We prove that the critical value 2π2 is isolated for the Möbius cross energy of two-component links in the round three dimensional sphere S3. Specifically, there exists ε0>0 such that any non-split, regular H2 pair of curves with disjoint images, having vanishing first variation and energy at most 2π2+ε0, must lie in the Möbius-reparametrization orbit of the standard Hopf link.
Introduction and main theorem
Set S1=R/(2πZ), and equip S3={z∈R4:∣z∣=1} with the round metric g0. For a pair Γ=(γ1,γ2) of rectifiable closed curves with disjoint component images, the Möbius cross energy is
E(Γ)=∫γ1∫γ2∣x−y∣2dℓ1(x)dℓ2(y),(1)
where ∣x−y∣ is the Euclidean chordal distance in R4. Integration along each parametrized curve counts multiplicity. O’Hara [6] introduced the related knot energy. Freedman, He, and Wang [2] studied the cross interaction (1), which is invariant under Möbius transformations.
A pair of disjoint compact connected sets is called non-split if no smooth embedded two dimensional sphere disjoint from them separates the two sets. Hopf links are non-split, and the standard Hopf link is
Its components are unit-speed great circles in orthogonal 2-planes. Since ∣x(s)−y(t)∣2=2,
E(H)=2π2.
Freedman, He, and Wang [2] conjectured that this is the minimum cross energy among all non-split two-component links. An important step was taken by He [4], who proved that a global minimizer exists and that every such minimizer is ambiently isotopic to the Hopf link. Agol, Marques, and Neves subsequently proved the sharp lower bound and classified all equality configurations [1].
Theorem 1.1 (Agol–Marques–Neves).Every non-split two-component rectifiable link in S3 satisfies
E(Γ)≥2π2.
Equality holds precisely when a conformal transformation carries the component images to the standard Hopf circles, each counted with multiplicity one.
Key words and phrases. Möbius cross energy, Hopf link, critical-value gap. For a pair of curves Γ=(γ1,γ2) in R3, the Gauss linking number is defined by
It is conformal invariant, therefore the linking number of a pair of curves Γ in S3 is defined to be that of its stereographic projection in R3. A nonzero linking number implies non-splitness, but the converse fails, as illustrated by the Whitehead link. A direct calculus implies
E(Γ)≥4π∣lk(Γ)∣.
Allowing a choice of component orientations, He [4] shows that a pair of non-split curves is link-homotopic to the Hopf link provided E(Γ)<8π.
As a continuation of the study of He, Agol, Marques, and Neves [4, 1], and motivated by Mondino and Nguyen [7], we study critical points rather than minimizers of the cross energy. The equality case in Theorem 1.1 characterizes the minimizers, but does not rule out non-minimizing critical pairs with energy arbitrarily close to 2π2. Our main result shows that such critical pairs do not exist.
To state the result precisely, we fix the following notation.
Definition 1.2. A closed curve γ is regular if ∣γ′∣>0 everywhere, we fix the extrinsic Sobolev space
(ii) A pair is weakly critical if the first variation δEΓ=0 on every H2 tangent variation.
(iii) Let Diff2(S1) denote the H2 circle maps that are C1 diffeomorphisms, allowing both orientations, and set
G=Mo¨b(S3)×Diff2(S1)×Diff2(S1).(3)
We write G⋅H for the set of pairs obtained from H by a common orientation-preserving Möbius transformation and separate reparametrizations.
We can now state our main gap theorem.
Theorem 1.3.There existsε0>0such that every non-split, weakly critical pairΓ∈Asatisfying
E(Γ)≤2π2+ε0
belongs toG⋅H. In particular,E(Γ)=2π2.
Remark 1.4. For fixed γ2, set
ρ2(x)=∫γ2∣x−y∣2dℓ2(y).
Then E(γ1,γ2) is the length of γ1 in the conformal metric ρ22g0 on S3∖γ2; see [2]. Thus a critical pair consists of unparametrized closed geodesics in the metrics generated by the opposite components. We derive the corresponding equation in § 2.
Exact criticality is essential here, since non-Möbius perturbations of the Hopf link can have energies arbitrarily close to 2π2. Although we work with regular H2 pairs, the critical equation implies smoothness after constant-speed reparametrization.
