Limits of Stochastic Semigroups and Block-Triangular Majorisation

arXiv:2609.04057 · math-ph, math.CO, math.MP, math.PR, quant-ph · Submitted 2026-09-03 · Read on arXiv

Listen

Radio episode about this paper

Transcript

Introduction to the show: ident: Mathematical Physics. The day in mathematical physics: rigorous models of physical systems.

Kai: I'm Kai.

Mira: And I'm Mira, here with Lev.

Kai: We're looking at the paper: "Limits of Stochastic Semigroups and Block-Triangular Majorisation".

Mira: Here's the gist of the paper. Given probability vectors gamma(beta) depending on a parameter beta, we introduce a notion of convergence as beta → ∞ for the corresponding semigroups of gamma(beta)-preserving stochastic matrices and investigate the structure of the resulting limit In general, the limiting semigroup differs from the semigroup preserving the limiting distribution, showing that these two operations do not commute.

Kai: The Non-Commutativity of Limits The paper investigates whether taking the zero-temperature limit commutes with passing from a Gibbs state to the semigroup of stochastic matrices that preserve its limit.

Mira: The authors show that this operation is generally non-commuting, meaning different families of invariant distributions may converge to the same limiting distribution while giving rise to different limiting semigroups.

Kai: The Nature of the Limiting Semigroup When the invariant distributions are Gibbs vectors at an inverse temperature beta, the limiting semigroup consists of block-upper-triangular stochastic matrices subject to additional substochasticity constraints.

Paper summary: Kai: So we've been talking about how these limits of probability vectors affect stochastic matrices, and now we’re going to wrap up what this paper is all about, "Limits of Stochastic Semigroups and Block-Triangular Majorisation."

Mira: It boils down to the fact that taking the zero-temperature limit of a system doesn't just give you a final distribution; it gives you a specific type of matrix structure.

Lev: And this structure isn't always what you expect from just looking at the limit of the distributions themselves. It involves these block-upper-triangular matrices with those extra rules we call substochasticity constraints.

Lev: We need to make sure that when we build things, whether we use standard majorisation or this block-triangular one, our error budgets match up correctly for the real hardware constraints.

Conclusion: ---: Conclusion ---

Kai: So we’re wrapping up on this paper, "Limits of Stochastic Semigroups and Block-Triangular Majorisation." Basically, they’re looking at how probability vectors change as you increase a parameter, that beta, and what kind of stochastic matrices you end up with in the limit.

Mira: The big takeaway is that when you try to take the zero-temperature limit of these systems—the Gibbs states—you get a very specific matrix structure. It’s not just any stochastic matrix; it’s one that has these block-upper-triangular constraints they call substochasticity constraints.

Lev: That means if we're designing hardware to model a physical system cooling down, we can't just look at the final distribution and assume the evolution of the system is simple. We have to account for these specific block structures and those extra rules when analyzing the long-term behavior of our operations.

Kai: Right, so this paper shows that taking these limits—looking at distributions versus looking at matrix semigroups—they don't always give you the same result. That non-commutativity is key here.

Mira: It’s crucial because it means that two different starting distributions can lead to different limiting matrix evolutions, even if both of them settle down to the exact same final state distribution in the long run. That's a big distinction for understanding how these systems behave over time.

Lev: From an engineering standpoint, that tells us if we build hardware based on one mathematical approach and then use another approach for our analysis, we’re going to be dealing with fundamentally different constraints on our final system evolution.

Kai: Exactly. It shows that just having the limiting distribution isn't enough to tell you how the whole semigroup behaves in the long run.

Mira: And this Block-Triangular majorisation concept is what connects those two ways of looking at convergence, showing how they actually relate to each other.

Lev: We need to make sure that when we build things, whether we use standard majorisation or this block-triangular one, our error budgets match up correctly for the real hardware constraints.

FABIO DEELAN CUNDEN, JAKUB CZARTOWSKI, GIOVANNI GRAMEGNA, MARILENA LIGABO

Gruppo Nazionale di Fisica Matematica INdAM · Istituto Nazionale di Fisica Nucleare INFN · PNRR MUR project PE0000023-NQSTI · PNRR MUR project CN0000013 'Italian National Centre on HPC, Big Data and Quantum Computing', 'PRIN 2022 project 2TBE52WPE1- 'The charm of integrability: from nonlinear waves to random matrices', the project “Patto territoriale sistema universitario pugliese · Nanyang Technological University · Research Ireland

math-ph, math.CO, math.MP, math.PR, quant-ph

Submitted: 2026-09-03

Updated: 2026-11-09

Comments: 19 pages, 1 figure, Major revision with presentation improved and typos fixed; Comments welcome!

