Next Article in Journal
Definite Integral of Arctangent and Polylogarithmic Functions Expressed as a Series
Previous Article in Journal
Evaluating the Suitability of a Smart Technology Application for Fall Detection Using a Fuzzy Collaborative Intelligence Approach
 
 
Font Type:
Arial Georgia Verdana
Font Size:
Aa Aa Aa
Line Spacing:
Column Width:
Background:
Article

A Solution to Qi’s Conjecture on a Double Inequality for a Function Involving the Tri- and Tetra-Gamma Functions

by
Ladislav Matejíčka
Faculty of Industrial Technologies in Púchov, Trenčín University of Alexander Dubček in Trenčín, I. Krasku 491/30, 02001 Púchov, Slovakia
Mathematics 2019, 7(11), 1098; https://doi.org/10.3390/math7111098
Submission received: 10 October 2019 / Revised: 4 November 2019 / Accepted: 8 November 2019 / Published: 13 November 2019
(This article belongs to the Section Mathematics and Computer Science)

Abstract

:
In the paper, the author gives a solution to a conjecture on a double inequality for a function involving the tri- and tetra-gamma functions, which was first posed in Remark 6 of the paper “Complete monotonicity of a function involving the tri- and tetragamma functions” (2015) and repeated in the seventh open problem of the paper “On complete monotonicity for several classes of functions related to ratios of gamma functions” (2019).

1. Introduction

It is common knowledge that the classical Euler’s gamma function [1,2] is defined by
Γ ( x ) = 0 t x 1 e t d t
for x > 0 and the digamma function [3] is defined as the logarithmic derivative of the gamma function.
ψ ( x ) = Γ ( x ) Γ ( x ) .
The functions ψ , ψ , ψ , ψ , … are known as polygamma functions [4].
Very recently, in the paper [5], F. Qi and R. P. Agarwal surveyed some results related to the function ψ 2 + ψ . They also posed eight open problems. The goal of the paper is to find a solution of the seventh open problem which was first posed as a conjecture in Remark 6 of the paper [6].
The seventh open problem states that the double inequality
1 x 4 x 2 + 4 x + 12 12 ( 1 + x ) 2 α < ψ ( x ) 2 + ψ ( x ) < 1 x 4 x 2 + 4 x + 12 12 ( 1 + x ) 2 β
holds on ( 0 , ) if and only if α 6 / 5 and β 1 .

