betaFeedback

WorksMathematics

Generic finiteness of rotating relative equilibria in the planar NN-vortex problem

Abstract

For every N≥2N\ge2, we prove that Zariski-generic nonzero real circulations, with any prescribed signs, admit at most N2NN^{2N} labeled, collision-free rotating relative equilibria modulo translations, rotations in SO(2)SO(2), and positive scaling. The excluded parameters lie in an explicit finite union of quadratic hypersurfaces. The proof combines a logarithmic-residue obstruction to positive-dimensional complex solution sets with Bézout degree bounds.

Introduction

We consider the following planar NN-vortex problem. Fix N≥2N \ge2 and Γ=(Γ1,…,ΓN)∈(R∗)N\Gamma= (\Gamma_1,\ldots,\Gamma_N) \in(\mathbb{R}^*)^N. A labeled, collision-free configuration q=(q1,…,qN)q = (q_1,\ldots,q_N) in R2\mathbb{R}^2 is a rotating relative equilibrium if there are c∈R2c \in\mathbb{R}^2 and λ∈R∗\lambda\in\mathbb{R}^* such that

λ(qi−c)=Fi(q),Fi(q):=∑j≠iΓjqi−qj∣qi−qj∣2,1≤i≤N.(1)\lambda(q_i-c)=F_i(q),\qquad F_i(q):=\sum_{j\ne i}\Gamma_j\frac{q_i-q_j}{\lvert q_i-q_j\rvert^2},\qquad1\le i\le N. \tag*{(1)}

The sign convention is fixed by this equation. We identify configurations under translations, rotations in SO(2)SO(2), and positive dilations.

For k=(kij)1≤i<j≤N∈Z(N2)k=(k_{ij})_{1\le i<j\le N}\in\mathbb{Z}^{\binom{N}{2}}, write ∥k∥∞=max⁡i<j∣kij∣\lVert k\rVert_\infty=\max_{i<j}\lvert k_{ij}\rvert. Put

DN=N2N,Qk(Γ)=∑i<jkijΓiΓj,EN=⋃k∈Z(N2)0<∥k∥∞≤2DN{Γ∈CN:Qk(Γ)=0}.(2)\begin{aligned} D_N&=N^{2N},\\ Q_k(\Gamma)&=\sum_{i<j}k_{ij}\Gamma_i\Gamma_j,\\ \mathcal{E}_N&=\bigcup_{\substack{k\in\mathbb{Z}^{\binom{N}{2}}\\0<\lVert k\rVert_\infty\le2D_N}}\{\Gamma\in\mathbb{C}^N:Q_k(\Gamma)=0\}. \tag*{(2)} \end{aligned}

Theorem 1.1. The set EN\mathcal{E}_N is a proper algebraic subset of CN\mathbb{C}^N, defined over Q\mathbb{Q}. For every Γ∈(R∗)N∖EN\Gamma\in(\mathbb{R}^*)^N\setminus\mathcal{E}_N, the set of labeled, collision-free rotating relative equilibria satisfying (1), modulo translations, rotations and positive scaling, is finite and has at most DND_N elements. The conclusion also applies when ∑iΓi=0\sum_i\Gamma_i=0.

For each prescribed sign vector ε∈{1,−1}N\varepsilon\in\{1,-1\}^N, the allowed parameters form a nonempty Zariski-open intersection with

Oε={Γ∈RN:εiΓi>0 for every i}.\mathcal{O}_\varepsilon=\{\Gamma\in\mathbb{R}^N:\varepsilon_i\Gamma_i>0\text{ for every }i\}.

This intersection is open, dense, and of full Lebesgue measure in Oε\mathcal{O}_\varepsilon.

Corollary 1.2 (Zero total circulation). Let

H={Γ∈CN:∑iΓi=0}.H=\left\{\Gamma\in\mathbb{C}^{N}:\sum_i\Gamma_i=0\right\}.

Then H∩ENH\cap\mathcal{E}_N is a proper algebraic subset of HH. Consequently, the same bound DND_N holds for relatively Zariski-generic nonzero real circulations on HH. The good set is relatively open, dense, and of full (N−1)(N-1)-dimensional Lebesgue measure in every nonempty intersection H∩OεH\cap\mathcal{O}_{\varepsilon}.

The generic assumption is necessary. An example with a continuum of relative equilibria can be found in [9].

Literature review

The finiteness problem for planar Newtonian central configurations is a longstanding question in celestial mechanics and appears on Smale’s list of problems for the twenty-first century [10 Problem 6]. A related finiteness problem arises for planar point vortices, whose interaction Hamiltonian is logarithmic [2 Introduction]. The logarithmic interaction comes from the fundamental solution of the planar Laplacian. Logarithmic interactions also occur in the plasma analogy for the Laughlin wave function in the fractional quantum Hall effect [1 Section 3.1.2].

