On the Flucher–Rumpf isoperimetric question for the interior Bernoulli free boundary problem: a conformal reformulation, quantitative bounds and the second variation at the disc
Let c(Ω) be the Bernoulli constant of a bounded domain Ω⊂RN, that is, the infimum of the parameters λ for which the interior Bernoulli free boundary problem Δu=0 in Ω∖K, u=0 on ∂Ω, u=1 and ∣∇u∣=λ on ∂K admits a classical solution. Flucher and Rumpf asked whether c(Ω)≥c(Ω∗), where Ω∗ is the ball of the same volume. The variational analogue was settled by Daners and Kawohl, but the variational constant differs from c already for the ball, and the original question remains open. We prove the following. (i) Through a conformal parametrisation of Ω∖K by an annulus, every classical solution in a simply connected planar domain with connected, filled free set satisfies λ≥μ/R, where πR2=∣Ω∣ and μ≈2.485268; hence c(Ω)≥0.9142c(Ω∗) for every bounded planar convex domain of class C1. (ii) In this class, the Flucher–Rumpf inequality is equivalent to an area inequality for analytic functions in an annulus, which also provides an explicit generator of exact solutions. (iii) A Rellich–Pohozaev identity yields, in every dimension N≥2 and for every strictly star-shaped domain, c(Ω)≥max{θ(Ω),κN}c(Ω∗), where θ(Ω)∈(0,1] is an explicit geometric quantity equal to 1 only for balls and κN is an explicit constant with κ3≈0.650 and κN→1 as N→∞.
Introduction
The interior Bernoulli problem
Let Ω⊂RN, N≥2, be a bounded domain and λ>0. The interior Bernoulli problem consists in finding a compact set K⊂Ω and a function u such that
Δu=0 in Ω∖K,u=0 on ∂Ω,u=1 and ∣∇u∣=λ on ∂K.(1)
A classical solution is a pair (K,u) such that ∂K is of class C1,α, u is continuous on Ω∖K∘, u∈C1(N∖K∘) for a neighbourhood N of ∂K, and (1.1) holds pointwise, the gradient on ∂K being the limit from Ω∖K. Problems of this type arise in ideal fluid flow, electrochemical machining and shape optimisation; see Flucher and Rumpf [11] and the survey [21].
For K⊂Ω compact we denote by
cap(K;Ω)=inf{∫Ω∣∇v∣2dx:v∈W01,2(Ω),v≥1a.e. on K}
the condenser capacity of K relative to Ω, and by uK the capacitary potential, i.e. the minimiser. If (K,u) is a classical solution, then u=uK on Ω∖K and, by Green’s formula,
cap(K;Ω)=∫∂K∣∇u∣dσ=λPer(K).
For convex Ω there is a threshold: by the work of Beurling [3], Acker [1], Henrot and Shahgholian [13, 14], and Cardaliaguet and Tahraoui [7], there exists c(Ω)>0 such that (1) has no solution for λ<c(Ω) and admits a classical solution with K convex for every λ≥c(Ω); moreover, the solution at λ=c(Ω) is unique [7]. In general we set
c(Ω)=inf{λ>0:(1.1) admits a classical solution},(2)
the Bernoulli constant of Ω (with c(Ω)=+∞ if no classical solution exists). Since u is invariant and ∣∇u∣ scales like t−1 under x↦tx, one has c(tΩ)=c(Ω)/t, and c is invariant under rigid motions.
Ω∗ being the ball with ∣Ω∗∣=∣Ω∣. For the ball BR⊂RN the radial solutions are explicit. With K=Bs one has u(r)=log(R/r)/log(R/s) if N=2 and u(r)=(r2−N−R2−N)/(s2−N−R2−N) if N≥3, so that on ∂Bs
in agreement with [4] (case p=2). For λ>c(BR) two radial solutions coexist (the elliptic and the hyperbolic one) and they merge at λ=c(BR): the Bernoulli constant is a turning point (fold) of the solution branch.
Several partial answers are known. Ly and Seck [15] proved, for the p-Laplacian and star-shaped domains, the existence of classical solutions for all parameters above an explicit threshold, which yields an explicit upper bound for the Bernoulli constant. Cardaliaguet and Tahraoui [7] gave an upper bound in terms of the harmonic radius and the lower bound c(Ω)≥c(BR), where BR is the smallest ball containing Ω; this lower bound degenerates for elongated domains. Bianchini and Salani [4] established a Brunn–Minkowski inequality for c in the convex class, from which an Urysohn-type inequality follows: among convex sets of given mean width, the ball minimises c. As they observe, this does not imply (3). Daners and Kawohl [9] considered the minimal parameter Λ(Ω) for which the Alt–Caffarelli functional
Jλ(v)=∫Ω∣∇v∣2dx+λ2∣{v<1}∣,v∈W01,2(Ω),
has a nontrivial minimiser, and proved by Schwarz symmetrisation that Λ(Ω)≥Λ(Ω∗), with equality essentially only for balls. However, Λ and c are different: for the disc Λ(BR)=2e/R≈3.30/R, whereas c(BR)=e/R≈2.72/R; for c(BR)≤λ<Λ(BR) solutions exist, but none of them has lower energy than v≡0. Inequality (3) itself remains open, even for convex planar domains; see also [4, 7].
Main results
In the plane we use conformal mapping. Throughout Sections 2–3, Ω⊂C is a bounded simply connected domain, and we only consider solutions whose free set K is connected and filled (C∖K connected); we call them admissible solutions. For convex Ω of class C1, the solution at λ=c(Ω) is admissible by [14].
