---
title: Peripheral Poisson Boundary in Quantum Dynamics
url: https://www.emergentmind.com/topics/peripheral-poisson-boundary
type: topic
---

# Peripheral Poisson Boundary in Quantum Dynamics

Peripheral Poisson boundary is a noncommutative boundary object associated with a normal unital completely positive map on a von Neumann algebra. It is defined from the operator space generated by all peripheral eigenvectors, meaning eigenvectors \(x\) satisfying \(\tau(x)=\lambda x\) for some \(\lambda\in\mathbb T=\{z\in\mathbb C:|z|=1\}\), and it extends the usual noncommutative Poisson boundary, which retains only the fixed-point space \(F(\tau)=\{x\in A:\tau(x)=x\}\). The construction introduced for unital maps equips the norm closure of the peripheral eigenspaces with a \(C^*\)-algebra structure via dilation theory and an extended Choi–Effros product, thereby incorporating all “persistent modes” of the discrete quantum dynamics rather than only the stationary ones [2209.07731].

## 1. Conceptual placement and basic definition

For a countable group \(G\) with probability measure \(\mu\), the classical Poisson boundary is the canonical measure space encoding all bounded \(\mu\)-harmonic functions, and in that setting it is fundamentally measure-theoretic rather than topological [2506.14029]. The peripheral Poisson boundary belongs to a different framework: it is defined for a normal unital completely positive map \(\tau:A\to A\) on a von Neumann algebra \(A\subseteq B(H)\), viewed as a discrete-time quantum Markov map or quantum channel [2209.07731].

The usual noncommutative Poisson boundary starts from the fixed-point space
\[
F(\tau)=\{x\in A:\tau(x)=x\}.
\]
In general, \(F(\tau)\) need not be closed under the original product of \(A\), but Choi–Effros showed that it becomes a von Neumann algebra after changing the product, and Izumi identified this as the noncommutative Poisson boundary. The peripheral refinement replaces the eigenvalue \(1\) by the entire peripheral point spectrum. For \(\lambda\in\mathbb C\),
\[
E_\lambda(\tau)=\{x\in A:\tau(x)=\lambda x\},
\]
and the peripheral point spectrum consists of those \(\lambda\in\mathbb T\) for which \(E_\lambda(\tau)\neq\{0\}\). The operator space generated by all peripheral eigenvectors is
\[
E(\tau)=\operatorname{span}\{x:\ x\in E_\lambda(\tau)\text{ for some }\lambda\in\mathbb T\},
\]
and its norm closure is
\[
P(\tau)=\overline{E(\tau)}^{\|\cdot\|}.
\]
The central theorem is that \(P(\tau)\) admits a natural \(C^*\)-algebra structure after modifying the product. This enlarges the usual Poisson boundary exactly by including all eigenvalues on the unit circle, not only \(1\) [2209.07731].

## 2. Dilation-theoretic construction

The decisive technical tool is the minimal dilation of \(\tau\). There exists a triple \((K,B,\theta)\) in which \(K\) is a Hilbert space containing \(H\), \(B\subseteq B(K)\) is a von Neumann algebra with \(A=pBp\) for the projection \(p\) onto \(H\), and \(\theta:B\to B\) is a normal unital \(^*\)-endomorphism such that
\[
\tau^n(x)=p\,\theta^n(x)\,p,\qquad x\in A,\ n\in\mathbb Z_+.
\]
The dilation is minimal and unique up to the natural unitary equivalence. Two structural properties are repeatedly used:
\[
p\le \theta^n(p)\quad\text{and}\quad \theta^n(p)\uparrow 1\ \text{strongly},
\]
together with the compatibility relation
\[
p\theta(z)p=\tau(pzp),\qquad z\in B.
\]
This moves the analysis from a UCP map to a multiplicative endomorphism, where peripheral spectral data are easier to organize [2209.07731].

Peripheral eigenvectors of \(\tau\) lift canonically to peripheral eigenvectors of \(\theta\). If \(x\in A\) satisfies
\[
\tau(x)=\lambda x,\qquad \lambda\in\mathbb T,
\]
then there exists a unique \(\widehat x\in B\) such that
\[
\theta(\widehat x)=\lambda \widehat x,\qquad p\widehat x p=x,
\]
and moreover
\[
\widehat x=s\!-\!\lim_{n\to\infty}\lambda^{-n}\theta^n(x).
\]
Conversely, if \(\theta(\widehat x)=\lambda\widehat x\), then \(p\widehat x p\in E_\lambda(\tau)\). The restriction \(|\lambda|=1\) is essential: the strong-limit argument uses the boundedness of \(\lambda^{-n}\theta^n(x)\) along the orbit [2209.07731].

