Bull. Math. Sci. (2015) 5:517–607DOI 10.1007/s13373-015-0074-xThe quantitative isoperimetric inequality and relatedtopicsNicola Fusco1Received: 17 May 2015 / Revised: 25 August 2015 / Accepted: 29 August 2015 /Published online: 15 September 2015© The Author(s) 2015. This article is published with open access at SpringerLink.comAbstract We present some recent stability results concerning the isoperimetricinequality and other related geometric and functional inequalities. The main tech-niques and approaches to this field are discussed.Contents1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5182 Notation and preliminary results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5213 The Fuglede’s approach . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5263.1 Nearly spherical sets . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5263.2 Convex sets . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5304 The quantitative isoperimetric inequality: the approach via symmetrization . . . . . . . . . . . 5344.1 Reduction to n-symmetric sets . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5374.2 Reduction to axially symmetric sets . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5465 Other proofs and extensions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5545.1 Isoperimetric inequality via mass transport . . . . . . . . . . . . . . . . . . . . . . . . . . 5545.2 Quantitative isoperimetric inequality via regularity . . . . . . . . . . . . . . . . . . . . . . 5595.3 An improved version of the quantitative isoperimetric inequality . . . . . . . . . . . . . . 5625.4 The isoperimetric inequality in higher codimension . . . . . . . . . . . . . . . . . . . . . 5715.5 The isoperimetric inequality on the sphere . . . . . . . . . . . . . . . . . . . . . . . . . . 5735.6 The fractional perimeter . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5775.7 The isoperimetric inequality in Gauss space . . . . . . . . . . . . . . . . . . . . . . . . . 5806 Other functional inequalities . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5846.1 Stability of Sobolev inequality . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 584Communicated by Ari Laptev.B Nicola Fuscon.fusco@unina.it1 Dipartimento di Matematica e Applicazioni “R. Caccioppoli”, Università degli Studi di Napoli“Federico II”, Naples, Italy123518 N. Fusco6.2 The Faber–Krahn inequality . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5956.3 Further results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 600References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6031 IntroductionThe isoperimetric inequality is probably one of the most beautiful and at the sametime one of the most powerful tools of modern mathematics. Despite the fact that theisoperimetric property of the circle among all planar figures was already well knownin ancient times, the first rigorous proofs of this fact have been only obtained at thebeginning of last century. In particular Hurwitz proved in 1901 that given a simple,closed curve γ ⊂ R2 of length L enclosing an area A, thenL2 ≥ 4π A, (1.1)with equality holding if and only if γ is a circle. His proof, see [86], used the Gauss–Green formula andWirtinger inequality that is the one-dimensional Sobolev–Poincaréinequality with the sharp constant. This is not surprising since in higher dimensionthe connections of the isoperimetric inequality with the divergence theorem and withthe Sobolev inequality are even more evident.Passing to higher dimensions was not an easy task for the mathematicians of lastcentury, also because it was clear that proving the isoperimetric inequality in fullgenerality would require a deeper understanding of the concept of surface measure.In the end, three main directions were followed to this purpose. The first one, andmaybe the simplest, was based on the Brunn–Minkowski inequality, see for instance[29, Th. 8.1.1], stating that if H, K ⊂ Rn are compact sets then|H + K | 1n ≥ |H | 1n + |K | 1n . (1.2)From this inequality one gets easily that if K ⊂ Rn is compact thenσ(∂K ) ≥ σ(∂ Br ), (1.3)where Br is a ball of the same volume as K and σ(∂K ) denotes the outer Minkowskicontent of the boundary of K , defined asσ(∂K ) := lim infε↓0|K + ε B̄| − |K |ε,where B̄ is the closed unit ball. Note that if K is the closure of a bounded smooth openset, then σ(∂K ) coincides with the usual measure of the surface ∂K . Therefore (1.3)can be viewed as a generalized version of the classical isoperimetric inequality.Another approach to the isoperimetric inequality comes from Calculus of Varia-tions. Provided that an isoperimetric set exists and is a smooth bounded open set, bytaking small variations of this set one immediately sees that its boundary must have123The quantitative isoperimetric inequality. . . 519constant mean curvature. This observation led people to prove that the ball is the onlysmooth open set inRn whose boundary has constant mean curvature, a result that waseventually obtained by Aleksandrov in 1958, see [2].A third approachwas followed byDeGiorgi in his celebrated paper [53] of 1958.Heused an observation going back to Steiner: the surface area of a set E , ormore generallyits perimeter P(E), decreases under Steiner symmetrization, see Theorem 2.7. Theisoperimetric property of balls then follows, since they are the only sets that areinvariant under Steiner symmetrization in every direction. The formulation given byDe Giorgi in the framework of the theory of sets of finite perimeter is the most generalone. It states that if E ⊂ Rn is a Lebesgue measurable set with the same volume as aball Br thenP(E) ≥ P(Br ), (1.4)with the equality holding if and only if E is a ball.Other proofs of the isoperimetric inequality have been proposed more recently.Among them we shall discuss the one given by Gromov in [100], see Sect. 5.1, whichuses ideas related to mass transportation and ultimately reduces the isoperimetricinequality to the divergence theorem and the arithmetic-geometric mean inequality.Finally, an elegant and short proof has been given by Cabré in [30] using the propertiesof solutions of the Neumann problem for the Laplacian. We cannot pursue here thehistory of all proofs and formulations of the isoperimetric inequality, but we referthe interested reader to the survey paper of Ossermann [104] and to the book ofChavel [37].Let us go back to the isoperimetric inequality in the plane. In the 1920’s Bonnesenobtained a series of improved versions of (1.1), later on named by Osserman [105]Bonnesen type inequalities, of the formL2 − 4π A ≥ d,where the quantity d on the right hand side is a nonnegative geometrically significantexpression vanishing if and only if γ is a circle. In particular in 1924 Bonnesen [21]proved the following sharp and elegant estimate.Theorem 1.1 (Bonnesen) Given a closed, simple curve γ ⊂ R2 enclosing a convexset C of area A, there exist two concentric circles C1 ⊂ C ⊂ C2 of radii r1 and r2,respectively, such that(r2 − r1)2 ≤ L2 − 4π A4π.This inequality has the feature of being sharp, since the constant 4π at the denomina-tor cannot be increased, and of having an elementary proof. It provides a nice exampleof quantitative isoperimetric inequality, i.e., an inequality where a suitable distanceof a set E from a ball of the same volume is controlled in terms of the difference ofthe perimeter of E and the perimeter of the ball.It was only in 1989 that Fuglede extended this result to convex sets in higherdimension. Indeed, he proved in [72] that given a convex body K with the same123520 N. Fuscovolume as the unit ball B one can always measure the Hausdorff distance of a translateof K from B by a suitable power of the isoperimetric deficit of K , i.e., the differenceP(K ) − P(B), see Theorem 3.2.His result was the starting point of modern investigations on the stability of isoperi-metric inequality. In fact only a fewyears later Fuglede’s result was extended to generalsets of finite perimeter in Rn in two papers by Hall et al. [84] and by Hall [82]. Pre-cisely, in the last paper it was proved that if E ⊂ Rn is a measurable set with the samevolume as the unit ball B, thenminx∈Rn{|(x + E)B|}4 ≤ C(n)[P(E) − P(B)], (1.5)where C(n) is a positive constant depending only on the dimension. Note that indealing with general sets one has to replace the Hausdorff distance with the weakerone given by the measure of the symmetric difference between a translated of theset E and B. This is unavoidable if one thinks that by adding to B some long thinspikes the volume and the perimeter of the resulting set change very little while itsHausdorff distance from B goes to infinity. Note also that in (1.5) the power 4 on theleft hand side is independent of the dimension, though Hall suggested in [82] that itshould be replaced with the optimal power 2, see the discussion at the beginning ofSect. 4.Inequality (1.5) with the power 2 was proved in [76] by Maggi, Pratelli and theauthor using symmetrization arguments. This result was not the end of the story. Onthe contrary it initiated a series of papers where other proofs of the same inequalitywere given or similar stability results were obtained for other inequalities such as theSobolev, the Faber–Krahn, the Brunn–Minkowski and the isodiametric inequalities,see Sects. 5 and 6.The aim of this paper is to present the results, the techniques and the main ideasdeveloped in this context. Three different approaches are now available if one wantsto study these stability issues: a first one using symmetrization, another one via masstransportation and a third one based on the regularity theory for minimal surfaces andother PDEs.The paper is organized as follows. After quickly recalling some basic tools fromDe Giorgi’s theory of sets of finite perimeter and from Geometric Measure Theory,in Sect. 3 we present Fuglede’s proof of the quantitative isoperimetric inequality forconvex sets and for nearly spherical sets, that are sets close to a ball in C1 sense.In rewriting the original proofs of [72] we have tried to simplify some unnecessar-ily complicated parts and to highlight some ideas contained therein that have beensuccessfully used later by various authors in different contexts.In Sect. 4 we present the proof of the quantitative isoperimetric inequality via sym-metrization given in [76], with some simplifications introduced by Maggi in [94].Although a shorter proof has been given later by Cicalese and Leonardi [45] using theregularity theory for area minimizing sets of finite perimeter, the proof by symmetriza-tion uses no deep results from geometric and functional analysis, is geometricallyintuitive and can be adapted to other situations where a complete regularity theory isnot available.123The quantitative isoperimetric inequality. . . 521The stability proofs via mass transportation and via regularity are presented inSect. 5. Not all the details of the proofs are given, but we have tried to explain the mainideas and the novel techniques. In this sectionwediscuss also several generalizations ofthe quantitative isoperimetric inequality to higher codimension, to non local perimetersand to non Euclidean settings such as the Gauss space.Finally, Sect. 6 gives an account on how the techniques developed for the isoperi-metric inequality have been successfully applied to study the stability of other relatedinequalities. The quantitative versions of the Sobolev inequality and of the Faber–Krahn inequality for the first eigenvalue of the Laplacian are discussed with somedetails. Reference to other inequalities are given in the last section. But at that pointwe had to stop since for some of them new developments are foreseen in the nextyears.2 Notation and preliminary resultsIn the following we denote by Br (x) the ball with radius r > 0 and center x and weuse the following simplified notationBr := Br (0), B(x) := B1(x), B := B(0).The measure of the unit ball B will be denoted by ωn .Though most of the results that we are going to present could be given first forsmooth sets and then extended to sets of finite perimeters via the approximation The-orem 2.4 below, we have preferred to state them in full generality. However the readerwho is not familiar with the theory of sets of finite perimeter may still read the state-ments and the proofs without problems by simply pretending that all sets involved aresmooth. For the basic properties of sets of finite perimeter and related results we referto [8,59]. We start by recalling the definition. Let be an open set in Rn and E ⊂ Rna measurable set. The perimeter of E in is defined asP(E; ) := sup{∫Edivϕ dx : ϕ ∈ C∞c ( ; Rn), ||ϕ||∞ ≤ 1}. (2.1)If E is smooth we can use the classical divergence theorem to find that∫Edivϕ dx =∫∂ E∩ ϕ · νdHn−1,thus, taking the supremum over all ϕ as in the definition, we getP(E; ) = Hn−1(∂ E ∩ ).123522 N. FuscoNote that if P(E; ) < +∞ the mapϕ ∈ C1c ( ,Rn) →∫Edivϕ dxis linear and continuous, so by means of Riesz’s theorem, we can find a vector valuedRadon measure μ = (μ1, . . . , μn) such that∫Edivϕ dx =∫ ϕ · dμ =n∑i=1∫ ϕi dμi .Thusμ = −Dχ E , where Dχ E is the distributional derivative ofχ E and the aboveformula can be rewritten as∫Edivϕ dx =∫ χ E divϕ dx = −∫ ϕ · d Dχ E . (2.2)So, E has finite perimeter in if and only if Dχ E is a Radon measure with valuesin Rn and finite total variation. In fact, from the definition we immediately get thatP(E; ) = |Dχ E |( ).Recalling Besicovitch differentiation theorem, we also have that for |Dχ E |-a.e.x ∈ supp|Dχ E |, there existsνE (x) := − limr→0Dχ E (Br (x))|Dχ E |(Br (x))and |νE (x)| = 1. (2.3)We shall denote by ∂∗E the set of all points in supp|Dχ E |where (2.3) holds. The set∂∗E is called the reduced boundary of E , while the vector νE (x) is the generalizedexterior normal at x . From (2.3) we have that the measure Dχ E is obtained byintegrating −νE with respect to |Dχ E |, i.e.,Dχ E = −νE |Dχ E |.Thus (2.2) can be rewritten as∫E∩ divϕdx =∫∂∗ E∩ ϕ · νE d|Dχ E |, ∀ϕ ∈ C1c ( ,Rn). (2.4)Since ∂∗E ⊂ supp|Dχ E | ⊂ ∂ E , the reduced boundary of E is a subset of thetopological boundary. The next result (see e.g. [59, Sect. 5.7] or [8, Th. 3.59]) describesthe structure of sets of finite perimeter. Here and in the sequel byHk , k = 0, 1, . . . , n,we denote the k-dimensionalHausdorff measure inRn . For the definition and the basicproperties of rectifiable sets see [8, Ch. 2].123The quantitative isoperimetric inequality. . . 523In the following the perimeter of E in Rn will be denoted by P(E). If P(E) < ∞we say that E is a set of finite perimeter.Theorem 2.1 (De Giorgi) Let E ⊂ Rn be a measurable set of finite perimeter, thenthe following hold:(i) ∂∗E is countably (n−1)-rectifiable, i.e., ∂∗E = ⋃i Ki ∪N0, whereHn−1(N0) =0 and Ki are compact subsets of C1 manifolds Mi of dimension n − 1;(ii) |Dχ E | = Hn−1 ∂∗E, in particular P(E; ) = Hn−1(∂ E∗ ∩ ) for any openset ⊂ Rn;(iii) for Hn−1-a.e. x ∈ Ki , the generalized exterior normal νE (x) is orthogonal tothe tangent hyperplane to the manifold Mi at x;(iv) for all x ∈ ∂∗E, |E∩Br (x)|Br (x) → 12 as r → 0;(v) for all x ∈ ∂∗E, limr→0 Hn−1(∂∗ E∩Br (x))ωn−1rn−1 = 1.From (ii) above we have that (2.4) can we rewritten as∫Edivϕ dx =∫∂∗ E∩ ϕ · νE dHn−1 for all ϕ ∈ C1c ( ,Rn).Another characterization of the reduced boundary is contained in the followingresult. To this aim we recall that if E ⊂ Rn is a measurable set E has density t ∈ [0, 1]at a point x ∈ Rn ifD(E; x) := limr→0+|E ∩ Br (x)||Br | = t.We shall denote by E (t) := {x ∈ Rn : D(E; x) = t}. Then the measure theoreticboundary of E is defined as ∂M E := Rn\(E (0) ∪ E (1)). For the proof of the nexttheorem see [8, Th. 3.61].Theorem 2.2 (Federer) Let E be a set of finite perimeter in Rn. Then∂∗E ⊂ E (1/2) ⊂ ∂ M E and Hn−1(Rn\(E (0) ∪ ∂∗E ∪ E (1)) = 0.Let Eh be a sequence of measurable subsets ofRn . We say that the sets Eh convergein measure to a set E ⊂ Rn in an open set ⊂ Rn if χ Eh → χ E in L1( ). Thelocal convergence in measure is defined accordingly. Note that from the definition(2.1) of perimeter it follows immediately that the perimeter is lower semicontinuouswith respect to the local convergence in measure, i.e. if the sets Eh converge locallyin measure in to a set E , then P(E; ) ≤ lim infh P(Eh; ). Note also, as aconsequence of Rellich–Kondrachov theorem, that sets of finite perimeter are compactwith respect to the convergence in measure, see [8, Th. 3.39].Theorem 2.3 Let Eh be a sequence of measurable subsets of Rn such thatsuph P(Eh; ) < ∞. Then, there exists a subsequence Ehk locally converging inmeasure in to a set E of finite perimeter in . Moreover,123524 N. FuscoP(E; ) ≤ limk→∞ P(Ehk ; ).Another useful property of sets of finite perimeter is stated in the next approximationresult.Theorem 2.4 Let E be a set of finite perimeter in . Then there exists a sequenceof smooth, bounded open sets Eh ⊂ Rn such that Eh → E in measure in andP(Eh; ) → P(E; ).Note also that in Theorem 2.4 one may replace the smooth sets Eh with polyhedra,i.e. bounded open sets obtained as the intersection of finitely many half-spaces.The next result is a special case of the coarea formula for rectifiable sets, see [8,Th. 2.93]. To this aim, given k = 1, . . . , n − 1, we set x = (x ′, y) ∈ Rn−k × Rk .Similarly, given a vector v = (v1, . . . , vn) ∈ Rn we set vx ′ = (v1, . . . , vn−k), vy =(vn−k+1, . . . , vn), and given a measurable set E ⊂ Rn we denote the section of Eover x ′ ∈ Rn−k by Ex ′ := {y ∈ Rk : (x ′, y) ∈ E}.Theorem 2.5 Let E be a set of finite perimeter in Rn. Given k ∈ {1, . . . , n − 1},for Hn−k-a.e. x ′ ∈ Rn−k the slice Ex ′ is a set of finite perimeter in Rk andHk−1(∂∗(Ex ′)(∂∗E)x ′) = 0. Moreover if g : Rn → [0,+∞] is a Borel function wehave∫∂∗ Eg(x)νEy (x)dHn−1(x) =∫Rn−kdx ′∫(∂∗ E)x ′g(x ′, y)dHk−1(y).Fix k and a measurable set E ⊂ Rn . Let vE : Rn−k → [0,∞) be the functionmeasuring the k-dimensional slices of E , i.e.,vE (x′) := Hk(Ex ′) for Hn−k-a.e. x ′ ∈ Rn−k . (2.5)Note that vE is Hn−k-measurable. Moreover, if E is a set of finite perimeter, vEis a function of bounded variation (BV -function) and even a Sobolev function if thereduced boundary of E has no vertical parts. For the definition and themain propertiesof BV functions we refer to [8, Ch. 3]. The following result is a consequence ofLemma 3.1 and Propositions 3.4 and 3.5 of [11].Theorem 2.6 Let E be a set of finite perimeter with finite measure and k ∈{1, . . . , n −1}. Then the function vE defined in (2.5) belongs to BV (Rn−k). Moreoverif Hn−1({(x ∈ ∂∗E : νEy (x) = 0}) = 0, then vE ∈ W 1,1(Rn−k). In addition, ifU ⊂ Rn−k is a Borel setP(E; U × Rk) ≥∫U√pE (x ′)2 + |∇vE (x ′)|2 dx ′ + |DsvE |(U ),where ∇vE and DsvE are the absolutely continuous part and the singular part, respec-tively, of the gradient measure DvE with respect to Lebesgue measure in Rn−k andfor Hn−k-a.e. x ′ ∈ Rn−k we have set pE (x ′) := Hk−1(∂∗(Ex ′)).123The quantitative isoperimetric inequality. . . 525Given a measurable set E ⊂ Rn and k ∈ {1, . . . , n − 1} we define the essentialprojection of E over the first n − k coordinate hyperplanes by settingπ(E)+ := {x ′ ∈ Rn−k : vE (x ′) > 0}.If x ′ ∈ π(E)+ we denote by rE (x ′) the radius of a k-dimensional ball with Hk-measure equal to the one of Ex ′ . Then the Steiner symmetral of E of codimension kwith respect to the subspace {y = 0} is defined by settingE S := {(x ′, y) ∈ Rn−k × R : x ′ ∈ π(E)+, |y| < rE (x ′)}.Note that by definition vE (x ′) = vE S (x ′) for Hn−k-a.e. x ′ ∈ Rn−k , hence |E | =|E S|. On the other hand perimeter decreases under Steiner symmetrization. Precisely,we have the following result whose proof can be found for instance in [11, Section 3].Theorem 2.7 Let E ⊂ Rn be a set of finite perimeter and E S its Steiner symmetralwith respect to to first n−k coordinate hyperplanes. Then E S is a set of finite perimeterand for any Borel set U ⊂ Rn−k we haveP(E S; U × Rk) ≤ P(E; U × Rk). (2.6)Moreover,P(E S; U × Rk) =∫U√pEs (x ′)2 + |∇vE S (x ′)|2 dx ′ + |DsvE S |(U ).Let us apply the above formula to a set of finite perimeter E , axially symmetricwith respect to the xn axis and let us set v(t) := Hn−1({x ′ ∈ Rn−1 : (x ′, t) ∈ E}).For any open set U ⊂ R we haveP(E;Rn−1 × U ) =∫U√(n − 1)2ω2n−1n−1v(t)2(n−2)n−1 + v′(t)2 + |Dsv|, (2.7)provided n ≥ 3 or n = 2 and v > 0 in U .A characterization of the equality cases in the perimeter inequality (2.6), undersuitable assumptions on the set E , has been given in [38] for the case k = 1 and in[11] for the higher codimension case. The complete characterization of the equalitycases for the codimension 1 case has been recently proven in [33].We conclude this section by recalling the coarea formula for functions. Thoughit holds more generally for functions of bounded variations, we are going to apply itonly to Sobolev functions. In this case we have, see for instance [8, Th. 3.40].Theorem 2.8 Let ⊂ Rn be an open set and f ∈ W 1,1loc ( ). Then, for every Borelfunction g : → [0,∞] we have∫ g(x)|∇ f | dx =∫ ∞−∞dt∫{ f =t}g(x) dHn−1(x). (2.8)123526 N. Fusco3 The Fuglede’s approach3.1 Nearly spherical setsIn this section, following Fuglede [72], we consider a nearly spherical set, that is aset E ⊂ Rn whose boundary can be written as a graph over the boundary of the ballwith the same volume and the same barycenter as E . Thus, up to a translation and adilation, we shall assume that there exists a Lipschitz function u : Sn−1 → (−1, 1)such thatE = {y ∈ Rn : y = t x(1 + u(x)), with x ∈ Sn−1, 0 ≤ t < 1}, (3.1)that |E | = |B| and that E has barycenter at the origin.Fuglede’s result states that if E is sufficiently close to B, then the isoperimetricdeficit, that is the difference between P(E) and P(B), controls the L2 norm of thetangential gradient ∇τ u.Theorem 3.1 There exists ε(n) > 0 such that if E is as in (3.1), |E | = |B|, thebarycenter of E is the origin and ‖u‖W 1,∞(Sn−1) ≤ ε, thenP(E) − P(B) ≥ 14‖∇τ u‖2L2(Sn−1) ≥18ωn|EB|2. (3.2)Proof Step 1 Let us prove that if E is as in (3.1) thenHn−1(∂ E) =∫Sn−1√(1 + u)2(n−1) + (1 + u)2(n−2)|∇τ u|2 dHn−1. (3.3)In fact, from the area formula we have thatHn−1(∂ E) =∫Sn−1Jn−1ϕ dHn−1, (3.4)where Jn−1ϕ is the (n − 1)-dimensional Jacobian of the map ϕ(x) = x(1 + u(x)),x ∈ Sn−1. Recall that, if TxSn−1 is the tangential plane to Sn−1 at x , then Jn−1ϕ =√det((dxϕ)∗ ◦ dxϕ), where the linear map dxϕ : TxSn−1 → Rn is the tangentialdifferential ofϕ at x and (dxϕ)∗ : Rn → TxSn−1 denotes the adjoint of the differential.Since for any τ ∈ TxSn−1 we have dxϕ(τ) = τ(1+u(x))+ x Dτ u(x), the coefficientsof the matrix dxϕ relative to an orthonormal base {τ1, . . . , τn−1} of TxSn−1 and tothe standard base {e1, . . . , en} are τi · eh(1 + u(x)) + xh Dτi u, for i = 1, . . . , n −1, h = 1, . . . , n. Thus, for all i, j ∈ {1, . . . , n − 1} the coefficients ai j of the matrix(dxϕ)∗ ◦ dxϕ are given byai j =n∑h=1(τi · eh(1 + u) + xh Dτi u)(τ j · eh(1 + u) + xh Dτ j u)= δi j (1 + u)2 + Dτi u Dτ j u,123The quantitative isoperimetric inequality. . . 