1 Introduction

Let sinht, cosht, and cotht be the hyperbolic sine, hyperbolic cosine, and hyperbolic cotangent, respectively. It is well known that (see [16])

1< sinh t t < e t coth t 1 <cosht
(1.1)

holds for all t0.

In the recent paper [7], we have established the following Cusa-type inequalities of exponential type for the triple {1,(sinht)/t,cosht} described as follows.

Theorem 1.1 (Cusa-type inequalities [[7], Part (i) of Theorem 1.1])

Let p4/5, and t0. Then the double inequality

(1λ)+λ ( cosh t ) p < ( sinh t t ) p <(1η)+η ( cosh t ) p
(1.2)

holds if and only if η1/3 and λ0.

On the other hand, the author of this paper [8] obtains the following inequalities of exponential type for the triple {1,exp(tcotht1),cosht}.

Theorem 1.2 ([[8], Theorem 2])

Let p>0, and t0. Then

(1) if 0<p6/5, the double inequality

α ( cosh t ) p +(1α)< e p ( t coth t 1 ) <β ( cosh t ) p +(1β)
(1.3)

holds if and only if α2/3 and β ( 2 / e ) p ;

(2) if p2, the double inequality

α ( cosh t ) p +(1α)< e p ( t coth t 1 ) <β ( cosh t ) p +(1β)
(1.4)

holds if and only if α ( 2 / e ) p and β2/3.

Next, we do the work for each of the triples {(sinht)/t,exp(tcotht1),cosht} and {1,(sinht)/t,exp(tcotht1)}, and obtain the following two new results.

Theorem 1.3 Let 0<p8/5, and t0. Then

α ( cosh t ) p +(1α) ( sinh t t ) p < e p ( t coth t 1 ) <β ( cosh t ) p +(1β) ( sinh t t ) p
(1.5)

holds if and only if α1/2 and β ( 2 / e ) p .

Theorem 1.4 Let p286/693, and t0. Then

α+(1α) e p ( t coth t 1 ) < ( sinh t t ) p <β+(1β) e p ( t coth t 1 )
(1.6)

holds if and only if β1/2 and α1.

In this paper, we shall give the elementary proofs of Theorem 1.3 and Theorem 1.4. In the last section, we apply Theorems 1.1-1.4 to obtain some new results for four classical means.

2 Lemmas

Lemma 2.1 ([911])

Let f,g:[a,b]R be two continuous functions which are differentiable on (a,b). Further, let g 0 on (a,b). If f / g is increasing (or decreasing) on (a,b), then the functions (f(x)f( b ))/(g(x)g( b )) and (f(x)f( a + ))/(g(x)g( a + )) are also increasing (or decreasing) on (a,b).

Lemma 2.2 Let t(0,+). Then the inequality

D(t)t sinh 5 t+2t sinh 3 t+ t 4 cosht sinh 4 tcosht t 3 sinh 3 t2 t 3 sinht>0

holds.

Proof Using the power series expansions of the functions sinh5t, sinh3t, cosht, sinh 4 tcosht, and sinht, we have

D ( t ) = 1 16 t ( sinh 5 t 5 sinh 3 t + 10 sinh t ) + 1 2 t ( sinh 3 t 3 sinh t ) + t 4 cosh t 1 16 ( cosh 5 t 3 cosh 3 t + 2 cosh t ) 1 4 t 3 ( sinh 3 t 3 sinh t ) 2 t 3 sinh t = 1 16 n = 0 5 2 n + 1 5 3 2 n + 1 + 10 ( 2 n + 2 ) ! t 2 n + 1 + 1 2 n = 0 3 2 n + 1 3 ( 2 n + 1 ) ! t 2 n + 2 + n = 0 1 ( 2 n ) ! t 2 n + 4 1 16 n = 0 5 2 n 3 3 2 n + 2 ( 2 n ) ! t 2 n 1 4 n = 0 3 2 n + 1 3 ( 2 n + 1 ) ! t 2 n + 4 2 n = 0 1 ( 2 n + 1 ) ! t 2 n + 4 = 1 16 n = 3 l n ( 2 n + 4 ) ! t 2 n + 4 ,

