betaFeedback

WorksMathematics

The polyharmonic Hénon–Lane–Emden conjecture

Abstract

We study positive classical solutions of the polyharmonic Hénon–Lane–Emden system (−Δ)mu=∣x∣αvp(-Δ)^m u=|x|^αv^p, (−Δ)mv=∣x∣βuq(-Δ)^m v=|x|^βu^q in Rn\mathbb{R}^n, where m≥2m\ge2, p,q≥1p,q\ge1, pq>1pq>1, and α,β≥0α,β\ge0. We prove that there is no such solution whenever n+αp+1+n+βq+1>n−2m\frac{n+α}{p+1}+\frac{n+β}{q+1}>n-2m, in every dimension and without any assumption on the behavior of the solutions at infinity. Combined with an existence result of Villavert, this proves the polyharmonic Hénon–Lane–Emden conjecture of Phan for m≥2m\ge2: for these exponents and weights, a positive classical solution exists if and only if n+αp+1+n+βq+1≤n−2m\frac{n+α}{p+1}+\frac{n+β}{q+1}\le n-2m. The proof has two parts. We first show that positivity forces (−Δ)iu>0(-Δ)^i u>0 and (−Δ)iv>0(-Δ)^i v>0 for 1≤i≤m−11\le i\le m-1. For mm even this needs pq>1pq>1, and explicit radial examples show that it fails when pq<1pq<1 and when p=q=1p=q=1. The system is then equivalent to a weighted integral system with a Riesz kernel of order 2m2m. For such integral systems we prove the corresponding Liouville theorem for every Riesz order μ∈(1,n)μ\in(1,n), adapting a recent argument for the second-order case that combines a localized virial identity with an interval comparison along source layers.

Introduction

We are concerned with positive solutions of the system

(−Δ)mu=∣x∣αvp,(−Δ)mv=∣x∣βuqin Rn,(1)(-\Delta)^m u=|x|^\alpha v^p,\qquad(-\Delta)^m v=|x|^\beta u^q\qquad\text{in }\mathbb{R}^n, \tag*{(1)}

where m≥2m\ge2 is an integer and p,q>0p,q>0, α,β∈R\alpha,\beta\in\mathbb{R}. The scaling uR(x)=Rσ1u(Rx)u_R(x)=R^{\sigma_1}u(Rx), vR(x)=Rσ2v(Rx)v_R(x)=R^{\sigma_2}v(Rx) with

σ1=2m+α+p(2m+β)pq−1,σ2=2m+β+q(2m+α)pq−1(2)\sigma_1=\frac{2m+\alpha+p(2m+\beta)}{pq-1},\qquad\sigma_2=\frac{2m+\beta+q(2m+\alpha)}{pq-1} \tag*{(2)}

leaves (1) invariant when pq≠1pq\ne1, and it singles out the critical hyperbola

n+αp+1+n+βq+1=n−2m.(3)\frac{n+\alpha}{p+1}+\frac{n+\beta}{q+1}=n-2m. \tag*{(3)}

Phan [14] conjectured that, for p,q≥1p,q\ge1, pq≠1pq\ne1, and α,β≥0\alpha,\beta\ge0, the system (1) has no positive classical solution if and only if (p,q)(p,q) lies below the critical hyperbola. The nonexistence half, which is [14 Conjecture A] and is also stated by Fazly [6 Conjecture 1], asserts that there is no positive classical solution when

n+αp+1+n+βq+1>n−2m.(4)\frac{n+\alpha}{p+1}+\frac{n+\beta}{q+1}>n-2m. \tag*{(4)}

Fazly proved it for radial solutions in every dimension and for bounded solutions in dimension n=2m+1n=2m+1 [6]. Phan proved it for bounded solutions when, in addition, max⁡{2m(p+1),2m(q+1)}≥(n−2m−1)(pq−1)\max\{2m(p+1),2m(q+1)\}\ge(n-2m-1)(pq-1), and without the boundedness assumption in a smaller range of exponents [14]. Liouville theorems for integral and polyharmonic systems have also been obtained by the method of scaling spheres [5, 7]. Arthur and Yan obtained further results under a slow-decay assumption [2]. More recently, Xu and Luo proved a super polyharmonic property and Liouville theorems for radial solutions subject to the growth condition u,v=O(∣x∣2)u,v=O(|x|^2) [20]. These restrictions are absent from the result below. In higher dimensions the general unrestricted case has remained open.