Palmore announced almost-everywhere finiteness for positive circulations [5 Corollary 5.2]; his note does not supply a proof of the generic nondegeneracy theorem used to obtain it. No part of the proof below depends on that assertion. O’Neil’s general-NN work treats static equilibria, translations, and collinear relative equilibria [3]; the scope is summarized in [2 Introduction]. It should not be read as a theorem about all planar rotating configurations with arbitrary signs.

For four vortices, Xiang Yu proves finiteness of rotating relative equilibria for arbitrary nonzero real circulations [8]. For five vortices, Xiang Yu–Shuqiang Zhu obtain finiteness outside a codimension-two algebraic exceptional set, and finiteness for every same-sign circulation vector [9 Theorems 1.2 and 1.4]. O’Neil’s more recent generic theorem applies for arbitrary NN under a mirror-symmetry restriction, with at most six vortices off the symmetry line, together with a special eight-vortex case [4]. The point of the present argument is the unrestricted number of vortices and the direct signed extension of the full complex-fiber finiteness statement.

Acknowledgment

The author is supported by the National Natural Science Foundation of China (Grant No. 12271285, 11790273), and by the New Cornerstone Investigator program and Xiaomi Foundation. Moreover, the author is a member of IAS supported by the Shiing-Shen Chern Member Fund this semester.

AI declaration: The work is done with the help of ChatGPT 6.1. The author was inspired by [1 Section 3.1.2] and suggested GPT the method of complexification which is specially adapted to the current problem and utilize the fact that the log potential is the plane Green function. GPT found the key argument, in particular the residue argument, and also assisted in writing.

A normalization that retains both rotation signs

Put T=∑iΓiT=\sum_i\Gamma_i, L=∑i<jΓiΓjL=\sum_{i<j}\Gamma_i\Gamma_j and rij=qi−qjr_{ij}=q_i-q_j. Pairing the terms indexed by (i,j)(i,j) and (j,i)(j,i) gives

∑iΓiFi=∑i<jΓiΓjrij−rij‾∣rij∣2=0,\sum_i\Gamma_iF_i=\sum_{i<j}\Gamma_i\Gamma_j\frac{r_{ij}-\overline{r_{ij}}}{\lvert r_{ij}\rvert^2}=0,
∑iΓiqi⋅Fi=∑i<jΓiΓj(qi−qj)⋅rij∣rij∣2=L.(3)\sum_i\Gamma_iq_i\cdot F_i=\sum_{i<j}\Gamma_i\Gamma_j\frac{(q_i-q_j)\cdot r_{ij}}{\lvert r_{ij}\rvert^2}=L. \tag*{(3)}

For (1), translating the rotation center to 0 therefore gives

∑iΓi(qi−c)=0,λIR=L,IR=∑iΓi∣qi−c∣2.(4)\sum_i \Gamma_i(q_i-c)=0,\qquad\lambda I_{\mathbb{R}}=L,\qquad I_{\mathbb{R}}=\sum_i\Gamma_i\lvert q_i-c\rvert^2. \tag*{(4)}

If T≠0T\ne0, then c=T−1∑iΓiqic=T^{-1}\sum_i\Gamma_iq_i; if T=0T=0, the first identity imposes ∑iΓiqi=0\sum_i\Gamma_iq_i=0. The actual rotation data are unique:

λ(qi−c)=λ′(qi−c′) (∀i)⟹(λ−λ′)(q2−q1)=0⟹λ=λ′,c=c′.\lambda(q_i-c)=\lambda'(q_i-c')\ (\forall i)\quad\Longrightarrow\quad(\lambda-\lambda')(q_2-q_1)=0\quad\Longrightarrow\quad\lambda=\lambda',\quad c=c'.

For fixed Γ∈(C∗)N\Gamma\in(\mathbb{C}^*)^N, define the complex fiber XΓX_\Gamma by

zi=∑j≠iΓjwi−wj,wi=∑j≠iΓjzi−zj,1≤i≤N,(5)z_i=\sum_{j\ne i}\frac{\Gamma_j}{w_i-w_j},\qquad w_i=\sum_{j\ne i}\frac{\Gamma_j}{z_i-z_j},\qquad1\le i\le N, \tag*{(5)}

with

z2−z1=1,sij:=(zi−zj)(wi−wj)≠0(1≤i<j≤N).(6)z_2-z_1=1,\qquad s_{ij}:=(z_i-z_j)(w_i-w_j)\ne0\quad(1\le i<j\le N). \tag*{(6)}

Here zi,wiz_i,w_i are independent complex variables. The unsliced equations are invariant under (z,w)↦(tz,t−1w)(z,w)\mapsto(tz,t^{-1}w), t∈C∗t\in\mathbb{C}^*, and z2−z1=1z_2-z_1=1 meets each orbit exactly once.

