Skip to main content

Proof of an open inequality with double power-exponential functions

Abstract

Cîrtoaje (J. Nonlinear Sci. Appl. 4(2):130-137, 2011) conjectured that the inequality a ( 2 b ) x + b ( 2 a ) x 1 with double power-exponential functions holds for all nonnegative real numbers a, b with a+b=1 and all x1. In this paper, we shall prove the conjecture affirmatively.

MSC:26D10.

1 Introduction

The study of inequalities with power-exponential functions is one of the active areas of research in the mathematical analysis. Cîrtoaje [1, 2] studied inequalities with power-exponential functions and conjectured some open inequalities. He posed the open inequality as Conjecture 4.8 in [1],

a 2 b + b 2 a 1,

which holds for all nonnegative real numbers a, b with a+b=1. He proved in [2] that this inequality holds. Moreover, he conjectured the more generalized inequality containing double power-exponential functions in [2]:

a ( 2 b ) x + b ( 2 a ) x 1
(1.1)

holds for all nonnegative real numbers a, b with a+b=1 and all x1, which is Conjecture 5.1 in [2] and still an open problem. Cîrtoaje’s open inequality (1.1) is an interesting and new problem of great importance in the power exponential inequality theory. In this paper, we shall prove the conjecture affirmatively. The following is our main theorem.

Theorem 1.1 For all nonnegative real numbers a, b with a+b=1 and all x1, inequality (1.1) holds.

We shall show this theorem by using differentiation mainly.

Let

b=1a

and

F(x,a)= a ( 2 b ) x + b ( 2 a ) x .

Since F(x,0)=F(x, 1 2 )=1 and F(x,a)=F(x,1a), it suffices to show that F(x,a)1 for 0<a< 1 2 and x1. To prove Theorem 1.1, we shall investigate the sign of

F x (x,a)= x F(x,a)= b ( 2 a ) x ( 2 a ) x ln(2a)lnb+ a ( 2 b ) x ( 2 b ) x ln(2b)lna.

We set

G(x,a)=ln [ b ( 2 a ) x ( 2 a ) x ln ( 2 a ) ln b ] ln [ a ( 2 b ) x ( 2 b ) x ln ( 2 b ) ( ln a ) ] .

Then F x (x,a) clearly has the same sign as G(x,a).

Since G(x,a) has the both signs, in order to get the sign of it, we need to investigate the signs of G(1,a) and G x (1,a). We shall describe the results in Sections 2.1 and 2.2. Moreover, for fixed a(0, 1 2 ), we show in Section 2.3 that G x =G/x is strictly increasing for x1; that is, G(x,a) is convex on [1,), which is the main idea of our proof.

In Section 3, we consider the cases of G x (1,a)0 and G x (1,a)<0 to prove Theorem 1.1. Using three propositions given in Section 2, we notice the following (1) and (2).

  1. (1)

    From Proposition 2.8, for fixed a, if G x (1,a)0, then G x (x,a)>0 for x>1 and if G x (1,a)<0, then there exists uniquely a number x ˜ >1 such that G x ( x ˜ ,a)=0.

  2. (2)

    From Propositions 2.5 and 2.6, we notice that G(1,a)<0 when G x (1,a)<0.

(1) and (2) play an important role in the proof of Theorem 1.1.

We shall use the functions F(x,a) and G(x,a) defined here throughout this paper.

2 Preliminaries

2.1 The sign of G(1,a)

From the definition of G(x,a), we have

G(x,a)= ( 2 a ) x lnb+xln(2a) ( 2 b ) x lnaxln(2b)+R(a),

where

R ( a ) = ln [ ln ( 2 a ) ln b ] ln [ ln ( 2 b ) ( ln a ) ] = ln ( 1 + ln 2 ln a ) ln ( 1 ln 2 ln b ) .

Then we have

G(1,a)=2alnb+ln(2a)2blnaln(2b)+R(a),b=1a.

In this subsection, we shall show that G(1,a)<0 for 15 100 a< 1 2 .

