Diffusion limits of cyclic finite-velocity random motions along vector fields

arXiv:2608.22514 · math.PR, math.DS · Submitted 2026-08-23 · Read on arXiv

Listen

Radio episode about this paper

Transcript

Introduction to the show: ident: Mathematics. The day in mathematics: new theorems, new proofs, and what they open.

Ada: We're looking at the paper: "Diffusion limits of cyclic finite-velocity random motions along vector fields".

Emmy: Here's the gist of the paper. Diffusion limits of cyclic finite-velocity random motions along vector fields investigate diffusion approximations for a class of multivariate inhomogeneous finite-velocity random motions motivated by models of active particles <ref:2608.22514#pg2>.

Ada: Under a Kac-type scaling that couples vanishing run times with diverging particle speed, the authors prove weak convergence of both processes to multidimensional diffusions <ref:2608.22514#pg4>.

Emmy: It turns out that the order of successive cyclic motions affects the diffusion limit <ref:2608.22514#pg4>.

Ada: In addition to the second-order term generated by fluctuations of the random run times, the limiting drift contains directional derivatives DVirVj s of the underlying vector fields <ref:2608.22514#pg4>.

Title and authors: Ada: Let's go over the title and the authors of this paper, "Diffusion limits of cyclic finite-velocity random motions along vector fields." It’s written by Olga Aryasova, Ilya Pavlyukevich, and Yusif Sualah.

Emmy: I see it's a heavy theoretical piece that bridges stochastic processes with differential geometry because it deals with the vector fields V1 through Vp.

Kolya: I don't see the connection immediately; what is the research actually trying to achieve by studying these finite-velocity random motions in this specific way?

Ada: The goal is to investigate diffusion approximations for a class of multivariate inhomogeneous finite-velocity random motions that are motivated by models of active particles.

Emmy: So, they are taking microscopic particle behaviors—things like motile bacteria or artificial microswimmers—and finding the macroscopic diffusion behavior.

Ada: They do this by alternating these particles cyclically among prescribed velocity fields V1 through Vp with random run times and seeing what happens under the right scaling.

Kolya: What kind of models are they comparing these to? Are they comparing it to something more standard, like a simple Brownian motion?

Ada: They compare them rigorously to multidimensional diffusions, which is a step up from simpler results that might only look at single-particle telegraph processes.

Emmy: The structure of the paper suggests this work is building on older results in the field, like the weak diffusion limits established by Watanabe and Watanabe in one thousand nine hundred seventy <ref:2608.22514#pg1>.

Ada: That’s right, they are taking those established methods and applying them to a much more complex setup involving multiple vector fields cycling together.

Kolya: So if you were someone trying to use this result, what’s the immediate implication for modeling these active systems?

Ada: The immediate implication is that you need a more sophisticated model than just assuming simple random steps; you need to account for the cyclic structure and its orientation when scaling up.

The paper's summary: Emmy: Now, let's look at the actual summary of this paper, "Diffusion limits of cyclic finite-velocity random motions along vector fields." It outlines the main steps they took to get to their conclusion.

Ada: The summary says they establish weak convergence for both flight and gliding dynamics under a Kac-type scaling that links vanishing run times with diverging particle speed.

Emmy: That specific scaling is crucial because it’s what allows them to rigorously prove the convergence of these processes into multidimensional diffusions.

Kolya: What about the distinction between flight and gliding dynamics in the summary? Is that where they found something new?

Ada: Yes, they point out that even though both converge to a diffusion, the limiting drift terms are different because one comes from fluctuations of run times while the other has those directional derivatives.

Emmy: It highlights that the second-order term is determined by fluctuations of run times, but the drift term itself captures how the vector fields interact through their Lie brackets when there's cyclic reorientation.

Kolya: So, in plain terms, it’s saying that microscopic ordering survives in the macroscopic diffusion limit because the non-commutativity of moving between vector fields is what matters.

Ada: Exactly. The key finding is that this generic non-commutativity generates a macroscopic drift dependent on the cyclic order of operations.

Emmy: It confirms that you need to look at how these vector fields relate to each other geometrically, not just their individual properties, when you’re looking for the long-term behavior.

The paper's improvements: Ada: Now we talk about what the paper suggests as improvements or extensions of this research. They aren't necessarily proposing a new method but rather showing what else is possible with these tools.

Kolya: What kind of mathematical challenges are they pointing toward when they discuss these potential extensions?

Emmy: One thing is that when the vector fields commute, meaning the Lie brackets between them are zero, the order-dependent Lie bracket contribution vanishes entirely.

