betaFeedback

WorksMathematics

A Sharp Perimeter Inequality for Mean Distances on Convex Curves

Abstract

For a planar convex body KK, let θ(K)θ(K) denote the expected Euclidean distance between two independent points uniformly distributed on ∂K\partial K with respect to arclength. We prove the sharp inequality θ(K)per⁡K>16.\frac{θ(K)}{\operatorname{per}K}>\frac{1}{6}. The constant is approached by rectangles degenerating to a segment.

Introduction

Let K⊂R2K \subset\mathbb{R}^2 be a convex body, and let B1,B2B_1,B_2 be independent random points uniformly distributed on ∂K\partial K with respect to arclength. Put

θ(K)=E∣B1−B2∣,\theta(K) = \mathbb{E}|B_1-B_2|,

and denote the perimeter of KK by per⁡K\operatorname{per} K.

According to Tokmachev [3 Eq. (5)], Gusakova and Zaporozhets proposed the pair of sharp inequalities

16<θ(K)per⁡K≤2π2.\frac{1}{6} < \frac{\theta(K)}{\operatorname{per} K} \le\frac{2}{\pi^2}.

The upper bound, with equality only for the disk, was proved by Tokmachev [2]. The purpose of the present paper is to establish the complementary lower bound.

Theorem 1.1. For every planar convex body KK,

θ(K)>per⁡K6.\theta(K) > \frac{\operatorname{per} K}{6}.

The constant 1/6 is sharp.

Our result is closely related to the Zaporozhets–Tarasov conjecture, which concerns the comparison of mean distances between independent uniform points inside a convex body and on its boundary. Bonnet, Gusakova, Thäle and Zaporozhets [1] proved that the mean distance Δ(K)\Delta(K) between two independent uniform points in a planar convex body satisfies

Δ(K)<per⁡K6.\Delta(K) < \frac{\operatorname{per} K}{6}.

Consequently, Theorem 1.1 gives another proof of the planar Zaporozhets–Tarasov inequality Δ(K)<θ(K)\Delta(K) < \theta(K). This inequality was previously established by a different approach in [5], where a stronger directional comparison was obtained.

Our proof is based on a comparison of one-dimensional distributions associated with portions of ∂K\partial K visible from a certain direction, which we call half-boundaries. We consider the tangent angle of a uniformly sampled point on a half-boundary and the median of three independent such angles. Using the concavity of the Gini mean difference under mixtures, we first bound θ(K)\theta(K) from below by an integral involving the median distribution. A probabilistic comparison then allows us to replace the median by a single tangent angle, after which Cauchy’s perimeter formula gives the desired constant 1/61/6. Finally, we examine the equality conditions to establish strictness.

We first recall the central notion and its basic properties.

Definition 1.2. For a probability measure ρ\rho on R\mathbb{R}, write

D(ρ)=∬R2∣x−y∣ dρ(x)dρ(y).D(\rho) = \iint_{\mathbb{R}^{2}} |x-y|\,d\rho(x)d\rho(y).

This is the Gini mean difference of ρ\rho.

We use the following standard identities for the Gini mean difference; see also [4 Sec. 2.1.1].

Lemma 1.3. Let ρ,σ\rho,\sigma be probability measures on R\mathbb{R} with finite first moments, and let F,GF,G be their distribution functions. Then

D(ρ)=2∫RF(x)(1−F(x)) dx.(1)D(\rho) = 2\int_{\mathbb{R}} F(x)(1-F(x))\,dx. \tag*{(1)}

Moreover, for 0≤λ≤10 \leq\lambda\leq1,

D(λρ+(1−λ)σ)=λD(ρ)+(1−λ)D(σ)+2λ(1−λ)∫R(F−G)2 dx.(2)D(\lambda\rho+ (1-\lambda)\sigma) = \lambda D(\rho) + (1-\lambda)D(\sigma) + 2\lambda(1-\lambda)\int_{\mathbb{R}}(F-G)^2\,dx. \tag*{(2)}

Finally, if XX has an atomless distribution ρ\rho, then

D(ρ)=4Cov⁡(X,F(X)).(3)D(\rho) = 4\operatorname{Cov}(X,F(X)). \tag*{(3)}

Proof. Let X,X′X,X' be independent with law ρ\rho. Since

∣X−X′∣=∫R∣1{X≤x}−1{X′≤x}∣ dx,|X-X'| = \int_{\mathbb{R}} \left|\mathbf{1}_{\{X\leq x\}}-\mathbf{1}_{\{X'\leq x\}}\right|\,dx,

taking expectations and applying Tonelli’s theorem gives (1.1). The distribution function of λρ+(1−λ)σ\lambda\rho+ (1-\lambda)\sigma is H=λF+(1−λ)GH = \lambda F + (1-\lambda)G. The identity

H(1−H)=λF(1−F)+(1−λ)G(1−G)+λ(1−λ)(F−G)2H(1-H) = \lambda F(1-F) + (1-\lambda)G(1-G) + \lambda(1-\lambda)(F-G)^2

then proves (1.2).