Consider first the case 15 100 a 1 4 . We have

R (a)=ln2 [ 1 P ( a ) + 1 P ( b ) ]

and

R (a)=ln2 [ Q ( a ) P 2 ( a ) Q ( b ) P 2 ( b ) ] ,

where

P(x)=xln(2x)lnx

and

Q(x)= ( ln x ) 2 +(2+ln2)lnx+ln2.

Lemma 2.1 If 15 100 a 1 4 , then

G(1,a)<0.

Proof First, we show that

R (a)<0

for 15 100 a 1 4 . Let λ 1 and λ 2 ( λ 1 < λ 2 ) be the solutions of

x 2 +(2+ln2)x+ln2=0,

then we have

λ 1 = ( 2 + ln 2 ) ( ln 2 ) 2 + 4 2 , λ 2 = ( 2 + ln 2 ) + ( ln 2 ) 2 + 4 2

and

e λ 1 0.0903, e λ 2 0.7495.

Since Q(a)0 for e λ 1 a e λ 2 , we have Q(a)<0 for 15 100 a 1 4 . Since Q(b)0 for b e λ 2 , that is, for a1 e λ 2 0.2505, we have Q(b)>0 for 15 100 a 1 4 . Therefore, from Q(a)<0 and Q(b)>0, we get R (a)<0 for 15 100 a 1 4 .

Next, we show that

R(a)<5 ( a 1 4 ) 1

for 15 100 a 1 4 . If we set

f(a)=R(a)5 ( a 1 4 ) +1,

then

f (a)= R (a)5

and

f (a)= R (a).

Since R (a)<0 for 15 100 a 1 4 , f is strictly decreasing on the interval [ 15 100 , 1 4 ] and we have f (a) f ( 1 4 )= R ( 1 4 )5(0.0377)>0. Therefore, f is strictly increasing on the interval [ 15 100 , 1 4 ], so we have f(a)f( 1 4 )=R( 1 4 )+1(0.0363)<0. Thus, we get R(a)<5(a 1 4 )1 for 15 100 a 1 4 .

In order to complete the proof of this lemma, it suffices to show from the above inequality with respect to R(a) that g(a)<0 for 15 100 a 1 4 , where

g(a)=2alnb+ln(2a)2blnaln(2b)+5 ( a 1 4 ) 1,b=1a.

We have

g (a)=2lnb 2 a b + 1 a +2lna 2 b a + 1 b +5

and

g (a)= 2 ( 1 2 a ) a b + 1 2 a a 2 b 2 .

Since g (a)>0, g is strictly increasing on the interval [ 15 100 , 1 4 ]. Since g ( 15 100 )(2.9624)<0 and g ( 1 4 )(0.3187)>0, there exists uniquely a number c( 15 100 , 1 4 ) such that g (c)=0. Then we have g (a)<0 for 15 100 <a<c and g (a)>0 for c<a< 1 4 . Hence, g is strictly decreasing on the interval [ 15 100 ,c] and strictly increasing on the interval [c, 1 4 ]. Therefore, g(a)max{g( 15 100 ),g( 1 4 )}. Since g( 15 100 )0.0582 and g( 1 4 )0.1630, we can get g(a)<0 for 15 100 a 1 4 . □

It still remains to show that G(1,a)<0 for 1 4 a< 1 2 . Since

G(1,a)=2alnb+ln(2a)2blnaln(2b)+ln ( ln ( 2 a ) ln ( 2 b ) ) +ln ( ln b ln a ) ,b=1a,

using the substitution

a= 1 t 2 ,

we need to prove that A(t)<0 for 0<t 1 2 , where

A ( t ) = ( 1 t ) ln ( 1 + t 2 ) + ln ( 1 t ) ( 1 + t ) ln 1 t 2 ln ( 1 + t ) + ln S 1 ( t ) + ln S 2 ( t ) , S 1 ( t ) = ln ( 1 t ) ln ( 1 + t ) , S 2 ( t ) = ln 2 ln ( 1 + t ) ln 2 ln ( 1 t ) .

Lemma 2.2 If 0<t 1 2 , then

S 1 (t)<1+t+ t 2 .

