Causal Nonseparability in Quantum Processes

• Causal nonseparability is a quantum property defined within the process-matrix framework, where events occur in superpositions of causal orders.
• It enables information-processing advantages, exemplified by the quantum switch which demonstrates enhanced computational and thermodynamic capabilities.
• Certification relies on semidefinite programming and witness operators to differentiate causally nonseparable processes from those with fixed causal order.

Causal nonseparability is a central concept in the quantum theory of processes without definite causal order, formalizing the possibility that events can occur in superpositions of alternative causal structures. It arises in the process-matrix framework, which generalizes the quantum formalism for scenarios where local quantum operations are performed in “laboratories” without assuming a global time or causal structure. Causally nonseparable processes cannot be decomposed, even probabilistically, into mixtures of fixed causal orders, implying an intrinsic indeterminacy in the sequence of events that transcends classical and standard quantum separability. This phenomenon, exemplified by the quantum switch, underpins the operational resource theory of indefinite causal order, enables novel information-processing capabilities, and demands new tools for characterizing and certifying causal structure in quantum processes.

1. Formal Framework and Definition

In the process-matrix formalism, each party (e.g., Alice, Bob) is modeled as a local quantum instrument acting on input and output Hilbert spaces. A process matrix $W$ is a positive semidefinite operator on the tensor product of all parties’ input and output spaces that generates the joint probability of outcomes for arbitrary local operations via a generalized Born rule. The process matrix must satisfy linear trace and “no forbidden signalling” constraints to ensure normalized, valid probability distributions (Taddei et al., 2019, Milz et al., 2017, Branciard, 2016, Oreshkov et al., 2015, Letertre, 2022).

A process is causally separable if it admits a convex decomposition into processes corresponding to fixed causal orders, for instance: $$W = p\, W^{A \prec B} + (1-p) W^{B \prec A},\quad 0\leq p \leq 1$$ where $W^{A \prec B}$ admits no signalling from $B$ to $A$, and $W^{B \prec A}$ vice versa. If no such decomposition exists, $W$ is causally nonseparable.

Formally, for multipartite processes, causal separability is defined inductively: for every ancilla extension, $W \otimes \rho$ must decompose as a convex mixture of processes, each compatible with some party being last and the conditional reductions being themselves causally separable (Wechs et al., 2018, Oreshkov et al., 2015).

2. Resource Theory of Causal Nonseparability

Causal nonseparability serves as an operational resource for information-processing advantages. The quantum switch—a coherent control of order between two operations—embodies quantum control of causal order and is causally nonseparable: $$|w_{\mathrm{qs}}\rangle = \frac{1}{\sqrt{2}} \left( |0_C\rangle \otimes |\mathbb{1}_0\rangle + |1_C\rangle \otimes |\mathbb{1}_1\rangle \right)$$ where the control qubit $C$ coherently determines whether the “target” passes through $$W = p\, W^{A \prec B} + (1-p) W^{B \prec A},\quad 0\leq p \leq 1$$0 then $$W = p\, W^{A \prec B} + (1-p) W^{B \prec A},\quad 0\leq p \leq 1$$1 or $$W = p\, W^{A \prec B} + (1-p) W^{B \prec A},\quad 0\leq p \leq 1$$2 then $$W = p\, W^{A \prec B} + (1-p) W^{B \prec A},\quad 0\leq p \leq 1$$3.

A resource theory is constructed with:

• Objects: all valid process matrices.
• Free operations: concatenations with causally separable processes, such as Local Operations with Ancillary Entanglement (LOAE) and Probabilistic Lab Swaps (PLS), implemented via the link product (Taddei et al., 2019). These preserve causal separability and its absence.
• Monotones: Functionals that are nonincreasing under free operations, e.g., the generalized robustness, relative entropy distance to the set of causally separable processes, or witness-based measures.

Distillation protocols allow for concentrating nonseparability via conversion of multiple copies of weaker nonseparable processes into fewer, more “maximally” nonseparable ones (such as the quantum switch) using operations free within this resource theory (Taddei et al., 2019).

3. Mathematical Characterization and Certification

Causal (non)separability is tested via convex decomposition and semidefinite programming (SDP). For a process matrix $$W = p\, W^{A \prec B} + (1-p) W^{B \prec A},\quad 0\leq p \leq 1$$4, the question of separability is formulated as:

• Primal SDP: Attempts to construct a decomposition into fixed-order processes.
• Dual SDP: Searches for a witness operator $$W = p\, W^{A \prec B} + (1-p) W^{B \prec A},\quad 0\leq p \leq 1$$5 such that:

$$W = p\, W^{A \prec B} + (1-p) W^{B \prec A},\quad 0\leq p \leq 1$$6

where $$W = p\, W^{A \prec B} + (1-p) W^{B \prec A},\quad 0\leq p \leq 1$$7 is the target process. Such a witness certifies causal nonseparability (Branciard, 2016, Araújo et al., 2015).

An information-theoretic witness based on Shannon entropy can also be used: if the total joint output entropy exceeds the maximal value achievable by any causally separable process with the same nonsignalling part, then the process is certified as causally nonseparable (Francica, 2021).

In device-dependent scenarios, witnesses are constructed and measured via process tomography; in semi-device-independent and network-device-independent scenarios, frameworks utilizing trusted quantum inputs or self-testing protocols allow certification even for processes like the quantum switch, which do not violate causal inequalities (Dourdent et al., 2021, Dourdent et al., 2023).