Ada: That simplifies things considerably because in that case, the limiting operators A and B take a much simpler form.

Kolya: So if you have commuting vector fields, does it mean the resulting diffusion becomes just a standard Brownian motion covariance matrix for homogeneous velocities?

Emmy: In a sense, yes. When they commute, the complexity arising from the ordering disappears because there's no interaction between them through those non-commutative terms.

Ada: And another thing they look at is run-and-reverse motion along two opposite vector fields, V1 and V2 being opposite to each other.

Kolya: What happens in that specific case? Does it simplify the whole picture down to something even simpler than the general case?

Emmy: In that scenario, the limiting dynamics becomes degenerate, meaning it’s governed by a one-dimensional Brownian motion.

Ada: And there’s this interesting detail about how they describe that one-dimensional motion. The flight regime diffusion component of X is tangent to a specific line generated by V, while the additional drift acts in the direction determined by the directional derivative of their Lie bracket structure.

Conclusion: Emmy: So, wrapping up this paper, we’ve seen that the main conclusion is that both limiting diffusions X and Y have the same second-order part.

Ada: But they definitely have different drifts, and those differences are driven by the cyclic ordering of the vector fields through their Lie brackets in Stratonovich form.

Kolya: So if I’m left with just one big idea to take away from this, it's that microscopic ordering remains visible in the diffusion limit.

Emmy: That’s a strong statement, Kolya. It means you can't ignore the structure of how those vector fields are arranged when you zoom out to the macroscopic scale.

Ada: So we’re done with this paper on "Diffusion limits of cyclic finite-velocity random motions along vector fields." It confirms that the non-commutativity is a fundamental feature of these systems.

Kolya: I just want to say that even in these complicated setups, you have to be careful about those constants and whether they actually survive the application.

Emmy: That’s a fair point, Kolya. The structure of the limiting generator A and B shows exactly how the microscopic realization of the motion remains visible in the diffusion limit.

Institute of Mathematics, Friedrich Schiller University Jena · Institute of Geophysics, National Academy of Sciences of Ukraine

math.PR, math.DS

Submitted: 2026-08-23

Updated: 2026-08-31

Comments: 26 pages, 3 figures

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

Importance score: 79/100

The gist: Diffusion limits of cyclic finite-velocity random motions along vector fields investigate diffusion approximations for a class of multivariate inhomogeneous finite-velocity random motions motivated…

Key concepts

Flight vs. Gliding Dynamics
These are two types of motion modeled: flights involve straight motion with a constant velocity during a run, while gliding involves movement along the flow generated by the current vector field. The analysis shows that even when driven by the same fields, these two motions converge to different limiting diffusions.
Kac-type Scaling
This is a mathematical scaling technique used to analyze the random motions. It couples vanishing run times with diverging particle speeds. This specific scaling allows the authors to prove weak convergence of both flight and gliding processes to multidimensional diffusions.
Lie Brackets (rVi, Vj s)
These terms represent the non-commutativity of the vector fields V1...Vp. In Stratonovich form, they appear in the limiting drift. This indicates that the order in which vector fields are applied matters for the final diffusion limit, generating macroscopic drift dependent on cyclic ordering.

Terminology

Summary

Diffusion limits of cyclic finite-velocity random motions along vector fields investigate diffusion approximations for a class of multivariate inhomogeneous finite-velocity random motions motivated by models of active particles <ref:2608. p. 0

Where the paper says it

Diffusion limits of cyclic finite-velocity random motions along vector fields Olga Aryasova∗ , Ilya Pavlyukevich† , and Yusif Sualah‡ September 1, 2026 Abstract We investigate diffusion approximations for a class of multivariate inhomogeneous finite-velocity random motions motivated by models of active particles. A particle alternates cyclically among prescribed velocity fields V1, .

22514#pg2>. Under a Kac-type scaling that couples vanishing run times with diverging particle speed, the authors prove weak convergence of both processes to multidimensional diffusions <ref:2608. p. 0
Where the paper says it

, Vp with random run times and evolves according to either a flight dynamics, consisting of piecewise-linear motion, or a gliding dynamics, in which the particle follows the corresponding velocity flow. Under a Kac-type scaling that couples vanishing run times with diverging particle speed, we prove weak convergence of both processes to multidimensional diffusions.

22514#pg4>. It turns out that the order of successive cyclic motions affects the diffusion limit <ref:2608. p. 0
Where the paper says it

Under a Kac-type scaling that couples vanishing run times with diverging particle speed, we prove weak convergence of both processes to multidimensional diffusions. It turns out that the order of successive cyclic motions affects the diffusion limit.

