betaFeedback

WorksMathematics

pp-harmonic polynomials

Abstract

We prove that every real polynomial pp-harmonic function on \Rn is affine when n≥3n\ge3 and p≥n−1/13p\ge n-1/13. The proof uses the Bernstein–Szegő inequality on great circles and a deficit estimate at the first Chebyshev node.

Introduction and statement of the result

Let 1<p<∞1 < p < \infty, p≠2p \ne2, and consider weak solutions of the pp-Laplace equation

Δpu=div⁡(∣Du∣p−2Du)=0in Rn.\Delta_p u = \operatorname{div}\left(|Du|^{p-2}Du\right) = 0 \qquad\text{in }\mathbb{R}^n.

Whether this equation admits a homogeneous polynomial solution of degree at least two is a longstanding open question, raised by Lewis [2] and recorded in its present form by Tkachev [6]. Lewis [2] ruled out homogeneous polynomial solutions of degree at least two in the plane, as well as solutions of degree two in every dimension. Tkachev [5] treated degree three by algebraic methods, and Lewis and Vogel [3] treated degree four, as well as degree five in three variables. In this paper we give an explicit range of pp, depending only on the dimension, in which nonexistence holds in every degree.

Write Sn−1={x∈Rn:∣x∣=1}S^{n-1} = \{x \in\mathbb{R}^n : |x| = 1\} and set

Ep(u)=∣Du∣2Δu+(p−2)⟨D2u Du,Du⟩.(1)\mathcal{E}_p(u) = |Du|^2 \Delta u + (p - 2)\langle D^2u\,Du,Du\rangle. \tag*{(1)}

If a nonconstant polynomial uu is a weak solution of Δpu=0\Delta_pu = 0, then uu is a classical solution on the open set {Du≠0}\{Du \ne0\}, so that Ep(u)\mathcal{E}_p(u) vanishes there. At points where Du=0Du = 0, both terms in (1.1) vanish. Thus Ep(u)=0\mathcal{E}_p(u) = 0 throughout Rn\mathbb{R}^n.

Theorem 1.1. Let n≥3n \ge3 and p≥n−1/13p \ge n - 1/13. Every real polynomial uu on Rn\mathbb{R}^n satisfying Ep(u)=0\mathcal{E}_p(u) = 0 is affine. In particular, every polynomial pp-harmonic function on Rn\mathbb{R}^n is affine.

Remark 1.2. The theorem rules out non-affine polynomial pp-harmonic functions throughout the range p≥np \ge n, with no restriction on the degree. This gives a finite-pp counterpart of the nonexistence of ∞\infty-harmonic homogeneous polynomials of degree at least two. See, for example [8]. The deficit estimate also allows the argument to cover the interval n−1/13≤p<nn - 1/13 \le p < n. The hypotheses ensure p>2p > 2, as required in the proof.

Remark 1.3. The constant 1/131/13 is chosen to keep the estimates explicit and the proof simple. A sharper evaluation of the final numerical estimates gives only a small improvement, with a threshold of approximately 0.0789. We do not claim that 1/131/13 is optimal for the problem.

2020 Mathematics Subject Classification. Primary 35J92; Secondary 26D05, 41A17, 42A05.

Key words and phrases. pp-harmonic functions, homogeneous polynomials, Bernstein inequality, Chebyshev nodes, nonexistence.

The plan of the paper is as follows. In § 2 we prove a deficit estimate at the first Chebyshev node (Lemma 2.1). In § 3, we prove Theorem 1.1 by evaluating the equation at a well-chosen point on the sphere, θ=π/m\theta= \pi/m.

A deficit estimate

The Bernstein–Szegő inequality states that a real trigonometric polynomial TT of degree at most m≥1m \ge1, with ∥T∥∞≤1\|T\|_\infty\le1, satisfies T2+(T′/m)2≤1T^2 + (T'/m)^2 \le1; it goes back to Szegő and to van der Corput and Schaake [7]. A proof can be found in [4 Theorem 3.7]. Applying it also to T′/mT'/m gives

T2+(T′)2m2≤1,(T′)2m2+(T′′)2m4≤1.(2)T^2 + \frac{(T')^2}{m^2} \le1,\qquad\frac{(T')^2}{m^2} + \frac{(T'')^2}{m^4} \le1. \tag*{(2)}

In particular, ∥T′′∥∞≤m2\|T''\|_\infty\le m^2.

The proof of the following lemma is an interpolation argument at the Chebyshev nodes, in the spirit of the proof by Duffin and Schaeffer of the Markov brothers’ inequality [1].