Code: https://github.com/jack-point13/deg_ext_gp_stoch

License: http://creativecommons.org/licenses/by/4.0/

Importance score: 87/100

The gist: Given probability vectors gamma(beta) depending on a parameter beta, we introduce a notion of convergence as beta → ∞ for the corresponding semigroups of gamma(beta)-preserving stochastic…

Key concepts

Semigroup Convergence
This refers to how a family of stochastic matrices associated with $\gamma(\beta)$ behaves as the parameter $\beta$ becomes very large. The study examines the resulting limiting semigroup, which describes the long-term behavior of these transformations. It shows that this limit is not simply determined by the limiting probability distribution alone.
Block-Triangular Majorisation
This is a structural property characterizing the extreme points of the limiting semigroups. It describes how standard majorisation (when $\mu = (n)$) smoothly transitions into upper triangular or unordered majorisation (when $\mu = (1,..., 1)$). This structure depends on the multiscale profile of $\gamma(\beta).
BT-stochastic Matrices
These are the specific matrices that form the zero-temperature limit of the semigroups. They are row-stochastic block-upper-triangular matrices where each diagonal block has column substractivity constraints. The number of extreme points for these matrices is determined by a formula involving Laguerre polynomials.
Non-Commutativity of Limits
The paper demonstrates that the operation of taking the zero-temperature limit and the operation of finding the semigroup preserving the limiting distribution do not commute. This means different invariant distributions can converge to the same limit, but they lead to different resulting semigroups.

Terminology

Summary

Given probability vectors gamma(beta) depending on a parameter beta, we introduce a notion of convergence as beta → ∞ for the corresponding semigroups of gamma(beta)-preserving stochastic matrices and investigate the structure of the resulting limit In general, the limiting semigroup differs from the semigroup preserving the limiting distribution, showing that these two operations do not commute. p. 0

Where the paper says it

Given probability vectors 𝛾(𝛽) depending on a parameter 𝛽, we introduce a notion of convergence as 𝛽 → ∞ for the corresponding semigroups of 𝛾(𝛽)-preserving stochastic matrices and investigate the structure of the resulting limit. In general, the limiting semigroup differs from the semigroup preserving the limiting distribution, showing that these two operations do not commute.

The Non-Commutativity of Limits

The paper investigates whether taking the zero-temperature limit commutes with passing from a Gibbs state to the semigroup of stochastic matrices that preserve its limit. The authors show that this operation is generally non-commuting, meaning different families of invariant distributions may converge to the same limiting distribution while giving rise to different limiting semigroups. p. 1

Where the paper says it

In this paper we show that the answer is negative: taking the limit of invariant distributions and taking the semigroup of stochastic matrices preserving them are, in general, non-commuting operations. Different families of invariant (not necessarily Gibbs) distribution 𝛾(𝛽) may converge to the same limiting distribution while giving rise to different limiting semigroups.

The Nature of the Limiting Semigroup

When the invariant distributions are Gibbs vectors at an inverse temperature beta, the limiting semigroup consists of block-upper-triangular stochastic matrices subject to additional substochasticity constraints. This structure is characterized by a notion called Block-Triangular majorisation, which interpolates between ordinary majorisation and upper triangular (aka unordered) majorisation. p. 0

Where the paper says it

We characterise and enumerate their extremal elements and determine the preorder on probability vectors induced by the action of the semigroup. The resulting notion of Block-Triangular majorisation interpolates between ordinary majorisation and upper triangular (aka unordered) majorisation.

