Pith. sign in

REVIEW 2 major objections 4 minor 1 cited by

Non-uniqueness of stationary measures for stochastic systems with almost surely invariant manifolds

T0 review · 2 major / 4 minor · reviewed 2026-08-07 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗

Pith's one-line read This paper proves that a positive transverse Lyapunov exponent forces a new stationary measure off an almost surely invariant submanifold, and in the degenerate-forced Lorenz 96 model small damping yields exactly two ergodic stationary…

desk verdict The abstract compact-manifold theorem is the real contribution and holds; the L96 application is credible but Theorem 1.1 is not fully established for N=9 and N=12 because the required Lie-algebra computation is explicitly omitted. read the letter →

arxiv 2505.22903 v2 pith:LW7XA5CT submitted 2025-05-28 math.DS math.PR

Jacob Bedrossian , Alex Blumenthal , Sam Punshon-Smith This is my paper
classification math.DSmath.PR MSC 37H1560H1037D2535R6037A50
keywords non-uniquenessofstationarymeasurestransverseLyapunovexponentalmostsurelyinvariantsubmanifoldLorenz96modeldegeneratestochasticforcingFeynman–KacsemigroupHörmanderconditiongeometricergodicity
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

The paper establishes a general criterion for non-uniqueness of stationary measures: a stochastically forced system with an almost surely invariant submanifold must admit an additional stationary measure supported off that submanifold whenever the transverse Lyapunov exponent is positive and the transverse projective process is uniformly geometrically ergodic. The proof converts exponential transverse growth into a Lyapunov–Foster drift condition using an eigenfunction of a tilted Feynman–Kac semigroup, and the drift condition produces the new measure by standard Markov-process stability theory. The criterion is implemented in the Lorenz 96 model with forcing on every third mode, where the invariant subspace carries an independent Ornstein–Uhlenbeck process. For sufficiently small damping the paper proves there are exactly two ergodic stationary measures, the Gaussian on the invariant subspace and a smooth, geometrically ergodic measure equivalent to Lebesgue measure on the whole space, so generic long-time statistics are governed by the off-subspace measure. This matters because it gives a computable, mechanism-driven answer to when stochastic high-dimensional systems have multiple statistical equilibria.

What carries the argument

The machinery is the transverse projective process $(y_t, v_t)$ on the bundle of unit vectors perpendicular to the invariant submanifold, together with the tilted semigroup $\widehat{P}^{\perp,p}_t$ obtained by twisting the projective Markov semigroup with the factor $|A^\perp_t v|^{-p}$. For small $p$, this semigroup has a spectral gap, and its positive dominant eigenfunction $\psi_p$ satisfies the Feynman–Kac eigenfunction equation whose eigenvalue $\Lambda(p)$ has derivative $\Lambda'(0)=\lambda^\perp_\epsilon$; the combination $V(u)=|\Pi^\perp u|^{-p}\,\psi_p(\Pi u, \Pi^\perp u/|\Pi^\perp u|) + e^{\eta |u|^2}$ then satisfies the drift condition $LV \le -\lambda V + C$, producing recurrence off the invariant subspace. The second engine is the algebraic generation result $\mathrm{Lie}(\{M_k : k\in I\}) = \mathfrak{sl}(H_I^\perp)$, proved by exact symbolic computation for a base case and extended to all $N\ge 15$ by shift invariance, with $N=9,12$ left to direct computation; this bracket-spanning statement is what supplies Hörmander's condition, topological irreducibility, the strong Feller property, and ultimately the smooth densities and geometric ergodicity used for uniqueness.

What would settle it

Perform the exact symbolic Lie-bracket computation for $N=9$ and $N=12$ and check whether $\mathrm{Lie}(\{M_k:k\in I\})=\mathfrak{sl}(H_I^\perp)$; a failure in either case would overturn Theorem 1.1 for that size, while a numerical check that $\lambda^\perp_\epsilon$ is negative for small $\epsilon$ in those sizes would contradict the positivity step.

