SHARP BOUNDS FOR GENERALIZED ELLIPTIC INTEGRALS

OF THE FIRST KIND

arXiv:1502.02225v1 [math.CA] 8 Feb 2015

MIAO-KUN WANG, YU-MING CHU, AND SONG-LIANG QIU

Abstract. In this paper, we prove that the double inequality

Ka (r)
1 + αr ′2 < < 1 + βr ′2
sin(πa) log(eR(a)/2 /r ′ )
holds for all a ∈ (0, 1/2] and √
r ∈ (0, 1) if and only if α ≤ π/[R(a) sin(πa)] − 1
and β ≥ a(1 − a), where r ′ = 1 − r 2 , Ka (r) is the generalized elliptic integral
of the first kind and R(x) is the Ramanujan constant function. Besides, as
the key tool, the series expression for the Ramanujan constant function R(x)
is given.

1. Introduction
For r ∈ (0, 1), Legendre’s complete elliptic integrals [1] of the first kind and the
second kind are given by
Z π/2
dt
K = K(r) = q ,
0 1 − r2 sin2 (t)
Z π/2 q
E = E(r) = 1 − r2 sin2 (t)dt,
0
respectively. They are the particular cases of Gaussian hypergeometric function
∞
X (a)n (b)n xn
F (a, b; c; x) = (−1 < x < 1),
n=0
(c)n n!
R∞
where (a)n = Γ(a + n)/Γ(a) and Γ(x) = 0 tx−1 e−t dt (x > 0) is the gamma
function. Indeed, we have
   
π 1 1 2 π 1 1 2
K(r) = F , ; 1; r , E(r) = F − , ; 1; r .
2 2 2 2 2 2
It is well known that the complete elliptic integrals and Gaussian hypergeometric
function have important applications in quasiconformal mappings, number theory,
and other fields of the mathematical and mathematical physics. For instance, the
Gaussian arithmetic-geometric mean AGM and the modulus of the plane Grötzsch
ring can be expressed in terms of the complete elliptic integrals of the first kind,
and the complete elliptic integrals of the second kind gives the formula of the
perimeter of an ellipse. Moreover, Ramanujan modular equation and continued

2010 Mathematics Subject Classification. 33E05, 33C05.

Key words and phrases. Gaussian hypergeometric function, generalized elliptic integral, Ra-
manujan constant function.
This work was supported by Natural Science Foundation of China (Grant Nos. 61374086,
11171307) and Natural Science Foundation of the Zhejiang Province (Grant No. LY13A010004).
1
2 MIAO-KUN WANG, YU-MING CHU, AND SONG-LIANG QIU

fraction in number theory are both related to the Gaussian hypergeometric function
F (a, b; c; x). For these, and the properties for the complete elliptic integrals and
Gaussian hypergeometric function see [2, 3, 5, 6, 9, 11, 16, 19-21].
For r ∈ (0, 1) and a ∈ (0, 1), the generalized elliptic integrals (see [4, 10]) are
defined by
π
Ka = Ka (r) = F (a, 1 − a; 1; r2 ) (1.1)
2
and
π
Ea = Ea (r) = F (a − 1, 1 − a; 1; r2 ). (1.2)
2
Clearly Ka (0) = Ea (0) = π/2, Ka (1− ) = ∞ and Ea (1) = [sin(πa)]/[2(1 − a)]. In the
particular case a = 1/2, the generalized elliptic integrals reduce to the complete
elliptic integrals. By symmetry of (1.1), we assume that a ∈ (0, 1/2] in the sequence.
The generalized elliptic integrals satisfy the following derivative formulas:
dKa 2(1 − a) 2 dEa 2(1 − a)
= 2 (Ea − r′ Ka ), =− (Ka − Ea ), (1.3)
dr rr ′ dr r
d 2(1 − a)rEa d 2
(Ka − Ea ) = 2 , (Ea − r′ Ka ) = 2arKa . (1.4)
dr r √
′ dr
Here and in what follows we set r′ = 1 − r2 for r ∈ (0, 1).
In 2000, Anderson, Qiu, Vamanamurthy and Vuorinen [4] reintroduced the gen-
eralized elliptic integrals in geometry function theory, found that the generalized
elliptic integrals of the first kind Ka arises from the Schwarz-Christoffel trans-
formation of the upper half-plane onto a parallelogram, and established several
monotonicity theorems for the generalized elliptic integrals Ka and Ea .
Recently, the generalized elliptic integrals have attracted the attention of many
mathematicians. In particular, many remarkable properties and inequalities for the
generalized elliptic integrals can be found in the literature [8, 12-14, 22].
Very recently, Takeuchi [18] discussed the generalized trigonometric function and
found a new form of the generalized complete elliptic integrals.
In [16], Qiu and Vamanamurthy proved that the inequality
K(r) 1
< 1 + r′2 (1.5)
log(4/r′ ) 4
holds for all r ∈ (0, 1).
Alzer [3] proved that the inequality
 
