Mixed Volumes, the Wigner Caustic, and Isoperimetric Inequalities

Michał Zwierzyński Email address: Michal.Zwierzynski@pw.edu.pl Email address: ORCID: 0000-0002-9627-1563 Address: Warsaw University of Technology
Faculty of Mathematics and Information Science
ul. Koszykowa 75
00-662 Warsaw, Poland

Abstract.

Let $K\subset\mathbb{R}^{n}$ be a convex body and $C=\frac{1}{2}(K+(-K))$ its central symmetral. We study the defects $\mathcal{A}_{k}(K)=W_{n-k}(C)-W_{n-k}(K)$ through the even–odd decomposition of the support function. We derive exact even mixed-volume expansions, identify their Krawtchouk transform with the mixed difference-body coefficients, and obtain a Kubota formula expressing each defect as the averaged symmetrization gain of projections. These identities yield strengthened isoperimetric inequalities. The quadratic defect is the average of the absolute oriented areas of projected Wigner caustics and obeys a sharp spherical-harmonic stability estimate. We give explicit formulas in dimensions $2$ , $3$ , and $4$ . The four-dimensional expansion includes a sign-indefinite quartic Wigner volume; on a plane of cubic harmonics we find nonspherical constant-width bodies for which it vanishes. Finally, we prove a projection-transfer bound and an arbitrary-dimensional expansion for constant-width bodies.

Key words and phrases:

central symmetral, Wigner caustic, projective hedgehog, mixed volume, quermassintegral, isoperimetric inequality, constant width

2020 Mathematics Subject Classification

Primary 52A40; Secondary 52A39, 53A07

1. Introduction

The planar improved isoperimetric inequality

(1.1)

L(K)^{2}\geqslant 4\uppi\left(\operatorname{Area}(K)+2\left|A^{*}_{\mathcal{W}}(K)\right|\right)

relates the perimeter and area of a convex body to the oriented area of the Wigner caustic of its boundary. Equality holds precisely for bodies of constant width. This inequality was proved in [42, Theorem 3.4] and is the starting point of this paper. Exact planar relations involving the same oriented area, the Constant Width Measure Set, and the Spherical Measure Set appear in [43, Theorem 2.17] and [44, Theorem 5.1]. A normed-plane analogue of the Wigner caustic and the Constant Width Measure Set, together with corresponding improved isoperimetric inequalities for admissible curves, was developed in [37, Sections 4–5]. The singularities and global geometry of planar Wigner caustics and affine equidistants are treated in [18, Section 2] and [17, Sections 3–6]. The related Centre Symmetry Set was introduced by Janeczko as the bifurcation set of a family of ratios [22] and reformulated by Giblin and Holtom as the envelope of chords joining parallel-tangent pairs [19]; its planar geometry is developed further in [31]. Schneider’s middle hedgehog extends the planar construction beyond smooth ovals [40, Sections 2–3], while recent support-function visualizations are given in [10]. Convex polygons with parallel opposite sides, together with their discrete equidistants and centre symmetry sets, were introduced in [8]; a discrete counterpart of (1.1) for this class was subsequently obtained in [24, Theorem 4.4]. Related Fourier decompositions and isoperimetric refinements occur in the work on Hurwitz-type deficits [9], higher-order discrete and smooth inequalities [27], mixed isoperimetric deficits [41], and higher-order preserving harmonic and midpoint sets [36, Theorem 5.2].

The Wigner caustic itself goes back to Berry’s phase-space construction [2, Section 6]. The role of odd functions in the singularity theory of higher-dimensional symplectic Wigner caustics was studied in [13]. Affine equidistants and global centre symmetry sets of Lagrangian submanifolds are developed further in [14, 15]; algebraic symmetry-defect loci are studied in [11, 23]. The convex-hypersurface setting considered here is different, since the antipodal Gauss directions select one distinguished pair of parallel tangent hyperplanes.

The purpose of this paper is to identify the higher-dimensional mixed-volume mechanism behind (1.1). It is elementary at its core: one applies classical inequalities to the central symmetral

(1.2)

C=\frac{1}{2}(K+(-K)),

whose support function is the even part of the support function of $K$ . The odd part is not the support function of a convex body in general, but it is a virtual convex body and hence determines a hedgehog. For smooth strictly convex bodies its geometric image is exactly the Wigner caustic. The formal-difference viewpoint was initiated in [28]; projective hedgehogs and their offset constructions are also treated in [34]; algebraic mixed volumes of hedgehogs are developed systematically in [29, Chapter 3, pp. 67–83]. In the planar smooth case the identity $x_{h}(u)-x_{h}(-u)=2x_{h_{\mathrm{even}}}(u)$ shows that the secant caustic of $\partial K$ is $2\partial C$ . Thus the central symmetral is, up to this homothety and filling the bounded component, a special case of the secant-caustic construction; we use the normalization of [16, Definition 2.4 and Corollary 4.3]. This decomposition turns the planar correction in (1.1) into a general mixed-volume mechanism.

Our main points are as follows.

(i)

For every $k=1,\ldots,n$ , the difference between $W_{n-k}(K)$ and the same quermassintegral of $C$ has an exact expansion containing only even numbers of copies of the Wigner support function. The same defect is the mean symmetrization gain of the $k$ -dimensional projections of $K$ , which yields positivity and a sharp Rogers–Shephard bound. Moreover, the mixed volumes of $K$ and $-K$ are, at top degree, exactly the Krawtchouk transform of these even Wigner terms. The relevant results are Theorems 3.2 and 3.3, and Propositions 4.1, 4.2, and 4.3.
(ii)

Applying the Alexandrov–Fenchel inequalities to $C$ gives a family which genuinely strengthens the generalized Urysohn inequalities for $K$ . Its $k=2$ member has the explicit Wigner correction

$\frac{2\kappa_{n}}{\uppi}\int_{\operatorname{Gr}(n,2)}\left|A^{*}_{\mathcal{W}}(K\mathbin{|}E)\right|\,\mathrm{d}E.$

The family is stated in Theorem 5.1; its Alexandrov–Fenchel input is clarified in Proposition 5.2, while the explicit Wigner interpretation of the $k=2$ correction follows from Theorem 4.5.
(iii)

The quadratic Wigner correction is a positive spectral energy of the odd part of the support function. It controls, with a sharp constant, the squared $L^{2}$ -distance from central symmetry after translations are removed. These statements are proved in Theorem 6.1 and Corollary 6.2.
(iv)

The cases $n=2,3,4$ are made explicit. In dimension four, a pure quartic mixed volume $V(h_{\mathrm{odd}}^{[4]})$ appears. It is the Gauss-cover algebraic volume of the projective Wigner hedgehog and assumes both signs arbitrarily close to a ball. On an explicit plane of cubic harmonics we determine the quartic form, its zero cone, and the optimal quartic-to-quadratic ratios. A termwise difference-body estimate sharpens the lower boundary of the quadratic–quartic feasible region, while projection to three dimensions transfers Nishioka’s bound [33, Theorem 1] to $\mathcal{A}_{2}\leqslant\kappa_{4}r^{2}/11$ at width $2r$ . The dimension-specific formulas are developed in Sections 7, 8, and 9. The four-dimensional sign, cubic-harmonic, and feasible-region results are given in Lemma 9.1, Proposition 9.2, and Theorem 9.3, respectively; the projection argument uses Theorem 4.4.
(v)

For constant-width bodies in arbitrary dimension, every quermassintegral has an even Wigner expansion. In even dimensions the top-degree projective Wigner volume survives as the residual term in the volume–surface relation. These results are stated in Theorem 10.1 and Corollary 10.2.

There is substantial recent activity around constant-width bodies and difference-body methods. Mixed-volume reformulations of the Blaschke–Lebesgue problem were obtained by Bogosel [3]; variational methods and Cheeger-type extremal questions were studied in [4, 5]. In dimension three, the support-function formula of Anciaux and Guilfoyle [1] underlies recent spectral progress, including Nishioka’s improved lower bound for the Blaschke–Lebesgue problem [33]. Higher-order difference bodies and their affine inequalities are developed in [21]. A termwise generalization of the higher-order difference-body inequality, together with its anti-blocking case and valuation interpretation, appears in [25]; a recent preprint proves the original Godbersen mixed-volume inequalities in full [26, Theorem 1.1]. Refined mean-width inequalities for origin-symmetric bodies appear in [6]. These works address different extremal or normalized problems. The contribution here is the unified Wigner interpretation of central-symmetral corrections and the resulting all-dimensional family. General background on constant-width bodies may be found in [30].

Algebraic volumes and mixed volumes of hedgehogs are established notions; the novelty claim concerns the identities and consequences above, not the classical Brunn–Minkowski, Alexandrov–Fenchel, Kubota, or Rogers–Shephard ingredients; see [29, Chapter 3, pp. 67–83] for the algebraic mixed-volume framework for hedgehogs.

Throughout, mixed-volume conventions and the differential formulas are those of [39, Sections 5.1 and 5.5]; the precise inequality and integral-geometric references are given where they are used.

2. Support functions, hedgehogs, and mixed volumes

2.1. Convex bodies and quermassintegrals

Let $K\subset\mathbb{R}^{n}$ be a convex body, i.e. a compact convex set with non-empty interior. Its support function is

h_{K}(u)=\max_{x\in K}\langle x,u\rangle,\qquad u\in\mathbb{S}^{n-1}.

We write $B^{n}$ for the Euclidean unit ball and $\kappa_{n}=\operatorname{Vol}_{n}(B^{n})$ . The mixed volume of $n$ convex bodies is the symmetric coefficient defined by the volume polynomial. Equivalently,

(2.1)

V(K_{1},\ldots,K_{n})=\left.\frac{1}{n!}\frac{\partial^{n}}{\partial t_{1}\cdots\partial t_{n}}\operatorname{Vol}_{n}(t_{1}K_{1}+\cdots+t_{n}K_{n})\right|_{t_{1}=\cdots=t_{n}=0}.

See [39, Section 5.1]. Repeated arguments are written as $V(K^{[r]},L^{[s]},B^{[t]})$ . The quermassintegrals are normalized by

(2.2)

W_{j}(K)=V(K^{[n-j]},B^{[j]}),\qquad j=0,\ldots,n.

Thus

(2.3)

\operatorname{Vol}_{n}(K+tB^{n})=\sum_{j=0}^{n}\binom{n}{j}W_{j}(K)t^{j},

and

(2.4)

W_{0}(K)=\operatorname{Vol}_{n}(K),\qquad W_{n}(K)=\kappa_{n},\qquad nW_{1}(K)=S(K),

where $\mathcal{H}^{n-1}$ denotes $(n-1)$ -dimensional Hausdorff measure and $S(K)=\mathcal{H}^{n-1}(\partial K)$ .

We first work with a convex body of class $C^{2}_{+}$ , meaning that $\partial K$ is $C^{2}$ and has positive Gauss curvature. Every convex body is a Hausdorff limit of convex bodies of class $C^{\infty}_{+}$ : one applies the smoothing procedure from [39, Theorem 3.4.1] and then adds a Euclidean ball of vanishing radius. Since mixed volumes, and hence quermassintegrals, are continuous under Hausdorff convergence [39, pp. 280–281], the inequalities involving only convex bodies and quermassintegrals extend to arbitrary convex bodies.

For $f\in C^{2}(\mathbb{S}^{n-1})$ , put

(2.5)

Q_{f}=\nabla^{2}_{\mathbb{S}^{n-1}}f+f\operatorname{Id}.

The associated hedgehog is parametrized by

(2.6)

x_{f}(u)=f(u)u+\nabla_{\mathbb{S}^{n-1}}f(u).

If $f=h_{K}$ , then $Q_{f}$ is positive definite and (2.6) is the inverse Gauss-map parametrization of $\partial K$ ; see [39, Section 2.5]. Conversely, for every smooth $f$ , the operator $Q_{f+R}=Q_{f}+R\operatorname{Id}$ is positive definite when $R$ is sufficiently large. Thus $f+R$ is a support function and $f=(f+R)-R$ is a formal difference of support functions. This gives a direct multilinear extension of mixed volume to all smooth functions.

Let $D_{n-1}$ be the polarized determinant on symmetric endomorphisms of an $(n-1)$ -dimensional vector space, normalized by $D_{n-1}(A,\ldots,A)=\det A$ . The mixed-volume form extends from support functions of convex bodies to arbitrary smooth functions by

(2.7)

V(f_{1},\ldots,f_{n})=\frac{1}{n}\int_{\mathbb{S}^{n-1}}f_{1}D_{n-1}(Q_{f_{2}},\ldots,Q_{f_{n}})\,\mathrm{d}\sigma.