Lemma 2.1. For Γ∈(R∗)N\Gamma\in(\mathbb{R}^*)^N, physical similarity classes satisfying (1), with either sign of λ\lambda, are in bijection with

{(z,w)∈XΓ:wi=bziˉ (∀i),b=w2−w1∈R∗}.\{(z,w)\in X_\Gamma:w_i=b\bar{z_i}\ (\forall i),\quad b=w_2-w_1\in\mathbb{R}^*\}.

Proof. Represent qi−cq_i-c by ζi∈C\zeta_i\in\mathbb{C}. The force equations become

λζi=∑j≠iΓjζiˉ−ζjˉ,λζiˉ=∑j≠iΓjζi−ζj.\lambda\zeta_i=\sum_{j\ne i}\frac{\Gamma_j}{\bar{\zeta_i}-\bar{\zeta_j}},\qquad \lambda\bar{\zeta_i}=\sum_{j\ne i}\frac{\Gamma_j}{\zeta_i-\zeta_j}.

For a=ζ2−ζ1≠0a=\zeta_2-\zeta_1\ne0, set

zi=ζia,wi=λaζiˉ,b=λ∣a∣2.(7)z_i=\frac{\zeta_i}{a},\qquad w_i=\lambda a\bar{\zeta_i},\qquad b=\lambda\lvert a\rvert^2. \tag*{(7)}

Then

∑j≠iΓjwi−wj=1λa∑j≠iΓjζiˉ−ζjˉ=ζia=zi,\sum_{j\ne i}\frac{\Gamma_j}{w_i-w_j} =\frac{1}{\lambda a}\sum_{j\ne i}\frac{\Gamma_j}{\bar{\zeta_i}-\bar{\zeta_j}} =\frac{\zeta_i}{a}=z_i,
∑j≠iΓjzi−zj=a∑j≠iΓjζi−ζj=λaζiˉ=wi,\sum_{j\ne i}\frac{\Gamma_j}{z_i-z_j} =a\sum_{j\ne i}\frac{\Gamma_j}{\zeta_i-\zeta_j} =\lambda a\bar{\zeta_i}=w_i,

and

z2−z1=1,wi=bziˉ,sij=λ∣ζi−ζj∣2≠0.(8)z_2-z_1=1,\qquad w_i=b\bar{z_i},\qquad s_{ij}=\lambda\lvert\zeta_i-\zeta_j\rvert^2\ne0. \tag*{(8)}

For a similarity ζi′=αζi\zeta_i'=\alpha\zeta_i, λ′=λ/∣α∣2\lambda'=\lambda/\lvert\alpha\rvert^2, one has

a′=αa,zi′=zi,b′=λ∣α∣2∣αa∣2=b,wi′=wi.a'=\alpha a,\qquad z_i'=z_i,\qquad b'=\frac{\lambda}{\lvert\alpha\rvert^2}\lvert\alpha a\rvert^2=b,\qquad w_i'=w_i.

Conversely,

zi′=zi (∀i)⟹ζi′=a′aζi (∀i),z_i'=z_i\ (\forall i)\quad\Longrightarrow\quad\zeta_i'=\frac{a'}{a}\zeta_i\ (\forall i),

so the map is injective on similarity classes. Finally, any point in the stated locus satisfies

bzi=wiˉ=∑j≠iΓjziˉ−zjˉ=∑j≠iΓjzi−zj∣zi−zj∣2,b z_i=\bar{w_i}=\sum_{j\ne i}\frac{\Gamma_j}{\bar{z_i}-\bar{z_j}} =\sum_{j\ne i}\Gamma_j\frac{z_i-z_j}{\lvert z_i-z_j\rvert^2},

which is (1) with positions ziz_i, center 0 and angular parameter bb. □

Isolation and the Bézout bound

We first prove an isolation criterion by logarithmic residues, then bound the number of isolated solutions by Bézout and use the same estimate to obtain an explicit finite union of exceptional quadratic hypersurfaces.

