---
title: Two-Time Pseudo-Density Operator
url: https://www.emergentmind.com/topics/two-time-pseudo-density-operator
type: topic
---

# Two-Time Pseudo-Density Operator

A two-time pseudo-density operator is a Hermitian, unit-trace operator that represents quantum correlations between an earlier and a later event in a sequential measurement process. It is the temporal analogue of a bipartite density operator, except that positivity is not required: each one-time marginal is an ordinary density matrix, but the joint two-time object can have negative eigenvalues. In the qubit literature the same construction is often called a pseudo-density matrix, and in more general finite-dimensional settings it coincides with the canonical state over time \(\mathcal E\star\rho=\frac12\{\rho\otimes\mathds 1_B,\mathscr J[\mathcal E]\}\). Recent work has sharpened its interpretation by showing that the partial transpose of any bipartite density operator is always a two-time pseudo-density operator, so that partial transposition can be understood as a mathematically precise conversion of spatial correlations into temporal correlations—a “space-time swap” [2304.03954, 2405.17555, 2508.12256].

## 1. Formal definition and state-over-time viewpoint

In the state-over-time approach, a pseudo-density operator is defined on a tensor-product algebra \(A=A_0\otimes\cdots\otimes A_n\). A self-adjoint element \(\tau\in A\) is a pseudo-density operator with respect to this factorization if and only if, for every \(i\in\{0,\dots,n\}\), the corresponding partial trace is an ordinary density matrix. Because all marginals have unit trace, \(\tau\) itself also has unit trace. The decisive departure from standard density operators is that \(\tau\) need not be positive [2304.03954].

For the two-time case, the basic construction associates a single operator to a process \((\rho,\mathscr E)\) consisting of an initial state and a quantum channel. With \(\mathscr J[\mathscr E]\) the Jamiołkowski operator of the channel, the two-time pseudo-density operator is
\[
(\rho,\mathscr E)=\frac12\Big((\rho\otimes \mathds{1})\mathscr J[\mathscr E]+\mathscr J[\mathscr E](\rho\otimes \mathds{1})\Big).
\]
The same formula appears in the canonical-state-over-time literature as
\[
\mathcal E\star\rho=\frac12\{\rho\otimes\mathds 1_B,\mathscr J[\mathcal E]\}.
\]
This symmetrized form is the quantum analogue of a classical joint distribution built from a prior state and a transition rule, but quantum temporality breaks the classical equivalence between static and dynamical descriptions. The pseudo-density operator therefore functions as a “state over time” rather than as an ordinary state at one time [2304.03954, 2405.17555].

This viewpoint was developed in response to the observation, emphasized in work associated with Fitzsimons, Jones and Vedral, that a quantum process should be representable by a single operator carrying both the endpoint states and the correlations between them. The resulting object is density-matrix-like in its marginals and trace, but its non-positivity records specifically temporal structure [2304.03954].

## 2. Sequential measurement construction and operator representation

The operational core of the two-time pseudo-density operator is a sequential measurement experiment. In a two-point sequential measurement scenario, Alice prepares a state \(\rho_A\), measures a Pauli observable \(\sigma_\alpha\), sends the post-measurement system through a channel \(E:A\to B\), and Bob measures \(\sigma_\beta\). If \(\Pi_\alpha^\pm\) are the projectors onto the \(\pm1\) eigenspaces of \(\sigma_\alpha\), the two-time correlator is
\[
\langle \sigma_\alpha,\sigma_\beta\rangle
=
\operatorname{Tr}\!\big[E(\Pi_\alpha^+\rho_A\Pi_\alpha^+)\sigma_\beta\big]
-
\operatorname{Tr}\!\big[E(\Pi_\alpha^-\rho_A\Pi_\alpha^-)\sigma_\beta\big].
\]
Because Pauli operators form an orthogonal basis, there is a unique operator \(\varrho_{AB}\) such that
\[
\langle \sigma_\alpha,\sigma_\beta\rangle
=
\operatorname{Tr}\!\big[\varrho_{AB}(\sigma_\alpha\otimes\sigma_\beta)\big].
\]
That operator is the pseudo-density operator, given by
\[
\varrho_{AB}=E\star \rho_A\equiv \frac12\{\rho_A\otimes \mathds 1,\,J[E]\},
\]
with \(J[E]=\sum_{i,j}|i\rangle\langle j|\otimes E(|j\rangle\langle i|)\) [2508.12256].

In finite dimensions, this representation has a sharp domain of validity. There is no general operator \(\varrho_{AB}\in A\otimes B\) such that \(\langle O_A,O_B\rangle=[\varrho_{AB}(O_A\otimes O_B)]\) for all observables \(O_A,O_B\): a no-go theorem shows that the two-time expectation function is not universally bilinear in the first argument. The obstruction is removed by restricting earlier-time observables to the maximal class of light-touch observables, whose spectrum is either \(\{\lambda\}\) or \(\{\pm\lambda\}\). For that class, there exists a unique operator representation, and the representing operator is exactly the canonical state over time
\[
\varrho_{AB}=\mathcal E\star \rho=\frac12\{\rho\otimes \mathds 1_B,\mathscr J[\mathcal E]\}.
\]
Pauli observables are the canonical qubit example, so the original pseudo-density-matrix formalism is recovered as a special case of a more general operator representation theorem [2405.17555].