A plausible implication is that the peripheral Poisson boundary is not merely a spectral subspace construction. Its defining product is imported from the dilation, where unimodular eigenspaces multiply exactly because \(\theta\) is multiplicative.

## 3. Extended Choi–Effros product and algebraic structure

Inside the dilation, the peripheral part is already multiplicative. If
\[
P(\theta)=\overline{\operatorname{span}\{y\in B:\theta(y)=\lambda y,\ \lambda\in\mathbb T\}},
\]
then \(P(\theta)\) is a \(C^*\)-algebra under the ordinary product, since
\[
\theta(xy)=\lambda\mu\,xy
\]
whenever \(\theta(x)=\lambda x\) and \(\theta(y)=\mu y\). The compression map
\[
T:B\to A,\qquad T(z)=pzp,
\]
restricts to an isometric bijection \(P(\theta)\to P(\tau)\), indeed a complete isometry. Therefore one defines, for \(x,y\in P(\tau)\),
\[
x\circ y:=T\!\left(T^{-1}(x)\,T^{-1}(y)\right),
\]
and \((P(\tau),\circ)\) becomes a unital \(C^*\)-algebra [2209.07731].

The explicit multiplication formula is one of the main structural results. If \(x,y\in A\) satisfy
\[
\tau(x)=\lambda x,\qquad \tau(y)=\mu y,\qquad \lambda,\mu\in\mathbb T,
\]
then
\[
x\circ y=s\!-\!\lim_{n\to\infty}(\lambda\mu)^{-n}\tau^n(xy).
\]
This is the peripheral analogue of the Choi–Effros/Izumi product for fixed points. It says that one multiplies peripheral eigenvectors by iterating the ordinary product under \(\tau\), removing the accumulated phase \((\lambda\mu)^n\), and taking a strong limit. If \(\tau\) preserves a faithful state, then the new product coincides with the original one:
\[
x\circ y=xy.
\]
In that case, the peripheral boundary embeds as an actual subalgebra of \(A\) [2209.07731].

Several formal consequences follow. If \(x\in E_\lambda(\tau)\) and \(y\in E_\mu(\tau)\), then
\[
x\circ y\in E_{\lambda\mu}(\tau).
\]
If \(A\) is abelian, then \(P(\tau)\) is abelian. For any subgroup \(G\subseteq\mathbb T\),
\[
E_G(\tau)=\operatorname{span}\{E_\lambda(\tau):\lambda\in G\},
\]
and the associated \(C^*\)-algebra \(P_G(\tau)\) includes the \(k\)-cyclic Poisson boundary obtained from the subgroup of \(k\)-th roots of unity. Each eigenspace \(E_\lambda(\tau)\) becomes a two-sided Hilbert \(C^*\)-module over the fixed-point algebra \((F(\tau),\circ)\) [2209.07731].

## 4. Dynamical invariance and spectral rigidity

The peripheral Poisson boundary is stable under discrete-time iteration. For every \(k\ge 1\),
\[
P(\tau^k)=P(\tau).
\]
This invariance is stronger than equality of individual eigenspaces: the eigenvalues of \(\tau\) and \(\tau^k\) may differ, but the full peripheral boundary remains unchanged. Moreover,
\[
P_{\{1\}}(\tau^k)=P_{G_k}(\tau),
\]
where
\[
G_k=\{1,\omega,\omega^2,\dots,\omega^{k-1}\},\qquad \omega=e^{2\pi i/k}.
\]
This yields a discrete Fourier-type decomposition of the fixed-point space of \(\tau^k\) into peripheral eigenspaces of \(\tau\). If the group generated by the peripheral spectrum is finite, then the peripheral boundary coincides with a \(k\)-cyclic Poisson boundary and is therefore a von Neumann algebra [2209.07731].

The action of the dynamics on the boundary is reversible:
\[
\tau|_{P(\tau)}
\]
is an automorphism of the peripheral Poisson boundary. The interpretation given in the original formulation is that the peripheral spectrum records the part of the dynamics that survives on the unit circle, whereas eigenvalues with modulus \(<1\) correspond to decaying behavior. This suggests that the peripheral Poisson boundary isolates the non-decaying oscillatory or recurrent component of the evolution [2209.07731].