π K(r)
1+ − 1 r′2 < (1.6)
4 log 2 log(4/r′ )
holds for all r ∈ (0, 1). Moreover, Alzer also proved that the constant factors 1/4
in (1.5) and π/(4 log 2) − 1 in (1.6) are best possible.
The main purpose of this paper is to generalize inequalities (1.5) and (1.6) to
Ka . Our main result is the following Theorem 1.1.
Theorem 1.1. Let R(x) be the Ramanujan constant function defined by (2.1).
Then the double inequality
Ka (r)
1 + αr′2 < < 1 + βr′2 (1.7)
sin(πa) log(eR(a)/2 /r′ )
holds for all a ∈ (0, 1/2] and r ∈ (0, 1) if and only if α ≤ α0 = π/[R(a) sin(πa)] − 1
and β ≥ β0 = a(1 − a).
 SHARP BOUNDS FOR GENERALIZED ELLIPTIC INTEGRALS OF THE FIRST KIND 3

2. Some properties for Ramanujan constant function R(x)

For x ∈ (0, 1/2], the Ramanujan constant function R(x) ([17]) is given by
R(x) = −2γ − Ψ(x) − Ψ(1 − x), R (1/2) = 4 log 2, (2.1)
Pn
where γ = lim ( k=1 1/k−log n) = 0.577215 · · · is the Euler-Mascheroni constant,
n→∞
and
Z∞
Γ′ (x)
Ψ(x) = , Γ(x) = tx−1 e−t dt, x > 0.
Γ(x)
0
It is well known that Ψ(n) (x) (n ≥ 0) has the series expansion as follows:
 ∞
 1 X x

 −γ − + , n=0
x k(k + x)


(n) k=1
Ψ (x) = ∞ (2.2)
 n+1
X 1
 (−1) n! , n ≥ 1.


(x + k)n+1

k=0

The purpose of this section is to present the series expansion formula for R(x)(x ∈
(0, 1/2]) (Theorem 2.2) and establish two important inequalities involving R(x)
(Corollaries 2.4 and 2.5), which will be used in the proof of our main result.
Lemma 2.1. The function ξ(x) = 1/[x(1 − x)] − R(x) is strictly increasing from
(0, 1/2] onto (1, 4 − 4 log 2].
Proof. Differentiating ξ yields
1 1
ξ ′ (x) = − 2 + + Ψ′ (x) − Ψ′ (1 − x)
x (1 − x)2
∞ ∞
1 1 X 1 X 1
=− 2 + + −
x (1 − x)2 (k + x)2 (k + 1 − x)2
k=0 k=0
∞ ∞
X 1 X 1
= 2
− >0
(k + x) (k + 1 − x)2
k=1 k=1
for x ∈ (0, 1/2]. Moreover, ξ(1/2) = 4 − R(1/2) = 4 − 4 log 2 and
 
1 1
lim ξ(x) = lim + + Ψ(x) + Ψ(1 − x) + 2γ
x→0− x→0− x 1−x
 ∞
1 1 1 X x 1
= lim + −γ− + −γ−
x→0− x 1−x x k(k + x) 1−x
k=1
∞ 
X 1−x
+ + 2γ
k(k + 1 − x)
k=1
∞ ∞
! ∞
X x X 1−x X 1
= lim + = = 1.
x→0− k(k + x) k(k + 1 − x) k(k + 1)
k=1 k=1 k=1

Theorem 2.2. The Ramanujan constant function R(x) has the following series
expansion:
1
R(x) = + 2ζ(3)x2 + 2ζ(5)x4 + · · · + 2ζ(2k + 1)x2k + · · ·
x
4 MIAO-KUN WANG, YU-MING CHU, AND SONG-LIANG QIU

∞
1 X
= + 2ζ(2k + 1)x2k , (2.3)
x
k=1
P∞
where ζ(x) = n=1 n−x is the well-known Riemann zeta function.
Proof. Let f (x) = xR(x). Then simple computations lead to
∞ ∞
!
1 1 X x X 1−x
lim f (x) = lim x + − − = 1, (2.4)
x→0 x→0 x 1−x k(k + x) k(k + 1 − x)
k=1 k=1

f ′ (x) =R(x) + xR′ (x) = −Ψ(x) − Ψ(1 − x) − 2γ + x [−Ψ′ (x) + Ψ′ (1 − x)]

∞ ∞
1 1 X x X 1−x
= + − −
x 1−x k(k + x) k(k + 1 − x)
k=1 k=1
" ∞ ∞
#
X 1 X 1
+x − +
(k + x)2 (k + 1 − x)2
k=0 k=0
∞ 2 ∞ ∞
1 X x X 1−x X x
= − − + ,
1−x k(k + x)2 k(k + 1 − x) (k + 1 − x)2
k=1 k=1 k=0
′
lim f (x) = 0. (2.5)
x→0
For n ≥ 2, it follows from (2.2) that
R(n) (x) = − Ψ(n) (x) − (−1)n Ψ(n) (1 − x)
∞ ∞
X 1 X 1
= − n!(−1)n+1 n+1
− (−1) n
n!(−1)n+1
(x + k) (1 − x + k)n+1
k=0 k=0
"∞ #
X 1 1
=n! + (−1)n .
(1 − x + k)n+1 (x + k)n+1
k=0
Therefore, we get
f (n) (x) =nR(n−1) (x) + xR(n) (x)
∞  
X 1 n−1 1
=n! + (−1)
(1 − x + k)n (x + k)n
k=0
∞  
X x n x
+ n! + (−1)
(1 − x + k)n+1 (x + k)n+1
k=0
∞ 
(−1)n−1 k