Integration by parts makes (2.7) symmetric in all arguments. This is the usual mixed volume on convex support functions and its multilinear extension to virtual convex bodies or hedgehogs; see [39, Sections 5.1 and 5.5]. In particular, in expressions such as $V(f^{[r]},g^{[s]},1^{[t]})$ , the function $1$ is the support function of $B^{n}$ .

If $\ell(u)=\langle a,u\rangle$ is linear, then

(2.8)

Q_{\ell}=0.

Consequently, every mixed volume with one linear argument vanishes. This is the support-function form of translation invariance.

2.2. The even–odd decomposition

Write $h=h_{K}$ and decompose

(2.9)

\displaystyle h=h_{\mathrm{odd}}+h_{\mathrm{even}},\quad\text{where}\quad h_{\mathrm{even}}(u)=\frac{h(u)+h(-u)}{2},\qquad h_{\mathrm{odd}}(u)=\frac{h(u)-h(-u)}{2}.

Thus $h_{\mathrm{even}}$ is even and $h_{\mathrm{odd}}$ is odd. The notation records the parity explicitly and will be used throughout.

Definition 2.1 (Wigner caustic).

For a convex body $K\subset\mathbb{R}^{n}$ of class $C^{2}_{+}$ , its Wigner caustic is the midpoint locus

(2.10)

\mathrm{E}_{1/2}(\partial K)=\left\{\frac{x_{h}(u)+x_{h}(-u)}{2}:u\in\mathbb{S}^{n-1}\right\}.

The two boundary points in each pair have opposite outer normals and hence parallel tangent hyperplanes.

Proposition 2.2 (Central symmetral and Wigner hedgehog).

Let $K\subset\mathbb{R}^{n}$ be a convex body of class $C^{2}_{+}$ with support function $h=h_{\mathrm{odd}}+h_{\mathrm{even}}$ . Then $h_{\mathrm{even}}$ is the support function of the central symmetral

(2.11)

C=\frac{1}{2}\bigl(K+(-K)\bigr).

Moreover,

(2.12)		$\displaystyle x_{h_{\mathrm{even}}}(u)$	$\displaystyle=\frac{x_{h}(u)-x_{h}(-u)}{2},$
(2.13)		$\displaystyle x_{h_{\mathrm{odd}}}(u)$	$\displaystyle=\frac{x_{h}(u)+x_{h}(-u)}{2}.$

In particular, $x_{h_{\mathrm{odd}}}(-u)=x_{h_{\mathrm{odd}}}(u)$ , and the image of $x_{h_{\mathrm{odd}}}$ is $\mathrm{E}_{1/2}(\partial K)$ , parametrized through the double cover $\mathbb{S}^{n-1}\to\mathbb{RP}^{n-1}$ . We call the resulting parametrized object the projective Wigner hedgehog; its geometric image is the Wigner caustic.

Proof.

The support function of $-K$ is $u\mapsto h(-u)$ , so Minkowski linearity of support functions proves (2.11). The antipodal map has differential $-\operatorname{Id}$ . Differentiating the even and odd identities for $h_{\mathrm{even}}$ and $h_{\mathrm{odd}}$ , and identifying $T_{u}\mathbb{S}^{n-1}$ and $T_{-u}\mathbb{S}^{n-1}$ with $u^{\perp}$ , and using (2.6), gives (2.12) and (2.13). The last assertion follows immediately from the oddness of $h_{\mathrm{odd}}$ . ∎

The Wigner caustic can be singular and self-intersecting. Formula (2.7), rather than an unsigned Hausdorff measure of its image, is therefore the appropriate source of its algebraic geometric quantities.

Translations leave $h_{\mathrm{even}}$ fixed and add a linear function to $h_{\mathrm{odd}}$ . Thus all mixed-volume expressions below are translation invariant. Also, $K$ has constant width $2r$ if and only if

(2.14)

h_{\mathrm{even}}\equiv r,

or equivalently if and only if $C=rB^{n}$ .

2.3. Parity

Lemma 2.3 (Parity of mixed volumes).

Suppose that $f_{1},\ldots,f_{n}$ are formal differences of support functions and that each is either even or odd on $\mathbb{S}^{n-1}$ . If an odd number of these functions are odd, then $V(f_{1},\ldots,f_{n})=0$ .

Proof.

For $f\in C^{2}(\mathbb{S}^{n-1})$ , let $\widehat{f}(u)=f(-u)$ . Simultaneous reflection of all arguments leaves mixed volume unchanged, hence

V(\widehat{f}_{1},\ldots,\widehat{f}_{n})=V(f_{1},\ldots,f_{n}).

If exactly $s$ functions are odd, multilinearity makes the left-hand side equal to $(-1)^{s}V(f_{1},\ldots,f_{n})$ . For odd $s$ , the mixed volume equals its negative. ∎

2.4. Regularity and approximation

The differential expressions in (2.5)–(2.7) and the parametrized Wigner hedgehog require the stated smoothness. The convex-geometric defects, projection identities, and inequalities do not. Indeed, support functions identify Hausdorff convergence of convex bodies with uniform convergence on the sphere, and mixed volumes are continuous under this convergence; see [39, Sections 1.8 and 5.1]. Convolution with a smooth rotational approximate identity gives smooth support functions and commutes with the antipodal map, so the even and odd parts converge separately. Adding a vanishing Euclidean-ball summand makes the approximants $C^{2}_{+}$ .

If $K$ has constant width $2r$ , rotational convolution preserves the identity $h(u)+h(-u)=2r$ . A subsequent Minkowski interpolation with $rB^{n}$ preserves the same width and yields positive curvature. Thus every constant-width body can be approximated by $C^{\infty}$ bodies of class $C^{2}_{+}$ and the same width. Consequently, all statements below that involve only $W_{j}$ , $\mathcal{A}_{k}$ , projections, or mixed volumes extend by continuity to arbitrary convex bodies. Pointwise differential or algebraic-volume interpretations are asserted only in the regularity class specified in their statements.

3. Symmetrization defects and their exact expansion

For $t\in[-1,1]$ , consider the Minkowski symmetrization path

(3.1)

K_{t}=\frac{1+t}{2}K+\frac{1-t}{2}(-K).

Its support function is

(3.2)

h_{t}=h_{\mathrm{even}}+th_{\mathrm{odd}}.

Thus $K_{1}=K$ , $K_{-1}=-K$ , and $K_{0}=C$ .

Definition 3.1.

For $k=1,\ldots,n$ , the $k$ -dimensional symmetrization defect is

(3.3)

\mathcal{A}_{k}(K)=W_{n-k}(C)-W_{n-k}(K).

We also put $\mathcal{A}_{0}(K)=0$ .

Notice that $\mathcal{A}_{1}=0$ , because $W_{n-1}$ is Minkowski linear and reflection invariant. The volume defect is $\mathcal{A}_{n}$ , while the surface-area defect is $n\mathcal{A}_{n-1}$ .

Theorem 3.2 (Exact Wigner expansion).

Let $K\subset\mathbb{R}^{n}$ be a convex body, let $C$ be its central symmetral, and write $h=h_{\mathrm{odd}}+h_{\mathrm{even}}$ as in (2.9). For $k=1,\ldots,n$ ,

(3.4)

\displaystyle W_{n-k}(K_{t})

\displaystyle=W_{n-k}(C)+\sum_{r=1}^{\lfloor k/2\rfloor}\binom{k}{2r}t^{2r}V\left(h_{\mathrm{even}}^{[k-2r]},h_{\mathrm{odd}}^{[2r]},1^{[n-k]}\right).

Consequently,

(3.5)

\mathcal{A}_{k}(K)=-\sum_{r=1}^{\lfloor k/2\rfloor}\binom{k}{2r}V\left(h_{\mathrm{even}}^{[k-2r]},h_{\mathrm{odd}}^{[2r]},1^{[n-k]}\right).

Proof.

By (2.2), (3.2), and multilinearity,

W_{n-k}(K_{t})=\sum_{s=0}^{k}\binom{k}{s}t^{s}V\left(h_{\mathrm{even}}^{[k-s]},h_{\mathrm{odd}}^{[s]},1^{[n-k]}\right).

Lemma 2.3 removes every odd value of $s$ . The term $s=0$ is $W_{n-k}(C)$ . Setting $t=1$ and using (3.3) proves (3.5). ∎

Formula (3.4) gives a variational meaning to every mixed term: it is an even response of a quermassintegral when one moves from the central symmetral in the Wigner direction $h_{\mathrm{odd}}$ .

Theorem 3.3 (Positivity, rigidity, and the quadratic response).

Let $K\subset\mathbb{R}^{n}$ be a convex body and $C$ its central symmetral. For $k=2,\ldots,n$ ,

(3.6)

\mathcal{A}_{k}(K)\geqslant 0.

Equality holds if and only if $K$ is centrally symmetric up to translation. Moreover,

(3.7)

V\left(h_{\mathrm{even}}^{[k-2]},h_{\mathrm{odd}}^{[2]},1^{[n-k]}\right)\leqslant 0.

Proof.

The Brunn–Minkowski inequality for quermassintegrals states that

(3.8)

W_{n-k}((1-s)L+sM)^{1/k}\geqslant(1-s)W_{n-k}(L)^{1/k}+sW_{n-k}(M)^{1/k};

see [39, Theorem 7.4.5] for the inequality and [39, Theorems 7.4.6 and 7.6.9] for the equality case. Apply (3.8) to $L=K$ , $M=-K$ , and $s=1/2$ . Reflection invariance gives $W_{n-k}(-K)=W_{n-k}(K)$ , and therefore (3.6).

For $k\geqslant 2$ , equality in (3.8) holds precisely when $K$ and $-K$ are homothetic. They have the same size, so the homothety ratio is one and they differ by a translation. This is equivalent to central symmetry up to translation.

Finally, $t\mapsto W_{n-k}(K_{t})^{1/k}$ is concave by (3.8) and even by reflection invariance. Hence its second derivative at $t=0$ is non-positive. The first derivative of $W_{n-k}(K_{t})$ vanishes at zero, while (3.4) gives

\left.\frac{\,\mathrm{d}^{2}}{\,\mathrm{d}t^{2}}\right|_{t=0}W_{n-k}(K_{t})=k(k-1)V\left(h_{\mathrm{even}}^{[k-2]},h_{\mathrm{odd}}^{[2]},1^{[n-k]}\right).

This proves (3.7). ∎

Remark 3.4.

The total sum in (3.5) has a prescribed sign, but the higher individual terms generally do not. This point is already essential in dimension four, where a quartic Wigner term occurs.

4. Projection formulas and Rogers–Shephard bounds

Let $\operatorname{Gr}(n,k)$ be the Grassmannian of $k$ -dimensional linear subspaces of $\mathbb{R}^{n}$ , equipped with invariant probability measure $\,\mathrm{d}E$ . We write $K\mathbin{|}E$ for the orthogonal projection of $K$ onto $E$ .

The Kubota formula, with the normalization (2.2), reads [39, (5.72)]

(4.1)

W_{n-k}(K)=\frac{\kappa_{n}}{\kappa_{k}}\int_{\operatorname{Gr}(n,k)}\operatorname{Vol}_{k}(K\mathbin{|}E)\,\mathrm{d}E.

Orthogonal projection commutes with Minkowski addition and reflection, so

C\mathbin{|}E=\frac{K\mathbin{|}E+(-(K\mathbin{|}E))}{2}.

Subtract (4.1) for $K$ from the same formula for $C$ and we obtain the following proposition.

Proposition 4.1 (Projection formula).

Let $K\subset\mathbb{R}^{n}$ be a convex body and $C=\frac{1}{2}(K+(-K))$ its central symmetral. For $k=1,\ldots,n$ ,

(4.2)

\mathcal{A}_{k}(K)=\frac{\kappa_{n}}{\kappa_{k}}\int_{\operatorname{Gr}(n,k)}\left[\operatorname{Vol}_{k}\left(\frac{K\mathbin{|}E+(-(K\mathbin{|}E))}{2}\right)-\operatorname{Vol}_{k}(K\mathbin{|}E)\right]\,\mathrm{d}E.

Thus $\mathcal{A}_{k}$ is an averaged $k$ -dimensional asymmetry deficit. The pointwise integrand in (4.2) is non-negative by the ordinary Brunn–Minkowski inequality in $E$ .

Proposition 4.2 (Two-sided universal bound).

Let $K\subset\mathbb{R}^{n}$ be a convex body. For $k=1,\ldots,n$ ,

(4.3)

0\leqslant\mathcal{A}_{k}(K)\leqslant\left(2^{-k}\binom{2k}{k}-1\right)W_{n-k}(K).

