Quantum many-body mathematics studies systems in which simple local degrees of freedom combine into a Hilbert space whose dimension grows exponentially. The hard part is not writing the Hamiltonian. It is extracting stable collective structure from an object too large to represent naively.
A chain of N qubits has Hilbert-space dimension 2N. At N=50 that dimension already exceeds 1015. Yet physically meaningful Hamiltonians are often local: each term acts on only one site, neighbouring sites or a small bounded region. Locality, symmetry, spectral gaps, correlation decay and entanglement structure create the compression that makes many-body physics intelligible.
This guide develops local Hamiltonians, ground-state expectation values, connected correlation functions, spectral gaps, order parameters, spontaneous symmetry breaking, quantum critical points, finite-size scaling, correlation length and locality bounds. It complements Guide 16 on tensor networks: here the owner is the physics and mathematics of phases; tensor networks remain one computational representation for selected states.
Local interactions → global ground state → correlations and entanglement → gap and length scales → phase structure → critical scaling.
1. Exponential Hilbert-space growth
Suppose each of N sites has local dimension d. The full Hilbert space is
H_total=⊗_{j=1}^N H_j
with dimension
dim H_total=d^N.
For spin-1/2 particles, d=2. Exact state-vector storage therefore scales as 2N complex amplitudes before symmetry reductions.
The many-body problem is not difficult because local laws are complicated. It is difficult because composition creates a vast configuration space.
2. Local Hamiltonians
A k-local Hamiltonian has the form
H=Σ_X h_X
where each hX acts nontrivially on at most k sites.
Geometric locality is stronger: hX couples sites that are nearby in a physical lattice or interaction graph.
Example: a nearest-neighbour spin chain may use
H=JΣ_{j=1}^{N-1} Z_jZ_{j+1}+hΣ_{j=1}^{N} X_j.
Only one- and two-site operators appear, yet the eigenstates live in the full 2N-dimensional space.
3. Ground-state energy
The ground-state energy is the smallest eigenvalue E₀ of H:
H|ψ₀⟩=E₀|ψ₀⟩.
The Rayleigh–Ritz principle gives
E₀=min_{||ψ||=1}⟨ψ|H|ψ⟩.
This connects many-body ground-state physics to optimisation. Guide 14’s variational eigensolver machinery is one algorithmic implementation of the same mathematical principle.
4. Product-state upper bound
Any trial state produces an upper bound on E₀.
For the transverse-field Ising Hamiltonian
H=-JΣ Z_jZ_{j+1}-hΣ X_j
take the product state |0…0⟩. Then each ⟨Z_jZ_{j+1}⟩=1 while each ⟨X_j⟩=0. For open boundaries,
⟨H⟩=−J(N−1).
Therefore E₀≤−J(N−1).
The bound can be crude because the true ground state may entangle sites and exploit both interaction terms simultaneously.
5. Energy density
For extensive systems it is often better to study energy per site
e₀(N)=E₀(N)/N.
If the thermodynamic limit exists,
e₀=lim_{N→∞}E₀(N)/N.
Boundary energies are usually O(1), so dividing by N removes them in the large-system limit.
6. Expectation values are local probes of a global state
For observable O, ground-state expectation is
⟨O⟩₀=⟨ψ₀|O|ψ₀⟩.
A local magnetisation may be m_z=(1/N)Σ_j⟨Z_j⟩. A bond energy may be ⟨Z_jZ_{j+1}⟩.
Different phases can share similar energy density while differing sharply in symmetry, correlations or topology. One scalar energy is not a complete phase label.
7. Two-point correlation functions
For local observables Ai and Bj, define
C_AB(i,j)=⟨A_iB_j⟩.
The connected correlator subtracts independent one-point contributions:
C_conn(i,j)=⟨A_iB_j⟩−⟨A_i⟩⟨B_j⟩.
If the distant sites become statistically independent in the state, the connected correlator tends to zero.
8. Worked product-state correlation
Take |0…0⟩ and choose A=Zi, B=Zj.
⟨Z_i⟩=1;⟨Z_j⟩=1;⟨Z_iZ_j⟩=1.
Therefore
C_conn=1−1·1=0.
The raw two-point expectation is large, but it contains no connected fluctuation correlation. This is why subtracting the product of means matters.
