Random system of lines in the Euclidean plane E2 Giuseppe Caristi Abstract. In this paper we consider a random variable arising from a problem of geometrical intersection between a fixed convex body K and a system of random lines in E2 . M.S.C. 2000: 60D05,52A22. Key words: Geometric probability, stochastic geometry, random sets, random convex sets, integral geometry. 1 Introduction Let E2 be the Euclidean plane and let K be a convex non empty and bounded domain of area SK and with boundary ∂K of length L. We consider a family F of random, uniformly distributed n-lines {G1 , ..., Gn } with n ≥ 2. We assume that if Gh , Gk ∈ F , f . It is possible that this points belongs to K or not. In this way then Gh ∩ Gk 6= ¡ we have a random variable X(n,K) . In this paper we give the following result ³ ´ ¡ ¢ k Theorem 1. The expression of the mean value E X(n,K) , the k-moments E X(n,K) ¡ ¢ and the variance σ 2 X(n,K) of the random variable X(n,K) can be calculated as follows # " ³ ´ X ¡ ¢ SK πSK k! k E X(n,K) = απ 2 , E X(n,K) = , L J1 !...Jα ! L2 J1 +...+Jα =k ¡ ¢ πSK σ 2 X(n,K) = L2 µ ¶ πSK 1− 2 α2 , L where α= n (n − 1) , 2 and k is a positive integer. Other results about the computation of the variance are investigated in [2] and an extension in the 3-dimensional Euclidean space of the same problem is studied in [6]. Applied Sciences, Vol.10, 2008, pp. 48-53. c Balkan Society of Geometers, Geometry Balkan Press 2008. ° Random system of lines 2 49 Main Results Let N be the set of natural numbers, n ≥ 2 a fixed natural integer, {G1 , ..., Gn } and K as in introduction. We can state the following Theorem 2. Let us consider the random variable X(n,K) . Then ¢ ¡ SK E X(n,K) = απ 2 , L where α = K. n(n−1) ,L 2 is the length of ∂K and SK is the area of the domain defined by Proof. It is easy to see that, denoting with L the length of ∂K and dG the elementary measure of the lines in the Euclidean plane E2 , we have Z ½ ¾ dG = L. f G∩K6= ¡ Since G1 , ..., Gn are stochastically independent, we get Z ¾ dG1 ∧ ... ∧ dGn = Ln . ½ f G∩K6= ¡ If we consider the lines G1 , ..., Gn , then the intersection points might belong to K or not. Hence we obtain a random variable which we denote by X(n,K) . In order to compute the variance, we have the following integral Z ¾ X(n,K) dG1 ∧ ... ∧ dGn ½ f G∩K6= ¡ We define the application ²hk = 1 if Gh ∩ Gk ∈ K (with h 6= k = 1, ..., n) and zero otherwise. Then X X(n,K) = ²hk . h,k=1 Further, let us consider Z I2 := ½ f Gh ,Gk ∩K6= ¡ ¾ ²hk dGh ∧ dGk . If (Gh ∩ Ek ) ∈ K, then we have ²hk = 1. We denote with λk the chord intercepted by Gk on K (and its length). We have Z Z Z ½ ¾ ² dG ∧ dGk = ½ ¾ dGh dGk , ¾ ½ f hk h f f Gh ,Gk ∩K6= ¡ Gk ∩K6= ¡ Gk ∩K6= ¡ but it is well known that Z ½ f Gh ∩λk 6= ¡ ¾ dGh = λk , 50 Giuseppe Caristi Z ½ f Gk ∩K6= ¡ ¾ Moreover, Z ½ f G∩K6= ¡ λk dGk = πSK . Z ¾ X(n,K) dG1 ∧ ... ∧ dGn = X ½ f G∩K6= ¡ ¾ ²hk dG1 ∧ ... ∧ dGn = h,k=1 πSK Ln−2 + ... + πSK Ln−2 . Taking into account that the number of the different sets {Gh , Gk } is (the binomial coefficient) n (n − 1) α= , 2 we have Z ½ ¾ X(n,K) dG1 ∧ ... ∧ dGn = απSK Ln−2 . f G∩K6= ¡ By definition, R½ ¡ ¢ E X(n,K) := f G∩K6= ¡ R½ ¾ X(n,K) dG1 ∧ ... ∧ dGn f G∩K6= ¡ ¾ dG1 ∧ ... ∧ dGn , ³ ´ 2 and hence mathbf E X(n,K) = απ SLK2 . Now we compute ³ ´ R½ 2 E X(n,K) = We put f G∩K6= ¡ R½ ¾ 2 X(n,K) dG1 ∧ ... ∧ dGn f G∩K6= ¡ ¾ dG1 ∧ ... ∧ dGn . Z J= ½ f G∩K6= ¡ ¾ We can prove that 2 X(n,K) dG1 ∧ ... ∧ dGn 2 X(n,K) = X 2 ²hk , h,k=1 and then 2 X(n,K) = X ²2hk + 2 h,k=1 X ²hk ²sn . (h,k)6=(s,n) With this observations we can compute the integral Z X X ¾ ²2hk + 2 ²hk ²sn dG1 ∧ ... ∧ dGn . J= ½ f G∩K6= ¡ h,k=1 (h,k)6=(s,n) Random system of lines 51 Considering that Z Z ½ f G∩K6= ¡ ¾ ²2 hk dGh ∧ dGk = ½ f Gk ∩K6= ¡ Z ¾ ½ f Gh ∩λk 6= ¡ ¾ dGh dGk , where λk is the chord intercepted by Gk on K, we obtain J = πSK Ln−2 + ... + πSK Ln−2 + 2πSK Ln−2 + ... + 2πSK Ln−2 . Taking into account that the number of different sets {Gh , Gk } is α and that G1 , ..., G2 are independent, we infer ³ ´ πS αLn−2 πSK α2 K 2 E X(n,K) = = . n L L2 We obtain Theorem 3. Let us consider the random variable X(n,K) . Hence ¡ ¢ πSK σ 2 X(n,K) = L2 where α = K. n(n−1) ,L 2 µ ¶ πSK 1− 2 α2 , L is the length of ∂K and SK is the area of the domain defined by Moreover, we note that k X(n,K) = X k ²hk , h,k=1 implies ³ ´ " k E X(n,K) = X J1 +...+Jα =k 3 # k! πSK . J1 !...Jα ! L2 Applications 1. As first case we consider in the plane a square Q of side a. We can compute the following values ¡ ¢ απ E X(n,Q) = ≈ 0, 196345α. 16 " # ³ ´ X k! π k E X(n,Q) = . J1 !...Jα ! 16 J1 +...+Jα =k and the variance is σ 2 X(n,Q) = π´ 2 π ³ 1− α ≈ 0, 15779α2 . 16 16 52 Giuseppe Caristi 2. Taking in plane a rectangle R of sides a and b we have ¢ ¡ E X(n,R) = ³ E k X(n,R) ´ " απab 2, 4 (a + b) # k! π = , J1 !...Jα ! 4 (a + b)2 J1 +...+Jα =k X and the variance σ X(n,Q) = Ã π 2 2 4 (a + b) 1− ! π 4 (a + b) 2 α2 . 3. Let C be a circle of radius δ. We have ³ ´ α k E X(n,C) = , 4 # " ³ ´ X k! α k E X(n,C) = J1 !...Jα ! 4 J1 +...+Jα =k and the variance is ¡ ¢ π³ α´ 2 σ 2 X(n,C) = 1− α ≈ 0, 16855α2 . 4 4 4. As last case we consider an equilateral triangle T of side a, obtaining, ³ ´ απ √3 k ≈ 0, 3023α, E X(n,T ) = 18 " # √ ³ ´ X k! π 3 k E X(n,T ) = , J1 !...Jα ! 18 J1 +...+Jα =k and for the variance the expression σ ¡ 2 X(n,T ) ¢ √ Ã √ ! π 3 π 3 = 1− α2 ≈ 0, 2109α2 . 18 18 Remark 4. We observe that in examples 1, 3, 4 the functions are independent of the dimension of the convex body. References [1] W. Blaschke, Vorlesungen über Integralgeometrie, III Auflage, V.E.B. Deutscher Verlag der Wiss., Berlin 1955. [2] G. Caristi amd G. Moltica Bisci, On the variance associated to a family of ovaloids in the Euclidean Sapce E3 , Bollettino U.M.I. (8) 10-B (2007), 87-98. Random system of lines 53 [3] E. Czuber, Zur Theorie der geometrischen Wahrscheinlichketein, Sitz. Akad. Wiss. Wien 90 (1884), 719-742. [4] R. Deltheil, Probabilitiés géométriques, Gauthier-Villars, Paris 1926. [5] A. Duma and M. Stoka, Probabilità geometriche per sistemi di piani nello spazio euclideo E3 , Rend. Sem. Mat. Messina 7 (2000), 5-11. [6] G. Molica Bisci, Random systems of planes in the Euclidean space E3 , Atti Acc. Scienze di Torino 2005, 43-48. [7] H. Poincaré, Calcul des probabilitiés, Ed. 2, Carré, Paris, 1912. Author’s address: Giuseppe Caristi University of Messina, 75 Via dei Verdi, 98122 Messina, Italy. E-mail: gcaristi@unime.it