Time-Symmetric Causal Dilation Theorems

• The paper introduces time-symmetric causal dilation theorems that represent operational processes as compressed parts of larger physical structures.
• It establishes equivalence between time-symmetric and time-forward formulations using tester positivity and double causality constraints.
• The work extends dilation theory to complex causal diagrams, offering insights into structures like ladders, snakes, and signalling constraints.

Searching arXiv for the primary paper and closely related time-symmetric causal work. Search query: "time symmetric causal dilation theorem operational probabilistic theories arXiv" Time symmetric causal dilation theorems are representation theorems in which an operational process, formulated without privileging a time-forward perspective, is recovered as the visible or compressed part of a larger mathematically physical structure. In the formulation developed for operational probabilistic theories (OPTs), a time-symmetric operation carries inputs and outputs for physical systems together with incomes and outcomes for classical pointer data, with both temporal directions treated on the same footing. Within that setting, physical admissibility is expressed by tester positivity and a pair of causality constraints running in opposite temporal directions, and the main dilation results show that physical operations admit enlargements to natural maxometries; corresponding results extend, for specified classes of causal diagrams, to causally complex operations such as ladders and snakes (Hardy, 12 Mar 2026).

1. Time-symmetric operational setting

The time-symmetric framework is built by reformulating OPTs so that the standard asymmetry between preparation and observation is not taken as primitive. In the usual time-forward presentation, operations have inputs before the operation and outputs after it, while outcomes are treated as future-directed classical readouts. In the time-symmetric presentation, an operation instead has inputs and outputs for systems together with incomes and outcomes for pointer information. The point is not merely notational. In the simple causal case, the framework is presented as empirically equivalent to the standard time-forward formalism, but it packages the probabilistic content in a temporally symmetric way (Hardy, 12 Mar 2026).

A basic distinction is drawn between simple and complex operations. Simple operations have simple causal structure, with all inputs before all outputs. Complex operations are equipped with a causal diagram and may have more intricate temporal relations, including cases in which some outputs are before some inputs. This distinction is structurally important because the strongest equivalence results hold in the simple case, whereas the dilation theory for complex operations is established only for particular diagram classes rather than for arbitrary causal graphs (Hardy, 12 Mar 2026).

Within this perspective, a causal dilation theorem is not merely a theorem about embedding a map into a larger map. It is a theorem about embedding a physical operation into a larger causal structure that still satisfies the required positivity and causality constraints. This places the subject closer to Stinespring-style representation theory than to ordinary time-directed circuit decompositions, while preserving symmetry between past-directed and future-directed classical conditioning.

2. Physicality, tester positivity, and double causality

The foundational constraint is physicality. The starting assumption is that “Every circuit has a probability depending only on the specification of that circuit.” This permits a linear extension of circuit probabilities and supports a notion of equivalence defined by testing against all complementary networks. Physicality then has two components: positivity, ensuring nonnegative probabilities, and causality, ensuring that forbidden signalling directions do not occur except when one conditions on the relevant classical data (Hardy, 12 Mar 2026).

In the time-symmetric formulation, positivity is expressed through tester positivity, while causality becomes double causality. The double causality theorem states that deterministic operations satisfy both a forward and a backward causality condition. Conceptually, forward causality says that choices in the future cannot influence the past unless one conditions on future outcomes; backward causality says that choices in the past cannot influence the future unless one conditions on past incomes. This is the symmetric replacement for the usual one-way no-signalling requirement (Hardy, 12 Mar 2026).

The framework also proves uniqueness statements for deterministic “ignore” structures. For each pointer type $x$, there is a unique deterministic pointer result/preparation $R$ up to equivalence, and for each system type $a$, there is a unique deterministic preparation/result $I$. These objects serve as canonical ignore boxes. The flatness assumption,

$\Pr(R_x)=\frac{1}{N_x},$

where $N_x$ is the number of pointer values, later becomes essential for the equivalence between time-symmetric and time-forward descriptions (Hardy, 12 Mar 2026).

Closure under composition is a central structural fact. If each component in a network satisfies forward causality, then the whole network satisfies forward causality; similarly for backward causality. If pure preparations and pure results satisfy tester positivity, then any network built by wiring together operations that satisfy tester positivity also satisfies tester positivity. From this, the circuit positivity theorem yields nonnegative circuit probabilities, and the circuit subunity theorem yields probabilities $\le 1$ when general double causality holds. The simple physicality composition theorem combines these results: any network built from physical operations is again physical (Hardy, 12 Mar 2026).

3. Equivalence with the time-forward formalism

A major result in the simple case is that the time-symmetric, time-forward, and time-backward perspectives are operationally equivalent under flatness. In the time-forward picture, operations are handled through conditional probabilities and do not explicitly include incomes. In the time-symmetric picture, the same content is encoded as joint probabilities with both incomes and outcomes retained. The paper’s central equivalence theorem states that, under the flatness assumption, any physical operation in the time-symmetric theory can be modeled by a set of physical operations in the time-forward theory, and conversely any physical time-forward operation can be modeled by a physical time-symmetric operation (Hardy, 12 Mar 2026).

A key technical device is the forward seatbelt identity, derived from the midcome identity and flatness. It permits conversion between the two descriptions by inserting or removing preselection normalization factors. In the relevant conditional-probability context, the replacement rule is

$\text{preselection box} \sim N_x \times R.$

The time-forward formalism is thereby recovered by conditioning the time-symmetric joint-probability calculus on incomes.

This equivalence is conceptually significant because it implies that, in the simple case, standard time-forward operational quantum theory is not more fundamental than the symmetric formulation. The asymmetry of the conventional framework is instead attributable to conditionalization. A plausible implication is that the time-symmetric dilation theorems should be read not as modifications of ordinary operational quantum theory, but as a more symmetric representation theory from which the standard formalism is derivable.