Watch

Extended reading notes

Core claim

The central discovery is that exponential growth transverse to an almost surely invariant submanifold, measured by $\lambda^\perp_\epsilon$, is not just an instability indicator but a constructive tool: it forces the existence of a stationary measure that lives on the complement of the submanifold. The abstract result (Theorem 2.3) says that if the projective process of unit normal vectors is uniformly geometrically ergodic and $\lambda^\perp_\epsilon > 0$, then there is a stationary measure $\mu$ with $\mu(\mathcal{N})=0$, hence at least two stationary measures; the Lyapunov function is built as $V(x) = |w(x)|^{-p}\psi_q(y(x), v(x))$ near the submanifold, where $\psi_q$ is the positive dominant eigenfunction of the tilted semigroup. In the Lorenz 96 application every hypothesis is checked from the structure of the equations: the computer-assisted Lie algebra generation $\mathrm{Lie}(\{M_k : k\in I\}) = \mathfrak{sl}(H_I^\perp)$ gives Hörmander's condition and irreducibility for the projective process and for the full process on $H \setminus H_I$, the Fisher-information identity together with the absence of an invariant density for the $\epsilon=0$ flow gives $\lambda^\perp_\epsilon>0$ for small $\epsilon$, and Harris's theorem upgrades existence to uniqueness and geometric ergodicity. The final statement is the exact bifurcation picture: for every $N\ge 9$ and $\epsilon$ below a threshold, exactly two ergodic stationary measures exist.

Load-bearing premise

The Lorenz 96 result rests on the computer-checked claim that the forced modes' linearizations generate the full special linear algebra on the transverse space for every $N\ge 9$; for $N=9$ and $N=12$ the appendix explicitly skips that check, so the theorem covers those sizes only if the omitted computation passes.

Editorial extensions

If this is right

  • For the Lorenz 96 model with degenerate forcing, once $\epsilon$ is below the threshold, Lebesgue-generic initial data — including data starting arbitrarily close to the invariant subspace — have long-time statistics given by the off-subspace measure $\mu$, not by the Gaussian $\mu_I$.
  • The two-measure description is exact for small damping: the non-uniqueness is not merely a lower bound, so the stationary statistical behavior of the model is completely classified in this regime.
  • In the abstract compact setting, the criterion gives a simple sufficient condition — uniform geometric ergodicity of the projective process plus $\lambda^\perp_\epsilon>0$ — for at least two stationary measures, which can be checked in other stochastic systems with invariant submanifolds.
  • The second measure inherits a smooth density and exponential convergence from the Hörmander/irreducibility package, so observables converge to their $\mu$-expectations at rate $e^{-\gamma t}$ with a weight that blows up near the invariant subspace.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

  • If the omitted Lie-algebra checks for $N=9$ and $N=12$ pass, the same theorem should hold uniformly across all $N\ge9$; since generation already fails for $N=3,6$, the $N\ge 9$ threshold is probably sharp, not a technical artifact.
  • A concrete numerical prediction follows from the proof: just below the threshold, sample paths should show long transient trapping near $H_I$ followed by ejection, with the escape rate determined by the Feynman–Kac eigenvalue $\Lambda(p)$; this could be tested by simulating the transverse linearized process.
  • The sparsity-and-shift-invariance route to Lie algebra generation suggests the same proof template applies to other high-dimensional equations with local-in-frequency interactions, such as shell models of turbulence, where verifying $\mathrm{Lie}(\{M_k\})=\mathfrak{sl}$ would be the first step toward a two-measure theorem.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Referee Report

2 major / 4 minor

Summary. The paper develops a general framework for proving non-uniqueness of stationary measures for stochastic dynamical systems with an almost surely invariant submanifold. The abstract result, Theorem 2.3, gives sufficient conditions—uniform geometric ergodicity of the transverse projective process and a positive transverse Lyapunov exponent—for the existence of a stationary measure supported off the invariant submanifold on a compact manifold. The authors then apply this machinery to the Lorenz 96 model with forcing on every third mode, asserting in Theorem 1.1 that for N >= 9 and sufficiently small damping parameter, there are exactly two ergodic stationary measures: the Gaussian measure µ_I on the invariant subspace H_I and a second measure µ with smooth density on H \ H_I, which is geometrically ergodic. The proof combines the abstract Lyapunov-function construction with hypoellipticity, irreducibility, geometric ergodicity, a Fisher-information identity for the transverse Lyapunov exponent, and a computer-assisted Lie algebra generation result.

