About two decades ago Lutwak introduced the concept of p-affine surface area. More recently, the results of Lutwak have been generalized by Ma to the entire class of so-called ith p-affine surface areas. In this paper, we further research this new notion and give its integral representation. Affine isoperimetric and Blaschke-Santaló inequalities, which generalize the inequalities obtained by Lutwak, are established. Furthermore, we prove the ith p-affine area ratio of convex body K for the ith p-affine surface area, which does not exceed the generalized Santaló product of convex body K.
Hinweise
Competing interests
The authors declare that they have no competing interests.
Authors’ contributions
The authors completed the paper and read and approved the final manuscript.
1 Introduction
During the past three decades, the investigations of the classical affine surface area have received great attention from many articles (see articles [1‐14] or books [15, 16]). Based on the classical affine surface area, Lutwak [17] introduced the notion of p-affine surface area and obtained some isoperimetric inequalities for p-affine surface area. Regarding the studies of p-affine surface area also see [18‐25]. In particular, Ma [26] studied the ith p-geominimal surface area.
The setting for this paper is n-dimensional Euclidean space \(\mathbb {R}^{n}\). Let \(\mathcal{K}^{n}\) denote the set of convex bodies (compact, convex subsets with nonempty interiors) and \(\mathcal{K}^{n}_{o}\) denote the subset of \(\mathcal{K}^{n}\) that contains the origin in their interiors in \(\mathbb{R}^{n}\). Let \(\mathcal{K}^{n}_{c}\) denote the set of convex bodies whose centroids lie at the origin. As usual, \(V_{i}(K)\) denotes the i-dimensional volume (i.e., Lebesgue measure) of a compact convex set K in \(\mathbb{R}^{n}\). Instead of \(V_{n}(K)\) we usually write \(V(K)\). Let \(\mathbb{S}^{n-1}\) denote the unit sphere with unit ball \(B_{n}\), \(\omega_{n}\) is the volume of \(B_{n}\), and \(\omega_{i}:=V_{i}(B_{n})\) denotes the i-dimensional intrinsic volume of \(B_{n}\). For \(K\in\mathcal{K}^{n}_{o}\), let \(K^{*}\) denote the polar body of K. Let \(\mathcal{S}^{n}_{o}\) denote the set star bodies in \(\mathbb{R}^{n}\) containing the origin in their interiors.
Anzeige
In [3], Leichtweiß defined the affine surface area \(\Omega(K)\) by
In [17], Lutwak generalized the affine surface area \(\Omega(K)\) to the p-affine surface area \(\Omega_{p}(K)\) by using the Brunn-Minkowski-Fiery theory as follows:
with equality if and only if\(K\in\mathcal{E}^{n}\), where\(\mathcal{E}^{n}=\{K\in\mathcal{F}^{n}_{o}: K^{*} \textit{ and } \Lambda K \textit{ are dilates}\}\)and ΛKdenotes the curvature image ofK.
It is easily seen that the p-affine surface area belongs to the Brunn-Minkowski-Fiery theory. Recently, Ma [24] further extended the p-affine surface area \(\Omega_{p}(K)\) to the ith p-affine surface area \(\Omega_{p}^{(i)}(K)\) of \(K\in\mathcal{K}^{n}_{o}\) (also called the \((i,0)\) type p-affine surface area, \(i\in\{0,1,\ldots,n-1\} \)) by using the Brunn-Minkowski-Fiery theory as follows:
It is the aim of this paper to establish several generalized forms of inequalities (1.3), (1.4), and (1.6). Our main results can be stated as follows.
Theorem 1.4
If\(p\geq1\), \(i\in\{0,1,\ldots, n-1\}\), and\(K\in\mathcal {F}^{n}_{i,o}\), then the integral expressions ofithp-affine surface area\(\Omega_{p}^{(i)}(K)\)are as follows:
where \(K\in\mathcal{F}^{n}_{o}\subset\mathcal{K}^{n}_{o}\) is a convex body with a positive continuous curvature function, \(f_{p}(K,\cdot)\) denotes a p-curvature function of \(K\in\mathcal{K}^{n}_{o}\).
Theorem 1.5
Suppose\(K\in\mathcal{K}^{n}_{c}\), \(i\in\{0,1,\ldots, n-1\}\)and\(p\geq 1\). Then
with equality for\(i=0\)if and only ifKis an ellipsoid, and for\(0< i\leq n-1\)if and only if all\((n-i)\)-dimensional sub-convex bodies contained inKare\((n-i)\)-ball centered at the origin.
Taking \(i=0\), inequality (1.10) reduces to Lutwak’s result (see [17], this is also Theorem 1.1 in our article).
Theorem 1.7
Suppose\(K\in\mathcal{K}^{n}_{o}\), \(i\in\{0,1,\ldots, n-1\}\)and\(p\geq 1\). Then
with equality in inequality for\(i=0\)if and only ifKis an ellipsoid centered at the origin, and for\(0< i\leq n-1\)if and only ifKis a ball centered at the origin.
Corolloary 1.8
Suppose\(K\in\mathcal{K}^{n}_{c}\), \(i\in\{0,1,\ldots, n-1\}\)and\(p\geq 1\). Then
with equality in inequality for\(i=0\)if and only ifKis an ellipsoid centered at the origin, and for\(0< i\leq n-1\)if and only if all\((n-i)\)-dimensional sub-convex bodies contained inKare an\((n-i)\)-ball centered at the origin.