9. GHZ state shows long-range connected correlation
For
|GHZ⟩=(|0…0⟩+|1…1⟩)/√2,
each local ⟨Z_i⟩=0, but for any i≠j,
⟨Z_iZ_j⟩=1.
Thus C_conn(i,j)=1 independent of separation. The state has long-range order even though each one-site magnetisation vanishes in the symmetry-preserving superposition.
10. Correlation length
In many gapped phases, connected correlations decay approximately exponentially:
C_conn(r)∼e^{-r/ξ}
possibly multiplied by oscillatory or power-law factors.
ξ is the correlation length. It gives the characteristic distance over which local fluctuations remain strongly linked.
At many continuous quantum critical points, ξ diverges as a control parameter approaches its critical value.
11. Spectral gap
Let E₀≤E₁≤… be Hamiltonian eigenvalues. The finite-size spectral gap is
Δ_N=E₁−E₀
when the ground state is unique. With exact ground-state degeneracy, one defines the gap to the first excitation above the entire ground-state manifold.
A gapped phase has a nonzero thermodynamic excitation gap under the appropriate boundary and sector conventions. At a quantum critical point, the gap often closes as system size grows.
12. Gap and response timescale
Energy gaps set natural dynamical timescales through
τ∼1/Δ
in ℏ=1 units.
A small gap means slow adiabatic evolution and enhanced low-frequency response. Guide 20’s adiabatic analysis depends on this same gap structure.
The inverse-gap rule is an order-of-magnitude relation, not a universal relaxation formula for every open system.
13. Order parameters
An order parameter is an observable designed to distinguish phases by symmetry or structure.
For a ferromagnet, one may use magnetisation
m_z=(1/N)Σ_j⟨Z_j⟩.
In a finite system with exact Z₂ symmetry, the unique ground state may satisfy mz=0 even in the ordered regime. One then studies symmetry-broken boundary conditions, absolute magnetisation, long-range correlators or the thermodynamic limit taken in the correct order.
Finite-size symmetry can therefore hide spontaneous order if the observable is interpreted too literally.
14. Spontaneous symmetry breaking
A Hamiltonian can be symmetric while its infinite-system ground states are not.
For example, a Z₂-symmetric ferromagnet can possess two thermodynamic ground states with opposite magnetisations ±m. The Hamiltonian is invariant under flipping all spins, but each chosen phase selects one orientation.
In finite volume, quantum tunnelling can combine these into symmetric and antisymmetric states separated by an exponentially small splitting. The order of limits—system size versus symmetry-breaking field—matters.
15. Quantum phase transition
A zero-temperature quantum phase transition occurs when the ground-state structure changes nonanalytically as a Hamiltonian parameter g crosses a critical value gc.
The control parameter changes the Hamiltonian itself rather than thermal occupation.
For
H(g)=H_0+gV,
one studies E₀(g), derivatives of E₀, order parameters, correlation length, gap and entanglement as g varies.
16. First-order versus continuous transitions
At a first-order quantum transition, competing phases exchange stability and the first derivative of ground-state energy can be discontinuous in the thermodynamic limit.
At a continuous transition, lower derivatives may remain smooth while correlation length diverges and scale invariance emerges.
Finite systems have analytic eigenvalues except at exact symmetry-protected crossings, so numerical phase-transition signatures are rounded. Finite-size scaling is needed to infer the thermodynamic singularity.
17. Critical exponents
Near a continuous critical point, correlation length often scales as
ξ∼|g−g_c|^{-ν}.
The gap scales as
Δ∼ξ^{-z}∼|g−g_c|^{zν}
where ν is the correlation-length exponent and z is the dynamical critical exponent.
An order parameter can scale as m∼|g−g_c|^β on the ordered side.
These exponents organise universality classes.
18. Universality
Microscopically different systems can share the same long-distance critical exponents and scaling functions.
The universal behaviour depends on coarse features such as spatial dimension, symmetry, range of interactions and order-parameter structure rather than every lattice-scale coupling.
This is one of the deepest compression principles in many-body physics: near criticality, many microscopic details become irrelevant in the renormalisation-group sense.
19. Finite-size scaling
At criticality an infinite correlation length cannot fit inside a finite system. The system size L becomes the largest scale.
A generic finite-size scaling form for observable O is
O(g,L)=L^{-x_O}F((g−g_c)L^{1/ν}).
