state |psi(theta)>, measures the energy expectation, and hands that scalar to a classical optimizer that adjusts theta. It rests on the variational principle: the measured energy is always an upper bound on the true ground-state energy.

# VQE outer loop (classical optimizer drives quantum evaluations)
def vqe(H_paulis, ansatz, theta0, optimizer):
    def energy(theta):
        E = 0.0
        for coeff, pauli in H_paulis:      # H = sum coeff * Pauli
            E += coeff * measure_expectation(ansatz(theta), pauli)
        return E                            # <psi(theta)| H |psi(theta)>
    return optimizer.minimize(energy, theta0)

# Variational bound:  E(theta) = <psi(theta)|H|psi(theta)>  >=  E_ground

The first-wall problem

The breeder first wall faces 14 MeV neutrons at high fluence. Understanding displacement damage, helium and hydrogen retention, and vacancy clustering in tungsten requires accurate energetics of open-shell d-electron configurations, where classical mean-field methods are least reliable. A small correlated active space around a defect is the natural target for VQE benchmarking.

The obstacles we do not hide

• Measurement cost: the number of Pauli terms grows as O(N^4) in orbitals, so accurate energies need enormous shot counts.
• Barren plateaus: for many ansatze the cost-function gradient vanishes exponentially in qubit count, stalling optimization.
• Noise: gate errors bias the energy, requiring error mitigation that itself costs more samples.

Because of these, VQE has not beaten classical coupled-cluster or DMRG on any real first-wall problem. Kronos uses it as a calibrated benchmark: run VQE and a classical solver on the same small active space, and only extend trust as hardware and ansatze improve. Genuine advantage waits on error correction.