Taking \(i=0\), inequality (1.12) reduces to Lutwak’s result (see [17], this is also Theorem 1.2 in our article).
with equality if and only if\(K\in\mathcal{E}^{n}_{i}\), where symbol\(\mathcal{E}^{n}_{i}\)is defined in (3.18).
Taking \(i=0\), inequality (1.13) reduces to Lutwak’s result (see [17], this is also Theorem 1.3 in our article).
The paper is organized as follows. For the sake of convenience, in Section 2 we introduce the basic knowledge about the convex geometric analysis. In Section 3 we discuss some of the properties of the ith p-affine surface area \(\Omega_{p}^{(i)}\) and ith p-curvature image \(\Lambda_{p,i}\). Meanwhile, we prove Theorems 1.4-1.7 stated at the beginning of this paper. In Section 4 we establish the cyclic inequalities of ith p-affine surface area \(\Omega _{p}^{(i)}(K)\) and the monotonicity of ith p-affine area ratio and ith p-curvature ratio; these results are a generalization of Lutwak’s conclusions (see [17]). At the same time, we complete the proof of the monotonicity theorem (Theorem 1.9 stated at the beginning of this paper). In Section 5, we further define the concept of \(\Omega_{\infty}^{(i)}\) and discuss its interesting properties. In addition, we give a daisy chain of inequalities for ith p-affine area ratio with monotone nondecreasing in p, which does not exceed the generalized Santaló product of convex body.
2 Notation and preliminaries
2.1 Support function, radial function, and polar of convex body
As usual, \(GL(n)\) denotes a nonsingular linear transformation group in \(\mathbb{R}^{n}\). For \(\phi\in GL(n)\), let \(\phi^{t}\), \(\phi^{-1}\), and \(\phi^{-t}\) denote the transpose, inverse, and inverse of the transpose of ϕ, respectively. For \(K\in\mathcal {K}^{n}\), let \(h(K,\cdot): \mathbb{R}^{n}\rightarrow \mathbb{R}\) denote the support function of \(K\in\mathcal{K}^{n}\). Namely,
where \(x \cdot y\) denotes the standard inner product of x and y in \(\mathbb{R}^{n}\). For \(\phi\in GL(n)\), then obviously \(h(\phi K,x)=h(K,\phi^{t}x)\). For the sake of convenience, we write \(h_{K}\) rather than \(h(K,\cdot)\) for the support function of K. Apparently, for \(K, L\in \mathcal{K}^{n}\), \(K\subseteq L\) if and only if \(h_{K}\leq h_{L}\). The set \(\mathcal{K}^{n}\) will be viewed as equipped with the Hausdorff metric δ defined by \(\delta (K,L)=\|h_{K}-h_{L}\|_{\infty}\) is the sup (or max) norm on the space of continuous functions on the unit sphere \(C(S^{n-1})\).
For a compact subset L of \(\mathbb{R}^{n}\), which is star-shaped with respect to the origin, we shall use \(\rho(L,\cdot)\) to denote its radial function; i.e., for \(x\in\mathbb{R}^{n}\backslash\{0\}\),
If \(\rho(L,\cdot)\) is continuous and positive, L will be called a star body, and \(\mathcal{S}^{n}_{o}\) will be used to denote the class of star bodies in \(\mathbb{R}^{n}\) containing the origin in their interiors. Apparently, for \(K, L\in \mathcal{S}^{n}_{o}\), \(K\subseteq L\) if and only if \(\rho_{K}\leq\rho _{L}\). Two star bodies K and L are said to be dilates (of one another) if \(\rho(K,u)/\rho(L,u)\) is independent of \(u\in\mathbb{S}^{n-1}\). Let \(\widetilde{\delta}\) denote the radial Hausdorff metric as follows: if \(K,L\in\mathcal {S}^{n}_{o}\), then \(\widetilde{\delta}(K,L)=\|\rho_{K}-\rho_{L}\|_{\infty}\).
For \(K\in\mathcal{K}_{o}^{n}\), the polar body \(K^{*}\) of K is defined by
Obviously, we have \((K^{*})^{*}=K\). If \(\lambda>0\), then \((\lambda K)^{*}=\lambda^{-1}K^{*}\). More generally, if \(\phi\in GL(n)\), then \((\phi K)^{*}=\phi^{-t}K^{*}\). For \(K\in\mathcal{K}_{o}^{n}\), the support and radial function of the polar body \(K^{*}\) of K are defined respectively by (see [16, 27])
$$\begin{aligned} h_{K^{*}}(u)=\frac{1}{\rho_{K}(u)} \quad\mbox{and}\quad \rho_{K^{*}}(u)=\frac {1}{h_{K}(u)} \quad\mbox{for all } u\in\mathbb{S}^{n-1}. \end{aligned}$$
(2.1)
Define the Santaló product of \(K\in\mathcal {K}^{n}_{o}\) by \(V(K)V(K^{*})\). The Blaschke-Santaló inequality (see [16, 27]) is one of the fundamental affine isoperimetric inequalities. It states that if \(K\in\mathcal {K}^{n}_{c}\) then
for arbitrary convex bodies \(K_{1},\ldots,K_{m}\in\mathcal{K}^{n}\) and numbers \(\lambda_{1},\ldots,\lambda_{m}\geq0\).
