Quantum chemistry turns molecular electronic structure into a many-body eigenvalue problem. The mathematics begins with electrons moving in the electrostatic field of nuclei, passes through antisymmetric fermionic wavefunctions and second quantisation, then reaches a quantum computer only after fermionic creation and annihilation operators are mapped to qubit Pauli operators.
Guide 14 owns the variational eigensolver machinery. Guide 46 owns Jordan–Wigner as an exact one-dimensional spin-chain transformation. This guide owns a different job: how molecular electronic structure becomes a qubit Hamiltonian suitable for quantum algorithms.
The central object is the electronic Hamiltonian. In a finite spin-orbital basis it becomes
H = Σpq hpqap†aq + (1/2)Σpqrs hpqrsap†aq†asar + ENN.
The exact index convention and factor in the two-electron term depend on how the integrals are defined. This guide states its convention explicitly and keeps the algebra internally consistent.
Molecular geometry → one-electron basis → fermionic Hamiltonian → occupation-number encoding → qubit mapping → Pauli Hamiltonian → quantum eigensolver or phase-estimation layer.
1. Born–Oppenheimer separation
Nuclei are much heavier than electrons. A standard first approximation fixes nuclear positions RA and solves the electronic problem at that geometry.
The nonrelativistic electronic Hamiltonian in atomic units is
H_e = −(1/2)Σ_i ∇_i² − Σ_{i,A} Z_A/|r_i−R_A| + Σ_{i<j}1/|r_i−r_j|.
The classical nuclear-repulsion term is
E_NN=Σ_{A<B} Z_AZ_B/|R_A−R_B|.
At fixed geometry the total Born–Oppenheimer electronic energy is Eelectronic+ENN.
2. Why basis sets are needed
The electronic wavefunction depends continuously on electron positions and spins. Digital computation starts by choosing a finite one-electron basis.
Spatial atomic-orbital basis functions χμ(r) are combined into molecular orbitals
φ_p(r)=Σ_μ C_{μp}χ_μ(r).
Each spatial orbital can be paired with spin α or β, producing spin orbitals.
The finite basis introduces basis-set error. Increasing basis size systematically improves flexibility but also increases the number of spin orbitals—and therefore qubits in direct occupation encodings.
3. Antisymmetry
Electrons are identical fermions. Exchanging electron labels changes the many-electron wavefunction sign:
Ψ(...,x_i,...,x_j,...)=−Ψ(...,x_j,...,x_i,...).
This antisymmetry is not optional bookkeeping. It enforces Pauli exclusion and creates exchange structure.
Slater determinants build antisymmetry automatically from one-electron spin orbitals.
4. Slater determinants
For N occupied spin orbitals φp1,…,φpN, the determinant state is
Ψ(x_1,…,x_N)=1/√(N!) det[φ_{p_j}(x_i)].
Swapping two electron coordinates swaps two determinant rows and changes sign.
Hartree–Fock selects one determinant variationally. Correlated methods superpose many determinants.
5. Occupation-number language
Instead of labelling electrons individually, label whether each spin orbital is occupied:
|n_0 n_1 … n_{M−1}⟩, n_p∈{0,1}.
This is fermionic Fock space. A fixed-electron-number sector satisfies
Σ_p n_p=N_e.
The occupation representation aligns naturally with qubits because each fermionic mode is either empty or occupied.
6. Fermionic creation and annihilation operators
Define ap† to create an electron in spin orbital p and ap to remove one.
They satisfy canonical anticommutation relations
{a_p,a_q}=0;{a_p†,a_q†}=0;{a_p,a_q†}=δ_pq.
The number operator is
n_p=a_p†a_p.
Because (ap†)²=0, one spin orbital cannot contain two identical electrons.
7. Sign from fermionic ordering
Choose an ordered list of spin-orbital modes.
When ap† acts on an occupation state, moving it past occupied lower-index creation operators produces a sign
(−1)^{Σ_{j<p} n_j}.
This parity string is the same algebraic reason Jordan–Wigner qubit mappings require chains of Z operators.
8. One-electron integrals
For spin orbitals φp, define
h_pq=∫ dx φ_p*(x)[−(1/2)∇²−Σ_A Z_A/|r−R_A|]φ_q(x).
These terms contain kinetic energy and electron–nuclear attraction.
They form a Hermitian matrix: hpq=hqp*.