where

l n = ( 2 n + 4 ) ( 5 2 n + 3 5 3 2 n + 3 + 10 ) + 8 ( 2 n + 4 ) ( 3 2 n + 3 3 ) + 16 ( 2 n + 4 ) ( 2 n + 3 ) ( 2 n + 2 ) ( 2 n + 1 ) ( 5 2 n + 4 3 3 2 n + 4 + 2 ) 4 ( 2 n + 4 ) ( 2 n + 3 ) ( 2 n + 2 ) ( 3 2 n + 1 3 ) 32 ( 2 n + 4 ) ( 2 n + 3 ) ( 2 n + 2 ) = ( 250 n 125 ) 25 n + ( 279 462 n 432 n 2 96 n 3 ) 9 n + 256 n 4 + 1 , 120 n 3 + 1 , 520 n 2 + 532 n 154 , n = 3 , 4 , .

Using a basic differential method, we can easily prove

f ( x ) ( 250 x 125 ) 25 x + ( 279 462 x 432 x 2 96 x 3 ) 9 x + 256 x 4 + 1 , 120 x 3 + 1 , 520 x 2 + 532 x 154 > 0

on [3,). This leads to l n >0 for n=3,4, , and D(t)>0. So, the proof of Lemma 2.2 is complete. □

3 Proof of Theorem 1.3

Let

F(t) ( t sinh t e t coth t 1 ) p 1 ( t coth t ) p 1 = f 1 ( t ) f 1 ( 0 + ) g 1 ( t ) g 1 ( 0 + ) ,

where f 1 (t)= ( t sinh t e t coth t 1 ) p and g 1 (t)= ( t coth t ) p . Then

k 1 (t) f 1 ( t ) g 1 ( t ) = e p ( t coth t 1 ) ( cosh t ) p 1 sinh 2 t t 2 sinh t ( sinh t cosh t t ) .

We compute

k 1 (t)= e p ( t coth t 1 ) ( cosh t ) p u 1 ( t ) ( sinh t ) 3 ( sinh t cosh t t ) 2 ,

where

u 1 ( t ) = 2 t 2 sinh 4 t cosh t + sinh 4 t cosh t 4 t sinh 5 t 3 t sinh 3 t + 3 t 2 sinh 2 t cosh t t 3 sinh t p ( t sinh 5 t + 2 t sinh 3 t + t 4 cosh t sinh 4 t cosh t t 3 sinh 3 t 2 t 3 sinh t ) = 2 t 2 sinh 4 t cosh t + sinh 4 t cosh t 4 t sinh 5 t 3 t sinh 3 t + 3 t 2 sinh 2 t cosh t t 3 sinh t p D ( t ) .

If 0<p8/5, by Lemma 2.2 we have

5 u 1 ( t ) 10 t 2 sinh 4 t cosh t + 13 sinh 4 t cosh t 28 t sinh 5 t 46 t sinh 3 t + 30 t 2 sinh 2 t cosh t + 6 t 3 sinh t 8 t 4 cosh t + 8 t 3 sinh 3 t = n = 3 h n 16 ( 2 n + 4 ) ! t 2 n + 4 ,

where

h n = 10 ( 2 n + 4 ) ( 2 n + 3 ) ( 5 2 n + 2 3 3 2 n + 2 + 2 ) + 13 ( 5 2 n + 4 3 3 2 n + 4 + 2 ) 28 ( 2 n + 4 ) ( 5 2 n + 3 5 3 2 n + 3 + 10 ) 184 ( 2 n + 4 ) ( 3 2 n + 3 3 ) + 120 ( 2 n + 4 ) ( 2 n + 3 ) ( 3 2 n + 2 1 ) + 96 ( 2 n + 4 ) ( 2 n + 3 ) ( 2 n + 2 ) ( 2 n + 1 ) 2 n 128 ( 2 n + 4 ) ( 2 n + 3 ) ( 2 n + 2 ) ( 2 n + 1 ) + 96 ( 2 n + 4 ) ( 2 n + 3 ) ( 2 n + 2 ) ( 3 2 n 1 ) = ( 1 , 000 n 2 3 , 500 n 2 , 875 ) 25 n + ( 768 n 3 + 6 , 696 n 2 + 13 , 956 n + 4 , 113 ) 9 n + ( 2 n + 4 ) ( 2 n + 3 ) ( 2 n + 2 ) ( 2 n + 1 ) ( 192 n 128 ) 96 ( 2 n + 4 ) ( 2 n + 3 ) ( 2 n + 2 ) 100 ( 2 n + 4 ) ( 2 n + 3 ) + 272 ( 2 n + 4 ) + 26 > 0