Finally, if ρ\rho is atomless, then F(X)F(X) is uniform on [0,1][0,1]. By independence and symmetry,

2Cov⁡(X,F(X))=2E[X1{X′≤X}]−EX=E[(X−X′)1{X′≤X}]=12E∣X−X′∣.2\operatorname{Cov}(X,F(X)) = 2\mathbb{E}[X\mathbf{1}_{\{X'\le X\}}] - \mathbb{E}X = \mathbb{E}[(X-X')\mathbf{1}_{\{X'\le X\}}] = \frac{1}{2}\mathbb{E}|X-X'|.

This proves (1.3).

The lower bound

The argument we use to prove Theorem 1.1 is a comparison of one-dimensional distributions. The relevant laws are obtained by choosing either one random tangent direction or the median of three random tangent directions on a moving half-boundary.

Orient ∂K\partial K counterclockwise and parametrize it by arclength s∈[0,P]s\in[0,P], where P=per⁡KP=\operatorname{per}K:

r:[0,P]⟶∂K.r:[0,P]\longrightarrow\partial K.
r′(s)=(cos⁡α(s),sin⁡α(s))for almost every s.r'(s)=(\cos\alpha(s),\sin\alpha(s)) \qquad\text{for almost every }s.

Its tangent angle α(s)\alpha(s) is defined almost everywhere and admits a nondecreasing lift. Let μ\mu be the pushforward of arclength under s↦α(s)s\mapsto\alpha(s), and put

P=μ(S1),S(t)=μ([t,t+π]).P=\mu(S^1), \qquad S(t)=\mu([t,t+\pi]).

For each tt, let It∈[0,π]I_t\in[0,\pi] have the normalized restriction of μ\mu to [t,t+π][t,t+\pi], shifted by −t-t, and let MtM_t be the median of three independent copies of ItI_t. If ∂K\partial K is C2C^2 with positive curvature, then μ(dα)=τ(α) dα\mu(d\alpha)=\tau(\alpha)\,d\alpha, where τ\tau is the radius of curvature.

Proposition 2.1. With the preceding notation,

θ(K)≥16P∫02πS(t)2Esin⁡Mt dt.(4)\theta(K)\ge\frac{1}{6P}\int_0^{2\pi}S(t)^2\mathbb{E}\sin M_t\,dt. \tag*{(4)}

Proof. Fix tt such that μ({t})=μ({t+π})=0\mu(\{t\})=\mu(\{t+\pi\})=0. Such tt form a set of full Lebesgue measure, since μ\mu has at most countably many atoms. Project the body onto

ut=(−sin⁡t,cos⁡t).u_t=(-\sin t,\cos t).

The boundary ∂K\partial K splits into two arcs joining the two points where the projection onto utu_t attains its minimum and maximum. Let Γt\Gamma_t and Γt+π\Gamma_{t+\pi} be the two boundary arcs whose tangent angles range over [t,t+π][t,t+\pi] and [t+π,t+2π][t+\pi,t+2\pi]. Their arclengths are S(t)S(t) and S(t+π)=P−S(t)S(t+\pi)=P-S(t). Let μt,μt+π\mu_t,\mu_{t+\pi} be the pushforwards of normalized arclength measures on these arcs under projection onto utu_t. Finally, let νt\nu_t be the pushforward of normalized arclength measure on the whole boundary ∂K\partial K under projection onto utu_t. Since

∂K=Γt∪Γt+π,S(t)+S(t+π)=P,\partial K=\Gamma_t\cup\Gamma_{t+\pi}, \qquad S(t)+S(t+\pi)=P,

::: we have

νt=S(t)Pμt+S(t+π)Pμt+π.\nu_t = \frac{S(t)}{P}\mu_t + \frac{S(t+\pi)}{P}\mu_{t+\pi}.

