---
title: Quantum State Over Time (QSOT)
url: https://www.emergentmind.com/topics/quantum-state-over-time-qsot
type: topic
---

# Quantum State Over Time (QSOT)

Searching arXiv for the listed QSOT-related papers to ground the article with current records.
arxiv_search(query="all: \"quantum state over time\" OR all: \"past quantum states\" OR all: \"Quantum State Smoothing\" OR id:1503.02799 OR id:1305.0681 OR id:2202.03607 OR id:2308.12752 OR id:2311.00162 OR id:2304.03954 OR id:2410.22630 OR id:2504.04856", max_results=10, sort_by="relevance")
Quantum State Over Time (QSOT) denotes a family of formalisms that attempt to represent temporal quantum structure by an operator on a tensor product of Hilbert spaces labeled by times, so that quantum correlations across both space and time may be treated with a common mathematical formalism. In the basic two-time setting, one starts from an initial state \(\rho_A\) and a channel \(\mathcal E_{B|A}\), and asks for an operator on \(A\otimes B\) that plays the role of a temporal analogue of a joint state. The literature shows both that such objects are generally not ordinary positive density operators and that their admissible form depends sensitively on the axioms imposed: some constructions are Hermitian but indefinite, some are non-Hermitian but associative, and some recent results identify a unique multipartite extension once appropriate operational assumptions are adopted [1607.03637][2202.03607][2308.12752][2410.22630].

## 1. Problem setting and desiderata

The point of departure for QSOT is the structural asymmetry of standard quantum theory. A composite system at one time is described by a density operator on a tensor-product Hilbert space, whereas a single system at two times is described by an initial state together with a channel. In the simplest setting one seeks an operator
\[
\rho_{AB}=f(\rho_A,\mathcal E_{B|A}),
\]
or equivalently a star product
\[
\rho_{AB}=E_{B|A}\star \rho_A,
\]
where \(E_{B|A}\) is the channel state associated with \(\mathcal E_{B|A}\) [1607.03637].

The natural desiderata for such a construction were formulated explicitly in the two-time literature. A temporal joint state should be a Hermitian operator on the tensor product Hilbert space, preserve probabilistic mixtures, reduce to the classical joint distribution in the commuting limit, have the correct single-time marginals, and compose associatively across multiple time-steps. In the classical case, these requirements are straightforward because one writes a joint law as \(P(XY)=P(Y|X)P(X)\). QSOT asks whether an analogous operator-level construction exists in the quantum case [1607.03637].

This basic program already contains a tension. If a temporal state is to resemble an ordinary spatial joint state too closely, then noncommutativity obstructs the classical product rule. If, instead, one relaxes some of the single-time density-operator intuitions, then a broader class of temporal operators becomes available. Much of the modern QSOT literature is devoted to making that trade-off precise [1607.03637][2308.12752].

## 2. Bipartite constructions and star products

A central strand of the literature defines QSOT through a star product that combines the initial state with a channel-state operator. In the matrix-algebra formulation of “On quantum states over time,” the density associated to the temporal state is
\[
\mathscr D[F\star\omega] = \frac12\Big([F](\rho\otimes 1_B)+(\rho\otimes 1_B)[F]\Big),
\]
which is exactly the Jordan product of the channel density with \(\rho\otimes 1_B\) [2202.03607].

The same construction appears in the later uniqueness literature as the Fullwood–Parzygnat state over time:
\[
\mathcal E_{B|A}\star_{FP}\rho_A := \frac{1}{2}\left\{\rho_A\otimes \mathbf 1_B,\; \mathcal D[\mathcal E]\right\}.
\]
Here \(\mathcal D[\mathcal E]\) is the channel state, and the anticommutator is the Jordan symmetrization [2308.12752]. This construction is Hermitian, has the correct temporal marginals, and reduces to the ordinary product in the commuting classical limit. It is not generally positive, so it is not, in full generality, an ordinary density operator [2202.03607][2308.12752].