Eliminate z2z_2 by z2=z1+1z_2=z_1+1. The remaining coordinates are

(z1,z3,…,zN,w1,…,wN)∈C2N−1;(z_1,z_3,\ldots,z_N,w_1,\ldots,w_N)\in\mathbb{C}^{2N-1};

we continue to abbreviate such a point by (z,w)(z,w), with z2=z1+1z_2=z_1+1 understood in every formula. All vortex indices range from 11 to NN; an empty product equals 11. For 1≤i≤N1\leq i\leq N, define

Ai=zi∏j≠i(wi−wj)−∑j≠iΓj∏k∉{i,j}(wi−wk),A_i=z_i\prod_{j\ne i}(w_i-w_j)-\sum_{j\ne i}\Gamma_j\prod_{k\notin\{i,j\}}(w_i-w_k),
Bi=wi∏j≠i(zi−zj)−∑j≠iΓj∏k∉{i,j}(zi−zk).(9)B_i=w_i\prod_{j\ne i}(z_i-z_j)-\sum_{j\ne i}\Gamma_j\prod_{k\notin\{i,j\}}(z_i-z_k). \tag*{(9)}

Clearing the nonzero denominators in (5) gives

YΓ:={(z,w)∈C2N−1:Ai=Bi=0 for 1≤i≤N},Y_\Gamma:=\{(z,w)\in\mathbb{C}^{2N-1}:A_i=B_i=0\text{ for }1\leq i\leq N\},
XΓ={(z,w)∈YΓ:sij≠0 for every i<j},sij=(zi−zj)(wi−wj).(10)X_\Gamma=\{(z,w)\in Y_\Gamma:s_{ij}\ne0\text{ for every }i<j\},\qquad s_{ij}=(z_i-z_j)(w_i-w_j). \tag*{(10)}

An isolated solution means an isolated point of XΓX_\Gamma in the complex Euclidean topology. Since XΓX_\Gamma is an open subset of the algebraic set YΓY_\Gamma, its isolated points are precisely the zero-dimensional irreducible components of YΓY_\Gamma that satisfy all the inequalities sij≠0s_{ij}\ne0. An algebraic set has finitely many irreducible components; thus, if every solution is isolated, XΓX_\Gamma is finite. All counts below concern distinct points; no Jacobian nondegeneracy is assumed.

Isolation by logarithmic residues

Proposition 3.1. Fix Γ∈(C∗)N\Gamma\in(\mathbb{C}^{*})^N. If XΓX_\Gamma has a nonisolated point, then there exists

0≠k∈Z(N2)such thatQk(Γ)=∑i<jkijΓiΓj=0.0\ne k\in\mathbb{Z}^{\binom{N}{2}}\qquad\text{such that}\qquad Q_k(\Gamma)=\sum_{i<j}k_{ij}\Gamma_i\Gamma_j=0.

Consequently, if the numbers ΓiΓj\Gamma_i\Gamma_j, 1≤i<j≤N1\leq i<j\leq N, are linearly independent over Q\mathbb{Q}, every point of XΓX_\Gamma is isolated.

Proof. Let x∈XΓx\in X_\Gamma be nonisolated, and let ZZ be a positive-dimensional irreducible component of YΓY_\Gamma containing xx. Put d=dim⁡Z≥1d=\dim Z\geq1.

*An algebraic curve through x.∗x.* Let Z‾\overline{Z} denote the projective closure of ZZ in complex projective space P2N−1(C)\mathbb{P}^{2N-1}(\mathbb{C}), with C2N−1\mathbb{C}^{2N-1} identified with its usual affine coordinate chart. If d=1d=1, take C=Z‾C=\overline{Z}. If d≥2d\geq2, choose projective hyperplanes

H1,…,Hd−1⊂P2N−1(C)H_1,\ldots,H_{d-1}\subset\mathbb{P}^{2N-1}(\mathbb{C})

successively through xx, each containing no irreducible component of the preceding intersection. Such choices exist because every component at that stage is positive-dimensional, and the hyperplanes through xx containing it form a proper closed subset of the hyperplane space. Each intersection has pure dimension one less than the preceding one. Choose a reduced irreducible component CC through xx of

Z‾∩H1∩⋯∩Hd−1;dim⁡C=1.(11)\overline{Z}\cap H_1\cap\cdots\cap H_{d-1};\qquad\dim C=1. \tag*{(11)}

For d=1d=1, this intersection means Z‾\overline{Z} itself. Since Z‾∩C2N−1=Z⊂YΓ\overline{Z}\cap\mathbb{C}^{2N-1}=Z\subset Y_\Gamma,

C∩XΓ={(z,w)∈C∩C2N−1:sij≠0 for every i<j}∋xC\cap X_\Gamma=\{(z,w)\in C\cap\mathbb{C}^{2N-1}:s_{ij}\ne0\text{ for every }i<j\}\ni x

is a nonempty open subset of CC, with finite complement. Thus CC is algebraic by construction. errs_mid=1

Let ν:C^→C\nu:\widehat{C}\to C be its normalization: C^\widehat{C} is a nonsingular projective curve, and ν\nu has finite fibers and is an isomorphism away from the finitely many singular points of CC [7 I, Proposition 6.8 and Corollary 6.11]. Over C\mathbb{C}, C^\widehat{C} is a compact connected Riemann surface. We use the same symbols ziz_i, wiw_i, sijs_{ij} for their pullbacks to meromorphic functions on C^\widehat{C}. The force equations (5) hold as identities of meromorphic functions, and sijs_{ij}, zi−zjz_i-z_j, wi−wjw_i-w_j are not identically zero for i<ji<j.

