Quantum Value Functionals

• Quantum value functionals are mathematical constructions that assign values to quantum states, tensors, and operators to quantify optimality and performance.
• They are applied in variational mechanics, tensor asymptotics, scoring rules, and quantum control, bridging theory with practical quantum estimation and algorithm design.
• Their formulations use convex analysis, spectral invariants, and semidefinite programming to transform complex quantum data into actionable scalar or structured outputs.

Quantum value functionals are a heterogeneous class of constructions that assign values to quantum states, observables, tensors, or stochastic objects in order to encode optimality, admissibility, or truthful evaluation. In the recent literature, the phrase appears in several technically distinct senses: as a stationary action functional on wave functions, as an entropy-optimized spectral invariant of tensors, as a convex spectral potential on density operators generating proper scoring rules, as an estimand for nonlinear quantum algorithms, and as a linear or lattice-valued value map in quantum control and quantum probability (Knorst et al., 2024, Sakabe et al., 29 Jan 2026, AlMasri, 6 May 2026, Chen et al., 22 May 2025, Saldi et al., 2024, Doering et al., 2012). The shared theme is not a single canonical definition but the use of functionals to convert quantum structure into a scalar, operatorial, or order-theoretic object that can be optimized, compared, or interpreted probabilistically.

1. Variational quantum value functionals in stationary mechanics

A direct variational use of the term appears in "On the quantum Guerra-Morato Action Functional" (Knorst et al., 2024). For a smooth potential $W:\mathbb T^n\to\mathbb R$, the functional is

$$I(\psi)=\int_{\mathbb T^n}\left(\frac{D v\, D v^*}{2}(x)-W(x)\right)a(x)^2\,dx,$$

with

$$\psi=a\,e^{i u/\hbar},\qquad u=\frac{v+v^*}{2},\qquad a=e^{(v^*-v)/(2\hbar)},$$

and normalization $\int a^2\,dx=1$. The paper treats this as a quantum value functional in the variational sense because it assigns a scalar cost or value to each admissible wave function and studies critical points under constraints.

The central constraint is

$\operatorname{div}(a^2 Du)=0,$

interpreted as zero flux. In this formulation the current is $j=a^2Du$, so $\operatorname{div}(j)=0$ expresses stationary conservation of probability flow. Under criticality, the amplitude $a$ and phase $u$ satisfy a Hamilton–Jacobi equation with quantum potential,

$$-\frac{\hbar^2}{2m}\,\frac{\Delta a}{a}+\frac{|Du|^2}{2m}+W=E,$$

equivalently

$$I(\psi)=\int_{\mathbb T^n}\left(\frac{D v\, D v^*}{2}(x)-W(x)\right)a(x)^2\,dx,$$0

with

$$I(\psi)=\int_{\mathbb T^n}\left(\frac{D v\, D v^*}{2}(x)-W(x)\right)a(x)^2\,dx,$$1

The same critical points also satisfy the stationary Schrödinger-type eigenvalue equation

$$I(\psi)=\int_{\mathbb T^n}\left(\frac{D v\, D v^*}{2}(x)-W(x)\right)a(x)^2\,dx,$$2

The paper’s main structural result is the second variation formula at a critical point: $$I(\psi)=\int_{\mathbb T^n}\left(\frac{D v\, D v^*}{2}(x)-W(x)\right)a(x)^2\,dx,$$3 Its sign determines whether the critical point is a local minimum, local maximum, or saddle. The paper explicitly remarks that the second variation may be positive or negative, and notes special cases in which it is negative, including $$I(\psi)=\int_{\mathbb T^n}\left(\frac{D v\, D v^*}{2}(x)-W(x)\right)a(x)^2\,dx,$$4 and the case where $$I(\psi)=\int_{\mathbb T^n}\left(\frac{D v\, D v^*}{2}(x)-W(x)\right)a(x)^2\,dx,$$5 is constant. This makes the functional a genuine local stability criterion rather than merely a device for recovering Euler–Lagrange equations.

