Skip to main content

Three families of two-parameter means constructed by trigonometric functions

Abstract

In this paper, we establish three families of trigonometric functions with two parameters and prove their monotonicity and bivariate log-convexity. Based on them, three two-parameter families of means involving trigonometric functions, which include Schwab-Borchardt mean, the first and second Seiffert means, Sándor’s mean and many other new means, are defined. Their properties are given and some new inequalities for these means are proved. Lastly, two families of two-parameter hyperbolic means, which similarly contain many new means, are also presented without proofs.

MSC: 26E60, 26D05, 33B10, 26A48.

1 Introduction

Let R + denote the set of positive real numbers and a,b R + . A two-variable continuous function M: R + 2 R + is called a mean on R + if

min(a,b)M(a,b)max(a,b)

holds. For convenience, however, we assume that ab in what follows unless otherwise stated.

There exist many elementary means. They can be divided into three classes according to main categories of basic elementary functions by their composition. The first class is mainly constructed by power functions, like the Stolarsky means [1] defined by

S p , q (a,b)= { ( q ( a p b p ) p ( a q b q ) ) 1 / ( p q ) , p q , p q 0 , ( a p b p p ( ln a ln b ) ) 1 / p , p 0 , q = 0 , ( a q b q q ( ln a ln b ) ) 1 / q , p = 0 , q 0 , exp ( a p ln a b p ln b a p b p 1 p ) , p = q 0 , a b , p = q = 0
(1.1)

and Gini means [2] defined by

G p , q (a,b)= { ( a p + b p a q + b q ) 1 / ( p q ) , p q ; exp ( a p ln a + b p ln b a p + b p ) , p = q .
(1.2)

It is well known that the Stolarsky and Gini means are very important, they contain many famous means, for instance, S 1 , 0 (a,b)=L(a,b) - the logarithmic mean, S 1 , 1 (a,b)=I(a,b) - the identric (exponential) mean, S 2 , 1 (a,b)= G 1 , 0 (a,b)=A(a,b), S 2 p , p (a,b)= G p , 0 (a,b)= A 1 / p ( a p , b p )= A p - the p-order power mean, S p , 0 (a,b)= L 1 / p ( a p , b p )= L p - the p-order logarithmic mean, S p , p (a,b)= I 1 / p ( a p , b p )= I p - the p-order identric (exponential) mean; G 2 , 0 (a,b)=Q(a,b) - the quadratic mean, G 1 , 1 (a,b)=Z(a,b) - the power-exponential mean, G p , p (a,b)= Z 1 / p ( a p , b p )= Z p - the p-order power-exponential mean, etc. The second class is mainly made up of exponential and logarithmic functions, such as the second part of Schwab-Borchardt mean (see [3], [[4], Section 3, equation (2.3)], [5]) defined by

SB=SB(a,b)= { b 2 a 2 arccos ( a / b ) , 0 a < b , a 2 b 2 arccosh ( a / b ) , b < a ,
(1.3)

the logarithmic mean L(a,b), the exponential mean defined by

E(a,b)= b e b a e a e b e a 1

given in [6] (also see [7, 8]) by Sándor and Toader, and the Neuman-Sándor mean defined in [5] by

NS=NS(a,b)= a b 2 arcsinh a b a + b .
(1.4)

It should be noted that NS is actually a Schwab-Borchardt mean since NS(a,b)=SB(Q,A) mentioned by Neuman and Sándor in [5].

The third class is mainly composed of trigonometric functions and their inverses, for example, the first part of Schwab-Borchardt mean defined by (1.3), the first and second Seiffert means [9, 10] defined by

P=P(a,b)= a b 2 arcsin a b a + b ,
(1.5)
T=T(a,b)= a b 2 arctan a b a + b ,
(1.6)

respectively, and the new mean presented recently by Sándor in [11, 12] defined as

X=X(a,b)=A e G / P 1 ,
(1.7)

where A=(a+b)/2, G= a b , P is defined by (1.5). As Neuman and Sándor pointed out in [5], the first and second Seiffert means are generated by the Schwab-Borchardt mean, because SB(G,A)=P(a,b), SB(A,Q)=T(a,b).

From the published literature, the first and second classes have been focused on and investigated, and there are a lot of references (see [1, 1324]). While the third class is relatively little known.

The aim of this paper is to define three families of two-parameter means constructed by trigonometric functions, which include the Schwab-Borchardt mean SB, the first and second Seiffert means P, T, and Sándor’s mean X.

The paper is organized as follows. In Section 2, some useful lemmas are given. Three families of trigonometric functions and means with two parameters and their properties are presented in Sections 3-5. In Section 6, we establish some new inequalities for two-parameter trigonometric means. In the last section, two families of two-parameter hyperbolic means are also presented in the same way without proofs.

2 Lemmas

For later use, we give the following lemmas.

Lemma 2.1 [[25], p.26]

Let f be a differentiable function defined on an interval I. Then the divided differences function F defined on I 2 by

F(x,y)= f ( x ) f ( y ) x y if xyandF(x,x)= f (x)
(2.1)

is increasing (decreasing) in both variables if and only if f is convex (concave).

Lemma 2.2 [[26], Theorem 1]

Let f be a differentiable function defined on an interval I, and let F be defined on I 2 by (2.1). Then the following statements are equivalent:

  1. (i)

    f is convex (concave) on I,

  2. (ii)

    F(x,y)() f ( x ) + f ( y ) 2 for all x,yI,

  3. (iii)

    F is bivariate convex (concave) on I 2 .

Lemma 2.3 If f:(m,m)R is a differentiable even function such that f is convex in (0,m), then the function xF(c+x,cx) defined by (2.1) increases for positive x if c0 and decreases if c0 provided c+x,cx(m,m).

Proof Differentiation yields

F (c+x,cx)= f ( c + x ) + f ( c x ) 2 f ( c + x ) f ( c x ) 2 x x := x 1 g c (x).

Since f is an even and differentiable function, it is easy to verify that g c (x)= g c (x), g c (x)= g c (x). From this we only need to prove that for c0, F (c+x,cx)>0 for x(0,m) provided c+x,cx(m,m) if f is convex on (0,m).

To this end, we first show two facts. Firstly, application of Lemma 2.2 leads to

f ( c + x ) f ( c x ) 2 x = | c + x | | c x | 2 x f ( | c + x | ) f ( | c x | ) | c + x | | c x | < 2 c | c + x | + | c x | f ( | c + x | ) + f ( | c x | ) 2 .

The second one states that if h is a continuous and odd function on [d,d] (d>0) and is convex on [0,d], then, for u,v(0,d] with u>v, the inequality

h ( u ) u > h ( v ) v

holds. Indeed, using the fact h(0)=0 and the property of a convex function, the second one easily follows.

Now we can prove the desired result. When x(0,m), application of the two previous facts and notice that f is odd on (m,m) lead to

g c ( x ) > f ( c + x ) + f ( c x ) 2 2 c | c + x | + | c x | f ( | c + x | ) + f ( | c x | ) 2 = { 0 if  c x [ 0 , m ) , ( c x ) ( c + x ) 2 x ( f ( c + x ) x + c f ( c x ) c x ) > 0 if  c x ( m , 0 ) .

Hence, F (c+x,cx)= x 1 g c (x)>0 for c,x0.

This completes the proof. □

Lemma 2.4 The following inequalities are true:

sinxx<(>)0for x>(<)0,
(2.2)
sinxxcosx>(<)0for x(0,π) ( x ( π , 0 ) ) ,
(2.3)
x+sinx>(<)0for x>(<)0,
(2.4)
cosx+xsinx>1for 0<|x|<π/2,
(2.5)
cosx sin 3 x x 3 <0for 0<|x|<π,
(2.6)
sin x x < 2 + cos x 3 for x0.
(2.7)

Proof Inequalities (2.2)-(2.5) easily follow by the elementary differential method, and we omit all details here. Inequality (2.6) can be derived from a well-known inequality given in [[27], p.238]) by Adamović and Mitrinović for 0<|x|<π/2, while it is obviously true for π/2|x|<π. Inequality (2.7) can be found in [[28], Problem 5.11, 5.12]. This lemma is proved. □

Lemma 2.5 [[29], pp.227-229]

Let 0<|x|<π. Then we have

1 sin x = 1 x + n = 1 2 2 n 2 ( 2 n ) ! | B 2 n | x 2 n 1 ,
(2.8)

where B n is the Bernoulli number.