Multiscale Structure and Degeneracy

The convergence of the semigroups depends on the multiscale profile of gamma(beta) as beta → ∞. Specifically, for Gibbs vectors, the limiting semigroup only depends on the degeneracy structure of gamma(beta), resulting in a set of block-upper-triangular matrices dubbed BT-stochastic matrices. p. 3

Where the paper says it

1 to the 𝛽 → ∞ limit of S 𝛾(𝛽) , when 𝛾(𝛽) is a Gibbs vector with arbitrary multiplicities of the energies. In this case the limit semigroup only depends on the degeneracy structure of 𝛾(𝛽), and it is a set of block-upper-triangular matrices that we dub BT-stochastic matrices (‘BT’ for ‘block triangular’).

Extreme Operations and Majorisation

The extreme points of the limiting semigroups, denoted ext(Tmu), are characterized by specific conditions related to row and column constraints on the diagonal blocks. The BT-majorisation relation interpolates between standard majorisation (when mu = (n)) and upper triangular or unordered majorisation (when mu = (1,. p. 7

Where the paper says it

, 𝑛, ∑︁ ℓ 𝑖=1 𝑥 𝜇,↓ 𝑖 ≤ ∑︁ ℓ 𝑖=1 𝑦 𝜇,↓ 𝑖 , with equality for ℓ = 𝑛. BT-majorisation interpolates between standard majorisation (when 𝜇 = (𝑛)) and upper triangular (UT) or unordered majorisation (when 𝜇 = (1, .

p. 7
Where the paper says it

, 𝑛, ∑︁ ℓ 𝑖=1 𝑥 𝜇,↓ 𝑖 ≤ ∑︁ ℓ 𝑖=1 𝑦 𝜇,↓ 𝑖 , with equality for ℓ = 𝑛. BT-majorisation interpolates between standard majorisation (when 𝜇 = (𝑛)) and upper triangular (UT) or unordered majorisation (when 𝜇 = (1, .

.., 1)).

Finite-Temperature Corrections

The zero-temperature limit captures the leading-order behavior of Gibbs-preserving transformations at low temperatures in a way that is analogous to how permutations capture leading-order behavior at high temperatures. Finite-temperature corrections can be described by asymptotic expansions in powers of e−beta. p. 3

Where the paper says it

4), and we study the zero-temperature limit of the Renyi entropic inequalities ` (Proposition 3. The limiting semigroups and their extreme points are illustrated in Section 4 for 𝑛 = 3 and 𝑛 = 4, followed by a discussion of finite-temperature corrections.

These corrections reveal how BT-majorisation deforms into thermomajorisation away from the limiting regimes. p. 0
Where the paper says it

Among these, the resource theory of athermality exhibits the richest majorisation structure, described by thermomajorisation [GMN+15, dOJCZK22 ˙ ]. While much recent work has focused on coherent systems, especially in asymptotic [ST25] and catalytic settings [SN24], the energy-incoherent regime continues to reveal nontrivial structural aspects of finitesize thermodynamics.

Gibbs Critical Temperatures

In the Gibbs case, critical inverse temperatures beta∗ are determined by conditions derived from the multiscale structure of the invariant distribution. These critical points suggest a constructive procedure for determining the finite-temperature extremal structure by iterating through temperature intervals. p. 0

Where the paper says it

The convex geometry of this set, its extreme points and its relation to other notions such as transportation polytopes has been extensively studied in the literature [Har74,GKP92, vD22,CCGdOJ25]. A natural question is: what becomes of the finite-temperature semigroupsS 𝛾(𝛽) in the zero-temperature limit, 𝛽 → ∞?

The zero-temperature limit yields the BT-stochastic matrices Tmu, which are proper subsemigroups of Sgamma(∞). p. 6
Where the paper says it

Indeed, Theorem 3. 2 claims that the zero-temperature limit of S 𝛾(𝛽) is the set of row-stochastic upper triangular matrices T(1,1,1) , a proper subsemigroup of S 𝛾(∞) .

Proof Methodology

The proof of Theorem 3.1 relies on showing two inclusions: LsSgamma(beta) ⊆ Tq1,. p. 0

Where the paper says it

Given probability vectors 𝛾(𝛽) depending on a parameter 𝛽, we introduce a notion of convergence as 𝛽 → ∞ for the corresponding semigroups of 𝛾(𝛽)-preserving stochastic matrices and investigate the structure of the resulting limit. In general, the limiting semigroup differs from the semigroup preserving the limiting distribution, showing that these two operations do not commute.

