We prove that the distance between the centers of mass c(K) and c(∂K) of a convex body K and its boundary ∂K, respectively, does not exceed Dn times the width of K in the direction of the vector joining c(∂K) and c(K). The constant Dn is optimal and coincides with the sharp constant in the problem of bounding the distance between the orthogonal projection of the centroid of a convex body onto a hyperplane and the centroid of its orthogonal projection onto that hyperplane.
Introduction
Let K⊂Rn be a convex body, i.e., a convex compact set with non-empty interior. Write
c(K)=∣K∣1∫Kxdx,c(∂K)=∣∂K∣1∫∂KxdS(x),
for the centroid of K and its boundary, respectively, where ∣K∣ is the volume of K, ∣∂K∣ is its surface area, and dS denotes the surface area measure on ∂K.
Let Sn−1 denote the unit sphere in Rn. For u∈Sn−1, let
wK(u)=x∈Kmax⟨x,u⟩−x∈Kmin⟨x,u⟩
be the width of K in the direction u.
In their book Unsolved Problems in Geometry, Croft, Falconer, and Guy [1 p. 36, Problem A25] ask how far apart the centroid of a planar convex body and the centroid of its boundary can be, relative to such quantities as the diameter and the width. They also point out the analogous questions in higher dimensions.
The planar problem was solved by Nazarov, Ryabogin, and Yaskin [5]. They proved that for every planar convex body K and every u∈S1,
∣⟨c(∂K)−c(K),u⟩∣≤61wK(u),
and therefore
∣c(∂K)−c(K)∣≤61diamK.
The constant 1/6 in both inequalities is best possible.
The purpose of the present paper is to solve the corresponding problem in all dimensions. In [5], it was conjectured that the supremum of
Key words and phrases. Convex body, center of mass, extremal problem.
The first, third and fourth authors are supported in part by NSERC. The second author is supported in part by U.S. National Science Foundation Grant DMS-2247771. over all convex bodies K⊂Rn and all directions u∈Sn−1 is the constant Dn defined below. For an affine function ℓ(t)=(1−t)α+tβ on [0,1] such that α,β≥0 are not both zero, set
Ij(ℓ)=∫01tjℓ(t)n−2dt,j=0,1,2,(1)
and let
Dn=ℓ≥0affine on [0,1]ℓ≡0max{I1(ℓ)I2(ℓ)−I0(ℓ)I1(ℓ)}.(2)
Multiplying ℓ by a positive constant does not change the expression in (2). Hence we may normalize ℓ so that ℓ(0)+ℓ(1)=1. The resulting parameter set is compact, so the maximum is attained. The constants Dn form an increasing sequence such that
Theorem 1.1.For every convex body K⊂Rn, n≥2, and every u∈Sn−1,
∣⟨c(∂K)−c(K),u⟩∣≤DnwK(u).(3)
Consequently,
∣c(∂K)−c(K)∣≤DndiamK.(4)
The constant Dn is optimal in both inequalities.
The constants Dn first appeared in the following problem studied by Myroshnychenko, Tatarko, and Yaskin [4]. For a convex body K⊂Rn and a hyperplane H in Rn, consider PHc(K), the orthogonal projection of c(K) onto H, and c(PHK), the centroid of the orthogonal projection of K onto H. If these two points are different, let
There is a geometric reason to expect the constants Dn from [4] to appear in Theorem 1.1. Fix a hyperplane H and compress a convex body K in the direction orthogonal to H, preserving its projection onto H. The centroid of the compressed body projects to PHc(K), while the projection of its boundary centroid approaches c(PHK). Indeed, as the body becomes thinner, its surface area becomes concentrated on the two boundary graphs over PHK. Thus the inequality of [4] is recovered as a limiting case of the inequality in Theorem 1.1. In particular, the optimal constant in (3) cannot be smaller than Dn.
The paper is organized as follows. Section 2 outlines the proof and states its main ingredient, Proposition 2.1. In Section 3, we derive bounds (3) and (4) from this proposition, and in Section 4 we establish the sharpness of the constant Dn. Section 5 contains one-dimensional auxiliary inequalities. Finally, Section 6 uses localization to prove Proposition 2.1.
Outline of the proof
The proof of Theorem 1.1 is based on decomposing K into parallel chords. The coordinate of c(K) in a fixed direction is expressed as an average of midpoint coordinates of chords weighted by chord length; see (6). To obtain the corresponding coordinate of c(∂K), we first take the unweighted average of these midpoint coordinates for each chord direction and then average the resulting values over all directions, with weights given by the volumes of the corresponding projections; see (10). The following proposition allows us to compare these averages.