Theorem A (quantitative isoperimetric bound in the plane). Let Ω⊂C be bounded and simply connected with ∣Ω∣=πR2, and let (K,u) be an admissible classical solution of (1). Then
where ρ∗≈0.450764 is the unique root in (0,1) of log(1/ρ)=1−ρ2. In particular, for every bounded convex domain Ω⊂R2 of class C1,
c(Ω)≥eμc(Ω∗)≈0.914279c(Ω∗).
To state the converse we introduce, for 0<ρ<1, the annulus Aρ={ρ<∣w∣<1} and the following class.
Definition 1.1. Let A be the set of triples (ρ,m,φ) with ρ∈(0,1), m>0 and φ analytic in Aρ, continuous on Aρ∪{∣w∣=ρ}, with Reφ=0 on {∣w∣=ρ}, such that meφ has a primitive F on Aρ (equivalently, Resw=0eφ=0 in the Laurent sense) which is bounded and extends to an injective continuous map on {ρ≤∣w∣<1}. We write KF for the compact set bounded by the Jordan curve F({∣w∣=ρ}) and ΩF=F(Aρ)∪KF.
Proposition B (equivalent extremal problem). The Flucher–Rumpf inequality λ≥e/R holds for all admissible solutions in all bounded simply connected planar domains if and only if, for every (ρ,m,φ)∈A,
∣ΩF∣≥πm2(eρlog(1/ρ))2.(6)
Moreover, every element of A produces an explicit admissible solution in ΩF.
Proposition B is a reformulation rather than a new inequality: all the difficulty of (3) is transferred to the univalence constraint in Definition 1.1, which is global and non-convex. Its interest lies in the converse part, which generates exact solutions on which (3) can be tested (§ 6).
Proposition C (second variation at the disc; formal). Let k≥1 and Ωε={r<(1+εcoskθ)/1+ε2/2}, so that ∣Ωε∣=π. At the level of a formal expansion (see Remark 3.1),
where xk=e−2k. One has A1=0 (translations), Ak>0 for all k≥2, and Ak∼e(k+1)/2 as k→∞.
The expansion is even in ε, because the rotation θ↦θ+π/k maps Ωε onto Ω−ε. Proposition C is derived by two independent computations (the conformal parametrisation and a domain expansion in the physical plane), which agree identically and match the numerical values to four digits.
Our last main result is not restricted to the plane. For a bounded domain Ω⊂RN of class C1,α which is strictly star-shaped with respect to the origin, i.e. x⋅ν(x)>0 on ∂Ω, set
I(Ω)=∫∂Ωx⋅νdσ,θ(Ω)=I(Ω)I(Ω∗),
where Ω∗ is the ball centred at the origin with ∣Ω∗∣=∣Ω∣. We show in § 4 that θ(Ω)∈(0,1], with θ(Ω)=1 if and only if Ω is a ball centred at the origin. Define also κ2=0 and, for N≥3,
κN=NN−2(N−1)N−2N−1N(N/2)N−21.(7)
Theorem D (Rellich–Pohozaev bounds in every dimension). Let N≥2, let Ω⊂RN be a bounded domain of class C1,α, strictly star-shaped with respect to the origin, and let (K,u) be a classical solution of (1.1) with parameter λ (no assumption is made on the topology of K). Then
λ2N∣K∣=∫∂Ω∣∇u∣2(x⋅ν)dσ+(N−2)cap(K;Ω),(8)
and
λ≥max{θ(Ω),κN}c(Ω∗).
Consequently c(Ω)≥max{θ(Ω),κN}c(Ω∗).
N
3
4
5
10
20
100
κN
0.6495
0.7698
0.8278
0.9234
0.9636
0.9930
Table 1. The constant κN of Theorem D; κN→1 as N→∞.
For N≥3, Theorem D gives a uniform bound of the type of Theorem A for all strictly star-shaped domains, convex or not. For N=2 the coefficient N−2 in (1.8) vanishes; this is why a uniform constant in the plane requires a different tool, here the conformal parametrisation. In the plane, the two results are complementary.
Corollary E. Let Ω⊂R2 be a bounded convex domain of class C1,α and x0∈Ω. Then
c(Ω)≥max{eμ,θx0(Ω)}c(Ω∗),
where θx0 is computed with respect to x0. For an ellipse of aspect ratio q, centred at x0, θ=2/(q+q−1), and the second bound is the better one for q<1.8529.
Finally, § 5 explains why the natural symmetrisation route through polarization fails at the level of subsolutions, while it does work at the level of energies.
Organisation.§ 2 contains the conformal dictionary and the proofs of Theorem A and Proposition B. § 3 is devoted to Proposition C. § 4 proves Theorem D and Corollary E. § 5 deals with polarization. § 6 reports the numerical experiments, and § 7 lists open problems.
Organisation
The conformal dictionary and the proofs of Theorem A and Proposition B
From solutions to analytic functions
2.1. From solutions to analytic functions. Let (K,u) be an admissible classical solution in a bounded simply connected domain Ω⊂C. Since K is connected and filled and ∂K is a C1,α curve, K has nonempty interior and ∂K is a single C1,α Jordan curve. As ∂Ω is a nondegenerate continuum, Ω∖K is a doubly connected domain with two nondegenerate boundary components. Hence there exist a unique ρ∈(0,1) and a conformal map
F:Aρ={ρ<∣w∣<1}⟶Ω∖K
such that F(w)→∂Ω as ∣w∣→1 and F(w)→∂K as ∣w∣→ρ.
