Spacetime Density Matrix in Quantum Systems

• Spacetime density matrix is an operator that generalizes the standard density matrix by encoding multi-time correlations on tensor product Hilbert spaces.
• It encompasses various constructions including correlator-based, action-based, pseudo-density, and doubled-density operators to recover multi-time quantum statistics.
• Non-Hermiticity and non-positivity in these frameworks enable capturing temporal causality and inter-slice dynamics with applications in holography and quantum gravity.

A spacetime density matrix is an operator-valued generalization of the ordinary density matrix in which the relevant data are not confined to a single Cauchy surface. In the most direct formulations, it is defined so that traces against operators supported on different time slices reproduce multitime correlation functions, while ordinary density matrices on individual slices are recovered as marginals or partial traces (Guo, 28 Aug 2025, Diaz et al., 10 Jun 2026). The term is not used uniformly, however. Current literature includes correlator-defined multitime operators, action-based “spacetime states,” pseudo-density matrices, doubled density operators for space-time processes, holographic analyses of bulk regions encoded by reduced density matrices, and several gravitational or phase-space analogs that are suggestive of a density-matrix description without defining the same object (Fullwood, 2023, Jia et al., 2023, Czech et al., 2012).

1. Definition and conceptual scope

The most literal use of the term appears in constructions that promote the equal-time density matrix to an operator on a tensor product of Hilbert spaces associated with multiple times. For two Cauchy surfaces $C_0$ and $C_1$, one such definition requires an operator $T_{C_0C_1}$ satisfying

$$\mathrm{Tr}(\rho_0\, \mathcal O_0 \mathcal O_1(t_1)) = \mathrm{tr}\!\left(T_{C_0C_1}\,\mathcal O_0 \mathcal O_1\right),$$

so that $T_{C_0C_1}$ is the object encoding cross-time correlators between the two surfaces (Guo, 28 Aug 2025). In this sense, the spacetime density matrix is not simply a time-evolved state but an operator carrying inter-slice correlation data.

A closely related formulation defines a “spacetime state” on

$$\mathcal H=\bigotimes_{t=0}^{N-1} h_t \simeq h^{\otimes N},$$

with one copy of the system Hilbert space for each time slice. The fundamental operator is

$\mathcal R=\rho_0\,e^{i\tilde{\mathcal S}},$

where the time history is encoded by a quantum action rather than by a single-slice state (Diaz et al., 10 Jun 2026). This construction treats time as another tensor-factor label and is designed so that ordinary Schrödinger-picture density matrices arise as one-time marginals.

Other formalisms adopt different structural requirements. The pseudo-density-matrix approach represents a “state over time” by a Hermitian, unit-trace operator whose marginals are density matrices but which need not be positive (Fullwood, 2023). The spatiotemporal doubled density operator instead places the process in

$W\in \mathcal B(\mathcal H_L\otimes \mathcal H_R),$

with doubled left/right copies of the local Hilbert spaces and a single Born rule for both spatial and temporal processes (Jia et al., 2023). A recent synthesis argues that path integrals, quantum states over time, pseudo-density matrices, Page–Wootters states, superdensity operators, and timelike-entanglement proposals are different manifestations of one underlying spacetime-state object (Diaz et al., 10 Jun 2026).

A persistent source of confusion is terminological. Some papers using “spacetime density matrix” actually study a boundary reduced density matrix whose dual is a spacetime region, or they use density-matrix language in phase space, generalized geometry, or spacetime algebra. Those usages are related but not equivalent to the multitime operator constructions discussed above (Czech et al., 2012, Perepelkin et al., 2019, Wang, 2018).

2. Principal operator constructions

In the correlator-based construction, the operator $T_{C_0C_1}$ is written explicitly in a basis as a doubled-space object whose matrix elements combine forward evolution, backward evolution, and the initial state. Its defining feature is that the usual density matrices on the individual slices are obtained as partial traces: $$\mathrm{tr}_{C_1} T_{C_0C_1} = \rho_0, \qquad \mathrm{tr}_{C_0} T_{C_0C_1} = \rho(t_1)=U\rho_0 U^\dagger.$$ The many-slice generalization $C_1$0 is defined analogously by requiring it to reproduce the full multi-time anti-time-ordered correlator data (Guo, 28 Aug 2025).

The action-based spacetime-state formalism uses two distinct generators: translation across slices by a SWAP chain and translation within slices by the Hamiltonian. These are combined into the quantum action

$C_1$1

and a related state quantum action $C_1$2. The resulting spacetime state $C_1$3 has unit trace,

$C_1$4

but is generally non-Hermitian because it is built to encode time ordering and causal asymmetry (Diaz et al., 10 Jun 2026). Partial trace over all but one slice gives the ordinary evolved density matrix on that slice.

