PROCEEDINGS OF THE

AMERICAN MATHEMATICAL SOCIETY

Volume 145, Number 4, April 2017, Pages 16611670
http://dx.doi.org/10.1090/proc/13337
Article electronically published on October 13, 2016

INEQUALITIES FOR THE RATIO

OF COMPLETE ELLIPTIC INTEGRALS

HORST ALZER AND KENDALL RICHARDS

(Communicated by Mourad Ismail)

Abstract. We present various inequalities for the complete elliptic integral

of the rst kind,
 /2
1
K(r) =  dt (0 < r < 1).
0 1 r 2 sin2 (t)
Among others, we prove that the inequalities

1 K(r) K( 1 r 2 ) 2
< and <
1
1 + 4r K( r) K( 1 r) 1+ r
are valid for all r (0, 1). These estimates rene results published by Ander-
son, Vamanamurthy, and Vuorinen in 1990.

1. Introduction
The complete elliptic integrals of the rst and second kind are dened for r
(0, 1) by
 /2
1 
(1.1) K(r) =  dt = an r 2n
0 1 r sin (t)
2 2 2 n=0

and
 
/2
 an
E(r) = 1 r 2 sin2 (t)dt = r 2n
0 2 n=0 1 2n
with
 (1/2) 2 1  (n + 1/2) 2
n
(1.2) an = = .
n! (n + 1)
Both functions can be expressed in terms of the Gaussian hypergeometric function,
namely

K(r) = F (1/2, 1/2; 1; r 2 ) and E(r) = F (1/2, 1/2; 1; r 2 ).
2 2
The associate complete elliptic integrals are given by

K (r) = K(r  ) and E  (r) = E(r  ), where r  = 1 r 2 .

Received by the editors April 18, 2016 and, in revised form, June 9, 2016, June 10, 2016 and
June 12, 2016.
2010 Mathematics Subject Classication. Primary 33C75, 39B62; Secondary 26E60.
Key words and phrases. Complete elliptic integrals, functional inequalities, means.

2016
c American Mathematical Society

1661

Licensed to AMS.
License or copyright restrictions may apply to redistribution; see http://www.ams.org/journal-terms-of-use
 1662 HORST ALZER AND KENDALL RICHARDS

These functions have remarkable applications in the theory of quasiconformal map-

pings as well as in geometry, mechanics, geodesy and other elds. The main prop-
erties of K and E are collected, for instance, in the books [6] and [8]. We also refer
to the research articles [2], [9], [10], [11], [14], [15], [16] and the references cited
therein.
In the next sections we make use of the formulas
1 r 1 + r
(1.3) K = K (r)
1+r 2
and
 2r 
(1.4) K = (1 + r)K(r)
1+r
which are known as Landen identities and we apply the limit relations
 1 
(1.5) lim K(r) + log(1 r 2 ) = log(4)
r1 2
and
E(r) (1 r 2 )K(r)
(1.6) lim 2
= ;
r0 r 4
see [8, (164.02), (112.01)] [1, (17.3.27)].
The work on this paper has been inspired by the following elegant inequalities
which were published by Anderson et al. [4] in 1990:

1 K(r)
(1.7) <
1+r K( r)
and
K (r) 2
(1.8) < .
K ( r) 1+r

Both inequalities are valid for all r (0, 1).

Here we present various new inequalities for the ratio of complete elliptic inte-
grals. Among others, we prove that the following renements of (1.7) and (1.8)
hold for all r (0, 1):

1 K(r) K (r) 2

1 < K( r) and 
< .
1 + 4r K ( r) 1+ r

An application of the second inequality leads to an improvement of the well-known

arithmetic mean - geometric mean inequality
1+r
0< r (0 < r < 1).
2
We show that the lower bound 0 can be replaced by a positive expression written
in terms of K.
In the next section we provide three lemmas. They play a role in the proofs of
our main results which are given in Section 3.

Licensed to AMS.
License or copyright restrictions may apply to redistribution; see http://www.ams.org/journal-terms-of-use
 INEQUALITIES FOR THE RATIO OF COMPLETE ELLIPTIC INTEGRALS 1663

2. Lemmas
The rst lemma provides properties of the logarithmic derivative of the gamma
function; see [1, (6.3.8), (6.4.10)].
Lemma 2.1. The function =  / is strictly concave on (0, ) and satises for
x > 0 the duplication formula
1 1  1
(2x) = (x) + x + + log(2).
2 2 2
Moreover, we need several limit relations. They are collected in the following
lemmas.
Lemma 2.2. We have
F (a, b; a + b; r 2 ) F (a, b; a + b; 1 r 2 )
(2.1) lim =1 and lim = 2.
r1 F (a, b; a + b; r) r0 F (a, b; a + b; 1 r)

Proof. Using [12, eq. (2), p. 74] gives the asymptotic relation
1
F (a, b; a + b; x) log(1 x) (x 1),
B(a, b)
where B(a, b) denotes Eulers beta function. This leads to (2.1). 