Proof We need to prove that f(t)>0 for 0<t 1 2 , where

f(t)= ( 1 + t + t 2 ) ln(1+t)+ln(1t).

We have

f ( t ) = ( 1 + 2 t ) ln ( 1 + t ) + t + 1 1 + t 1 1 t , f ( t ) = 2 ln ( 1 + t ) 1 1 + t 1 ( 1 + t ) 2 1 ( 1 t ) 2 + 3

and

f (t)= 2 1 + t + 1 ( 1 + t ) 2 + 2 ( 1 + t ) 3 2 ( 1 t ) 3 .

If we set g(t)= f (t)× ( 1 + t ) 3 ( 1 t ) 3 , then we have

g ( t ) = 2 t 5 + t 4 + 2 t 3 4 t 2 16 t + 3 , g ( t ) = 10 t 4 + 4 t 3 + 6 t 2 8 t 16 .

From

g (t)<4 t 3 +6 t 2 16< 1 2 + 3 2 16<0,

it follows that g is strictly decreasing on (0, 1 2 ). Since g(0)=3 and g( 1 2 )= 23 4 , there exists uniquely a number c 1 (0, 1 2 ) such that g( c 1 )=0. Since g(t)>0 for 0<t< c 1 and g(t)<0 for c 1 <t< 1 2 , we have f (t)>0 for 0<t< c 1 and f (t)<0 for c 1 <t< 1 2 . It follows that f is strictly increasing on the interval (0, c 1 ) and strictly decreasing on the interval ( c 1 , 1 2 ). Since f (0)=0 and f ( 1 2 )=2ln 3 2 19 9 (1.3001)<0, there exists uniquely a number c 2 (0, 1 2 ) such that f ( c 2 )=0. Since f (t)>0 for 0<t< c 2 and f (t)<0 for c 2 <t< 1 2 , f is strictly increasing on the interval (0, c 2 ) and strictly decreasing on the interval ( c 2 , 1 2 ). Since f (0)=0 and f ( 1 2 )=2ln 3 2 5 6 (0.0224)<0, there exists uniquely a number c 3 (0, 1 2 ) such that f ( c 3 )=0. Hence, f (t)>0 for 0<t< c 3 and f (t)<0 for c 3 <t< 1 2 . Thus, f is strictly increasing on the interval (0, c 3 ) and strictly decreasing on the interval ( c 3 , 1 2 ). Since f(0)=0 and f( 1 2 )= 7 4 ln 3 2 ln20.0164, we can get f(t)>0 for 0<t 1 2 . □

Lemma 2.3 If 0<t 1 2 , then

ln S 2 (t)< 2 ln 2 t.

Proof We need to show that f(t)<0, where

f(t)=ln S 2 (t)+ 2 ln 2 t.

We have

f (t)= ( ln S 2 ( t ) ) + 2 ln 2

and

f (t)= ( ln S 2 ( t ) ) ,

where

ln S 2 ( t ) = ln [ ln 2 ln ( 1 + t ) ] ln [ ln 2 ln ( 1 t ) ] , ( ln S 2 ( t ) ) = 1 [ ln 2 ln ( 1 + t ) ] ( 1 + t ) 1 [ ln 2 ln ( 1 t ) ] ( 1 t ) ,

and

( ln S 2 ( t ) ) = 1 + ln 2 ln ( 1 + t ) [ ln 2 ln ( 1 + t ) ] 2 ( 1 + t ) 2 + 1 ln 2 + ln ( 1 t ) [ ln 2 ln ( 1 t ) ] 2 ( 1 t ) 2 .

We see that ( ln S 2 ( t ) ) has the same sign as

B ( t ) = ( ln S 2 ( t ) ) × [ ln 2 ln ( 1 + t ) ] 2 ( 1 + t ) 2 [ ln 2 ln ( 1 t ) ] 2 ( 1 t ) 2 = ( 1 ln 2 ) B 1 ( t ) + B 2 ( t ) + B 3 ( t ) ,

where

B 1 ( t ) = [ ln 2 ln ( 1 + t ) ] 2 ( 1 + t ) 2 [ ln 2 ln ( 1 t ) ] 2 ( 1 t ) 2 , B 2 ( t ) = ln ( 1 + t ) [ ln 2 ln ( 1 t ) ] 2 ( 1 t ) 2

and

B 3 (t)=ln(1t) [ ln 2 ln ( 1 + t ) ] 2 ( 1 + t ) 2 .

