47. Quantum Simulation
47.1Why Simulate Nature with Quantum Systems
A molecule of n interacting orbitals lives in a Hilbert space of dimension exponential in n; classical simulation must truncate, approximate, or surrender — quantum Monte Carlo and DFT are brilliant approximations with known failure modes (strongly correlated electrons, transition-metal chemistry, high-Tc superconductivity). A quantum computer represents the state natively: 2ⁿ amplitudes in n qubits, evolved by gates that are the physical dynamics. This is the least speculative application: the speedup argument requires no hidden algebra or unproven assumptions — only that quantum states are hard to write down classically (Part V proved it). The bottleneck is engineering: useful chemistry needs error-corrected machines (Ch. 35's costs), which is why simulation leads every serious roadmap.
47.2Hamiltonians
The Hamiltonian H is the energy operator — the program and the data of quantum simulation: it encodes the system's physics, and evolution is generated by it. The relevant form for computation: H = Σₖ hₖ Pₖ, a weighted sum of Pauli strings (measurements of each term are exactly what Part V's operator machinery gives you). Examples to internalize: the transverse-field Ising model H = −J Σ ZᵢZⱼ − h Σ Xᵢ (the workhorse of everything from magnetism to QAOA), the Hubbard model (electrons on a lattice — the high-Tc mystery), molecular Hamiltonians (Ch. 48). Reading a physics paper's Hamiltonian and seeing "circuit to prepare and evolve this" is the core literacy of quantum simulation.
47.3Time Evolution
The simulation task: given H, prepare a state, compute e^(−iHt)|ψ⟩ — dynamics, thermal states, or (via phase estimation) energies. Classical circuits get e^(−iHt) for free from linear algebra when H is small; quantumly, we build it from the exponential of each Pauli term, which has an exact circuit: for a Pauli string P with support S, conjugate the qubits in S into the X basis, apply a two-qubit "phase rotation" CNOT-ladder, conjugate back. Cost: O(|S|) two-qubit gates per term per time step. The engineering is in the bookkeeping — Pauli-string circuits compose, commute, and route; your Part XII compiler is already the right tool.
47.4Trotterization
H = Σₖ hₖPₖ's terms don't commute, so e^(−iHt) ≠ Π e^(−ihₖt) — but Trotterization bridges the gap: Πₖ e^(−ihₖt/r), repeated r times, approximates the true evolution with error O(t²/r) (first order; second-order symmetric formulas halve the constant). The tuning problem: more steps = more accuracy but more gates = more noise — the same tradeoff everywhere in quantum engineering. Known improvements: commutator-aware orderings, randomized products (qDRIFT — sample terms by weight, unbiased), and the asymptotic champion qubitization (28.7), which achieves optimal T-count for structured H. Implement Trotter evolution of the 2-site Ising model on your simulator (Part VI's!) and watch observables oscillate — a one-evening experiment that makes "dynamics" concrete.
47.5Phase Estimation
The readout stage: given a state |ψ⟩ with H|ψ⟩ = E|ψ⟩ (or U|ψ⟩ = e^(iφ)|ψ⟩), phase estimation (Ch. 24's machinery) extracts E to k bits of precision with O(1/ε) controlled-U applications — the quantum side of "diagonalize the Hamiltonian." Applications: ground-state energies (chemistry, Ch. 48), quasiparticle spectra, and the theoretical underpinning of Shor itself. Honest engineering note: textbook QPE needs deep circuits and clean ancillas; the fault-tolerant era uses iterated/sampled variants (ML-QPE, Kitaev) with better resource profiles. Precision budgeting is the real design skill: chemical accuracy (~1.6 mHartree) sets both the phase bits and the simulation error tolerance — every layer must agree on ε.
47.6Fermionic Systems
Electrons are fermions: antisymmetric under exchange, which makes their Hilbert space awkward for qubits. Second quantization handles it: label orbitals, represent configurations as occupation numbers — occupation number n is a natural bit! — and represent fermionic operators (creation/annihilation a†, a) that obey the anticommutation algebra {aᵢ, aⱼ†} = δᵢⱼ. The mismatch: qubit operators (Paulis) commute tensor-wise, fermions anticommute — the mappings of Ch. 48 fix this at the price of non-local qubit operators. Before the mappings, learn the physics that needs them: the Jordan–Wigner string intuition (ordering orbitals, and paying for it in string length) will make the transformations feel inevitable rather than magical.
47.7Molecular Simulation
End to end: molecular geometry → basis set → integrals → second-quantized Hamiltonian → fermion-to-qubit mapping → circuit → energy. Every step has knobs and costs: basis size sets qubit count (H₂ in a minimal basis: 4 qubits; FeMoco, nitrogenase's active site, the famous target: ~100+ logical qubits with 10⁹–10¹⁰ T gates in modern estimates — millisecond-scale evolution on a fault-tolerant machine). Interim landmarks that exist today: exact simulation of H₂/LiH on simulators, analog-digital experiments on hardware. This section's experiment: reproduce the H₂ dissociation curve (Ch. 48's pipeline) on your laptop — the entire application stack, miniaturized, verifiable against textbook chemistry.