When do cross-diffusion systems have an entropy structure?

Ansgar J\"ungel; Xiuqing Chen

REVIEW 2 major objections 3 minor 28 references

When do cross-diffusion systems have an entropy structure?

T0 review · 2 major / 3 minor · reviewed 2026-08-14 · deepseek-v4-flash

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

Pith's one-line read The paper proves that normal ellipticity plus a symmetric Onsager matrix guarantees an entropy structure.

desk verdict A clean, correct paper on entropy structure for cross-diffusion; the n-species SKT normal ellipticity result is the real news, but the general characterization only runs inside the detailed-balance class and Remark 12 has a false determinant identity. read the letter →

arxiv 1908.06873 v1 pith:ELEFTF7T submitted 2019-08-19 math.AP

Xiuqing Chen , Ansgar Jüngel This is my paper · ORCID

classification math.AP MSC 35K4035K5535Q9235Q7915A2315A24

keywords CrossdiffusionentropymethodnormalellipticitymatrixfactorizationLyapunovequationpopulationmodelvolume-fillingfluidmixture

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

Cross-diffusion systems couple the gradient of one species into the flux of another, and their diffusion matrices are typically neither symmetric nor positive definite, which blocks the usual energy estimates. This paper asks when such a system can be rewritten in entropy variables so that the transformed diffusion matrix becomes positive definite, yielding a Lyapunov functional and the gradient estimates needed for global existence. The answer is a matrix-factorization criterion: normal ellipticity, meaning all eigenvalues of the diffusion matrix have positive real part, is necessary for an entropy structure, and it becomes sufficient once there is a strictly convex entropy density making the Onsager matrix symmetric. For constant or nearly constant diffusion matrices, the paper shows that entropy structure and normal ellipticity are equivalent. This matters because the entropy structure is the main analytical tool for proving global-in-time existence of solutions to these systems.

What carries the argument

The central mechanism is the factorization A(u) = A1A2 or A(u) = A2A1, with A1 symmetric positive definite. A matrix is normally elliptic when all eigenvalues have positive real part, and Lyapunov's matrix theorem says such matrices are exactly those admitting a factorization with A2 positive definite. If A2 can be chosen symmetric, the factorization forces A2 to be positive definite whenever A is normally elliptic, and taking A1 = h''(u)^{-1} turns A2 into h''(u)A(u). This factorization is the bridge between normal ellipticity and entropy structure; the detailed-balance condition pi_i partial p_i/partial u_j = pi_j partial p_j/partial u_i supplies the symmetry of the Onsager matrix in the model applications.

What would settle it

Take the 3-species SKT-type matrix from Lemma 21, A(u) = [[u3, 0, u1], [u2, u1, 0], [0, u3, u2]], and determine whether the system has an entropy structure. A proof that an entropy exists would show that Theorem 7(ii)'s symmetry assumption is not necessary; a proof that none exists would confirm the gap between normal ellipticity and entropy structure.

Watch

Extended reading notes

Core claim

The central claim is Theorem 7: if the system has an entropy structure with strictly convex entropy density h, then the diffusion matrix A(u) is normally elliptic for every u; conversely, if A(u) is normally elliptic and there exists a strictly convex h in $C^{2}$(D) with h''(u)A(u) symmetric for all u, then h''(u)A(u) is positive definite, so the system has an entropy structure. If in addition h''(u)A(u) is symmetric, A(u) is diagonalizable with positive eigenvalues. The proof factorizes A(u) = A1A2 with A1 = h''(u)^{-1} symmetric positive definite and A2 = h''(u)A(u); Lyapunov's matrix theorem and inertia arguments convert normal ellipticity of A into positive definiteness of A2. Applications show that the criterion reproduces the known Boltzmann entropies for the SKT population model and the volume-filling model, and it produces two entropies, one Boltzmann-type and one quadratic, for fluid mixtures with detailed balance. The paper also proves normal ellipticity of the n-species SKT diffusion matrix without detailed balance, a property not previously established.

Load-bearing premise

The sufficient direction of the main theorem assumes that a strictly convex entropy density exists making the Onsager matrix symmetric on the whole domain; in the model examples this symmetry is the detailed-balance condition on the coefficients, a structural restriction not derived from the equations.