For $n=k$ , the upper constant is sharp and equality is attained by simplices.

Proof.

Let $L=K\mathbin{|}E\subset E$ . The Rogers–Shephard difference-body inequality [35, Theorem 1] gives

\operatorname{Vol}_{k}(L-L)\leqslant\binom{2k}{k}\operatorname{Vol}_{k}(L).

For full-dimensional $L\subset E$ , equality holds if and only if $L$ is a simplex. This includes the one-dimensional case, in which every interval is a simplex. Since $(L-L)/2$ is the central symmetral of $L$ ,

0\leqslant\operatorname{Vol}_{k}\left(\frac{L-L}{2}\right)-\operatorname{Vol}_{k}(L)\leqslant\left(2^{-k}\binom{2k}{k}-1\right)\operatorname{Vol}_{k}(L).

Integrating and using (4.1) proves (4.3). When $n=k$ , this is exactly the sharp Rogers–Shephard inequality. ∎

Proposition 4.3 (Difference-body coefficients in Wigner variables).

Let $K\subset\mathbb{R}^{n}$ be a convex body, write $h_{K}=h_{\mathrm{even}}+h_{\mathrm{odd}}$ , and set

I_{2r}=V\left(h_{\mathrm{even}}^{[n-2r]},h_{\mathrm{odd}}^{[2r]}\right),\qquad r=0,\ldots,\left\lfloor\frac{n}{2}\right\rfloor.

For $j=0,\ldots,n$ , define

G_{j}(K)=V\left((-K)^{[j]},K^{[n-j]}\right)

and the binary Krawtchouk polynomial

(4.4)

\operatorname{Kr}_{m}(j;n)=\sum_{\ell=0}^{m}(-1)^{\ell}\binom{j}{\ell}\binom{n-j}{m-\ell},

where binomial coefficients outside their natural range are understood to be zero. Then

(4.5)

G_{j}(K)=\sum_{r=0}^{\lfloor n/2\rfloor}\operatorname{Kr}_{2r}(j;n)I_{2r}.

Conversely,

(4.6)

I_{2r}=\frac{1}{2^{n}\binom{n}{2r}}\sum_{j=0}^{n}\binom{n}{j}\operatorname{Kr}_{2r}(j;n)G_{j}(K).

Along the symmetrization path (3.1), the same volume polynomial therefore has the two representations

(4.7)

\displaystyle\operatorname{Vol}_{n}(K_{t})=\sum_{r=0}^{\lfloor n/2\rfloor}\binom{n}{2r}t^{2r}I_{2r}=2^{-n}\sum_{j=0}^{n}\binom{n}{j}(1+t)^{n-j}(1-t)^{j}G_{j}(K).

Proof.

The support functions of $K$ and $-K$ are, respectively, $h_{\mathrm{even}}+h_{\mathrm{odd}}$ and $h_{\mathrm{even}}-h_{\mathrm{odd}}$ . Expanding $G_{j}(K)$ by multilinearity shows that the coefficient of $V(h_{\mathrm{even}}^{[n-m]},h_{\mathrm{odd}}^{[m]})$ is exactly $\operatorname{Kr}_{m}(j;n)$ . Lemma 2.3 removes the odd values of $m$ and proves (4.5). The standard orthogonality relation

\sum_{j=0}^{n}\binom{n}{j}\operatorname{Kr}_{m}(j;n)\operatorname{Kr}_{s}(j;n)=2^{n}\binom{n}{m}\,\delta_{ms}

follows by extracting the coefficient of $z^{m}w^{s}$ from

\sum_{j=0}^{n}\binom{n}{j}\bigl((1-z)(1-w)\bigr)^{j}\bigl((1+z)(1+w)\bigr)^{n-j}=2^{n}(1+zw)^{n}.

It gives (4.6). Finally, the first equation of (4.7) is the case $k=n$ of (3.4), while the second follows by expanding $K_{t}=\frac{1+t}{2}K+\frac{1-t}{2}(-K)$ directly. ∎

Thus the Wigner expansion and the mixed difference-body expansion are the power-basis and Bernstein-basis descriptions of one polynomial. The termwise estimates

(4.8)

G_{j}(K)\leqslant\binom{n}{j}\operatorname{Vol}_{n}(K),\qquad j=0,\ldots,n,

are the Godbersen inequalities. Their higher-order difference-body extension and its proof for anti-blocking bodies are developed in [25]; the full family (4.8) is proved in the recent preprint [26, Theorem 1.1]. Summing (4.8) with weights $\binom{n}{j}$ gives

\operatorname{Vol}_{n}(K-K)=2^{n}I_{0}\leqslant\sum_{j=0}^{n}\binom{n}{j}^{2}\operatorname{Vol}_{n}(K)=\binom{2n}{n}\operatorname{Vol}_{n}(K),

so Proposition 4.2 is their aggregated Rogers–Shephard consequence when $k=n$ . Applying the same argument to every $k$ -dimensional projection and integrating gives the full bound (4.3).

4.1. Transfer of lower volume bounds through projections

The projection formula also transfers any dimension-specific lower volume bound for constant-width bodies to all higher dimensions.

Theorem 4.4 (Projection-transfer principle).

Fix $2\leqslant k\leqslant n$ and suppose that every $k$ -dimensional convex body $L$ of constant width $2r$ satisfies

(4.9)

\operatorname{Vol}_{k}(L)\geqslant\mu_{k}\kappa_{k}r^{k}

for some constant $0<\mu_{k}\leqslant 1$ . If $K\subset\mathbb{R}^{n}$ has constant width $2r$ , then

(4.10)

\mathcal{A}_{k}(K)\leqslant\kappa_{n}(1-\mu_{k})r^{k}.

Proof.

For every $E\in\operatorname{Gr}(n,k)$ , the projection $K\mathbin{|}E$ has constant width $2r$ in $E$ , while its central symmetral is $rB_{E}$ . Hence (4.9) gives

\operatorname{Vol}_{k}(rB_{E})-\operatorname{Vol}_{k}(K\mathbin{|}E)\leqslant(1-\mu_{k})\kappa_{k}r^{k}.

Insert this pointwise estimate into (4.2). The invariant measure on the Grassmannian has total mass one. ∎

For $k=2$ , the Blaschke–Lebesgue theorem gives $\mu_{2}=2(\uppi-\sqrt{3})/\uppi$ ; see [3, Theorem 1]. Nishioka’s recent three-dimensional bound [33, Theorem 1], written at width $2r$ , gives $\mu_{3}=8/11$ . Thus every constant-width body in $\mathbb{R}^{n}$ satisfies

(4.11)		$\displaystyle\mathcal{A}_{2}(K)$	$\displaystyle\leqslant\kappa_{n}\left(\frac{2\sqrt{3}}{\uppi}-1\right)r^{2},$
(4.12)		$\displaystyle\mathcal{A}_{3}(K)$	$\displaystyle\leqslant\frac{3}{11}\kappa_{n}r^{3}.$

The second estimate will materially sharpen the four-dimensional outer bound in Section 9.

4.2. The projected Wigner-area formula

Assume temporarily that $K$ is of class $C^{2}_{+}$ . For $E\in\operatorname{Gr}(n,2)$ , let $h_{\mathrm{odd},E}$ be the restriction of $h_{\mathrm{odd}}$ to the unit circle of $E$ . The Wigner caustic of $\partial(K\mathbin{|}E)$ is parametrized by $x_{h_{\mathrm{odd},E}}$ . Its full-cover algebraic area is

(4.13)

a(h_{\mathrm{odd},E})=\frac{1}{2}\int_{0}^{2\uppi}\left(h_{\mathrm{odd},E}^{2}-(h_{\mathrm{odd},E}^{\prime})^{2}\right)\,\mathrm{d}\theta.

Because $x_{h_{\mathrm{odd},E}}$ is $\uppi$ -periodic, the oriented area counted once is

(4.14)

A^{*}_{\mathcal{W}}(K\mathbin{|}E)=\frac{1}{2}a(h_{\mathrm{odd},E}).

The function $h_{\mathrm{odd},E}$ is $\uppi$ -antiperiodic. Wirtinger’s inequality therefore implies $a(h_{\mathrm{odd},E})\leqslant 0$ . The first harmonic is the harmless translation mode.

Theorem 4.5 (Quadratic Wigner projection formula).

Let $K\subset\mathbb{R}^{n}$ be a convex body of class $C^{2}_{+}$ and let $C$ be its central symmetral. Then

(4.15)

\mathcal{A}_{2}(K)=\frac{2\kappa_{n}}{\uppi}\int_{\operatorname{Gr}(n,2)}\left|A^{*}_{\mathcal{W}}(K\mathbin{|}E)\right|\,\mathrm{d}E.

In particular,

(4.16)

0\leqslant\mathcal{A}_{2}(K)\leqslant\frac{1}{2}W_{n-2}(K).

Proof.

Formula (3.5) with $k=2$ gives

\mathcal{A}_{2}(K)=-V(h_{\mathrm{odd}},h_{\mathrm{odd}},1^{[n-2]}).

Kubota’s formula in the precise normalization (4.1) – that is, [39, (5.72)] – is polynomial under Minkowski addition and hence extends by polarization from convex support functions to virtual ones. Therefore

(4.17)

V(h_{\mathrm{odd}},h_{\mathrm{odd}},1^{[n-2]})=\frac{\kappa_{n}}{\uppi}\int_{\operatorname{Gr}(n,2)}a(h_{\mathrm{odd},E})\,\mathrm{d}E.

Now use (4.14) and $a(h_{\mathrm{odd},E})\leqslant 0$ . Inequality (4.16) is the case $k=2$ of Proposition 4.2. Equivalently, it follows by applying the sharp planar estimate $|A^{*}_{\mathcal{W}}(L)|\leqslant\operatorname{Area}(L)/4$ inside (4.15). This planar estimate is recorded and derived from Rogers–Shephard in Corollary 7.2 below. ∎

The extension of (4.15) to non-smooth planar projections can be phrased using Schneider’s middle hedgehog [40]. The mixed volume and projection formulas themselves remain valid without smoothness.

5. Strengthened isoperimetric inequalities from the central symmetral

The Alexandrov–Fenchel inequality [39, Theorem 7.3.1], iterated along the quermassintegrals, gives the generalized Urysohn inequalities: for every convex body $L\subset\mathbb{R}^{n}$ and $k=1,\ldots,n$ ,

(5.1)

W_{n-1}(L)^{k}\geqslant\kappa_{n}^{\,k-1}W_{n-k}(L).

For $k\geqslant 2$ , equality holds if and only if $L$ is a Euclidean ball. The equality characterization follows from [39, Theorems 7.4.6 and 7.6.9].

Theorem 5.1 (Central-symmetral family of strengthened isoperimetric inequalities).

Let $K\subset\mathbb{R}^{n}$ be a convex body and let $C$ be its central symmetral. For every $k=2,\ldots,n$ ,

(5.2)

W_{n-1}(K)^{k}-\kappa_{n}^{\,k-1}W_{n-k}(K)\geqslant\kappa_{n}^{\,k-1}\mathcal{A}_{k}(K).

Equality holds if and only if $K$ has constant width.

Proof.

Minkowski linearity and reflection invariance give

W_{n-1}(C)=W_{n-1}(K).

Apply (5.1) to $C$ and use

W_{n-k}(C)=W_{n-k}(K)+\mathcal{A}_{k}(K).

Equality holds precisely when $C$ is a ball. By (2.14), this is equivalent to constant width of $K$ . ∎

Theorem 5.1 is a genuine strengthening of (5.1), because $\mathcal{A}_{k}\geqslant 0$ . For $k=2$ , it takes the especially geometric form

(5.3)

W_{n-1}(K)^{2}\geqslant\kappa_{n}\left(W_{n-2}(K)+\frac{2\kappa_{n}}{\uppi}\int_{\operatorname{Gr}(n,2)}\left|A^{*}_{\mathcal{W}}(K\mathbin{|}E)\right|\,\mathrm{d}E\right).

5.1. The corrected Alexandrov–Fenchel chain

For convenience, set $\mathcal{A}_{0}=\mathcal{A}_{1}=0$ . Since $W_{j}(C)=W_{j}(K)+\mathcal{A}_{n-j}(K)$ , the entire Alexandrov–Fenchel chain for $C$ becomes the following.

Proposition 5.2.

Let $K\subset\mathbb{R}^{n}$ be a convex body and let $C$ be its central symmetral. For $j=0,\ldots,n-2$ ,

(5.4)

\displaystyle\left(W_{j+1}(K)+\mathcal{A}_{n-j-1}(K)\right)^{2}

\displaystyle\geqslant\left(W_{j}(K)+\mathcal{A}_{n-j}(K)\right)\cdot\left(W_{j+2}(K)+\mathcal{A}_{n-j-2}(K)\right).