We have

B 1 (t)= f 1 (t) f 2 (t),

where

f 1 (t)= [ ln 2 ln ( 1 + t ) ] (1+t)+ [ ln 2 ln ( 1 t ) ] (1t)

and

f 2 (t)= [ ln 2 ln ( 1 + t ) ] (1+t) [ ln 2 ln ( 1 t ) ] (1t).

Since ln2>ln(1+t) and ln2>ln(1t) for 0<t 1 2 , we have f 1 (t)>0. Since f 2 (t)=2ln22ln(1 t 2 )4ln22ln30.3260 for 0<t 1 2 , f 2 is strictly decreasing on the interval (0, 1 2 ]. Therefore, we can get f 2 (t)< lim t 0 f 2 (t)=0 for 0<t 1 2 . From f 1 (t)>0 and f 2 (t)<0, it follows that B 1 (t)<0. Since B 1 (t)<0, B 2 (t)<0 and B 3 (t)<0, we get B(t)<0, hence ( ln S 2 ( t ) ) <0 for 0<t 1 2 . Thus, f is strictly decreasing on the interval (0, 1 2 ] and we have f (t)< lim t 0 f (t)=0 for 0<t 1 2 . Since f is strictly decreasing on the interval (0, 1 2 ], we have f(t)< lim t 0 f(t)=0 for 0<t 1 2 . Therefore, we get ln S 2 (t)< 2 ln 2 t. □

Lemma 2.4 If 1 4 a 1 2 , then

G(1,a)<0.

Proof We need to show that A(t)<0 for 0<t 1 2 . By Lemmas 2.2 and 2.3, it suffices to show that f(t)<0, where

f(t)=(1t)ln 1 + t 2 +ln(1t)(1+t)ln 1 t 2 ln(1+t)+ln ( 1 + t + t 2 ) 2 ln 2 t.

We have

f (t)=ln(1+t)ln(1t)+ 1 1 + t + 1 1 t + 1 + 2 t 1 + t + t 2 +2ln22 2 ln 2

and

f ( t ) = 2 t ( 1 + t ) ( 1 t ) 2 t ( 1 + t ) ( 1 + t + t 2 ) 2 + 4 t ( 1 + t ) 2 ( 1 t ) 2 + 1 ( 1 + t + t 2 ) 2 = 2 t 2 ( 1 + 4 t + 3 t 2 + t 3 ) ( 1 + t ) ( 1 t ) ( 1 + t + t 2 ) 2 + 4 t ( 1 + t ) 2 ( 1 t ) 2 + 1 ( 1 + t + t 2 ) 2 .

Since f (t)>0 for 0<t 1 2 , f is strictly increasing on the interval (0, 1 2 ]. Since lim t 0 f (t)=1+2ln2 2 ln 2 (0.4990)<0 and f ( 1 2 )=4ln2+ 38 21 ln3 2 ln 2 (0.5981)>0, there exists uniquely a number c(0, 1 2 ) such that f (c)=0. Hence, f is strictly decreasing on the interval (0,c) and strictly increasing on the interval (c, 1 2 ]. Since lim t + 0 f(t)=0 and f( 1 2 )= 1 2 ln3+ln7 1 ln 2 0.0460, we get f(t)<0. Therefore, A(t)<f(t)<0 for 0<t 1 2 . □

From Lemmas 2.1 and 2.4, we get the following result.

Proposition 2.5 If 15 100 a< 1 2 , then

G(1,a)<0.

We notice that lim a 1 2 0 G(1,a)=0.

2.2 The sign of G x (1,a)

We have

G x (x,a)= x G(x,a)= ( 2 a ) x ln(2a)lnb+ln(2a) ( 2 b ) x ln(2b)lnaln(2b),

hence

G x (1,a)=(2a)(ln2+lna)lnb+lna2b(ln2+lnb)lnalnb,

where b=1a.