3 Two-parameter sine means

3.1 Two-parameter sine functions

We begin with the form of hyperbolic functions of Stolarsky means defined by (1.1) to introduce the two-parameter sine functions. Let t=ln b / a . Then the Stolarsky means can be expressed in hyperbolic functions as

S p , q (a,b)= a b Sh(p,q,t),

where

Sh(p,q,t)= { ( q p sinh p t sinh q t ) 1 / ( p q ) , p q ( p q ) 0 , ( sinh p t p t ) 1 / p , p 0 , q = 0 , ( sinh q t q t ) 1 / q , p = 0 , q 0 , e t coth p t 1 / p , p = q , p q 0 , 1 , p = q = 0 .
(3.1)

We call Sh(p,q,t) two-parameter hyperbolic sine functions. Accordingly, for suitable p, q, t, we can give the definition of sine versions of Sh(p,q,t) as follows.

Definition 3.1 The function S ˜ is called a sine function with parameters if S ˜ is defined on [ 2 , 2 ] 2 ×(0,π/2) by

S ˜ (p,q,t)= { ( q p sin p t sin q t ) 1 / ( p q ) , p q ( p q ) 0 , ( sin p t p t ) 1 / p , q = 0 , p 0 , ( sin q t q t ) 1 / q , p = 0 , q 0 , e t cot p t 1 / p , p = q 0 , 1 , p = q = 0 .
(3.2)

S ˜ is said to be a two-parameter sine function for short.

Now let us observe its properties.

Proposition 3.1 Let the two-parameter sine function S ˜ be defined by (3.2). Then

  1. (i)

    S ˜ is decreasing in p, q on [2,2], and is log-concave in (p,q) for (p,q) [ 0 , 2 ] 2 and log-convex for (p,q) [ 2 , 0 ] 2 ;

  2. (ii)

    S ˜ is decreasing and log-concave in t on (0,π/2) for p+q>0, and is increasing and log-convex for p+q<0.

Proof We have

ln S ˜ (p,q,t)= f ( p ) f ( q ) p q if pqandln S ˜ (p,p,t)= f (p),

where

f(x)=ln sin x t x if |x|(0,2]andf(0)=lnt.
(3.3)
  1. (i)

    We prove that S ˜ is decreasing in p, q on [2,2], and is log-concave in (p,q) for (p,q) [ 0 , 2 ] 2 and log-convex for (p,q) [ 2 , 0 ] 2 . By Lemmas 2.1 and 2.2, it suffices to check that f is concave in p,q[2,2] and that f is concave for p,q[0,2] and convex for p,q[2,0].

Differentiation and employing (2.2), (2.6) yield that for t(0,π/2),

f (x)= t cos t x sin t x 1 x if |x|(0,2]and f (0)=0,
(3.4)
f (x)= sin 2 t x t 2 x 2 x 2 sin 2 t x <0for |x|(0,2]and f (0)= t 2 3 <0,
(3.5)
f (x)= 2 t 3 sin 3 t x ( cos t x sin 3 t x t 3 x 3 ) { < 0 if  x ( 0 , 2 ] , > 0 if  x [ 2 , 0 ) ,
(3.6)

which prove part one.

  1. (ii)

    Now we show that S ˜ is decreasing and log-concave in t on (0,π/2) for p+q>0, and increasing and log-convex for p+q<0. It is easy to verify that f is an odd function on [2,2], and so ln S ˜ (p,q,t) can be written in the form of integral as

    ln S ˜ (p,q,t)= 1 p q q p f (x)dx= p + q | p | + | q | 1 | p | | q | | q | | p | f (x)dx.
    (3.7)

Differentiation and application of (2.2) and (2.3) give

f t = 1 2 sin 2 t x ( sin 2 t x 2 t x ) < 0 for  x ( 0 , 2 ] , 2 f t 2 = 2 x sin 3 t x ( t x cos t x sin t x ) < 0 for  x ( 0 , 2 ] .

It follows from (3.7) that

ln S ˜ t = p + q | p | + | q | 1 | p | | q | | q | | p | f t d x { < 0 if  p , q [ 2 , 2 ]  with  p + q > 0 , > 0 if  p , q [ 2 , 2 ]  with  p + q < 0 , 2 ln S ˜ t 2 = p + q | p | + | q | 1 | p | | q | | q | | p | 2 f t 2 d x { < 0 if  p , q [ 2 , 2 ]  with  p + q > 0 , > 0 if  p , q [ 2 , 2 ]  with  p + q < 0 ,

which proves part two and, consequently, the proof is completed. □

From the proof of Proposition 3.1, we see that f defined by (3.3) is an even function and f (x)<0 for x[0,2] given by (3.6). Let m=2 and cx=p(2,2). Then by Lemma 2.3 we immediately obtain the following.

Proposition 3.2 For fixed c(2,2), let min(2,22c)pmin(2,2c+2) and t(0,π/2), and let S ˜ (p,q,t) be defined by (3.2). Then the function p S ˜ (p,2cp,t) is decreasing on [2,c) and increasing on (c,2c+2] for c(2,0], and is increasing on [2c2,c) and decreasing on (c,2] for c(0,2).

By Propositions 3.1 and 3.2 we can obtain some new inequalities for trigonometric functions.

Corollary 3.1 For t(0,π/2), we have

sin t t cos t < 2 cos t + 1 3 <exp(tcott1)< 1 + cos t 2 < sin t t < ( 1 + cos t 2 ) 1 / 3 ,
(3.8)
cos 1 / 3 t < ( 2 3 cos t + 1 3 ) 1 / 2 < ( 1 2 cos t + 1 2 cos t 3 ) 3 / 5 cos 1 / 3 t < sin t t < ( cos t 5 cos 2 t 5 ) 5 / 3 < ( 2 3 cos t 2 + 1 3 ) 2 cos 1 / 3 t < cos 3 t 3 < exp ( t 1 + cos t sin t 2 ) < cos 4 t 4 .
(3.9)

Proof (i) By Proposition 3.2, S ˜ (p,2p,t) is increasing in p on [0,1) and decreasing on [1,2], we have

S ˜ (2,0,t)< S ˜ ( 3 2 , 1 2 , t ) < S ˜ (1,1,t).

Due to S ˜ (1,q,t) is decreasing in q on [2,2], we get

S ˜ (1,1,t)< S ˜ ( 1 , 1 2 , t ) < S ˜ (1,0,t)< S ˜ ( 1 , 1 2 , t ) .

Simplifying leads to (3.8).

  1. (ii)

    Similarly, since p S ˜ (p,1p,t) is increasing on [1,1/2) and decreasing on (1/2,2], we get

    S ˜ ( 2 , 1 , t ) < S ˜ ( 3 2 , 1 2 , t ) < S ˜ ( 4 3 , 1 3 , t ) < S ˜ ( 1 , 0 , t ) < S ˜ ( 4 5 , 1 5 , t ) < S ˜ ( 3 4 , 1 4 , t ) < S ˜ ( 2 3 , 1 3 , t ) < S ˜ ( 1 2 , 1 2 , t ) ,

while S ˜ ( 1 2 , 1 2 ,t)< S ˜ ( 1 2 , 1 4 ,t) follows by the monotonicity of S ˜ (1/2,q,t) in q on [2,2]. Simplifying yields inequalities (3.9). □

3.2 Definition of two-parameter sine means and examples

Being equipped with Propositions 3.1, 3.2, we can easily establish a family of means generated by (3.2). To this end, we have to prove the following statement.

Theorem 3.1 Let p,q[2,2], and let S ˜ (p,q,t) be defined by (3.2). Then, for all a,b>0, S p , q (a,b) defined by

S p , q ( a , b ) = max ( a , b ) × S ˜ ( p , q , arccos ( min ( a , b ) max ( a , b ) ) ) if  a b and  S p , q ( a , a ) = a
(3.10)

is a mean of a and b if and only if 0p+q3.

Proof Without lost of generality, we assume that 0<ab. Let t=arccos(a/b). Then the statement in question is equivalent to that the inequalities

cost S ˜ (p,q,t)1
(3.11)

hold for t(0,π/2) if and only if 0p+q3, where S ˜ (p,q,t) is defined by (3.2).

Necessity. We prove that the condition 0p+q3 is necessary. If (3.11) holds, then we have

lim t 0 + ln S ˜ ( p , q , t ) ln cos t t 2 0and lim t 0 + ln S ˜ ( p , q , t ) t 2 0.

Using power series extension gives

ln S ˜ ( p , q , t ) ln cos t = 1 6 t 2 ( p + q 3 ) + O ( t 4 ) , ln S ˜ ( p , q , t ) = 1 6 t 2 ( p + q ) + O ( t 4 ) .

Hence we have

1 6 (p+q3)0and 1 6 (p+q)0,

which implies that 0p+q3.

Sufficiency. We show that the condition 0p+q3 is sufficient. Clearly, max(p,q)0. Now we distinguish two cases to prove (3.11).

Case 1: p,q0 and p+q3. This case can be divided into two subcases. In the first subcase of (p,q)[0,2]×[0,1] or [0,1]×[0,2], by the monotonicity of S ˜ (p,q,t) in p, q on [2,2], we get

cost= S ˜ (2,1,t) S ˜ (p,q,t) S ˜ (0,0,t)=1.