Remark 2.3. Setting a = b = 1/2 in (2.1) yields

K(r)
(2.2) lim =1
r1 K( r)
and
K (r)
(2.3) lim = 2.
r0 K ( r)

Lemma 2.4. Let

d 1
u(r) = K ( r), v(r) = r u(r), w(r) = u(r) + log(r)
dr 2
and
(2.4) A(r) = v(r)w(r 2 ) 2v(r 2 )w(r), B(r) = v(r 2 ) v(r),
 w(r 2 ) 
C(r) = 1 2w(r) w(r 2 ) .
log(r)
Then,

lim A(r) = log(2), lim B(r) = 0, lim C(r) = log(4), lim v(r) = .
r0 r0 r0 r1 8
Proof. From (1.5) we obtain
(2.5) lim w(r) = log(4).
r0

Thus,

E( 1 r) rw(r) + 12 r log(r) 10+0 1
(2.6) lim v(r) = lim = = .
r0 r0 2(r 1) 2 (0 1) 2

Licensed to AMS.
License or copyright restrictions may apply to redistribution; see http://www.ams.org/journal-terms-of-use
 1664 HORST ALZER AND KENDALL RICHARDS

Using (2.5) and (2.6) leads to

1 1
lim A(r) = log(4) + 2 log(4) = log(2),
2
r0 2
1 1 
lim B(r) = + = 0, lim C(r) = (0 1) 2 log(4) log(4) = log(4).
r0 2 2 r0
Applying (1.6) gives
 1 E(1 r) rK(1 r)  1
lim v(r) = lim = = .
r1 r1 2 1r 2 4 8


3. Inequalities
Our rst theorem oers an improvement of inequality (1.7).
Theorem 3.1. For all r (0, 1) we have
1 K(r) 1
(3.1) < <
1 + r K( r) 1 + r
with the best possible constant factors = 1/4 and = 0.
Proof. Let r (0, 1). We dene
2 1  2
f (r) = 1 + r K(r) and g(r) = K( r).
4
Applying (1.1) and (1.2) yields the series representations
  
1
f (r) = an + an r r 2n
n=0
4
and

 
g(r) = a2n + a2n+1 r r 2n .
n=0
Let
an + 14 an r
Pn (r) = .
a2n + a2n+1 r
Then, P0 (r) = 1. We obtain
1 + 14 r an 1 + 14 an
Pn (r) = 4n+1 2 a > 4n+1 2 a .
1 + ( 4n+2 ) r 2n 1 + ( 4n+2 ) 2n
Let x > 0 and
(x + 1/2)(2x + 1)
Q(x) = .
(x + 1)(2x + 1/2)
Using Lemma 2.1 and Jensens inequality gives
Q (x)
= (x + 1/2) + 2(2x + 1) (x + 1) 2(2x + 1/2)
Q(x)


= 2(x + 1/2) (x + 1/4) + (x + 3/4) 2(x + 1/2) 2(x + 1/2) = 0.
It follows that Q is increasing on (0, ). Thus, for n 1,
an 16
= Q2 (n) Q2 (1) = .
a2n 9

Licensed to AMS.
License or copyright restrictions may apply to redistribution; see http://www.ams.org/journal-terms-of-use
 INEQUALITIES FOR THE RATIO OF COMPLETE ELLIPTIC INTEGRALS 1665

This yields
16 1 + 14 32n2 + 104n + 35
Pn (r) > 4n+1 = 1 + > 1.
9 1 + ( 4n+2 )2 9(32n2 + 24n + 5)
Hence, we have f (r) > g(r) which is equivalent to the left-hand side of (3.1) with
= 1/4.
Since K is strictly increasing on (0, 1), we obtain K(r)/K( r) < 1. This leads
to the right-hand side of (3.1) with = 0.
It remains to show that the given constants are best possible. Double-inequality
(3.1) can be written as
 K(r) 1
(3.2) < S(r) = 1 < .
K(r) r
Since
a1 + (a2 a1 )r + . . .
S(r) = ,
a0 + a1 r 2 + . . .
we obtain
a1 1
(3.3) lim S(r) = = .
r0 a0 4
Using (2.2) gives
(3.4) lim S(r) = 0.
r1

From (3.2), (3.3) and (3.4) we conclude that the constants = 1/4 and = 0 are
sharp.