As the Gini functional is concave under mixtures,

D(νt)≥S(t)PD(μt)+S(t+π)PD(μt+π).(5)\mathcal{D}(\nu_t) \ge\frac{S(t)}{P}\mathcal{D}(\mu_t) + \frac{S(t+\pi)}{P}\mathcal{D}(\mu_{t+\pi}). \tag*{(5)}

Now

∣z∣=12∫0π∣⟨z,ut⟩∣ dt,|z| = \frac{1}{2}\int_0^\pi|\langle z,u_t\rangle|\,dt,

and hence

θ(K)=12∫0πD(νt) dt.\theta(K) = \frac{1}{2}\int_0^\pi\mathcal{D}(\nu_t)\,dt.

Using 2.2,

θ(K)≥12P∫0π[S(t)D(μt)+S(t+π)D(μt+π)] dt.\theta(K) \ge\frac{1}{2P}\int_0^\pi\left[S(t)\mathcal{D}(\mu_t)+S(t+\pi)\mathcal{D}(\mu_{t+\pi})\right]\,dt.

Projection of Γt+π\Gamma_{t+\pi} onto ut+π=−utu_{t+\pi}=-u_t reflects the law obtained by projection onto utu_t. Since the Gini mean difference is invariant under reflection, the change of variable r=t+πr=t+\pi gives

θ(K)≥12P[∫0πS(t)D(μt) dt+∫π2πS(t)D(μt) dt]=12P∫02πS(t)D(μt) dt.(6)\theta(K) \ge\frac{1}{2P}\left[\int_0^\pi S(t)\mathcal{D}(\mu_t)\,dt+\int_\pi^{2\pi}S(t)\mathcal{D}(\mu_t)\,dt\right]=\frac{1}{2P}\int_0^{2\pi}S(t)\mathcal{D}(\mu_t)\,dt. \tag*{(6)}

Parametrize Γt\Gamma_t by arclength as

rt:[0,S(t)]⟶Γt,r_t:[0,S(t)]\longrightarrow\Gamma_t,

starting at the point minimizing the projection onto utu_t, and put

xt(s)=⟨rt(s),ut⟩.x_t(s)=\langle r_t(s),u_t\rangle.

If αt(s)\alpha_t(s) denotes the tangent angle at rt(s)r_t(s), then

rt′(s)=(cos⁡αt(s),sin⁡αt(s))r'_t(s)=(\cos\alpha_t(s),\sin\alpha_t(s))

and therefore

xt′(s)=⟨rt′(s),ut⟩=sin⁡(αt(s)−t)≥0.x'_t(s)=\langle r'_t(s),u_t\rangle=\sin(\alpha_t(s)-t)\ge0.

Thus xtx_t is nondecreasing on [0,S(t)][0,S(t)]. Using that xtx_t is nondecreasing and applying Fubini’s theorem we get

S(t)2D(μt)=∫0S(t)∫0S(t)∣xt(s)−xt(r)∣ dr ds=2∬0≤r≤s≤S(t)xt(s)−xt(r) dr ds=2∬0≤r≤s≤S(t)∫rsxt′(z) dz dr ds=2∫0S(t)z(S(t)−z)xt′(z) dz.(7)\begin{aligned} S(t)^2\mathcal{D}(\mu_t) &=\int_0^{S(t)}\int_0^{S(t)}|x_t(s)-x_t(r)|\,dr\,ds =2\iint_{0\le r\le s\le S(t)}x_t(s)-x_t(r)\,dr\,ds\\ &=2\iint_{0\le r\le s\le S(t)}\int_r^s x'_t(z)\,dz\,dr\,ds =2\int_0^{S(t)}z\bigl(S(t)-z\bigr)x'_t(z)\,dz. \tag*{(7)} \end{aligned}

Observe that the median Z(2)Z_{(2)} of three independent uniform points Z1,Z2,Z3Z_1,Z_2,Z_3 on [0,S(t)][0,S(t)] has density 6z(S(t)−z)S(t)3 dz\frac{6z(S(t)-z)}{S(t)^3}\,dz, so

S(t)2D(μt)=S(t)33Ext′(Z(2)).S(t)^2\mathcal{D}(\mu_t)=\frac{S(t)^3}{3}\mathbb{E}x'_t(Z_{(2)}).

Since s↦αt(s)s \mapsto\alpha_t(s) is nondecreasing on the half-boundary, taking the median commutes with change of variable, i.e.

αt(Z(2))−t=Mt.\alpha_t(Z_{(2)})-t=M_t.

and, by definition xt′(s)=sin⁡(αt(s)−t)x'_t(s)=\sin(\alpha_t(s)-t). Hence

S(t)2D(μt)=S(t)33Esin⁡Mt.S(t)^2D(\mu_t)=\frac{S(t)^3}{3}\mathbb{E}\sin M_t.

