The moments of the distance between two random points

in a regular polygon

Uwe Bäsel

Abstract
In this paper, we derive formulas for the analytical calculation of the moments of the distance
between two uniformly and independently distributed random points in an n-sided regular
polygon. A number of closed form expressions is provided, e.g. the expected distances for
arXiv:2101.03815v1 [math.PR] 11 Jan 2021

n = 3, 4, 5, 6, 8, 10, 12, where the results for n = 5, 8, 10, 12 are new to the best of the au-
thor’s knowledge. Applying results of Voss, remarkably short formulas for the second and
the fourth moments are derived.
2010 Mathematics Subject Classification: 60D05, 52A22, 53C65
Keywords: Geometric probability, random distance, mean distance, moments of random
distances, chord length distribution function, distance distribution function, regular poly-
gons, chord power integrals

1 Introduction
Let Pn, r , n = 3, 4, . . ., be the regular polygon with n sides and circumscribed circle with radius r.
The distance between two points chosen independently and uniformly from Pn, r is a random
variable, which we denote by ∆n,r . The moments of ∆n,r are given by
Z d
Mm (Pn, r ) := E[∆m
n,r ] = xm dGn,r (x) ,
0

where Gn,r is the cumulative distribution function of ∆n,r , and d is the diameter of Pn, r . The
first moment M1 (Pn, r ) is the mean distance between two (random) points in Pn, r .
Czuber (1884) found the mean values M1 (P3,r ) (p. 206) and M1 (P4,r ) (pp. 202-204, Prob-
lem XV) for the equilateral triangle and the square, respectively, in terms of the side length
a = 2r sin(π/n). (Czuber even found the mean distance in a rectangle. Concerning equilateral
triangle, square and rectangle see also Santaló (1976, p. 49).) Ghosh (1951) derived the distance
distribution for a rectangle. Sulanke (1961) found the distribution function G3,r for the equilat-
eral triangle in terms of the side length (see also Duma and Stoka (2008)). The density function
gn,r and the distribution function Gn,r for any regular polygon were obtained by Bäsel (2014).
An example for the simulation of point distances is shown in Fig. 1.
Hammersley (1950) derived the distribution of the distance in a hypersphere. Bailey, Borwein,
and Crandall (2007) studied the mean distance in the unit n-cube giving closed form expressions
for the cases n = 1, . . . , 4 (see also Borwein and Crandall (2013), especially Example 14). The
density function of the distance between two points in a three-dimensional box was found by
Philip (2007). Furthermore, Philip (n.d.) found the density function for a 4-cube and a 5-cube.
Although the density function and the distribution function provide much more information
than the mean value, there is a constant interest in mean values; see e. g. Dunbar (1997), and
Burgstaller and Pillichshammer (2009). Recently, Bonnet et al. (2020) obtained sharp inequalities
for the mean distance in convex bodies.
 Fig. 1: Simulation of 1000 line segments
each as a connection of two random points
in P5,1 using PolygonV111.jar (Achtelstädter
and Hübener, 2013)
Here the empirical values of M1 (P5,1 ) and the
variance Var[∆5,1 ] are 0.7911 (cf. Tab. 2) and
0.1316, respectively. Range: 1.6795

Without loss of generality one may assume that the centroid of Pn, r is located at the origin O
of a cartesian x, y-coordinate system, and the vertices of Pn, r are given by

Xk = xk + iyk = re2ikα , k = 0, 1, . . . , n − 1, (1)

with angle
1  # » # » π
α=] OX0 , OX1 = .
2 n
The distance `k between vertex X0 and vertex Xk is for k = 1, 2, . . . , n − 1 given by
q
`k = r (cos(2kα) − 1)2 + sin2 (2kα) = 2r sin(kα) . (2)

Clearly, `1 = 2r sin α is the side length, and, denoting by b·c the integer part of · ,

d := `K+1 with K := bn/2c − 1 (3)

is the diameter of Pn, r . The length L(∂ Pn, r ) of the boundary ∂ Pn, r and the area A(Pn, r ) of
Pn, r are respectively

`1 1
L(∂ Pn, r ) = n `1 = 2nr sin α and A(Pn, r ) = n r cos α = nr2 sin(2α) . (4)
2 2
For Pn, r with odd n we denote by λ the distance between the vertex X0 and the side XK+1 XK+2 ,

λ = r + r cos α = 2r cos2 (α/2) . (5)

We will also use (5) for Pn, r with even n. Here λ is the sum of r = OX0 and the distance
between O and side XK XK+1 (and also XK+1 XK+2 ). For Pn, r , n = 3, 4, 5, . . ., we have

`K < λ < `K+1 = d .

The paper is organized as follows:

• In Section 2 we provide a simple and clear form of the chord length distribution function Fn,r
derived in Bäsel (2014). This simple form is completely given in Theorem 1.

2
• For the point distance density function gn,r and the moments Mm (Pn, r ) required antideriva-
tives are provided in Section 3.
• From Fn,r we conclude in Section 4 a simple and clear form of the point distance density
function gn,r derived in Bäsel (2014).
• Formulas for analytically calculating the distance moments Mm (Pn, r ), m = −1, 0, 1, 2, . . .,
are derived from Fn,r in Section 5.
• Closed form expressions for the mean distances M1 (Pn, r ) and a number of further moments
are given for n = 3, 4, 5, 6, 8, 10, 12 in Section 6. According to the author’s knowledge, cases
n = 5, 8, 10, 12 are new results.
• High precision values with respectively 75 and 76 digits for M1 (Pn, r )/r, n = 3, 4, . . . , 30, and
Mm (P5,r )/rm , m = −1, 0, 1, 2, . . . , 10, are given in Section 7.
• Using a result from Czuber (1884), in Section 8 a short formula for the second moments
M2 (Pn, r ) is derived. This makes it easier to calculate the variances Var[∆n,r ]. Especially in
all cases, where a (short) closed form expression for the mean distance M1 (Pn, r ) is found, one
has therefore also a closed form expression for the variance.
• Applying a result from Voss (1982), we obtain in Section 9 a short formula for the fourth
moments M4 (Pn, r ). For this purpose it is used that the value of the second area moment of
Pn, r with respect to an axis through the centroid of Pn, r is independent from the direction
of this axis.
• Since clearly the moments for the circle Cr with radius r are the limit values of the moments
for Pn, r as n tends to infinity, limn→∞ Mm (Pn, r ) = Mm (Cr ), in Section 10 the already known
moments are derived as values for comparison. (Comparisons have already been done in
previous sections.)
• Finally, we point out the possibility of the simple derivation of chord power integrals for Pn, r
from the present results (Section 11).