For Markov operators arising from symmetric, generating probability measures on countable discrete groups, the peripheral spectrum becomes especially rigid. If
\[
P_\mu(T)=\sum_{g\in G}\mu(g)\,\lambda_g T\lambda_g^*
\]
on \(B(\ell^2(G))\), then every peripheral eigenvalue is a root of unity when the support of \(\mu\) generates \(G\) as a semigroup, and in the symmetric case the only peripheral eigenvalues are
\[
\pm 1.
\]
In this group setting, the peripheral boundary algebra is actually a von Neumann algebra [2307.11295].

## 5. Group-theoretic form and jointly bi-harmonic functions

The group case ties peripheral boundary theory to classical harmonic analysis on \(G\). If \(\lambda\) is a peripheral eigenvalue of \(P_\mu\) on \(B(\ell^2(G))\), then there exists a nonzero \(f\in\ell^\infty(G)\) such that
\[
f*\mu=\lambda f.
\]
In particular, when \(\mu\) is symmetric,
\[
-1 \text{ is an eigenvalue } \iff \text{there exists a nonzero anti-harmonic function.}
\]
If \(e\in\operatorname{supp}(\mu)\) and \(P_\mu(T)=\lambda T\) for some nonzero \(T\), then \(\lambda=1\). Applying this to \(P_{\mu^{*2}}=P_\mu^2\) yields the symmetric-case restriction \(\lambda=\pm1\) [2307.11295].

These spectral facts lead to a characterization of jointly bi-harmonic functions. For a symmetric, generating probability measure \(\mu\) on a countable discrete group \(G\), a bounded function satisfying
\[
\mu*f*\mu=f
\]
need not be constant, but there exists \(c\in\mathbb C\) such that
\[
f-c
\]
is separately anti-harmonic under both left and right convolution by \(\mu\). This answers a question of Kaimanovich in the form recorded by the paper [2307.11295].

Anti-harmonic functions themselves are rigid. If there exists a nonzero \(f\in\ell^\infty(G)\) such that
\[
f*\mu=-f,
\]
then there exists a multiplicative character
\[
\chi:G\to\mathbb T
\]
such that
\[
\chi|_{\operatorname{supp}(\mu)}=-1,
\]
hence
\[
\chi*\mu=-\chi.
\]
Moreover, every anti-harmonic function factors as
\[
F=h\chi
\]
for some harmonic \(h\in\operatorname{Har}(\mu)\). In this sense, the \(-1\)-part of the peripheral spectrum is governed by characters and by harmonic functions twisted by those characters [2307.11295].

## 6. Full Fock space model

A concrete and highly structured example is provided by the full Fock space over a separable Hilbert space \(H\). Writing
\[
\mathcal F(H)=\mathbb C\Omega\oplus \bigoplus_{m\ge1} H^{\otimes m},
\]
with orthonormal basis \(\{e_i:i\in\mathcal O\}\), the left and right creation operators are defined by
\[
l_i(\xi)=e_i\otimes\xi,\qquad r_i(\xi)=\xi\otimes e_i,
\]
and satisfy identities including
\[
r_i^*r_j=l_i^*l_j=\delta_{ij}I,\qquad r_i^*l_j=l_jr_i^*=\delta_{ij}p_\Omega.
\]
For positive weights \(w_i>0\) with \(\sum_i w_i=1\), the normal unital completely positive map
\[
P_w(x)=\sum_{i\in\mathcal O} w_i\,l_i^*xl_i
\]
acts on \(B(\mathcal F(H))\) [2406.11167].

The corresponding peripheral Poisson boundary is independent, up to \(^*\)-isomorphism, of the choice of orthonormal basis of \(H\). It strictly contains the usual Poisson boundary:
\[
\mathcal F(P_w)\subsetneq P(P_w).
\]
For \(\lambda\in\mathbb T\setminus\{1\}\), the diagonal phase operator
\[
x_\lambda=\sum_{I\in\mathcal A}\lambda^{|I|}\,|e_I\rangle\langle e_I|
\]
satisfies
\[
P_w(x_\lambda)=\lambda x_\lambda,
\]
so \(x_\lambda\) belongs to the peripheral boundary but not to the fixed-point algebra [2406.11167].