Further, there is a symmetric map S from \((\mathcal{K}^{n})^{n-1}\) into the space of finite Borel measures on \(\mathbb{S}^{n-1}\), the mixed area measure such that, for \(m\in\mathbb{N}\),
for \(K_{1},\ldots,K_{m}\in\mathcal{K}^{n}\) and \(\lambda_{1},\ldots ,\lambda_{m}\geq0\) (where we write \(S(K_{1},\ldots,K_{n-1},\cdot )=S(K_{1},\ldots,K_{n-1})(\cdot)\)). Taking \(K_{1}=\cdots=K_{n-i-1}=K\) and \(K_{n-i}=\cdots=K_{n-1}=B_{n}\) in \(S(K_{1},\ldots,K_{n-1},\cdot)\), we write \(S_{i}(K,\cdot)\) for \(S(K, \ldots, K, B_{n}, \ldots,B_{n}, \cdot)\).
It turns out that the ith surface area measure \(S_{i}(K,\cdot)\) of K, \(i\in\{0,1,\ldots,n-1\}\), on \(\mathbb {S}^{n-1}\) is absolutely continuous with respect to the ordinary surface area measure \(S(K,\cdot )\) of K and has the Radon-Nikodym derivative (see [29])
For real \(p\geq1\), \(K,L\in\mathcal{K}^{n}_{o}\), and \(\alpha, \beta\geq 0\) (not both zero), the Firey p-linear combination \(\alpha\circ K +_{p}\beta\circ L\), is defined by (see [30])
For \(K,L\in\mathcal {K}_{o}^{n}\), \(\varepsilon>0\), and real \(p\geq1\), the p-mixed quermassintegrals \(W_{p,i}(K,L)\) of K and L, \(i\in\{0,1,\ldots,n-1\}\) are defined by (see [28])
Obviously, for \(p=1\), \(W_{1,i}(K,L)\) is just the classical mixed quermassintegral \(W_{i}(K,L)\). For \(i=0\), the p-mixed quermassintegral \(W_{p,0}(K,L)\) is just the p-mixed volume \(V_{p}(K,L)\).
For \(p\geq1\), \(i\in\{0,1,\ldots,n-1\}\), and each \(K\in\mathcal{K}_{o}^{n}\), there exists a positive Borel measure \(S_{p,i}(K,\cdot)\) on \(\mathbb{S}^{n-1}\) such that the p-mixed quermassintegral \(W_{p,i}(K,L)\) has the following integral representation (see [28]):
for all \(L\in\mathcal{K}_{o}^{n}\). It turns out that the measure \(S_{p,i}(K,\cdot)\), \(i\in\{0,1,\ldots,n-1\}\), on \(\mathbb{S}^{n-1}\) is absolutely continuous with respect to \(S_{i}(K,\cdot)\) and has the Radon-Nikodym derivative
For \(K,L\in\mathcal{S}_{o}^{n}\), \(p\geq1\), and \(\lambda,\mu\geq0\) (not both zero), the p-harmonic radial combination \(\lambda\ast K+_{-p}\mu\ast L\in\mathcal{S}_{o}^{n}\) is defined by (see [17])
Note that here ‘\(\varepsilon\ast L\)’ is different from ‘\(\varepsilon\circ L\)’ in Firey p-linear combination.
For \(K,L\in\mathcal{S}_{o}^{n}\), \(\varepsilon>0\), \(p\geq1\), and real \(i\neq n\), the dual p-mixed quermassintegral \(\widetilde{W}_{-p,i}(K,L)\) of K and L is defined by (see [31])
If \(i=0\), we easily see that (2.9) is just the definition of dual p-mixed volume, i.e., \(\widetilde{W}_{-p,0}(K,L)=\widetilde{V}_{-p}(K,L)\).
From (2.9), the integral representation of the dual p-mixed quermassintegrals is given by Wang and Leng [31]: If \(K,L\in \mathcal{S}_{o}^{n}\), \(p\geq1\), and real \(i\neq n\), \(i\neq n+p\), then
Together with (2.8) and (2.10), for \(K\in \mathcal{S}_{o}^{n}\), \(p\geq1\), and \(i\neq n, n+p\), it follows that \(\widetilde{W}_{-p,i}(K,K)=\widetilde{W}_{i}(K)\).
Further, Wang and Leng [31] proved the following analog of the Minkowski inequality for the dual p-mixed quermassintegrals: If \(K,L\in \mathcal{S}_{o}^{n}\), \(p\geq1\), then, for \(i< n\) or \(i>n+p\),
and for \(n< i< n+p\), inequality (2.11) is reverse, with equality in every inequality if and only if K and L are dilates of each other.
Another consequence of inequality (2.11) will be needed (see [32]): Suppose \(K,L\in \mathcal{S}_{o}^{n}\), \(p\geq1\) and \(\lambda, \mu>0\). If real \(i< n\) or \(n< i< n+p\), then
with equality in every inequality if and only if K and L are dilates of each other, and for \(n>n+p\) inequality (2.12) is reverse.
The following result will be needed.
Lemma 2.1
If\(p\geq1\), \(i\in\mathbb{R}\), \(\mathcal{M}\subset\mathcal{S}^{n}_{o}\)is a class of bodies such that\(K,L\in\mathcal{M}\). If
$$\begin{aligned} \widetilde{W}_{-p,i}(K,Q)/\widetilde{W}_{i}(K)= \widetilde {W}_{-p,i}(L,Q)/\widetilde{W}_{i}(L) \quad\textit{for all } Q\in\mathcal{M}, \end{aligned}$$
(2.13)
then\(K=L\).