527where in the last equality we have used the fact that τi · τ j = δi j and τi · x = 0 for alli, j = 1, . . . , n−1.Hence, recalling that fora, b ∈ Rk onehas det(I+a⊗b) = 1+a·b,we have thatJn−1ϕ =√det(ai j ) =√(1 + u)2(n−1) + (1 + u)2(n−2)|∇τ u|2and thus (3.3) follows immediately from (3.4). Let us extend u by setting u(x) :u(x/|x |) for all x = 0. Then E = (B), where : B → E is the map defined bysetting (x) := x(1 + u(x)). Note that D(x) = (1 + u(x))I + x ⊗ Du and thatx · Du(x) = 0 for all x = 0. Thus, we conclude that the n-dimensional Jacobian of is given by (1 + u(x))n . Hence, the assumptions |E | = |B| and ∫E y dy = 0 canbe rewritten as follows1n|B|∫Sn−1(1 + u(x))n dHn−1 = 1,∫Sn−1x(1 + u(x))n+1 dHn−1 = 0. (3.5)Step 2 Using (3.3) we haveP(E) − P(B) =∫Sn−1⎡⎣(1 + u)n−1√1 + |∇τ u|2(1 + u)2 − 1⎤⎦ dHn−1=∫Sn−1[(1 + u)n−1 − 1]dHn−1+∫Sn−1(1 + u)n−1⎡⎣√1 + |∇τ u|2(1 + u)2 − 1⎤⎦ dHn−1.From the Taylor expansion of the square root it follows that for t > 0 sufficientlysmall√1 + t ≥ 1+ t2 − t27 . Hence from the smallness assumption ‖u‖W 1,∞(Sn−1) < εwe getP(E) − P(B) ≥∫Sn−1[(1 + u)n−1 − 1]dHn−1+∫Sn−1(1 + u)n−1[12|∇τ u|2(1 + u)2 −17|∇τ u|4(1 + u)4]dHn−1≥∫Sn−1[(1 + u)n−1 − 1]dHn−1 +(12− Cε)∫Sn−1|∇τ u|2dHn−1.(3.6)Note that the first equation in (3.5), that is the assumption |E | = |B|, implies that∫Sn−1[(1 + u)n − 1]dHn−1 = 0,123528 N. Fuscothat is ∫Sn−1(nu +n∑h=2(nh)uh)dHn−1 = 0. (3.7)From this identity, recalling again that ‖u‖L∞(Sn−1) < ε, we have∫Sn−1udHn−1 ≥ −n − 12∫Sn−1u2dHn−1 − Cε∫Sn−1u2dHn−1.Therefore, using this last inequality and the smallness assumption, wemay estimate∫Sn−1[(1 + u)n−1−1]dHn−1 = (n − 1)∫Sn−1udHn−1+n−1∑h=2(n − 1h)∫Sn−1uhdHn−1≥ (n − 1)∫Sn−1udHn−1 + (n − 1)(n − 2)2∫Sn−1u2dHn−1 − Cε∫Sn−1u2dHn−1≥ −n − 12∫Sn−1u2dHn−1 − Cε∫Sn−1u2dHn−1.In conclusion, recalling (3.6), we have proved that if ‖u‖W 1,∞(Sn−1) ≤ ε, thenP(E)−P(B) ≥(12− Cε)∫Sn−1|∇τ u|2dHn−1−(n − 12+ Cε)∫Sn−1u2dHn−1.(3.8)Step 3 Now, for any integer k ≥ 0, let us denote by yk,i , i = 1, . . . , G(n, k), thespherical harmonics of order k, i.e., the restriction to Sn−1 of the homogeneous har-monic polynomials of degree k, normalized so that ||yk,i ||L2(Sn−1) = 1, for all kand for i ∈ {1, . . . , G(n, k)}. For instance, since ∫Sn−1 1dHn−1 = nωn and, fori = 1, . . . n, ∫Sn−1 x2i dHn−1 = n−1∫Sn−1 |x |2dHn−1 = ωn , we have y0 = 1/√nωnand y1,i = xi/√ωn , respectively. The functions yk,i are eigenfunctions of theLaplace–Beltrami operator on Sn−1 and for all k and i−Sn−1 yk,i = k(k + n − 2)yk,i .Therefore if we writeu =∞∑k=0G(n,k)∑i=1ak,i yk,i , where ak,i =∫Sn−1uyk,i dHn−1,we have||u||2L2(Sn−1) =∞∑k=0G(n,k)∑i=1a2k,i , ||∇τ u||2L2(Sn−1) =∞∑k=1k(k +n −2)G(n,k)∑i=1a2k,i . (3.9)123The quantitative isoperimetric inequality. . . 529Observe that from formula (3.7) we havea0 = 1√nωn∫Sn−1udHn−1 = − 1n√nωnn∑h=2(nh)∫Sn−1uhdHn−1,hence|a0| ≤ C ||u||22 ≤ Cε||u||2.Similarly, from the second equality in (3.5), that is the assumption that the barycen-ter of E is at the origin, using the equality∫Sn−1 xi = 0, we immediately get that forall i = 1, . . . , n,|a1,i | =∣∣∣∣ 1√ωn∫Sn−1uxi dHn−1∣∣∣∣ ≤ Cε||u||2.Therefore, from (3.9) we get||u||22 ≤ Cε2||u||22 +∞∑k=2G(n,k)∑i=1|ak,i |2 ⇒ ||u||22 ≤11 − Cε∞∑k=2G(n,k)∑i=1|ak,i |2.But since for k ≥ 2, k(k + n − 2) ≥ 2n, from (3.9) we have||u||22 ≤12n(1 − Cε) ||∇τ u||22and thus, recalling (3.8) and choosing ε sufficiently small, in dependence on n, we getP(E) − P(B) ≥(12− Cε)∫Sn−1|∇τ u|2dHn−1−(n − 12+Cε)12n(1 − Cε) ||∇τ u||22≥ 14∫Sn−1|∇τ u|2dHn−1 ≥ n3||u||2L2(Sn−1) ≥13ωn||u||2L1(Sn−1).(3.10)This proves the first inequality in (3.2). To get the second inequality we observethat, choosing again ε sufficiently small|EB| = 1n∫Sn−1|(1 + u(x))n − 1| dHn−1 ≤ n + 1n∫Sn−1|u| dHn−1.Therefore, from the last inequality of (3.10) we conclude thatP(E) − P(B) ≥ 13ωn||u||2L1(Sn−1) ≥n23(n + 1)2ωn |EB|2 ≥ 18ωn|EB|2.123530 N. Fusco3.2 Convex setsWe shall now discuss the case of convex sets for which Fuglede in [72] proved aquantitative estimate where the distance from a ball is measured in terms of the Haus-dorff distance. To this aim we recall that if E, F are any two sets in Rn the Hausdorffdistance between E, F is defined asdH (E, F) := inf{ε > 0 : E ⊂ F + Bε, F ⊂ E + Bε}.In this section we shall always assume that K is a closed convex set with non emptyinterior and |K | = |B|. For such a set we define the isoperimetric deficit D(K ) andthe asymmetry index A(E) asD(K ) := P(K ) − P(B), A(K ) := minx∈Rn dH (K , B(x)),respectively. With these definitions in hand the result proved in [72, Th. 2.3] reads asfollows.Theorem 3.2 (Fuglede) Let n ≥ 2. There exist δ, C, depending only on n, such thatif K is convex, |K | = |B|, and D(K ) ≤ δ, then:A(K ) ≤⎧⎪⎪⎪⎪⎨⎪⎪⎪⎪⎩C√D(K ), n = 2√D(K ) log(1D(K )), n = 3CD(K ) 2n+1 , n ≥ 4.(3.11)To be precise, in his paper [72] Fuglede only deals with the case n ≥ 3, since whenn = 2 the result was already known with a better estimate and a more elementaryproof. Indeed, the case of planar convex sets was already studied by Bernstein [13]in 1905 and by Bonnesen [21] in 1924, see Theorem 1.1 in Sect. 1. Note that theestimates (3.11) are sharp with respect to the order of magnitude as D(K ) → 0, see[72, Sect. 3].The proof of Theorem 3.2 is based on the quantitative estimate proved in the pre-vious section. The key point is the observation that a convex sets K with the samevolume as the unit ball and small isoperimetric deficit is nearly spherical. Indeed, theboundary of K can be written as the graph of a Lipschitz function u over the boundaryof the sphere centered at the barycenter of K with ‖u‖W 1,∞ small. This is preciselythe content of the next lemma.Lemma 3.3 For all ε > 0, there exists δε > 0 such that if K is a closed convex,|K | = |B|, the barycenter of K is the origin and D(K ) < δε, then there exists aLipschitz function u : Sn−1 → (0,∞), with ‖u‖W 1,∞(Sn−1) ≤ ε, such thatK = {t x(1 + u(x)) : x ∈ Sn−1, 0 ≤ t ≤ 1}. (3.12)123The quantitative isoperimetric inequality. . . 531Proof Note that if K is any closed convex set with barycenter at the origin and non-empty interior, then K can be always represented as in (3.12) for some Lipschitzfunction u : Sn−1 → (0,∞). The point here is to show that the W 1,∞ norm of u issmall when D(K ) is small.Step 1 We start by proving a weaker statement, namely that for any ε > 0 there existsδε > 0 such that if K is a closed convex set with barycenter at 0, |K | = |B| andD(K ) < δε, then K can be represented as in (3.12) for some Lipschitz function u with‖u‖L∞(Sn−1) ≤ ε.To prove this we argue by contradiction assuming that there exist ε > 0 and asequence of closed convex sets K j such that |K j | = |B|, the barycenter of K j is theorigin, D(K j ) → 0, but ||u j ||L∞(Sn−1) ≥ ε0, where u j is the function representingK j as in (3.12).The following fact is well known, see for instance [58, Lemma 4.1]. If n ≥ 2 thereexists C(n) such that for any convex set K with non empty interiordiam(K ) ≤ C(n) P(K )n−1|K |n−2 .Using this inequality we deduce that the sets K j are equibounded and so, up to a notrelabeled subsequence, wemay assume that they converge in the Hausdorff distance toa closed set K . Note that K is convex and that |K jK | → 0 as j → ∞. Therefore, bythe lower semicontinuity of the perimeter we have that P(K ) ≤ lim inf j→∞ P(K j ).Therefore, since D(K j ) → 0, we conclude that P(K ) = P(B) and thus that K is theclosed unit ball centered at the origin. This gives a contradiction, since it is not possiblethat the sets K j converge in the Hausdorff sense to B̄, while ||u j ||L∞(Sn−1) ≥ ε0 forall j .Step 2 Let us now assume that ‖u‖L∞(Sn−1) ≤ ε for some ε < 1/2. To concludethe proof of the lemma we show that if this is the case, then the following strongerinequality holds‖∇τ u‖∞ ≤ 2√‖u‖∞ 1 + ‖u‖∞1 − ‖u‖∞ . (3.13)To prove this estimate we observe that if K is represented as in (3.12), then it is nottoo hard to show that for Hn−1-a.e. x ∈ Sn−1 the exterior normal νK (y(x)) to ∂K atthe point y(x) = x(1 + u(x)) is given byνK (y(x)) = x(1 + u(x)) − ∇τ u(x)√(1 + u(x))2 + |∇τ u(x)|2. (3.14)Thus, if x ∈ Sn−1 is a point where u is differentiable from the previous equality wehave thatx · νK (y(x)) = 1 + u(x)√(1 + u(x))2 + |∇τ u(x)|2,123532 N. FuscoOz(x)νK(y(x))νB(x)xFig. 1 The construction in Lemma 3.3where we used the fact that x · ∇τ u(x) = 0. Then, denoting by z(x) the projection ofthe origin on the tangent plane to K at y(x), the triangle with vertices 0, z(x), y(x) isrectangle at z(x) and the angle at the origin is equal to the angle formed by the vectorsx and νK (y(x)), see Fig. 1. Therefore |z(x)| = |y(x)|(x · νK (y(x))). Observe alsothat|y(x)| ≤ 1 + ||u||∞, |z(x)| ≥ 1 − ||u||∞,where the last inequality follows from the convexity of K , and thus1 − ||u||∞1 + ||u||∞ ≤ x · νK (y(x)) = 1 + u(x)√(1 + u(x))2 + |∇τ u(x)|2,from which we get|∇τ u(x)|2(1 + u(x))2 ≤(1 + ||u||∞1 − ||u||∞)2− 1 = 4||u||∞(1 − ||u||∞)2 ,thus concluding|∇τ u(z)|2 ≤ 4||u||∞(1 + ||u||∞1 − ||u||∞)2,whence (3.13) follows. The proof of Theorem 3.2 now follows quite easily from the lemma we have justproved and from the next interpolation result, whose proof is given in [72, Lemma 1.4].123The quantitative isoperimetric inequality. . . 533Lemma 3.4 If v ∈ W 1,∞(Sn−1) and ∫Sn−1 v = 0, then||v||n−1L∞(Sn−1) ≤⎧⎪⎪⎪⎨⎪⎪⎪⎩π ||∇τ v||2, n = 24||∇τ v||22 log8e||∇τ v||∞||∇τ v||22, n = 3C(n)||∇τ v||22||∇τ v||n−3∞ , n ≥ 4,where the constant C(n) depends only on the dimension and e is the Neper number.Proof of Theorem 3.2 Fix ε > 0. Thanks to Lemma 3.3 we may assume that K is aconvex set represented as in (3.12) by means of a function u such that ‖u‖W 1,∞ < ε.Set v := (1+u)n−1n . From the volume constraint |K | = |B| we have, see the firstequality in (3.5),∫Sn−1v dHn−1 = 1n∫Sn−1[(1 + u)n − 1]dHn−1 = 0.Moreover, sincev = u + 1nn∑h=2(nh)uh,if ε > 0 is small enough we have12|u| ≤ |v| ≤ 2|u|, 12|∇τ u| ≤ |∇τ v| ≤ 2|∇τ u|.Assume now, to fix the ideas, that n ≥ 4. If ε is smaller than or equal to the oneprovided by Theorem 3.1, from the interpolation Lemma 3.4 and (3.13) we have‖u‖∞ ≤ 2‖v‖∞ ≤ C(n)‖∇τ v‖2n−12 ‖∇vτ‖n−3n−1∞≤ C(n)‖∇uτ‖2n−12 ‖∇τ u‖n−3n−1∞ ≤ C(n)‖∇τ u‖2n−12 ‖u‖n−32(n−1)∞ ,hence ‖u‖∞ ≤ C(n)‖∇τ u‖4n+12 . Thus, recalling the first inequality in (3.2), we mayconclude thatA(K ) ≤ dH (K , B) = ‖u‖∞ ≤ CD(E) 2n+1 .The cases n = 2, 3 are proved in the same way. When passing from a convex set to a general set E of finite perimeter one cannotexpect to estimate the isoperimetric deficit with the Hausdorff distance from a ball,see the discussion at the beginning of the next section. However, this is still possibleif some additional structure is imposed on the set E . At this regard we mention two123534 N. Fuscoresults that have been recently obtained in this direction. The first one, proved in[73], deals with sets satisfying a uniform interior cone condition at the boundary. Tosimplify the statement we shall assume as before that E is a set with the same volumeas the unit ball. The general case can be recovered by suitably rescaling all the relevantquantities, see [73, Theorem 1.1].Given x ∈ Rn , r > 0 and ν ∈ Sn−1 the spherical sector with vertex at x , axis ofsymmetry parallel to ν, radius r and aperture π/2 is defined asSν,r (x) = {y ∈ Rn : |y − x | < r,√2〈y − x, ν〉 > |y − x |}.Then, we say that a closed set E satisfies the interior cone condition at the boundarywith radius r and aperture π/2 if for any x ∈ ∂ E , there exists νx ∈ Sn−1 such thatSνx ,r (x) ⊂ E and we denote by Cr the family of all closed sets E , with |E | = |B|,satisfying the interior cone condition at the boundary with radius r . It can be easilychecked [73, Lemma 2.9], that if E ∈ Cr and D(E) is sufficiently small then E iscompact.The interior cone condition at the boundary is a rather mild regularity condition.In fact, given any > 0 and any θ ∈ (0, π/2) one may construct a set E ∈ Cr whichdoes not satisfy the standard interior cone condition with height and aperture θ ,see [73, Example 2.3]. On the other hand, a set in Cr has always finite perimeter [73,Prop. 2.4], but the Hn−1-measure of its topological boundary may be strictly largerthan its perimeter, [73, Example 2.6]. The main result proved in [73] then reads asfollows.Theorem 3.5 For any r > 0 there exist δ, C, depending only on n and r, such that ifE ∈ Cr and D(E) ≤ δ, then:A(E) ≤⎧⎪⎪⎪⎪⎨⎪⎪⎪⎪⎩C√D(E), n = 2√D(E) log(1D(E)), n = 3CD(E) 2n+1 , n ≥ 4.(3.15)We observe that the powers appearing in (3.15) are the same as in Theorem 3.2.Another estimate with the same powers has been also obtained by Rajala and Zhongin [106, Th. 1.5] for John domains whose complement with respect to a suitable ballis also a John domain. Note that though the sets considered in [106] do not necessarilybelong to Cr , they cannot have singularities such as inward cusps, which are insteadadmissible for sets in Cr .4 The quantitative isoperimetric inequality: the approach viasymmetrizationWe now discuss the quantitative isoperimetric inequality for general sets of finiteperimeter. In this case it is clear that we cannot use the Hausdorff distance to measurethe asymmetry of a set. Think for instance of a set which is the union of a large ball123The quantitative isoperimetric inequality. . . 535and a far away tiny one. For this reason we replace the asymmetry index consideredin the previous section with an L1-type distance, the so-called Fraenkel asymmetry,which is defined for a measurable set E , as the minimum of the symmetric differencebetween E and any ball with the same measureα(E) := minx∈Rn{ |EBr (x)||Br | : |E | = |Br |}. (4.1)We shall refer to a ball minimizing α(E) as to an optimal ball. Note that an optimalball needs not to be unique. Note also that α(E) is scaling invariant. It is convenientto define the isoperimetric deficit of a set E also in a scaling invariant way by settingD(E) := P(E) − P(Br )P(Br ),where again r is the radius of a ball with the same measure of E .The first quantitative estimate for sets of finite perimeter was obtained in 1992by Hall [82]. Using some previous results proved in collaboration with Hayman andWeitsman [84], he showed that there exists a constantC(n) such that for all measurablesets of finite measureα(E)4 ≤ C(n)D(E). (4.2)It is interesting to observe that the power 4 on the left hand side of this estimatedoes not depend on the dimension, while in Fuglede’s Theorem 3.2 it does. However,in his paper Hall conjectured that the right power should be 2 and actually provedthat (4.2) holds with the exponent 2 if E is an axially symmetric set, see [82, Th. 2].Moreover he observed that one cannot expect a power smaller than 2. To see this, takeε > 0 and consider in any dimension n ≥ 2 the ellipsoidEε ={x211 + ε + x22 (1 + ε) + x23 + · · · + x2n ≤ 1}.It can be proved, see Proposition 4.3, that α(Eε) = |EεB|. Then one can showthatD(Eε)α2(Eε)→ γ > 0, as ε → 0+.In 2008 Maggi, Pratelli and the author proved in [76] the following quantitativeisoperimetric inequality with the sharp exponent.Theorem 4.1 There exists a constant γ (n) such that for any measurable set E of finitemeasureα(E)2 ≤ γ (n)D(E). (4.3)123536 N. FuscoNote that inequality (4.3) can be rewritten in the following equivalent way: if|E | = |Br |,P(E) ≥ P(Br )(1 + α(E)2γ (n)).Thus the asymmetry index α(E) can be viewed as a quantity estimating the secondorder term in the Taylor expansion of P(E) around P(Br ). This point of view hasbeen exploited in the proof of Theorem 3.1 where the idea is precisely to view theperimeter of E as an integral functional over the sphere depending on the function uand then to write the second order Taylor expansion of that functional.In this section we will present the proof of the quantitative isoperimetric inequalitygiven in [76] which is mostly based on symmetrization arguments. Though differentand even shorter proofs can be given, based on completely different strategies, the onewe present here has the advantage of not using any sophisticate technical tool, it isgeometrically intuitive and can be adapted to other contexts where symmetry playsa fundamental role. Other proofs of the quantitative isoperimetric inequality will bediscussed in the Sect. 5.Let us now give a short description of how the proof goes. The main idea is toreduce the problem, by means of suitable geometric constructions, to the case of moreand more symmetric sets, namely to axially symmetric sets that are also symmetricwith respect to n orthogonal hyperplanes. For these sets one may either use the alreadymentioned result by Hall [82, Th. 2], or simpler ad hoc arguments.Observe that both the asymmetry and the isoperimetric deficit are scaling invariant.Therefore, throughout all this sectionwewill only dealwith setswith the samemeasureof the unit ball B and with small isometric deficit. In fact, since α(E) is always smallerthan 2 inequality (4.3) becomes trivial when D(E) is large.The first step in the proof is to reduce to sets that are confined in a sufficiently largecube. This step is needed in order to take advantage of the compactness properties ofbounded sets with equibounded perimeter. This first reduction is achieved by using theconcavity property of the isoperimetric function which is the function that associatesto every t > 0 the perimeter nω1/nt (n−1)/n of the ball of measure t , see Lemma 4.2.The second step in the proof is to reduce to the case of n-symmetric sets. We saythat E is n-symmetric if E is symmetric with respect to all coordinate hyperplanes.This reduction is the most delicate part the proof. On the other hand, when restrictedto n-symmetric sets, the proof of (4.3) will be relatively easy. The reason is that whilein general it can be be quite hard to find an optimal ball, when E is n-symmetric theball centered at the center of symmetry of E plays the same role of the optimal ball,even if it is not optimal. Indeed, if E is n-symmetric and |E | = |B|, Lemma 4.4 statesthat if B(x0) is an optimal ball then|EB(x0)| ≤ |EB| ≤ 3|EB(x0)|,a simple, but very useful property.The next step is to pass from an n-symmetric to an axially symmetric set. Thisfurther reduction is achieved by proving, see Proposition 4.9, that if E is n-symmetric123The quantitative isoperimetric inequality. . . 537and E∗ denotes its Steiner symmetral with respect to an axis passing through its centerof symmetry, then we haveα(E) ≤ α(E∗) + C√D(E) (4.4)for some positive constant C depending only on n, provided that n = 2 or n ≥ 3 andthe quantitative isoperimetric inequality holds true in dimension n − 1. To concludethe proof one has only to show that (4.3) holds true for the axially symmetric andn-symmetric set E∗, i.e.α(E∗)2 ≤ C(n)D(E∗).Then, (4.3) for E follows at once by combining this inequality with (4.4) andrecalling that D(E∗) ≤ D(E), see Theorem 2.7.4.1 Reduction to n-symmetric setsAs we said before the first step is to reduce to the case where the set E is contained ina cube of fixed edge length. To this aim, given l > 0 we denote be Ql the open cube(−l, l)n . Thus, we may state the following result, proved in [76, Lemma 5.1].Lemma 4.2 There exist two constants, l, C > 0 depending only on n with the follow-ing property: given a measurable set E ⊂ Rn, with |E | = |B|, it is possible to find anew set F ⊂ Ql, such that |F | = |B| andD(F) ≤ C D(E), α(E) ≤ α(F) + C D(E). (4.5)Proof By rotating E a little if necessary, we may assume thatHn−1({x ∈ ∂∗E : νE (x) = ±en})= 0. (4.6)Thus, Theorem 2.6 applied with k = n − 1 yields that the functionvE (t) := Hn−1({x ′ ∈ Rn−1 : (x ′, t) ∈ E}) for t ∈ R,belongs to W 1,1(R). Hence, vE is continuous. Set E−t = {x ∈ E : x1 < t} for allt ∈ R and recall that if F is a set of finite perimeter, then by Theorem 2.2 we haveHn−1(F (1/2)\∂∗F) = 0. Therefore from (4.6) we have that for H1-a.e. t ∈ RP(E−t ) ≤ P(E; {xn < t}) + vE (t), P(E\E−t ) ≤ P(E; {xn > t}) + vE (t). (4.7)Let us now define a function g : R → R+ asg(t) := |E−t |ωn.123538 N. FuscoFrom Fubini’s theorem it follows that g is a non-decreasing absolutely continuousfunction such that g′(t) = vE (t)/ωn for H1-a.e. t ∈ R. Therefore there exist −∞ ≤a < b ≤ +∞ such that {t : 0 < g(t) < 1} = (a, b). Fix any t ∈ (a, b) and note thatby definition|g(t)− 1n E−t | = |B|.Therefore, by the isoperimetric inequality (1.4) P(g(t)−1/n E−t ) ≥ P(B), henceP(E−t ) = g(t)n−1n P(g(t)−1n E−t)≥ g(t) n−1n P(B).Similarly,P(E\E−t ) ≥ (1 − g(t))n−1n P(B).Therefore, from (4.7) we get thatP(E) + 2vE (t) ≥ P(B)(g(t)n−1n + (1 − g(t)) n−1n)for all t ∈ (a, b). On the other hand, from the definition of isoperimetric deficit wehave P(E) = P(B)(1 + D(E)), and thusvE (t) ≥ 12P(B)(g(t)n−1n + (1 − g(t)) n−1n − 1 − D(E)). (4.8)Let us now define a function ψ : [0, 1] → R+ asψ(t) := t n−1n + (1 − t) n−1n − 1.Note thatψ(0) = ψ(1) = 0 and thatψ is a concave function attaining its maximumat 1/2, ψ(1/2) = 21/n − 1. Therefore,ψ(t) ≥ 2(21/n − 1) t for all 0 ≤ t ≤ 12. (4.9)We may assume, without loss of generality, that 2D(E) < ψ(1/2). Otherwise,since α(E) ≤ 2, the assertion immediately follows by choosing F = B and C =4/ψ(1/2). Then, let a < t1 < t2 < b be two numbers such that g(t1) = 1− g(t2) andψ(g(t1)) = ψ(g(t2)) = 2D(E); therefore,ψ(g(t)) ≥ 2D(E) for all t ∈ (t1, t2) (4.10)and by (4.9)g(t1) = 1 − g(t2) ≤ D(E)21/n − 1 . (4.11)123The quantitative isoperimetric inequality. . . 539Thanks to (4.8) and (4.10), for H1-a.e. t1 ≤ t ≤ t2 we havevE (t) ≥ 12P(B)(ψ(g(t)) − D(E))≥ 14P(B)ψ(g(t)) + 14P(B)(ψ(g(t)) − 2D(E)) ≥ nωn4ψ(g(t)).(4.12)Therefore, recalling that vE (t) = ωng′(t),t2 − t1 ≤ 4n∫ t2t1g′(t)ψ(g(t))dt = 4n∫ g(t2)g(t1)1ψ(s)ds ≤ 4n∫ 101ψ(s)ds = α (4.13)for some constant α = α(n). Let us now setτ1 = max{t ∈ (a, t1] : vE (t) ≤ nωn D(E)2},τ2 = min{t ∈ [t2, b) : vE (t) ≤ nωn D(E)2}.Note that τ1 and τ2 are well defined since vE is continuous and vE (a) = vE (b) = 0and that by (4.10) and (4.12) vE (τ1) = vE (τ2) = (nωn D(E))/2. Moreover,from (4.11), we havet1 − τ1 ≤ 2nωn D(E)∫ t1τ1vE (t) dt = 2nD(E)∫ t1τ1g′(t) dt ≤ 2g(t1)nD(E)≤ 2n(21/n − 1)and a similar estimate holds for τ2 − t2.Let us now set Ẽ = E ∩ {x : τ1 < x1 < τ2}. From the above estimate and (4.13),we have that τ2 − τ1 < β(n). Moreover, (4.11), the definition of τ1, τ2 and (4.7)immediately yield|Ẽ | ≥ |B|(1 − 2 D(E)21/n − 1), P(Ẽ) ≤ P(E) + nωn D(E). (4.14)Let us now assume that D(E) < (21/n − 1)/4 and setσ :=( |B||Ẽ |)1/n, F := σ Ẽ .Clearly, |F | = |B| and F is contained in a strip {τ ′1 < x1 < τ ′2}, with τ ′2 − τ ′1 ≤ β ′,where β ′ is a constant depending only on n. Let us now show that F satisfies (4.5) fora suitable constant C depending only on n.123540 N. FuscoFig. 2 The optimal ball of an-symmetric set is not alwayscentered at the originTo this aim, note that since we are assuming D(E) small, from (4.14) we get that1 ≤ σ ≤ 1 + C0D(E), with C0 = C0(n). Thus, from (4.14), we getP(F) = σ n−1P(Ẽ) ≤ σ n−1(P(E) + P(B)D(E))= σ n−1P(B)(1 + 2D(E)) ≤ P(B)(1 + C(n)D(E)).Hence, the first inequality in (4.5) follows. To prove the second inequality, let usdenote by B(x) an optimal ball for F . From the first inequality in (4.14), we then get|B|α(E) ≤ |EB(x/σ)| ≤ |EẼ | + |ẼB1/σ (x/σ)| + |B1/σ (x/σ)B(x/σ)|= |E\Ẽ | + |B|α(F)σ n+ |B\B1/σ |≤ C(n)D(E) + |B|α(F) + C(n)(σ − 1) ≤ |B|α(F) + C D(E).Thus, the set F satisfies (4.5). Starting from this set, we may repeat the sameconstruction with respect to the xn−1 axis, thus getting a new set, still denoted byF , uniformly bounded with respect to the first two coordinate directions and satisfy-ing (4.5) with a new constant, still depending only on n. Thus, the assertion followsby repeating this argument with respect to all remaining coordinate directions. The next step is to show that we may reduce to a n-symmetric set. To this aim,following the terminology introduced in [76] we say that E ⊂ Rn is an n-symmetricset if it is symmetric with respect to the n coordinate hyperplanes. Note that for an n-symmetric set it is not true in general that the optimal ball for the Fraenkel asymmetryis the one centered at the center of symmetry, see Fig. 2. However, this is true if theset E is convex as shown in the next result that was proved in [11, Lemma 5.9].123The quantitative isoperimetric inequality. . . 