Editorial extensions

If this is right

For any system satisfying Theorem 7(ii), the function t maps to integral of h(u(t)) is nonincreasing and the production term supplies gradient estimates, opening the boundedness-by-entropy route to global existence.
For constant diffusion matrices, normal ellipticity alone is equivalent to an entropy structure, with an explicit Lyapunov function h(u) = (1/2) u^T H u and H given by an integral of exp(-A^T t) exp(-A t).
A constant normally elliptic matrix perturbed by a small bounded nonlinear term still admits an entropy structure, so the criterion is stable under perturbation.
When the entropy density is a sum of single-variable functions and h''A is symmetric, positive definiteness is equivalent to positivity of all leading principal minors of A, avoiding eigenvalue computations.
For fluid mixtures with detailed balance and an invertible pressure Jacobian, two distinct entropy densities exist, one Boltzmann-type and one quadratic, giving complementary gradient estimates.

Reading between the lines

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

The determinant formula det H = 1/(2 tr A) for the constant-matrix Lyapunov function suggests that entropy production rates for nearly constant diffusion matrices could be estimated from the trace alone; a testable extension would compare such bounds with numerical solutions for slowly varying A(u).
The Poincare-lemma construction behind the second entropy indicates that any flux function F with invertible Jacobian and curl-free pi_i F_i field admits an entropy, which could be checked for reaction-cross-diffusion systems with lower-order terms.
The paper leaves open whether a normally elliptic matrix that is not diagonalizable with positive eigenvalues can still support an entropy structure with a nonsymmetric Onsager matrix; the 3-species SKT-type example from Lemma 21 is a concrete test case.

Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

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

Desk Editor's Note

read the letter

Short version: this is a solid, useful paper. The two results I didn't know were in the literature are the normal ellipticity of the n-species SKT diffusion matrix without detailed balance (Lemma 20) and the existence of a second, quadratic entropy for fluid mixtures under a curl-free condition (Proposition 17). The matrix-factorization framework around Theorem 7 is clean, and the central proofs are correct. The paper is honest about its main limitation: Theorem 7(ii) requires h''(u)A(u) to be symmetric, and in every application that symmetry comes from the detailed-balance condition, not from the equations themselves. The gap between normal ellipticity and entropy structure for nonsymmetric Onsager matrices is left open, and the paper says so explicitly in the discussion after Lemma 21. That is a limitation of scope, not a hidden flaw.

The one genuine error is in Remark 12. The formula det H = ∫ det(e^{-A^T t}e^{-A t})dt is not valid; determinants do not pass through integrals. For a concrete check, take A = diag(λ1, λ2), which gives det H = 1/(4λ1λ2) while the integral equals 1/(2(λ1+λ2)). This is a localized mistake in a remark, and nothing downstream depends on it, but it should be corrected or deleted.

The weakest part of the paper is the sufficiency side. The general sufficient condition rests on a structural symmetry assumption, so the title's "when" is answered only within the symmetric-Onsager class. The authors are upfront about this, and it doesn't undermine the main theorems. The citation pattern is appropriate: the factorization results are partly known, but the paper credits them and applies them to genuinely new cases.

My recommendation: send it to peer review. A competent referee can verify the main theorems quickly, and the paper will be useful to anyone working on entropy methods for cross-diffusion systems. I'd ask for a correction to Remark 12 and maybe a slightly more modest title, but the body holds up.

Referee Report

2 major / 3 minor

Summary. The paper studies when the diffusion matrix A(u) of a quasilinear cross-diffusion system (1) admits an entropy structure, i.e., a strictly convex h with h''(u)A(u) positive definite. The main results are: an entropy structure implies normal ellipticity of A(u) (Theorem 7(i)); if A(u) is normally elliptic and h''(u)A(u) is symmetric, then h''(u)A(u) is positive definite (Theorem 7(ii)); for constant or nearly constant diffusion matrices, normal ellipticity is equivalent to existence of an entropy structure (Propositions 11 and 13); and for entropies that are sums of single-variable functions, positive definiteness reduces to positivity of all leading principal minors of A(u) (Proposition 14). The paper also constructs entropies for SKT, volume-filling, and fluid-mixture models, and proves normal ellipticity of the SKT diffusion matrix without the detailed-balance condition (Lemma 20), together with related results.