Lemma 2.1 (boundary regularity at the inner circle). F extends to an injective map of class C1,α on {ρ≤∣w∣<1}, and F′ does not vanish on {∣w∣=ρ}.
Proof. The Kellogg–Warschawski theorem [19 Thm. 3.6] concerns simply connected domains, so we reduce to that case. Fix z0∈K∘ and let T(z)=1/(z−z0). Then W=T(C∖K) is a bounded simply connected domain containing 0=T(∞), whose boundary T(∂K) is a C1,α Jordan curve. Let g:D→W be a Riemann map. By the Kellogg–Warschawski theorem, g∈C1,α(D) and g′=0 on D. Moreover T(Ω∖K)=W∖E, where E=T(C∖Ω) is a compact connected subset of W.
The map h=g−1∘T∘F is a conformal map of Aρ onto the doubly connected domain D∖g−1(E), and ∣h(w)∣→1 as ∣w∣→ρ. By the Schwarz reflection principle, h extends analytically to a neighbourhood of {∣w∣=ρ} by h(w)=1/h(ρ2/w). This extension is injective near the circle: the two half-collars are mapped into D and into C∖D respectively, and the boundary map is a homeomorphism of circles. Hence h′=0 on {∣w∣=ρ}.
Near the circle, T∘F=g∘h stays away from 0, so F=z0+1/(g∘h) is of class C1,α up to {∣w∣=ρ}. Finally, F′=−(g′∘h)h′/(g∘h)2=0 there. □
The function w↦log∣w∣/logρ is harmonic in Aρ, equal to 0 on ∣w∣=1 and to 1 on ∣w∣=ρ. The function u∘F is bounded and harmonic in Aρ, with boundary limits 0 as ∣w∣→1 (since u is continuous up to ∂Ω) and 1 as ∣w∣→ρ. By uniqueness for the Dirichlet problem,
u∘F(w)=logρlog∣w∣,w∈Aρ.(9)
Set L=log(1/ρ). By conformal invariance, ∣∇u(F(w))∣∣F′(w)∣=∣∇(u∘F)(w)∣=1/(∣w∣L). Letting ∣w∣→ρ and using Lemma 2.1, we obtain on ∂K
∣∇u(F(w))∣=ρL∣F′(w)∣1,∣w∣=ρ.(10)
The Bernoulli condition ∣∇u∣=λ is therefore equivalent to
∣F′(w)∣=m on ∣w∣=ρ,λ=mρL1.(11)
Lemma 2.2. Let h=log∣F′∣, which is harmonic in Aρ. Then for every r∈(ρ,1)
2π1∫02πh(reiθ)dθ=logm.
Proof. For a harmonic function in an annulus, the circular mean M(r) satisfies (rM′(r))′=0, so M(r)=a+βlogr with
by the Cauchy–Riemann equations (r∂rlog∣f∣=∂θargf). Thus β is the winding number of F′ along ∣w∣=r.
The curve γr=F({∣w∣=r}) is an analytic Jordan curve which separates K from ∂Ω. Since F preserves orientation and maps the bounded region {ρ<∣w∣<r} onto the region between ∂K and γr, the curve γr is positively oriented. Its tangent vector is ∂θF(reiθ)=ireiθF′(reiθ). By the Umlaufsatz, the argument of the tangent increases by 2π along γr. The argument of ireiθ also increases by 2π, hence the winding number of F′ is 0.
Thus M is constant on (ρ,1). By Lemma 2.1, h is uniformly continuous on {ρ≤∣w∣≤r1} for every r1<1, so M(r)→2π1∫02πh(ρeiθ)dθ=logm as r→ρ.
Lemma 2.2 shows that logF′ has a single-valued branch in Aρ. Writing F′=meφ, the function φ is analytic in Aρ, continuous up to ∣w∣=ρ, and Reφ=0 there. In Laurent series this means
φ(w)=iγ+n≥1∑(cnwn−cnρ2nw−n),γ∈R,(12)
since on ∣w∣=ρ one has w−n=wnρ−2n and each bracket equals 2iIm(cnwn). Single-valuedness of the primitive F is the condition Resw=0eφ=0. Since F is bounded and, by Lemma 2.1, injective on {ρ≤∣w∣<1}, we have (ρ,m,φ)∈A, K=KF and Ω=ΩF.
From analytic functions to solutions
Conversely, let (ρ,m,φ)∈A with primitive F. Since Reφ=0 on ∣w∣=ρ, the Schwarz reflection φ(w)=−φ(ρ2/w) extends φ analytically across {∣w∣=ρ}. Hence F is analytic across that circle with ∣F′∣=m=0 there, and Γρ=F({∣w∣=ρ}) is an analytic Jordan curve.
Let KF be the closed region it bounds and ΩF=F(Aρ)∪KF. Then ΩF is open, bounded (as F is bounded) and simply connected: its complement is the unbounded component of C∖F(Aρ). Define u=log∣F−1∣/logρ in F(Aρ) and u=1 on KF. Then u is harmonic in ΩF∖KF and of class C1 up to Γρ, since F−1 is analytic up to Γρ. As z→∂ΩF, F−1(z) leaves every compact subset of {ρ≤∣w∣<1}, so ∣F−1(z)∣→1 and u(z)→0. By (10)–(11), (KF,u) is an admissible classical solution of (1.1) in ΩF with λ=1/(mρL). Moreover
∣ΩF∣=∣KF∣+∬Aρ∣F′(w)∣2dA(w).(13)
This gives an explicit way to generate exact solutions, which we use in Section 6.