Proof
Taking \(Q=L\) gives \(\widetilde {W}_{-p,i}(K,L)/\widetilde{W}_{i}(K)=\widetilde {W}_{-p,i}(L,L)/\widetilde{W}_{i}(L)=1\). Now inequality (2.11) gives \(\widetilde{W}_{i}(L)\geq\widetilde {W}_{i}(K)\) with equality if and only if K and L are dilates. Take \(Q=K\) and get \(\widetilde{W}_{i}(K)\geq \widetilde{W}_{i}(L)\) with equality if and only if L and K are dilates. Hence, \(\widetilde{W}_{i}(K)= \widetilde{W}_{i}(L)\), and K and L must be dilates. Thus, \(K=L\). □
By inequality (2.12), for convex bodies, we introduce the following definition: Suppose \(K\in\mathcal{K}^{n}_{o}\) and \(L\in\mathcal{S}^{n}_{o}\). For \(p\geq1\) and \(i\in\{0,1, \ldots,n-1\}\), define \(W_{p,i}(K,L^{*})\) by
Since \(h_{Q^{*}}=1/\rho_{Q}\) for \(Q\in\mathcal{K}^{n}_{o}\), it follows from the integral representation (2.6) that, if L happens to belong to \(\mathcal{K}^{n}_{o}\) (rather than just \(\mathcal{S}^{n}_{o}\)), the new definition of \(W_{p,i}(K,L^{*})\) agrees with the old definition.
2.3 The ith p-curvature function and ith p-curvature image
A convex body \(K\in\mathcal{K}^{n}\) is said to have a continuous ith curvature function \(f_{i}(K,\cdot):\mathbb{S}^{n-1}\rightarrow\mathbb {R}\) if its mixed surface area measure \(S_{i}(K,\cdot)\) is absolutely continuous with respect to the spherical Lebesgue measure S and has the Radon-Nikodym derivative (see [28])
Let \(\mathcal{F}^{n}_{i}\), \(\mathcal{F}^{n}_{i,o}\), \(\mathcal{F}^{n}_{i,c}\) denote a set of all bodies in \(\mathcal{K}^{n}\), \(\mathcal{K}^{n}_{o}\), \(\mathcal{K}^{n}_{c}\), respectively, that have an ith positive continuous curvature function. In particular, \(\mathcal{F}^{n}_{0}:=\mathcal{F}^{n}\), \(\mathcal{F}^{n}_{0,o}:=\mathcal{F}^{n}_{o}\), \(\mathcal{F}^{n}_{0,c}:=\mathcal{F}^{n}_{c}\).
A convex body \(K\in\mathcal{K}^{n}_{o}\) is said to have a p-curvature function \(f_{p}(K,\cdot): \mathbb{S}^{n-1}\rightarrow\mathbb{R}\) if its p-surface area measure \(S_{p}(K,\cdot)\) is absolutely continuous with respect to the spherical Lebesgue measure S and has the Radon-Nikodym derivative (see [17])
Lutwak [17] showed the notion of p-curvature image as follows: For each \(K\in\mathcal{F}^{n}_{o}\) and \(p\geq1\), define \(\Lambda _{p}K\in \mathcal{S}^{n}_{o}\), the p-curvature image of K, by
It should be noted that, for \(p=1\), this definition of curvature image differs from the definition used by the author in [8, 10], and [33].
Recently, Liu et al. [34], Lu and Wang [35], as well as Ma and Liu [24, 36, 37] independently introduced the concept of ith p-curvature function of \(K\in\mathcal{K}^{n}_{o}\) as follows: Let \(p\geq1\), \(i\in\{0,1,\ldots, n-1\}\), a convex body \(K\in\mathcal{K}^{n}_{o}\) is said to have an ith p-curvature function \(f_{p,i}(K,\cdot): \mathbb{S}^{n-1}\rightarrow \mathbb{R}\) if its ith p-surface area measure \(S_{p,i}(K,\cdot)\) is absolutely continuous with respect to the spherical Lebesgue measure S and has the Radon-Nikodym derivative
According to the concept of ith p-curvature function of convex body, Lu and Wang [35] as well as Ma [24] introduced independently the concept of ith p-curvature image of convex body as follows: For each \(K\in\mathcal{F}^{n}_{i,o}\), \(i\in\{ 0,1,\ldots, n-1\}\), and real \(p\geq1\), define \(\Lambda_{p,i}K \in \mathcal{S}^{n}_{o}\), the ith p-curvature image of K, by
For the case \(p=1\) or \(i=0\), the subscript p or i in \(\Lambda _{p,i}\) will often be suppressed. If \(\Lambda_{p,i}K\in\mathcal {K}^{n}_{o}\), write \(\Lambda_{p,i}^{*}K\) for \((\Lambda_{p,i}K)^{*}\). The unusual normalization of definition (2.21) is chosen so that, for the unit ball \(B_{n}\), it follows that \(\Lambda _{p,i}B_{n}=B_{n}\). From definitions (2.17), (2.19), and (2.21), if \(i=0\), then \(\Lambda_{p,0}K=\Lambda_{p}K\). In particular, we note that if \(p=1\) in (2.21), then
An immediate consequence of the definition of the ith p-curvature image and the integral representations of \(W_{p,i}\) and \(\widetilde{W}_{-p,i}\) is the following results.
Proposition 2.2
If\(p\geq1\), \(i\in\{0,1,\ldots,n-1\}\), and\(K\in\mathcal {F}^{n}_{i,o}\), then, for all\(Q\in\mathcal{S}^{n}_{o}\),
The ordinary surface area measure of a polytope is concentrated on a finite set of points of \(\mathbb{S}^{n-1}\) (see, for example, Lutwak [17]). From this, (2.4) and (2.7), it follows that the ith p-surface area measure \(S_{p,i}(P,\cdot)\) of a polytope \(P\in\mathcal{K}^{n}_{o}\) is concentrated on a finite set of points of \(\mathbb{S}^{n-1}\). A direct consequence of this fact and the definition of ith p-affine surface area is as follows.
