Large Poisson-Voronoi Cells and Crofton Cells
|
|
- Carol Lawson
- 6 years ago
- Views:
Transcription
1 Large Poisson-Voronoi Cells an Crofton Cells Daniel Hug Matthias Reitzner Rolf Schneier Abstract. It is prove that the shape of the typical cell of a stationary Poisson-Voronoi tessellation in Eucliean space, uner the conition that the volume of the typical cell is large, must be close to spherical shape, with high probability. The same hols if the volume is replace by the surface area or other suitable functionals. Similar results are establishe for the zero cell of a stationary an isotropic Poisson hyperplane tessellation. Keywors: Poisson-Voronoi tessellation; typical cell; Poisson hyperplane tessellation; Crofton cell; asymptotic shape; D.G. Kenall s conjecture; intrinsic volume AMS 2000 Subject Classification: Primary 60D05 Seconary 52A22, 52A40 1 Introuction In this paper, we continue a line of research that began with a problem of D.G. Kenall (see the forewor to the first eition of [17]). He conjecture that the shape of the zero cell (or Crofton cell) of a stationary an isotropic Poisson line process in the plane, given that the area of the cell tens to infinity, must become circular. Contributions to Kenall s question are ue to Miles [11] an Golman [3], an the conjecture was prove by Kovalenko [8], [10]. In [7], Kovalenko s result was extene to higher imensions an to stationary, but not necessarily isotropic Poisson hyperplane processes. It was also strengthene, by estimating the probability of large eviations from spherical shape, given that the volume of the zero cell lies in a prescribe interval. In the present paper, we prove an analogous result for the typical cell of a stationary Poisson-Voronoi tessellation (mosaic) of -imensional space. Thus we exten an strengthen a result of Kovalenko [9] in the planar case. We further exten this result by consiering, in aition to the volume functional, also the kth intrinsic volume, k = 1,..., 1. This inclues cells of large surface area or of large mean with. The result from [7] on Crofton cells of stationary Poisson hyperplane processes with large volume is also extene to the kth intrinsic volume, but only for k 2 an uner the aitional assumption of isotropy. For both types of ranom polytopes, Poisson-Voronoi cells an isotropic Crofton cells, we can also replace (somewhat easier) the conition of large volume by the conition of Postal aress: Mathematisches Institut, Albert-Luwigs-Universität, Eckerstr. 1, D Freiburg, Germany. aress: aniel.hug@math.uni-freiburg.e Postal aress: Institut für Analysis un Technische Mathematik, Technische Universität Wien, Wiener Hauptstrasse 8-10, A-1040 Wien, Austria. aress: mreitzne@mail.zserv.tuwien.ac.at Postal aress: Mathematisches Institut, Albert-Luwigs-Universität, Eckerstr. 1, D Freiburg, Germany. aress: rolf.schneier@math.uni-freiburg.e 1
2 large inraius. This is suggeste by consierations of Miles [11] on Crofton cells in the plane an by the work of Calka [1] on planar Poisson-Voronoi tessellations. Finally, we mention here that cells of large volume in Poisson-Delaunay tessellations were treate in [6]; such cells ten to be regular simplices. Let A be a locally finite point set in Eucliean space R (with scalar prouct, an norm ), where 2. For x A, the Voronoi cell of x with respect to A is efine by C(x, A) := y R : y x y a for all a A. Let X be a stationary Poisson point process of intensity λ > 0 in R. (In treating simple point processes, we conveniently ientify a simple counting measure with its support.) Then X := C(x, X) : x X is the Poisson-Voronoi tessellation erive from X. Let Z enote the typical cell of X (we recall its efinition in Section 2). For a convex boy (non-empty, compact, convex set) K R, we enote the volume by v (K). The (conveniently renormalize) intrinsic volumes v 0 (K),..., v 1 (K) can be efine by means of the Steiner formula v (K + ɛb ) = ( ) ɛ k v k (K), ɛ 0. k k=0 Here B := x R : x 1 is the unit ball. Equivalently, v i (K) is the mixe volume V (K,..., K, B,..., B ), where K appears i times an B appears i times. The functional W j = v j is known as the jth quermassintegral. In particular, v 1 is the surface area, an (2/κ )v 1 is the mean with; here κ := v (B ). More information is foun in [15]. Let K R be a compact set with o K an containing more than one point. In orer to measure the eviation of K from a ball with centre o, we efine ϑ(k) := R o ρ o R o + ρ o, where R o is the raius of the smallest ball with centre o containing K an ρ o is the raius (possibly zero) of the largest ball with centre o containe in K. By P we enote the unerlying probability, an P( ) is a conitional probability. Theorem 1. Let X enote the Poisson-Voronoi tessellation erive from a stationary Poisson point process with intensity λ > 0 in R ; let Z be its typical cell. Let k 1,...,. There is a positive constant c 0 epening only on the imension such that the following is true. If ɛ (0, 1) an I = [a, b) is any interval (possibly b = ) with a /k λ σ 0, for some constant σ 0 > 0, then P(ϑ(Z) ɛ v k (Z) I) c exp c 0 ɛ (+3)/2 a /k λ, where c is a constant epening only on, ɛ an σ 0. In particular, lim P(ϑ(Z) ɛ v k(z) a) = 0 for every ɛ > 0, (1) a 2
3 but Theorem 1 provies much stronger information. The relation (1) can equivalently be formulate as follows (see Section 2 for further explanation). Corollary. The conitional law for the shape of Z, given a lower boun for v k (Z), converges weakly, as that lower boun tens to infinity, to the law concentrate at the shape of a ball. Theorem 1 will be prove in Section 6, after preliminary explanations in Section 2 an preparations in Sections 3 to 5. In [7], a similar result was obtaine for the volume of the zero cell (also calle Crofton cell) of the tessellation generate by a stationary Poisson hyperplane process. We will inicate in Section 7 how, uner the aitional assumption of isotropy, this result can be extene to the kth intrinsic volume, k = 2,...,. As in [7], we measure the eviation of the shape of a convex boy K R with interior points from spherical shape by r B, which we abbreviate by r, thus r (K) := mins/r 1 : rb + z K sb + z, z R, r, s > 0. Theorem 2. Let Z o be the zero cell of the tessellation inuce by a stationary isotropic Poisson hyperplane process in R with intensity λ > 0. Let k 2,...,. There is a positive constant c 0 epening only on the imension such that the following is true. If ɛ (0, 1) an I = [a, b) is any interval (possibly b = ) with a 1/k λ σ 0, for some constant σ 0 > 0, then P(r (Z o ) ɛ v k (Z o ) I) c exp c 0 ɛ (+3)/2 a 1/k λ, where c is a constant epening only on, ɛ an σ 0. The case of the volume is inclue here for k =. We remark that in this case the inequality of Theorem 2 is sharper (in its epenence on ɛ) than Theorem 1 of [7], specialize to the isotropic case. The reason for this improvement lies in the fact that in the isotropic case sharper stability estimates from convex geometry are available. Unfortunately, our metho of proof oes not permit us to treat the case k = 1, which in the plane is the case of the perimeter, alreay stuie by Miles [11] in his heuristic approach. In aition to ρ o (K) efine above, we enote by ρ(k) the raius of the largest ball containe in the convex boy K. Theorem 3. Let Z an Z o be efine as in Theorems 1 an 2, respectively. There is a positive constant c 0 epening only on the imension such that the following is true. If ɛ (0, 1) an I = [a, b) is any interval (possibly b = ) with a λ σ 0 in the case of Z, respectively aλ σ 0 in the case of Z o, with some constant σ 0 > 0, then P(ϑ(Z) ɛ ρ o (Z) I) c exp c 0 ɛ (+1)/2 a λ an P(r (Z o ) ɛ ρ(z o ) I) c exp c 0 ɛ (+1)/2 aλ, where c is a constant epening only on, ɛ an σ 0. 3