Proposition 2.1. Let Ω⊂Rn−1 be a bounded, open, and convex set. Let a,b be bounded functions on Ω, with a convex, b concave, and 0≤a≤b≤1. Put
h=b−a,m=2a+b.
If h≡0, then
∣Ω∣1∫Ωm(y)dy−∫Ωh(y)dy∫Ωh(y)m(y)dy≤Dn.(5)
The appearance of Dn in Proposition 2.1 can be understood through localization. Applied in Rn−1, the localization lemma of Kannan, Lovász, and Simonovits reduces the proof of (5) to one-dimensional integrals with weights proportional to ℓn−2, where ℓ is nonnegative and affine. These are precisely the weights occurring in the definition of Dn. For affine functions, the required one-dimensional bound follows from this definition. The one-dimensional auxiliary inequalities extend this bound to general pairs of convex and concave functions.
Proof of the bounds in Theorem 1.1
Proof. Assuming Proposition 2.1, we now prove inequalities (3) and (4). Fix u∈Sn−1. Since the inequalities are invariant under translations and dilations, we may assume that
x∈Kmin⟨x,u⟩=0,x∈Kmax⟨x,u⟩=1.
Thus it suffices to prove
∣⟨c(∂K)−c(K),u⟩∣≤Dn.
Fix v∈Sn−1 with ⟨u,v⟩=0. Replacing v by −v if necessary, we may assume ⟨u,v⟩>0. Put
Ωv=int(Pv⊥K),
where Pv⊥ denotes the orthogonal projection onto the hyperplane v⊥={x∈Rn:⟨x,v⟩=0}, and the interior is taken relative to v⊥.
We next express ⟨c(∂K),u⟩ using the same chord midpoint coordinates mv(y). Let νK(x) denote the outer unit normal to K, defined dS-almost everywhere on ∂K. For every nonnegative measurable function f, the area formula (see [2 Theorem 3.2.3]) gives
For y∈Ωv, the fiber on the right consists of the two endpoints of the corresponding chord. The boundary of Pv⊥K has (n−1)-dimensional Lebesgue measure zero and therefore does not contribute to the integral. We may apply (7) with f=1 and f(x)=⟨x,u⟩, obtaining
∫∂K∣⟨νK(x),v⟩∣dS(x)=2∣Ωv∣(8)
and
∫∂K⟨x,u⟩∣⟨νK(x),v⟩∣dS(x)=2∫Ωvmv(y)dy.(9)
Since v and −v determine the same chords, the functions av, bv, hv, mv, as well as Ωv, are unchanged under v↦−v. Consequently, (6), (8), and (9) hold for every v with ⟨u,v⟩=0.
To apply Proposition 2.1, we now average (8) and (9) over all chord directions v to obtain a suitable representation of the boundary centroid. Let dσ be the normalized rotation-invariant measure on Sn−1. Since
∫Sn−1∣⟨u,v⟩∣dσ(v)=γn>0
is independent of the unit vector u, Fubini’s theorem together with (9) gives
Here the directions satisfying ⟨u,v⟩=0 may be disregarded, since they form a set of spherical measure zero.
The last inequality is precisely (3) since wK(u)=1. If c(∂K)=c(K), choosing u in the direction of their difference and using wK(u)≤diamK gives (4).
Sharpness
To prove the sharpness of the constant Dn in (3), it is enough to exhibit a sequence of bodies for which the quotient in (3) tends to Dn in the direction e1.
Assume first n≥3. Choose an affine function ℓ≥0 attaining the maximum in (2), and for ε>0 define
The body Kε is convex and wKε(e1)=1. Fubini’s theorem gives
⟨c(Kε),e1⟩=I1(ℓ)I2(ℓ).(12)
The bottom and top faces, z=0 and z=ε2t, have total area
(1+1+ε4)(2ε)n−2I0(ℓ),
and the first coordinates of their centroids are both I1(ℓ)/I0(ℓ). The 2(n−2) side faces have total area O(εn−1). The part of the boundary at t=1 has area O(εn), and the part at t=0 has zero (n−1)-dimensional area. Hence
When n=2, let Kε be the triangle with vertices (0,0),(1,0),(1,ε2). Then ⟨c(Kε),e1⟩=2/3, ⟨c(∂Kε),e1⟩→1/2, and diamKε→1 as ε→0+. Thus we obtain the same conclusions as above.