The two-time pseudo-density operator is therefore not merely a formal rewriting of a channel-state pair. It is the unique operator reproducing the experimentally defined two-time correlators for the relevant class of earlier-time observables.

## 3. Structural properties, marginals, and non-positivity

The standard structural properties are fixed across the literature: a pseudo-density operator is Hermitian, has unit trace, and has ordinary density-matrix marginals, but it need not be positive semidefinite. Negative eigenvalues are not treated as defects of the formalism. Rather, they are interpreted as signatures of temporal quantum correlations or, in the state-over-time language, of correlations that imply causation [1911.02441, 2304.03954].

A basic qubit example is the two-time operator
\[
R_{12}=\frac14\left(I_{12}+X_1X_2+Y_1Y_2+Z_1Z_2\right).
\]
Its marginals are both maximally mixed,
\[
\operatorname{Tr}_2(R_{12})=\operatorname{Tr}_1(R_{12})=\frac{I}{2},
\]
but \(R_{12}\) is not positive: the singlet state is an eigenstate with eigenvalue \(-\tfrac12\). The point of the example is that labels \(1,2\) need not denote two spatial subsystems; they can denote two times of the same system. The resulting operator looks density-matrix-like while encoding time-ordered correlations that no ordinary density operator can capture [1911.02441].

The same phenomenon persists beyond qubits. For a qutrit process with \(\rho=\mathrm{diag}(1,0,0)\) and \(\mathcal E=\mathrm{id}\), the canonical state over time is a \(9\times 9\) operator with eigenvalues
\[
\left\{1,-\tfrac12,-\tfrac12,\tfrac12,\tfrac12,0,0,0,0\right\},
\]
illustrating again that non-positivity is expected rather than exceptional [2405.17555].

A related quantitative witness is the PDM negativity or causal monotone
\[
f(R):=\|R\|_{\rm tr}-1=\operatorname{Tr}\sqrt{R^\dagger R}-1.
\]
If \(R\) has negative eigenvalues, then \(f(R)>0\). This supplies a compact diagnostic of temporality within the PDM formalism [2306.05958].

## 4. Partial transpose theorem and the space-time swap

The strongest recent interpretation of the two-time pseudo-density operator comes from the partial transpose theorem. For any bipartite density operator \(\rho_{AB}\in A\otimes B\), there exists a quantum channel \(E:A\to B\) such that
\[
\rho_{AB}^{T_B}=E\star \rho_A,\qquad
\rho_{AB}^{T_A}=\overline{E}\star \rho_A^T,
\]
where \(\rho_A=\operatorname{Tr}_B[\rho_{AB}]\) and \(\overline{E}=T\circ E\circ T\) is the complex-conjugate channel. If \(\rho_A\) is full rank, the channel \(E\) is unique. Thus the partial transpose of a bipartite state is always a two-time pseudo-density operator for some sequential measurement process [2508.12256].

This theorem makes precise the idea of a space-time swap. The ordinary bipartite state \(\rho_{AB}\) describes correlations across space, while \(\rho_{AB}^{T_B}\) or \(\rho_{AB}^{T_A}\) reinterprets the same correlation data as arising from a causal sequence in which Alice measures first and Bob measures later after a channel. For Pauli observables,
\[
\langle \sigma_\alpha,\sigma_\beta\rangle_{(\rho_A,E)}
=
\operatorname{Tr}\!\big[\rho_{AB}^{T_B}(\sigma_\alpha\otimes\sigma_\beta)\big],
\]
and analogously for \(T_A\). Partial transposition is therefore not merely an algebraic operation on matrix entries; it maps a spatial description into a temporal one [2508.12256].

The theorem also recasts the Peres-Horodecki criterion. If \(\rho_{AB}^{T_B}\ge 0\), then the pseudo-density operator obtained by the space-time swap is itself positive, so the correlations admit both a spatial and a temporal description. Such PPT states are described as dual states. Conversely, a negative partial transpose corresponds to temporal correlations that cannot be modeled by ordinary spacelike-separated quantum systems [2508.12256].

The Bell-state example makes the point sharply. A maximally entangled two-qubit state written as
\[
\rho_{AB}=\frac14\Big(\mathds{1}\otimes\mathds{1}+\sum_{i=1}^3 a_i\,\sigma_i\otimes\sigma_i\Big)
\]
violates the CHSH inequality with the value \(\pm2\sqrt2\) for a standard choice of observables \(Q,R,S,T\). After partial transpose, the operator \(\rho_{AB}^{T_B}\) yields the same CHSH-type value,
\[
E_{QS}+E_{RS}+E_{RT}-E_{QT}
=
\operatorname{Tr}[\rho_{AB}^{T_B}O]
=
\pm2\sqrt2,
\]
but now the interpretation is temporal rather than spacelike. In particular, the resulting two-time operator has a negative eigenvalue, so its correlations cannot be realized by any ordinary spacelike-separated pair of qubits. The Bell-violating nonlocality of the original state is thus converted into Bell-violating causal correlations [2508.12256].