X 1+k
=n! + .
(1 − x + k)n+1 (x + k)n+1
k=0
Furthermore, if n is even, then
∞  
(n)
X 1+k k
f (x) =n! −
(1 − x + k)n+1 (x + k)n+1
k=0
∞  
X 1+k 1+k
=n! − ,
(1 − x + k)n+1 (x + 1 + k)n+1
k=0

lim f (n) (x) = 0. (2.6)

x→0
 SHARP BOUNDS FOR GENERALIZED ELLIPTIC INTEGRALS OF THE FIRST KIND 5

If n is odd, then
∞  
X 1+k k
f (n) (x) =n! n+1
+
(1 − x + k) (x + k)n+1
k=0
∞  
X 1+k 1+k
=n! + ,
(1 − x + k)n+1 (x + 1 + k)n+1
k=0
∞
X 1
lim f (n) (x) = 2n! = 2n!ζ(n). (2.7)
x→0 (k + 1)n
k=0
Equations (2.4)-(2.7) implies that f (x) has the following Taylor series expansion
∞
X
3 5
f (x) = 1 + 2ζ(3)x + 2ζ(5)x + · · · = 1 + 2ζ(2k + 1)x2k+1 , x ∈ (0, 1).
k=1

Therefore, (2.3) follows. 

Theorem 2.3. The function
π/ sin(πx) − R(x)
η(x) =
x(1 − x)
is strictly decreasing from (0, 1/2] onto [4π − 16 log 2, π 2 /6).
Proof. Clearly (2.1) gives η(1/2) = 4π − 16 log 2. Simple computations lead to
 2   
π cos(πx) π
x2 (1 − x)2 η ′ (x) = − + R ′
(x) x(1 − x) − − R(x) (1 − 2x),
sin2 (πx) sin(πx)
 
1
η′ = 0. (2.8)
2
Making use of the well-known formulas (see [1, p.75, 4.3.68] and [7, p.16 and
p.56])
∞
1 1 X 22k − 2
= + |B2k |x2k−1 , |x| < π,
sin x x (2k)!
k=1
ζ(2k)(2k)!
B2k = 2(−1)k+1 ,
(2π)2k
where Bk is the Bernouli number, we can rewrite η(x) as
∞ ∞
X 22k − 2 X
ζ(2k)x2k − 2ζ(2k + 1)x2k+1
πx/ sin(πx) − xR(x) 22k−1
η(x) = = k=1 k=1
x2 (1 − x) x2 (1 − x)
∞ ∞
X 22k+2 − 2 2k
X
ζ(2k + 2)x − 2ζ(2k + 3)x2k+1
22k+1
= k=0 k=0
.
(1 − x)
Therefore, η(0+ ) = ζ(2) = π 2 /6 and
(1 − x)2 η ′ (x)
"∞ ∞
#
X 22k+2 − 2 X
= (2k)ζ(2k + 2)x2k−1 − 2(2k + 1)ζ(2k + 3)x2k (1 − x)
22k+1
k=1 k=0
6 MIAO-KUN WANG, YU-MING CHU, AND SONG-LIANG QIU

∞ ∞
X 22k+2 − 2 2k
X
+ ζ(2k + 2)x − 2ζ(2k + 3)x2k+1
22k+1
k=0 k=0
∞ ∞
X 22k+4 − 2 X
= (2k + 2)ζ(2k + 4)x2k+1 − 2(2k + 1)ζ(2k + 3)x2k
22k+3
k=0 k=0
∞ 2k+4 ∞
X 2 −2 X
− (2k + 2)ζ(2k + 4)x2k+2 + 2(2k + 1)ζ(2k + 3)x2k+1
22k+3
k=0 k=0
∞ ∞
X 22k+2 − 2 2k
X
+ ζ(2k + 2)x − 2ζ(2k + 3)x2k+1
22k+1
k=0 k=0
∞ ∞
X 22k+2 − 2 X
= ζ(2k + 2)x2k + 2(2k + 1)ζ(2k + 3)x2k+1
22k+1
k=0 k=0
∞ 2k+4 ∞
X 2 −2 X
+ (2k + 2)ζ(2k + 4)x2k+1 − 2(2k + 1)ζ(2k + 3)x2k
22k+3
k=0 k=0
∞ ∞
X X 22k+4 − 2
− 2ζ(2k + 3)x2k+1 − (2k + 2)ζ(2k + 4)x2k+2
22k+3
k=0 k=0
= : g(x). (2.9)

