Boltzmann counting in Hilbert space

AI summary

Overview

Research area: Quantum physics — foundations of quantum statistical mechanics, quantum state-space geometry, and coarse-grained descriptions of quantum states.

Technical level: Advanced. The paper assumes familiarity with Hilbert spaces, density matrices and partial traces, Haar measures on pure-state manifolds, CPTP channels, and Dirac-delta / Heaviside-function manipulations.

Scope (one sentence): The paper defines a Boltzmann-style entropy for quantum preparations as the logarithm of the Haar-measure volume of pure states compatible with a given constraint, derives closed-form volume expressions for subspace restrictions, fixed expectation values, and coarse-grained subsystem descriptions, and illustrates them with spin, projector, and partial-trace examples.

What This Paper Is About

In classical statistical mechanics a macroscopic preparation picks out a region of phase space, and Boltzmann's entropy is the logarithm of that region's volume. Quantum mechanically, a preparation instead picks out a set of pure states in Hilbert space, but the standard entropies (von Neumann, observational entropy) are defined on density matrices rather than directly on that set. This paper asks whether one can assign both a statistical state and an entropy directly from the set of pure states compatible with a preparation, and answers the entropy half of that question by replacing "number of compatible microstates" with the Hilbert-space volume of compatible pure states.

Key Contributions

1. A geometric ("Boltzmann quantum") entropy for preparations. The authors define $S_Q(\Omega_{\vec{f}}^{\epsilon}) = \log(V_0,|\Omega_{\vec{f}}^{\epsilon}|)$, where $|\Omega_{\vec{f}}^{\epsilon}|$ is the Haar-measure volume of the set of pure states satisfying the preparation constraints to within tolerance $\epsilon$, and $V_0>1$ is an arbitrary scale for the unconstrained volume. Because the constraints are equalities, the compatible sets have zero Haar measure, so the paper introduces $\epsilon$-regularized versions and a general integral formula involving a product of Heaviside step functions.

2. Closed-form volumes for subspace restrictions. For a subspace of dimension $M$ inside a total Hilbert space of dimension $N$, the constraint-manifold measure is $\mathcal{V}_R = \frac{1}{\pi^{N-M}}\frac{(N-1)!}{(M-1)!}$, giving the regularized volume $|\Omega_R^{\epsilon}| \approx \left(\frac{4\epsilon^2}{\pi}\right)^{N-M}\frac{(N-1)!}{(M-1)!}$ with $K = 2(N-M)$ constraints. This is connected to the Venuti–Zanardi distribution: the projected norm $x$ follows a Beta law with parameters $M$ and $N-M$, $\mathcal{V}'_R(x) = \frac{(N-1)!}{(M-1)!(N-M-1)!}x^{M-1}(1-x)^{N-M-1}$.

3. Volumes for fixed expectation values, including a spin example. For three fixed spin expectation values $\langle \psi | J_i/j | \psi \rangle = r_i$ ($i \in [3]$) on an irreducible spin-$j$ block of dimension $D = 2j+1$, the paper reports that $\mathcal{V}_{\vec J}$ depends only on $r = |\vec r|$, is piecewise polynomial with both the number of pieces and the polynomial degree increasing with $D$, and approaches a Gaussian with $\sigma^2 = 1/[3(D-1)]$, obtained from $\overline{r^2} = 1/(D-1)$.

4. A closed-form volume for the partial trace. For a system of dimension $d_S$ and environment of dimension $d_E$, the volume of pure states whose reduced state is $\varrho$ scales as $\mathcal{V}{\Lambda{\text{PT}}}(\varrho) \propto \det[\varrho]^{,d_E - d_S}$, which is maximal for the maximally mixed state and vanishes for states approaching rank deficiency.

Main Findings

• Volume replaces counting. Because compatible pure states need not be orthogonal, the number of states is replaced by a volume in state space; the volume is normalized to lie between 0 and 1, with maximum 1 when no constraint is imposed, and the entropy is $\log(V_0)$ in that unconstrained case. Only densities of states matter for entropy differences at fixed constraint number, so the precise size of $\epsilon$ and the value of $V_0$ drop out.

