1 Introduction

Wilker [1] proposed two open problems, the first of which states that the inequality

( sin x x ) 2 + tan x x >2
(1.1)

holds for all x(0,π/2). Inequality (1.1) was proved by Sumner et al. in [2].

Recently, the Wilker inequality (1.1) and its generalizations, improvements, refinements and applications have attracted the attention of many mathematicians (see [317] and related references therein).

In [9], Wu and Srivastava established the following Wilker-type inequality:

( x sin x ) 2 + x tan x >2for x(0,π/2)
(1.2)

and its weighted and exponential generalization.

Theorem Wu ([[9], Theorem 1])

Let λ>0, μ>0 and p2qμ/λ. If q>0 or qmin(1,λ/μ), then the inequality

λ λ + μ ( sin x x ) p + μ λ + μ ( tan x x ) q >1
(1.3)

holds for x(0,π/2).

As an application of inequality (1.3), an open problem was proposed, answered and improved by Sándor and Bencze in [18]. Recently, inequality (1.3) and its related inequalities in [9] were extended to Bessel functions [3], and the hyperbolic version of Theorem Wu was presented in [12].

In 2009, Zhu [16] gave another exponential generalization of Wilker inequality (1.1) as follows.

Theorem Zh1 ([[16], Theorems 1.1 and 1.2])

Let 0<x<π/2. Then the inequalities

( sin x x ) 2 p + ( tan x x ) p > ( x sin x ) 2 p + ( x tan x ) p >2
(1.4)

hold if p1, while the first one in (1.4) holds if and only if p>0.

Theorem Zh2 ([[16], Theorems 1.3 and 1.4])

Let x>0. Then the inequalities

( sinh x x ) 2 p + ( tanh x x ) p > ( x sinh x ) 2 p + ( x tanh x ) p >2
(1.5)

hold if p1, while the first one in (1.5) holds if and only if p>0.

In [16], Zhu also proposed an open problem: find the respectively largest range of p such that inequalities (1.4) and (1.5) hold. It was solved by Matejička in [19].

Another inequality associated with the Wilker inequality is the following:

2 sin x x + tan x x >3
(1.6)

for x(0,π/2), which is known as the Huygens inequality [20]. The following refinement of Huygens inequality is due to Neuman and Sándor [7]:

2 sin x x + tan x x >2 x sin x + x tan x >3
(1.7)

for x(0,π/2). Very recently, the generalizations of (1.7) were given by Neuman in [8]. In [21], Zhu proved that the inequalities

(1 ξ 1 ) sin x x + ξ 1 tan x x >1>(1 η 1 ) sin x x + η 1 tan x x ,
(1.8)
(1 ξ 2 ) x sin x + ξ 2 x tan x >1>(1 η 2 ) x sin x + η 2 x tan x
(1.9)

hold for all x(0,π/2) with the best constants ξ 1 =1/3, η 1 =0, ξ 2 =1/3, η 2 =12/π. Later, Zhu [15] generalized inequalities (1.8) and (1.9) to the exponential form as follows.

Theorem Zh3 ([[15], Theorems 1.1 and 1.2])

Let 0<x<π/2. Then we have

  1. (i)

    If p1, then the double inequality

    (1λ) ( x sin x ) p +λ ( x tan x ) p <1<(1η) ( x sin x ) p +η ( x tan x ) p
    (1.10)

holds if and only if η1/3 and λ1 ( 2 / π ) p .

  1. (ii)

    If 0p4/5, then double inequality (1.10) holds if and only if λ1/3 and η1 ( 2 / π ) p .

  2. (iii)

    If p<0, then the second inequality in (1.10) holds if and only if η1/3.

The hyperbolic version of inequalities (1.7) was given in [7] by Neuman and Sándor. Later, Zhu showed the following.

Theorem Zh4 ([[17], Theorem 4.1])

Let x>0. Then one has

  1. (i)

    If p4/5, then the double inequality

    (1λ) ( x sinh x ) p +λ ( x tanh x ) p <1<(1η) ( x sinh x ) p +η ( x tanh x ) p
    (1.11)

holds if and only if η1/3 and λ0.