Differentiating g yields

∞ ∞
X 22k+2 − 2 2k−1
X
g ′ (x) = (2k)ζ(2k + 2)x + 2(2k + 1)2 ζ(2k + 3)x2k
22k+1
k=1 k=0
∞ 2k+4
X 2 −2
+ (2k + 1)(2k + 2)ζ(2k + 4)x2k
22k+3
k=0
X∞ ∞
X
2k−1
− 2(2k)(2k + 1)ζ(2k + 3)x − 2(2k + 1)ζ(2k + 3)x2k
k=1 k=0
∞
X 22k+4 − 2
− (2k + 2)2 ζ(2k + 4)x2k+1
22k+3
k=0
∞ ∞
X 22k+4 − 2 2k+1
X 22k+4 − 2
= 2k+3
(2k + 2)ζ(2k + 4)x − (2k + 2)2 ζ(2k + 4)x2k+1
2 22k+3
k=0 k=0
X∞ ∞
X
+ 2(2k + 1)2 ζ(2k + 3)x2k − 2(2k + 1)ζ(2k + 3)x2k
k=0 k=0
∞
X 22k+4 − 2
+ (2k + 1)(2k + 2)ζ(2k + 4)x2k
22k+3
k=0
X∞
− 2(2k + 2)(2k + 3)ζ(2k + 5)x2k+1
k=0
∞
X 22k+4 − 2
=− (2k + 1)(2k + 2)ζ(2k + 4)x2k+1
22k+3
k=0
 SHARP BOUNDS FOR GENERALIZED ELLIPTIC INTEGRALS OF THE FIRST KIND 7

∞
X
+ 2(2k + 2)(2k + 3)ζ(2k + 5)x2k+2
k=0
∞
X 22k+4 − 2
+ (2k + 1)(2k + 2)ζ(2k + 4)x2k
22k+3
k=0
∞
X
− 2(2k + 2)(2k + 3)ζ(2k + 5)x2k+1
k=0
=(1 − x)g1 (x), (2.10)
where
∞
X 22k+4 − 2
g1 (x) = (2k + 1)(2k + 2)ζ(2k + 4)x2k
22k+3
k=0
X∞
− 2(2k + 2)(2k + 3)ζ(2k + 5)x2k+1 , x ∈ (0, 1/2]. (2.11)
k=0

Note that
X 22k+4 − 2 ∞
7 93
g1 (x) = ζ(4) + ζ(6)x2 + (2k + 1)(2k + 2)ζ(2k + 4)x2k
2 4 22k+3
k=2
∞
X
− 12ζ(5)x − 40ζ(7)x3 − 2(2k + 2)(2k + 3)ζ(2k + 5)x2k+1
k=2
∞
127 X
>3.788 − 12.444x + 23.653x2 − 40.334x3 + (2k + 1)(2k + 2)x2k
64
k=2
∞
X
− 2ζ(9) (2k + 2)(2k + 3)x2k+1
k=2
12x8 − 34x6 + 30x4
 
>3.788 − 12.444x + 23.653x2 − 40.334x3 + 1.984
(1 − x2 )3
20x9 − 54x7 + 42x5
 
− 2.005
(1 − x2 )3
1
3788 − 12444x + 12289x2 − 3002x3 − 75x4 − 540x5

= 2 3
1000(1 − x )
− 285x6 − 288x7 + 155x8 + 234x9 .

(2.12)
It is not difficult to verify that the polynomial function x → 3788 − 12444x +
12289x2 − 3002x3 − 75x4 − 540x5 − 285x6 − 288x7 + 155x8 + 234x9 is strictly
decreasing and positive on (0, 1/2]. Therefore, g(x) is strictly increasing on (0, 1/2],
g(x) < g(1/2) = η ′ (1/2)/4 = 0 for x ∈ (0, 1/2] and η(x) is strictly decreasing on
(0, 1/2] follows from (2.8) and (2.9) together with (2.10)-(2.12). 

Corollary 2.4. The inequality

π sin(πx) − πx(1 − x)
−1> (2.13)
R(x) sin(πx) sin(πx) [R(x) − 1]
holds for all x ∈ (0, 1/2].
8 MIAO-KUN WANG, YU-MING CHU, AND SONG-LIANG QIU

Proof. We clearly see that inequality (2.13) is equivalent to

π[1 + x(1 − x)]R(x) − π − sin(πx)R(x)2
>0
x(1 − x)
or
π/ sin(πx) − R(x) 1 − x(1 − x)R(x)
R(x) sin(πx) −π > 0.
x(1 − x) x(1 − x)
It follows from Lemma 2.1 and Theorem 2.3 together with the fact that x →
R(x) sin(πx) is strictly decreasing from (0, 1/2] onto [4 log 2, π) (see [15, Theorem
2]) that
π/ sin(πx) − R(x) 1 − x(1 − x)R(x)
R(x) sin(πx) −π
x(1 − x) x(1 − x)
>4 log 2(4π − 16 log 2) − π(4 − 4 log 2) = (20 log 2 − 4)π − 64 log2 2 = 0.236 · · ·
for x ∈ (0, 1/2]. 

Corollary 2.5. The inequality

π − R(x) sin(πx)
x(1 − x) > (2.14)
R(x) sin(πx)
holds for all x ∈ (0, 1/2].
Proof. It follows from Theorem 2.3 that
 
π − R(x) sin(πx) π/ sin(πx) − R(x)
R(x) sin(πx) − > log 16 − sin(πx)
x(1 − x) x(1 − x)
2 2
π π
> log 16 − sin(πx) > log 16 − = 1.127 · · · > 0 (2.15)
6 6
for x ∈ (0, 1/2]. Therefore, inequality (2.14) follows easily from (2.15). 

