Positive Recurrent Subspace in Quantum and Markov Processes

• Positive recurrent subspace is a key construct where all states exhibit recurrence with finite expected return times and well-defined invariant supports.
• In quantum dynamics and Markov semigroups, it is characterized via projective measurement protocols and rational spectral functions that ensure state stability.
• This concept extends to operator theory and stochastic processes, linking recurrence with spectral properties, stability analyses, and improved numerical conditioning.

Positive recurrent subspace is not a single universal construction across the cited literature; rather, it denotes or borders several field-specific notions that isolate the recurrent part of a dynamics by support, return-time finiteness, or spectral structure. In discrete-time quantum dynamics, a finite-dimensional subspace $V$ is positive recurrent when all states in $V$ are recurrent with finite expected return time (Bourgain et al., 2013). In weak-coupling-limit quantum Markov semigroups, the adjacent object is the fast recurrent subspace, defined as the largest support among all invariant states, together with the hereditary subalgebra on which faithful invariant states exist (Bolaños-Servín et al., 2023). Related recurrence-subspace formalisms also appear for linear operators on Banach spaces, for jumping-in diffusions with large jumps, for semimartingale reflecting Brownian motion, and in matrix-analytic treatments of null recurrent and positive recurrent regimes in nonsymmetric algebraic Riccati equations (López-Martínez, 2022, Yamato, 2022, Bramson, 2010, Iannazzo et al., 2010).

1. Domain-specific meanings and recurrent structures

The cited works use recurrence-subspace language in several technically distinct ways. In each case, the relevant object is a distinguished subspace or sector on which the long-time dynamics exhibits recurrence, invariant-state support, or finite expected return structure (Bourgain et al., 2013, Bolaños-Servín et al., 2023, López-Martínez, 2022, Yamato, 2022).

Domain	Object	Defining criterion or characterization
Discrete-time quantum dynamics	Finite-dimensional subspace $V$	All states in $V$ are recurrent with finite expected return time
Weak coupling limit QMS	Fast recurrent subspace $\mathcal{R}_c$	Largest support among all invariant states
Banach-space linear dynamics	Recurrent subspace $Z\subset X$	Infinite-dimensional closed subspace with $Z\subset \operatorname{Rec}(T)$
Jumping-in diffusions	Positive recurrent subspace	Set of jumping-in diffusions for which $\frac{1}{t}\eta(t)\xrightarrow[t\to\infty]{P} b$

In this comparison, the quantum and stochastic-process settings are the ones in which positive recurrence is attached most directly to invariant states or finite return times. The Banach-space literature uses the exact term recurrent subspace rather than positive recurrent subspace, but it supplies a precise infinite-dimensional subspace formalism. The SRBM and NARE literatures do not define a positive recurrent subspace as such; instead, they analyze positive recurrent regimes and invariant subspaces that control stability, recurrence, or numerical conditioning (Bramson, 2010, Iannazzo et al., 2010).

2. Finite-dimensional quantum subspaces with finite expected return time

For a discrete-time quantum process with unitary evolution operator $U$ on a Hilbert space $\mathcal{H}$ and a finite-dimensional subspace $V$0, subspace recurrence is defined by a monitored protocol: after each application of $V$1, a projective measurement asks whether the system is in $V$2. With $V$3 the orthogonal projector onto $V$4 and $V$5, the first return probability that a normalized state $V$6 returns to $V$7 at the $V$8-th step is $V$9, where

$V$0

The total first return probability is

$V$1

A subspace $V$2 is recurrent if every unit vector $V$3 is $V$4-recurrent (Bourgain et al., 2013).

Positive recurrence is the stronger condition that all states in $V$5 are recurrent with finite expected return time. The paper gives several equivalent characterizations: the spectral measure $V$6 is a sum of finitely many mass points; the operator-valued Schur function $V$7 is rational inner; $V$8 is rational inner; and $V$9 is contained in a finite sum of eigenspaces of $V$0. The generating function of first return amplitudes,

$V$1

encodes the return problem directly, and the expected return time of a $V$2-recurrent state is

$V$3

When $V$4 is recurrent and $V$5 is rational inner, $V$6 is identified with minus the Aharonov-Anandan geometric phase along the loop $V$7, and the averaged expected return time is always a rational number of the form $V$8, where $V$9 is a positive integer (Bourgain et al., 2013).

The same work also records a nonclassical feature: state recurrence can occasionally give higher return probabilities than subspace recurrence. Thus return probabilities are not monotonic with respect to enlarging the target subspace. This distinguishes quantum monitored recurrence from classical return theory and makes the positive recurrent subspace sensitive to interference and to the measurement protocol rather than only to set inclusion (Bourgain et al., 2013).