2 The chord length distribution function

A straight line g in the plane is in Hesse normal form determined by the angle ϕ, 0 ≤ ϕ < 2π,
between the positive x-axis and the direction perpendicular to g, and by its distance p, 0 ≤ p <
∞, from the origin O,

g = g(p, ϕ) = (x, y) ∈ R2 : x cos ϕ + y sin ϕ = p



(see Santaló (1976, p. 2)). We assume p and ϕ to be independent random variables uniformly
distributed in [0, r] and [0, 2π), respectively, and consider only lines g with g ∩ Pn, r 6= ∅. Such a
line g produces a chord of length Λn,r , 0 ≤ Λn,r ≤ d = `K+1 (see (2), (3)). We denote by Fn,r
the distribution function of the random variable Λn,r , Fn,r (x) = P (Λn, r ≤ x).
Theorem 1. With

α = π/n , K = bn/2c − 1 , `k = 2r sin(kα) , λ = 2r cos2 (α/2) ,

(
0 if n = 3,
hk = 2r sin(kα) sin((k + 1)α) , pk =
cot(2kα) − cot(2(k + 1)α) if n > 3,
(
1/2 if n = 3 ,
qk = tan(kα) − tan((k + 1)α) , sk =
k α pk if n > 3 ,
a p
σµ,a (x) = xµ arcsin , τµ,a (x) = xµ x2 − a2 ,
x

3
the chord length distribution function Fn,r of the regular polygon Pn, r , n = 3, 4, . . ., is given by


 0 if −∞ < x < `0 = 0 ,
Fn,r (x) = P (Λn, r ≤ x) = 1 − Hk (x)/`1 if `k ≤ x < `k+1 for k = 0, . . . , K,

`K+1 ≤ x < ∞

1 if

with


 `1 − [1 + α (tan α − cot α)] x/2 if k = 0 ∧ x < λ ,



pk σ1,hk (x) − [kα cot(2kα) − (k + 1)α cot(2(k + 1)α)] x





− (hk pk + 2rqk cos α) τ−1,hk (x) if (n is even ∧ 1 ≤ k ≤ K − 1) ∨






(n is odd ∧ 1 ≤ k ≤ K ∧ x < λ) ,





Hk (x) = [1/2 − kα cot(2kα)] x + cot(2kα) σ1,hk (x) − (hk − 2r cos α) hk /x


 − [hk cot(2kα) + 2r cos α tan(kα)] τ−1,h (x) if n is even ∧ k = K ,


 k






 pk σ1,hk (x) + 2 cot(2(k + 1)α) pk σ1,λ (x) − [(π − α) cot(2(k + 1)α) + sk ] x

− (hk pk + 2rqk cos α) τ−1,hk (x) − 2 cos α [2r cos(α/2) sec((k + 1)α)






 − λ csc(2(k + 1)α)] τ−1,λ (x) if n is odd ∧ k = K ∧ x ≥ λ .

Proof. This is the chord length distribution function from Bäsel (2014, Theorem 1), simplified
by inserting, rearranging and combining terms. Furthermore, the former case

(n is even ∧ k ∈ {0, . . . , K − 1}) ∨ (n is odd ∧ x < λ)

is split into two cases.

Examples for graphs of distribution functions Fn,r are shown in Fig. 2.

1.0

0.8 7

n=3 6
0.6 10
4
0.4 �
5 11
0.2 Fig. 2: Graphs of distribution
functions Fn,r , n = 3, 4, 5, 6,
7, 10, 11, and limit distribution
0.5 1.0 1.5 x/r 2.0 function (circle)

3 Some antiderivatives
Lemma 1. For µ = −1, 0, 1, 2, . . ., an antiderivative ψeµ,a (x) of the function

xµ
ψµ,a : [a, ∞) → R , x 7→ ψµ,a (x) := √ , a > 0,
x2 − a2

4
is given by
 1 a
 − arcsin if µ = −1 ,
a x




 √
x + x2 − a2



 ln if µ = 0 ,


a

ψµ,a (x) =
e
µ−1
√
 x x2 − a2 γµ,a (x)
if µ = 1, 3, 5, . . . ,


µ



√ √


µ−1 x − a2 γµ,a (x) (µ − 1)!! aµ
 2
 x

 x + x2 − a2
 + ln if µ = 2, 4, 6, . . . ,
µ µ!! a
where
b µ−1
2
c ν
X a 2ν Y µ + 1 − 2j
γµ,a (x) = 1 + .
x µ − 2j
ν=1 j=1

Proof. In the case µ = −1 we have

Z Z
dx dx
ψe−1,a (x) = √ = p .
x x2 − a2 x2 1 − (a/x)2
The substitution u = a/x, du = −a dx/x2 gives
Z
1 du 1 1 a
ψ−1,a (x) = −
e √ = − arcsin u = − arcsin .
a 1−u 2 a a x
For µ = 0, 1, 2, . . ., the substitution
p
x = a cosh u , x2 − a2 = a sinh u , dx = a sinh u du
yields
xµ
Z Z
ψeµ,a (x) = √ dx = aµ coshµ u du .
x2 − a2
For µ = 0 it follows that
√
x2 − a2
Z
x x+
ψe0,a (x) = du = u = arcosh = ln .
a a
For µ = 1, 3, 5, . . . from Eq. 2.412.4 in Gradstein and Ryshik (1981, Vol. 1, p. 127) we get
 µ−1   
2 ν
µ + 1 − 2j 
Z
sinh u  X Y
coshµ u du = coshµ−1 u +  coshµ−1−2ν u
µ µ − 2j
ν=1 j=1
 µ−1 
µ−1 2 ν
cosh u sinh u  X 1 Y µ + 1 − 2j 
= 1+ , (6)
µ
ν=1
cosh2ν u j=1 µ − 2j

whereas for µ = 2, 4, 6 . . . from Eq. 2.412.1 in (ibid., Vol. 1, p. 127) we get

 µ 
−1 ν
µ−1 2
µ + 1 − 2j  (µ − 1)!!
Z
cosh u sinh u  X 1 Y
coshµ u du = 1+ 2ν + u. (7)
µ cosh u µ − 2j µ!!
ν=1 j=1

The back substitution

√ r  !
x x2 − a2 x x x 2
cosh u = , sinh u = , u = arcosh = ln + −1
a a a a a

in (6) and (7) yields ψeµ,a (x) for the remaining cases in Lemma 1.

5
Examples 1. From Lemma 1 one gets the following special cases
p
ψe1,a (x) = x2 − a2 ,
√ !
1 p x + x 2 − a2
ψe2,a (x) = x x2 − a2 + a2 ln ,
2 a
1 2
p
ψe3,a (x) = x + 2a2 x2 − a2 ,
3
" √ #
1 2 2

p
2 2 4 x + x2 − a2
ψ4,a (x) =
e x 2x + 3a x − a + 3a ln ,
8 a
1
p
ψe5,a (x) = 3x4 + 4a2 x2 + 8a4 x2 − a2 ,
15
" √ #
1
p x + x 2 − a2
4 2 2 4 6
ψe6,a (x) = x 8x + 10a x + 15a x2 − a2 + 15a ln .
48 a

Remark 1. The integral

xµ−1 xµ
Z Z
p dx = √ dx
1 − (a/x)2 x2 − a2

is also used and discussed in the context of computing the moments of a point distance associated
with triangles in Li and Qiu (2016, p. 4).
Lemma 2. For µ = 0, 1, 2, . . ., an antiderivative σ
eµ,a (x) of the function
a
σµ,a : [a, ∞) → R , x 7→ σµ,a (x) := xµ arcsin , a ≥ 0,
x
(from Theorem 1) is given by
 1  a 
 xµ+1 arcsin + a ψeµ,a (x) if a > 0 ,
σ
eµ,a (x) = µ+1 x

0 if a = 0 ,

with ψeµ,a (x) according to Lemma 1.

Proof. Clearly, σ
eµ,0 (x) ≡ 0. If a > 0, then integration by parts gives
Z
a 1  µ+1 a 
eµ,a (x) = xµ arcsin dx =
σ x arcsin + a ψeµ,a (x) .
x µ+1 x

Lemma 3. For µ = −1, 0, 1, 2, . . ., an antiderivative τeµ,a (x) of the function

p
τµ,a : [a, ∞) → R , x 7→ τµ,a (x) := xµ x2 − a2 , a ≥ 0 ,

(from Theorem 1) is given by

 1  p 
µ+1 2 − a2 − a2 ψ

 µ+2
 x x eµ,a (x) if a > 0 ,
τeµ,a (x) =
 xµ+2

 if a = 0 ,
µ+2

with ψeµ,a (x) according to Lemma 1.