22514#pg4>. In addition to the second-order term generated by fluctuations of the random run times, the limiting drift contains directional derivatives DVirVj s of the underlying vector fields <ref:2608. p. 0
Where the paper says it

It turns out that the order of successive cyclic motions affects the diffusion limit. In addition to the second-order term generated by fluctuations of the random run times, the limiting drift contains directional derivatives DVirVj s of the underlying vector fields.

22514#pg4>. In Stratonovich form, part of this drift is expressed through their Lie brackets rVi, Vj s <ref:2608. p. 0
Where the paper says it

In addition to the second-order term generated by fluctuations of the random run times, the limiting drift contains directional derivatives DVirVj s of the underlying vector fields. In Stratonovich form, part of this drift is expressed through their Lie brackets rVi, Vj s.

22514#pg4>. Thus, the generic non-commutativity of the microscopic motions survives the diffusive scaling and generates a macroscopic drift which depends on the cyclic order <ref:2608. p. 0
Where the paper says it

In Stratonovich form, part of this drift is expressed through their Lie brackets rVi, Vj s. Thus, the generic non-commutativity of the microscopic motions survives the diffusive scaling and generates a macroscopic drift which depends on the cyclic order.

22514#pg4>.

Model Setup and Distinctions

The study considers an idealized model of an active particle performing cyclic motions in a spatially inhomogeneous environment determined by vector fields V1,..., Vp and changing their directions according to a prescribed cyclic reorientation rule <ref:2608. p. 2

Where the paper says it

The cyclic and inhomogeneous motions considered here may be viewed as models of active particles moving in spatially inhomogeneous environments determined by vector fields V1, . , Vp and changing their directions according to a prescribed cyclic reorientation rule.

22514#pg5>. The authors distinguish between two types of motion: flights, corresponding to straight motion with velocity frozen at the beginning of each run, and gliding, corresponding to motion along the flow generated by the current vector field <ref:2608. p. 3
Where the paper says it

, Vp are determined by independent copies of positive random variables τ1, . We distinguish between two types of motion: flights, corresponding to straight motion with velocity frozen at the beginning of each run, and gliding, corresponding to motion along the flow generated by the current vector field, see Fig.

22514#pg6>.

The dynamics are formally defined as follows:

  1. Flights assume a particle moves with constant velocity Vιpxq during a random time interval T1d“ τι, and then changes its velocity at T1 to Vn1pXT1qpt ´ T1q, t P rT2q, (2. p. 3

    Where the paper says it

    Assume that the particle starts at t “ 0 at position X0 “ x with the initial “direction” n0 “ ι P t1, . Then, it moves with the constant velocity Vιpxq during the random time interval T1 d“ τι .

    2) <ref:2608. p. 0
    Where the paper says it

    sualah@uni-jena. de 1 arXiv:2608.

    22514#pg6>.

  2. Gliding dynamics have the form Yt “ YTk ż t Tk VnkpYsq ds, t P rTk, Tk’1q, k P N0 (3. p. 3

    Where the paper says it

    More precisely, the gliding dynamics has the form Y0 “ y P R d , n0 “ ι P t1, . , pu, T0 “ 0, nk “ 1 ` pι ` k ´ 1q mod p, Tk`1 ´ Tk d“ τnk are independent random variables, Yt “ YTk ` ż t Tk Vnk pYsq ds, t P rTk, Tk`1q, k P N0.

    1) <ref:2608. p. 0
    Where the paper says it

    sualah@uni-jena. de 1 arXiv:2608.

    22514#pg7>.

Key Assumptions and Scaling

The analysis relies on several assumptions to ensure the existence of the diffusion limit <ref:2608.22514#pg8>. These include:

**> AV: The vector fields V1,..., Vp are bounded with bounded first and second derivatives <ref:2608. p. 4

Where the paper says it

The vector fields V1, . , Vp are bounded with bounded first and second derivatives, i.

22514#pg9>. ∇V: We denote the Hessian matrix of a real-valued function f are denoted by ∇2f respectively. p. 3
Where the paper says it

For a function f “ fpxq, }f} denotes its supremum norm. The gradient and the Hessian matrix of a real-valued function f are denoted by ∇f and ∇2f, respectively.

∇VirVj s “ DVipxqVj pxq of the vector fields Vi in direction Vj defined as vector fields with coordinates pDVirVj sqα “ ÿd ββ1 Vβ j¨ BβV αi, α = 1,. p. 5
Where the paper says it

εq, ε, δ Ó 0. 16) For the formulation of the main result we will need the directional derivatives DVirVj s “ DVipxqVj pxq of the vector fields Vi in direction Vj defined as vector fields with coordinates pDVirVj sqα “ ÿ d β“1 V β j ¨ BβV α i , α “ 1, .

