betaFeedback

WorksPhysics and Astronomy

The Maxwell Conjecture is False

Abstract

We exhibit a configuration of five point charges in Euclidean space whose electrostatic potential admits at least 24 critical points all of which are non-degenerate. Maxwell's conjecture that the field of nn point charges has at most (n−1)2(n-1)^2 critical points which are all non-degenerate is therefore false.

Introduction

In J. C. Maxwell’s 1873 treatise on electricity and magnetism he discusses the number of equilibria of the electric field generated by nn point charges [5 §113]. Apparently unaware of this, M. Morse and S. S. Cairns in 1969 posed the problem of finding an upper bound for the number of equilibria [6 p. 293]. The first general bounds were supplied by A. Gabrielov, D. Novikov, and B. Shapiro in [3] who, based on their reading of [5 §113], formulated the ‘Maxwell conjecture’ which states that if the critical points of the electrostatic potential generated by nn point charges are all non-degenerate then their number cannot exceed (n−1)2(n-1)^2. These bounds were later improved by V. Zolotov in 2023 [8] and further improved by H. Edelsbrunner, C. Fillmore, and G. Oliveira in 2026 [2]. Maxwell’s bound is trivially achieved for n=2n=2 but it is not known even for n=3n=3 if 4 is the maximum number, except in the case of equal charges [7]. Further related problems in classical electrostatics are discussed in [1].

In this note we produce a counterexample to Maxwell’s conjecture: a configuration of five point charges with at least 24 critical points all of which are non-degenerate. The counterexample starts with three unit charges placed at the vertices of an equilateral triangle. This configuration has four equilibria: one at the centre and three displaced inwards around the edges of the triangle, as shown in Figure 1(a). Two small charges are then added to the centre of the triangle and moved slightly apart along the orthogonal axis to form a shallow triangular bipyramid. The three edge equilibria persist under the addition of the small axial charges whereas the central equilibrium bifurcates into a family of 21 equilibria.

Figure 1. (a) The electric field generated by an equilateral triangle with no axial charges. The next three figures give numerical evidence for 24 equilibria when two axial charges are added with ε=1/6\varepsilon= 1/6: (b) two off-planar equilibria in the positive (x,z)(x,z)-quadrant of the y=0y = 0 plane; (c) the potential along the zz-axis showing 3 axial equilibria; (d) the potential along the ray z=0z = 0, θ=π\theta= \pi (with split axes) showing 3 planar equilibria.

Acknowledgments

Kvalheim was supported in part by the Air Force Office of Scientific Research under award number FA9550-24-1-0299.

Tool and computational resource disclosure

The idea behind this construction was suggested by an LLM (OpenAI’s GPT-5.6 Sol). The authors have verified the mathematical details and have written the argument in their own words. Computer algebra software (Mathematica, Maple) was used to verify computations and produce visualisations.

Counterexample

Place three unit charges at the vertices of an equilateral triangle

a1=(100),a2=(−1/23/20),a3=(−1/2−3/20).\mathbf{a}_1 = \begin{pmatrix} 1 \\ 0 \\ 0 \end{pmatrix}, \quad\mathbf{a}_2 = \begin{pmatrix} -1/2 \\ \sqrt{3}/2 \\ 0 \end{pmatrix}, \quad\mathbf{a}_3 = \begin{pmatrix} -1/2 \\ -\sqrt{3}/2 \\ 0 \end{pmatrix}.

The Coulomb potential generated by these charges has the following Taylor expansion around the origin

V△=∑j=131∥x−aj∥=3+34H2+158H3+964H4+O(∥x∥5)V_{\triangle} = \sum_{j=1}^{3} \frac{1}{\lVert\mathbf{x} - \mathbf{a}_j \rVert} = 3 + \frac{3}{4}H_2 + \frac{15}{8}H_3 + \frac{9}{64}H_4 + O(\lVert\mathbf{x}\rVert^5)

where

H2=r2−2z2,H_2 = r^2 - 2z^2,
H3=x3−3xy2,H_3 = x^3 - 3xy^2,
H4=3r4−24r2z2+8z4.H_4 = 3r^4 - 24r^2z^2 + 8z^4.

are harmonic polynomials with r2=x2+y2r^2=x^2+y^2. Place two additional qεq_\varepsilon-charges along the triangle’s axis of symmetry at

a4=(00ε),a5=(00−ε).\mathbf{a}_4=\begin{pmatrix}0\\0\\\varepsilon\end{pmatrix},\qquad\mathbf{a}_5=\begin{pmatrix}0\\0\\-\varepsilon\end{pmatrix}.