541Proposition 4.3 Let E ⊂ Rn be an n-symmetric bounded convex set with |E | = |Br |.Thenminx∈Rn|EBr (x)| = |EBr |. (4.15)Proof Given F, G ⊂ Rn , consider their Steiner symmetral F S , GS with respect to anycoordinate hyperplane. Then, it is easily checked that |F S ∩ GS| ≥ |F ∩ G|. Since Eis n-symmetric and convex, E is stable under the Steiner symmetrization with respectto any coordinate hyperplane. Hence for any x ∈ Rn we have|E ∩ Br (x)| ≤ |E ∩ Br (x1, . . . , xn−1, 0)| ≤ · · · ≤ |E ∩ Br |.from which (4.15) immediately follows. On the other hand, though for n symmetric sets the ball centered at the origin is notalways a minimizer for the Fraenkel asymmetry, yet it is ‘optimal’ up to a constant.Lemma 4.4 Let E be n-symmetric, |E | = |Br |. Thenminx∈Rn|EBr (x)| ≤ |EBr | ≤ 3 minx∈Rn|EBr (x)|Proof Let Br (x0) be an optimal ball for E . Since E is n-symmetric, also Br (−x0) isoptimal. Thus, using twice the triangular inequality and the inequality |Br (x0)Br | ≤|Br (x0)Br (−x0)|, we have|EBr | ≤ |EBr (x0)| + |Br (x0)Br | ≤ |EBr (x0)| + |Br (x0)Br (−x0)|≤ |EBr (x0)| + |Br (x0)E | + |EBr (−x0)| = 3|EBr (x0)|.Remark 4.5 The same argument used to prove the above lemma, shows that if E issymmetric with respect to k orthogonal hyperplanes H1, . . . , Hk, k = 1, . . . , n, andH = ∩ki=1Hi , thenminx∈Rn|EBr (x)| ≤ miny∈H |EBr (y)| ≤ 3 minx∈Rn |EBr (x)|. (4.16)The next theorem is the key point of thewhole proof of the quantitative isoperimetricinequality via symmetrization.Theorem 4.6 There exist δ and C, depending only on n, such that if E ⊂ Ql, |E | =|B|, D(E) ≤ δ, then there exists an n-symmetric set F such that F ⊂ Q2l , |F | = |E |andα(E) ≤ Cα(F), D(F) ≤ 2n D(E). (4.17)The proof of Theorem4.6 is quite tricky. In order to explain the strategywe startwithsome definitions. Given a hyperplane H , we consider the two half spaces H+, H−in which Rn is divided by H and denote by rH the reflection about H . Let E be123542 N. FuscoFig. 3 Construction of E+ and E−Fig. 4 Counterexample to (4.20)a measurable set divided by the hyperplane H in two parts of equal volume. Thenconsider the two halves in which E is divided by H : E ∩ H+ and E ∩ H− and thesets obtained by adding to each half its symmetral with respect to H , i.e.,E+ := (E ∩ H+) ∪ rH (E ∩ H+), E− := (E ∩ H−) ∪ rH (E ∩ H−). (4.18)The construction is illustrated in Fig. 3. By construction, |E+| = |E−| = |E |.Moreover, using Theorem 2.2 it is not too hard to show thatP(E+) + P(E−) ≤ 2P(E), hence D(E±) ≤ 2D(E), (4.19)with the first inequality possibly being strict. Thus, if for some universal constantC(n)one hadα(E) ≤ C(n)α(E+) or α(E) ≤ C(n)α(E−), (4.20)iterating this estimate we would immediately get (4.17). Unfortunately, (4.20) is false,as shown by the example in Fig. 4, where we have α(E) > 0, but α(E+) = α(E−) =0.The following proposition provides the right strategy in order to deal with suchunpleasant situation.Proposition 4.7 There exist δ0, C0 > 0 such that if E is a measurable set with finitemeasure and D(E) ≤ δ0, given any two orthogonal hyperplanes H1, H2 dividing E123The quantitative isoperimetric inequality. . . 543Fig. 5 A different symmetrization may give a bigger asymmetryin four parts of equal measure and the four sets E±1 , E±2 defined as in (4.18) withH replaced by H1 and H2, respectively, we have that at least one of them, call it Ẽ ,satisfies the estimateα(E) ≤ C0α(Ẽ). (4.21)In other words, even if some of the sets E±i have a small asymmetry (or even zeroasymmetry as in Fig. 4), at least one of themmust have a sufficiently large asymmetry,compared to the one of E . To convince oneself that indeed this is the case, observethat if we symmetrize the set of the previous example in the horizontal direction weget a larger asymmetry, see Fig. 5.Before proving Proposition 4.7 let us show the continuity of the asymmetry indexwith respect to the isoperimetric deficit which is a simple consequence of the com-pactness properties of sets of finite perimeter.Lemma 4.8 Let l > 0. For any ε > 0 there exists δ > 0 such that if E ⊂ Ql,|E | = |B|, and D(E) ≤ δ then α(E) ≤ ε.Proof We argue by contradiction. If the assertion is not true, there exist ε > 0 anda sequence of sets E j ⊂ Ql , with |E j | = |B|, D(E j ) → 0 and α(E j ) ≥ ε > 0for all j ∈ N. Since the sets E j are all contained in Ql , thanks to Theorem 2.3 wemay assume that up to a subsequence the E j converge in measure to some set E∞ offinite perimeter. Thus |E∞| = |B|, and by the lower semicontinuity of the perimetersP(E∞) ≤ P(B), so E∞ is a ball. However the convergence in measure of E j to E∞immediately implies that |E jE∞| → 0, against the assumption α(E j ) ≥ ε. Thecontradiction concludes the proof. Following [94, Th. 6.1] we may now give the proof of Proposition 4.7.Proof of Proposition 4.7 Step 1 Without loss of generality we may assume |E | = |B|.Let E+1 , E−1 , E+2 , E−2 , be the four sets obtained by reflecting E around the twoorthogonal hyperplanes H1 , H2 dividing E in parts of equal measure. For i = 1, 2 letB±i four balls of radius one such that|E±i B±i | = minx∈Hi |E±i B(x)|.Observe that for i = 1, 2, from the definition of B±i and the triangular inequalitywe have123544 N. Fuscominx∈Rn|EB(x)| ≤ |EB+i | = |(EB+i ) ∩ H+i | + |(EB+i ) ∩ H−i |≤ |(EB+i ) ∩ H+i | + |(EB−i ) ∩ H−i | + |(B+i B−i ) ∩ H−i |= 12|E+i B+i | +12|E−i B−i | +12|B+i B−i |. (4.22)To conclude the proof it is enough to show that if D(E) is sufficiently small at leastone of the following two inequalities|B+1 B−1 | ≤ 16[|E+1 B+1 | + |E−1 B−1 |]or |B+2 B−2 | ≤ 16[|E+2 B+2 | + |E−2 B−2 |] (4.23)holds. Indeed, if for instance the first inequality is true, from (4.22) applied with i = 1,recalling the definition of B±1 and (4.16), we getminx∈Rn|EB(x)| ≤ 9[|E+1 B+1 | + |E−1 B−1 |]≤ 27[minx∈Rn|E+1 B(x)| + minx∈Rn |E−1 B(x)|],thus proving (4.21) with C0 = 54 and Ẽ equal to E+1 or E−1 .Step 2 To prove (4.23) we argue by contradiction assuming that|E+1 B+1 | + |E−1 B−1 | <|B+1 B−1 |16and |E+2 B+2 | + |E−2 B−2 |<|B+2 B−2 |16(4.24)and introduce the following unions of half ballsS1 = (B+1 ∩ H+1 ) ∪ (B−1 ∩ H−1 ), S2 = (B+2 ∩ H+2 ) ∪ (B−2 ∩ H−2 ).Then, by (4.24) we get that|S1S2| ≤ |S1E | + |ES2|=122∑i=1(|E+i B+i | + |E−i B−i |)<1322∑i=1|B+i B−i |.The contradiction will be achieved if we show thatmax{|B+1 B−1 |, |B+2 B−2 |} < 16|S1S2|, (4.25)provided D(E) is sufficiently small. To this aim, recall that given ε > 0, Lemma 4.8,(4.16) and the second inequality in (4.19) yield that there exists δ0 > 0 such that ifD(E) < δ0, thenmax{α(E), |E±i B±i | : i = 1, 2} < ε.123The quantitative isoperimetric inequality. . . 545Note that this inequality tells us that E is close in the sense of convergence inmeasure to its optimal balls. Similarly, also the balls B±i are close to the sets E±i inthe same sense. Hence the centers of these balls are all close to each other and to thecenter of an optimal ball for E which in turn is close to the ball of radius 1 centered atthe intersection of H1 and H2. Therefore all the four regions H±1 ∩ H±2 contain almostone quarter of the balls B±i . Therefore we may choose ε so small, and δ0 accordingly,that for all possible choices of the σ, τ ∈ {+,−} we have|(Bσ1 Bτ2 ) ∩ (Hσ1 ∩ H τ2 )| >|Bσ1 Bτ2 |8.Thus, recalling the definition of S1 and S2, we have|S1S2| ≥ |(S1S2) ∩ (Hσ1 ∩ H τ2 )| = |(Bσ1 Bτ2 ) ∩ (Hσ1 ∩ H τ2 )| >|Bσ1 Bτ2 |8.From this inequality we have in particular that|B+1 ∩ B−1 | ≤ |B+1 ∩ B+2 | + |B+2 ∩ B−1 | < 16|S1S2|and in a similar way that |B+2 ∩ B−2 | < 16|S1S2|, thus proving the desired contra-diction (4.25). At this point we have all the ingredients for the proof of Theorem 4.6.Proof of Theorem 4.6 Take δ1 = δ02−(n−1), where δ0 is chosen as in Proposition 4.7.By applying this proposition n − 1 times to different pairs of orthogonal directionsand recalling (4.19) we find a set Ẽ , with |Ẽ | = |B| such thatα(E) ≤ Cn−10 α(Ẽ), D(Ẽ) ≤ 2n−1D(E).Moreover, by translating Ẽ and relabeling the coordinate axes, if needed, we mayassume without loss of generality that Ẽ is symmetric about all the coordinate hyper-planes {x1 = 0}, . . . , {xn−1 = 0}. In order to get the last symmetry we take ahyperplane H orthogonal to en and dividing Ẽ into two parts of equal measure andconsider the corresponding sets Ẽ+, Ẽ−. Again, by translating Ẽ in the direction ofen , if necessary, we may assume that H = {xn = 0}. As before we haveD(Ẽ±) ≤ 2D(Ẽ) ≤ 2n D(E).To control the asymmetry of Ẽ± observe that since Ẽ is symmetric with respect tothe first n − 1 coordinate hyperplanes, Ẽ+ and Ẽ− are both n-symmetric so we canuse Lemma 4.4 to get|B|α(Ẽ) ≤ |ẼB| = |(ẼB) ∩ {xn > 0}| + |(ẼB) ∩ {xn < 0}|= 12[|Ẽ+B| + |Ẽ−B|] ≤ 3|B|2[α(Ẽ+) + α(Ẽ−)].123546 N. FuscoThus, at least one of the sets Ẽ± has asymmetry index greater than 13α(Ẽ). There-fore, denoting by F this set, we haveD(F) ≤ 2D(Ẽ) ≤ 2n D(E) and α(E) ≤ Cn−10 α(Ẽ) ≤ 3Cn−10 α(F).Finally, the inclusion F ⊂ Q2l follows immediately from the construction per-formed in the proof of Proposition 4.7 and the one performed here. 4.2 Reduction to axially symmetric setsFrom the results obtained in the previous section it is clear that in order to prove thequantitative isoperimetric inequality (4.3) we may assume without loss of generalitythat there exist δ0 ∈ (0, 1) and l > 0 such that|E | = |B|, E ⊂ Ql , D(E) ≤ δ0, E is n-symmetric. (4.26)In fact, since α(E) ≤ 2 it is clear that if D(E) ≥ δ0 (4.3) follows immediatelywith γ (n) = 4/δ0. Thus, if δ0 is sufficiently small Lemma 4.2 and Theorem 4.6 tellus that we may assume without loss of generality that E is contained in some cubeof fixed size and that it is n-symmetric. Therefore, throughout this section we shallalways assume that E satisfies the above assumptions (4.26).The next step consists in reducing the general case to the case of anaxially symmetricset, i.e., a set E having an axis of symmetry such that every non-empty cross section ofE perpendicular to this axis is an (n − 1)-ball. To this aim we recall that the Schwarzsymmetral of a measurable set E with respect to the xn axis is defined asE∗ = {(x ′, t) ∈ Rn−1 × R : t ∈ R, |x | < rE (t)},where rE (t) is the radius of the (n − 1)-dimensional ball having the same measure ofthe section Et , that is Hn−1(Et ) = ωn−1rE (t)n−1. Note that E∗ coincides with theSteiner symmetral of E with respect to the line {x ′ = 0}. Therefore, from Theorem 2.7we have that P(E∗) ≤ P(E). Moreover the properties stated in Theorems 2.6 and2.7 apply to the functions defined for H1-a.e. t ∈ R as vE (t) := Hn−1(Et ) andpE (t) := Hn−2(∂∗Et ).Proposition 4.9 Let E satisfy assumptions (4.26). If n = 2 or if n ≥ 3 and the quan-titative isoperimetric inequality (4.3) holds true in Rn−1, then there exist a constantC(n) such that|EE∗| ≤ C√D(E) and D(E∗) ≤ D(E). (4.27)Proof The second inequality in (4.27) follows immediately from the fact that |E∗| =|E | and P(E∗) ≤ P(E). In order to prove the first inequality in (4.27) we start byassuming thatHn−1({x = (x ′, t) ∈ ∂∗E : νEx ′(x) = 0}) = 0. (4.28)123The quantitative isoperimetric inequality. . . 547Thus, from Theorem 2.6 we have that vE is an absolutely continuous function inR. We use Theorem 2.6 again and Theorem 2.7 to estimateP(E) − P(B) ≥ P(E) − P(E∗) ≥∫R(√p2E + v′2E −√p2E∗ + v′2E)dt=∫Rp2E − p2E∗√p2E + v′2E +√p2E∗ + v′2Edt≥(∫R√p2E − p2E∗ dt)2 1∫R(√p2E + v′2E +√p2E∗ + v′2E)dt≥(∫R√p2E − p2E∗ dt)2 1P(E) + P(E∗) ,where the inequality before the last one follows from Hölder’s inequality. SinceD(E) ≤ δ0 < 1, we have P(E∗) ≤ P(E) ≤ 2P(B). Therefore we get, observ-ing that pE ≥ pE∗ ,√D(E) ≥ c∫R√p2E − p2E∗ dt = c∫R√pE + pE∗√pE∗√(pE − pE∗)/pE∗ dt≥ √2c∫RpE∗√(pE − pE∗)/pE∗ dt, (4.29)for some constant c depending only on n. Now assume that n ≥ 3 and observe thatsince (E∗)t is a (n − 1)-dimensional ball of radius rE (t) withHn−1 measure equal toHn−1(Et ), the ratiopE (t) − pE∗(t)pE∗(t)is precisely the isoperimetric deficit in Rn−1 of Et . Since by assumption, the quanti-tative isoperimetric inequality (4.3) holds true in Rn−1, we have√γ (n − 1)√pE (t) − pE∗(t)pE∗(t)≥ αn−1(Et ),where αn−1(Et ) is the (n − 1)-dimensional Fraenkel asymmetry of Et . But Et is an(n −1)-symmetric set inRn−1 and (E∗)t is the ball centered at the center of symmetryof Et . Therefore from Lemma 4.4 we get√γ (n − 1)√pE (t) − pE∗(t)pE∗(t)≥ αn−1(Et ) ≥ 13Hn−1(Et(E∗)t )Hn−1((E∗)t ) .Inserting this inequality in (4.29) we then get123548 N. Fusco√D(E) ≥ c′∫RHn−1(EtE∗t )rE (t)dt ≥ c′l∫ l−lHn−1(EtE∗t ) dt =c′l|EtE∗t |,where the inequality before the last one follows from the inclusion E ⊂ Ql and thelast equality is just Fubini’s theorem. This proves (4.27). Hence the assertion followswhen n ≥ 3.If n = 2, since E is 2-symmetric, either Et is a symmetric interval (hence Et = E∗t )or Et is the union of at least two intervals with no endpoints in common, hencepE (t) ≥ 4, while pE∗(t) = 2. Moreover, since E ⊂ Ql , then H1(EtE∗t ) ≤ 2l forall t ∈ R. Therefore, from (4.29) we get√D(E) ≥ 2c∫{t : Et =E∗t }√pE − pE∗ dt ≥ 2c∫{t : Et =E∗t }√2 dt≥√2cl∫{t : Et =E∗t }H1(EtE∗t )dt =√2cl|EE∗|,thus concluding the proof also in this case.Finally, if (4.28) does not hold, we approximate E in measure by a sequence ofsets Eh obtained by rotating a little E so that (4.28) holds true for each Eh . Then,the conclusion follows by observing that also E∗h converge to E∗ in measure and that(4.27) holds for all the sets Eh . For an axially symmetric set the proof of the quantitative isoperimetric inequalitywas given by a direct argument in [82, Th. 2]. A different proof can be also foundin [76, Th. 4.2]. However both proofs are technically a bit complicate. Here we shallgive a third proof due to Maggi [94, Sec. 7.2]. This last proof has two interestingfeatures. First, it is based on a further and very natural reduction argument, whichwas not exploited in the original proof of (4.3) given in [76], the reduction to aconnected set. This further reduction, is again a consequence of the strict concavity ofthe isoperimetric function. Another interesting feature is the use of some ideas fromthe mass transportation argument of Gromov’s proof of the isoperimetric inequality,to which we will come back in the next section. So, let us start with the reductionlemma proved in [94, Th. 4.4].Lemma 4.10 There exists two positive constants, δ and C, depending only on thedimension n, such that if E is a bounded open set of class C∞ with D(E) ≤ δ thereexists a connected component F of E such that |F | > |E |/2,α(E) ≤ α(F) + C D(E), D(F) ≤ C D(E). (4.30)Proof Without loss of generality we may assume |E | = |B|. Since E is a smoothbounded open set, if E is not connected we may decompose it as the union of finitelymany connected components Eh , h ∈ I ⊂ N. Moreover, P(E) = ∑h∈I P(Eh). Fromthe isoperimetric inequality (1.4) we have123The quantitative isoperimetric inequality. . . 549D(E) = 1n|B|(∑h∈IP(Eh) − P(B))≥ 1n|B|(n|B|1/n∑h∈I|Eh | n−1n − n|B|)=∑h∈Ian−1nh − 1,where ah := |Eh |/|B|. Observe that∑h∈I ah = 1 and assume that the ah are arrangedin decreasing order. If a1 > 1/2, from inequality (4.9) we haveD(E) ≥ an−1n1 + (1 − a1)n−1n − 1 ≥ 2(21/n − 1)(1 − a1), (4.31)hence |E\E1| ≤ C0D(E), with C0 = ωn/2(21/n − 1). Thus, we set F = E1 and wehave, denoting by r the radius of the ball with the same measure of F ,P(F) − P(Br )≤P(E) − P(B) + P(B) − P(Br )≤P(E) − P(B) + nωn(1 − rn−1).From this inequality, since ωn(1 − rn) = |E\E1| ≤ C0D(E) and ωnrn = |F | >|B|/2, we easily getD(F) ≤ C D(E)for some possibly bigger constant, still depending only on n. Similarly, denoting byBr (x0) an optimal ball for α(F) we getωnα(E) ≤ |EB(x0)| ≤ |EF | + |FBr (x0)| + |Br (x0)B(x0)|= |FBr (x0)| + 2|E\F | ≤ |FBr (x0)| + 2C0D(E).thus completing the proof of (4.30).To conclude the proof of the lemma it is enough to show that if δ is sufficientlysmall, in dependence on n, then a1 > 1/2. In fact, suppose that a1 ≤ 1/2 and denoteby N ≥ 2 the smallest integer such that∑hNah) n−1n− 1 ≥ 2(21/n − 1)∑h>NahAdding up these two inequalities we then conclude thatD(E) ≥ (21/n − 1)∑h =Nah = (21/n − 1)(1 − aN ) ≥ (21/n − 1)(1 − a1)>21/n − 12,123550 N. Fuscowhich is impossible if we choose δ < 2/(21/n − 1). We may now proceed to the proof of the quantitative isoperimetric inequality. Aswe already said before, for the final argument, i.e. the proof of (4.3) for an axially andn-symmetric set, we follow the argument of Maggi [94].Proof of Theorem 4.1 Step 1 As we observed at the beginning of this section, it isenough to prove the quantitative inequality (4.3) only when E satisfies the assump-tions (4.26) for a sufficiently small δ0, depending only on the dimension, that will bespecified during the proof. Observe that wemay also assumewithout loss of generalitythat E is a bounded open set of class C∞. Otherwise, by a standard approximationprocedure, see for instance the proof of [8, Th. 3.42], we may find a sequence ofn-symmetric smooth open sets Eh converging in measure to E , |Eh | = |E | for allh, P(Eh) → P(E), satisfying the assumptions (4.26) with l possibly replaced by2l. Then the quantitative isoperimetric inequality for E will follow from the sameinequality for Eh .Finally, observe that we may also assume that E is connected. In fact if E were notconnected, taking δ0 sufficiently small we could apply Lemma 4.10 and consider theopen connected component F of E such that |F | > |E |/2, satisfying (4.30). Observethat also this connected component is n-symmetric. Indeed, consider for instance thehyperplane H = {x1 = 0} and assume by contradiction that there exists x ∈ F suchthat its reflection rH (x) /∈ F . Then F ∩ H = ∅, since otherwise x and rH (x) wouldbelong to the same connected component of E , hence both would belong to F . SinceE is n-symmetric, also rH (F) is contained in E . However, this is not possible since Fis contained either in H+ or in H− and thus |F ∪ rH (F)| = 2|F | > |E |. This provesthat rH (F) = F . By repeating this argument for all the coordinate hyperplanes, weconclude that F is n-symmetric. Thus, we may replace E by λF , where λ > 0 issuch that λn|F | = |B|. The set λF is now a bounded, smooth, open and connected setsatisfying all the assumptions in (4.26), with l possibly replaced by 2l.Consider the strips S = {x : |xn| <√2/2} and S′ = {x : |x1| <√2/2}. SinceB ⊂ S ∪ S′ one of the two strips must contain at least one half of the measure ofB\E . To fix the ideas, let us assume that S does it, that is |(B\E) ∩ S| ≥ |B\E |/2.Therefore,|EB| ≤ 4|(B\E) ∩ S| (4.32)Denote by E∗ is the Schwartz symmetral of E with respect to the xn axis. We claimthat if D(E∗) is sufficiently small, then|(BE∗) ∩ S| ≤ C(n)√D(E∗). (4.33)Note that the conclusion of the proof will follow immediately from (4.33). Indeed,assume n = 2 or n ≥ 3 and that (4.3) holds true inRn−1. Then, if D(E) is sufficientlysmall, from Lemma 4.4, and (4.27), (4.32) and (4.33) we get|B|α(E) ≤ 3|EB| ≤ 12|(B\E) ∩ S|≤ 12(|(BE∗) ∩ S| + |E∗E |) ≤ C(n)√D(E∗) ≤ C(n)√D(E).123The quantitative isoperimetric inequality. . . 551The rest of the proof will be devoted to prove (4.33).Step 2 Set v(t) := Hn−1(E∗ ∩ {xn = t}). From our assumptions and Theorem 2.6we have that v ∈ BV (R), v is symmetric with respect to 0 and {v > 0} = (−a, a)for some a > 0. Observe that we may also assume without loss of generality thatv ∈ W 1,1(R). Otherwise, we split v as v = va + vs where va and vs are nonnegative,symmetric with respect to 0, va ∈ W 1,1(R) and vs is purely singular, see [8, Sect. 3.2],and we approximate vs in L1(R) with a sequence wh ∈ C1c (R), with wh nonnegativeand symmetric with respect to 0, wh > 0 in an interval (−a′, a′) and such that∫ a−a |w′h | → |Dvs |(R) as h → ∞. Then denoting by E∗h the axially symmetric setssuch that vE∗h = va + wh , from (2.7) we get immediately that the E∗h converge to E∗in measure and that P(E∗h ) → P(E∗). Thus we prove (4.33) for E∗h and then pass tothe limit as h → ∞.So, let us assume that v ∈ W 1,1(R) and in particular that it is continuous. We firstshow that if D(E∗) is sufficiently small, then the strip S contains a fairly large amountof the mass of E∗. Precisely, we are going to show that there exist c0 and δ0, dependingonly on the dimension n such that if D(E∗) ≤ δ0, thenv(t) ≥ c0 for all t ∈(−√22,√22). (4.34)In order to prove this inequality we compare the distribution function v of E∗ withthe distribution function w of B, defined as w(t) := Hn−1(B ∩ {xn = t}).We set c0(n) = w(4/5)/3 and then argue by contradiction. If (4.34) is not true,there exists t0 ∈ (−√2/2,√2/2) ⊂ (−4/5, 4/5) such that v(t0) < c0. Then, thereexist t1 < t0 < t2 such that (t1, t2) is the largest interval containing t0, contained in(−4/5, 4/5) and such thatv(t) ≤ 2c0 for all t ∈ (t1, t2).By the definition of c0, recalling Lemma 4.4, we immediately havec0(t2 − t1) ≤∫R|w − v| dx = |E∗B| ≤ 3|B|α(E∗). (4.35)Then Lemma 4.8 implies that t2−t1 is small if D(E∗) is small. Therefore the closedinterval [t1, t2] is contained in (−4/5, 4/5) and thus in particular v(t1) = v(t2) = 2c0.Now we compare E∗ with the axially symmetric set whose distribution function vFcoincides with v outside the interval (t1, t2) and such that vF ≡ 2c0 in (t1, t2). Then,setting := Rn−1 × (t1, t2),P(F) = P(E∗) + P(F;) − P(E∗;).Clearly P(F;) = c1(n)(t2 − t1) with c1(n) depending only on c0(n) and from(2.7) we have123552 N. FuscoP(E∗;) ≥∫ t2t1|v′(t)| dt ≥ |v(t2) − v(t0)| + |v(t0) − v(t1)| = 2c0.Thus, from the above inequalities, observing that |F | ≥ |E∗| = |B| and using theisoperimetric inequality, we getP(E∗) ≥ P(F) + 2c0 − c1(n)(t2 − t1) ≥ P(B) + 2c0 − c1(n)(t2 − t1).Since by (4.35) t2 − t1 → 0 as D(E∗) → 0+, from this inequality we get acontradiction if D(E∗) is sufficiently small.Step 3 Recall that {v > 0} = (−a, a) for some a > 0. Observe that (4.34) impliesthat a >√2/2. Then, we define a function τ : (−a, a) → (−1, 1) which associatesto every sublevel E∗ ∩ {xn < t} the sublevel B ∩ {xn < τ(t)} with the same measure.Precisely, for all t ∈ (−a, a) we define τ(t) as the unique number in (−1, 1) suchthat∫ t−∞v(s) ds =∫ τ(t)−∞w(s) ds.Note that τ is a strictly increasing C1(−a, a) function and for all t ∈ (−a, a)τ ′(t) = v(t)w(τ(t)). (4.36)Since v ∈ W 1,1(R) the above equality implies that τ ∈ W 2,1loc (−a, a). Moreover,since w is locally Lipschitz in (−1, 1), setting I := (−√2/2,√2/2), we get from(4.36)|(E∗B) ∩ S| =∫I|w(t) − v(t)| dt =∫I|w(t) − w(τ(t))τ ′(t)| dt≤∫I[|w(t) − w(τ(t))| + |w(τ(t)) − w(τ(t))τ ′(t)|] dt≤ C(n)∫I[|t − τ(t)| + |1 − τ ′(t)|] dt ≤ C(n)∫I|1 − τ ′(t)| dt,for some constant C(n) depending only on the Lipschitz constant of w in I . Note thatin the last inequality we have used the fact that since τ(0) = 0 by the n-symmetry ofE , then |t − τ(t)| ≤ ∫I |1 − τ ′|. Thus, to achieve the proof of (4.33) we only need toshow that ∫I|τ ′(t) − 1| dt ≤ C(n)√D(E∗). (4.37)To this aim we introduce the map T : (−a, a) × Rn−1 → Rn defined asT (x) =n−1∑i=1(w(τ(xn))v(xn)) 1n−1xi ei + τ(xn)en .123The quantitative isoperimetric inequality. . . 