We next claim that there is at least one sijs_{ij} that is non constant on CC.

Proof of the claim. Suppose all sijs_{ij} are constant. They are nonzero on C∩XΓC\cap X_\Gamma. Put

ui=zi−z1,vi=wi−w1,u1=v1=0,u2=1,v2=s12=:b∈C∗.u_i=z_i-z_1,\qquad v_i=w_i-w_1,\qquad u_1=v_1=0,\qquad u_2=1,\qquad v_2=s_{12}=:b\in\mathbb{C}^*.

For i≥3i\ge3,

uivi=s1i,(1−ui)(b−vi)=s2i,u_i v_i=s_{1i},\qquad(1-u_i)(b-v_i)=s_{2i},
vi=s1i+b−s2i−bui,v_i=s_{1i}+b-s_{2i}-bu_i,
0=bui2−(s1i+b−s2i)ui+s1i=b(ui−αi)(ui−βi).0=bu_i^2-(s_{1i}+b-s_{2i})u_i+s_{1i}=b(u_i-\alpha_i)(u_i-\beta_i).

where αi,βi∈C\alpha_i,\beta_i\in\mathbb{C} are the roots of this quadratic. The identity theorem implies that one of the two factors vanishes identically. Hence

ui∈{αi,βi}⊂C,vi=s1i/ui∈C.u_i\in\{\alpha_i,\beta_i\}\subset\mathbb{C},\qquad v_i=s_{1i}/u_i\in\mathbb{C}.

Thus every ui,viu_i,v_i is constant, and

zi=∑j≠iΓjvi−vj∈C,wi=∑j≠iΓjui−uj∈C.z_i=\sum_{j\ne i}\frac{\Gamma_j}{v_i-v_j}\in\mathbb{C},\qquad w_i=\sum_{j\ne i}\frac{\Gamma_j}{u_i-u_j}\in\mathbb{C}.

All affine coordinates of CC would be constant, contradicting dim⁡C=1\dim C=1. Hence at least one sijs_{ij} is nonconstant. □

The residue relation. Pairing (i,j)(i,j) with (j,i)(j,i) in the force equations gives

∑iΓiziwi=∑i<jΓiΓj(wiwi−wj+wjwj−wi)=∑i<jΓiΓj.(12)\sum_i\Gamma_i z_iw_i=\sum_{i<j}\Gamma_i\Gamma_j\left(\frac{w_i}{w_i-w_j}+\frac{w_j}{w_j-w_i}\right)=\sum_{i<j}\Gamma_i\Gamma_j. \tag*{(12)}

Since the circulations Γi\Gamma_i are fixed, differentiation on C^\widehat{C} yields

0=∑iΓi(wi dzi+zi dwi)=∑i∑j≠iΓiΓj(dzizi−zj+dwiwi−wj)=∑i<jΓiΓj(dzi−dzjzi−zj+dwi−dwjwi−wj)=∑i<jΓiΓjdsijsij.(13)\begin{aligned} 0=\sum_i\Gamma_i(w_i\,dz_i+z_i\,dw_i) \\ &=\sum_i\sum_{j\ne i}\Gamma_i\Gamma_j\left(\frac{dz_i}{z_i-z_j}+\frac{dw_i}{w_i-w_j}\right) \\ &=\sum_{i<j}\Gamma_i\Gamma_j\left(\frac{dz_i-dz_j}{z_i-z_j}+\frac{dw_i-dw_j}{w_i-w_j}\right) \\ &=\sum_{i<j}\Gamma_i\Gamma_j\frac{ds_{ij}}{s_{ij}}. \tag*{(13)} \end{aligned}

Choose a nonconstant sijs_{ij}. It has a pole p∈C^p\in\widehat{C}: otherwise it would be holomorphic on the compact Riemann surface C^\widehat{C}, hence constant by the maximum principle. Choose a local holomorphic coordinate tt with t(p)=0t(p)=0. For any nonzero meromorphic function gg near pp, write

g=tmh(t),m∈Z,h holomorphic,h(0)≠0.g=t^m h(t),\qquad m\in\mathbb{Z},\qquad h\text{ holomorphic},\qquad h(0)\ne0.