Finally, plugging this into 2.3 gives

θ(K)≥16P∫02πS(t)2Esin⁡Mt dt.(8)\theta(K)\geq\frac{1}{6P}\int_{0}^{2\pi}S(t)^2\mathbb{E}\sin M_t\,dt. \tag*{(8)}

Let us replace MtM_t in the integral above with ItI_t. Then

∫02πS(t)2Esin⁡It dt=∫02πS(t)∫tt+πsin⁡(ω−t) μ(dω) dt.\int_{0}^{2\pi}S(t)^2\mathbb{E}\sin I_t\,dt=\int_{0}^{2\pi}S(t)\int_t^{t+\pi}\sin(\omega-t)\,\mu(d\omega)\,dt.

The inner integral is the width w(t)w(t) of KK in the direction utu_t. Hence

∫02πS(t)2Esin⁡It dt=∫02πS(t)w(t) dt=∫0π(S(t)+S(t+π))w(t) dt=P∫0πw(t) dt=P2.(9)\int_{0}^{2\pi}S(t)^2\mathbb{E}\sin I_t\,dt=\int_{0}^{2\pi}S(t)w(t)\,dt=\int_{0}^{\pi}(S(t)+S(t+\pi))w(t)\,dt=P\int_{0}^{\pi}w(t)\,dt=P^2. \tag*{(9)}

where in the last step we used Cauchy’s perimeter formula. Therefore, the following Proposition together with Proposition 2.1 would conclude the proof of Theorem 1.1 with non-strict inequality.

Proposition 2.2. For every 0≤a≤π/20\leq a\leq\pi/2,

∫02πS(t)2P{Mt∈[a,π−a]} dt≥∫02πS(t)2P{It∈[a,π−a]} dt.(10)\int_{0}^{2\pi}S(t)^2\mathbb{P}\{M_t\in[a,\pi-a]\}\,dt\geq\int_{0}^{2\pi}S(t)^2\mathbb{P}\{I_t\in[a,\pi-a]\}\,dt. \tag*{(10)}

Consequently,

∫02πS(t)2Esin⁡Mt dt≥∫02πS(t)2Esin⁡It dt.(11)\int_{0}^{2\pi}S(t)^2\mathbb{E}\sin M_t\,dt\geq\int_{0}^{2\pi}S(t)^2\mathbb{E}\sin I_t\,dt. \tag*{(11)}

Proof. Put b=π−ab=\pi-a and let, as before, μ(dα)\mu(d\alpha) be the tangent-angle measure. Set

C:=∫02πS(t)2 dt,fT(t):=S(t)2C,C:=\int_{0}^{2\pi}S(t)^2\,dt,\qquad f_T(t):=\frac{S(t)^2}{C},

and choose T∈[0,2π)T\in[0,2\pi) with density fTf_T. Conditional on T=tT=t, choose three independent points X,Y,ZX,Y,Z of the angle circle from normalized μ\mu on the half-arc [t,t+π][t,t+\pi]. Write m=med⁡(X,Y,Z)m=\operatorname{med}(X,Y,Z), taking real representatives in this interval. Conditional on T=tT=t, the laws of X−tX-t and m−tm-t are those of ItI_t and MtM_t, respectively. Define the events

Em:={m∈[T+a,T+b]},EX:={X∈[T+a,T+b]}.E_m:=\{m\in[T+a,T+b]\},\qquad E_X:=\{X\in[T+a,T+b]\}.

Total probability, applied to the start TT, gives

P(Em)=∫02πP(Em∣T=t)fT(t) dt=1C∫02πS(t)2P{Mt∈[a,b]} dt\mathbb{P}(E_m)=\int_{0}^{2\pi}\mathbb{P}(E_m\mid T=t)f_T(t)\,dt=\frac{1}{C}\int_{0}^{2\pi}S(t)^2\mathbb{P}\{M_t\in[a,b]\}\,dt

:::`

and similarly

P(EX)=1C∫02πS(t)P{It∈[a,b]} dt.\mathbb{P}(E_X)=\frac{1}{C}\int_0^{2\pi}S(t)\mathbb{P}\{I_t\in[a,b]\}\,dt.

Thus (10) is equivalent to P(Em)≥P(EX)\mathbb{P}(E_m)\ge\mathbb{P}(E_X). We will prove the stronger conditional inequality

P(Em∣X,Y)≥P(EX∣X,Y)almost surely,(12)\mathbb{P}(E_m\mid X,Y)\ge\mathbb{P}(E_X\mid X,Y)\quad\text{almost surely}, \tag*{(12)}

and then take expectations over the pair.