One-dimensional comparison inequalities
The remainder of the paper is devoted to proving Proposition 2.1. In this section we obtain a series of auxiliary one-dimensional integral inequalities that will be used in the proof of Proposition 5.5. The latter will be the main ingredient in the localization argument of Proposition 2.1.
Preliminary estimates
For any nonzero finite positive measure ν and bounded measurable real-valued function f, define
Aν(f)=∫dν∫fdν
and
Vν(f)=Aν(f2)−Aν(f)2.
Thus Aν denotes averaging, and Vν is the corresponding variance. Notice that
Vν(f)=Aν((f−Aν(f))2)≥0.
For the remainder of the paper, put C=Dn. Fix a nonzero nonnegative affine function ℓ(t) on [0,1] and define the normalized weight
ω(t)=I0(ℓ)ℓ(t)n−2,0≤t≤1.(14)
Let dμ(t)=ω(t)dt. Throughout this section, we omit the subscript μ and write A and V for Aμ and Vμ, respectively. Since μ([0,1])=1, we have
A(f)=∫01f(t)ω(t)dt.
Let T(t)=t. Then
A(T)=∫01tω(t)dt=I0(ℓ)I1(ℓ),
and
A(T2)=∫01t2ω(t)dt=I0(ℓ)I2(ℓ).
Therefore
V(T)=A(T2)−A(T)2=I0I2−I02I12.
From the definition of Dn,
A(T)V(T)=I1(ℓ)I2(ℓ)−I0(ℓ)I1(ℓ)≤C,
and hence
V(T)≤CA(T).
Since A is linear and V(1−T)=V(T), applying the same inequality after the reflection t↦1−t gives
In the lemmas below, we establish integral inequalities for bounded convex or concave functions on [0,1]. Since changing their endpoint values does not affect these integrals, we may assume that such functions are continuous on [0,1], replacing their endpoint values by their one-sided limits if necessary.
The following estimate will be used in Lemma 5.3.
Lemma 5.1. If 0≤h≤1 is a concave function on [0,1], then
V(h)≤31A(h).(16)
Proof. Suppose first that n≥3, and consider the convex body
Pℓ={(t,y)∈[0,1]×Rn−2:∥y∥∞≤ℓ(t)}.
For fixed t, the y-section of Pℓ has (n−2)-dimensional volume (2ℓ(t))n−2. Note that the explicit construction of Pℓ is not important. All we need is that the (n−2)-dimensional volume of its cross-sections is proportional to ℓn−2(t). Hence, for every integrable function f on [0,1],
When n=2, we take Pℓ=[0,1], and the same identity holds.
If A(h)=0 or A(h)=1, then V(h)=0 and there is nothing to prove. Hence for the remainder of the proof we assume 0<A(h)<1.
Let
F(s)=A(1{h>s})=∣Pℓ∣∣{(t,y)∈Pℓ:h(t)>s}∣,0≤s≤1.
Since h is nonnegative, concave, and not identically zero, it is positive on (0,1), so F(0)=1.
Since (t,y)↦h(t) is concave on Pℓ, the Brunn–Minkowski inequality implies that F is log-concave. By the layer cake decomposition,
A(h)=∫01F(s)ds,andA(h2)=2∫01sF(s)ds.(17)
To bound V(h), we compare the function F with an exponential function that has the same integral. In particular, this will allow us to bound A(h2) in terms of A(h). Choose λ>0 so that
A(h)=∫01e−λsds=λ1−e−λ,
which is always possible since λ1−e−λ decreases continuously from 1 to 0 as λ increases from 0 to ∞.
Since logF(s)+λs is concave and vanishes at s=0, the difference F(s)−e−λs changes sign at most once, from nonnegative to nonpositive. Thus there exists s0∈[0,1] such that
For the last inequality, multiplication by the positive denominator reduces it to
Φ(λ)=(λ−3)eλ+3e−λ+5λ≥0.
Now Φ(0)=Φ′(0)=0 and
Φ′′(λ)=(λ−1)eλ+3e−λ>0.
For λ≥1 this is immediate. For 0≤λ≤1, it follows from (1−λ)e2λ≤e/2<3. Thus Φ is increasing and so (18) holds.