6
Proof. We have
xµ+2
Z
τeµ,0 (x) = xµ+1 dx = .
µ+2
In the case a > 0 we have
Z µ 2
x (x − a2 ) xµ+2
Z p Z
τeµ,a (x) = x µ 2 2
x − a dx = √ dx = √ dx − a2 ψeµ,a (x) . (8)
2
x −a 2 2
x −a 2

Integration by parts with

x
u = xµ+1 , v0 = √ ,
x − a2
2
p
u0 = (µ + 1)xµ , v = x2 − a2

yields

xµ+2
Z p Z p
√ dx = xµ+1 x2 − a2 − (µ + 1) xµ x2 − a2 dx
x2 − a2
p
= xµ+1 x2 − a2 − (µ + 1) τeµ,a (x) .

Putting this into (8) provides

p
τeµ,a (x) = xµ+1 x2 − a2 − (µ + 1) τµ,a (x) − a2 ψeµ,a (x) ,

hence
1  µ+1 p 2 
τeµ,a (x) = x x − a2 − a2 ψeµ,a (x) .
µ+2

Remark 2. I do not use τeµ,a (x) in the closer form

xµ+2
 Z 
1 p 1  µ+1 p 2 
τeµ,a (x) = xµ+1 x2 − a2 − √ dx = x x − a2 − ψeµ+2,a (x) ,
µ+1 x2 − a2 µ+1

which follows immediately by partial integration from the first integral in (8), because it does
not hold for the case µ = −1, which is also needed in the following.

4 The point distance density function

Theorem 2. With L(∂ Pn, r ) and A(Pn, r ) according to (4), α, K, `k , λ, hk , pk , qk , sk as in
Theorem 1,
1 2 a p  p a
σ
e1,a (x) = x arcsin + a x2 − a2 , τe−1,a (x) = x2 − a2 + a arcsin ,
2 x x
the density function gn,r of the distance ∆n,r between two uniformly and independently distributed
random points in Pn, r , n = 3, 4, . . ., is given by
 " #
 2x L(∂ Pn, r ) φ[ (x)
π− if x ∈ [0, `K+1 ) ,


gn,r (x) = A(Pn, r ) A(Pn, r ) `1


 0 if x ∈ R \ [0, ` ), K+1

where
k−1
X
[
φ (x) = Jν[ (`ν+1 ) + Jk[ (x) if `k ≤ x < `k+1 , k = 0, . . . , K
ν=0

7
with
 [
 H0,k (x) if k = 0 ∧ x < λ ,


 [ (x) − H [ (` )
H1,k if (n is even ∧ 1 ≤ k ≤ K − 1) ∨
1,k k





 (n is odd ∧ 1 ≤ k ≤ K ∧ x < λ) ,
Jk[ (x) = [ (x) − H [ (` )

 H2,k 2,k k if n is even ∧ k = K ,

[ (x) − H [ (λ) + H [ (λ)




 H3,k 3,k 0,k if n = 3 ∧ k = 0 ∧ x ≥ λ ,

 [ [ [ [
H3,k (x) − H3,k (λ) + H1,k (λ) − H1,k (`k ) if n is odd ∧ n > 3 ∧ x ≥ λ ,
and
[
H0,k (x) = `1 x − [1 + α (tan α − cot α)] x2 /4 ,
[
H1,k e1,hk (x) − [kα cot(2kα) − (k + 1)α cot(2(k + 1)α)] x2 /2
(x) = pk σ
− (hk pk + 2rqk cos α) τe−1,hk (x) ,
[
H2,k (x) = [1/2 − kα cot(2kα)] x2 /2 + cot(2kα) σ
e1,hk (x) − (hk − 2r cos α) hk ln x
− [hk cot(2kα) + 2r cos α tan(kα)] τe−1,hk (x) ,
[
H3,k (x) = pk σ e1,λ (x) − [(π − α) cot(2(k + 1)α) + sk ] x2 /2
e1,hk (x) + 2 cot(2(k + 1)α) σ
− (hk pk + 2rqk cos α) τe−1,hk (x)
− 2 cos α [2r cos(α/2) sec((k + 1)α) − λ csc(2(k + 1)α)] τe−1,λ (x) .

Proof. From Piefke’s (1978, p. 130) result it immediately follows that the density functions fn,r
and gn,r are connected by
2L(∂ Pn, r ) x d
Z
gn,r (x) = (s − x)fn,r (s) ds .
A(Pn, r )2 x
In Bäsel (2014) (see pp. 10-11) we concluded that
L(∂ Pn, r ) φ\ (x)
 
2x
gn,r (x) = π− (9)
A(Pn, r ) A(Pn, r )
where Z x Z x
φ\ (x) := x − Fn,r (s) ds = (1 − Fn,r (s)) ds . (10)
0 0
So for `k ≤ x ≤ `k+1 , k = 0, . . . , K, we have
k−1
X
`1 φ\ (x) = φ[ (x) := Jν[ (`ν+1 ) + Jk[ (x) (11)
ν=0
where Z x
Jν[ (x) = Hν (s) ds (12)
`ν
with Hν according to Theorem 1 (the sum in (11) is empty if k = 0). Considering the necessary
case distinctions, the result of Theorem 2 follows from evaluating the integral (12) for ν =
0, . . . , K. Note that the condition k = 0 ∧ x < λ is an abbreviation for the equivalent, more
detailed condition
(n = 3 ∧ k = 0 ≤ x < λ) ∨ (n > 3 ∧ k = 0) .
The expression for σ e1,a (x) follows from Lemma 2 with ψe1,a (x) in Examples 1, whereas the
expression for τe−1,a (x) follows from Lemma 3 with ψe−1,a (x) according to Lemma 1.

Examples for graphs of density functions are shown in Figures 3 and 4. fn,r (x) = dFn,r (x)/dx
denotes the chord length density function.

8
 2.0

1.5

1.0

0.5
Fig. 3: Graphs of density func-
tions for P7,r : r × f7,r (x), and
0.5 1.0 1.5 x/r 2.0 r × g7,r (x) (dashed)

2.0

1.5

1.0

0.5
Fig. 4: Graphs of density func-
tions for P8,r : r × f8,r (x), and
0.5 1.0 1.5 x/r 2.0 r × g8,r (x) (dashed)

5 Moments of the distance

Theorem 3. With
• L(∂ Pn, r ), A(Pn, r ) according to (4),
• α, K, `k , λ, hk , pk , qk , sk as in Theorem 1,
• σ
eµ,a (x) according to Lemma 2,
• τeµ,a (x) according to Lemma 3,
the moments of the distance between two uniformly and independently distributed random points
in Pn, r are given by
K
2L(∂ Pn, r ) X
Mm (Pn, r ) = Jem,k (`k+1 ) , m = −1, 0, 1, 2, . . .
(m + 2) A(Pn, r )2 `1
k=0

where

 e (0) (x)
H if k = 0 ∧ x < λ ,

 m,k

e (1) (x) e (1) (`k )

− if (n is even ∧ 1 ≤ k ≤ K − 1) ∨



 Hm,k Hm,k

(n is odd ∧ 1 ≤ k ≤ K ∧ x < λ) ,



Jem,k (x) = e (2) (x) − H
e (2) (`k )