Significance. If all steps are completed, this is a substantial contribution: it gives a general, conceptually clean mechanism for producing multiple stationary measures from transverse instability, and it demonstrates the mechanism in a genuinely high-dimensional, non-compact example with a sharp characterization of the stationary measures. The compact-manifold theorem is proved in the paper, and the computer-assisted algebraic verification for the base case is a concrete reproducible artifact. The strategy of using a Feynman-Kac eigenfunction to build a Lyapunov function repelling from the invariant manifold is elegant and likely to be influential. The main caveat is that the L96 application is currently conditional on an omitted computer check for N = 9 and N = 12, and on some sketched adaptations of known quantitative hypoelliptic estimates.

major comments (2)
  1. Theorem 1.1 is stated for all N >= 9, and Proposition 4.3 is stated for all N = 3K with K >= 3, but the computer-assisted proof in Appendix C is carried out only for N = 15 and then extended to N >= 15 by shift invariance. Lemma C.3 explicitly says that the cases N = 9 and N = 12 'can be treated by direct computation' and then states 'These cases are omitted'. This is load-bearing: Proposition 4.3 is used in Propositions 4.4, 4.5, and 4.7 to establish Hörmander's condition for the linearized, projective, and full processes, which in turn drives irreducibility, the strong Feller property, geometric ergodicity, and the uniqueness statement in Corollary 4.8. The referenced notebook [36] is described for N = 15 only. Because Remark 1.2 reports that generation genuinely fails for N = 3 and N = 6, the threshold is meaningful, and the theorem as stated is not established for N = 9 and N = 12 without the omitted verification.
  2. The positivity of the transverse Lyapunov exponent λ^ε_⊥ is a central hypothesis for the Lyapunov-function construction, but its proof is presented as a contradiction sketch. Lemma 3.8 is quoted as a 'straightforward adaptation' of [10, Theorem B] and depends on an ε-uniform quantitative hypoellipticity estimate; Lemma 4.17 is stated as 'essentially the same as [Lemma B.2; [10]]' rather than proved, and Corollary 4.16 asserts uniformity from ε-independence of the vector fields but does not supply the full quantitative control needed for the W^{s,1} bound. Lemma 3.7 is only sketched in Section 5.2, with integrability and integration-by-parts justifications deferred to a 'straightforward adaptation' of [10, Proposition 3.2]. The argument may be correct, but as written the positivity claim is supported by unproven adaptations of external results, and this is load-bearing for the existence of the second stationary measure. The manuscript should either give complete proofs of these estimates or state the adapted theorems with all hypotheses explicitly verified.
minor comments (4)
  1. The text refers to 'Theorem 3.4 above', but the displayed item is Proposition 3.4; the cross-reference should be corrected.
  2. The claim that N = 9 and N = 12 can be handled by direct computation should be supported by an explicit record of that computation, either in the appendix or in the public repository, rather than left as an assertion in a remark.
  3. The proof uses the transitivity of the SL(H^⊥_I)-action on the unit sphere without proof or citation; this is standard, but a one-line justification would improve readability.
  4. The phrase 'By standard semigroup arguments (see Section 6.1 for details)' hides the verification that H_p is an eigenfunction of the generator L_ξ; since this is a main step in the drift calculation, it would be better to make the semigroup-to-generator passage explicit.

Circularity Check

0 steps flagged · score 2.0 of 10

No significant circularity: the positivity of lambda_perp, the Feynman-Kac drift construction, and the two-measure scenario are derived independently of the conclusions. One reviewer-flagged gap: the load-bearing algebraic Proposition 4.3 is proved only for N>=15, with N=9,12 explicitly omitted (Lemma C.3, Remark 1.2) — a completeness gap in a computer-assisted computation, not a circular step.