The potential generated by the axial charges has the expansion

V±ε=∑j=45qε∥x−aj∥=qεε−2qεε3H2+qε4ε5H4+O(qε∥x∥6ε7).V_{\pm}^{\varepsilon}=\sum_{j=4}^{5}\frac{q_\varepsilon}{\lVert\mathbf{x}-\mathbf{a}_j\rVert}=\frac{q_\varepsilon}{\varepsilon}-\frac{2q_\varepsilon}{\varepsilon^3}H_2+\frac{q_\varepsilon}{4\varepsilon^5}H_4+O\left(\frac{q_\varepsilon\lVert\mathbf{x}\rVert^6}{\varepsilon^7}\right).

Theorem 1. There is an ε0>0\varepsilon_0>0 such that for all 0<ε<ε00<\varepsilon<\varepsilon_0 the potential Vε=V△+V±εV_\varepsilon=V_\triangle+V_{\pm}^{\varepsilon} generated by five charges located at a1,…,a5\mathbf{a}_1,\ldots,\mathbf{a}_5 with respective strengths 1,1,1,qε,qε1,1,1,q_\varepsilon,q_\varepsilon and

qε=34ε3−532ε5(1)q_\varepsilon=\frac{3}{4}\varepsilon^3-\frac{5}{32}\varepsilon^5 \tag*{(1)}

admits at least 24 non-degenerate critical points. Moreover, a small perturbation to the charge strengths gives a potential with a finite number k≥24k\geq24 of critical points all of which are non-degenerate.

To prove Theorem 1, we introduce a deformed potential Φε\Phi_\varepsilon, defined as

Φε(X)=Vε(ε2X)−Vε(0)ε6,for ε>0.\Phi_\varepsilon(\mathbf{X})=\frac{V_\varepsilon(\varepsilon^2\mathbf{X})-V_\varepsilon(0)}{\varepsilon^6},\qquad\text{for }\varepsilon>0.

Lemma 1. Let qε=34ε3−532ε5q_\varepsilon=\frac{3}{4}\varepsilon^3-\frac{5}{32}\varepsilon^5 as in (1) and let

Φ0=532H2+158H3+316H4.\Phi_0=\frac{5}{32}H_2+\frac{15}{8}H_3+\frac{3}{16}H_4.

Then there exists an open neighbourhood UU of {0}×R3\{0\}\times\mathbb{R}^3 and a function Ψ\Psi real-analytic on UU such that

Φε(X)=Φ0(X)+ε2Ψ(ε2,X),\Phi_\varepsilon(\mathbf{X})=\Phi_0(\mathbf{X})+\varepsilon^2\Psi(\varepsilon^2,\mathbf{X}),

whenever ε>0\varepsilon>0 and (ε2,X)∈U(\varepsilon^2,\mathbf{X})\in U. Consequently, for every k≥0k\geq0, and every K⊂R3K\subset\mathbb{R}^3 compact we have ∥Φε−Φ0∥Ck(K)≤CK,kε2\lVert\Phi_\varepsilon-\Phi_0\rVert_{C^k(K)}\leq C_{K,k}\varepsilon^2 for all sufficiently small ε>0\varepsilon>0.

Proof. Let K⊂R3K\subset\mathbb{R}^3 be any compact set. Write t=ε2t=\varepsilon^2. Since V△V_\triangle is real analytic near 00 and the HjH_j are homogeneous, we have

V△(tX)−V△(0)=34t2H2+158t3H3+964t4H4+t5R△(t,X)V_\triangle(t\mathbf{X})-V_\triangle(0)=\frac{3}{4}t^2H_2+\frac{15}{8}t^3H_3+\frac{9}{64}t^4H_4+t^5R_\triangle(t,\mathbf{X})

where RΔR_{\Delta} is real-analytic near {0}×K\{0\} \times K. For the axial pair,

V±ε(tX)−V±ε(0)=t(34−532t)F(ε,X).V^{\varepsilon}_{\pm}(tX)-V^{\varepsilon}_{\pm}(0)=t\left(\frac{3}{4}-\frac{5}{32}t\right)F(\varepsilon,X).

where

F(ε,X)=11−2εZ+ε2∣X∣2+11+2εZ+ε2∣X∣2−2.F(\varepsilon,X)=\frac{1}{\sqrt{1-2\varepsilon Z+\varepsilon^{2}\lvert X\rvert^{2}}}+\frac{1}{\sqrt{1+2\varepsilon Z+\varepsilon^{2}\lvert X\rvert^{2}}}-2.