553Note that T maps every level set E∗ ∩ {xn = t} into the level set B ∩ {xn = τ(t)}, andit is a one-to-one continuous map between the open sets E∗ and B. Moreover for all0 < b < a, T ∈ W 1,1(E∗ ∩ (Rn−1 × (−b, b)); B). It is also easily checked that fora.e. x ∈ E∗ ∩ (Rn−1 × (−a, a))div T (x) = n − 1τ ′(xn)1/(n−1)+ τ ′(xn) ≥ n,wherewe used that the function s → s+(n−1)s−1/(n−1) has a strictminimumat s = 1for s > 0. Therefore, using the divergence theorem in the sets E∗ ∩ (Rn−1 × (−b, b))and observing that limb→a− v(±b) = 0, we haveP(E∗) = Hn−1(∂∗E∗) ≥∫∂ E∗T · νE∗ dHn−1= limb→a−∫∂ E∗∩(Rn−1×(−b,b))T · νE∗ dHn−1= limb→a−[∫∂(E∗∩(Rn−1×(−b,b))T · νE∗ dHn−1 − τ(b)v(b) + τ(−b)v(−b)]= limb→a−∫E∗∩(Rn−1×(−b,b))div T dx =∫E∗div T dx .Therefore we have by Fubini’s theoremP(E∗) − P(B) ≥∫E∗(div T − n) dx =∫E∗(n − 1τ ′(xn)1/(n−1)+ τ ′(xn) − n)dx=∫ a−av(t)(n − 1τ ′(t)1/(n−1)+ τ ′(t) − n)dx=∫ a−av(t)τ ′(t)1/(n−1)(n − 1 + τ ′(t)n/(n−1) − nτ ′(t)1/(n−1))dx≥ n2∫ a−av(t)τ ′(t)1/(n−1)(τ ′(t)1/(n−1) − 1)2dt,where we used the elementary inequality n − 1 + tn − nt ≥ n(t − 1)2/2 for allt > 0 and n ≥ 2. Then from the previous chain of inequalities we get, settingσ(t) := τ ′(t)1/(n−1),∫ a−av|σ − 1| dt ≤√∫ a−avσ(σ − 1)2 dt√∫ a−avσ dt ≤ C(n)√D(E∗), (4.38)where the last inequality follows by observing that, since∫ a−a τ′dt = 2, ∫ a−a vdt =|E∗| = |B| and v ≤ (2l)n−1 by assumption (4.26), if n ≥ 3 we have123554 N. Fusco∫ a−avσ dt ≤(∫ a−aτ ′dt) 1n−1 (∫ a−av(n−1)/(n−2)dt) n−2n−1 ≤ C(n).A similar estimate holds also when n = 2. To conclude the proof we observe thatthere exists a constant λ(n) ∈ (0, 1) such thatsupt∈I|τ(t)| ≤ λ(n). (4.39)To this aim, note that|E∗ ∩ {xn >√2/2}| ≥ |B ∩ {xn >√2/2}| − 3|B|α(E∗) ≥ c(n) − 3|B|α(E∗) ≥ c2(n),for some positive constant c2(n), provided that D(E∗), hence α(E∗) is sufficientlysmall. On the other hand, by the very definition of τ , we have|E∗ ∩ {xn >√2/2}| = |B ∩ {xn > τ(√2/2)}| ≤ c(n)(1 − τ(√2/2)).Combining the twoprevious inequalitiesweget that 1−τ(√2/2) ≥ c3(n) for a suitablepositive constant depending only on n, hence τ(√2/2) ≤ 1 − c3(n). Since a similarinequality can be proved for τ(−√2/2), (4.39) follows. Observe that (4.39), togetherwith (4.36) implies that supt∈I |τ ′(t)| ≤ C(n) for a suitable constant depending onlyon n. Then, recalling (4.34) and (4.38) we may conclude that∫I|τ ′ − 1| dt ≤ C∫I∣∣(τ ′)1/(n−1) − 1∣∣ dt = C∫I|σ − 1| dt≤ C∫Iv|σ − 1| dt ≤ C√D(E∗),thus proving (4.37). 5 Other proofs and extensions5.1 Isoperimetric inequality via mass transportIn the previous section we have presented the proof of the the quantitative isoperimet-ric inequality given in [76]. As we have seen, that proof was based on symmetrizationarguments. The same approach has been used in several other papers to obtain quan-titative versions of the Sobolev inequality, of the isoperimetric inequality in Gaussspace and of other relevant geometric and functional inequalities, see for instance[43,44,75,78]. On the other hand there are situations where one considers inequalitieswhich are realized by non symmetric sets or functions.This is the case of the anisotropic isoperimetric inequality. In order to state it let usfix some notation. Let ϕ : Rn → [0,∞) be a positively 1-homogeneous, continuous123The quantitative isoperimetric inequality. . . 555function such that ϕ(x) > 0 for all x = 0. To the function ϕ we may associate theanisotropic perimeter, defined for any set E of finite perimeter asPϕ(E) :=∫∂∗ Eϕ(νE (x))dHn−1.It is well known that the isoperimetric sets for Pϕ are the homothetic and translatedof the so called Wulff shape set associated to ϕ which is given byWϕ := {x ∈ Rn : 〈x, ν〉 − ϕ(ν) < 0 for all ν ∈ Sn−1}.Then, the anisotropic isoperimetric inequality states thatPϕ(E) ≥ Pϕ(Wϕ)for all sets of finite perimeter such that |E | = |Wϕ |, with equality holding if and onlyif E is a translated of the Wulff shape set Wϕ (see [69,70] and also [50] for a proof intwo dimensions). Note that as in the case of the standard euclidean perimeter we havePϕ(Wϕ) = n|Wϕ |. However, differently from the usual perimeter, Pϕ is not invariantunder the action of either O(n) or SO(n). Moreover, in general Pϕ(E) = Pϕ(Rn\E)unless ϕ, hence Wϕ , is symmetric with respect to the origin.The quantitative version of the anisotropic isoperimetric inequality is a remarkableresult proved by Figalli et al. in [67]. It states that there exists a constant C , dependingonly on n, such that for any set of finite perimeter E such that |E | = rn|Wϕ |αϕ(E)2 ≤ C Dϕ(E), (5.1)whereαϕ(E) := minx∈Rn{ |E(x + r Wϕ)|E |}, Dϕ(E) := Pϕ(E) − Pϕ(r Wϕ)Pϕ(r Wϕ)denote the anisotropic asymmetry index and the anisotropic isoperimetric deficit,respectively.Since the Wulff shape Wϕ can be any bounded open convex set, it is clear that nosymmetrization argument can be used to prove the anisotropic isoperimetric inequalityor its quantitative counterpart (5.1). And in fact the strategy used in [67] is completelydifferent from the one we have seen in the last section, since it relies on an optimalmass transportation argument and on the proof of a very general trace inequality.To simplify even further the presentationof themain ideas used in the proof ofFigalliet al. we shall only consider the case of the standard perimeter where inequality (5.1)reduces to the more familiar quantitative isoperimetric inequality (4.3). The startingpoint is a variant of Gromov’s proof of the classical isoperimetric inequality where theKnothe map originally used in [100, App. 1] is replaced by the Brenier map. Its mainproperties are stated in the next theorem, which follows from the results in [26,93].123556 N. FuscoTheorem 5.1 Let μ and ν be two probability measures on Rn with μ absolutelycontinuous with respect to Lebesgue measure. Then there exists a convex functionϕ : Rn → R, such that the map T := ∇ϕ transports μ onto ν, that is for every Borelfunction h : Rn → [0,∞]∫Rnh(y) dν(y) =∫Rnh(T (x)) dμ(x). (5.2)Observe that (5.2) can be restated by saying that the map T pushes forward themeasure μ onto ν (for the definition of the push forward of a measure see for instance[8, Sect. 1.5]). In the particular case that E is a set of finite perimeter with |E | = |B|,setting μ = 1ωnχ E dx and ν = 1ωn χ B dx , from the above theorem we may conclude(see also [97, Sect. 2.1]), that there exists a convex function ϕ : Rn → R such that,setting T := ∇ϕ, then T (x) ∈ B for a.e. x ∈ Rn and for every Borel functionh : B → [0,∞] ∫Bh(y) dy =∫Eh(T (x)) dx . (5.3)Note that there is a regularity issue here, since ϕ is convex, hence T is defined only a.e.and it is just a BV map. However, in order to avoid technical complications, see thediscussion in [67, Sect. 2.2], let us assume that T is a Lipschitz map. Under this extraassumption, by a change of variable, one easily gets from (5.3) that det∇T (x) = 1for a.e. x ∈ E . Then we can give theGromov’s proof of the isoperimetric inequality For every x ∈ E denote by λi (x),i = 1, . . . , n, the eingenvalues of the symmetric matrix ∇T (x). Using the geometric–arithmetic mean inequality and the divergence theorem, we haveP(B) = nωn = n∫Bdy = n∫E(det∇T )1/ndx = n∫E(λ1 · · · λn)1/ndx≤∫E(λ1 + · · · + λn)dx =∫Ediv T dx =∫∂ ET · νE dHn−1 ≤ P(E).Observe that if P(E) = P(B) then λ1(x) = λ2(x) = · · · = λn(x) = 1 for a.e. x ∈ E ,since det∇T (x) = 1. Hence, up to a translation, T is the identity map and E is a ball.Let us try to exploit this argument to prove the quantitative isoperimetric inequality.Since, by definition,P(E) = P(B) + |B|D(E),from the inequalities above we immediately get that∫E[(λ1 + · · · λn)/n − (λ1 · · · λn)1/n]dx ≤ ωn D(E), (5.4)∫∂∗ E(1 − T · νE )dHn−1 ≤ nωn D(E). (5.5)123The quantitative isoperimetric inequality. . . 557It is not too difficult to show (see [67, Corollary 2.4]) that (5.4) implies that thereexists a constant C depending only on n such that, if D(E) ≤ 1, then∫E|∇T − I | dx ≤ C(n)√D(E), (5.6)where I stands for the identity matrix. Let us assume for a moment that the set E is sogood to satisfy a Poincaré inequality for some universal constant depending only onn and let us see what information we may deduce from (5.6). Indeed, it can be provedthat if D(E) is smaller than some positive δ depending only on n one may removefrom E a small critical set, whose measure is controlled by D(E), such that outsidethis critical set a Poincaré inequality holds true with a constant depending only on n.Therefore, by a reduction argument, we may assume thatinfc∈RN∫E| f − c| dx ≤ C(n)∫E|∇ f | dx for all f ∈ C1c (Rn).Therefore, by translating E we have from (5.6) that∫E|T (x) − x | dx ≤ C(n)√D(E).Therefore, given ε > 0, from the previous inequality we getωnα(E) ≤ |EB| = 2|E\B| ≤ 2(|E\B1+ε| + |B1+ε\B|)≤ C(ε−1∫E|T x − x | + ε)≤ C(n)(ε + ε−1√D(E)) .Minimizing the right hand side of this inequality with respect to ε we then getα(E)4 ≤ C(n)D(E),that is the quantitative isoperimetric inequality with the not optimal exponent 4. Notehowever that this argument can never lead to a proof of the quantitative isoperimetricinequality: firstly, because even if E is a connected open set the constant of the Poincaréinequality may blow up in presence of small cusps; secondly because in the aboveargument we are not taking into account the information contained in the inequality(5.5) derived from Gromov’s proof.Indeed the strategy followed in [67] is more subtle. Namely, one can show that ifE has small deficit, then (see [67, Th. 3.4 and Lemma 3.1]), up to removing a smallcritical set from E whose measure is controlled by D(E), one may assume that Esatisfies the following trace inequalityinfc∈Rn∫∂∗ E| f − c| dHn−1 ≤ τ(n)∫E|∇ f | dx for all f ∈ C1c (Rn),123558 N. FuscoFig. 6 |EB1| ≤ C∫∂∗ E∣∣1−|x |∣∣for some constant τ depending only on n. Then, by applying the previous inequalityto the map T − I d, up to translating E conveniently, we have, recalling (5.6),∫∂∗ E|T (x) − x | dHn−1 ≤ τ(n)∫E|∇T − I | dx ≤ C(n)√D(E).Combining this inequality with (5.5), we obtain, assuming D(E) ≤ 1,∫∂∗ E|1 − |x || dHn−1 ≤∫∂∗ E[∣∣1 − |T (x)|∣∣ + ∣∣|T (x)| − |x |∣∣] dHn−1≤∫∂∗ E[(1 − T (x) · νE (x))+ |T (x) − x |]dHn−1≤ C[D(E) + √D(E)] ≤ C(n)√D(E).The proof of the quantitative isoperimetric inequality (4.3) then immediately followsfrom this estimate since, see [67, Lemma 3.5], there exists a constant C(n) such that,see Fig. 6,|EB| ≤ C(n)∫∂ E∣∣1 − |x |∣∣ dHn−1.Beside providing an alternative proof of the quantitative isoperimetric inequality in thewider framework of anisotropic perimeter, the paper by Figalli et al. contains severalinteresting results. In particular, Theorem 3.4 which states that given any set of finiteperimeter E with small deficit one may always extract from E a maximal set forwhich a trace inequality holds with a universal constant. This is a new and deep resultthat may have several applications. Moreover, the mass transportation approach usedin [67] has been also successfully used to obtain the quantitative versions of otherimportant inequalities, see [43,66,68].123The quantitative isoperimetric inequality. . . 5595.2 Quantitative isoperimetric inequality via regularityAnother very interesting proof of the quantitative isoperimetric inequality has beenrecently given by Cicalese and Leonardi in [45]. The starting point is the quantitativeinequality (3.2) proved by Fuglede for nearly spherical sets and the observation thatall known examples suggest that the quantitative inequality becomes really criticalonly when the set E is a small perturbation of a ball. Therefore their idea is to reducethe general case to the case of nearly spherical sets via a contradiction argument.Precisely, they start by assuming that there exists a sequence of sets E j , convergingin measure to the unit ball, for which the quantitative inequality does not hold. Thenthey replace it with a different sequence of sets Fj , still not satisfying the quantitativeinequality, but converging to B in C1, thus contradicting Fuglede’s Theorem 3.1 fornearly spherical sets. The sets Fj are constructed as the solutions of certain minimumproblems and their convergence in C1 to the unit ball is a consequence of the apriori estimates for perimeter almost minimizers established in the theory of minimalsurfaces.Though the approach of Cicalese and Leonardi to the quantitative isoperimetricinequality is based on the results of a difficult and deep theory, it has the advantageof providing a short proof that has been successfully applied to several other inequal-ities, see [1,18,19,24,47,48,74]. The proof we are going to present here is a furthersimplification of the original proof by Cicalese and Leonardi which has been devel-oped in a more general context by Acerbi et al. in [1]. To this aim, let us quicklyrecall the definition and the regularity properties of the perimeter almost minimiz-ers.Let ω, r0 be positive numbers. A set F of finite perimeter is an (ω, r0)-almostminimizer if, for all balls Br (x) with r < r0 and all measurable sets G such thatFG ⊂⊂ Br (x), we haveP(F) ≤ P(G) + ωrn, (5.7)see (Fig. 7). Thus, an almost minimizer locally minimizes the perimeter up to a higherorder error term. The main properties of almost minimizers are contained in the fol-lowing statement which is essentially due to Tamanini, see [114, Sect. 1.9 and 1.10]and also [95, Th. 26.5 and 26.6].Fig. 7 A perimeter almostminimizer F123560 N. FuscoTheorem 5.2 If E is an (ω, r0)-almost minimizer, then ∂∗E is a C1,1/2 manifold,∂ E\∂∗E is relatively closed in ∂ E and Hs(∂ E\∂∗E) = 0 for all s > n−8. Moreover,if E j is a sequence of equibounded (ω, r0)-almost minimizers converging in measureto a C2 open set E, then for j large each E j is of class C1,1/2 and the sequence E jconverges to E in C1,α for all 0 < α < 1/2.As we said above the starting point of the proof is the Fuglede estimate for nearlyspherical sets. Recall that Theorem 3.1 states that there exist two positive constantsε(n), c0(n) such that if E is an open set with |E | = |B|, the barycenter of E is atthe origin and ∂ E = {x + u(x) : x ∈ ∂ B} for a Lipschitz function u such that‖u‖W 1,∞(Sn−1) < ε, then the following estimate holdsP(E) − P(B) ≥ c0(n)|EB|2. (5.8)We also need the following simple lemma.Lemma 5.3 If > n, the unique solution up to translations of the problemmin{P(F) + ∣∣|F | − |B|∣∣ : F ⊂ Rn} (5.9)is the unit ball.Proof By the isoperimetric inequality it follows that in order tominimize the functionalin (5.9), we may restrict to the balls Br . Thus the above problem is equivalent tominimizing in [0,∞) the function r → nrn−1 + |rn − 1|, which has a uniqueminimum for r = 1, if > n. We are now ready to give the proof of the quantitative isoperimetric inequality (4.3)via regularity. Before that we need also to introduce the non-rescaled asymmetry indexby setting for any measurable set E of finite measureA(E) := minx∈Rn{|EB(x)|}.Proof of the quantitative isoperimetric inequality via regularity Step 1 Thanks toLemma 4.2 we may assume, without loss of generality, that |E | = |B| and thatE ⊂ BR0 , for some R0 depending only on n. As we have already observed in theprevious section, in order to prove (4.3) it is enough to show thatClaim There exists δ0 > 0 such that if E ⊂ BR0 , |E | = |B| and P(E)− P(B) ≤ δ0,then c02A(E)2 ≤ P(E) − P(B), (5.10)where c0 is the constant in (5.8).To this aim we argue by contradiction assuming that there exist a sequence E j ⊂BR0 , |E j | = |B|, with P(E j ) → P(B) andP(E j ) < P(B) + c02A(E j )2. (5.11)123The quantitative isoperimetric inequality. . . 561Since P(E j ) → P(B), by the compactness Theorem 2.3 we may assume that up toa subsequence the sets E j converge in measure to some set E . Then, by the lowersemicontinuity of the perimeter we get that P(E) = P(B). Thus E is a ball of radius1 and we may conclude that A(E j ) → 0 as j → ∞. Now, to achieve the proof of(5.10), we would like to replace the E j with a sequence of sets converging to B in C1and contradicting inequality (5.8).To build this new sequence, for every j we consider a minimizer Fj of the problem:min{P(F) + |A(F) − A(E j )| + ∣∣|F | − |B|∣∣ : F ⊂ BR0}, (5.12)where > n is a fixed constant. Using again the compactness Theorem 2.3 we mayassume that the sets Fj converge in measure to a set F . Moreover, the lower semicon-tinuity of the perimeter immediately yields that F is a minimizer of the problem:min{P(E) + A(E) + ∣∣|E | − |B|∣∣ : E ⊂ BR0}.Therefore by Lemma 5.3 we may conclude that the sequence Fj converge in measureto a ball B(x0). Let us now show that this convergence holds indeed in C1. To this aim,by Theorem 5.2 it is enough to prove that each Fj is an (ω, R0)-almost minimizerfor some ω > 0. To prove this take a ball Br (x) with r < R0 and a set G such thatFjG ⊂⊂ Br (x). Two cases may occur.Case 1 G ⊂ B̄R0 . Then, by the minimality of Fj we getP(Fj ) ≤ P(G) + |A(G) − A(E j )| − |A(Fj ) − A(E j )|+ [∣∣|G| − |B|∣∣ − ∣∣|Fj | − |B|∣∣]≤ P(G) + |A(G) − A(Fj )| + ∣∣|G| − |Fj |∣∣≤ P(G) + ( + 1)|GFj | ≤ P(G) + ( + 1)ωnrn .Case 2 |G\B̄R0 | > 0. In this case we split G as followsP(Fj ) − P(G) = [P(Fj ) − P(G ∩ BR0)] + [P(G ∩ BR0) − P(G)].Since G ∩ BR0 ⊂ BR0 , as before we haveP(Fj ) − P(G ∩ BR0) ≤ ( + 1)ωnrn,whileP(G ∩ BR0) − P(G) = P(BR0) − P(G ∪ BR0) ≤ 0by the isoperimetric inequality. Therefore we may conclude that the sets Fj are all(( + 1)ωn, R0)-almost minimizers and that they converge to a ball B(x0) in C1,αfor all α < 1/2.123562 N. FuscoStep 2 By the minimality of the Fj , recalling (5.11) and using Lemma 5.3, we getP(Fj ) + ∣∣|Fj | − |B|∣∣ + |A(Fj ) − A(E j )| ≤ P(E j )≤ P(B) + c02A(E j )2 ≤ P(Fj ) + ∣∣|Fj | − |B|∣∣ + c02A(E j )2. (5.13)Therefore, we have that |A(Fj ) − A(E j )| ≤ c02 A(E j )2 and since A(E j ) → 0 we getthat A(Fj )/A(E j ) → 1 as j → ∞.To conclude the proof we need only to adjust the volumes of the sets Fj . For thisreason we set F̃j = λ j Fj , where λ j is chosen so that |F̃j | = |B|. Note that λ j → 1since the Fj are converging in C1 to B(x0). Observe also that, since P(Fj ) → P(B)and > n, for j large we have P(Fj ) < |Fj |. Therefore for j large we have|P(F̃j ) − P(Fj )| = P(Fj )|λn−1j − 1| ≤ P(Fj )|λnj − 1|≤ |λnj − 1||Fj | = ∣∣|F̃j | − |Fj |∣∣.From this estimate, recalling (5.13) we get thatP(F̃j ) ≤ P(Fj ) + ∣∣|F̃j | − |Fj |∣∣ = P(Fj ) + ∣∣|Fj | − |B|∣∣ ≤ P(B) + c02A(E j )2.(5.14)However, since A(Fj )/A(E j ) → 1 as j → ∞ we have A(E j )2 < 2A(F̃j )2 for jlarge. Therefore, from (5.14) we obtainP(F̃j ) − P(B) < c0A(F̃j )2,which is a contradiction to (5.8) since, up to translations, the sets F̃j have all barycenterat the origin and are converging in C1 to the unit ball. This contradiction proves theClaim, thus concluding the proof of the quantitative inequality. 5.3 An improved version of the quantitative isoperimetric inequalityLet E be a nearly spherical set and let us look back at the estimate (3.2) stated inFuglede’s Theorem 3.1. Observe that in the previous argument we have only used partof the information provided by (3.2), since we have not exploited the presence of thefull norm of u in H1(Sn−1).The fact that in the quantitative isoperimetric inequality (4.3) we are throwingaway some valuable information encoded in the isoperimetric deficit D(E) can beunderstood by looking at the two sets E and F in Fig. 8. Indeed, E and F have thesame measure, the same Fraenkel asymmetry, but D(E) << 1, while D(F) >> 1.Therefore the quantitative inequality (4.3) gives a sharp information on E while it isuseless when applied to F . The reason is that the isoperimetric deficit depends stronglyon the oscillation of the boundary of the set, while the Fraenkel asymmetry only looksat the distance in measure of a set from a ball.123The quantitative isoperimetric inequality. . . 563Fig. 8 E and F have the same measure and the same asymmetryThis suggests that we should introduce a more precise index which takes intoaccount also the oscillation of the normals. To this aim, given a set of finite perimeterE and a ball Br (y) with the same volume as E , we are going to measure the distancefrom E to the ball in the following way, see Fig. 9. For every point x ∈ ∂∗E wetake the projection πy,r (x) of x on the boundary of Br (y) and consider the distance|νE (x) − νr,y(πy,r (x))| from the exterior normal to E at the point x to the exteriornormal to Br (y) at the projection point πy,r (x). Then, we take the L2 norm of thisdistance and minimize the resulting norm among all possible balls, thus gettingβ(E) := miny∈Rn{(12∫∂∗ E|νE (x) − νr,y(πy,r (x))|2 dHn−1(x))1/2}. (5.15)We shall refer to β(E) as to the oscillation index (or excess) of the set E .Observe that Fuglede’s Theorem 3.1 provides indeed an estimate for the oscillationindex. In fact, if E is a nearly spherical set satisfying (3.1), |E | = |B| and the W 1,∞norm of u is sufficiently small, by (3.14) at every point x ∈ ∂∗E the exterior normalto E is given byνE (x) = z(1 + u(z)) − ∇τ u(z)√(1 + u(z))2 + |∇τ u(z)|2,where z = x/|x | and thus x = z(1 + u(z)). Hence, from (3.2) we haveβ(E)2 ≤ 12∫∂∗ E∣∣∣∣νE (x) − x|x |∣∣∣∣2dHn−1 =∫∂∗ E(1 − νE (x) · x|x |)dHn−1≤ C∫Sn−1(1 − 1 + u(z)√(1 + u)2 + |∇τ u|2)dHn−1= C∫Sn−1√(1 + u)2 + |∇τ u|2 − (1 + u)√(1 + u)2 + |∇τ u|2dHn−1 (5.16)123564 N. Fuscoyπy,r(x)xBr(y)EFig. 9 The construction of the oscillation index≤ C∫Sn−1|∇τ u|2 dHn−1 ≤ C̃(P(E) − P(B)),for some constant C̃ depending only on n. The next result, proved by Julin and theauthor in [74], is an improved version of the quantitative isoperimetric inequality.Theorem 5.4 There exists a constant κ(n), such that for any set of finite perimeter E,with |E | = |Br | for some r > 0,β(E)2 ≤ κ(n)(P(E) − P(Br )). (5.17)Inequality (5.17) is stronger than (4.3), since a Poincaré type inequality shows thatthe asymmetry index α is always controlled by β, see Proposition 5.5. Before provingthis, let us give a closer look to the new index. Observe that by the divergence theoremwe immediately have12∫∂∗ E|νE (x) − νr,y(πy,r (x))|2 dHn−1 =∫∂∗ E(1 − νE (x) · x − y|x − y|)dHn−1= P(E) −∫En − 1|x − y| dx .Therefore, we may writeβ(E)2 = P(E) − (n − 1)γ (E), (5.18)123The quantitative isoperimetric inequality. . . 565where we have setγ (E) := maxy∈Rn∫E1|x − y| dx . (5.19)If y ∈ Rn is the center of an optimal ball for β(E) then y maximizes the integral on theright hand side of (5.19). Such a point will be called a center for the set E . Note thatin general such a center is not unique. The next result is contained in [74, Prop. 1.2].Proposition 5.5 There exists a constant C(n) such that if E is a set of finite perimeter,with |E | = |Br | for some r > 0, thenα(E)2 + D(E) ≤ Cr1−nβ(E)2.Proof Let E be a set of finite perimeter. Without loss of generality we may assumethat |E | = |B| and that E is centered at the origin, i.e.,β(E)2 =∫∂∗ E(1 − νE · x|x |)dHn−1.Since β(B) = 0, hence P(B) = (n − 1) ∫E 1|x |dx , from the equality above and (5.18)we haveβ(E)2 = P(B)D(E) −∫E\Bn − 1|x | dx +∫B\En − 1|x | dx . (5.20)Let us estimate the last two terms in (5.20). Since |E | = |B| we have|E\B| = |B\E | =: a. (5.21)Denote by A(R, 1) = BR\B1 and A(1, r) = B1\Br two annuli such that |A(R, 1)| =|A(1, r)| = a. Hence,R =(1 + aωn)1/nand r =(1 − aωn)1/n.By construction |A(R, 1)| = |E\B|. Therefore, since 1|x | is radially decreasing,∫E\Bn − 1|x | dx ≤∫A(R,1)n − 1|x | dx .Similarly, we have∫B\En − 1|x | dx ≥∫A(1,r)n − 1|x | dx .123566 N. FuscoTherefore we may estimate (5.20) byβ(E)2 ≥ P(B)D(E) −∫A(R,1)n − 1|x | dx +∫A(1,r)n − 1|x | dx= P(B)D(E) − n[ωn(Rn−1 − 1) − ωn(1 − rn−1)]= P(B)D(E) + nωn(2 −(1 + aωn) n−1n −(1 − aωn) n−1n).