The order ord⁡pg:=m\operatorname{ord}_p g := m is positive at a zero and negative at a pole. The residue Res⁡p\operatorname{Res}_p of a meromorphic differential is the coefficient of dt/tdt/t in its local expansion. Thus

dgg=mdtt+h′(t)h(t) dt,Res⁡pdgg=m=ord⁡pg.(14)\frac{dg}{g}=m\frac{dt}{t}+\frac{h'(t)}{h(t)}\,dt,\qquad\operatorname{Res}_p\frac{dg}{g}=m=\operatorname{ord}_p g. \tag*{(14)}

Taking the residue of (13) at this fixed point yields

k:=(ord⁡psij)i<j∈Z(N2)∖{0},0=∑i<jΓiΓjord⁡psij=Qk(Γ).(15)k:=(\operatorname{ord}_p s_{ij})_{i<j}\in\mathbb{Z}^{\binom{N}{2}}\setminus\{0\},\qquad0=\sum_{i<j}\Gamma_i\Gamma_j\operatorname{ord}_p s_{ij}=Q_k(\Gamma). \tag*{(15)}

The vector kk is nonzero because the chosen sijs_{ij} has a pole at pp. Rational linear independence of the pair products rules out (15), proving the isolation assertion. □

Bézout: the number of isolated solutions

For an irreducible projective algebraic set W⊂PnW\subset\mathbb{P}^n of dimension dd, its degree deg⁡W\deg W is the number of points of W∩LW\cap L, counted with multiplicity, where L⊂PnL\subset\mathbb{P}^n is a general linear subspace of dimension n−dn-d [6 Definition 18.1(iii), p. 225]. For an irreducible affine algebraic set W⊂CnW\subset\mathbb{C}^n, set deg⁡W:=deg⁡W‾\deg W:=\deg\overline{W}, where W‾⊂Pn\overline{W}\subset\mathbb{P}^n is its projective closure under x↦[1:x]x\mapsto[1:x]. In particular, a point has degree 1. All sums below run over irreducible components of all dimensions; an empty sum is zero. The degree of a polynomial is its total degree.

Lemma 3.2. Let Y⊂CnY\subset\mathbb{C}^n be an algebraic set and PP a polynomial of degree at most an integer e≥1e\ge1. Let Z1,…,ZrZ_1,\ldots,Z_r be the irreducible components of YY, and W1,…,WsW_1,\ldots,W_s those of Y∩{P=0}Y\cap\{P=0\}. Then

∑b=1sdeg⁡Wb≤e∑a=1rdeg⁡Za.\sum_{b=1}^{s}\deg W_b\le e\sum_{a=1}^{r}\deg Z_a.

The same inequality holds for a projective algebraic set cut by a homogeneous polynomial. In particular, for every Γ∈(C∗)N\Gamma\in(\mathbb{C}^*)^N, if Z1,…,ZrZ_1,\ldots,Z_r are the irreducible components of YΓY_\Gamma, then

∑a=1rdeg⁡Za≤N2N=DN.(16)\sum_{a=1}^{r}\deg Z_a\le N^{2N}=D_N. \tag*{(16)}

Proof. Consider one component ZaZ_a. If PP vanishes identically on ZaZ_a, its degree remains unchanged and deg⁡Za≤edeg⁡Za\deg Z_a\le e\deg Z_a. If PP does not vanish identically and either dim⁡Za=0\dim Z_a=0 or PP is a nonzero constant, the intersection is empty. In the remaining case, homogenize PP and apply projective Bézout [7 I, Theorem 7.7]. Discarding intersection multiplicities and components at infinity gives

∑Wdeg⁡W≤edeg⁡Za,\sum_W\deg W\le e\deg Z_a,

where WW runs over the irreducible components of Za∩{P=0}Z_a\cap\{P=0\}. Summing over aa proves the inequality; discarding repeated components or components contained in others can only decrease the sum. The same argument applies to projective intersections. Since deg⁡Ai,deg⁡Bi≤N\deg A_i,\deg B_i\le N, apply the inequality successively to the 2N2N polynomials in (9), starting with C2N−1\mathbb{C}^{2N-1}, of degree 1. This gives (16). □

Proposition 3.3. For every Γ∈(C∗)N\Gamma\in(\mathbb{C}^*)^N,

#{x∈XΓ:x is isolated in XΓ}≤DN=N2N.\#\{x\in X_\Gamma:x\text{ is isolated in }X_\Gamma\}\le D_N=N^{2N}.

This counts isolated points even when XΓX_\Gamma also has positive-dimensional components. In particular, if every point of XΓX_\Gamma is isolated, then #XΓ≤DN\#X_\Gamma\le D_N.

Proof. Let Z1,…,ZrZ_1,\ldots,Z_r be the irreducible components of YΓY_\Gamma. Since XΓX_\Gamma is open in YΓY_\Gamma, each isolated point of XΓX_\Gamma is a zero-dimensional component of YΓY_\Gamma. Hence

#{x∈XΓ:x is isolated}≤#{a:dim⁡Za=0}≤∑a=1rdeg⁡Za≤DN.\#\{x\in X_\Gamma:x\text{ is isolated}\}\le\#\{a:\dim Z_a=0\}\le\sum_{a=1}^{r}\deg Z_a\le D_N.

