Quantum Channels: CPTP Maps

• Quantum channels are linear maps that are completely positive and trace-preserving, described via the Kraus theorem to ensure valid quantum system evolutions.
• They are characterized using tools such as the Choi matrix, categorical frameworks, and convex geometry to study structural and extremal properties.
• Quantum channels underpin practical applications in quantum control, tomography, and error correction by linking abstract theory with operational protocols.

A quantum channel is a linear map between finite-dimensional C*-algebras that is completely positive and trace-preserving (CPTP). In quantum information theory, CPTP maps serve as the fundamental mathematical representation of physical transformations applied to quantum systems, encompassing both reversible (unitary) and irreversible (noisy, decohering, dissipative) evolutions. The abstract, categorical, algebraic, topological, and information-theoretic properties of quantum channels carry crucial implications for quantum computation, quantum communication, and quantum statistical mechanics.

1. Fundamental Definitions and Structure

Let $H$ and $K$ be finite-dimensional Hilbert spaces, with $B(H)$ the algebra of all linear operators acting on $H$. A quantum channel $\Phi: B(H) \to B(K)$ is a linear map satisfying:

• Complete positivity (CP): For every integer $n$, the map $$id_n \otimes \Phi: B(\mathbb{C}^n \otimes H)\to B(\mathbb{C}^n \otimes K)$$ sends positive semidefinite operators to positive semidefinite operators.
• Trace preservation (TP): For all $\rho \in B(H)$, $$\mathrm{Tr}\left[\Phi(\rho)\right] = \mathrm{Tr}[\rho]$$.

By the Kraus theorem, any CPTP map admits an operator-sum representation:

$$\Phi(\rho) = \sum_{i} K_i \rho K_i^\dagger, \qquad \sum_{i} K_i^\dagger K_i = I_H$$

where $\{K_i\}$ are Kraus operators. Complete positivity is equivalent to the existence of such a decomposition; trace preservation is enforced by the normalization condition on the Kraus operators.

The condition can also be expressed via the Choi matrix $$C_\Phi = (\Phi \otimes id)(|\Omega\rangle\langle \Omega|)$$, where $$|\Omega\rangle = \sum_j |j\rangle \otimes |j\rangle$$, such that $\Phi$ is CP iff $C_\Phi \geq 0$, and TP iff $\mathrm{Tr}_K C_\Phi = I_H$ (Coecke et al., 2013).

2. Categorical and Algebraic Organization

Quantum channels fit naturally into the categorical framework of dagger-compact categories. The CP*-construction introduced by Coecke–Heunen–Kissinger formalizes this: given the category FHilb of finite-dimensional Hilbert spaces and linear maps, the CP*-construction produces a new category whose objects are finite-dimensional C*-algebras and whose morphisms are completely positive maps (CP); imposing the counit-preservation (trace-preserving) condition selects out CPTP maps (Coecke et al., 2013).

• Objects: finite-dimensional C*-algebras $A \cong \bigoplus_i M_{n_i}(\mathbb{C})$.
• Morphisms: linear maps $\Lambda: A \to B$ that are (a) completely positive, (b) trace-preserving.
• Tensor structure: The category is monoidal, and the tensor product of normalisable Frobenius algebras (C*-algebras) is again a normalisable Frobenius algebra.
• Dagger structure: The adjoint $f^\dagger$ of a CP map $f$ is also CP.

Classical channels appear as the subcategory of commutative Frobenius algebras, which correspond to stochastic maps on finite sets.

3. Convex Geometry, Extremal Points, and Tensor Product

The set of CPTP maps $CPTP(X,Y)$ forms a compact convex set. Its extreme points (the "atomic" channels) are characterized by Choi's criterion: a CPTP map $\epsilon$ with minimal Kraus operators $\{E_k\}$ is extremal iff the set $\{E_j^\dagger E_k : 1 \leq j,k \leq r\}$ is linearly independent (Silva, 2023). Every CPTP map is a convex combination of extreme points.

The tensor product of two CPTP maps $\epsilon$ and $\epsilon'$ gives a CPTP map $\epsilon \otimes \epsilon'$ whose Kraus operators are the tensor products of the original Kraus operators. Extremality is preserved under tensor product for CPTP and unital CP maps but not always for the unital+TP subclass (UCPT), where rank proliferation can cause loss of extremality.