Proof of Theorem A
By Jensen’s inequality and Lemma 2.2, for every r∈(ρ,1),
It remains to compute μ. Let f(ρ)=1−ρ2/(ρlog(1/ρ)), which tends to +∞ as ρ→0 and as ρ→1. A direct computation gives
ff′(ρ)=−1−ρ2ρ−ρ1+ρlog(1/ρ)1,
which vanishes if and only if log(1/ρ)=1−ρ2. The function ψ(ρ)=log(1/ρ)−1+ρ2 satisfies ψ′(ρ)=2ρ−1/ρ. It is therefore decreasing on (0,2−1/2) and increasing on (2−1/2,1); it tends to +∞ at 0, is negative at 2−1/2 and vanishes at ρ=1. Hence it has a unique root ρ∗∈(0,2−1/2), and at this root f(ρ∗)=1/(ρ∗1−ρ∗2). Numerically, ρ∗=0.4507636… and μ=2.4852675….
If Ω is convex of class C1, then by [14] problem (1.1) has a classical solution with K convex at λ=c(Ω). This solution is admissible, so c(Ω)≥μ/R=(μ/e)c(Ω∗) by (5). □
Remark 2.3. Two inequalities are lossy in this argument: Jensen’s inequality, which is an equality only if ∣F′∣ is constant on each circle, and the bound ∣K∣≥0. For the disc (F(w)=mw,ρ=1/e) one has ∣K∣=πm2ρ2. Any lower bound of the form ∣K∣≥πm2ρ2Θ, with Θ>0 depending on how far 2π1∫∣F′∣2 is from m2, would improve the constant μ/e.
Proof of Proposition B
Assume that (1.6) holds on A, and let (K,u) be an admissible solution in Ω with ∣Ω∣=πR2. The dictionary above gives (ρ,m,φ)∈A with Ω=ΩF and λ=1/(mρL). Then (1.6) reads πR2≥π(e/λ)2, i.e. λ≥e/R.
Conversely, assume the Flucher–Rumpf inequality for admissible solutions. Each (ρ,m,φ)∈A produces, by § 2.2, the admissible solution (KF,u) in ΩF with λ=1/(mρL). Hence λ≥e/R with πR2=∣ΩF∣, that is, ∣ΩF∣≥πe2/λ2=πm2(eρL)2. □
Remark 2.4. The disc corresponds to φ≡iγ and ρ=1/e, where (1.6) is an equality. Inequality (1.6) with the factor (eρlog(1/ρ))2 replaced by 1 is false for ρ close to 1 (see (15) below, where Qk(ρ)<0 for ρ near 1). Thus the factor (eρlog(1/ρ))2≤1, which vanishes as ρ→1, is essential.
Second variation at the disc
Reduction at the turning point
Let Ωε be as in Proposition C. Along the branch of solutions in Ωε, parametrised by a smooth parameter s with s=s0 at the turning point of the disc, we write
λ(s,ε)=λ0(s)+ελ1(s)+ε2λ2(s)+O(ε3).
The rotation θ↦θ+π/k maps Ωε onto Ω−ε, so λ1≡0 and the expansion is even in ε. Since λ0′(s0)=0 and λ0′′(s0)>0, the minimum in s moves by O(ε2), which changes the minimal value only by O(ε4). Hence
c(Ωε)=sminλ(s,ε)=e+ε2λ2(s0)+O(ε4).(14)
For the same reason λ2(s0) does not depend on the choice of parametrisation of the branch. This is the simplest instance of a Lyapunov–Schmidt reduction at a fold.
First derivation: conformal parametrisation
We use the conformal modulus ρ as the branch parameter (ρ0=1/e). Take k≥2, a>0 and
φε(w)=εa(wk−ρ2kw−k),
which is of the form (2.4). All exponents in eφε=∑jgjwj are multiples of k, so g−1=0 and
F(w)=mj∑j+1gjwj+1
is single-valued and analytic in C∖{0}. For ε small it is a C1-small perturbation of mw on a neighbourhood of Aρ, hence univalent there. Thus (ρ,m,φε)∈A and, by § 2.2, we obtain an exact solution with λ=1/(mρL).
Area. If F(eiθ)=∑naneinθ, the area enclosed by the Jordan curve F({∣w∣=1}) is ∣ΩF∣=π∑nn∣an∣2. Here aj+1=mgj/(j+1). From eφ=1+φ+φ2/2+O(ε3) we get
Up to an O(ε) tangential reparametrisation, ∂ΩF is therefore the curve r=m(1+δcoskψ+O(ε2)), with
δ=εaSk(ρ),Sk(ρ)=k+11+k−1ρ2k.
Matching. The O(ε2) corrections to φ needed to match a prescribed outer boundary exactly involve only the modes 0 (with imaginary coefficient, i.e. a rotation) and ±2k. They do not affect (3.2) at order ε2. A mode 2k of amplitude O(δ2) in the boundary changes c only at order δ4: by rotational invariance, the cubic terms of the expansion of c couple it only to the square of the k-th mode. Normalising ∣ΩF∣=π and using (2.3), the solution of conformal modulus ρ in the domain with boundary amplitude δ satisfies
λ(ρ,δ)2=(ρlog(1/ρ))21+δ2Qk(ρ)/Sk(ρ)2+O(δ4).
By (3.1), c=e(1+δ2Qk(1/e)/(2Sk(1/e)2))+O(δ4), that is,
Ak=2Sk(1/e)2eQk(1/e).(16)
With xk=e−2k one has Qk(1/e)=[(k−1)−2(k2−1)xk−(k+1)xk2]/(k2−1) and Sk(1/e)=[(k−1)+(k+1)xk]/(k2−1), which gives the formula of Proposition C.