.., d, 1 ď i, j ď p (2.17). rVi, Vj s defined as DVj rVis ´ DVirVj s, 1 ď i, j ď p prVi, Vj sqα = ÿd ββ1´ Vβi¨ BβV αj ´ Vβj¨ BβV αi¯, α = 1,. p. 5
Where the paper says it

, d, 1 ď i, j ď p, (2. 17) and the Lie bracket rVi , Vj s defined as rVi , Vj s “ DVj rVis ´ DVirVj s, 1 ď i, j ď p, prVi , Vj sqα “ ÿ d β“1 ´ V β i ¨ BβV α j ´ V β j ¨ BβV α i ¯ , α “ 1, .

.., d, 1 ď i, j ď p (2.18). **

**> Aτ: The random variables τ1,. p. 4

Where the paper says it

, }∇Vp}u. Aτ : The random variables τ1, .

.., τp are positive and the Lyapunov condition Eτ2ρi ă 8 for some ρ P p0, 1q is satisfied. p. 4
Where the paper says it

, τp are positive and the Lyapunov condition Eτ 2`ρ i ă 8, i “ 1, . , p, is satisfied for some ρ P p0, 1q.

Eτi “ µi P p0, 8q, Eτ2i = λ2i P p0, 8q, Var τi = λ2i – µ2i P r0, 8q for i = 1. p. 4
Where the paper says it

, p, is satisfied for some ρ P p0, 1q. We denote Eτi “ µi P p0, 8q, Eτ 2 i “ λ 2 i P p0, 8q, Var τi “ λ 2 i ´ µ 2 i “ σ 2 i P r0, 8q, i “ 1, .

.., p (2.7). **

**> AV,µ: The vector fields V1,. p. 0

Where the paper says it

Diffusion limits of cyclic finite-velocity random motions along vector fields Olga Aryasova∗ , Ilya Pavlyukevich† , and Yusif Sualah‡ September 1, 2026 Abstract We investigate diffusion approximations for a class of multivariate inhomogeneous finite-velocity random motions motivated by models of active particles. A particle alternates cyclically among prescribed velocity fields V1, .

.., Vp satisfy µ1V1pxq ¨ ¨ ¨ µpVppxq = 0, x P R d (2. p. 4
Where the paper says it

AV,µ: The vector fields V1, . , Vp satisfy µ1V1pxq ` ¨ ¨ ¨ ` µpVppxq “ 0, x P R d .

9). This condition eliminates the first-order averaged motion. p. 2
Where the paper says it

The short successive run times introduce a rapidly varying microscopic time scale, while Kac’s scaling simultaneously increases the magnitude of the velocities. The centering condition eliminates the firstorder averaged motion, and the resulting diffusion can therefore be interpreted as a second-order effective dynamics of the underlying cyclic random evolution.

**

**> Aκ: There is κ P p0, 8q such that there is a limit lim ε,δÓ0 εδ2 = κ2. p. 5

Where the paper says it

This leads to the following assumption. Aκ: There is κ P p0, 8q such that there is a limit lim ε,δÓ0 ε δ 2 “ κ 2 .

(2.15). This scaling relates the small parameters responsible for space-time scaling of the evolutions X and Y. p. 5
Where the paper says it

, ξp independently of the spatial dependence encoded by V pxq. Let ε, δ P p0, 1s be the small parameters responsible for the space-time scaling of the evolutions X and Y .

**

Limiting Diffusion Properties

The main results establish weak convergence of both flight and gliding dynamics to diffusions X and Y, respectively.

  1. The flight evolutions tXε,δu converge weakly in CpR’, R d q to a diffusion X starting at x with the generator Af “ κ2µ ÿi σ2i x∇2fVi, Viy ´ κ2µ ÿjăi µiµj x∇f, DVirVj sy, f P C3b pR d, Rq (2. p. 6

    Where the paper says it

    Then, for any initial values x P R d and ι P t1, . , pu, the flight evolutions tXε,δu converge weakly in CpR`, R d q to a diffusion X starting at x with the generator Af “ κ 2 2µ ÿ i σ 2 i x∇2 fVi , Viy ` κ 2 µ ÿ jăi µiµj x∇f, DVirVj sy, f P C 3 b pR d , Rq.