Proposition 3.4
If\(p\geq1\), and\(P\in\mathcal{K}^{n}_{o}\)is a polytope, then\(\Omega _{p}^{(i)}(P)=0\)for any\(i\in\{0,1,\ldots, n-1\}\).
If \(i=0\), Proposition 3.4 reduces to the isotropy of the p-surface area measures, which was essentially proved in [17] by Lutwak.
Lemma 3.5
If\(p\geq1\)and\(K_{j}\)is a sequence of convex bodies in\(\mathcal {K}^{n}_{o}\)such that\(K_{j}\rightarrow K_{0}\in\mathcal{K}^{n}_{o}\), then, for\(i=0,1,\ldots,n-1\), \(S_{p,i}(K_{j},\cdot)\rightarrow S_{p,i}(K_{0},\cdot)\)weakly.
Proof
Suppose \(f\in C(S^{n-1})\). Since \(K_{j}\rightarrow K_{0}\), by the definition of support function, \(h_{K_{j}}\rightarrow h_{K_{0}}\) uniformly on \(\mathbb{S}^{n-1}\). Since the continuous function \(h_{K_{0}}\) is positive, \(h_{K_{j}}\) are uniformly bounded away from 0. It follows that \(h_{K_{j}}^{1-p}\rightarrow h_{K_{0}}^{1-p}\) uniformly on \(\mathbb{S}^{n-1}\), and thus that
$$f h_{K_{j}}^{1-p}\rightarrow f h_{K_{0}}^{1-p} \quad\mbox{uniformly on } \mathbb{S}^{n-1}. $$
But \(K_{j}\rightarrow K_{0}\) also implies that
$$S_{i}(K_{j},\cdot)\rightarrow S_{i}(K_{0}, \cdot) \quad\mbox{weakly on } \mathbb{S}^{n-1} $$
follows from the weak continuity of surface area measures (see, for example, Schneider [17, 39]). Hence,
Since \(h_{L_{j}}\rightarrow h_{L_{0}}\) uniformly on \(\mathbb{S}^{n-1}\) and \(h_{L}\) is continuous, then \(h_{L_{i}}\) are uniformly bounded on \(\mathbb{S}^{n-1}\). Hence,
$$h_{L_{j}}^{p}\rightarrow h_{L_{0}}^{p} \quad\mbox{uniformly on } \mathbb{S}^{n-1}. $$
By Lemma 3.5\(K_{j}\rightarrow K_{0}\) implies that
$$S_{p,i}(K_{j},\cdot)\rightarrow S_{p,i}(K_{0}, \cdot) \quad\mbox{weakly on } \mathbb{S}^{n-1}. $$
Recall that \(\Lambda_{p,i}K\in\mathcal{S}^{n}_{o}\) is defined by \(f_{p,i}(K,\cdot)=\omega_{n}\rho(\Lambda_{p,i}K,\cdot )^{n+p-i}/\widetilde{W}_{i}(\Lambda_{p,i}K)\). From this and the formula for the ith dual quermassintegrals, it follows that the quantity on the right in (3.1) is just \(\omega_{n}\widetilde {W}_{i}(\Lambda_{p,i}K)^{\frac{p}{n-i}}\). By Proposition 2.2,
The fact that the quantity on the left in (3.2) is no less than the quantity on the right is a simple consequence of the dual p-mixed quermassintegrals inequality (2.11). To see that the quantity on the right in (3.2) is no less than the quantity on the left, take \(Q=\Lambda_{p,i}K\) and note that
An immediate consequence of the definition of the ith p-curvature image and the integral representations of \(\Omega_{p}^{(i)}\) as well as \(\widetilde{W}_{i}\) is as follows.
Proposition 3.10
If\(p\geq1\), \(i\in\{0,1,\ldots,n-1\}\), and\(K\in\mathcal {F}^{n}_{i,o}\), then
By the equality condition of the Blaschke-Santaló inequality (2.2) and Lemma 3.11, equality holds in the inequality of Theorem 1.5 for \(i=0\) if and only if K is an ellipsoid, for \(0< i\leq n-1\) if and only if K is an n-ball centered at the origin. □
In the proof process, we can easily know that for \(i=0\) equality of inequality (1.10) holds if and only if K is an ellipsoid, and for \(0< i\leq n-1\) if and only if all \((n-i)\)-dimensional sub-convex bodies contained in K are \((n-i)\)-ball centered at the origin. □
Remark 3.13
More recently, the author in [26] defined the notion of ith p-geominimal surface area: For \(K\in\mathcal{K}^{n}_{o}\), \(p\geq1\), \(i\in\{0,1,\ldots,n-1\}\), then
The inequality above is (1.8) of the article [26], and from the proof in [26] we know that equality holds in (3.8) if and only if \(K\in\mathcal{W}_{p,i}^{n}\), where symbol \(\mathcal {W}_{p,i}^{n}\) is defined in (3.17).
Some results of this paper can immediately be given by (3.8). For example, Theorem 4.2 of [26] implies our Theorem 1.5; Theorem 1.4 of [26] implies our Theorem 1.6. But we need to note in particular that the condition of equality holds in inequalities (1.9) and (1.10) cannot be determined.