 Hm,k m,k if n is even ∧ k = K ,


e (3) (x) − H
e (3) (λ) + H
e (0) (λ)

H if n = 3 ∧ k = 0 ∧ x ≥ λ ,




 m,k m,k m,k


 e (3) (x) − H
H e (3) (λ) + H
e (1) (λ) − H
e (1) (`k ) if n is odd ∧ n > 3 ∧ x ≥ λ
m,k m,k m,k m,k

9
with
m+3 [1 + α (tan α − cot α)] xm+4
e (0) (x) = `1 x
H − ,
m,k m+3 2(m + 4)

e (1) (x) = pk σ xm+4

Hm,k em+3,hk (x) − [kα cot(2kα) − (k + 1)α cot(2(k + 1)α)]
m+4
− (hk pk + 2rqk cos α) τem+1,hk (x) ,
 m+4
xm+2

(2) 1 x
H
e (x) =
m,k − kα cot(2kα) + cot(2kα) σ
e m+3,hk (x) − (hk − 2r cos α) h k
2 m+4 m+2
− [hk cot(2kα) + 2r cos α tan(kα)] τem+1,hk (x) ,
e (3) (x) = pk σ
H em+3,hk (x) + 2 cot(2(k + 1)α) σ
em+3,λ (x)
m,k

xm+4
− [(π − α) cot(2(k + 1)α) + sk ] − (hk pk + 2rqk cos α) τem+1,hk (x)
m+4
− 2 cos α [2r cos(α/2) sec((k + 1)α) − λ csc(2(k + 1)α)] τem+1,λ (x) .

Proof. From
Z d
Mm (Pn, r ) = E[∆m
n,r ] = xm gn,r (x) dx
0
with (9) and (10) we get
d
2L(∂ Pn, r ) d m+1 x
Z Z Z
2π
Mm (Pn, r ) = xm+1 dx − x (1 − Fn,r (s)) ds dx
A(Pn, r )
0 A(Pn, r )2 x=0 s=0

2πdm+2 2L(∂ Pn, r ) d m+1 x

Z Z
= − x (1 − Fn,r (s)) ds dx .
(m + 2)A(Pn, r ) A(Pn, r )2 x=0 s=0

Integration by parts with

Rx
u= 0 (1 − F (s)) ds , v 0 = xm+1 ,
u0 = 1 − F (x) , v = xm+2 /(m + 2)
yields
  m+2 Z x d
2πdm+2

2L(∂ Pn, r ) x
Mm (Pn, r ) = − (1 − Fn,r (s)) ds
(m + 2)A(Pn, r ) A(Pn, r )2 m+2 0 x=0
Z d 
1 m+2
− x (1 − Fn,r (x)) dx
m+2 0
2πdm+2 2L(∂ Pn, r ) dm+2 d
Z
= − (1 − Fn,r (s)) ds
(m + 2)A(Pn, r ) (m + 2)A(Pn, r )2 0
Z d
2L(∂ Pn, r )
+ xm+2 (1 − Fn,r (x)) dx .
(m + 2)A(Pn, r )2 0
Now, integration by parts with
u = 1 − Fn,r (s) , v0 = 1 ,
u0 = −fn,r (s) , v =s
gives
Z d d
Z d
(1 − Fn,r (s)) ds = s (1 − Fn,r (s)) + s fn,r (s) ds .
0 | {z }0 | 0
{z }
0 E[Λn,r ]

10
With
πA(Pn, r )
E[Λn,r ] = (Santaló, 1976, p. 30)
L(∂ Pn, r )
we have found
d
2L(∂ Pn, r )
Z
Mm (Pn, r ) = xm+2 (1 − Fn,r (x)) dx . (13)
(m + 2)A(Pn, r )2 0
It follows that
K Z `k+1
2L(∂ Pn, r ) X
Mm (Pn, r ) = xm+2 (1 − Fn,r (x)) dx
(m + 2)A(Pn, r )2 `k k=0
K Z `k+1
2L(∂ Pn, r ) X
= xm+2 Hk (x) dx .
(m + 2)A(Pn, r )2 `1
k=0 | `k {z }
=: Jem,k (`k+1 )
with Hk (x), k ∈ {0, 1, . . . , N }, from Theorem 1. So it remains to determine the antiderivatives
Z
e (q) (x) := xm+2 Hk (x) dx ,
Hm,k

where the q is used to indicate the necessary case distinctions.

The antiderivatives σ
em+3,a (x) and τem+1,a (x) are used because of
Z Z
m+2 a
x σ1,a (x) dx = xm+2 x arcsin dx = σ em+3,a (x)
x
and √
x2 − a2
Z Z
m+2 m+2
x τ−1,a (x) dx = x dx = τem+1,a (x) ,
x
respectively.

Remark 3. Note that, although the expressions are in general quite long, Theorem 3 always
provides closed form expressions for the moments of every Pn, r .
Remark 4. The relation (13) also follows from
2
Tm = Sm+3 , m = −1, 0, 1, 2, . . . , (14)
(m + 2)(m + 3)
where Sm and Tm are the m-th chord power integral and the m-th distance power integral,
respectively (Blaschke (1955, pp. 19-20), Santaló (1976, pp. 46-47), see also Schneider and Weil
(2008, p. 364, proof of Thm. 8.6.6; p. 374)). With
Sm (Pn, r ) Sm (Pn, r ) Tm (Pn, r ) Tm (Pn, r )
E[Λm
n,r ] = = and Mm (Pn, r ) = E[∆m
n,r ] = = (15)
S0 (Pn, r ) L(∂ Pn, r ) T0 (Pn, r ) A(Pn, r )2
one gets
L(∂ Pn, r ) d m+3
Z
2
Mm (Pn, r ) = x dFn,r (x)
(m + 2)(m + 3) A(Pn, r )2 0
Z d
2L(∂ Pn, r )
= xm+2 (1 − Fn,r (x)) dx
(m + 2)A(Pn, r )2 0
(see also Aharonyan and Ohanyan (2016, Eq. (2.3))).
Heinrich (2009) derived a general formula for the second-order chord power integrals, S2 , for
Pn, r . This gives T−1 and hence E[∆−1
n,r ] (see also Section 11).

11
Corollary 1. The mean distance M1 (Pn, r ) between two uniformly and independently distributed
random points in Pn, r is given by Theorem 3 with m = 1. If a > 0, then
" √ #
1 a
p x + x 2 − a2
5 2 2 5
σe4,a (x) = 8 x arcsin + a x 2x + 3a x2 − a2 + 3 a ln ,
40 x a
" √ #
1
p x + x 2 − a2
2 2 4
τe2,a (x) = x 2x − a x2 − a2 − a ln .
8 a

6 Examples of closed form expressions for the moments

For abbreviation we denote the moments in the following with Mm instead of Mm (Pn, r ). From
the context it will be clear for which polygon Pn, r the moments Mm are valid.
• 3-gon (equilateral triangle)
√ √ √
M−1 4 3 ln 3 M0 M1 3 3 3 ln 3 M2 1
−1
= , 0
= 1, = + , 2
= ,
r 3 r r 5 20 r 2
√ √ √ √
M3 51 3 81 3 ln 3 M4 9 M5 383 3 405 3 ln 3 M6 747
3
= + , 4
= , 5
= + , 6
= ,
r 280 1120 r 20 r 1792 7168 r 1400
√ √
M7 6669 3 5103 3 ln 3 M8 261
7
= + , 8
= ,
r 22528 90112 r 350
√ √
M9 977913 3 1240029 3 ln 3 M10 2511
= + , =
r9 2129920 18743296 r10 2156
√
With side length `1 = 3 r we have
 