4 The proof will be sketche in Section 8. A few wors about the choices of the shape parameters ϑ(k) an r (K) seem in orer. For the formulation of our estimates, we want a simple, similarity invariant measure for the eviation of the shape of a convex boy from spherical shape. Such a measure appears implicitly in relation (2) of Miles [11], an explicitly in the paper of Kovalenko [10]. The number r (K) use above is the extension of the latter to higher imensions. For an interior point z of the compact convex set K, let Rz be the raius of the smallest ball with centre z containing K, an ρz the raius of the largest ball with centre z containe in K. Then r (K) is the minimum, over all z in the interior of K, of the quotient (Rz ρz)/ρz. In the case of the typical cell of a Poisson-Voronoi tessellation, for which the nucleus o is a istinguishe point, we get a sharper result if we use instea the eviation measure (Ro ρo)/ρo, or, which for boies close to a ball with centre o is essentially equivalent, (Ro ρo)/(ro +ρo) = ϑ(k). Here we have chosen the latter, since for this the crucial stability estimate (13) below takes a simple form. Finally, we mention that Golman [3], in the planar case, consiers the first eigenvalue of the Laplacian for the Dirichlet problem an fins the same asymptotic behaviour for large Crofton cells an their incircles, but this oes apparently not lea to explicit geometric estimates for the eviation from circular shape. 2 The typical cell of a Poisson-Voronoi tessellation In general, the notion of the typical cell of a stationary ranom tessellation requires the choice of a centroi function (e.g., see Møller [12, Section 3.2]), but for Voronoi cells there is a canonical choice, the nucleus. Let X be the Poisson-Voronoi tessellation generate by a stationary Poisson point process X. Let K o enote the space of convex boies K in R with o K, equippe with the Hausorff metric an corresponing Borel structure. The istribution Q of the typical cell of X can be efine by Q(A) = 1 λ E x X 1 A (C(x, X) x)1 B (x) (E enotes mathematical expectation) for Borel sets A Ko, where B R is an arbitrary Borel set with λ (B) = 1; here λ enotes Lebesgue measure in R. The istribution Q is commonly interprete as the conitional istribution of C(o, X) given that o X. An intuitive interpretation follows from the fact that stationary Poisson-Voronoi tessellations are mixing ([16, Satz 6.4.1]) an hence ergoic. This implies that hols with probability one. Q(A) = lim r car x X rb : C(x, X) x A car ( X rb ) Since X is a stationary Poisson process, it follows from Slivnyak s theorem that the typical cell of the Poisson-Voronoi tessellation X is equal in istribution to the ranom polytope Z = C(o, X o) (see [12, Remark 4.1.1]; see also [13]). Hence, we can consier Z as the typical cell of X, an for this we obtain a convenient representation. For x R, we efine H(x) := y R : y, x = x 2 /2, H (x) := y R : y, x x 2 /2, 4
5 so that H(x) is the mi hyperplane of o an x. Then Z = H (x), thus Z is the zero cell of the tessellation inuce by the Poisson hyperplane process x X Y := H(x) : x X. The intensity measure E Y ( ) of this process can be represente as follows (recall that Y enotes a ranom simple counting measure on the space of hyperplanes as well as its support). For a Borel set A in the space of hyperplanes, we have Writing E Y (A) = E car (A Y ) = E car x X : H(x) A = λ λ (x R : H(x) A). H(u, t) := y R : y, u = t, H (u, t) := y R : y, u t for u S 1 := x R : x = 1 an t R, an introucing polar coorinates, we get E Y ( ) = 2 λ 1H(u, t) t 1 t σ(u), (2) S 1 where σ enotes spherical Lebesgue measure on the unit sphere S 1. 0 In particular, for K K o let H K be the set of all hyperplanes H R with H K. Then (2) gives E Y (H K ) = 2 λu(k) (3) with U(K) := 1 where h(k, ) is the support function of K. Writing S 1 h(k, u) σ(u), (4) Φ(K) := y R : H(2y) K, we have U(K) = λ (Φ(K)). The star boy Φ(K) is the union of all close balls having a iameter segment [o, x] with x K. The relation (3), together with the Poisson property of the hyperplane process Y, implies that P(Y (H K ) = n) = [2 U(K)λ] n exp 2 U(K)λ (5) n! for K K o an n N 0. It is now clear that we are in a similar situation as in [7]. There, the zero cell of a stationary (not necessarily isotropic) Poisson hyperplane process, with intensity measure given by [7, (2)], was stuie. This process is now replace by the isotropic, non-stationary Poisson hyperplane process Y, with intensity measure given by (2). The functional U(K), efine by (4), will play the role of the mixe volume V 1 (B, K) in [7] (up to imensional factors). In aition to the volume functional consiere in [7], we now treat general intrinsic volumes. All these ifferences require a number of changes an new arguments, but also some parallel 5
6 reasoning is possible. In the latter cases, we will be brief an just list how the arguments of [7] have to be moifie. In analogy to Kenall s original conjecture, of which a more general version was treate in [7], we have formulate the Corollary after Theorem 1. We make this more precise. As a space of shapes, suitable for our purpose, we may take the space S (with the Hausorff metric) of convex boies containing o an with circumraius one. The ilate version of Z containe in S is enote by Z. The conitional law for the shape of Z, given the lower boun a for v k (Z), can be efine as the probability measure on S given by µ a ( ) := P(Z v k (Z) a). We are asserting that lim a µ a = δ B weakly, where δ B is the Dirac measure on S concentrate at the ball B (as we are prescribing the centre of the ball, this is a slightly stronger assertion than formulate in the Corollary). Thus, we have to prove that lim sup µ a (C) δ B (C) (6) a for every close set C S. This is trivial if B C. Let B / C. Since the eviation measure ϑ is continuous an positive on C, there exists ɛ > 0 such that C K S : ϑ(k) ɛ. This implies µ a (C) µ a (K S : ϑ(k) ɛ) = P(ϑ(Z) ɛ v k (Z) a) 0 for a, by (1). This proves the Corollary. Conversely, the assertion of the Corollary implies (1): given ɛ > 0, let C := K S : ϑ(k) ɛ, then C is close an B / C, hence (6) yiels (1). As the etails of the proofs for Theorems 1 3 are a bit technical, we want to sketch here the main lines of the reasoning, taking Theorem 1 as an example. Let k 1,...,. In the first instance, we are intereste in bouning the probability P(ϑ(Z) ɛ v k (Z) a) from above, for given ɛ > 0 an large a. This conitional probability is a quotient. A lower boun for the enominator P(v k (Z) a) follows immeiately from (5): the ball B a := (a/κ ) 1/k B satisfies v k (B a ) = a, hence P(v k (Z) a) P(Y (H Ba ) = 0) = exp 2 U(B a )λ = exp 2 κ 1 /k a /k λ. (7) To obtain an upper boun for the numerator P(ϑ(Z) ɛ, v k (Z) a), we use the geometric inequality U(K) κ 1 /k v k (K) /k, (8) in a strengthene form. Equality in (8) hols if an only if K is a ball with centre o. Let K K o be a convex boy satisfying ϑ(k) ɛ > 0. It can be prove (Lemma 1) that with f(ɛ) > 0. If now K satisfies this gives U(K) (1 + f(ɛ))κ 1 /k v k (K) /k ϑ(k) ɛ an v k (K) a, (9) P(Y (H K ) = 0) = exp 2 U(K)λ exp (1 + f(ɛ))2 κ 1 /k a /k λ. Since a convex boy containe in the interior of the cell Z oes not meet any hyperplane of Y, one might now hope that this estimate remains essentially true if the fixe convex boy K 6
7 is replace by the cell Z, in the cases where the latter satisfies the inequalities (9). Although this replacement is not legitimate, a similar an slightly weaker inequality of the form P(ϑ(Z) ɛ, v k (Z) a) c exp (1 + g(ɛ))2 κ 1 /k a /k λ, (10) with a constant c > 0 an g(ɛ) > 0, might be true. If this hols, then iviing (10) by (7) immeiately gives P(ϑ(Z) ɛ v k (Z) a) c exp g(ɛ)2 κ 1 /k a /k λ, which is of the require type. An estimate of type (10) can inee be prove, but at first not for v k (Z) in intervals [a, ), but in intervals a(1, 2). This is one in Lemmas 4 an 6. The istinction of two cases is necessary since it turns out that cells which are in a sense too elongate nee an extra treatment. The convex boies are classifie by an integer parameter m such that increasing m means increasing elongation. The auxiliary geometric Lemma 3 provies inner an outer inclusion estimates for boies of given elongation. For large m, the estimate of Lemma 4 can be use; here the conition ϑ(z) ɛ oes not play a role. Its proof uses only elementary geometric arguments. For small m, the conition ϑ(z) ɛ is essential. Lemma 6 contains the relevant estimate. It is here that the geometric stability result of Lemma 1 is neee. Moreover, the approximation result for polytopes expresse in Lemma 5 is require for the reuction to a situation involving only a fixe number of hyperplanes. The further Lemmas 7, 8, 9 are neee to pass from intervals a(1, 2) to intervals a(1, 1 + h), with fixe h; the extension is achieve by a transformation. The obtaine estimates are then combine into Lemma 10, which gives an upper estimate for the probability P(ϑ(Z) ɛ, v k (Z) a(1, 1+h)). Since this upper boun contains h as a factor, it is necessary to estimate the enominator P(v k (Z) a(1, 1 + h)) from below by a suitable boun which is also linear in h. This is achieve in Lemma 2. Its proof is essentially constructive, exhibiting sufficiently many realizations of Y for which the event v k (Z) a(1, 1 + h) occurs. In both, Lemmas 2 an 10, the number h must be sufficiently small. The final proof of Theorem 1 extens the estimates from the special intervals a(1, 1 + h), with small h, to general intervals [a, b), by a covering argument, as in the proof of Theorem 1 in [7]. 3 A stability estimate For K K o, we trivially have K Φ(K), hence U(K) v (K). Here equality hols if an only if K is a ball with centre o (this follows from the consierations below, but can also be shown irectly). A similar inequality can be obtaine for the other intrinsic volumes. In the following, we write h(k, ) =: h K for the support function of K. Integrations with respect to σ exten over S 1. By Höler s inequality, U(K) 1 (κ ) 1 ( h K σ), an since h K σ = v 1 (K), we get U(K) 1/ κ (1 )/ v 1 (K). 7