A second constraint,

$$I(\psi)=\int_{\mathbb T^n}\left(\frac{D v\, D v^*}{2}(x)-W(x)\right)a(x)^2\,dx,$$6

introduces prescribed average momentum or flux. With the Bloch-type decomposition $$I(\psi)=\int_{\mathbb T^n}\left(\frac{D v\, D v^*}{2}(x)-W(x)\right)a(x)^2\,dx,$$7, the variational problem becomes coupled to a parameter $$I(\psi)=\int_{\mathbb T^n}\left(\frac{D v\, D v^*}{2}(x)-W(x)\right)a(x)^2\,dx,$$8 and leads to the dual eigenvalue problems

$$I(\psi)=\int_{\mathbb T^n}\left(\frac{D v\, D v^*}{2}(x)-W(x)\right)a(x)^2\,dx,$$9

$$\psi=a\,e^{i u/\hbar},\qquad u=\frac{v+v^*}{2},\qquad a=e^{(v^*-v)/(2\hbar)},$$0

with $$\psi=a\,e^{i u/\hbar},\qquad u=\frac{v+v^*}{2},\qquad a=e^{(v^*-v)/(2\hbar)},$$1 and $$\psi=a\,e^{i u/\hbar},\qquad u=\frac{v+v^*}{2},\qquad a=e^{(v^*-v)/(2\hbar)},$$2. Via the Cole–Hopf transforms, the paper derives eikonal-type equations, the transport equation $$\psi=a\,e^{i u/\hbar},\qquad u=\frac{v+v^*}{2},\qquad a=e^{(v^*-v)/(2\hbar)},$$3, and an averaged identity showing that the effective Hamiltonian $$\psi=a\,e^{i u/\hbar},\qquad u=\frac{v+v^*}{2},\qquad a=e^{(v^*-v)/(2\hbar)},$$4 arises as a variational spectral quantity. In this setting, a quantum value functional is thus a stationary action principle whose first variation yields quantum Hamilton–Jacobi and spectral equations, while the second variation governs local optimality.

2. Entropic quantum functionals in tensor asymptotics

A different meaning of quantum functional is developed in "Strassen's support functionals coincide with the quantum functionals" (Sakabe et al., 29 Jan 2026). For a $$\psi=a\,e^{i u/\hbar},\qquad u=\frac{v+v^*}{2},\qquad a=e^{(v^*-v)/(2\hbar)},$$5-tensor

$$\psi=a\,e^{i u/\hbar},\qquad u=\frac{v+v^*}{2},\qquad a=e^{(v^*-v)/(2\hbar)},$$6

the quantum functional is

$$\psi=a\,e^{i u/\hbar},\qquad u=\frac{v+v^*}{2},\qquad a=e^{(v^*-v)/(2\hbar)},$$7

where $$\psi=a\,e^{i u/\hbar},\qquad u=\frac{v+v^*}{2},\qquad a=e^{(v^*-v)/(2\hbar)},$$8 is the entanglement polytope of $$\psi=a\,e^{i u/\hbar},\qquad u=\frac{v+v^*}{2},\qquad a=e^{(v^*-v)/(2\hbar)},$$9 and $\int a^2\,dx=1$0 are Shannon entropies. Christandl–Vrana–Zuiddam introduced these as universal spectral points: they satisfy Strassen’s axioms of the asymptotic spectrum of tensors, including monotonicity under restriction, multiplicativity under tensor product, additivity under direct sum, and normalization on unit tensors.