Lemma 2.1. Let m≥2m \ge2 be an integer, and let TT be a real trigonometric polynomial of degree at most mm, with ∥T∥∞≤1\|T\|_\infty\le1 and T(0)=1T(0) = 1. If d=m2+T′′(0)d = m^2 + T''(0), then

0≤1+T(±π/m)≤d(1−cos⁡(π/m))≤π2d2m2.(3)0 \le1 + T(\pm\pi/m) \le d(1-\cos(\pi/m)) \le\frac{\pi^2d}{2m^2}. \tag*{(3)}

Proof. Write P(cos⁡t)=(T(t)+T(−t))/2P(\cos t) = (T(t) + T(-t))/2, where PP is a polynomial of degree at most mm, and let Cm(cos⁡t)=cos⁡(mt)C_m(\cos t) = \cos(mt). Then ∣P∣≤1|P| \le1 on [−1,1][-1,1] and P(1)=1P(1) = 1. Differentiating P(cos⁡t)P(\cos t) twice at t=0t = 0 gives P′(1)=−T′′(0)P'(1) = -T''(0), and Cm′(1)=m2C_m'(1) = m^2, so that d=Cm′(1)−P′(1)d = C_m'(1) - P'(1). At the nodes xj=cos⁡(jπ/m)x_j = \cos(j\pi/m), 0≤j≤m0 \le j \le m, differentiation of the Lagrange interpolant of Cm−PC_m-P gives

d=∑j=1maj(1−(−1)jP(xj)),aj={2/(1−xj),1≤j<m,1/2,j=m.d = \sum_{j=1}^{m} a_j(1-(-1)^jP(x_j)),\qquad a_j = \begin{cases} 2/(1-x_j), & 1 \le j < m,\\ 1/2, & j=m. \end{cases}

The j=0j=0 term vanishes because (Cm−P)(1)=0(C_m-P)(1)=0. Indeed, for w(x)=(x2−1)Cm′(x)w(x)=(x^2-1)C_m'(x), the Lagrange cardinal polynomial ℓj\ell_j satisfies

ℓj′(1)=w′(1)w′(xj)(1−xj)=(−1)jaj,1≤j≤m.\ell_j'(1) = \frac{w'(1)}{w'(x_j)(1-x_j)} = (-1)^j a_j,\qquad1 \le j \le m.

The last equality follows by differentiating w(cos⁡t)=−msin⁡tsin⁡(mt)w(\cos t)=-m\sin t\sin(mt), taking limits at the endpoints. All summands are nonnegative. Taking j=1j=1 and writing h=π/mh=\pi/m, we get

d≥(1+T(h))+(1+T(−h))1−cos⁡h.d \ge\frac{(1+T(h))+(1+T(-h))}{1-\cos h}.

Each term in the numerator is nonnegative. The conclusion follows from 1−cos⁡h≤h2/21-\cos h \le h^2/2. □

Proof of the theorem

Proof of Theorem 1.1. Suppose that uu has degree m≥2m \ge2. The component of degree 3m−43m-4 in Ep(u)\mathcal{E}_p(u) is Ep(um)\mathcal{E}_p(u_m), where umu_m is the highest homogeneous part of uu. We may therefore assume that uu is homogeneous of degree mm and, after replacing uu by −u-u if necessary, normalize it so that

u(e)=max⁡Sn−1∣u∣=1.(4)u(e)=\max_{S^{n-1}}|u|=1. \tag*{(4)}

For a unit vector v⊥ev \perp e, put Tv(t)=u(ecos⁡t+vsin⁡t)T_v(t)=u(e\cos t+v\sin t). This is a trigonometric polynomial of degree at most mm. At ee, homogeneity and maximality give Du(e)=meDu(e)=me and D2u(e)[e,e]=m(m−1)D^2u(e)[e,e]=m(m-1). Since Tv′′(0)=D2u(e)[v,v]−⟨Du(e),e⟩T_v''(0)=D^2u(e)[v,v]-\langle Du(e),e\rangle, we obtain

dv:=m2+Tv′′(0)=D2u(e)[v,v]+m(m−1)≥0,d_v:=m^2+T_v''(0)=D^2u(e)[v,v]+m(m-1)\geq0,

by (2). Evaluating Ep(u)=0\mathcal{E}_p(u)=0 at ee and summing over any orthonormal basis v1,…,vn−1v_1,\ldots,v_{n-1} of e⊥e^\perp gives

∑i=1n−1dvi=(n−p)m(m−1).(5)\sum_{i=1}^{n-1}d_{v_i}=(n-p)m(m-1). \tag*{(5)}

Since every dvid_{v_i} is nonnegative, p≤np\leq n, and the theorem is proved for p>np>n. With δ=n−p\delta=n-p, we have 0≤δ≤1/130\leq\delta\leq1/13 and dv≤δm(m−1)d_v\leq\delta m(m-1) for every vv.