• Thin-shell approximation. For $\epsilon \ll 1$, $\Theta(\epsilon - |x|) \sim 2\epsilon,\delta(x)$, so $|\Omega_{\vec f}^{\epsilon}| \approx (2\epsilon)^K \int d\psi \prod_{i=1}^{K}\delta(f_i(\psi))$: the product of deltas measures the compatible surface and $(2\epsilon)^K$ is the shell thickness.

• Subspace restriction scaling. The number of real equality constraints is $2(D-D_R)$ in the notation of the preparation section; the volume falls off rapidly as the co-dimension $N-M$ grows, which the authors note makes numerical rejection sampling impractical (shown in their Fig. 2 for two values of co-dimension).

• Two inequivalent descriptions of the same subspace. $\mathcal{V}_R$ is not the $x \to 1^-$ limit of $\mathcal{V}'_R(x)$; the latter limit is 0. The paper attributes this to the two constraints inducing different measures (a point-localizing delta versus a radial delta in the orthogonal sector) and refers to Appendix C for when two constraint descriptions of a compatible set are equivalent, tied to a critical value of $\langle \psi | \Pi_R | \psi \rangle$.

• Gaussianity of spin-vector volumes. In Fig. 1 the exact radial profiles for $j = 3/2, 4/2, 5/2, 6/2, 11/2$ (i.e., $D = 4, 5, 6, 7, 12$) are compared with a Gaussian of width $\sigma^2 = 1/[3(D-1)]$; the fit improves as $D$ grows.

• Partial-trace volume is a determinant law. Distinguishability of the environment and system dimensions controls the behavior: when $d_E = d_S$ all full-rank states have the same volume of uncertainty; since realistic environments have $d_E > d_S$, the determinant factor is the relevant one. The authors argue that almost rank-deficient (e.g., pure) reduced states would require infinite resources to prepare to arbitrary precision, citing Refs. [22–27].

• Numerical support for the determinant law. Random local qutrit states were tested by sampling 100,000 microscopic states from $\mathcal{H}_3 \otimes \mathcal{H}_E$ and counting those satisfying $|\mathrm{tr}(|\psi\rangle\langle\psi|,g_i \otimes \mathbb{1}_E) - \mathrm{tr}(\varrho, g_i)| < \epsilon$ for all $i \in [8]$ Gell-Mann matrices with $\epsilon = 0.2$. For $d_E = 3$ the volume is constant, for $d_E = 4$ it grows linearly with $\det(\varrho)$; the sentence describing the $d_E = 5$ case is cut off in the available text ("grows quadrat…"), so the full statement is not reported here.

• Relation to prior entropy notions. The construction is positioned as complementary to observational entropy (Refs. [4–7]), which is defined relative to a chosen coarse-graining, and as related in spirit to the purification-volume approach of Ref. [3], but it assigns volume directly to states compatible with a preparation rather than to purifications of a reduced state. It also complements the authors' earlier average assignment map (Refs. [1, 2]), which selects a statistical state as the convex combination of compatible pure states and does not generally agree with the maximum-entropy assignment.

• Gap in the literature. The authors state that, to the best of their knowledge, there are no general results for the volume of sets defined by several simultaneous fixed expectation values; the spin example is presented as an illustrative tractable case.

Methodology in Plain English

The authors proceed in three steps. First, they write down, for each kind of preparation, the exact set of pure states consistent with it: states living in a subspace, states with prescribed expectation values for one or several observables, and states whose image under a coarse-graining (dimension-reducing CPTP) map equals a target reduced state; they show the third case is a special case of the second via $d^2-1$ constraints built from an orthonormal traceless operator basis. Second, because exact equalities define zero-measure sets under Haar-random sampling, they thicken each set by a tolerance $\epsilon$ and define its volume as the probability that a Haar-random pure state satisfies all constraints — a quantity computable as an integral of a product of step functions, and, for small $\epsilon$, as a product of delta functions times $(2\epsilon)^K$. Third, they evaluate these integrals in closed form where possible, using rotational symmetry (to reduce the spin problem to a radial profile), unitary invariance of the Haar measure (to diagonalize the target reduced state), and a rescaling of the coefficient vectors by $\sqrt{\lambda_i}$ that pulls out powers of $\det[\varrho]$. Numerical rejection sampling is used as a cross-check, and the paper notes that regularization is also what makes such sampling well defined.