9. Two-electron integrals
Use the chemists’ unsymmetrised two-electron integral
(pq|rs)=∫∫ dx_1dx_2 φ_p*(x_1)φ_q*(x_2) [1/r_12] φ_r(x_1)φ_s(x_2).
Then one consistent second-quantised Hamiltonian is
H=Σ_pq h_pq a_p†a_q + (1/2)Σ_pqrs (pq|rs)a_p†a_q†a_s a_r + E_NN.
If antisymmetrised integrals ⟨pq||rs⟩ are used instead, the prefactor and index convention change. Mixing these conventions causes factor-of-two and exchange errors.
10. Why second quantisation is efficient conceptually
The first-quantised N-electron wavefunction must enforce antisymmetry explicitly.
Second quantisation moves antisymmetry into the operator algebra. The Fock basis itself is indexed by occupations, while anticommutation ensures correct exchange signs.
This makes particle-number-conserving electronic Hamiltonians compact to write and straightforward to map into qubit operators.
11. Hartree–Fock reference
A Hartree–Fock calculation chooses orbitals so one Slater determinant minimises the mean-field energy.
In the occupation basis, the Hartree–Fock state is simply one computational-basis bitstring after fermion-to-qubit mapping.
Example: with four spin orbitals and two electrons occupying the first two modes,
|HF⟩=|1100⟩
under a direct occupation convention.
Correlation means the true state generally needs a superposition of many such determinants.
12. Configuration-interaction expansion
A correlated state may be written
|Ψ⟩=Σ_I c_I|D_I⟩
over determinants DI.
Full configuration interaction in M spin orbitals and Ne electrons uses
C(M,N_e)=M!/[N_e!(M−N_e)!]
determinants before additional symmetries.
This combinatorial growth motivates approximate classical methods and quantum algorithms.
13. Worked determinant count
With M=20 spin orbitals and Ne=10 electrons,
C(20,10)=184756.
At M=100 and Ne=50, the number is roughly 1029.
Quantum hardware does not remove that Hilbert-space dimension; it represents amplitudes implicitly across qubits instead of storing every coefficient classically.
14. Jordan–Wigner mapping
Map fermionic mode p to qubit p with occupation encoded in Z.
A common convention is
a_p = (Z_0Z_1…Z_{p−1})(X_p+iY_p)/2
and
a_p† = (Z_0Z_1…Z_{p−1})(X_p−iY_p)/2.
The exact ±i sign depends on whether qubit |0⟩ denotes empty and on raising/lowering convention. The parity Z string is essential.
15. Number operator under Jordan–Wigner
Under the common empty=|0⟩, occupied=|1⟩ convention,
n_p=a_p†a_p=(I−Z_p)/2.
Therefore a diagonal orbital-energy term
ε_p n_p
becomes
(ε_p/2)(I−Z_p).
Even simple fermionic terms therefore become sums of Pauli strings.
16. Worked hopping term
For neighbouring modes p and p+1, the Hermitian hopping term
t(a_p†a_{p+1}+a_{p+1}†a_p)
maps, under a standard Jordan–Wigner convention, to
(t/2)(X_pX_{p+1}+Y_pY_{p+1}).
For nonadjacent modes, a chain of Z operators appears between p and q to carry fermionic parity.
17. Qubit Hamiltonian
After mapping every one- and two-body fermion term, the molecular Hamiltonian becomes
H_q=Σ_j α_j P_j
where Pj are tensor products of I,X,Y,Z.
This is the form consumed by VQE, qubitization, Trotter simulation, phase estimation and many measurement strategies.
The Pauli-string count can be large—naively O(M⁴) from two-electron integrals—although symmetry, integral factorisation and low-rank structure can reduce practical cost.
18. Bravyi–Kitaev mapping
Jordan–Wigner stores occupation locally but parity nonlocally, producing O(M)-length parity strings.
Bravyi–Kitaev redistributes occupation and parity information across qubits using a binary-tree structure.
The typical operator-support scaling becomes O(log M) for update/parity information, at the price of a less direct bit interpretation.
No mapping is universally optimal. Circuit architecture, tapering opportunities, measurement grouping and compiler connectivity all matter.
19. Parity mapping and symmetry tapering
Alternative encodings make selected global parities local.
If the Hamiltonian commutes with known Z₂ symmetries such as fixed total particle-number parity or spin-sector parity, corresponding qubits can sometimes be removed after fixing the symmetry eigenvalues.
This is often called qubit tapering by symmetries.
One must verify that the prepared state and all applied ansatz operations remain inside the chosen symmetry sector.