Proposition 2.6 There exists a number c( 15 100 , 1 2 ) such that G x (1,a)>0 for 0<a<c.

Proof Let us denote G x (1,a) by f(a). We have

f (a)=2ln(2a)lnb+2lnb 2 a b ln(2a)+ 1 a +2ln(2b)lna+2lna 2 b a ln(2b)+ 1 b

and

f ( a ) = 2 ( b ln b a ln a ) a b 2 ln ( 2 a ) b 2 ln ( 2 a ) b 2 + 2 ln ( 2 b ) a + 2 ln ( 2 b ) 1 a 2 + 4 ( 1 2 a ) a b + 1 b 2 .

Since 0<a< 1 2 , we have blnbalna>0, ln(2a)<0, ln(2b)>0 and 12a>0. Therefore, if 2ln(2b)10, then f (a)>0. The condition 2ln(2b)10 is true for 0<a< a 0 , where

a 0 =1 1 2 e 0.1756.

Consequently, f is strictly increasing on (0, a 0 ]. Since f ( a 0 )2.5412, it follows that f (a)<0 on (0, a 0 ], and f is strictly decreasing on (0, a 0 ]. Since lim a + 0 f(a)= and f( a 0 )0.0413, there exists uniquely a number c(0, a 0 ) such that f(c)=0. Then we have f(a)>0 for 0<a<c and f(a)<0 for c<a< a 0 . Since 15 100 < a 0 and f( 15 100 )0.0354, we can get 15 100 <c. □

2.3 The convexity of G(x,a)

In order to investigate the convexity of G(x,a) with respect to x, we need the following lemma.

Lemma 2.7 If 0<a< 1 2 , then

( ln 2 + ln a ) 2 lnb> ( ln 2 + ln b ) 2 lna,

where b=1a.

Proof We first show that the inequality

( ln 2 ) 2 >lnalnb

holds for 0<a< 1 2 . We denote

f(a)=lnalnb.

Then we have

f (a)= b ln b a ln a a b .

We set

g(a)=blnbalna,

then we have

g (a)=lnblna2

and

g (a)= 2 a 1 a b <0.

Therefore, g is strictly decreasing on the interval (0, 1 2 ). Since lim a + 0 g (a)= and g ( 1 2 )=2ln22<0, there exists uniquely a number c(0, 1 2 ) such that g (c)=0. Since g (a)>0 for 0<a<c and g (a)<0 for c<a< 1 2 , g is strictly increasing on the interval (0,c) and strictly decreasing on the interval (c, 1 2 ). Since lim a + 0 g(a)=0 and g( 1 2 )=0, we get g(a)>0 for 0<a< 1 2 . Therefore, we have f (a)>0 for any a(0, 1 2 ). Since f is strictly increasing on the interval (0, 1 2 ), we can get f(a)<f( 1 2 )= ( ln 2 ) 2 . Hence, we have ( ln 2 ) 2 >lnalnb for 0<a< 1 2 . Also, the inequality

( ln 2 + ln a ) 2 lnb> ( ln 2 + ln b ) 2 lna

is equivalent to

(lnblna) ( ( ln 2 ) 2 ln a ln b ) >0.

From lnblna>0 and ( ln 2 ) 2 lnalnb>0, it follows that (lnblna)( ( ln 2 ) 2 lnalnb)>0. This completes the proof of the lemma. □

Proposition 2.8 If 0<a< 1 2 and x1, then G x is strictly increasing with respect to x.

Proof For fixed a(0, 1 2 ), let us denote f(x)= G x (x,a); that is,

f(x)= ( 2 a ) x ln(2a)lnb+ln(2a) ( 2 b ) x ln(2b)lnaln(2b),b=1a.

Clearly, we need to show that f (x)>0 for x1. From

f (x)= ( 2 a ) x ( ln 2 + ln a ) 2 lnb ( 2 b ) x ( ln 2 + ln b ) 2 lna,

we can write the inequality f (x)>0 in the form

( a b ) x ( ln 2 + ln a ) 2 < ( ln 2 + ln b ) 2 ln a ln b .