The joint law of T,X,YT,X,Y is

S(t)2C dt 1{x∈[t,t+π]} dμ(x)S(t)1{y∈[t,t+π]} dμ(y)S(t)=1C1{x,y∈[t,t+π]} dt dμ(x) dμ(y).\frac{S(t)^2}{C}\,dt\, \frac{\mathbf{1}_{\{x\in[t,t+\pi]\}}\,d\mu(x)}{S(t)} \frac{\mathbf{1}_{\{y\in[t,t+\pi]\}}\,d\mu(y)}{S(t)} = \frac{1}{C}\mathbf{1}_{\{x,y\in[t,t+\pi]\}}\,dt\,d\mu(x)\,d\mu(y).

Hence, conditional on X,YX,Y, the start TT is uniform among the half-windows containing both points.

If X=YX=Y, then m=Xm=X, and the conditional comparison (12) is an equality. Fix a pair X=xX=x, Y=yY=y, order its points along their shorter arc, rotate the coordinates and μ\mu, and lift the points to 0,d0,d, where 0<d<π0<d<\pi. Thus X,YX,Y take the values 0,d0,d in either order. Using the corresponding lift of the start, conditional on this fixed pair, TT is uniform on

J=[d−π,0].J=[d-\pi,0].

In particular, events with XX and YY antipodal occur with probability zero, so we assume d<πd<\pi.

For each t∈Jt\in J, let ZtZ_t have normalized μ\mu-law on [t,t+π][t,t+\pi], and put

mt=med⁡(0,d,Zt),p(t)=P{mt∈[t+a,t+b]}.m_t=\operatorname{med}(0,d,Z_t),\qquad p(t)=\mathbb{P}\{m_t\in[t+a,t+b]\}.

Conditional on T=tT=t, the third sample is independent of X,YX,Y, so conditioning on the pair leaves its normalized μ\mu-law on [t,t+π][t,t+\pi] unchanged. Consequently,

p(t)=1S(t)∫tt+π1{med⁡(0,d,z)∈[t+a,t+b]} μ(dz)=P(Em∣X,Y,T=t).p(t)=\frac{1}{S(t)}\int_t^{t+\pi}\mathbf{1}_{\{\operatorname{med}(0,d,z)\in[t+a,t+b]\}}\,\mu(dz) =\mathbb{P}(E_m\mid X,Y,T=t).

Conditional total probability gives

P(Em∣X,Y)=∫JP(Em∣X,Y,T=t)P{T∈dt∣X,Y}=1π−d∫Jp(t) dt.\mathbb{P}(E_m\mid X,Y) =\int_J\mathbb{P}(E_m\mid X,Y,T=t)\mathbb{P}\{T\in dt\mid X,Y\} =\frac{1}{\pi-d}\int_Jp(t)\,dt.

For either fixed point q∈{0,d}q\in\{0,d\}, the event q∈[T+a,T+b]q\in[T+a,T+b] is equivalent to T∈[q−b,q−a]T\in[q-b,q-a]. Hence

P{q∈[T+a,T+b]∣X,Y}=1π−d∫J1{q∈[t+a,t+b]} dt=∣J∩[q−b,q−a]∣π−d=δπ−d,\mathbb{P}\{q\in[T+a,T+b]\mid X,Y\} =\frac{1}{\pi-d}\int_J\mathbf{1}_{\{q\in[t+a,t+b]\}}\,dt =\frac{\lvert J\cap[q-b,q-a]\rvert}{\pi-d} =\frac{\delta}{\pi-d},

where the two interval intersections have the same length δ=max⁡{0,min⁡{b−a,b−d}}\delta=\max\{0,\min\{b-a,b-d\}\}. Since XX is one of these two fixed points, P(EX∣X,Y)=δ/(π−d)\mathbb{P}(E_X\mid X,Y)=\delta/(\pi-d). Thus the conditional comparison reduces to proving

∫Jp(t) dt≥δ.\int_Jp(t)\,dt\ge\delta.

If δ=0\delta= 0, this follows from p(t)≥0p(t) \ge0. Assume δ>0\delta> 0.

We estimate the probability that mtm_t falls outside [t+a,t+b][t+a,t+b], on either its left or its right. The key observation is that the law of mtm_t increases stochastically with tt. Indeed, for 0<c≤d0 < c \le d,

P{mt<c}=μ[t,c)μ[t,c)+μ[c,t+π].\mathbb{P}\{m_t < c\} = \frac{\mu[t,c)}{\mu[t,c)+\mu[c,t+\pi]}.