8 A well-known inequality (e.g., [15, p. 334]) says that for k = 1,...,. Hence, v 1 (K) k κ k 1 v k (K) (11) U(K) κ 1 /k v k (K) /k. (12) Equality for a number k 1,..., hols if an only if K is a ball with centre o. We will nee an improve version of (12), in the form of a stability estimate. The following proof combines techniques evelope for obtaining stability results relate to Höler s inequality an to isoperimetric inequalities such as (11). Lemma 1. For K K o an k 1,...,, U(K) (1 + γϑ(k) (+3)/2 )κ 1 /k v k (K) /k, (13) where γ is a positive constant epening only on the imension. Proof. We assume that K contains more than one point; otherwise the assertion is trivial. In orer to improve Höler s inequality, we use Lemma 4.2 of Garner an Vassallo [2]. There we put m = 2, f 0 = f 2 = 1, f 1 = h K, w 1 = 1/, w 2 = ( 1)/ (hence w = 1/), an obtain hk σ 1 ( h K σ ) 1/ ( ) ( 1)/ 1 σ 1 [ h /2 K ( h K σ ) 1/2 1 ( 1 σ ) 1/2 ] 2 σ =: β(k). Using (1 β)(1 + β) 1, we euce that U(K) 1/ (1 + β(k))κ (1 )/ v 1 (K). (14) Next, we establish an estimate of the form β(k) cϑ(k) α with α > 0. For this, we argue similarly as in the proof of Lemma 1 in [5] (which is reprouce in [4], see inequality (2.3.3)). From now on in this paper, c 1, c 2,... enote constants epening only on the imension, except in those cases where other epenences are explicitly inicate. Let K Ko be given. Since β an ϑ are invariant uner ilatations, we can normalize K an assume that h K σ = 1 σ = κ, (15) which permits us to obtain the following relations. From (12) we get κ = U(K) κ 1 v 1 (K) c 1 D(K), where D enotes the iameter, hence D(K) c 2 an therefore h K c 2. (16) Moreover, β(k) = c 3 ( h /2 K 1 ) 2 σ. 8
9 Suppose that ρ o B K R o B, where ρ o is maximal an R o is minimal. It follows from (15) that ρ o 1 R o. Case 1: R o 1 1 ρ o. We put R o = 1 + h, then ϑ(k) = R o ρ o R o + ρ o 2h. (17) There exists a vector u 0 S 1 such that h K (u 0 ) = R o = 1 + h, an the point p = (1 + h)u 0 belongs to K. For u S 1, let E(u) be the hyperplane through p an orthogonal to u. Let ω enote the angle between u an u 0, an let ω 0 be the angle between u 0 an any v such that E(v) is tangent to B. Then cos ω 0 = 1/(1 + h) an (with σ 1 := ( 1)κ 1 ) ( h /2 K 1 ) 2 σ σ 1 ω0 0 [(1 + h) cos ω] /2 1 2 (sin ω) 2 ω. We set (1 + h) /2 1 =: α an substitute [(1 + h) cos ω] /2 1 = αx, to obtain ( ) h /2 2 K 1 σ 2σ 1α 3 (1 + h) 2 By (16), we can estimate an, for 0 x 1, This gives The function 1 0 x 2 [(1 + h) 2 (αx + 1) 4/ ] ( 3)/2 (αx + 1) (2 )/ x. (1 + h) 2 c 4 (αx + 1) (2 )/ (α + 1) (2 )/ = (1 + h) (2 )/2 c 1/ β(k) c 5 α 3 x 2 [(1 + h) 2 (αx + 1) 4/ ] ( 3)/2 x. 0 f(x) := (1 + h) 2 (αx + 1) 4/, 0 x 1, satisfies f(0) = h(h + 2), f(1) = 0, an f (1) = (4/)α(1 + h) (4 )/2. It is convex for 4 an concave for = 2, 3. For 4 we euce that f(x) 4 α(1 + h)(4 )/2 (1 x) c 6 α(1 x), an for = 2, 3 we get Together with this yiels From (17) we now get f(x) h(h + 2)(1 x) h(1 x). α = (1 + h) /2 1 2 h β(k) c 7 h (+3)/2. β(k) c 8 ϑ(k) (+3)/2. 9
10 Case 2: R o 1 < 1 ρ o. We put ρ o = 1 h, then R o < 1 + h, hence an K (1 + h)b (18) ϑ(k) 2h. (19) There is a vector u 0 S 1 such that h K (u 0 ) = ρ o = 1 h, an the hyperplane G through ρ o u 0 orthogonal to u 0 supports K. Let p ( (1 + h)b ) G. Let u 1 S 1 be the vector, positively spanne by u 0 an p, that is orthogonal to a support plane of B through p. Let ω 0 be the angle between u 0 an u 1. For u S 1, let ω u be the angle between u 0 an u. If ω u ω 0, then h K (u) 1, hence 1 h K (u) /2 1 h K (u) 0 an thus β(k) c 3 (1 h K (u)) 2 1ω u ω 0 σ(u). Now exactly the argument of Case 2 in [5, pp ] leas to which together with (19) gives β(k) c 9 h (+3)/2, β(k) c 10 ϑ(k) (+3)/2. In each case, we conclue from (14) an (11) the inequality (13). 4 Probabilities involving small intervals In this section we prove an inequality which replaces the easily obtaine estimate (7) in the case where v k (Z) is containe in a boune interval. For A R we efine Z(A) := x A H (x) an set Z(x 1,..., x n ) := Z(x 1,..., x n ). For the ranom polytope Z( X), the typical cell of X, we retain the notation Z. Let k 1,..., be fixe. A main feature of the following estimate is the linear epenence of the lower boun on the length h of the interval (1, 1 + h). A similar result, for the zero cell of a stationary Poisson hyperplane process an the function v, is given in [7] as Lemma 3.2. It is an important avantage of the present argument that it merely uses the monotoneity, continuity an homogeneity properties of the function v k. Lemma 2. For each β > 0, there are constants h 0 > 0, N N an c 11 > 0, epening only on β an, such that for a > 0 an 0 < h < h 0, ( ) N P(v k (Z) a(1, 1 + h)) c 11 h a /k λ exp (1 + β) 2 κ 1 /k a /k λ. 10
11 Proof. Let β > 0 an a > 0 be given. For n 2 we efine ( Q n := (x 1,..., x n 1, u) (1 + β/2)b ) n 1 S 1 : Z(x 1,..., x n 1, u) 2 1 (1 + β/2)b, v k (Z(x 1,..., x n 1, u)) 2 k κ. A continuity argument shows that we can choose N = N(β, ) N (sufficiently large) so that... 1(x 1,..., x N 1, u) Q N x 1... x N 1 σ(u) =: c 12 (, β) > 0. S 1 R R If (x 1,..., x N 1, u) Q N, then We efine h 0 = h 0 (β) > 0 by v k (Z(x 1,..., x N 1, u)) (1 + β/2) k 2 k κ. (20) (1 + h 0 ) 1/k (1 + β/2) = 1 + β (21) an suppose that 0 < h < h 0. As before, we put B a := (a/κ ) 1/k B. We start with the trivial estimate (again we use that a realization of X enotes a simple counting measure an also its support) P(v k (Z) a(1, 1 + h)) P( X(2(1 + β)ba ) = N, Z( X 2(1 + β)b a ) (1 + β)b a, v k (Z( X ) 2(1 + β)b a )) a(1, 1 + h) ) ( = P( X(2(1 + β)ba ) = N P Z( X 2(1 + β)b a ) (1 + β)b a, v k (Z( X 2(1 + β)b a )) a(1, 1 + h) ) X(2(1 + β)ba ) = N. Since X is a stationary Poisson process with intensity λ, we obtain (using Satz 3.2.3(b) of [16]) P(v k (Z) a(1, 1 + h)) λn N! exp v (2(1 + β)b a )λ... R 1 i : x i 2(1 + β)b a 1 Z(x 1,..., x N ) (1 + β)b a R 1 v k (Z(x 1,..., x N )) a(1, 1 + h) x 1... x N. Assume that x 1,..., x N R satisfy the conitions (i) (x 1,..., x N ) x N Q N ; (ii) v k (Z(x 1,..., x N )) a(1, 1 + h). Then, using (ii), (i) an the efinition of Q N, we get a(1 + h) x N k v k(z(x 1,..., x N )) x N k 2 k κ, 11
12 hence By (21) an (22), x N 2(a/κ ) 1/k (1 + h) 1/k. (22) x N 2(1 + β)(a/κ ) 1/k. Further, using (i), the efinition of Q N, (22) an (21), we fin that, for i = 1,..., N 1, x i x N (1 + β/2) 2(a/κ ) 1/k (1 + h) 1/k (1 + β/2) 2(1 + β)(a/κ ) 1/k, thus x i 2(1 + β)b a. Finally, (i), the efinition of Q N, (22) an (21) imply that Z(x 1,..., x N ) 2 1 x N (1 + β/2)b (a/κ ) 1/k (1 + h) 1/k (1 + β/2)b (1 + β)b a. Hence, introucing polar coorinates, we obtain P(v k (Z) a(1, 1 + h)) λn N! exp v (2(1 + β)b a )λ... 1 (x 1,..., x N ) x N Q N R R 1v k (Z(x 1,..., x N )) a(1, 1 + h)x 1... x N = λn N! exp v (2(1 + β)b a )λ... 1 (x 1,..., x N 1, ru) rq N S 1 0 R R 1v k (Z(x 1,..., x N 1, ru)) a(1, 1 + h)r 1 x 1... x N 1 r σ(u). Substituting x i = ry i for i = 1,..., N 1, we get P(v k (Z) a(1, 1 + h)) λn N! exp v (2(1 + β)b a )λ... 1(y 1,..., y N 1, u) Q N S 1 0 R R 1r k v k (Z(y 1,..., y N 1, u)) a(1, 1 + h)r N 1 y 1... y N 1 r σ(u) = λn N! exp v (2(1 + β)b a )λ S 1... R 1(y 1,..., y N 1, u) Q N R an/k ) N v k(z(y 1,..., y N 1, u)) N/k (1 + h) N/k 1 y 1... y N 1 σ(u) λn N! exp v (2(1 + β)b a )λ an/k ( ) N/k N 2 k (1 + β/2) k κ 1 N k h c 12(, β) ( N h a λ) /k exp v (2(1 + β)b a )λ c 13 (, β), where also (20) was use. assertion. Since v (2(1 + β)b a ) = (1 + β) 2 κ 1 /k a /k, this proves the 12
13 5 Probabilities involving elongate cells We will later nee estimates showing that typical cells which, compare to their value of v k, are too long, occur only with small probability. This requires the following preparations, in the course of which convex boies are classifie accoring to their egree of elongation. We enote by P o K o the subset of convex polytopes an by G(, k) the Grassmannian of k-imensional linear subspaces of R. For K K o an L G(, k), the set K L is the image of K uner orthogonal projection to L. For k 1,...,, we efine η k (K) := mind(k)/ (K, L) : L G(, k), where D(K) enotes the iameter of K an (K, L) is the with of K L evaluate in L. The simplest cases are k =, where η (K) is the ratio of the iameter an the with of K, an k = 1, where η 1 (K) = 1. Let a > 0 be given. For m N, we set K,k a (m) := K K o : v k (K) a(1, 2), η k (K) [m k, (m + 1) k ) The information that K Ka,k (m) can be use to erive estimates for the size of K from above an below.. Lemma 3. Let m N an k 1,...,. Then (a) K K,k a (m) implies that K c 14 m k a 1/k B =: C; (b) there exists a measurable map Ka,k (m) Po P v(p ) such that v(p ) is a vertex of P with v(p ) c 15 ma 1/k. Proof. We use repeately that v k (K L) is a constant multiple of the k-imensional volume of K L. (a) A special case of equation (5.3.23) in [15] an the monotoneity of mixe volumes imply that v k (K L) c 16 v k (K) hols for all K Ko an L G(, k). Let K Ka,k (m), an choose L 0 G(, k) such that η k (K) = D(K)/ (K, L 0 ). Then 2a v k (K) c 1 16 v k(k L 0 ) c 17 (K, L 0 ) k 1 D(K L 0 ), where the estimate (16) from [7] was use. Thus an therefore Since o K, this implies (a). 2a c 17 (K, L 0 ) k c 17 (m + 1) k2 D(K) k D(K) c 18 m k a 1/k. (b) For any L G(, k), we enclose K L in a rectangular parallelepipe in L with one ege length equal to (K, L) an the other ege lengths at most D(K L). Then v k (K L) c 19 (K, L)D(K L) k 1 c 19 m k D(K)D(K) k 1 13