3. Proof of Theorem 1.1

Lemma 3.1. (see [5, Theorem 1.25]) For −∞ < a < b < ∞, let f, g : [a, b] → R
be continuous on [a, b], and be differentiable on (a, b), let g ′ (x) 6= 0 on (a, b). If
f ′ (x)/g ′ (x) is increasing (decreasing) on (a, b), then so are
f (x) − f (a) f (x) − f (b)
and .
g(x) − g(a) g(x) − g(b)
If f ′ (x)/g ′ (x) is strictly monotone, then the monotonicity in the conclusion is also
strict.
Lemma 3.2. The inequality
πx(1 − x)
sin(πx) > [2 + x(1 − x)] (3.1)
2
holds for all x ∈ (0, 1/2].
 SHARP BOUNDS FOR GENERALIZED ELLIPTIC INTEGRALS OF THE FIRST KIND 9

Proof. Let
πx(1 − x)
h(x) = sin(πx) − [2 + x(1 − x)] . (3.2)
2
Then simple computations lead to
h(0+ ) = 0, (3.3)
2
′
 
h (x) = π cos(πx) − (1 − 2x)(1 + x − x ) ,
 
1
h′ (0+ ) = 0, h′ = 0, (3.4)
2
h′′ (x) = π −π sin(πx) + 1 + 6x − 6x2 ) ,
 
   
′′ + ′′ 1 5
h (0 ) = π, h =π − π < 0, (3.5)
2 2
h′′′ (x) = π 6 − 12x − π 2 cos(πx) ,
 
 
1
h′′′ (0+ ) = 6 − π 2 < 0, h′′′ = 0, (3.6)
2
h(4) (x) = π −12 + π 3 sin(πx) ,
 
(3.7)
 
1
h(4) (0+ ) = −12π, h(4) = π π 3 − 12 > 0.


(3.8)
2
From (3.7) and (3.8) we clearly see that there exists x0 ∈ (0, 1/2] such that
h(4) (x) < 0 for x ∈ (0, x0 ) and h(4) (x) > 0 for x ∈ (x0 , 1/2]. Thus h′′′ (x) is strictly
decreasing on (0, x0 ] and strictly increasing on [x0 , 1/2].
Equation (3.6) and the piecewise monotonicity of h′′′ (x) implies that h′′′ (x) < 0
for x ∈ (0, 1/2]. Hence h′′ (x) is strictly decreasing on (0, 1/2]. Then (3.5) leads to
the conclusion that there exists x1 ∈ (0, 1/2] such that h′ (x) is strictly increasing
on (0, x1 ] and strictly decreasing on [x1 , 1/2].
It follows from (3.3) and (3.4) together with the piecewise monotonicity of h′ (x)
that h(x) is strictly increasing on (0, 1/2] and h(x) > h(0+ ) = 0 for x ∈ (0, 1/2].
Therefore, inequality (3.1) follows easily from (3.2) and h(x) > 0 for x ∈ (0, 1/2].

Lemma 3.3. For a ∈ (0, 1/2], defined the function F on (0, 1) by
2 2
r4 sin(πa) − 2a(1 − a)r2 r′2 Ka + 2(1 − a)(r′ − r2 )(Ea − r′ Ka )
F (r) = .
r′ 2 r4
Then F (r) is strictly increasing from (0, 1) onto (sin(πa) − πa(1 − a) − (π/2)a2(1 −
a)2 , a(1 − a) sin(πa)).
Proof. Let F1 (r) = sin(πa)−2a(1−a)r′2 Ka /r2 +2(1−a)(r′2 /r2 −1)(Ea −r′2 Ka )/r2 ,
F2 (r) = r′2 . Then F (r) = F1 (r)/F2 (r), F1 (1) = F2 (1) = 0,
 
2 r′2 2(1 − a) ′2
F1′ (r) = − 2a(1 − a) − 3 Ka + 2 (E a − r K a )
r r rr′ 2
2ar2 Ka − 2(Ea − r′2 Ka )
   
2 Ea − r′2 Ka r′2
+ 2(1 − a) − 3 2
+ 2(1 − a) 2
− 1
r r r r3
′ 2 2
Ka 2 Ea − r Ka Ea − r′ Ka
=4a(1 − a) − 4a(1 − a) − 4(1 − a)
r3 r3 r5
2 ′2
ar Ka − (Ea − r Ka )
+ 4(1 − a)(r′2 − r2 ) ,
r5
10 MIAO-KUN WANG, YU-MING CHU, AND SONG-LIANG QIU

F2′ (r) = −2r,

−ar2 Ka + a(1 − a)r2 + 1 (Ea − r′ 2 Ka )

  
F1′ (r)
=2(1 − a)
F2′ (r) r6
(r′2 − r2 ) ar2 Ka − (Ea − r′2 Ka )
 
−
r6
F3 (r)
=2(1 − a) 4 , (3.9)
r
where
 Ea − r′2 Ka
F3 (r) = 2 − 2r2 + a(1 − a)r2