Equality holds if and only if $K$ has constant width.

Proof.

This is $W_{j+1}(C)^{2}\geqslant W_{j}(C)W_{j+2}(C)$ , followed by the equality characterization for the Alexandrov–Fenchel inequality with the unit ball among the fixed arguments. The inequality is [39, Theorem 7.3.1]. The equality statement follows from [39, Theorems 7.4.6 and 7.6.9]. ∎

5.2. The classical isoperimetric inequality applied to $C$

The volume and surface area of the central symmetral are

(5.5)		$\displaystyle\operatorname{Vol}_{n}(C)$	$\displaystyle=\operatorname{Vol}_{n}(K)+\mathcal{A}_{n}(K),$
(5.6)		$\displaystyle S(C)$	$\displaystyle=S(K)+n\mathcal{A}_{n-1}(K).$

Theorem 5.3 (Symmetral surface–volume inequality).

For every convex body $K\subset\mathbb{R}^{n}$ ,

(5.7)

\left(S(K)+n\mathcal{A}_{n-1}(K)\right)^{n}\geqslant n^{n}\kappa_{n}\left(\operatorname{Vol}_{n}(K)+\mathcal{A}_{n}(K)\right)^{n-1}.

Equality holds if and only if $K$ has constant width.

Proof.

Substitute (5.5) and (5.6) into the Euclidean isoperimetric inequality

S(C)^{n}\geqslant n^{n}\kappa_{n}\operatorname{Vol}_{n}(C)^{n-1}.

For this normalization and the equality case, see [39, Section 7.2]. Equality holds precisely when $C$ is a ball, equivalently when $K$ has constant width. ∎

For $n=2$ , $\mathcal{A}_{n-1}=\mathcal{A}_{1}=0$ , so (5.7) is a genuine one-sided strengthening. For $n\geqslant 3$ , both sides are corrected, and (5.7) should be understood as an exact Wigner rewriting of the isoperimetric inequality of $C$ .

6. The universal quadratic Wigner energy

The defect $\mathcal{A}_{2}$ admits a direct analytic formula in every dimension.

Theorem 6.1 (Spectral formula and rigidity).

Let $K\subset\mathbb{R}^{n}$ be a convex body of class $C^{2}_{+}$ , let $C$ be its central symmetral, and let $h_{\mathrm{odd}}$ be the odd part of its support function. Then

(6.1)

\mathcal{A}_{2}(K)=\frac{1}{n(n-1)}\int_{\mathbb{S}^{n-1}}\left(|\nabla h_{\mathrm{odd}}|^{2}-(n-1)h_{\mathrm{odd}}^{2}\right)\,\mathrm{d}\sigma.

It is non-negative and vanishes if and only if $h_{\mathrm{odd}}$ is the restriction of a linear function, equivalently if and only if $K$ is centrally symmetric up to translation.

Proof.

For a smooth function $g$ , the operator $Q_{g}=\nabla^{2}_{\mathbb{S}^{n-1}}g+g\operatorname{Id}$ is the spherical radius-of-curvature operator defined in (2.5). The polarized determinant satisfies

D_{n-1}(Q_{g},\operatorname{Id}^{[n-2]})=\frac{\operatorname{tr}Q_{g}}{n-1}=\frac{\Delta g+(n-1)g}{n-1}.

Using (2.7) and integrating by parts gives, for arbitrary smooth $f,g$ ,

(6.2)

V(f,g,1^{[n-2]})=\frac{1}{n(n-1)}\int_{\mathbb{S}^{n-1}}\left((n-1)fg-\langle\nabla f,\nabla g\rangle\right)\,\mathrm{d}\sigma.

Now use (3.5) with $k=2$ . Since $h_{\mathrm{odd}}$ is odd, it has zero spherical mean. The sharp Poincaré inequality on $\mathbb{S}^{n-1}$ is

(6.3)

\int_{\mathbb{S}^{n-1}}|\nabla h_{\mathrm{odd}}|^{2}\,\mathrm{d}\sigma\geqslant(n-1)\int_{\mathbb{S}^{n-1}}h_{\mathrm{odd}}^{2}\,\mathrm{d}\sigma.

Its equality space is the first eigenspace of $-\Delta_{\mathbb{S}^{n-1}}$ , namely the restrictions of linear functions. Applying (6.3) to (6.1) proves non-negativity and rigidity. A linear odd term is exactly a translation of $K$ . The sharp inequality, sphere spectrum, and harmonic eigenspaces are recalled in [32, Chapter 2]. ∎

Let

h_{\mathrm{odd}}=\sum_{\begin{subarray}{c}\ell\geqslant 1\\ \ell\text{ odd}\end{subarray}}h_{\mathrm{odd},\ell}

be the spherical-harmonic decomposition, where

-\Delta h_{\mathrm{odd},\ell}=\lambda_{\ell}h_{\mathrm{odd},\ell},\qquad\lambda_{\ell}=\ell(\ell+n-2).

Then (6.1) becomes

(6.4)

\mathcal{A}_{2}(K)=\frac{1}{n(n-1)}\sum_{\begin{subarray}{c}\ell\geqslant 1\\ \ell\text{ odd}\end{subarray}}\left(\lambda_{\ell}-(n-1)\right)\|h_{\mathrm{odd},\ell}\|_{L^{2}}^{2}.

The $\ell=1$ summand is zero. Translate $K$ so that $h_{\mathrm{odd},1}=0$ . Since the next admissible degree is $\ell=3$ and

\lambda_{3}-(n-1)=2(n+2),

we obtain the following sharp estimate.

Corollary 6.2 (Quantitative central-asymmetry estimate).

Let $K\subset\mathbb{R}^{n}$ be a convex body of class $C^{2}_{+}$ and let $h_{\mathrm{odd}}$ be the odd part of its support function. After the translation mode has been removed,

(6.5)

\mathcal{A}_{2}(K)\geqslant\frac{2(n+2)}{n(n-1)}\|h_{\mathrm{odd}}\|_{L^{2}(\mathbb{S}^{n-1})}^{2}.

The constant is sharp, and equality holds precisely when $h_{\mathrm{odd}}$ belongs to the degree-three spherical-harmonic eigenspace (including $h_{\mathrm{odd}}=0$ ). Consequently,

(6.6)

\displaystyle W_{n-1}(K)^{2}-\kappa_{n}W_{n-2}(K)

\displaystyle\geqslant\kappa_{n}\mathcal{A}_{2}(K)\geqslant\frac{2\kappa_{n}(n+2)}{n(n-1)}\|h_{\mathrm{odd}}\|_{L^{2}}^{2}.

Proof.

Estimate (6.5) follows term by term from (6.4). Equality in the spectral estimate requires that only the degree-three component be present. The first inequality in (6.6) is the $k=2$ case of Theorem 5.1. ∎

7. Dimension two

Let $K\subset\mathbb{R}^{2}$ be a smooth strictly convex body. Write $L(K)$ and $A(K)$ for its perimeter and area. Then $W_{1}(K)=\frac{1}{2}L(K)$ , $W_{0}(K)=A(K)$ , and $\kappa_{2}=\uppi$ . The full-cover algebraic area of the Wigner hedgehog is

V(h_{\mathrm{odd}},h_{\mathrm{odd}})=\frac{1}{2}\int_{0}^{2\uppi}\left(h_{\mathrm{odd}}^{2}-(h_{\mathrm{odd}}^{\prime})^{2}\right)\,\mathrm{d}\theta=2A^{*}_{\mathcal{W}}(K).

It is non-positive. Consequently,

(7.1)

\mathcal{A}_{2}(K)=A(C)-A(K)=-V(h_{\mathrm{odd}},h_{\mathrm{odd}})=2\left|A^{*}_{\mathcal{W}}(K)\right|.

For an arbitrary planar convex body, we use

A^{*}_{\mathcal{W}}(K):=-\frac{1}{2}\bigl(A(C)-A(K)\bigr)

as the definition of its projectively reduced algebraic Wigner area. Schneider’s middle-hedgehog construction [40, Sections 2–3] shows that this convention agrees with the oriented area above in the smooth strictly convex case. Thus (7.1) holds for every planar convex body with this interpretation.

Figure 1. The convex body

K

(blue), its central symmetral

C

(dashed), and the Wigner caustic

x_{q}(\mathbb{S}^{1})

(red). The black chord joins two boundary points with opposite Gauss directions, and the short black segments indicate their parallel tangent lines. The marked red midpoint belongs to the Wigner caustic, which is traversed twice by the parametrization

x_{q}

Figure 1 illustrates the decomposition for the explicit support function

(7.2)

h(\theta)=2+0.10\cos(2\theta)+0.068\sin(5\theta),\qquad\theta\in\mathbb{R}/2\uppi\mathbb{Z}.

Its even and odd parts are, respectively, $c(\theta)=2+0.10\cos(2\theta)$ , $q(\theta)=0.068\sin(5\theta)$ . Thus $c$ is the support function of the central symmetral $C$ , whereas $x_{q}(\mathbb{S}^{1})$ parametrizes the Wigner caustic. Moreover,

(7.3)

h(\theta)+h^{\prime\prime}(\theta)=2-0.30\cos(2\theta)-1.632\sin(5\theta)\geqslant 0.068,

so $h$ is the support function of a smooth strictly convex body. The marked points in the figure are $p_{+}=x_{h}(u_{0})$ , $p_{-}=x_{h}(-u_{0})$ , $u_{0}=(\cos\theta_{0},\sin\theta_{0})$ for $\theta_{0}=0.73$ . Their outer normals are opposite, hence their tangent lines are parallel, and linearity of (2.6) gives

\frac{1}{2}(p_{+}+p_{-})=x_{q}(u_{0})\in\mathrm{E}_{1/2}(\partial K).

Since $L(C)=L(K)$ , Theorem 5.1 gives the following result for $n=k=2$ .

Theorem 7.1.

[42, Theorem 3.4] Let $K$ be a convex planar body. Then

(7.4)

L(K)^{2}\geqslant 4\uppi\left(A(K)+2\left|A^{*}_{\mathcal{W}}(K)\right|\right).

Equality holds if and only if $K$ has constant width.

The improved isoperimetric inequality, stated here as Theorem 7.1, is the main result of [42]. Its original proof relies on the Fourier series expansion of the support function of a planar convex body.

Proposition 4.2, applied with $n=k=2$ , gives a sharp global bound for the algebraic Wigner area.

Corollary 7.2 (Sharp planar Wigner-caustic area bound).

Let $K\subset\mathbb{R}^{2}$ be a convex body. Then the algebraic area of its projective Wigner hedgehog satisfies

(7.5)

-\frac{1}{4}A(K)\leq A^{*}_{\mathcal{W}}(K)\leq 0.

The constant $\frac{1}{4}$ is sharp. Equality in the left-hand inequality holds if and only if $K$ is a triangle, whereas equality in the right-hand inequality holds if and only if $K$ is centrally symmetric up to translation.

For smooth strictly convex bodies, the left-hand inequality in (7.5) is strict, since such a body cannot be a triangle. Nevertheless, the constant $\frac{1}{4}$ remains optimal, as triangles can be approximated by smooth strictly convex bodies.

8. Dimension three

Let $K\subset\mathbb{R}^{3}$ be of class $C^{2}_{+}$ . Denote by

M(K)=\int_{\partial K}H\,\mathrm{d}S

the integral of the normalized mean curvature. With the standard normalization, $S(K)=3W_{1}(K)$ , $M(K)=3W_{2}(K)$ , and $\kappa_{3}={4\uppi}/{3}$ .

8.1. The quadratic surface correction

On $\mathbb{S}^{2}$ , let $R_{h_{\mathrm{odd}}}=\det Q_{h_{\mathrm{odd}}}$ . Define the non-negative full-cover algebraic Wigner surface defect by

\mathscr{S}_{\mathcal{W}}(K)=-\int_{\mathbb{S}^{2}}R_{h_{\mathrm{odd}}}\,\mathrm{d}\sigma.

Because

3V(h_{\mathrm{odd}},h_{\mathrm{odd}},1)=\int_{\mathbb{S}^{2}}R_{h_{\mathrm{odd}}}\,\mathrm{d}\sigma,

we have

(8.1)

\mathscr{S}_{\mathcal{W}}(K)=3\mathcal{A}_{2}(K).

The projection formula (4.15) becomes

(8.2)

\mathscr{S}_{\mathcal{W}}(K)=8\int_{\operatorname{Gr}(3,2)}\left|A^{*}_{\mathcal{W}}(K\mathbin{|}E)\right|\,\mathrm{d}E.