Significance. The matrix-factorization approach is attractive and, for the most part, cleanly executed. Theorems 7, Propositions 2, 3, 6, 11, 13, 14, and Lemmas 20-22 appear correct and provide useful, checkable criteria. The new normal-ellipticity result for the n-species SKT model without detailed balance is a genuine contribution, and the paper is honest about the remaining open gap between normal ellipticity and entropy structure (see the example after Lemma 21). However, the 'second entropy' results in Section 6 are not justified as stated: the proofs establish positivity of h''(u)A(u) but do not verify that the constructed h is strictly convex, and the stated assumptions do not imply convexity.

major comments (2)

[Section 6, Propositions 17 and 19] The proofs show only that z^T h''(u)A(u)z > 0 for z != 0, but the definition of entropy structure also requires h to be strictly convex. The assumptions in Proposition 17 do not imply that h'' = diag(pi) Q is positive definite. For example, take n = 2, pi = (1,1), p1 = u1 + 2u2, p2 = 2u1 + u2. The detailed-balance condition (10) holds, Q = [[1,2],[2,1]] is invertible, and h''(u)A(u) = Q diag(u) Q is positive definite for u in R_+^2. However, the Poincare potential h = (1/2)u1^2 + 2u1u2 + (1/2)u2^2 has Hessian Q with eigenvalues 3 and -1, so h is not convex on any convex open domain. Proposition 19 has the same defect. The propositions need an additional assumption that ensures h'' is positive definite, such as normal ellipticity of Q as in Proposition 15, or a modified conclusion.
[Remark 12] The displayed identity det H = integral over [0,infinity) of det(e^{-A^T t}) det(e^{-A t}) dt is false, because the determinant does not commute with integration. For A = diag(lambda1, lambda2), the Lyapunov solution is H = diag(1/(2 lambda1), 1/(2 lambda2)), so det H = 1/(4 lambda1 lambda2), whereas the formula in the remark gives 1/(2(lambda1 + lambda2)). The formula H = integral_0^infty e^{-A^T t} e^{-A t} dt is correct, but the determinant consequence should be removed or corrected.

minor comments (3)

[Section 3, Case 1.1, Eq. (7)] For the system as written, the diffusion matrix is A = [[1, -u1], [1, delta]], but the displayed factorization A1 A2 equals [[1, -u1], [delta, 1]]. The reported eigenvalues 1 ± i sqrt(delta u1) correspond to the factorized matrix, not to the stated system. Please correct either the system or the factorization and entropy computation.
[Section 7, after Lemma 21] The example matrix [[u3,0,u1],[u2,u1,0],[0,u3,u2]] satisfies the conditions of Lemma 21 with the triple (3,1,2), not with (1,2,3), since b11 = b22 = b33 = 0 in this example.
[Throughout] There are several typographical errors: 'elipticity' in the abstract should be 'ellipticity'; 'symmetrix' in Case 2.1 should be 'symmetric'; 'tripel' in Lemma 21 should be 'triple'.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the paper's main sufficient condition is explicitly conditional on the symmetry of h''A (provided in applications by the detailed-balance assumption), and the proofs rest on external matrix theorems rather than on the conclusions.