14 an hence, by an integral-geometric projection formula ([15], (5.3.27)), a v k (K) c 20 m k D(K) k. Therefore, K has a point at istance at least 2 1 c 1/k 20 ma 1/k from the origin. If K is a polytope, such a point can be chosen as a vertex. That a measurable selection is possible, follows as in [7, Lemma 4.3(c)]. Remark. We have η 1 (K) = 1; moreover, Ka,1 (m) only for m = 1. Therefore, some of the subsequent arguments simplify consierably, or can be omitte, in the case k = 1. Let a > 0, ɛ > 0 be given. For m N, we efine an K,k a,ɛ (m) := K K,k a (m) : ϑ(k) ɛ (23) q k a,ɛ(m) := P(Z K,k a,ɛ (m)). (24) Similarly as in [7], we prove two estimates concerning the ecay of q k a,ɛ(m) as a /k λ. The epenence on ɛ will not play a role until Lemma 6. For u 1,..., u n S 1 an t 1,..., t n (0, ) we introuce the abbreviation n H (u i, t i ) =: P (u (n), t (n) ). i=1 Our next lemma correspons to Lemma 5.1 in [7]; its proof is a moifie an slightly simplifie version of the proof of the latter. Lemma 4. For m N an a /k λ σ 0, where σ 0 > 0 is a constant, q k a,ɛ(m) c 21 (, σ 0 ) exp c 22 m a /k λ. (25) Proof. Let C be the ball efine in Lemma 3(a). Then Here, q k a,ɛ(m) = N=+1 P(Y (H C ) = N)P(Z K,k a,ɛ (m) Y (H C ) = N). (26) p N := P(Z Ka,ɛ,k (m) Y (H C ) = N) = 1 U(C) N... S 1 0 S P (u (N), t (N) ) Ka,ɛ,k (m) (27) 1 i : H(u i, t i ) C (t 1 t N ) 1 t 1 σ(u 1 )... t N σ(u N ). Suppose that u 1,..., u N, t 1,..., t N are such that the inicator functions occurring in the multiple integral are all equal to one; then P := P (u (N), t (N) ) Ka,ɛ,k (m) has a vertex v(p ) 14
15 accoring to Lemma 3(b). This vertex is the intersection of facets of P. Hence, there exists an inex set J 1,..., N with elements such that v(p ) = i J H(u i, t i ). The segment S := [o, v(p )] satisfies relint S H(u j, t j ) = for j 1,..., N \ J, where relint enotes the relative interior. Since S C, we have 1H(u, t) C, H(u, t) S = t 1 t σ(u) S 1 0 = U(C) U(S) = U(C) 2 κ S, (28) where S enotes the length of S. Similarly as in the proof of [7, Lemma 5.1] we obtain ( ) N 1 p N U(C) N... 1H(u i, t i ) C, i = 1,..., S 1 0 S 1 [U(C) c 23 m a /k ] N (t 1 t ) 1 t 1 σ(u 1 )... t σ(u ) 0 = ( N ) ( ) N 1 c 23m a /k. (29) U(C) This leas to the estimate qa,ɛ(m) k [2 U(C)λ] N N! N=+1 ( ) ( N exp 2 U(C)λ 1 c 23m a /k U(C) ) N = 1! [2 U(C)λ] exp 2 U(C)λ N=+1 1 [ ] N 2 U(C)λ c 24 m a /k λ (N )! 1! [2 U(C)λ] exp c 24 m a /k λ ( ) c 25 m k2 a /k λ exp c 24 m a /k λ c 27 (, σ 0 ) exp c 26 m a /k λ, which completes the proof. The following result allows us to approximate a given convex polytope P by a polytope L P with a restricte number of vertices such that U(L) is not much smaller than U(P ). For a polytope P, let extp be the set of vertices an f 0 (P ) the number of vertices of P. Lemma 5. Let α > 0 be given. There is a number ν N epening only on an α such that the following is true. For P P o there exists a polytope L = L(P ) P o satisfying 15
16 extl extp, f 0 (L) ν, an U(L) (1 α)u(p ). Moreover, there exists a measurable selection P L(P ). Proof. The following can be extracte from the proof of Lemma 4.2 in [7]. There exist numbers k 0 = k 0 () an b 0 = b 0 () such that the following is true. Let P P o be a polytope an let P RB, where R is minimal. Let k k 0. There is a measurable map P L(P ) such that L = L(P ) is the convex hull of at most (k + 1) vertices of P, o L, an P L + κrb with κ = b 0 k 2/( 1). There is a unit vector u such that R[o, u] P an hence U(P ) U(R[o, u]) = 2 κ R. By a suitable choice of k, epening only on an α, we can achieve that κ 1 an 4 κ α. Since h L R, we get U(P ) 1 [h L + κr] σ U(L) + κ 2 κ R U(L) + 4 κu(p ), S 1 thus U(L) (1 α)u(p ), an ν = (k + 1) is the require number. For the probabilities q k a,ɛ, we now state another upper boun, which is base on the stability estimate in Lemma 1 an on the preceing approximation result. Lemma 6. For m N, ɛ (0, 1) an a, λ > 0, ( qa,ɛ(m) k c 28 (, ɛ)m k2ν exp where ν epens only on an ɛ. 1 + c 29 ɛ (+3)/2) 2 κ 1 /k a /k λ, Proof. We efine C as in Lemma 3(a) an use (26) an (27) again. Assume that u 1,..., u N, t 1,..., t N are such that the inicator functions in (27) are all equal to one. Then, by Lemma 1, U ( P (u (N), t (N) ) ) (1 + γɛ (+3)/2 )κ 1 /k a /k. (30) Let α := γɛ (+3)/2 /(2 + γɛ (+3)/2 ); then (1 α)(1 + γɛ (+3)/2 ) = 1 + α. Put c 30 := γ/(2 + γ); then α > c 30 ɛ (+3)/2. By Lemma 5, there are ν = ν(, ɛ) vertices of P (u (N), t (N) ) such that the convex hull L = L ( P (u (N), t (N) ) of these vertices satisfies The inequalities (30) an (31) imply that U(L) (1 α)u ( P (u (N), t (N) ) ). (31) U(L) (1 + α) κ 1 /k a /k. (32) Now the same argument as in the proof of Lemma 5.2 in [7], with the obvious moifications, yiels P(Z K,k a,ɛ (m) Y (H C ) = N)[U(C)] N ν j=+1 ( N j )( ) j ν [ U(C) (1 + α)κ 1 /k a /k] N j [U(C)] j. 16
17 Here j enotes the number of hyperplanes generating the vertices of L, an ( j ) bouns the number of points of intersection of these hyperplanes; thus ( j ν ) estimates the possibilities to choose the vertices of L. The probability that the other N j hyperplanes intersecting C o not meet L is given by [U(C) U(L)] N j U(C) N+j, which is estimate using (32). Inserting the inequality in (26), we can continue as in [7], finally using U(C) = (c 14 m k a 1/k ) κ an α > c 30 ɛ (+3)/2. This completes the proof. 6 Proof of Theorem 1 From now on, the proofs follow essentially the lines of those given in [7]. We will, therefore, state only the necessary lemmas in their moifie forms an refer to the corresponing proofs in [7]. thus with Let a > 0 an ɛ (0, 1) be given. For h (0, 1] an m N we efine K,k a,ɛ,h K (m) := Ko : v k (K) a(1, 1 + h), η k (K) [m k, (m + 1) k ), ϑ(k) ɛ ; (33) Moreover, we put P(v k (Z) a(1, 1 + h), ϑ(z) ɛ) = m=1 q k,h a,ɛ (m) ( ) qa,ɛ k,h (m) := P Z K,k a,ɛ,h (m). (34) ( ) qa,ɛ k,h (m, n) := P Z K,k a,ɛ,h (m), f 1(Z) = n for n N; here f 1 (P ) enotes the number of facets of a polytope P. Then we have P(v k (Z) a(1, 1 + h), ϑ(z) ɛ) = m=1 n=+1 q k,h a,ɛ (m, n). (35) Finally, we efine R k,h a,ɛ (m, n) := (u 1,..., u n, t 1,..., t n ) (S 1 ) n (0, ) n : P (u (n), t (n) ) K,k a,ɛ,h (m), f ( 1 P (u(n), t (n) ) ) = n, H(u i, t i ) C for i = 1,..., n, where the ball C is again efine as in Lemma 3(a), for the given a, ɛ, m. Lemma 7. For m, n N, n + 1 an h (0, 1], qa,ɛ k,h (m, n) = (2 λ) n n! R k,h a,ɛ (m,n) exp 2 U ( P (u (n), t (n) ) ) λ (t 1 t n ) 1 t 1... t n σ(u 1 )... σ(u n ). 17
18 The proof is the same as that for Lemma 6.1 in [7], with the obvious necessary changes. Lemma 8. Let w > 0, h (0, 1/2) an r 1. Then k 1+h 1 x r exp wx x c 20 hw[1 + (expw/2 1) 1 ] k 2 1 x r exp wx x. After the substitution x = y, one can imitate the proof of Lemma 6.2 in [7] to obtain the result. The next (technical) lemma states that each boun for qa,ɛ k,1 yiels a boun for qa,ɛ k,h which is linear in h. This shoul be compare to the boun in Lemma 2 which is also linear in h. Lemma 9. For m N, h (0, 1/2) an a /k λ σ 0 > 0, q k,h a,ɛ (m) c 31 (, σ 0 )h a /k λ m k q k,1 a,ɛ (m). Again, the proof is obtaine by aapting the corresponing one from [7], namely that of Lemma 6.3. After applying Lemmas 7 an 8, we arrive at the inequality qa,ɛ k,h (m, n) (2 λ) n t(ζ) n c 32 hu(k(ζ, t(ζ)))λ n! U(m,n) ( ( ) ) exp 2 1 U(K(ζ, t(ζ)))λ 1 k 2 1 s n 1 exp 2 U(K(ζ, t(ζ)))λs s (t 1 t n 1 ) 1 t 1... t n 1 σ(u 1 )... σ(u n ), where U(m, n), t(ζ) an K(, ) are efine as in [7], with the obvious changes. Now we have to observe that v k (K(ζ, t(ζ))) = a, hence K(ζ, t(ζ)) K,k a,ɛ (m), which implies that c 33 m a /k U(K(ζ, t(ζ))) c 34 m k a /k, by Lemma 3. The estimation can now be complete as in [7]. The following lemma establishes an upper estimate for an unconitional probability. Lemma 10. Let ɛ (0, 1), h (0, 1/2) an a /k λ σ 0 > 0. Then P(v k (Z) a(1, 1 + h), ϑ(z) ɛ) ( c 35 (, ɛ, σ 0 )h exp 1 + (c 29 /2)ɛ (+3)/2) 2 κ 1 /k a /k λ. Here c 29 is the constant appearing in Lemma 6. The proof of Lemma 10 follows the one of Proposition 7.1 in [7] an uses Lemmas 9, 4 an 6, in this orer. 18
19 The choice (1 + β) = 1 + (c 29 /4)ɛ (+3)/2 in Lemma 2 immeiately proves Theorem 1 with b = a(1 + h) in the case h min(h 0, 1/2). As to arbitrary b a, we observe that Lemmas 2 an 10 have the same structure as Lemma 3.2 an Proposition 7.1, respectively, in [7]; they iffer only by the values of some parameters. It is, therefore, clear that Theorem 1 now follows precisely in the same way as Theorem 1 of [7] was prove. 7 Proof of Theorem 2 In this section, Y enotes a stationary isotropic Poisson hyperplane process in R with intensity λ > 0. For a convex boy K R an for n N 0, we have P(Y (H K ) = n) = [2κ 1 v 1(K)λ] n n! exp 2κ 1 v 1(K)λ, (36) by [7, (4)], where B is now the ball with surface area 1, thus B = (κ ) 1/( 1) B. Let Z o be the zero cell of the tessellation inuce by Y. Lemma 11. Let k 1,...,. For each β > 0, there are constants h 0 > 0, N N an c 36 > 0, epening only on β an, such that for a > 0 an 0 < h < h 0, ( ) N P(v k (Z o ) a(1, 1 + h)) c 36 h a 1/k λ exp (1 + β)2κ 1/k a 1/k λ. Proof. In orer to be able to essentially copy the proof of Lemma 2, we efine a measure ψ on R by ψ(a) := 1 1 A (tu) t σ(u) κ S 1 0 for Borel sets A R. Let X be the Poisson process in R with intensity measure λψ, an let Y be the hyperplane process efine by Y := H(x) : x X. Then Y is a Poisson process in the space H of hyperplanes, an for a Borel set A H we have E Y (A) = λψ(x R : H(x) A) = 2λ 1 A (H(u, t)) t σ(u). κ S 1 This shows that Y has the same intensity measure as Y. Since Y an Y are Poisson processes, they are stochastically equivalent. We can now repeat the proof of Lemma 2, where we replace Z by Z o an x by ψ(x). Further, we observe that ψ((ru)) = (κ ) 1 r σ(u) for u = 1 an ψ((ry)) = rψ(y) for r > 0. In the exponential, v (2(1 + β)b a ) (where B a = (a/κ ) 1/k B ) has to be replace by ψ(2(1 + β)b a ) = 2(1 + β)(a/κ ) 1/k. With these changes, the proof of Lemma 2 yiels the assertion of Lemma We will nee a stability version of the inequality (11). Lemma 12. There is a positive constant γ, epening only on the imension, such that for ɛ (0, 1) an every convex boy K R with r (K) ɛ, the inequality ( v 1 (K) 1 + γɛ (+3)/2) κ 1 1/k v k (K) 1/k (37) 19