The entanglement polytope is

$\int a^2\,dx=1$1

so $\int a^2\,dx=1$2 is an entropy-maximization problem over local spectra in orbit closure. This geometric construction contrasts with Strassen’s support functional, which depends on the support polytope

$\int a^2\,dx=1$3

and the basis-dependent optimization

$\int a^2\,dx=1$4

The main theorem states that these two quantities coincide: $\int a^2\,dx=1$5 This resolves Strassen’s 1991 question about whether support functionals are universal spectral points. The proof proceeds through a general minimax theorem for convex optimization on moment polytopes and support polytopes,

$\int a^2\,dx=1$6

and uses Hirai’s Fenchel-type duality theorem on Hadamard manifolds together with the Kempf–Ness function $\int a^2\,dx=1$7.

This usage of quantum functional is neither variational mechanics nor state-estimation risk. It is an entropy-based asymptotic invariant of tensors. Its significance lies in the bridge it establishes between entanglement geometry and combinatorial tensor support. The paper further derives formulas for asymptotic slice rank and weighted slice rank, showing that the same functional formalism governs asymptotic restriction problems across tensor complexity, quantum information, and additive combinatorics.

3. Convex spectral functionals and proper quantum scoring rules

In "Quantum Proper Scoring Rules: Minimax Estimation and Resource-Theoretic Advantages" (AlMasri, 6 May 2026), the phrase "Quantum Value Functional" is defined directly. The formal definition is: $\int a^2\,dx=1$8 Here $\int a^2\,dx=1$9 is the convex set of density operators. By the spectral theorem, such a functional depends only on the eigenvalues of $\operatorname{div}(a^2 Du)=0,$0, and the paper writes

$\operatorname{div}(a^2 Du)=0,$1

for a convex generator $\operatorname{div}(a^2 Du)=0,$2.

The central result is a quantum McCarthy-type duality between convex value functionals and proper scoring rules. If $\operatorname{div}(a^2 Du)=0,$3 is closed, convex, and unitarily invariant, then the induced quantum scoring rule $\operatorname{div}(a^2 Du)=0,$4 satisfies

$\operatorname{div}(a^2 Du)=0,$5

Properness follows from the subgradient inequality

$\operatorname{div}(a^2 Du)=0,$6

and the score gap dominates the associated quantum Bregman divergence: $\operatorname{div}(a^2 Du)=0,$7

For spectral functionals of the form

$\operatorname{div}(a^2 Du)=0,$8

the scoring rule is given explicitly by functional calculus: $\operatorname{div}(a^2 Du)=0,$9 The example $j=a^2Du$0 yields

$j=a^2Du$1

When $j=a^2Du$2 is operator convex, the paper states that the resulting divergence aligns with Petz’s quantum $j=a^2Du$3-divergence under a symmetry or commutativity condition,

$j=a^2Du$4

The minimax estimation theory is organized through the Quantum Cramér-Rao-McCarthy Bound. For

$j=a^2Du$5

the paper proves, for twice differentiable $j=a^2Du$6 with $j=a^2Du$7,

$j=a^2Du$8

linking minimax regret to the curvature of $j=a^2Du$9 and the Quantum Fisher Information.

The same framework supports resource-theoretic statements. The forecasting gap between classical fixed-basis and quantum joint-measurement strategies is stated as

$\operatorname{div}(j)=0$0

and for the logarithmic score the paper gives the coherence-risk tradeoff

$\operatorname{div}(j)=0$1

where $\operatorname{div}(j)=0$2 is the relative entropy of coherence. In the qubit example $\operatorname{div}(j)=0$3, the paper computes a score advantage

$\operatorname{div}(j)=0$4

matching the coherence $\operatorname{div}(j)=0$5. In this line of work, a quantum value functional is a convex spectral potential that simultaneously generates truthful scoring rules, divergences, and asymptotic minimax risk bounds.