− 2ar′2 Ka .
r2
Making use of series expansion, we get
∞ ∞
2 X a(a)n (1 − a)n 2n X (a)n (1 − a)n 2n
F3 (r) = 2 + [a(1 − a) − 2]r2 r − 2a(1 − r2 )

r
π n=0
(n + 1)!n! n=0
(n!)2
 X ∞ ∞
(a)n (1 − a)n 2n X (a)n (1 − a)n 2n+2
=a 2 r + [a(1 − a) − 2] r
n=0
(n + 1)!n! n=0
(n + 1)!n!
∞ ∞ 
X (a)n (1 − a)n 2n X (a)n (1 − a)n 2n+2
−2 r + 2 r
n=0
(n!)2 n=0
(n!)2
 X ∞ ∞
(a)n+1 (1 − a)n+1 2n+2 X (a)n (1 − a)n 2n+2
=a 2 r + [a(1 − a) − 2] r
n=0
(n + 2)!(n + 1)! n=0
(n + 1)!n!
∞ ∞ 
X (a)n+1 (1 − a)n+1 2n+2 X (a)n (1 − a)n 2n+2
−2 r +2 r
n=0
[(n + 1)!]2 n=0
(n!)2
∞
X (a)n+1 (1 − a)n+1 2n+4
=a [2 + a(1 − a)] r . (3.10)
n=0
(n + 3)!n!
From (3.9) and (3.10) one has
∞
F1′ (r) X (a)n+1 (1 − a)n+1 2n
= πa(1 − a) [2 + a(1 − a)] r . (3.11)
F2′ (r) n=0
(n + 3)!n!
Therefore, the monotonicity of F (r) follows from Lemma 3.1 and (3.11). More-
over, by l’Hôptial’s rule we get
lim F (r) = a(1 − a) sin(πa),
r→1−

2(1 − a)(−r2 )(Ea − r′2 Ka )

lim F (r) = sin(πa) + lim
r→0+ r→0+ r4
2(1 − a)r (Ea − r Ka ) − 2a(1 − a)r2 r′2 Ka
′2 ′2
+ lim
r→0+ r4
πa (Ea − r′2 Ka ) − ar2 Ka
= sin(πa) − 2(1 − a) + lim+ 2(1 − a)
2 r→0 r4
−2(1 − a)ar2 (Ea − r′2 Ka )/(rr′2 )
= sin(πa) − πa(1 − a) + 2(1 − a) lim
r→0+ 4r3
2 2
= sin(πa) − πa(1 − a) − (π/2)a (1 − a) . 
 SHARP BOUNDS FOR GENERALIZED ELLIPTIC INTEGRALS OF THE FIRST KIND 11

Lemma 3.4. Let a ∈ (0, 1/2], λ1 = [sin(πa) − a(1 − a)π]/{sin(πa)[R(a) − 1]},

λ2 = 1 − [πa(1 − a)]/ sin(πa) − [πa2 (1 − a)2 ]/[2 sin(πa)], β = a(1 − a), and the
function Gλ (·) be defined on (0, 1) by
 R(a)/2  2
1 e r2 sin(πa) − 2(1 − a)(Ea − r′ Ka )
Gλ (r) = − log + .
2 r′ 2λr2 r′ 2 sin(πa)
Then the following statements are true:
(1) G′λ (r) > 0 for all r ∈ (0, 1) if 0 < λ ≤ λ2 ;
(2) G′λ (r) < 0 for all r ∈ (0, 1) if λ ≥ β;
(3) There exists r0 ∈ (0, 1) such that G′λ (r) < 0 for r ∈ (0, r0 ) and G′λ (r) > 0
for r ∈ (r0 , 1) if λ2 < λ < β.
Moreover,
 Gλ (0+ ) > 0,

0 < λ < λ1 ,
Gλ (0+ ) = 0, λ = λ1 , (3.12)
Gλ (0+ ) < 0, λ > λ1

and
−
 Gλ (1 ) = +∞, 0 < λ < β,


− 1
Gλ (1 ) = 1 − < 0, λ = β, (3.13)

 2a(1 − a)
Gλ (1− ) = −∞, λ > β.
Proof. It is apparent from Lemma 3.2 that λ2 > 0 for all a ∈ (0, 1/2]. Therefore,
parts (1)-(3) follows from Lemma 3.3 and the fact that
r 1
G′λ (r) = − ′2 +
r 2λ sin(πa)
[2r sin(πa) − 4a(1 − a)rKa ]r2 r′2 − [r2 sin(πa) − 2(1 − a)(Ea − r′2 Ka )](2rr′2 − 2r3 )
×
r4 r′4
r sin(πa) − 2a(1 − a)r2 r′2 Ka + 2(1 − a)(r′2 − r2 )(Ea − r′2 Ka )
4
 
r 1
= ′2 −λ +
λr sin(πa) r′2 r4
 
r 1
= ′2 −λ + F (r) ,
λr sin(πa)
where F is defined as in Lemma 3.3.
Next, we calculate the limit values of Gλ (r) at 0 and 1. Simple computations
lead to
1 R(a) sin(πa) − 2(1 − a)[(πa)/2]
Gλ (0+ ) = − +
2 2 2λ sin(πa)
 
sin(πa)[R(a) − 1] sin(πa) − πa(1 − a)
= −λ +
2λ sin(πa) sin(πa)[R(a) − 1]
sin(πa)[R(a) − 1]
= (−λ + λ1 ),
2λ sin(πa)
which implies (3.12).
It follows from
Ka (r)  
lim− − log eR(a)/2 /r′ = 0
r→1 sin(πa)
in [17, p. 635, (2.26)] that
   2 
− 1 Ka R(a)/2 ′ r sin(πa) − sin(πa)
Gλ (1 ) = + lim− − log(e /r ) + lim−
2 r→1 sin(πa) r→1 2 sin(πa)λr2 r′2
12 MIAO-KUN WANG, YU-MING CHU, AND SONG-LIANG QIU

 
sin(πa) − 2(1 − a)(Ea − r′2 Ka ) Ka (r)
+ lim− −
r→1 2 sin(πa)λr2 r′2 sin(πa)
 ′2