The second-order case m=1m=1 has a longer history. For α=β=0\alpha=\beta=0 it is the Lane–Emden conjecture. Mitidieri introduced a Rellich identity for systems and obtained the first nonexistence regions [10, 11]. Serrin and Zou settled dimension three under a growth condition [16]. Poláčik, Quittner, and Souplet removed that condition [15], and Souplet proved the conjecture in dimension four [17]. Phan formulated and partly proved the weighted version [13], and Li and Zhang proved it in dimension three [8]. In each of these works the decisive step is an estimate for the boundary terms of a Rellich–Pohozaev identity, and the available Sobolev-type estimates control them only in low dimensions. A recent preprint [12] removes this restriction for m=1m = 1 and every nn. Its main estimate does not bound the boundary term of the localized Pohozaev identity. Instead, it shows that this term is dominated by an interaction term of favorable sign, up to an error εE+Cε\varepsilon E + C_{\varepsilon} relative to the localized energy EE. The domination comes from an elementary comparison between pairs of intervals on parallel lines, applied to the level sets of the sources.

That argument uses the equation only through the Newton potential representation u=Γ∗(∣x∣αvp)u = \Gamma* (|x|^{\alpha}v^{p}), v=Γ∗(∣x∣βuq)v = \Gamma* (|x|^{\beta}u^{q}), together with the fact that the kernel is a power of ∣x∣|x|. Our first observation is that it goes through for a Riesz kernel of any order μ∈(1,n)\mu\in(1,n). Mei and Wei carried out this extension for the unweighted fractional system, that is, for Riesz kernels of order 2s∈(0,2)2s \in(0,2) and α=β=0\alpha= \beta= 0 [9]. When 2s≤12s \le1 the kernel gradient is not locally integrable, and they combine the two equations before removing the diagonal. The present problem requires weights and orders μ=2m≥4\mu= 2m \ge4. The natural route for m≥2m \ge2 is therefore to prove that positive solutions of (1) are Riesz potentials of order 2m2m.

This is where the polyharmonic problem differs from the second-order one. When m=1m = 1, a positive solution is superharmonic and the representation follows from the Riesz decomposition of superharmonic functions. When m≥2m \ge2, positivity of uu says nothing directly about the sign of −Δu-\Delta u, and the representation can fail. In fact, for mm even and pq<1pq < 1, or for p=q=1p = q = 1, the system (1) with α=β=0\alpha= \beta= 0 has positive radial solutions with Δu>0\Delta u > 0 everywhere, although (4) holds (Proposition 1.4 below). The property that has to be proved is the super polyharmonic property

(−Δ)iu>0,(−Δ)iv>0in Rn,1≤i≤m−1.(-\Delta)^i u > 0,\qquad(-\Delta)^i v > 0 \qquad\text{in } \mathbb{R}^n,\quad1 \le i \le m - 1.

For the scalar equation it goes back to Wei and Xu [21]. For systems (−Δ)kui=fi(u1,…,ul)(-\Delta)^k u_i = f_i(u_1,\ldots,u_l) without weights, Chen and Li proved it and deduced the equivalence with the corresponding integral system when 2k<n2k < n [4], and Xu and Luo obtained it for radial solutions of weighted systems under a growth condition [20]. The weights ∣x∣α|x|^\alpha and ∣x∣β|x|^\beta are not centered at the points where one averages, and we give a self-contained proof, without symmetry or growth assumptions, that also covers the endpoint cases p=1p = 1 or q=1q = 1. Spherical averages and a parity argument give the property at once except in one case: mm even and (−Δ)m−1u(-\Delta)^{m-1}u negative somewhere. There the averages of uu and vv grow, but the parity count produces no contradiction. We show instead that the growth exponents feed each other and increase geometrically when pq>1pq > 1, which is impossible for a finite function. The example above shows that pq>1pq > 1 cannot be dropped here.

Our main result settles the nonexistence half of the conjecture.

