Time Scale Entanglement in Quantum Systems

• Time scale entanglement is the study of ordered temporal correlations in quantum systems that resist description by a single density operator.
• It employs techniques like tensor-train decompositions to compress multi-time correlators, with bond dimensions indicating phase transitions and crossover phenomena.
• Research spans influence matrices, quantum-clock history states, and timelike operator constructions to provide varied operational views of temporal dynamics.

Time scale entanglement denotes a family of research concepts in which temporal structure itself is treated as an entanglement-bearing object rather than merely a parameter of evolution. In the narrow sense introduced through correlator compression, it is the intrinsic, non-separable correlation among the ordered times $\tau_1,\ldots,\tau_n$ in multi-point imaginary-time correlators, diagnosed by the bond dimension required in a quantics tensor-train representation (Rohshap et al., 15 Jul 2025). Closely related literatures define temporal entanglement for influence matrices in space-time duality (Foligno et al., 2023), system-time entanglement for history states of quantum clocks (Boette et al., 2015), and entanglement in time for timelike-separated subsystems or non-Hermitian reduced operators (Narayan et al., 2023, Milekhin et al., 17 Feb 2025). The shared theme is that quantum correlations may be organized across ordered times, histories, or causal separations, but the underlying states, entropy functionals, and operational interpretations are not identical.

1. Terminological scope and conceptual variants

The modern literature uses several near-neighbor notions that are sometimes conflated. A recurrent source of confusion is terminological rather than physical: the same broad phrase can refer to compression complexity of imaginary-time correlators, entanglement of influence matrices along the temporal direction, bipartite entanglement between a system and a quantum clock, or entropy assigned to timelike operator insertions. These constructions agree that time can be treated as part of the quantum information structure, but they do not share a single universal density operator or entanglement monotone (Rohshap et al., 15 Jul 2025, Foligno et al., 2023, Boette et al., 2015, Milekhin et al., 17 Feb 2025).

Setting	Central object	Representative quantity
Imaginary-time correlator compression	$C(\tau_1,\ldots,\tau_n)$ in QTT form	$E_{\text{time}}=\max_k \chi_k$
Space-time dual circuit dynamics	Influence matrix fixed point	$S_\alpha(t)$ across temporal cuts
Quantum-clock history state	$$|\Psi\rangle=\frac{1}{\sqrt N}\sum_t |\psi_t\rangle\otimes|t\rangle$$	$E(S{:}T)=S(\rho_S)$
Timelike operator construction	Reduced non-Hermitian $\rho_A(t)$ or spacetime matrix $T_{AB}$	$S_A(t)$, $\mathrm{Tr}\,T_{AB}^n$, $C(\tau_1,\ldots,\tau_n)$0

In the correlator-based usage, low time-scale entanglement means that a multi-time correlator factorizes into fewer independent time functions, whereas high time-scale entanglement means that many ordered times are irreducibly linked and require a large tensor-train bond dimension (Rohshap et al., 15 Jul 2025). In the influence-matrix usage, temporal entanglement measures the entanglement of a state defined on temporal sites after a space-time swap in a folded circuit (Foligno et al., 2023). In the system-time usage, entanglement quantifies how distinguishable a discrete evolution is when encoded into a joint state of the system and a reference clock (Boette et al., 2015). In operatorial timelike constructions, the entropy can be complex because the reduced operator is non-Hermitian or because Lorentzian continuation introduces a branch cut (Narayan et al., 2023).

A plausible implication is that “time scale entanglement” is best understood as an umbrella label for several technically distinct programs rather than as a single invariant.

2. Correlator compression and the QTT definition

The most explicit definition of time-scale entanglement is formulated for multi-point imaginary-time correlators

$C(\tau_1,\ldots,\tau_n)$1

In this framework, time-scale entanglement is the intrinsic, non-separable correlation among the ordered times $C(\tau_1,\ldots,\tau_n)$2 that forces the correlator to have a high compressed rank when treated as a function of its $C(\tau_1,\ldots,\tau_n)$3 time arguments (Rohshap et al., 15 Jul 2025). After discretizing each $C(\tau_1,\ldots,\tau_n)$4 on a grid of $C(\tau_1,\ldots,\tau_n)$5 points, one obtains an $C(\tau_1,\ldots,\tau_n)$6-dimensional tensor

$C(\tau_1,\ldots,\tau_n)$7

which is represented in quantics tensor-train form as

$C(\tau_1,\ldots,\tau_n)$8

The bond dimension $C(\tau_1,\ldots,\tau_n)$9 measures the effective number of independent correlations linking $E_{\text{time}}=\max_k \chi_k$0 to $E_{\text{time}}=\max_k \chi_k$1, and a scalar diagnostic is