full rationale

We find no derivation step that reduces by construction to its own inputs, and no fitted quantity relabeled as a prediction. Positivity of the transverse Lyapunov exponent (Lemma 3.6) is established by an independent contradiction argument: assuming liminf_{eps->0} lambda_perp_eps/eps < inf, the Fisher-information identity eps*FI(f_eps)=|T|lambda_perp_eps+eps*N (Lemma 3.7, proved in Section 5.2) and the hypoelliptic bound of Lemma 3.8 force L1-local precompactness of the stationary densities f_eps; any limit would be an invariant density for the eps=0 deterministic projective ODE, whose nonexistence is proved in-paper (Lemma 5.6) via the explicit spectrum of DB(y_{a,b}) on the invariant subspace span{e1,e2,e4,e5} (Lemma 5.7). The drift condition (Lemma 3.10) uses Lambda(p)=p*lambda_perp_eps+o(p) with lambda_perp_eps>0 established independently, so the Lyapunov-function construction does not presuppose the existence of mu; the eigenfunction psi_p is constructed by spectral perturbation of the Markov semigroup (Lemma 6.1, Corollary 6.2, Section 6.1), and its positivity and C^1_Veta regularity are proved within the paper (Lemma 6.3, Lemma 6.5). The paper leans heavily on the same authors' prior [10] (Fisher-information identity, Theorem B hypoelliptic W^{s,1} estimate, quantitative hypoellipticity in Lemma 6.5), but those are published, peer-reviewed, parameter-free results whose assumptions do not include the target theorem; under the review rules this is real independent evidence and does not raise the circularity score. We do flag, as the reviewing rule on asserted limitations requires: the entire L96 application rests on Proposition 4.3, Lie({M_k : k in I}) = sl(H^perp_I) for all N>=9, which drives every Hormander/irreducibility/ergodicity statement (Propositions 4.4, 4.5, 4.7; Lemma 3.5; Corollaries 4.6, 4.8, 4.16). Appendix C verifies generation only for N=15 by exact Sympy computation (Proposition C.2) and extends to N>=15 by shift invariance (Lemma C.1); for N=9,12, Lemma C.3 states 'The cases N = 9, 12 can be treated by direct computation, either computer-assisted or by hand. These cases are omitted', and Remark 1.2 merely asserts 'the result can be extended to N = 9 and N = 12 by direct computation'.

Assumptions & free parameters 0 free parameters · 6 assumptions · 0 invented entities

The paper introduces no fitted parameters and no new postulated entities. Its central claims rest on standard tools from ergodic theory and PDE (multiplicative ergodic theorem, Hörmander's theorem, Harris' theorem, spectral perturbation) plus two paper-specific inputs: a quantitative hypoelliptic estimate adapted from the authors' earlier work [10], and a computer-assisted Lie algebra generation argument that is shipped as code for N=15 but asserted without displayed computation for N=9 and N=12.

assumptions (6)
  • standard math Multiplicative ergodic theorem for continuous-time random cocycles, used to define the transverse Lyapunov exponent λ⊥.
    Proposition 3.4(a) invokes the MET (Arnold) to get existence of λ⊥; log-integrability is established in Corollary A.1.
  • standard math Hörmander's hypoellipticity theorem, used to obtain smooth densities and the strong Feller property.
    Sections 4.2 through 4.4 use Hörmander's theorem to conclude hypoellipticity and existence of smooth densities for the projective and full processes.
  • standard math Harris ergodic theorem with weighted norms, used for geometric ergodicity.
    Sections 3.4 and 4.5 apply Harris' theorem to get geometric ergodicity of the projective and full processes in weighted spaces.
  • domain assumption Lemma 3.8, a quantitative hypoelliptic estimate adapted from Theorem B in [10].
    The lemma is stated with a proof sketch ('straightforward adaptation of [10, Section 4]') and is load-bearing for the lower bound on λ⊥. The adaptation is not fully written out.
  • ad hoc to paper The Lie algebra generation result Lie({M_k})=sl(H⊥_I) for N=9 and N=12, asserted without a displayed computation.
    Appendix C proves the N=15 case via code and extends by shift invariance to N≥15; the cases N=9 and N=12 are declared 'omitted' in Lemma C.3, yet Theorem 1.1 includes N≥9.
  • domain assumption Fisher information identity Lemma 5.5, proven only by sketch with integration by parts formally justified by Lemma 5.4.
    The identity underpins the contradiction proof of positivity of λ⊥; the proof is labeled 'proof sketch' and refers to [10, Proposition 3.2] for details.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Non-uniqueness of stationary measures for stochastic systems with almost surely invariant manifolds." pith.science (2026). https://pith.science/paper/LW7XA5CT

@misc{pith2026250522903,
  author       = {Pith},
  title        = {Pith review of: Non-uniqueness of stationary measures for stochastic systems with almost surely invariant manifolds},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/LW7XA5CT}},
  note         = {Machine review of arXiv:2505.22903}
}
read the original abstract