If the correct gc, ν and scaling dimension xO are used, data from different L values can collapse onto one scaling curve F.
Data collapse is evidence, not proof. Corrections to scaling and overfitting can produce misleading visual agreement.
20. Worked gap scaling
Suppose a one-dimensional critical model has z=1 and finite-size gaps
- L=20: Δ≈0.157;
- L=40: Δ≈0.0785;
- L=80: Δ≈0.0393.
Doubling L approximately halves Δ, consistent with
Δ_L∝L^{-1}.
A log–log fit of Δ versus L would have slope approximately −1.
A few sizes cannot establish asymptotic scaling by themselves; irrelevant operators and boundaries can distort small systems.
21. Susceptibility from derivatives
If H(g)=H₀−gM, the Hellmann–Feynman theorem gives
dE₀/dg=−⟨M⟩
for a nondegenerate differentiable ground state.
A second derivative is related to response susceptibility and can become large near a continuous transition.
This gives a practical bridge between energy curves and order-parameter response.
22. Structure factor
Spatial correlations can be transformed into momentum space. For spin observable Z, define a static structure factor
S(k)=(1/N)Σ_{i,j}e^{ik(r_i-r_j)}⟨Z_iZ_j⟩.
A peak at k=0 signals ferromagnetic ordering; a peak near k=π in a one-dimensional bipartite lattice signals antiferromagnetic alternation.
Fourier space converts real-space correlation patterns into order-wavevector signatures.
23. Entanglement entropy
Split the many-body system into region A and complement B. For a pure ground state, define
S_A=−Tr(ρ_A log ρ_A).
In many gapped local ground states, entanglement scales with the boundary area of A rather than its volume.
In one-dimensional critical systems described by conformal field theory, interval entropy grows logarithmically with interval size rather than saturating.
This difference helps explain why matrix-product-state methods are powerful for many one-dimensional gapped systems yet become more demanding near criticality.
24. One-dimensional area law
Hastings proved an area law for entanglement entropy of one-dimensional gapped local Hamiltonians under standard finite-range assumptions. [2]
In one dimension, the boundary of an interval has constant size, so the entropy is bounded independently of interval length in the gapped setting, up to model-dependent constants.
The theorem does not say every gapped state in every dimension is trivial, nor that every tensor-network bond dimension needed in practice is small.
25. Lieb–Robinson bounds
Local lattice Hamiltonians do not transmit operator influence instantaneously across arbitrary distances.
Lieb and Robinson proved that for local observables A and B separated by distance r, a commutator under Heisenberg time evolution obeys a bound schematically like
||[A(t),B]|| ≤ C exp[-μ(r−v|t|)].
Outside an effective light cone r≈v|t|, the commutator is exponentially suppressed. [3]
This emergent velocity is not necessarily the speed of light. It is a bound set by lattice interactions and norms.
26. Correlation decay from a gap
For many local Hamiltonians with a unique gapped ground state, exponential clustering theorems connect a nonzero gap and locality to exponentially decaying connected correlations.
This explains the frequent pairing:
- gapped phase → finite correlation length;
- critical gap closing → diverging correlation length and power-law correlations.
There are important qualifications involving degeneracy, topological order and long-range interactions. The implication should always be stated with its hypotheses.
27. Topological phases need nonlocal diagnostics
Not every phase transition is captured by a local Landau order parameter.
Topologically ordered phases can share all local symmetry properties yet differ in ground-state degeneracy on nontrivial manifolds, anyonic excitations or long-range entanglement.
Guide 26 introduced surface-code topology from an error-correction perspective. In many-body physics, the same toric-code Hamiltonian is a paradigmatic topologically ordered phase.
28. Frustration
A Hamiltonian H=Σh_X is frustration free when a ground state minimises every local term simultaneously.
In a frustrated system, local energy preferences compete: no global state can minimise all hX at once.
Geometric frustration appears, for example, when antiferromagnetic interactions are placed around an odd cycle.
Frustration is not synonymous with computational hardness, but it often enriches the phase diagram and complicates variational optimisation.
29. Worked triangular frustration
Take three classical Ising spins si∈{±1} with antiferromagnetic energy
H=J(s₁s₂+s₂s₃+s₃s₁), J>0.