Outline of the paper.§ 2 derives the first variation and the Euler–Lagrange equation for weakly critical pairs. § 3 establishes a Möbius normalization, proves regularity of weakly critical pairs, and develops the compactness theory near the minimum energy. In § 4, we compute the second variation at the Hopf link, identify its kernel, and establish transverse coercivity. § 5 develops the nonlinear estimates and proves the renormalization lemma. Finally, in § 6, we combine the transverse analysis of § 4 with the renormalization argument of § 5 to prove the local gap theorem, and then use the compactness result of § 3 to complete the proof of the main theorem.
Acknowledgments. The authors grateful to Professor Jian Ge for useful discussions. They also thank Xingwang Xu, Xingdong Tang and Jie Zhou for their interest in this problem. The first author is supported by the China Postdoctoral Science Foundation (Grant No. 2026M793350). The second author is supported by the Open Project Program of the Key Laboratory of Mathematics and Complex System at Beijing Normal University (Grant No. 202501). The principal arguments and the core of the paper were developed by the second author while visiting SIMIS.
AI assistance declaration: The main ideas and most proofs are solely the work of the authors. ChatGPT (version 5.6sol) was used only to assist in proving Proposition 3.1 and to aid in manuscript preparation. The authors have verified all contributions and assume complete accountability for the paper.
First variation and the Euler–Lagrange equation
In this section, we establish the basic preliminaries for the Möbius cross energy and derive its first variation and Euler–Lagrange equation.
We denote D the Euclidean connection on R4 and ∇ the Levi-Civita connection of g0 on S3. Let X,Y be tangent vector fields on S3, then for each x∈S3, the Gauss formula gives
DXY=∇XY+II(X,Y),II(X,Y)=−⟨X,Y⟩g0x,(4)
where II is the second fundamental form of S3⊂R4.
Thus, for a unit-speed curve, γ′′=κ−γ, where κ=∇γ′γ′ is its curvature vector in S3.
For a variation Γτ with initial velocity ξ, we write δEΓ[ξ]=dτdE(Γτ)τ=0 and δ2 for the second differential in a fixed chart. Ordinary coordinate derivatives are denoted by ∂ or d.
The parameter torus is T2=Ss1×St1. Throughout the paper,
∬F(s,t)dsdt:=∫02π∫02πF(s,t)dsdt
always refers to this fixed torus with the product parameter measure.
Lemma 2.1.Every regular closed H2 curve admits a constant-speed parametrization obtained by an orientation-preserving H2 circle diffeomorphism. Separate such reparametrizations preserve the cross energy, non-splittness, and weak criticality.
Proof. For a lift of the parameter to [0,2π], set
θ(s)=L(γ)2π∫0s∣γ′(r)∣dr,s∈[0,2π].
Since γ′∈H1↪C0 and γ is a regular closed curve, composition with the Euclidean norm gives ∣γ′∣∈H1. Consequently θ∈H2, and θ′=2π∣γ′∣/L(γ) is continuous and bounded below by a positive constant. The identity θ(s+2π)=θ(s)+2π makes θ a degree-one C1 circle diffeomorphism. Denote its inverse by ψ=θ−1, change of variables gives ψ∈H2. Thus we can define γˉ:=γ∘ψ satisfies γˉ∈H2 and ∣γˉ′∣=L(γ)/(2π).
Changing variables separately in each of the two integrals proves the energy invariance, since the curve images remain unchanged, the non-split property is also preserved. Composition by each fixed H2 diffeomorphism preserves H2 tangent fields. Pulling back a variation and differentiating energy invariance proves preservation of weak criticality in both directions. A fixed smooth Möbius transformation preserves weak criticality by the same argument. □
Next, we derive the first variation in the spherical model. Let Γ=(γ1,γ2)∈A be a regular H2 pair with disjoint component images in S3. Write Ti=∂ℓiγi and κi=∇TiTi. The projection onto the curve-normal plane inside TγiS3 is
In arclength coordinates, Ti∈H1 and κi∈L2, so the integration by parts is valid in the Sobolev sense. Since Fi∈L2, the right-hand side extends continuously to H1 sphere-tangent test fields. Summing the two partial variations proves (8).