.., Tqk and Tq1,..., Tqk ⊆ LiSgamma(beta). The latter involves constructing a family of matrices P(beta) that converges to the target matrix P. p. 1
Where the paper says it

We denote by S the semigroup of 𝑛×𝑛 stochastic matrices and by S 𝑞 the subsemigroup of 𝑞-preserving stochastic matrices. Both the identity matrix and the rank-one matrix e 𝑇 𝑞 (all rows equal to 𝑞) belong to S 𝑞 , so the latter is nonempty.

p. 1
Where the paper says it

We denote by S the semigroup of 𝑛×𝑛 stochastic matrices and by S 𝑞 the subsemigroup of 𝑞-preserving stochastic matrices. Both the identity matrix and the rank-one matrix e 𝑇 𝑞 (all rows equal to 𝑞) belong to S 𝑞 , so the latter is nonempty.

Extreme Point Enumeration

The number of extreme points of Tmu is given by a formula involving Laguerre polynomials, which corresponds to the Laguerre representation. This count is computed by considering the number of ways to place ones within each block structure. p. 6

Where the paper says it

, 1), the limiting semigroup T𝜇 is the set of upper triangular row-stochastic matrices discovered by Narasimhachar and Gour [GMN+15]. The difference between T𝜇 and S 𝛾(∞) can be understood in terms of allowed probability flows between energy levels: S 𝛾(∞) inhibits only transitions from the “lowest” or “ground” energy level to the “excited” ones, but still allows transitions from lower to upper energy levels among the excited ones, while this is not possible within T𝜇, which is reminiscent of the level hierarchy in the low-temperature Gibbs weights.

Majorisation Collapse

In the zero-temperature limit, entropic inequalities collapse to much simpler conditions, showing that the mean energy is the primary monotone in the free energy. When U(x) = U(y), entropy provides a subleading order within a fixed energy sector. p. 8

Where the paper says it

, ∥𝑥 𝑘 ∥1 , where ∥𝑥 𝑎 ∥1 = Í𝑚𝑎 𝑗=1 𝑥 𝑎 𝑗 and ∥𝑦 𝑎 ∥1 = Í𝑚𝑎 𝑗=1 𝑦 𝑎 𝑗 , for all 𝑎 = 1, . Thus, in the low-temperature limit, the mean energy is the primary monotone in the free energy, while entropy provides the subleading order within a fixed energy sector.

The alpha-entropic inequalities collapse to a reverse lexicographic condition on the block sums when 1 − alpha > 0, and a lexicographic condition when 1 − alpha < 0. p. 8
Where the paper says it

Decreasing entropy therefore requires population transfer to lower energy sectors. The 𝛼-entropic inequalities all collapse to a reverse lexicographic condition on the block sums when 1 − 𝛼 > 0, and a lexicographic condition when 1 − 𝛼 < 0.

Low-Dimensional Examples

For low dimensions like n = 3, specific limiting semigroups Tmu are explicitly described based on the energy structure of E1 ≥ E2 ≥ E3. The structure of these limiting semigroups is illustrated in Section 4 for n = 3 and n = 4. p. 3

Where the paper says it

4), and we study the zero-temperature limit of the Renyi entropic inequalities ` (Proposition 3. The limiting semigroups and their extreme points are illustrated in Section 4 for 𝑛 = 3 and 𝑛 = 4, followed by a discussion of finite-temperature corrections.

Finite-Temperature Extremal Flow

The construction of the complete graph of extremal Gibbs-preserving operations across the entire temperature range involves starting from the permutation vertices at infinite temperature and propagating branches towards lower temperatures. This procedure is used to determine which branches terminate at subsequent critical temperatures. p. 1

Where the paper says it

GRAMEGNA, AND M. LIGABO` At low temperatures, the Gibbs state becomes increasingly concentrated on the ground-state sector.

The resulting structure may then be verified by performing the same construction in the opposite direction. p. 0
Where the paper says it

We investigate limits of semigroups of stochastic matrices defined by their invariant distribution. Given probability vectors 𝛾(𝛽) depending on a parameter 𝛽, we introduce a notion of convergence as 𝛽 → ∞ for the corresponding semigroups of 𝛾(𝛽)-preserving stochastic matrices and investigate the structure of the resulting limit.