The bipartite literature also contains competing constructions. The Leifer–Spekkens prescription uses the square-root sandwich
\[
(\sqrt{\rho}\otimes \mathds 1)\,[F]\,(\sqrt{\rho}\otimes \mathds 1),
\]
the left and right products use ordinary left or right multiplication by \(\rho\otimes \mathds 1\), and Wigner-inspired constructions expand both state and channel in phase-point operators [1607.03637][2410.22630]. These alternatives differ sharply in which axioms they satisfy. The Leifer–Spekkens proposal is Hermitian and locally positive but fails convex-bilinearity and associativity; the Fitzsimons–Jones–Vedral proposal is the Jordan product for qubits and fails full associativity as an abstract binary operation; the Wigner construction is Hermitian, convex-bilinear, and associative, but fails the classical-limit requirement as formulated in the no-go analysis [1607.03637].

A later transport-theoretic application renames the Jordan-product temporal operator a state over time or “stote,” defined by
\[
Q = (\rho\otimes I)\star J = \frac{1}{2}\bigl((\rho\otimes I)J + J(\rho\otimes I)\bigr),
\]
with \(J\) the Jamiołkowski matrix of a CPTP map. In that setting the stote functions as a coupling object for quantum transport costs, with correct initial and final marginals but, again, without positivity in general [2504.04856].

## 3. No-go theorems and uniqueness

The modern theory of QSOT is shaped by a no-go theorem. “Can a quantum state over time resemble a quantum state at a single time?” proves that there is no function
\[
\star \colon \mathbb H_n \times \mathbb H_n \to \mathbb H_n
\]
satisfying convex-bilinearity, product on commuting pairs, product when traced, and associativity. Since Hermiticity is built into the codomain, this means that no Hermitian-valued star product can satisfy all the natural axioms simultaneously [1607.03637].

The same work shows that if Hermiticity is dropped, then the only functions
\[
\star \colon \mathbb M_n \times \mathbb M_n \to \mathbb M_n
\]
satisfying convex-bilinearity, product on commuting pairs, product when traced, and associativity are ordinary matrix multiplication and its reversed-order variant:
\[
x\star y=xy
\qquad\text{or}\qquad
x\star y=yx.
\]
This identifies Hermiticity as the obstructing requirement. A plausible implication is that temporal composition is naturally represented by a broader class of operators than ordinary density matrices [1607.03637].

“On quantum states over time” responds by changing the domain of the construction. Instead of demanding a binary operation on an enlarged operator domain, it defines the state-over-time function only on the physically faithful domain of a channel and a state, and proves that the Jordan-product assignment satisfies Hermiticity, normalization, bilinearity, classical limit, correct marginals, and compositionality on that restricted domain [2202.03607].

The remaining issue was uniqueness. “Uniqueness of quantum state over time function” proves that the earlier axioms do not uniquely determine a QSOT; one can add nontrivial commutator-like contributions and still satisfy weaker conditions. The paper then proposes an alternative set of operational axioms—Completeness, Compositionality, Classical Conditionability, and Time reversal symmetry—and proves that the Fullwood–Parzygnat construction is the only state over time function satisfying them [2308.12752]. In that sense, the Jordan product is not merely an admissible choice; under those operational axioms it is the unique one.

## 4. Multipartite extension and quantum Markovianity

The two-time problem does not settle the multi-time case. “Unique multipartite extension of quantum states over time” shows that fixing the bipartite rule does not by itself determine a unique \(n\)-time extension, because many different multipartite products reduce to the same one-step formula [2410.22630].

The paper considers a Markovian chain of channels
\[
\bm{E}=(E_1,\dots,E_n),\qquad E_i:A_{i-1}\to A_i,
\]
with initial state \(\rho\in\mathfrak{S}(A_0)\), and a spatiotemporal product assigning an operator
\[
\bm{E}\star \rho \in A_0\cdots A_n
\]
satisfying temporal marginal conditions. The main theorem states that two assumptions—linearity in the initial state and a quantum analog of conditionability—uniquely force the iterative formula
\[
\bm{E}\star \rho = E_n \star \big(E_{n-1}\star(\cdots \star (E_1\star \rho))\big).
\]
Thus, once the one-step rule is fixed, the multipartite extension is uniquely determined [2410.22630].