According to the conditions of equality holds in the inequality of Lemma 3.12, we see that equality holds in inequality (1.11) for \(i=0\) if and only if K is an ellipsoid, and for \(0< i\leq n-1\) if and only if K is a ball centered at the origin. □
Suppose\(K,L\in\mathcal{K}^{n}_{o}\), and\(\mathcal{B}\subset\mathcal {K}^{n}_{o}\)is a class of bodies such that\(K,L\in\mathcal{B}\). If\(0\leq i< n\)and\(n-i\neq p>1\), and if
$$\begin{aligned} W_{p,i}(K,Q)=W_{p,i}(L,Q) \quad\textit{for all } Q\in \mathcal{B}, \end{aligned}$$
Take \(Q=\Lambda_{p,i}\phi K\) and note that \(\widetilde{W}_{-p,i}(\phi^{-t}\Lambda_{p,i}K, \phi^{-t}\Lambda_{p,i}K)=\widetilde{W}_{i}(\phi^{-t}\Lambda_{p,i}K)\), it follows that
Together with (3.15) and Lemma 3.17, we immediately get the result. □
Recall that \(\Lambda_{p,i}\) maps \(B_{n}\), the centered unit ball, into \(B_{n}\); i.e., \(\Lambda_{p,i}B_{n}=B_{n}\). Since \((\phi Q)^{*}=\phi^{-t}Q^{*}\), for \(\phi\in O(n)\) and \(Q\in\mathcal {K}^{n}_{o}\), Proposition 3.18 shows that if E is a centered ellipsoid and \(E=\phi B_{n}\), then
It follows from Proposition 3.16 that for \(K\in\mathcal {F}^{n}_{i,o}\) and \(p>1\), the body \(\Lambda_{p,i}K\) is a centered ellipsoid if and only if K is a centered ellipsoid. Define
$$\begin{aligned} \mathcal{W}_{p,i}^{n}=\bigl\{ K\in \mathcal{F}^{n}_{i,o}: \mbox{ there exists } Q\in \mathcal{K}^{n}_{o} \mbox{ with } f_{p,i}(K, \cdot)=h(Q,\cdot )^{-(n+p-i)}\bigr\} . \end{aligned}$$
(3.17)
An immediate consequence of the definition of \(\mathcal{W}_{p,i}^{n}\) and the definition of \(\Lambda_{p,i}\) is the following.
Proposition 3.19
If\(p\geq1\), \(i\in\{0,1,\ldots,n-1\}\)and\(K\in\mathcal{F}^{n}_{i,o}\), then
$$K\in\mathcal{W}_{p,i}^{n} \quad\textit{if and only if}\quad \Lambda _{p,i}K\in\mathcal{K}^{n}_{o}. $$
It follows from Propositions 3.18 and 3.19 that \(\mathcal{W}^{n}_{p,i}\) is an orthogonal transformation invariant class.
Proposition 3.20
Suppose\(K\in\mathcal{F}^{n}_{i,o}\)and\(i\in\{0,1,\ldots,n-1\}\). If\(p\geq1\)and\(\phi\in O(n)\), then\(K\in\mathcal{W}^{n}_{p,i}\)if and only if\(\phi K\in\mathcal{W}^{n}_{p,i}\).
Define
$$\begin{aligned} \mathcal{E}^{n}_{i}=\bigl\{ K\in \mathcal{F}^{n}_{i,o}: K^{*} \mbox{ and } \Lambda_{i}K \mbox{ are dilates}\bigr\} . \end{aligned}$$
(3.18)
For the case \(i=0\), the subscript i in \(\mathcal{E}^{n}_{i}\) will often be suppressed. Namely, \(\mathcal{E}^{n}_{0}=\mathcal{E}^{n}\). Obviously, \(\mathcal{E}^{n}_{i}\subset\mathcal{W}^{n}_{p,i}\) for all \(p\geq1\) and \(i\in\{0,1,\ldots,n-1\}\). From Proposition 3.18 and (3.16) it follows that all centered ellipsoids belong to \(\mathcal{E}^{n}_{i}\). If \(K\in\mathcal {E}^{n}_{i}\), then from definition (2.21) of the ith p-curvature image, (2.19) and (2.22), and noting that \(\Lambda_{i}K=\lambda K^{*}\) with arbitrary \(\lambda >0\), we have
for all \(p\geq1\) and \(i\in\{0,1,\ldots,n-1\}\).
On the other hand, if \(p\geq1\), \(i\in\{0,1,\ldots,n-1\}\) and the body \(K\in\mathcal{F}^{n}_{i,o}\) is such that \(\Lambda_{p,i}K\) and \(K^{*}\) are dilates, then let \(\Lambda _{p,i}K=\lambda K^{*}\) with \(\lambda>0\). From definition (2.21) of the ith p-curvature image, (2.19), (2.20), and (2.1) as well as (3.14), it follows that
Comparing to (2.22) and (3.19) and using Proposition 3.19, we let \(Q=\lambda^{\frac{1-p}{n-i-1}}K\) and \(\lambda K^{*}=\Lambda_{i}Q\). Then from (3.14) we get
Accordingly, \(K\in\mathcal{E}^{n}_{i}\). Thus, the sets defined for \(p\geq1\) and \(i\in\{0,1,\ldots,n-1\}\) by \(\mathcal{E}^{n}_{p,i}=\{K\in\mathcal{F}^{n}_{i,o}: K^{*} \mbox { and } \Lambda_{p,i}K \mbox{ are dilates}\}\) are one and the same. Namely, \(\mathcal{E}^{n}_{p,i}=\mathcal{E}^{n}_{i}\) for all \(p\geq1\) and \(i\in\{0,1,\ldots,n-1\}\).
