Quantum Channels and the Reversible/Irreversible Dichotomy

Introduction

The companion article Quantum Physics in Biquaternionic Form developed the qubit in the biquaternion algebra $\mathbb{B} = \mathbb{C}\otimes_\mathbb{R}\mathbb{H}$: states and observables are elements of the Hermitian subspace $\mathbb{B}_{\Psi}$; reversible evolution is rotor conjugation $\tilde{Ϛ}_{\Psi} \mapsto \tilde Q\tilde{Ϛ}_{\Psi}\tilde Q^{*}$ by a $*$-unitary biquaternion; measurement is the two-sided operation $\tilde{Ϛ}_{\Psi} \mapsto \tilde\Pi\tilde{Ϛ}_{\Psi}\tilde\Pi$. It listed the dynamics of open systems — dissipation, decoherence, and the Lindblad equation — among its open questions.

This article takes up that question. Its subject is the general state map on $\mathbb{B}_{\Psi}$: the most general process carrying a state of the informational sector to a state of the informational sector — in quantum information theory, a quantum channel or quantum operation. We develop the requirements that define it — linearity, complete positivity, and trace preservation — in the biquaternion language, and then examine the reversible/irreversible dichotomy they produce.

The dichotomy has a sharp algebraic form here. A state map is reversible if and only if it is conjugation by a single $*$-unitary biquaternion, $\tilde{Ϛ}_{\Psi} \mapsto \tilde Q\tilde{Ϛ}_{\Psi}\tilde Q^{*}$ with $\tilde Q\tilde Q^{*} = e_t$. Each one-parameter subgroup is generated by a Hamiltonian $\tilde H_\Psi \in \mathbb{B}_{\Psi}$ through the exponential $\tilde Q(r'_t) = \exp(-i\theta_{\mathrm{evo}})$. The irreversible maps are the rest: the completely positive, trace-preserving maps, which are sums of conjugations rather than a single conjugation. The canonical example is dephasing, worked out below in the $\mathbb{B}_{\Psi}$ formulation.

The angle is $\theta_{\mathrm{evo}}$, the scalar part of $\tilde H_\Psi$ times the interval over $\hbar\,c$: the generator is anti-Hermitian, so its scalar part is purely imaginary and the angle is real.

The relation to the two subspaces is the conceptual thread. States live in $\mathbb{B}_{\Psi}$; the reversible motion of those states is generated by elements of the complementary subspace $\mathbb{B}_{\Phi}$; and the irreversible maps are exactly those that are not inner automorphisms of the algebra. Irreversibility is thus a statement about which kinds of map the algebra admits, not an extra postulate imposed on it.

Two cautionary notes, in the spirit of the companion articles. First, the mathematics is standard quantum information theory transcribed into biquaternion notation; no result below depends on the physical hypothesis that $\mathbb{B}_{\Psi}$ is a distinct sector of the world. Second, the reformulation reframes the dichotomy without resolving the conceptual problems that attend it in standard quantum physics — the position taken in Entangled Subsystems in the Biquaternion Framework, that the framework changes the language and not the physics, applies here too.

The conventions are those of the companion articles: the quaternion basis $e_t = 1, \vec e_x, \vec e_y, \vec e_z$ with $\vec e_\eta^2 = -e_t$; the central scalar imaginary $i$; the fixed-point subspaces $\mathbb{B}_{C}$ (the complex subspace, the centre), $\mathbb{B}_{H}$ (the real-quaternion subspace), $\mathbb{B}_{\Psi}$ (Hermitian), and $\mathbb{B}_{\Phi}$ (anti-Hermitian); a general element $\tilde{Q} = Q_t e_t + Q_x \vec e_x + Q_y \vec e_y + Q_z \vec e_z$ with $Q_\mu \in \mathbb{C}$; and the scalar part $\mathrm{Sc}(\tilde{Q})$, which every statement below uses directly.

The State Space and the Plain Form

A state of the informational sector is an element $\tilde{Ϛ}_{\Psi} = \tfrac{1}{2}(e_t + i\vec s_{\Psi})$ of $\mathbb{B}_{\Psi}$ with $\vec s_{\Psi} \in \mathbb{R}^3$, positive if and only if $|\vec s_{\Psi}| \leq 1$. The states form the Bloch ball, whose boundary is the set of pure states; these are exactly the idempotents $\tilde\Pi_{1,2}(\vec u) = \tfrac{1}{2}(e_t \pm i\vec u)$, with $\vec u$ a unit pure real quaternion. Every state is normalised by the Hermitian form, $\langle\tilde{Ϛ}_{\Psi},e_t\rangle_{*}=\tfrac12\langle e_t\rangle_{*}$, that is $\mathrm{Sc}(\tilde{Ϛ}_{\Psi})=\tfrac12$.

An observable is a general Hermitian element $\tilde H_\Psi = h_t e_t + i\vec h_\Psi$, and the Hermitian form

$$ \langle\tilde{Ϛ}_{\Psi},\tilde H_\Psi\rangle_{*} = \mathrm{Sc}(\tilde{Ϛ}_{\Psi}\tilde H_\Psi^{*}) = \tfrac12(h_t + \vec s_{\Psi}\cdot\vec h_\Psi) $$

is the Born pairing. A state is normalised by $\langle\tilde{Ϛ}_{\Psi},e_t\rangle_{*}=\tfrac12\langle e_t\rangle_{*}$, so the Born weight is $h_t + \vec s_{\Psi}\cdot\vec h_\Psi$, and the same number is $2\,\mathrm{Sc}(\tilde{Ϛ}_{\Psi}\tilde H_\Psi^{*}) = h_t + \vec s_{\Psi}\cdot\vec h_\Psi$. On $\mathbb{B}_{\Psi}$ this form is the symmetric, positive-definite bilinear form, the Euclidean pairing $\langle\tilde H_\Psi,\tilde{Q}\rangle_{*} = h_t\,Q_t + \vec h_\Psi\cdot\vec Q$, whose doubled value $2\langle\tilde H_\Psi,\tilde{Q}\rangle_{*}$ is the number the matrix picture reads as a trace. It plays a second, equally central role here: it is the operation that pairs a state map with its dual, and therefore the operation that makes the Heisenberg picture of a channel possible.

Two structural facts will be used repeatedly. First, $\mathbb{B}$ is the complexification of the real subspace $\mathbb{B}_{\Psi}$, with the central scalar $i$ as the complex structure: $i\,\mathbb{B}_{\Psi} = \mathbb{B}_{\Phi}$ and $\mathbb{B} = \mathbb{B}_{\Psi} \oplus \mathbb{B}_{\Phi}$, so every element decomposes uniquely into a Hermitian and an anti-Hermitian part. Second, positivity is the biquaternion-norm condition $\langle\tilde{Ϛ}_{\Psi}\rangle_{\natural} = \tfrac{1}{4}(1 - |\vec s_{\Psi}|^2)e_t \geq 0$; the state space is the state slice of the future light cone of the biquaternion norm, normalised by $\langle\tilde{Ϛ}_{\Psi},e_t\rangle_{*}=\tfrac12$.