In the Fullwood–Parzygnat case, the resulting multi-time operator is a Hermitian unit-trace quasi-state. It need not be positive; the paper interprets the resulting negativity as a witness of nonclassical temporal correlations. The same theorem yields a canonical multipartite extension of Kirkwood–Dirac and Margenau–Hill quasi-probability distributions, and it provides an operator-level characterization of quantum Markovianity through the iterative factorization of the multi-time state [2410.22630].

This makes the multipartite QSOT program more rigid than the earlier two-time literature. Before this result, multi-time constructions were underdetermined; after it, conditionability and state-linearity single out a canonical extension of the already distinguished bipartite product [2410.22630].

## 5. Covariance, pseudo-density matrices, interferometry, and spacetime-state unification

A different strand of the literature formulates QSOT through broadcast-like maps and pseudo-density matrices. “General covariance for quantum states over time” defines a canonical broadcasting map
\[
\mathfrak{B}_A=\frac{1}{2}\left(\mu_A^*+\widetilde{\mu}_A^*\right),
\]
a bloom of a channel
\[
(E)=(id_A\otimes E)\circ \mathfrak{B}_A,
\]
and a canonical multi-time state
\[
(E_1,\dots,E_n)\star \rho=(E_1,\dots,E_n)(\rho).
\]
It then proves covariance under arbitrary \(*\)-isomorphisms:
\[
(\phi_0\otimes \cdots \otimes \phi_n)\Big((E_1,\dots,E_n)\star \rho\Big)=(E_1',\dots,E_n')\star \rho',
\]
with \(E_k'=\phi_k\circ E_k\circ \phi_{k-1}^{-1}\) and \(\rho'=\phi_0(\rho)\) [2311.00162].

“Quantum dynamics as a pseudo-density matrix” develops the corresponding pseudo-state viewpoint. A pseudo-density matrix is Hermitian, unit trace, has density-matrix marginals, and need not be positive. For one channel the state over time is
\[
\psi(\rho,\mathscr E)=\frac{1}{2}\Big((\rho\otimes \mathds 1)\mathscr J[\mathscr E] + \mathscr J[\mathscr E](\rho\otimes \mathds 1)\Big),
\]
and for arbitrary finite chains the paper proves that the recursively defined multi-time pseudo-density matrix is well defined, independent of parenthesization, and reducible under partial traces to coarser processes. It also gives an inverse reconstruction theorem: on a large subclass, the initial state and each intermediate channel can be recovered from the pseudo-density matrix via Sylvester-equation inversion and the inverse Jamiołkowski map [2304.03954].

A more explicitly operational turn appears in “Probing Quantum States Over Spacetime Through Interferometry.” There the spacetime state \(\rho_{A_1\dots A_n}\) is defined by the interference term
\[
I=\Tr[(V_{A_1}\otimes \cdots \otimes V_{A_n})\rho_{A_1\dots A_n}]
\]
for local unitary interventions in one arm of an interferometer. The paper proves that a quantum measurement is causally agnostic if and only if it can be implemented by multi-arm interferometry. In the temporal bipartite case this yields
\[
I=\Tr[(V_A\otimes W_B)(E\star_L \rho)],
\]
and under time-reversal symmetry
\[
I=\Tr[(V_A\otimes W_B)(E\star_{FP} \rho)].
\]
The same work identifies QSOT as the first-order approximation of a process matrix and uses mixed temporal states to model non-Markovianity [2507.19258].