  1. (ii)

    If p<0, then the inequality

    (1η) ( x sinh x ) p +η ( x tanh x ) p >1
    (1.12)

holds if and only if η1/3.

The main aim of this paper is to present the best possible parameter p such that the inequalities

2 k + 2 ( sin x x ) k p + k k + 2 ( tan x x ) p >1for x(0,π/2),
(1.13)
2 k + 2 ( sinh x x ) k p + k k + 2 ( tanh x x ) p >1for x(0,)
(1.14)

or their reversed inequalities hold for certain fixed k with k(k+2)0. As applications, we also present several new analytic inequalities.

2 Lemmas

In order to establish our main results, we need several lemmas, which we present in this section.

Lemma 1 Let A, B and C be defined on (0,π/2) by

A=A(x)=cosx ( sin x x cos x ) 2 (xcosxsinx),
(2.1)
B=B(x)= ( x cos x sin x ) 2 (sinxxcosx),
(2.2)
C=C(x)= sin 2 x ( 2 x 2 cos x + x sin x + cos x sin 2 x ) .
(2.3)

Then, for fixed k1, the function xC(x)/(kA(x)+B(x)) is increasing on (0,π/2). Moreover, we have

5 12 ( k + 2 ) < C ( x ) k A ( x ) + B ( x ) <1.
(2.4)

Proof We clearly see that A,B>0 for x(0,π/2) because of sinxxcosx>0 and xcosxsinx=(2xsin2x)/2>0, and C>0 because of

( 2 x 2 cos x + x sin x + cos x sin 2 x ) = x 2 cosx ( ( sin x x ) 2 + tan x x 2 ) >0

by Wilker inequality (1.1).

Let D=(kA+B)/C, then simple computations lead to

D ( x ) = x sin 2 x ( 2 x 2 cos x + x sin x + cos x sin 2 x ) ( sin x x cos x ) ( x cos x sin x ) ( ( 1 k cos 2 x ) x + ( k 1 ) cos x sin x ) = 2 x 2 cos x + x sin x + cos x sin 2 x ( sin x x cos x ) ( x cos x sin x ) × x sin 2 x k ( sin x x cos x ) cos x + ( x cos x sin x ) : = D 1 ( x ) × D 2 ( x ) .

It follows from [[16], Lemma 2.9] that the function D 1 is positive and increasing on (0,π/2). Hence it remains to prove that the function D 2 is also positive and increasing. Clearly, D 2 (x)>0, we only need to show that D 2 (x)>0 for x(0,π/2). Indeed,

D 2 ( x ) = ( k 1 ) sin x ( 2 x 2 cos x + cos x sin 2 x + x sin x ) ( k ( sin x x cos x ) cos x + ( x cos x sin x ) ) 2 = ( k 1 ) x 2 sin x cos x ( k ( sin x x cos x ) cos x + ( x cos x sin x ) ) 2 ( ( sin x x ) 2 + tan x x 2 ) ,

which is clearly positive due to Wilker inequality (1.1). Therefore, C/(kA+B) is increasing on (0,π/2), and

5 12 ( k + 2 ) = lim x 0 C ( x ) k A ( x ) + B ( x ) <D(x)< lim x π / 2 C ( x ) k A ( x ) + B ( x ) =1.

This completes the proof. □

Lemma 2 Let E, F and G be defined on (0,) by

E=E(x)=coshx ( sinh x x cosh x ) 2 (xcoshxsinhx),
(2.5)
F=F(x)=(sinhxxcoshx) ( x cosh x sinh x ) 2 ,
(2.6)
G=G(x)=x sinh 2 x ( 2 x 2 cosh x x sinh x cosh x sinh 2 x ) .
(2.7)

Then, for fixed k1 (k<2), the function xG(x)/(kE(x)+F(x)) is decreasing (increasing) on (0,). Moreover, we have

min ( 0 , 12 5 ( k + 2 ) ) < G ( x ) k E ( x ) + F ( x ) <max ( 0 , 12 5 ( k + 2 ) ) .
(2.8)

Proof It is easy to verify that E,F<0 for x(0,) due to

( x cosh x sinh x ) = ( 2 x sinh 2 x ) / 2 < 0 , ( sinh x x cosh x ) = x ( sinh x x cos x ) < 0 .