State Maps on $\mathbb{B}_{\Psi}$

Linearity, Positivity, and Complete Positivity

A state map is a map $Ξ$ sending states to states. We require it to be linear. This is the standard framework of quantum operations and a genuine restriction: nonlinear modifications are excluded by the requirement that the map act consistently on mixtures, since a state prepared as a convex combination of ensembles must be transformed in a way that respects the preparation. In the biquaternion language, $Ξ$ is a real-linear map on $\mathbb{B}_{\Psi}$,

$$ Ξ : \mathbb{B}_{\Psi} \longrightarrow \mathbb{B}_{\Psi}, \qquad Ξ(a\tilde{Ϛ}_{\Psi,1} + b\tilde{Ϛ}_{\Psi,2}) = a\,Ξ(\tilde{Ϛ}_{\Psi,1}) + b\,Ξ(\tilde{Ϛ}_{\Psi,2}), \qquad a, b \in \mathbb{R}. $$

Because $\mathbb{B} = \mathbb{B}_{\Psi} \oplus i\mathbb{B}_{\Psi}$ is the complexification of $\mathbb{B}_{\Psi}$, every real-linear map on $\mathbb{B}_{\Psi}$ extends uniquely to a complex-linear map on $\mathbb{B}$, $Ξ(x + iy) = Ξ(x) + iΞ(y)$, and it is standard to work with this extension, regarding a quantum operation as a map on $\mathbb{B} \cong M_2(\mathbb{C})$.

The first requirement is positivity: $\tilde{Ϛ}_{\Psi} \geq 0$ must imply $Ξ(\tilde{Ϛ}_{\Psi}) \geq 0$. Positivity alone is not enough. The system may be entangled with a reference system on which $Ξ$ does not act, and a map that is positive on the system alone can fail to be positive on the joint system. Complete positivity is the condition that the extended map

$$ Ξ \otimes \mathrm{id}_n : \mathbb{B}\otimes M_n(\mathbb{C}) \longrightarrow \mathbb{B}\otimes M_n(\mathbb{C}) $$

be positive for every $n$. A positive but not completely positive map is not a physical process, because acting on half of an entangled pair it can produce an operator with a negative eigenvalue.

The standard illustration is the transpose $\tilde{Q} \mapsto \tilde{Q}^{\mathsf{T}}$ in a fixed matrix representative. It is positive and complex-linear, but not completely positive: transposing one factor of a maximally entangled two-qubit state produces a non-positive operator. Complete positivity is strictly stronger than positivity, and it is the condition the algebra imposes on physical operations.

Normalisation Preservation and the Dual Map

Because a state map must send normalised elements to normalised elements, we require $\langleΞ(\tilde{Ϛ}_{\Psi}),e_t\rangle_{*} = \langle\tilde{Ϛ}_{\Psi},e_t\rangle_{*}$ for all states, equivalently $\langleΞ(\tilde{Q}),e_t\rangle_{*} = \langle\tilde{Q},e_t\rangle$ for all $\tilde{Q} \in \mathbb{B}$; in the matrix picture this is trace preservation, and in the algebra it is preservation of the scalar part, $\mathrm{Sc}(Ξ(\tilde{Ϛ}_{\Psi})) = \mathrm{Sc}(\tilde{Ϛ}_{\Psi})$. The plain form defines a dual map on observables, the Heisenberg-picture representative of $Ξ$, by

$$ \langleΞ(\tilde{Ϛ}_{\Psi}),\tilde H_\Psi\rangle_{*} = \langle\tilde{Ϛ}_{\Psi},Ξ^{*}(\tilde H_\Psi)\rangle_{*} \qquad \text{for all } \tilde{Ϛ}_{\Psi}, \tilde H_\Psi. $$

Normalisation preservation of $Ξ$ is equivalent to unitality of the dual, $Ξ^{*}(e_t) = e_t$: setting $\tilde H_\Psi = e_t$ gives $\langleΞ(\tilde{Ϛ}_{\Psi}),e_t\rangle_{*} = \langle\tilde{Ϛ}_{\Psi},Ξ^{*}(e_t)\rangle_{*}$, which equals $\langle\tilde{Ϛ}_{\Psi},e_t\rangle_{*}$ for every $\tilde{Ϛ}_{\Psi}$ precisely when $Ξ^{*}(e_t) = e_t$. The duality is mediated throughout by the plain form that gives the Born rule, $2\langle\tilde{Ϛ}_{\Psi},\tilde H_\Psi\rangle_{*} = 2\,\mathrm{Sc}(\tilde{Ϛ}_{\Psi}\tilde H_\Psi^{*})$.

The Kraus Representation

The structure theorem for completely positive maps is the Kraus representation (the Choi–Kraus theorem). A map $Ξ$ on $\mathbb{B}$ is completely positive if and only if there exist finitely many Kraus operators $\tilde M_{\mathrm{kraus},l} \in \mathbb{B}$ with

$$ \boxed{\;Ξ(\tilde{Q}) = \sum_l \tilde M_{\mathrm{kraus},l}\,\tilde{Q}\,\tilde M_{\mathrm{kraus},l}^{*}\;} , $$

and it is trace preserving if and only if $\sum_l \tilde M_{\mathrm{kraus},l}^{*} \tilde M_{\mathrm{kraus},l} = e_t$. A completely positive, trace-preserving map is a quantum channel; its dual is $Ξ^{*}(\tilde H_\Psi) = \sum_l \tilde M_{\mathrm{kraus},l}^{*} \tilde H_\Psi\tilde M_{\mathrm{kraus},l}$, and the normalization is exactly $Ξ^{*}(e_t) = e_t$. The representation is not unique: two Kraus sets define the same channel precisely when they differ by a partial isometry. The Kraus rank is the minimal number of Kraus operators, equal to the rank of the Choi matrix $J(Ξ) = \sum_{j\eta} \tilde Q_{j\eta}\otimesΞ(\tilde Q_{j\eta})$, written on a basis $\{\tilde Q_{jk}\}$ of $\mathbb{B}$, and complete positivity is the condition $J(Ξ) \geq 0$. For a qubit channel the Choi matrix is $4\times4$, so the Kraus rank is at most four.

The companion article already introduced this form for generalized measurements. The point here is that it describes every physical process, not only measurement.

The Cartan Factor of a Kraus Operator

The Kraus representation is a sum, but each term has its own generalized polar decomposition. Write

$$ \tilde M_{\mathrm{kraus},l}=\tilde H_{\Psi,\mathrm{pos},l}\,\tilde U_{\mathrm{p},l}, \qquad \tilde H_{\Psi,\mathrm{pos},l}=\sqrt{\tilde M_{\mathrm{kraus},l}\tilde M_{\mathrm{kraus},l}^{*}}, $$

with $\tilde H_{\Psi,\mathrm{pos},l}$ Hermitian positive semidefinite and $\tilde U_{\mathrm{p},l}$ a partial isometry. The companion article The Polar Decomposition of a Kraus Operator reads the factorization physically: the positive factor is the strength of the outcome, the partial isometry its backaction, and the effect is their conjugated square.