We develop a general framework for establishing non-uniqueness of stationary measures for stochastically forced dynamical systems possessing an almost surely invariant submanifold. Our main abstract result provides sufficient conditions for the existence of multiple stationary measures on compact manifolds, though the underlying methodology extends to non-compact settings. The key insight is to construct additional stationary measures by exploiting the linear instability of the invariant submanifold, as quantified by a positive transverse Lyapunov exponent. To demonstrate the practical applicability of our framework, we apply it to the Lorenz 96 model with degenerate stochastic forcing, which serves as an example of both non-compact and high-dimensional dynamics. We prove that as the damping parameter becomes sufficiently small, the unique stationary measure bifurcates, giving rise to exactly two distinct stationary measures. The proof combines our general theory with computer-assisted verification of certain Lie algebra generation properties that ensure the required hypoellipticity and irreducibility conditions.

Figures

Figures reproduced from arXiv: 2505.22903 by the authors.

Figure 1
Figure 1. The submanifold N (shown as a circle) contains a point p that is simultaneously a sink along N and a source in the transverse direction. Trajectories starting off N may initially be attracted to N away from p, but once near N, they flow toward p and then are repelled transversely, preventing permanent collapse onto N. Further comments on Example 2.7 are given in Remark 2.16 at the end of Section 2. 2.3 Preliminaries… view at source ↗

Discussion (0). Sign in to comment.

Forward citations

Cited by 1 Pith paper

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score. Full citation record

  1. Asymptotics of Lyapunov Exponents and Phase Transitions for Fluids with Degenerate Forcing

    math.PR 2026-07 accept novelty 7.5 of 10

    Under a Lie-algebra non-degeneracy condition, the top Lyapunov exponent of the slow-fast system (2.3) converges to E[λ(A(Z))] as the time-scale separation vanishes, implying ergodicity phase transitions in truncated f...

Reference graph

Works this paper leans on

41 extracted references · 40 canonical work pages · cited by 1 Pith paper

  1. [36]

    Punshon-Smith, L-96 cap: Computer-assisted proof for lorenz 96 model hypoellipticity , GitHub, 2025

    S. Punshon-Smith, L-96 cap: Computer-assisted proof for lorenz 96 model hypoellipticity , GitHub, 2025. Available at: https://github.com/punshs/L-96_CAP

    work page 2025

  2. [10]

    Bedrossian, A

    J. Bedrossian, A. Blumenthal, and S. Punshon-Smith, A regularity method for lower bounds on the lyapunov exponent for stochastic differential equations, Inventiones mathematicae 227 (2022), no. 2, 429–516

    work page 2022

  3. [1]

    Albritton, S

    D. Albritton, S. Armstrong, J.-C. Mourrat, and M. Novack, Variational methods for the kinetic Fokker–Planck equation 17 (July 19, 2024), 1953–2010 (en)

    work page 2024

  4. [2]

    Arendt, A

    W. Arendt, A. Grabosch, G. Greiner, U. Moustakas, R. Nagel, U. Schlotterbeck, U. Groh, H. P Lotz, and F. Neubrander, One-parameter semigroups of positive operators, V ol. 1184, Springer, 1986

    work page 1986

  5. [3]

    Arnold, A formula connecting sample and moment stability of linear stochastic systems, SIAM Journal on Applied Mathe- matics 44 (1984), no

    L. Arnold, A formula connecting sample and moment stability of linear stochastic systems, SIAM Journal on Applied Mathe- matics 44 (1984), no. 4, 793–802

    work page 1984

  6. [4]

    Arnold, Random dynamical systems , 1st ed., Springer Monographs in Mathematics, Springer Berlin, Heidelberg, 1998

    L. Arnold, Random dynamical systems , 1st ed., Springer Monographs in Mathematics, Springer Berlin, Heidelberg, 1998. Hardcover ISBN: 978-3-540-63758-5, Softcover ISBN: 978-3-642-08355-6, eBook ISBN: 978-3-662-12878-7

    work page 1998

  7. [5]

    Ayyer and M

    A. Ayyer and M. Stenlund, Exponential decay of correlations for randomly chosen hyperbolic toral automorphisms , Chaos: An Interdisciplinary Journal of Nonlinear Science 17 (2007), no. 4

    work page 2007

  8. [6]