□
The goal of this section is to establish Proposition 5.5, i.e., we need to prove the inequality
−2CA(b−a)≤V(a)−V(b)≤2CA(b−a)
for continuous functions 0≤a≤b≤1 on [0,1], with a convex and b concave. Consider the functional
J(f)=V(f)+2CA(f).
Then the right-hand inequality above is equivalent to
J(a)≤J(b).
The advantage of this formulation is that J satisfies certain comparison inequalities for elementary piecewise affine functions, which can then be extended to general convex functions using the convexity of J.
To see that the functional J is convex, take 0≤λ≤1. Then
Lemma 5.2. The function s↦J(Rs) is nonincreasing on [0,1]. In particular,
J(Rs)≤J(R0)=J(T).(19)
Proof. We first show that J(Rs) is nonincreasing on s∈[0,1) by proving that dsdJ(Rs)≤0. For 0≤s<1, let μs be the probability measure on [s,1] defined by
dμs(t)=∫s1ω(r)drω(t)dt.
Also, for s≤t≤1,
Rs(t)=1−st−s.
Under the change of variables
r=1−st−s,
Rs(t) becomes the coordinate function T(r)=r on [0,1], and the measure μs is now defined on [0,1] and is given by
∫01ℓ(s+(1−s)ρ)n−2dρℓ(s+(1−s)r)n−2dr.
Since r↦ℓ(s+(1−s)r) is again a nonnegative affine function on [0,1], this is a weight of the form (14). Hence the second inequality in (15), applied to μs, gives
Thus all three coefficients are nonnegative, proving the claim.
Affine majorants
Lemma 5.4. Let b be affine and let a be convex on [0,1]. If 0≤a≤b≤1, then
J(a)≤J(b).(22)
Proof. The idea of the proof is to approximate a by convex piecewise affine functions, and then use convexity of J and Lemma 5.3.
We first reduce to the case where a vanishes at some point. Subtracting min[0,1]a from both a and b preserves all the assumptions of the lemma and leaves J(b)−J(a) unchanged, since V is invariant under addition of constants and A is linear. Thus we may assume that
[0,1]mina=0.
If b≡0, then a≡0 and there is nothing to prove. Reflecting both the functions and the weight if necessary, we may assume b(1)≥b(0). We may further assume that b(1)=1. Indeed, if (22) holds for the functions a/b(1) and b/b(1), then
V(a)−V(b)≤2Cb(1)A(b−a)≤2CA(b−a),
which is equivalent to J(a)≤J(b). Here we used the linearity of A, the fact that V is homogeneous of degree two, and the inequalities 0<b(1)≤1 and A(b−a)≥0.
These reductions preserve the condition min[0,1]a=0. We therefore have
b(t)=b(0)(1−t)+t,0≤b(0)≤1,
and a(t0)=0 for some t0∈[0,1].
For each N, subdivide [0,t0] and [t0,1] into N equal subintervals:
0=u0<⋯<uN=t0=s0<⋯<sN=1
(with the degenerate side omitted if t0=0 or 1), and let pN be the piecewise affine function equal to a at all partition points. Then pN is convex,
Since pN≥0 and pN(t0)=0, both functions are continuous and convex. Moreover, pN− vanishes on [t0,1], so it is a linear combination with nonnegative coefficients of the functions (uj−t)+, where uj≤t0. The coefficients are the jumps in its slope. Similarly, pN+ is a linear combination with nonnegative coefficients of the functions (t−si)+, where si≥t0. Thus we may write
pN−=j∑αjb(0)Luj,αj≥0,j∑αj≤1,uj≤t0,
and
pN+=i∑βiRsi,βi≥0,i∑βi≤1,si≥t0.
Here ∑iβi≤1 follows from pN+(1)≤1. If b(0)>0, then ∑jαj≤1 follows from pN−(0)≤b(0). If b(0)=0, then t0=0 and pN−≡0; in this case we take a single term with α0=1 and u0=0.
In order to apply Lemma 5.3 and use convexity of J, we now express pN as a convex combination. Since L0=0 and R1=0, we may add them to the preceding sums so that
Proposition 5.5. Let a be convex and b be concave on [0,1] with 0≤a≤b≤1. Then
∣V(a)−V(b)∣≤2CA(b−a).(23)
Proof. We first prove that J(a)≤J(b). It suffices to do so when b>0 on [0,1]: the general case follows by applying the result to (a+δ)/(1+δ) and (b+δ)/(1+δ) and letting δ→0+, using the continuity of J under uniform convergence.