Note added in proof. After the paper was accepted for publication we dis-
covered that G.D. Anderson, M.K. Vamanamurthy and M. Vuorinen [5] proved the
inequalities

1 K(r) min{ 4 2, 1/ r  }
(3.5) < < , (0 < r < 1).
4
1+r K( r) 4
1+r
The lower bound in (3.5) improves the left-hand side of (3.1) with = 1/4, whereas
the upper bounds given in (3.5) and (3.1) cannot be compared in general. We also
refer to S.-L. Qiu and M.K. Vamanamurthy [13] who showed that the function

r  4 1 + rK(r)/K( r) is strictly increasing from [0, 1) onto [1, 4 2). 

An application of (3.1) leads to the following result.

Corollary 3.2. For all r (0, 1) we have

K( r) K(r) 1
(3.6) 0<   < 3.
2 r
K 1+r K( r)

Both bounds are sharp.

Proof. Let r (0, 1). We denote the ratio in(3.6) by
T (r). Since K is strictly
increasing on (0, 1), we conclude from 0 < r < r < 2 r/(1 + r) < 1 that
 2r 
0 < K( r) K(r) and 0 < K K( r).
1+r

Licensed to AMS.
License or copyright restrictions may apply to redistribution; see http://www.ams.org/journal-terms-of-use
 1666 HORST ALZER AND KENDALL RICHARDS

Thus, T (r) > 0. Using (1.4) and the left-hand side of (3.1) with = 1/4 gives
  2r  
K K( r) 1 3T (r) = (4 + r)K(r) 4K( r)
1+r
 1 
> (4 + r)K(r) 4 1 + r K(r) = 0.
4
This leads to T (r) < 1/3.
Applying (2.2) gives
K(r)
1
K( r) 11
(3.7) lim T (r) = lim = = 0.
r1 r1 (1 + K(r)
r) K(r) 1 21

We have
S(r)
, T (r) =
1 S(r)
where S is dened in (3.2). Using (3.3) leads to
1
4 1
(3.8) lim T (r) = = .
r0 1 1
4
3
From (3.7) and (3.8) we conclude that the constant bounds given in (3.6) are best
possible. 
Now, we present a renement of (1.8).
Theorem 3.3. Let , R. The inequalities
2 K (r) 2
(3.9) < <
1 + r K ( r) 1 + r
are valid for all r (0, 1) if and only if 0 and 1/2.
Proof. First, we prove that if = 0 and = 1/2, then (3.9) holds for all r (0, 1).
Since K is strictly decreasing on (0, 1), we obtain
2 K (r)
(3.10) = 1 < .
1 + r0 K ( r)
Let
1 r
. (r) =
1+ r
We replace in (1.4) r by 2 (r) and obtain
1 r  
(3.11) K = 1 + 2 (r) K 2 (r) .
1+r
Using (1.3) and (3.11) gives
 1r 
K (r) 1 + r K 1+r 1 + r K 2 (r)
(3.12) = = 2
1 + (r) .
K ( r) 1 + r K((r)) 1+r K((r))
Since K is strictly increasing on (0, 1), we get

K (r) 1 + r 2
(3.13) < 1 + 2 (r) = .
K ( r) 1+r 1+ r
If r (0, 1), then the function t  2/(1 + r t ) is increasing on R. It follows from
(3.10) and (3.13) that if 0 and 1/2, then (3.9) is valid for all r (0, 1).

Licensed to AMS.
License or copyright restrictions may apply to redistribution; see http://www.ams.org/journal-terms-of-use
 INEQUALITIES FOR THE RATIO OF COMPLETE ELLIPTIC INTEGRALS 1667

Next, we assume that (3.9) holds for all r (0, 1). Then, we obtain
(3.14) < F (r) <
with
log(G(r)) K ( r)
F (r) = and G(r) = 2  1.
log(r) K (r)
Using (2.3) leads to
lim G(r) = 0.
r0
Thus,
rG1 (r)
(3.15) lim F (r) = lim
r0 r0 G(r)
with G1 (r) = (d/dr)G(r). We have
A(r)
rG1 (r) log(r) + B(r)
=2 ,
G(r) C(r)
where A, B and C are dened in (2.4). Applying Lemma 2.4 gives
rG1 (r) 0+0
(3.16) lim =2 = 0.
r0 G(r) log(4)
From (3.15) and (3.16) we conclude that
(3.17) lim F (r) = 0.
r0

Since G(1) = 1, we obtain

rG1 (r)
(3.18) lim F (r) = lim = lim G1 (r).
r1 r1 G(r) r1

We have
2u(r 2 )v(r) 4u(r)v(r 2 )
G1 (r) = ,
ru2 (r 2 )

with u(r) = K ( r). Using u(1) = /2 and Lemma 2.4 yields
1
(3.19) lim G1 (r) = .
r1 2
From (3.18) and (3.19) we get
1
(3.20) lim F (r) = .
r1 2
Applying (3.14) and the limit relations (3.17) and (3.20) leads to 0 and 1/2.
This completes the proof of Theorem 3.3. 