While G<0 because of

( 2 x 2 cosh x x sinh x cosh x sinh 2 x ) = x 2 coshx ( ( sinh x x ) 2 + tanh x x 2 ) <0

by Wilker inequality (1.5).

Denote G/(kE+F) by H and simple computations give

H ( x ) = x sinh 2 x ( 2 x 2 cosh x x sinh x cosh x sinh 2 x ) cosh x ( sinh x x cosh x ) 2 ( x sinh x cosh x ) k + ( sinh x x cosh x ) ( x sinh x cosh x ) 2 = 2 x 2 cosh x + x sinh x + cosh x sinh 2 x ( x cosh x sinh x ) ( sinh x cosh x x ) × x sinh 2 x ( k ( x cosh x sinh x ) cosh x + sinh x cosh x x ) : = H 1 ( x ) × H 2 ( x ) .

Clearly, H 1 (x)>0, and it was proved in [[19], Proof of Lemma 2.2] that H 1 is decreasing on (0,). In order to prove the monotonicity of H, we only need to deal with the sign and monotonicity of H 2 .

  1. (i)

    Clearly, H 2 (x)>0 for k1. And we claim that H 2 is also decreasing on (0,). Indeed,

    H 2 ( x ) = ( k 1 ) sinh x ( 2 x 2 cosh x + cosh x sinh 2 x + x sinh x ) ( x cosh x sinh x ) 2 ( cosh x sinh x x ) 2 = ( k 1 ) x 2 sinh x cosh x ( x cosh x sinh x ) 2 ( cosh x sinh x x ) 2 ( ( sinh x x ) 2 + tanh x x 2 ) < 0 .

Consequently, H= H 1 × H 2 is positive and decreasing on (0,), and so

0= lim x G ( x ) k E ( x ) + F ( x ) < G ( x ) k E ( x ) + F ( x ) < lim x 0 G ( x ) k E ( x ) + F ( x ) = 12 5 ( k + 2 ) .
  1. (ii)

    For k<2, by the previous proof we clearly see that H 2 is decreasing on (0,), and so

    0< 1 k = lim x ( H 2 ( x ) ) < H 2 (x)< lim x 0 ( H 2 ( x ) ) = 3 k + 2 ,

which implies that H 2 is positive and decreasing on (0,), and so is the function H= H 1 ×( H 2 ). That is, H is negative and increasing on (0,), and inequality (2.8) holds true.

This completes the proof. □

Remark 1 It should be noted that kE(x)+F(x)<0 for k1 and kE(x)+F(x)>0 for k<2. In fact, it suffices to notice (2.8) and G(x)<0 for x(0,).

Lemma 3 For k1, we have

1> ln ( k + 2 ) ln 2 k ( ln π ln 2 ) > 12 5 ( k + 2 ) .

Proof It suffices to show that

δ 1 ( k ) = ln ( k + 2 ) ln 2 ln π ln 2 k < 0 , δ 2 ( k ) = ln ( k + 2 ) ln 2 ln π ln 2 12 k 5 ( k + 2 ) > 0

for k1.

Differentiation gives

δ 1 ( k ) = 1 ( ln π ln 2 ) ( k + 2 ) 1 < 0 , δ 2 ( k ) = 1 5 5 k + 24 ln 2 24 ln π + 10 ( k + 2 ) 2 ( ln π ln 2 ) > 0

for k1. Therefore, Lemma 3 follows from δ 1 (k) δ 1 (1)=(ln3ln2)/(ln3lnπ)<0 and δ 2 (k) δ 2 (1)=(ln3ln2)/(lnπln2)4/5>0. □

3 Main results

Theorem 1 For fixed k1, inequality (1.13) holds for x(0,π/2) if and only if p>0 or p ln ( k + 2 ) ln 2 k ( ln π ln 2 ) .