4. Algorithmic estimation of nonlinear quantum and distributional functionals

Quantum algorithms use the term functional in yet another sense: the target of estimation. "Quantum Multi-Level Estimation of Functionals of Discrete Distributions" (Chen et al., 5 May 2026) studies

$\operatorname{div}(j)=0$6

with purified quantum query access

$\operatorname{div}(j)=0$7

The core method partitions the probability range into logarithmically many exponentially shrinking intervals, applies non-destructive singular value discrimination to isolate the relevant $\operatorname{div}(j)=0$8, and uses local polynomial approximation plus amplitude estimation. The paper emphasizes two structural features: it avoids variable-time overhead and requires only four additional ancilla qubits.

As an application, it gives efficient quantum estimators for the $\operatorname{div}(j)=0$9-Tsallis entropy of discrete distributions. The abstract states that for $a$0 the query complexity is

$a$1

improving the prior best $a$2, and that for $a$3 it obtains

$a$4

The paper presents these as the first near-optimal quantum estimators for parameterized $a$5-entropy for non-integer $a$6.

A state-estimation analogue appears in "Simultaneous Estimation of Nonlinear Functionals of a Quantum State" (Chen et al., 22 May 2025). The target family is

$a$7

which is nonlinear in $a$8 for $a$9. The main theorem states that all $u$0 values can be estimated simultaneously using

$u$1

samples, equivalently $u$2 up to logarithmic factors, while estimating only $u$3 already requires

$u$4

samples.

The estimator is built from commuting collective observables derived by weighted permutations and symmetrization: $u$5 with

$u$6

The method extends to general functionals $u$7 by polynomial approximation, with applications to entanglement spectroscopy and virtual cooling. In this algorithmic literature, a quantum value functional is not primarily a potential or action; it is the nonlinear quantity to be accessed from copies of a state or from query access to a distribution.

5. Operator, lattice, and control-theoretic formulations

An order-theoretic formulation is given in "Self-adjoint Operators as Functions II: Quantum Probability" (Doering et al., 2012). The paper proves a bijection

$u$8

between self-adjoint operators affiliated with a von Neumann algebra $u$9 and q-observable functions on the projection lattice $$-\frac{\hbar^2}{2m}\,\frac{\Delta a}{a}+\frac{|Du|^2}{2m}+W=E,$$0. For a self-adjoint operator $$-\frac{\hbar^2}{2m}\,\frac{\Delta a}{a}+\frac{|Du|^2}{2m}+W=E,$$1 with spectral family $$-\frac{\hbar^2}{2m}\,\frac{\Delta a}{a}+\frac{|Du|^2}{2m}+W=E,$$2, the associated q-observable function is

$$-\frac{\hbar^2}{2m}\,\frac{\Delta a}{a}+\frac{|Du|^2}{2m}+W=E,$$3

The paper interprets q-observable functions as generalized quantile functions for quantum observables. More generally, if $$-\frac{\hbar^2}{2m}\,\frac{\Delta a}{a}+\frac{|Du|^2}{2m}+W=E,$$4 is a complete meet-semilattice and $$-\frac{\hbar^2}{2m}\,\frac{\Delta a}{a}+\frac{|Du|^2}{2m}+W=E,$$5 is an $$-\frac{\hbar^2}{2m}\,\frac{\Delta a}{a}+\frac{|Du|^2}{2m}+W=E,$$6-valued cumulative distribution function, then its left adjoint

$$-\frac{\hbar^2}{2m}\,\frac{\Delta a}{a}+\frac{|Du|^2}{2m}+W=E,$$7

is an $$-\frac{\hbar^2}{2m}\,\frac{\Delta a}{a}+\frac{|Du|^2}{2m}+W=E,$$8-quantile function, with Galois connection

$$-\frac{\hbar^2}{2m}\,\frac{\Delta a}{a}+\frac{|Du|^2}{2m}+W=E,$$9

The paper then extends this picture to the spectral presheaf $$I(\psi)=\int_{\mathbb T^n}\left(\frac{D v\, D v^*}{2}(x)-W(x)\right)a(x)^2\,dx,$$00, whose clopen subobjects form a complete bi-Heyting algebra and play the role of a generalized quantum sample space.