Write F(θ,v)=u(ecos⁡θ+vsin⁡θ)F(\theta,v)=u(e\cos\theta+v\sin\theta) and θ0=π/m\theta_0=\pi/m. By Lemma 2.1,

−1≤F(θ0,v)≤−1+π2δ2(6)-1\leq F(\theta_0,v)\leq-1+\frac{\pi^2\delta}{2} \tag*{(6)}

for every vv.

Choose v0v_0 maximizing F(θ0,⋅)F(\theta_0,\mathord\cdot) on the unit sphere in e⊥e^\perp. At this point DvF=0D_vF=0 and ΔvF≤0\Delta_vF\leq0, where these operators act on that sphere. Set

a=−F(θ0,v0),b=Fθm,c=Fθθm2,q=a2+b2.a=-F(\theta_0,v_0),\qquad b=\frac{F_\theta}{m},\qquad c=\frac{F_{\theta\theta}}{m^2},\qquad q=a^2+b^2.

All derivatives below are evaluated at (θ0,v0)(\theta_0,v_0). Since π2<10\pi^2<10, (2) and (6) imply

1≥a≥1−π2δ2>813,q≤1,b2+c2≤1.1\geq a\geq1-\frac{\pi^2\delta}{2}>\frac{8}{13},\qquad q\leq1,\qquad b^2+c^2\leq1.

Put s=1−b2s=\sqrt{1-b^2}, so that c≤sc\leq s and s≥a>0s\geq a>0.

In the coordinates x=r(ecos⁡θ+vsin⁡θ)x=r(e\cos\theta+v\sin\theta), with r>0r>0, 0<θ<π0<\theta<\pi, and vv in the unit sphere of e⊥e^\perp, the Laplacian reads

Δu=urr+n−1rur+1r2(uθθ+(n−2)cot⁡θ uθ+Δvusin⁡2θ).(7)\Delta u=u_{rr}+\frac{n-1}{r}u_r+\frac{1}{r^2}\left(u_{\theta\theta}+(n-2)\cot\theta\,u_\theta+\frac{\Delta_vu}{\sin^2\theta}\right). \tag*{(7)}

Since DvF=0D_vF=0, the gradient at x0=ecos⁡θ0+v0sin⁡θ0x_0=e\cos\theta_0+v_0\sin\theta_0 has only radial and polar components, and ∣Du∣2=m2q>0|Du|^2=m^2q>0; in particular, Du(x0)Du(x_0) lies in the plane Π=span⁡{e,v0}\Pi=\operatorname{span}\{e,v_0\}, so that only the restriction of D2uD^2u to Π\Pi enters ⟨D2u Du,Du⟩\langle D^2u\,Du,Du\rangle. Because Π\Pi is an affine plane, the second derivatives of uu along Π\Pi are those of the restriction rmF(θ,v0)r^mF(\theta,v_0), computed in polar coordinates: in the orthonormal frame {er,eθ}\{e_r,e_\theta\} at x0x_0,

Du=mFer+Fθeθ,D2u∣Π=(m(m−1)F(m−1)Fθ(m−1)FθFθθ+mF).(8)Du=mF e_r+F_\theta e_\theta,\qquad D^2u\big|_\Pi= \begin{pmatrix} m(m-1)F & (m-1)F_\theta\\ (m-1)F_\theta& F_{\theta\theta}+mF \end{pmatrix}. \tag*{(8)}

Substituting F=−aF=-a, Fθ=mbF_\theta=mb, Fθθ=m2cF_{\theta\theta}=m^2c in (7) and (8), and dividing Ep(u)=0\mathcal{E}_p(u)=0 by m2∣Du∣2=m4qm^2|Du|^2=m^4q, we obtain

0=Ep(u)m2∣Du∣2=c−(p−1+δm)a+(n−2)cot⁡θ0mb+(p−2)b2(c−a)q+ΔvFm2sin⁡2θ0.(9)0=\frac{\mathcal{E}_p(u)}{m^2|Du|^2} =c-\left(p-1+\frac{\delta}{m}\right)a+\frac{(n-2)\cot\theta_0}{m}b +(p-2)\frac{b^2(c-a)}{q}+\frac{\Delta_vF}{m^2\sin^2\theta_0}. \tag*{(9)}

Since p>2p > 2, the coefficient of cc is positive. Using c≤sc \le s, ΔvF≤0\Delta_v F \le0, and 0≤cot⁡θ0/m≤1/π0 \le\cot\theta_0/m \le1/\pi, we obtain

0≤s−(p−1)a+(n−2)∣b∣π+(p−2)b2(s−a)q0 \le s - (p - 1)a + \frac{(n - 2)|b|}{\pi} + (p - 2)\frac{b^2(s-a)}{q}

Moreover, s2−a2=1−qs^2-a^2=1-q and q+a2/q≥2aq+a^2/q \ge2a give

0≤b2(s−a)q=1+a2−q−a2/qs+a≤(1−a)22a=:A.0 \le\frac{b^2(s-a)}{q}=\frac{1+a^2-q-a^2/q}{s+a}\le\frac{(1-a)^2}{2a}=:A.

