207
Part Five · Convergence

Build project: simulate a small biomolecule end to end (Hamiltonian, mapping, VQE, analysis)

This is the project that ties Part 5 together: take a molecule from a biological question to a quantum-computed answer and back, doing every step yourself. The plate runs the whole chain for the hydrogen molecule, the smallest case where every number can be checked. Then the chapter gives you code to repeat it with real tools and a slightly bigger, biologically flavoured molecule, and a rubric for the two-page report you write at the end. It is educational: it shows no quantum advantage, and claims none.

On the plate: the pipeline, one step at a time.

The seven steps

(1) Question and molecule: a bond-breaking energy, the simplest model of what enzymes do to bonds. (2) Integrals: STO-3G integrals (Chapter 184). (3) Mapping and tapering: four spin orbitals; with particle number, spin and spatial symmetry, the problem reduces to the two configurations σg2\sigma_g^2 and σu2\sigma_u^2, a single qubit (Chapter 186). (4) VQE: one rotation, Ry(θ)R_y(\theta), with shot noise (Chapter 187). (5) Phase estimation: ideal phase estimation from the Hartree–Fock state and from the exact state (Chapter 188). (6) Resource estimate for scaling the method to a real target (Chapter 189). (7) Critique against classical methods. Established

Run it

Bond length 1.40 bohr; phase-estimation bits 5. Show step . The one-qubit Hamiltonian is g0I+gzZ+gxXg_0I + g_zZ + g_xX, with gxg_x the exchange integral coupling the two configurations. Phase estimation from the Hartree–Fock state returns the ground energy with probability equal to the overlap squared, high near equilibrium and falling to one half as the bond breaks, where the two configurations mix equally: the overlap problem of Chapter 188 in miniature. To repeat with real tools:

from pyscf import gto, scf, ao2mo
import numpy as np
mol = gto.M(atom="H 0 0 0; H 0 0 0.7408", basis="sto-3g")   # 1.4 bohr
mf = scf.RHF(mol).run()
C = mf.mo_coeff
h = C.T @ mf.get_hcore() @ C
eri = ao2mo.kernel(mol, C, compact=False).reshape(2, 2, 2, 2)
Enuc = mol.energy_nuc()
Ehf = 2*h[0, 0] + eri[0, 0, 0, 0] + Enuc
E2  = 2*h[1, 1] + eri[1, 1, 1, 1] + Enuc
K   = eri[0, 1, 0, 1]
print(np.linalg.eigvalsh([[Ehf, K], [K, E2]])[0])   # about -1.1373

from qiskit import QuantumCircuit
from qiskit.circuit import Parameter
from qiskit.quantum_info import SparsePauliOp
from qiskit.primitives import StatevectorEstimator
op = SparsePauliOp(["I", "Z", "X"], [(Ehf + E2)/2, (Ehf - E2)/2, K])
t = Parameter("t"); qc = QuantumCircuit(1); qc.ry(t, 0)
est = StatevectorEstimator()
E = [float(est.run([(qc, op, [x])]).result()[0].data.evs)
     for x in np.linspace(-np.pi, np.pi, 201)]
print(min(E))

Then choose a biologically flavoured target, formaldehyde (a carbonyl model) in a small active space such as CAS(4e, 3o), and repeat with a general mapping. Established

Report and rubric

Write two pages. Your report is graded on: (a) a correct chain, every number reproducible; (b) a resource estimate for scaling to a meaningful target, with assumptions stated; (c) a comparison with CCSD(T) and, for the larger case, DMRG or selected CI, run or cited; (d) an honest status label for any advantage; (e) a limitations paragraph naming the basis set, active space and noise model. A strong report says plainly what the quantum computation added, which for these molecules is nothing a laptop could not do, and what would have to change for that to be otherwise. Established as method.

What you now have

A complete, checkable pipeline from molecule to quantum result: integrals, symmetry reduction to one qubit, VQE, phase estimation with the overlap effect, the path to resource estimation, runnable PySCF and Qiskit code, and a rubric for reporting honestly.

· · ·

Three questions to carry with you.

Why does minimal-basis H₂ need only one qubit?tap
Symmetry leaves two configurations, σg² and σu², coupled by one exchange integral.
What happens to the HF overlap as H₂ dissociates?tap
It falls toward 1/√2, so phase estimation from HF succeeds only half the time.
What must a good report admit?tap
That for small molecules the quantum computation adds nothing a classical computer cannot do.

And one to work out. At 1.4 bohr, what is the coupling gxg_x (the exchange integral KguK_{gu}) in hartree, to three decimals?

g_x ≈

Sources: A. Szabo and N. S. Ostlund, Modern Quantum Chemistry (Dover, 1996). Q. Sun et al., J. Chem. Phys. 153, 024109 (2020) (PySCF). Qiskit 2.x documentation (2026). Checked by the plate: one-qubit Hamiltonian from Book.H2; VQE minimum equals FCI; ideal phase-estimation probabilities from exact overlaps.

Next · Chapter 208
Reading and critiquing a research paper

Annotate a real paper (for example Robledo-Moreno 2025 ✓ or Xu 2021 ✓) with claims, evidence, controls and overreach

Turn the page