1 1 sin(πa) − 2(1 − a)(Ea − r Ka ) Ka (r)
= − + lim− − . (3.14)
2 2λ r→1 2 sin(πa)λr2 r′2 sin(πa)
We divide the proof of (3.13) into two cases.
Case 1 λ 6= β. Then from (3.14) we have
 
− 1 1 sin(πa) − 2(1 − a)(Ea − r′2 Ka ) 1 Ka (r)
Gλ (1 ) = − + lim −1 +
2 2λ r→1− r′2 Ka 2λr2 sin(πa)
 
1 1 a(1 − a) Ka (r)
= − + lim −1 +
2 2λ r→1− λ sin(πa)

+∞, 0 < λ < β,
=
−∞, λ > β.
Case 2 λ = β. Then equation (3.14) leads to
1 1 sin(πa) − 2(1 − a)(Ea − r′2 Ka ) − 2a(1 − a)r2 r′2 Ka
Gβ (1− ) = − + lim−
2 2a(1 − a) r→1 2a(1 − a)r2 r′2 sin(πa)
sin(πa)
− (Ea − r′2 Ka ) − ar2 r′2 Ka
1 1 1 2(1 − a)
= − + lim
2 2a(1 − a) a sin(πa) r→1− r′2
1 1
= −
2 2a(1 − a)
1 −2arKa − 2arr′2 Ka + 2ar3 Ka − 2a(1 − a)r(Ea − r′2 Ka )
+ lim−
a sin(πa) r→1 −2r
1 1 1  ′2 
= − + lim 2r Ka + (1 − a)(Ea − r′2 Ka )
2 2a(1 − a) sin(πa) r→1−
1
=1 − .
2a(1 − a)

Proof of Theorem 1.1. Let
Hλ (r) = sin(πa)(1 + λr′2 ) log(eR(a)/2 /r′ ) − Ka (r), λ ∈ R+ . (3.15)
Then simple computations lead to
Hλ (0+ ) = [sin(πa)(1 + λ) − π]/2, (3.16)
−
Hλ (1 ) = 0, (3.17)
 r 
Hλ′ (r) = sin(πa)(−2λr) log(eR(a)/2 /r′ ) + sin(πa)(1 + λr′2 ) ′2
r
(Ea − r′2 Ka )
− 2(1 − a)
rr′2
=2λr sin(πa)Gλ (r), (3.18)
where Gλ is defined as in Lemma 3.4.
We divide the proof of inequality (1.7) into two cases.
Case I λ = α0 = π/[R(a) sin(πa)] − 1. Then equation (3.16) reduces to
Hα0 (0+ ) = 0. (3.19)
 SHARP BOUNDS FOR GENERALIZED ELLIPTIC INTEGRALS OF THE FIRST KIND 13

From Corollaries 2.4 and 2.5 we know that β > α0 > λ1 , then (3.12) and (3.13)
lead to the conclusion that Gα0 (0+ ) < 0 and Gα0 (1− ) = +∞. Moreover, wether
α0 ∈ (0, λ2 ] or α0 ∈ (λ2 , β), it follows from part (1) or (3) in Lemma 3.4 that there
exists r0∗ ∈ (0, 1) such that Gα0 (r) < 0 for r ∈ (0, r0∗ ) and Gα0 (r) > 0 for r ∈ (r0∗ , 1).
Hence, from (3.18) we clearly see that Hα0 (r) is strictly decreasing on (0, r0∗ ) and
strictly increasing on (r0∗ , 1).
Equations (3.17) and (3.19) together with the piecewise monotonicity of Hα0 (r)
lead to the conclusion that Hα0 (r) < 0 for all r ∈ (0, 1). Therefore, the first
inequality in (1.7) for α = α0 follows easily from (3.15).
Case II λ = β0 = a(1 − a). Then from (3.12), (3.13), Lemma 3.4(2) and the
fact that β0 > λ1 for all a ∈ (0, 1/2] we know that Gβ0 (r) is strictly decreasing on
(0, 1), Gβ0 (0+ ) < 0 and Gβ0 (1− ) < 0. Thus Gβ0 (r) < 0 for r ∈ (0, 1). It follows
from (3.17) and (3.18) that Hβ0 (r) is strictly decreasing on (0, 1) and Hβ0 (r) >
Hβ0 (1− ) = 0 for r ∈ (0, 1). Therefore, the second inequality in (1.7) for β = β0
follows from (3.15).
Finally, we prove that α = α0 and β = β0 are the best possible parameters such
that inequality (1.7) holds for all a ∈ (0, 1/2] and r ∈ (0, 1). In fact, if λ > α0 ,
then from (3.16) we know that Hλ (0+ ) > 0. Hence there exists r1 ∈ (0, 1) such
that Hλ (r) > 0 for r ∈ (0, r1 ). That is, 1 + λr′2 > Ka (r)/[sin(πa) log(eR(a)/2 /r′ )]
for r ∈ (0, r1 ).
On the other hand, if 0 < λ < β0 , then (3.13) and (3.18) imply that there exists
r1∗ ∈ (0, 1) such that Hλ′ (r) > 0 for r ∈ (r1∗ , 1). Thus Hλ (r) is strictly increasing on
(r1∗ , 1) and Hλ (r) < Hλ (1− ) = 0. That is, 1 + λr′2 < Ka (r)/[sin(πa) log(eR(a)/2 /r′ )]
for r ∈ (r1∗ , 1). 