    19). The SDE in Itô form is dXt “ κ2µ ÿi σiVipXtq dWi t ´ κ2µ ÿjăi µiµjDVirVj spXtq dt, where W1,. p. 6
    Where the paper says it

    19) The diffusion X is a solution to an Itˆo-SDE dXt “ κ ? µ ÿ i σiVipXtq dWi t ` κ 2 µ ÿ jăi µiµjDVirVj spXtq dt, (2.

    .., Wp are independent standard one-dimensional Brownian motions (2. p. 6
    Where the paper says it

    20) where W1 , . , Wp are independent standard one-dimensional Brownian motions.

    20).

  2. The gliding evolutions tYε,δu converge weakly in CpR’, R d q to a diffusion Y starting at y with the generator Bf “ κ2µ ÿi σ2i x∇2fVi, Viy ´ κ2µ ÿi λ2i x∇f, DVirVisy, DVirVj sy, f P C3b pR d, Rq (2. p. 6

    Where the paper says it

    Then, for any initial values y P R d and ι P t1, . , pu, the gliding evolutions tY ε,δu converge weakly in CpR`, R d q to a diffusion Y starting at y with the generator Bf “ κ 2 2µ ÿ i σ 2 i x∇2 fVi , Viy ` κ 2 2µ ÿ i λ 2 i x∇f, DVirVisy ` κ 2 µ ÿ jăi µiµj x∇f, DVirVj sy, f P C 3 b pR d , Rq.

    22). The SDE in Stratonovich form is dYt “ κ2µ ÿi σiVipYtq ˝ dWi t ´ κ2µ ÿjăi µiµj rVi, Vj spYtq dt (2. p. 6
    Where the paper says it

    23) can be rewritten in the Stratonovich form as dYt “ κ ? µ ÿ i σiVipYtq ˝ dWi t ` κ 2 2µ ÿ jăi µiµj rVi , Vj spYtq dt.

    24).

Influence of Cyclic Ordering

The two limiting diffusions X and Y have the same second-order part, determined by the fluctuations of the run times, but generally different drifts. The cross terms DVirVj s, j ≠ i, reflect the prescribed ordering of the vector fields within one cycle. p. 6

Where the paper says it

24) The two limiting diffusions X and Y have the same second-order part, determined by the fluctuations of the run times, but generally different drifts. The cross terms DVirVj s, j ă i, reflect the prescribed ordering of the vector fields within one cycle.

In Stratonovich form, the order-dependent part of the drift is expressed through their Lie brackets. p. 6
Where the paper says it

The cross terms DVirVj s, j ă i, reflect the prescribed ordering of the vector fields within one cycle. In Stratonovich form, the order-dependent part of the drift is expressed through their Lie brackets.

Example 2.5 illustrates that opposite orientations produce Liebracket drift terms with opposite signs, showing that the orientation of the microscopic cyclic motion remains visible in the macroscopic diffusion limit. p. 7
Where the paper says it

27) In particular, the orderings (3) and (5), corresponding to opposite orientations of the cycle, produce Liebracket drift terms with opposite signs. Thus, the orientation of the microscopic cyclic motion remains visible in the macroscopic diffusion limit.

Invariance and Degeneracy

When vector fields commute, i.e., rVi, Vj s = 0, the order-dependent Lie-bracket contribution vanishes. p. 8

Where the paper says it

, rVi , Vj s “ DVj rVis ´ DVirVj s ” 0, i, j “ 1, . 32) so that the order-dependent Lie-bracket contribution in (2.

In this case, the limiting operators A and B take a simpler form. p. 8
Where the paper says it

33) and the operators A and B take the following simpler form: Af “ κ 2 2µ ÿ i σ 2 i x∇2 fVi , Viy ´ κ 2 2µ ÿ i µ 2 i x∇f, DVirVisy, Bf “ κ 2 2µ ÿ i σ 2 i x∇2 fVi , Viy ` κ 2 2µ ÿ i σ 2 i x∇f, DVirVisy, f P C 3 b pR d , Rq. 34) Consequently, the limiting diffusions satisfy the following SDEs: dXt “ κ ?

Furthermore, for run-and-reverse motion along two opposite vector fields V1 = V and V2 = -V, the limiting dynamics is degenerate and is governed by a one-dimensional Brownian motion W. p. 8
Where the paper says it

37) The limiting dynamics is degenerate and is governed by a one-dimensional Brownian motion W. In the gliding regime, the diffusion Y occurs along the integral curves of the vector field V , dYt “ κ ?

The flight regime diffusion component of X is tangent to the one-dimensional foliation generated by V, whereas the additional drift acts in the direction -DV rV s.