Given an arbitrary smooth periodic R4-valued field φ, take an admissible sphere-tangent field
ξi=(Id−γi⊗γi)φ∈H2.
Since Fi⊥γi, weak criticality gives ∫⟨Fi,φ⟩dℓi=0. The speed is bounded away from zero implies Fi=0 almost everywhere. Conversely, Fi=0 makes every first variation vanish. Dividing by the positive potentials proves (9).
Remark 2.3. He [4 Lemma 3.1, equations (3-2)–(3-3)] derived the first variation for smooth links and smooth variation fields. We repeat the computation here in order to justify the formula for regular H2 links and H1 variation fields, which obtain the weak Euler–Lagrange equation used below.
Global normalization and compactness of critical pairs
In this section, we establish a global Möbius gauge for non-split pairs in A. Combined with the critical equation, this yields regularity and uniform derivative estimates for weakly critical pairs. We then prove a compactness theorem for critical sequences approaching the Hopf energy, which will be used in the proof of the gap theorem.
We first give a global Möbius gauge. Let N=(0,0,0,1) and let σ:S3∖{N}→R3 be stereographic projection. Its inverse satisfies
Proposition 3.1. Fix 0<Λ<∞. Every non-split pair Γ=(γ1,γ2)∈A satisfying E(Γ)≤Λ has an orientation-preserving Möbius image for which
32π≤L(γi)≤π6Λ,i=1,2,(11)
and
distR4(γ1,γ2)≥dΛ>0.(12)
The constant dΛ depends only on Λ.
Proof. Send a point of γ1 to the north pole N, and then apply a translation and dilation in the stereographic chart σ so that
N∈γ1,0∈σ(γ2),diamR3σ(γ2)=1.
Thus σ(γ2)⊂B1(0). Non-splitness forces γ1 to meet σ−1(B1(0)); otherwise the inverse image of a slightly larger round sphere would separate the components. We also use the elementary estimate
L(η)≥2diamS3η(S1)(13)
for every closed rectifiable curve η:S1→S3 by considering the two arcs joining a diameter-realizing pair. The first component contains N and a point q∈γ1(S1) with ∣σ(q)∣≤1, so its spherical diameter is at least π/2. Applying (13) gives
L(γ1)≥π.(14)
The compact set σ(γ2(S1)) realizes its diameter, so there exist z,w in this set such that
∣z−w∣=1.
Note that ∣z∣,∣w∣≤1. If d denotes their spherical distance, then formula (10) gives
2sin2d=∣σ−1(z)−σ−1(w)∣=1+∣z∣21+∣w∣22∣z−w∣≥1,
and hence d≥π/3. Another application of (13) yields
Combining E(Γ)≤Λ with (14)–(15) gives the stronger individual estimates
L(γ1)≤L(γ2)4Λ≤π6Λ,L(γ2)≤L(γ1)4Λ≤π4Λ.
Together with (14)–(15), these imply the symmetric bounds (11).
To obtain a uniform lower bound for the distance between the two components, we set
d:=distR4(γ1,γ2)=x∈γ1(S1),y∈γ2(S1)min∣x−y∣.
Using unit-speed arclength coordinates on R/L(γi)Z with the two origins chosen at a closest pair, so that
∣γ1(0)−γ2(0)∣=d.
Put ℓ:=32π. The lower length bounds ensure that both curves contain arclength intervals [0,ℓ]. For 0≤s,t≤ℓ, the triangle inequality and the fact that chordal distance is bounded above by arclength give
The function ρˉ2 is bounded below by 1/4, and all of its ambient derivatives are bounded by constants depending only on m and d. Since logρ2=logL(γ2)+logρˉ2, every derivative of positive order is independent of the constant factor L(γ2). The ordinary chain rule for log proves (17). □
Proposition 3.4. Every constant-speed weakly critical pair in A is smooth. Moreover, let Γn=(γ1,n,γ2,n) be sequences of constant-speed weakly critical pairs, under the bounds
0<L−≤L(γi,n)≤L+<∞,dist(γ1,n,γ2,n)≥d0>0.
one has ∣γi,n∣Cm≤Cm(L−,L+,d) for every $m\ge0.