For a channel with a single Kraus operator, trace preservation is $\tilde M_{\mathrm{kraus}}^{*}\tilde M_{\mathrm{kraus}}=e_t$. In finite dimension this makes $\tilde M_{\mathrm{kraus}}$ invertible, so $\tilde M_{\mathrm{kraus}}\tilde M_{\mathrm{kraus}}^{*}=e_t$, hence $\tilde H_{\Psi,\mathrm{pos}}=e_t$ and $\tilde Q_{\mathrm{p}}=\tilde M_{\mathrm{kraus}}$ is $*$-unitary. The reversible case is therefore exactly the case of a trivial positive factor, and the one-term channel is conjugation by a $*$-unitary. This is the same statement as §The Reversible Case, read on the operator rather than on the map.

A single term of a multi-term channel need not be trace preserving. Its $\tilde H_{\Psi,\mathrm{pos}}$ can be non-trivial, and its $\tilde Q_{\mathrm{p}}$ can be a proper partial isometry. The channel's reversibility is not the value of one $\tilde H_{\Psi,\mathrm{pos}}$; it is the minimal Kraus rank of the sum. The dephasing channel is the witness: one of its Kraus operators is a scalar multiple of the identity, with scalar strength and $\tilde Q_{\mathrm{p}}=e_t$, and the channel is irreversible for $p>0$ because there is a second term. The polar factor classifies a term; the Kraus rank classifies the map.

The Reversible Case: $*$-Unitary Rotor Conjugation

$*$-Unitary Elements and Their Conjugation Action

The simplest channel is conjugation by a single $*$-unitary biquaternion:

$$ Ξ_{\tilde Q}(\tilde{Ϛ}_{\Psi}) = \tilde Q\,\tilde{Ϛ}_{\Psi}\,\tilde Q^{*}, \qquad \tilde Q \in \mathbb{B}, \quad \tilde Q\tilde Q^{*} = e_t . $$

This is a channel with the single Kraus operator $\tilde M_{\mathrm{kraus}} = \tilde Q$, and it maps states to states because $\tilde Q\tilde{Ϛ}_{\Psi}\tilde Q^{*}$ is Hermitian, positive, and of trace one.

Here "$*$-unitary" means $*$-unitary in the matrix sense, $\tilde Q\tilde Q^{*} = e_t$, and the $*$-unitary biquaternions form the group $\mathrm{U}(2) \subset \mathbb{B}$. This must be distinguished from the condition $\tilde Q\tilde Q^{\natural} = e_t$ of unit biquaternion norm, which defines $SL(2,\mathbb{C})$ and describes the Lorentz rotors acting on the material sector $\mathbb{B}_{\Phi}$. A Lorentz boost rotor is Hermitian and hence is not matrix-unitary: for a boost, $\tilde Q\tilde Q^{*} = \tilde Q^2 \neq e_t$. Conjugation by a boost does not preserve the scalar part, so a boost is not a channel. The reversible processes of the informational sector are the matrix-unitary conjugations. The distinction is one of sector, not of group type. On $\mathbb{B}_{\Phi}$ the group $\{\tilde Q:\tilde Q\tilde Q^{\natural}=e_t\}=SL(2,\mathbb{C})$ gives the whole proper orthochronous Lorentz group, up to the double cover; on $\mathbb{B}_{\Psi}$ only the compact factor acts, so the image is the maximal compact subgroup $\mathrm{U}(2)$. The $\mathbb{B}_{\Psi}$ transformation is therefore the compact half of the same group, not a second group: Lorentz is boost times rotation, and on $\mathbb{B}_{\Psi}$ only the rotation survives.

Under $\tilde{Ϛ}_{\Psi} = \tfrac{1}{2}(e_t + i\vec s_{\Psi})$, the state $Ξ_{\tilde Q}(\tilde{Ϛ}_{\Psi})$ is again of the form $\tfrac{1}{2}(e_t + i\vec s_{\Psi,2})$ with $|\vec s_{\Psi,2}| = |\vec s_{\Psi}|$; the scalar part is fixed and the vector part is rotated. Purity, $2\,\mathrm{Sc}(\tilde{Ϛ}_{\Psi}^2) = \tfrac{1}{2}(1 + |\vec s_{\Psi}|^2)$, and the spectrum $\lambda_\pm = (1\pm|\vec s_{\Psi}|)/2$ are invariant. The $*$-unitary channels are exactly the rotations of the Bloch ball. The short name for this action is the Bloch rotation, the conjugation $Ξ_{\tilde Q_0}$ by the compact factor $\tilde Q_0$ of the polar decomposition. Its kernel is the central circle $\mathrm{U}(1)$ and its image is $\mathrm{U}(2)/\mathrm{U}(1)\cong SO(3)$: the name records what the states see, a rotation, not the group element that produces it.

The One-Parameter Group Generated by a Hamiltonian

The $*$-unitary channels describing continuous evolution are the one-parameter subgroups generated by Hamiltonians. For Hermitian $\tilde H_\Psi = h_t e_t + i\vec h_\Psi \in \mathbb{B}_{\Psi}$, the Schrödinger equation $\frac{i\,\hbar\,d}{dr'_{t}}\tilde Q(r'_t) = \tilde H_\Psi\tilde Q(r'_t)$ with $\tilde Q(0) = e_t$ has the solution

$$ \tilde Q(r'_t) = \exp\!\left(-\frac{i\,\tilde{H}_{\Psi}\,r'_{t}}{\hbar\,c}\right) = \exp(-i\theta_{\mathrm{evo}})\left(\cos\!\left(\frac{|\vec h_\Psi|r'_t}{\hbar\,c}\right)e_t + \sin\!\left(\frac{|\vec h_\Psi|r'_t}{\hbar\,c}\right)\vec u_\Psi\right), \qquad \vec u_\Psi = \frac{\vec h_\Psi}{|\vec h_\Psi|}, $$

the trigonometric form following from $\vec u_\Psi^2 = -e_t$. The elements satisfy $\tilde Q(r'_t)\tilde Q(r'_t)^{*} = e_t$ and form a one-parameter group, $\tilde Q(r'_t+s) = \tilde Q(r'_t)\tilde Q(s)$, because they are functions of the single element $\tilde H_\Psi$ and therefore commute. The state evolves by rotor conjugation $\tilde{Ϛ}_{\Psi}(r'_t) = \tilde Q(r'_t)\tilde{Ϛ}_{\Psi}(0)\tilde Q(r'_t)^{*}$, or in differential form $\frac{i\,\hbar\,d}{dr'_{t}}\tilde{Ϛ}_{\Psi}(r'_t) = [H, \tilde{Ϛ}_{\Psi}(r'_t)]$, the von Neumann equation.

The Generator Lies in $\mathbb{B}_{\Phi}$

The generator of the group is $-\frac{i\,\tilde{H}_{\Psi}}{\hbar\,c} \in \mathbb{B}_{\Phi}$, since $\tilde H_\Psi \in \mathbb{B}_{\Psi}$ implies $(-i H)^{*} = i\tilde H_\Psi^{*} = i\tilde H_\Psi = -(-i H)$. Because $i\mathbb{B}_{\Psi} = \mathbb{B}_{\Phi}$, every element of $\mathbb{B}_{\Phi}$ is of this form, and the exponential map sends $\mathbb{B}_{\Phi}$ onto the $*$-unitary group. This is the cleanest expression of the complementarity that organizes the subject: states and observables live in $\mathbb{B}_{\Psi}$, while the generators of their reversible evolution live in $\mathbb{B}_{\Phi}$. The subspace $\mathbb{B}_{\Phi}$ is the Lie algebra of the $*$-unitary group under the commutator, and reversible dynamics is its action on $\mathbb{B}_{\Psi}$ by inner derivations.