4. Hilbert objects, operator tensors, and the simple dilation theorem

The dilation results are formulated only after introducing a substantial representational apparatus. The paper develops Hilbert objects by “doubling up” wires, leading to left Hilbert objects, right Hilbert objects, and general Hilbert objects. These serve as the substrate for operator tensors, which in the simple quantum case correspond to Hermitian operator tensors. This doubled formalism is paired with a transformation group

$$\mathcal C=\{I,\ \bar I,\ H,\ \bar H,\ V,\ \bar V,\ T,\ \bar T\},$$

called the normal conjuposition group when built from orthonormal bases and the natural conjuposition group when built from the physically natural ortho-physical basis. In that setting, the natural transpose is identified as the correct time-reversal map, and mirror notation is introduced to express the transformations diagrammatically (Hardy, 12 Mar 2026).

Before the dilation theorem proper, the paper proves a maximal representation theorem. Assuming double maximality, any operation $B$ can be rewritten as

$R$0

with $R$1 of the same physical type and with three preserved properties: $R$2 satisfies tester positivity iff $R$3 does, $R$4 satisfies general double causality iff $R$5 does, and $R$6 has the same physical norm as $R$7. This reduction places the object to be dilated into a canonical form (Hardy, 12 Mar 2026).

The simple time-symmetric causal dilation theorem is then a Stinespring-style result. Any physical operator or operation can be realized as a compressed part of a larger object built from a natural maxometry together with appropriate ignore and preparation structure. The crucial point is that the time-symmetric analogue of the Stinespring isometry or unitary is not, in general, a unitary but a natural maxometry. A natural unitary is a special case of a natural maxometry, but a general natural maxometry need not be physically realizable as a laboratory transformation. The theorem is therefore mathematically strong while remaining nonconstructive in an implementation sense (Hardy, 12 Mar 2026).

A refined version, the dilation theorem with sufficient pairs, packages the auxiliary dilation data as sufficient pairs of Hilbert objects. The paper presents this as an analogue of Stinespring or Naimark-type representation in which the additional states and effects required to reconstruct the original operation are identified explicitly.

5. Complex-causal generalizations: ladders, snakes, and signalling constraints

The theory extends beyond simple causal order to complex operations endowed with explicit causal diagrams. In this regime, physicality requires tester positivity together with double causality for every synchronous partition of the causal diagram, that is, for every admissible “time-slice” of the causal graph. This sharply constrains which causally complex operations count as physical (Hardy, 12 Mar 2026).

The book proves dilation theorems for several specific diagram classes: 2-causal ladders, $R$8-causal ladders, causal snakes, and many ancestors/descendants diagrams. For each such class, the theorem states, roughly, that any physical operation with that causal diagram can be represented by a larger dilation object, typically using natural maxometries and sufficient pairs, from which the original operation is recovered by suitable compression or closure. The various theorems differ in the allowed diagram class and in how tightly the sufficient pairs can be chosen, including versions with minimal sufficient pairs and versions using $R$9isometries when a tighter dilation is available (Hardy, 12 Mar 2026).

The theory does not claim a general dilation theorem for arbitrary causal diagrams. That limitation is explicit and is part of the current state of the subject. The achieved results nonetheless show that nontrivial causal interleaving can be handled without abandoning the positivity and causality requirements that define physicality.

An important application is the treatment of Sorkin’s impossible measurements. Within the complex operational framework, the conclusion is that if operations are physical, then such “impossible measurements” are unphysical because physicality prevents anomalous signalling. This gives the causal dilation theorems a regulatory role: they are not only representation results but also constraints on admissible signalling patterns in complex causal networks (Hardy, 12 Mar 2026).

6. Relation to other time-symmetric causal frameworks

Time symmetric causal dilation theorems belong to a broader family of research programs in which causal asymmetry is not treated as fundamental. One nearby line uses stochastic bridge models with both initial and final boundary conditions,

$a$0

and factorizes the bridge-conditioned density into forward and backward components, for example

$a$1

In that setting, the principal theorem states that the direct-time and reverse-time formulations are fully time-symmetric and equivalent iff Doob’s transform and Anderson’s reversal correction drifts coincide,

$a$2

This is not an operational dilation theorem, but it is a closely related equivalence theorem linking forward and reverse descriptions under time-symmetric boundary conditions (Klimenko, 13 Jun 2025).

Other related work frames time symmetry in explicitly two-boundary quantum terms. One paper advocates a collapse-less two-boundary quantum mechanics with no intrinsic time-ordered causal structure at the microscopic level and attributes ordinary macroscopic causality to cosmological asymmetry, decoherence, and correspondence transition rules (Bopp, 2016). Another presents a toy model in which Bell-inequality-violating correlations arise from retarded and advanced light-cone effects rather than instantaneous influences, with future outcomes feeding back into earlier hidden variables through zigzag causal effects (Lazarovici, 2014). These programs differ sharply in formalism from the OPT-based dilation theorems, but they share the broader thesis that familiar one-way causal order may be emergent rather than primitive.

Taken together, these developments suggest a common structural pattern. In the operational setting, the pattern appears as dilation into natural maxometries constrained by tester positivity and double causality. In the stochastic bridge setting, it appears as equivalence between forward and reverse conditioned diffusions when the correction drifts match. In two-boundary quantum proposals, it appears as effective macroscopic causality emerging from temporally symmetric microscopic dynamics. The time symmetric causal dilation theorems are the most explicitly representation-theoretic members of this family, and their distinctive contribution is to place temporal symmetry, causal admissibility, and dilation theory inside a unified OPT framework (Hardy, 12 Mar 2026).