Bounding the residue coefficients

The same degree estimate bounds the integers in the residue obstruction. This reduces the possible relations to the finite list in (2) and proves finiteness outside EN\mathcal{E}_N.

Corollary 3.4. For every Γ∈(C∗)N∖EN\Gamma\in(\mathbb{C}^{*})^{N} \setminus\mathcal{E}_N, every point of XΓX_{\Gamma} is isolated and

#XΓ≤DN.\#X_{\Gamma} \le D_N.

Proof. Suppose XΓX_{\Gamma} has a nonisolated point. Take the component ZZ, the curve CC with normalization ν:C^→C\nu:\widehat{C} \to C, and the pole pp used to construct the residue vector kk in the proof of Proposition 3.1. Applying Lemma 3.2 with e=1e=1 at each hyperplane intersection in (11) gives

δ:=deg⁡C≤deg⁡Z‾=deg⁡Z≤DN.(17)\delta:= \deg C \le\deg\overline{Z} = \deg Z \le D_N. \tag*{(17)}

Consider a coordinate difference f=zi−zjf=z_i-z_j or f=wi−wjf=w_i-w_j with i<ji<j, as a meromorphic function on C^\widehat{C}. If ff is constant, it is nonzero, and ord⁡pf=0\operatorname{ord}_p f=0. Otherwise choose a∈Ca \in\mathbb{C} avoiding the finite values of ff at the finitely many points of C^\widehat{C} mapped by ν\nu to singular points or points at infinity of CC. Every zero of f−af-a then lies over a smooth affine point of CC. Its order is the local intersection multiplicity with the hyperplane f=af=a. Bézout therefore gives

∑q∈C^max⁡{ord⁡q(f−a),0}≤δ.\sum_{q\in\widehat{C}}\max\{\operatorname{ord}_q(f-a),0\}\le\delta.

For any nonzero meromorphic function gg on C^\widehat{C}, the residue theorem and (14) give

∑q∈C^ord⁡qg=∑q∈C^Res⁡qdgg=0;\sum_{q\in\widehat{C}}\operatorname{ord}_q g=\sum_{q\in\widehat{C}}\operatorname{Res}_q\frac{dg}{g}=0;

only finitely many terms are nonzero. Since f−af-a and ff have the same poles with the same orders,

∑q∈C^max⁡{ord⁡qf,0}=∑q∈C^max⁡{−ord⁡qf,0}=∑q∈C^max⁡{ord⁡q(f−a),0}≤δ.\sum_{q\in\widehat{C}}\max\{\operatorname{ord}_q f,0\} =\sum_{q\in\widehat{C}}\max\{-\operatorname{ord}_q f,0\} =\sum_{q\in\widehat{C}}\max\{\operatorname{ord}_q(f-a),0\}\le\delta.

Thus ∣ord⁡pf∣≤δ|\operatorname{ord}_p f|\le\delta in both the constant and nonconstant cases. The order of a product is the sum of the orders, so

∣kij∣=∣ord⁡psij∣=∣ord⁡p(zi−zj)+ord⁡p(wi−wj)∣≤2δ≤2DN.(18)|k_{ij}|=|\operatorname{ord}_p s_{ij}|=|\operatorname{ord}_p(z_i-z_j)+\operatorname{ord}_p(w_i-w_j)|\le2\delta\le2D_N. \tag*{(18)}

Together with (15), this gives

0<∥k∥∞≤2DN,Qk(Γ)=0,0<\lVert k\rVert_{\infty}\le2D_N,\qquad Q_k(\Gamma)=0,

contrary to Γ∉EN\Gamma\notin\mathcal{E}_N. Thus every point is isolated; Proposition 3.3 gives the count. □

Proof of Theorem 1.1 and Corollary 1.2. The monomials ΓiΓj\Gamma_i\Gamma_j, i<ji<j, are distinct, so Qk≢0Q_k\not\equiv0 for every k≠0k\ne0. Since the index set in (2) is finite, the polynomial

PN(Γ):=∏k∈Z(N2)0<∥k∥∞≤2DNQk(Γ)P_N(\Gamma):=\prod_{\substack{k\in\mathbb{Z}^{\binom{N}{2}}\\0<\lVert k\rVert_{\infty}\le2D_N}}Q_k(\Gamma)

is nonzero and has integer coefficients. Thus EN={Γ∈CN:PN(Γ)=0}\mathcal{E}_N=\{\Gamma\in\mathbb{C}^{N}:P_N(\Gamma)=0\} is a proper algebraic subset of CN\mathbb{C}^{N}, defined over Q\mathbb{Q}.