The fixed-point subspaces $\mathbb{B}_{C}$ and $\mathbb{B}_{H}$ mark two extremes. The scalar part $h_t e_t$ lies in $\mathbb{B}_{C}\cap\mathbb{B}_{\Psi} = \mathbb{R}e_t$, the centre; it generates the central phase $\exp(-i\theta_{\mathrm{evo}})$, which commutes with everything and cancels in the conjugation, affecting only the unobservable global phase. The scalar-free part $i\vec h_\Psi$ has $h_t = 0$, and then

$$ \tilde Q(r'_t) = \cos\!\left(\frac{|\vec h_\Psi|r'_t}{\hbar\,c}\right)e_t + \sin\!\left(\frac{|\vec h_\Psi|r'_t}{\hbar\,c}\right)\vec u_\Psi \in \mathbb{B}_{H} $$

is invertible real quaternion, whose conjugation is an ordinary spatial rotation of the Bloch sphere, the $SU(2)$ subgroup of $\mathrm{U}(2)$.

The exponential is onto the invertibles. The statement that the exponential of $\mathbb{B}_{\Phi}$ is the $*$-unitary group is the compact case of a wider statement. The exponential is surjective onto the whole group of invertibles,

$$ \exp:\mathbb{B}\longrightarrow\mathbb{B}^{\times}, $$

not only onto the $*$-unitary slice (Biquaternion Elementary Functions, §The Logarithm; the algebra is $M_2(\mathbb{C})$, and the isomorphism carries one exponential to the other). Every invertible element therefore has a generator, but the generator is unique only up to the kernel of the exponential, and there exists a logarithm in $\mathbb{B}_{\Phi}$ exactly when the element is $*$-unitary. A boost has a logarithm, and that logarithm is not a Hamiltonian: it lies in $\mathbb{B}_{\Psi}$. The reversible/irreversible dichotomy is therefore not a dichotomy of generators but of Kraus rank.

The kernel is explicit. For an element with $B\neq0$ the logarithms differ by the parity-matched lattice $\pi i m\,e_t+\pi n\,\vec U$ with $m\equiv n\pmod2$, so two generators of the same reversible process differ by a gauge shift and not by a physical parameter. The Exponential, the Logarithm and the Generator reads that ambiguity and its limits.

Reversibility

The map $Ξ_{\tilde Q}$ is invertible, with inverse $Ξ^{-1}_{\tilde Q} = Ξ_{\tilde Q^{*}} = Ξ_{\tilde Q(-r'_t)}$, again a $*$-unitary channel: $*$-unitary evolution is reversible.

The converse is the precise statement of the dichotomy. A completely positive, trace-preserving map on $\mathbb{B} \cong M_2(\mathbb{C})$ is invertible, with an inverse that is again completely positive and trace preserving, if and only if it is conjugation by a $*$-unitary element; equivalently, a channel preserves the purity of every state if and only if it is a $*$-unitary conjugation. In the Kraus language the criterion is transparent: a channel with Kraus rank one has the form $Ξ(\tilde{Q}) = \tilde M_{\mathrm{kraus}}\tilde{Q}\tilde M_{\mathrm{kraus}}^{*}$ with $\tilde M_{\mathrm{kraus}}^{*}\tilde M_{\mathrm{kraus}} = e_t$, so $\tilde M_{\mathrm{kraus}}$ is $*$-unitary; and a $*$-unitary conjugation has Kraus rank one. Hence

$$ \boxed{\;\text{reversible} \quad\Longleftrightarrow\quad \text{Kraus rank one} \quad\Longleftrightarrow\quad Ξ = Ξ_{\tilde Q} \text{ for some }{}^{*}\text{-unitary } \tilde Q\;} . $$

The kernel of $\tilde Q \mapsto Ξ_{\tilde Q}$ on $\mathbb{B}_{\Psi}$ is the group of central $*$-unitary elements, the unit circle $\mathrm{U}(1) \subset \mathbb{B}_{C}$; unitaries differing by a global phase define the same channel. The effective group is $PU(2) \cong SO(3)$, acting on the Bloch ball by rotations — the biquaternion form of the unobservability of global phase.

The Irreversible Case: General Completely Positive Maps

Sums of Conjugations and Stinespring Dilation

A general quantum channel is a sum of conjugations, $Ξ(\tilde{Ϛ}_{\Psi}) = \sum_l \tilde M_{\mathrm{kraus},l}\tilde{Ϛ}_{\Psi}\tilde M_{\mathrm{kraus},l}^{*}$ with $\sum_l \tilde M_{\mathrm{kraus},l}^{*}\tilde M_{\mathrm{kraus},l} = e_t$, involving two or more Kraus operators. But a trace-preserving channel with a single Kraus operator is necessarily $*$-unitary, since $\tilde M_{\mathrm{kraus}}^{*}\tilde M_{\mathrm{kraus}} = e_t$ is unitarity. It follows that a channel is irreversible precisely when its Kraus rank is at least two. Irreversibility is not a matter of degree in the action of a single element; it is the presence of a sum, and the choice among the terms is what the map forgets.

The information-theoretic picture is that the Kraus operators describe the interaction of the system with an environment that is not observed. This is made precise by the Stinespring dilation: every channel can be written

$$ \boxed{\;Ξ(\tilde{Ϛ}_{\Psi}) = 2\,\mathrm{Sc}_E\!\left(\tilde{V}_{\Phi}\,\tilde{Ϛ}_{\Psi}\,\tilde{V}_{\Phi}^{*}\right)\;} $$

for an isometry $\tilde{V}_{\Phi}$ into the system tensor an environment, with $2\,\mathrm{Sc}_E$ the reduction over the environment factor, written as in the companion article on entangled subsystems, and the Kraus operators the environment components of $\tilde{V}_{\Phi}$. This is the deep structural statement of the dichotomy: every irreversible process on the system is the reduction of a reversible evolution on a larger system, followed by discarding the environment. The irreversibility is entirely the discarding.

Contraction of the Bloch Ball

Geometrically, a qubit channel is an affine map of the Bloch ball into itself, $\vec s_{\Psi} \mapsto M\vec s_{\Psi} + \vec t$. The $*$-unitary channels are the rotations $M \in SO(3)$, $\vec t = 0$. Two canonical types are distinguished by whether the origin is fixed:

  • Unital channels have $\vec t = 0$ and fix the maximally mixed state, contracting the ball toward the centre. The depolarizing channel $\mathcal{E}_p(\tilde{Ϛ}_{\Psi}) = (1-p)\tilde{Ϛ}_{\Psi} + p\,e_t/2$, with Bloch map $\vec s_{\Psi} \mapsto (1-p)\vec s_{\Psi}$, is completely positive and trace preserving for $0 \leq p \leq 1$; in the biquaternion algebra it has Kraus operators $\{\sqrt{1-\tfrac{3p}{4}}\,e_t,\ \sqrt{\tfrac{p}{4}}\,i\vec e_x,\ \sqrt{\tfrac{p}{4}}\,i\vec e_y,\ \sqrt{\tfrac{p}{4}}\,i\vec e_z\}$, whose normalization uses $(i\vec e_\eta)^{*}(i\vec e_\eta) = e_t$.
  • Non-unital channels have $\vec t \neq 0$ and move the maximally mixed state off centre. The amplitude-damping channel is the standard example, acting on the Bloch vector as $(s_x, s_y, s_z) \mapsto (\sqrt{1-\gamma}\,s_x,\ \sqrt{1-\gamma}\,s_y,\ \gamma + (1-\gamma)s_z)$, $0 \leq \gamma \leq 1$, contractive for $\gamma > 0$ and the identity only at $\gamma = 0$.