Remark 3.4. From (3.12) and the left-hand side of (3.1) with = 1/4 we obtain
for r (0, 1):

K (r) 1 + r 1 + 2 (r) 2(1 + r)
> = > 1.
K ( r) 1 + r 1 + 14 2 (r) (1 + r)2 + 14 (1 r)2
This is a renement of the rst inequality in (3.9) with = 0.

Licensed to AMS.
License or copyright restrictions may apply to redistribution; see http://www.ams.org/journal-terms-of-use
 1668 HORST ALZER AND KENDALL RICHARDS

Remark 3.5. The following companion of (3.9) holds for r (0, 1):
2 K (r)
0<  0.2254 . . . .
1 + r K ( r)
The constant bounds are sharp. Does there exist a closed expression for the given
numerical upper bound?
In the literature we can nd many renements of the classical arithmetic mean
- geometric mean inequality; see, for example, [7, II.3]. The second inequality in
(3.9) plays a key role in the proof of the following new renement.
Corollary 3.6. For all r (0, 1) we have
   
1r
K 1r
1+r K 1+ r 1+r
(3.21) 0< < r.
K( 1 r) 2
Proof. Let r (0, 1). Since

1 r 1r
0< < < 1,
1+ r 1+r
we obtain
 1 r  1 r
K <K .
1+ r 1+r
This leads to the left-hand side of (3.21). Applying the second inequality in (3.9)
with = 1/2 gives

K (r) 2 2+r r
< < .
K ( r) 1+ r 1+r
It follows that

(1 + r)K (r) < (2 + r r)K ( r) = (1 + r)K ( r) + (1 + r 2 r)K ( r).
Using (1.3) we nd
1 r  1 r 
2K < 2K + (1 + r 2 r)K( 1 r).
1+r 1+ r
This is equivalent to the right-hand side of (3.21). 

4. Concluding remarks
The generalized elliptic integrals are dened for a (0, 1/2] and r (0, 1) by

Ka (r) = F (a, 1 a; 1; r 2 ) and Ea (r) = F (a 1, 1 a; 1; r 2 ).
2 2
The special case a = 1/2 gives
K1/2 = K and E1/2 = E.
For more information on these functions we refer the reader to [3].
The referee pointed out that Theorems 3.1 and 3.3 seem to hold not only for the
complete elliptic integral K but also for Ka . Indeed, using similar methods as in
the proof of Theorem 3.1 we obtain the following result.

Licensed to AMS.
License or copyright restrictions may apply to redistribution; see http://www.ams.org/journal-terms-of-use
 INEQUALITIES FOR THE RATIO OF COMPLETE ELLIPTIC INTEGRALS 1669

Theorem 4.1. Let a (0, 1/2]. For all r (0, 1) we have

1 Ka (r) 1
< <
1 + a r Ka ( r) 1 + a r
with the best possible factors a = a(1 a) and a = 0.
It remains an open problem to nd a corresponding generalization of Theorem
3.3. M.E.H. Ismail even asked whether our inequalities can be extended to the
zero-balanced hypergeometric function F (a, b; a + b; r 2 ).

Acknowledgements
We are grateful to the referee for helpful comments and for completing our list
of references. In particular, we thank Professor Ismail for the proof of Lemma 2.2.