Moreover,

(8.3)

S(C)=S(K)+\mathscr{S}_{\mathcal{W}}(K).

The quotient parameter space $\mathbb{RP}^{2}$ is non-orientable, so (8.2), rather than an unsigned area of the image, is the invariant that enters naturally.

Figure 2 is based on the axisymmetric support function

(8.4)

h(u)=2+q(u),\qquad q(u)=\varepsilon f(u_{3}),\qquad\varepsilon=0.078,

where $f(t)=16t^{5}-20t^{3}+5t$ . Since $f$ is odd, we have $h(u)+h(-u)=4$ . Thus $K$ has constant width $4$ , its even support function is the constant $c=2$ , and its central symmetral is the ball $C=2B^{3}$ .

Both $K$ and its Wigner hedgehog $x_{q}(\mathbb{S}^{2})$ are invariant under rotations about the $e_{3}$ -axis. Indeed, the two principal eigenvalues of $Q_{f}=\nabla^{2}_{\mathbb{S}^{2}}f+f\operatorname{Id}$ are $-24f(t)$ and $f(t)-tf^{\prime}(t)=-64t^{5}+40t^{3}$ , and both belong to $[-24,24]$ . Consequently, $Q_{h}=2\operatorname{Id}+\varepsilon Q_{f}\geqslant(2-24\varepsilon)\operatorname{Id}=0.128\operatorname{Id}$ , so the pictured body is smooth and strictly convex.

Let $E=\operatorname{span}\{e_{1},e_{3}\}$ be a meridional plane. Along the great circle $u(\theta)=(\cos\theta,0,\sin\theta)$ the odd part reduces to $q(u(\theta))=0.078\sin(5\theta)$ . Therefore the Wigner caustic of the orthogonal projection $K_{E}=K\mathbin{|}E$ has the five-cusped pentagram shape shown in Figure 2(B). The middle part of panel (A) is a magnified three-dimensional view of the full axisymmetric hedgehog $x_{q}(\mathbb{S}^{2})$ ; the dark red curve is its meridional generating curve.

This surface of revolution is deliberately highly nongeneric: axial symmetry forces degeneracies that are absent after a generic perturbation. Generic singularities of Wigner caustics and affine equidistants in the related Lagrangian and higher-dimensional surface settings are classified or analyzed in [7, 12, 13, 14, 15, 20].

Refer to caption — Figure 2. An axisymmetric constant-width body and its Wigner hedgehog in dimension three. (A) Left: the body $K$ (blue), its spherical central symmetral $C=2B^{3}$ (gray wireframe), the singular Wigner hedgehog $x_{q}(\mathbb{S}^{2})$ (red), and one antipodal Gauss chord. Centre: a magnified three-dimensional view of the axisymmetric hedgehog and its meridional generating curve. (B) The orthogonal projection onto the meridional plane $E=\operatorname{span}\{e_{1},e_{3}\}$ , showing $K_{E}$ , the central disk $C_{E}$ , and the pentagram-shaped planar Wigner caustic $\mathrm{E}_{1/2}(\partial K_{E})$

8.2. The volume correction

Parity gives $V(h_{\mathrm{odd}},h_{\mathrm{odd}},h_{\mathrm{odd}})=0$ and

\operatorname{Vol}_{3}(K)=\operatorname{Vol}_{3}(C)+3V(h_{\mathrm{even}},h_{\mathrm{odd}},h_{\mathrm{odd}}).

Therefore

(8.5)

\mathcal{A}_{3}(K)=-3V(h_{\mathrm{even}},h_{\mathrm{odd}},h_{\mathrm{odd}})=-\int_{\mathbb{S}^{2}}h_{\mathrm{even}}R_{h_{\mathrm{odd}}}\,\mathrm{d}\sigma.

Thus $\mathcal{A}_{3}$ is not the algebraic volume of the Wigner hedgehog. That volume vanishes by parity. It is instead the Wigner area density weighted by the even support function $h_{\mathrm{even}}$ .

Theorem 5.1 now yields two genuine refinements:

(8.6)		$\displaystyle M(K)^{2}$	$\displaystyle\geqslant 4\uppi\left(S(K)+\mathscr{S}_{\mathcal{W}}(K)\right),$
(8.7)		$\displaystyle M(K)^{3}$	$\displaystyle\geqslant 48\uppi^{2}\left(\operatorname{Vol}_{3}(K)+\mathcal{A}_{3}(K)\right).$

The corrected surface–volume inequality is

(8.8)

\left(S(K)+\mathscr{S}_{\mathcal{W}}(K)\right)^{3}\geqslant 36\uppi\left(\operatorname{Vol}_{3}(K)+\mathcal{A}_{3}(K)\right)^{2}.

All three equalities characterize constant width.

Proposition 4.2 gives

(8.9)

0\leqslant\mathscr{S}_{\mathcal{W}}(K)\leqslant\frac{1}{2}S(K),\qquad 0\leqslant\mathcal{A}_{3}(K)\leqslant\frac{3}{2}\operatorname{Vol}_{3}(K).

8.3. Constant width and the Blaschke formula

If $K$ has constant width $2r$ , then $h_{\mathrm{even}}=r$ and (8.1), (8.5) give $\mathcal{A}_{3}(K)=r\mathscr{S}_{\mathcal{W}}(K)$ . Since $C=rB^{3}$ ,

(8.10)		$\displaystyle S(K)$	$\displaystyle=4\uppi r^{2}-\mathscr{S}_{\mathcal{W}}(K),$
(8.11)		$\displaystyle\operatorname{Vol}_{3}(K)$	$\displaystyle=\frac{4\uppi}{3}r^{3}-r\mathscr{S}_{\mathcal{W}}(K).$

Eliminating $\mathscr{S}_{\mathcal{W}}$ gives the classical Blaschke relation

(8.12)

\operatorname{Vol}_{3}(K)=rS(K)-\frac{8\uppi}{3}r^{3}.

The preceding constant-width formulas are exactly formulas of Theorem 2 in [1] of Anciaux and Guilfoyle. They prove them using the different support-function calculations. In the notation used in [1, 33], the quadratic functional denoted there by $E$ is exactly $\mathscr{S}_{\mathcal{W}}(K)$ when evaluated on $h_{\mathrm{odd}}$ .

9. Dimension four

Now let $K\subset\mathbb{R}^{4}$ be of class $C^{2}_{+}$ . Since $\kappa_{4}=\uppi^{2}/2$ , define the mean width by

\overline{w}(K)=\frac{2W_{3}(K)}{\kappa_{4}}=\frac{4}{\uppi^{2}}W_{3}(K).

The three nontrivial defects are

(9.1)	$\displaystyle\mathcal{A}_{2}(K)$	$\displaystyle=-V(h_{\mathrm{odd}},h_{\mathrm{odd}},1,1),$
(9.2)	$\displaystyle\mathcal{A}_{3}(K)$	$\displaystyle=-3V(h_{\mathrm{even}},h_{\mathrm{odd}},h_{\mathrm{odd}},1),$
(9.3)	$\displaystyle\mathcal{A}_{4}(K)$	$\displaystyle=-6V(h_{\mathrm{even}},h_{\mathrm{even}},h_{\mathrm{odd}},h_{\mathrm{odd}})-V(h_{\mathrm{odd}},h_{\mathrm{odd}},h_{\mathrm{odd}},h_{\mathrm{odd}}).$

In particular,

(9.4)		$\displaystyle S(C)$	$\displaystyle=S(K)+4\mathcal{A}_{3}(K),$
(9.5)		$\displaystyle\operatorname{Vol}_{4}(C)$	$\displaystyle=\operatorname{Vol}_{4}(K)+\mathcal{A}_{4}(K).$

9.1. Projected Wigner areas and curvature

The three defects already have complementary integral-geometric meanings. Formula (4.15) specializes to

(9.6)

\mathcal{A}_{2}(K)=\uppi\int_{\operatorname{Gr}(4,2)}\left|A^{*}_{\mathcal{W}}(K\mathbin{|}E)\right|\,\mathrm{d}E.

whereas Proposition 4.1 gives

(9.7)

\mathcal{A}_{3}(K)=\frac{3\uppi}{8}\int_{\operatorname{Gr}(4,3)}\left[\operatorname{Vol}_{3}(C\mathbin{|}E)-\operatorname{Vol}_{3}(K\mathbin{|}E)\right]\,\mathrm{d}E,

and $\mathcal{A}_{4}(K)=\operatorname{Vol}_{4}(C)-\operatorname{Vol}_{4}(K)$ . Thus $\mathcal{A}_{2}$ is a mean projectively reduced Wigner area, $\mathcal{A}_{3}$ is a mean three-dimensional symmetrization gain, and $\mathcal{A}_{4}$ is the full four-dimensional gain. If $H$ denotes normalized mean curvature of $\partial K$ , then

W_{2}(K)=\frac{1}{4}\int_{\partial K}H\,\mathrm{d}S.

The $k=2$ case of the central-symmetral family in Theorem 5.1, equivalently of (5.2), therefore becomes

(9.8)

\overline{w}(K)^{2}\geqslant\frac{2}{\uppi^{2}}\int_{\partial K}H\,\mathrm{d}S+\frac{8}{\uppi}\int_{\operatorname{Gr}(4,2)}\left|A^{*}_{\mathcal{W}}(K\mathbin{|}E)\right|\,\mathrm{d}E.

The remaining two members are

(9.9)		$\displaystyle\overline{w}(K)^{3}$	$\displaystyle\geqslant\frac{4}{\uppi^{2}}\left(S(K)+4\mathcal{A}_{3}(K)\right),$
(9.10)		$\displaystyle\overline{w}(K)^{4}$	$\displaystyle\geqslant\frac{32}{\uppi^{2}}\left(\operatorname{Vol}_{4}(K)+\mathcal{A}_{4}(K)\right).$

Equality in (9.8)–(9.10) is equivalent to constant width.

The symmetral surface–volume inequality becomes

(9.11)

\left(S(K)+4\mathcal{A}_{3}(K)\right)^{4}\geqslant 128\uppi^{2}\left(\operatorname{Vol}_{4}(K)+\mathcal{A}_{4}(K)\right)^{3}.

9.2. The quartic Wigner volume

The first three quadratic mixed volumes in (9.1)–(9.3) have fixed signs:

	$\displaystyle V(h_{\mathrm{odd}},h_{\mathrm{odd}},1,1)$	$\displaystyle\leqslant 0,$
	$\displaystyle V(h_{\mathrm{even}},h_{\mathrm{odd}},h_{\mathrm{odd}},1)$	$\displaystyle\leqslant 0,$
	$\displaystyle V(h_{\mathrm{even}},h_{\mathrm{even}},h_{\mathrm{odd}},h_{\mathrm{odd}})$	$\displaystyle\leqslant 0,$

by (3.7) with $k=2,3,4$ . In contrast, $V(h_{\mathrm{odd}},h_{\mathrm{odd}},h_{\mathrm{odd}},h_{\mathrm{odd}})$ has no separate universal sign forced by the argument.

The Gauss-cover algebraic volume of the Wigner hedgehog is

(9.12)

V_{\mathcal{W}}^{\mathrm{cov}}(K):=V(h_{\mathrm{odd}},h_{\mathrm{odd}},h_{\mathrm{odd}},h_{\mathrm{odd}})=\frac{1}{4}\int_{\mathbb{S}^{3}}h_{\mathrm{odd}}\det Q_{h_{\mathrm{odd}}}\,\mathrm{d}\sigma.

Since $x_{h_{\mathrm{odd}}}(-u)=x_{h_{\mathrm{odd}}}(u)$ , it is natural to introduce the projectively reduced normalization

(9.13)

V^{*}_{\mathcal{W}}(K)=\frac{1}{2}V_{\mathcal{W}}^{\mathrm{cov}}(K).

Here $\mathbb{RP}^{3}$ is orientable and the antipodal covering $\mathbb{S}^{3}\to\mathbb{RP}^{3}$ is orientation preserving. Thus the factor $1/2$ is the natural projective normalization of the algebraic volume after fixing the induced orientation. This is an algebraic quantity with multiplicity. No regular embedded hypersurface is assumed. Formula (9.3) becomes

(9.14)

\mathcal{A}_{4}(K)=-6V(h_{\mathrm{even}},h_{\mathrm{even}},h_{\mathrm{odd}},h_{\mathrm{odd}})-2V^{*}_{\mathcal{W}}(K).

Thus the four-dimensional volume correction splits into a quadratic term weighted twice by the central symmetral and a pure projective Wigner volume.