20. Active spaces
Not every molecular orbital contributes equally to the chemistry of interest.
Freeze deeply occupied core orbitals and discard sufficiently high virtual orbitals, then treat a selected active set explicitly.
An active-space model reduces qubit count and Hamiltonian size while attempting to preserve the dominant correlation physics.
The approximation must be chemically validated. A small active space can miss bond breaking, near-degeneracy or transition-metal correlations even if the quantum algorithm itself is exact on that reduced model.
21. Frozen-core energy shift
When core orbitals are frozen at fixed occupation, their one- and two-electron contributions can be absorbed into:
- a scalar core-energy offset;
- modified effective one-electron integrals for active orbitals.
The active Hamiltonian therefore remains a fermionic operator of the same general form but in fewer modes.
22. Basis-set error versus correlation error
Two independent approximations are easy to confuse:
- basis-set error: the finite one-electron basis cannot represent the exact continuum orbitals;
- correlation error: the many-electron method does not represent enough determinants/configurations within the chosen basis.
A quantum computer solving the finite-basis Hamiltonian exactly removes correlation-method error for that model but does not remove basis-set or Born–Oppenheimer error.
23. Symmetries of molecular Hamiltonians
Typical nonrelativistic electronic Hamiltonians conserve total electron number.
Under suitable approximations they can also conserve spin projection Sz and point-group quantum numbers.
Using symmetries can:
- reduce Hilbert-space dimension;
- improve ansatz design;
- enable error detection/symmetry verification;
- remove qubits by tapering.
But spin–orbit coupling and relativistic terms can change the symmetry set.
24. Excitation operators
Starting from a reference determinant, a single excitation replaces occupied orbital i by virtual orbital a:
a_a†a_i.
A double excitation is
a_a†a_b†a_j a_i.
Unitary coupled-cluster-inspired ansätze exponentiate anti-Hermitian combinations such as T−T†.
Guide 14 owns the VQE optimisation loop; here the point is that chemical excitation structure generates physically motivated variational operators.
25. Measurement cost
If
H_q=Σ_j α_jP_j,
then energy is
E=Σ_j α_j⟨P_j⟩.
Each Pauli expectation must be estimated from quantum measurements unless commuting terms can be grouped or more advanced estimators are used.
For chemistry VQE, measurement shot cost can dominate circuit execution. Hamiltonian factorisation, classical shadows, covariance-aware allocation and low-rank decompositions therefore matter alongside qubit count.
26. Phase estimation for chemistry
Fault-tolerant algorithms can instead simulate e−iHt and use quantum phase estimation to extract an eigenenergy.
This shifts cost from repeated Pauli expectation measurement to coherent Hamiltonian simulation, long logical depth and preparation of a state with sufficient overlap onto the target eigenstate.
Neither VQE nor QPE removes the need to construct the molecular Hamiltonian correctly.
27. Resource estimates depend on representation
A headline such as “molecule requires M qubits” is incomplete.
Resource counts depend on:
- basis set and active space;
- fermion-to-qubit mapping;
- symmetry tapering;
- Hamiltonian factorisation;
- target chemical accuracy;
- algorithm and fault-tolerance code;
- state-preparation method.
Comparisons are meaningful only when these choices are aligned.
28. Chemical accuracy is a scale, not a universal theorem
The phrase “chemical accuracy” often refers to roughly 1 kcal/mol, about 1.6 millihartree, for energy differences.
That scale is useful historically but not sufficient for every chemical decision. Spectroscopy, reaction barriers, weak interactions and materials properties can require different tolerances.
Algorithmic precision should be set by the downstream chemical question rather than by a slogan.
29. Worked two-mode mapping
Consider two spin orbitals with Hamiltonian
H=ε_0 n_0+ε_1 n_1+t(a_0†a_1+a_1†a_0).
Under the Jordan–Wigner convention above,
n_0=(I−Z_0)/2;n_1=(I−Z_1)/2;a_0†a_1+a_1†a_0=(X_0X_1+Y_0Y_1)/2.
Therefore
H=((ε_0+ε_1)/2)I−(ε_0/2)Z_0−(ε_1/2)Z_1+(t/2)(X_0X_1+Y_0Y_1).
The fermionic hopping problem has become a sum of five simple Pauli strings.
30. Common misconception: one qubit represents one electron
In occupation encodings, one qubit usually represents one spin orbital’s occupation bit, not one labelled electron. Electron number is the Hamming weight of occupied modes.