Second derivation: domain expansion
We check (3.3) by a direct expansion in the physical plane. Let Ωε={r<1+εcoskθ}, whose area is π(1+ε2/2), and write c=coskθ. Let
∂K={r=s+εbc+ε2g2(θ)},∫02πg2dθ=0,
so that s is the mean radius, and set ℓ=logs<0. We look for
u=u0+εu1+ε2u2,u0=ℓlogr,u1=(αrk+βr−k)c,
where u2 has a radial part c0+d0logr plus modes that average to zero on circles. Below, ⟨⋅⟩ denotes the mean in θ; note ⟨c2⟩=21.
First order. The conditions u=0 on the outer boundary, u=1 on ∂K and ∣∇u∣ constant on ∂K give
α+β=−ℓ1,αsk+βs−k=−sℓb,k(αsk−βs−k)=sℓb.
With B=b/(sℓ) and D(s)=(k−1)s−k+(k+1)sk, the solution is
The right-hand side is exactly (k2−1)[(k−1)−2(k2−1)xk−(k+1)xk2], so both derivations agree identically.
Positivity
For k=1 the factor k2−1 vanishes, so A1=0, as it must be, since translations leave c invariant. For k≥2, (k+1)xk=(k+1)e−2k≤3e−4<0.055, because k↦(k+1)e−2k is decreasing. Hence
By rotational invariance, the second variation is invariant under rotations; on each two-dimensional space spanned by coskθ and sinkθ the rotation group acts irreducibly, so the second variation is a multiple of the identity there. Thus it is the diagonal quadratic form
which is positive definite modulo translations and, since Ak≍k, coercive in the H1/2 norm of the boundary perturbation.
Remark 3.1. To turn Proposition C into a rigorous local minimality statement, one needs: (a) the implicit function theorem for the branch of solutions near the fold, in Hölder spaces, to justify the expansions; (b) a stability argument of Dambrine–Pierre type [8] to pass from H1/2-coercivity of the second variation to strict local minimality in a C2,α neighbourhood (the usual two-norm discrepancy). We do not carry this out here. The numerical values in Table 4 confirm the formula independently.
k
ε=0.03
ε=0.06
extrapolated
Ak (Prop. C)
2
3.25860
3.26208
3.2574
3.257430
3
5.27815
5.28396
5.2762
5.276246
4
6.76165
6.75119
6.7651
6.765339
5
8.13388
8.08991
8.1485
8.149293
Table 4. Second variation coefficients.
A Rellich–Pohozaev identity and the proof of Theorem D
Throughout this section Ω⊂RN, N≥2, is a bounded domain of class C1,α, strictly star-shaped with respect to the origin, and (K,u) is a classical solution of (1.1). Set D=Ω∖K. By the Schauder estimates, u∈C1,α up to ∂Ω, and u∈C1 up to ∂K by assumption, so u∈C2(D)∩C1(D). We denote by ν the outer unit normal to Ω on ∂Ω and to K on ∂K.
Two integral identities
Lemma 4.1.
∫∂Ω∣∇u∣dσ=∫∂K∣∇u∣dσ=cap(K;Ω)=λPer(K).
Proof. By the strong maximum principle 0<u<1 in D, so ∣∇u∣=−∂νu on ∂Ω and ∣∇u∣=∂νu on ∂K (one-sided derivatives from D). Since u is harmonic, ∫∂D∂νDudσ=0, which gives the first equality. Green’s formula ∫D∣∇u∣2=∫∂Du∂νDudσ=∫∂K∣∇u∣dσ and the minimality of uK=u give the second; the third follows from ∣∇u∣=λ on ∂K. □
so divV=22−N∣∇u∣2. Since V∈C(D)∩C1(D), divV∈L1(D) and ∂D is of class C1, the divergence theorem applies (e.g. by exhausting D by smooth subdomains).
On each component of ∂D, u is constant, so ∇u=(∂νDu)νD and V⋅νD=21(x⋅νD)∣∇u∣2. On ∂K, νD=−ν and ∣∇u∣=λ, while ∫∂Kx⋅νdσ=N∣K∣ by the divergence theorem in K. Using Lemma 4.1,
Lemma 4.3.I(Ω)≥N∣Ω∣∣∂Ω∣2≥I(Ω∗)=R∣∂Ω∗∣, where ∣Ω∗∣=∣BR∣. Hence θ(Ω)∈(0,1], with equality if and only if Ω is a ball centred at the origin.
Proof. By Cauchy–Schwarz, ∣∂Ω∣2≤I(Ω)∫∂Ωx⋅νdσ=I(Ω)N∣Ω∣. The isoperimetric inequality gives N∣Ω∣∣∂Ω∣2≥N∣BR∣∣∂BR∣2=R∣∂BR∣=I(BR). Equality in the isoperimetric inequality forces Ω to be a ball; equality in Cauchy–Schwarz forces x⋅ν to be constant on ∂Ω, so the ball is centred at the origin. □
Proof of Theorem D
Step 1: a lower bound by the capacity. By Lemma 4.1 and Cauchy–Schwarz,
Extend u by 1 on K and by 0 outside Ω, and let u∗ be its Schwarz symmetrisation. By the Pólya–Szegő inequality [18, 6], u∗∈W01(BR) and ∫∣∇u∗∣2≤∫∣∇u∣2. Since 0<u<1 in D, the set {u∗=1} is the closed ball of measure ∣{u=1}∣=∣K∣, namely Bs. Thus u∗ is admissible for cap(Bs;BR), and
Step 3: the bound θ. For the concentric radial solution K=Bs in Ω=BR, the gradient ∣∇u∣=C(s)/∣∂BR∣ is constant on ∂BR and x⋅ν=R. Hence both Cauchy–Schwarz steps above are equalities, and (8) reads
λrad(s)2N∣Bs∣=I(BR)C(s)2+(N−2)C(s).(20)
Since I(Ω)=I(BR)/θ with θ=θ(Ω)≤1 (Lemma 4.3), (19) gives
The function t↦t2−tN is maximal on (0,1) at tN−2=2/N, where it equals t2(N−2)/N. Hence
λ2≥NN−2ΛN(R)2,ΛN(R)=RN(N/2)N−21.