(5.22)The function f (t) = (1 + t) n−1n is uniformly concave in [−1, 1] and there exists apositive constant c(n) such that for all s, t ∈ [−1, 1]12( f (t) + f (s)) ≤ f(t2+ s2)− c(n)|t − s|2.Therefore, recalling (5.21), we may estimate (5.22) byβ(E)2 ≥ P(B)D(E) + 8nc(n)ωna2 = P(B)D(E) + 2nc(n)ωn|EB|2.Hence, the assertion follows. The strategy for proving inequality (5.17) is the same that we have seen in theprevious section and the starting point is again Fuglede’s inequality (3.2)which impliesin particular (5.17) if E is a nearly spherical set sufficiently close in W 1,∞ to the unitball. However, beside using the regularity properties of perimeter almost minimizers,we need to introduce another notion of minimality. We say that a set of locally finiteperimeter E is an area (K , r0)-quasi-minimizer if for every F , such that FE ⊂⊂Br (x), with r ≤ r0, the following inequality holdsP(E; Br (x)) ≤ K P(F; Br (x)).We observe that this definition is the counterpart in the framework of sets of finiteperimeter of the notion of quasi-minimum introduced by Giaquinta and Giusti in [79]in the context of variational integrals. Therefore it is not surprising that although quasi-minimality is a very weak property, yet area quasi-minimizers have some kind of mildregularity. This is indeed the content of the next result which was first proved by Davidand Semmes in [51] and then extended by Kinnunen et al. [88] to the metric spacessetting.Theorem 5.6 Suppose that E is an area (K , r0)-quasi-minimizer. Then there existsC > 1 such that for any 0 < r < r0 and every x ∈ ∂ E there are points y, z ∈ Br (x)for whichBr/C (y) ⊂ E and Br/C (z) ⊂ Rn\E .123The quantitative isoperimetric inequality. . . 567We are now ready to prove the main result of this section.Proof of Theorem 5.4 Step 1 As in the previous section, we start by observing thatin order to prove (5.17) it is enough to assume |E | = |B| and that D(E) is smallerthan a given constant δ0 to be chosen later. Moreover, the same reduction argumentused to prove Lemma 4.2 shows that also in this case we may assume without loss ofgenerality that E ⊂ BR0 , where the radius R0 depends only on the dimension n, see[74, Lemma 3.2].Let c0 > 0 be a constant which will be chosen at the end of the proof. From what wehave just observed it is enough to show that there exists δ0 > 0 such that, if D(E) ≤ δ0,E ⊂ BR0 and |E | = |B|, thenc0β(E)2 ≤ P(E) − P(B).We argue by contradiction assuming that there exists a sequence of sets E j ⊂ BR0such that |E j | = |B|, D(E j ) → 0 andP(E j ) < P(B) + c0β(E j )2. (5.23)By the compactness Theorem 2.3 it follows that, up to a subsequence, E j → E∞ inmeasure and by the lower semicontinuity of the perimeter we immediately concludethat E∞ is a ball of radius one. It is easily checked that the functional γ defined in(5.19) is continuous with respect to the convergence in measure. Therefore, since theE j are converging to a ball of radius one and P(E j ) → P(B), we have thatβ(E j )2 = P(E j ) − (n − 1)γ (E j ) → 0.As in the previous section we replace each set E j by a minimizer Fj of the followingproblemmin{P(F) + ∣∣|F | − |B|∣∣ + 14|β(F)2 − β(E j )2|, F ⊂ BR0}for some fixed > n. It is not too hard to show that the above functional is lowersemicontinuous with respect to the convergence in measure, see [74, Lemma 3.4], andtherefore a minimizer exists. Moreover, up to a subsequence, we may assume thatFj → F∞ in measure. From the minimality of Fj , (5.23) and Lemma 5.3 we havethatP(Fj ) + ∣∣|Fj | − |B|∣∣ + 14|β(Fj )2 − β(E j )2| ≤ P(E j ) < P(B) + c0β(E j )2≤ P(Fj )+∣∣|Fj |−|B|∣∣+c0β(E j )2.Hence |β(Fj )2 − β(E j )2| ≤ 4c0β(E j )2, which implies β(Fj ) → 0 andβ(E j )2 ≤ 11 − 4c0 β(Fj )2. (5.24)123568 N. FuscoTherefore F∞ is a minimizer of the problemmin{P(F) + ∣∣|F | − |B|∣∣ + 14β(F)2 : F ⊂ BR0}.Thus from Lemma 5.3 we conclude that F∞ is a ball B(x0), for some x0 ∈ Rn .Step 2 We claim that for any ε > 0, B1−ε(x0) ⊂ Fj ⊂ B1+ε(x0) for j large enough.To this aim we show that the sets Fj are area (K , r0)-quasi-minimizers and use The-orem 5.6. Let G ⊂ Rn be such that GFj ⊂⊂ Br (x), r ≤ r0.Case 1 Suppose that Br (x) ⊂ BR0 . By the minimality of Fj we obtainP(Fj ) ≤ P(G) + 14|β(Fj )2 − β(G)2| + ∣∣|Fj | − |G|∣∣ (5.25)Assume that β(Fj ) ≥ β(G) (otherwise the argument is similar) and denote by yG acenter of G. Then we get|β(Fj )2 − β(G)2| ≤∫∂∗ Fj(1 − νFj · z − yG|z − yG |)dHn−1(z)−∫∂∗G(1 − νG · z − yG|z − yG |)dHn−1(z)=∫∂∗ Fj ∩Br (x)(1 − νFj · z − yG|z − yG |)dHn−1(z)−∫∂∗G∩Br (x)(1 − νG · z − yG|z − yG |)dHn−1(z)≤ 2[P(Fj ; Br (x)) + P(G; Br (x))].Therefore, from (5.25) we getP(Fj ; Br (x)) ≤ 3P(G; Br (x)) + 2∣∣|Fj | − |G|∣∣.From the above inequality the (K , r0)-quasi-minimality immediately follows observ-ing that|FjG| ≤ ω1/nn r1/n|FjG| n−1n ≤ C(n)r1/n[P(Fj ; Br (x)) + P(G; Br (x))]and choosing r0 sufficiently small.Case 2 If |Br (x)\BR0 | > 0, we may writeP(Fj ; Br (x)) − P(G; Br (x)) = P(Fj ; Br (x)) − P(G ∩ BR0; Br (x))+ P(G ∩ BR0; Br (x)) − P(G; Br (x))= P(Fj ; Br (x)) − P(G ∩ BR0; Br (x))+ P(BR0) − P(G ∪ BR0)≤ P(Fj ; Br (x)) − P(G ∩ BR0; Br (x)).123The quantitative isoperimetric inequality. . . 569FromCase 1we have that this term is less than (K −1)P(G∩BR0 ; Br (x))which in turnis smaller than (K −1)P(G; Br (x)). Hence, all Fj are (K , r0)-quasi-minimizers withuniform constants K and r0. The claim then follows from the theory of (K , r0)-quasi-minimizers and the fact that Fj → B(x0) in L1. Indeed, arguing by contradiction,assume that there exists 0 < ε0 < 2r0 such that for infinitely many j one can findx j ∈ ∂ Fj for whichx j /∈ B1+ε0(x0)\B1−ε0(x0).Let us assume that x j ∈ B1−ε0(x0) for infinitely many j (otherwise, the argumentis similar). From Theorem 5.6 it follows that there exist y j ∈ B ε02(x j ) such thatB ε02C(y j ) ⊂ B(x0)\Fj . This implies|B(x0)\Fj | ≥ |B ε02C| > 0,which contradicts the fact that Fj → B(x0) in L1, thus proving the claim.Step 3 Let us now translate Fj , for j large, so that the resulting sets, still denoted byFj , are contained in BR0 , have their barycenters at the origin and converge to B. Weare going to use Theorem 5.2 to show that the Fj are C1,1/2 and converge to B in C1,αfor all α < 1/2. To this aim, fix a small ε > 0. From Step 2 we have that for j largeB1−ε ⊂ Fj ⊂ B1+ε. (5.26)We want to show that when j is large Fj is a (′, r0)-almost minimizer for someconstants ′, r0 independent of j . To this aim, fix a set G ⊂ Rn such that GFj ⊂⊂Br (y), with r < r0.If Br (y) ⊂ B1−ε, from (5.26) it follows that GFj ⊂⊂ Fj for j large enough. Thisimmediately yields P(Fj ) ≤ P(G).If Br (y) ⊂ B1−ε, choosing r0 and ε sufficiently small we have thatBr (y) ∩ B1/2 = ∅. (5.27)Denote by yFj and yG the centers of Fj and G, respectively. If ε is sufficiently small,from (5.26) and Lemma 5.7 below we have that for j large|yFj | ≤14and |yG | ≤ 14. (5.28)By the minimality of Fj we haveP(Fj ) ≤ P(G) + 14|P(Fj ) − P(G)| + ∣∣|Fj | − |G|∣∣ + n − 14|γ (Fj ) − γ (G)|,which immediately impliesP(Fj ) ≤ P(G) + 2|FjG| + (n − 1)|γ (Fj ) − γ (G)|. (5.29)123570 N. FuscoWe may estimate the last term simply byγ (Fj ) − γ (G) ≤∫Fj1|x − yFj |dx −∫G1|x − yFj |dx ≤∫Fj G1|x − yFj |dxandγ (G) − γ (Fj ) ≤∫G1|x − yG | dx −∫F1|x − yG | dx ≤∫Fj G1|x − yG | dx .Therefore, recalling (5.27) and (5.28), we have|γ (Fj ) − γ (G)| ≤ 4|FjG|.From this estimate and inequality (5.29) we may then conclude thatP(Fj ) ≤ P(G) + (2 + 4(n − 1)) |FjG| ≤ P(G) + ′ rn .Hence, the sets Fj are (′, r0)- almost minimizers with uniform constants ′ and r0.Thus, Theorem 5.2 yields that the Fj are C1,1/2 and that, for j large,∂ Fj = {z(1 + u j (z)) : z ∈ ∂ B}for some u j ∈ C1,1/2(∂ B) such that u j → 0 in C1(∂ B1).Step 4 By the minimality of Fk , (5.23) and (5.24) we haveP(Fj ) + ∣∣|Fj | − |B|∣∣ ≤ P(E j ) < P(B) + c0β(E j )2 ≤ P(B) + c01 − 4c0 β(Fj )2.(5.30)We are almost in a position to use (5.16) to obtain a contradiction. We only need torescale the Fj so that the volume constraint is satisfied. Thus, set F ′j := λ j Fj , whereλ j is such that λnj |Fj | = |B|. Then λ j → 1 and also the sets F ′j converge to B in C1and have their barycenters at the origin. Therefore, since > n, P(Fj ) → n|B| and|Fj | → |B|, we have that for j sufficiently large|P(F ′j ) − P(Fj )| = |λn−1j − 1| P(Fj ) ≤ |λnj − 1| |Fj | = ∣∣|F ′j | − |Fj |∣∣. (5.31)Then (5.30) and (5.31) yieldP(F ′j ) ≤ P(Fj ) + ∣∣|Fj | − |B|∣∣ < P(B) + c01 − 4c0 β(Fj )2= P(B) + c0 λ1−nj1 − 4c0 β(F′j )2.which contradicts (5.16) if 2c0/(1 − 4c0) < 1/C̃ and j is large. 123The quantitative isoperimetric inequality. . . 571Lemma 5.7 For every ε > 0 there exists δ > 0 such that if F ⊂ BR0 and |FB| < δ,then |yF | < ε for every center yF of F.Proof We argue by contradiction and assume that there exist Fj ⊂ BR0 such thatFj → B in measure and yFj → y0 with |y0| ≥ ε, for some ε > 0. Then we wouldhave∫Fj1|x | dx ≤∫Fj1|x − yFj |dx .Letting j → ∞, by the dominated convergence theorem the left hand side convergesto∫B1|x | dx , while the right hand side converges to∫B1|x−y0| dx . Thus we have∫B1|x | dx ≤∫B1|x − y0| dx .By the divergence theorem we conclude that∫∂ B1 dx ≤∫∂ Bx · x − y0|x − y0| dxand this inequality may only hold if y0 = 0, thus leading to a contradiction. We conclude this section quoting a recent paper by Neumayer [103] dealing withthe same kind of improved quantitative estimate that we have discussed here, but inthe case of the anisotropic perimeter. It turns out that if the function ϕ which definesthe anisotropic perimeter is of class C2 and satisfies a suitable form of ellipticity, thenone can prove an inequality of the type (5.17) with β replaced by a suitable anisotropicoscillation index. In this case, see [103, Th. 1.5] one may get the stability estimatewith the exponent 2 and a constant depending on the function ϕ. In the general case,i.e., when no assumptions on ϕ are made, a stability estimate is still obtained with aconstant depending only on the dimension, but with a not optimal power [103, Th. 1.1].5.4 The isoperimetric inequality in higher codimensionThe isoperimetric inequality in higher codimension goes back to a celebrated paperby Almgren [3]. In that paper he proved the manifold counterpart of the classicalisoperimetric inequality established by De Giorgi in [53]. Though stated in the frame-work of currents, in the particular case of a smooth (n − 1)-dimensional manifolds ⊂ Rn+k without boundary, spanning an area minimizing smooth surface M , hisinequality states thatHn−1() ≥ Hn−1(∂ D),where D is an n-dimensional flat disk with the same area as M . Moreover equalityoccurs if and only if is the boundary of a flat disk. The stability of this inequality has123572 N. Fuscobeen addressed, in the context of currents, in a recent paper by Bögelein et al. [18].In order to describe the result we restrict to the case of smooth (n − 1)-dimensionalclosed surfaces in Rn+k and give the definitions of the quantities that come intoplay.Denoting by Q() an area minimizing n-dimensional surface with boundary theisoperimetric deficit is defined asD() := Hn−1() − Hn−1(∂ Dr )Hn−1(∂ Dr ) ,where Dr is an n-dimensional flat disk in Rn+k with the same area as Q(), i.e.,Hn(Dr ) = Hn(Q()). Note that the area minimizing surface Q() may have singu-larities even if is smooth. It is then clear why the use of currents with finite massbecomes unavoidable even to give the basic definitions. This is even more true inthe case of the asymmetry index α() whose precise definition is more technical andrequires the use of a certain seminorm defined for integer multiplicity currents [18,Sect. 2]. However, the underlying geometric idea is easy to describe. Given any flatdisk Dr with the same area as Q(), first one considers an area minimizing cylindrictype surface (Dr ) spanned by the boundary components and ∂ Dr , and afterwardsone takes the infimum of the surface area Hn((Dr )) among all possible disks Dr :α() := r−n inf {Hn((Dr )) : Hn(Dr ) = Hn(Q())}.With these definitions in hand we may state the following quantitative isoperimetricinequality proved in [18, Th. 2.1].Theorem 5.8 Let n ≥ 2 and k ≥ 0. There exists a constant C = C(n, k) > 0 suchthat for any (n − 1)-dimensional closed surface ⊂ Rn+k the following inequalityholds:α()2 ≤ CD(). (5.32)Note that if is the boundary of a smooth open set E contained in an n-dimensionalhyperplane, then the asymmetry index α() coincides with the Fraenkel asymmetryindex α(E) defined in (4.1). Hence, (5.32) reduces to (4.3). In particular this showsthat the exponent 2 on the left hand side of the inequality cannot be improved, sincewe know that it is already optimal for (4.3).As in the two previous sections the strategy to prove Theorem 5.8 is to show firsta Fuglede type inequality and then to reduce the general case to it via a regularityargument. However, here the situation is more delicate and involved due to the highercodimension. First of all, the analogue of Fuglede’s result deals with a sphericalgraph over Sn−1 in Rn+k , i.e., a manifold which can be parametrized by a mapX : Sn−1 → Rn+k of the formX (x) := (1 + u(x))(x, 0) + (0, v(x)) x ∈ Sn−1,where u ∈ C1(Sn−1) and v ∈ C1(Sn−1,Rk) have both small C1-norms. In the presentcase a substantial difficulty arises from the fact that, beside imposing the volume123The quantitative isoperimetric inequality. . . 573constraintHn(Q()) = ωn and that the barycenter of is at the origin, we have alsoto fix some second order quantities. This can be done for instance by assuming thatthey are all equal to zero, i.e.,∫zi z j dHn−1 = 0 (5.33)for any choice of i = 1, . . . , n and j = n + 1, . . . , n + k. Differently from the casek = 0 considered in Sect. 3, in which v does not appear, the conditions (5.33) play acrucial role in the estimation of the n · k first order Fourier coefficients of v. Instead,the bounds on the first order Fourier coefficients of u and the zero order Fouriercoefficient of v follow from the barycenter condition, while the zero order Fouriercoefficient of u is controlled by using the constraint Hn(Q()) = ωn . Then, underthe above assumptions on u and v one has the following inequality, see [18, Th. 4.1],Hn−1() − Hn−1(Sn−1) ≥ c(n)[‖u‖2H1(Sn−1) + ‖v‖2H1(Sn−1,Rk )], (5.34)where c(n) is a constant depending only on n, provided the C1 norms of u and v aresufficiently small. Note also thatα() ≤ c(n)[‖u‖2L2(Sn−1) + ‖v‖2L2(Sn−1,Rk )].Combining this inequality with (5.34) one then gets the quantitative isoperimetricinequality (5.32) in this case.The next step is to reduce the general case to the previous one by the same contra-diction argument that we have seen in Sect. 5.2. Beside the technical complicationsdue to the fact that one is now dealing with higher codimension, the main ingredientscome from the theory of currents that ‘almost’ minimize certain elliptic variationalintegrals. The regularity results needed in the proof were obtained in [20,54].5.5 The isoperimetric inequality on the sphereThe isoperimetric property of geodesic balls on the sphere and in hyperbolic spacesgoes back to Schmidt [108]. Before stating it, let us fix some basic notation. Recallthat if x, y are two points in the n-dimensional sphere Sn ⊂ Rn+1, n ≥ 1, the geodesicdistance between them is given bydistSn (x, y) := arccos(x · y).Thus, the open geodesic ball with center at a point p ∈ Sn and radius ϑ ∈ (0, π) isdefined by settingBϑ(p) := {x ∈ Sn : distSn (x, p) < ϑ}.123574 N. FuscoWe just write Bϑ if the center of the geodesic ball is at the north pole en+1. Then, theisoperimetric inequality on the sphere states that if E ⊂ Sn is a Borel set such thatHn(E) = Hn(Bϑ) thenP(Bϑ) ≤ P(E), (5.35)with the equality occurring if and only if E is a geodesic ball. Here, P(E) denotesthe perimeter of E on the sphere, see the definition below, which coincides with theHn−1-measure of the (relative) boundary of E on the sphere if E is a C1 manifold.Note that from (5.35) we immediately have that if E ⊂ Sn is a set of finite perimetersuch that P(E) = P(Bϑ) for some 0 < ϑ < π , then|E | ≤ |Bϑ | or |Sn\E | ≤ |Bϑ |.Also in this case equality occurs if and only if E is a geodesic ball.We recall that if E ⊂ Sn is a Borel set, the perimeter of E on the sphere can bedefined similarly to the euclidean case by settingP(E) = sup{∫Edivτ ϕ dx : ϕ ∈ C∞(Sn; Rn+1), ϕ(x) · x = 0 for all x ∈ Sn, ||ϕ||∞≤1}.If P(E) < ∞ by Riesz’s representation theorem it follows that the tangential gradientof χ E can be represented as a vector valued Radon measure Dτχ E . Thus, as in theeuclidean case one may consider the set ∂∗E of all points in Sn such that the followinglimit existsνE (x) := − limϑ↓0Dτχ E (Bϑ(x))|Dτχ E |(Bρ(x))and satisfies |νE (x)| = 1 and νE (x) ∈ TxSn , where TxSn stands for the tangentialspace to Sn at the point x . Then the De Giorgi structure theorem on the sphere statesthat ∂∗E is countably (n − 1)-rectifiable and that the total variation measure |Dτχ E |coincides with Hn−1 ∂∗E . Moreover, if g is a C1 tangential vector-field on Sn wehave the following Gauss–Green formula, see [110, (7.6)]∫EdivSn g dHn =∫∂∗ Eg · νE dHn−1. (5.36)Note that when E is a smooth open subset of Sn , the unit vector νE (x) ∈ TxSncoincides with the outer unit normal vector to ∂ E at x . In the case of a geodesicball Bϑ(p) it can be easily checked that the outer unit normal vector field along itsboundary is given byνBϑ (p)(x) =(x · p)x − p√1 − (x · p)2 . (5.37)123The quantitative isoperimetric inequality. . . 575In view of the results that we have seen in Sect. 4, the natural counterpart of (4.3)would be the inequalityP(E) − P(Bϑ)P(Bϑ)≥ c(n)α(E)2, (5.38)where now the Fraenkel asymmetry index is defined byα(E) := minp∈Sn{Hn(EBϑ(p))Hn(Bϑ) : Hn(E) = Hn(Bϑ)}.When compared with (4.3) inequality (5.38), even if it looks similar, has a completelydifferent nature. In fact (4.3) is scaling invariant while there is no scaling at all on Sn . Itwould be quite easy to adapt one of the different arguments that we have seen in Sect. 4and in Sect. 5.2 in order to prove (5.38) with a constant depending additionally on thevolume of the set E , but possibly blowing up as ϑ ↓ 0. Indeed, the difficult case iswhen the set E has a small volume sparsely distributed over the sphere. In this situationa localization argument aimed to reduce the problem to the flat Euclidean estimate(4.3) does not work. However, as in Sect. 5.3 one can prove a stronger inequality byreplacing α(E) with an oscillation index which is the counterpart in our new settingof the one defined in (5.15).The idea is the same as in the euclidean case. Given a geodesic ball Bϑ(p) withthe sameHn-measure of E , for every point x ∈ ∂∗E we want to measure the distancebetweenνE (x) and the unit normal vector νBϑ (p)(πϑ,p(x)),whereπϑ,p(x) is the closestpoint in ∂ Bϑ(p) in the sense of geodesic distance. However, since the two normalsνE (x) and νBϑ (p)(πϑ,p(x)) lie in two different tangent spaces in order to measuretheir distance correctly we have to parallel transport the normal νBϑ (p)(πϑ,p(x)) to thetangent space TxSn . Since the parallel transport on the sphere is just translation alonggreat circles, the normal νBϑ (p)(πϑ,p(x)) is thus carried into the normal νBϑ(x)(p)(x) atthe point x to the geodesic ball with center p and radius ϑ(x) := distSn (x, p). So, asin Sec. 5.3, we measure the L2 distance between the normals νE (x) and νBϑ(x)(p)(x)and then minimize the resulting norm among all possible geodesic balls having thesame measure of E . This leads to the following definition of oscillation index for theset Eβ(E) := minp∈Sn[12∫∂∗ E∣∣νE (x) − νBϑ(x)(p)(x)∣∣2 dHn−1] 12.The quantitative isoperimetric inequality proved in [19, Th. 1.1] reads as follows.Theorem 5.9 There exists a constant c(n) such that for any Borel set E ⊂ Sn withHn(E) = Hn(Bϑ) for some ϑ ∈ (0, π), the following inequality holdsP(E) − P(Bϑ) ≥ c(n)β2(E). (5.39)123576 N. FuscoAs in the euclidean case, see the discussion made in Sect. 5.3, inequality aboveimplies (5.38), since a Poincaré type inequality still holds in the case, allowing tocontrol α(E) by the oscillation index β(E), see [19, Lemma 2.7].Lemma 5.10 There exists a constant c = c(n) > 0 such that for any Borel set E ⊂ Snwith Hn(E) = Hn(Bϑ) for some ϑ ∈ (0, π) there holdsβ(E)2 ≥ c(n)P(Bϑ)α(E)2.As in Sect. 5.3 the starting point for the proof of Theorem 5.9 is a Fuglede-typestability result aimed to establish (5.39) in the special case of sets E ⊂ Sn whoseboundary can be written as a radial graph over the boundary of a geodesic ball Bϑ(p)with the same volume. To establish such a result one could follow in principle thestrategy used in the euclidean case in Theorem 3.1. However, to deduce (5.39) forradial graphs with a constant not depending on the volume needs much more care inthe estimations, see the proof of Theorem 3.1 in [19]. The main difficulty arises whenpassing from the special situation of radial graphs to arbitrary sets. To deal with thisissue one needs to change significantly the strategies that we have seen in the previoussections.To explain where the major difficulties come from, we observe that as in the euclid-ean case the oscillation index can be re-written as the difference of P(E) and a suitablepotential on E . To this aim, note that12∫∂∗ E∣∣νE (x) − νBϑ(x)(p)(x)∣∣2 dHn−1 =∫∂∗ E1 − νE (x) · νBϑ(x)(po)(x) dHn−1= P(E) −∫∂∗ EνE (x) · νBϑ(x)(po)(x) dHn−1.To proceed further, we recall (5.37) and setX (x) := νBϑ(x)(p)(x) =(x · p)x − p√1 − (x · p)2 .Note that X is a tangential vector field, that is X (x) · x = 0 for all x ∈ ∂∗E . Hence,by the Gauss–Green formula (5.36) we have12∫∂∗ E∣∣νE (x) − νBϑ(x)(p)(x)∣∣2 dHn−1 = P(E) −∫EdivSn X dHn .Computing the tangential divergence divSn X we obtain, denoting by {τ1, . . . , τn} anorthonormal base for TxSn ,123The quantitative isoperimetric inequality. . . 577divSn X =n∑i=1τi ·⎡⎢⎣ (τi · p)x + (x · p)τi√1 − (x · p)2 + ((x · p)x − p)(x · p)(τi · p)(√1 − (x · p)2)3⎤⎥⎦=n∑i=1⎡⎢⎣ x · p√1 − (x · p)2 −(x · p)(τi · p)2(√1 − (x · p)2)3⎤⎥⎦ = (n − 1)(x · p)√1 − (x · p)2 .Inserting this equality in the formula above, we conclude thatβ2(E) = P(E) − γ (E),whereγ (E) := (n − 1)maxp∈Sn∫Ex · p√1 − (x · p)2 .From this formula it is clear that the core of the proof is to provide estimates indepen-dent of the volume of E for the potential∫Ex · p√1 − (x · p)2 dHn−1 (5.40)and for its maximum with respect to p. This requires some technically involved ideasand strategies. In particular, in the contradiction argument used to deduce (5.39) forgeneral sets from the case of a radial graph we need to show that all the constants areindependent of the volume of E . The arguments become particularly delicate whenthe volume of E is small. In this case inequality (5.39) shows a completely differentnature depending on the size of the ratio β2(E)/P(Bϑ). In fact, if Hn(E) → 0 andalso β2(E)/P(Bϑ) → 0, then E behaves asymptotically like a flat set, i.e., a set inRnand inequality (5.39) can be proven by reducing to the euclidean case, rescaling andthen arguing as when E has large volume. However, the most difficult situation to dealwith is when Hn(E) → 0 and β2(E)/P(Bϑ) converge to a strictly positive number.This case has to be treated with ad hoc estimates for the potential (5.40).5.6 The fractional perimeterThe fractional s-perimeter has recently attracted the interest of several authors. Inparticular, Caffarelli et al. have initiated in [31] the study of Plateau-type problems inthe fractional setting, see also [32]. Recall that for s ∈ (0, 1) and a Borel set E ⊂ Rn ,n ≥ 2, the fractional s-perimeter is defined by settingPs(E) :=∫E∫Ec1|x − y|n+s dxdy.123578 N. FuscoNote that if Ps(E) < ∞, thenPs(E) = 12[χ E ]pW σ,p(Rn),for p ≥ 1 and σ p = s, where [χ E ]pW σ,p(Rn) denotes the Gagliardo W σ,p-seminormof the characteristic function of E . The functional Ps(E) can be thought as a (n −s)-dimensional perimeter in the sense that Ps(λE) = λn−s Ps(E) for any λ > 0.Note that since for s ∈ (0, 1) BV (Rn) is imbedded in W s(n−1)n−s , n−sn−1 (Rn), see [81,Theorem 1.4.4.1], the s-perimeter of E is finite if E has finite (standard) perimeterand finite measure. On the other hand, Ps(E) can be finite even if the Hausdorffdimension of ∂ E is strictly greater than n − 1, see for instance [107]. Observe alsothat if E, F are two bounded sets of non trivial and finite s-perimeter in generalPs(E ∪ F) < Ps(E) + Ps(F).However, the s-perimeter provides an approximation of the standard euclidean perime-ter since it can be shown that for any bounded set of finite perimeter Elims↑1(1 − s) Ps(E) = ωn−1 P(E). (5.41)Formula (5.41) originates from a paper by Bourgain, Brezis andMironescu [23, Th. 3’and Rem 4]. It was completed by Dávila [52], see also [115] for a simpler proof. Sub-sequently, the same formula was proved in [32] and [7] under additional smoothnessassumptions on E . On the other hand, as a consequence of [98, Th. 3], we have that,for any set E of finite measure and finite s-perimeter for all s ∈ (0, 1),lims↓0 s Ps(E) = nωn |E |.Also for the s-perimeter balls are the isoperimetric sets, since it can be proved that ifE ⊂ Rn is a measurable set with |E | = |Br | for some r > 0, thenPs(Br ) ≤ Ps(E) (5.42)with equality holding if and only if E is a ball. Inequality (5.42) can be deduced froma symmetrization result due to Almgren and Lieb [4], while the cases of equality havebeen determined in [71]. Note also, see for instance [62, (2.11)], that the s-perimeterof the unit ball is given byPs(B) = 21−sπ n−12 P(B)s(n − s) ·( 1−s2)( n−s2) ,where is the so-called gamma function. In order to state a quantitative version ofthe isoperimetric inequality (5.42), if E is a measurable set such that |E | = |Br | ands ∈ (0, 1), we define the s-isoperimetric deficit as123The quantitative isoperimetric inequality. . . 