$E_{\text{time}}=\max_k \chi_k$2

The algorithmic realization is a standard TT-SVD or quantics-TT-SVD. The correlator tensor is reshaped successively into matrices, singular-value decompositions are performed, and singular values below a cutoff $E_{\text{time}}=\max_k \chi_k$3 are discarded. Stable plateaus of $E_{\text{time}}=\max_k \chi_k$4 over a range such as $E_{\text{time}}=\max_k \chi_k$5 indicate genuine many-time structure, whereas spurious bonds disappear once $E_{\text{time}}=\max_k \chi_k$6 is moved above the noise floor (Rohshap et al., 15 Jul 2025).

This construction was verified on the Hubbard dimer, the four-site Hubbard ring with and without next-nearest-neighbor hopping, and the single-impurity Anderson model. In these examples, peaks in the maximum bond dimension diagnose ground-state crossings, thermal crossovers, and the Kondo-crossover line. The paper reports that time-scale entanglement is a system-inherent property becoming maximal at phase transitions and crossover, and that this becomes visible as peaks in the QTT bond dimension of correlator functions (Rohshap et al., 15 Jul 2025). The same study also relates these peaks to quantum Fisher information, negativity, mutual-information-type quantities, and charge or spin susceptibilities.

A common misconception is that this notion is simply another entanglement entropy. It is not: the primary observable is a compression rank or bond dimension extracted from a tensor-network factorization of correlators rather than a von Neumann entropy of a physical density matrix.

3. Temporal entanglement of influence matrices

A second major line of work studies temporal entanglement through space-time duality in one-dimensional quantum circuits. After folding the doubled circuit and exchanging space and time, the infinite-volume dynamics is encoded in dominant fixed points of a space transfer matrix, known as influence matrices (Foligno et al., 2023). One then bipartitions the temporal sites of the influence-matrix state and defines

$E_{\text{time}}=\max_k \chi_k$7

with $E_{\text{time}}=\max_k \chi_k$8 obtained in the $E_{\text{time}}=\max_k \chi_k$9 limit.

In generic chaotic circuits, temporal entanglement follows a volume law in time. The 2023 analysis found that the von Neumann entropy grows linearly and that, for generic space-like evolution, higher Rényi entropies also grow linearly (Foligno et al., 2023). Two marginal cases were isolated. For pure space evolution in generic chaotic systems, one has

$S_\alpha(t)$0

For any space-like evolution in dual-unitary circuits, one has

$S_\alpha(t)$1

The mechanism is a large overlap of the influence matrix with a suitable product state, which constrains the largest Schmidt value and suppresses higher Rényi growth even though the von Neumann entropy remains extensive (Foligno et al., 2023).

The later study of temporal entanglement transitions sharpened the simulation-theoretic interpretation. In a random-unitary bath model, temporal entanglement is extensive for low enough bath growth rate and reflects genuine non-Markovianity, but this memory is concentrated in highly complex temporal correlations whose effect on few-point temporal correlators is negligible (Vilkoviskiy et al., 5 Nov 2025). An influence-matrix coarse-graining that reduces the allowed measurement frequency produces a transition from volume-law to area-law temporal entanglement. The same transition was shown in the kicked Ising model analytically at dual-unitary points and numerically away from them, and the compact influence-matrix MPS obtained through standard compression algorithms was found to accurately describe local evolution (Vilkoviskiy et al., 5 Nov 2025).

This literature directly addresses a common misunderstanding: volume-law temporal entanglement does not automatically imply that local observables are intractable. The coarse-grained influence matrix can obey an area law even when the exact influence matrix shows volume-law scaling.

4. Quantum clocks, timelike operators, and pseudo-entropy

A distinct program treats time as a quantum subsystem. In the discrete-time model, the joint history state

$S_\alpha(t)$2

encodes the evolution of a system $S_\alpha(t)$3 relative to a quantum clock $S_\alpha(t)$4 (Boette et al., 2015). The corresponding system-time entanglement is

$S_\alpha(t)$5

This quantity vanishes for stationary evolution, reaches $S_\alpha(t)$6 when the system jumps onto a new orthogonal state at each time step, and for a constant-Hamiltonian cyclic evolution reduces to the Shannon entropy of the initial energy distribution. The same work gives the entropic energy-time uncertainty relation

$S_\alpha(t)$7

and shows that for a qubit clock the linear entropy satisfies

$S_\alpha(t)$8

in the two-step case (Boette et al., 2015).