Dephasing, treated next, is unital: it contracts the ball toward an axis rather than its centre.

The Lindblad Form

The infinitesimal version of the dichotomy is the Lindblad equation (the Gorini–Kossakowski–Sudarshan–Lindblad form), governing a one-parameter semigroup $Ξ_{r'_{t}}$ of channels:

$$ \boxed{\;\frac{d\tilde{Ϛ}_{\Psi}}{dr'_t} = -\frac{i}{\hbar\,c}\,[H, \tilde{Ϛ}_{\Psi}] + \sum_\eta \left(\tilde M_{\mathrm{kraus},k}\,\tilde{Ϛ}_{\Psi}\,\tilde M_{\mathrm{kraus},k}^{*} - \tfrac{1}{2}\left\{\tilde M_{\mathrm{kraus},k}^{*}\tilde M_{\mathrm{kraus},k}, \tilde{Ϛ}_{\Psi}\right\}\right)\;}, $$

with $\tilde H_\Psi \in \mathbb{B}_{\Psi}$ Hermitian, the jump operators $\tilde M_{\mathrm{kraus},k} \in \mathbb{B}$ arbitrary, and $\{A, B\} = AB + BA$. The two terms separate exactly along the reversible/irreversible split. The Hamiltonian term is generated by $-\frac{i\,H}{\hbar\,\in\,\mathbb{B}}_{\Phi}$ and alone would give a reversible group. The dissipator involves arbitrary jump operators $\tilde M_{\mathrm{kraus},k} \in \mathbb{B}$; each contributes a completely positive term, and the anticommutator keeps the scalar part at one half, since $2\,\mathrm{Sc}(\tilde M_{\mathrm{kraus},k}\tilde{Ϛ}_{\Psi}\tilde M_{\mathrm{kraus},k}^{*} - \tfrac{1}{2}\{\tilde M_{\mathrm{kraus},k}^{*}\tilde M_{\mathrm{kraus},k},\tilde{Ϛ}_{\Psi}\}) = 0$ by cyclicity of the scalar part. The GKSL theorem states that every semigroup of completely positive, trace-preserving maps with a bounded generator has this form. The reversible part of the generator is an inner derivation by an element of $\mathbb{B}_{\Phi}$ (equivalently, by a Hamiltonian in $\mathbb{B}_{\Psi}$); the irreversible part is a sum of completely positive dissipators that are not derivations at all.

Dephasing in $\mathbb{B}_{\Psi}$

Definition and Kraus Form

Dephasing destroys the phase coherence between the two eigenstates of a chosen direction without changing their populations. Let $\vec u$ be a unit pure real quaternion and $\tilde\Pi_1(\vec u) = \tfrac{1}{2}(e_t + i\vec u)$, $\tilde\Pi_2(\vec u) = \tfrac{1}{2}(e_t - i\vec u)$ the complementary idempotents along $\vec u$. For $p \in [0,1]$, the dephasing channel along $\vec u$ is

$$ \boxed{\;Ξ_{\mathrm{deph},p}(\tilde{Ϛ}_{\Psi}) = (1-p)\,\tilde{Ϛ}_{\Psi} + p\left(\tilde\Pi_1\,\tilde{Ϛ}_{\Psi}\,\tilde\Pi_1 + \tilde\Pi_2\,\tilde{Ϛ}_{\Psi}\,\tilde\Pi_2\right)\;}. $$

It is completely positive and trace preserving with the three Kraus operators $\tilde M_{\mathrm{kraus},0} = \sqrt{1-p}\,e_t$, $\tilde M_{\mathrm{kraus},1} = \sqrt{p}\,\tilde\Pi_1(\vec u)$, $\tilde M_{\mathrm{kraus},2} = \sqrt{p}\,\tilde\Pi_2(\vec u)$: because the idempotents are Hermitian, $\tilde M_{\mathrm{kraus},1}^{*}\tilde M_{\mathrm{kraus},1} + \tilde M_{\mathrm{kraus},2}^{*}\tilde M_{\mathrm{kraus},2} = p(\tilde\Pi_1 + \tilde\Pi_2) = p\,e_t$, and the normalization sums to $(1-p)e_t + p\,e_t = e_t$.

The Bloch Vector Formula

Put $\alpha = i\vec u$, so that $\tilde\Pi_{1,2} = \tfrac{1}{2}(e_t \pm \alpha)$ with $\alpha^{*} = \alpha$ and $\alpha^2 = e_t$. A direct expansion gives

$$ \tilde\Pi_1\,\tilde{Q}\,\tilde\Pi_1 + \tilde\Pi_2\,\tilde{Q}\,\tilde\Pi_2 = \tfrac{1}{2}\left(\tilde{Q} + \alpha\,\tilde{Q}\,\alpha\right), $$

since the cross terms cancel between the two idempotents. For a state $\tilde{Ϛ}_{\Psi} = \tfrac{1}{2}(e_t + i\vec s_{\Psi})$, the elementary products are $\alpha\,e_t\,\alpha = \alpha^2 = e_t$ and $\alpha\,(i\vec s_{\Psi})\,\alpha = -i\,\vec u\,\vec s_{\Psi}\,\vec u = -i(\vec s_{\Psi} - 2(\vec u\cdot\vec s_{\Psi})\vec u)$, using $\vec u\,\vec s_{\Psi}\,\vec u = \vec s_{\Psi} - 2(\vec u\cdot\vec s_{\Psi})\vec u$. Combining,

$$ \tilde\Pi_1\tilde{Ϛ}_{\Psi}\tilde\Pi_1 + \tilde\Pi_2\tilde{Ϛ}_{\Psi}\tilde\Pi_2 = \tfrac{1}{2}\left(e_t + i(\vec u\cdot\vec s_{\Psi})\,\vec u\right), $$

so the fully dephasing map projects the Bloch vector onto the $\vec u$ axis. Including the mixture with weight $1-p$,

$$ \boxed{\;Ξ_{\mathrm{deph},p}(\tilde{Ϛ}_{\Psi}) = \tfrac{1}{2}\left(e_t + i\,\vec s_{\Psi,2}\right), \qquad \vec s_{\Psi,2} = (1-p)\,\vec s_{\Psi} + p\,(\vec u\cdot\vec s_{\Psi})\,\vec u\;}. $$

The transverse components of the Bloch vector are multiplied by $1-p$, while the longitudinal component is untouched. In the eigenbasis of $\vec u$, the off-diagonal entries of the density matrix — the coherences — are scaled by $1-p$, while the diagonal entries — the populations — are unchanged. This is the standard phase-damping channel, expressed entirely in the objects of $\mathbb{B}_{\Psi}$.

