⚛️ quantum physics

Correlations and quantum circuits with dynamical causal order

This paper introduces a new intermediate class of correlations and quantum processes characterized by "non-influenceable causal order," where the causal structure between parties is dynamical yet independent of past actions, thereby refining the understanding of how indefinite and dynamical causal orders manifest in quantum circuits.

Original authors: Raphaël Mothe, Alastair A. Abbott, Cyril Branciard

Published 2026-08-19

📖 8 min read🧠 Deep dive

CC BY 4.0

Original authors: Raphaël Mothe, Alastair A. Abbott, Cyril Branciard

Original paper licensed under CC BY 4.0 (http://creativecommons.org/licenses/by/4.0/). ✨ This is an AI-generated explanation of the paper below. It is not written or endorsed by the authors. For technical accuracy, refer to the original paper. Read full disclaimer

In the physical world, events usually happen in a sequence. One thing causes another, which causes a third, creating a clear line of time from past to future. For centuries, this idea of a fixed order was considered a fundamental rule of reality. However, the strange laws of quantum mechanics have recently challenged this certainty. Scientists have discovered that under specific conditions, the order in which events occur can be blurred, existing in a state where it is not yet decided which event happened first. This concept, known as indefinite causal order, suggests that the very structure of cause and effect can be fluid. While this sounds like a theoretical curiosity, understanding how these fluid structures work is crucial for developing future technologies, such as quantum computers that could solve problems impossible for today's machines. The question researchers have been trying to answer is whether this fluidity is a chaotic free-for-all, or if there are hidden rules that govern how the order of events can change.

A team of physicists has now uncovered a new layer of complexity in this puzzle, revealing that the way events influence one another is more nuanced than previously thought. For a long time, scientists assumed that if the order of events was changing, it was because an earlier action was directly deciding the order of later events. Imagine a traffic light that changes based on how many cars are waiting; the cars in the past influence the signal for the future. This paper demonstrates that this is not the only way things can work. The researchers found that with four or more parties involved, it is possible for the order of events to be dynamic—changing as the process unfolds—without any of the earlier participants having any control over that change. The order shifts, but it shifts on its own, completely independent of the choices made by the people acting before it.

To understand this discovery, one must first look at how scientists describe the relationships between different people or machines, which they call "parties." In a standard scenario, if Party A acts before Party B, A can send a message to B, but B cannot send a message back to A. This is a fixed causal order. Scientists have known for some time that if you add a third party, the order doesn't have to be fixed in advance. It can be determined on the fly. For example, Party A might choose an action that decides whether Party B or Party C goes next. This is what is known as a dynamical causal order, where the sequence is established as the process happens. The prevailing assumption was that this dynamism always required influence: the past had to be pulling the strings to determine the future sequence.

The authors of this study, working with a group of collaborators, realized that this assumption was incomplete. They set out to define a new category of behavior where the sequence of events is dynamic but not influenced by the past. They began by analyzing simple scenarios involving classical information, where parties exchange inputs and outputs. By creating a specific mathematical framework, they identified a class of correlations where the causal order between parties changes, yet the probability of that change remains completely unaffected by the inputs chosen by the earlier parties. In their simulations, they showed that for three parties, this new class of behavior is impossible; the order either stays fixed or is influenced by the past. However, as soon as they introduced a fourth party, a new possibility emerged. They found that the order could be dynamic, shifting between different sequences, but the likelihood of each sequence was fixed from the start, immune to the actions of the participants.

This discovery led the researchers to propose a new classification system for these interactions. They defined a middle ground between processes that have a completely fixed order and those that are fully dynamic and influenced by the past. They called this intermediate state "non-influenceable causal order." To prove that this was a distinct and real phenomenon, they designed a specific game involving four parties. In this game, the players try to guess each other's inputs based on the rules of the game. The researchers calculated the maximum score possible for different types of causal structures. They found that the score achievable with non-influenceable orders was strictly higher than what was possible with fixed orders, but strictly lower than what could be achieved if the past could influence the future. This gap in performance proved that the new class of behavior is not just a theoretical variation, but a fundamentally different way the world can operate.

The team then extended these findings into the realm of quantum physics, where the rules are even more flexible. They looked at quantum circuits, which are the building blocks of quantum computers, specifically those where the order of operations is controlled by a quantum system. In these circuits, a control system can exist in a superposition, effectively putting the order of operations into a state where multiple sequences happen at once. The researchers applied their new definitions to these quantum circuits. They identified a class of quantum processes where the control system is dynamic but non-influenceable. In these processes, the quantum control system evolves in a way that determines the order of operations, but this evolution is not altered by the specific operations performed by the parties.