The broadest synthesis appears in “Unifying spacetime approaches to quantum mechanics.” There the parent object is the spacetime state
\[
\mathcal R=\rho_0\, e^{i\tilde{\mathcal S}},
\]
and the two-time channel case becomes
\[
\mathcal R=(\rho\otimes \mathbbm 1)J(\mathcal E).
\]
QSOT then appears as one manifestation of this more general object: either the non-Hermitian spacetime-state form \(\mathcal R\), the symmetrized operator \((\mathcal R+\mathcal R^\dag)/2\), or a similarity-transformed Hermitian version. This unifying perspective treats QSOT, pseudo-density matrices, Page–Wootters states, superdensity operators, and timelike-entanglement proposals as different manifestations of the same underlying spacetime state [2606.12539].

## 6. Related temporal formalisms, scope, and common misconceptions

Not every temporally extended quantum formalism is a QSOT in the strict sense of a joint operator on multiple temporal Hilbert spaces. “Past quantum states” defines the past quantum state at time \(t\) as
\[
\Xi(t)=\bigl(\rho(t),E(t)\bigr),
\]
where \(\rho(t)\) is the forward-conditioned density operator and \(E(t)\) is a backward effect matrix. Its key retrodictive rule is
\[
p_{\mathrm p}(m) = \frac{\mathrm{Tr}\!\left(\Omega_m \rho(t)\Omega_m^\dagger E(t)\right)} {\sum_{m'}\mathrm{Tr}\!\left(\Omega_{m'} \rho(t)\Omega_m^\dagger E(t)\right)}.
\]
This is a temporally extended inference formalism, but it is not a single density matrix over multiple times [1305.0681].

“Quantum State Smoothing” is similarly QSOT-adjacent rather than QSOT proper. It defines, for a partially monitored open quantum system, a smoothed single-time state
\[
\rho_{\mathrm S}(t) = \mathbb E_{\overleftarrow{\mathbf U}_t \mid \overleftrightarrow{\mathbf O}} \!\left[ \rho_{\overleftarrow{\mathbf O}_t,\overleftarrow{\mathbf U}_t}(t) \right],
\]
namely a retrospective state assignment at one time conditioned on past and future observed data through a smoothed distribution over unobserved records. It is time-symmetric and all-time-conditioned, but it is not a joint quantum state living across multiple times. In the driven two-level-atom example, smoothing recovers about \(26\%\) of the purity lost due to unobserved radiation for \(Y\)-homodyne detection and about \(12\%\) for \(X\)-homodyne detection [1503.02799].

A further development is the observable-side dual formalism “Quantum observables over time for information recovery.” There the temporal object is
\[
\mathcal{E}^\dagger \star O \in \mathcal{L}(\mathcal{H}_A \otimes \mathcal{H}_B),
\]
with observable marginals
\[
\operatorname{Tr}_B[\mathcal{E}^\dagger \star O] = O,
\qquad
\operatorname{Tr}_A[\mathcal{E}^\dagger \star O] = \mathcal{E}^\dagger(O).
\]
Unlike QSOT, such a QOOT is not always definable: a necessary and sufficient condition is
\[
\operatorname{Tr}[O] = \operatorname{Tr}[\mathcal{E}^\dagger(O)].
\]
The paper’s Jordan-product QOOT,
\[
\mathcal{E}^\dagger \star O = \frac{1}{2}\left\{ O\otimes \mathcal{I}, \mathcal{D}[\mathcal{E}^\dagger]\right\},
\]
therefore exposes a genuine observable-side obstruction that has no direct analogue on the state side [2412.11659].

These neighboring theories clarify a persistent misconception. QSOT is not a synonym for every formalism that uses future data, retrodiction, sequential measurements, or trajectories through time. In the strict literature surveyed here, QSOT refers to state-like operators on tensor products of time-labeled Hilbert spaces, usually Hermitian or at least trace-one, with temporal marginals and explicitly static operator representations of dynamics. Retrodictive two-object formalisms, smoothed single-time states, and observable-over-time duals are closely related, but they solve different problems [1305.0681][1503.02799][2412.11659].

Source: https://www.emergentmind.com/topics/quantum-state-over-time-qsot