Positivity of $\mathcal{A}_{4}$ also gives the quartic control

(9.15)

V(h_{\mathrm{odd}},h_{\mathrm{odd}},h_{\mathrm{odd}},h_{\mathrm{odd}})\leqslant-6V(h_{\mathrm{even}},h_{\mathrm{even}},h_{\mathrm{odd}},h_{\mathrm{odd}}).

Equality in (9.15) holds precisely when $K$ is centrally symmetric up to translation.

9.3. A four-dimensional Blaschke-type identity

Suppose that $K$ has constant width $2r$ . Then $h_{\mathrm{even}}=r$ and

(9.16)		$\displaystyle\mathcal{A}_{3}(K)$	$\displaystyle=3r\mathcal{A}_{2}(K),$
(9.17)		$\displaystyle\mathcal{A}_{4}(K)$	$\displaystyle=6r^{2}\mathcal{A}_{2}(K)-V(h_{\mathrm{odd}},h_{\mathrm{odd}},h_{\mathrm{odd}},h_{\mathrm{odd}}).$

Since $C=rB^{4}$ , one obtains

(9.18)		$\displaystyle S(K)$	$\displaystyle=4\kappa_{4}r^{3}-12r\mathcal{A}_{2}(K),$
(9.19)		$\displaystyle\operatorname{Vol}_{4}(K)$	$\displaystyle=\kappa_{4}r^{4}-6r^{2}\mathcal{A}_{2}(K)+V(h_{\mathrm{odd}},h_{\mathrm{odd}},h_{\mathrm{odd}},h_{\mathrm{odd}}).$

Eliminating $\mathcal{A}_{2}$ yields

(9.20)

\operatorname{Vol}_{4}(K)=\frac{r}{2}S(K)-\frac{\uppi^{2}}{2}r^{4}+V(h_{\mathrm{odd}},h_{\mathrm{odd}},h_{\mathrm{odd}},h_{\mathrm{odd}}).

Equivalently, the final term is $2V^{*}_{\mathcal{W}}(K)$ . Formula (9.20) shows exactly why the three-dimensional Blaschke relation does not extend to a linear volume–surface identity in dimension four: a genuine quartic Wigner invariant remains.

9.4. The quartic invariant assumes both signs

The absence of a sign assertion after (9.3) is essential, not merely a limitation of the argument. The following explicit computation also shows that the phenomenon persists arbitrarily close to a ball.

Lemma 9.1 (Explicit quartic Wigner volumes of opposite signs).

On $\mathbb{S}^{3}\subset\mathbb{R}^{4}$ , let

(9.21)

\displaystyle q_{-}(u)=u_{1}^{3}-\frac{1}{2}u_{1},\qquad q_{+}(u)=(u_{1}^{2}-u_{2}^{2})u_{3}-2u_{1}u_{2}u_{4}.

Both are odd spherical harmonics of degree three, and

(9.22)

\begin{array}[]{c|cc}q&-V(q,q,1,1)=\|q\|_{L^{2}(\mathbb{S}^{3})}^{2}&V(q,q,q,q)\\ \hline\cr q_{-}&\displaystyle\frac{\uppi^{2}}{32}&\displaystyle-\frac{3\uppi^{2}}{128}\\[5.0pt] q_{+}&\displaystyle\frac{\uppi^{2}}{12}&\displaystyle\frac{\uppi^{2}}{7}\end{array}

Consequently, for every $r>0$ and all sufficiently small nonzero $\varepsilon$ , each function

h_{\varepsilon,\pm}=r+\varepsilon q_{\pm}

is the support function of a $C^{\infty}$ body of class $C^{2}_{+}$ and constant width $2r$ . Its quartic invariant is $\varepsilon^{4}V(q_{\pm},q_{\pm},q_{\pm},q_{\pm})$ . Hence both signs occur among constant-width bodies arbitrarily close to $rB^{4}$ .

Proof.

The functions in (9.21) are the restrictions of the harmonic cubics

H_{-}(x)=x_{1}^{3}-\frac{1}{2}x_{1}|x|^{2},\qquad H_{+}(x)=\operatorname{Re}\bigl((x_{1}+\mathrm{i}x_{2})^{2}(x_{3}+\mathrm{i}x_{4})\bigr).

For any homogeneous cubic $H$ and $q=H\mathbin{|}\mathbb{S}^{3}$ , the spherical Hessian formula gives

(9.23)

Q_{q}=\left.P_{u}\bigl(D^{2}H(u)-2H(u)\operatorname{Id}\bigr)P_{u}\right|_{u^{\perp}},\qquad V(q^{[4]})=\frac{1}{4}\int_{\mathbb{S}^{3}}q\det Q_{q}\,\mathrm{d}\sigma,

where $P_{u}$ is orthogonal projection onto $u^{\perp}$ . This is also a direct specialization of (2.5) and (2.7).

For $q_{-}$ , put $t=u_{1}$ . The eigenvalues of $Q_{q_{-}}$ are $6t-8t^{3}$ , $-2t^{3}$ , and $-2t^{3}$ . Consequently,

\displaystyle V(q_{-}^{[4]})

\displaystyle=\int_{\mathbb{S}^{3}}\left(10t^{10}-8t^{12}-3t^{8}\right)\,\mathrm{d}\sigma=10\frac{21\uppi^{2}}{256}-8\frac{33\uppi^{2}}{512}-3\frac{7\uppi^{2}}{64}=-\frac{3\uppi^{2}}{128}.

Here we used the standard moment identity

\int_{\mathbb{S}^{3}}t^{2m}\,\mathrm{d}\sigma=2\uppi^{2}\frac{1\cdot 3\cdot\ldots\cdot(2m-1)}{4\cdot 6\cdot\ldots\cdot(2m+2)},\qquad m\geqslant 1.

The same identity gives $\|q_{-}\|_{2}^{2}=\uppi^{2}/32$ .

For completeness, the positive example can be evaluated without a computer algebra system. Use Hopf coordinates

u_{1}+\mathrm{i}u_{2}=a\,\mathrm{e}^{\mathrm{i}\alpha},\qquad u_{3}+\mathrm{i}u_{4}=b\,\mathrm{e}^{\mathrm{i}\beta},\qquad a=\cos\eta,\quad b=\sin\eta,

and put $\psi=2\alpha+\beta$ , $C_{\psi}=\cos\psi$ , $S_{\psi}=\sin\psi$ . Then $q_{+}=a^{2}bC_{\psi}$ . In the orthonormal frame $(\partial_{\eta},a^{-1}\partial_{\alpha},b^{-1}\partial_{\beta})$ , away from the two harmless coordinate circles represented by $a=0$ and $b=0$ , direct differentiation gives

(9.24)

Q_{q_{+}}=\begin{pmatrix}2b(4b^{2}-3)C_{\psi}&(4b^{2}-2)S_{\psi}&2abS_{\psi}\\ (4b^{2}-2)S_{\psi}&2b(b^{2}-2)C_{\psi}&-2aC_{\psi}\\ 2abS_{\psi}&-2aC_{\psi}&-2ba^{2}C_{\psi}\end{pmatrix}.

Writing $\det Q_{q_{+}}=C_{\psi}^{3}F(b)+C_{\psi}S_{\psi}^{2}G(b)$ , one obtains $F(b)=8b(3-4b^{2})(1-b^{2})^{3}$ and $G(b)=24b(1-b^{2})^{3}$ , and hence $3F(b)+G(b)=96b(1-b^{2})^{4}$ . With angle brackets denoting angular mean, $\langle C_{\psi}^{4}\rangle=3/8$ , $\langle C_{\psi}^{2}S_{\psi}^{2}\rangle=1/8$ , and $\,\mathrm{d}\sigma=ab\,\,\mathrm{d}\eta\,\mathrm{d}\alpha\,\mathrm{d}\beta=b\,\,\mathrm{d}b\,\mathrm{d}\alpha\,\mathrm{d}\beta$ , (9.12) gives

\displaystyle V(q_{+}^{[4]})

\displaystyle=\uppi^{2}\int_{0}^{1}12b^{3}(1-b^{2})^{5}\,\mathrm{d}b=\frac{\uppi^{2}}{7}.

The same coordinates give, explicitly,

\|q_{+}\|_{2}^{2}=2\uppi^{2}\int_{0}^{1}b^{3}(1-b^{2})^{2}\,\mathrm{d}b=\frac{\uppi^{2}}{12}.

Finally, the first equality in the middle column of (9.22) follows from (6.4), because degree three has eigenvalue $15$ on $\mathbb{S}^{3}$ .

The continuous fields $Q_{q_{\pm}}$ are bounded. Put

M_{\pm}=\max_{\mathbb{S}^{3}}\|Q_{q_{\pm}}\|_{\mathrm{op}}.

Then $Q_{h_{\varepsilon,\pm}}=r\operatorname{Id}+\varepsilon Q_{q_{\pm}}$ is positive definite whenever $|\varepsilon|<r/M_{\pm}$ . Oddness gives $h_{\varepsilon,\pm}(u)+h_{\varepsilon,\pm}(-u)=2r$ , completing the convexity and constant-width assertions. ∎

Proposition 9.2 (The exact cubic-harmonic plane).

Let $f=q_{-}$ and $g=q_{+}$ be given by (9.21), put $q_{\alpha,\beta}=\alpha f+\beta g$ , and write $\mathscr{E}(q)=-V(q,q,1,1)$ . Then

(9.25)		$\displaystyle\mathscr{E}(q_{\alpha,\beta})$	$\displaystyle=\uppi^{2}\left(\frac{\alpha^{2}}{32}+\frac{\beta^{2}}{12}\right),$
(9.26)		$\displaystyle V(q_{\alpha,\beta}^{[4]})$	$\displaystyle=\uppi^{2}\left(-\frac{3\alpha^{4}}{128}+\frac{3\alpha^{2}\beta^{2}}{70}+\frac{\beta^{4}}{7}\right).$

For every nonzero $q_{\alpha,\beta}$ ,

(9.27)

-\frac{24}{\uppi^{2}}\mathscr{E}(q_{\alpha,\beta})^{2}\leqslant V(q_{\alpha,\beta}^{[4]})\leqslant\frac{144}{7\uppi^{2}}\mathscr{E}(q_{\alpha,\beta})^{2}.

Both constants are optimal on this plane: equality on the left occurs exactly on the $f$ -axis and equality on the right exactly on the $g$ -axis. Furthermore,

(9.28)

V(q_{\alpha,\beta}^{[4]})=0,\qquad\frac{\beta^{2}}{\alpha^{2}}=\frac{\sqrt{1194}-12}{80},

describes the two nonzero zero lines. Consequently, for every $r>0$ there are nonspherical $C^{\infty}$ bodies of constant width $2r$ , arbitrarily close to $rB^{4}$ , whose quartic Wigner volume vanishes.

Proof.

The orthogonal map $(u_{1},u_{2},u_{3},u_{4})\mapsto(u_{1},u_{2},-u_{3},-u_{4})$ fixes $f$ and sends $g$ to $-g$ . Orthogonal invariance and multilinearity of mixed volume therefore eliminate $V(f^{[3]},g)$ and $V(f,g^{[3]})$ . The same symmetry gives $\langle f,g\rangle_{L^{2}}=0$ . Since both functions have degree three, (6.4) and (9.22) prove (9.25).

It remains to compute the even cross term. Let $F$ and $G$ be the homogeneous harmonic cubics restricting to $f$ and $g$ , respectively, and set

A_{s}(u)=D^{2}(F+sG)(u)-2(F+sG)(u)\operatorname{Id}.

For every symmetric $4\times 4$ matrix $A$ and every unit vector $u$ , the determinant of its bilinear restriction to $u^{\perp}$ is $u^{\top}\operatorname{adj}(A)u$ . Thus (9.23), followed by direct polynomial expansion, gives

(9.29)

\frac{1}{4\uppi^{2}}\int_{\mathbb{S}^{3}}(F+sG)\,u^{\top}\operatorname{adj}(A_{s})u\,\mathrm{d}\sigma=-\frac{3}{128}+\frac{3}{70}s^{2}+\frac{1}{7}s^{4}.

For transparency, every coefficient in this expansion follows from the elementary moment formula

(9.30)

\frac{1}{\uppi^{2}}\int_{\mathbb{S}^{3}}u_{1}^{2a_{1}}u_{2}^{2a_{2}}u_{3}^{2a_{3}}u_{4}^{2a_{4}}\,\mathrm{d}\sigma=\frac{2}{(a_{1}+a_{2}+a_{3}+a_{4}+1)!}\prod_{j=1}^{4}\frac{(2a_{j})!}{4^{a_{j}}a_{j}!}.