3. Fast recurrent support and positive recurrence in quantum Markov semigroups

In the theory of Quantum Markov Semigroups generated by a Gorini-Kossakowski-Sudarshan-Lindblad operator of weak coupling limit type, the fast recurrent subspace is defined as

$\mathcal{R}_c$0

It is the largest support among all invariant states and is also called the fast recurrent projection in the terminology cited by the paper; its range is the fast recurrent subspace (Bolaños-Servín et al., 2023).

For the $\mathcal{R}_c$1-level quantum transport model studied there, the main explicit result is

$\mathcal{R}_c$2

The system consists of $\mathcal{R}_c$3 levels, with level-to-level transition operators $\mathcal{R}_c$4 that are scalar multiples of DFT-type operators. Invariant states are characterized completely: any invariant state can be written as a convex combination

$\mathcal{R}_c$5

where $\mathcal{R}_c$6 is supported on $\mathcal{R}_c$7, $\mathcal{R}_c$8 is supported on the interaction-free subspace $\mathcal{R}_c$9, $Z\subset X$0 is the pure state at the top level, and $Z\subset X$1 with $Z\subset X$2. Consequently, every invariant state is supported in $Z\subset X$3, and no invariant state has support outside this space (Bolaños-Servín et al., 2023).

The relation to positive recurrence is explicit at the hereditary-subalgebra level. The paper states that in the QMS restricted to the hereditary subalgebra

$Z\subset X$4

there exists a faithful invariant state. This hereditary subalgebra is identified as the natural setting for quantum positive recurrence in this context. On that subalgebra, for any initial state $Z\subset X$5, the long-term limit $Z\subset X$6 exists and is an invariant state, and the domains of attraction are made explicit in terms of projections onto $Z\subset X$7 for a chosen subspace $Z\subset X$8 (Bolaños-Servín et al., 2023).

A central structural point is that the generalized DFT operators govern both support and spectrum. The paper states that the structure of invariant states and their spectra is determined in terms of a natural generalization of the Discrete Fourier Transform operator, so the recurrent support is not merely geometric but also spectrally organized by iterated transport through the $Z\subset X$9 and $Z\subset \operatorname{Rec}(T)$0 operators (Bolaños-Servín et al., 2023).

4. Recurrent subspaces in Banach-space linear dynamics

For an operator $Z\subset \operatorname{Rec}(T)$1 on a Banach space, a recurrent subspace is defined as an infinite-dimensional closed subspace $Z\subset \operatorname{Rec}(T)$2 such that $Z\subset \operatorname{Rec}(T)$3, where $Z\subset \operatorname{Rec}(T)$4 is the set of recurrent vectors $Z\subset \operatorname{Rec}(T)$5 for which there exists an increasing sequence of integers $Z\subset \operatorname{Rec}(T)$6 such that $Z\subset \operatorname{Rec}(T)$7. An operator is recurrent if $Z\subset \operatorname{Rec}(T)$8 is dense in $Z\subset \operatorname{Rec}(T)$9 (López-Martínez, 2022).

A main sufficient criterion is formulated through quasi-rigidity. If $\frac{1}{t}\eta(t)\xrightarrow[t\to\infty]{P} b$0 is quasi-rigid with respect to an increasing sequence $\frac{1}{t}\eta(t)\xrightarrow[t\to\infty]{P} b$1, and if there exists a non-increasing sequence $\frac{1}{t}\eta(t)\xrightarrow[t\to\infty]{P} b$2 of infinite-dimensional closed subspaces of $\frac{1}{t}\eta(t)\xrightarrow[t\to\infty]{P} b$3 such that

$\frac{1}{t}\eta(t)\xrightarrow[t\to\infty]{P} b$4

then $\frac{1}{t}\eta(t)\xrightarrow[t\to\infty]{P} b$5 has a recurrent subspace: there exists an infinite-dimensional closed subspace $\frac{1}{t}\eta(t)\xrightarrow[t\to\infty]{P} b$6 and a subsequence $\frac{1}{t}\eta(t)\xrightarrow[t\to\infty]{P} b$7 of $\frac{1}{t}\eta(t)\xrightarrow[t\to\infty]{P} b$8 so that $\frac{1}{t}\eta(t)\xrightarrow[t\to\infty]{P} b$9 for all $U$0. The paper also states equivalent forms in terms of boundedness or convergence of $U$1 on an infinite-dimensional closed subspace (López-Martínez, 2022).