Critical Temperature Determination

The critical temperature beta∗ is defined by an equation derived from the multiscale structure, which provides a lower bound for all nontrivial Gibbs critical inverse temperatures. This threshold is realized by balancing a selected population in the block of energy Er against the cumulative Gibbs weight of higher-energy blocks. p. 8

Where the paper says it

Decreasing entropy therefore requires population transfer to lower energy sectors. The 𝛼-entropic inequalities all collapse to a reverse lexicographic condition on the block sums when 1 − 𝛼 > 0, and a lexicographic condition when 1 − 𝛼 < 0.

The sequence of critical temperatures suggests a constructive procedure for determining the finite-temperature extremal structure. p. 3
Where the paper says it

In this case the limit semigroup only depends on the degeneracy structure of 𝛾(𝛽), and it is a set of block-upper-triangular matrices that we dub BT-stochastic matrices (‘BT’ for ‘block triangular’). We identify and enumerate the extremal BTstochastic matrices (Proposition 3.

Extremal Operation Flow

The flow graph for extreme points of Sgamma(beta) for n = 3 illustrates how permutations and block-upper-triangular matrices interpolate between different regimes based on energy degeneracies. The structure matches the flowchart given in [MH18]. p. 0

Where the paper says it

We investigate limits of semigroups of stochastic matrices defined by their invariant distribution. Given probability vectors 𝛾(𝛽) depending on a parameter 𝛽, we introduce a notion of convergence as 𝛽 → ∞ for the corresponding semigroups of 𝛾(𝛽)-preserving stochastic matrices and investigate the structure of the resulting limit.

The extreme operations are summarized in Table 1 for n = 4, depending on the underlying degeneracy structure μ. p. 0
Where the paper says it

Given probability vectors 𝛾(𝛽) depending on a parameter 𝛽, we introduce a notion of convergence as 𝛽 → ∞ for the corresponding semigroups of 𝛾(𝛽)-preserving stochastic matrices and investigate the structure of the resulting limit. In general, the limiting semigroup differs from the semigroup preserving the limiting distribution, showing that these two operations do not commute.

Finite-Temperature Corrections Details

The analysis of finite-temperature corrections involves defining block masses w a(beta) and relative scale functions ρ ar(beta). The criticality condition simplifies to an exponential form for the Gibbs case, leading to a unique positive solution betar for Rr(betar) = 1. p. 1

Where the paper says it

A stochastic matrix 𝑃 is called 𝑞-subpreserving if 𝑞𝑃 ≤ 𝑞, where the inequality is understood componentwise. In the special case 𝑞 = e, the condition e𝑃 ≤ e is equivalent to requiring that all column sums of 𝑃 are at most 1.

This implies that every nontrivial Gibbs critical inverse temperature beta∗ must satisfy beta∗ ≤ max r βr =: βlow. p. 7
Where the paper says it

17) 𝑥 ≺ 𝛾(𝛽) 𝑦 ⇐⇒ ∑︁ 𝑗 𝛾(𝛽)𝑗 𝑔 𝑥 𝑗 𝛾(𝛽)𝑗 ≤ ∑︁ 𝑗 𝛾(𝛽)𝑗 𝑔 𝑦 𝑗 𝛾(𝛽)𝑗 for every continuous convex function 𝑔 defined on a domain containing the points 𝑥 𝑗/𝛾(𝛽)𝑗 and 𝑦 𝑗/𝛾(𝛽)𝑗 . Assume that 𝛾(𝛽) is a Gibbs vector parametrised by the inverse temperature 𝛽.

The construction of extreme operations is available in [Cza26]. p. 0
Where the paper says it

Given probability vectors 𝛾(𝛽) depending on a parameter 𝛽, we introduce a notion of convergence as 𝛽 → ∞ for the corresponding semigroups of 𝛾(𝛽)-preserving stochastic matrices and investigate the structure of the resulting limit. In general, the limiting semigroup differs from the semigroup preserving the limiting distribution, showing that these two operations do not commute.

The resulting branching of extreme operations is summarized in graphs shown in Figs. p. 0
Where the paper says it

Given probability vectors 𝛾(𝛽) depending on a parameter 𝛽, we introduce a notion of convergence as 𝛽 → ∞ for the corresponding semigroups of 𝛾(𝛽)-preserving stochastic matrices and investigate the structure of the resulting limit. In general, the limiting semigroup differs from the semigroup preserving the limiting distribution, showing that these two operations do not commute.