full rationale

The paper does not fit parameters to data, rename a known result, or smuggle in its conclusion through a self-citation chain. The central results are proved from Lyapunov's theorem, Sylvester's inertia theorem, Gershgorin-type diagonal dominance, the Routh-Hurwitz criterion, the Bauer-Fike theorem, and the Poincaré lemma, all of which are external mathematical facts. Theorem 7(ii) assumes a strictly convex h with h''(u)A(u) symmetric and concludes positive definiteness under normal ellipticity; this is a conditional sufficient condition, not an unconditional derivation of an entropy structure from normal ellipticity alone. The symmetry requirement is supplied in the applications by the detailed-balance condition (10), which is stated as a structural assumption and is not claimed to follow from the equations. The paper explicitly identifies the remaining gap: after Lemma 21 it states, 'It is an open question whether (1) with this diffusion matrix has an entropy structure,' showing that no circular reduction is hiding the incompleteness. Self-citations, including [9], [17], [19], and [28], are used for context or for previously established existence results, while the new entropy-structure and normal-ellipticity statements are proved in the paper from the external matrix theory; consequently the self-citations are not load-bearing. In sum, the characterization is incomplete for nonsymmetric Onsager matrices, but incompleteness of a sufficiency condition is a limitation, not circularity.

Assumptions & free parameters 0 free parameters · 7 assumptions · 0 invented entities

The paper introduces no fitted parameters and no new postulated entities. It relies on standard matrix-analysis theorems and on structural assumptions about the diffusion coefficients, especially the detailed-balance condition, which are given as model hypotheses rather than derived. The only notable flaw is the determinant identity in Remark 12, which is an internal mathematical error and not an input assumption.

assumptions (7)

standard math Lyapunov theorem for matrix equations (Theorem 1), including existence and uniqueness of H solving HA + A^T H = G.
Invoked throughout Section 2 to derive the factorizations in Propositions 2, 3, and 6.
standard math Matrix factorization results from Bosch [5], Sylvester's inertia theorem [7], and eigenvalue bounds from [23].
Used to state Propositions 3 and 6 and the remark on eigenvalues in Section 2.
standard math Gershgorin circle theorem, Routh-Hurwitz criterion, and Bauer-Fike theorem from Horn and Johnson [15,16].
Used in Lemma 20, Lemma 21, and Proposition 9 for eigenvalue localization and perturbation bounds.
standard math Poincaré lemma for closed differential forms on simply connected domains.
Used in Propositions 17 and 19 to construct entropy densities from curl-free vector fields.
domain assumption Existence of a strictly convex entropy h in C^2(D) with positive definite Hessian and invertible derivative h'.
The entropy structure definition and the change of variables w = h'(u) both require this regularity; it is stated in the introduction and Section 3.
domain assumption Detailed-balance condition with positive weights pi_i for the SKT, fluid mixture, and general SKT models.
Used in Section 3 and Section 6 to guarantee symmetry of h''(u)A(u), which is the key sufficient condition in Theorem 7(ii).
domain assumption For volume-filling models in Lemma 22, positivity of pi, qi, q_i' on the Gibbs simplex D.
Needed to ensure the factor matrices A1 and A2 are symmetric positive definite and hence that A(u) is normally elliptic.

how reviews work

0 comments

Cite this review

Pith. "Pith review of When do cross-diffusion systems have an entropy structure?." pith.science (2026). https://pith.science/paper/ELEFTF7T

@misc{pith2026190806873,
  author       = {Pith},
  title        = {Pith review of: When do cross-diffusion systems have an entropy structure?},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/ELEFTF7T}},
  note         = {Machine review of arXiv:1908.06873}
}

abstract

Necessary and sufficient conditions for the existence of an entropy structure for certain classes of cross-diffusion systems with diffusion matrix $A(u)$ are derived, based on results from matrix factorization. The entropy structure is important in the analysis for such equations since $A(u)$ is typically neither symmetric nor positive definite. In particular, the normal ellipticity of $A(u)$ for all $u$ and the symmetry of the Onsager matrix implies its positive definiteness and hence an entropy structure. If $A$ is constant or nearly constant in a certain sense, the existence of an entropy structure is equivalent to the normal ellipticity of $A$. Several applications and examples are presented, including the $n$-species population model of Shigesada, Kawasaki, and Teramoto, a volume-filling model, and a fluid mixture model with partial pressure gradients. Furthermore, the normal elipticity of these models is investigated and some extensions are discussed.