Proof Structure

The proof proceeds by establishing weak relative compactness of the family tZε,δu in Dpr0, 8q, R d q. This is followed by determining the limit of the p-step pseudo-characteristic operator Aε,δ (or equivalently Xε,δ). p. 0

Where the paper says it

Keywords: finite-velocity random motion; inhomogeneous telegraph process; active particles; diffusion approximation; random evolution; weak convergence; Itˆo diffusion; Stratonovich diffusion. 2020 Mathematics Subject Classification: 60J60, 60F17, 60G50, 82C41, 82C70 Contents 1 Introduction 1 2 Inhomogeneous multivariate finite-velocity motions: setting and main results 4 3 Weak relative compactness of tZε,δu 11 4 Asymptotics of the pseudo-characteristic operator of Zε,δ 16 5 Martingale problem and proof of Theorem 2.

The final step involves showing that every weak limit solves the martingale problem for A, thereby establishing convergence to the diffusion X. p. 0
Where the paper says it

Keywords: finite-velocity random motion; inhomogeneous telegraph process; active particles; diffusion approximation; random evolution; weak convergence; Itˆo diffusion; Stratonovich diffusion. 2020 Mathematics Subject Classification: 60J60, 60F17, 60G50, 82C41, 82C70 Contents 1 Introduction 1 2 Inhomogeneous multivariate finite-velocity motions: setting and main results 4 3 Weak relative compactness of tZε,δu 11 4 Asymptotics of the pseudo-characteristic operator of Zε,δ 16 5 Martingale problem and proof of Theorem 2.

The proof for the gliding dynamics Y follows a similar scheme using operator B and its associated martingale problem <ref:2608. p. 0
Where the paper says it

2020 Mathematics Subject Classification: 60J60, 60F17, 60G50, 82C41, 82C70 Contents 1 Introduction 1 2 Inhomogeneous multivariate finite-velocity motions: setting and main results 4 3 Weak relative compactness of tZε,δu 11 4 Asymptotics of the pseudo-characteristic operator of Zε,δ 16 5 Martingale problem and proof of Theorem 2. 3 19 6 Diffusion approximation for the gliding dynamics 22 1 Introduction The present work is motivated by the physical paradigm of active particles, i.

22514#pg8>.

Conclusion

The authors conclude that the two limiting diffusions X and Y have the same second-order part, but generally different drifts. The order-dependent part of the drift is expressed through their Lie brackets in Stratonovich form. p. 6

Where the paper says it

The cross terms DVirVj s, j ă i, reflect the prescribed ordering of the vector fields within one cycle. In Stratonovich form, the order-dependent part of the drift is expressed through their Lie brackets.

This confirms that microscopic ordering remains visible in the diffusion limit. p. 7
Where the paper says it

27) In particular, the orderings (3) and (5), corresponding to opposite orientations of the cycle, produce Liebracket drift terms with opposite signs. Thus, the orientation of the microscopic cyclic motion remains visible in the macroscopic diffusion limit.

p. 7
Where the paper says it

27) In particular, the orderings (3) and (5), corresponding to opposite orientations of the cycle, produce Liebracket drift terms with opposite signs. Thus, the orientation of the microscopic cyclic motion remains visible in the macroscopic diffusion limit.

p. 7
Where the paper says it

27) In particular, the orderings (3) and (5), corresponding to opposite orientations of the cycle, produce Liebracket drift terms with opposite signs. Thus, the orientation of the microscopic cyclic motion remains visible in the macroscopic diffusion limit.

The convergence results are established by classical weak-convergence methods. p. 2
Where the paper says it

, Lopusanschi and Simon (2020); Diamantakis and Woodfield (2025). We mention this connection only as an interpretation of the order-dependent drift; the convergence results of the present paper are established by classical weak-convergence methods.

p. 2
Where the paper says it

, Lopusanschi and Simon (2020); Diamantakis and Woodfield (2025). We mention this connection only as an interpretation of the order-dependent drift; the convergence results of the present paper are established by classical weak-convergence methods.

p. 2
Where the paper says it

, Lopusanschi and Simon (2020); Diamantakis and Woodfield (2025). We mention this connection only as an interpretation of the order-dependent drift; the convergence results of the present paper are established by classical weak-convergence methods.

How the training signal is built

The derivation of the SDE (5.17) yields dXt “ κ2µ ÿi σiVipXtq ˝ dWi t ´ κ2µ ÿjăi µiµj rVi, Vj spXtq dt. p. 6

Where the paper says it

20) can be rewritten in the Stratonovich form as dXt “ κ ? µ ÿ i σiVipXtq ˝ dWi t ´ κ 2 2µ ´ÿ i λ 2 i DVirVispXtq ` ÿ jăi µiµj rVi , Vj spXtq ¯ dt.