The product is concrete on words and creation operators. If \(x\in P(P_w)\) and \(I,J\in\mathcal A\), then
\[
x\circ r_I=xr_I,\qquad r_I\circ x=r_Ix,\qquad r_I\circ x\circ r_J=r_Ixr_J.
\]
For \(i\in\mathcal O\),
\[
r_i\circ x=r_ix+w_i\,p_\Omega x\,l_i.
\]
If \(x\in E_\lambda(P_w)\) with \(\lambda\in\mathbb T\setminus\{1\}\), then
\[
E_\lambda(P_w)=\{ax_\lambda:a\in \mathcal F(P_w)\}=\{x_\lambda a:a\in \mathcal F(P_w)\}.
\]
The ordinary Poisson boundary has trivial relative commutant inside the peripheral boundary,
\[
\mathcal F(P_w)'\cap P(P_w)=\mathbb C I,
\]
and therefore the center is trivial:
\[
Z(P(P_w))=\mathbb C I.
\]
There is also a conditional expectation
\[
E:P(P_w)\to \mathcal F(P_w),
\]
which yields an \(\mathcal F(P_w)\)-valued inner product
\[
\langle x,y\rangle = E(x\circ y^*),
\]
making \(P(P_w)\) a pre-Hilbert \(C^*\)-bimodule over the Poisson boundary \(\mathcal F(P_w)\) [2406.11167].

## 7. Extensions to contractive maps and semigroups, and limits of the theory

The original construction has been extended from normal UCP maps to normal contractive completely positive maps
\[
\tau:\mathcal A\to\mathcal A,\qquad \tau(1)\le 1,
\]
on von Neumann algebras, and further to unital and non-unital contractive quantum dynamical semigroups. The extension proceeds through the unitization
\[
\widetilde{\mathcal A}=\mathcal A\oplus\mathbb C,
\]
with unitized map
\[
\widetilde\tau(x\oplus c)=\big(\tau(x)+c(1-\tau(1))\big)\oplus c.
\]
For a quantum dynamical semigroup \(\tau=\{\tau_t:t\in\mathbb R_+\}\), the peripheral eigenspaces are indexed by \(a\in\mathbb R\):
\[
E_a(\tau)=\{x\in\mathcal A:\tau_t(x)=e^{iat}x\ \forall t\ge0\}.
\]
The same dilation mechanism then produces a peripheral boundary \((\mathcal P(\tau),\circ)\) in continuous time [2507.20668].

The strong-operator limit formula remains intact. In discrete time, if \(x\in E_\lambda(\tau)\) and \(y\in E_\mu(\tau)\), then
\[
x\circ y=\mathrm{s.}\lim_{n\to\infty}\frac{\tau^n(xy)}{\lambda^n\mu^n}.
\]
In continuous time, if \(x\in E_a(\tau)\) and \(y\in E_b(\tau)\), then
\[
x\circ y=\mathrm{s.}\lim_{t\to\infty}\frac{\tau_t(xy)}{e^{iat}e^{ibt}}.
\]
In both cases one also has
\[
x\circ y=p(\widehat x\,\widehat y)p
\]
in the dilation picture [2507.20668].

A central structural theorem states that for a normal contractive CP map,
\[
q_\tau=\mathrm{s-}\lim_{n\to\infty}\tau^n(1)
\]
satisfies the equivalences:
\(q_\tau\neq 0\), \(\tau\) has a nontrivial fixed point, \(\mathcal P(\tau)\) is nontrivial, and \(\mathcal P(\tau)\) is a unital \(C^*\)-algebra. In that case, \(q_\tau\) is the unit of \(\mathcal P(\tau)\) and the unique maximal Perron–Frobenius eigenvector of \(\tau\). The continuous-time analogue uses
\[
\mathrm{s-}\lim_{t\to\infty}\tau_t(1).
\]
Thus, even for non-unital contractive semigroups, the peripheral Poisson boundary is unital whenever it is nontrivial [2507.20668].

The theory does not extend naively to arbitrary \(C^*\)-algebra settings. The stated obstacles are that peripheral eigenvectors may lift in the von Neumann dilation but not in a smaller \(C^*\)-dilation, the extended Choi–Effros product relies on strong operator limits, and the closed span of peripheral eigenvectors may fail to admit any \(C^*\)-algebra structure compatible with the given operator space structure. The paper provides counterexamples showing that contractivity is essential and that the boundary construction is fundamentally von Neumann algebraic in character [2507.20668].

Source: https://www.emergentmind.com/topics/peripheral-poisson-boundary