Monomials with an odd exponent integrate to zero. The identity follows directly by writing a Gaussian integral in Cartesian and polar coordinates. In particular, the coefficient of $s^{2}$ in (9.29) says $6V(f,f,g,g)=3\uppi^{2}/70$ . Polarization and the two pure values in (9.22) now prove (9.26).

For the normalized estimate, put $x=\beta^{2}/\alpha^{2}\geqslant 0$ when $\alpha\neq 0$ . After cancelling the common scale, the ratio $\uppi^{2}V(q_{\alpha,\beta}^{[4]})/\mathscr{E}(q_{\alpha,\beta})^{2}$ equals

R(x)=\frac{-3/128+3x/70+x^{2}/7}{(1/32+x/12)^{2}}.

Its derivative has the sign of ${47}/{8960}+{3x}/{560}$ , so $R$ is strictly increasing on $[0,\infty)$ . Since $R(0)=-24$ and $R(\infty)=144/7$ , this proves (9.27) and its equality statements. Finally, the numerator of (9.26) vanishes when $640x^{2}+192x-105=0$ . Its unique positive root is the value in (9.28). The strict-convexity argument at the end of Lemma 9.1 applies to any fixed linear combination $q_{\alpha,\beta}$ and proves the final assertion. ∎

9.5. An outer bound for the quadratic–quartic feasible region

Although Lemma 9.1 rules out an independent sign for $V(h_{\mathrm{odd}}^{[4]})$ , convexity still couples it nontrivially to the quadratic energy.

Theorem 9.3 (Necessary outer regions at constant width).

Let $K\subset\mathbb{R}^{4}$ be a convex body of constant width $2r$ , let $C$ be its central symmetral, and set

(9.31)

a=\mathcal{A}_{2}(K),\qquad b=V(h_{\mathrm{odd}}^{[4]}),\qquad R=\kappa_{4}r^{2}.

Then

(9.32)

\displaystyle W_{2}(K)=R-a,\quad W_{1}(K)=r(R-3a),\quad\text{and}\quad\operatorname{Vol}_{4}(K)=r^{2}(R-6a)+b.

The classical Alexandrov–Fenchel and Rogers–Shephard inequalities alone give the necessary outer region

(9.33)

0\leqslant a<x_{*}R,\qquad 6r^{2}a-\frac{27}{35}\kappa_{4}r^{4}<b\leqslant\frac{r^{2}a(R+3a)}{R-a}\leqslant 2r^{2}a,

where

(9.34)

x_{*}=\frac{101-4\sqrt{106}}{315}=0.189896\ldots.

The $j=1$ Godbersen inequality gives the additional strict lower bound

(9.35)

\frac{24}{5}r^{2}a-\frac{3}{5}\kappa_{4}r^{4}<b.

Combining (9.35) with the three-dimensional projection bound (4.12) strengthens the outer region to

(9.36)

0\leqslant a\leqslant\frac{R}{11},\qquad\frac{24}{5}r^{2}a-\frac{3}{5}\kappa_{4}r^{4}<b\leqslant\frac{r^{2}a(R+3a)}{R-a}\leqslant\frac{7}{5}r^{2}a.

Equality in the Alexandrov–Fenchel upper bound occurs only for a ball. Moreover, $a=0$ holds if and only if $K$ is a Euclidean ball up to translation, in which case $b=0$ .

Proof.

Equations (9.32) follow from Lemma 2.3 and $V(h_{\mathrm{odd}},h_{\mathrm{odd}},1,1)=-a$ , after expanding $V\bigl((r+h_{\mathrm{odd}})^{[4-j]},1^{[j]}\bigr)$ for $j=0,1,2,3$ .

First, (4.3) with $k=3$ , which is the projected Rogers–Shephard inequality [35, Theorem 1], gives

3ra=\mathcal{A}_{3}(K)\leqslant\frac{3}{2}W_{1}(K)=\frac{3}{2}r(R-3a),

so $a\leqslant R/5$ . Next, the adjacent Alexandrov–Fenchel inequality

(9.37)

W_{1}(K)^{2}\geqslant W_{0}(K)W_{2}(K),

which is [39, Theorem 7.3.1] with two unit-ball arguments, together with (9.32), yields

(9.38)

b\leqslant\frac{r^{2}a(R+3a)}{R-a}\leqslant 2r^{2}a.

The second inequality uses $a\leqslant R/5$ .

Likewise, (4.3) with $k=4$ gives

6r^{2}a-b=\mathcal{A}_{4}(K)\leqslant\frac{27}{8}\operatorname{Vol}_{4}(K).

Since $\operatorname{Vol}_{4}(K)=\kappa_{4}r^{4}-\mathcal{A}_{4}(K)$ , this is equivalent to the lower bound for $b$ in (9.33). Put $x=a/R$ . Compatibility of that lower bound with the first upper bound in (9.38) implies

(9.39)

315x^{2}-202x+27\geqslant 0.

Together with $0\leqslant x\leqslant 1/5$ , this forces $x\leqslant x_{*}$ , where $x_{*}$ is the smaller root in (9.34). Equality $x=x_{*}$ would make the lower and upper bounds for $b$ coincide and hence force equality in (9.37). The equality case of the Alexandrov–Fenchel inequality [39, Theorems 7.4.6 and 7.6.9] would then make $K$ a ball, which has $a=0$ . Therefore $a<x_{*}R$ , proving (9.33).

The equality statement in the Rogers–Shephard difference-body inequality says that equality is possible only for a simplex [35, Theorem 1]. A body of constant width $2r$ satisfies $K-K=2rB^{4}$ and is strictly convex: otherwise a non-singleton support face $F(K,u)$ , after subtracting any point of $F(K,-u)$ , would give a non-singleton support face of the ball $K-K$ . A simplex is not strictly convex. Hence the lower bound for $b$ is strict. Likewise, the equality case cited for (9.37) proves that the upper bound is strict unless $K$ is a ball.

The case $j=1$ of (4.8) reads in dimension four

V(-K,K,K,K)\leqslant 4\operatorname{Vol}_{4}(K).

Indeed, after translating the centroid of $K$ to the origin, the classical inclusion $-K\subset 4K$ and monotonicity of mixed volume give this estimate. Schneider’s stability theorem [38] supplies the equality characterization: a full-dimensional equality case is a simplex. Since $K$ has constant width and $h_{K}=r+h_{\mathrm{odd}}$ , parity gives

V(-K,K,K,K)=\kappa_{4}r^{4}-b.

Substitution of the last identity and $\operatorname{Vol}_{4}(K)=\kappa_{4}r^{4}-6r^{2}a+b$ yields (9.35). Equality in the Godbersen inequality would force $K$ to be a simplex, so the lower bound is strict for a constant-width body.

For the stronger projection-derived bound, combine (9.16) with (4.12):

3ra=\mathcal{A}_{3}(K)\leqslant\frac{3}{11}\kappa_{4}r^{3},

and hence $a\leqslant R/11$ . Substitution into the last quotient in (9.38) gives $(R+3a)/(R-a)\leqslant 7/5$ , proving the upper bounds in (9.36). On the same range, the difference between the Godbersen lower bound and the Rogers–Shephard lower bound is

\frac{6}{35}r^{2}(R-7a)>0.

Thus (9.35) supplies the lower bound in (9.36). Finally, Theorem 3.3 shows that $a=0$ is equivalent to central symmetry up to translation. Central symmetry together with constant width forces a ball. Its odd support component is linear, so (2.8) gives $b=0$ . ∎

10. Constant-width expansions in arbitrary dimension

The preceding low-dimensional identities are instances of one general formula. Suppose that $K\subset\mathbb{R}^{n}$ has constant width $2r$ , so that $h=r+h_{\mathrm{odd}}$ , where $h_{\mathrm{odd}}(-u)=-h_{\mathrm{odd}}(u)$ . For $p=0,\ldots,\lfloor n/2\rfloor$ , define the even Wigner quermassintegrals

(10.1)

\mathcal{Q}_{2p}(h_{\mathrm{odd}})=V\left(h_{\mathrm{odd}}^{[2p]},1^{[n-2p]}\right),\qquad\mathcal{Q}_{0}(h_{\mathrm{odd}})=\kappa_{n}.

The quadratic term satisfies $\mathcal{Q}_{2}(h_{\mathrm{odd}})=-\mathcal{A}_{2}(K)\leqslant 0$ , whereas no independent sign is available for all higher terms: Lemma 9.1 gives both signs already for $\mathcal{Q}_{4}$ in dimension four.

Theorem 10.1 (All quermassintegrals at constant width).

Let $K\subset\mathbb{R}^{n}$ be a convex body of constant width $2r$ , and write $h_{K}=r+h_{\mathrm{odd}}$ with $h_{\mathrm{odd}}$ odd. For $j=0,\ldots,n$ ,

(10.2)

W_{j}(K)=\sum_{p=0}^{\lfloor(n-j)/2\rfloor}\binom{n-j}{2p}r^{\,n-j-2p}\mathcal{Q}_{2p}(h_{\mathrm{odd}}).

For $n\geqslant 3$ , volume and surface area satisfy

(10.3)

\displaystyle\operatorname{Vol}_{n}(K)=\frac{r}{n-2}S(K)-\frac{2\kappa_{n}}{n-2}r^{n}+\frac{2}{n-2}\sum_{p=2}^{\lfloor n/2\rfloor}(p-1)\binom{n}{2p}r^{\,n-2p}\mathcal{Q}_{2p}(h_{\mathrm{odd}}).

Proof.

Expand

W_{j}(K)=V\left((r+h_{\mathrm{odd}})^{[n-j]},1^{[j]}\right)

by multilinearity. Lemma 2.3 removes the odd powers of $h_{\mathrm{odd}}$ , giving (10.2).

For $j=0$ and $j=1$ , formula (10.2) gives

\displaystyle\operatorname{Vol}_{n}(K)=\sum_{p=0}^{\lfloor n/2\rfloor}\binom{n}{2p}r^{n-2p}\mathcal{Q}_{2p}(h_{\mathrm{odd}})\quad\text{and}\quad S(K)=n\sum_{p=0}^{\lfloor(n-1)/2\rfloor}\binom{n-1}{2p}r^{n-1-2p}\mathcal{Q}_{2p}(h_{\mathrm{odd}}).

Use $n\binom{n-1}{2p}=(n-2p)\binom{n}{2p}$ . The constant term and the quadratic term are isolated by the first two terms on the right-hand side of (10.3). The remaining coefficient is

1-\frac{n-2p}{n-2}=\frac{2(p-1)}{n-2}.

This proves (10.3). ∎

For $n=3$ , the sum in (10.3) is empty and one recovers (8.12). For $n=4$ , only $p=2$ remains, and (10.3) is exactly (9.20).

Corollary 10.2 (The even-dimensional residual).

Let $K\subset\mathbb{R}^{n}$ be a convex body of constant width $2r$ , and write $h_{K}=r+h_{\mathrm{odd}}$ with $h_{\mathrm{odd}}$ odd. If $n=2m\geqslant 4$ , then

(10.4)

\displaystyle\operatorname{Vol}_{n}(K)=\frac{r}{n-2}S(K)-\frac{2\kappa_{n}}{n-2}r^{n}+\frac{2}{n-2}\sum_{p=2}^{m-1}(p-1)\binom{n}{2p}r^{n-2p}\mathcal{Q}_{2p}(h_{\mathrm{odd}})+V(h_{\mathrm{odd}}^{[n]}).

The final summand is the Gauss-cover algebraic $n$ -volume of the Wigner hedgehog and occurs with coefficient one. Its projectively reduced normalization is $\frac{1}{2}V(h_{\mathrm{odd}}^{[n]})$ , because $\mathbb{RP}^{n-1}$ is orientable and the antipodal cover is orientation preserving. If $n$ is odd, then $V(h_{\mathrm{odd}}^{[n]})=0$ by parity and $\mathbb{RP}^{n-1}$ is non-orientable.

Proof.

In (10.3), the coefficient of the last term $p=m$ is $\frac{2(m-1)}{2m-2}\binom{2m}{2m}=1$ . The orientation assertions use the standard criterion that $\mathbb{RP}^{d}$ is orientable exactly when $d$ is odd. Equivalently, the antipodal map on $\mathbb{S}^{d}$ has degree $(-1)^{d+1}$ . For odd $n$ , Lemma 2.3 annihilates $V(h_{\mathrm{odd}}^{[n]})$ . ∎

The four-dimensional term in (9.20) is the first non-planar example of this residual, and Lemma 9.1 shows that it is sign-indefinite even on arbitrarily small constant-width perturbations of a ball.

Declarations

Funding

The author received no financial support for the research, authorship, or publication of this article.