Dividing by c(BR)=(N−1)N−2N−1/R gives λ≥κNc(BR) with κN as in (7). Since c(Ω∗)=c(BR), Theorem D follows by taking the infimum over all classical solutions. □
Remark 4.4.ΛN(R) is the Daners–Kawohl constant of BR (for N=2, Λ2(R)=2e/R). Comparing Jλ(v) with Jλ(0) shows that Λ(Ω)2=infKcap(K;Ω)/∣K∣. Dropping only the boundary term in (1.8), Step 4 yields more precisely
c(Ω)≥NN−2Λ(Ω)(N≥3).
so that, for star-shaped domains, the two constants c and Λ are comparable. Combined with Λ(Ω)≥Λ(Ω∗)[9], this gives the bound κN again.
Remark 4.5. As N→∞, Rc(BR)=N−1+logN+o(1) and RΛN(R)=N+log(N/2)+o(1), so κN→1. For Ω close to a ball, I(Ω)/I(Ω∗)−1 is quadratic in the deformation (it is minimal at the ball), so the bound θ deviates from 1 only at second order. This is the same order as in Proposition C, but with the opposite sign: it does not prove local minimality.
Proof of Corollary E
A bounded convex domain of class C1,α is strictly star-shaped with respect to any interior point x0, since (x−x0)⋅ν≥dist(x0,∂Ω)>0. By translation invariance of c, Theorem D applies with the origin at x0, and Theorem A gives the other bound. For the ellipse x=(acost,bsint) centred at x0=0, one has dσ=a2sin2t+b2cos2tdt and x⋅ν=ab/a2sin2t+b2cos2t, so
I=abπ(a2+b2),θ=a2+b22ab=q+q−12,q=ba.
The inequality θ>μ/e is equivalent to q+q−1<2e2/μ2≈2.39261, i.e. q<1.8529. □
Polarization: an obstruction and what survives
For a closed half-space H with reflection σ, the polarization of a set E is
EH=((E∪σE)∩H)∪((E∩σE)∩Hc),
and the polarization of a function v is max(v,v∘σ) on H and min(v,v∘σ) on Hc; see [5]. A natural strategy towards (1.3) is to show that c(ΩH)≤c(Ω) and to conclude by approximating Ω∗ by iterated polarizations. If c admits the Beurling-type characterisation
c(Ω)=K⊂Ωinf∂Ksup∣∇uK∣,
where uK is the capacitary potential of K in Ω, it would suffice to transport subsolutions, i.e. to deduce sup∂KH∣∇uKH∣≤λ in ΩH from sup∂K∣∇uK∣≤λ. For convex domains, such a characterisation is closely related to the subsolution methods of Beurling [3], Acker [1] and Henrot–Shahgholian [14]; our numerical tests (§ 6.5) support it. The next lemma shows that the polarized free set itself is never a good competitor. Note that it does not disprove c(ΩH)≤c(Ω): other free sets in ΩH may still do better.
Lemma 5.1 (corner obstruction). Assume that ∂K and ∂(σK) are C1,α curves that cross transversally at a point P∈int(Hc)∩ΩH. Then sup∂KH∣∇uKH∣=+∞.
Proof. Near P, KH coincides with K∩σK, which has at P a convex corner of interior angle ϑ∈(0,π). Hence ΩH∖KH has at P an opening angle ω=2π−ϑ>π. The function v=1−uKH is positive and harmonic in ΩH∖KH and vanishes on both arcs of ∂KH through P. By the corner asymptotics [12 Ch. 4–5], after straightening the two arcs by a conformal map, v=arπ/ωsin(πϕ/ω)+o(rπ/ω), with a>0 by the Hopf lemma at corners. Hence ∣∇v∣∼a(π/ω)rπ/ω−1 along ∂KH, which is unbounded since π/ω<1. □
On the side H, the union K∪σK creates reentrant corners of KH, at which the gradient vanishes. The obstruction is therefore entirely located in Hc, where the polarized function is a minimum of harmonic functions and is only superharmonic. The critical free sets of nonsymmetric convex domains are smooth and convex, and for most H they cross their reflections transversally. What survives is the energy inequality.
Proposition 5.2. For K⊂Ω set Eλ(K;Ω)=cap(K;Ω)−λ2∣K∣. Then, for every λ,
Eλ(KH;ΩH)≤Eλ(K;Ω).
Proof. Extend uK by 0 outside Ω; it is continuous on RN and {uK>0}=Ω, {uK≥1}=K. Its polarization uKH is continuous, vanishes outside {uK>0}H=ΩH, and therefore belongs to W01,2(ΩH). It satisfies uKH≥1 on {uK≥1}H=KH, and has the same Dirichlet integral as uK[5]. Hence cap(KH;ΩH)≤cap(K;Ω), while ∣KH∣=∣K∣. □
By the Hadamard formula δcap(K;Ω)=∫∂K∣∇uK∣2Vndσ, solutions of (1) are exactly the critical points of Eλ. For the disc, the elliptic solution is a local minimum and the hyperbolic one a saddle, and they merge into an inflection point at λ=c. Proposition 5.2 implies that polarization decreases the Daners–Kawohl constant Λ (the threshold for infEλ<0). It also decreases the capacitary profile CΩ(v)=inf{cap(K;Ω):∣K∣=v}. But c is the threshold of existence of a critical point, not an energy level, and it is not monotone under a decrease of Eλ.