Proof Inequality (1.13) is equivalent to

f(x)= 2 k + 2 ( sin x x ) k p + k k + 2 ( tan x x ) p 1>0
(3.1)

for x(0,π/2). Differentiation yields

f ( x ) = 2 k p k + 2 sin x x cos x x 2 ( sin x x ) k p 1 + k p k + 2 x sin x cos x x 2 cos 2 x ( tan x x ) p 1 = k p k + 2 x sin x cos x x 2 cos 2 x ( tan x x ) p 1 g ( x ) ,
(3.2)

where

g(x)=14 sin x x cos x 2 x sin 2 x ( sin x x ) ( k 1 ) p ( cos x ) p + 1 .
(3.3)

A simple computation leads to g( 0 + )=0.

Differentiation again and simplifying give

g (x)=8 ( sin x x ) ( k 1 ) p ( cos x ) p x sin x ( 2 x sin 2 x ) 2 h(x),
(3.4)

where

h ( x ) = cos x ( sin x x cos x ) 2 ( x cos x sin x ) k p + ( x cos x sin x ) 2 ( sin x x cos x ) p + x sin 2 x ( 2 x 2 cos x + x sin x + cos x sin 2 x ) = k p A ( x ) + p B ( x ) + C ( x ) = ( k A + B ) ( p + C k A + B ) ,
(3.5)

where A(x), B(x) and C(x) are defined as in (2.1), (2.2) and (2.3), respectively.

By (3.2), (3.4) we easily get

sgn f (x)=sgnpsgng(x),
(3.6)
sgn g (x)=sgnh(x).
(3.7)

Necessity. We first present two limit relations:

lim x 0 + x 4 f(x)= k p 36 ( p + 12 5 ( k + 2 ) ) ,
(3.8)
lim x ( π / 2 ) f(x)= { if  p > 0 , 2 k + 2 ( 2 π ) k p 1 if  p < 0 .
(3.9)

In fact, using power series extension yields

f(x)= k p 36 k p + 2 p + 12 / 5 k + 2 x 4 +o ( x 4 ) ,

which implies the first limit relation (3.8). From the fact that lim x π / 2 tanx=, the second one (3.9) easily follows.

Now we can derive that the necessary condition of (1.13) holds for x(0,π/2) from the simultaneous inequalities lim x 0 + x 4 f(x)0 and lim x ( π / 2 ) f(x)0. Solving for p yields p>0 or

pmin ( 12 5 ( k + 2 ) , ln ( k + 2 ) ln 2 k ( ln π ln 2 ) ) = ln ( k + 2 ) ln 2 k ( ln π ln 2 ) ,

where the equality holds due to Lemma 3.

Sufficiency. We prove that the condition p>0 or p ln ( k + 2 ) ln 2 k ( ln π ln 2 ) is sufficient. We divide the proof into three cases.

Case 1 p>0. Clearly, h(x)>0, then g (x)>0 and g(x)>g( 0 + )=0, which together with sgnp=1 yields f (x)>0 and f(x)>f( 0 + )=0.

Case 2 p1. By Lemma 1 it is easy to get

p+ C k A + B <p+10,

which reveals that h(x)<0, g (x)<0 and g(x)<g( 0 + )=0, which in combination with sgnp=1 implies f (x)>0 and f(x)>f( 0 + )=0.

Case 3 1<p ln ( k + 2 ) ln 2 k ( ln π ln 2 ) . Lemma 1 reveals that C k A + B is increasing on (0,π/2), so is the function xp+ C k A + B :=λ(x). Since

λ ( 0 + ) =p+ 12 5 ( k + 2 ) <0,λ ( π 2 ) =p+1>0,

there exists x 1 (0,π/2) such that λ(x)<0 for x(0, x 1 ) and λ(x)>0 for x( x 1 ,π/2), and so is g (x). Therefore, g(x)<g( 0 + )=0 for x(0, x 1 ) but g(π/ 2 )=1, which implies that there exists x 0 ( x 1 ,π/2) such that g(x)<0 for x(0, x 0 ) and g(x)>0 for x( x 0 ,π/2). Due to sgnp=1, it is deduced that f (x)>0 for x(0, x 0 ) and f (x)<0 for x( x 0 ,π/2), which reveals that f is increasing on (0, x 0 ) and decreasing on ( x 0 ,π/2). It follows that

0 = f ( 0 + ) < f ( x ) < f ( x 0 ) = 0 for  x ( 0 , x 0 ) , f ( x 0 ) > f ( x ) > f ( π / 2 ) = 2 k + 2 ( 2 π ) k p 1 0 for  x ( x 0 , π / 2 ) ,

that is, f(x)>0 for x(0,π/2).