It is known that if ∂K is a regular \(C^{2}\) hypersurface and \(K\in\mathcal{E}^{n}\), then K must be an ellipsoid. It is known that if \(K\in\mathcal{E}^{n}\) and K is a body of revolution, then K must be an ellipsoid. It is also known that \(\mathcal{E}^{2}\) consists only of centered ellipses. For all these facts, see Petty [12]. It has been conjectured that \(\mathcal{E}^{n}\) is exactly the class of centered ellipsoids (see [17]). Therefore, we conjecture that \(\mathcal{E}^{n}_{i}\), \(i=0,1,\ldots,n-1\), are exactly the class of centered ellipsoids. None of the facts stated in this paragraph will be used in this article.
For \(K\in\mathcal{K}^{n}_{o}\), define the ith p-curvature ratio of K as
Since \(K\in\mathcal{E}^{n}_{i}\) implies that \(\Lambda _{p,i}K=[W_{i}(K)/\omega_{n}]^{1/p}K^{*}\), it follows immediately that the ith p-curvature ratio of \(K\in\mathcal{E}^{n}_{i}\) equals \(W_{i}(K)\widetilde{W}_{i}(K^{*})\) of K. Namely, if \(K\in\mathcal{E}^{n}_{i}\), then
with equality if and only if \(\Lambda_{p,i}K\) and \(K^{*}\) are dilates. □
For bodies with ith continuous curvature functions, the equality conditions for the inequality of Proposition 3.14 are easily obtained by combining Propositions 3.10 and 3.21.
Theorem 3.22
If\(p\geq1\), \(i\in\{0,1,\ldots,n-1\}\), and\(K\in\mathcal{F}^{n}_{i,o}\), then
with equality if and only if there exists a constant\(c>0\)such that\(\rho_{L}=c /h_{K}\)almost everywhere with respect to\(S_{i}(K,\cdot)\).
Suppose \(1\leq p< q\) and \(K\in\mathcal{K}^{n}_{o}\) with \(L\in\mathcal {S}^{n}_{o}\). From the integral representation of \(W_{p,i}(K,L^{*})\) the easy estimate follows
Suppose\(K\in\mathcal{K}^{n}_{o}\), \(L\in\mathcal{S}^{n}_{o}\), and\(i\in \{0,1,\ldots,n-1\}\). The function defined on\([1,\infty)\)by
$$p\mapsto W_{p,i}\bigl(K,L^{*}\bigr) $$
is continuous.
From the equality conditions of Proposition 3.21 it follows that if \(K\in\mathcal{E}^{n}_{i}\), then the ith p-curvature ratios are independent of p. The next proposition provides a strong converse by showing that unless \(K\in\mathcal {E}^{n}_{i}\), the ith p-curvature ratios are (strictly) monotone increasing in p.
Proposition 4.4
If\(K\in\mathcal{F}^{n}_{i,o}\), \(i\in\{0,1,\ldots,n-1\}\), and\(1\leq p< q\), then
The dual ith p-mixed quermassintegrals inequality (2.11) now gives the desired inequality and shows that equality implies that \(\Lambda_{p,i}K\) and \(\Lambda _{q,i}K\) must be dilates. But definition (2.21) of ith p-curvature images and definition (2.22) of ith curvature images, together with (2.19), show that \(\Lambda_{p,i}K\) and \(\Lambda_{q,i}K\) can be dilates if and only if \(K\in\mathcal{E}^{n}_{i}\). □
The following cyclic inequality will be needed.
Proposition 4.5
If\(K\in\mathcal{F}^{n}_{i,o}\), \(i\in\{0,1,\ldots,n-1\}\), and\(1\leq p< q< r\), then
The Hölder inequality and formula (2.8) for ith dual quermassintegrals now yield the desired inequality and show that equality is possible if and only if \(\Lambda_{p,i}K\) and \(\Lambda_{r,i}K\) are dilates, or equivalently, if and only if \(K\in\mathcal{E}^{n}_{i}\). □
In contrast to the inequality of Proposition 4.4, there is the following proposition.
Proposition 4.6
If\(K\in\mathcal{F}^{n}_{i,o}\), \(i\in\{0,1,\ldots,n-1\}\), and\(1\leq p< q\), then
The Hölder inequality, together with formula (2.8) for ith dual quermassintegrals, now yields the desired inequality and shows that equality can occur if and only if \(\Lambda_{p,i}K\) and \(K^{*}\) are dilates, or equivalently, if and only if \(K\in\mathcal{E}^{n}_{i}\). □
It turns out that there is an inequality between the ith p-affine surface areas of a convex body that is similar to the classical cyclic inequality between the quermassintegrals of the convex body.
Theorem 4.7
Suppose\(K\in\mathcal{K}^{n}_{o}\), \(i\in\{0,1,\ldots, n-1\}\)and\(1\leq p< q< r\). Then
Note that if K is a polytope, then there is equality in the inequality of Theorem 4.7. For bodies with ith continuous curvature functions, the equality conditions of inequality of Theorem 4.7 are easily obtained from Propositions 3.10 and 4.5.
Proposition 4.8
Suppose\(K\in\mathcal{F}^{n}_{i,o}\), \(i\in\{0,1,\ldots, n-1\}\)and\(1\leq p< q< r\). Then
Note that if K is a polytope, then there is equality in inequality (4.11). For bodies with ith continuous curvature functions, the equality conditions of the inequality of Proposition 4.9 follow directly from Propositions 3.10 and 4.4. This completes the proof of Theorem 1.9. □
In contrast to the inequality of Proposition 4.9, we have the following proposition.