In the convex class, Cardaliaguet and Tahraoui [7] studied the profile
FΩ(σ)=inf{cap(D;Ω):D⊂⊂Ω convex,∣D∣=σ}.
They showed that, at points where FΩ is differentiable, convex minimisers are solutions of level FΩ′(σ). This suggests the following question.
Question 5.3. Find a characterisation of c(Ω) in terms of Eλ (or of FΩ) which is monotone under pointwise decrease of the energy. For instance: is
c(Ω)=inf{λ:Eλ(⋅;Ω) has a strict local minimiser},
and does a strict local minimiser for Ω produce one for ΩH?
Numerical experiments
Method
We solve (1) in the plane by the method of fundamental solutions. The potential is represented as u(x)=a0+∑jajlog∣x−zj∣, with source points zj placed outside Ω and inside K. The Dirichlet conditions u=0 on ∂Ω and u=1 on ∂K are imposed in the least-squares sense at 240–320 collocation points. The free boundary is parametrised in polar form, ρ(θ)=s+∑k≤M(akcoskθ+bksinkθ), M=12–20. For fixed mean radius s, the equation ∣∇u∣=λ on ∂K is solved for (ak,bk,λ) by a Levenberg–Marquardt method. The Bernoulli constant is then the minimum of λ(s) along the branch. Every domain is normalised to area π, so that c(Ω∗)=e.
Validation. For the unit disc the method returns c=2.718281828459045 at s=0.36787943, i.e. e and 1/e to machine precision. Capacities of concentric annuli are reproduced to 10−14. Residuals of the free boundary condition are below 10−5 in all reported runs, and usually below 10−10.
Ellipses and Steiner symmetrisation
Table 2 shows that c increases with the aspect ratio of ellipses. Values are stable to 10−9 between M=12 and M=20. For aspect ratio 3 the polar parametrisation of K is no longer adequate, and we do not report it. The last column gives the lower bound of Corollary E.
q
c(Ω)
c(Ω)/e
max{μ/e,θ}
1
2.718282
1
1
1.2
2.745282
1.009933
0.991769
1.25
2.758675
1.014860
0.987730
1.5
2.850462
1.048626
0.960769
2
3.095436
1.138747
0.914279
Table 2. Ellipses of area π and aspect ratio q.
To test the Steiner route we consider Ωf={(x,y):∣y−f(x)∣<0.81−x2}, whose Steiner symmetral with respect to y is the ellipse with semi-axes 1 and 0.8. Table 3 shows that symmetrisation decreases c in all cases, including a non-convex one.
Domain
f(x)
c/e
Steiner symmetral (ellipse)
0
1.014860
Mild banana
0.25x2
1.024310
S-shape
0.4x3
1.032890
Strong banana (non-convex)
0.5x2
1.049881
Table 3. Steiner symmetrisation; values normalised to area π.
Second variation
For Ωε as in Proposition C, we computed (c(Ωε)−e)/ε2 at ε=0.03 and ε=0.06 and Richardson-extrapolated in ε2, which is legitimate since the expansion is even in ε. The agreement with the closed formula is about 10−4 (Table 4).
A capacitary Cheeger inequality fails
Since cap(K;Ω)=∫∂K∣∇uK∣dσ, one has sup∂K∣∇uK∣≥cap(K;Ω)/Per(K), with equality for concentric discs. One might hope to prove (1.3) via infKcap(K;Ω)/Per(K)≥c(Ω∗). This is false. Oscillating boundaries make Per(K) arbitrarily large while the capacity stays bounded, so the infimum is 0 for every Ω. In the unit disc, already ρ(θ)=s(1+0.2cos4θ) gives cap/Per=2.5169<e, while sup∣∇uK∣=3.9453>e. The supremum cannot be replaced by a mean.
Testing the Beurling characterisation
In the unit disc, 129 random compact sets K (five Fourier modes, mean radius in [0.2,0.45]) all satisfy sup∂K∣∇uK∣≥3.65>e. Random perturbations of the critical free set Kc, with 40 samples for each amplitude, always increase the supremum. The smallest increments are 0.12, 0.64 and 1.43 for amplitudes 0.002, 0.01 and 0.03 in the disc, and 0.18, 0.64 and 1.67 for the ellipse q=2.
Testing Proposition B on exact solutions
The converse part of the dictionary turns every element of A into an exact solution. We sampled ρ∈[0.15,0.8] and 2 to 5 modes cn in (2.4), and adjusted c1 to enforce Reseφ=0. We computed F and the areas spectrally on each circle (with 1024 Fourier modes), and checked univalence (simple, positively oriented, nested boundary curves). Among 313 univalent random samples, the minimum of λR/e is 1.000032. We also ran 24 Nelder–Mead minimisations of λR/e under the univalence constraint, starting from random configurations. All stopped above 1, the lowest value being 1.0012 near the disc (ρ=0.366). No violation of (1.6) was found.