References
[1] M. Abramowitz and I. A. Stegun, Handbook of Mathematical Functions with Formulas,
Graphs, and Mathematical Tables, U. S. Government Printing Office, Washington, 1964.
[2] H. Alzer, Sharp inequalities for the complete elliptic integral of the first kind, Math. Proc.
Cambridge Philos. Soc., 1998, 124(2): 309-314.
[3] G. D. Anderson, R. W. Barnard, K. C. Richards, M. K. Vamanamurthy and M. Vuori-
nen, Inequalities for zero-balanced hypergeometric functions, Trans. Amer. Math. Soc., 1995,
347(5): 1713-1723.
[4] G. D. Anderson, S.-L. Qiu, M. K. Vamanamurthy and M. Vuorinen, Generalized elliptic
integrals and modular equations, Pacific J. Math., 2000, 192(1): 1-37.
[5] G. D. Anderson, M. K. Vamanamurthy and M. Vuorinen, Conformal Invariants, Inequalities,
and Quasiconformal Maps, John Wiley & Sons, New York, 1997.
[6] S. András and Á. Baricz, Bounds for complete elliptic integrals of the first kind, Expo. Math.,
2010, 28(4): 357-364.
[7] G. E. Andrews, R. Askey and R. Roy, Special Functions, Cambridge University Press, Cam-
bridge, 1999.
[8] Á. Baricz, Turán type inequalities for generalized complete elliptic integrals, Math. Z., 2007,
256(4): 895-911.
[9] R. W. Barnard, K. Pearce and K. C. Richards, A monotonicity property involving 3 F2 and
comparisons of the classical approximations of elliptical arc length, SIAM J. Math. Anal.,
2000, 32(2): 403-419.
[10] P. F. Byrd and M. D. Friedman, Handbook of Elliptic Integrals for Engineers and Scientists,
Springer-Verlag, New York, 1971.
[11] N. J. Fine, Basic Hypergeometric Series and Applications, American Mathematical Society,
Providence, 1988.
14 MIAO-KUN WANG, YU-MING CHU, AND SONG-LIANG QIU

[12] V. Heikkala, H. Lindén, M. K. Vamanamurthy and M. Vuorinen, Generalized elliptic integrals

and the Legendre M-function, J. Math. Anal. Appl., 2008, 338(1): 223-243.
[13] V. Heikkala, M. K. Vamanamurthy and M. Vuorinen, Generalized elliptic integrals, Comput.
Methods Funct. Theory, 2009, 9(1): 75-109.
[14] E. Neuman, Inequalities and bounds for generalized complete elliptic integrals, J. Math. Anal.
Appl., 2011, 373(1): 203-213.
[15] S.-L. Qiu, B.-P. Feng, Some properties of the Ramanujan constant, J. Hangzhou Dianzi Univ.,
2007, 27(3): 88-91.
[16] S.-L. Qiu and M. K. Vamanamurthy, Sharp estimates for complete elliptic integrals, SIAM
J. Math. Anal., 1996, 27(3): 823-834.
[17] S.-L. Qiu and M. Vuorinen, Special functions in geometric function theory, in: Handbook of
Complex Analysis: geometric function theory, Vol. 2, 621-659, Elsevier Sci. B. V., Amster-
dam, 2005.
[18] S. Takeuchi, A new form of the generalized complete elliptic integrals, arXiv:1411.4778
[math.CA], Available Online at http://arxiv-web3.library.cornell.edu/pdf/1411.4778v2.pdf.
[19] M. K. Vamanamurthy and M. Vuorinen, Inequalities for means, J. Math. Anal. Appl., 1994,
183(1): 155-166.
[20] M. Vuorinen, Singular values, Ramanujan modular equations, and Landen transformations,
Studia Math., 1996, 121(3): 221-230.
[21] M.-K. Wang, Y.-M. Chu, S.-L. Qiu and Y.-P. Jiang, Convexity of the complete elliptic in-
tegrals of the first kind with respect to Hölder means, J. Math. Anal. Appl., 2012, 388(2):
1141-1146.
[22] X.-H. Zhang, G.-D. Wang and Y.-M. Chu, Remark on generalized elliptic integrals, Proc.
Roy. Soc. Edinburgh Sect. A, 2009, 139(2): 417-426.

Miao-Kun Wang, Department of Mathematics, Huzhou University, Huzhou 313000,

China
E-mail address: wmk000@126.com

Yu-Ming Chu (Corresponding author), Department of Mathematics, Huzhou Univer-

sity, Huzhou 313000, China
E-mail address: chuyuming@hutc.zj.cn

Department of Mathematics, Zhejiang Sci-Tech University, Hangzhou, Zhejiang, 310018,

China
E-mail address: sl qiu@zstu.edu.cn