This function is real analytic and even in ε\varepsilon and its Taylor expansion is

F(ε,X)=−ε2H2(X)+ε44H4(X)+ε6R±(ε2,X).F(\varepsilon,X)=-\varepsilon^{2}H_{2}(X)+\frac{\varepsilon^{4}}{4}H_{4}(X)+\varepsilon^{6}R_{\pm}(\varepsilon^{2},X).

Substituting ε2=t\varepsilon^{2}=t and adding the VΔV_{\Delta} and V±εV^{\varepsilon}_{\pm} terms gives

Vε(tX)−Vε(0)=t3(532H2+158H3+316H4)+t4R(t,X).V_{\varepsilon}(tX)-V_{\varepsilon}(0)=t^{3}\left(\frac{5}{32}H_{2}+\frac{15}{8}H_{3}+\frac{3}{16}H_{4}\right)+t^{4}R(t,X).

for some real-analytic RR. Dividing by t3=ε6t^{3}=\varepsilon^{6} proves the formula for Φε\Phi_{\varepsilon}. The CkC^{k}-estimate follows because the derivatives of RR are bounded on a neighbourhood of {0}×K\{0\}\times K.

Remark 1. The charge qεq_{\varepsilon} is judiciously chosen to have the form in (1) for two reasons. Firstly, in order for lim⁡ε→0Φε\lim_{\varepsilon\to0}\Phi_{\varepsilon} to exist the order of qεq_{\varepsilon} must be at least ε3\varepsilon^{3}, the coefficient of ε3\varepsilon^{3} must equal 3/43/4, and there must be no ε4\varepsilon^{4} term. Secondly, the specific coefficient of ε5\varepsilon^{5} is chosen out of convenience to simplify the form of Φ0\Phi_{0}; it may be replaced with any negative coefficient greater than −45/256-45/256 without affecting our results.

Lemma 2. The polynomial Φ0\Phi_{0} has exactly 21 critical points all of which are non-degenerate.

Proof. In cylindrical coordinates

Φ0=132(18r4+60r3cos⁡3θ−144r2z2+5r2+48z4−10z2)\Phi_{0}=\frac{1}{32}\left(18r^{4}+60r^{3}\cos3\theta-144r^{2}z^{2}+5r^{2}+48z^{4}-10z^{2}\right)

with partial derivatives

∂Φ0∂r=r16(36r2+90rcos⁡3θ−144z2+5),\frac{\partial\Phi_{0}}{\partial r}=\frac{r}{16}\left(36r^{2}+90r\cos3\theta-144z^{2}+5\right),
∂Φ0∂θ=−45r38sin⁡3θ,\frac{\partial\Phi_{0}}{\partial\theta}=-\frac{45r^{3}}{8}\sin3\theta,
∂Φ0∂z=z8(48z2−72r2−5).\frac{\partial\Phi_{0}}{\partial z}=\frac{z}{8}\left(48z^{2}-72r^{2}-5\right).

The critical points can now be found and their Hessians evaluated with a routine calculation. These points are classified as follows:

  • The origin r=z=0r=z=0 with signature (++−)(++-).

  • Two axial equilibria at r=0r=0 and z=±1512z=\pm\frac{\sqrt{15}}{12} with signature (+−−)(+--).

  • Six planar equilibria at z=0z=0, θ=π3,π,5π3\theta=\frac{\pi}{3},\pi,\frac{5\pi}{3}, and r=15±20512r=\frac{15\pm\sqrt{205}}{12} with signature (++−)(++-) for the outer radius and (+−−)(+--) for the inner.

  • Twelve off-planar equilibria at (r,z)=(13,±3912)(r,z)=\left(\frac{1}{3},\pm\frac{\sqrt{39}}{12}\right) with signature (+−−)(+--) and (r,z)=(16,±2112)(r,z)=\left(\frac{1}{6},\pm\frac{\sqrt{21}}{12}\right) with signature (++−)(++-) for θ=0,2π3,4π3\theta=0,\frac{2\pi}{3},\frac{4\pi}{3}.