Limiting Cases, Measurement, and Irreversibility

At $p = 0$ the channel is the identity, a reversible $*$-unitary evolution. At $p = 1$ it is full dephasing, $Ξ_{\mathrm{deph},1}(\tilde{Ϛ}_{\Psi}) = \tfrac{1}{2}(e_t + i(\vec u\cdot\vec s_{\Psi})\vec u)$, whose image is the diameter of the Bloch ball along $\vec u$; it is idempotent, $(Ξ_{\mathrm{deph},1})^2 = Ξ_{\mathrm{deph},1}$, and non-$*$-unitary. Full dephasing is measure-and-forget: measuring $\alpha = i\vec u$ and discarding the outcome leaves the average state

$$ p_+\,\tilde\Pi_1 + p_-\,\tilde\Pi_2 = Ξ_{\mathrm{deph},1}(\tilde{Ϛ}_{\Psi}), \qquad p_\pm = \frac{2\langle\tilde\Pi_{1,2},\tilde{Ϛ}_{\Psi}\rangle_{*}}{2\langle\tilde{Ϛ}_{\Psi},e_t\rangle_{*}} = \frac{2\,\mathrm{Sc}(\tilde\Pi_{1,2}\tilde{Ϛ}_{\Psi})}{2\,\mathrm{Sc}(\tilde{Ϛ}_{\Psi})} = \tfrac{1}{2}\left(1 \pm \vec u\cdot\vec s_{\Psi}\right), $$

as a direct computation confirms. The star is inert on $\mathbb{B}_{\Psi}$, and the scalar part of the product is real and non-negative there, so $p_\pm\in[0,1]$. The three-Kraus form shows that $Ξ_{\mathrm{deph},p}$ is a convex mixture, with weight $p$, of this measurement channel and the identity; since the identity is reversible and full dephasing is not, the channel set is convex with its reversible elements among its extreme points.

Write $s_\parallel = \vec u\cdot\vec s_{\Psi}$ and $s_\perp^2 = |\vec s_{\Psi}|^2 - s_\parallel^2$. Under dephasing,

$$ |\vec s_{\Psi,2}|^2 = (1-p)^2\,s_\perp^2 + s_\parallel^2 \leq |\vec s_{\Psi}|^2 , $$

with equality only if $p = 0$ or $s_\perp = 0$. The purity $2\,\mathrm{Sc}(\tilde{Ϛ}_{\Psi,2}^2) = \tfrac{1}{2}(1 + |\vec s_{\Psi,2}|^2)$ therefore strictly decreases for any off-axis state when $0 < p < 1$, and the von Neumann entropy

$$ S(\tilde{Ϛ}_{\Psi}) = -\lambda_+\log\lambda_+ - \lambda_-\log\lambda_-, \qquad \lambda_\pm = \tfrac{1}{2}\left(1 \pm |\vec s_{\Psi}|\right), $$

strictly increases, since $S$ is a decreasing function of $|\vec s_{\Psi}|$. A pure state on the equator is driven into the interior of the Bloch ball; only the two pure states on the $\vec u$ axis are fixed. No channel can restore the lost coherence, because the inverse map is not completely positive. $*$-Unitary evolution, by contrast, preserves $|\vec s_{\Psi}|$ exactly, and with it the purity and the spectrum of every state.

The Dichotomy, the Two Subspaces, and the Informational Sector

The Algebraic Statement of the Split

Collecting the results, the dichotomy has the following exact algebraic form.

  • States and observables live in $\mathbb{B}_{\Psi}$, the Hermitian subspace, and are paired by the Hermitian form $\langle\tilde{Ϛ}_{\Psi},\tilde H_\Psi\rangle_{*}=\mathrm{Sc}(\tilde{Ϛ}_{\Psi}\tilde H_\Psi^{*})$, the Born rule, evaluated as $2\,\mathrm{Sc}(\tilde{Ϛ}_{\Psi}\tilde H_\Psi^{*})$.
  • Reversible processes are the inner automorphisms $\tilde{Ϛ}_{\Psi} \mapsto \tilde Q\tilde{Ϛ}_{\Psi}\tilde Q^{*}$ by matrix-unitary biquaternions $\tilde Q \in \mathrm{U}(2)$. They are exactly the channels of Kraus rank one, and they preserve purity. Their one-parameter subgroups have generators $-\frac{i\,H}{\hbar\,c}$ lying in the complementary subspace $\mathbb{B}_{\Phi}$: the states live in $\mathbb{B}_{\Psi}$, but their reversible motion is generated from $\mathbb{B}_{\Phi}$.
  • Irreversible processes are the remaining channels: completely positive, trace-preserving maps of Kraus rank at least two. Their Kraus operators are arbitrary elements of $\mathbb{B} = \mathbb{B}_{\Psi} \oplus \mathbb{B}_{\Phi}$, and their generators contain, besides a derivation by an $\mathbb{B}_{\Phi}$ element, completely positive dissipators that are not derivations.

The dichotomy is therefore not a dichotomy between the two subspaces. It is a dichotomy within the set of maps on $\mathbb{B}$, and the two subspaces appear asymmetrically. The Hermitian subspace $\mathbb{B}_{\Psi}$ is the arena: it contains the states acted upon and the observables labelling outcomes. The anti-Hermitian subspace $\mathbb{B}_{\Phi}$ is the source of reversibility: it is the Lie algebra whose exponential is the group of reversible maps, and its elements are the Hamiltonians of motion. The irreversible maps are those needing more than a single element of this Lie algebra; by Stinespring dilation they are the reversible evolutions of a larger system with information discarded in the reduction. Dephasing cannot be written as conjugation by any element at all — its three Kraus operators have no common $*$-unitary representative — and this is why its effect cannot be undone.

The Informational-Sector Reading

Under the interpretive hypothesis of the companion article, that $\mathbb{B}_{\Psi}$ is the arena of an informational sector, this article is a theory of the operations of that sector. The states of $\mathbb{B}_{\Psi}$ are informational configurations; a channel transforms them; the reversible channels are the deterministic, information-preserving transformations, and the irreversible channels are those in which information is lost to an environment the sector does not describe. Dephasing is the canonical such loss: it is decoherence, the destruction of phase relations, expressed as a map on $\mathbb{B}_{\Psi}$.

In this interpretation, reversibility is the regime in which the informational sector is closed and its evolution is generated by the material Hamiltonian; irreversibility is the regime in which the sector is open and its states are entangled with degrees of freedom outside the description, with Stinespring dilation making the discarded environment precise. Nothing here requires the material sector $\mathbb{B}_{\Phi}$ to be "accessed" by the channel: the generator of the reversible part of the motion is an $\mathbb{B}_{\Phi}$ element, but irreversibility itself is a property of the map on $\mathbb{B}_{\Psi}$. As elsewhere in this series, the interpretive reading is a hypothesis and not a result; the channel formalism above is standard quantum information theory, and whether the Hermitian subspace is a distinct physical sector is the open question identified in The Informational Sector: States, Observables and Probability.