579Ds(E) := Ps(E) − Ps(Br )Ps(Br ).The next result has been proven in [62, Th. 1.1].Theorem 5.11 For every n ≥ 2 and s0 ∈ (0, 1) there exists a positive constantC(n, s0) such thatα(E)2 ≤ C(n, s0)Ds(E), (5.43)whenever s ∈ [s0, 1] and E has positive, finite measure.The interesting feature of inequality (5.43) is that it holds uniformly with respectto s, at least if s is bounded away from zero. Therefore, thanks to (5.41), it implies inparticular the quantitative inequality (4.3) for the standard perimeter. As we have seenin the previous sections, the starting point is a Fuglede type result, see [62, Th. 2.1],which now states that there exist two positive constants ε, c0, depending only on n,such that if E is a nearly spherical set as in Theorem 3.1, with ‖u‖W 1,∞(Sn−1) < ε,then for all s ∈ (0, 1)Ps(E) − Ps(B) ≥ c0([u]2H1+s2 (Sn−1)+ s Ps(B) ‖u‖2L2(Sn−1)), (5.44)where the Gagliardo seminorm [u]H1+s2 (Sn−1)is given by[u]H1+s2 (Sn−1):=(∫∫Sn−1×Sn−1|u(x) − u(y)|2|x − y|n+s dHn−1x dHn−1y) 12.Since it can be proved, see [23, Cor. 2] and also [62, (8.4)] for a different proof, thatlims↑1(1 − s)[u]2H1+s2 (Sn−1)= ωn−1∫Sn−1|∇τ u|2,one immediately recovers the Fuglede’s estimate (3.2) from (5.44). As we have seenalready in different situations, in order to pass from a nearly spherical set to a generalset of finite perimeter one has to use the right notion of perimeter almost minimizer andthe right regularity results. The definition of perimeter almost minimizer given in (5.7)can be immediately extended to the s-perimeter in the obvious way and the counterpartin this new setting of Theorem 5.2 has been proved by Caputo and Guillen in [34].However, in order to prove Theorem 5.11 one needs to show that all the regularityestimates needed in the proof are in fact independent of s, at least if s is bounded awayfrom zero. This requires the use of more technically involved arguments and of moredelicates estimates, see Sect. 4 in [62].123580 N. Fusco5.7 The isoperimetric inequality in Gauss spaceThe Gauss space is the space Rn , n ≥ 1, endowed with the measure γ given byγ (E) = 1(2π)n/2∫Ee−|x |22 dx,for any Lebesgue measurable set E ⊂ Rn . Note that (Rn, γ ) is a probability space,since γ (Rn) = 1. Clearly, γ is invariant under a rotation around the origin, while isnot invariant under translation. Similarly to what we do in the euclidean space, if E ismeasurable and ⊂ Rn is an open set we define the Gaussian perimeter of E in asPγ (E; ) =√2π sup{∫E(divϕ − x · ϕ)dγ : ϕ ∈ C∞c ( ; Rn), ||ϕ||∞ ≤ 1}.It turns out that if E has finite Gaussian perimeter in then it has also locally finiteperimeter in in the Euclidean sense andPγ (E; ) = 1(2π)(n−1)/2∫∂∗ Ee−|x |22 dHn−1.The perimeter of E in Rn will be simply denoted by Pγ (E).The isoperimetric inequality in the Gauss space asserts that among all subsets ofRn with prescribed Gaussian measure, half-spaces have the least Gaussian perimeter.Precisely, for any ν ∈ Sn−1 and s ∈ R denote by Hν,s the half spaceHν,s := {x ∈ Rn : x · ν > s}.Then we have γ (Hν,s) = (s) and Pγ (Hν,s) = e− s22 , where is the strictly decreas-ing function defined for all s ∈ R by setting(s) = 1√2π∫ ∞se−t22 dt.Then the analytic form of the Gaussian isoperimetric inequality states that if γ (E) =γ (Hν,s) = (s) for some s ∈ R, thenPγ (E) ≥ Pγ (Hν,s) or equivalently Pγ (E) ≥ e− s22 = e− {−1(γ (E))}22 , (5.45)with the equality holding if and only if E = Hν,s for some ν ∈ Sn−1. This inequalitywas independently established by Borell in [22] and by Sudakov and Cirel’son in[112]. However, the characterization of half-spaces as the unique minimizers in theisoperimetric Gaussian problem is a relatively recent result of Carlen and Kerce [36],relying on the theory of rearrangements aswell as on probabilistic techniques involvingtheMehler semigroup of contractions.Adifferent proof has been recently given in [44],123The quantitative isoperimetric inequality. . . 581where the Ehrhard symmetrization [55] plays the same role of Steiner symmetrizationin De Giorgi’s proof [53] of the isoperimetric inequality.The stability of the isoperimetric property of half spaces in (5.45) has been addressedfor the first time in [44]. In order to state the result proved therein we introduce theGaussian isoperimetric deficit Dγ (E) and the Gaussian asymmetry αγ (E) by setting,respectively,Dγ (E) := Pγ (E) − e− s22 , αγ (E) := minν∈Sn−1γ (EHν,s),whenever E is a measurable set with γ (E) = (s), n ≥ 1. With this definitions inhand, the stability result proved in [44, Th. 1.1] reads as follows.Theorem 5.12 Let n ≥ 2. For any 0 < λ < 1 there exists a positive constant C(n, λ)depending only on n and λ such thatαγ (E)2 ≤ C(n, λ)Dγ (E). (5.46)for every measurable set E ⊂ Rn such that γ (E) = λ.Note that also in (5.46) the power 2 is sharp. To see this one can consider indimension 2 the family of sets constructed in [44, Lemma 6.1]. When n = 1, thequantitative version of the isoperimetric inequality (5.46) takes a somewhat differentform. In fact one can prove that if n = 1, for any 0 < λ < 1 there exists a positiveconstant C(λ) such thatαγ (E)√log(1αγ (E))≤ C(λ)Dγ (E) (5.47)for every measurable set E ⊂ R such that γ1(E) = λ. Also in this case the result issharp, in the sense that the left hand side of (5.47) cannot be replaced by any functionof αγ (E) decaying slower as αγ (E) → 0, see the sets defined in [44, Lemma 3.5].As in the proof of (4.3) that we discussed in Sect. 4, the strategy to prove (5.46)is to reduce, via subsequent simplifications, to classes of sets E enjoying specialadditional geometric properties. A first reduction consists in showing that it sufficesto prove (5.46) for (n − 1)-symmetric sets E , namely sets which are symmetric about(n − 1) mutually orthogonal hyperplanes containing the origin. This is obtained byproving a counterpart of Theorem 4.6 in the Gaussian setting. Once this reduction isperformed, in order to prove (5.46) in dimension n = 2 it is enough to deal only withsets that are symmetric with respect to one of the two coordinate axes. In this casethe proof follows by using ad hoc geometric constructions, together with a carefulstudy of a particular class of almost optimal sets. Finally, by combining Ehrhardsymmetrization with the reduction result to (n − 1)-symmetric sets, one may passfrom a generic set E to a Cartesian product of Rn−2 times an epigraph in R2. Thanksto the tensorial properties of the Gaussian density, this class of sets can be handled bythe two-dimensional result.123582 N. FuscoThe fact that the Gaussian isoperimetric inequality (5.45) is independent of thedimension suggests that also the constant in the quantitative inequality (5.46) shouldonly depend on λ. Beside its intrinsic mathematical interest, having a constant inde-pendent of n in (5.45) would have some interesting applications in probability, aspointed out in [101, Sect. 1.4]. Indeed in [101] a stability estimate with a non sharplogarithmic dependence on the isoperimetric deficit Dγ (E) was proven, but with adimension free constant. This last result was later on greatly improved by Eldan [56]who proved inequality (5.46), actually an even stronger one, with a dimensionless con-stant and an almost optimal dependence on Dγ (E), actually just a logarithm belowthe optimal growth. However, it was only a very recent paper by Barchiesi et al. [10]that settled the issue in a complete and satisfactory way.To state properly their result we have to introduce a sharper index, which in thegaussian context plays the role of the one that we have introduced in Sect. 5.3. To thisaim, for any measurable set E ⊂ Rn we define the non-renormalized barycenter of Eby setting b(E) := ∫E x dγ . Then, following [10] we define a strong asymmetry bysettingβγ (E) := minν∈Sn−1|b(E) − b(Hν,s)|.In [10] the following result is proved.Theorem 5.13 There exists an absolute constant C such that for every s ∈ R and forevery set E ⊂ Rn with γ (E) = (s) the following estimate holdsβγ (E) ≤ C(1 + s2)Dγ (E). (5.48)Observe that this inequality is extremely interesting for several reasons. First, theconstant C is dimension free, second, the dependence on the mass of E through s isalso optimal, see [10, Rem. 1]. Moreover, a Poincaré type inequality shows that forany measurable set E with γ (E) = (s) thenβγ (E) ≥ es224αγ (E)2.Finally, the new asymmetry index is clearly related to the Gaussian counterpart of theoscillation index defined in (5.15) since one can prove that2Dγ (E) + 2√2πβγ (E) = minν∈Sn−11(2π)(n−1)/2∫∂∗ E|νE (x) − ν|2 dHn−1.The key point in the proof of Theorem 5.13 is to study the minimization problemmin{Pγ (E) + ε2|b(E)|2 + |γ (E) − (s)| : E ⊂ Rn}and to show that for ε sufficiently small, and large, the only minima are the halfspaces Hν,s . A part from the penalization term that eventually forces the minimizers to123The quantitative isoperimetric inequality. . . 583have mass equal to(s), the other two terms in the energy functional are somewhat ofopposite nature. Indeed, for a givenmass, half spacesminimize theGaussian perimeterwhile maximize the absolute value of the barycenter. The reason why proving that thehalf spaces Hν,s minimize the above functional leads to the quantitative inequality(5.48) can be understood on observing that b(Hν,s) = e− s22 ν/√2π . Therefore, fromthe minimality of Hν,s one has that if γ (E) = (s), thenPγ (E) − Pγ (Hν,s) ≥ ε2(|b(Hν,s)|2 − |b(E)|2) = ε2(|b(Hν,s)| + |b(E)|)βγ (E)≥ ε2√2πe−s22 βγ (E),from which is not difficult to deduce (5.48), see the argument at the beginning ofSection 4 of [10].The proof that half spaces are minimizers of the above problem is based on a PDErigidity argument that is certainly new in this context. The idea in [10] is to derive thesecond variation for the above functional and then to deduce the minimality of a halfspace by a powerful and elegant choice of the test function, see Step 3 of the proof ofthe main result contained in Sect. 4 of [10].We conclude this section by observing that while the result by Barchiesi et al.proves the quantitative Gaussian isoperimetric inequality with an optimal dependenceon the dimension and on the mass, the dependence on the dimension of the constantγ (n) which appears in (4.3) is not known. However, inequality (3.2) implies that fora nearly spherical set E sufficiently close to a ball one hasα(E)2 ≤ 8nD(E).This suggests that γ (n) should grow like n as n → ∞. Unfortunately, looking morecarefully at the proof of the quantitative isoperimetric inequality (4.3) given in Sect. 4one gets a constant growing exponentially fastwithn, while the contradiction argumentused in the Sect. 5.2 gives no clue about γ (n). However, in [67, (1.12)] it is observedthat γ (n) has at most a polynomial growth, though the power that one obtains withtheir methods, even if one may be possibly improve it a bit, seems still far away frombeing optimal. The value of the constant in dimension n = 2 for small values ofthe asymmetry was first given by Hall et al. in [83,84] in the convex case, and thenextended by Cicalese and Leonardi in [45] for general sets of finite perimeter. Theirresult states thatD(E) ≥ π8(4 − π)α(E)2 + o(α(E)2).Other, interesting estimates of the quantitative isoperimetric constant in dimension 2for special classes of convex sets, not necessarily with small asymmetry, are studiedin [6,46].123584 N. Fusco6 Other functional inequalities6.1 Stability of Sobolev inequalityIn this section we shall discuss the stability of Sobolev inequality, which states that ifn ≥ 2, 1 ≤ p < n and p∗ = np/(n − p), thenS(p, n)‖ f ‖L p∗ (Rn) ≤ ‖∇ f ‖L p(Rn) (6.1)for every function f in the homogeneous Sobolev space W 1,p(Rn). Throughout thissection by the homogeneous Sobolev space we mean the completion of C1c (Rn) withrespect to the L p(Rn) norm of the gradient.If p = 1 it is well known that (6.1) is equivalent to the isoperimetric inequality,see [59, p. 192], and thus S(1, n) = nω1/nn . Indeed, a simple approximation argumentshows that if f is a function in BV (Rn), the homogeneous space of all functions inLnn−1 (Rn) whose distributional gradient is a measure with finite total variation, thennω1/nn ‖ f ‖L nn−1 (Rn) ≤ |D f |(Rn),where |D f |(Rn) denotes the total variation, with equality holding if and only if f isthe characteristic function of a ball. If p > 1 the best constant in (6.1) was foundindependently by Aubin [9] and Talenti [113] and it turns out to be equal toS(p, n) = √π n1/p(n − pp − 1)(p−1)/p ((n/p)(1 + n − n/p)(1 + n/2)(n))1/n.In this case the family of extremals in (6.1) is given by the functions ga,b,x0 defined asga,b,x0(x) =a(1 + b|x − x0|p′)(n−p)/p for x ∈ Rn (6.2)for some a = 0, b > 0, x0 ∈ Rn , where p′ = p/(p − 1). When p = 2, as aconsequence of a celebrated result by Gidas et al. [80] applied to the Euler equationof the functional ‖∇ f ‖L2(Rn)/‖ f ‖L2 (Rn), one can show that the above functionsare the only ones for which equality holds in (6.1). However if p > 1, p = 2 thecharacterization of the functions of the form (6.2) as the only ones attaining equalityin (6.1) was shown much later by Cordero-Erausquin et al. [49] with a proof of theSobolev inequality (6.1) via mass transportation in the same spirit of Gromov’s proofof the isoperimetric inequality that we have seen in Sect. 5.1. We reproduce here theirproof.Proof of the sharp Sobolev inequality via mass transport Observe that in order toprove (6.1) it is enough to assume that f is nonnegative since |∇| f || = |∇ f |.Therefore, if f and g are two nonnegative functions from W 1,p(Rn) such that‖ f ‖p∗ = ‖g‖p∗ = 1, by applying Theorem 5.1 with μ = f p∗dx and ν = g p∗dy we123The quantitative isoperimetric inequality. . . 585find that there exists a convex function ϕ : Rn → R such that, setting T := ∇ϕ onehas ∫Rnh(y)g p∗(y) dy =∫Rnh(T (x)) f p∗(x) dx, (6.3)for all nonnegative Borel functions h. To avoid some technical details that could makethe idea of the proof less clear, let us assume that f and g have compact support andthat T is Lipschitz. Then from the above formula we deduce thatf p∗(x) = g p∗(T (x)) det∇T (x) for a.e. x ∈ Rn .From this inequality, using the geometric-arithmetic mean inequality, we have that fora.e. x in the support of fg−p∗n (T (x)) = f − p∗n (x)(det∇T (x))1/n ≤ 1nf −p∗n (x)Tr (∇T (x))= 1nf −p∗n (x) div T (x),where Tr (T ) denotes the trace of T . Multiplying both sides of this inequality byf p∗(x) and integrating by parts we get∫Rng−p∗n (T (x)) f p∗(x) dx ≤ 1n∫Rnf p∗(1−1/n)(x) div T (x) dx= − p∗(n − 1)n2∫Rnfp∗p′ (x)∇ f (x) · T (x) dx . (6.4)Using Hölder inequality and recalling (6.3) again we have∫Rnfp∗p′ (x)∇ f (x) · T (x) dx ≤ ‖∇ f ‖p(∫Rnf p∗(x)|T (x)|p′ dx)1/p′= ‖∇ f ‖p(∫Rng p∗(y)|y|p′ dy)1/p′.Combining this inequality with (6.4) and recalling (6.3) we then conclude that∫Rng p∗(1−1/n)(y) dy ≤ p(n − 1)n(n − p)‖∇ f ‖p(∫Rng p∗(y)|y|p′dy)1/p′. (6.5)Now we take b = 1 and a > 0 such that the function h p := ga,1,0 has L p∗ norm inRn equal to 1 and we argue as before, choosing f = g = h p. Since in this case themap T is just the identity map, the above argument simplifies a lot and one can easilycheck that all previous inequalities are indeed equalities. Therefore we conclude bygetting that123586 N. Fusco∫Rnh p∗(1−1/n)p (y) dy = p(n − 1)n(n − p)‖∇h p‖p(∫Rnh p∗p (y)|y|p′dy)1/p′.From this equality, recalling (6.5) one gets that if ‖ f ‖p∗ = ‖h p‖p∗ = 1, then‖∇ f ‖p ≥ ‖∇h p‖p.This inequality, as one can check with a few computations is precisely the Sobolevinequality (6.1) with the sharp constant S(n, p). In the argument above we have chosen to put ourself in the simple situation whenthe functions f and g have compact support and T is Lipschitz. With a bit of technicaladjustments, see [49, Sect. 4] the above argument can be justified also in the generalcase. Thus, if f and g are two functions for which inequality (6.5) holds as an equalityone can prove [49, Prop. 6], that there exist two positive constants α and λ suchthat f (x) = αg(λ(x − x0)) for some x0 ∈ Rn . From this the characterization of thefunctions in (6.2) as the only extremals in the Sobolev inequality immediately follows.In [27, (c) p. 75] Brezis and Lieb raised the following stability problem: ‘is therea way to bound ‖∇ f ‖2 − S(2, n)‖ f ‖2∗ from below in terms of a ‘distance’ of ffrom the set of optimal functions (6.2)?’ The answer to this question was provided afew years later in a nice paper by Bianchi and Egnell [16]. Given a function f in thehomogeneous Sobolev space W 1,2(Rn) they considered the distance μ from f to themanifold M made up by all the functions ga,b,x0 given by (6.2) with p = 2:μ( f ) = infa,b,x0‖∇ f − ∇ga,b,x0‖L2(Rn). (6.6)They proved the following result.Theorem 6.1 There exist a positive constant c(n) depending only on n such that forall f ∈ W 1,2(Rn)‖∇ f ‖22 − S(2, n)2‖ f ‖22∗ ≥ c(n)μ( f )2. (6.7)Note that in inequality (6.7) the power 2 is optimal, since it cannot be replaced,up to a rescaling, by any smaller power. The key point in the proof is a lemma, see[16, Lemma 1], where they show the inequality when μ( f ) is sufficiently small.The proof of this lemma is elegant and simple. Given a function f , they considerthe function ga,b,x0 minimizing μ( f ). Due to the Hilbert structure of W1,2(Rn), oneimmediately has that f −ga,b,x0 is orthogonal to the tangent space TMga,b,x0 . Then theproof consists in relating TMga,b,x0 to the first and second eigenspace of the operatorg2−2∗a,b,x0 on the weighted L2 space in Rn with weight g2−2∗a,b,x0.The situation is completely different when dealing with p = 2, where one cannottake advantage of the Hilbert structure of the space. Indeed, the results proved in thiscase in [43,75] are weaker and not completely satisfactory. Let us first describe thecontent of the last paper, where the case 1 < p < n is consider. The first big differencebetween the result proved by Bianchi and Egnell and the one proved in [43] is thatone has to replace the distance considered in (6.6) with a weaker kind of distance or123The quantitative isoperimetric inequality. . . 587asymmetry. Indeed, given a function f in the homogeneous Sobolev space W 1,p(Rn)one setsλ( f ) = infa,b,x0⎧⎨⎩‖ f − ga,b,x0‖p∗L p∗ (Rn)‖ f ‖p∗L p∗ (Rn): ‖ga,b,x0‖L p∗ (Rn) = ‖ f ‖L p∗ (Rn)⎫⎬⎭ . (6.8)A part from the rescaling factor at the denominator and the constraint ‖ga,b,x0‖p∗ =‖ f ‖p∗ , the new distance λ( f ) is clearly controlled from above by μ( f ) due to theSobolev inequality itself. Indeed, if one goes back to the asymmetry indices definedin Sects. 4 and 5, it is clear that λ( f ) and μ( f ) play the role of the indices α(E), see(4.1), and β(E), see (5.15), respectively. Moreover in the quantitative estimate provedin [43, Th. 1] the power is far from being optimal.Theorem 6.2 Let n ≥ 2 and let 1 < p < n. There exist two positive constants α andκ , depending only on p and n, such that for every f ∈ W 1,p(Rn)S(p, n)‖ f ‖L p∗ (Rn)(1 + κλ( f )α) ≤ ‖∇ f ‖L p(Rn). (6.9)The power α in (6.9) can be retrieved from [43, (2.54), Cor. 4, (4.1) and the proof ofTh. 1], but it is clearly non optimal. A comparison with the statement of Theorem 6.1above and the statement of Theorem 3.3 in [43] suggests that the optimal power couldbe α = max{2, p}, but we have no examples in this direction. Observe also that (6.9)can be rewritten in equivalent way asκ(n)λ( f )α ≤ ‖∇ f ‖L p(Rn)S(p, n)‖ f ‖L p∗ (Rn)− 1 := δ( f ). (6.10)We shall refer to the term δ( f ) at the right hand side of the previous equation as to theSobolev deficit of the function f . Note also that in order to have an inequality of thetype (6.10) both the asymmetry λ( f ) and the deficit δ( f ) must be rescaling invariant.However we observe that the constraint in definition of λ( f ) is not really needed.Indeed one could define a smaller asymmetry by settingλ̃( f ) := infa, b, x0‖ f − ga,b,x0‖p∗L p∗ (Rn)‖ f ‖p∗L p∗ (Rn).Observe that λ̃( f ) ≤ λ( f ) ≤ 2p∗ λ̃( f ), hence we could indifferently use one of thetwo asymmetries.The proof of the inequality follows a strategy very similar to the one used in Sect. 4aimed to reduce the general case in (6.9) to the case where f is a more and moresymmetric function. But before discussing that proof in more detail, let us recall thenotion of spherically symmetric decreasing rearrangement of a function f : Rn →[0,∞) such that |{x : f (x) > t}| < ∞ for all t > 0. If f has this property, for everyx ∈ Rn we set123588 N. Fuscof (x) =: sup{t ≥ 0 : |{y ∈ Rn : f (y) > t}| > ωn|x |n}.Note that f is a decreasing function depending on the norm of x and that from thedefinition it follows that |{ f > t}| = |{ f > t}|. Thus, in particular, one has that‖ f ‖L p(Rn) = ‖ f ‖L p(Rn) for all p ≥ 1. (6.11)Moreover, the Pólya–Szegö principle states that if f is a nonnegative function fromW 1,p(Rn), with p ≥ 1, then also f belongs to W 1,p(Rn) and moreover, see forinstance [28,87], ∫Rn|∇ f |p dx ≤∫Rn|∇ f |p dx . (6.12)The proof of this inequality is a simple application of the coarea formula for functions(2.8) and of the isoperimetric inequality.Proof of the Pólya–Szegö inequality (6.12) Let f be a nonnegative function from thehomogeneous Sobolev space W 1,p(Rn) and assume without proving it that also f isin W 1,p(Rn) (see [92, Th. 73]). We set for all t > 0μ(t) := |{x ∈ Rn : | f (x)| > t}|. (6.13)The function μ is called the distribution function of f . Since by construction the levelsets of f and f have the same measure, the two functions share the same distributionfunction μ. Note that μ is a decreasing function. Moreover, an easy application of thecoarea formula (2.8) shows that for a.e. t > 0− μ′(t) = Hn−1({ f = t})|∇ f ||{ f =t} , (6.14)and that− μ′(t) ≥∫{ f =t}dHn−1|∇ f | , (6.15)see for instance [41, Lemmas 2.4 and 2.6]. Note that in (6.14) we have implicitly usedthe fact that |∇ f | is constant on { f = t}. Then from coarea formula (2.8), (6.14),the fact that |∇ f | is constant on { f = t} and the isoperimetric inequality applied tothe sublevel sets of f , we have∫Rn|∇ f |p dx =∫ ∞0dt∫{ f =t}|∇ f |p−1 dHn−1 =∫ ∞0Hn−1({ f = t})p(∫{ f =t}dHn−1|∇ f |)p−1 dt=∫ ∞0Hn−1({ f = t})p(−μ′(t))p−1 dt ≤∫ ∞0Hn−1({ f = t})p(−μ′(t))p−1 dt.123The quantitative isoperimetric inequality. . . 