This completes the proof. □

Theorem 2 For fixed k1, the reversed inequality of (1.13), that is,

2 k + 2 ( sin x x ) k p + k k + 2 ( tan x x ) p <1,
(3.10)

holds for x(0,π/2) if and only if 12 5 ( k + 2 ) p<0.

Proof Necessity. If inequality (3.10) holds for x(0,π/2), then we have

lim x 0 + f ( x ) x 4 = k p 36 ( p + 12 5 ( k + 2 ) ) 0.

Solving the inequality for p yields 12 5 ( k + 2 ) p<0.

Sufficiency. We prove that the condition 12 5 ( k + 2 ) p<0 is sufficient. It suffices to show that f(x)<0 for x(0,π/2). By Lemma 1 it is easy to get

p+ C k A + B p+ 12 5 ( k + 2 ) 0,

which reveals that h(x)>0, g (x)>0 and g(x)>g( 0 + )=0. In combination with sgnp=1, it implies f (x)<0. Thus, f(x)<f( 0 + )=0, which proves the sufficiency and the proof is completed. □

Theorem 3 For fixed k1, inequality (1.14) holds for x(0,) if and only if p>0 or p 12 5 ( k + 2 ) .

Proof Let

u(x)= 2 k + 2 ( sinh x x ) k p + k k + 2 ( tanh x x ) p 1.
(3.11)

Then inequality (1.14) is equivalent to u(x)>0. Differentiation leads to

u (x)= k p 2 ( k + 2 ) sinh 2 x 2 x x 2 cosh 2 x ( tanh x x ) p 1 v(x),
(3.12)

where

v(x)=14 sinh x x cosh x 2 x sinh 2 x ( sinh x x ) k p p ( cosh x ) p + 1 .
(3.13)

Differentiation again gives

v (x)= 2 cosh p x ( sinh x x ) k p p x sinh x ( x cosh x sinh x ) 2 w(x),
(3.14)

where

w ( x ) = cosh x ( sinh x x cosh x ) 2 ( x cosh x sinh x ) k p + ( sinh x x cosh x ) ( x cosh x sinh x ) 2 p + x sinh 2 x ( 2 x 2 cosh x x sinh x cosh x sinh 2 x ) = k p E ( x ) + p F ( x ) + G ( x ) = ( k E + F ) ( p + G k E + F ) ,
(3.15)

where E(x), F(x) and G(x) are defined as in (2.5), (2.6) and (2.7), respectively.

By (3.12) and (3.14) we easily get

sgn u (x)=sgn k k + 2 sgnpsgnv(x),
(3.16)
sgn v (x)=sgnw(x).
(3.17)

Necessity. If inequality (1.14) holds for x(0,), then we have lim x 0 + x 4 u(x)0. Expanding u(x) in power series gives

u(x)= k 36 p ( p + 12 5 p ( k + 2 ) ) x 4 +o ( x 4 ) .

Hence we get

lim x 0 + x 4 u(x)= k 36 p ( p + 12 5 ( k + 2 ) ) 0.

Solving the inequality for p yields p>0 or p 12 5 ( k + 2 ) .

Sufficiency. We prove that the condition p>0 or p 12 5 ( k + 2 ) is sufficient for (1.14) to hold.

If p>0, then w(x)<0 due to E,F,G<0. Hence, from (3.17) we have v (x)<0 and v(x)< lim x 0 + v(x)=0. It is derived by (3.16) that u (x)>0, and so u(x)> lim x 0 + u(x)=0.

If p 12 5 ( k + 2 ) , then by Lemma 2 we have

p+ G k E + F 12 5 ( k + 2 ) + G k E + F <0

and

w(x)=(kE+F) ( p + G k E + F ) >0.

From (3.17) we have v (x)>0 and v(x)> lim x 0 + v(x)=0. It follows by (3.16) that u (x)>0, which implies that u(x)> lim x 0 + u(x)=0.