This is obtained by substituting the identity (5. p. 2
Where the paper says it

1 of Ethier and Kurtz (1986). Quantitative convergence estimates in Wasserstein distance were recently obtained by Barrera and Lukkarinen (2023), while weak convergence in the more general semi-Markov setting was established by Pedicone and Cinque (2026).

16) into the Itô–Stratonovich correction of SDE (2. p. 6
Where the paper says it

The SDE (2. 20) can be rewritten in the Stratonovich form as dXt “ κ ?

20).

Where the gains come from

The presence of order-dependent Lie-bracket drift in Stratonovich form is a key finding. This term, which arises from the non-commutativity and orientation of the microscopic cyclic dynamics, generates a macroscopic drift that depends on the cyclic order. p. 0

Where the paper says it

Thus, the generic non-commutativity of the microscopic motions survives the diffusive scaling and generates a macroscopic drift which depends on the cyclic order. The limiting drift also distinguishes the flight and gliding dynamics, despite their being driven by the same vector fields and run times.

This effect is illustrated by Example 2. p. 2
Where the paper says it

Thus, the non-commutativity and orientation of the microscopic cyclic dynamics may remain visible at the macroscopic scale. This effect is illustrated in Example 2.

5 where opposite orientations produce Liebracket drift terms with opposite signs. p. 7
Where the paper says it

26) and B p1q f “ B p2q f “ B p4q f “ B p6q f “ κ 2 4c ´ x∇2 fV, V y ` x∇2 fU, Uy ` x∇f, DUrUs ` DV rV sy¯ , B p3q f “ κ 2 4c ´ x∇2 fV, V y ` x∇2 fU, Uy ` x∇f, DUrUs ` DV rV s ` rU, V sy¯ , B p5q f “ κ 2 4c ´ x∇2 fV, V y ` x∇2 fU, Uy ` x∇f, DUrUs ` DV rV s ´ rU, V sy¯ . 27) In particular, the orderings (3) and (5), corresponding to opposite orientations of the cycle, produce Liebracket drift terms with opposite signs.

The structure of the limiting generator A and B demonstrates how the microscopic realization of the motion remains visible in the diffusion limit. p. 7
Where the paper says it

27) In particular, the orderings (3) and (5), corresponding to opposite orientations of the cycle, produce Liebracket drift terms with opposite signs. Thus, the orientation of the microscopic cyclic motion remains visible in the macroscopic diffusion limit.

Pruning redundant reasoning tokens

The second-order part of the limiting generator is determined by fluctuations of run times. The appearance of order-dependent Lie-bracket drift also bears a connection to second-order (area) corrections arising in Wong–Zakai approximations and rough-path invariance principles. p. 2

Where the paper says it

In contrast to a purely averaging regime, the effective dynamics retains information about the ordered composition of the spatially dependent vector fields through the drift terms described above. The appearance of order-dependent Lie-bracket drift also bears a connection to second-order (area) corrections arising in Wong–Zakai approximations and rough-path invariance principles.

The convergence results are established by classical weak-convergence methods. The convergence is equivalent to weak convergence with respect to the topology of uniform convergence on compact time intervals. p. 0
Where the paper says it

, Vp with random run times and evolves according to either a flight dynamics, consisting of piecewise-linear motion, or a gliding dynamics, in which the particle follows the corresponding velocity flow. Under a Kac-type scaling that couples vanishing run times with diverging particle speed, we prove weak convergence of both processes to multidimensional diffusions.

p. 0
Where the paper says it

, Vp with random run times and evolves according to either a flight dynamics, consisting of piecewise-linear motion, or a gliding dynamics, in which the particle follows the corresponding velocity flow. Under a Kac-type scaling that couples vanishing run times with diverging particle speed, we prove weak convergence of both processes to multidimensional diffusions.

This confirms that microscopic ordering remains visible in the diffusion limit.

The proof for Theorem 2.4 follows the same scheme as that of Theorem 2. p. 0

Where the paper says it

A particle alternates cyclically among prescribed velocity fields V1, . , Vp with random run times and evolves according to either a flight dynamics, consisting of piecewise-linear motion, or a gliding dynamics, in which the particle follows the corresponding velocity flow.

3. The only essential modification concerns the expansion of the p-step pseudo-characteristic operator, since in the gliding model the particle follows the flow of the current vector field during each run. p. 3
Where the paper says it

We distinguish between two types of motion: flights, corresponding to straight motion with velocity frozen at the beginning of each run, and gliding, corresponding to motion along the flow generated by the current vector field, see Fig. Assume that the particle starts at t “ 0 at position X0 “ x with the initial “direction” n0 “ ι P t1, .

