Quantum Causal Order

Quantum causal order asks whether correlations among local laboratories must admit one classical order of influence. Process matrices answer this operational question without assuming a global background time, while quantum-gravity proposals ask a harder one: can dynamical geometries place events in a coherent superposition of causal relations? The two questions intersect, but causal nonseparability in a laboratory protocol is not by itself evidence that spacetime geometry was quantized.

Required background. Quantum Information supplies channels, instruments, complete positivity, entanglement witnesses, and operational probabilities.

Helpful background. Flat-Spacetime Quantum Field Theory supplies relativistic locality; Einstein Equations supplies dynamical causal cones; Quantum Gravity supplies programme comparison; and Causality and Global Structure distinguishes event order, signaling, and global causal structure.

Process matrices and causal separability

Let laboratories $A$ and $B$ receive finite-dimensional quantum systems, apply completely positive maps, and return outputs. Fix local input/output bases and use the transposed Choi convention $M_{a|x}=J(\mathcal M_{a|x})^T$, where $J(\mathcal M)=\sum_{ij}|i\rangle\langle j|\otimes\mathcal M(|i\rangle\langle j|)$ and $T$ is the full matrix transpose in those bases. With the matching convention for $W$, outcomes $a,b$ and settings $x,y$ have probabilities

$$p(a,b|x,y)= \operatorname{Tr}\!\left[ \left(M^A_{a|x}\otimes M^B_{b|y}\right)W \right].$$

Complete positivity requires $W\ge0$, while normalization for all local trace-preserving operations imposes additional affine linear constraints on $W$. A process compatible with $A$ before $B$ forbids signaling from $B$ to $A$; one compatible with $B$ before $A$ forbids the reverse. A causally separable process is a convex mixture

$$W_{\mathrm{sep}}=qW^{A\prec B}+(1-q)W^{B\prec A}, \qquad 0\le q\le1.$$

A process outside this convex set is causally nonseparable. Oreshkov, Costa, and Brukner introduced the framework and exhibited correlations incompatible with a fixed causal order Oreshkov, Costa, and Brukner 2012, pp. 2–6.

Three notions that must not be conflated

Causal nonseparability is a property of the process operator relative to specified laboratories and allowed operations. It can be certified by a witness $S$ satisfying $\operatorname{Tr}(SW_{\mathrm{sep}})\ge0$ for all separable processes but $\operatorname{Tr}(SW)\lt0$ for the target.

Violation of a causal inequality is stronger and device independent: observed probabilities lie outside the polytope generated by definite-order causal models. Some causally nonseparable processes, including standard implementations of the quantum switch, need not violate a causal inequality under the usual assumptions.

Indefinite spacetime causal structure says that the physical metric or relational geometry does not assign one order to the relevant events. A quantum-controlled circuit can coherently route a target operation through $A$ and $B$ in different orders while every component propagates on an ordinary background. Such an experiment demonstrates coherent control of order, not necessarily a superposition of gravitational light cones.

Witness constructions make the first distinction experimentally meaningful Araújo et al. 2015, §§ II–IV. The inference to quantum geometry requires additional locality and mediation assumptions.

Coupling order to dynamical geometry

A gravitational proposal should define events relationally—for example by clock readings or localized interactions—because coordinate labels are gauge. It should specify the joint state of geometry and matter, the constraint algebra or path-integral contour, and the map from branch geometries to laboratory operations. If branch 1 places $A$ in the causal past of $B$ while branch 2 reverses them, coherence must survive tracing out gravitational and environmental degrees of freedom.

Several checks follow. First, the local operations must be well defined in each branch and reduce to ordinary QFT in a semiclassical limit. Second, probabilities must remain positive and normalized. Third, the claimed order must be invariant under diffeomorphisms and not disappear after redefining the event labels. Fourth, an ordinary controlled-path explanation on a fixed geometry must be excluded if the claim is specifically gravitational.

Process matrices more general than quantum theory on acyclic spacetime can permit unusual signaling correlations. Not every mathematically valid process has a known realization in a unitary, local, dynamical spacetime. Conversely, a superposition of semiclassical metrics may yield an effective process that is causally separable after gravitational degrees of freedom decohere. The mapping is a calculation, not an identity.

Results and open questions through 2026-08-20

This discussion reflects results available through 2026-08-20. Operational indefinite order is an active and experimentally developed quantum-information subject. Its quantum-gravitational interpretation remains unresolved. Competing positions treat the process framework as, respectively, a theory-independent window on quantum spacetime, a resource theory of laboratory order on fixed backgrounds, or an effective language whose admissible subset must be derived from gravity.

Process, result, or proposal	Evidence at the cutoff	Scope and limitation
General process matrices	Consistent probability rule and explicit causally nonseparable examples	The full allowed set is not known to arise from dynamical spacetime
Quantum switch	Operational witnesses certify causal nonseparability in controlled-order implementations	Standard realizations can use fixed background geometry and do not by themselves quantize gravity
Causal-inequality violations	Provide device-independent separation from definite causal order in theory	Physical realization under ordinary quantum mechanics and gravity is restricted and unsettled
Superposed gravitational geometries	Models can correlate metric branches with reversed relational order	Event localization, gauge invariance, coherence, and measurement protocols remain formulation-dependent

The relevant formulation axis includes laboratory partition, allowed local operations, background assumptions, event definition, and whether the gravitational field is a mediator or merely a control label. Changing that axis can change a process from nonseparable to separable.

The conclusions could change if an experiment rules out every fixed-background controlled-route model while certifying a gravity-linked causal witness, or if a gravitational theory derives a sharply restricted process set; the interpretation could also change with a no-go theorem showing that physical metric superpositions cannot violate a causal inequality, or by a protocol that does so under closed locality assumptions.

Independent checks and limitations

Tomographic reconstruction of $W$ is device dependent and must include state-preparation and measurement errors. A witness should be optimized before data collection or corrected for the selection procedure. No-signaling conditions within each claimed order provide normalization checks. For gravitational interpretations, repeat the analysis with the control system treated explicitly; if tracing or relabeling it restores a definite geometry explanation, the stronger claim has not been established.

Exercise

Suppose $W=\tfrac12W^{A\prec B}+\tfrac12W^{B\prec A}$, where the order label is stored in an unobserved classical register. Is $W$ causally nonseparable? Would replacing the classical mixture by coherent control automatically prove indefinite spacetime geometry?

Solution

The displayed $W$ is explicitly a convex mixture of definite orders, so it is causally separable. An unobserved classical label changes run-to-run order without producing operational coherence. Coherent control can instead produce a causally nonseparable effective process, but that still does not automatically prove indefinite geometry: a control qubit may route systems along fixed-background paths. A gravitational conclusion additionally requires relational events whose order is tied to metric degrees of freedom, exclusion of the fixed-background routing model, and a witness robust to tracing out the geometry and environment.

References

• Araújo, Mateus, Cyril Branciard, Fabio Costa, Adrien Feix, Christina Giarmatzi, and Časlav Brukner (2015), “Witnessing Causal Nonseparability,” New Journal of Physics 17, 102001, arXiv:1506.03776.
• Oreshkov, Ognyan, Fabio Costa, and Časlav Brukner (2012), “Quantum Correlations with No Causal Order,” Nature Communications 3, 1092, arXiv:1105.4464.