    P. H Baxendale, Statistical equilibrium and two-point motion for a stochastic flow of diffeomorphisms , Spatial Stochastic Processes: A Festschrift in Honor of Ted Harris on his Seventieth Birthday (1991), 189–218

    work page 1991

Show all 41 references
  1. [7]

    H Baxendale, A stochastic hopf bifurcation, Probability Theory and Related Fields 99 (1994), 581–616

    P. H Baxendale, A stochastic hopf bifurcation, Probability Theory and Related Fields 99 (1994), 581–616

    1994

  2. [8]

    H Baxendale, Invariant measures for nonlinear stochastic differential equations , Lyapunov exponents: Proceedings of a conference held in oberwolfach, may 28–june 2, 1990, 2006, pp

    P. H Baxendale, Invariant measures for nonlinear stochastic differential equations , Lyapunov exponents: Proceedings of a conference held in oberwolfach, may 28–june 2, 1990, 2006, pp. 123–140

    1990

  3. [9]

    Baxendale and D

    P. Baxendale and D. Stroock, Large deviations and stochastic flows of diffeomorphisms, Probability Theory and Related Fields 80 (1988), no. 2, 169–215

    1988

  4. [11]

    Bedrossian, A

    J. Bedrossian, A. Blumenthal, and S. Punshon-Smith, Almost-sure exponential mixing of passive scalars by the stochastic navier–stokes equations, The Annals of Probability 50 (2022), no. 1, 241–303

    2022

  5. [12]

    Bedrossian and K

    J. Bedrossian and K. Liss, Quantitative spectral gaps for hypoelliptic stochastic differential equations with small noise, Prob- ability and Mathematical Physics 2 (2021), no. 3, 477–532

    2021

  6. [13]

    Bedrossian and K

    J. Bedrossian and K. Liss, Stationary measures for stochastic differential equations with degenerate damping, arXiv preprint arXiv:2206.02240 (2022)

    2022 arXiv

  7. [14]

    Bedrossian and S

    J. Bedrossian and S. Punshon-Smith, Chaos in stochastic 2d Galerkin-Navier–Stokes 405 (Apr. 16, 2024), 107 (en)

    2024

  8. [15]

    Benaim, Stochastic persistence, arXiv preprint arXiv:1806.08450 (2018)

    M. Benaim, Stochastic persistence, arXiv preprint arXiv:1806.08450 (2018)

    2018 arXiv

  9. [16]

    Blumenthal, M

    A. Blumenthal, M. Coti Zelati, and R. S Gvalani, Exponential mixing for random dynamical systems and an example of pierrehumbert, The Annals of Probability 51 (2023), no. 4, 1559–1601

    2023

  10. [17]

    Blumenthal, J

    A. Blumenthal, J. Xue, and L.-S. Young, Lyapunov exponents for random perturbations of some area-preserving maps includ- ing the standard map, Annals of Mathematics 185 (2017), no. 1, 285–310

    2017

  11. [18]

    Chemnitz and M

    D. Chemnitz and M. Engel, Positive Lyapunov exponent in the Hopf normal form with additive noise , Communications in Mathematical Physics 402 (2023), no. 2, 1807–1843

    2023

  12. [19]

    Coti Zelati and M

    M. Coti Zelati and M. Hairer, A noise-induced transition in the Lorenz system, Communications in Mathematical Physics 383 (2021), 2243–2274

    2021

  13. [20]

    Dolgopyat, V

    D. Dolgopyat, V . Kaloshin, and L. Koralov, Sample path properties of the stochastic flows , The Annals of Probability 32 (2004), no. 1A, 1–27

    2004

  14. [21]

    Ruelle, Ergodic theory of chaos and strange attractors , Reviews of modern physics 57 (1985), no

    J-P Eckmann and D. Ruelle, Ergodic theory of chaos and strange attractors , Reviews of modern physics 57 (1985), no. 3, 617

    1985

  15. [22]

    S Ellis, Entropy, large deviations, and statistical mechanics, Springer, 2007

    R. S Ellis, Entropy, large deviations, and statistical mechanics, Springer, 2007

    2007

  16. [23]