What the Reformulation Does and Does Not Claim

The framework does provide a natural algebraic home for the standard theory of quantum operations. Complete positivity is a property of maps on $\mathbb{B} \cong M_2(\mathbb{C})$; preservation of the scalar part is the statement that the dual fixes $e_t$; the Kraus operators are elements of $\mathbb{B}$; and the reversible/irreversible split is the split between Kraus rank one and higher rank, equivalently between inner automorphisms and general completely positive maps. The plain form $\langle\tilde{Ϛ}_{\Psi},\tilde H_\Psi\rangle_{*}=\mathrm{Sc}(\tilde{Ϛ}_{\Psi}\tilde H_\Psi^{*})$ makes the state/observable duality and the Schrödinger/Heisenberg duality of channels the same operation; the doubled value $2\langle\tilde{Ϛ}_{\Psi},\tilde H_\Psi\rangle_{*}$ is the number the matrix picture reads as a trace.

The framework does not claim to explain why a given physical process is irreversible, to derive the environment a channel reduces over, or to resolve the measurement problem. The reformulation changes the vocabulary of the dichotomy, not its physics; the selection problem — why one outcome, or one channel, rather than another — is untouched.

The Dichotomy and the Two Other Two-Class Splits

The dichotomy of this article is the dichotomy of Kraus rank: a channel is reversible exactly when its rank is one, and then it is conjugation by a $*$-unitary biquaternion; it is irreversible exactly when its rank is at least two. That cut must be kept apart from two other two-class splits of the algebra with which it is easily conflated, because they cut on different quantities.

The invertibles and the zero divisors. The multiplication monoid of the algebra is cut by the light cone (The Operators: One-Sided, Two-Sided, Two-Sided Operator): the invertible maps are left multiplications by invertibles, the elements with $\langle\tilde Q\rangle_{\natural}\neq0$, and the non-invertible ones are left multiplications by zero divisors, the elements with $\langle\tilde Q\rangle_{\natural}=0$. This cut is by the norm, and it is not the channel dichotomy — a Kraus operator of a rank-two channel can be a perfectly good unit. The two cuts agree only on the reversible side, where a rank-one channel is built from a single $*$-unitary and the monoid's inverse exists; on the irreversible side they differ, since the monoid's non-invertibles are the null elements and the channel's irreversible maps are the sums of two or more conjugations by invertibles.

The sector splits. Neither cut is the sector split $\mathbb{B}_{\Psi}\oplus\mathbb{B}_{\Phi}$, which is the distinction of states and observables, not of reversible and irreversible maps.

The arrow. The channel cut is genuinely one-way: a channel of Kraus rank at least two cannot be inverted by any channel, and the passage from an invertible element to a sum of conjugations is the algebraic image of tracing out an environment. This is a physical arrow of the map, and it is an interpretation of this article's dichotomy. It must not be confused with the temporal arrow, which the framework does not supply: the exchange of the two sectors is a rotation of order four and carries no orientation, so the algebra offers both senses of the tick and no rule that selects one (Relational Time from the Sector Exchange: The Material and Informational Clocks, which records exactly that absence). Two senses of "irreversible" therefore live in the corpus — the non-invertibility of a map and the missing orientation of time — and they are different questions about different objects.

Open Questions

1. Empirical content. As for the rest of the framework, the central question is whether the biquaternion formulation predicts anything standard quantum information theory does not. A reformulation of quantum channels is, by itself, empirically empty.

2. The geometry of the channel set. The general qubit channel is an affine contraction of the Bloch ball with a specific positivity structure. Which affine contractions are completely positive, and how does the biquaternion description of the Choi matrix read in terms of $\mathbb{B}_{C}$, $\mathbb{B}_{H}$, $\mathbb{B}_{\Psi}$ and $\mathbb{B}_{\Phi}$? A complete classification in the biquaternion language has not been carried out here.

3. Departures from complete positivity. The framework is built on completely positive maps. Physically motivated modifications sometimes relax complete positivity to positivity while retaining trace preservation. How would such a relaxation read in the biquaternion algebra, and is there a natural algebraic home for it?

4. Entropy and the environment. The increase of entropy under dephasing is proved state by state; is there a monotonicity theorem for a general biquaternion channel, in the form of data-processing inequalities for a suitable entropy functional on $\mathbb{B}_{\Psi}$? And is there a dynamics of the "informational environment" of Stinespring dilation, or is it purely formal?

5. The relation to measurement. Dephasing is measure-and-forget for the observable $i\vec u$; more generally every channel is a generalized measurement followed by a state transformation. Whether this unification has consequences beyond the standard one is open.

Summary

A general state map on the informational sector is a linear map on $\mathbb{B}_{\Psi}$, extended complex-linearly to $\mathbb{B} = \mathbb{B}_{\Psi} \oplus \mathbb{B}_{\Phi}$ and required to be completely positive and trace preserving. Complete positivity guarantees that the map remains positive when the system is entangled with a reference system; trace preservation guarantees that it sends states to states. Every such map has a Kraus representation $Ξ(\tilde{Ϛ}_{\Psi}) = \sum_l \tilde M_{\mathrm{kraus},l}\tilde{Ϛ}_{\Psi}\tilde M_{\mathrm{kraus},l}^{*}$ with $\sum_l \tilde M_{\mathrm{kraus},l}^{*}\tilde M_{\mathrm{kraus},l} = e_t$, and the minimal number of Kraus operators is the Choi rank.

The reversible/irreversible dichotomy is the dichotomy of Kraus rank. A channel is reversible if and only if it has Kraus rank one, which happens if and only if it is conjugation by a $*$-unitary biquaternion, $\tilde{Ϛ}_{\Psi} \mapsto \tilde Q\tilde{Ϛ}_{\Psi}\tilde Q^{*}$ with $\tilde Q\tilde Q^{*} = e_t$. This is exactly the class of inner automorphisms of the algebra, and exactly the class of channels that preserve purity. The one-parameter subgroups are generated by Hamiltonians $\tilde H_\Psi \in \mathbb{B}_{\Psi}$ through $\tilde Q(r'_t) = \exp(-i\theta_{\mathrm{evo}})$, whose generators $-\frac{i\,H}{\hbar\,c}$ lie in the complementary subspace $\mathbb{B}_{\Phi}$. A channel is irreversible if and only if its Kraus rank is at least two; the general irreversible channel is a sum of conjugations, and by Stinespring dilation it is the reduction of a reversible evolution on a larger system with the environment discarded.

Dephasing is the canonical irreversible channel. Along a direction $\vec u$, it maps the Bloch vector by $\vec s_{\Psi} \mapsto (1-p)\vec s_{\Psi} + p(\vec u\cdot\vec s_{\Psi})\vec u$, scaling the coherences by $1-p$ and leaving the populations fixed. At $p=1$ it is the idempotent "measure-and-forget" channel of the observable $i\vec u$, whose image is the diameter of the Bloch ball. It strictly decreases purity and strictly increases entropy for off-axis states, and it cannot be inverted by any channel.

The relation of the split to the two subspaces is asymmetric. States and observables live in $\mathbb{B}_{\Psi}$; the generators of reversible evolution live in $\mathbb{B}_{\Phi}$; and the irreversible maps are those that are not inner automorphisms, requiring Kraus operators that are general elements of $\mathbb{B} = \mathbb{B}_{\Psi} \oplus \mathbb{B}_{\Phi}$. The dichotomy is thus an algebraic fact about the space of maps on the biquaternion algebra rather than an additional postulate — but, as the companion articles emphasize, it is a reformulation of standard quantum theory, not a new physical theory.