Theorem 1.1. Let m≥2m \ge2 and n≥1n \ge1 be integers, let p,q≥1p,q \ge1 with pq>1pq > 1, and let α,β≥0\alpha,\beta\ge0. If (4) holds, then (1) has no solution u,v∈C2m(Rn)u,v \in C^{2m}(\mathbb{R}^n) with u>0u > 0 and v>0v > 0 in Rn\mathbb{R}^n.

Together with the existence theorem of Villavert [19], this settles the conjecture of Phan for m≥2m \ge2.

Corollary 1.2. Let m≥2m \ge2 and n≥1n \ge1 be integers, let p,q≥1p,q \ge1 with pq≠1pq \ne1, and let α,β≥0\alpha,\beta\ge0. Then (1) has a solution u,v∈C2m(Rn)u,v \in C^{2m}(\mathbb{R}^n) with u>0u > 0 and v>0v > 0 in Rn\mathbb{R}^n if and only if

n+αp+1+n+βq+1≤n−2m.\frac{n+\alpha}{p+1}+\frac{n+\beta}{q+1}\le n-2m.

In that case the solution can be chosen radial.

The radial solutions of [19] are constructed by shooting and are stated there to be of class C2mC^{2m} away from the origin. We check in § 3 that they are classical solutions on all of Rn\mathbb{R}^n.

No boundedness, radial symmetry, decay, or energy assumption is made in Theorem 1.1. Taking u=vu=v, p=qp=q, and α=β\alpha=\beta gives the scalar statement: for m≥2m\ge2, p>1p>1, α≥0\alpha\ge0, and n≥1n\ge1, the equation (−Δ)mu=∣x∣αup(-\Delta)^m u=|x|^\alpha u^p has no positive classical solution in Rn\mathbb{R}^n when p<n+2m+2αn−2mp<\frac{n+2m+2\alpha}{n-2m}, with the convention that the bound is +∞+\infty when n≤2mn\le2m. For n>2mn>2m this scalar statement was proved earlier, for α>−2m\alpha>-2m and 0<p0<p below the same bound, by Dai and Qin with the method of scaling spheres [5]. When n≤2mn\le2m, condition (4) holds automatically, and Theorem 1.1 says that there is no positive solution at all. The hypotheses p,q≥1p,q\ge1 and α,β≥0\alpha,\beta\ge0 are those of the conjecture. They enter the proof through the super polyharmonic property, which is the next statement, and α,β≥0\alpha,\beta\ge0 also makes the right sides of (1) continuous at the origin.

Theorem 1.3. Let m≥2m\ge2 and n≥1n\ge1 be integers, let p,q≥1p,q\ge1 and α,β≥0\alpha,\beta\ge0, and let u,v∈C2m(Rn)u,v\in C^{2m}(\mathbb{R}^n) be positive solutions of (1). If mm is even, assume also pq>1pq>1. Then

(−Δ)iu>0and(−Δ)iv>0in Rn,1≤i≤m−1.(-\Delta)^i u>0\quad\text{and}\quad(-\Delta)^i v>0\quad\text{in }\mathbb{R}^n,\qquad1\le i\le m-1.

Proposition 1.4. Let m≥2m\ge2 be even, n≥1n\ge1, and α=β=0\alpha=\beta=0. Suppose that either p,q>0p,q>0 with pq<1pq<1, or p=q=1p=q=1. Then (1) has a positive radial solution u,v∈C2m(Rn)u,v\in C^{2m}(\mathbb{R}^n) with Δu≥1\Delta u\ge1 and Δv≥1\Delta v\ge1 in Rn\mathbb{R}^n. Moreover, (p,q)(p,q) satisfies (4).

We do not know whether Theorem 1.3 holds when pq>1pq>1 and min⁡{p,q}<1\min\{p,q\}<1. The proof below uses Jensen’s inequality for the averages of uqu^q and vpv^p, and this is the essential reason for the restriction p,q≥1p,q\ge1.

Once Theorem 1.3 is available, uu and vv solve an integral system, and Theorem 1.1 follows from the next result with μ=2m\mu=2m. We state it for a general Riesz order because the proof is the same.