This completes the proof. □

Remark 2 For k1, since lim x u(x)= for p0 and lim x u(x)=0 for p=0, there does not exist p such that the reverse inequality of (1.14) holds for all x>0. But we can show that there exists x 0 (0,) such that u(x)<0, that is, the reverse inequality of (1.14) holds for 12 5 ( k + 2 ) <p<0. The details of the proof are omitted.

Theorem 4 For fixed k<2, the reverse of (1.14), that is,

2 k + 2 ( sinh x x ) k p + k k + 2 ( tanh x x ) p <1,
(3.18)

holds for x(0,) if and only if p<0 or p 12 5 ( k + 2 ) .

Proof Necessity. If inequality (3.18) holds for x(0,), then we have

lim x 0 + u ( x ) x 4 = k 36 p ( p + 12 5 ( k + 2 ) ) 0.

Solving the inequality for p yields p<0 or p 12 5 ( k + 2 ) .

Sufficiency. We prove that the condition p<0 or p 12 5 ( k + 2 ) is sufficient for (3.18) to hold.

If p<0, then w(x)=(kE+F)(p+ G k E + F )<0 due to kE+F>0 and G<0. Hence, from (3.17) we have v (x)<0 and v(x)< lim x 0 + v(x)=0. It is derived by (3.16) that u (x)<0, and so u(x)< lim x 0 + u(x)=0.

If p 12 5 ( k + 2 ) , then by Lemma 2 we have

p+ G k E + F p+ 12 5 ( k + 2 ) >0

and

w(x)=(kE+F) ( p + G k E + F ) >0.

From (3.17) we have v (x)>0 and v(x)> lim x 0 + v(x)=0. It follows by (3.16) that u (x)<0, which implies that u(x)< lim x 0 + u(x)=0.

This completes the proof. □

4 Applications

4.1 Huygens-type inequalities

Letting k=1 in Theorems 1 and 2, we have the following proposition.

Proposition 1 For x(0,π/2), the double inequality

2 3 ( sin x x ) p + 1 3 ( tan x x ) p >1> 2 3 ( sin x x ) q + 1 3 ( tan x x ) q
(4.1)

holds if and only if p>0 or p ln 3 ln 2 ln π ln 2 0.898 and 4/5q<0.

Let M r (a,b;w) denote the r th weighted power mean of positive numbers a,b>0 defined by

M r (a,b;w):= ( w a r + ( 1 w ) b r ) 1 / r if r0 and  M 0 (a,b;w)= a w b 1 w ,
(4.2)

where w(0,1).

Since

2 3 ( sin x x ) p + 1 3 ( tan x x ) p = 2 3 + 1 3 ( cos x ) p ( sin x x ) p ,

by Proposition 1 the inequality

sin x x > ( 2 3 + 1 3 ( cos x ) p ) 1 / p = M p ( 1 , cos x ; 2 3 )

holds for x(0,π/2) if and only if p4/5. Similarly, its reversed inequality holds if and only if p ln 3 ln 2 ln π ln 2 . The facts can be stated as a corollary.

Corollary 1 Let M r (a,b;w) be defined by (4.2). Then, for x(0,π/2), the inequalities

M α ( 1 , cos x ; 2 3 ) < sin x x < M β ( 1 , cos x ; 2 3 )
(4.3)

hold if and only if α4/5 and β ln 3 ln 2 ln π ln 2 0.898.

Remark 3 The Cusa-Huygens inequality [20] refers to

sin x x < 2 3 + 1 3 cosx
(4.4)

holds for x(0,π/2), which is equivalent to the second inequality in (1.7). As an improvement and generalization, Corollary 1 was proved in [22] by Yang. Here we provide a new proof.

Remark 4 Let a>b>0 and let x=arcsin a b a + b (0,π/2). Then sinx/x=P/A, cosx=G/A and inequalities (4.3) can be rewritten as

M α ( A , G ; 2 3 ) <P< M β ( A , G ; 2 3 ) ,
(4.5)

where P is the first Seiffert mean [23] defined by

P=P(a,b)= a b 2 arcsin a b a + b ,

A and G denote the arithmetic and geometric means of a and b, respectively.