As tt increases, μ[t,c)\mu[t,c) decreases and μ[c,t+π]\mu[c,t+\pi] increases, so the displayed probability is nonincreasing. For c≤0c \le0 or c>dc > d, the probability is constant.

Write

A(t)=P{mt<t+a},B(t)=P{mt>t+b}.A(t) = \mathbb{P}\{m_t < t+a\}, \qquad B(t) = \mathbb{P}\{m_t > t+b\}.

These are the probabilities of the two disjoint ways to miss the interval, so

p(t)=1−A(t)−B(t).p(t) = 1-A(t)-B(t).

Since 0≤mt≤d0 \le m_t \le d and δ≤b−d\delta\le b-d,

A(t)=0(d−π≤t≤d−π+δ),B(t)=0(−δ≤t≤0).\begin{aligned} A(t) &= 0 \qquad(d-\pi\le t \le d-\pi+\delta),\\ B(t) &= 0 \qquad(-\delta\le t \le0). \end{aligned}

Moreover, δ≤b−a\delta\le b-a and stochastic monotonicity give, for d−π≤t≤−δd-\pi\le t \le-\delta,

A(t+δ)≤P{mt<t+δ+a}≤P{mt≤t+b}=1−B(t).A(t+\delta) \le\mathbb{P}\{m_t < t+\delta+a\} \le\mathbb{P}\{m_t \le t+b\} = 1-B(t).

To integrate this comparison, first remove the intervals on which AA and BB vanish. For AA, this gives

∫d−π0A(u) du=∫d−π+δ0A(u) du=∫d−π−δA(t+δ) dt.\int_{d-\pi}^{0} A(u)\,\mathrm{d}u = \int_{d-\pi+\delta}^{0} A(u)\,\mathrm{d}u = \int_{d-\pi}^{-\delta} A(t+\delta)\,\mathrm{d}t.

For BB, we have

∫d−π0B(t) dt=∫d−π−δB(t) dt.\int_{d-\pi}^{0} B(t)\,\mathrm{d}t = \int_{d-\pi}^{-\delta} B(t)\,\mathrm{d}t.

Summing and using A(t+δ)+B(t)≤1A(t+\delta)+B(t) \le1, we obtain

∫J(A(t)+B(t)) dt=∫d−π−δ(A(t+δ)+B(t)) dt≤∫d−π−δ1 dt=π−d−δ.\int_J (A(t)+B(t))\,\mathrm{d}t = \int_{d-\pi}^{-\delta} (A(t+\delta)+B(t))\,\mathrm{d}t \le \int_{d-\pi}^{-\delta} 1\,\mathrm{d}t = \pi-d-\delta.

Since p(t)=1−A(t)−B(t)p(t)=1-A(t)-B(t), it follows that

∫Jp(t) dt=∣J∣−∫J(A(t)+B(t)) dt≥(π−d)−(π−d−δ)=δ.\int_J p(t)\,\mathrm{d}t = |J|-\int_J (A(t)+B(t))\,\mathrm{d}t \ge (\pi-d)-(\pi-d-\delta) = \delta.

Dividing by π−d\pi-d proves the conditional comparison:

P(Em∣X,Y)=1π−d∫Jp(t) dt≥δπ−d=P(EX∣X,Y).\mathbb{P}(E_m\mid X,Y) = \frac{1}{\pi-d}\int_J p(t)\,\mathrm{d}t \ge \frac{\delta}{\pi-d} = \mathbb{P}(E_X\mid X,Y).

Finally, taking expectations with respect to the joint law of X,YX,Y and using total probability yields

∫02πS(t)2P{Mt∈[a,b]} dt=CP(Em)=CE[P(Em∣X,Y)]≥CE[P(EX∣X,Y)]=CP(EX)=∫02πS(t)2P{It∈[a,b]} dt.\begin{aligned} \int_0^{2\pi} S(t)^2\mathbb{P}\{M_t\in[a,b]\}\,\mathrm{d}t = C\mathbb{P}(E_m) = C\mathbb{E}\left[\mathbb{P}(E_m\mid X,Y)\right] \\ &\ge C\mathbb{E}\left[\mathbb{P}(E_X\mid X,Y)\right]=C\mathbb{P}(E_X)=\int_0^{2\pi}S(t)^2\mathbb{P}\{I_t\in[a,b]\}\,dt. \end{aligned}

As b=π−ab=\pi-a, this is (2.7). This argument also applies to non-smooth tangent-angle measures; choices at window endpoints affect only a Lebesgue-null set of starts.

