Bipartite Temporal Bell Inequality Overview

• Bipartite temporal Bell inequality is a quantum test framework where time labels replace spatial measurement settings to probe nonlocal correlations.
• It encompasses multiple formulations including continuous-variable tests with time-dependent correlators, many-body spin-chain setups obeying Lieb–Robinson bounds, and time-resolved biphoton experiments.
• The framework builds a bridge between spatial Bell tests and temporal correlations through state–evolution mapping, highlighting constraints such as finite-speed information propagation.

A bipartite temporal Bell inequality is a Bell-type construction in which bipartiteness is retained while time enters the operational definition of the test. In the arXiv literature, the expression is used in several non-equivalent ways. One line of work replaces the usual local setting choices in a CHSH experiment by choices of measurement times on two spatially separated subsystems (Ando et al., 2020). Another defines the two effective parties as an early and a later measurement event connected only through a many-body medium with finite propagation speed, leading to a temporal Clauser–Horne inequality (Tononi et al., 2024). A looser but experimentally important usage applies standard CHSH directly to time-resolved biphoton coincidence distributions, so that Bell nonlocality is read out from a nonlocal temporal correlation function (Guo et al., 2016). A foundational bridge between spatial Bell tests and temporal Bell-in-time or Leggett–Garg-type inequalities is provided by a correlation-preserving state–evolution map (Marcovitch et al., 2011).

1. Terminological scope and distinct formulations

The term does not denote a single formalism. In the continuous-variable formulation, the system remains bipartite in space and time takes over the role usually played by detector settings. The same dichotomic observable is measured on subsystem $1$ at either $t_a$ or $t_a'$, and on subsystem $2$ at either $t_b$ or $t_b'$, so the Bell combination is still of CHSH type but the “settings” are temporal labels (Ando et al., 2020). In the many-body formulation, Alice measures at time $T=0$ and Bob at a later time $T=t$, so the bipartition is operationally between an early and a later measurement event; the nontriviality comes from the assumption that Alice-to-Bob influence can propagate only through a spin chain obeying a Lieb–Robinson bound (Tononi et al., 2024). In the narrowband biphoton experiment, the relevant object is the arrival-time-resolved two-photon coincidence function, and the paper explicitly interprets the result as a standard bipartite CHSH test on a two-photon entangled state rather than as a distinct Leggett–Garg-type temporal Bell framework (Guo et al., 2016). The state–evolution correspondence of Marcovitch and Reznik shows how an ordinary bipartite Bell inequality can be represented as a temporal one, with the two spatial parties mapped to past and future of a single evolving system (Marcovitch et al., 2011).

Formulation	Operational bipartition	Bell structure
Time as setting	Two spatial subsystems	CHSH
Early/later events through a chain	Two measurement events at different times	CH
Time-resolved biphoton readout	Two photons, correlation resolved in delay time	CHSH

These formulations overlap conceptually but are not interchangeable. In particular, “temporal” may refer to unequal-time setting choices, to sequential measurements constrained by a propagation medium, or to a time-resolved observable used inside an otherwise standard spatial Bell test.

2. Bell forms, correlators, and hidden-variable premises

The continuous-variable construction keeps the ordinary CHSH algebra,

$$B=E(t_a,t_b)+E(t_a,t_b')+E(t_a',t_b)-E(t_a',t_b')\le 2,$$

but defines each correlator from projective unequal-time measurements of the same dichotomic observable on the two subsystems. Because the Heisenberg-picture operators at different times need not commute, the correlator is written as

$$E(t_a,t_b)=\frac{1}{2}\left\langle \psi \left| \left\{ \hat S_1(t_a),\hat S_2(t_b)\right\}\right|\psi\right\rangle .$$

The cosmological squeezed-coherent-state analysis uses the same anti-commutator structure, writing a temporal Bell operator $t_a$0 whose expectation value obeys the classical bound $t_a$1 and the Tsirelson bound $t_a$2 (Ando et al., 2020, Mondal, 8 May 2026).

The many-body construction is explicitly a temporal Clauser–Horne inequality,

$t_a$3

with

$t_a$4

Here Alice measures $t_a$5 or $t_a$6 at time $t_a$7, Bob measures $t_a$8 or $t_a$9 at time $t_a'$0, and Bob’s operators evolve as $t_a'$1 (Tononi et al., 2024).

The time-resolved biphoton experiment instead evaluates the CHSH parameter as a function of relative arrival-time delay $t_a'$2,

$t_a'$3

with local realistic bound $t_a'$4 (Guo et al., 2016).