for n=3,4, .

We have u 1 (t)>0 for 0<p8/5. So, k 1 (t)>0 for t>0, and f 1 (t)/ g 1 (t)= k 1 (t) is increasing on (0,+). Hence, F(t) is increasing on (0,+) by Lemma 2.1. At the same time, lim t 0 + F(t)=1/2 and lim t + F(t)= ( 2 / e ) p . So, the proof of Theorem 1.3 is complete.

4 Proof of Theorem 1.4

Let

S(t) ( sinh t t e 1 t coth t ) p 1 e p ( 1 t coth t ) 1 = f 2 ( t ) f 2 ( 0 + ) g 2 ( t ) g 2 ( 0 + ) ,

where f 2 (t)= ( sinh t t e 1 t coth t ) p and g 2 (t)= e p ( 1 t coth t ) . Then

k 2 (t) f 2 ( t ) g 2 ( t ) = ( sinh t t ) p 1 ( sinh t ) 3 t 2 sinh t t 2 ( sinh t cosh t t ) ,

and

k 2 (t)= ( sinh t t ) p 2 u 2 ( t ) t 4 ( sinh t cosh t t ) 2 ,

where

where e n =1( d n / c n ) and

Let

Then

e n =1 d n c n = j ( n ) i ( n ) .

Let Δ(n)=286i(n)693j(n). Then

Δ ( n ) = ( 741 , 313 n 5 , 759 , 424 ) 36 n + 16 n [ 2 , 275 , 328 ( 2 n + 5 ) + 4 , 009 , 984 + 70 , 400 ( 2 n + 5 ) ( 2 n + 4 ) ( 2 n + 3 ) 532 , 224 ( 2 n + 5 ) ( 2 n + 4 ) ] + 4 n [ 9 , 152 ( 2 n + 5 ) ( 2 n + 4 ) ( 2 n + 3 ) ( 2 n + 2 ) 62 , 656 ( 2 n + 5 ) ( 2 n + 4 ) ( 2 n + 3 ) + 133 , 056 ( 2 n + 5 ) ( 2 n + 4 ) 610 , 016 ( 2 n + 5 ) 156 , 640 ] .

First, we check that Δ(n)>0 for n=3,4,5,6,7; second, we can easily obtain that Δ(n)>0 for n8. So, we have that Δ(n)>0 for n=3,4, .

So, we have u 2 (t)>0 for p286/693. So, k 2 (t)>0 for t>0, and f 2 (t)/ g 2 (t)= k 2 (t) is increasing on (0,+). Hence, S(t) is increasing on (0,+) by Lemma 2.1 when p286/693. At the same time, lim t 0 + S(t)=1/2 and lim t + S(t)=1. So, the proof of Theorem 1.4 is complete.

5 Applications of theorems

In this section, we assume that x and y are two different positive numbers. Let A(x,y), G(x,y), L(x,y), and I(x,y) be the arithmetic, geometric, logarithmic, and identric means, respectively. Without loss of generality, we set 0<x<y. By the transformation t=(log(y/x))/2, we can compute and obtain

where t>0.

Now, the four results in Section 1 are equivalent to the following ones for four classical means.

Theorem 5.1 Let p4/5, and x and y be positive real numbers with xy. Then

α A p (x,y)+(1α) G p (x,y)< L p (x,y)<β A p (x,y)+(1β) G p (x,y)
(5.1)

holds if and only if α0 and β1/3.