If two neighbouring bonds are satisfied with opposite spins, the third pair must be equal and therefore unsatisfied.
The best possible bond values are two −1 terms and one +1 term, giving
E_min=J(-1-1+1)=−J.
No configuration reaches −3J. The obstruction is geometric.
30. Common misconception: a phase is just one ground-state vector
A phase is an equivalence class of systems connected without closing the relevant gap or changing the defining structure, subject to the symmetry and locality conditions being considered. One finite-size eigenvector is only one representative.
31. Common misconception: zero finite-size magnetisation means no ferromagnetic order
Exact symmetry can force finite-system magnetisation to zero. Long-range correlations, nearly degenerate symmetry sectors and the thermodynamic limit can still reveal spontaneous order.
32. Common misconception: a small gap at one system size proves criticality
Finite-size gaps can be small because of boundary modes, symmetry splitting or accidental near-degeneracy. Criticality is inferred from systematic size scaling and other observables, not one number.
33. Worked synthesis problem
A one-dimensional model is simulated at sizes L=16,32,64 near a suspected critical coupling gc. The lowest excitation gaps are 0.196,0.0985,0.0494.
Step 1: Ratio test. Each doubling of L approximately halves the gap.
Step 2: Candidate exponent. This is consistent with Δ_L∝L^{-z} and z≈1.
Step 3: Correlation test. If connected correlations also cross from exponential decay away from gc to approximate power-law decay at gc, the critical interpretation strengthens.
Step 4: Order parameter. A symmetry-breaking observable should be checked across the same size sequence with an appropriate finite-size scaling form.
Step 5: Conclusion. The gap data are consistent with critical scaling but do not by themselves determine universality class. Additional exponents, symmetries and correlation structure are required.
34. Practice set
- How does Hilbert-space dimension scale for N local d-level systems?
- What is a k-local Hamiltonian?
- State the Rayleigh–Ritz formula for E₀.
- Why is connected correlation different from raw two-point expectation?
- Define the finite-size spectral gap.
- What does correlation length measure?
- Why can finite symmetric systems hide spontaneous order?
- What distinguishes a quantum phase transition from a thermal one?
- Write the common critical scaling of ξ.
- What does finite-size scaling replace the divergent correlation length with?
- What does a Lieb–Robinson bound constrain?
- Why can topological phases evade local order-parameter classification?
Answers
d^N.- A sum of terms each acting nontrivially on at most k sites.
E₀=min_{||ψ||=1}⟨ψ|H|ψ⟩.- It subtracts the product of one-point means and isolates correlated fluctuations.
- The energy difference from the ground-state manifold to the first excitation above it.
- The characteristic distance over which connected correlations remain appreciable.
- The exact finite-size eigenstate can preserve the Hamiltonian symmetry even when the infinite system has symmetry-broken phases.
- It is driven by changing a Hamiltonian/control parameter at zero temperature rather than by thermal population.
ξ∼|g−g_c|^{-ν}.- The finite system length L.
- The rate at which local operator influence can spread through a local lattice system.
- They can differ by global entanglement and excitation structure even when all local symmetry order parameters agree.
Sources and further study
[1] Subir Sachdev, Quantum Phase Transitions, Cambridge University Press. A standard reference for quantum criticality, scaling and many-body phases.
[2] Matthew B. Hastings, An Area Law for One Dimensional Quantum Systems. The one-dimensional gapped-ground-state entanglement area-law theorem.
[3] Elliott H. Lieb and Derek W. Robinson, The finite group velocity of quantum spin systems, Communications in Mathematical Physics 28, 251–257 (1972). The foundational locality/light-cone bound for lattice quantum systems.
[4] Matthew B. Hastings and Tohru Koma, Spectral Gap and Exponential Decay of Correlations. A rigorous connection between gaps, locality and correlation decay under stated assumptions.
Continue through Quantum Mathematics
Guide 46: Quantum Spin Chains, Jordan–Wigner Transformation, Free Fermions and Criticality solves an important many-body family by mapping spins to fermions. Guide 47: Quantum Chaos, Random Matrices, Level Statistics and Scrambling studies complex spectral and dynamical behaviour away from integrability. Guide 48: Quantum Thermodynamics, Gibbs States, Free Energy, Work and Resource Theories turns many-body energy and entropy into thermodynamic resources.