M1 1 3 ln 3 3 1 1
= + = + ln 3 ≈ 0.36479184330021645371 .
`1 5 20 5 3 4

This result is due to Czuber (1884, p. 206) (see also Santaló (1976, p. 49), Weisstein (n.d.[b]),
OEIS A093064 (2004), Zhuang and Pan (2012), and Aharonyan and Ohanyan (2016)).
• 4-gon (square)
√ √ √ √

M−1 4 2 2 √  √  M0 M1 2 2 2 2 ln 1 + 2
=− + + 2 2 ln 1 + 2 , = 1, = + + ,
r−1 3 3 r0 r 15 15 3
√ √ √

M2 2 M3 34 8 2 2 ln 1 + 2 M4 34
2
= , 3
= + + , 4
= ,
r 3 r 105 105 5 r 45
√ √ √

M5 73 4 2 5 2 ln 1 + 2 M6 116
= + + , = ,
r5 126 63 28 r6 105
√ √ √

M7 3239 32 2 7 2 ln 1 + 2 M8 2992
7
= + + , 8
= ,
r 2970 495 36 r 1575
√ √ √

M9 1721 32 2 21 2 ln 1 + 2 M10 7648
9
= + + , 10
=
r 780 429 88 r 2079
√
With side length `1 = 2 r we have
M1 1 h√  √ i
= 2 + 2 + 5 ln 1 + 2 ≈ 0.52140543316472067833 .
`1 15

12
This result is due to Czuber (1884, pp. 202-204, Problem XV). It can also be found in Ghosh
(1951), Santaló (1976, p. 49), Finch (2003, p. 479), Weisstein (n.d.[a]), OEIS A091505 (2004),
and Aharonyan and Ohanyan (2016).
• 5-gon (pentagon)
√ q
M1 2 √ h √  √   √  √ i
= 5 + 5 24 5 − 8 − 35 + 9 5 ln 5 − 2 25 + 11 5 ln 5−2 ,
r 480
√ √ √
M2 7 5 M4 47 11 5 M6 167 2 5
= + , = + , = + ,
r2 12 12 r4 72 72 r6 168 7
√ √
M8 65 145 5 M10 427375 625 5
= + , = +
r8 36 252 r10 116424 504
• 6-gon (hexagon)
√

M−1 16 8 22 ln 3 4 ln 2 + 3 M0
=− + √ + − , = 1,
r−1 9 3 3 9 9 r0
√

M1 7 7 19 ln 3 ln 2 + 3 M2 5
=− + √ + − , 2
= ,
r 90 10 3 40 60 r 6
√ √

M3 817 117 3 867 ln 3 3 ln 2 + 3 M4 209
3
= + + − , 4
= ,
r 5040 560 2240 1120 r 180
√ √

M5 146431 963 3 7045 ln 3 5 ln 2 + 3 M6 573
= + + − , = ,
r5 290304 3584 14336 7168 r6 280
√ √

M7 7886969 8541 3 139797 ln 3 21 ln 2 + 3 M8 13037
7
= + + − , 8
= ,
r 6082560 20480 180224 90112 r 3150
√ √

M9 1395486403 34152111 3 52256421 ln 3 1701 ln 2 + 3 M10 76273
9
= + + − , 10
=
r 421724160 46858240 37486592 18743296 r 8316
The mean distance was obtained by Zhuang and Pan (2011) in the form
√  √  √
M1 7 3 7 1 h  i
= − + 28 ln 2 3 + 3 + 29 ln 2 3 − 3
`1 30 90 60
with side length `1 = r. (Only the numerical approximation there is not correct.) Zhuang and
Pan (ibid.) also derived the second moment.
• 8-gon (octagon)
√ " √ √ !q
√ 1 13 2 √
q
M1 4 2π 1 2 2π
= +2 2− 2 + − + + 2+ 2
r 3 5 120 20 60 3
√ ! √ ! 
1 91 2  √  1 7 2 √ 
− 14 + ln 2 + 2 + 1+ ln 2 − 2
48 10 48 10
√ !  √ !  #
√ √
q  q
1 1 2 1 29 2
+ − ln 2 + 2 + 2 + 21 + ln 2 + 2 − 2 ,
20 3 4 40 2
√ √ √
M2 2 2 M4 77 16 2 M6 157 23 2
= + , = + , = + ,
r2 3 6 r4 90 45 r6 105 30
√ √
M8 326 928 2 M10 2132 3002 2
= + , = +
r8 105 525 r10 297 693

13
• 10-gon (decagon)
√ √ √ 
 q  
M1 1
= 4 5125 + 2110 5 − 24 − 11 5 − 505 + 239 5 ln 2
r 600
 √   √   √ 
− 30 4 + 3 5 ln 5 − 205 + 107 5 ln 1 + 5
√   √   √  √ √
  q 
+ 705 + 343 5 ln 3 + 5 − 105 + 47 5 ln 5 + 10 + 2 5

√  √ √
  q 
− 65 − 29 5 ln 5 + 10 − 2 5

√  √ √
  q 
+ 5 + 3 5 ln 5 − 5 + 2 10 − 2 5 ,
√ √ √
M2 3 5 M4 121 73 5 M6 1513 101 5
= + , = + , = + ,
r2 4 12 r4 120 360 r6 840 210
√ √
M8 23983 1073 5 M10 149279 223 5
8
= + , = +
r 6300 900 r10 16632 72
• 12-gon (dodecagon)
√ √ √ √  √
 q
M1 1 
= √ 504 − 660 2 + 396 3 − 4 6 + 2 3 + 3 2+ 3
r 1080 2
√  √ √  √   √ 
q
 27  
+ 6 1 + 47 3 2− 3− 15 + 11 3 ln 3 − 4 27 − 3 ln 1 + 2
2
√  √   √   √ 
− 4 3 ln 2 + 3 + 2 666 + 397 3 ln 2 − 3
√  √ √  √
  q    q 
− 33 − 19 3 ln 2 + 2 + 3 + 3 899 + 523 3 ln 2 + 2 − 3 ,

M2 2 1 M4 163 16 M6 71 661
2
= + √ , 4
= + √ , 6
= + √ ,
r 3 2 3 r 180 15 3 r 42 280 3
M8 2357 424 M10 19099 521
8
= + √ , 10
= + √
r 630 75 3 r 2079 36 3

7 Numerical values
Numerical values for mean distances are given in Table 1; as an example, numerical values of
moments for P5,r may be found in Table 2. Using Mathematica, all of these values are obtained
in two different ways:
1) from the analytical solution in Theorem 3 together with Corollary 1,
2) with numerical integration of
Rd
Mm (Pn, r ) = 0 xm gn,r (x) dx , m = −1, 0, 1, 2, . . . ,

with the density function gn,r from Theorem 2.

The integration with NIntegrate required the use of WorkingPrecision and PrecisionGoal
with suitable parameters and the splitting of the interval [0, d] into subintervals corresponding
to the different cases.
Fig. 5 shows how the expected distances converge against the expected distance for the circle Cr
with radius r (see (22), (23)).