589Fig. 10 An example showing the necessity of condition (6.16) to deduce that f is a translated of f From this inequality, using (6.15), Hölder inequality and coarea formula again we get∫Rn|∇ f |p dx ≤∫ ∞0Hn−1({ f = t})p(−μ′(t))p−1 dt ≤∫ ∞0Hn−1({ f = t})p(∫{ f =t}dHn−1|∇ f |)p−1 dt≤∫ ∞0dt∫{ f =t}|∇ f |p−1 dHn−1 =∫Rn|∇ f |p dx,thus concluding the proof. Note that from the above proof it is clear that when equality holds in (6.12) thenfor a.e. t > 0{ f > t} is equivalent to a ball and |∇ f | = |∇ f ||{ f =t} Hn−1-a.e. on { f = t}.However this information is not enough to conclude that f coincides with f ∗ up to atranslation, see Fig. 10. The equality cases in the Pólya–Szegö inequality have beenfully characterized first by Brothers and Ziemer in [28], see also [41,61]. Their resultreads as follows.Theorem 6.3 Let p > 1 and f ∈ W 1,p(Rn) a nonnegative function such that|{∇ f = 0} ∩ {0 < f < ess sup f }| = 0. (6.16)If equality holds in (6.12), then there exists x0 ∈ Rn such that f (x) = f (x − x0) forall x ∈ R.We observe that if |{∇ f = 0} ∩ {0 < f < ess sup f }| = 0, then (6.16) holds. Thereverse implication is not in general true.As we already observed, it may happen that ‖∇ f ‖p almost agrees with ‖∇ f ‖pwithout f being close to any translated of f . The presence of large sets where |∇ f |123590 N. Fuscois zero or very small is responsible of this phenomenon. Therefore the quantitativeversions of (6.12) available require either a control on the size of the set where |∇ f | issmall, see [40,42], or further assumptions on the class of functions under consideration,see [12, Th. 1.3].Going back to the Sobolev inequality, it is clear the role played by the Pólya–Szegö inequality (6.12). Indeed observe that we may always assume, without loss ofgenerality, that f ≥ 0. In fact, replacing f by | f | changes neither the L p norm of thefunction nor the L p norm of the gradient, since |∇| f |(x)| = |∇ f (x)| for a.e. x ∈ Rn .Therefore, if one wants to find the equality cases in the Sobolev inequality it is clearthat these will be realized by spherically symmetric and decreasing functions. ThusPólya–Szegö inequality plays in the proof of the quantitative Sobolev inequality givenin [43] the same role played by the perimeter inequality for Schwartz symmetrizationin the proof of (4.3) that we have seen in Sect. 4.Let us describe briefly how the proof of inequality (6.9) goes. The idea is to reduceto spherically symmetric decreasing functions exactly as in Sect. 4 we reduced theproof of the quantitative isoperimetric inequality to axially symmetric sets. Observethat if f ∈ W 1,p(Rn) is nonnegative, spherically symmetric and decreasing, then thereexists a decreasing function u ∈ W 1,ploc (R) such that f (x) = u(|x |). In this case theSobolev inequality (6.1) is equivalent to the one-dimensional Bliss inequality [17]S(p, n)(nωn∫ ∞0u(r)p∗rn−1 dr)1/p∗≤(nωn∫ ∞0(−u′(r))prn−1 dr)1/p(6.17)for every decreasing, locally absolutely continuous function u : [0,∞) → [0,∞).The extremals in (6.17) are the one dimensional profiles of the functions defined in(6.2), see for instance [17,49,113], that isva,b(r) = a(1 + br p′)(n−p)/p for r ≥ 0,for somea > 0,b > 0.Moreover, ifwe set for ameasurable functionw : (0,∞) → R,‖w‖p∗,rn−1 :=(∫ ∞0|w|p∗(r)rn−1dr) 1p∗,and for every nonnegative function u in (0,∞) we defineλ(u) = infa,b>0{‖u − va,b‖p∗p∗,rn−1‖u‖p∗p∗,rn−1: ‖u‖p∗,rn−1 = ‖va,b‖p∗,rn−1},Theorem 6.2 for spherically symmetric functions is equivalent to the following quan-titative Bliss inequality.123The quantitative isoperimetric inequality. . . 591Theorem 6.4 Let n ≥ 2 and let 1 < p < n. Then there exist constants β and κ ,depending only on p, n, such thatS(p, n)(nωn∫ ∞0u(r)p∗rn−1 dr)1/p∗ (1 + κλ(u)β)≤(nωn∫ ∞0(−u′(r))prn−1 dr)1/p(6.18)for every decreasing, locally absolutely continuous function u : [0,∞) → [0,∞).The proof of this theorem is a bit long and technically involved, but the underlyingidea is quite simple. First of all, by an approximation, rescaling and normalizationargument one may always assume that u is a nonnegative decreasing function withsupport in [0, 1] such thatnωn∫ ∞0u p∗(r)rn−1 dr = 1.Then, one tries to compare this function with the function v := va,1, where a is chosenin such a way that alsonωn∫ ∞0v p∗(r)rn−1 dr = 1.In order to compare the two functions, following the argument that we have seen in theproof of the Sobolev inequality given above, one defines a transport map T : [0, 1) →[0,∞) by setting∫ r0u(s)p∗sn−1ds =∫ T (r)0v(s)p∗sn−1ds.From this definition one gets immediately that T is a C1(0, 1) strictly increasingfunction such that T (0) = 0, limr→1−T (r) = ∞, andu(r)p∗ = v(T (r))p∗ T (r)n−1T ′(r)r1−n for all r ∈ (0, 1). (6.19)In particular, Eq. (6.19) implies that∫ 10h(T (r))u(r)p∗rn−1 dr =∫ ∞0h(r)v(r)p∗rn−1 dr,for every Borel function h : [0,∞) → [0,∞]. Therefore the function T can beregarded as a transport map carrying the measure u(r)p∗rn−1dr into v(r)p∗rn−1dr .Note that when T (r) = kr for some k > 0, one has u(r) = k(n−p)/pv(kr), thusproving that u is an extremal function in the Bliss inequality (6.17). Thus, the idea is123592 N. Fuscoto show that when the difference between the two sides of the Bliss inequality is small,then an interval [r1, r2] ⊂ [0, 1] can be chosen in such a way that T (r) is close tosome linear function kr for r ∈ [r1, r2], and simultaneously the integral of u(r)p∗rn−1outside [r1, r2] is small. These facts lead to conclude that u is close to k(n−p)/pv(kr)in in the quantitative way stated in (6.18).As observed in [43, (2.54)] one can in fact choose β in (6.18) to be the (non optimal)exponentβ = 3 + 4p − 3p + 1n. (6.20)The next step in the proof of Theorem 6.2 is to take advantage of the quantitative Blissinequality (6.18) to reduce the general case to the case of a spherically symmetricand decreasing function, via the Pólya–Szegö inequality (6.12). In fact, from thatinequality, recalling also (6.11), one has‖∇ f ‖L p(Rn) − ‖∇ f ‖L p(Rn) ≤ ‖∇ f ‖L p(Rn) − S(p, n)‖ f ‖L p∗ (Rn). (6.21)for every f ∈ W 1,p(Rn). The idea at this point is to estimate from below the differencebetween the L p norm of∇ f and∇ f in terms of the L p∗ distance of f from a suitabletranslated of f . Unfortunately, as we already observed, it is not possible in generalto control the distance between f and f in terms of the left hand side of (6.21),since if the measure of the set {∇ f = 0} is large the L p norms of ∇ f and ∇ f maybe equal, without f or f ∗ being close. However, if f belongs to a sufficiently rigidclass of functions, this estimate is still possible. Indeed, this is the case of n-symmetricfunctions, for which it is not too hard to show, see [43, Th. 3], the following stabilityresult for the Pólya-Szegö inequality.Theorem 6.5 Let n ≥ 2 and let 1 < p < n. Set q = max{p, 2}. Then a positiveconstant C exists such that∫Rn| f − f |p∗ ≤ C(∫Rn| f |p∗)p/n (∫Rn|∇ f |p)1/q ′ (∫Rn|∇ f |p −∫Rn|∇ f |p)1/qfor every nonnegative f ∈ W 1,p(Rn) which is symmetric with respect to the coordinatehyperplanes.The proof of this theorem consists in a careful revisitation of the proof of thePólya–Szegö inequality that we have given above, using the quantitative isoperimetricinequality (4.3) in place of the standard isoperimetric inequality and taking advantageof the fact that since f is n-symmetric then also the level sets { f > t} are n-symmetricand thus their asymmetryα({ f > t}) is equivalent to the symmetric differencewith theball centered at the origin with the same measure, see Lemma 4.4. As a consequenceof this theorem and of the quantitative Bliss inequality (6.18) one can easily deduce(6.9).Proposition 6.6 Let n ≥ 2 and let 1 < p < n. Then there exists a constant κ > 0such that (6.9) holds for every nonnegative n-symmetric function f ∈ W 1,p(Rn), withα = β, β as in (6.20).123The quantitative isoperimetric inequality. . . 593Proof As we already observed (6.9) is equivalent to (6.10). To prove this last one,since both δ( f ) and λ( f ) are invariant by rescaling and multiplication by a constant,we may assume, without loss of generality, that ‖ f ‖L p∗ (Rn) = 1. Observe also thatλ( f ) ≤ 2p∗ , hence wemay also assume that δ( f ) ≤ 1/S(p, n), since otherwise (6.10)follows with κ(n) = 1/(2αp∗ S(n, p)). Then, from (6.11) and (6.12) we haveS(p, n) ≤ ‖∇ f ‖p ≤ ‖∇ f ‖p ≤ 1 + S(p, n). (6.22)We haveλ( f ) ≤ 2p∗−1(λ( f ) + ‖ f − f ‖p∗p∗)≤ C[(‖∇ f ‖p − S(p, n))1/β + ‖∇ f ‖p/q ′p (‖∇ f ‖pp − ‖∇ f ‖pp)1/q],(6.23)for some constant C , where the first inequality is just a consequence of the triangleinequality, and the second one follows from Theorems 6.4 and 6.5. Inequalities (6.22)ensure that‖∇ f ‖pp − ‖∇ f ‖pp ≤ C(‖∇ f ‖p − ‖∇ f ‖p), (6.24)for some constant C . Combining (6.23), (6.24), and (6.21) yieldsλ( f ) ≤ C(δ( f )1/β + δ( f )1/q)for some constant C . Hence, inequality (6.9) follows with α = β, since by (6.20) onehas β > q. At this point the full proof of Theorem 6.2 follows from a reduction argument of thesame kind of the onewe stated in Theorem 4.6. However, the reduction to n-symmetricfunctions although related to a similar construction employed in Sect. 4, entails theovercoming of new serious obstacles in the present setting, mainly due to the nonlineargrowth of the functional ‖∇ f ‖pL p(Rn). The precise statement of the reduction theoremreads as follows, see [43, Th. 6].Theorem 6.7 Let n ≥ 2 and let 1 < p < n. There exists a positive constant C,depending only on n and p such that for every f ∈ W 1,p(Rn) there exists a nonnegativen-symmetric function f̃ with the property thatλ( f ) ≤ Cλ( f̃ ), δ( f̂ ) ≤ Cδ( f )1/β,where β is given by (6.20).We conclude this discussion on the quantitative Sobolev inequality for 1 < p < nby noting that, in view of the results of [16,75], the question arises of the optimalexponent α in equality (6.9). Furthermore, the result of [16] also leaves open theproblem of whether the distance of f from the family of extremals in L p∗(Rn) can bereplaced by the distance in W 1,p(Rn) in Theorem 6.2.123594 N. FuscoIn the special case p = 1 the Sobolev inequality (6.1) takes the formnω1/nn ‖ f ‖Ln′ (Rn) ≤∫Rn|∇ f | dxfor all f in the homogeneous Sobolev space W 1,1(Rn). Here n′ := n/(n − 1) standsfor the Sobolev exponent 1∗. By approximation it can be readily extended to thehomogenous space BV (Rn). For these functions it takes the formnω1/nn ‖ f ‖Ln′ (Rn) ≤ |D f |(Rn). (6.25)Aswe already observed, the above inequality is equivalent to the isoperimetric inequal-ity, Indeed if E is a set of finite measure, taking f = χ E in the inequality above onegets exactly the isoperimetric inequality. It is well-known that equality holds in (6.25)if and only if f = aχ Br (x0) for some a ∈ R, x0 ∈ Rn and r > 0. Therefore, inanalogy with (6.8) it is natural to introduce the asymmetry of a function f ∈ BV (Rn)by settingλ( f ) := inf⎧⎨⎩‖ f − aχ Br (x)‖n′Ln′‖ f ‖n′Ln′: |a|n′rnωn = ‖ f ‖n′Ln′ , a ∈ R, x ∈ Rn⎫⎬⎭ .Indeed, it can be proved that the above infimum is attained, see [75, Lemma B.1].In the case p = 1 the quantitative version of the Sobolev inequality (6.25) has beenobtained with the sharp exponent 2 [75, Th. 1.1].Theorem 6.8 There exists a dimensional constant C = C(n) such thatnω1/nn ‖ f ‖Ln′ (Rn)(1 + λ( f )2C(n))≤ |D f |(Rn) (6.26)for every f ∈ BV (Rn).The proof of this result follows the same pattern of the proof of the quantitativeisoperimetric inequality that we have seen in Sect. 4. A part from a series of technicaldifficulties due to the fact that one deals with functions instead of sets, the only pointwhere the proof really differs from the one we have already seen is when one has toshow inequality (6.26) for a spherically symmetric and decreasing function, see [75,Sect. 3].I would like also to point out that very recently Figalli et al. in [68] proved asimilar stability estimate for the anisotropic Sobolev (and log-Sobolev) inequality forfunctions of bounded variation. Also in this case the stability estimate is obtainedwith the sharp exponent 2, see [68, Th. 1.1]. The proof combines symmetrizationarguments with the Gromov’s idea for proving the isoperimetric inequality, and thusalso the Sobolev inequality for BV functions, with a mass transport argument.123The quantitative isoperimetric inequality. . . 595The Morrey–Sobolev imbedding theorem states that if p > n a function from thehomogeneous Sobolev space W 1,p(Rn) is essentially bounded and actually Höldercontinuous. In particular, if the support of f is bounded the following inequality holdsS(p, n)‖ f ‖L∞(Rn) ≤ |supp f |1n − 1p ‖∇ f ‖L p(Rn), (6.27)where supp f denotes the support of f and the sharp constant S(n, p) is given byS(p, n) = n1/pω1/nn(p − np − 1)1/p′.In this case the extremals are given by the family of functionsga,b,x0 :={a(bp−np−1 − |x − x0|p−np−1)if |x − x0| ≤ b,0 otherwise,for some a ∈ R, b ≥ 0 and x0 ∈ Rn . The stability for (6.27) has been proved byCianchi and stated in terms of the L∞ distance from the extremals. Precisely, see [39,Th. 1.1], we have the following quantitative estimate, where δ( f ) is defined as in(6.10) for p = ∞.Theorem 6.9 Let n ≥ 2 and p > n. There exist two positive constants α and C,depending only on p and n, such that for every f ∈ W 1,p(Rn)(infa,b,x0‖ f − ga,b,x0‖L∞(Rn)‖ f ‖L∞(Rn))α≤ Cδ( f ).6.2 The Faber–Krahn inequalityThe Faber–Krahn inequality goes back to Lord Rayleigh who in 1877 in his book‘The theory of sound’ [111] conjectured that the gravest principal tone of a vibratingmembrane is obtained by a circular one. In other words, if ⊂ Rn is an open set with| | = |Br | for some r > 0 thenλ( ) ≥ λ(Br ), (6.28)with the equality holding if and only if is a ball. Here by λ( ) we denote the firsteigenvalue of the Laplacian which is defined asλ( ) := min{∫ |∇ f |2 dx : ‖ f ‖L2( ) = 1, f ∈ H10 ( )}. (6.29)Note that a function f realizing the minimum on the right hand side is called aneigenfunction for the Laplacian and satisfies the equation− f = λ f in , f = 0 on ∂ . (6.30)123596 N. FuscoInequality (6.28) was proved in two dimensions by Faber [60] and extended for anyn ≥ 2 by Krahn [91]. Indeed, it can be proved as an immediate consequence of thePólya–Szegö inequality (6.12) by observing that if f is a minimizer in H10 ( ) of theabove variational problem thenλ( ) =∫ |∇ f |2 dx ≥∫Br|∇ f |2 dx ≥ λ(Br ). (6.31)The characterization of the equality cases in (6.28) can be also easily recovered usingTheorem 6.3. Observing that λ(r ) = r−2λ( ), inequality (6.28) may be rewrittenin the following scaling invariant form| | 2n λ( ) ≥ |B| 2n λ(B). (6.32)The stability of balls with respect to this inequality has been investigated by severalauthors. As for the quantitative isoperimetric inequality people tried first to understandthe convex case. In particular, it was proved by Melas in [99] that if ⊂ Rn is abounded, convex open set then one can find two balls B ′ ⊂ ⊂ B ′′ such thatmax{| \B ′|, |B ′′\ |}| |is controlled by a suitable power, depending on n, of the Faber–Krahn deficitδ( ) := | | 2n λ( ) − |B| 2n λ(B). (6.33)A similar result was later on obtained by Hansen and Nadirashivili in [85] where theinner and the outer radius of a convex set are used to control the distance from aball in terms of the Faber–Krahn deficit. For a general open set, it was conjectured byBhattacharya and Weitsman in [15] and by Nadirashvili in [102] that the followingquantitative Faber–Krahn inequality holdsα( )2 ≤ C(n)δ( ), (6.34)for any open set ⊂ Rn of finite measure, where α is the Fraenkel asymmetryindex defined in (4.1). An inequality of this kind, dealing with general open sets, wasobtained first in dimension two by Bhattacharya in [14] with the exponent 3 in place of2 and in any dimension n by Maggi et al. in [77] with the exponent 4. Indeed, in [77] amore general inequality was proved, related to the first eigenvalue of the p-Laplacian,p > 1 which is defined, similarly to (6.29), by setting for any open set with finitemeasureλp( ) := min{∫ |∇ f |p dx : ‖ f ‖L p( ) = 1, f ∈ W 1,p0 ( )}.Note that the same argument used to show Faber–Krahn inequality (6.32) implies thatfor any open set of finite measure one has123The quantitative isoperimetric inequality. . . 597| | pn λ( ) ≥ |B| pn λ(B).Then, the following quantitative estimate was proved in [77, Th. 1].Theorem 6.10 Let n ≥ 2 and p > 1. There exists a positive constant c(n, p) suchthat for every open set ⊂ Rn of finite measurec(n, p)α( )2+p ≤ | | pn λp( ) − |B| pn λp(B).The proof of the above inequalitymakes use of the following observation.Assumingfor the sake of simplicity that p = 2 and that has the same measure of the unit ballB, let f ∈ H10 ( ) be the function minimizing the right hand side of (6.29). Then,recalling (6.31), one has that∫ |∇ f |2 dx −∫Br|∇ f |2 dx ≤ λ( ) − λ(B).In turn, denoting as in the previous section by μ the distribution function of f definedas in (6.13) one has, see [77, (32)],c(n)∫ ∞0α({ f > t})2μ(t)2(n−1)/n|μ′(t)| dt ≤∫ |∇ f |2 dx −∫Br|∇ f |2 dx,for some positive constant c(n). Therefore, one may conclude thatc(n)∫ ∞0α({ f > t})2μ(t)2(n−1)/n|μ′(t)| dt ≤ λ( ) − λ(B).This inequality shows indeed that the Faber–Krahn deficit controls in an integral formthe asymmetry of the level sets of the eigenfunctions. However, to recover from thisinformation the sharp quantitative estimate of the asymmetry for does not seemeasy.The above conjecture on the optimal power in inequality (6.34) has been proved ina recent paper by Brasco et al. [24]. Here is their result.Theorem 6.11 There exists a positive constant C(n) such that (6.34) holds true forevery open set ⊂ Rn of finite measure.The proof is deep and technically involved. However, we shall try to explain themain ideas. The first key point of their strategy is to prove a sharp quantitative estimatefor a weaker inequality related to the torsional rigidity. To explain this inequality wehave to introduce, for an open set of finite measure, the following quantityE( ) := min{12∫ |∇ f |2 dx −∫ f dx : f ∈ H10 ( )}.123598 N. FuscoNote that the function f minimizing the integral functional on the right hand sidecoincides with the unique solution of the Dirichlet problem− f = 1 in , f = 0 on ∂ .Therefore, multiplying the above equation by f and integrating by parts one getsE( ) = −12∫ |∇ f |2 dx = −12∫ f dx .From this equality one easily gets a reformulation of E( ) asE( ) := −12max{(∫ f dx)2: ‖∇ f ‖L2( ) = 1}.The maximum of the integral on the right hand side of the previous formula is thetorsional rigidity of . Observe that E satisfies the scaling law: E(r ) = r−n−2E( )for all r > 0 and all open sets with finite measure. Then, the same argument usedto show inequality (6.28) immediately yields the following inequality| |− n+2n E( ) ≥ |B|− n+2n E(B). (6.35)with the equality holding if and only if is a ball. A deeper inequality, due to Kohler-Jobin, see [89,90], relates the first eigenvalue of the Laplacian to the torsional rigidity.Theorem 6.12 For every open set ⊂ Rn of finite measure we haveλ( )(−E( )) 2n+2 ≥ λ(B)(−E(B)) 2n+2 , (6.36)with the equality holding if and only if B is a ball.As we already said, the first key point is to observe that in view of (6.36) one candeduce (6.34) from a quantitative version of (6.35), see [24, Prop. 2.1].Proposition 6.13 Assume that there exists a constant c0(n) such that for every openset ⊂ Rn of finite measure| |− n+2n E( ) − |B|− n+2n E(B) ≥ c0(n)α( )2. (6.37)Then, there exists another constant c1(n) such that one has also| | 2n λ( ) − |B| 2n λ(B) ≥ c1(n)α( )2.Proof Assume | | = |B| and observe that from the Kohler-Jobin inequality (6.36)we haveλ( )λ(B)− 1 ≥(E(B)E( )) 2n+2 − 1. (6.38)123The quantitative isoperimetric inequality. . . 599Since E( ) is a negative quantity, from (6.35) we have that E(B)/E( ) ≥ 1. There-fore, if the ratio E(B)/E( ) ≥ 2, from (6.38) we getλ( )λ(B)− 1 ≥ 2 2n+2 − 1 ≥ 22n+2 − 14α( )2.On the other hand, if 1 ≤ E(B)/E( ) ≤ 2, from (6.38) and the assumption (6.37)we haveλ( )λ(B)− 1 ≥ c(n)(E(B)E( )− 1)≥ c(n)c0|E(B)|α( )2,where c(n) is such that t2n+2 − 1 ≥ c(n)(t − 1) for all t ∈ [1, 2]. With this proposition in hands it is now clear that the strategy followed in [24] is toprove a stability inequality for E( ) of the type (6.37). Namely they follow the patternthat we have discussed in Sect. 5.2 of proving first the stability estimate for nearlyspherical sets and then to extend it to general open sets by a contradiction argumentvia regularity. The proof for nearly spherical sets is essentially a second variationargument and it leads to the following Fuglede type result, see [24, Th. 3.3].Theorem 6.14 Let σ ∈ (0, 1]. There exists a positive constant δ, depending only onn and γ , such that if is a nearly spherical set as in (3.1), | | = |B|, the barycenterof is at the origin and ‖u‖C2,σ (Sn−1) ≤ δ, thenE( ) − E(B) ≥ 132n2‖u‖H1/2(Sn−1). (6.39)As for Fuglede’s estimate (3.2) this inequality immediately implies (6.37) for nearlyspherical sets. Note however that the perimeter deficit controls the H1 norm of u, whilethe difference E( ) − E(B), hence the Faber–Krahn deficit (6.33), only controls theweaker H1/2 norm of u. This is not surprising if one thinks that one may easily findopen sets h for which the perimeter P( h) goes to ∞ while the first eigenvalue ofthe Laplacian λ( h) remains bounded. Indeed it is enough to take h := {y ∈ Rn : y = t x(1 + uh(x)), with x ∈ Sn−1, 0 ≤ t < 1},with uh : Sn−1 → (−1, 1) chosen so that ‖uh‖H1/2(Sn−1) is bounded and‖uh‖H1(Sn−1) → ∞.Coming to the proof of the stability of the functional E , after proving Theorem 6.14one would like to argue as we did in Sect. 5.2. So let us try to imitate the contradictionargument used in Sect. 5.2 by assuming that there exists a sequence j of open setswith | j | = |B|, E( j ) → E(B) andE( j ) − E(B) ≤ C0(n)α( j )2, (6.40)123600 N. Fuscofor some big constant C0 to be chosen later. Then, one would like to replace the sets j with a sequence of sets, say U j , minimizing a functional of the type we introducedin (5.12)E( ) + |α( ) − α( j )| + ∣∣| | − |B|∣∣,to show that they still satisfy a ‘wrong’ inequality as (6.40) and that they converge inC2,σ to the a ball, so to get a contradiction to (6.39). This is the point where the proof in[24] becomes really complicate and requires some new ideas and deep arguments. Tounderstand why observe that minimizing the functional above is equivalent to findinga minimizer f j of the functional12∫ |∇ f |2 dx −∫ f dx + |α({ f > 0}) − α( j )| + ∣∣|{ f > 0}| − |B|∣∣ (6.41)among all functions with a support of finite measure. It turns out that to get the desiredregularity one has to modify in a non trivial way the theory developed by Alt andCaffarelli in [5]. But even this is not enough. In fact the functional in (6.41) is notsufficiently smooth to ensure that the support of the minimizers f j are of class C2,σ .Well, one could replace it with an essentially equivalent and smoother one, such as12∫ |∇ f |2 dx −∫ f dx +√α( j )2 + (α({ f > 0}) − α( j ))2 + |{ f > 0}|,but even this functional would notwork. And the reason is that the Fraenkel asymmetryis not smooth enough. So, another delicate point in [24] is the replacement of theasymmetry index with a new distance which on one side dominates the square of theFraenkel asymmetry, on the other side is much smoother since it behaves like thesquare of an L2 norm.6.3 Further resultsIn the previous sections we have seen the natural extension of the ideas developed forthe study of the stability of the quantitative isoperimetric inequality to two importantinequalities. Indeed, several other inequalities have been investigated in the last yearsin this direction. For instance, in [77] the same argument used to prove Theorem 6.10is also used to derive a quantitative inequality for the isocapacitary inequality statingthat for any open set of finite measure one has| | p−nn Capp( ) ≥ |B|p−nn Capp(B),where the p capacity of , for p ∈ (1, n), n ≥ 2 is defined asCapp( ) := inf{∫Rn|∇ f |p dx : f ≥ χ , f ∈ L p∗(Rn), |∇ f | ∈ L p(Rn)}.123The quantitative isoperimetric inequality. . . 