A control-theoretic value-functional theory appears in "Quantum Markov Decision Processes: Dynamic and Semi-Definite Programs for Optimal Solutions" (Saldi et al., 2024). The q-MDP state is a density operator $$I(\psi)=\int_{\mathbb T^n}\left(\frac{D v\, D v^*}{2}(x)-W(x)\right)a(x)^2\,dx,$$01, and the infinite-horizon discounted value functional is

$$I(\psi)=\int_{\mathbb T^n}\left(\frac{D v\, D v^*}{2}(x)-W(x)\right)a(x)^2\,dx,$$02

For open-loop policies, the Bellman operator is

$$I(\psi)=\int_{\mathbb T^n}\left(\frac{D v\, D v^*}{2}(x)-W(x)\right)a(x)^2\,dx,$$03

and for classical-state-preserving closed-loop policies the paper defines an analogous contraction operator. The main structural result is that, under additional assumptions and SDP duality, the optimal value functions are linear in the density operator: $$I(\psi)=\int_{\mathbb T^n}\left(\frac{D v\, D v^*}{2}(x)-W(x)\right)a(x)^2\,dx,$$04 The same duality implies the existence of stationary optimal policies in both the open-loop and classical-state-preserving closed-loop settings.

These two works exhibit a common pattern despite very different goals. In the first, the value map is lattice-valued and quantile-like; in the second, it is a linear functional recovered from dual SDP variables. Both replace a purely state-vector description with a function on a richer order or control structure.

6. Adjacent usages, extensions, and terminological limits

Several neighboring literatures use closely related language while attaching it to different mathematical objects. "Quantum Geometry of Expectation Values" (Song, 2023) recasts quantum theory as a geometry of expectation values rather than wave functions. The fundamental map

$$I(\psi)=\int_{\mathbb T^n}\left(\frac{D v\, D v^*}{2}(x)-W(x)\right)a(x)^2\,dx,$$05

has image $$I(\psi)=\int_{\mathbb T^n}\left(\frac{D v\, D v^*}{2}(x)-W(x)\right)a(x)^2\,dx,$$06, the expectation value moduli space. The paper states that the boundary $$I(\psi)=\int_{\mathbb T^n}\left(\frac{D v\, D v^*}{2}(x)-W(x)\right)a(x)^2\,dx,$$07 corresponds to the ground state, singular points correspond to eigenstates of Hamiltonian families, and the defining equations obtained by Jacobian minors and elimination provide an explicit construction of the density functional. Here the functional viewpoint is geometric and semialgebraic rather than convex-analytic or variational.

In perturbative algebraic quantum field theory, "A novel class of functionals for perturbative algebraic quantum field theory" (Hawkins et al., 2023) argues that microcausal functionals are not closed under the Peierls bracket and introduces equicausal functionals as the stable class. The new condition is equicontinuity of derivative maps on compact sets of configurations, and the paper shows closure under the $$I(\psi)=\int_{\mathbb T^n}\left(\frac{D v\, D v^*}{2}(x)-W(x)\right)a(x)^2\,dx,$$08-product and Peierls bracket together with the time-slice axiom. This is a case where a controversy is explicit: microcausal functionals are said not to be a suitable basis for pAQFT, and equicausal functionals are proposed as the remedy.

In quantum probability on easy compact quantum groups, "Gaussian Generating functionals on easy quantum groups" (Franz et al., 2024) studies generating functionals as infinitesimal generators of convolution semigroups of states. Gaussianity means vanishing on the third power of the augmentation ideal, and on $$I(\psi)=\int_{\mathbb T^n}\left(\frac{D v\, D v^*}{2}(x)-W(x)\right)a(x)^2\,dx,$$09 every Gaussian generating functional has the form

$$I(\psi)=\int_{\mathbb T^n}\left(\frac{D v\, D v^*}{2}(x)-W(x)\right)a(x)^2\,dx,$$10

with anti-hermitian drift $$I(\psi)=\int_{\mathbb T^n}\left(\frac{D v\, D v^*}{2}(x)-W(x)\right)a(x)^2\,dx,$$11 and covariance tensor $$I(\psi)=\int_{\mathbb T^n}\left(\frac{D v\, D v^*}{2}(x)-W(x)\right)a(x)^2\,dx,$$12 subject to

$$I(\psi)=\int_{\mathbb T^n}\left(\frac{D v\, D v^*}{2}(x)-W(x)\right)a(x)^2\,dx,$$13