2 and 3.

Final Structure Description

The zero-temperature limit yields the BT-stochastic matrices Tmu, which are row-stochastic block-upper-triangular matrices where each diagonal block is column-substochastic. The last diagonal block Pkk is both row-stochastic and column-substochastic, hence bistochastic. p. 5

Where the paper says it

The last diagonal block 𝑃 𝑘𝑘 is both row-stochastic and column-substochastic. Since its entries are nonnegative, it is automatically column-stochastic, hence bistochastic.

p. 5
Where the paper says it

The last diagonal block 𝑃 𝑘𝑘 is both row-stochastic and column-substochastic. Since its entries are nonnegative, it is automatically column-stochastic, hence bistochastic.

The number of extreme points in the zero-temperature limit is given by Proposition 3. p. 3
Where the paper says it

4), and we study the zero-temperature limit of the Renyi entropic inequalities ` (Proposition 3. The limiting semigroups and their extreme points are illustrated in Section 4 for 𝑛 = 3 and 𝑛 = 4, followed by a discussion of finite-temperature corrections.

p. 3
Where the paper says it

4), and we study the zero-temperature limit of the Renyi entropic inequalities ` (Proposition 3. The limiting semigroups and their extreme points are illustrated in Section 4 for 𝑛 = 3 and 𝑛 = 4, followed by a discussion of finite-temperature corrections.

3. The resulting values in these two asymptotic regimes exhibit substantial variation with the underlying degeneracy structure. p. 0
Where the paper says it

Given probability vectors 𝛾(𝛽) depending on a parameter 𝛽, we introduce a notion of convergence as 𝛽 → ∞ for the corresponding semigroups of 𝛾(𝛽)-preserving stochastic matrices and investigate the structure of the resulting limit. In general, the limiting semigroup differs from the semigroup preserving the limiting distribution, showing that these two operations do not commute.

p. 0
Where the paper says it

Given probability vectors 𝛾(𝛽) depending on a parameter 𝛽, we introduce a notion of convergence as 𝛽 → ∞ for the corresponding semigroups of 𝛾(𝛽)-preserving stochastic matrices and investigate the structure of the resulting limit. In general, the limiting semigroup differs from the semigroup preserving the limiting distribution, showing that these two operations do not commute.

Conclusion on Majorisation

BT-majorisation interpolates between standard majorisation (when mu = (n)) and upper triangular (UT) or unordered majorisation (when mu = (1,..., 1)). The paper concludes by showing that the limiting distribution alone does not determine the limiting semigroup. p. 3

Where the paper says it

Two families of probability vectors, 𝛾(𝛽) and e𝛾(𝛽), converge to the same boundary point of Δ2 as 𝛽 → ∞, while approaching it along different asymptotic trajectories. Consequently, their associated semigroups of preserving stochastic matrices have different Kuratowski limits, illustrating that the limiting distribution alone does not determine the limiting semigroup.

The map Δn-1 ∋ q ↦ Sq is not continuous at the boundary of Δn-1. p. 3
Where the paper says it

In particular, the limit depends on the multiscale profile of 𝛾(𝛽) (see Sec. 1 below), and this shows that the map Δ𝑛−1 ∋ 𝑞 ↦→ S𝑞 is not continuous at the boundary of Δ𝑛−1.

The resulting notion of Block-Triangular majorisation interpolates between ordinary majorisation and upper triangular (aka unordered) majorisation. p. 0
Where the paper says it

We characterise and enumerate their extremal elements and determine the preorder on probability vectors induced by the action of the semigroup. The resulting notion of Block-Triangular majorisation interpolates between ordinary majorisation and upper triangular (aka unordered) majorisation.

References

[BBI01] D. Burago, Y. Burago, and S. Ivanov. A Course in Metric Geometry, American Mathematical Society, (2001)

[BCP14] T. Baumgratz, M. Cramer, and M. B. Plenio, Quantifying coherence Phys Rev Lett. p. 0

Where the paper says it

Introduction Majorisation governs a number of fundamental state-conversion problems in quantum information theory and statistical physics. Examples include bipartite pure-state entanglement transformations [Nie99], coherence manipulation [BCP14], and thermodynamic state transitions under Gibbs-preserving constraints [HO13].