The inequality of Proposition 4.10 follows immediately from the definition of ith p-affine surface area once the following fact is established: Given \(Q\in\mathcal{S}^{n}_{o}\), there exists \(\overline{Q}\in\mathcal{S}^{n}_{o}\) such that
Together with (4.15) and (4.16), we show that (4.13), and this completes the argument. □
If K is a polytope there is equality in the inequality of Proposition 4.10. For bodies with ith continuous curvature functions, the equality conditions in inequality (4.12) follow immediately from Propositions 3.10 and 4.6.
The cyclic inequality of Theorem 4.7 shows that the function defined on \([1,\infty)\) by
$$p\mapsto(n+p-i)\log\Omega_{p}^{(i)}(K) $$
is convex. The continuity of this function on \([1,\infty)\) follows from this and Proposition 4.9. The continuity of this function immediately gives the following.
Proposition 4.13
If\(K\in\mathcal{K}^{n}_{o}\)and\(i\in\{0,1,\ldots,n-1\}\), then the function defined on\([1,\infty)\)by
$$p\mapsto\Omega_{p}^{(i)}(K) $$
is continuous.
5 Extremal ith affine surface area
Define the generalized Santaló product of \(K\in \mathcal{K}^{n}_{o}\) by \(W_{i}(K)\widetilde{W}_{i}(K^{*})\). Proposition 4.9 states that, for \(K\in\mathcal{K}^{n}_{o}\), the ith p-affine area ratio
is monotone nondecreasing in p, and Theorem 3.22 states that this ratio is bounded by the generalized Santaló product of K.
In order to facilitate the formulation of the ith p-affine area ratio for the case \(p=\infty\), it will be helpful to introduce a quermassintegrals-normalized version of ith p-mixed quermassintegrals. If K, L are convex bodies that contain the origin in their interiors, then for each real \(p > 0\) define
Note that \(\frac{1}{n}h_{K}\,dS_{i}(K,\cdot)/W_{i}(K)=\frac {1}{n}h_{K}^{1-i}\,dS(K,\cdot)/W_{i}(K)\) is a probability measure on supp \(S_{i}(K,\cdot)\) (or \(S(K,\cdot)\)).
According to (5.1), we can define ith ∞-mixed quermassintegrals, \(W_{\infty,i}(K,L)\), of \(K, L\in\mathcal {S}^{n}_{o}\) by
From definition (1.7) of ith p-affine surface area \(\Omega _{p}^{(i)}(K)\), definition (5.1) of \(\overline{W}_{\infty ,i}(K,L)\) and definition (5.2) of \(W_{\infty,i}(K,L)\), we have
An immediate consequence of Proposition 3.3 and the definition of \(\Omega_{\infty}^{(i)}(K)\) is that \(\Omega_{\infty}^{(i)}(K)\) is invariant under orthogonal transformations of K.
Proposition 5.1
If\(i\in\{0,1,\ldots n-1\}\)and\(K\in\mathcal{K}^{n}_{o}\), then
If K has the ith continuous curvature function, then the equality conditions in Proposition 5.3 are easily obtained. Note that from Theorem 1.9 it follows that if \(K\in\mathcal{F}^{n}_{i,o}\setminus\mathcal {E}^{n}_{i}\), then the limit
The inequality of Proposition 5.5 compares \(W_{i}(K)\widetilde {W}_{i}(K^{*})\) and \(\Omega_{\infty}^{(i)}(K)\) for \(K\in\mathcal{K}^{n}_{o}\). The next proposition shows that for an important class of bodies, these quantities are the same.
Proposition 5.7
If\(i\in\{0,1,\ldots,n-1\}\)and\(K\in\mathcal{F}^{n}_{i,o}\), then
Since \(f_{p,i}(K,\cdot)=h_{K}^{1-p}f_{i}(K,\cdot)\), and \(h_{K}\) and \(f_{i}(K,\cdot)\) are positive continuous functions, it is easily seen that
$$\lim_{p\rightarrow\infty}f_{p,i}(K,\cdot)^{\frac {n-i}{n+p-i}}=h_{K}^{-(n-i)} \quad\mbox{uniformly on } \mathbb{S}^{n-1}. $$
The formula for the ith dual quermassintegrals, together with the integral representation of Theorem 1.4, now yields the desired result. □
Proposition 5.7 shows that when restricted to \(\mathcal {F}^{n}_{i,o}\), the function \(\Omega_{\infty}^{(i)}: \mathcal {F}^{n}_{i,o}\rightarrow(0,\infty)\) is continuous.
When Theorem 1.9 and Proposition 5.4 are combined with Proposition 5.7, result is that for \(K\in\mathcal{F}^{n}_{i,o}\) and \(1\leq p\leq q\),
with equality in inequality for\(i=0\)if and only ifKis an ellipsoid, and for\(0< i\leq n-1\)if and only ifKis a ball.
Obviously, the case \(i=0\) of Conjecture 5.8 is just the p-affine isoperimetric inequality by Lutwak (see [17]).
Acknowledgements
This work is supported by the National Natural Science Foundation of China (Grant No. 11161019) and is supported by the Science and Technology Plan of the Gansu Province (Grant No. 145RJZG227), and is partly supported by the National Natural Science Foundation of China (Grant No. 11371224). The referee of this paper proposed many very valuable comments and suggestions to improve the accuracy and readability of the original manuscript. We would like to express our most sincere thanks to the anonymous referee.
Open Access This article is distributed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made.
Competing interests
The authors declare that they have no competing interests.
Authors’ contributions
The authors completed the paper and read and approved the final manuscript.