14
 Tab. 1: Numerical values for the mean distance

n M1 (Pn, r )/r
3 0.631838006782679248439363765946266548228352566630216491351433566313728372374
4 0.737378635076566348769571883395005078909753124947971245022700482462877666826
5 0.793698195033753381760971632749605393899757772205869281310580089360628333136
6 0.826258949490232082314283750323326010149318430219325083679749121623475379023
7 0.846561326216093164027700615562504932765674717512492654162687095709688638441
8 0.860007978015497247289475476698227488793382581154266797596077525063820243821
9 0.869349677996368661272424554005308059692018424046355884971707394951786624055
10 0.876093016045821455437851478188806272714728045761204545124286721555684616290
11 0.881115231029789401140263676940968133233965295741220353166084636303706353850
12 0.884953782114214064113523711258487868874291600958495392058472857817523925124
13 0.887952286709934145445459685691276130847157335822373829281308109648830663248
14 0.890338490756764575735944950351064496029659712504470764018294183941844198540
15 0.892268061893522241535511225799568091834470923360328845899768230281232859177
16 0.893850267566140981445305963668775169957199844033103927893285119318229687792
17 0.895163602871345666236825912146252831958826216118785135695365798856133878291
18 0.896265616099618345601553161509629568460084862108878882774989021025191298962
19 0.897199265437573395493137709163413742006864641423080106927798068769440815162
20 0.897997136941287061624499349289436760861985366382141463318416018681802183354
21 0.898684307244840900050480580043224097039202683773713372510772133748302247016
22 0.899280326175269024941354083121964386591039853255838396881868160285060306349
23 0.899800615121290023756007063614270208943463748969153209258516013892113626863
24 0.900257469758660200325240434610054253924497221077824646172921619542988501865
25 0.900660789967557175495408085191424058901073528518182128993867898837236409609
26 0.901018618516579218784735836535111473157040513720977689621453066648252571482
27 0.901337543658947219455759999490321561595657334895730647458502404432253617671
28 0.901623003533134180530119238163701029410880699933247882599362042456633796468
29 0.901879518798576670046389739873503980147299817776988523533576670581933050086
30 0.902110872199677767358604753115452862213333195981189704602574660701374264234
∞ 0.905414787367226799040760964963637259573814873545707797320063112868390670986

Tab. 2: Numerical moments for P5,r

m Mm (P5, r )/rm
−1 1.941532747349740237286829163767369291397446081205725687375578118779045940892
0 1.000000000000000000000000000000000000000000000000000000000000000000000000000
1 0.793698195033753381760971632749605393899757772205869281310580089360628333136
2 0.769672331458315808034097805727606352953384863300960477022574770450876743803
3 0.840599732769508183798403754882490669375942298135780251528211484773754115102
4 0.994399274340245648062512643833944980414538916051760874541387079159940696972
5 1.246445987635706663907926935585552909367630737102937503163072379765170177658
6 1.632924184047558960878811524399412257744938578936626397410732546307767883516
7 2.215335008390427051007858142854877211061180001285436481704877646702725059467
8 3.092181971180434547537024531611250214836863738665362023886032145176688627847
9 4.419361573049694692989749824140763338295447870581682927297936340850651262824
10 6.443751410804749342223892396954906260665664087034580725980990086289287076882

15
 0.90
0.85
0.80
0.75
0.70
0.65 Fig. 5: M1 (Pn, r )/r,
n = 3, 4, . . . , 25, and
5 10 15 20 n 25 M1 (Cr )/r

8 Second moment and variance

Theorem 4. The second moment of the point distance in Pn, r , r = 3, 4, . . ., is given by

r2
 
2π
M2 (Pn, r ) = 2 + cos .
3 n

Proof. Due to Czuber (1884, p. 217, Problem XXVII) we have

2Ip (Pn, r )
M2 (Pn, r ) = (16)
A(Pn, r )

where Ip (Pn, r ) is the polar moment of area of Pn, r defined by

Z
x2 + y 2 dA


Ip (Pn, r ) =
Pn, r

whereby it is again assumed that the centroid of Pn, r is located at the origin O of the x, y-
coordinate system, and the vertex X0 lies on the positive x-axis (see (1)). Now we rotate Pn, r
about O with angle −α so that the side X0 X1 is parallel to the y-axis. Using polar coordinates
ϕ and % we have %2 = x2 + y 2 and dA = % dϕ d%, we get
Z 2π Z R(ϕ)
Ip (Pn, r ) = %3 d% dϕ
ϕ=0 %=0

where R(ϕ) is the equation of ∂ Pn, r , and further

Z π/n Z R(ϕ) Z π/n
3 n
Ip (Pn, r ) = 2n dϕ % d% = R(ϕ)4 dϕ
ϕ=0 %=0 2 ϕ=0
π/n
nr4
Z

2
= cos4 (π/n) 1 + tan2 ϕ dϕ
2 ϕ=0
π/n
nr4
Z
dϕ
= cos4 (π/n) .
2 ϕ=0 cos4 ϕ

With Eq. 2.526.12 in Gradstein and Ryshik (1981, Vol. 1, p. 173) we get

nr4
 
sin(π/n) 2
Ip (Pn, r ) = cos4 (π/n) + tan(π/n)
2 3 cos3 (π/n) 3

16
 nr4 sin(π/n) + 2 sin(π/n) cos2 (π/n)
= cos4 (π/n)
6 cos3 (π/n)
nr4
cos(π/n) sin(π/n) 1 + 2 cos2 (π/n)
 
=
6
nr4
= cos(π/n) sin(π/n) [2 + cos(2π/n)] . (17)
6
Now from (16) with the area

A(Pn, r ) = nr2 cos(π/n) sin(π/n) (18)

(see also (4)) it follows that

r2
 
2π
M2 (Pn, r ) = 2 + cos .
3 n

Remark 5. We get
r2
lim M2 (Pn, r ) =(2 + 1) = r2
n→∞ 3
which is the result for the circle with radius r (see Table 3).
Corollary 2. The variance of the distance ∆n, r between two random points in a regular polygon
is given by
Var[∆n, r ] = E ∆2n, r − E[∆n, r ]2 = M2 (Pn, r ) − M1 (Pn, r )2
 

with M1 (Pn, r ) according to Corollary 1, and M2 (Pn, r ) according to Theorem 4.

Fig. 6 shows how the variances converge against the variance for the circle Cr with radius r (see
(25)).

0.18

0.16

0.14

0.12
Fig. 6: Var[∆n,r ]/r2 ,
n = 3, 4, . . . , 25, and
5 10 15 20 n 25 Var[∆r ]/r2

9 Fourth moment
Theorem 5. The fourth moment of the point distance in Pn, r , r = 3, 4, . . ., is given by

r4
 
2π 4π
M4 (Pn, r ) = 77 + 64 cos + 9 cos .
90 n n

Proof. From (14) together with the second formula in (15) we get

S7 (Pn, r )
M4 (Pn, r ) = .
21A(Pn, r )2

17
According to Voss (1982, Eq. (16)) we can compute S7 with
h
i
S7 (Pn, r ) = 42 A(Pn, r )Ip(4) (Pn, r ) + Ip (Pn, r )2 + 2 Ix (Pn, r )2 + Iy (Pn, r )2

where
Z Z 2π

2 1
Ip(4) (Pn, r ) = x2 + y 2 dA = R(ϕ)6 dϕ ,
Pn, r 6 0
Z Z
Ix (Pn, r ) = y 2 dA , Iy (Pn, r ) = x2 dA
Pn, r Pn, r

whereby the centroid of Pn, r is located at the origin O of the x, y-coordinate system and the
axes are the principal axes of intertia. (Ix und Iy are the second moments of area with respect
to the x-axis and the y-axis, respectively.) So we have
(4)
2Ip (Pn, r ) 2Ip (Pn, r )2 4 Ix (Pn, r )2 + Iy (Pn, r )2