20 hols for k = 2,...,. Proof. Without loss of generality, we assume that K has interior points, mean with 2 (the same as the unit ball), an Steiner point o. We put v i := v i (K) for i = 0,..., (v 0 = κ ) an use the Aleksanrov-Fenchel inequalities v 2 i v i 1v i+1 for i = 1,..., 1 (see [15]). Theorem an Lemma of [15] provie an improvement of the first of these inequalities, namely v 2 1 v 0 v 2 c 37 δ(k, B ) (+3)/2, where δ is the Hausorff istance. This can be rewritten as v 1 v ( c ) 37 δ(k, B ) (+3)/2 v 2 (1 + α) v 0 v 1 v 0 v 2 v 1 with α := c 38 δ(k, B ) (+3)/2, since v 0 v 2 v1 2 = c 39 (v 1 being a constant multiple of the mean with). From v 1 (1 + α) v 2 (1 + α) v k v 0 v 1 v k 1 we get an thus ( v1 If δ := δ(k, B ) 1/2, then (since ɛ < 1) v 0 ) k 1 (1 + α) k 1 v k v 1 v k 1 v k 1 0 v k 1 + c 38 δ(k, B ) (+3)/2. v k 1 v k 1 0 v k 1 + c 40 ɛ (+3)/2. If δ 1/2, let rb K sb where r is maximal an s is minimal. Then (by the efinition of the Hausorff metric) s 1 + δ an r 1 δ 1/2, hence ɛ r (K) (s/r) 1 4δ an thus v1 k v0 k c 41 ɛ (+3)/2. v k Both cases together give ( v c 42 ɛ (+3)/2) 1/k ( 1 1/k v 0 v 1/k k 1 + c 43 ɛ (+3)/2) κ 1 1/k 0 v 1/k k. Theorem 2 can now be prove in essentially the same way as Theorem 1, an we list only the necessary changes in Sections 5 an 6, in aition to those alreay mentione in the proof of Lemma 11. Definitions (23) an (24) are replace by K,k a,ɛ (m) := K K,k a (m) : r (K) ɛ, q k a,ɛ(m) := P(Z o K,k a,ɛ (m)). Lemma 13. Let k 1,...,. For m N an a 1/k λ σ 0, where σ 0 > 0 is a given constant, qa,ɛ(m) k c 44 (, σ 0 ) exp c 45 ma 1/k λ. (38) 20
21 Proof. In the proof of Lemma 4, the number U(C) is replace by v 1 (C). In the integrations, t 1 t is replace by t. Equation (28) now reas 1H(u, t) C, H(u, t) S = t σ(u) S 1 0 = v 1 (C) v 1 (S) = v 1 (C) κ 1 S, hence (29) reas p N ( N ) ( ) N 1 c 46ma 1/k. v 1 (C) Continuing as in the proof of Lemma 4, we arrive at (38). Lemma 14. Let k 2,...,. For m N, ɛ (0, 1) an a, λ > 0, ( qa,ɛ(m) k c 47 (, ɛ)m kν exp 1 + c 48 ɛ (+3)/2) 2κ 1/k a 1/k λ, where ν epens only on an ɛ. Proof. We use (26) an (27), with the changes alreay mae for the proof of Lemma 13. Assume that u 1,..., u N, t 1,..., t N are such that the inicator functions in (27) are all equal to one. By Lemma 12, ( v 1 (P (u (N), t (N) )) 1 + γɛ (+3)/2) κ 1 1/k a 1/k. We efine α as in the proof of Lemma 6 an put c 49 := γ/(2 + γ), so that α > c 49 ɛ (+3)/2. By Lemma 4.2 of [7] (with B = (κ ) 1/( 1) B ), there are ν = ν(, ɛ) vertices of P (u (N), t (N) ) such that their convex hull L = L(P (u (N), t (N) )) satisfies This gives v 1 (L) (1 α)v 1 (P (u (N), t (N) )). v 1 (L) (1 + α)κ 1 1/k a 1/k. As in the proof of Lemma 6 (an of Lemma 5.2 in [7]) we euce that By Lemma 3, P(Z o K,k a,ɛ (m) Y (H C ) = N)[v 1 (C)] N ν j=+1 ( N j )( ) j ν [ v 1 (C) (1 + α)κ 1 1/k a 1/k] N j [v1 (C)] j. v 1 (C) = c 50 m k a 1/k. The assertion is now obtaine as in the proofs quote above. The further proof again follows the lines of [7] an of the proof of Theorem 1. Definitions (33), (34) an (35) are replace by K,k a,ɛ,h K (m) := Ko : v k (K) a(1, 1 + h), η k (K) [m k, (m + 1) k ), r (K) ɛ, 21
22 Lemma 7 is replace by q k,h a,ɛ (m, n) = q k,h a,ɛ (m) := P q k,h a,ɛ (m, n) := P (2λ) n (κ ) n n! ( Z o K,k a,ɛ,h (m) ), ( Z o K,k a,ɛ,h (m), f 1(Z o ) = n R k,h a,ɛ (m,n) t 1... t n σ(u 1 )... σ(u n ). ). exp 2κ 1 v ( 1 P (u(n), t (n) ) ) λ This is a special case of Lemma 6.1 in [7]. Instea of Lemma 8, we use Lemma 6.2 from [7], with replace by k 2. Lemma 9 is replace by the following assertion. For m N, h (0, 1/2) an a 1/k λ σ 0 > 0, q k,h a,ɛ (m) c 51 (, σ 0 )h a 1/k λ m k q k,1 a,ɛ (m). The proof is obtaine by aapting the proof of Lemma 6.3 in [7]. In the course of the proof, one has to use that Lemma 3 implies c 52 ma 1/k v 1 (K) c 53 m k a 1/k for K K,k a,ɛ (m). The counterpart of Lemma 10 now reas as follows. Let ɛ (0, 1), h (0, 1/2) an a 1/k λ σ 0 > 0. Then P(v k (Z o ) a(1, 1 + h), r (Z o ) ɛ) ( c 54 (, ɛ, σ 0 )h exp 1 + (c 48 /2)ɛ (+3)/2) 2κ 1/k a 1/k λ. With these preliminaries, the proof of Theorem 2 can now be complete in the same way as that of Theorem 1. 8 Proof of Theorem 3 In large parts of the proofs of Theorems 1 an 2, only the following properties of the functional v k are use: it is monotone uner set inclusion, i.e., v k (K 1 ) v k (K 2 ) if K 1 K 2, positively homogeneous of egree k, i.e., v k (rk) = r k v k (K) for r 0, an continuous with respect to the Hausorff metric. These properties are share, with k = 1, by the inraius functionals ρ o an ρ. Hence, the corresponing parts of the proofs apply also to Theorem 3. Aitional properties of the function v k were only neee for the stability estimates of Lemmas 1 an 12 an for Lemma 3. We replace these lemmas by the following ones. Lemma 15. There is a positive constant γ, epening only on the imension, such that for ɛ (0, 1) an every convex boy K K o with ϑ(k) ɛ the inequality hols. U(K) (1 + γɛ (+1)/2 )κ ρ o (K) 22
23 Proof. Without loss of generality, we assume that ρ o (K) = 1. Let K R o B, where R o is minimal. First we assume that R o 3. Put R o = 1 + h, then ɛ ϑ(k) h 2. Proceeing similarly as in Case 1 of the proof of Lemma 1, we fin that U(K) U(B ) c 55 h (+1)/2 c 55 ɛ (+1)/2. Now suppose that R o > 3. Then there is a spherical cap A S 1 with σ(a) = c 56 > 0 on which h K 2. It follows that U(K) U(B ) 1 ( ) h K 1 σ c 57 ɛ (+1)/2, since ɛ < 1. Hence, both cases yiel U(K) U(B ) + c 58 ɛ (+1)/2 = A ( 1 + c 59 ɛ (+1)/2) κ ρ o (K). Lemma 16. There is a positive constant γ, epening only on the imension, such that for ɛ (0, 1) an every convex boy K K o with r (K) ɛ the inequality hols. v 1 (K) (1 + γɛ (+1)/2 )κ ρ(k) The proof is similar to that of Lemma 15. For a replacement of Lemma 3, let a > 0 be given. For m N we set K a(m) := K K o : ρ(k) a(1, 2), D(K)/ρ(K) [m, m + 1). Lemma 17. Let m N. Then (a) K K a(m) implies that K c 60 mab ; (b) there exists a measurable map K a(m) P o P v(p ) such that v(p ) is a vertex of P with v(p ) c 61 ma. A similar result hols with ρ o instea of ρ(k). The proof is immeiate. With these changes, the proofs of Theorems 1 an 2 yiel the proof of Theorem 3. References [1] Calka, P. (2002) The istributions of the smallest isks containing the Poisson-Voronoi typical cell an the Crofton cell in the plane. Av. Appl. Prob. 34, [2] Garner, R.J. an Vassallo, S. (1999) Stability of inequalities in the ual Brunn- Minkowski theory. J. Math. Anal. Appl. 231, [3] Golman, A. (1998) Sur une conjecture e D.G. Kenall concernant la cellule e Crofton u plan et sur sa contrepartie brownienne. Ann. Probab. 26,
24 [4] Groemer, H. (1996) Geometric Applications of Fourier Series an Spherical Harmonics. Encyclopeia of Mathematics an Its Applications 61, Cambrige University Press, Cambrige. [5] Groemer, H. an Schneier, R. (1991) Stability estimates for some geometric inequalities. Bull. Lonon Math. Soc. 23, [6] Hug, D. an Schneier, R. (2002). Large cells in Poisson-Delaunay tessellations. Discrete Comput. Geom. (to appear.) [7] Hug, D., Reitzner, M. an Schneier, R. (2002). The limit shape of the zero cell in a stationary Poisson hyperplane tessellation. Ann. Probab. (to appear). [8] Kovalenko, I.N. (1997) A proof of a conjecture of Davi Kenall on the shape of ranom polygons of large area. (Russian) Kibernet. Sistem. Anal. 1997, 3 10, 187; Engl. transl. Cybernet. Systems Anal. 33, [9] Kovalenko, I.N. (1998) An extension of a conjecture of D.G. Kenall concerning shapes of ranom polygons to Poisson Voronoï cells. In: Engel, P. et al. (es.), Voronoï s impact on moern science. Book I. Transl. from the Ukrainian. Kyiv: Institute of Mathematics. Proc. Inst. Math. Natl. Aca. Sci. Ukr., Math. Appl. 212, [10] Kovalenko, I.N. (1999) A simplifie proof of a conjecture of D.G. Kenall concerning shapes of ranom polygons. J. Appl. Math. Stochastic Anal. 12, [11] Miles, R.E. (1995) A heuristic proof of a long-staning conjecture of D.G. Kenall concerning the shapes of certain large ranom polygons. Av. Appl. Prob. 27, [12] Møller, J. (1994) Lectures on Ranom Voronoi Tessellations. Lect. Notes Statist. 87, Springer, New York. [13] Møller, J. (1998) A review on probabilistic moels an results for Voronoi tessellations. In: Engel, P. et al. (es.), Voronoï s impact on moern science. Book I. Transl. from the Ukrainian. Kyiv: Institute of Mathematics. Proc. Inst. Math. Natl. Aca. Sci. Ukr., Math. Appl. 212, [14] Okabe, A., Boots, B., Sugihara, K. an Chiu, S.N. (2000) Spatial Tessellations; Concepts an Applications of Voronoi Diagrams., 2n e., Wiley, Chichester. [15] Schneier, R. (1993) Convex Boies: the Brunn-Minkowski Theory. Encyclopeia of Mathematics an Its Applications 44, Cambrige University Press, Cambrige. [16] Schneier, R. an Weil, W. (2000) Stochastische Geometrie. Teubner Skripten zur Mathematischen Stochastik, Teubner, Stuttgart. [17] Stoyan, D., Kenall, W.S. an Mecke, J. (1995) Stochastic Geometry an its Applications. 2n e., Wiley, Chichester. 24
