---
title: Quantum Effect Labelled Transition Systems
url: https://www.emergentmind.com/topics/quantum-effect-labelled-transition-systems
type: topic
---

# Quantum Effect Labelled Transition Systems

Quantum effect labelled transition systems are a coalgebraic model for quantum-capable, concurrent, non-deterministic systems in which transition weights are quantum effects rather than probabilities. In this formulation, a process configuration does not carry a concrete quantum state. Instead, the transition structure records the observable probabilistic behavior that would arise after later instantiation with an input state, yielding a symbolic semantics that is parametric in the quantum input. The model is introduced as the quantum analogue of probabilistic labelled transition systems and is developed together with a resource-sensitive account of composition and a comparison of behavioural equivalences, with kernel bisimilarity singled out as the equivalence that matches the intended observational semantics [2509.20933].

## 1. Conceptual role and semantic scope

A qLTS is the paper’s quantum analogue of a probabilistic labelled transition system, with probabilities replaced by quantum effects. The intended use is to describe processes that manipulate quantum data while avoiding the need to store a concrete global quantum state in the syntax or in the state space. The transition system is therefore not a state-explicit model of quantum memory; it is a symbolic semantics for how probabilities would be observed once a state is supplied.

This design addresses a specific semantic problem in concurrent quantum process theory: defining a behavioural equivalence that matches the observational properties of a quantum-capable, concurrent, non-deterministic system. The difficulty, as formulated in the source, is that observational equivalence in such systems must respect both non-determinism and the constraints imposed by quantum theory. In particular, quantum data cannot be copied arbitrarily, and measurements consume the systems they inspect. The qLTS formalism responds by combining effect-valued transition weights, coalgebraic semantics, and grading by quantum resources.

Operationally, the model is “purely quantum” in the sense stated in the source: it does not attach a concrete state to each process configuration. Instead, transitions are annotated by quantum effects that encode the probabilities that will be observed when an actual state is later used as input. A plausible implication is that the formalism separates control-flow structure from the choice of input state, making the semantics explicitly parametric rather than state-specific.

## 2. Effect distributions and the role of quantum effects

The semantic basis is the construction of effect distributions. For any effect algebra $\langle \ealg,0,+,'\rangle$, the paper defines
$$
D_{\ealg}X = \left\{\, \Delta \in \ealg^{X} \;\middle|\; \supp(\Delta)\text{ is finite},\; \sum_{x\in \supp(\Delta)} \Delta(x)\sqsubseteq 1_{\ealg} \right\},
$$
where $\supp(\Delta)=\{x\mid \Delta(x)\neq 0\}$. Thus $D_{\ealg}X$ consists of finite-weight subdistributions valued in $\ealg$. For a map $f:X\to Y$, the action is by pushforward:
$$
(D_{\ealg}f)(\Delta)(y)=\sum_{x\in f^{-1}(y)}\Delta(x).
$$

This construction subsumes ordinary subprobability distributions when $\ealg=[0,1]$. It also subsumes quantum distributions when $\ealg=\ef[H]$, the effect algebra of quantum effects on a finite-dimensional Hilbert space $H$:
$$
\ef[H]=\{\,L\in\mathbb{C}^{d\times d}\mid 0_d\sqsubseteq L\sqsubseteq I_d\,\}.
$$
Here $\sqsubseteq$ is the Löwner order. Effects are interpreted operationally through the Born rule,
$$
\Pr(\text{yes on }L\text{ in state }\rho)=\tr(L\rho),
$$
so an effect denotes a probability-valued observable on states.

A central structural fact is the duality
$$
\ef[H]\;\cong\;\mathbf{Conv}(\dm[H],[0,1]), \qquad L\mapsto \big(\rho\mapsto \tr(L\rho)\big),
$$
which identifies a quantum effect with a convex map from density operators to probabilities. In the paper’s interpretation, this means that quantum effects can be regarded as probability-valued observables parametrized by an unknown state [2509.20933].

The significance of this duality is semantic rather than merely representational. It permits effect-valued transitions to be read as symbolic observables whose concrete probabilistic content emerges only after instantiation with a density operator. This is the formal mechanism by which qLTSs remain parametric in their quantum input.

## 3. Resource grading and compatibility with quantum theory

To compose weighted systems in a way that respects quantum theory, the paper introduces graded monads over a partial commutative monoid (PCM). A PCM is given as $\langle M,0,+\rangle$, where $+$ is partial, commutative and associative where defined, and $0$ is the unit. The grading records which quantum resources a computation consumes, and the partiality expresses that two computations can be combined only when their resources are disjoint.

The source explicitly connects this to the no-cloning theorem. Qubits cannot be duplicated or shared arbitrarily, and measurements consume the systems they inspect. The grading therefore tracks resource ownership and allows composition only along orthogonal, non-overlapping resources. This is not an external constraint added afterward; it is built into the monadic structure.