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

28 extracted references · 27 canonical work pages

[1]

H. Amann. Dynamic theory of quasilinear parabolic equations. II. Reaction-diﬀusion systems. Diﬀ. Int. Eqs. 3 (1990), 13–75

work page 1990
[2]

H. Amann. Nonhomogeneous linear and quasilinear elliptic and parab olic boundary value problems. In: H.J. Schmeisser and H. Triebel (editors), Function Spaces, Diﬀerential Operators and Nonlinear Analysis, pp. 9–126. Teubner, Stuttgart, 1993

work page 1993
[3]

Boillat, C

G. Boillat, C. Dafermos, P. Lax, and T.-P. Liu. Recent Mathematical Methods in Nonlinear Wave Propagation. Lect. Notes Math. 1640. Springer, Berlin, 1996

work page 1996
[4]

D. Bothe. On the Maxwell-Stefan equations to multicomponent diﬀ usion. In: J. Escher et al. (eds.), Parabolic Problems. Progress in Nonlinear Diﬀerential Equ ations and their Applications , pp. 81–93. Springer, Basel, 2011

work page 2011
[5]

A. Bosch. Note on the factorization of a square matrix into two H ermitian or symmetric matrices. SIAM Review 29 (1987), 463–468. 22 X. CHEN AND A. J ¨UNGEL

work page 1987
[6]

Calvez and L

V. Calvez and L. Corrias. The parabolic-parabolic Keller–Segel mo del in R2. Commun. Math. Sci. 6 (2008), 417–447

work page 2008
[7]

Carlson and H

D. Carlson and H. Schneider. Inertia theorems for matrices: th e semi-deﬁnite case. J. Math. Anal. Appl. 6 (1963), 430–446

work page 1963
[8]

L. Chen, E. Daus, and A. J¨ ungel. Rigorous mean-ﬁeld limits and cr oss-diﬀusion. Z. Angew. Math. Phys. 70 (2019), no. 122, 21 pages

work page 2019

Show all 28 references

[9]

X. Chen, E. Daus, and A. J¨ ungel. Global existence analysis of cr oss-diﬀusion population systems for multiple species. Arch. Ration. Mech. Anal. 227 (2018), 715–747

2018
[10]

Corrias, B

L. Corrias, B. Perthame, and H. Zaag. Global solutions of some chemotaxis and angiogenesis systems in high space dimensions. Milan J. Math. 72 (2004), 1–28

2004
[11]

Desvillettes, T

L. Desvillettes, T. Lepoutre, A. Moussa, and A. Trescases. O n the entropic structure of reaction-cross diﬀusion systems. Commun. Partial Diﬀ. Eqs. 40 (2015), 1705–1747

2015
[12]

F. R. Gantmacher. The Theory of Matrices . Volume 2. Chelsea Publishing Company, New York, 1959

1959
[13]

B. Hall. Lie Groups, Lie Algebras, and Representations . Second edition. Springer, Cham, 2015

2015
[14]

Hittmeir and A

S. Hittmeir and A. J¨ ungel. Cross diﬀusion preventing blow up in th e two-dimensional Keller–Segel model. SIAM J. Math. Anal. 43 (2011), 997–1022

2011
[15]

Horn and C

R. Horn and C. Johnson. Topics in Matrix Analysis . First edition. Cambridge University Press, Cam- bridge, 1991

1991
[16]

Horn and C

R. Horn and C. Johnson. Matrix Analysis . Second edition. Cambridge University Press, Cambridge, 2013

2013
[17]

J¨ ungel

A. J¨ ungel. The boundedness-by-entropy method for cross -diﬀusion systems. Nonlinearity 28 (2015), 1963–2001

2015
[18]

J¨ ungel.Entropy Methods for Diﬀusive Partial Diﬀerential Equation s

A. J¨ ungel.Entropy Methods for Diﬀusive Partial Diﬀerential Equation s. SpringerBriefs in Mathemat- ics, Springer, 2016

2016
[19]

J¨ ungel

A. J¨ ungel. Cross-diﬀusion systems with entropy structure. In: K. Mikula, D. ˇSevˇ coviˇ c, and J. Urb´ an (eds.). Proceedings of Equadiﬀ 2017 Conference , Spektrum Stu Publishing, Bratislava, 2017, pp. 181– 190

2017
[20]