Large Cells in Poisson-Delaunay Tessellations
Large Cells in Poisson-Delaunay Tessellations Daniel Hug an Rolf Schneier Mathematisches Institut, Albert-Luwigs-Universität, D-79104 Freiburg i. Br., Germany aniel.hug, rolf.schneier}@math.uni-freiburg.e
More informationLARGE TYPICAL CELLS IN POISSON DELAUNAY MOSAICS
LARGE TYPICAL CELLS IN POISSON DELAUNAY MOSAICS DANIEL HUG an ROLF SCHNEIDER Deicate to Tuor Zamfirescu on the occasion of his sixtieth birthay It is prove that the shape of the typical cell of a Poisson
More informationLecture Introduction. 2 Examples of Measure Concentration. 3 The Johnson-Lindenstrauss Lemma. CS-621 Theory Gems November 28, 2012
CS-6 Theory Gems November 8, 0 Lecture Lecturer: Alesaner Mąry Scribes: Alhussein Fawzi, Dorina Thanou Introuction Toay, we will briefly iscuss an important technique in probability theory measure concentration
More informationarxiv: v1 [math.mg] 10 Apr 2018
ON THE VOLUME BOUND IN THE DVORETZKY ROGERS LEMMA FERENC FODOR, MÁRTON NASZÓDI, AND TAMÁS ZARNÓCZ arxiv:1804.03444v1 [math.mg] 10 Apr 2018 Abstract. The classical Dvoretzky Rogers lemma provies a eterministic
More informationMonotonicity of facet numbers of random convex hulls
Monotonicity of facet numbers of ranom convex hulls Gilles Bonnet, Julian Grote, Daniel Temesvari, Christoph Thäle, Nicola Turchi an Florian Wespi arxiv:173.31v1 [math.mg] 7 Mar 17 Abstract Let X 1,...,
More informationGaussian polytopes: variances and limit theorems
Gaussian polytopes: variances an limit theorems Daniel Hug an Matthias Reitzner October 9, 4 Abstract The convex hull of n inepenent ranom points in R chosen accoring to the normal istribution is calle
More informationDiscrete Mathematics
Discrete Mathematics 309 (009) 86 869 Contents lists available at ScienceDirect Discrete Mathematics journal homepage: wwwelseviercom/locate/isc Profile vectors in the lattice of subspaces Dániel Gerbner
More informationTopics in Stochastic Geometry. Lecture 4 The Boolean model
Institut für Stochastik Karlsruher Institut für Technologie Topics in Stochastic Geometry Lecture 4 The Boolean model Lectures presented at the Department of Mathematical Sciences University of Bath May
More informationAgmon Kolmogorov Inequalities on l 2 (Z d )
Journal of Mathematics Research; Vol. 6, No. ; 04 ISSN 96-9795 E-ISSN 96-9809 Publishe by Canaian Center of Science an Eucation Agmon Kolmogorov Inequalities on l (Z ) Arman Sahovic Mathematics Department,
More informationAcute sets in Euclidean spaces
Acute sets in Eucliean spaces Viktor Harangi April, 011 Abstract A finite set H in R is calle an acute set if any angle etermine by three points of H is acute. We examine the maximal carinality α() of
More informationRandom Mosaics. 1 General Results. Daniel Hug
Random Mosaics Daniel Hug Mathematisches Institut, Albert-Ludwigs-Universität Eckerstr. 1, D-79104 Freiburg, Germany daniel.hug@math.uni-freiburg.de Mosaics arise frequently as models in materials science,
More informationSTABILITY RESULTS INVOLVING SURFACE AREA MEASURES OF CONVEX BODIES DANIEL HUG AND ROLF SCHNEIDER
STABILITY RESULTS INVOLVING SURFACE AREA MEASURES OF CONVEX BODIES DANIEL HUG AND ROLF SCHNEIDER Abstract We strengthen some known stability results from the Brunn-Minkowski theory an obtain new results
More informationLower bounds on Locality Sensitive Hashing
Lower bouns on Locality Sensitive Hashing Rajeev Motwani Assaf Naor Rina Panigrahy Abstract Given a metric space (X, X ), c 1, r > 0, an p, q [0, 1], a istribution over mappings H : X N is calle a (r,
More informationOn the cells in a stationary Poisson hyperplane mosaic
On the cells in a stationary Poisson hyperplane mosaic Matthias Reitzner and Rolf Schneider Abstract Let X be the mosaic generated by a stationary Poisson hyperplane process X in R d. Under some mild conditions
More informationTractability results for weighted Banach spaces of smooth functions
Tractability results for weighte Banach spaces of smooth functions Markus Weimar Mathematisches Institut, Universität Jena Ernst-Abbe-Platz 2, 07740 Jena, Germany email: markus.weimar@uni-jena.e March
More information6 General properties of an autonomous system of two first order ODE
6 General properties of an autonomous system of two first orer ODE Here we embark on stuying the autonomous system of two first orer ifferential equations of the form ẋ 1 = f 1 (, x 2 ), ẋ 2 = f 2 (, x
More informationSharp Thresholds. Zachary Hamaker. March 15, 2010
Sharp Threshols Zachary Hamaker March 15, 2010 Abstract The Kolmogorov Zero-One law states that for tail events on infinite-imensional probability spaces, the probability must be either zero or one. Behavior
More informationFLUCTUATIONS IN THE NUMBER OF POINTS ON SMOOTH PLANE CURVES OVER FINITE FIELDS. 1. Introduction
FLUCTUATIONS IN THE NUMBER OF POINTS ON SMOOTH PLANE CURVES OVER FINITE FIELDS ALINA BUCUR, CHANTAL DAVID, BROOKE FEIGON, MATILDE LALÍN 1 Introuction In this note, we stuy the fluctuations in the number
More informationSmall faces in stationary Poisson hyperplane tessellations
Small faces in stationary Poisson hyperplane tessellations Rolf Schneider Abstract We consider the tessellation induced by a stationary Poisson hyperplane process in d- dimensional Euclidean space. Under
More informationIterated Point-Line Configurations Grow Doubly-Exponentially
Iterate Point-Line Configurations Grow Doubly-Exponentially Joshua Cooper an Mark Walters July 9, 008 Abstract Begin with a set of four points in the real plane in general position. A to this collection
More informationLECTURE NOTES ON DVORETZKY S THEOREM
LECTURE NOTES ON DVORETZKY S THEOREM STEVEN HEILMAN Abstract. We present the first half of the paper [S]. In particular, the results below, unless otherwise state, shoul be attribute to G. Schechtman.
More informationA new proof of the sharpness of the phase transition for Bernoulli percolation on Z d
A new proof of the sharpness of the phase transition for Bernoulli percolation on Z Hugo Duminil-Copin an Vincent Tassion October 8, 205 Abstract We provie a new proof of the sharpness of the phase transition
More informationConvergence of Random Walks
Chapter 16 Convergence of Ranom Walks This lecture examines the convergence of ranom walks to the Wiener process. This is very important both physically an statistically, an illustrates the utility of
More informationLATTICE-BASED D-OPTIMUM DESIGN FOR FOURIER REGRESSION
The Annals of Statistics 1997, Vol. 25, No. 6, 2313 2327 LATTICE-BASED D-OPTIMUM DESIGN FOR FOURIER REGRESSION By Eva Riccomagno, 1 Rainer Schwabe 2 an Henry P. Wynn 1 University of Warwick, Technische
More informationA Sketch of Menshikov s Theorem
A Sketch of Menshikov s Theorem Thomas Bao March 14, 2010 Abstract Let Λ be an infinite, locally finite oriente multi-graph with C Λ finite an strongly connecte, an let p
More informationFunction Spaces. 1 Hilbert Spaces
Function Spaces A function space is a set of functions F that has some structure. Often a nonparametric regression function or classifier is chosen to lie in some function space, where the assume structure
More informationLeast-Squares Regression on Sparse Spaces
Least-Squares Regression on Sparse Spaces Yuri Grinberg, Mahi Milani Far, Joelle Pineau School of Computer Science McGill University Montreal, Canaa {ygrinb,mmilan1,jpineau}@cs.mcgill.ca 1 Introuction
More informationA Weak First Digit Law for a Class of Sequences
International Mathematical Forum, Vol. 11, 2016, no. 15, 67-702 HIKARI Lt, www.m-hikari.com http://x.oi.org/10.1288/imf.2016.6562 A Weak First Digit Law for a Class of Sequences M. A. Nyblom School of
More information12.11 Laplace s Equation in Cylindrical and
SEC. 2. Laplace s Equation in Cylinrical an Spherical Coorinates. Potential 593 2. Laplace s Equation in Cylinrical an Spherical Coorinates. Potential One of the most important PDEs in physics an engineering
More informationLower Bounds for the Smoothed Number of Pareto optimal Solutions
Lower Bouns for the Smoothe Number of Pareto optimal Solutions Tobias Brunsch an Heiko Röglin Department of Computer Science, University of Bonn, Germany brunsch@cs.uni-bonn.e, heiko@roeglin.org Abstract.