The hidden-variable assumptions differ correspondingly. In the continuous-variable and cosmological CHSH-type formulations, the subsystems remain spatially separated, so the derivation is based on realism and locality rather than realism plus non-invasiveness (Ando et al., 2020, Mondal, 8 May 2026). In the many-body case, the paper proves that without a propagation-medium constraint a local hidden-variable description exists for sequential probabilities; the relevant assumption is therefore that information can propagate only through the chain and not faster than the effective many-body propagation bound (Tononi et al., 2024). In the biphoton experiment, the authors explicitly take the fair-sampling assumption, so the detection loophole remains open even though locality separation is discussed (Guo et al., 2016).

3. Continuous-variable CHSH with time as the measurement setting

For two-mode squeezed states, the relevant observable on each subsystem is a coarse-grained sign-binned position operator,

$t_a'$5

which becomes the sign operator in the $t_a'$6 limit. Operationally, one measures $t_a'$7, identifies the interval $t_a'$8, and outputs $t_a'$9. This makes the Bell test depend only on position measurements, a feature emphasized as useful for continuous-variable systems and for cosmological situations in which momentum-like quadratures are effectively inaccessible (Ando et al., 2020).

The quantum state is a two-mode squeezed state generated by a squeezing operator $2$0 and a rotation operator $2$1. The three state parameters are the squeezing amplitude $2$2, squeezing angle $2$3, and rotation angle $2$4. A central point is that $2$5 drops out of the single-time state vector because the vacuum is rotation invariant, but it re-enters unequal-time correlators through the propagator $2$6. The correlator depends only on the relative rotation angle $2$7, and the paper identifies this as one of the main conceptual novelties of the multiple-time setting (Ando et al., 2020).

The analysis yields exact Gaussian-integral expressions for $2$8, together with asymptotic formulas in the large-$2$9, small-$t_b$0, and large-squeezing regimes. Several numerical conclusions are explicit. No Bell violation was found in the large-$t_b$1 or pure sign-binning limit. Violations appear at intermediate $t_b$2, roughly when $t_b$3. Sufficiently large squeezing appears necessary: at $t_b$4, the maximum is reported as $t_b$5 up to numerical precision, whereas at $t_b$6 clear violations occur. In representative parameter slices, the reported maxima are $t_b$7 for $t_b$8 and $t_b$9 for $t_b'$0, with violation regions appearing as islands in the space of rotation-angle differences (Ando et al., 2020).

These results establish a specifically bipartite temporal alternative to standard continuous-variable Bell tests. Relative to ordinary CHSH, only one observable per wing is required. Relative to Leggett–Garg inequalities, the system remains bipartite and locality remains a meaningful premise.

4. Temporal Clauser–Horne inequality in non-relativistic many-body systems

The many-body formulation considers an open XX spin-$t_b'$1 chain of length $t_b'$2, with Alice acting on site $t_b'$3 at time $t_b'$4 and Bob on site $t_b'$5 at time $t_b'$6. The Hamiltonian is

$t_b'$7

The initial state places the end spins in the Bell state

$t_b'$8

with all intermediate spins down. The measurement settings are chosen as

$t_b'$9

$T=0$0

For these settings, the Bell pair maximally violates the temporal CH inequality at $T=0$1,

$T=0$2

(Tononi et al., 2024).

The XX chain is solved by a Jordan–Wigner transformation to free fermions. In terms of the single-particle propagator $T=0$3, the exact Bell parameter is

$T=0$4

This makes the Bell quantity directly sensitive to two propagation amplitudes: survival at Bob’s edge and transport between the two edges (Tononi et al., 2024).

The principal physical result is that the inequality remains violated for a finite nonzero interval,

$T=0$5

with the short-time estimate

$T=0$6

The paper argues that this estimate effectively holds for any $T=0$7, since the short-time dynamics is dominated by local processes near Bob’s end. Violation revivals also occur at later times, roughly around $T=0$8 and $T=0$9, associated respectively with peaks in $T=t$0 and $T=t$1. As $T=t$2 grows, the initial violation window remains essentially unchanged, but long-time revivals become less frequent and are suppressed. In the thermodynamic regime $T=t$3, the Bell parameter tends to flatten around $T=t$4 (Tononi et al., 2024).

The paper’s conceptual claim is deliberately limited. Temporal correlations between causally connected events are ordinarily compatible with a local hidden-variable description. The nontriviality here comes from the assumption that the chain is the sole medium of propagation. Under that assumption, temporal CH violation becomes an operational probe of constrained quantum information spreading, with the relevant velocity scale identified as the group velocity $T=t$5, linked qualitatively to the Lieb–Robinson bound (Tononi et al., 2024).