J¨ ungel and I

A. J¨ ungel and I. Stelzer. Existence analysis of Maxwell-Stefa n systems for multicomponent mixtures. SIAM J. Math. Anal. 45 (2013), 2421–2440

2013
[21]

Kawashima and Y

S. Kawashima and Y. Shizuta. On the normal form of the symmet ric hyperbolic-parabolic systems associated with the conservation laws. Tohoku Math. J. 40 (1988), 449–464

1988
[22]

Lepoutre and A

T. Lepoutre and A. Moussa. Entropic structure and duality fo r multiple species cross-diﬀusion systems. Nonlin. Anal. 159 (2017), 298–315

2017
[23]

Lu and C

L.-Z. Lu and C. E. M. Pearce. Some new bounds for singular value s and eigenvalues of matrix products. Ann. Oper. Research 98 (2000), 141–148

2000
[24]

Liero and A

M. Liero and A. Mielke. Gradient structures and geodesic conve xity for reaction-diﬀusion systems. Phil. Trans. Royal Soc. A. 371 (2005), 20120346, 28 pages, 2013

2005
[25]

Mielke, M

A. Mielke, M. Peletier, and D. Renger. A generalization of Onsage r’s reciprocity relations to gradient ﬂows with nonlinear mobility. J. Non-Equil. Thermodyn. 41 (2016), 141–149

2016
[26]

D. Serre. The structure of dissipative viscous system of cons ervation laws. Physica D 239 (2000), 1381–1386

2000
[27]

Shigesada, K

N. Shigesada, K. Kawasaki, and E. Teramoto. Spatial segrega tion of interacting species. J. Theor. Biol. 79 (1979), 83–99

1979
[28]

Zamponi and A

N. Zamponi and A. J¨ ungel. Analysis of degenerate cross-diﬀu sion population models with volume ﬁlling. Ann. Inst. H. Poincar´ e – Anal. non lin. 34 (2017), 1–29. (Erratum: 34 (2017), 789–792.) CROSS-DIFFUSION SYSTEMS WITH ENTROPY STRUCTURE 23 School of Mathematics (Zhuhai), ...

2017

Pith tools

Reviewed August 14, 2026 · model on record in the stance chip above.

[1] [1]

H. Amann. Dynamic theory of quasilinear parabolic equations. II. Reaction-diﬀusion systems. Diﬀ. Int. Eqs. 3 (1990), 13–75

work page 1990

[2] [2]

H. Amann. Nonhomogeneous linear and quasilinear elliptic and parab olic boundary value problems. In: H.J. Schmeisser and H. Triebel (editors), Function Spaces, Diﬀerential Operators and Nonlinear Analysis, pp. 9–126. Teubner, Stuttgart, 1993

work page 1993

[3] [3]

Boillat, C

G. Boillat, C. Dafermos, P. Lax, and T.-P. Liu. Recent Mathematical Methods in Nonlinear Wave Propagation. Lect. Notes Math. 1640. Springer, Berlin, 1996

work page 1996

[4] [4]

D. Bothe. On the Maxwell-Stefan equations to multicomponent diﬀ usion. In: J. Escher et al. (eds.), Parabolic Problems. Progress in Nonlinear Diﬀerential Equ ations and their Applications , pp. 81–93. Springer, Basel, 2011

work page 2011

[5] [5]

A. Bosch. Note on the factorization of a square matrix into two H ermitian or symmetric matrices. SIAM Review 29 (1987), 463–468. 22 X. CHEN AND A. J ¨UNGEL

work page 1987

[6] [6]

Calvez and L

V. Calvez and L. Corrias. The parabolic-parabolic Keller–Segel mo del in R2. Commun. Math. Sci. 6 (2008), 417–447

work page 2008

[7] [7]

Carlson and H

D. Carlson and H. Schneider. Inertia theorems for matrices: th e semi-deﬁnite case. J. Math. Anal. Appl. 6 (1963), 430–446

work page 1963

[8] [8]

L. Chen, E. Daus, and A. J¨ ungel. Rigorous mean-ﬁeld limits and cr oss-diﬀusion. Z. Angew. Math. Phys. 70 (2019), no. 122, 21 pages

work page 2019