Let x=arctan a b a + b . Then sinx/x=T/Q, cosx=A/Q, and inequalities (4.3) can be rewritten as

M α ( Q , A ; 2 3 ) <T< M β ( Q , A ; 2 3 ) ,
(4.6)

where T is the second Seiffert mean [24] defined by

T=T(a,b)= a b 2 arctan a b a + b ,

Q denotes the quadratic mean of a and b.

Obviously, by Corollary 2, the two double inequalities (4.5) (see [22]) and (4.6) hold if and only if α4/5 and β ln 3 ln 2 ln π ln 2 0.898, (4.6) seems to be a new inequality.

In the same way, taking k=1 in Theorem 3, we get the following.

Proposition 2 For x(0,), the inequality

2 3 ( sinh x x ) p + 1 3 ( tanh x x ) p >1
(4.7)

holds if and only if p>0 or p 4 5 .

Similar to Corollary 1, we have the following.

Corollary 2 Let M r (a,b;w) be defined by (4.2). Then, for x(0,), the inequalities

M α ( 1 , cosh x ; 2 3 ) < sinh x x < M β ( 1 , cosh x ; 2 3 )
(4.8)

hold if and only if α0 and β4/5.

Remark 5 Let a>b>0 and x=ln a / b . Then sinhx/x=L/G, coshx=A/G, and (4.8) can be rewritten as

M α ( G , A ; 2 3 ) <L< M β ( G , A ; 2 3 ) ,
(4.9)

where L is the logarithmic means of a and b defined by

L=L(a,b)= a b ln a ln b .

Making use of x=arcsinh b a a + b yields sinhx/x=NS/A and coshx=Q/A, where NS is the Nueman-Sándor mean defined by

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

Thus, (4.8) is equivalent to

M α ( A , Q ; 2 3 ) <NS< M β ( A , Q ; 2 3 ) .
(4.10)

Corollary 2 implies that inequalities (4.9) and (4.10) hold if and only if α0 and β4/5. The second inequality in (4.10) is a new inequality.

Remark 6 It should be pointed out that all inequalities involving sinx/xand cosx or sinhx/xand coshx in this paper can be rewritten as the equivalent inequalities for bivariate means mentioned previously. In what follows we no longer mention this.

4.2 Wilker-Zhu-type inequalities

Letting k=2 in Theorems 1 and 2, we have the following.

Proposition 3 For x(0,π/2), the double inequality

( sin x x ) 2 p + ( tan x x ) p >2> ( sin x x ) 2 q + ( tan x x ) q
(4.11)

holds if and only if p>0 or p ln 2 2 ( ln π ln 2 ) 0.767 and 3/5q<0.

Note that

( sin x x ) 2 p + ( tan x x ) p 2 ( sin x x ) p + 8 + cos 2 p x + cos p x 2 = ( x sin x ) p 8 + cos 2 p x cos p x 2 .

By Proposition 3 the inequality

x sin x > ( 8 + cos 2 p x cos p x 2 ) 1 / p

or

sin x x < ( 8 + cos 2 p x + cos p x 4 ) 1 / p := H p (cosx)

holds for x(0,π/2) if and only if p ln 2 2 ( ln π ln 2 ) , where H r is defined on (0,) by

H r (t)= ( 8 + t 2 r + t r 4 ) 1 / r if r0 and  H 0 (t)= t 3 .
(4.12)

Likewise, its reversed inequality holds if and only if p3/5. This result can be stated as a corollary.

Corollary 3 Let H r (t) be defined by (4.12). Then, for x(0,π/2), the inequalities

H α (cosx)< sin x x < H β (cosx)
(4.13)

are true if and only if α3/5 and β ln 2 2 ( ln π ln 2 ) 0.767.

Taking k=2 in Theorem 3, we have the following.

Proposition 4 For x(0,), the inequality

( sinh x x ) 2 p + ( tanh x x ) p >2

holds if and only if p>0 or p3/5.

In a similar way, we get Corollary 4.

Corollary 4 Let H r (t) be defined by (4.12). Then, for x(0,), the inequalities

H α (coshx)< sinh x x < H β (coshx)
(4.14)

are true if and only if α0 and β3/5.

Now we give a generalization of inequalities (1.4) given by Zhu [15].