In the second subcase of (p,q) [ 1 , 2 ] 2 and p+q3, it is derived that

S ˜ (p,3p,t) S ˜ (p,q,t) S ˜ (0,0,t)=1.

From Proposition 3.2 it is seen that S ˜ (p,3p,t) is increasing on [1,3/2] and decreasing on [3/2,2], which reveals that S ˜ (p,3p,t)> S ˜ (2,1,t)=cost, that is, the desired result.

Case 2: p0, q0 or p0, q0 and p+q3. Because of the symmetry of p and q, we assume that pq. Then p0, q0. Due to p[0,2] and p+q3, we have pmin(3q,2)=2. Using the monotonicity of S ˜ (p,q,t) in p, q on [2,2] again leads us to

S ˜ (p,q,t) S ˜ (2,0,t)= sin 2 t 2 t = sin t t cos t >cost.

On the other hand, from p+q0, that is, pq, it is acquired that

S ˜ (p,q,t) S ˜ (q,q,t)=1,

which proves Case 2 and the sufficiency is complete. □

Now we can give the definition of the two-parameter sine means as follows.

Definition 3.2 Let a,b>0 and p,q[2,2] such that 0p+q3, and let S ˜ (p,q,t) be defined by (3.2). Then S p , q (a,b) defined by (3.10) is called a two-parameter sine mean of a and b.

As a family of means, the two-parameter sine means contain many known and new means.

Example 3.1 Clearly, for 0<a<b, all the following

S 1 , 0 (a,b)= b 2 a 2 arccos ( a / b ) =SB(a,b),
(3.12)
S 1 , 1 (a,b)=bexp ( a S B ( a , b ) 1 ) ,
(3.13)
S 1 / 2 , 1 / 2 (a,b)=bexp ( a + b S B ( a , b ) 2 ) ,
(3.14)
S 3 / 2 , 1 / 2 (a,b)= b 1 / 2 ( 2 3 a + 1 3 b ) 1 / 2
(3.15)

are means of a and b, where SB(a,b) is the Schwab-Borchardt mean defined by (1.3).

To generate more means involving a two-parameter sine function, we need to note a simple fact: If M 1 , M 2 , M are means of distinct positive numbers x and y with M 1 < M 2 , then M( M 1 , M 2 ) is also a mean and satisfies inequalities

M 1 <M( M 1 , M 2 )< M 2 .

Applying the fact to Definition 3.2, we can obtain more means involving a two-parameter sine function, in which, as mentioned in Section 1, G, A and Q denote the geometric, arithmetic and quadratic means, respectively, and we have G<A<Q.

Example 3.2 Let (a,b)(G,A). Then both the following

S 1 , 0 ( G , A ) = S B ( G , A ) = a b 2 arcsin a b a + b = P ( a , b ) , S 1 , 1 ( G , A ) = A exp ( G P 1 ) = X ( a , b )

are means of a and b, where P=P(a,b) is the first Seiffert mean defined by (1.5) and X(a,b) is Sándor’s mean defined by (1.7). Also, they lie between G and A.

Example 3.3 Let (a,b)(G,Q). Then both the following

S 1 , 0 ( G , Q ) = S B ( G , Q ) = a b 2 arctan a b 2 a b : = U ( a , b ) , S 1 , 1 ( G , Q ) = Q exp ( G U 1 )
(3.16)

are means of a and b, and between G and Q.

It is interesting that the new mean U(a,b) is somewhat similar to the second Seiffert mean T(a,b).

Example 3.4 Let (a,b)(A,Q). Then both the following

S 1 , 0 ( A , Q ) = S B ( A , Q ) = a b 2 arctan a b a + b = T ( a , b ) , S 1 , 1 ( A , Q ) = Q exp ( A T 1 )

are means of a and b, where T=T(a,b) is the second Seiffert mean defined by (1.6). Moreover, they are between A and Q.

3.3 Properties of two-parameter sine means

From Propositions 3.1, 3.2 and Theorem 3.1, we easily obtain the properties of two-parameter sine means.

Property 3.1 The two-parameter sine means S p , q (a,b) are symmetric with respect to parameters p and q.

Property 3.2 The two-parameter sine means S p , q (a,b) are decreasing in p and q.

Property 3.3 The two-parameter sine means S p , q (a,b) are log-concave in (p,q) for p,q>0.

Property 3.4 The two-parameter sine means S p , q (a,b) are homogeneous and symmetric with respect to a and b.

Now we prove the monotonicity of two-parameter sine means in a and b.

Property 3.5 Suppose that 0<a<b. Then, for fixed b>0, the two-parameter sine means S p , q (a,b) are increasing in a on (0,b). For fixed a>0, they are increasing in b on (a,).

Proof (i) Let t=arccos(a/b). Then ln S p , q (a,b):=lnb+ln S ˜ (p,q,t). Differentiation yields

a ln S p , q (a,b)= t ln S ˜ (p,q,t)× t a = 1 b 2 a 2 t ln S ˜ (p,q,t),

which, by part two of Proposition 3.1, reveals that (ln S p , q (a,b))/a>0, that is, S p , q (a,b) is increasing in a on (0,b).

(ii) Now we prove the monotonicity of S p , q (a,b) in b. We have ln S p , q (a,b):=lnalncost+ln S ˜ (p,q,t). Differentiation leads to

b ln S p , q ( a , b ) = ( t ln S ˜ ( p , q , t ) + sin t cos t ) × t b = a b b 2 a 2 ( H ( p , q ) + tan t ) ,

where

H(p,q):= h ( p ) h ( q ) p q if pqandH(p,p)= h (p),
(3.17)

here

h(x)=xcottxif |x|(0,2]andh(0)= 1 t

is an even function on [2,2]. Hence, to prove (ln S p , q (a,b))/b>0, it suffices to prove that for p,q[2,2] with 0p+q3, the inequality H(p,q)+tant>0 is valid for t(0,π/2). Differentiation again gives

h ( x ) = 2 t sin t x t x cos t x sin 3 t x < 0 for  | x | ( 0 , 2 ]  by (2.3) and h ( 0 ) = 2 t 3 < 0 , h ( x ) = 6 t 3 x sin 4 t x ( 2 + cos 2 t x 3 sin 2 t x 2 t x ) < 0 for  x ( 0 , 2 ]  by (2.7) .

It follows by Lemmas 2.1 and 2.3 that H(p,q) is decreasing in p and q on [2,2] and H(p,3p) is increasing on [1,3/2) and decreasing on (3/2,2].

Next we distinguish two cases to prove H(p,q)>0 for p,q[2,2] with 0p+q3.

Case 1: p,q0 and p+q3. This case can be divided into two subcases. In the first subcase of (p,q)[0,2]×[0,1] or [0,1]×[0,2], by the monotonicity of H(p,q) in p, q on [2,2], we have

H(p,q)+tantH(2,1)+tant=2cot2tcott+tant=0.

In the second subcase of (p,q) [ 1 , 2 ] 2 and p+q3, it is derived from the monotonicities of H(p,q) and H(p,3p) that

H(p,q)+tantH(p,3p)+tantH(2,1)+tant=0.

Case 2: p0, q0 or p0, q0 and p+q3. Because of the symmetry of p and q, we assume that pq. Then p0, q0. This together with p,q[2,2] with p+q3 gives pmin(3q,2)=2. Therefore, we have

H(p,q)+tantH(2,0)+tant= 1 2 ( 2 cot 2 t 1 t ) +tant= 2 t sin 2 t 2 t sin 2 t >0,

which proves the monotonicity of S p , q (a,b) in b on (a,) and the proof is complete. □

Remark 3.1 Suppose that 0<a<b. Then, by the monotonicity of S p , q (a,b) in a and b, we see that

S 1 , 0 (G,A)< S 1 , 0 (G,Q)< S 1 , 0 (A,Q),

which implies that

P(a,b)<U(a,b)<T(a,b).
(3.18)

Similarly, we have

X(a,b)=Aexp ( G P 1 ) <Qexp ( G U 1 ) <Qexp ( A T 1 ) .

4 Two-parameter cosine means

4.1 Two-parameter cosine functions

In the same way, the Gini means defined by (1.2) can be expressed in hyperbolic functions by letting t=ln b / a :

G p , q (a,b)= a b Ch(p,q,t),

where

Ch(p,q,t)= { ( cosh p t cosh q t ) 1 / ( p q ) , p q , e t tanh p t , p = q 0 , 1 , p = q = 0 .
(4.1)

We call Ch(p,q,t) two-parameter hyperbolic cosine functions. Analogously, we can define the two-parameter cosine functions as follows.