Theorem 5.1 can deduce the following one, which is from Zhu [8].

Corollary 5.2 ([[8], Theorem 1])

Let p1, and x and y be positive real numbers with xy. Then

α A p (x,y)+(1α) G p (x,y)< L p (x,y)<β A p (x,y)+(1β) G p (x,y)
(5.2)

holds if and only if α0 and β1/3.

When letting p=1 in Theorem 5.1, one can obtain the result (see [1214], [[15], Theorem 1]).

Corollary 5.3 Let x and y be positive real numbers with xy. Then

αA(x,y)+(1α)G(x,y)<L(x,y)<βA(x,y)+(1β)G(x,y)
(5.3)

holds if and only if α0 and β1/3.

When letting β=1/3 in the right-hand inequality of (5.3), one can obtain the well-known inequality by Carlson [16]

L(x,y)< 1 3 A(x,y)+ 2 3 G(x,y).
(5.4)

Theorem 5.4 Let p>0. Then

(1) if 0<p6/5, the double inequality

α A p (x,y)+(1α) G p (x,y)< I p (x,y)<β A p (x,y)+(1β) G p (x,y)
(5.5)

holds if and only if α2/3 and β ( 2 / e ) p ;

(2) if p2, the double inequality

α A p (x,y)+(1α) G p (x,y)< I p (x,y)<β A p (x,y)+(1β) G p (x,y)
(5.6)

holds if and only if α ( 2 / e ) p and β2/3.

The part (2) of Theorem 5.4 is a result of Trif [17].

When letting p=2 and β=2/3 in the right-hand inequality of (5.6), one can obtain the following result, which is from Sándor and Trif [18].

I 2 (x,y)< 2 3 A 2 (x,y)+ 1 3 G 2 (x,y).
(5.7)

When letting p=1 in the double inequality (5.5), one can obtain the following result (see [12], [[15], Theorem 2]).

Corollary 5.5 Let x and y be positive real numbers with xy. Then

αA(x,y)+(1α)G(x,y)<I(x,y)<βA(x,y)+(1β)G(x,y)
(5.8)

holds if and only if α2/3 and β2/e.

When letting α=2/3 in the left-hand inequality in (5.8), one can obtain the following result, which is from Sándor [19].

2 3 A(x,y)+ 1 3 G(x,y)<I(x,y).
(5.9)

Theorem 5.6 Let 0<p8/5, x and y be positive real numbers with xy. Then

α A p (x,y)+(1α) L p (x,y)< I p (x,y)<β A p (x,y)+(1β) L p (x,y)
(5.10)

holds if and only if α1/2 and β ( 2 / e ) p .

Theorem 5.6 can deduce the following result (see Zhu [15]).

Corollary 5.7 ([[15], Theorem 3])

Let x and y be positive real numbers with xy. Then

αA(x,y)+(1α)L(x,y)<I(x,y)<βA(x,y)+(1β)L(x,y)
(5.11)

holds if and only if α1/2 and β2/e.

When letting α=1/2 in the left-hand inequality of (5.11), one can obtain the following result, which is from Sándor [4, 19].

I(x,y)> A ( x , y ) + L ( x , y ) 2 .
(5.12)

Finally, we give the bounds for L p (x,y) in terms of G p (x,y) and I p (x,y), and obtain the following new result.

Theorem 5.8 Let x and y be positive real numbers with xy, and p286/693. Then

α G p (x,y)+(1α) I p (x,y)< L p (x,y)<β G p (x,y)+(1β) I p (x,y)
(5.13)

holds if and only if β1/2 and α1.

Theorem 5.8 can deduce a result of Zhu [15]:

Corollary 5.9 ([[15], Theorem 4])

Let x and y be positive real numbers with xy. Then

αG(x,y)+(1α)I(x,y)<L(x,y)<βG(x,y)+(1β)I(x,y)
(5.14)

holds if and only if β1/2 and α1.

Obviously, the right-hand side of (5.14) is an extension of the following inequality:

L(x,y)< 1 2 ( G ( x , y ) + I ( x , y ) ) ,
(5.15)

which was given by Alzer [5].