Why This Matters

Impact on research. The work gives a preparation-level, geometric counterpart to density-matrix entropies, supplying an explicit quantitative meaning to "quantum ignorance" about the microscopic state given a macroscopic or coarse-grained description. It also collects and unifies scattered results on the geometry of expectation-value-constrained state manifolds (Brody et al., Venuti–Zanardi, Müller–Gross–Eisert, Bartsch–Gemmer, Reimann) under a single volume-based framework, and it clarifies a subtle pitfall: different constraint parameterizations of the same set can carry different measures and hence different volumes.

Potential applications (the paper itself does not develop applied use cases; these are plausible implications of the formalism):

• Quantifying how much microscopic information a tomographic or coarse-grained preparation actually fixes, e.g., in state-preparation and benchmarking protocols.
• Bounding the resources needed to prepare nearly pure or nearly rank-deficient states, in line with the paper's argument that such preparations require infinite resources at finite precision.
• Assessing how much freedom remains in the global state when only a subsystem is controlled, which is the standard setting of open quantum systems.
• Characterizing the action of imperfect measurement maps (mentioned in the abstract as an example; the corresponding calculation lies beyond the truncated text available here).

Industry relevance. The framework is foundational rather than device-specific, but its ingredients — partial trace, CPTP coarse-graining, finite-precision preparation, and Haar-random state sampling — are standard in quantum information, quantum control, and the validation of quantum hardware, where knowing the size of the state set consistent with a calibration or readout record is directly useful.

Future Directions

• Multivariate central-limit behavior for the spin vector. The authors conjecture that if higher-order joint cumulants of $\vec r = (x,y,z)$ decay fast enough with Hilbert-space dimension, $\mathcal{V}(x,y,z)$ becomes asymptotically Gaussian, and they explicitly leave this calculation for a future publication.

• Volumes for general multiple fixed expectation values. Since no general results exist for $\Omega_{\vec a}$ with several observables, deriving generic scaling laws — including the dimension reduction per constraint and the conditions under which the compatible set becomes empty — remains open.

• Equivalence classes of constraint descriptions. Appendix C is said to give the general criterion for when two constraint descriptions define the same compatible set with the same measure; extending this into a systematic calculus of preparation constraints would clarify when volumes are comparable across protocols.

• The imperfect detector map and combined preparations. The abstract lists an imperfect detector map among the worked examples and mentions combinations of the three constraint types; these extensions are not covered in the available text and are natural next steps.

Target Audience

Advanced graduate students and researchers in quantum information and quantum statistical mechanics, particularly those working on foundations of statistical mechanics, typicality and concentration of measure in high-dimensional Hilbert spaces, coarse-graining and open quantum systems, and the geometry of quantum state space. Readers specifically interested in alternatives or complements to von Neumann and observational entropy will find the most value, while beginners will need substantial background in Hilbert-space measure theory to follow the derivations.

Authors’ abstract

We introduce a geometric entropy for quantum preparations, defined as the logarithm of the Hilbert-space volume of pure states compatible with a given set of constraints. This construction extends Boltzmann's counting perspective to the quantum setting, where compatible states need not be orthogonal and the relevant notion of "number of states" is naturally replaced by a volume in state space. We analyze three classes of constraints: restriction to a subspace, fixed expectation values, and coarse-grained subsystem descriptions. For representative examples, including subspace projection, spin expectation values, partial trace, and an imperfect detector map, we obtain explicit scaling laws and closed-form expressions for the associated volumes. The resulting framework provides a geometric measure of quantum ignorance at the level of the preparation and complements entropy notions based on density matrices and coarse graining.