Another construction promotes the unitary time-evolution operator itself to a normalized, generally non-Hermitian operator,

$S_\alpha(t)$9

or alternatively $$|\Psi\rangle=\frac{1}{\sqrt N}\sum_t |\psi_t\rangle\otimes|t\rangle$$0, and then traces over a subsystem to obtain $$|\Psi\rangle=\frac{1}{\sqrt N}\sum_t |\psi_t\rangle\otimes|t\rangle$$1 or $$|\Psi\rangle=\frac{1}{\sqrt N}\sum_t |\psi_t\rangle\otimes|t\rangle$$2 (Narayan et al., 2023). The formal entropy

$$|\Psi\rangle=\frac{1}{\sqrt N}\sum_t |\psi_t\rangle\otimes|t\rangle$$3

is generically complex because $$|\Psi\rangle=\frac{1}{\sqrt N}\sum_t |\psi_t\rangle\otimes|t\rangle$$4 need not be Hermitian and may have complex eigenvalues. In a 2D CFT, the replica continuation for a timelike interval yields

$$|\Psi\rangle=\frac{1}{\sqrt N}\sum_t |\psi_t\rangle\otimes|t\rangle$$5

so the imaginary part is a universal shift coming from $$|\Psi\rangle=\frac{1}{\sqrt N}\sum_t |\psi_t\rangle\otimes|t\rangle$$6 (Narayan et al., 2023). The same paper identifies real subfamilies when only a single nontrivial phase survives and shows that projecting $$|\Psi\rangle=\frac{1}{\sqrt N}\sum_t |\psi_t\rangle\otimes|t\rangle$$7 onto an initial state reproduces pseudo-entropy for the pair $$|\Psi\rangle=\frac{1}{\sqrt N}\sum_t |\psi_t\rangle\otimes|t\rangle$$8 and $$|\Psi\rangle=\frac{1}{\sqrt N}\sum_t |\psi_t\rangle\otimes|t\rangle$$9.

A more operational formulation defines a spacetime density matrix $E(S{:}T)=S(\rho_S)$0 through the Wightman correlator

$E(S{:}T)=S(\rho_S)$1

This leads to timelike analogues of Rényi data,

$E(S{:}T)=S(\rho_S)$2

and to the entanglement $E(S{:}T)=S(\rho_S)$3-imagitivity

$E(S{:}T)=S(\rho_S)$4

These quantities bound timelike correlators and commutators, agree with analytic continuation from spacelike to timelike separation in relativistic QFT, admit explicit measurement protocols based on controlled-SWAP tests, and were implemented on the IBM device ibm_sherbrooke for a simple qubit system (Milekhin et al., 17 Feb 2025).

Taken together, these constructions show that temporal entanglement need not be associated with an ordinary Hermitian reduced density matrix, and that complex-valued entropic quantities are not pathologies but expected features of Lorentzian or non-Hermitian formulations.

5. Time scales of entanglement generation, freezing, and relaxation

Adjacent literature shifts attention from entanglement of temporal degrees of freedom to the characteristic time scales governing entanglement dynamics. For two initially unentangled subsystems coupled by

$E(S{:}T)=S(\rho_S)$5

the short-time purity decay of subsystem $E(S{:}T)=S(\rho_S)$6 defines the entanglement timescale

$E(S{:}T)=S(\rho_S)$7

Thus $E(S{:}T)=S(\rho_S)$8, where $E(S{:}T)=S(\rho_S)$9 is the correlated quantum uncertainty (Yang, 2017). For a single coupling channel $\rho_A(t)$0, this reduces to $\rho_A(t)$1.

In Markovian open systems with two subsystems $\rho_A(t)$2 and $\rho_A(t)$3 coupled only through a common reservoir, a different time-entanglement relation emerges. For pure steady states, the more steady-state entanglement one engineers, the slower the dissipative preparation dynamics must be. Writing the scaled entanglement as

$\rho_A(t)$4

the dissipative gap obeys a bound of the form

$\rho_A(t)$5

so the relaxation time $\rho_A(t)$6 must grow at least like $\rho_A(t)$7, diverging as $\rho_A(t)$8 (Pocklington et al., 2024). The physical explanation is an emergent strong symmetry when the target approaches a maximally entangled dark state.