More informationWitt#5: Around the integrality criterion 9.93 [version 1.1 (21 April 2013), not completed, not proofread]
Witt vectors. Part 1 Michiel Hazewinkel Sienotes by Darij Grinberg Witt#5: Aroun the integrality criterion 9.93 [version 1.1 21 April 2013, not complete, not proofrea In [1, section 9.93, Hazewinkel states
More informationThe effect of dissipation on solutions of the complex KdV equation
Mathematics an Computers in Simulation 69 (25) 589 599 The effect of issipation on solutions of the complex KV equation Jiahong Wu a,, Juan-Ming Yuan a,b a Department of Mathematics, Oklahoma State University,
More informationTHE GENUINE OMEGA-REGULAR UNITARY DUAL OF THE METAPLECTIC GROUP
THE GENUINE OMEGA-REGULAR UNITARY DUAL OF THE METAPLECTIC GROUP ALESSANDRA PANTANO, ANNEGRET PAUL, AND SUSANA A. SALAMANCA-RIBA Abstract. We classify all genuine unitary representations of the metaplectic
More informationA LIMIT THEOREM FOR RANDOM FIELDS WITH A SINGULARITY IN THE SPECTRUM
Teor Imov r. ta Matem. Statist. Theor. Probability an Math. Statist. Vip. 81, 1 No. 81, 1, Pages 147 158 S 94-911)816- Article electronically publishe on January, 11 UDC 519.1 A LIMIT THEOREM FOR RANDOM
More informationPDE Notes, Lecture #11
PDE Notes, Lecture # from Professor Jalal Shatah s Lectures Febuary 9th, 2009 Sobolev Spaces Recall that for u L loc we can efine the weak erivative Du by Du, φ := udφ φ C0 If v L loc such that Du, φ =
More informationThe Press-Schechter mass function
The Press-Schechter mass function To state the obvious: It is important to relate our theories to what we can observe. We have looke at linear perturbation theory, an we have consiere a simple moel for
More informationSINGULAR PERTURBATION AND STATIONARY SOLUTIONS OF PARABOLIC EQUATIONS IN GAUSS-SOBOLEV SPACES
Communications on Stochastic Analysis Vol. 2, No. 2 (28) 289-36 Serials Publications www.serialspublications.com SINGULAR PERTURBATION AND STATIONARY SOLUTIONS OF PARABOLIC EQUATIONS IN GAUSS-SOBOLEV SPACES
More informationarxiv: v4 [cs.ds] 7 Mar 2014
Analysis of Agglomerative Clustering Marcel R. Ackermann Johannes Blömer Daniel Kuntze Christian Sohler arxiv:101.697v [cs.ds] 7 Mar 01 Abstract The iameter k-clustering problem is the problem of partitioning
More informationSYMPLECTIC GEOMETRY: LECTURE 3
SYMPLECTIC GEOMETRY: LECTURE 3 LIAT KESSLER 1. Local forms Vector fiels an the Lie erivative. A vector fiel on a manifol M is a smooth assignment of a vector tangent to M at each point. We think of M as
More informationNOTES ON EULER-BOOLE SUMMATION (1) f (l 1) (n) f (l 1) (m) + ( 1)k 1 k! B k (y) f (k) (y) dy,
NOTES ON EULER-BOOLE SUMMATION JONATHAN M BORWEIN, NEIL J CALKIN, AND DANTE MANNA Abstract We stuy a connection between Euler-MacLaurin Summation an Boole Summation suggeste in an AMM note from 196, which
More informationarxiv: v1 [math.mg] 27 Jan 2016
On the monotonicity of the moments of volumes of random simplices arxiv:67295v [mathmg] 27 Jan 26 Benjamin Reichenwallner and Matthias Reitzner University of Salzburg and University of Osnabrueck Abstract
More informationOn the Orlicz-Brunn-Minkowski theory
On the Orlicz-Brunn-Minkowski theory C J Zhao 2 3 4 5 6 7 8 9 0 2 3 4 5 Abstract Recently, Garner, Hug an Weil evelope an Orlicz-Brunn- Minkowski theory Following this, in the paper we further consier
More informationMath 342 Partial Differential Equations «Viktor Grigoryan
Math 342 Partial Differential Equations «Viktor Grigoryan 6 Wave equation: solution In this lecture we will solve the wave equation on the entire real line x R. This correspons to a string of infinite
More informationSome Examples. Uniform motion. Poisson processes on the real line
Some Examples Our immeiate goal is to see some examples of Lévy processes, an/or infinitely-ivisible laws on. Uniform motion Choose an fix a nonranom an efine X := for all (1) Then, {X } is a [nonranom]
More informationLower Bounds for k-distance Approximation
Lower Bouns for k-distance Approximation Quentin Mérigot March 21, 2013 Abstract Consier a set P of N ranom points on the unit sphere of imension 1, an the symmetrize set S = P ( P). The halving polyheron
More informationCHAPTER 1 : DIFFERENTIABLE MANIFOLDS. 1.1 The definition of a differentiable manifold
CHAPTER 1 : DIFFERENTIABLE MANIFOLDS 1.1 The efinition of a ifferentiable manifol Let M be a topological space. This means that we have a family Ω of open sets efine on M. These satisfy (1), M Ω (2) the
More informationOn the number of isolated eigenvalues of a pair of particles in a quantum wire
On the number of isolate eigenvalues of a pair of particles in a quantum wire arxiv:1812.11804v1 [math-ph] 31 Dec 2018 Joachim Kerner 1 Department of Mathematics an Computer Science FernUniversität in
More informationRamsey numbers of some bipartite graphs versus complete graphs
Ramsey numbers of some bipartite graphs versus complete graphs Tao Jiang, Michael Salerno Miami University, Oxfor, OH 45056, USA Abstract. The Ramsey number r(h, K n ) is the smallest positive integer
More informationComputing Exact Confidence Coefficients of Simultaneous Confidence Intervals for Multinomial Proportions and their Functions
Working Paper 2013:5 Department of Statistics Computing Exact Confience Coefficients of Simultaneous Confience Intervals for Multinomial Proportions an their Functions Shaobo Jin Working Paper 2013:5
More informationLinear First-Order Equations
5 Linear First-Orer Equations Linear first-orer ifferential equations make up another important class of ifferential equations that commonly arise in applications an are relatively easy to solve (in theory)
More informationGLOBAL SOLUTIONS FOR 2D COUPLED BURGERS-COMPLEX-GINZBURG-LANDAU EQUATIONS
Electronic Journal of Differential Equations, Vol. 015 015), No. 99, pp. 1 14. ISSN: 107-6691. URL: http://eje.math.txstate.eu or http://eje.math.unt.eu ftp eje.math.txstate.eu GLOBAL SOLUTIONS FOR D COUPLED
More informationLecture XII. where Φ is called the potential function. Let us introduce spherical coordinates defined through the relations
Lecture XII Abstract We introuce the Laplace equation in spherical coorinates an apply the metho of separation of variables to solve it. This will generate three linear orinary secon orer ifferential equations:
More informationSection 2.7 Derivatives of powers of functions
Section 2.7 Derivatives of powers of functions (3/19/08) Overview: In this section we iscuss the Chain Rule formula for the erivatives of composite functions that are forme by taking powers of other functions.
More informationSturm-Liouville Theory
LECTURE 5 Sturm-Liouville Theory In the three preceing lectures I emonstrate the utility of Fourier series in solving PDE/BVPs. As we ll now see, Fourier series are just the tip of the iceberg of the theory
More informationSeparation of Variables
Physics 342 Lecture 1 Separation of Variables Lecture 1 Physics 342 Quantum Mechanics I Monay, January 25th, 2010 There are three basic mathematical tools we nee, an then we can begin working on the physical
More informationSurvival exponents for fractional Brownian motion with multivariate time
Survival exponents for fractional Brownian motion with multivariate time G Molchan Institute of Earthquae Preiction heory an Mathematical Geophysics Russian Acaemy of Science 84/3 Profsoyuznaya st 7997
More information1. Aufgabenblatt zur Vorlesung Probability Theory
24.10.17 1. Aufgabenblatt zur Vorlesung By (Ω, A, P ) we always enote the unerlying probability space, unless state otherwise. 1. Let r > 0, an efine f(x) = 1 [0, [ (x) exp( r x), x R. a) Show that p f
More informationThe Principle of Least Action
Chapter 7. The Principle of Least Action 7.1 Force Methos vs. Energy Methos We have so far stuie two istinct ways of analyzing physics problems: force methos, basically consisting of the application of
More informationGeneralized Tractability for Multivariate Problems
Generalize Tractability for Multivariate Problems Part II: Linear Tensor Prouct Problems, Linear Information, an Unrestricte Tractability Michael Gnewuch Department of Computer Science, University of Kiel,
More informationUnit vectors with non-negative inner products
Unit vectors with non-negative inner proucts Bos, A.; Seiel, J.J. Publishe: 01/01/1980 Document Version Publisher s PDF, also known as Version of Recor (inclues final page, issue an volume numbers) Please
More informationarxiv: v1 [math.co] 15 Sep 2015
Circular coloring of signe graphs Yingli Kang, Eckhar Steffen arxiv:1509.04488v1 [math.co] 15 Sep 015 Abstract Let k, ( k) be two positive integers. We generalize the well stuie notions of (k, )-colorings
More informationTime-of-Arrival Estimation in Non-Line-Of-Sight Environments
2 Conference on Information Sciences an Systems, The Johns Hopkins University, March 2, 2 Time-of-Arrival Estimation in Non-Line-Of-Sight Environments Sinan Gezici, Hisashi Kobayashi an H. Vincent Poor
More information7.1 Support Vector Machine
67577 Intro. to Machine Learning Fall semester, 006/7 Lecture 7: Support Vector Machines an Kernel Functions II Lecturer: Amnon Shashua Scribe: Amnon Shashua 7. Support Vector Machine We return now to
More informationMonotonicity for excited random walk in high dimensions
Monotonicity for excite ranom walk in high imensions Remco van er Hofsta Mark Holmes March, 2009 Abstract We prove that the rift θ, β) for excite ranom walk in imension is monotone in the excitement parameter
More informationEXPLICIT BOUNDS ON MONOMIAL AND BINOMIAL EXPONENTIAL SUMS
EXPLICIT BOUNDS ON MONOMIAL AND BINOMIAL EXPONENTIAL SUMS TODD COCHRANE AND CHRISTOPHER PINNER Abstract. Let p be a prime an e p = e 2πi /p. First, we make explicit the monomial sum bouns of Heath-Brown
More informationOn colour-blind distinguishing colour pallets in regular graphs
J Comb Optim (2014 28:348 357 DOI 10.1007/s10878-012-9556-x On colour-blin istinguishing colour pallets in regular graphs Jakub Przybyło Publishe online: 25 October 2012 The Author(s 2012. This article
More informationII. First variation of functionals
II. First variation of functionals The erivative of a function being zero is a necessary conition for the etremum of that function in orinary calculus. Let us now tackle the question of the equivalent
More informationThe chromatic number of graph powers
Combinatorics, Probability an Computing (19XX) 00, 000 000. c 19XX Cambrige University Press Printe in the Unite Kingom The chromatic number of graph powers N O G A A L O N 1 an B O J A N M O H A R 1 Department
More informationINRADII OF SIMPLICES
INRADII OF SIMPLICES ULRICH BETKE, MARTIN HENK, AND LYDIA TSINTSIFA Abstract. We stuy the following generalization of the inraius: For a convex boy K in the -imensional Eucliean space an a linear k-plane
More informationMany problems in physics, engineering, and chemistry fall in a general class of equations of the form. d dx. d dx
Math 53 Notes on turm-liouville equations Many problems in physics, engineering, an chemistry fall in a general class of equations of the form w(x)p(x) u ] + (q(x) λ) u = w(x) on an interval a, b], plus
More informationLEGENDRE TYPE FORMULA FOR PRIMES GENERATED BY QUADRATIC POLYNOMIALS
Ann. Sci. Math. Québec 33 (2009), no 2, 115 123 LEGENDRE TYPE FORMULA FOR PRIMES GENERATED BY QUADRATIC POLYNOMIALS TAKASHI AGOH Deicate to Paulo Ribenboim on the occasion of his 80th birthay. RÉSUMÉ.