    Hairer and J

    M. Hairer and J. C Mattingly, Yet another look at harris’ ergodic theorem for markov chains, Seminar on stochastic analysis, random fields and applications vi: Centro stefano franscini, ascona, may 2008, 2011, pp. 109–117

    2008

  17. [24]

    Z. Hani, Y . Li, A. Nahmod, and G. Staffilani, Non-equilibrium steady state for a three-mode energy cascade model , arXiv preprint arXiv:2505.16018 (2025)

    2025 arXiv

  18. [25]

    Hening and D

    A. Hening and D. H Nguyen, Coexistence and Extinction for Stochastic Kolmogorov Systems, The Annals of Applied Proba- bility 28 (2018), no. 3, 1893–1942

    2018

  19. [26]

    P Herzog and J

    D. P Herzog and J. C Mattingly, A practical criterion for positivity of transition densities 28 (Aug. 1, 2015), 2823–2845 (en)

    2015

  20. [27]

    H ¨ormander, Hypoelliptic second order differential equations, Acta Mathematica 119 (1967), no

    L. H ¨ormander, Hypoelliptic second order differential equations, Acta Mathematica 119 (1967), no. 1, 147–171. 47

    1967

  21. [28]

    Jacobson, Lie algebras, Interscience Tracts in Pure and Applied Mathematics, vol

    N. Jacobson, Lie algebras, Interscience Tracts in Pure and Applied Mathematics, vol. No. 10, Interscience Publishers (a division of John Wiley & Sons, Inc.), New York-London, 1962. MR143793

    1962

  22. [29]

    1, 1985), 361–368

    V Jurdjevic and I Kupka, Polynomial control systems272 (Sep. 1, 1985), 361–368

    1985

  23. [30]

    Jurdjevic,Geometric control theory, Cambridge university press, 1997

    V . Jurdjevic,Geometric control theory, Cambridge university press, 1997

    1997

  24. [31]

    Kifer,Ergodic theory of random transformations, V ol

    Y . Kifer,Ergodic theory of random transformations, V ol. 10, Springer Science & Business Media, 2012

    2012

  25. [32]

    Lian and M

    Z. Lian and M. Stenlund, Positive Lyapunov exponent by a random perturbation , Dynamical Systems 27 (2012), no. 2, 239–252

    2012

  26. [33]

    N Lorenz, Predictability: A problem partly solved, Proc

    E. N Lorenz, Predictability: A problem partly solved, Proc. seminar on predictability, 1996

    1996

  27. [34]

    C Mattingly, Ergodicity of 2d navier–stokes equations with random forcing and large viscosity, Communications in mathe- matical physics 206 (1999), 273–288

    J. C Mattingly, Ergodicity of 2d navier–stokes equations with random forcing and large viscosity, Communications in mathe- matical physics 206 (1999), 273–288

    1999

  28. [35]

    P Meyn and R

    S. P Meyn and R. L Tweedie, Markov chains and stochastic stability, Springer Science & Business Media, 2012

    2012

  29. [37]

    Punshon-Smith, L96-matrix-cap, Google Colab, 2025

    S. Punshon-Smith, L96-matrix-cap, Google Colab, 2025. Available at: https://colab.research.google.com/ drive/1QbzpjwPjui2ebCE07Zsd-pLkZ0aHHMRt

    2025

  30. [38]

    Revuz and M

    D. Revuz and M. Yor, Continuous Martingales and Brownian Motion , Grundlehren der mathematischen Wissenschaften, Springer, 2010 (en)

    2010

  31. [39]

    D. W Stroock and S R S Varadhan, On the support of diffusion processes with applications to the strong maximum principle, Proceedings of the Sixth Berkeley Symposium on Mathematical Statistics and Probability (Univ. California, Berkeley, Calif., 1970/1971), Vol. III: Probabili...

    1970

  32. [40]

    Triebel, Theory of function spaces II, Birkhauser, 1992

    H. Triebel, Theory of function spaces II, Birkhauser, 1992

    1992

  33. [41]

    W Warner, Foundations of Differentiable Manifolds and Lie Groups, Graduate texts in mathematics, Springer, 2010 (en)

    F. W Warner, Foundations of Differentiable Manifolds and Lie Groups, Graduate texts in mathematics, Springer, 2010 (en). 48

    2010

Pith tools

Reviewed August 7, 2026 · model on record in the stance chip above.