Proof. By Lemma 2.1, take
∣γi′∣≡ai=2πLi>0,Ti=aiγi′.
Thus the curvature and γi′′ satisfy
γi′′=ai2(κi−γi).
Combining this identity with (9) gives the following Euler–Lagrange equation:
Here PNi is the projection onto the normal space. By Lemma 3.3, Gj extends smoothly near γi. Initially H2↪C1,1/2, so γi∈C1,1/2, which implies γi∈C2,1/2. Repeating this argument yields γi∈C∞.
Under the stated bounds, ∣γi,n∣=1 and ∣γi,n′∣=ai,n give a uniform C1 bound. Since Lemma 3.3 gives common bounds for every ambient derivative of Gj,n in a fixed neighborhood of the opposite component, (18) gives the C2 bound. The higher-order estimates follow inductively. Indeed, if
∣γi,n∣Cm≤Cm
for some m≥2, then differentiating (18)m−1 times expresses γi,n(m+1) as a finite sum involving only derivatives of γi,n of order at most m and ambient derivatives of Gj,n of order at most m−1. All these terms are uniformly bounded by the induction hypothesis and (17). Hence
∣γi,n∣Cm+1≤Cm+1(L−,L+,d).
Induction proves the result. □
Lemma 3.5. Suppose that non-split curve pairs Γn converge uniformly to a curve pair Γ∞ whose component images are disjoint. Then Γ∞ is non-split.
Proof. If a smooth embedded sphere S separated the two limiting images, compactness would give a positive distance from S to each image. Uniform convergence then places each approximating image, for large n, on the same side of S as its limit. The same fixed sphere would split Γn, a contradiction. □
Combining the preceding normalization, regularity, and uniform estimates, we obtain the following compactness theorem for sequences of weakly critical pairs.
Theorem 3.6. Let Γn∈A be non-split and weakly critical, with E(Γn)→2π2 as n→∞. Up to Möbius transformations and separate reparametrizations of the components, a subsequence converges to H in Cm for every m.
Proof. Discarding finitely many terms, fix Λ with E(Γn)≤Λ. Apply Proposition 3.1 and Lemma 2.1 to obtain two-sided length bounds, a uniform positive chordal separation, and constant-speed parametrizations. The curves are smooth and have uniform Cm bounds for every m by Proposition 3.4. From Arzelà–Ascoli theorem and a diagonal subsequence, we get a limit Γ∞ in every Cm norm.
The lower speed bound persists under C1 convergence, and the separation bound persists under uniform convergence. Thus Γ∞∈A. It is non-split by Lemma 3.5. On the fixed parameter torus the energy densities converge uniformly, since their denominators stay away from zero. Hence
E(Γ∞)=n→∞limE(Γn)=2π2.
The equality case of Theorem 1.1 implies that after a conformal map Φ∈Mob+(S3) (labeling and orientation fixed as explained after (3)), each Φ∘γi,∞ is a regular covering of a standard Hopf circle of some degree mi=0, and
2π2=E(Γ∞)=E(ΦΓ∞)=2π2∣m1m2∣.
So equality forces ∣m1∣=∣m2∣=1, both are circle diffeomorphisms. Thus, g∞⋅Γ∞=H for some g∞∈G. These preserve convergence in every Cm norm and give the stated conclusion. □
The second variation and infinitesimal Möbius action at the Hopf Link
In this section, we firstly compute the normal directional second variation (the Hessian) of the Möbius cross energy at the Hopf Link, and we obtain its kernel. Then we identify the kernel with the infinitesimal Möbius action at the Hopf Link in the normal direction.
At first, we calculate the second variation of the Möbius cross energy at Hopf Link. Set
P1=span{e1,e2},P2=span{e3,e4}.
For the Hopf pair (2), the curve-normal bundles are the constant bundles Nxx=P2 and Nyy=P1. Thus normal fields have the form
U:S1→P2,V:S1→P1.