5. Time-resolved CHSH on frequency-bin entangled biphotons

The photonic realization uses spontaneous four-wave mixing in laser-cooled $T=t$6 atoms held in a 2D MOT. Counter-propagating pump and coupling lasers generate paired Stokes and anti-Stokes photons into two symmetric spatial paths. The biphotons are narrowband, with measured coherence time about $T=t$7 and bandwidth about $T=t$8. In path $T=t$9, the Stokes photon is shifted by $$B=E(t_a,t_b)+E(t_a,t_b')+E(t_a',t_b)-E(t_a',t_b')\le 2,$$0 and given a phase $$B=E(t_a,t_b)+E(t_a,t_b')+E(t_a',t_b)-E(t_a',t_b')\le 2,$$1; in path $$B=E(t_a,t_b)+E(t_a,t_b')+E(t_a',t_b)-E(t_a',t_b')\le 2,$$2, the anti-Stokes photon is shifted by the same $$B=E(t_a,t_b)+E(t_a,t_b')+E(t_a',t_b)-E(t_a',t_b')\le 2,$$3 and given a phase $$B=E(t_a,t_b)+E(t_a,t_b')+E(t_a',t_b)-E(t_a',t_b')\le 2,$$4. After recombination at two $$B=E(t_a,t_b)+E(t_a,t_b')+E(t_a',t_b)-E(t_a',t_b')\le 2,$$5 beam splitters, the output state is a frequency-bin entangled superposition in which the entanglement resides in the degree of freedom “which photon carries the frequency shift” (Guo et al., 2016).

Because the photons are long-lived in time, the interference between the two frequency-bin amplitudes appears as a temporal beating in the coincidence rate. The second-order correlation function contains the oscillatory factor

$$B=E(t_a,t_b)+E(t_a,t_b')+E(t_a',t_b)-E(t_a',t_b')\le 2,$$6

with $$B=E(t_a,t_b)+E(t_a,t_b')+E(t_a',t_b)-E(t_a',t_b')\le 2,$$7. The normalized visibility is

$$B=E(t_a,t_b)+E(t_a,t_b')+E(t_a',t_b)-E(t_a',t_b')\le 2,$$8

so accidental coincidences $$B=E(t_a,t_b)+E(t_a,t_b')+E(t_a',t_b)-E(t_a',t_b')\le 2,$$9 directly suppress Bell violation. The detector-port mapping to dichotomic outcomes is defined by assigning $$E(t_a,t_b)=\frac{1}{2}\left\langle \psi \left| \left\{ \hat S_1(t_a),\hat S_2(t_b)\right\}\right|\psi\right\rangle .$$0 to equal-sign port pairs $$E(t_a,t_b)=\frac{1}{2}\left\langle \psi \left| \left\{ \hat S_1(t_a),\hat S_2(t_b)\right\}\right|\psi\right\rangle .$$1 or $$E(t_a,t_b)=\frac{1}{2}\left\langle \psi \left| \left\{ \hat S_1(t_a),\hat S_2(t_b)\right\}\right|\psi\right\rangle .$$2, and $$E(t_a,t_b)=\frac{1}{2}\left\langle \psi \left| \left\{ \hat S_1(t_a),\hat S_2(t_b)\right\}\right|\psi\right\rangle .$$3 to opposite-sign pairs $$E(t_a,t_b)=\frac{1}{2}\left\langle \psi \left| \left\{ \hat S_1(t_a),\hat S_2(t_b)\right\}\right|\psi\right\rangle .$$4 or $$E(t_a,t_b)=\frac{1}{2}\left\langle \psi \left| \left\{ \hat S_1(t_a),\hat S_2(t_b)\right\}\right|\psi\right\rangle .$$5 (Guo et al., 2016).

For the phase choices

$$E(t_a,t_b)=\frac{1}{2}\left\langle \psi \left| \left\{ \hat S_1(t_a),\hat S_2(t_b)\right\}\right|\psi\right\rangle .$$6

the measured $$E(t_a,t_b)=\frac{1}{2}\left\langle \psi \left| \left\{ \hat S_1(t_a),\hat S_2(t_b)\right\}\right|\psi\right\rangle .$$7 oscillates sinusoidally with envelope $$E(t_a,t_b)=\frac{1}{2}\left\langle \psi \left| \left\{ \hat S_1(t_a),\hat S_2(t_b)\right\}\right|\psi\right\rangle .$$8. The reported maximum violation occurs at

$$E(t_a,t_b)=\frac{1}{2}\left\langle \psi \left| \left\{ \hat S_1(t_a),\hat S_2(t_b)\right\}\right|\psi\right\rangle .$$9

equivalently $t_a$00. By adjusting phase settings, the paper reports Bell violation over the range

$t_a$01

where $t_a$02. At fixed $t_a$03, the phase-scan visibilities are

$t_a$04

leading to

$t_a$05

In a normalized temporal-beating analysis over a $t_a$06 interval, the reported visibilities are $t_a$07 and $t_a$08, corresponding to $t_a$09 and $t_a$10 (Guo et al., 2016).