Since ( a b ) x <1, it is enough to show that

( ln 2 + ln a ) 2 < ( ln 2 + ln b ) 2 ln a ln b ,

which follows immediately from Lemma 2.7. □

3 Proof of Theorem 1.1

In this section, we shall prove the main theorem.

Proof We shall show that if 0a 1 2 and x1, then F(x,a)1. The inequality F(1,a)1 is proved in Cîrtoaje [2]. Since F(x,0)=F(x, 1 2 )=1, we may show this inequality for 0<a< 1 2 . We note that G x (1,a) has the both signs, so we consider the cases of G x (1,a)0 and G x (1,a)<0.

  1. (1)

    We first assume that G x (1,a)0. We have G x (x,a)> G x (1,a) for x>1 from Proposition 2.8. Since we have G x (x,a)>0 on the assumption, G is strictly increasing with respect to x. Therefore, G(x,a)>G(1,a) for x>1. We note that G(1,a) has the both signs, so we consider the cases of G(1,a)0 and G(1,a)<0.

    1. (i)

      If G(1,a)0, then G(x,a)>0, so we have F x (x,a)>0 for x>1. Hence, F is strictly increasing with respect to x, and we have F(x,a) lim x F(x,a)=1.

    2. (ii)

      If G(1,a)<0, then since lim x G(x,a)=, there exists uniquely a number x 1 >1 satisfying G( x 1 ,a)=0. Since G(x,a)<0 for 1<x< x 1 and G(x,a)>0 for x> x 1 , we have F x (x,a)<0 for 1<x< x 1 and F x (x,a)>0 for x> x 1 . Hence, F is strictly decreasing for 1<x< x 1 and strictly increasing for x> x 1 . Therefore, we get

      F(x,a)max { F ( 1 , a ) , lim x F ( x , a ) } =1.
  2. (2)

    We next assume that G x (1,a)<0. Since G x is strictly increasing with respect to x from Proposition 2.8 and lim x G x (x,a)=, there exists uniquely a number x 2 >1 satisfying G x ( x 2 ,a)=0. Since G x (x,a)<0 for 1<x< x 2 and G x (x,a)>0 for x> x 2 , G is strictly decreasing for 1<x< x 2 and strictly increasing for x> x 2 . By the assumption G x (1,a)<0 and Proposition 2.6, it follows that a> 15 100 . Hence, G(1,a)<0 by Proposition 2.5. From lim x G(x,a)=, there exists uniquely a number x 3 > x 2 satisfying G( x 3 ,a)=0. If 1<x< x 3 , then G(x,a)<0, so F x (x,a)<0. If x> x 3 , then G(x,a)>0, so F x (x,a)>0. Hence F is strictly decreasing for 1<x< x 3 and strictly increasing for x> x 3 . So, we get

    F(x,a)max { F ( 1 , a ) , lim x F ( x , a ) } =1.

This completes the proof of Theorem 1.1. □

References

  1. Cîrtoaje V: On some inequalities with power-exponential functions. JIPAM. J. Inequal. Pure Appl. Math. 2009., 10(1): Article ID 21

    Google Scholar 

  2. Cîrtoaje V: Proofs of three open inequalities with power-exponential functions. J. Nonlinear Sci. Appl. 2011, 4(2):130–137.

    MATH  MathSciNet  Google Scholar 

Download references

Acknowledgements

We would like to thank the referees for their valuable comments according to which various parts of this paper have been cleared and improved.

Author information

Authors and Affiliations

Authors

Corresponding author

Correspondence to Yusuke Nishizawa.

Additional information

Competing interests

The authors declare that they have no competing interests.

Authors’ contributions

MM had the first drift of the paper and YN added some contents and re-organized the paper. All authors read and approved the final manuscript.

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

Miyagi, M., Nishizawa, Y. Proof of an open inequality with double power-exponential functions. J Inequal Appl 2013, 468 (2013). https://doi.org/10.1186/1029-242X-2013-468

Download citation

  • Received:

  • Accepted:

  • Published:

  • DOI: https://doi.org/10.1186/1029-242X-2013-468

Keywords