Summary of Notation

Symbol Meaning
$\mathbb{B} = \mathbb{C}\otimes_\mathbb{R}\mathbb{H}$ Biquaternion algebra, $\cong M_2(\mathbb{C})$
$\mathbb{B}_{\Psi}$ Hermitian subspace (states, observables)
$\mathbb{B}_{\Phi}$ Anti-Hermitian subspace (generators of reversible evolution)
$\mathbb{B}_{C}$ Complex subspace (centre); trace part of a Hamiltonian
$\mathbb{B}_{H}$ Real-quaternion subspace (rotation rotors)
$e_t = 1, \vec e_x, \vec e_y, \vec e_z$ Quaternion basis, $\vec e_\eta^2 = -e_t$
$i$ Scalar imaginary, $i^2 = -1$
$\tilde{Ϛ}_{\Psi} = \tfrac{1}{2}(e_t + i\vec s_{\Psi})$ State of the informational sector
$\tilde H_\Psi = h_t e_t + i\vec h_\Psi$ Observable / Hamiltonian
$\tilde\Pi_{1,2}(\vec u) = \tfrac{1}{2}(e_t \pm i\vec u)$ Idempotents (pure states)
$2\langle\tilde{Ϛ}_{\Psi},\tilde H_\Psi\rangle_{*} = 2\,\mathrm{Sc}(\tilde{Ϛ}_{\Psi}\tilde H_\Psi)$ Born weight at normalisation
$Ξ(\tilde{Ϛ}_{\Psi}) = \sum_l \tilde M_{\mathrm{kraus},l}\tilde{Ϛ}_{\Psi}\tilde M_{\mathrm{kraus},l}^{*}$ Kraus representation of a channel
$\sum_l \tilde M_{\mathrm{kraus},l}^{*}\tilde M_{\mathrm{kraus},l} = e_t$ Trace-preservation condition
$\tilde M_{\mathrm{kraus},l}=\tilde H_{\Psi,\mathrm{pos},l}\tilde U_{\mathrm{p},l}$, $\tilde H_{\Psi,\mathrm{pos},l}=\sqrt{\tilde M_{\mathrm{kraus},l}\tilde M_{\mathrm{kraus},l}^{*}}$ Generalized Cartan factorization of a Kraus operator; $\tilde H_{\Psi,\mathrm{pos},l}\ge0$, $\tilde U_{\mathrm{p},l}$ a partial isometry
$Ξ^{*}(\tilde H_\Psi) = \sum_l \tilde M_{\mathrm{kraus},l}^{*}\tilde H_\Psi\tilde M_{\mathrm{kraus},l}$ Dual (Heisenberg-picture) map
$\tilde Q(r'_t) = \exp(-i\theta_{\mathrm{evo}})$ One-parameter $*$-unitary group
$-\frac{i\,H}{\hbar\,\in\,\mathbb{B}}_{\Phi}$ Generator of reversible evolution
$Ξ_{\mathrm{deph},p}$ Dephasing channel along $\vec u$
$\vec s_{\Psi} \mapsto (1-p)\vec s_{\Psi} + p(\vec u\cdot\vec s_{\Psi})\vec u$ Dephasing action on the Bloch vector
$J(Ξ) = \sum_{j\eta}\tilde Q_{j\eta}\otimesΞ(\tilde Q_{j\eta})$ Choi matrix; Kraus rank $=\mathrm{rank}\,J(Ξ)$
$\langle\tilde{Q}_1,\tilde{Q}_2\rangle_{\natural}$ the general quaternionic bilinear form, $\mathrm{Sc}(\tilde{Q}_1^{\natural}\tilde{Q}_2)$; on the diagonal $\langle\tilde{Q}_2,\tilde{Q}_2\rangle_{\natural}=\langle\tilde{Q}_2\rangle_{\natural}$
$\langle\tilde{Q}_1,\tilde{Q}_2\rangle$ the general plain bilinear form, the scalar part of the general plain bilinear product, $\mathrm{Sc}(\tilde{Q}_1\tilde{Q}_2)$
$\langle\tilde{Q}_1,\tilde{Q}_2\rangle_{*}$ the general plain sesquilinear form, $\mathrm{Sc}(\tilde{Q}_1\tilde{Q}_2^{*})$; on the diagonal $\langle\tilde{Q}_2,\tilde{Q}_2\rangle_{*}=\sum_\mu\lvert Q_{2\mu}\rvert^{2}$

Further Reading

  • The Cartan Hermitian Pairing of the Complete Polar Decomposition of an N-Regular Biquaternion, for the Cartan projection $\tilde Q\mapsto\tilde Q_0$ and the factorization $\tilde Q=\tilde Q_H\tilde Q_0$.
  • Biquaternion Elementary Functions, for the logarithm, its domain and the kernel of the exponential.
  • The Polar Decomposition of a Kraus Operator: Effect, Backaction and the Measurement Boundary, for the strength and backaction of a single outcome and the null-cone boundary.
  • Michael A. Nielsen and Isaac L. Chuang, Quantum Computation and Quantum Information (Cambridge, 2000), for quantum operations, the Kraus representation, and the depolarizing and dephasing channels.
  • K. Kraus, States, Effects, and Operations (Springer, 1983), for the original formulation of completely positive maps in quantum theory.
  • M.-D. Choi, "Completely positive linear maps on complex matrices," Linear Algebra and its Applications 10 (1975) 285–290, for the complete-positivity criterion and the Choi matrix.
  • W. F. Stinespring, "Positive functions on $C^*$-algebras," Proceedings of the American Mathematical Society 6 (1955) 211–216, for the dilation theorem.
  • G. Lindblad, "On the generators of quantum dynamical semigroups," Communications in Mathematical Physics 48 (1976) 119–130, and V. Gorini, A. Kossakowski, and E. C. G. Sudarshan, "Completely positive dynamical semigroups of $N$-level systems," Journal of Mathematical Physics 17 (1976) 821–825, for the GKSL form.
  • M. B. Ruskai, S. Szarek, and E. Werner, "An analysis of completely-positive trace-preserving maps on $2\times2$ matrices," Linear Algebra and its Applications 347 (2002) 159–187, for the affine Bloch-ball picture of qubit channels.
  • P. Arrighi and C. Patricot, "A note on the correspondence between qubit quantum operations and special relativity," Journal of Physics A: Mathematical and General 36 (2003) L287–L296, for the same dictionary read from the other side: unnormalised qubit states as a Minkowski future cone whose cross-sections are Bloch spheres, pure states as null vectors, unitary operations as rotations about the cone's axis, and positive operations as pure Lorentz boosts, so that the generalized measurement formalism is the restricted Lorentz group together with future-directed null boosts. That is the external counterpart of the dichotomy drawn here. The difference is the sector: their boosts act on the state space, while here a boost is a congruence on $\mathbb{B}_{\Phi}$ and is not a channel on $\mathbb{B}_{\Psi}$.
  • The companion articles of this series: Introduction to the Biquaternion Universe, Four-Vectors and the Interval, States, Observables and the Born Weight, Quantum Physics in Biquaternionic Form, and Entangled Subsystems in the Biquaternion Framework.