Let L(t)=(1−t)b(0)+tb(1). Then
0≤a≤L≤b≤1,
and Lemma 5.4 gives J(a)≤J(L). Since L>0 on [0,1], for sufficiently small ε>0 the function
g=(1+ε)L−εb
is nonnegative. It is also convex and satisfies g≤L, so Lemma 5.4 gives J(g)≤J(L). Since
L=1+εg+εb,
the convexity of J yields
J(L)≤1+εJ(g)+εJ(b)≤1+εJ(L)+εJ(b).
Thus J(L)≤J(b), and hence J(a)≤J(b). Equivalently,
V(a)−V(b)≤2CA(b−a).(24)
Applying the same argument to the convex function 1−b and the concave function 1−a, satisfying 0≤1−b≤1−a≤1, and using
Since h≥0 is concave and not identically zero, h>0 in Ω. Throughout the proof, A and V denote the functionals associated with Lebesgue measure on Ω. Thus,
We consider the planar case and the higher-dimensional case separately. If n=2, then Ω⊂R is an interval. After an affine reparametrization of Ω by (0,1), normalized Lebesgue measure on Ω becomes the constant weight, which is precisely the weight ℓn−2≡1 from § 5. Extending a and b continuously to [0,1] and applying Proposition 5.5, we obtain
We may therefore assume n≥3 in the localization argument below. To reach a contradiction, suppose that
A(m)>A(h)A(hm)+C.
Then the continuous functions
φ=m−A(h)A(hm)−C,ψ=h(m−A(h)A(hm))
satisfy
∫Ωφ>0,∫Ωψ=0.
By the localization lemma of Kannan, Lovász, and Simonovits [3 Corollary 2.4], there exist a line segment I⊂Ω and a nonnegative affine function ℓ on I such that
∫Iφ(x)ℓn−2(x)dx>0,∫Iψ(x)ℓn−2(x)dx=0.(26)
where dx denotes length measure along I.
Let ν denote the normalized measure on I with density proportional to ℓn−2. Using (26) together with the definitions of φ and ψ, we get
Aν(m)>A(h)A(hm)+C,Aν(h)Aν(hm)=A(h)A(hm).
Here Aν(h)>0 because h>0 on Ω. After an affine parametrization of I by [0,1], the measure ν has a density of the form (14). The restrictions of a and b remain respectively convex and concave and satisfy 0≤a≤b≤1. Thus Proposition 5.5 applies.
The same algebra as in (25), now applied to ν, together with Proposition 5.5, yields
To obtain the opposite inequality, apply the same argument to the functions
a=1−b,b=1−a,
where a is convex, b is concave, and 0≤a≤b≤1. For this pair,
h=b−a=h,m=2a+b=1−m.
The inequality already proved therefore gives
A(m)≤A(h)A(hm)+C.
Since
A(m)=1−A(m),A(h)A(hm)=1−A(h)A(hm),
we obtain
A(1−m)≤1−A(h)A(hm)+C,
or equivalently
A(m)≥A(h)A(hm)−C.
This proves the opposite inequality and hence (5). □
Acknowledgments. The project was made possible by SQuaRE at the American Institute of Mathematics. The authors thank AIM for providing a supportive and mathematically rich environment.
AI Disclaimer. ChatGPT was used in exploratory discussions and in the preparation of this manuscript.
S. Myroshnychenko, Department of Mathematics and Statistics, University of the Fraser Valley, Abbotsford, BC V2S 7M8, Canada
[1]H. T. Croft, K. J. Falconer, and R. K. Guy, Unsolved Problems in Geometry, Problem Books in Mathematics, Springer-Verlag, New York, 1991.
[2]H. Federer, Geometric Measure Theory, Grundlehren der mathematischen Wissenschaften, vol. 153, Springer-Verlag, New York, 1969.
[3]R. Kannan, L. Lovász, and M. Simonovits, Isoperimetric problems for convex bodies and a localization lemma, Discrete Comput. Geom. 13 (1995), 541–559.DOI
[4]S. Myroshnychenko, K. Tatarko, and V. Yaskin, How far apart can the projection of the centroid of a convex body and the centroid of its projection be?, Math. Ann. 390 (2024), no. 1, 1155–1169.arxiv.org/abs/2212.14456
[5]F. Nazarov, D. Ryabogin, and V. Yaskin, On the maximal distance between the centers of mass of a planar convex body and its boundary, Discrete Comput. Geom. 73 (2025), no. 4, 1016–1036.arxiv.org/abs/2305.15646