References

  1. [1]D. H. Armitage and S. J. Gardiner, Classical Potential Theory, Springer Monographs in Mathematics, Springer, London, 2001. 5, 6DOI
  2. [2]F. Arthur and X. Yan, A Liouville-type theorem for higher order elliptic systems of Hénon–Lane–Emden type, Commun. Pure Appl. Anal. 15 (2016), 807–830. 1DOI
  3. [3]W. Chen and C. Li, An integral system and the Lane–Emden conjecture, Discrete Contin. Dyn. Syst. 24 (2009), 1167–1184. 3DOI
  4. [4]W. Chen and C. Li, Super polyharmonic property of solutions for PDE systems and its applications, Commun. Pure Appl. Anal. 12 (2013), 2497–2514. 2arxiv.org/abs/1110.2539
  5. [5]W. Dai and G. Qin, Liouville type theorems for fractional and higher order Hénon–Hardy type equations via the method of scaling spheres, Int. Math. Res. Not. IMRN 2023 (2023), 9001–9070. 1, 3DOI
  6. [6]M. Fazly, Liouville theorems for the polyharmonic Hénon–Lane–Emden system, Methods Appl. Anal. 21 (2014), 265–282. 1, 5arxiv.org/abs/1308.0073
  7. [7]P. Le, Method of scaling spheres for integral and polyharmonic systems, J. Differential Equations 298 (2021), 132–158. 1DOI
  8. [8]K. Li and Z. Zhang, Proof of the Hénon–Lane–Emden conjecture in R³, J. Differential Equations 266 (2019), 202–226. 2DOI
  9. [9]L. Mei and J. Wei, A proof of the fractional Lane–Emden conjecture, preprint, arXiv:2610.10188, 2026. 2, 3arxiv.org/abs/2610.10188
  10. [10]E. Mitidieri, A Rellich type identity and applications, Comm. Partial Differential Equations 18 (1993), 125–151. 2
  11. [11]E. Mitidieri, Nonexistence of positive solutions of semilinear elliptic systems in ℝⁿ, Differential Integral Equations 9 (1996), 465–479. 2
  12. [12]OpenAI, The subcritical Hénon–Lane–Emden conjecture, preprint, September 24, 2026, https://github.com/openai/math/blob/main/preprints/The-Subcritical-Henon-Lane-Emden-Conjecture-September-24-2026/paper.pdf. 2, 3, 6, 7, 8, 9, 10, 11, 12DOI
  13. [13]Q. H. Phan, Liouville-type theorems and bounds of solutions for Hardy–Hénon elliptic systems, Adv. Differential Equations 17 (2012), 605–634. 2arxiv.org/abs/1108.1312
  14. [14]Q. H. Phan, Liouville-type theorems for polyharmonic Hénon–Lane–Emden system, Adv. Nonlinear Stud. 15 (2015), 415–432. 1DOI
  15. [15]P. Poláčik, P. Quittner, and P. Souplet, Singularity and decay estimates in superlinear problems via Liouville-type theorems, I: Elliptic equations and systems, Duke Math. J. 139 (2007), 555–579. 2DOI
  16. [16]J. Serrin and H. Zou, Non-existence of positive solutions of Lane–Emden systems, Differential Integral Equations 9 (1996), 635–653. 2DOI
  17. [17]P. Souplet, The proof of the Lane–Emden conjecture in four space dimensions, Adv. Math. 221 (2009), 1409–1427. 2DOI
  18. [18]E. M. Stein, Singular Integrals and Differentiability Properties of Functions, Princeton University Press, Princeton, NJ, 1970. 6
  19. [19]J. Villavert, Shooting with degree theory: analysis of some weighted poly-harmonic systems, J. Differential Equations 257 (2014), 1148–1167. 2, 7DOI
  20. [20]Y. Xu and H. Luo, Liouville type theorems for higher order elliptic systems, J. Differential Equations 453 (2026), 113823. 1, 2DOI
  21. [21]J. Wei and X. Xu, Classification of solutions of higher order conformally invariant equations, Math. Ann. 313 (1999), 207–228. 2DOI

Paper details

vlnc:W_mg93csrw
Version
v1 · latest
Published
Fresh — added to the Hub within the last 30 days
Added
License
CC BY 4.0
Usage

views · 0

downloads · 0

A view counts once per browser per day, a download each time the PDF is requested.

  • Mathematics

AI-assisted tools

View statement
Looking for related papers…
Contents