References
[1] Handbook of mathematical functions with formulas, graphs, and mathematical tables, Dover
Publications, Inc., New York, 1992. Edited by Milton Abramowitz and Irene A. Stegun;
Reprint of the 1972 edition. MR1225604
[2] Horst Alzer and Song-Liang Qiu, Monotonicity theorems and inequalities for the com-
plete elliptic integrals, J. Comput. Appl. Math. 172 (2004), no. 2, 289312, DOI
10.1016/j.cam.2004.02.009. MR2095322
[3] G. D. Anderson, S.-L. Qiu, M. K. Vamanamurthy, and M. Vuorinen, Generalized el-
liptic integrals and modular equations, Pacic J. Math. 192 (2000), no. 1, 137, DOI
10.2140/pjm.2000.192.1. MR1741031
[4] G. D. Anderson, M. K. Vamanamurthy, and M. Vuorinen, Functional inequalities for complete
elliptic integrals and their ratios, SIAM J. Math. Anal. 21 (1990), no. 2, 536549, DOI
10.1137/0521029. MR1038906
[5] G. D. Anderson, M. K. Vamanamurthy, and M. Vuorinen, Functional inequalities for hyper-
geometric functions and complete elliptic integrals, SIAM J. Math. Anal. 23 (1992), no. 2,
512524, DOI 10.1137/0523025. MR1147875
[6] Glen D. Anderson, Mavina K. Vamanamurthy, and Matti K. Vuorinen, Conformal invariants,
inequalities, and quasiconformal maps, Canadian Mathematical Society Series of Monographs
and Advanced Texts, John Wiley & Sons, Inc., New York, 1997. With 1 IBM-PC oppy disk
(3.5 inch; HD); A Wiley-Interscience Publication. MR1462077
[7] P. S. Bullen, D. S. Mitrinovi c, and P. M. Vasic, Means and their inequalities, Mathematics
and its Applications (East European Series), vol. 31, D. Reidel Publishing Co., Dordrecht,
1988. Translated and revised from the Serbo-Croatian. MR947142
[8] Paul F. Byrd and Morris D. Friedman, Handbook of elliptic integrals for engineers and
physicists, Die Grundlehren der mathematischen Wissenschaften in Einzeldarstellungen mit
besonderer Ber ucksichtigung der Anwendungsgebiete. Bd LXVII, Springer-Verlag, Berlin-
Gottingen-Heidelberg, 1954. MR0060642
[9] Y. M. Chu, S. L. Qiu, and M. K. Wang, Sharp inequalities involving the power mean and
complete elliptic integral of the rst kind, Rocky Mountain J. Math. 43 (2013), no. 5, 1489
1496, DOI 10.1216/RMJ-2013-43-5-1489. MR3127833
[10] Yu-Ming Chu, Miao-Kun Wang, Song-Liang Qiu, and Yue-Ping Jiang, Bounds for complete
elliptic integrals of the second kind with applications, Comput. Math. Appl. 63 (2012), no. 7,
11771184, DOI 10.1016/j.camwa.2011.12.038. MR2900091
[11] Yu-Ming Chu and Tie-Hong Zhao, Convexity and concavity of the complete elliptic inte-
grals with respect to Lehmer mean, J. Inequal. Appl., posted on 2015, 2015:396, 6, DOI
10.1186/s13660-015-0926-7. MR3434489
[12] A. Erd elyi, W. Magnus, F. Oberhettinger, and F. G. Tricomi, Higher Transcendental Func-
tions, Vol. 1, McGraw-Hill, New York, 1953.
[13] S.-L. Qiu and M. K. Vamanamurthy, Sharp estimates for complete elliptic integrals, SIAM
J. Math. Anal. 27 (1996), no. 3, 823834, DOI 10.1137/0527044. MR1382835

Licensed to AMS.
License or copyright restrictions may apply to redistribution; see http://www.ams.org/journal-terms-of-use
 1670 HORST ALZER AND KENDALL RICHARDS

[14] Miao-Kun Wang and Yu-Ming Chu, Asymptotical bounds for complete elliptic inte-
grals of the second kind, J. Math. Anal. Appl. 402 (2013), no. 1, 119126, DOI
10.1016/j.jmaa.2013.01.016. MR3023241
[15] Miao-Kun Wang, Yu-Ming Chu, Song-Liang Qiu, and Yue-Ping Jiang, Convexity of the com-
plete elliptic integrals of the rst kind with respect to H
older means, J. Math. Anal. Appl.
388 (2012), no. 2, 11411146, DOI 10.1016/j.jmaa.2011.10.063. MR2869813
[16] Gendi Wang, Xiaohui Zhang, and Yuming Chu, A power mean inequality involving the
complete elliptic integrals, Rocky Mountain J. Math. 44 (2014), no. 5, 16611667, DOI
10.1216/RMJ-2014-44-5-1661. MR3295648

l, Germany
Morsbacher Str. 10, D-51545 Waldbro
E-mail address: h.alzer@gmx.de

Department of Mathematics and Computer Science, Southwestern University,

Georgetown, Texas
E-mail address: richards@southwestern.edu

Licensed to AMS.
License or copyright restrictions may apply to redistribution; see http://www.ams.org/journal-terms-of-use