qfoldit-quantum

qfoldit-quantum (VQE simulator)

A from-scratch classical statevector VQE simulator (numpy only -- no Qiskit, no real quantum hardware, no network access to quantum cloud backends). Implements: Pauli-string Hamiltonian construction, a hardware-efficient ansatz (Kandala et al. 2017 style: RY+RZ rotations per qubit per layer, CNOT entanglers between layers), and a gradient-free classical optimization loop (COBYLA with multi-restart) that minimizes the ansatz's energy expectation value.

Read references/model_documentation.md before answering -- it has the full validation results (toy Hamiltonians, and a real H2 molecular benchmark matching the well-known -1.1373 Hartree literature value) and, critically, the scope limits below.

Critical scope limits -- read before promising anything

1. This is a classical simulator, not a real quantum backend. Statevector simulation scales as 2^n in memory/compute -- practical here only for small systems (2-6 qubits demonstrated; tens of qubits would already be infeasible on this classical hardware). It cannot provide any quantum computational advantage; it is useful for algorithm development/testing/education, matching what a "simulator" backend option would offer, not a "real quantum hardware" backend.
2. No arbitrary-molecule Hamiltonian generation. Building a qubit Hamiltonian for a molecule the user names (beyond H2, whose coefficients were sourced from a published paper for validation) requires an actual quantum chemistry integral/mapping pipeline (PySCF + Jordan-Wigner/parity mapping, e.g. via OpenFermion or Qiskit Nature) that is NOT implemented in this skill. If the user asks to fold/simulate an arbitrary molecule, this skill can only proceed if they supply the qubit Hamiltonian coefficients directly (as a Pauli-string dict) -- it cannot derive them from a SMILES string or molecule name on its own.
3. VQE energies are electronic-structure energies only. For a real molecule's total energy, nuclear repulsion must be added separately (see nuclear_repulsion_energy -- point-charge model, correct for diatomics; polyatomic systems need a sum over all nuclear pairs, not implemented here beyond the 2-body case).
4. Quantum annealing is a separate backend mentioned in the qFoldIT architecture (quantum-adapter unifies VQE and annealing) -- this skill covers VQE only. Do not answer annealing questions as if this skill's code applies.