4. Generalizations: Positive, Hermitian-Preserving, and Non-Markovian Maps

Beyond CPTP maps, physical processes may be described by more general trace-preserving maps (Cao et al., 2023):

• Hermitian-preserving trace-preserving (HPTP): Maps that preserve Hermiticity and trace but not necessarily positivity or complete positivity.
• Semi-positive trace-preserving (SP-TP): HPTP maps such that there exists at least one invertible input state mapped to a valid density operator.
• Semi-nonnegative trace-preserving (SN-TP): HPTP maps for which some (not necessarily full-rank) input is mapped to a valid density operator.

Set inclusions establish the hierarchy:

$$CPTP \subseteq P\,TP \subseteq SP \subseteq SN \subseteq HPTP$$

The convex structure is nontrivial: CPTP and PTP are convex and compact; SP-TP is open and star-shaped; SN-TP is closed and star-shaped (Cao et al., 2023).

Physical realization of non-CP maps is linked to quantum non-Markovian dynamics (breakdown of CP-divisibility), and error correction extends to SP-TP noise provided the Knill–Laflamme conditions are satisfied.

5. Topological and Geometric Structure

Quantum channels possess a rich topological geometry. The Kraus operator-sum representation provides an embedding into a complex Stiefel manifold $St(n, mk)$ (Russkikh et al., 2024):

• Channels correspond to equivalence classes of Kraus operator tuples up to unitary mixing: $Chan(n,m) \cong St(n, mk)/U(k)$.
• The Stiefel manifold inherits a natural Riemannian metric, inducing a distance on quantum channels via minimization over unitary mixing.
• This geometric structure enables powerful Riemannian optimization techniques for quantum control, tomography, and discrimination.

Bayesian quantum process tomography leverages this structure, defining continuous probability distributions over the Stiefel manifold that respect CPTP constraints, admitting tractable conjugate priors and efficient sampling and estimation algorithms for full Bayesian characterization of channel uncertainty (Schultz, 2017).

6. Tomographic and Probability Representations

Quantum channels can be represented as linear integral maps acting on tomograms—probability distributions associated with quantum states (Amosov et al., 2017, Avanesov et al., 2019). For finite dimensions, the action of a CPTP map on a qudit density matrix translates to an affine transformation on a vector of classical coin probabilities, subject to linear and polynomial constraints induced by trace preservation and complete positivity on the associated dynamical matrix.

This representation allows quantum processes to be described in classical probabilistic terms, enabling reformulation and simulation of quantum information tasks as classical inference problems.

7. Channels in Quantum Dynamics, Spectral Structure, and Asymptotics

The Kraus-operator algebra $\mathcal{A}(K_i)$ generated by the channel's Kraus operators determines the asymptotic dynamics via its spectral properties (Białończyk et al., 2017):

• The peripheral spectrum (eigenvalues on the unit circle) is constrained by the structure of $\mathcal{A}$; peripheral eigenvalues must be roots of unity of bounded order linked to the block sizes in the algebra's decomposition.
• Criterion for primitivity (channel with unique, attractive fixed point): The algebra is irreducible (generates $M_n(\mathbb{C})$).
• Shemesh and Amitsur–Levitzki theorems offer effective combinatorial and algebraic tests for determining mixing properties, convergence speed, and the possible cycle lengths in quantum dynamical semigroups.

The connection to Ruelle transfer operators reveals further analogies between the mixing properties of CPTP maps and the exponential decay of correlations in classical dynamical systems; the barycenter theorem identifies invariant measures as corresponding to channel fixed points (Lardizabal, 2012).

---

Quantum channels, as formalized by CPTP maps, are a unifying object in quantum information theory with categorical, convex-geometric, algebraic, topological, and probabilistic facets. They encompass classical and quantum processes in a single theoretical framework and invite investigation into generalized dynamical maps that probe the boundaries of physically meaningful quantum operations. Their mathematical organization underpins the diverse phenomena of quantum transport, control, measurement, tomography, and open-system dynamics, motivating ongoing research in their extremal structure, representation, resource convertibility, and non-Markovian extensions (Coecke et al., 2013, Cao et al., 2023, Russkikh et al., 2024, Silva, 2023, Huot et al., 2019, Mitra et al., 8 Dec 2025, Amosov et al., 2017, Avanesov et al., 2019, Białończyk et al., 2017, Lardizabal, 2012).