One of the most striking examples they constructed involves four parties. In their setup, two parties act first, and their actions are part of a complex quantum process that is causally non-separable, meaning their order is not fixed. However, the outcome of a measurement performed on these two parties, which then decides the order in which the remaining two parties act, is statistically independent of what the first two parties actually did. The measurement result is random, but it is not influenced by the choices of the earlier parties. This creates a scenario where the causal order between the last two parties is determined dynamically by the measurement, yet the probability of that order is fixed and cannot be manipulated by the past. This stands in contrast to other known quantum processes, like the famous "quantum switch," where the order is fixed in a superposition from the very beginning and does not change dynamically, and other processes where the past actively controls the future order.

The researchers used advanced computational techniques to test the limits of these new classes. They calculated the maximum possible scores for their four-party game using these non-influenceable quantum circuits. The results showed that while these circuits could outperform those with fixed orders, they could not reach the maximum theoretical limit allowed by general causal correlations. This indicates that while non-influenceable dynamical order is a powerful resource, it is still more restricted than the full range of possibilities allowed by quantum mechanics. The study confirms that the universe allows for a subtle form of dynamism where the sequence of events can be fluid without being controlled by the past.

This work reshapes our understanding of how cause and effect can operate in the quantum realm. It shows that the relationship between time, order, and influence is not a simple binary choice between fixed and chaotic. Instead, there is a structured middle ground where the future can be uncertain and changing, yet remain completely independent of the past's choices. This distinction is vital for the future of quantum information science. As researchers develop more complex quantum networks and computers, knowing exactly which types of causal structures are possible—and which are impossible—will help them design more efficient systems. The ability to distinguish between processes that are dynamically changing because of past influence and those that are changing on their own provides a new tool for engineers and physicists. It suggests that there are new, untapped resources in the way quantum systems can be arranged, resources that do not rely on the past controlling the future, but rather on a more subtle, self-contained evolution of order.

The paper concludes by highlighting that this is just the beginning of understanding these structures. The researchers note that while they have defined these classes and shown they exist, there are still many open questions. For instance, it is not yet clear if there are other, more complex forms of non-influenceable order that they have not yet discovered, or if these new classes can provide a practical advantage in specific computational tasks. They also point out that their definitions rely on a specific framework of quantum circuits, and it remains to be seen how these ideas apply to the broader, more abstract theories of quantum gravity. Nevertheless, the identification of this new class of causal order marks a significant step forward. It moves the field beyond the simple question of whether the order is fixed or not, and into a deeper exploration of how that order is determined, revealing a hidden layer of independence in the fabric of cause and effect.

Technical Summary: Correlations and Quantum Circuits with Dynamical Causal Order

Problem Statement
The study of causal structures in quantum mechanics has established that well-defined causal orders constrain the correlations parties can share. While it is known that for $N \ge 3$ parties, causal correlations can exhibit "dynamical" causal orders—where the order of future parties is established "on the fly" based on the actions of past parties—the precise definition and classification of such dynamicality have remained incomplete. Specifically, the literature has largely equated "dynamical" with "influenceable," assuming that a causal order is dynamical only if it can be influenced by the choices of previous parties. Furthermore, while the process matrix formalism allows for "indefinite" causal orders (e.g., the quantum switch), a rigorous framework to distinguish between processes that are merely indefinite and those that are both indefinite and dynamically determined by party actions has been lacking. This paper addresses the need to precisely define (non-)dynamical causal structures for both classical correlations and quantum processes, and to characterize the interplay between dynamicality and influenceability.

Methodology
The authors employ a two-pronged approach, first analyzing classical causal correlations and then extending the analysis to quantum processes within the framework of quantum circuits with classical (QC-CCs) and quantum (QC-QCs) control of causal order.