Remark 2. For the original triangular configuration the central equilibrium has Morse index 1 and the three edge equilibria each have Morse index 2. These satisfy the Euler characteristic equation

m0−m1+m2−m3=0m_0-m_1+m_2-m_3=0

for m1=1m_1=1, m2=3m_2=3, and where m3=3m_3=3 represents the charge locations and m0=1m_0=1 the point at infinity. When the two small axial charges are added the three edge equilibria persist but the central equilibrium bifurcates into the critical points of Φ0\Phi_0, 10 of which have Morse index 1, and 11 have Morse index 2. The Morse inequalities therefore remain consistent with m0=1m_0=1, m1=10m_1=10, m2=14m_2=14, and m3=5m_3=5.

Proof of Theorem 1. By Lemma 1, the function (ε,X)↦Φε(X)(\varepsilon,\mathbf{X})\mapsto\Phi_\varepsilon(\mathbf{X}) is smooth, so the implicit function theorem implies that the 21 non-degenerate critical points of Φ0\Phi_0 provided by Lemma 2 persist to 21 non-degenerate critical points of Φε\Phi_\varepsilon for all sufficiently small ε>0\varepsilon>0. These yield 21 non-degenerate critical points of VεV_\varepsilon at distance O(ε2)O(\varepsilon^2) from the origin.

On the other hand, the implicit function theorem also implies that the three non-zero non-degenerate critical points of V△V_\triangle persist to three non-degenerate critical points of VεV_\varepsilon for all sufficiently small ε>0\varepsilon>0. The distance of these from the origin is bounded away from 0, so they are distinct from the preceding 21 and hence VεV_\varepsilon has at least 24 non-degenerate critical points for all sufficiently small ε>0\varepsilon>0.

Fix such an ε>0\varepsilon>0. We have not eliminated the possibility that VεV_\varepsilon has degenerate critical points elsewhere. To get rid of them, we perturb the vector q0=(1,1,1,qε,qε)\mathbf{q}_0=(1,1,1,q_\varepsilon,q_\varepsilon) of charges of the particles.4 Let VεqV_\varepsilon^{\mathbf{q}} be the electrostatic potential generated by particles at the same locations a1,…,a5\mathbf{a}_1,\ldots,\mathbf{a}_5 as those generating Vε=Vεq0V_\varepsilon=V_\varepsilon^{\mathbf{q}_0}, but with charge vector q\mathbf{q}. Consider the smooth map

F:R5×(R3∖{a1,…,a5})→R3,F(q,x)=−∇xVεq(x)\mathbf{F}: \mathbb{R}^5 \times\left(\mathbb{R}^3 \setminus\{\mathbf{a}_1,\ldots,\mathbf{a}_5\}\right) \to\mathbb{R}^3,\qquad\mathbf{F}(\mathbf{q},\mathbf{x})=-\nabla_{\mathbf{x}}V_\varepsilon^{\mathbf{q}}(\mathbf{x})

given by the resulting electric field. Since a1,…,a5\mathbf{a}_1,\ldots,\mathbf{a}_5 do not lie in a common plane, it is straightforward to show that F\mathbf{F} is a submersion (cf. [4 Ex. 1.7.22]), so the parametric transversality theorem implies that 0\mathbf{0} is a regular value for F(q,⋅)\mathbf{F}(\mathbf{q},\cdot) and hence VεqV_\varepsilon^{\mathbf{q}} is a Morse function for almost every q∈R5\mathbf{q}\in\mathbb{R}^5 [4 p. 68].

By the implicit function theorem, choosing any such q\mathbf{q} sufficiently close to q0\mathbf{q}_0 yields a Morse potential VεqV_\varepsilon^{\mathbf{q}} having at least 24 critical points. And since all of the charges are positive, all critical points of VεqV_\varepsilon^{\mathbf{q}} belong to the compact convex hull of {a1,…,a5}\{\mathbf{a}_1,\ldots,\mathbf{a}_5\}, so there can be only finitely many. □

Remark 3. The construction employed in this counterexample can be iterated in a manner similar to constructions in [2]. Suppose that, after finitely many steps, the resulting potential U(x)U(\mathbf{x}) enjoys the same D3×Z2\mathrm{D}_3\times\mathbb{Z}_2-symmetry as the original triangle configuration. The symmetry forces the Taylor expansion of U(x)U(\mathbf{x}) around the origin to be of the form