The same work proposes a geometric reading: \(T_B\) and \(T_A\) can be viewed as reflections about future or past light cones of an auxiliary system, or equivalently as a rotation of light-cone structure. The black-hole analogy is presented cautiously: because Schwarzschild geometry interchanges the roles of space and time inside the horizon, a partial transpose is suggested as a quantum-mechanical analogue of that interchange [2508.12256].

## 5. Relation to multiple-time states, process matrices, and transition matrices

The two-time pseudo-density operator sits within a broader family of spatiotemporal state formalisms. One major unification result shows that operational scenarios underlying two-time states can be represented as PDMs, and that this representation preserves the relevant probabilities. Since earlier work had already mapped bipartite process matrices to bipartite two-time states, the new mapping yields a route from process matrices with measurements to PDMs. The consequence is that PDMs can, like process matrices, model indefinite causal order [2306.05958].

Operationally, the mapping proceeds by representing a pre-/post-selected two-time scenario as a three-time or multi-time PDM whose trace reproduces the same outcome statistics as the ABL-style probability rule of the two-time-state formalism. For ensembles of pure two-time states, the corresponding PDM is obtained by linear mixing. The relation is therefore one of statistical equivalence for operationally realized scenarios, not of complete conceptual identity: the PDM still uses definite measurement times, while the process-matrix formalism is aimed at the absence of a global causal order [2306.05958].

A distinct but nearby construction is the reduced transition matrix
\[
\mathcal T^{\psi|\phi}=\frac{|\psi\rangle\langle\phi|}{\langle\phi|\psi\rangle},
\]
which underlies pseudoentropy and pseudo-Rényi entropy. Unlike the two-time pseudo-density operator, this object is generally non-Hermitian. Its reduced form \(\mathcal T_A^{\psi|\phi}\) supports entropic quantities such as
\[
S^{(n)}(\mathcal T_A)=\frac{\log \operatorname{tr}_A(\mathcal T_A)^n}{1-n},
\qquad
S(\mathcal T_A)=-\operatorname{tr}_A(\mathcal T_A\log \mathcal T_A),
\]
and is related by an operator sum rule to reduced density matrices of superposition states [2308.05261].

A closely related interpretation appears when the time evolution operator is treated as a density-operator-like object and projected onto an initial state. In that setting, the projected time evolution operator becomes essentially the reduced transition matrix of pseudo-entropy for the pair \((|I\rangle,U(t)|I\rangle)\). This motivates a two-time pseudo-density-operator viewpoint based on initial and final time slices, but the resulting entropy is generically complex, reflecting the non-Hermitian character of the transition-matrix framework [2303.01307].

The key distinction is therefore structural: the two-time pseudo-density operator is a Hermitian state-over-time object designed to encode sequential correlations, whereas transition-matrix and pseudoentropy constructions are non-Hermitian generalizations built from off-diagonal amplitudes between two states.

## 6. Applications, simulations, and conceptual significance

An important application arises in chronology-violating scenarios. In the open-timelike-curve setting studied with three qubits \(Q_1,Q_2,Q_3\), the chronology-violating region is modeled by the pseudo-density operator
\[
R_{123}=\frac18\left\{I_{123}-\Sigma_{12}+\Sigma_{23}-\Sigma_{13}\right\},
\]
with \(\Sigma_{ij}=X_iX_jI_k+Y_iY_jI_k+Z_iZ_jI_k\). Its marginals are chosen so that \(Q_1\) and \(Q_2\) are maximally entangled before the curve, \(Q_1\) and \(Q_3\) are maximally entangled after it, and \(Q_2\) and \(Q_3\) behave as the same qubit at two times. Because the two-time correlation operator \(R_{23}\) is temporal rather than spatial, the model allows violations of entanglement monogamy that ordinary density matrices cannot represent [1911.02441].

The same work gives an optical simulation with polarization-entangled photons and reports what it describes as the first tomographic reconstruction of a PDO. The reconstructed physical marginals agree well with theory, with fidelities approximately
\[
F_{12}=0.964,\qquad F_{13}=0.963.
\]
The experiment is explicitly a simulation rather than a literal realization of time travel, but it demonstrates that PDO tomography is operationally meaningful even when ordinary state tomography must be modified because sequential measurement disturbance matters [1911.02441].

Across these applications, the two-time pseudo-density operator serves a consistent interpretive role. It unifies spatially separated subsystems, temporally separated events, and mixed spacetime scenarios within a single operator formalism, while treating negativity as evidence that the relevant correlations are genuinely temporal. A plausible implication is that the formalism is best understood not as an extension of mixed-state ontology, but as a representation theory for quantum correlations when the tensor factors denote events in time as well as systems in space. In that sense, the two-time pseudo-density operator is the basic state-like object for temporal quantum correlations, just as the density operator is the basic state-like object for spatial ones [1911.02441, 2304.03954, 2508.12256].

Source: https://www.emergentmind.com/topics/two-time-pseudo-density-operator