For k=0,1,2, we set Hk=Hk(S1,P2)⊕Hk(S1,P1) and H−1=(H1)∗. In particular,
Proof. It suffices first to take smooth fields and use the pointwise spherical exponential variations xτ(s)=expx(s)(τU(s)) and yτ(t)=expy(t)(τV(t)). Orthogonality of the two planes gives
Let a(s,t)=⟨U(s),y(t)⟩+⟨x(s),V(t)⟩. For cτ=⟨xτ,yτ⟩, orthogonality of P1 and P2 gives c0=0, c˙0=a, and c¨0=2⟨U,V⟩=0. Therefore
∣xτ−yτ∣21=2(1−cτ)1=21+2τa+2τ2a2+O(τ3).
The first-order term is linear in x(s) and y(t). Since both Hopf circles have vanishing mean, ∫S1x(s)ds=∫S1y(t)dt=0, every such term integrates to zero on the parameter torus, hence δEH=0 and H is critical. Twice the coefficient of τ2 after integration is
Hence the pure squares in ∬a2dsdt equal π∫∣U∣2 and π∫∣V∣2 and cancel the negative zeroth-order terms. The remaining cross term is exactly (20). □
Remark 4.2. In fact, the first variation extends continuously to H1, while the second variation extends to a bounded bilinear form on H1×H1; see Lemma 5.1 for details. These extensions will be used in the local analysis.
Decompose the normal fields into their constant, first-harmonic, and higher-frequency parts:
U(s)=U0+Ax(s)+U≥2(s),V(t)=V0+By(t)+V≥2(t)
where
U0=2π1∫02πU(s)ds,A=π1∫02πU⊗xds,
and V0, B are defined analogously. Thus A∈Hom(P1,P2), B∈Hom(P2,P1), the Fourier expansion of U≥2 contains only frequencies ∣n∣≥2.
If U∈H1, the same orthogonal decomposition holds in H1, because Fourier projection commutes with differentiation.
Proposition 4.3.For every W∈H1 one has the exact identity
δ2EH(W)=π2∣B+AT∣F2+π∫∣U≥2′∣2+π∫∣V≥2′∣2.(22)
where∣A∣F2=tr(ATA). Consequently, the nullspace KH of the bilinear form is
Proof. Only the first Fourier modes contribute to the nonlocal term in (20). Indeed, put
I(U,V)=∫02π∫02π⟨y(t),U(s)⟩⟨x(s),V(t)⟩dsdt.
If U=U0 is constant, integration in s and ∫x(s)ds=0 give I(U0,V)=0; similarly I(U,V0)=0. If U=U≥2, then for each fixed t the function s↦⟨y(t),U≥2(s)⟩ contains only frequencies ∣n∣≥2, whereas s↦⟨x(s),V(t)⟩ has frequency ∣n∣=1. Their integral therefore vanishes. Interchanging s and t shows that every term containing V≥2 also vanishes. Hence
I(U,V)=I(Ax,By).
Now we compute this remaining term. Since ⟨y,Ax⟩=⟨ATy,x⟩, the moment identity (21) gives, after first integrating in s,
because ∫x′⊗x′ds=πIdP1. The second component gives ∫∣(By)′∣2dt=π∣B∣F2 in the same way. Thus the first-mode contribution is
π2(∣A∣F2+∣B∣F2+2tr(AB))=π2∣B+AT∣F2.
so (22) and (23) are proved. Orthogonality to the constant kernel removes U0,V0, while orthogonality to every (Cx,−CTy) gives B=AT, this implies (24).
On frequencies ∣n∣≥2, Poincaré’s inequality gives ∫∣f′∣2≥4∫∣f∣2, and hence
π∫∣f′∣2≥(4π/5)∣f∣H12.
The Fourier pieces are orthogonal, which proves (25). □
The kernel in (23) consists of constant normal fields and mixed rotational fields. We will construct an explicit family of Möbius transformations realizing these velocities.
For A∈Hom(P1,P2), the skew-symmetric matrix
(0A−AT0)
generates a rotation mixing the two orthogonal planes. Its normal trace along the Hopf pair is (Ax,−ATy). The constant normal fields are generated by gradients of height functions. For a=a1+a2∈P1⊕P2, set fa(z)=⟨a,z⟩. On the unit sphere,
∇fa(z)=a−⟨a,z⟩z,∇2fa=−fag0.
Thus ∇fa is a conformal vector field. Its normal trace along the Hopf pair is (a2,a1), since the curve-normal planes along x and y are P2 and P1, respectively.