The pseudo-density-matrix construction begins from a channel-state/Jamiołkowski object and defines the relevant two-time operator by the symmetric bloom

$C_1$5

This operator is Hermitian, classically reducible, and associative, and for finite chains of channels it extends to a canonical $C_1$6-step state-over-time operator with the expected marginals (Fullwood, 2023). The paper further proves that, for dynamically evolving qubit systems, this construction reproduces the Fitzsimons–Jones–Vedral pseudo-density matrix.

The doubled-density-operator formalism uses a doubled correlation tensor $C_1$7 and defines

$C_1$8

In the spatial case, partial trace recovers the usual density matrix on either side,

$C_1$9

whereas in the temporal case the left and right reductions need not be density operators (Jia et al., 2023). The formalism is explicitly designed so that the same operator-valued Born rule,

$T_{C_0C_1}$0

covers both equal-time and temporal measurement statistics.

These constructions agree on a common structural point: the spacetime object is larger than the ordinary density matrix because it stores inter-time correlations. They disagree on what properties should be retained from ordinary states—positivity, Hermiticity, symmetry between time slices, or direct causal ordering—so they should be regarded as distinct but overlapping definitions rather than as merely notational variants.

3. Path-integral, super-operator, and dynamical formulations

A central technical theme is the Schwinger–Keldysh or closed-time-path representation. In the multitime correlator formulation, matrix elements of $T_{C_0C_1}$1 admit a Schwinger–Keldysh path-integral representation with forward and backward branches and cuts at the selected Cauchy surfaces (Guo, 28 Aug 2025). The path-integral picture makes the marginal properties geometrically transparent: tracing out one slice corresponds to gluing the corresponding forward and backward boundaries together.

The same paper introduces a super-operator framework in which the spacetime density matrix appears as

$T_{C_0C_1}$2

with $T_{C_0C_1}$3 encoding the inter-slice evolution. In this language the spacetime density matrix satisfies a Liouville–von Neumann-type equation,

$T_{C_0C_1}$4

while a related dual object obeys a simpler commutator equation (Guo, 28 Aug 2025). The result is a genuine dynamical law for an operator living on the product of time-slice Hilbert spaces.

In quantum field theory, thermo field dynamics and the Schwinger–Keldysh formalism provide a different route to similar objects. The Hilbert space is doubled as

$T_{C_0C_1}$5

and the Liouville equation becomes a Schrödinger-like evolution equation for $T_{C_0C_1}$6. Density-matrix elements are then written directly as matrix elements in the doubled space and rewritten as closed-time-path path integrals over $T_{C_0C_1}$7 fields (Käding et al., 11 Mar 2025). In this formulation, a “spacetime density matrix” is not a finite-dimensional matrix in the first instance but a path-integral/closed-time-path functional encoding density-matrix elements across spacetime fields and branch doubling.

For closed systems, the resulting dynamical map is divisible because the unitary cocycle property holds. For open systems, tracing out the environment generates an influence action $T_{C_0C_1}$8, and the reduced map is generally non-divisible; the paper identifies this non-divisibility with non-Markovianity (Käding et al., 11 Mar 2025). The same formalism yields master equations that contain no time integrals over density matrix elements, a feature emphasized in contrast with many other derivations.

The action-based spacetime-state program places path integrals, channel-state duality, and multitime correlators in a single operator language. Path integrals arise by evaluating $T_{C_0C_1}$9 in a spacetime basis, while several other multitime formalisms emerge from reductions, linear maps, or quantum channels applied to the spacetime state (Diaz et al., 10 Jun 2026). This suggests a broad unification of operator and path-integral viewpoints.

4. Reduced spacetime states, moments, and temporal quantum structure

Reduction is the point at which spacetime density matrices begin to behave most differently from ordinary states. Given factorized Hilbert spaces

$$\mathrm{Tr}(\rho_0\, \mathcal O_0 \mathcal O_1(t_1)) = \mathrm{tr}\!\left(T_{C_0C_1}\,\mathcal O_0 \mathcal O_1\right),$$0

one defines reduced spacetime density matrices by tracing over complementary degrees of freedom on each slice: $$\mathrm{Tr}(\rho_0\, \mathcal O_0 \mathcal O_1(t_1)) = \mathrm{tr}\!\left(T_{C_0C_1}\,\mathcal O_0 \mathcal O_1\right),$$1 For causally disconnected regions these reduced operators are Hermitian and behave much like ordinary density matrices; for causally connected regions they are generally non-Hermitian and may have complex spectra, so the associated entropies are interpreted as pseudoentropy (Guo, 28 Aug 2025).