The difference between limiting generators A and B originates from the evolution of each vector field along its own flow. In contrast to the flight model, this within-run variation produces the additional DVirVis contribution to the effective drift. p. 2

Where the paper says it

The order of the vector fields together with the distributions of the run times determines the microscopic switching dynamics, whereas the diffusion approximation under Kac’s scaling and a centering condition provides an effective description on the macroscopic scale. Besides obtaining explicit formulas for the limiting diffusions, a principal feature of our results is that the ordering of the cyclic dynamics survives in the diffusion limit through an additional drift determined by directional derivatives DVirVj s of the underlying vector fields.

This shows that even for the same vector fields, run times, and cyclic switching rule, the microscopic realization of the motion remains visible in the diffusion limit. p. 2
Where the paper says it

The order of the vector fields together with the distributions of the run times determines the microscopic switching dynamics, whereas the diffusion approximation under Kac’s scaling and a centering condition provides an effective description on the macroscopic scale. Besides obtaining explicit formulas for the limiting diffusions, a principal feature of our results is that the ordering of the cyclic dynamics survives in the diffusion limit through an additional drift determined by directional derivatives DVirVj s of the underlying vector fields.

The second-order part of the limiting generator is determined by fluctuations of run times. The appearance of order-dependent Lie-bracket drift also bears a connection to second-order (area) corrections arising in Wong–Zakai approximations and rough-path invariance principles <ref:2608. p. 2

Where the paper says it

In contrast to a purely averaging regime, the effective dynamics retains information about the ordered composition of the spatially dependent vector fields through the drift terms described above. The appearance of order-dependent Lie-bracket drift also bears a connection to second-order (area) corrections arising in Wong–Zakai approximations and rough-path invariance principles.

22514#pg8>. The convergence results are established by classical weak-convergence methods. The convergence is equivalent to weak convergence with respect to the topology of uniform convergence on compact time intervals. This confirms that microscopic ordering remains visible in the diffusion limit.

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. Improved AI can perform diffusion approximation for multivariate inhomogeneous finite-velocity random motions motivated by active particles by proving weak convergence of both flight and gliding dynamics to multidimensional diffusions under Kac-type scaling.

  2. The improved system can account for how the order of successive cyclic motions affects the diffusion limit, specifically by generating a limiting drift determined by directional derivatives DVirVj s of the underlying vector fields, which are expressed in Stratonovich form through Lie brackets rVi, Vj s.

  3. The system can distinguish between flight and gliding dynamics even when driven by the same vector fields and run times, as the limiting drift for each motion is generally different.

  4. The AI can model run-and-reverse motion along two non-commutative vector fields to show how microscopic ordering survives in the macroscopic diffusion limit through Lie bracket drift terms with opposite signs depending on cyclic orientation.

  5. The system can handle cases where vector fields commute by showing that the order-dependent Lie bracket contribution vanishes, simplifying the limiting operators and resulting in a standard Brownian motion covariance matrix for homogeneous velocities.

  6. The AI can generate a specific diffusion process for run-and-reverse motion where one component is tangent to the one-dimensional foliation generated by the vector field V, while the other acquires an additional radial drift with absolute value κ squared c DV rV s.

Abstract

We investigate diffusion approximations for a class of multivariate inhomogeneous finite-velocity random motions motivated by models of active particles. A particle alternates cyclically among prescribed velocity fields V 1,,V p with random run times and evolves according to either a flight dynamics, consisting of piecewise-linear motion, or a gliding dynamics, in which the particle follows the corresponding velocity flow. Under a Kac-type scaling that couples vanishing run times with diverging particle speed, we prove weak convergence of both processes to multidimensional diffusions. It turns out that the order of successive cyclic motions affects the diffusion limit. In addition to the second-order term generated by fluctuations of the random run times, the limiting drift contains directional derivatives DV i[V j] of the underlying vector fields. In Stratonovich form, part of this drift is expressed through their Lie brackets [V i,V j]. Thus, the generic non-commutativity of the microscopic motions survives the diffusive scaling and generates a macroscopic drift which depends on the cyclic order. The limiting drift also distinguishes the flight and gliding dynamics, despite their being driven by the same vector fields and run times. Several examples, including run-and-reverse motion and cyclic dynamics generated by multiple vector fields, illustrate how the cyclic switching protocol and the geometry of the underlying velocity fields shape the limiting diffusions.

Related papers