Definition 4.1 The function C ˜ is called a two-parameter cosine function if C ˜ is defined on [ 1 , 1 ] 2 ×(0,π/2) by

C ˜ (p,q,t)= { ( cos p t cos q t ) 1 / ( p q ) , p q , e t tan p t , p = q 0 , 1 , p = q = 0 .
(4.2)

Similar to the proofs of Propositions 3.1 and 3.2, we give the following assertions without proofs.

Proposition 4.1 Let the two-parameter cosine function C ˜ be defined by (4.2). Then

  1. (i)

    C ˜ is decreasing in p, q on [1,1], and is log-concave in (p,q) for (p,q) [ 0 , 1 ] 2 and log-convex for (p,q) [ 1 , 0 ] 2 ;

  2. (ii)

    C ˜ is decreasing and log-concave in t on (0,π/2) for p+q>0, and is increasing and log-convex for p+q<0.

Proposition 4.2 For fixed c(1,1), let min(1,12c)pmin(1,1+2c) and t(0,π/2), and let C ˜ (p,q,t) be defined by (4.2). Then the function pC(p,2cp,t) is decreasing on [1,c) and increasing on (c,2c+1] for c(1,0], and is increasing on [2c1,c) and decreasing on (c,1] for c(0,1).

Propositions 4.1 and 4.2 also contain some new inequalities for trigonometric functions, as shown in the following corollary.

Corollary 4.1 For t(0,π/2), we have

cos 2 t 2 >exp ( t 1 cos t sin t ) > ( 2 cos t 2 1 ) 2 >cost.
(4.3)

Proof By Propositions 4.1 and 4.2, we see that C ˜ (1/2,q,t) is decreasing in q on [1,1] and C ˜ (p,1p,t) is decreasing in p on [0,1/2). It is obtained that

C ˜ ( 1 2 , 0 , t ) > C ˜ ( 1 2 , 1 2 , t ) > C ˜ ( 3 4 , 1 4 , t ) > C ˜ (1,0,t),

which by some simplifications yields the desired inequalities. □

4.2 Definition of two-parameter cosine means and examples

Similarly, by Propositions 4.1, 4.2, we can easily present a family of means generated by (4.2). Of course, we need to prove the following theorem.

Theorem 4.1 Let p,q[1,1], and let C ˜ (p,q,t) be defined by (4.2). Then, for all a,b>0, C p , q (a,b) defined by

C p , q ( a , b ) = max ( a , b ) × C ˜ ( p , q , arccos ( min ( a , b ) max ( a , b ) ) ) if  a b and  C p , q ( a , a ) = a
(4.4)

is a mean of a and b if and only if 0p+q1.

Proof We assume that 0<ab and let t=arccos(a/b). Then the desired assertion is equivalent to the inequalities

cost C ˜ (p,q,t)1
(4.5)

hold for t(0,π/2) if and only if 0p+q1, where C ˜ (p,q,t) is defined by (4.2).

Necessity. If (4.5) holds, then we have

lim t 0 + ln C ˜ ( p , q , t ) ln cos t t 2 0and lim t 0 + ln C ˜ ( p , q , t ) t 2 0.

Using power series extension gives

ln C ˜ ( p , q , t ) ln cos t = 1 2 t 2 ( p + q 1 ) + O ( t 4 ) , ln C ˜ ( p , q , t ) = 1 2 t 2 ( p + q ) + O ( t 4 ) .

Hence we have

1 2 (p+q1)0and 1 2 (p+q)0,

which yields 0p+q1.

Sufficiency. We show that the condition 0p+q1 is sufficient. Clearly, max(p,q)0. Now we distinguish two cases to prove (4.5).

Case 1: p,q0 and p+q1. By Proposition 4.1 it is obtained that

C ˜ (p,1p,t) C ˜ (p,q,t) C ˜ (0,0,t)=1.

From Proposition 4.2 it is seen that C ˜ (p,1p,t) is increasing on [0,1/2] and decreasing on [1/2,1], which yields C ˜ (p,1p,t)> C ˜ (1,0,t)=cost, which proves Case 1.

Case 2: p0, q0 or p0, q0 and p+q1. We assume that pq. Then p0, q0. Due to p[0,1] and p+q1, we have pmin(1q,1)=1. Using the monotonicity of C ˜ (p,q,t) in p, q on [1,1] gives

C ˜ (p,q,t) C ˜ (1,0,t)=cost.

At the same time, since p+q0, that is, pq, we have

C ˜ (p,q,t) C ˜ (q,q,t)=1,

which proves Case 2 and the sufficiency is complete. □

Thus the two-parameter cosine means can be defined as follows.

Definition 4.2 Let a,b>0 and p,q[1,1] such that 0p+q1, and let C ˜ (p,q,t) be defined by (4.2). Then C p , q (a,b) defined by (4.4) is called a two-parameter cosine mean of a and b.

The two-parameter cosine means similarly include many new means, for example, when 0<a<b,

C 1 / 2 , 1 / 2 (a,b)=bexp ( a b S B ( a , b ) )
(4.6)

is a mean, where SB(a,b) is the Schwab-Borchardt mean defined by (1.3).

Additionally, let (a,b)(G,A),(G,Q),(A,Q). Then all the following

C 1 / 2 , 1 / 2 ( G , A ) = A exp ( G A P ) , C 1 / 2 , 1 / 2 ( G , Q ) = Q exp ( G Q U ) , C 1 / 2 , 1 / 2 ( A , Q ) = Q exp ( A Q T )

are means of a and b, where P, T are the first and second Seiffert mean defined by (1.5) and (1.6), U is defined by (3.16), and they lie between G and A, G and Q, A and Q, respectively.

4.3 Properties of two-parameter cosine means

From Propositions 4.1, 4.2 and Theorem 4.1, we can deduce the properties of two-parameter cosine means as follows.

Property 4.1 C p , q (a,b) are symmetric with respect to parameters p and q.

Property 4.2 C p , q (a,b) are decreasing in p and q.

Property 4.3 C p , q (a,b) are log-concave in (p,q) for p,q>0.

Property 4.4 C p , q (a,b) are homogeneous and symmetric with respect to a and b.

Property 4.5 Suppose that 0<a<b. Then, for fixed b>0, the two-parameter cosine means C p , q (a,b) are increasing in a on (0,b). For fixed a>0, they are increasing in b on (a,).

The proof of Property 4.5 is similar to that of Property 3.5, which is left to readers.

Remark 4.1 Assume that 0<a<b. Then employing the monotonicity of C p , q (a,b) in a and b, we have

G<Aexp ( G A P ) <Qexp ( G Q U ) <Qexp ( A Q T ) <Q.

5 Two-parameter tangent means

5.1 Two-parameter tangent functions

Now we define the two-parameter tangent function and prove its properties, proofs of which are also the same as those of Propositions 3.1 and 3.2.

Definition 5.1 The function T ˜ is called a two-parameter tangent function if T ˜ is defined on [ 1 , 1 ] 2 ×(0,π/2) by

T ˜ (p,q,t)= { ( q p tan p t tan q t ) 1 / ( p q ) , p q ( p q ) 0 , ( tan p t p t ) 1 / p , q = 0 , p 0 , ( tan q t q t ) 1 / q , p = 0 , q 0 , exp ( 2 t sin 2 p t 1 p ) , p = q 0 , 1 , p = q = 0 .
(5.1)

Proposition 5.1 Let the two-parameter tangent function T ˜ be defined by (5.1). Then

  1. (i)

    T ˜ is increasing in p, q on [1,1], and is log-convex in (p,q) for p,q>0 and log-convex for p,q<0;

  2. (ii)

    T ˜ is increasing and log-convex in t for p+q>0, and is decreasing and log-concave for p+q<0.

Proof We have

ln T ˜ (p,q,t)= g ( p ) g ( q ) p q if pqandln T ˜ (p,p,t)= g (p),

where

g(x)=ln tan t x x if |x|(0,1]andg(0)=lnt.
(5.2)

(i) To prove part one, by Lemmas 2.1 and 2.2 it suffices to check that g is convex on [1,1] and g is convex on [0,1]. In fact, differentiation and application of (2.8) yield

g (x)= { 2 t sin 2 t x 1 x if  | x | ( 0 , 1 ] , 0 if  x = 0 = n = 1 ( 2 2 n 2 ) 2 2 n ( 2 n ) ! | B 2 n | t 2 n x 2 n 1 .
(5.3)

Differentiation again gives

g ( x ) = n = 1 ( 2 n 1 ) ( 2 2 n 2 ) 2 2 n ( 2 n ) ! | B 2 n | t 2 n x 2 n 2 > 0 for  x [ 1 , 1 ] , g ( x ) = n = 2 ( 2 n 1 ) ( 2 n 2 ) ( 2 2 n 2 ) 2 2 n ( 2 n ) ! | B 2 n | t 2 n x 2 n 3 { 0 if  x [ 0 , 1 ] , 0 if  x [ 1 , 0 ] .