The same work derives several universal moment formulas for the full two-slice operator: $$\mathrm{Tr}(\rho_0\, \mathcal O_0 \mathcal O_1(t_1)) = \mathrm{tr}\!\left(T_{C_0C_1}\,\mathcal O_0 \mathcal O_1\right),$$2 More generally,

$$\mathrm{Tr}(\rho_0\, \mathcal O_0 \mathcal O_1(t_1)) = \mathrm{tr}\!\left(T_{C_0C_1}\,\mathcal O_0 \mathcal O_1\right),$$3

showing that the global moments are fixed by the initial state and the Hilbert-space dimension rather than by the elapsed time (Guo, 28 Aug 2025). By contrast, moments of reduced spacetime density matrices are dynamical and interaction-sensitive. The paper derives a universal short-time expansion of the second moment and emphasizes that coupling between subsystems is crucial for nontrivial behavior.

In the action-based framework, the anti-Hermitian part of the spacetime state,

$$\mathrm{Tr}(\rho_0\, \mathcal O_0 \mathcal O_1(t_1)) = \mathrm{tr}\!\left(T_{C_0C_1}\,\mathcal O_0 \mathcal O_1\right),$$4

directly controls unequal-time commutators and is bounded by a norm identified with “imagitivity” in related work (Diaz et al., 10 Jun 2026). The Hermitian part controls Leggett–Garg correlators. This makes non-Hermiticity a witness of causal response rather than an incidental pathology. The same paper derives pseudo-density matrices as symmetrized or twirled reductions of a more detailed spacetime state, shows that superdensity operators arise by partial transpose and realignment of a doubled spacetime state, and identifies timelike-entanglement operators as partial traces of the spacetime state.

The doubled-density-operator program uses reduction differently. There, failure of the left or right partial trace to be a density operator is a criterion for temporality: purely spatial states have legitimate reduced density operators on both halves, while temporal processes generally do not (Jia et al., 2023). This is tied to the distinction between full non-signaling in spatial scenarios and one-way non-signaling in time-ordered scenarios. The same formalism is also stated to accommodate indefinite causal order.

The pseudo-density-matrix literature gives a complementary interpretation of non-positivity. The defining properties are Hermiticity, unit trace, and density-matrix marginals, with no requirement of positivity (Fullwood, 2023). In that setting, negative eigenvalues encode temporal quantum structure that cannot be represented as an ordinary joint probability distribution. A recurring theme across these approaches is therefore that deviation from standard density-matrix positivity or Hermiticity is not treated as a defect; it is the mechanism by which temporal and causal information is stored.

5. Holographic and gravitational realizations

In holography, the key question is not usually to define a density matrix of spacetime but to determine how much bulk spacetime is encoded by a boundary reduced density matrix

$$\mathrm{Tr}(\rho_0\, \mathcal O_0 \mathcal O_1(t_1)) = \mathrm{tr}\!\left(T_{C_0C_1}\,\mathcal O_0 \mathcal O_1\right),$$5

The bulk region dual to $$\mathrm{Tr}(\rho_0\, \mathcal O_0 \mathcal O_1(t_1)) = \mathrm{tr}\!\left(T_{C_0C_1}\,\mathcal O_0 \mathcal O_1\right),$$6, denoted $$\mathrm{Tr}(\rho_0\, \mathcal O_0 \mathcal O_1(t_1)) = \mathrm{tr}\!\left(T_{C_0C_1}\,\mathcal O_0 \mathcal O_1\right),$$7, is constrained to depend only on the boundary domain of dependence $$\mathrm{Tr}(\rho_0\, \mathcal O_0 \mathcal O_1(t_1)) = \mathrm{tr}\!\left(T_{C_0C_1}\,\mathcal O_0 \mathcal O_1\right),$$8, to be causally closed, and to avoid causal overlap with reconstructible regions of disjoint boundary subsets (Czech et al., 2012). Two geometric candidates are emphasized: the causal wedge $$\mathrm{Tr}(\rho_0\, \mathcal O_0 \mathcal O_1(t_1)) = \mathrm{tr}\!\left(T_{C_0C_1}\,\mathcal O_0 \mathcal O_1\right),$$9 and the entanglement wedge $T_{C_0C_1}$0. The causal wedge is presented as a conservative lower bound, while the entanglement wedge is a richer candidate because entanglement entropies accessible from $T_{C_0C_1}$1 probe extremal surfaces lying outside the causal wedge. The same paper also stresses that $T_{C_0C_1}$2 need not always be the final answer; in some star-like geometries it can be too small.