More informationImplicit Differentiation
Implicit Differentiation Thus far, the functions we have been concerne with have been efine explicitly. A function is efine explicitly if the output is given irectly in terms of the input. For instance,
More informationLectures - Week 10 Introduction to Ordinary Differential Equations (ODES) First Order Linear ODEs
Lectures - Week 10 Introuction to Orinary Differential Equations (ODES) First Orer Linear ODEs When stuying ODEs we are consiering functions of one inepenent variable, e.g., f(x), where x is the inepenent
More informationPOROSITY OF CERTAIN SUBSETS OF LEBESGUE SPACES ON LOCALLY COMPACT GROUPS
Bull. Aust. Math. Soc. 88 2013), 113 122 oi:10.1017/s0004972712000949 POROSITY OF CERTAIN SUBSETS OF EBESUE SPACES ON OCAY COMPACT ROUPS I. AKBARBAU an S. MASOUDI Receive 14 June 2012; accepte 11 October
More informationA stability result for mean width of L p -centroid bodies.
A stability result for mean with of L p -centroi boies. B. Fleury, O. Guéon an G. Paouris Abstract We give a ifferent proof of a recent result of lartag [1] concerning the concentration of the volume of
More informationTensors, Fields Pt. 1 and the Lie Bracket Pt. 1
Tensors, Fiels Pt. 1 an the Lie Bracket Pt. 1 PHYS 500 - Southern Illinois University September 8, 2016 PHYS 500 - Southern Illinois University Tensors, Fiels Pt. 1 an the Lie Bracket Pt. 1 September 8,
More informationChromatic number for a generalization of Cartesian product graphs
Chromatic number for a generalization of Cartesian prouct graphs Daniel Král Douglas B. West Abstract Let G be a class of graphs. The -fol gri over G, enote G, is the family of graphs obtaine from -imensional
More informationAn Optimal Algorithm for Bandit and Zero-Order Convex Optimization with Two-Point Feedback
Journal of Machine Learning Research 8 07) - Submitte /6; Publishe 5/7 An Optimal Algorithm for Banit an Zero-Orer Convex Optimization with wo-point Feeback Oha Shamir Department of Computer Science an
More informationWELL-POSEDNESS OF A POROUS MEDIUM FLOW WITH FRACTIONAL PRESSURE IN SOBOLEV SPACES
Electronic Journal of Differential Equations, Vol. 017 (017), No. 38, pp. 1 7. ISSN: 107-6691. URL: http://eje.math.txstate.eu or http://eje.math.unt.eu WELL-POSEDNESS OF A POROUS MEDIUM FLOW WITH FRACTIONAL
More informationA note on asymptotic formulae for one-dimensional network flow problems Carlos F. Daganzo and Karen R. Smilowitz
A note on asymptotic formulae for one-imensional network flow problems Carlos F. Daganzo an Karen R. Smilowitz (to appear in Annals of Operations Research) Abstract This note evelops asymptotic formulae
More informationWUCHEN LI AND STANLEY OSHER
CONSTRAINED DYNAMICAL OPTIMAL TRANSPORT AND ITS LAGRANGIAN FORMULATION WUCHEN LI AND STANLEY OSHER Abstract. We propose ynamical optimal transport (OT) problems constraine in a parameterize probability
More informationRobustness and Perturbations of Minimal Bases
Robustness an Perturbations of Minimal Bases Paul Van Dooren an Froilán M Dopico December 9, 2016 Abstract Polynomial minimal bases of rational vector subspaces are a classical concept that plays an important
More informationThe total derivative. Chapter Lagrangian and Eulerian approaches
Chapter 5 The total erivative 51 Lagrangian an Eulerian approaches The representation of a flui through scalar or vector fiels means that each physical quantity uner consieration is escribe as a function
More informationarxiv: v2 [math.cv] 2 Mar 2018
The quaternionic Gauss-Lucas Theorem Riccaro Ghiloni Alessanro Perotti Department of Mathematics, University of Trento Via Sommarive 14, I-38123 Povo Trento, Italy riccaro.ghiloni@unitn.it, alessanro.perotti@unitn.it
More informationSolutions to Math 41 Second Exam November 4, 2010
Solutions to Math 41 Secon Exam November 4, 2010 1. (13 points) Differentiate, using the metho of your choice. (a) p(t) = ln(sec t + tan t) + log 2 (2 + t) (4 points) Using the rule for the erivative of
More informationMulti-View Clustering via Canonical Correlation Analysis
Technical Report TTI-TR-2008-5 Multi-View Clustering via Canonical Correlation Analysis Kamalika Chauhuri UC San Diego Sham M. Kakae Toyota Technological Institute at Chicago ABSTRACT Clustering ata in
More informationMARTIN HENK AND MAKOTO TAGAMI
LOWER BOUNDS ON THE COEFFICIENTS OF EHRHART POLYNOMIALS MARTIN HENK AND MAKOTO TAGAMI Abstract. We present lower bouns for the coefficients of Ehrhart polynomials of convex lattice polytopes in terms of
More informationarxiv: v3 [math.mg] 3 Nov 2017
arxiv:702.0069v3 [math.mg] 3 ov 207 Random polytopes: central limit theorems for intrinsic volumes Christoph Thäle, icola Turchi and Florian Wespi Abstract Short and transparent proofs of central limit
More informationAffine surface area and convex bodies of elliptic type
Affine surface area and convex bodies of elliptic type Rolf Schneider Abstract If a convex body K in R n is contained in a convex body L of elliptic type (a curvature image), then it is known that the
More informationCOUNTING VALUE SETS: ALGORITHM AND COMPLEXITY
COUNTING VALUE SETS: ALGORITHM AND COMPLEXITY QI CHENG, JOSHUA E. HILL, AND DAQING WAN Abstract. Let p be a prime. Given a polynomial in F p m[x] of egree over the finite fiel F p m, one can view it as
More informationensembles When working with density operators, we can use this connection to define a generalized Bloch vector: v x Tr x, v y Tr y
Ph195a lecture notes, 1/3/01 Density operators for spin- 1 ensembles So far in our iscussion of spin- 1 systems, we have restricte our attention to the case of pure states an Hamiltonian evolution. Toay
More informationLecture 6 : Dimensionality Reduction
CPS290: Algorithmic Founations of Data Science February 3, 207 Lecture 6 : Dimensionality Reuction Lecturer: Kamesh Munagala Scribe: Kamesh Munagala In this lecture, we will consier the roblem of maing
More informationPerfect Matchings in Õ(n1.5 ) Time in Regular Bipartite Graphs
Perfect Matchings in Õ(n1.5 ) Time in Regular Bipartite Graphs Ashish Goel Michael Kapralov Sanjeev Khanna Abstract We consier the well-stuie problem of fining a perfect matching in -regular bipartite
More informationA simple tranformation of copulas
A simple tranformation of copulas V. Durrleman, A. Nikeghbali & T. Roncalli Groupe e Recherche Opérationnelle Créit Lyonnais France July 31, 2000 Abstract We stuy how copulas properties are moifie after
More informationu!i = a T u = 0. Then S satisfies
Deterministic Conitions for Subspace Ientifiability from Incomplete Sampling Daniel L Pimentel-Alarcón, Nigel Boston, Robert D Nowak University of Wisconsin-Maison Abstract Consier an r-imensional subspace
More information. ISSN (print), (online) International Journal of Nonlinear Science Vol.6(2008) No.3,pp
. ISSN 1749-3889 (print), 1749-3897 (online) International Journal of Nonlinear Science Vol.6(8) No.3,pp.195-1 A Bouneness Criterion for Fourth Orer Nonlinear Orinary Differential Equations with Delay
More informationOn non-antipodal binary completely regular codes
Discrete Mathematics 308 (2008) 3508 3525 www.elsevier.com/locate/isc On non-antipoal binary completely regular coes J. Borges a, J. Rifà a, V.A. Zinoviev b a Department of Information an Communications
More informationThe Three-dimensional Schödinger Equation
The Three-imensional Schöinger Equation R. L. Herman November 7, 016 Schröinger Equation in Spherical Coorinates We seek to solve the Schröinger equation with spherical symmetry using the metho of separation
More informationOn lower bounds for integration of multivariate permutation-invariant functions
arxiv:1310.3959v1 [math.na] 15 Oct 2013 On lower bouns for integration of multivariate permutation-invariant functions Markus Weimar October 16, 2013 Abstract In this note we stuy multivariate integration
More informationGeneralization of the persistent random walk to dimensions greater than 1
PHYSICAL REVIEW E VOLUME 58, NUMBER 6 DECEMBER 1998 Generalization of the persistent ranom walk to imensions greater than 1 Marián Boguñá, Josep M. Porrà, an Jaume Masoliver Departament e Física Fonamental,
More information