Thus part one is proved.

(ii) For proving part two, we have to check that g /t0 and 2 g / t 2 0 for x[0,1]. Differentiating g (x) given in (5.3) for t, we have

g t = n = 1 2 n ( 2 2 n 2 ) 2 2 n ( 2 n ) ! | B 2 n | t 2 n 1 x 2 n 1 0 for  x [ 0 , 1 ] , 2 g t 2 = n = 1 2 n ( 2 n 1 ) ( 2 2 n 2 ) 2 2 n ( 2 n ) ! | B 2 n | t 2 n 2 x 2 n 1 0 for  x [ 0 , 1 ] .

In the same method as the proof of part two in Proposition 3.1, part two in this proposition easily follows.

This completes the proof. □

The following proposition is a consequence of Lemma 2.3, the proof of which is also the same as that of Proposition 3.2 and is left to readers.

Proposition 5.2 For fixed c(1,1), let min(1,12c)pmin(1,1+2c) and t(0,π/2), and let T ˜ (p,q,t) be defined by (5.1). Then the function p T ˜ (p,2cp,t) is increasing on [1,c) and decreasing on (c,1+2c] for c(1,0], and is decreasing on [2c1,c) and increasing on (c,1] for c(0,1).

As an application of Propositions 5.1 and 5.2, we give the following corollary.

Corollary 5.1 For t(0,π/2), we have

( sin t t cos 2 t 2 ) 2 <exp ( 2 t sin t 2 ) < ( 2 3 cos t 2 + 1 3 ) 4 ( 2 3 cos t + 1 3 ) 2 < tan t t .
(5.4)

Proof Propositions 5.1 and 5.2 indicate that T ˜ (1/2,q,t) is increasing in q on [1,1] and T ˜ (p,1p,t) is decreasing in p on [0,1/2) and increasing on (1/2,1]. It follows that

T ˜ ( 1 2 , 0 , t ) < T ˜ ( 1 2 , 1 2 , t ) < T ˜ ( 3 4 , 1 4 , t ) < T ˜ (1,0,t),

which, by some simplifications, yields the required inequalities. □

5.2 Definition of two-parameter tangent means and examples

Before giving the definition of two-parameter tangent means, we firstly prove the following statement.

Theorem 5.1 Let p,q[1,1], and let T ˜ (p,q,t) be defined by (5.1). Then, for all a,b>0, T p , q (a,b) defined by

T p , q ( a , b ) = min ( a , b ) × T ˜ ( p , q , arccos ( min ( a , b ) max ( a , b ) ) ) if  a b and T p , q ( a , a ) = a
(5.5)

is a mean of a and b if 0p+q1.

Proof We assume that 0<ab and let t=arccos(a/b). Then T p , q (a,b) is a mean of a and b if and only if the inequalities

1 T ˜ (p,q,t) 1 cos t

hold for t(0,π/2), where T ˜ (p,q,t) is defined by (5.1). Similarly, it can be divided into two cases.

Case 1: p,q0 and p+q1. From the monotonicity of T p , q (a,b) in p, q on [1,1], it is deduced that

1= T ˜ (0,0,t) T ˜ (p,q,t) T ˜ (p,1p,t).

By Proposition 5.2 we can see that T ˜ (p,1p,t) is decreasing on [0,1/2] and increasing on [1/2,1], which yields

T ˜ (p,1p,t)< T ˜ (1,0,t)= tan t t = sin t t 1 cos t < 1 cos t ,

that is, the desired result.

Case 2: p0, q0 or p0, q0 and p+q1. We assume that pq. Analogously, there must be pmin(1q,1)=1. Using the monotonicity of T ˜ (p,q,t) in p, q on [1,1] gives

T ˜ (p,q,t) T ˜ (1,0,t)= tan t t < 1 cos t .

Noticing that p+q0, that is, pq, we have

T ˜ (p,q,t) T ˜ (q,q,t)=1,

which proves Case 2 and the proof is finished. □

We are now in a position to define the two-parameter tangent means by (5.1).

Definition 5.2 Let a,b>0 and p,q[1,1] such that 0p+q1, and let T ˜ (p,q,t) be defined by (5.1). Then T p , q (a,b) defined by (5.5) is called a two-parameter tangent mean of a and b.

Here are some examples of two-parameter tangent means.

Example 5.1 For 0<a<b, both the following

T 1 , 0 (a,b)= b 2 a 2 arccos ( a / b ) =SB(a,b),
(5.6)
T 1 / 2 , 1 / 2 (a,b)=aexp ( 2 b S B ( a , b ) 2 )
(5.7)

are means of a and b, where SB(a,b) is the Schwab-Borchardt mean defined by (1.3).

Example 5.2 Let (a,b)(G,A),(G,Q),(A,Q). Then all the following

T 1 / 2 , 1 / 2 ( G , A ) = G exp ( 2 A P 2 ) , T 1 / 2 , 1 / 2 ( G , Q ) = G exp ( 2 Q U 2 ) , T 1 / 2 , 1 / 2 ( A , Q ) = A exp ( 2 Q T 2 )

are means of a and b, where P, T are the first and second Seiffert mean defined by (1.5) and (1.6), U is defined by (3.16). Also, they lie between G and A, G and Q, A and Q, respectively.

5.3 Properties of two-parameter tangent means

From Propositions 5.1 and 5.2 and Theorem 5.1, we see that the properties of two-parameter tangent means are similar to those of sine ones.

Property 5.1 T p , q (a,b) is symmetric with respect to parameters p and q.

Property 5.2 T p , q (a,b) is increasing in p and q.

Property 5.3 T p , q (a,b) is log-convex in (p,q) for p,q>0.

Property 5.4 T p , q (a,b) is homogeneous and symmetric with respect to a and b.

Now we prove the monotonicity of two-parameter trigonometric means in a and b.

Property 5.5 Let 0<a<b. Then, for fixed a>0, the two-parameter tangent mean T p , q (a,b) is increasing in b on (a,). For fixed b>0, the two-parameter tangent mean T p , q (a,b) is increasing in a on (0,b).

Proof (i) Let t=arccos(a/b). Then ln T p , q (a,b):=lna+ln T ˜ (p,q,t). Differentiation yields

b ln T p , q (a,b)= t ln T ˜ (p,q,t)× t b = a b 2 a 2 t ln T ˜ (p,q,t).

Application of Proposition 5.1 yields (ln T p , q (a,b))/b>0, which proves part one.

  1. (ii)

    Now we prove the monotonicity of T p , q (a,b) in a. Since ln T p , q (a,b) can be written as ln T p , q (a,b)=lnb+ln T ˜ (p,q,t)+lncost, we have

    a ln T p , q ( a , b ) = ( t ln T ˜ ( p , q , t ) sin t cos t ) × t a : = 1 b 2 a 2 ( J ( p , q ) tan t ) ,

where

J(p,q)= j ( p ) j ( q ) p q if pqandJ(p,p)= j (p),
(5.8)

here

j(x)= 2 x sin 2 x t if |x|(0,1]andj(0)= 1 t

is even on [1,1]. Thus, to prove (ln T p , q (a,b))/a>0, it suffices to prove that for p,q[1,1] with 0p+q1, the inequality J(p,q)tant<0 holds for t(0,π/2).

Utilizing (2.8) and differentiating again give

j ( x ) = 1 t + n = 1 ( 2 2 n 2 ) 2 2 n ( 2 n ) ! | B 2 n | t 2 n 1 x 2 n , j ( x ) = n = 1 2 n ( 2 n 1 ) ( 2 2 n 2 ) 2 2 n ( 2 n ) ! | B 2 n | t 2 n 1 x 2 n 2 > 0 for  x [ 1 , 1 ] , j ( x ) = n = 2 2 n ( 2 n 1 ) ( 2 n 2 ) ( 2 2 n 2 ) 2 2 n ( 2 n ) ! | B 2 n | t 2 n 1 x 2 n 3 0 for  x [ 0 , 1 ] .

By Lemmas 2.1 and 2.3 we see that J(p,q) is increasing in p and q on [1,1], and J(p,1p) is decreasing on [0,1/2) and increasing on (1/2,1].

Now we distinguish two cases to prove J(p,q)tant<0 for p,q[1,1] with 0p+q1.

Case 1: p,q0 and p+q1. By the monotonicity of J(p,q) in p, q on [1,1] and of J(p,1p) in p on [0,1], we have

J ( p , q ) tan t J ( p , 1 p ) tan t J ( 1 , 0 ) tan t = 2 sin 2 t 1 t sin t cos t = cos t sin t 1 t < 0 .