Classical Correlations: The authors utilize a characterization of causal correlations where the causal order is treated as a hidden variable $\pi$ . They introduce the concept of "non-influenceable" causal orders, defined as correlations where the probability distribution of the causal order $\pi$ is independent of the parties' input choices. They define three specific classes of correlations:
- convFO: Convex mixtures of fixed causal orders (static).
- NIO: Causal correlations with non-influenceable orders (potentially dynamical but independent of inputs).
- NIO': Causal correlations with non-influenceable coarse-grained orders (where the unordered set of past parties is independent of inputs).
- causal: General causal correlations (potentially dynamical and influenceable).
To separate these classes, the authors construct a four-partite causal game (an extension of the LGYNI game) and derive specific bounds (inequalities) for the success probability $I_4$ within each polytope.
Quantum Processes: The authors map these correlation classes to subclasses of quantum circuits:
- QC-convFO: Convex mixtures of quantum circuits with fixed causal order.
- QC-NICC: Quantum circuits with non-influenceable classical control of causal order.
- QC-supFO: Quantum circuits with superposition of fixed causal orders (non-dynamical quantum control).
- QC-NIQC: Quantum circuits with non-influenceable quantum control of causal order.
- QC-CC / QC-QC: General classes with classical or quantum control, respectively.
They provide semidefinite programming (SDP) characterizations for these classes, specifically defining constraints on the process matrices (e.g., requiring that internal control systems or their diagonal weights remain independent of external operations). They utilize SDP techniques to compute the maximal values of the causal game quantity $I_4$ achievable by these quantum classes.

Key Contributions and Results

Identification of Non-Influenceable Dynamical Order: The paper demonstrates that for $N \ge 4$ , there exist causal correlations that are dynamical (not a convex mixture of fixed orders) yet non-influenceable (the causal order distribution is independent of party inputs). This reveals that "dynamical" and "influenceable" are inequivalent notions, correcting the assumption that dynamicality necessarily implies influenceability.
Strict Hierarchy of Correlation Polytopes: The authors establish a strict hierarchy for $N \ge 4$ :
$P_{convFO} \subsetneq P_{NIO} \subsetneq P_{NIO'} \subsetneq P_{causal}$
They prove this separation using the quantity $I_4$ , showing distinct upper bounds for each class (e.g., $15/16$ for $P_{convFO}$ , $31/32$ for $P_{NIO}$ , $47/48$ for $P_{NIO'}$ , and $1$ for $P_{causal}$ ).
Formalization of Quantum Circuit Classes: The paper defines and characterizes the quantum analogues of these correlation classes:
- QC-NICC: Circuits where the classical control system's state is independent of previous operations.
- QC-supFO: Circuits where the quantum control system encodes a superposition of fixed orders with weights fixed from the start (generalizing the static quantum switch).
- QC-NIQC: Circuits where the classical weights (diagonal elements) of the quantum control system in the basis defining the order are independent of previous operations.
Hierarchy of Quantum Circuits: A strict inclusion hierarchy is established for $N \ge 4$ :
$QC\text{-}convFO \subsetneq QC\text{-}NICC \subsetneq QC\text{-}CC$
$QC\text{-}supFO \subsetneq QC\text{-}NIQC \subsetneq QC\text{-}QC$
(With intersections and specific reductions for $N \le 3$ ).
Quantum vs. Classical Bounds: While QC-CC and QC-QC can saturate the general causal bound ( $I_4=1$ ), the authors show via SDP that QC-NICC and QC-NIQC cannot reach the theoretical polytope bounds of $P_{NIO}$ and $P_{NIO'}$ (achieving $\approx 0.9482$ and $\approx 0.9527$ respectively, compared to $0.96875$ and $0.9792$). This indicates that quantum circuits with non-influenceable control generate a strict subset of the corresponding classical correlation polytopes.
Explicit Examples: The authors construct explicit process matrices (e.g., $W_{NICC}$ and $W_{NIQC}$ ) that belong to the non-influenceable classes but not the static ones. These examples feature a structure where a measurement on an initial entangled pair determines the causal order of subsequent parties, a process that is dynamical (order not fixed a priori) but non-influenceable (measurement outcome probabilities are independent of the parties' operations).

Significance
The paper claims to provide the first precise definitions of (non-)dynamical causal orders for both correlations and quantum processes. By distinguishing between "dynamical" and "influenceable," the authors refine the understanding of causal structures, showing that dynamicality can exist without the causal order being influenced by past parties. This distinction is crucial for foundational questions, particularly regarding theories of quantum gravity where causal structures may be both indefinite and dynamical.

Furthermore, the formalization of classes like QC-supFO and QC-NIQC allows for a rigorous analysis of whether "indefinite causal order" combined with "dynamical causal order" offers computational advantages. The authors note that while advantages for classically controlled dynamical orders have been shown, the advantage for coherently controlled dynamical orders remains an open question, as their new formalism allows for precise comparison of these resources. The work also highlights that for $N \ge 4$ , the landscape of causal correlations and quantum processes is richer than previously understood, containing intermediate classes that are neither static nor fully influenceable.

More like this