4. Examples and Classification

• Quantum Switch: Realizes a superposition of “$$W = p\, W^{A \prec B} + (1-p) W^{B \prec A},\quad 0\leq p \leq 1$$8” and “$$W = p\, W^{A \prec B} + (1-p) W^{B \prec A},\quad 0\leq p \leq 1$$9” with a control system. It is causally nonseparable, robust to various noise models, but cannot violate any causal inequality under device-independent (purely classical input-output) assumptions (Taddei et al., 2019, Araújo et al., 2015, Letertre, 2022, Steffinlongo et al., 21 Feb 2025).
• Processes with Causal Models: Certain causally nonseparable bipartite processes admit a causal model: all the correlations they generate can be simulated by a (possibly different) causally separable process. These do not violate causal inequalities and are analogous to entangled-but-local (Werner) states in nonlocality theory (Feix et al., 2016).
• Cyclic Quantum Causal Models: For unitary processes, causal nonseparability is equivalent to the presence of cycles in the associated causal structure graph. Such processes cannot be decomposed into acyclic (fixed-order) quantum causal models (Barrett et al., 2020).
• Activation: In both multiparty and multi-round scenarios, a process may be causally separable in its native setting but becomes nonseparable (and even causally noncausal) when local memory or ancillary entanglement is exploited (“activation of nonseparability”) (Oreshkov et al., 2015, Hoffreumon et al., 2020, Wechs et al., 2018).

5. Operational and Foundational Implications

Causal nonseparability distinguishes quantum processes not just by the absence of definite order but by their functional role:

• Resource for information-processing: Superposing causal orders can outperform fixed-order circuits in certain communication and computation tasks (e.g., for quantum thermodynamics, the quantum switch enables heat extraction not possible in any fixed order) (Taddei et al., 2019, Cao et al., 2021).
• Relation to quantum nonseparability: Although both are convex-set properties and use witness techniques, causal nonseparability applies to process matrices (maps relating labs’ inputs/outputs), not states. Not all causally nonseparable processes exhibit operational device-independent “noncausality” (i.e., causal-inequality violation), deepening the analogy with entanglement versus nonlocality (Feix et al., 2016, Letertre, 2022, Dourdent et al., 2023).
• Philosophical consequences: Within the process-matrix formalism’s background-independence, causal nonseparability implies an indeterminacy of spatiotemporal relations. Depending on the interpretational stance, this can be seen as epistemic (undetermined) or metaphysical (intrinsically indefinite). It shapes discussions about space–time in quantum gravity, where superpositions of spacetimes and indefinite causal structure are anticipated (Letertre, 2022).

6. Experimental Realizability and Simulation

Physical implementation of causally nonseparable processes requires:

• Ancillary degrees of freedom and postselection on environmental subsystems for simulating arbitrary (even causally indefinite) processes, leveraging non-Markovianity and tripartite entanglement (Milz et al., 2017).
• The quantum switch has been experimentally realized in optical systems to demonstrate indefinite causal order and resource function, such as enhanced quantum refrigeration performance (Cao et al., 2021).
• Certain processes (e.g., the SHIFT measurement simulated by the quantum switch) illustrate that semi-device-independent certification witnesses causal nonseparability but not genuine device-independent noncausality in the sense of causal-inequality violation (Steffinlongo et al., 21 Feb 2025).

7. Multipartite and Multi-Round Generalizations

In multipartite and multi-round settings, causal nonseparability becomes more nuanced:

• For $W^{A \prec B}$0 parties, several inequivalent definitions exist; the robust one (extensible causal separability, ECS) demands stability under arbitrary ancillary extensions (Wechs et al., 2018, Oreshkov et al., 2015).
• In multi-round scenarios (where parties interact with the environment multiple times), processes may be causally separable as one-round objects but become nonseparable when side-channels (quantum memory) are enabled—the so-called activation phenomenon (Hoffreumon et al., 2020).
• Activation and extensibility inform resource-theoretic hierarchies and motivate ongoing studies of the structure and interconvertibility of causally nonseparable processes.

---

References

• (Taddei et al., 2019) Quantum superpositions of causal orders as an operational resource
• (Milz et al., 2017) Entanglement, non-Markovianity, and causal non-separability
• (Barrett et al., 2020) Cyclic Quantum Causal Models
• (Branciard, 2016) Witnesses of causal nonseparability: an introduction and a few case studies
• (Oreshkov et al., 2015) Causal and causally separable processes
• (Dourdent et al., 2021) Semi-Device-Independent Certification of Causal Nonseparability with Trusted Quantum Inputs
• (Araújo et al., 2015) Witnessing causal nonseparability
• (Steffinlongo et al., 21 Feb 2025) Simulating Noncausality with Quantum Control of Causal Orders
• (Francica, 2021) Information Exchange in Causally Nonseparable Processes
• (Feix et al., 2016) Causally nonseparable processes admitting a causal model
• (Cao et al., 2021) Quantum simulation of indefinite causal order induced quantum refrigeration
• (Dourdent et al., 2023) Network-Device-Independent Certification of Causal Nonseparability
• (Letertre, 2022) Causal nonseparability and its implications for spatiotemporal relations
• (Wechs et al., 2018) On the definition and characterisation of multipartite causal (non)separability
• (Hoffreumon et al., 2020) The Multi-round Process Matrix
• (Filatov et al., 2021) Ordering the processes with indefinite causal order