M4 (Pn, r ) = + + . (19)
A(Pn, r ) A(Pn, r )2 A(Pn, r )2

We get
π/n π/n
nr6
Z Z
n
3
Ip(4) (Pn, r ) = R(ϕ) dϕ = 6
cos6 (π/n) 1 + tan2 ϕ dϕ
3 0 3 0
6 Z π/n
nr 6 dϕ
= cos (π/n) ,
3 0 cos6 ϕ
and, with Eq. 2.526.14 in Gradstein and Ryshik (1981, Vol. 1, p. 174),

nr6
 
(4) 6 sin(π/n) 4 3 4
Ip (Pn, r ) = cos (π/n) + tan (π/n) + tan(π/n)
3 5 cos5 (π/n) 15 5
nr6
cos(π/n) sin(π/n) 3 + 4 sin2 (π/n) cos2 (π/n) + 12 cos4 (π/n)
 
=
45
nr6
= cos(π/n) sin(π/n) [8 + 6 cos(2π/n) + cos(4π/n)] ,
45
and, with the area (18),
(4)
2Ip (Pn, r ) 2r4
= [8 + 6 cos(2π/n) + cos(4π/n)] . (20)
A(Pn, r ) 45

With (17) we get

1 4 cos(π/n) sin(π/n)
2
2Ip (Pn, r )2 6 nr (2 + cos(2π/n)) r4
=2 = [2 + cos(2π/n)]2
A(Pn, r )2 (nr2 cos(π/n) sin(π/n))2 18
r4
= [9 + 8 cos(2π/n) + cos(4π/n)] . (21)
36
Now we compute Ix (Pn, r ) and Iy (Pn, r ). Since the axes of the coordinate system are principal
axes of inertia and the origin is located in the centroid of Pn, r , the second area moments Ix (Pn, r )
and Iy (Pn, r ) take extreme values compared to all other second area moments with respect to
axes through the centroid (Göldner and Holzweißig, 1989, p. 165). This means that one of the
two values Ix (Pn, r ) and Iy (Pn, r ) is the maximum and one is the minimum value. An axis of
symmetry and the axis perpendicular to it are always principal axes of inertia (ibid., p. 165).
Each polygon Pn, r , n = 3, 4, . . ., has more than two symmetry axes that are not perpendicular

18
to each other. Therefore it follows that the second area moment is equal for all axes through the
centroid. Thus with Ix (Pn, r ) + Iy (Pn, r ) = Ip (Pn, r ) we have found

1
Ix (Pn, r ) = Iy (Pn, r ) = Ip (Pn, r ) ,
2
hence
4 Ix (Pn, r )2 + Iy (Pn, r )2


2Ip (Pn, r )2
= ,
A(Pn, r )2 A(Pn, r )2
and (19) becomes
(4)
2Ip (Pn, r ) 4Ip (Pn, r )2
M4 (Pn, r ) = +
A(Pn, r ) A(Pn, r )2
(cp. Eq. (17) in (Voss, 1982)). So, with (20) and (21) we finally get

2r4 r4
M4 (Pn, r ) = [8 + 6 cos(2π/n) + cos(4π/n)] + [9 + 8 cos(2π/n) + cos(4π/n)]
45 18
r4
 
2π 4π
= 77 + 64 cos + 9 cos .
90 n n

Remark 6. From Theorem 5 we get

r4 5
lim M4 (Pn, r ) = (77 + 64 + 9) = r4
n→∞ 90 3
which is the result for the circle with radius r (see Table 3).
Remark 7. With the side length `1 = 2r sin(π/n) and the area A(Pn, r ) (see (18)), the second
area moment Ix (Pn, r ) with respect to an axis through the centroid of Pn, r can be written in the
form
`2
 
A(Pn, r )
Ix (Pn, r ) = 3r2 − 1 .
12 2
This result can already be found in Hütte (Ed.) (1872, p. 249). The independence of Ix (Pn, r )
from the direction of the x-axis through the centroid is also discussed in anderstood and ccorn
(n.d.).

10 Circle
Now, we consider the moments Mm (Cr ), m = 1, 2, . . ., of the distance ∆r between two random
points in a circle Cr with radius r.
Czuber (1884) found
128
M1 (Cr ) = r (p. 197) , M2 (Cr ) = r2 (p. 217) . (22)
45π
A numerical approximation of M1 (Cr ) is given by

M1 (Cr ) ≈ 0.90541478736722679904 r . (23)

A general formula for Mm (Cr ) can be found in Mathai (1999, p. 207, Theorem 2.3.15). Here we
calculate the moments using (13), with Cr instead of Pn, r , and 2r instead of d, in the form
2r
2L(∂ Cr )
Z
Mm (Cr ) = E[∆m
r ] = xm+2 (1 − Fr (x)) dx
(m + 2)A(Cr )2 0

19
where Fr (x) is the chord length distribution function of Cr . One easily finds
r  x 2
Fr (x) = 1 − 1 −
2r
(see also Piefke (1978)), hence
Z 2r r
4πr m+2
 x 2
Mm (Cr ) = x 1 − dx .
(m + 2)π 2 r4 0 2r
The substitution  x 2
t=
2r
yields Z 1
2m+4 rm
Mm (Cr ) = t(m+1)/2 (1 − t)1/2 dt .
(m + 2)π 0
The last integral is the Euler beta function B((m + 3)/2, 3/2). Using the relationship to the
gamma function Γ ,
Γ (x)Γ (y)
B(x, y) = ,
Γ (x + y)
√
and Γ (1/2) = π one finds
Γ m+3


2m+4 rm
Mm (Cr ) = √ 2
. (24)
π (m + 2)(m + 4) Γ m 2 +2
This is Mathai’s result. Special values of (24) are to be found in Table 3 .

Tab. 3: Special values of Mm (Cr )

m 1 2 3 4 5 6 7 8
128 2048 3 5 4 16384 5 7 6 524288 7 42 8
Mm (Cr ) r r2 r r r r r r
45π 525π 3 2205π 2 31185π 5
Eq. (22) (22)

For the variance we have

" #
128 2 2

 2 2 2
Var[∆r ] = E ∆r − E[∆r ] = M2 (Cr ) − M1 (Cr ) = 1 − r
45π (25)
≈ 0.18022406281675948280 r2 .

11 Chord power integrals

From (14) with the second formula in (15) one gets
m(m − 1) m(m − 1)
Sm (Pn, r ) = Tm−3 (Pn, r ) = A(Pn, r )2 Mm−3 (Pn, r ) , m = 2, 3, 4, . . . ,
2 2
(see also Santaló (1976, p. 47)). Thus it is easy to compute chord power integrals Sm (Pn, r ) with
the results of the present paper. One obtains e.g. for the second chord power integral of P6,r
√
!
27r 4 16 8 22 ln 3 4 ln 2 + 3 1
S2 (P6,r ) = A(P6,r )2 M−1 (P6,r ) = − + √ + −
4 9 3 3 9 9 r
√
!
27r3 16 8 22 ln 3 4 ln 2 + 3
= − + √ + −
4 9 3 3 9 9
≈ 12.568534 r3
(cf. Heinrich (2009)).

