Quantum de Finetti Theorems
The bosonic quantum de Finetti theorem represents an infinitely extendible hierarchy of symmetric density matrices as a mixture of pure tensor powers. A finite-dimensional quantitative version controls the error for fixed marginals of a large finite system. This is a representation of correlations by a probability measure; it does not assert that every bosonic state is a single condensate.
Required background. Symmetric tensor products supplies the bosonic subspace. Partial trace supplies consistency of marginals. Mean-field limits gives the dynamical use of factorized hierarchies.
Consistent bosonic hierarchies are mixtures of powers
Section titled “Consistent bosonic hierarchies are mixtures of powers”Let be a separable complex Hilbert space. Suppose that for every , is a positive trace-one operator supported on , and
The strong bosonic de Finetti theorem states that there is a unique probability measure on unit one-particle rays such that
The integral is a trace-class Bochner integral. Equivalently, use the unique phase-invariant measure on the unit sphere. Phase must be quotiented or averaged because all displayed projectors are unchanged by .
Consistency for every order is essential. A single finite entangled state need not be a mixture of powers. Also, support on the symmetric subspace is stronger than merely commuting with permutations. General exchangeable hierarchies use mixtures of powers of one-particle density matrices; the bosonic support restriction reduces them to pure one-particle projectors.
A finite-dimensional approximation has an explicit measure
Section titled “A finite-dimensional approximation has an explicit measure”Now let , and let be any trace-one state on . Denote its marginals by . Let be normalized unitary-invariant measure on the unit sphere and define
This is a probability measure. Indeed, unitary invariance and irreducibility of the symmetric-power representation give
Taking the trace against proves normalization.
The quantitative theorem gives, for ,
Here , without the factor sometimes used for trace distance. This bound is the convenient consequence (2.7) of Lewin, Nam, and Rougerie’s Theorem 2.1. Their proof, based on finite symmetric-power identities and positivity, is imported; the exact measure and error dependence are part of the statement.
The bound is useful at fixed with large. It does not tend to zero if the dimension or observed particle count grows too rapidly with .
Compactness turns the finite estimate into the infinite theorem
Section titled “Compactness turns the finite estimate into the infinite theorem”In finite dimension, apply the quantitative theorem to . Consistency ensures that its -body marginal is exactly whenever .
The unit-ray space is compact. Extract a weakly convergent subsequence of the probability measures , with limit . For each fixed , matrix elements of are continuous functions of the ray. The integral projectors therefore converge in finite-dimensional trace norm. The error bound tends to zero, proving the representation.
For uniqueness, the hierarchy determines every polynomial moment in the matrix entries of . These functions form a self-adjoint algebra containing constants and separating rays. Stone–Weierstrass makes them dense in the continuous functions on the compact ray space, so two representing measures must agree.
The separable infinite-dimensional theorem requires additional localization and tightness arguments; compactness of the unit sphere is no longer available. The stated trace-one consistency hypotheses still give the sphere representation, as explained in the cited treatment.
A rigorous mean-field ground-energy consequence
Section titled “A rigorous mean-field ground-energy consequence”In finite dimension take the Hamiltonian from mean-field limits, with one-particle and bounded exchange-symmetric pair operator . For any bosonic ,
Define the Hartree energy
The same measure approximates both marginals. Trace duality and the finite theorem imply
A pure tensor trial state gives the opposite variational upper bound with no error. Thus the ground energy per particle converges to the Hartree minimum, with the displayed finite-dimensional estimate. The proof does not require that the minimizer be unique or that all approximate ground states select one pure condensate.
For infinite-dimensional unbounded Hamiltonians, energy localization and control of the discarded modes are additional steps. One cannot insert into the finite bound.
Exercises
Section titled “Exercises”A cat state has mixed limiting marginals. Let be orthogonal unit vectors and . Find its fixed- marginal for .
Solution
Every cross term contains the factor . Hence the marginal is one half of each pure tensor projector. The de Finetti measure is , not one point mass. Bosonic symmetry alone does not establish complete condensation into a specified orbital.
Loss of mass in infinite dimension. Let be orthonormal and take . What do fixed marginals converge to against compact operators?
Solution
For each , the vectors are orthonormal, so the expectations of compact operators tend to zero. The weak-star trace-class limit is the zero operator, despite every marginal having trace one. There is no trace-norm convergence and no trace-one consistent limit hierarchy to which the strong theorem applies. Additional compactness or energy confinement is needed to prevent escape into new modes.
References
Section titled “References”- M. Lewin, P. T. Nam, and N. Rougerie, “Remarks on the quantum de Finetti theorem for bosonic systems,” arXiv:1310.2200, version 3, 2014, especially Theorem 2.1 and Section 1, author text.