Formally, an $M$-graded effect monoid is a family $\{\ealg_m\}_{m\in M}$ with effect-algebra objects and multiplication morphisms
$$
\nabla_{m,n}:\ealg_m\otimes\ealg_n\to \ealg_{m+n} \qquad\text{for }m\perp n,
$$
together with a unit $\eta:2\to \ealg_0$, satisfying graded associativity and unitality diagrams. If the braiding condition also holds, the graded effect monoid is commutative. From such a graded effect monoid, the paper constructs a PCM-graded monad $\{\emon[\ealg_m]\}$ with
$$
\eta(x)=1_{\ealg_0}\bullet x, \qquad \mu_{m,n}\!\left(\sum_i e_i\bullet \Delta_i\right)(x) = \sum_i \nabla_{m,n}\bigl(e_i,\Delta_i(x)\bigr).
$$

In the quantum specialization, the grading monoid is
$$
\mathcal{S}=\langle \mathcal{P}(Sys),\emptyset,\uplus\rangle,
$$
the PCM of finite sets of quantum systems with partial disjoint union. For $C\subseteq Sys$, the effect-algebra component is $\ef[H_C]$, and multiplication is given by the “sorted” Kronecker product
$$
L_1\boxtimes L_2 = Sort_{C,D}(L_1\otimes_k L_2),
$$
defined when $C\cap D=\emptyset$. This yields the graded monad $Q_C=\emon[\ef[H_C]]$, with the intended meaning that measurements may be combined only when they act on disjoint collections of qubits [2509.20933].

This suggests that the qLTS framework is resource-sensitive at the semantic level: the admissibility of composition is determined by ownership of quantum subsystems, not merely by syntactic side conditions.

## 4. Coalgebraic presentation of qLTSs

A quantum effect labelled transition system is defined as a coalgebra with labels and non-deterministic branching:
$$
X\xrightarrow{c}\mathcal{P}(\emon[\ealg]X)^L.
$$
Thus, for a state $x\in X$ and a label $\mu\in L$, the value $c(x)(\mu)$ is a finite set of possible effect distributions. The paper assumes a silent action $\tau\in L$ and an involution $\overline{\mu}$ for visible labels, following standard process-calculus conventions.

In the quantum specialization, the systems are qLTSs when $\ealg=\ef[H_C]$ and pLTSs when $\ealg=[0,1]$. The distinction is therefore entirely in the weight structure: both are labelled, non-deterministic transition systems, but qLTSs use quantum effects as symbolic weights and pLTSs use ordinary probabilities. The qLTS does not itself select an input state; it records all observable probabilities as functions of such a state.

The coalgebraic formulation is important because the paper’s analysis concerns not only the existence of such systems but also the appropriate notion of bisimilarity for coalgebras of the form $D_{\ealg}$. In that setting, the form of the weight object strongly influences which equivalences are behaviourally meaningful. For qLTSs, that question is decisive, because quantum effects do not behave like decomposable probabilistic weights [2509.20933].

A plausible implication is that qLTSs are best viewed not simply as a variant of weighted transition systems, but as a semantic interface between process-theoretic operational rules and the convex-observational structure of open quantum systems.

## 5. Bisimulation: Aczel–Mendler versus kernel equivalence

The paper studies both Aczel–Mendler bisimilarity and kernel bisimilarity. For $F$-coalgebras $X\xrightarrow{c}FX$ and $Y\xrightarrow{d}FY$, an AM-bisimulation is a relation $R\subseteq X\times Y$ equipped with an $F$-coalgebra structure making both projections coalgebra homomorphisms. Kernel bisimilarity is instead defined via cocongruences: a cospan
$$
X\xrightarrow{m_1}Z\xleftarrow{m_2}Y
$$
of coalgebra homomorphisms, whose kernel relation is
$$
xRy\iff m_1(x)=m_2(y).
$$
The largest such relation is denoted $\sim_k$.

For effect distributions, the paper proves the implication
$$
x\sim_{AM} y \implies x\sim_k y,
$$
and the converse
$$
x\sim_k y \implies x\sim_{AM} y \quad\text{iff the effect algebra is decomposable}.
$$
The decomposability condition is the splitting property
$$
a+b=c+d \implies \exists e_{11},e_{12},e_{21},e_{22}:\; a=e_{11}+e_{12},\; b=e_{21}+e_{22},\; c=e_{11}+e_{21},\; d=e_{12}+e_{22}.
$$

The probabilistic effect algebra $[0,1]$ is decomposable, so AM and kernel bisimilarity coincide in the probabilistic case. Quantum effects, however, are not decomposable once the Hilbert space has dimension at least $2$. The paper shows this using
$$
\ketbra{0}+\ketbra{1}=\ketbra{+}+\ketbra{-},
$$
and argues that no decomposition satisfying the matrix-splitting constraints exists. Consequently, the two bisimilarities diverge for qLTSs [2509.20933].