The identity

sin⁡r=∫0π/21[a,π−a](r)cos⁡a da,0≤r≤π,\sin r=\int_0^{\pi/2}\mathbf{1}_{[a,\pi-a]}(r)\cos a\,da,\qquad0\le r\le\pi,

and Tonelli’s theorem yield (2.8) by integrating (2.7) against cos⁡a da\cos a\,da.

Proof of Theorem 1.1. As we outlined before Proposition 2.2,

∫02πS(t)2Esin⁡It dt=∫02πS(t)(∫tt+πsin⁡(α−t) μ(dα)) dt=∫02πS(t)w(t) dt,(13)\begin{aligned} \int_0^{2\pi}S(t)^2\mathbb{E}\sin I_t\,dt &=\int_0^{2\pi}S(t)\left(\int_t^{t+\pi}\sin(\alpha-t)\,\mu(d\alpha)\right)\,dt \\ &=\int_0^{2\pi}S(t)w(t)\,dt, \tag*{(13)} \end{aligned}

where w(t)w(t) is the width of KK in direction utu_t. Indeed,

w(t)=⟨rt(S(t))−rt(0),ut⟩=∫0S(t)⟨rt′(s),ut⟩ ds=∫0S(t)sin⁡(αt(s)−t) ds=∫tt+πsin⁡(α−t) μ(dα).\begin{aligned} w(t)&=\langle r_t(S(t))-r_t(0),u_t\rangle\\ &=\int_0^{S(t)}\langle r_t'(s),u_t\rangle\,ds =\int_0^{S(t)}\sin(\alpha_t(s)-t)\,ds \\ &=\int_t^{t+\pi}\sin(\alpha-t)\,\mu(d\alpha). \end{aligned}

Since

S(t)+S(t+π)=P,w(t+π)=w(t),S(t)+S(t+\pi)=P,\qquad w(t+\pi)=w(t),

Cauchy’s perimeter formula gives

∫02πS(t)w(t) dt=P∫0πw(t) dt=P2.(14)\int_0^{2\pi}S(t)w(t)\,dt=P\int_0^\pi w(t)\,dt=P^2. \tag*{(14)}

Combining Propositions 2.1 and 2.2 with (2.11), we obtain

θ(K)≥P6.\theta(K)\ge\frac{P}{6}.

It remains to exclude equality for a nondegenerate convex body. Suppose therefore that

θ(K)=P6.\theta(K)=\frac{P}{6}.

Then both inequalities supplied by Propositions 2.2 and 2.1 must be equalities. Recall that we denote μ\mu the tangent-angle measure of ∂K\partial K, of total mass PP. We show that equality in (2.8) requires that no open semicircle contain three points of supp⁡μ\operatorname{supp}\mu. Suppose otherwise. Choose small arcs U,V,WU,V,W about three such points, lifted in this order to R\mathbb{R}, and a0>0a_0>0 so that

v−x>a0,y−v>a0,y−x<π−a0(x∈U, v∈V, y∈W).v-x>a_0,\qquad y-v>a_0,\qquad y-x<\pi-a_0\qquad(x\in U,\ v\in V,\ y\in W).

All three arcs have positive μ\mu-measure. Fix 0<a<a00<a<a_0 and a pair x∈Ux\in U, y∈Wy\in W. Translate the angular coordinates so that x=0x=0, y=dy=d, and use mtm_t, p(t)p(t), A(t)A(t), B(t)B(t) from the proof of Proposition 2.2. Here δ=π−d−a>0\delta=\pi-d-a>0. The difference between the two sides of its integral comparison is

∫d−π0p(t) dt−δ=∫d−πd−π+a[1−B(t)−A(t+δ)] dt.\int_{d-\pi}^{0}p(t)\,dt-\delta=\int_{d-\pi}^{d-\pi+a}\left[1-B(t)-A(t+\delta)\right]\,dt.

For t=d−π+st=d-\pi+s with 0<s<a0<s<a, stochastic monotonicity gives

1−B(t)−A(t+δ)=P{mt≤d+s−a}−P{mt+δ<s}≥P{s≤mt≤d+s−a}≥μ(V)P>0.1-B(t)-A(t+\delta)=\mathbb{P}\{m_t\le d+s-a\}-\mathbb{P}\{m_{t+\delta}<s\} \ge\mathbb{P}\{s\le m_t\le d+s-a\}\ge\frac{\mu(V)}{P}>0.