, 113, 140401, (2014)

[Bir46] G Birkhoff, Tres observaciones sobre el algebra lineal Univ. Nac. Tucuman, Ser. A, 5, 147–154, (1946)

[Brø83] A Brøndsted An Introduction to Convex Polytopes Springer New York, (1983)

[CCGdOJ25] F. D. Cunden, J. Czartowski, G. p. 0

Where the paper says it

LIMITS OF STOCHASTIC SEMIGROUPS AND BLOCK-TRIANGULAR MAJORISATION FABIO DEELAN CUNDEN, JAKUB CZARTOWSKI, GIOVANNI GRAMEGNA, AND MARILENA LIGABO` Abstract. We investigate limits of semigroups of stochastic matrices defined by their invariant distribution.

Gramegna, and A de Oliveira Junior Relative volume of comparable pairs under semigroup majorization Letters in Mathematical Physics, 115, 79, (2025)

[Cza26] J Czartowski Companion code for extreme points of Gibbs-preserving stochastic maps https://github.com/jack-point13/deg ext gp stoch, (2026)

[dOJCZK22] A de Oliveira Junior, J Czartowski, K Zyczkowski, and K Korzekwa Geometric structure of thermal cones (2022)

[GKP92] D A Gregory, S J Kirkland, and N J Pullman Row-stochastic matrices with a common left fixed vector Linear Algebra and its Applications, 169, 131–149, (1992)

[GMN+15] G Gour, M P Müller, V Narasimhachar R W Spekkens, and N Yunger Halpern The resource theory of informational nonequilibrium in thermodynamics Phys Rep, 583, 1-58, (2015)

[Har74] D J Hartfiel A study of convex sets of stochastic matrices induced by probability vectors Pacific Journal of Mathematics, 52, 405–418, (1974)

[HO13] M Horodecki and J Oppenheim Fundamental limitations for quantum and nanoscale thermodynamics Nat Commun., 4, (June 2013)

[Joe90] H Joe Majorization and divergence Journal of Mathematical Analysis and Applications, 148, 287–305, (May 1990)

[JR67] W Jurkat and H Ryser Term ranks and permanents of nonnegative matrices Journal of Algebra, 5, (1967)

[Kur66] K Kuratowski Topology Volume I Academic Press New York, (1966)

[MOA79] A W Marshall, I Olkin, and B C Arnold Inequalities theory of majorization and its applications volume 143 Springer, (1979)

[Nie99] M A Nielsen Conditions for a class of entanglement transformations Phys Rev Lett., 83, 436, (1999)

[SN24] J Son and N H Ng A hierarchy of thermal processes collapses under catalysis Quantum Science and Technology, 10, 015011, (2024)

[ST25] N Shiraishi and R Takagi Recovery of the second law in fully quantum thermodynamics arXiv preprint arXiv: 2510.05642, (2025)

[Sta11] R P Stanley Enumerative Combinatorics Cambridge Studies in Advanced Mathematics Cambridge University Press, 2 edition, (2011)

[vD22] F vom Ende and G Dirr The d-majorization polytope Linear Algebra Appl., 649, 152-185, (2022)

Summary

Given probability vectors gamma(beta) depending on a parameter beta, we introduce a notion of convergence as beta → ∞ for the corresponding semigroups of gamma(beta)-preserving stochastic matrices and investigate the structure of the resulting limit. The paper shows that taking the limit of invariant distributions and taking the semigroup of stochastic matrices preserving them are, in general, non-commuting operations. p. 1

Where the paper says it

2) asks whether taking the zero-temperature limit commutes with passing from a Gibbs state to the semigroup of stochastic maps that preserve its limit 𝛾(∞). In this paper we show that the answer is negative: taking the limit of invariant distributions and taking the semigroup of stochastic matrices preserving them are, in general, non-commuting operations.

How the training signal is built

The limiting semigroup Sgamma(beta) converges to Tq1,..., Tqk of row-stochastic block-upper-triangular matrices. p. 5

Where the paper says it

Then the semigroups S 𝛾(𝛽) converge, in the Kuratowski sense, to the set T 𝑞 1 ,. ,𝑞𝑘 of row-stochastic block-upper-triangular matrices (3.