601Quantitative inequalities for the so called Cheeger constant have been also obtainedin [77] as a limit case of Theorem 6.10 and in [65] with the optimal exponent. To thisaim we recall that if is an open set with finite measure, n ≥ 2, andm >1n′, where n′ = nn − 1 ,then the m-Cheeger constant of is defined by settingcm( ) := inf{P(E)|E |m : E ⊂ , |E | > 0}.When m = 1 a set E minimizing the right hand side is called a Cheger set of . TheCheeger inequality states that balls minimize the Cheeger constant among all opensets with the same measure. Thus, taking into account the right scaling factor, we havethat| |m− 1n′ cm( ) ≥ |B|m−1n′ cm(B),with the equality holding if and only if is a ball. Using the quantitative isoperimetricinequality it has been shown by Figalli et al. in [65] that the following stability estimateholds. Givenm > 1/n′, there exist a positive constant κ(m, n) such that for every openset of finite measure one hasκα( )2 ≤ | |m− 1n′ cm( ) − |B|m−1n′ cm(B).The results discussed in the previous section have been extended byBrasco and Pratelliin [25] to the second eigenvalue of the Laplacian. To this aim, we recall that if is anopen set with finite measure the second eigenvalue λ2( ) is defined byminimizing theDirichlet integral among all functions that are orthogonal to the first eigenfunctions.Precisely, denoting by f a non trivial eigenfunction, that is a nontrivial solution toEq. (6.30). We haveλ2( ) := min{∫ |∇ f |2 dx : ‖ f ‖L2( ) = 1, f ∈ H10 ( ),∫ f f dx = 0}.Then Krahn–Szegö inequality states that among all open sets of given measure theunique minimizer of λ2 is given by the union of two disjoint balls of equal measure.In other words, setting := B ′ ∪ B ′′ where the balls B ′ and B ′′ are disjoint and|B ′| = |B ′′| = |B|/2 we have, taking into account the rescaling law,| | 2n λ2( ) ≥ |B| 2n λ2(),with the equality holding if and only if is the union of two disjoint balls of equalmeasure. As a consequence of Theorem 6.11 and of [25, Th. 3.5 and Rem. 3.6] onehas123602 N. FuscoTheorem 6.15 There exists a constant κ(n) depending only on the dimension suchthat for any open set of finite measure one has| | 2n λ2( ) − |B| 2n λ2() ≥ κ(n)̃α( )n+1,whereα̃( ) := inf{ | ∩ (B ′r (x ′) ∪ B ′′r (x ′))|| | , B′r (x′) ∩ B ′′r (x ′) = ∅, |Br | =12| |}.It is not knownwhether the power n+1 in the statement of Theorem 6.15 is optimal,though it is clear that in this case the optimal power must depend on the dimension.Things are better understood with the second, or the first non trivial, eigenvalue ofthe Laplacian under Neumann boundary conditions μ2( )which is defined by settingfor every open set of finite measureμ2( ) := min{∫ |∇ f |2 dx : ‖ f ‖L2( ) = 1, f ∈ H10 ( ),∫ f dx = 0}.In this case the Szegö–Weinberger inequality states that among all open sets of givenmeasure the unique maximizer of λ2 is given by a ball. This inequality can be writtenin a scaling invariant form as|B| 2n μ2(B) ≥ | | 2n μ2( ).It turns out that a quantitative version of this inequality can be proved as well, see [25,Th. 4.1].Theorem 6.16 There exists a constant κ(n) depending only on the dimension suchthat for every open set with Lipschitz boundary one hasκ(n)α( )2 ≤ |B| 2n μ2(B) − | | 2n μ2( ).It is interesting to observe that in the inequality above the exponent 2 is sharp.The proof of this inequality uses a simple geometric argument. However proving thesharpness of the exponent 2 is harder, see [25, Sect. 6].We conclude this section by quickly quoting a few other inequalities for whichstability results have been obtained recently.We recall that the Brunn–Minkowski inequality (1.2) states that if H, K ⊂ Rn arecompact sets then|H + K | 1n ≥ |H | 1n + |K | 1n ,with the equality holding if and only if: |H + K | = 0, either H or K consists ofa single point, H and K are two convex bodies homothetic to each other, i.e., thereexist x0 ∈ Rn and λ > 0 such that H = x0 + λK , see [29, Th. 8.1.1]. A quantitative123The quantitative isoperimetric inequality. . . 603version of the inequality above, in the case when H and K are convex bodies, has beenobtained in [66] and with an improved constant in [109]. See also a recent paper byEldan and Klartag [57]. The general case of measurable sets is still open apart from avery recent result by Figalli and Jerison [64].The isodiametric inequality states that among all sets of given diameter the ball isthe one enclosing the biggest volume. Equivalently, if E is a measurable set of finitemeasure in Rn then|E | ≤ ωn(diam (E)2)n,with the equality holding if and only if E is a ball. The stability of this inequality hasbeen studied by Maggi et al. in [96] where the following result is proved.Theorem 6.17 If E ⊂ Rn is a set with diam (E) = 2, there exists a ball B(x) suchthatκ(n)|EB(x)|2 ≤ |B| − |E ||E | ,for some positive constant κ depending only on the dimension n.Paper [63] is devoted to the stability of the relative isoperimetric inequality ina cone while [11] contains a characterization of the equality cases in the perimeterinequality for Steiner symmetrization in any codimension, as well as a stability resultfor this inequality for a class of convex sets. A similar stability result, concerningconcave and log-concave functions has been proved in [12]. Finally, a quantitativeversion of the Gagliardo–Sobolev–Nirenberg inequality, has been recently proved byCarlen and Figalli in [35], starting from the Bianchi and Egnell stability result that wehave discussed in Sect. 6.1.Acknowledgments Part of this work has been carried on at the University of Jyväskylä and supported bythe FiDiPro project 2100002028. The friendly atmosphere of the Mathematics and Statistics Department ofJyväskylä is warmly acknowledged. I would also thankMarco Barchiesi, Laura Bufford andMatteo Rinaldiwho did some of the pictures and Marco Barchiesi for carefully reading the manuscript.Open Access This article is distributed under the terms of the Creative Commons Attribution 4.0 Interna-tional License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution,and reproduction in any medium, provided you give appropriate credit to the original author(s) and thesource, provide a link to the Creative Commons license, and indicate if changes were made.References1. Acerbi, E., Fusco, N., Morini, M.: Minimality via second variation for a nonlocal isoperimetricproblem. Commun. Math. Phys. 322, 515–557 (2013)2. Aleksandrov, A.D.: Uniqueness theorems for surfaces in the large. V (Russian). Vestnik Leningr.Univ. 13, 5–8 (1958)3. Almgren, F.: Optimal isoperimetric inequalities. Indiana Univ. Math. J. 35, 451–547 (1986)4. Almgren, F.J., Lieb, E.H.: Symmetric decreasing rearrangement is sometimes continuous. J. Am.Math. Soc. 2, 683–773 (1989)123604 N. Fusco5. Alt, H.W., Caffarelli, L.A.: Existence and regularity for a minimum problem with free boundary. J.Reine Angew. Math. 325, 105–144 (1981)6. Alvino, A., Ferone, V., Nitsch, C.: A sharp isoperimetric inequality in the plane. J. Eur. Math. Soc.13, 185–206 (2011)7. Ambrosio, L., De Philippis, G., Martinazzi, L.: Gamma-convergence of nonlocal perimeter function-als. Manuscr. Math. 134, 377–403 (2011)8. Ambrosio, L., Fusco, N., Pallara, D.: Functions of Bounded Variation and Free Discontinuity Prob-lems. Oxford Mathematical Monographs. The Clarendon Press, New York (2000)9. Aubin, T.: Problèmes isopérimétriques et espaces de Sobolev. J. Differ. Geom. 11, 573–598 (1976)10. Barchiesi, M., Brancolini, A., Julin, V.: Sharp dimension free quantitative estimates for the Gaussianisoperimetric inequality. Ann. Probab. arXiv:1409.2106v1 (2015, to appear)11. Barchiesi, M., Cagnetti, F., Fusco, N.: Stability of the Steiner symmetrization of convex sets. J. Eur.Math. Soc. 15, 1245–1278 (2013)12. Barchiesi, M., Capriani, G.M., Fusco, N., Pisante, G.: Stability of Pólya–Szegö inequality for log-concave functions. J. Funct. Anal. 267, 2264–2297 (2014)13. Bernstein, F.: Über die isoperimetriche Eigenschaft des Kreises auf der Kugeloberflache und in derEbene. Math. Ann. 60, 117–136 (1905)14. Bhattacharya, T.: ome observations on the first eigenvalue of the p-Laplacian and its connectionswith asymmetry. Electron. J. Differ. Equ. 35, 15 (2001)15. Bhattacharya, T.,Weitsman,A.: Estimates forGreens function in terms of asymmetry. AMSContemp.Math. Ser. 221, 31–58 (1999)16. Bianchi, G., Egnell, H.: A note on the Sobolev inequality. J. Funct. Anal. 100, 18–24 (1991)17. Bliss, G.A.: An integral inequality. J. Lond. Math. Soc. 5, 40–46 (1930)18. Bögelein, V., Duzaar, F., Fusco, N.: A sharp quantitative isoperimetric inequality in higher codimen-sion. Rend. Lincei Mat. Appl. 26, 309–362 (2015)19. Bögelein, V., Duzaar, F., Fusco, N.: A quantitative isoperimetric inequality on the sphere. http://cvgmt.sns.it/paper/2146/ (2013)20. Bombieri, E.: Regularity theory for almost minimal currents. Arch. Ration. Mech. Anal. 78, 99–130(1982)21. Bonnesen, T.: Über die isoperimetrische Defizit ebener Figuren. Math. Ann. 91, 252–268 (1924)22. Borell, C.: The Brunn–Minkowski inequality in Gauss space. Invent. Math. 30, 207–216 (1975)23. Bourgain, J., Brezis, H., Mironescu, P.: Another look at Sobolev spaces. In: Menaldi, J.L., Rofman,E., Sulem, A. (eds.) Optimal Control and Partial Differential Equations, pp. 429–455. IOS Press,Amsterdam (2001)24. Brasco, L., De Philippis, G., Velichkov, B.: Faber–Krahn inequalities in sharp quantitative form. DukeMath. J. 164, 1777–1831 (2015)25. Brasco, L., Pratelli, A.: Sharp stability of some spectral inequalities. Geom. Funct. Anal. 22, 107–135(2012)26. Brenier, Y.: Polar factorization and monotone rearrangement of vector-valued functions. Commun.Pure Appl. Math. 44, 375–417 (1991)27. Brezis, H., Lieb, E.H.: Sobolev inequalities with remainder terms. J. Funct. Anal. 62, 73–86 (1985)28. Brothers, J., Ziemer, W.: Minimal rearrangements of Sobolev functions. J. Reine. Angew. Math. 384,153–179 (1988)29. Burago, Y.D., Zalgaller, V.A.: Geometric Inequalities. Springer, Berlin (1988)30. Cabré,X.: Partial differential equations, geometry and stochastic control (Catalan).Butl. Soc.CatalanaMat. 15, 7–27 (2000)31. Caffarelli, L., Roquejoffre, J.M., Savin, O.: Non-local minimal surfaces. Commun. Pure Appl. Math.63, 1111–1144 (2010)32. Caffarelli, L., Valdinoci, E.: Uniform estimates and limiting arguments for nonlocal minimal surfaces.Calc. Var. Partial Differ. Equ. 41, 203–240 (2011)33. Cagnetti, F., Colombo,M.,DePhilippis,G.,Maggi, F.: Rigidity of equality cases in Steiner’s perimeterinequality. Anal. PDE 7, 1535–1593 (2014)34. Caputo, M.C., Guillen, N.: Regularity for non-local almost minimal boundaries and applications.arXiv:1003.2470 (2010)35. Carlen, E.A., Figalli, A.: Stability for a GNS inequality and the log-HLS inequality, with applicationto the critical mass Keller–Segel equation. Duke Math. J. 162, 579–625 (2013)123The quantitative isoperimetric inequality. . . 60536. Carlen, E.A., Kerce, C.: On the cases of equality in Bobkov’s inequality and Gaussian rearrangement.Calc. Var. Partial Differ. Equ. 13, 1–18 (2001)37. Chavel, I.: Isoperimetric Inequalities. Cambridge Tracts in Mathematics, vol. 145. Cambridge Uni-versity Press, Cambridge (2001)38. Chlebík, M., Cianchi, A., Fusco, N.: The perimeter inequality under Steiner symmetrization: casesof equality. Ann. Math. 162, 525–555 (2005)39. Cianchi, A.: Sharp Morrey–Sobolev inequalities and the distance from extremals. Trans. Am. Math.Soc. 360, 4335–4347 (2008)40. Cianchi, A., Esposito, L., Fusco, N., Trombetti, C.: A quantitative Pólya–Szegö principle. J. ReineAngew. Math. 614, 153–189 (2008)41. Cianchi, A., Fusco, N.: Functions of bounded variation and rearrangements. Arch. Ration. Mech.Anal. 165, 1–40 (2002)42. Cianchi, A., Fusco, N.: Dirichlet integrals and Steiner asymmetry. Bull. Sci. Math. 130, 675–696(2006)43. Cianchi, A., Fusco, N., Maggi, F., Pratelli, A.: The sharp Sobolev inequality in quantitative form. J.Eur. Math. Soc. 11, 1105–1139 (2009)44. Cianchi, A., Fusco, N., Maggi, F., Pratelli, A.: On the isoperimetric deficit in the Gauss space. Am.J. Math. 133, 131–186 (2011)45. Cicalese, M., Leonardi, G.P.: A selection principle for the sharp quantitative isoperimetric inequality.Arch. Ration. Mech. Anal. 206, 617–643 (2012)46. Cicalese, M., Leonardi, G.P.: Best constants for the isoperimetric inequality in quantitative form. J.Eur. Math. Soc. 15, 1101–1129 (2013)47. Cicalese, M., Leonardi, G.P., Maggi, F.: Sharp stability inequalities for planar double bubbles.arXiv:1211.3698v2 (2012)48. Cicalese, M., Leonardi, G.P., Maggi, F.: Improved convergence theorems for bubble clusters. I. Theplanar case. arXiv:1409.6652v2 (2014)49. Cordero-Erausquin, D., Nazaret, B., Villani, C.: A mass-transportation approach to sharp Sobolevand Gagliardo–Nirenberg inequalities. Adv. Math. 182, 307–332 (2004)50. Dacorogna, B., Pfister, C.E.: Wulff theorem and best constant in Sobolev inequality. J. Math. PuresAppl. 71, 97–118 (1992)51. David, G., Semmes, S.: Quasiminimal surfaces of codimension 1 and John domains. Pac. J. Math.183, 213–277 (1998)52. Dávila, J.: On an open question about functions of bounded variation. Calc. Var. Partial Differ. Equ.15, 519–527 (2002)53. De Giorgi, E.: Sulla proprietà isoperimetrica dell’ipersfera, nella classe degli insiemi aventi frontieraorientata di misura finita (Italian). Atti Accad. Naz. Lincei. Mem. Cl. Sci. Fis. Mat. Nat. Sez. I (8) 5,33–44 (1958)54. Duzaar, F., Steffen, K.: Optimal interior and boundary regularity for almost minimizers to ellipticvariational integrals. J. Reine Angew. Math. 546, 73–138 (2002)55. Ehrhard, A.: Symétrisation dans l’espace de Gauss. Math. Scand. 53, 281–301 (1983)56. Eldan, R.: A two-sided estimate for the Gaussian noise stability deficit. Invent. Math. 201, 561–624(2015)57. Eldan, R., Klartag, B.: Dimensionality and stability. Ann. Sc. Norm. Super. Pisa Cl. Sci. 13, 975–1007(2014)58. Esposito, L., Fusco, N., Trombetti, C.: A quantitative version of the isoperimetric inequality: theanisotropic case. Ann. Sc. Norm. Super. Pisa Cl. Sci. 4, 619–651 (2005)59. Evans, L.C., Gariepy, R.F.: Measure Theory and Fine Properties of Functions. CRC Press Inc, BocaRaton (1992)60. Faber, C.: Beweis dass unter allen homogen Membranen von gleicher Fläche und gleicher Spannungdie kreisförmige den tiefsten Grundton gibt, Sitzungsber, pp. 169–172 . Bayer. Akad. der Wiss.Math.-Phys., Munich (1923)61. Ferone, A., Volpicelli, R.: Minimal rearrangements of Sobolev functions: a new proof. Ann. Inst. H.Poincaré Anal. Non Linéaire 20, 333–339 (2003)62. Figalli, A., Fusco, N., Maggi, F., Millot, V., Morini, M.: Isoperimetry and stability properties of ballswith respect to nonlocal energies. Commun. Math. Phys. 336, 441–507 (2015)63. Figalli, A., Indrei, E.: A sharp stability result for the relative isoperimetric inequality inside convexcones. J. Geom. Anal. 23, 938–969 (2013)123606 N. Fusco64. Figalli, A., Jerison, D.: Quantitative stability for sumsets in Rn . J. Eur. Math. Soc. 17, 1079–1106(2015)65. Figalli, A., Maggi, F., Pratelli, A.: A note on Cheeger sets. Proc. Am. Math. Soc. 137, 2057–2062(2009a)66. Figalli, A., Maggi, F., Pratelli, A.: A refined Brunn–Minkowski inequality for convex sets. Ann. Inst.H. Poincaré Anal. Non Linéaire 26, 2511–2519 (2009b)67. Figalli, A., Maggi, F., Pratelli, A.: A mass transportation approach to quantitative isoperimetricinequalities. Invent. Math. 182, 167–211 (2010)68. Figalli, A., Maggi, F., Pratelli, A.: Sharp stability theorems for the anisotropic Sobolev and log-Sobolev inequalities on functions of bounded variation. Adv. Math. 242, 80–101 (2013)69. Fonseca, I.: The Wulff theorem revisited. Proc. R. Soc. Lond. 432, 125–145 (1991)70. Fonseca, I.,Müller, S.: A uniqueness proof for theWulff theorem. Proc. R. Soc. Edinb. 119A, 125–136(1991)71. Frank, R.L., Seiringer, R.: Non-linear ground state representations and sharp Hardy inequalities. J.Funct. Anal. 255, 3407–3430 (2008)72. Fuglede, B.: Stability in the isoperimetric problem for convex or nearly spherical domains in Rn .Trans. Am. Math. Soc. 314, 619–638 (1989)73. Fusco, N., Gelli, M.S., Pisante, G.: On a Bonnesen type inequality involving the spherical deviation.J. Math. Pures Appl. 98, 616–632 (2012)74. Fusco, N., Julin, V.: A strong form of the quantitative isoperimetric inequality. Calc. Var. PartialDiffer. Equ. 50, 925–937 (2014)75. Fusco, N., Maggi, F., Pratelli, A.: The sharp quantitative Sobolev inequality for functions of boundedvariation. J. Funct. Anal. 244, 315–341 (2007)76. Fusco, N., Maggi, F., Pratelli, A.: The sharp quantitative isoperimetric inequality. Ann. Math. 168,941–980 (2008)77. Fusco, N., Maggi, F., Pratelli, A.: Stability estimates for certain Faber–Krahn, isocapacitary andCheeger inequalities. Ann. Sc. Norm. Super. Pisa Cl. Sci. (5) 8, 51–71 (2009)78. Fusco, N., Millot, V., Morini, M.: A quantitative isoperimetric inequality for fractional perimeters. J.Funct. Anal. 261, 697–715 (2011)79. Giaquinta, M., Giusti, E.: Quasi-minima. Ann. Inst. H. Poincaré Anal. Non Linéaire 1, 79–107 (1984)80. Gidas, B., Ni, W.M., Nirenberg, L.: Symmetry and related properties via the maximum principle.Commun. Math. Phys. 68, 209–243 (1979)81. Grisvard, P.: Elliptic Problems in Non Smooth Domains. Monongraphs and Studies in Mathematics.Pitman, Boston (1985)82. Hall, R.R.: A quantitative isoperimetric inequality in n-dimensional space. J. Reine Angew. Math.428, 161–176 (1992)83. Hall, R.R., Hayman, W.K.: A problem in the theory of subordination. J. Anal. Math. 60, 99–111(1993)84. Hall, R.R., Hayman, W.K., Weitsman, A.W.: On asymmetry and capacity. J. Anal. Math. 56, 87–123(1991)85. Hansen, W., Nadirashvili, N.: Isoperimetric inequalities in potential theory. In: Proceedings from theInternational Conference on Potential Theory (Amersfoort, 1991). Potential Analysis, vol. 3, pp. 1–14(1994)86. Hurwitz, A.: Sur le problème des isopérimètres. C. R. Acad. Sci. Paris 132, 401–403 (1901)87. Kawohl, B.: Rearrangements and Convexity of Level Sets in PDE. Lecture Notes in Mathematics,vol. 1150. Springer, Berlin (1985)88. Kinnunen, J., Korte, R., Lorent, A., Shanmugalingam, N.: Regularity of sets with quasiminimalboundary surfaces in metric spaces. J. Geom. Anal. 23, 1607–1640 (2013)89. Kohler-Jobin, M.T.: Une méthode de comparaison isopérimétrique de fonctionnelles de domaines dela physique mathématique. I. Une démonstration de la conjecture isopérimétrique Pλ2 ≥ π j40 /2 dePolýa et Szegö. Z. Angew. Math. Phys. 29, 757–766 (1978)90. Kohler-Jobin,M.T.: Symmetrization with equal Dirichlet integrals. SIAM J.Math. Anal. 13, 153–161(1982)91. Krahn, E.: Über Minimaleigenschaft des Kugel in drei und mehr Dimensionen. Acta Comment. Univ.Tartu Math. A9, 1–44 (1926)92. Leoni, G.: A First Course in Sobolev Spaces. Graduate Studies in Mathematics. American Mathe-matical Society, Providence (2009)123The quantitative isoperimetric inequality. . . 60793. McCann, R.J.: Existence and uniqueness of monotone measure-preserving maps. Duke Math. J. 80,309–323 (1995)94. Maggi, F.: Some methods for studying stability in isoperimetric type problems. Bull. Am. Math. Soc.45, 367–408 (2008)95. Maggi, F.: Sets of Finite Perimeter and Geometric Variational Problems. An Introduction to Geo-metric Measure Theory. Cambridge Studies in Advanced Mathematics. Cambridge University Press,Cambridge (2012)96. Maggi, F., Ponsiglione, M., Pratelli, A.: Quantitative stability in the isodiametric inequality via theisoperimetric inequality. Trans. Am. Math. Soc. 366, 1141–1160 (2014)97. Maggi, F., Villani, C.: Balls have the worst best Sobolev inequalities. J. Geom. Anal. 15, 83–121(2005)98. Maz’ya, V., Shaposhnikova, T.: On the Bourgain, Brezis, andMironescu theorem concerning limitingembeddings of fractional Sobolev spaces. J. Funct. Anal. 195, 230–238 (2002) (Erratum, J. Funct.Anal. 201, 298–300 (2003))99. Melas, A.: The stability of some eigenvalue estimates. J. Differ. Geom. 36, 19–33 (1992)100. Milman, V.D., Schechtman, G.: Asymptotic Theory of Finite-Dimensional Normed Spaces. With anAppendix by M. Gromov. Lecture Notes in Mathematics, vol. 1200. Springer, Berlin (1986)101. Mossel, E., Neeman, J.: Robust dimension free isoperimetry in Gaussian space. Ann. Probab. 43,971–991 (2015)102. Nadirashvili, N.: Conformal maps and isoperimetric inequalities for eigenvalues of the Neumannproblem. In: Proceedings of the Ashkelon Workshop on Complex Function Theory, pp. 197–201(1996), Israel Mathematical Conference Proceedings, vol. 11. Bar-Ilan University, Ramat Gan (1997)103. Neumayer, R.: A strong form of the quantitative Wulff inequality. arXiv:1503.06705 (2015)104. Osserman, R.: The isoperimetric inequality. Bull. Am. Math. Soc. 84, 1182–1238 (1978)105. Osserman, R.: Bonnesen-style isoperimetric inequalities. Am. Math. Mon. 86, 1–29 (1979)106. Rajala, K., Zhong, X.: Bonnesen’s inequality for John domains inRn . J. Funct. Anal. 263, 3617–3640(2012)107. Runst, T., Sickel,W.: Sobolev Spaces of Fractional Order, Nemytskij Operators, andNonlinear PartialDifferential Equations. de Gruyter Series in Nonlinear Analysis and Applications. Walter de Gruyter& Co., Berlin (1996)108. Schmidt, E.: Beweis der isoperimetrischen Eigenschaft der Kugel im hyperbolischen und sphärischenRaum jeder Dimensionszahl. Math. Z. 49, 1–109 (1943)109. Segal, A.: Remark on stability of Brunn–Minkowski and isoperimetric inequalities for convex bodies.In: Geometric Aspects of Functional Analysis. Lecture Notes inMathematics, vol. 2050, pp. 381-391.Springer, Heidelberg (2012)110. Simon, L.: Lectures on geometric measure theory. In: Proceedings of the Centre for MathematicalAnalysis, Australian National University, vol. 3. Australian National University, Centre for Mathe-matical Analysis, Canberra (1983)111. Strutt, J.W. (Lord Rayleigh): The Theory of Sound. 2nd edn. Dover Publications, New York (1945)112. Sudakov, V.N., Cirel’son, B.S.: Extremal properties of half-spaces for spherically invariant measures(Russian). In: Problems in the Theory of Probability Distributions, II. Zap. Naučn. Sem. Leningrad.Otdel. Mat. Inst. Steklov. (LOMI) vol. 41, pp. 14–24 (1974)113. Talenti, G.: Best constant in Sobolev inequality. Ann. Mat. Pura Appl. 110, 353–372 (1976)114. Tamanini, I.: Regularity results for almostminimal oriented hypersurfaces.Quaderni delDipartimentodi Matematica dell’Università di Lecce, Lecce. http://cvgmt.sns.it/paper/1807/ (1984)115. Van Schaftingen, J.,Willem,M.: Set transformations, symmetrizations and isoperimetric inequalities,Nonlinear analysis and applications to physical sciences, 135–152. Springer Italia, Milan (2004)123

A note on Cheeger sets

Abstract

Extracted data

A note on Cheeger sets

Abstract

Extracted data

Related items

Related items