This divergence is a central point and also a common source of misunderstanding. The issue is not that one equivalence is universally stronger in a purely formal sense divorced from observation; rather, the source argues that AM-bisimilarity depends on a decomposability property that quantum effects do not satisfy, whereas kernel bisimilarity remains aligned with the intended behavioural reading after probabilistic instantiation. The controversy, as framed by the paper, is therefore about semantic adequacy rather than about categorical elegance alone.

## 6. Instantiation semantics, local parameterization, and process operators

For each density operator $\rho\in\dm[H_C]$, the Born-rule map
$$
m_\rho:\ef[H_C]\to [0,1], \qquad m_\rho(L)=\tr(L\rho),
$$
is an effect-algebra homomorphism. It therefore induces a natural transformation $Q_C\Rightarrow \emon[[0,1]]$ and hence a functor from qLTSs to pLTSs. This functor instantiates a quantum process at input state $\rho$, turning quantum weights into classical probabilities. The first correctness desideratum in the paper is that two qLTS states should be equivalent exactly when they are equivalent after every such instantiation.

AM-bisimilarity does not satisfy that requirement in the quantum case. Kernel bisimilarity does: under finite quantum effect algebras, the paper proves the reflection property
$$
x\sim_k y \text{ in }(X,c),(Y,d)\ \text{for all }\rho \;\Longrightarrow\; x\sim_k y \text{ in }(X,c),(Y,d),
$$
using a finite distinguishing density operator $\widehat{\rho}$ such that
$$
\tr(L\widehat{\rho})=\tr(L'\widehat{\rho}) \iff L=L'
$$
for all effects $L,L'$ in the finite effect algebra. Kernel bisimilarity is therefore complete for the intended observational semantics.

The paper also defines a stronger “locally parameterized probabilistic” interpretation based on the isomorphism
$$
\ef[H_C]\cong \mathbf{Conv}(\dm[H_C],[0,1]),
$$
with induced natural transformation
$$
\alpha_{\mathit{lpp}:Q_C\to \emon[f_C]
$$
where $f_C=\mathbf{Conv}(\dm[H_C],[0,1])$. The theorem stated is that kernel bisimilarity on $Q_C$-coalgebras is exactly preserved and reflected by this transformation, so coalgebraic kernel equivalence coincides with the lpp semantics.

The graded-monadic framework also supports process-calculus-style operators. Parallel composition is defined in two stages: first a synchronization operator on transition structures, then a combination of weights via the canonical transformation
$$
\alpha:T_mX\times T_nY\to T_{m+n}(X\times Y).
$$
For CCS-style synchronization, the source gives the rules
$$
\infer{s \parallel t \xrightarrow{\mu} \langle \Delta,\ \xi_{0 \preceq n} (\eta(t)) \rangle}{s \xrightarrow{\mu} \Delta}
\qquad
\infer{s \parallel t \xrightarrow{\mu} \langle \xi_{0 \preceq m} (\eta(s)) ,\ \Theta \rangle}{t \xrightarrow{\mu} \Theta}
\qquad
\infer{s \parallel t \xrightarrow{\tau} \langle \Delta,\ \Theta\rangle} {s \xrightarrow{\mu} \Delta \ {data}\ t \xrightarrow{\mu} \Theta}.
$$
The “idle” behavior of a process is represented by the monad unit $\eta$, extended to the appropriate grade by $\xi$; in the quantum case, this corresponds to leaving unused qubits untouched via identity measurement on the complement. Parallel composition becomes the functor
$$
\parallel:\coalgcat[P T_m]\times \coalgcat[P T_n]\to \coalgcat[P T_{m+n}], \qquad (X,c)\parallel(Y,d)=(X\times Y,\;P\alpha\circ(c|d)),
$$
and it preserves bisimilarity.

A quantum-specific partial evaluation operator is also defined. For $C'\subseteq C$ and $\rho\in\dm[H_{C'}]$,
$$
L \mapsto \tr_{C'}\!\bigl(L(\rho\boxtimes \mathbb{I})\bigr)
$$
induces a functor
$$
\#1{\rho}:\coalgcat[PQ_C]\to \coalgcat[PQ_{C\setminus C'}]
$$
that instantiates part of the quantum input and reduces the resource grade accordingly. This operator preserves kernel bisimilarity as well [2509.20933].

Taken together, these constructions give qLTSs a compositional and resource-sensitive semantics for quantum processes parameterized by quantum inputs. The paper’s conclusion is that kernel bisimilarity, rather than AM-bisimilarity, matches the observable behaviour obtained across probabilistic instantiations, while the graded structure ensures that composition respects the disjointness constraints imposed by quantum resources.

Source: https://www.emergentmind.com/topics/quantum-effect-labelled-transition-systems