31. Common misconception: exact quantum solution means exact molecular energy
An exact solution of the mapped finite-basis Hamiltonian remains approximate relative to the full physical molecule because of basis truncation, Born–Oppenheimer assumptions and omitted relativistic/QED/environmental effects.
32. Common misconception: Jordan–Wigner and Bravyi–Kitaev change chemistry
They are encodings of the same fermionic algebra into qubits. When implemented exactly, they preserve the spectrum. They change operator locality and circuit cost, not the underlying molecular physics.
33. Worked synthesis problem
A calculation uses M=12 spin orbitals and Ne=6 electrons.
Step 1: Direct occupation qubits. Jordan–Wigner requires 12 qubits before tapering.
Step 2: Fixed-electron-number determinant dimension. C(12,6)=924.
Step 3: Compare spaces. Twelve qubits span 4096 computational basis states, but the physical fixed-N sector uses only 924 of them before spin/point-group restrictions.
Step 4: Symmetry use. Fixing known Z₂ symmetries can reduce qubit count, while particle-number-preserving ansätze can avoid exploring much of the unphysical sector.
Step 5: Chemical modelling. If the 12-orbital active space omits an orbital essential to bond breaking, no quantum algorithm on the reduced Hamiltonian can recover that missing physics. The model choice precedes the solver choice.
34. Practice set
- What approximation fixes nuclei while solving the electronic problem?
- Why are Slater determinants used?
- What are the fermionic canonical anticommutation relations?
- Write the finite-basis second-quantised electronic Hamiltonian used here.
- What does the occupation number n_p represent?
- Why does Jordan–Wigner need Z strings?
- How does n_p map to qubits in the stated convention?
- What is the main idea of Bravyi–Kitaev mapping?
- What is an active space?
- What is symmetry tapering?
- Why can VQE measurement cost be large in chemistry?
- Why does exact solution of the finite Hamiltonian not imply exact molecular physics?
Answers
- The Born–Oppenheimer approximation at fixed nuclear geometry.
- They enforce fermionic antisymmetry automatically.
{a_p,a_q}=0,{a_p†,a_q†}=0,{a_p,a_q†}=δ_pq.Σh_pqa_p†a_q+(1/2)Σ(pq|rs)a_p†a_q†a_sa_r+E_NNunder this guide’s integral convention.- Whether spin orbital p is occupied, with eigenvalues 0 or 1.
- The strings reproduce fermionic parity signs when operators on different modes are exchanged.
(I−Z_p)/2.- Distribute occupation/parity/update information through a binary structure to reduce typical operator-support scaling from linear to logarithmic.
- A selected subset of orbitals treated explicitly while core/high virtual orbitals are frozen or removed approximately.
- Fix known commuting symmetry eigenvalues so redundant qubits can be removed.
- The energy is a sum of many Pauli expectations, each requiring statistical estimation unless grouped/factorised.
- Basis-set, Born–Oppenheimer and other physical-model approximations remain.
Sources and further study
[1] Sam McArdle and colleagues, Quantum computational chemistry, Reviews of Modern Physics 92, 015003 (2020). A broad review of electronic-structure encodings and quantum algorithms for chemistry.
[2] James D. Whitfield, Jacob Biamonte and Alán Aspuru-Guzik, Simulation of Electronic Structure Hamiltonians Using Quantum Computers, Molecular Physics 109, 735–750 (2011). An explicit second-quantised-to-qubit treatment.
[3] Jacob T. Seeley, Martin J. Richard and Peter J. Love, The Bravyi–Kitaev transformation for quantum computation of electronic structure, Journal of Chemical Physics 137, 224109 (2012). A detailed comparison of Jordan–Wigner and Bravyi–Kitaev encodings.
[4] Bela Bauer and colleagues, Quantum Algorithms for Quantum Chemistry and Quantum Materials Science, Chemical Reviews 120, 12685–12717 (2020). Resource and algorithm perspectives across chemistry/materials applications.
Continue through Quantum Mathematics
Guide 57: Quantum Amplitude Estimation, Approximate Counting, Monte Carlo and Quadratic Sampling Speedups develops a precision-estimation primitive. Guide 59: Measurement-Based Quantum Computation, Qubit Graph States, Cluster States and Feed-Forward develops a distinct architecture for universal quantum processing. Guide 60: Quantum Causal Models, Process Matrices, Causal Inequalities and Indefinite Causal Order generalises circuit-order assumptions.