The paper further shows that central Gaussian generating functionals are highly restricted: on $$I(\psi)=\int_{\mathbb T^n}\left(\frac{D v\, D v^*}{2}(x)-W(x)\right)a(x)^2\,dx,$$14 and $$I(\psi)=\int_{\mathbb T^n}\left(\frac{D v\, D v^*}{2}(x)-W(x)\right)a(x)^2\,dx,$$15 there are no nonzero central Gaussian generating functionals for $$I(\psi)=\int_{\mathbb T^n}\left(\frac{D v\, D v^*}{2}(x)-W(x)\right)a(x)^2\,dx,$$16, and on $$I(\psi)=\int_{\mathbb T^n}\left(\frac{D v\, D v^*}{2}(x)-W(x)\right)a(x)^2\,dx,$$17 and $$I(\psi)=\int_{\mathbb T^n}\left(\frac{D v\, D v^*}{2}(x)-W(x)\right)a(x)^2\,dx,$$18 the only central Gaussian functionals factor through the circle group.

A terminological boundary appears in Bell nonlocality. "Quantum value for a family of $$I(\psi)=\int_{\mathbb T^n}\left(\frac{D v\, D v^*}{2}(x)-W(x)\right)a(x)^2\,dx,$$19-like Bell functionals" (Gigena et al., 2022) studies the maximal quantum value of a Bell functional,

$$I(\psi)=\int_{\mathbb T^n}\left(\frac{D v\, D v^*}{2}(x)-W(x)\right)a(x)^2\,dx,$$20

not a value functional on states in the convex or variational sense. For one branch of its three-parameter family the paper proves the closed-form expression

$$I(\psi)=\int_{\mathbb T^n}\left(\frac{D v\, D v^*}{2}(x)-W(x)\right)a(x)^2\,dx,$$21

while the branch containing $$I(\psi)=\int_{\mathbb T^n}\left(\frac{D v\, D v^*}{2}(x)-W(x)\right)a(x)^2\,dx,$$22 is treated numerically. The proximity of the phrases "quantum value" and "quantum value functional" can therefore be misleading.

A still looser use occurs in decision theory. "Decision-making under uncertainty: a quantum value operator approach" (Xin et al., 2022) introduces a density operator

$$I(\psi)=\int_{\mathbb T^n}\left(\frac{D v\, D v^*}{2}(x)-W(x)\right)a(x)^2\,dx,$$23

as a value operator representing subjective belief, with quantum expected value

$$I(\psi)=\int_{\mathbb T^n}\left(\frac{D v\, D v^*}{2}(x)-W(x)\right)a(x)^2\,dx,$$24

The operator is then constructed approximately by quantum decision trees built from logic operations and quantum gates and optimized by genetic programming.

Taken together, these adjacent usages show that the expression "quantum value functional" has become a cross-disciplinary umbrella rather than a single technical term. In some works it denotes a convex spectral potential on density operators; in others, a stationary action, an entropy score on entanglement polytopes, a nonlinear estimation target, a quantile-like map on projection lattices, or a stochastic generator. A plausible implication is that any interpretation of the term requires immediate attention to the ambient framework—variational mechanics, tensor asymptotics, scoring theory, estimation, control, quantum probability, or quantum information—because the underlying object, codomain, and notion of optimality vary substantially across these literatures.