The many-body open-system literature also contains a distinct notion of frozen entanglement over a finite interval. For a one-dimensional spin chain under local Markovian noise, logarithmic negativity $\rho_A(t)$9 is said to be frozen on $T_{AB}$0 when

$T_{AB}$1

with $T_{AB}$2 (Chanda et al., 2016). In the paramagnetic phases of the ATXY model, the freezing interval satisfies

$T_{AB}$3

so the freezing time of a fixed nearest-neighbor link is scale-invariant with respect to chain length. For links sufficiently far from the noisy boundary, numerics yield

$T_{AB}$4

with the example values $T_{AB}$5, $T_{AB}$6, $T_{AB}$7, in contrast to the linear dependence suggested by a Lieb-Robinson estimate (Chanda et al., 2016).

Scale invariance in the Hamiltonian can also modify the time growth law of ordinary spatial entanglement after a quench. In one-dimensional integrable theories with dispersion $T_{AB}$8, $T_{AB}$9, entanglement grows linearly at early times because of fast modes and then crosses over to a slow-mode regime with

$S_A(t)$0

together with a logarithmic enhancement in the bosonic case (Mozaffar et al., 2021). Although this is not a definition of time-scale entanglement in the QTT sense, it directly concerns how distinct time scales are distributed across entangling quasiparticle sectors.

6. Relativistic, cosmological, and applied extensions

Several works embed temporal entanglement into relativistic or mode-entanglement settings. In the two-oscillator Lorentz-covariant construction, the longitudinal separation $S_A(t)$1 and time-separation $S_A(t)$2 of a bound state form a Gaussian state that becomes squeezed under a boost. The boosted wave function can be expanded as

$S_A(t)$3

so the spatial and time-like modes are entangled (Kim et al., 2014). Tracing out the unobserved time-separation produces a mixed reduced state with entropy

$S_A(t)$4

and comparison with a thermal oscillator yields

$S_A(t)$5

or equivalently $S_A(t)$6 in dimensionless units [(Kim et al., 2014); (Baskal et al., 2016)]. These analyses interpret the unobserved time-like coordinate as part of Feynman’s “rest of the universe.”

A broader conceptual extension appears in the Page–Wootters-style account of time as entanglement. There the stationary global state $S_A(t)$7 on $S_A(t)$8 satisfies $S_A(t)$9, and projection onto a clock state $\mathrm{Tr}\,T_{AB}^n$0 yields a relative system state obeying the Schrödinger equation. The same paper argues that if Schmidt weights acquire the form $\mathrm{Tr}\,T_{AB}^n$1, the reduced clock state becomes thermal with $\mathrm{Tr}\,T_{AB}^n$2, and it further extends this logic to inflationary two-mode squeezed states and modewise effective temperatures $\mathrm{Tr}\,T_{AB}^n$3 fixed by squeezing $\mathrm{Tr}\,T_{AB}^n$4 (Vedral, 2014).

In quantum-simulation settings, a continuous-time quantum walk model for one-dimensional Dirac dynamics uses the time interval $\mathrm{Tr}\,T_{AB}^n$5 as a direct control parameter for entanglement between internal and external spaces. The reduced coin state $\mathrm{Tr}\,T_{AB}^n$6 defines the entanglement entropy

$\mathrm{Tr}\,T_{AB}^n$7

and the study reports that $\mathrm{Tr}\,T_{AB}^n$8 yields nearest-neighbor coupling and the largest oscillation amplitude of $\mathrm{Tr}\,T_{AB}^n$9, whereas $C(\tau_1,\ldots,\tau_n)$00 gives longer-range but weaker nearest-step weight and smaller entanglement amplitude (Wang et al., 2024).

A technologically oriented extension uses high-dimensional time-bin entanglement. In an $C(\tau_1,\ldots,\tau_n)$01-dimensional time-bin encoding, a maximally entangled two-qudit state is written as

$C(\tau_1,\ldots,\tau_n)$02

and the corresponding high-dimensional Bell basis introduces a shift or phase index (Aktaş et al., 23 Dec 2025). In the proposed blockchain application, measurement times become cryptographic anchors, high-dimensional Bell-state measurements at successive times generate block-dependent key material, and perturbing the ordering of measurements modifies coincidence structure and reveals tampering (Aktaş et al., 23 Dec 2025). This suggests that in some applied contexts “time-entanglement” functions simultaneously as a correlation resource and as a causal-order resource.

Across these relativistic and applied directions, the literature consistently treats temporal structure not as a passive background but as a quantum degree of freedom, a compression axis, a causal label, or an operational resource. The resulting objects range from bond dimensions and Rényi entropies to complex pseudo-entropies, effective temperatures, and protocol-level witnesses, and the diversity of these constructions is itself a defining feature of the subject.