Proposition 5 For fixed k1, both chains of inequalities

2 k + 2 ( sin x x ) k p + k k + 2 ( tan x x ) p k k + 2 ( sin x x ) k p + 2 k + 2 ( tan x x ) p > 2 k + 2 ( x sin x ) k p + k k + 2 ( x tan x ) p > 1 ,
(4.15)
2 k + 2 ( sin x x ) k p + k k + 2 ( tan x x ) p > 2 k + 2 ( x tan x ) p + k k + 2 ( x sin x ) k p 2 k + 2 ( x sin x ) k p + k k + 2 ( x tan x ) p > 1
(4.16)

hold for x(0,π/2) if and only if k2 and p ln ( k + 2 ) ln 2 k ( ln π ln 2 ) .

Proof The first inequality in (4.15) is equivalent to

2 k + 2 ( sin x x ) k p + k k + 2 ( tan x x ) p k k + 2 ( sin x x ) k p 2 k + 2 ( tan x x ) p = k 2 k + 2 ( ( tan x x ) p ( sin x x ) k p ) > 0 .

Due to tan x x >1 and sin x x <1, it holds for x(0,π/2) if and only if

(k,p){k2,p>0}{1k2,p<0}:= Ω 1 .

The second one is equivalent to

k k + 2 ( sin x x ) k p + 2 k + 2 ( tan x x ) p 2 k + 2 ( x sin x ) k p + k k + 2 ( x tan x ) p >1,

which can be simplified to

( sin x x ) k p ( tan x x ) p = ( ( sin x x ) k + 1 1 cos x ) p >1.

It is true for x(0,π/2) if and only if (k,p){k+13,p0}:= Ω 2 .

By Theorem 1, the third one in (4.15) holds for x(0,π/2) if and only if

(k,p){k1,p>0} { k 1 , p ln ( k + 2 ) ln 2 k ( ln π ln 2 ) } := Ω 3 .

Hence, inequalities (4.15) hold for x(0,π/2) if and only if

(k,p) Ω 1 Ω 2 Ω 3 = { k 2 , p ln ( k + 2 ) ln 2 k ( ln π ln 2 ) } ,

which proves (4.15).

In the same way, we can prove (4.16), the details are omitted. □

Letting k=2 in Proposition 5, we have the following.

Corollary 5 For x(0,π/2), inequality (1.4) holds if and only if p ln 2 2 ( ln π ln 2 ) 0.767.

Similarly, using Theorem 3 we easily prove the following proposition.

Proposition 6 For fixed k1, the inequalities

k k + 2 ( sinh x x ) k p + 2 k + 2 ( tanh x x ) p > 2 k + 2 ( x sinh x ) k p + k k + 2 ( x tanh x ) p >1
(4.17)

hold for x(0,) if and only if k2 and p 12 5 ( k + 2 ) .

Letting k=2 in Proposition 6, we have the following.

Corollary 6 For x(0,), inequality (1.5) holds if and only if p3/5.

Remark 7 Clearly, Corollaries 5 and 6 offer another method for solving the problems posed by Zhu in [16].

4.3 Other Wilker-type inequalities

Taking k=3,4 in Theorems 1 and 2, we obtain the following.

Proposition 7 For x(0,π/2), the inequality

2 5 ( sin x x ) 3 p + 3 5 ( tan x x ) p >1
(4.18)

holds if and only if p>0 or p ln 5 ln 2 3 ( ln π ln 2 ) 0.676. It is reversed if and only if 12/25p<0.

Proposition 8 For x(0,π/2), the inequality

1 3 ( sin x x ) 4 p + 2 3 ( tan x x ) p >1
(4.19)

holds if and only if p>0 or p ln 3 4 ( ln π ln 2 ) 0.608. It is reversed if and only if 2/5p<0.

Putting k=3,4 in Theorem 3, we get the following.

Proposition 9 For x(0,), the inequality

( tanh x x ) p < 2 3 ( x sinh x ) 3 p + 1 3
(4.20)

holds if and only if p<0 or p12/5.

Proposition 10 For x(0,π/2), the inequality

2 ( tanh x x ) p < ( x sinh x ) 4 p +1
(4.21)

holds if and only if p<0 or p6/5.