Indeed, the translated arc VV lies in (a,d−a)(a,d-a), hence in [s,d+s−a][s,d+s-a], and a third sample in VV is the median. The excess is therefore at least aμ(V)/Pa\mu(V)/P for every pair in U×WU\times W. The pair disintegration in Proposition 2.2 now makes (10) strict for every 0<a<a00<a<a_0. Integration against cos⁡a da\cos a\,da makes (11) strict, a contradiction. Now, we have shown that

∣supp⁡μ∣≤4.|\operatorname{supp}\mu|\le4.

Indeed, if five points occurred in cyclic order and g1,…,g5g_1,\ldots,g_5 were the intervening angular gaps, then the absence of three points in an open semicircle would imply

gi+gi+1≥π(i(mod5)).g_i+g_{i+1}\ge\pi\qquad(i\pmod5).

Summing gives 4π≥5π4\pi\ge5\pi, since ∑igi=2π\sum_i g_i=2\pi, which is impossible. Hence KK is a polygon with at most four sides. We now show that the inequality of Proposition 2.1 is strict for every nondegenerate polygon. Choose an open interval JJ of angles tt so that for every t∈Jt\in J

ut=(−sin⁡t,cos⁡t)u_t=(-\sin t,\cos t)

is not normal to any edge of KK. For t∈Jt\in J the linear functional x↦⟨x,ut⟩x\mapsto\langle x,u_t\rangle attains its minimum and maximum at unique vertices vv and ww, respectively. Those vertices split ∂K\partial K in two complementary polygonal chains. Denote their lengths by S1,S2S_1,S_2, and let ν1,t,ν2,t\nu_{1,t},\nu_{2,t} be their normalized arclength measures projected onto utu_t.

The equality in the concavity step of Proposition 2.1 is equivalent to

ν1,t=ν2,t.\nu_{1,t}=\nu_{2,t}.

Let e1,e2e_1,e_2 be the unit vectors along the edges of the two chains incident to vv and directed away from vv. For t∈Jt\in J,

⟨ei,ut⟩>0.\langle e_i,u_t\rangle>0.

Near the projection of vv, the densities of ν1,t\nu_{1,t} and ν2,t\nu_{2,t} are respectively

1S1⟨e1,ut⟩,1S2⟨e2,ut⟩.\frac{1}{S_1\langle e_1,u_t\rangle},\qquad\frac{1}{S_2\langle e_2,u_t\rangle}.

Therefore equality of the two projected laws would imply

⟨S1e1−S2e2,ut⟩=0.\langle S_1e_1-S_2e_2,u_t\rangle=0.

As S1e1−S2e2≠0S_1 e_1-S_2 e_2 \ne0, this can hold only for isolated values of tt, whereas JJ is an interval. Hence the Gini concavity inequality is strict on a set of positive measure, and so Proposition 2.1 is strict for every nondegenerate polygon. Therefore,

θ(K)>per⁡K6\theta(K) > \frac{\operatorname{per} K}{6}

Sharpness follows from the rectangles Kε=[0,1]×[0,ε]K_\varepsilon=[0,1]\times[0,\varepsilon]. As ε↓0\varepsilon\downarrow0, normalized boundary arclength concentrates on the two horizontal sides, and hence θ(Kε)→E∣U−V∣=1/3\theta(K_\varepsilon)\to\mathbb{E}|U-V|=1/3, where U,V∼Unif⁡[0,1]U,V\sim\operatorname{Unif}[0,1], while per⁡Kε→2\operatorname{per} K_\varepsilon\to2.

References

  1. [1]G. Bonnet, A. Gusakova, C. Thäle, and D. Zaporozhets, Sharp inequalities for the mean distance of random points in convex bodies, Adv. Math. 386 (2021), 107813.DOI
  2. [2]A. S. Tokmachev, The Mean Distance Between Random Points on the Boundary of a Convex Figure, J. Math. Sci. 286 (2024), 798–806.DOI
  3. [3]A. S. Tokmachev, Inequalities for convex functions of random points inside and on the boundary of convex bodies, arXiv:2607.08869, 2026.arxiv.org/abs/2607.08869
  4. [4]S. Yitzhaki, Gini’s mean difference: a superior measure of variability for non-normal distributions, Metron 61 (2003), no. 2, 285–316.
  5. [5]E. Shen, The Zaporozhets–Tarasov Conjecture on Mean Distances, arXiv:2608.06470v3, 2026.arxiv.org/abs/2608.06470

Paper details

vlnc:W_prcbvjp9
Version
v1 · latest
Published
Fresh — added to the Hub within the last 30 days
Added
License
CC BY 4.0
Usage

views · 0

downloads · 0

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

  • Mathematics

Not recorded for this paper

View statement
Looking for related papers…
Contents