U(x)=U(0)+αH2(x)+βH3(x)+O(∥x∥4),(2)U(\mathbf{x})=U(0)+\alpha H_2(\mathbf{x})+\beta H_3(\mathbf{x})+O(\lVert\mathbf{x}\rVert^4), \tag*{(2)}

where, for induction, we suppose α\alpha and β\beta are both greater than zero (as is the case for the original potential V△V_\triangle of the triangular configuration). Place a pair of charges of strength qη=αη3−γη5q_\eta=\alpha\eta^3-\gamma\eta^5 at a distance ±η\pm\eta along the zz-axis. The contribution of this pair near the origin is

−qηη3H2+qη4η5H4+O(qη∥x∥6η7),-\frac{q_\eta}{\eta^3}H_2+\frac{q_\eta}{4\eta^5}H_4+O\left(\frac{q_\eta\lVert\mathbf{x}\rVert^6}{\eta^7}\right),

so its leading H2H_2 term cancels the term αH2\alpha H_2 in UU. On the scale x=λη2X\mathbf{x}=\lambda\eta^2\mathbf{X}, and after dividing by η6\eta^6, a repeat of the argument above shows that the new, deformed, potential converges to

γλ2H2+βλ3H3+14αλ4H4\gamma\lambda^2H_2+\beta\lambda^3H_3+\frac{1}{4}\alpha\lambda^4H_4

as η→0\eta\to0. For an appropriate choice of λ\lambda and γ\gamma, this potential has the same 21 nondegenerate critical points as found above. For example, the choice λ=2β5α, γ=β230α\lambda=\frac{2\beta}{5\alpha},\ \gamma=\frac{\beta^2}{30\alpha} makes the limit of the deformed potential a multiple of the Φ0\Phi_0 above. One of these critical points is the origin, which was already a critical point before the new pair of charges was added. The remaining 20 are new. Since the potential of the added pair tends to zero in C2C^2 on compact sets disjoint from the origin, the nondegenerate critical points of $U persist in the new potential. Thus, we have added 20 critical points at the expense of adding two small positive charges. Moreover, after this step the expansion (2) of the new potential retains the same form, with coefficient of H2H_2 equal to γη2>0\gamma\eta^2 > 0 and coefficient of H3H_3 unchanged, so the argument may be repeated. Applying this iteration to the initial potential V△V_{\triangle} and using the perturbation argument from the proof of Theorem 1 gives the following

Proposition 1. For every m≥0m \ge0, there exists a configuration of 3+2m3 + 2m positive point charges whose potential has a finite number k≥4+20mk \ge4 + 20m of critical points, all of which are non-degenerate.

Therefore, this construction produces an asymptotic critical point to charge ratio of 10, which is larger than the value of 25/7 achieved in [2].

References

  1. [1]A. Abanov, N. Hayford, D. Khavinson, and R. Teodorescu, Around a theorem of F. Dyson and A. Lenard: energy equilibria for point charge distributions in classical electrostatics, Expo. Math., 39 (2021), pp. 182–196.DOI
  2. [2]H. Edelsbrunner, C. Fillmore, and G. Oliveira, Counting equilibria of the electrostatic potential, Proc. Lond. Math. Soc. (3), 132 (2026), pp. Paper No. e70163, 32.arxiv.org/abs/2501.05315
  3. [3]A. Gabrielov, D. Novikov, and B. Shapiro, Mystery of point charges, Proc. Lond. Math. Soc. (3), 95 (2007), pp. 443–472.
  4. [4]V. Guillemin and A. Pollack, Differential topology, AMS Chelsea Publishing, Providence, RI, 2010. Reprint of the 1974 original.
  5. [5]J. C. Maxwell, A Treatise on Electricity and Magnetism, vol. 1, Clarendon Press, Oxford, 1873.
  6. [6]M. Morse and S. S. Cairns, Critical point theory in global analysis and differential topology: An introduction, vol. Vol. 33 of Pure and Applied Mathematics, Academic Press, New York-London, 1969.
  7. [7]Y.-L. Tsai, Maxwell’s conjecture on three point charges with equal magnitudes, Phys. D, 309 (2015), pp. 86–98.DOI
  8. [8]V. Zolotov, Upper bounds for the number of isolated critical points via the Thom-Milnor theorem, Anal. Math. Phys., 13 (2023), pp. Paper No. 81, 18.arxiv.org/abs/2307.00312

Paper details

vlnc:W_n376rkx3
Version
v1 · latest
Published
Fresh — added to the Hub within the last 30 days
Added
License
CC BY 4.0
Usage

views · 6

downloads · 0

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

  • Physics and Astronomy

Not recorded for this paper

View statement
Looking for related papers…
Contents