Conflict of interest

The author declares that there are no competing interests.

Data availability

No data were generated or analyzed in this theoretical study.

References

[1] H. Anciaux and B. Guilfoyle, On the three-dimensional Blaschke–Lebesgue problem, Proc. Amer. Math. Soc. 139 (2011), 1831–1839, doi:10.1090/S0002-9939-2010-10588-9.
[2] M. V. Berry, Semi-classical mechanics in phase space: a study of Wigner’s function, Philos. Trans. Roy. Soc. London Ser. A 287 (1977), 237–271, doi:10.1098/rsta.1977.0145.
[3] B. Bogosel, Mixed volumes and the Blaschke–Lebesgue theorem, Acta Math. Hungar. 173 (2024), 122–138, doi:10.1007/s10474-024-01435-w.
[4] B. Bogosel, New variational arguments regarding the Blaschke–Lebesgue theorem, Expo. Math. 43 (2025), article 125700, doi:10.1016/j.exmath.2025.125700.
[5] B. Bogosel, A geometric proof of the Blaschke–Lebesgue theorem for the Cheeger constant, Ann. Inst. Fourier (Grenoble) 76 (2026), no. 3, 1035–1053, doi:10.5802/aif.3728.
[6] K. J. Böröczky, F. Fodor and D. Hug, Strengthened inequalities for the mean width and the $\ell$ -norm of origin symmetric convex bodies, Math. Ann. 393 (2025), 271–316, doi:10.1007/s00208-025-03228-0.
[7] M. Craizer, W. Domitrz and P. de M. Rios, Singular improper affine spheres from a given Lagrangian submanifold, Adv. Math. 374 (2020), article 107326, doi:10.1016/j.aim.2020.107326.
[8] M. Craizer, R. C. Teixeira and M. A. H. B. da Silva, Polygons with parallel opposite sides, Discrete Comput. Geom. 50 (2013), 474–490, doi:10.1007/s00454-013-9520-2.
[9] J. Cufí, E. Gallego and A. Reventós, A note on Hurwitz’s inequality, J. Math. Anal. Appl. 458 (2018), no. 1, 436–451, doi:10.1016/j.jmaa.2017.09.017.
[10] I. Danielewska, D. Poławski, D. Sterczewska and M. Zwierzyński, Artistic aspects of the Wigner caustic and the centre symmetry set, J. Math. Arts 19 (2025), nos. 3–4, 119–135, doi:10.1080/17513472.2025.2576841.
[11] L. R. G. Dias, M. Farnik and Z. Jelonek, Generic symmetry defect set of an algebraic curve, Proc. Amer. Math. Soc. 152 (2024), no. 7, 2739–2749, doi:10.1090/proc/16741.
[12] W. Domitrz, S. Janeczko, P. de M. Rios and M. A. S. Ruas, Singularities of affine equidistants: extrinsic geometry of surfaces in $4$ -space, Bull. Braz. Math. Soc. (N.S.) 47 (2016), 1155–1179, doi:10.1007/s00574-016-0208-0.
[13] W. Domitrz, M. Manoel and P. de M. Rios, The Wigner caustic on shell and singularities of odd functions, J. Geom. Phys. 71 (2013), 58–72, doi:10.1016/j.geomphys.2013.04.005.
[14] W. Domitrz and P. de M. Rios, Singularities of equidistants and global centre symmetry sets of Lagrangian submanifolds, Geom. Dedicata 169 (2014), 361–382, doi:10.1007/s10711-013-9861-2.
[15] W. Domitrz, P. de M. Rios and M. A. S. Ruas, Singularities of affine equidistants: projections and contacts, J. Singul. 10 (2014), 67–81, doi:10.5427/jsing.2014.10d.
[16] W. Domitrz, M. C. Romero Fuster and M. Zwierzyński, The geometry of the secant caustic of a planar curve, Differ. Geom. Appl. 78 (2021), article 101797, doi:10.1016/j.difgeo.2021.101797.
[17] W. Domitrz and M. Zwierzyński, The geometry of the Wigner caustic and a decomposition of a curve into parallel arcs, Anal. Math. Phys. 12 (2022), article 7, doi:10.1007/s13324-021-00617-x.
[18] W. Domitrz and M. Zwierzyński, Singular points of the Wigner caustic and affine equidistants of planar curves, Bull. Braz. Math. Soc. (N.S.) 51 (2020), 11–26, doi:10.1007/s00574-019-00141-4.
[19] P. J. Giblin and P. A. Holtom, The centre symmetry set, Banach Center Publ. 50 (1999), 91–105, doi:10.4064/-50-1-91-105.
[20] P. J. Giblin and G. Reeve, Equidistants for families of surfaces, J. Singul. 21 (2020), 97–118, doi:10.5427/jsing.2020.21f.
[21] J. Haddad, D. Langharst, E. Putterman, M. Roysdon and D. Ye, Affine isoperimetric inequalities for higher-order projection and centroid bodies, Math. Ann. 393 (2025), 1073–1121, doi:10.1007/s00208-025-03271-x.
[22] S. Janeczko, Bifurcations of the center of symmetry, Geom. Dedicata 60 (1996), 9–16, doi:10.1007/BF00150864.
[23] S. Janeczko, Z. Jelonek and M. A. S. Ruas, Symmetry defect of algebraic varieties, Asian J. Math. 18 (2014), no. 3, 525–544, doi:10.4310/AJM.2014.v18.n3.a8.
[24] I. Konicer, B. Murawski, B. Rogala, T. Wieczorek and M. Zwierzyński, Convex polygons with parallel opposite sides: convergence, reconstruction, and isoperimetric inequalities, arXiv:2606.22612 (2026), doi:10.48550/arXiv.2606.22612.
[25] J. Kotrbatý, On a generalization of Godbersen’s conjecture, Int. Math. Res. Not. IMRN 2025 (2025), no. 22, article rnaf341, 11 pp., doi:10.1093/imrn/rnaf341.
[26] J. Kotrbatý and M. A. Mouamine, Godbersen’s conjecture and the $L_{p}$ -Rogers–Shephard inequality, arXiv:2607.20387 (2026), doi:10.48550/arXiv.2607.20387.
[27] K.-K. Kwong, A unified approach to higher order discrete and smooth isoperimetric inequalities, Mathematika 70 (2024), no. 1, article e12238, doi:10.1112/mtk.12238.
[28] R. Langevin, G. Levitt and H. Rosenberg, Hérissons et multihérissons (enveloppes paramétrées par leur application de Gauss), in: Singularities (Warsaw, 1985), Banach Center Publ. 20, PWN, Warsaw, 1988, 245–253.
[29] Y. Martinez-Maure, Hedgehog Theory, Lecture Notes in Math., vol. 2381, Springer, Cham, 2026, doi:10.1007/978-3-032-04298-9.
[30] H. Martini, L. Montejano and D. Oliveros, Bodies of Constant Width: An Introduction to Convex Geometry with Applications, Birkhäuser/Springer, Cham, 2019, doi:10.1007/978-3-030-03868-7.
[31] D. Miller and M. Zwierzyński, The geometry of the Centre Symmetry Set of a planar curve, J. Topol. Anal. 18 (2026), no. 3, 713–738, doi:10.1142/S1793525324500444.
[32] C. Müller, Spherical Harmonics, Lecture Notes in Math., vol. 17, Springer-Verlag, Berlin, 1966, doi:10.1007/BFb0094775.
[33] A. Nishioka, An improved lower bound for the three-dimensional Blaschke–Lebesgue problem from spectral and dual perspectives, arXiv:2606.01754 (2026), doi:10.48550/arXiv.2606.01754.
[34] D. Rochera, Offsets and front tire tracks to projective hedgehogs, Comput. Aided Geom. Design 97 (2022), article 102135, doi:10.1016/j.cagd.2022.102135.
[35] C. A. Rogers and G. C. Shephard, The difference body of a convex body, Arch. Math. (Basel) 8 (1957), 220–233, doi:10.1007/BF01899997.
[36] M. F. Safarewicz and M. Zwierzyński, The $k$ th order preserving sets and isoperimetric-type inequalities for planar ovals, J. Geom. Anal. 36 (2026), article 125, doi:10.1007/s12220-026-02372-6.
[37] R. S. dos Santos and M. Craizer, Isoperimetric inequalities in normed planes, J. Convex Anal. 29 (2022), no. 2, 321–331.
[38] R. Schneider, Stability for some extremal properties of the simplex, J. Geom. 96 (2009), nos. 1–2, 135–148, doi:10.1007/s00022-010-0028-0.
[39] R. Schneider, Convex Bodies: The Brunn–Minkowski Theory, 2nd expanded ed., Encyclopedia of Mathematics and its Applications, vol. 151, Cambridge University Press, Cambridge, 2014, doi:10.1017/CBO9781139003858.
[40] R. Schneider, The middle hedgehog of a planar convex body, Beitr. Algebra Geom. 58 (2017), 235–245, doi:10.1007/s13366-016-0315-5.
[41] D. Zhang, The lower bounds of the mixed isoperimetric deficit, Bull. Malays. Math. Sci. Soc. 44 (2021), no. 5, 2863–2872, doi:10.1007/s40840-020-01067-7.
[42] M. Zwierzyński, The improved isoperimetric inequality and the Wigner caustic of planar ovals, J. Math. Anal. Appl. 442 (2016), 726–739, doi:10.1016/j.jmaa.2016.05.016.
[43] M. Zwierzyński, Isoperimetric equalities for rosettes, Internat. J. Math. 31 (2020), no. 5, article 2050041, doi:10.1142/S0129167X2050041X.
[44] M. Zwierzyński, The Constant Width Measure Set, the Spherical Measure Set and isoperimetric equalities for planar ovals, Commun. Anal. Geom. 33 (2025), no. 6, 1271–1303, doi:10.4310/CAG.251115002100.

Mixed Volumes, the Wigner Caustic, and Isoperimetric Inequalities

Abstract.

Key words and phrases:

2020 Mathematics Subject Classification

1. Introduction

2. Support functions, hedgehogs, and mixed volumes

2.1. Convex bodies and quermassintegrals

2.2. The even–odd decomposition

Definition 2.1 (Wigner caustic).

Proposition 2.2 (Central symmetral and Wigner hedgehog).

Proof.

2.3. Parity

Lemma 2.3 (Parity of mixed volumes).

Proof.

2.4. Regularity and approximation

3. Symmetrization defects and their exact expansion

Definition 3.1.

Theorem 3.2 (Exact Wigner expansion).

Proof.

Theorem 3.3 (Positivity, rigidity, and the quadratic response).

Proof.

Remark 3.4.

4. Projection formulas and Rogers–Shephard bounds

Proposition 4.1 (Projection formula).

Proposition 4.2 (Two-sided universal bound).

Proof.

Proposition 4.3 (Difference-body coefficients in Wigner variables).

Proof.

4.1. Transfer of lower volume bounds through projections

Theorem 4.4 (Projection-transfer principle).

Proof.

4.2. The projected Wigner-area formula

Theorem 4.5 (Quadratic Wigner projection formula).

Proof.

5. Strengthened isoperimetric inequalities from the central symmetral

Theorem 5.1 (Central-symmetral family of strengthened isoperimetric inequalities).

Proof.

5.1. The corrected Alexandrov–Fenchel chain

Proposition 5.2.

Proof.

5.2. The classical isoperimetric inequality applied to CC

Theorem 5.3 (Symmetral surface–volume inequality).

Proof.

6. The universal quadratic Wigner energy

Theorem 6.1 (Spectral formula and rigidity).

Proof.

Corollary 6.2 (Quantitative central-asymmetry estimate).

Proof.

7. Dimension two

Theorem 7.1.

Corollary 7.2 (Sharp planar Wigner-caustic area bound).

8. Dimension three

8.1. The quadratic surface correction

8.2. The volume correction

8.3. Constant width and the Blaschke formula

9. Dimension four

9.1. Projected Wigner areas and curvature

9.2. The quartic Wigner volume

9.3. A four-dimensional Blaschke-type identity

9.4. The quartic invariant assumes both signs

Lemma 9.1 (Explicit quartic Wigner volumes of opposite signs).

Proof.

Proposition 9.2 (The exact cubic-harmonic plane).

Proof.

9.5. An outer bound for the quadratic–quartic feasible region

Theorem 9.3 (Necessary outer regions at constant width).

Proof.

10. Constant-width expansions in arbitrary dimension

Theorem 10.1 (All quermassintegrals at constant width).

Proof.

Corollary 10.2 (The even-dimensional residual).

Proof.

Declarations

Funding

Conflict of interest

Data availability

References

5.2. The classical isoperimetric inequality applied to $C$