Put k=n−2≥1k=n-2\ge1. Since p−1=k+1−δp-1=k+1-\delta, p−2=k−δp-2=k-\delta, and s−(1−δ)a≥0s-(1-\delta)a\ge0, it follows that

0≤s−(p−1)a+k∣b∣π+(p−2)A≤k(s+∣b∣π−(2−δ)a+A).\begin{aligned} 0 &\le s-(p-1)a+k\frac{|b|}{\pi}+(p-2)A\\ &\le k\left(s+\frac{|b|}{\pi}-(2-\delta)a+A\right). \end{aligned}

Finally, Cauchy–Schwarz and π>3\pi>3 give s+∣b∣/π≤1+π−2<17/16s+|b|/\pi\le\sqrt{1+\pi^{-2}}<17/16. Using δ≤1/13\delta\le1/13 and a>8/13a>8/13, we conclude that

0≤s+∣b∣π−(2−δ)a+A<1716−200169+25208=−11352<0,0\le s+\frac{|b|}{\pi}-(2-\delta)a+A<\frac{17}{16}-\frac{200}{169}+\frac{25}{208}=-\frac{1}{1352}<0,

a contradiction. □

Declaration of generative AI and AI-assisted technologies in the manuscript preparation process

During the preparation of this work, the author(s) used LLM models from OpenAI and Anthropic for assistance in the discovery and verification of proofs. The author(s) reviewed and edited the output as needed and take full responsibility for the content of the published article. All mathematical results presented in this paper have been formally verified in Lean 4 using the Mathlib library. The Lean code and an explanation of the formalization process are available at https://github.com/na0b/pharmonicpoly-lean.

Acknowledgments. JM was partially supported by a gift from the Simons Foundation (Gift Number 962828). This publication is based upon work supported by King Abdullah University of Science and Technology (KAUST). AS acknowledges the hospitality of the mathematics department at the University of Pittsburg, where part of this research was conducted.

Department of Mathematics, University of Pittsburgh, 301 Thackeray Hall, Pittsburgh, PA 15260, USA

Email address: [email protected]

Applied Mathematics and Computational Sciences (AMCS), Computer, Electrical and Mathematical Sciences and Engineering Division (CEMSE), King Abdullah University of Science and Technology (KAUST), Thuwal, 23955-6900, Kingdom of Saudi Arabia

Email address: [email protected]

References

  1. [1]R. J. Duffin and A. C. Schaeffer, A refinement of an inequality of the brothers Markoff, Trans. Amer. Math. Soc. 50 (1941), 517–528.DOI
  2. [2]J. L. Lewis, Smoothness of certain degenerate elliptic equations, Proc. Amer. Math. Soc. 80 (1980), no. 2, 259–265.DOI
  3. [3]J. L. Lewis and A. Vogel, On p Laplace polynomial solutions, J. Anal. 24 (2016), 143–166.DOI
  4. [4]H. Queffélec and R. Zaïrou, On Bernstein’s inequality for polynomials, Anal. Math. Phys. 9 (2019), no. 3, 1181–1207.
  5. [5]V. G. Tkachev, On the non-vanishing property for real analytic solutions of the p-Laplace equation, Proc. Amer. Math. Soc. 144 (2016), no. 6, 2375–2382.DOI
  6. [6]J. L. Lewis and V. G. Tkachev, Homogeneous polynomial p-harmonic functions, SciLag open problem P-181114.2 (2018). https://www.scilag.net/problem/P-181114.2.
  7. [7]J. G. van der Corput and G. Schaake, Ungleichungen für Polynome und trigonometrische Polynome, Compositio Math. 2 (1935), 321–361.
  8. [8]Y. Yu, A remark on C² infinity-harmonic functions, Electronic Journal of Differential Equations, No. 122, pp. 1–4, 2006.

Paper details

vlnc:W_ghv5br2y
Version
v1 · latest
Published
Fresh — added to the Hub within the last 30 days
Added
License
CC BY 4.0
Usage

views · 3

downloads · 0

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

  • Mathematics

LLM models from OpenAI and Anthropic

View statement
Looking for related papers…
Contents