Case 2: p0, q0 or p0, q0 and p+q1. We assume that pq. Then p0, q0 and pmin(1q,1)=1. Therefore, we get

J(p,q)tantJ(1,0)tant<0,

which proves the monotonicity of T p , q (a,b) in a on (0,b). □

Remark 5.1 Utilizing the monotonicity property, we have

T 1 / 2 , 1 / 2 (G,A)< T 1 / 2 , 1 / 2 (G,Q)< T 1 / 2 , 1 / 2 (A,Q),

which indicates that

Gexp ( 2 A P 2 ) <Gexp ( 2 Q U 2 ) <Aexp ( 2 Q T 2 ) .

Additionally, T p , q (a,b) has a unique property which shows the relation among two-parameter sine, cosine and tangent means.

Property 5.6 For 0<a<b, if 0p+q1, then T p , q (a,b)=a S p , q (a,b)/ C p , q (a,b). In particular, T 1 , 0 (a,b)= S 1 , 0 (a,b)=SB(a,b).

6 Inequalities for two-parameter trigonometric means

As shown in the previous sections-, by using Propositions 3.1-5.2 we can establish a series of new inequalities for trigonometric functions and reprove some known ones. However, we are more interested in how to establish new inequalities for two-parameter trigonometric means from these ones derived by using Propositions 3.1-5.2, as obtaining an inequality for bivariate mans from the corresponding one for hyperbolic functions (see [2224, 30]). In fact, Neuman also offered some successful examples (see [31]).

The inequalities involving Schwab-Borchardt mean SB are mainly due to Neuman and Sándor (see [3133]), and Witkowski [34] also has some contributions to them. More often, however, inequalities for means constructed by trigonometric functions seem to be related to the first and second Seiffert means, see [11, 33, 3555]. In this section, we establish some new inequalities for two-parameter trigonometric means by using their monotonicity and log-convexity. Our steps are as follows.

Step 1: Obtaining an inequality (I1) for trigonometric functions sint, cost and tant by using the monotonicity and log-convexity of two-parameter trigonometric functions and simplifying them.

Step 2: For 0<a<b, letting t=arccos(a/b) in inequalities (I1) obtained in Step 1 and next multiplying both sides by b or a and simplifying yield an inequality (I2) for means involving trigonometric functions.

Step 3: Let m=m(a,b) and M=M(a,b) be two means of a and b with m(a,b)<M(a,b) for all a,b>0. Making a change of variables am(a,b) and bM(a,b) leads to another inequality (I3) for means involving trigonometric functions.

Now we illustrate these steps.

Example 6.1

Step 1: For t(0,π/2), we have (3.8).

Step 2: For 0<a<b, letting t=arccos(a/b) and next multiplying each side of (3.8) by b and simplifying yield

a S B ( a , b ) < 2 a + b 3 <b e a / S B ( a , b ) 1 < a + b 2 <SB(a,b)< b 2 / 3 ( a + b 2 ) 1 / 3 .

Step 3: With (a,b)(G,A) yields

G P < 2 G + A 3 <A e G / P 1 < A + G 2 <P< A 2 / 3 ( A + G 2 ) 1 / 3 .

With (a,b)(G,Q) yields

G U < 2 G + Q 3 <Q e G / U 1 < G + Q 2 <U< Q 2 / 3 ( G + Q 2 ) 1 / 3 .

With (a,b)(A,Q) yields

A T < 2 A + Q 3 <Q e A / T 1 < A + Q 2 <T< Q 2 / 3 ( A + Q 2 ) 1 / 3 .

Example 6.2

Step 1: For t(0,π/2), from (3.9) it is derived that

cos 1 / 3 t < ( 2 3 cos t + 1 3 ) 1 / 2 < sin t t < ( 2 3 cos t 2 + 1 3 ) 2 < e t ( 1 + cos t ) / sin t 2 < ( 1 2 + 1 2 cos t 2 ) 2 .
(6.1)

Step 2: For 0<a<b, letting t=arccos(a/b) and next multiplying each side of (6.1) by b and simplifying yield

b 2 / 3 a 1 / 3 < b 1 / 2 ( 2 3 a + 1 3 b ) 1 / 2 < S B ( a , b ) < ( 2 3 a + b 2 + b 3 ) 2 < b e ( a + b ) / S B ( a , b ) 2 .
(6.2)

Step 3: With (a,b)(G,A) yields

A 2 / 3 G 1 / 3 < A 1 / 2 ( 2 3 G + 1 3 A ) 1 / 2 <P< ( 2 3 A + G 2 + 1 3 A ) 2 <A e ( A + G ) / P 2 .

With (a,b)(G,Q) yields

Q 2 / 3 G 1 / 3 < Q 1 / 2 ( 2 3 G + 1 3 Q ) 1 / 2 <U< ( 2 3 G + Q 2 + 1 3 Q ) 2 <Q e ( G + Q ) / U 2 .

With (a,b)(A,Q) yields

Q 2 / 3 A 1 / 3 < Q 1 / 2 ( 2 3 A + 1 3 Q ) 1 / 2 <T< ( 2 3 A + Q 2 + 1 3 Q ) 2 <Q e ( A + Q ) / T 2 .

Example 6.3

Step 1: For t(0,π/2), we have (4.3).

Step 2: For 0<a<b, letting t=arccos(a/b) and next multiplying each side of (4.3) by b and simplifying yield

a + b 2 >b e ( a b ) / S B ( a , b ) > ( 2 a + b 2 b ) 2 >a.
(6.3)

Step 3: With (a,b)(G,A) yields

A + G 2 >A e ( G A ) / P > ( 2 A + G 2 A ) 2 >G.
(6.4)

With (a,b)(G,Q),(A,Q) can yield corresponding inequalities.

Remark 6.1 From inequalities (6.3) it is derived that

L(a,b)<SB(a,b)<2L ( a + b 2 , b )
(6.5)

hold for 0<a<b, where L(x,y) is the logarithmic mean of positive numbers x and y. The first inequality of (6.5) follows from the relation between the second and fourth terms, that is, bexp((ab)/SB(a,b))>a, while the second one is obtained by the first one in (6.3).

Example 6.4

Step 1: For t(0,π/2), we have (5.4).

Step 2: For 0<a<b, letting t=arccos(a/b) and next multiplying each side of (5.4) by a and simplifying give

a ( S B ( a , b ) a + b 2 ) 2 <a e 2 b / S B ( a , b ) 2 <a ( 2 3 a + b 2 + 1 3 b ) 4 ( 2 3 a + 1 3 b ) 2 <SB(a,b).
(6.6)

Step 3: With (a,b)(G,A) yields

G ( 2 P A + G ) 2 <G e 2 A / P 2 <G ( 2 3 A + G 2 + 1 3 A ) 4 ( 2 3 G + 1 3 A ) 2 <P.
(6.7)

With (a,b)(G,Q),(A,Q), we can derive corresponding inequalities.

Remark 6.2 Applying our method in establishing inequalities for means to certain known ones involving trigonometric functions, we can obtain corresponding inequalities which are possibly related to means. For example, the Wilker inequality states that for t(0,π/2),

( sin t t ) 2 + tan t t >2.

If for 0<a<b, put t=arccos(a/b), then we have

( S B ( a , b ) b ) 2 + S B ( a , b ) a >2.

In a similar way, the following inequalities

( 1 + cos t 2 ) 2 / 3 < sin t t < 2 + cos t 3 ,t(0,π/2)
(6.8)

can be changed into

b 1 / 3 ( a + b 2 ) 2 / 3 <SB(a,b)< 2 b + a 3 ,

by letting t=arccos(a/b) for 0<a<b, where the left inequality in (6.8) is due to Neuman and Sándor [[56], (2.5)] (also see [5759]) and the right one is known as Cusa’s inequality.

Remark 6.3 The third inequality in (6.1) is clearly superior to Cusa’s inequality (the right one of (6.8)) because

( 2 3 cos t 2 + 1 3 ) 2 < 2 3 cos 2 t 2 + 1 3 = 2 + cos t 3 .

While the second one in (6.1) is weaker than the first one in (6.8) since

( 2 3 cos t + 1 3 ) 3 ( 1 + cos t 2 ) 4 = 1 432 ( 34 cos t + 27 cos 2 t + 11 ) ( cos t 1 ) 2 <0.

7 Families of two-parameter hyperbolic means

After three families of two-parameter trigonometric means have been successfully constructed, we are encouraged to establish further two-parameter means of a hyperbolic version. They are included in the following theorems.

Theorem 7.1 Let p,qR, and let Sh(p,q,t) be defined by (3.1). Then, for all a,b>0, S h p , q (b,a) defined by

S h p , q ( b , a ) = min ( a , b ) × S h ( p , q , arccosh ( max ( a , b ) min ( a , b ) ) ) if  a b and  S h p , q ( a , a ) = a
(7.1)

is a mean of a and b if and only if

p + q 3 and L ( p , q ) 1 ln 2 , if  p , q > 0 , 0 p + q 3 , otherwise.