Accordingly, for the uniquely decomposed kernel field
h=(U0+Ax,V0−ATy)∈KH,
define
a(h)=U0+V0,Ω(h)=(0A−AT0),(26)
and
Xh(z)=Ω(h)z+∇fa(h)(z)=Ω(h)z+a(h)−⟨a(h),z⟩z.(27)
This assignment is linear in h, and (PP2Xh(x),PP1Xh(y))=h. In particular, a(h) is chosen to realize the constant part of h, while Ω(h) realizes its first-harmonic part.
Every Möbius transformation of S3 is a composition of a rotation in SO(4) and a standard conformal diffeomorphism Tb(z) of S3 as in the following:
Tb(z)=(1−∣b∣2)∣z−b∣2z−b−b,z∈S3,(28)
where ∣b∣<1.
We see T0=Id and Tτb is a path of sphere diffeomorphisms for 0≤τ≤1.
dτdτ=0Tτb(z)=−2(b−⟨b,z⟩z)=−2∇fb(z).(29)
Consequently, T−τa/2 has initial velocity ∇fa. This fixes the normalization of the conformal parameter in the following construction.
Proposition 4.4. For h∈KH sufficiently close to zero, define
Φh=exp(Ω(h))∘T−a(h)/2.(30)
Then Φh∈Mob+(S3), the map (h,z)↦Φh(z) is smooth, and Φ0=Id. Moreover,
dτdτ=0Φτh(z)=Xh(z),(31)
so the normal differential of this family along H is the identity on KH. Thus KH is exactly the space of normal infinitesimal Möbius motions of the Hopf pair.
Proof. The exponential in (30) is the ordinary matrix exponential and belongs to SO(4). Linearity of a(h) and Ω(h), together with (29), gives
Φτh(z)=z+τ(Ω(h)z+a(h)−⟨a(h),z⟩z)+O(τ2).(32)
By combining (27), we have (31), i.e. every kernel field is realized as a normal Möbius velocity.
Conversely, let us eliminate the tangential velocity of a Möbius motion by reparametrization, then the energy invariance implies zero second variation in its remaining normal velocity, which therefore belongs to KH.
We now derive the linearized operator from the quadratic form. For W=(U,V) and W=(U,V) define the polarization
BH(W,W):=21(δ2EH(W+W)−δ2EH(W)−δ2EH(W)).
Substitution of (20) and integrating by parts gives
On H0 with domain H2, LH is a self-adjoint operator. It also admits the weak realization
LH:H1⟶H−1
where the second derivatives in (33) are understood distributionally. Moreover,
⟨LHW,W⟩=BH(W,W).
Its kernel is KH. Since KH is finite-dimensional and consists of smooth fields, the L2-orthogonal projections
PKH,P⊥=I−PKH(34)
extend boundedly to H1 and H−1.
Nonlinear estimate and renormalization lemma
In this section, we derive a nonlinear estimate for the first variation near the Hopf link. Motivated by the renormalization lemma of Mondino and Nguyen [7], we then construct a local Möbius gauge in which any sufficiently nearby link is represented as a normal graph over H with graph field orthogonal to KH.
Fix 0<r0<π/4. The normal exponential maps
Ψ1(s,ξ)=expx(s)ξ,(s,ξ)∈S1×Br0(0;P2),(35)
Ψ2(t,η)=expy(t)η,(t,η)∈S1×Br0(0;P1),(36)
are diffeomorphisms onto disjoint tubular neighborhoods of the Hopf circles. For a small field v=(U,V), set
Here v∈H2 has small C1 norm, and ζ∈H2. The notation δG(v)[ζ] denotes its linearization in the direction ζ and satisfies
⟨δG(v)[ζ],η⟩=δ2E(v)[ζ,η].
The graph variation in the first component is
∂ξΨ1(s,U(s))[ζ1(s)].