The Rellich–Pohozaev identity
For the critical solution in the ellipse q=1.5, the two sides of (1.8) (with N=2, i.e. 2λ2∣K∣=∫∂Ω∣∇u∣2x⋅νdσ) evaluate to 6.72954776011166 and 6.72954776011170. The bound of Theorem D, θ=0.960769, is compared with the computed values of c/e in Table 2.
Concluding remarks and open problems
Sharp constant in Theorem A. The constant 0.914 comes from neglecting ∣K∣ and from Jensen’s inequality. A quantitative stability estimate, saying that circle means of ∣F∣2 close to m2 force F to be close to a homothety and hence ∣K∣≈πm2ρ2, would improve it. Proving (1.6) would settle (1.3) for convex planar domains.
Higher dimensions. Improve the constant κN of Theorem D, in particular for N=3, where κ3≈0.650. The boundary term ∫∂Ω∣∇u∣2x⋅νdσ, which is only estimated through I(Ω), carries the missing information.
Local minimality. Make Proposition C rigorous along the lines of Remark 3.1. The domain expansion of Section 3 extends to N≥3 with spherical harmonics, u1=(αrℓ+βr2−N−ℓ)Yℓ; it remains to extend it to the p-Laplacian.
Polarization. Answer the Question of Section 5.
Riemannian setting. Bernoulli problems on manifolds have been studied by Djité and Seck [10] (exterior problem, p-Laplacian, compact manifolds) and, in the one-phase framework, by Peng, Zhang and Zhu [16]. The argument of [9] only uses the characterisation of Λ as a capacity/volume quotient and the Pólya–Szegő inequality, which hold for geodesic symmetrisation on SN and HN. We therefore expect Λ(Ω)≥Λ(B) for geodesic balls B of the same volume in these space forms. The classical constant c is open there as well.
Acknowledgements
Acknowledgements. The authors acknowledge the use of Claude (Anthropic), an AI assistant, for symbolic and numerical verifications, for implementing the numerical experiments and for help in drafting and revising the manuscript. All results, proofs and computations were checked by the authors, who take full responsibility for the content of this paper.
Université Cheikh Anta Diop de Dakar, Sénégal Email address:[email protected]
Université Cheikh Anta Diop de Dakar, Sénégal Email address:[email protected]
References
[1]A. Acker, Uniqueness and monotonicity of solutions for the interior Bernoulli free boundary problem in the convex n-dimensional case, Nonlinear Anal. 13 (1989), 1409–1425.DOI
[3]A. Beurling, On free-boundary problems for the Laplace equation, Seminars on Analytic Functions, vol. 1, Institute for Advanced Study, Princeton, 1957, 248–263.
[4]C. Bianchini and P. Salani, Concavity properties for elliptic free boundary problems, arXiv:0810.4842.arxiv.org/abs/0810.4842
[5]F. Brock and A. Yu. Solynin, An approach to symmetrization via polarization, Trans. Amer. Math. Soc. 352 (2000), 1759–1796.DOI
[6]J. E. Brothers and W. P. Ziemer, Minimal rearrangements of Sobolev functions, J. Reine Angew. Math. 384 (1988), 153–179.DOI
[7]P. Cardaliaguet and R. Tahraoui, Some uniqueness results for Bernoulli interior free-boundary problems in convex domains, Electron. J. Differential Equations 2002, No. 102, 1–16.
[8]M. Dambrine and M. Pierre, About stability of equilibrium shapes, M2AN Math. Model. Numer. Anal. 34 (2000), 811–834.DOI
[9]D. Danners and B. Kawohl, An isoperimetric inequality related to a Bernoulli problem, Calc. Var. Partial Differential Equations 39 (2010), 547–555.
[10]A. S. Djité and D. Seck, Shape optimization and Bernoulli free boundary problems for the Riemannian p-Laplacian, arXiv:2609.00571.arxiv.org/abs/2609.00571
[11]M. Flucher and M. Rumpf, Bernoulli’s free-boundary problem, qualitative theory and numerical approximation, J. Reine Angew. Math. 486 (1997), 165–204.DOI
[12]P. Grisvard, Elliptic Problems in Nonsmooth Domains, Pitman, Boston, 1985.
[13]A. Henrot and H. Shahgholian, Existence of classical solutions to a free boundary problem for the p-Laplace operator. I. The exterior convex case, J. Reine Angew. Math. 521 (2000), 85–97.DOI
[14]A. Henrot and H. Shahgholian, Existence of classical solutions to a free boundary problem for the p-Laplace operator. II. The interior convex case, Indiana Univ. Math. J. 49 (2000), 311–323.DOI
[15]I. Ly and D. Seck, Isoperimetric inequality for an interior free boundary problem with p-Laplacian operator, Electron. J. Differential Equations 2004, No. 109, 1–12.
[16]Yu. Peng, H.-C. Zhang and X.-P. Zhu, One-phase free boundary problems on Riemannian complexes, Tr. Mat. Inst. Steklova 330 (2025).
[17]S. I. Pohozaev, On the eigenfunctions of the equation Δu + λf(u) = 0, Soviet Math. Dokl. 6 (1965), 1408–1411.
[18]G. Pólya and G. Szegő, Isoperimetric Inequalities in Mathematical Physics, Princeton University Press, 1951.
[19]Ch. Pommerenke, Boundary Behaviour of Conformal Maps, Springer, Berlin, 1992.DOI
[20]F. Rellich, Darstellung der Eigenwerte von Δu + λu = 0 durch ein Randintegral, Math. Z. 46 (1940), 635–636.DOI
[21]D. Seck, On an isoperimetric inequality and various methods for the Bernoulli’s free boundary problems, São Paulo J. Math. Sci. 10 (2016), 36–59.DOI