Theorem 7.2 Let p,qR, and let Ch(p,q,t) be defined by (4.1). Then, for all a,b>0, C h p , q (b,a) defined by

C h p , q ( b , a ) = min ( a , b ) × C h ( p , q , arccosh ( max ( a , b ) min ( a , b ) ) ) if  a b and  C h p , q ( a , a ) = a
(7.2)

is a mean of a and b if and only if 0p+q1.

To prove the above theorems, it suffices to use comparison theorems given in [15, 16] by Páles because both S h p , q (b,a) and C h p , q (b,a) are means if and only if

S h ( 0 , 0 , t ) = 1 S h ( p , q , t ) cosh t = S h ( 2 , 1 , t ) , C h ( 0 , 0 , t ) = 1 C h ( p , q , t ) cosh t = C h ( 1 , 0 , t ) ,

respectively. Here we omit further details.

The monotonicities and log-convexities of S h p , q (b,a) and of C h p , q (b,a) in parameters p and q are clearly the same as those of Sh(p,q,t) and of Ch(p,q,t), which are in turn equivalent to those of Stolarsky means defined by (1.1) and of Gini means defined by (1.2), respectively. These properties can be found in [13, 14, 1719, 21].

The above theorems indicate that for 0<a<b, all the following

S h 1 , 0 ( b , a ) = a sinh t t = b 2 a 2 arccosh ( b / a ) = S B ( b , a ) , S h 1 , 1 ( b , a ) = a exp ( t coth t 1 ) = a exp ( b S B ( b , a ) 1 ) , C h 1 / 2 , 1 / 2 ( b , a ) = a exp ( t tanh t 2 ) = a exp ( t cosh t 1 sinh t ) = a exp ( b a S B ( b , a ) )

are means of a and b, where SB(b,a) is the Schwab-Borchardt mean defined by (1.3).

It is easy to verify that

S h 1 , 0 ( A , G ) = S B ( A , G ) = b a ln b ln a = L ( b , a ) , S h 1 , 1 ( A , G ) = G exp ( A L 1 ) = I ( b , a ) , C h 1 / 2 , 1 / 2 ( A , G ) = G exp ( A G L ) = Z 1 / 2 ( b , a ) ,

where L(b,a) and I(b,a) are logarithmic and identric means, respectively, while Z p (b,a)= Z 1 / p ( a p , b p ) is the p-order power-exponential mean. Also, all the following

S h 1 , 0 ( Q , G ) = S B ( Q , G ) = a b 2 arcsinh a b 2 a b : = V ( a , b ) , S h 1 , 1 ( Q , G ) = G exp ( Q V 1 ) , C h 1 / 2 , 1 / 2 ( Q , G ) = G exp ( Q G V )

are means lying in G and Q. Likewise, all the following

S h 1 , 0 ( Q , A ) = S B ( Q , A ) = a b 2 arcsinh a b a + b = N S ( a , b ) , S h 1 , 1 ( Q , A ) = A exp ( Q N S 1 ) , C h 1 / 2 , 1 / 2 ( Q , A ) = A exp ( Q A N S )

are also means between A and Q, where NS(a,b) is the Neuman-Sándor mean defined by (1.4).

It should be noted that the new mean V(a,b) is similar to NS(a,b).

Similar to (5.1), for p,qR, we can define the two-parameter hyperbolic tangent function as follows:

Th(p,q,t)= { ( q p tanh p t tanh q t ) 1 / ( p q ) , p q ( p q ) 0 , ( tanh p t p t ) 1 / p , q = 0 , p 0 , ( tanh q t q t ) 1 / q , p = 0 , q 0 , exp ( 2 t sinh 2 p t 1 p ) , p = q 0 , 1 , p = q = 0 .
(7.3)

By some verifications, however, Th(p,q,t) does not have good properties like monotonicity in parameters p and q, and therefore, we fail to define a family of two-parameter hyperbolic tangent means. However, for certain p, q and 0<a<b, it is showed that bTh(p,q,arccosh(b/a)) is a mean of a and b, for example,

b×Th ( 1 , 0 , arccosh ( b / a ) ) =SB(b,a)

is clearly a mean of a and b. It is also proved that

b×Th ( 1 2 , 1 2 , arccosh ( b / a ) )

is also a mean of a and b. For this reason, we pose an open problem as the end of this paper.

Problem 7.1 Let 0<a<b, and let Th(p,q,t) be defined by (7.3). Finding p, q such that b×Th(p,q,arccosh(b/a)) is a mean of a and b.

References

  1. Stolarsky KB: Generalizations of the logarithmic mean. Math. Mag. 1975, 48: 87–92. 10.2307/2689825

    Article  MathSciNet  MATH  Google Scholar 

  2. Gini C: Di una formula comprensiva delle medie. Metron 1938, 13: 3–22.

    MATH  Google Scholar 

  3. Brenner JL, Carlson BC: Homogeneous mean values: weights and asymptotes. J. Math. Anal. Appl. 1987, 123: 265–280. 10.1016/0022-247X(87)90308-8

    Article  MathSciNet  MATH  Google Scholar 

  4. Carlson BC: Algorithms involving arithmetic and geometric means. Am. Math. Mon. 1971, 78: 496–505. 10.2307/2317754

    Article  MathSciNet  MATH  Google Scholar 

  5. Neuman E, Sándor J: On the Schwab-Borchardt mean. Math. Pannon. 2003, 14(2):253–266.

    MathSciNet  MATH  Google Scholar 

  6. Toader G: Some mean values related to the arithmetic-geometric mean. J. Math. Anal. Appl. 1998, 218(2):358–368. 10.1006/jmaa.1997.5766

    Article  MathSciNet  MATH  Google Scholar 

  7. Sándor, J, Toader, G: On some exponential means. Seminar on Mathematical Analysis (Cluj-Napoca, 1989–1990), Preprint, 90, “Babes-Bolyai” Univ., Cluj, 35–40 (1990)

  8. Sándor J, Toader G: On some exponential means. Part II. Int. J. Math. Math. Sci. 2006., 2006: Article ID 51937

    Google Scholar 

  9. Seiffert HJ: Werte zwischen dem geometrischen und dem arithmetischen Mittel zweier Zahlen. Elem. Math. 1987, 42: 105–107.

    MathSciNet  MATH  Google Scholar 

  10. Seiffert HJ: Aufgabe 16. Die Wurzel 1995, 29: 221–222.

    Google Scholar 

  11. Sándor, J: Trigonometric and hyperbolic inequalities. Available online at arXiv:1105.0859 (2011). e-printatarXiv.org

  12. Sándor J: Two sharp inequalities for trigonometric and hyperbolic functions. Math. Inequal. Appl. 2012, 15(2):409–413.

    MathSciNet  MATH  Google Scholar 

  13. Leach EB, Sholander MC: Extended mean values. Am. Math. Mon. 1978, 85: 84–90. 10.2307/2321783

    Article  MathSciNet  MATH  Google Scholar 

  14. Leach EB, Sholander MC: Extended mean values II. J. Math. Anal. Appl. 1983, 92: 207–223. 10.1016/0022-247X(83)90280-9

    Article  MathSciNet  MATH  Google Scholar 

  15. Páles Z: Inequalities for sums of powers. J. Math. Anal. Appl. 1988, 131: 265–270. 10.1016/0022-247X(88)90204-1

    Article  MathSciNet  MATH  Google Scholar 

  16. Páles Z: Inequalities for differences of powers. J. Math. Anal. Appl. 1988, 131: 271–281. 10.1016/0022-247X(88)90205-3

    Article  MathSciNet  MATH  Google Scholar 

  17. Qi F: Logarithmic convexities of the extended mean values. Proc. Am. Math. Soc. 2002, 130(6):1787–1796. 10.1090/S0002-9939-01-06275-X

    Article  MATH  Google Scholar 

  18. Yang Z-H: On the homogeneous functions with two parameters and its monotonicity. J. Inequal. Pure Appl. Math. 2005., 6(4): Article ID 101. Available online at http://www.emis.de/journals/JIPAM/images/155_05_JIPAM/155_05.pdf

  19. Yang Z-H: On the log-convexity of two-parameter homogeneous functions. Math. Inequal. Appl. 2007, 10(3):499–516.

    MathSciNet  MATH  Google Scholar 

  20. Yang Z-H: On the monotonicity and log-convexity of a four-parameter homogeneous mean. J. Inequal. Appl. 2008., 2008: Article ID 149286. Available online at http://www.journalofinequalitiesandapplications.com/content/2008/1/149286

    Google Scholar 

  21. Yang Z-H: The log-convexity of another class of one-parameter means and its applications. Bull. Korean Math. Soc. 2012, 49(1):33–47. Available online at http://www.mathnet.or.kr/mathnet/thesis_file/BKMS-49–1-33–47.pdf 10.4134/BKMS.2012.49.1.033