Its projection to the normal plane of the current curve is an isomorphism at v=0, and remains so for small ∣v∣C1. The second component has the same property. For ∣v∣C1 small, the normal projection of the graph variations onto the normal bundles of the current curves is pointwise invertible. Hence, Proposition 2.2 gives
G(v)=0⟺GraphH(v) is weakly critical.(37)
Lemma 5.1. There are C,r1>0 such that, for v,w∈H2 with ∣v∣C1,∣w∣C1<r1, the linearization δG(v) extends to a uniformly bounded map H1→H−1 and
∣δG(v)−δG(w)∣Lin(H1,H−1)≤C∣v−w∣C1.(38)
Here Lin(X,Y) is the space of bounded linear maps with its operator norm. Moreover,
Let q=(U,U′,V,V′). For ∣v∣C1<r1, the speeds and the separation of the two components are uniformly bounded away from zero. Hence Dqkf is uniformly bounded for $k\leq 3.
To display the differentiation, abbreviate the tuple of values and first derivatives by
zv(s,t)=(U(s),U′(s),V(t),V′(t)).
We define zζ,zη analogously for ζ,η∈H2. Twice differentiating the finite-dimensional integrand gives
⟨δG(v)[ζ],η⟩=∬∂q2f(s,t,zv)[zζ,zη]dsdt.(41)
Since ∂q2f is uniformly bounded, each term in Dq2f(s,t,zv)[zζ,zη] is bounded by a constant times a product ac, where a is one of
∣ζ1(s)∣,∣ζ1′(s)∣,∣ζ2(t)∣,∣ζ2′(t)∣,
and c is the corresponding quantity associated with η. Hence, by Cauchy-Schwarz inequality
we obtain ∣D2E(v)[ζ,η]∣≤C∣ζ∣H1∣η∣H1. Thus D2E(v) extends uniquely to a bounded bilinear form on H1×H1.
Finally, the bound for the third derivatives of f and the mean-value theorem give
s,tsup∣∂q2f(s,t,zv)−∂q2f(s,t,zw)∣≤C∣v−w∣C1.
Applying the same product estimates to this difference proves (38). At v=0 the chart Hessian is the normal Hopf Hessian, so δG(0)=LH. Integration along τv now gives
Write Ψi−1=(πi,νi). For Z=(Z1,Z2) close to H in C1, let
θi=πi∘Zi,ηi=νi∘Zi.
Then θi is a degree-one circle diffeomorphism close to the identity. The field of its normal-graph representative is
ngH(Z):=(η1∘θ1−1,η2∘θ2−1).(42)
Lemma 5.2. On a sufficiently small C1 neighborhood of H, the functions θi,ηi satisfy θi′>1/2, after taking lifts close to the identity. They define degree-one circle diffeomorphisms and
ui=ηi∘θi−1,Zi(r)=Ψi(θi(r),ui(θi(r))).
The assignment Z↦ngH(Z)=(u1,u2) is continuous in C1, with ngH(H)=0. If Z∈H2, then θi,θi−1, and ui belong to H2.
Proof. At Z=H, θi(r)=r and ηi=0. Smoothness of the tubular coordinates gives the lower derivative bound and the identity θi(r+2π)=θi(r)+2π on a small C1 neighborhood. Thus each θi is a degree-one diffeomorphism.
For completeness, suppose Z(n)→Z in C1. Smooth pointwise composition gives θi(n)→θi and ηi(n)→ηi in C1. The inverse lifts are uniformly Lipschitz, with constant at most 2, so their uniform convergence follows from that of the original lifts. The formula
((θi(n))−1)′=(θi(n))′∘(θi(n))−11
and uniform continuity of the limiting derivative give convergence in C1. Applying the chain rule to ui(n) proves ui(n)→ui in C1.
If Z∈H2, the ordinary Sobolev chain rule gives θi,ηi∈H2. Write ψi=θi−1. Almost everywhere,
The derivative bounds for θi and change of variables imply ψi′′∈L2. The first term in ui′′ lies in L2 by the same change of variables, and the second does because ηi′ and ψi′ are bounded. This proves the H2 assertions. □
Let PKH denote the L2-orthogonal projection onto the kernel of the Hopf Hessian. We define M(Z) by
M(Z):=PKHngH(Z).
For h∈KH sufficiently close to zero, let Φh be the Möbius transformation defined in (30). Set
X1=C1(S1,P2)⊕C1(S1,P1),
and define, on sufficiently small zero neighborhoods in KH×X1,
F(h,w)=M(Φh(GraphH(w))).