A more literal gravitational density-matrix construction appears in elliptic de Sitter $T_{C_0C_1}$3. Because the Euclidean continuation is $T_{C_0C_1}$4, not a sphere, the manifold does not admit a reflection that splits it into two disconnected halves. As a result, the Euclidean path integral does not define a no-boundary wavefunction in the usual sense, and the proposal is instead that it defines a no-boundary density matrix (Dulac et al., 30 Nov 2025). For a region $T_{C_0C_1}$5, the matrix elements are given by a Euclidean path integral with boundary data on the two sides of a slit, and

$T_{C_0C_1}$6

In the free Dirac fermion example in two dimensions, the von Neumann and Rényi entropies can be computed exactly from twist-operator correlators on non-orientable surfaces. A striking accompanying result is that the global Hilbert space is one-dimensional, whereas the Hilbert space seen by any observer inside a static patch is a nontrivial Fock space (Dulac et al., 30 Nov 2025).

In loop-quantum-gravity-inspired black-hole physics, a different spacetime density-matrix idea is used. A coherent-state density operator is constructed for a semiclassical black-hole slice, split into outside, horizon, and inside sectors, and reduced by tracing over inaccessible interior degrees of freedom (Dasgupta, 2010). Time evolution is then generated not by a standard Hermitian Hamiltonian but by a non-Hermitian quasilocal-energy horizon Hamiltonian,

$T_{C_0C_1}$7

The resulting change in entropy is interpreted as the origin of Hawking radiation. Here the “density matrix” refers to the quantum geometry of a slice rather than to matter on a fixed background.

Taken together, these gravitational examples show that the phrase can refer either to an operator whose support is explicitly multitemporal, to a Euclidean mixed-state preparation of a spacetime region, or to a reduced state of quantum geometry itself. The common thread is that partial access, causal restriction, or nontrivial global spacetime structure obstructs a purely single-slice pure-state description.

6. Related but non-equivalent usages

Several papers deploy density-matrix language in ways that are adjacent to, but distinct from, the multitime operator formalism. One example is the “universal density matrix for the phase space,” in which the Wigner function of an arbitrary quantum system is written as

$T_{C_0C_1}$8

Here $T_{C_0C_1}$9 is system-specific and $$\mathcal H=\bigotimes_{t=0}^{N-1} h_t \simeq h^{\otimes N},$$0 is a universal phase-space kernel basis built from harmonic-oscillator functions (Perepelkin et al., 2019). The diagonal elements of $$\mathcal H=\bigotimes_{t=0}^{N-1} h_t \simeq h^{\otimes N},$$1 are the Wigner functions of harmonic-oscillator eigenstates, while the off-diagonal elements contain oscillatory factors associated in the paper with dissipation. The paper explicitly concerns phase space rather than spacetime, so it is best regarded as a phase-space analog rather than a literal spacetime density matrix.

A different neighboring concept is the “thermal entropy density of spacetime,” defined by

$$\mathcal H=\bigotimes_{t=0}^{N-1} h_t \simeq h^{\otimes N},$$2

Using Einstein’s equations and a $$\mathcal H=\bigotimes_{t=0}^{N-1} h_t \simeq h^{\otimes N},$$3 decomposition, this quantity is rewritten in terms of curvature and extrinsic-curvature data in arbitrary spacetime, without assuming a horizon or temperature in advance (Yang, 2011). The paper suggests that geometry and thermodynamics are interchangeable descriptions and that spacetime may admit a deeper statistical interpretation, but it does not derive a literal density matrix.

Similarly, the surface density of spacetime degrees of freedom has been reinterpreted as a projected phase space adapted to accelerating observers. The construction uses a foliation along the direction of acceleration, projects the resulting canonical pair along the observer’s velocity, and integrates over one Euclidean thermal period to recover Padmanabhan’s density formula (Hadad, 2016). This supports a density-matrix-style picture of observer-dependent spacetime microstates, but the object itself is a canonical phase-space density rather than a density matrix on multiple Cauchy surfaces.

A further terminological divergence appears in the multiparticle spacetime algebra formulation of qubit density operators. There, density operators are represented as multivectors in the algebra $$\mathcal H=\bigotimes_{t=0}^{N-1} h_t \simeq h^{\otimes N},$$4, with the complex structure encoded by a correlated pseudoscalar and local unitary operations represented as rotations (Wang, 2018). Despite the phrase “spacetime algebra,” the subject is not multitime quantum states; it is a geometric-algebra reformulation of ordinary qubit density operators and their invariants.

The broader literature therefore supports a restricted conclusion. A spacetime density matrix in the strict sense is a multitime operator that generalizes the single-slice density matrix and reproduces cross-time correlators. Closely related constructions may be Hermitian or non-Hermitian, positive or non-positive, doubled or undoubled, and may arise from channels, quantum actions, or Schwinger–Keldysh path integrals (Guo, 28 Aug 2025, Diaz et al., 10 Jun 2026, Fullwood, 2023, Jia et al., 2023). Beyond that core, the same phrase can also denote holographic, gravitational, thermodynamic, or phase-space analogs, and careful distinction among these usages is essential.