    Article  MathSciNet  MATH  Google Scholar 

  22. Yang Z-H: New sharp bounds for logarithmic mean and identric mean. J. Inequal. Appl. 2013. 10.1186/1029-242X-2013-116

    Google Scholar 

  23. Zhu L: Generalized Lazarevićs inequality and its applications - Part II. J. Inequal. Appl. 2009., 2009: Article ID 379142

    Google Scholar 

  24. Zhu L: New inequalities for hyperbolic functions and their applications. J. Inequal. Appl. 2012. 10.1186/1029-242X-2012-303

    Google Scholar 

  25. Bullen PS: Handbook of Means and Their Inequalities. Kluwer Academic, Dordrecht; 2003.

    Book  MATH  Google Scholar 

  26. Merkle M: Conditions for convexity of a derivative and some applications to the Gamma function. Aequ. Math. 1998, 55: 273–280. 10.1007/s000100050036

    Article  MathSciNet  MATH  Google Scholar 

  27. Mitrinović DS: Analytic Inequalities. Springer, Berlin; 1970.

    Book  MATH  Google Scholar 

  28. Mitrinović DS: Elementary Inequalities. Noordhoff, Groningen; 1964.

    MATH  Google Scholar 

  29. Group of compilation: Handbook of Mathematics. Peoples’ Education Press, Beijing; 1979. (Chinese)

    Google Scholar 

  30. Yang Z-H: New sharp bounds for identric mean in terms of logarithmic mean and arithmetic mean. J. Math. Inequal. 2012, 6(4):533–543. 10.7153/jmi-06-51

    Article  MathSciNet  MATH  Google Scholar 

  31. Neuman E: Inequalities for the Schwab-Borchardt mean and their applications. J. Math. Inequal. 2011, 5(4):601–609.

    Article  MathSciNet  MATH  Google Scholar 

  32. Neuman E, Sándor J: On certain means of two arguments and their extensions. Int. J. Math. Math. Sci. 2003, 16: 981–993.

    Article  MathSciNet  MATH  Google Scholar 

  33. Neuman E, Sándor J: On the Schwab-Borchardt mean. Math. Pannon. 2006, 17(1):49–59.

    MathSciNet  MATH  Google Scholar 

  34. Witkowski A: Interpolations of Schwab-Borchardt mean. Math. Inequal. Appl. 2012, 16(1):193–206.

    MathSciNet  MATH  Google Scholar 

  35. Jagers AA: Solution of problem 887. Nieuw Arch. Wiskd. 1994, 12(2):30–31.

    Google Scholar 

  36. Sándor J: On certain inequalities for means III. Arch. Math. 2001, 76: 34–40. 10.1007/s000130050539

    Article  MathSciNet  MATH  Google Scholar 

  37. Hästö PA: A monotonicity property of ratios of symmetric homogeneous means. J. Inequal. Pure Appl. Math. 2002., 3(5): Article ID 71

  38. Sándor J, Neuman E: On certain means of two arguments and their extensions. Int. J. Math. Math. Sci. 2003, 2003(16):981–993. 10.1155/S0161171203208103

    Article  MathSciNet  MATH  Google Scholar 

  39. Chu Y-M, Qiu Y-F, Wang M-K: Sharp power mean bounds for the combination of Seiffert and geometric means. Abstr. Appl. Anal. 2010., 2010: Article ID 108920

    Google Scholar 

  40. He D, Shen Z-J: Advances in research on Seiffert mean. Commun. Inequal. Res. 2010., 17(4): Article ID 26. Available online at http://old.irgoc.org/Article/UploadFiles/201010/20101026104515652.pdf

  41. Wang S-S, Chu Y-M: The best bounds of the combination of arithmetic and harmonic means for the Seiffert’s mean. Int. J. Math. Anal. 2010, 4(21–24):1079–1084.

    MathSciNet  MATH  Google Scholar 

  42. Wang M-K, Qiu Y-F, Chu Y-M: Sharp bounds for Seiffert means in terms of Lehmer means. J. Math. Inequal. 2010, 4(4):581–586.

    Article  MathSciNet  MATH  Google Scholar 

  43. Chu Y-M, Qiu Y-F, Wang M-K, Wang G-D: The optimal convex combination bounds of arithmetic and harmonic means for the Seiffert’s mean. J. Inequal. Appl. 2010., 2010: Article ID 436457

    Google Scholar 

  44. Liu H, Meng X-J: The optimal convex combination bounds for Seiffert’s mean. J. Inequal. Appl. 2011., 2011: Article ID 686834

    Google Scholar 

  45. Chu Y-M, Wang M-K, Gong W-M: Two sharp double inequalities for Seiffert mean. J. Inequal. Appl. 2011., 2011: Article ID 44 10.1186/1029-242X-2011-44

    Google Scholar 

  46. Chu Y-M, Hou S-W: Sharp bounds for Seiffert mean in terms of contraharmonic mean. Abstr. Appl. Anal. 2012., 2012: Article ID 425175

    Google Scholar 

  47. Chu Y-M, Long B-Y, Gong W-M, Song Y-Q: Sharp bounds for Seiffert and Neuman-Sándor means in terms of generalized logarithmic means. J. Inequal. Appl. 2013., 2013: Article ID 10. Available online at http://www.journalofinequalitiesandapplications.com/content/2013/1/10

    Google Scholar 

  48. Jiang W-D, Qi F: Some sharp inequalities involving Seiffert and other means and their concise proofs. Math. Inequal. Appl. 2012, 15(4):1007–1017.

    MathSciNet  MATH  Google Scholar 

  49. Hästö PA: Optimal inequalities between Seiffert’s mean and power mean. Math. Inequal. Appl. 2004, 7(1):47–53.

    MathSciNet  MATH  Google Scholar 

  50. Costin I, Toader G: A nice separation of some Seiffert type means by power means. Int. J. Math. Math. Sci. 2012., 2012: Article ID 430692

    Google Scholar 

  51. Yang, Z-H: Sharp bounds for the second Seiffert mean in terms of power means. Available online at arXiv:1206.5494v1 (2012). e-printatarXiv.org

  52. Yang, Z-H: The monotonicity results and sharp inequalities for some power-type means of two arguments. Available online at arXiv:1210.6478 (2012). e-printatarXiv.org

  53. Yang, Z-H: Sharp bounds for Seiffert mean in terms of weighted power means of arithmetic mean and geometric mean. Math. Inequal. Appl. (2013, in print)

  54. Chu Y-M, Wang M-K, Qiu Y-F: An optimal double inequality between power-type Heron and Seiffert means. J. Inequal. Appl. 2010., 2010: Article ID 146945

    Google Scholar 

  55. Costin I, Toader G: Optimal evaluations of some Seiffert-type means by power means. Appl. Math. Comput. 2013, 219: 4745–4754. 10.1016/j.amc.2012.10.091

    Article  MathSciNet  MATH  Google Scholar 

  56. Neuman E, Sándor J: On some inequalities involving trigonometric and hyperbolic functions with emphasis on the Cusa-Huygens, Wilker and Huygens inequalities. Math. Inequal. Appl. 2010, 13(4):715–723.

    MathSciNet  MATH  Google Scholar 

  57. Lv Y-P, Wang G-D, Chu Y-M: A note on Jordan type inequalities for hyperbolic functions. Appl. Math. Lett. 2012, 25: 505–508.

    Article  MathSciNet  MATH  Google Scholar 

  58. Yang, Z-H: Refinements of Mitrinovic-Cusa inequality. Available online at arXiv:1206.4911 (2012). e-printatarXiv.org

  59. Yang Z-H: Refinements of a two-sided inequality for trigonometric functions. J. Math. Inequal. 2013, 7(4):601–615. 10.7153/jmi-07-57

    Article  MathSciNet  MATH  Google Scholar 

Download references

Acknowledgements

The author would like to thank Ms. Jiang Yiping for her help. The author also wishes to thank the reviewer(s) who gave some important and valuable advice.

Author information

Authors and Affiliations

Authors

Corresponding author

Correspondence to Zhen-Hang Yang.

Additional information

Competing interests

The author declares that they have no competing interests.

Rights and permissions

Open Access This article is distributed under the terms of the Creative Commons Attribution 2.0 International License (https://creativecommons.org/licenses/by/2.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.

Reprints and permissions

About this article

Cite this article

Yang, ZH. Three families of two-parameter means constructed by trigonometric functions. J Inequal Appl 2013, 541 (2013). https://doi.org/10.1186/1029-242X-2013-541

Download citation

  • Received:

  • Accepted:

  • Published:

  • DOI: https://doi.org/10.1186/1029-242X-2013-541

Keywords