Based on the above argument, the following lemma, motivated by the renormalization lemma in [7], provides a local Möbius gauge near H: after a small Möbius transformation and reparametrization, a nearby link can be written as a normal graph over H, with its graph field orthogonal to KH.
Lemma 5.3.For every sufficiently small w∈X1, there is a unique sufficiently small h(w)∈KH such that
v=ngH(Φh(w)(GraphH(w)))satisfiesPKHv=0.(43)
Moreover, h(w) depends smoothly on w, v→0 in C1 as w→0, and H2-regularity is preserved.
Proof. Let k1,…,k8 be a smooth L2-orthonormal basis of KH with ka=(ka,1,ka,2). Define
The differentials of the inverse tubular coordinates give
dM(H)[ζ]=PKH(PP2ζ1,PP1ζ2).
The graph map and the finite-dimensional Möbius action are smooth in the C1 topology. Hence F is smooth near (0,0), and F(0,0)=0. By Proposition 4.4, for every k∈KH,
dhF(0,0)[k]=dMH[(Xk∘x,Xk∘y)]=PKHk=k.
Thus dhF(0,0) is the identity on KH. The implicit-function theorem gives a unique smooth small parameter h=h(w), with h(0)=0 and F(h(w),w)=0. This proves (43) and the asserted local uniqueness of the parameter. The remaining continuity and regularity statements follow from Lemma 5.2. □
Proof of the critical-value gap
In this section, we establish a local gap theorem near the Hopf orbit. Combining this local rigidity result with the compactness theorem in Theorem 3.6, we then prove the main theorem.
Theorem 6.1. There is a C1 neighborhood of the Hopf orbit such that every weakly critical pair in this neighborhood belongs to $\mathbf G\cdot\mathbf H.
Proof. By Lemma 5.2, a pair Γ∈A sufficiently close to H in C1 can be separately reparametrized as GraphH(w), with w∈H2 small in C1.
Apply Lemma 5.3 to obtain a Möbius-equivalent graph GraphH(v) with v⊥L2KH. The continuity statements in the two lemmas allow us to shrink the original neighborhood so that ∣v∣C1<r∗, where r∗<r1 and Cr∗<4π/5 for the constants in Lemma 5.1.
Weak criticality is preserved by both operations from Lemma 2.1. We write
R(v)=G(v)−LHv.
then Lemma 5.1 implies LHv=−R(v). Pairing with v gives
54π∣v∣H12≤δ2EH(v)=−⟨R(v),v⟩≤Cr∗∣v∣H12.(45)
Since Cr∗<4π/5, one has v=0. Undoing the transformations proves Γ∈G⋅H. Finally, each fixed element of G acts by a C1 homeomorphism and preserves weak criticality, which proves the statement on the union of translates. □
Proof of Theorem 1.3. Suppose that no positive gap exists. Then, there is a sequence of non-split, weakly critical pairs Γn∈A∖(G⋅H) such that
2π2≤E(Γn)≤2π2+n1.
By Theorem 3.6, a subsequence converges smoothly to H after common Möbius transformations and separate reparametrizations. For large n the resulting pairs lie in the neighborhood of Theorem 6.1, and hence belong to G⋅H. Invariance of this orbit gives a contradiction. □
References
[1]Ian Agol; Fernando C. Marques; André Neves. Min-max theory and the energy of links. J. Amer. Math. Soc. 29 (2016), no. 2, 561–578.DOI
[2]Michael H. Freedman; Zheng-Xu He; Zhenghan Wang. Möbius energy of knots and unknots. Ann. of Math. (2) 139 (1994), no. 1, 1–50.DOI
[3]Jian Ge. Gehring link problem, focal radius and over-torical width. Amer. J. Math. 146 (2024), no. 5, 1295–1304.arxiv.org/abs/2102.05901
[4]Zheng-Xu He. On the minimizers of the Möbius cross energy of links. Experiment. Math.11 (2002), no. 2, 244–248.DOI
[6]O’Hara Jun. Energy of a knot. Topology. 30 (1991), no.2, 241-247DOI
[7]Andrea Mondino; Huy The Nguyen. A gap theorem for Willmore tori and an application to the Willmore flow. Nonlinear Anal. 102 (2014), 220–225.DOI