On HH, substitute ΓN=−∑i<NΓi\Gamma_N=-\sum_{i<N}\Gamma_i. Then

Qk(Γ1,…,ΓN−1,−∑i<NΓi)=−∑i<NkiNΓi2+∑i<j<N(kij−kiN−kjN)ΓiΓj.(19)Q_k\left(\Gamma_1,\ldots,\Gamma_{N-1},-\sum_{i<N}\Gamma_i\right) =-\sum_{i<N}k_{iN}\Gamma_i^2+\sum_{i<j<N}(k_{ij}-k_{iN}-k_{jN})\Gamma_i\Gamma_j. \tag*{(19)}

If this polynomial vanished identically, the square coefficients would first give kiN=0k_{iN}=0 for i<Ni<N; the remaining mixed coefficients would then give kij=0k_{ij}=0 for i<j<Ni<j<N. This contradicts $k\ne 0.

Each factor, and hence PNP_N, therefore remains nonzero after the substitution. In particular, H∩ENH \cap\mathcal{E}_N is a proper subset of HH.

The zero set of a nonzero real polynomial is closed, has empty interior, and has Lebesgue measure zero. Apply this fact to PNP_N on RN\mathbb{R}^N and to the substituted polynomial in the N−1N-1 linear coordinates Γ1,…,ΓN−1\Gamma_1,\ldots,\Gamma_{N-1} on H∩RNH \cap\mathbb{R}^N. This proves the asserted nonemptiness, openness, density and full measure in each OεO_\varepsilon and each nonempty H∩OεH \cap O_\varepsilon, with relative topology and measure on H∩RNH \cap\mathbb{R}^N. The condition PN(Γ)≠0P_N(\Gamma) \ne0 also gives the Zariski-open intersections and relative genericity.

Finally, for Γ∈(R∗)N∖EN\Gamma\in(\mathbb{R}^*)^N \setminus\mathcal{E}_N, Lemma 2.1 and Corollary 3.4 give

#{labeled physical similarity classes for Γ}≤#XΓ≤DN=N2N.\#\{\text{labeled physical similarity classes for }\Gamma\} \le\#X_\Gamma\le D_N = N^{2N}.

The correspondence in Lemma 2.1 includes both signs of λ\lambda and is valid also when ∑iΓi=0\sum_i \Gamma_i = 0. □

References

Rm 305, Wenbei Building, New Cornerstone Sciences Laboratory, Department of Mathematics Sciences, Tsinghua University, 100084

Email address: [email protected]

References

  1. [1]D. Tong, The Quantum Hall Effect, TIFR Infosys Lectures, 2016. https://www.damtp.cam.ac.uk/user/tong/qhe/qhe.pdf.
  2. [2]M. Hampton and R. Moeckel, Finiteness of stationary configurations of the four-vortex problem, Transactions of the American Mathematical Society 361 (2009), 1317–1332. doi:10.1090/S0002-9947-08-04685-0.DOI
  3. [3]K. A. O’Neil, Stationary configurations of point vortices, Transactions of the American Mathematical Society 302 (1987), 383–425.
  4. [4]K. A. O’Neil, Finiteness of mirror-symmetric relative equilibria of point vortices, Physica D: Nonlinear Phenomena 481 (2025), 134823. doi:10.1016/j.physd.2025.134823.DOI
  5. [5]J. I. Palmore, Relative equilibria of vortices in two dimensions, Proceedings of the National Academy of Sciences of the USA 79 (1982), 716–718. doi:10.1073/pnas.79.2.716.DOI
  6. [6]J. Harris, Algebraic Geometry: A First Course, Graduate Texts in Mathematics, vol. 133, Springer-Verlag, New York, 1992. doi:10.1007/978-1-4757-2189-8.DOI
  7. [7]R. Hartshorne, Algebraic Geometry, Graduate Texts in Mathematics, vol. 52, Springer-Verlag, New York, 1977. doi:10.1007/978-1-4757-3849-0.DOI
  8. [8]X. Yu, Finiteness of stationary configurations of the planar four-vortex problem, Advances in Mathematics 435 (2023), 109378. doi:10.1016/j.aim.2023.109378.DOI
  9. [9]X. Yu and S. Zhu, On finiteness of stationary configurations of the planar five-vortex problem, Mathematische Annalen 392 (2025), 4023–4069. doi:10.1007/s00208-025-03183-w.DOI
  10. [10]S. Smale, Mathematical problems for the next century, The Mathematical Intelligencer 20 (1998), no. 2, 7–15. doi:10.1007/BF03025291.DOI

Paper details

vlnc:W_ahb52m5f
Version
v1 · latest
Published
Fresh — added to the Hub within the last 30 days
Added
License
CC BY 4.0
Usage

views · 7

downloads · 0

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

  • Mathematics

ChatGPT 6.1

View statement
Looking for related papers…
Contents