The conceptual status of this experiment is precise. The paper describes a CHSH Bell test on a two-photon entangled state, evaluated on time-resolved coincidence distributions. It is therefore genuinely temporal in the sense that $t_a$11 is measured bin by bin in relative arrival time, but it is not a separate formal Bell structure analogous to Leggett–Garg. The paper also notes several limitations: fair sampling is assumed, the detection loophole remains open, the visibility is reduced by stray light, dark counts, and spontaneous-pair-generation accidentals, and the analysis depends on selected coincidence events in time bins (Guo et al., 2016).

6. State–evolution correspondence and the Bell–Leggett–Garg bridge

Marcovitch and Reznik derive an exact correspondence between bipartite spatial correlations and temporal correlations in a single system undergoing an evolution. A pure bipartite state

$t_a$12

is mapped to a temporal operator

$t_a$13

and mixed states map by convexity to mixtures of evolutions. Under this map, maximally entangled states correspond to unitary evolutions, product states correspond to selective projector-like evolutions, mixed bipartite states correspond to mixtures of evolutions, and states with maximally mixed marginals correspond to non-selective environments (Marcovitch et al., 2011).

The main theorem identifies the spatial correlator

$t_a$14

with a temporal correlator,

$t_a$15

provided $t_a$16 and $t_a$17. The correspondence is operationally realized through weak measurements. The paper gives the temporal weak-pointer correlator

$t_a$18

and states explicitly that the mapping does not apply to strong measurements (Marcovitch et al., 2011).

This framework turns Bell inequalities into Bell-in-time or Leggett–Garg-type inequalities with the same classical and quantum bounds. For CHSH, the classical bound remains $t_a$19 and the quantum bound remains $t_a$20. The canonical example is the Bell state $t_a$21, which maps to the identity evolution $t_a$22. In the corresponding temporal test, the paper obtains the maximal qubit violation $t_a$23. The same correspondence is used to discuss decohering dynamics: for a Lindblad model with $t_a$24, the reported temporal quantity is

$t_a$25

which decays to the classical bound when $t_a$26 (Marcovitch et al., 2011).

This state–evolution map gives a rigorous sense in which a bipartite Bell inequality can become temporal. It also explains why the literature sometimes speaks of a temporal Bell inequality when the two “parts” are past and future rather than simultaneous spatial wings.

7. Cosmological squeezed coherent states and remaining interpretive issues

In inflationary cosmology, the two subsystems are the modes $t_a$27 and $t_a$28, while time enters through unequal-time pseudo-spin correlators. The temporal Bell operator is written as

$t_a$29

with

$t_a$30

The practical motivation is explicit: only one component of the pseudo-spin operator is observationally accessible, so the temporal construction avoids the need for two distinct sets of observables (Mondal, 8 May 2026).

The state is taken to be an initial two-mode coherent state that inflation evolves into a squeezed coherent state. The mode evolution is generated by

$t_a$31

and the quadrature wavefunction is a displaced Gaussian,

$t_a$32

The paper derives an analytical formula for the unequal-time correlator $t_a$33 in terms of normalization prefactors, Jacobians, matrices, and a coherent-state displacement vector $t_a$34, and then evaluates the Bell combination numerically in de Sitter space (Mondal, 8 May 2026).

The principal conclusion is negative in the strict Bell sense but not in the informational sense. For squeezed coherent states of inflationary perturbations, the bipartite temporal Bell inequality is not violated. The paper states that the Bell quantity still differs slightly from the squeezed-vacuum result, especially for large squeezing, so distinguishability of initial states does not rely on Bell violation. A further distinctive claim is that the temporal Bell correlator depends explicitly on a purely imaginary phase factor in the wavefunction normalization. According to the paper, that phase cancels in equal-time spatial Bell correlators but survives in unequal-time temporal correlators, making the phase sensitivity a feature unique to the temporal construction in this setting (Mondal, 8 May 2026).

Taken together, these developments show that the expression “bipartite temporal Bell inequality” designates a family of related but formally distinct objects. In some cases it is an ordinary CHSH scenario with times replacing settings; in others it is a temporal Clauser–Horne inequality made meaningful by a propagation constraint; in still others it is standard CHSH evaluated on a time-resolved nonlocal correlation function. The shared theme is that Bell-type nonclassicality is interrogated through unequal-time structure rather than through spatial settings alone, but the precise physical content depends on whether the relevant premises are locality, realism, fair sampling, weak sequential measurement, or finite-speed information propagation.