20
References
Achtelstädter, Martin and Tobias M. Hübener (2013). PolygonV111.jar. Java application for the
simulation of distances between random points and chord lengths of random lines in regular
polygons. url: https://fing.htwk-leipzig.de/fakultaet/professuren/prof-baesel/.
Aharonyan, Narine G. and Victor K. Ohanyan (2016). “Moments of the distance between two
random points”. In: Modeling of Artifical Intelligence 10.2, pp. 64–70. url: http://ysu.am/
files/9-1509965954-.pdf.
anderstood and ccorn (n.d.). Invariance of the second moment of area of a regular polygon. Math-
ematics Stack Exchange. url: https://math.stackexchange.com/questions/2109659/
invariance-of-the-second-moment-of-area-of-a-regular-polygon.
Bailey, D. H., Jonathan M. Borwein, and Richard E. Crandall (2007). “Box integrals”. In: Jour-
nal of Computational and Applied Mathematics 206, pp. 196–208. url: https : / / www .
sciencedirect.com/science/article/pii/S0377042706004250.
Bäsel, Uwe (2014). “Random chords and point distances in regular polygons”. In: Acta Math.
Univ. Comenianae LXXXIII.1, pp. 1–18. issn: 0862-9544. url: http : / / www . iam . fmph .
uniba.sk/amuc/ojs/index.php/amuc/article/view/71.
Blaschke, Wilhelm (1955). Vorlesungen über Integralgeometrie. Berlin: VEB Deutscher Verlag
der Wissenschaften.
Bonnet, Gilles et al. (2020). “Sharp inequalities for the mean distance of random points in convex
bodies”. In: arXiv:2010.03351 [math.MG] 7 Oct 2020. url: https://arxiv.org/abs/2010.
03351.
Borwein, Jonathan M. and Richard E. Crandall (2013). “Closed forms: What they are and why
we care”. In: Notices of the AMS 60.1.
Burgstaller, Bernhard and Friedrich Pillichshammer (2009). “The average distance between two
points”. In: Bull. Aust. Math. Soc. 80.3, pp. 353–359. url: https://www.cambridge.org/
core/journals/bulletin- of- the- australian- mathematical- society/article/the-
average-distance-between-two-points/F182A617B5EC6DB5AD31042A4BDF83AE.
Czuber, Emanuel (1884). Geometrische Wahrscheinlichkeiten und Mittelwerte. Mit 115 in den
Text gedruckten Figuren. Leipzig: Verlag von B. G. Teubner. url: https://books.google.
de/books/about/Geometrische_Wahrscheinlichkeiten_und_Mi.html?id=2jQAAAAAQAAJ&
redir_esc=y.
Duma, Andrei and Marius I. Stoka (2008). “Schnitte eines „kleinen“ gleichseitigen Dreiecks mit
den Gittern von Buffon und Laplace”. In: Seminarberichte aus der Fakultät für Mathematik
und Informatik der FernUniversität in Hagen 79, pp. 13–22.
Dunbar, Steven R. (1997). “The average distance between points in geometric figures”. In: The
College Mathematics Journal 28.3, pp. 187–197.
Finch, Steven R. (2003). Mathematical constants. Cambridge: Cambridge University Press.
Ghosh, Birendranath (1951). “Random distances within a rectangle and between two rectangles”.
In: Bull. Calcutta Mat. Soc. 43, pp. 17–24.
Göldner, Hans and Franz Holzweißig (1989). Leitfaden der technischen Mechanik. 11., verbesserte
Aufl. Leipzig: VEB Fachbuchverlag Leipzig.
Gradstein, Israil Solomonowitsch and Iossif Moissejewitsch Ryshik (1981). Tables of Series, Prod-
Izrail~ Solomonoviq Gradxten,
ucts, and Integrals (2 Volumes). 1st ed. Original edition:
Iosif Moiseeviq Ryik: Tablicy integralov, summ, rdov i proizvedeni, Izda-
tel~stvo ”Nauka“ Moskva 1971. Mir, Moscow, Licence edition for Verlag Harri Deutsch,
Thun, Frankfurt am Main.
Hammersley, John M. (1950). “The distribution of the distance in a hypersphere”. In: Ann. Math.
Statist. 21, pp. 447–452. url: https://projecteuclid.org/euclid.aoms/1177729805.
Heinrich, Lothar (2009). “On lower bounds for second-order chord power integrals of convex
discs”. In: Rend. Circ. Mat. Palermo, Serie II Suppl. 81, pp. 213–222.

21
Hütte (Ed.) (1872). Des Ingenieurs Taschenbuch. Herausgegeben von dem Verein „Hütte“. 9th ed.
Berlin: Verlag von Wilhelm Ernst & Korn. url: https : / / books . google . de / books ?
id = L7lAAQAAIAAJ & printsec = frontcover & dq = h % C3 % BCtte + 1872 + des + ingenieurs +
Taschenbuch&hl=de&sa=X&ved=2ahUKEwjeh5qy3YvtAhWBjKQKHQUgAIQQ6AEwAHoECAAQAg#
v=onepage&q=h%C3%BCtte%201872%20des%20ingenieurs%20Taschenbuch&f=false.
Li, Hongjun and Xing Qiu (2016). “Moments of distance from a vertex to a uniformly distributed
random point within arbitrary triangles”. In: Mathematical Problems in Engineering 2016,
pp. 1–10. url: https://www.hindawi.com/journals/mpe/2016/8371750/.
Mathai, A. M. (1999). An Introduction to Geometrical Probability. Distributional Aspects with
Applications. Amsterdam: Gordon and Breach.
OEIS A091505 (2004). Sequence A091505 in The On-Line Encyclopedia of Integer Sequences.
url: https://oeis.org/A091505.
OEIS A093064 (2004). Sequence A093064 in The On-Line Encyclopedia of Integer Sequences.
url: https://oeis.org/A093064.
Philip, Johan (2007). The probability distribution of the distance between two random points in
a box. url: https://people.kth.se/~johanph/habc.pdf.
— (n.d.). The distance between two random points in a 4- and 5-cube. url: https://people.
kth.se/~johanph/h45.pdf.
Piefke, Frank (1978). “Beziehungen zwischen der Sehnenlängenverteilung und der Verteilung
des Abstandes zweier zufälliger Punkte im Eikörper”. In: Z. Wahrscheinlichkeitstheorie verw.
Gebiete 43, pp. 129–134.
Santaló, Luis A. (1976). Integral Geometry and Geometric Probability. London: Addison-Wesley.
Schneider, Rolf and Wolfgang Weil (2008). Stochastic and Integral Geometry. (in der Reihe:
Probability and its Applications). Berlin, Heidelberg: Springer-Verlag.
Sulanke, Rolf (1961). “Die Verteilung der Sehnenlängen an ebenen und räumlichen Figuren”. In:
Math. Nachr. 23, pp. 51–74.
Voss, Klaus (1982). “Powers of chords of convex sets”. In: Biom. J. 24.5, pp. 513–516.
Weisstein, Eric W. (n.d.[a]). Square Line Picking. From MathWorld – A Wolfram Web Resource.
url: http://mathworld.wolfram.com/SquareLinePicking.html.
— (n.d.[b]). Triangle Line Picking. From MathWorld – A Wolfram Web Resource. url: http:
//mathworld.wolfram.com/TriangleLinePicking.html.
Zhuang, Yanyan and Jianping Pan (2011). “Random distances associated with hexagons”. In:
arXiv:1106.2200v1 [math.PR] 11 Jun 2011. url: https://arxiv.org/abs/1106.2200v1.
— (2012). “Random distances associated with equilateral triangles”. In: arXiv:1207.1511v1 [math.PR]
6 Jul 2012. url: https://arxiv.org/abs/1207.1511v1.

Uwe Bäsel, Leipzig University of Applied Sciences (HTWK Leipzig), Faculty of Engineering,
PF 30 11 66, 04251 Leipzig, Germany, uwe.baesel@htwk-leipzig.de

22