In the complex case, if $U$2 is quasi-rigid, having a recurrent subspace is equivalent to the essential spectrum intersecting the closed unit disk: $U$3 The corresponding real-case statement uses the complexification $U$4: $U$5 The same paper further states that a weakly-mixing operator on a real or complex separable Banach space has a hypercyclic subspace if and only if it has a recurrent subspace (López-Martínez, 2022).

This body of results does not define positive recurrent subspace in the probabilistic sense of finite expected return time. Instead, it supplies a spectral and spaceability framework for infinite-dimensional subspaces made entirely of recurrent vectors. A plausible implication is that, in operator theory, recurrence-subspace questions are organized less by invariant measures than by essential-spectrum placement and by rigidity properties of powers of the operator.

5. Positive recurrent sectors in stochastic processes

For unilateral or bilateral jumping-in diffusions, positive recurrence is identified by the asymptotic behavior of the inverse local time at $U$6, denoted $U$7. The positive recurrent subspace is described as the set of jumping-in diffusions specified by a speed measure $U$8 and a jumping-in measure $U$9 for which

$\mathcal{H}$0

In the large-jump regime, this corresponds to

$\mathcal{H}$1

The paper establishes fluctuation scaling limits for inverse local times and occupation times, and develops a hierarchy of modified Neumann boundary conditions of order $\mathcal{H}$2 through quantities $\mathcal{H}$3 and the index $\mathcal{H}$4. It states that if $\mathcal{H}$5, then the process is positive recurrent in a generalized sense, whereas when $\mathcal{H}$6, the process is not positive recurrent (Yamato, 2022).

The same work characterizes the Laplace exponent by

$\mathcal{H}$7

with $\mathcal{H}$8 expressed through the generalized eigenfunction $\mathcal{H}$9 and an explicit coefficient $V$00. The fluctuation result

$V$01

places positive recurrent sectors within a scaling-limit theory in which recurrence is tied to boundary singularity order and to continuity of Laplace exponents under scaling (Yamato, 2022).

A different stochastic-process perspective is provided by semimartingale reflecting Brownian motions in the nonnegative orthant. There, positive recurrence means that the expected time to hit any open neighborhood of the origin is finite, for every starting state. The standard fluid-path sufficiency theorem states: if every fluid path associated with $V$02 is attracted to the origin, then the SRBM is positive recurrent. However, the converse fails in dimension $V$03: the cited paper constructs a family of examples in $V$04 with $V$05, $V$06, and appropriate $V$07, that are positive recurrent even though a linear fluid path diverges to infinity (Bramson, 2010).

This counterexample is a central caution against identifying positive recurrence with deterministic fluid stability. In the paper’s construction, a divergent linear fluid path coexists with positive recurrence of the stochastic system. This suggests that, in high-dimensional reflected diffusions, a “positive recurrent sector” cannot always be read off from the fluid model alone (Bramson, 2010).

6. Invariant-subspace methods near null and positive recurrence in algebraic Riccati equations

In the matrix-analytic setting of nonsymmetric algebraic Riccati equations associated with an M-matrix,

$V$08

recurrence terminology enters through the Markov-chain interpretation of the linearizing matrix

$V$09

When the two critical eigenvalues satisfy $V$10, the problem is called critical or null recurrent. Close-to-critical cases arise when these eigenvalues are real and close to zero but not exactly zero. In positive recurrent cases, corresponding to the nonsingular M-matrix case, these problematic eigenvalues are strictly separated from zero, and standard techniques work well (Iannazzo et al., 2010).

The paper introduces a subspace shift technique that acts on the invariant subspace associated with the problematic eigenvalues as a whole. If $V$11 is the central invariant subspace associated with the $V$12 eigenvalues closest to the imaginary axis, and $V$13 is the corresponding left invariant subspace, the shifted matrix is

$V$14

with $V$15. This shifts every eigenvalue associated with $V$16 by scaling it by $V$17, while the minimal nonnegative solution of the shifted equation is the same as for the original NARE (Iannazzo et al., 2010).

The recurrence-theoretic significance is indirect but clear. The difficult regime is the neighborhood of null recurrence, where small eigenvalues near the imaginary axis create ill-conditioning and slow convergence. The subspace shift improves separation and conditioning by moving the entire central invariant subspace away from the critical region. In positive recurrent cases, by contrast, the paper states that the problematic eigenvalues are already strictly separated from zero. Thus invariant-subspace manipulations provide a computational analogue of recurrence classification, even though the object under study is not itself called a positive recurrent subspace (Iannazzo et al., 2010).