| Variational Quantum Eigensolver | |
|---|---|
| Name | Variational Quantum Eigensolver |
| Developer | IBM Research, Google Quantum AI, Rigetti Computing, Xanadu |
| Introduced | 2014 |
| Application | Quantum chemistry, Materials science |
| Programming language | Qiskit, Cirq, Forest |
Variational Quantum Eigensolver
The Variational Quantum Eigensolver (VQE) is a hybrid quantum algorithm that estimates the lowest eigenvalue of a Hamiltonian using a parameterized quantum circuit and a classical optimizer. It is important in quantum physics because it enables near-term noisy intermediate-scale quantum (NISQ) devices to tackle problems in quantum chemistry and materials science that are intractable for classical algorithms. VQE bridges quantum computing hardware constraints and variational methods from computational physics.
The VQE was proposed to leverage variational principles such as the Rayleigh–Ritz method to compute ground-state energies by minimizing the expectation value of a target Hamiltonian H. The algorithm interleaves short-depth quantum circuits run on devices from vendors like IBM, Google, and Rigetti Computing with classical routines including gradient descent, Nelder–Mead, and other optimization methods. VQE is a cornerstone of variational quantum algorithms research and a primary candidate for near-term demonstrations of quantum advantage in chemistry and materials modeling.
VQE uses the variational principle of quantum mechanics: for any normalized trial state |ψ(θ)⟩ parameterized by θ, ⟨ψ(θ)|H|ψ(θ)⟩ ≥ E0, where E0 is the ground-state energy. The trial state is prepared by a parameterized quantum circuit often called an ansatz; common examples include the unitary coupled cluster (UCC) ansatz and hardware-efficient ansätze. VQE maps fermionic problems to qubits using transformations such as the Jordan–Wigner transformation and the Bravyi–Kitaev transformation, producing a qubit Hamiltonian expressed as a sum of Pauli matrices and measured via quantum measurement primitives. Theoretical analysis connects VQE to concepts in variational methods, adiabatic quantum computing, and quantum approximate optimization algorithm (QAOA) frameworks.
A typical VQE run consists of: (1) encoding a problem Hamiltonian from models like the electronic structure problem or the Hubbard model; (2) selecting an ansatz such as UCCSD or a hardware-efficient ansatz; (3) initializing parameters and iteratively preparing the trial state on a quantum processor; (4) measuring expectation values of Pauli-term decompositions; and (5) updating parameters with classical optimizers (e.g., BFGS, COBYLA, SPSA). Efficient measurement strategies exploit grouping by commutativity, classical shadows, and error mitigation techniques. Cost-function landscapes and barren plateau phenomena influence optimizer choice and convergence.
Implementations target NISQ devices from companies and institutions including IBM Quantum, Google Quantum AI, Rigetti, IonQ, and research groups at MIT, Caltech, and Harvard University. Hardware considerations include gate fidelity, qubit connectivity, coherence times, and native gate sets (e.g., CZ gate, cross-resonance). Platforms vary: superconducting qubits, trapped ions, and photonic processors (e.g., Xanadu). Practical runs use software stacks such as Qiskit, Cirq, PyQuil, and OpenFermion for Hamiltonian construction and circuit transpilation. Techniques such as readout error mitigation, zero-noise extrapolation, and randomized compiling are often employed to improve measured energies.
VQE has been applied to compute ground-state energies and properties of small molecules like H2, LiH, and BeH2, and to model correlated materials using simplified lattice models. It is used to approximate electronic structure problems central to computational chemistry, complementing classical methods like density functional theory (DFT) and coupled cluster approaches. Research efforts at IBM Research, Google, and academic groups target scaling to larger active spaces, embedding techniques, and error-mitigated simulations relevant for catalysis, drug discovery, and materials design.
Performance depends on ansatz expressibility and trainability, noise characteristics, and optimizer robustness. Common error sources include depolarizing noise, gate errors, measurement errors, and decoherence. Phenomena such as barren plateau landscapes can cause vanishing gradients, impeding training as system size increases. Scalability is constrained by qubit counts, circuit depth, and required measurement overhead; strategies to improve scaling include problem-tailored ansätze, qubit reduction via symmetries (e.g., particle number conservation), and tensor factorization techniques inspired by quantum chemistry approximations.
Numerous variants extend VQE: subspace-expansion methods for excited states, quantum subspace expansion (QSE), adaptive schemes like ADAPT-VQE, and measurement-efficient variants using classical shadows or grouping heuristics. Hybrid quantum-classical workflows combine VQE with classical pre- and post-processing, embedding methods such as density matrix embedding theory (DMET), and machine learning approaches that use neural-network-based optimizers. Integration with error mitigation protocols and advances in hardware (e.g., improved qubit coherence and gate fidelities) continue to expand VQE applicability toward larger, chemically relevant systems.