This convergence is shown by proving two inclusions: LsSgamma(beta) ⊆ Tq1,. p. 0
Where the paper says it

Given probability vectors 𝛾(𝛽) depending on a parameter 𝛽, we introduce a notion of convergence as 𝛽 → ∞ for the corresponding semigroups of 𝛾(𝛽)-preserving stochastic matrices and investigate the structure of the resulting limit. In general, the limiting semigroup differs from the semigroup preserving the limiting distribution, showing that these two operations do not commute.

.., Tqk and Tq1,..., Tqk ⊆ LiSgamma(beta). The latter involves constructing a family of matrices P(beta) that converges to the target matrix P.

Where the gains come from

The BT-stochastic matrices Tmu are characterized by extreme points which are 0–1 matrices satisfying row-stochasticity and column-substochasticity constraints. The number of extreme points of Tmu is given by a formula involving Laguerre polynomials. p. 6

Where the paper says it

It is known [Joe90, Remark 4. 5] that, for 𝑞 ∈ Δ𝑛−1, the number of extreme points of 𝑆 𝑞 is at least 𝑛!

The BT-majorisation relation interpolates between standard majorisation (when mu = (n)) and upper triangular or unordered majorisation (when mu = (1,..., 1)).

Conclusion

The zero-temperature limit yields the BT-stochastic matrices Tmu, which are row-stochastic block-upper-triangular matrices where each diagonal block is column-substochastic. The last diagonal block Pkk is both row-stochastic and column-substochastic, hence bistochastic. The number of extreme points in the zero-temperature limit is given by Proposition 3.3. The resulting values in these two asymptotic regimes exhibit substantial variation with the underlying degeneracy structure.

Green — the paper says it, and the page is printed. Red — no passage carries it, so treat that sentence as unverified. The same check runs on your own text: check a summary against its source, $5.

Improvements for AI systems

  1. Automated determination of limiting semigroup structure: The improved system can determine if a given family of invariant distributions converges to a single limiting distribution or multiple distributions, which in turn dictates whether the resulting limit semigroup is determined by Kuratowski convergence or requires a more rigid structure like block-upper-triangular matrices.

  2. Characterization of extreme operations for Gibbs states: The system can identify the complete graph of extremal Gibbs-preserving operations across all temperature ranges by implementing the iterative construction procedure described in Section 3.4, allowing prediction of how extremal structures deform into Block-Triangular majorisation as temperature varies from zero to infinite.

  3. Finite-temperature correction analysis: The improved system can analyze asymptotic expansions of the semigroup and induced preorder in powers of exp negative temperature, revealing how Block-Triangular majorisation deforms into thermomajorisation by identifying critical inverse temperatures where extremal operations coalesce or branch.

  4. Automated identification of energy degeneracy effects: The system can automatically classify probability vectors based on their energy structure type mu, allowing it to predict the resulting limiting semigroup T mu and its extreme points based solely on the multiplicities of the energy levels.

  5. Constraint checking for majorisation types: The system can verify whether a given relationship between two probability vectors satisfies Block-Triangular majorisation by checking if it adheres to the conditions derived from Theorem 3.4, which involves comparing sums of rearranged entries according to the block structure mu.

Abstract

We investigate limits of semigroups of stochastic matrices defined by their invariant distribution. Given probability vectors q(β) depending on a parameter β, we introduce a notion of convergence as β to infinity for the corresponding semigroups of q(β) -preserving stochastic matrices and investigate the structure of the resulting limit. In general, the limiting semigroup differs from the semigroup preserving the limiting distribution, showing that these two operations do not commute. We develop a general framework for such limiting semigroups and study in detail the case in which the invariant distributions are Gibbs vectors at inverse temperature β. In that case, the limiting semigroup consists of block-upper-triangular stochastic matrices subject to additional substochasticity constraints. We characterise and enumerate their extremal elements and determine the preorder on probability vectors induced by the action of the semigroup. The resulting notion of Block-Triangular majorisation interpolates between ordinary majorisation and upper triangular (also known as unordered) majorisation. We show that it is completely characterised by a finite family of monotones and analyse the corresponding behaviour of R é nyi α-entropies as β to infinity.

Sources

Related papers