2. The Key Lemmas

In this section, we prove two important lemmas.
Lemma 1.
Let Δ ( x ) = ψ ( x ) 2 + ψ ( x ) and
ϕ ( α , x ) = Δ ( x ) 1 x 4 x 2 + 4 x + 12 12 ( 1 + x ) 2 α
for x > 0 and α R . Then
lim x + ϕ ( α , x ) = 0 ,
lim x 0 + x 4 ϕ ( α , x ) = 0 ,
lim x 0 + d [ x 4 ϕ ( α , x ) ] d x = 2 + 5 α 3 .
Proof of Lemma 1. 
To prove (1) it suffices to show lim x + Δ ( x ) = 0 which follows from the double inequality
x 2 + 12 12 x 4 ( 1 + x ) 2 < Δ ( x ) < x 2 + 4 x + 12 12 x 4 ( 1 + x ) 2
(see [5], p. 9). Making use of the previous double inequality yields
lim x 0 + x 4 Δ ( x ) x 2 + 4 x + 12 12 ( 1 + x ) 2 α = 0 = lim x 0 + x 4 ϕ ( α , x )
which is (2).
Simple calculation brings
d x 4 ϕ ( α , x ) d x = 4 x 3 Δ ( x ) + x 4 2 ψ ( x ) ψ ( x ) + ψ ( x ) + α x 2 + 4 x + 12 12 ( 1 + x ) 2 α 1 x + 10 6 ( 1 + x ) 3 .
Denote
δ ( x ) = 4 x 3 Δ ( x ) + x 4 2 ψ ( x ) ψ ( x ) + ψ ( x ) .
To prove (3) it suffices to show
lim x 0 + δ ( x ) = lim x 0 + 4 x 3 ψ ( x ) 2 + ψ ( x ) + x 4 2 ψ ( x ) ψ ( x ) + ψ ( x ) = 2 .
The function δ ( x ) can be rewritten as δ ( x ) = δ 1 ( x ) + δ 2 ( x ) where
δ 1 ( x ) = 2 x 3 ψ ( x ) 2 ψ ( x ) + x ψ ( x ) ,
δ 2 ( x ) = x 3 4 ψ ( x ) + x ψ ( x ) .
Making use of well known formulas:
for polygamma functions
ψ ( x ) = n = 0 1 ( n + x ) 2 , ψ ( x ) = 2 n = 0 1 ( n + x ) 3 , ψ ( x ) = 6 n = 0 1 ( n + x ) 4 ,
and for values of the Riemann zeta function [7]
ζ ( 2 ) = n = 1 1 n 2 = π 2 6 , ζ ( 3 ) = n = 1 1 n 3 = 1.20205 , ζ ( 4 ) = n = 1 1 n 4 = π 4 90
gives
2 x 3 ψ ( x ) < 2 x + 2 x 3 n = 1 1 n 2 = 2 x + x 3 π 2 3
and
2 ψ ( x ) + x ψ ( x ) = 2 n = 1 1 ( n + x ) 2 2 x n = 1 1 ( n + x ) 3 π 2 3 + 2 x ζ ( 3 ) .
lim x 0 + δ 1 ( x ) = lim x 0 + 2 x 3 ψ ( x ) + x ψ ( x ) = 0 .
The equality (4) can be rewritten as
δ 2 ( x ) = 4 x 3 2 x 3 2 n = 1 1 ( n + x ) 3 + x 4 6 x 4 + 6 n = 1 1 ( n + x ) 4 = 2 8 x 3 n = 1 1 ( n + x ) 3 + 6 x 4 n = 1 1 ( n + x ) 4 .
Because of
n = 1 1 ( n + x ) 3 < ζ ( 3 )
and
n = 1 1 ( n + x ) 4 < ζ ( 4 ) = π 4 90 ,
we obtain lim x 0 + δ 2 ( x ) = 2 . The proof of Lemma 1 is complete. □
Lemma 2.
Let
s ( x ) = ϕ 6 5 , x = Δ ( x ) 1 x 4 x 2 + 4 x + 12 12 ( 1 + x ) 2 6 / 5
for x > 0 . Then s ( x ) > 0 for x > 0 .
Proof of Lemma 2. 
Consider three cases
(a)
5.7 x < ,
(b)
1.27 x < 5.7 ,
(c)
0 < x 1.27 .
The case (a).
In the paper (see [5], p. 9), it was presented that
Δ ( x ) > 1 x 4 x 2 + 12 12 ( 1 + x ) 2
for x > 0 . So the case (a) will be done if we show
1 x 4 x 2 + 12 12 ( 1 + x ) 2 1 x 4 x 2 + 4 x + 12 12 ( 1 + x ) 2 6 / 5 ,
which is equivalent to
s 1 ( x ) = log ( x 2 + 12 ) 6 5 log ( x 2 + 4 x + 12 ) + 1 5 log ( 12 ) + 2 5 log ( 1 + x ) > 0
for 5.7 x < . In order to prove s 1 ( x ) > 0 it is sufficient to show s 1 ( 5.7 ) > 0 and s 1 ( x ) > 0 for 5.7 < x < . The inequality s 1 ( x ) > 0 is equivalent to
g ( x ) = 12 ( x 2 + 12 ) 5 ( 1 + x ) 2 ( x 2 + 4 x + 12 ) 6 > 0 .
Easy computation gives g ( 5.7 ) 1.06 10 09 > 0 , so s 1 ( 5.7 ) > 0 .
Differentiation yields
s 1 ( x ) = x ( 22 x 2 + 40 x 216 ) 5 ( x 2 + 12 ) ( x + 1 ) ( x 2 + 4 x + 12 ) .
Because of
22 x 2 + 40 x 216 = 2 ( 11 x 2 + 20 x 108 ) = 2 x 2 11 322 5 x + 2 11 322 + 5 = 2 ( x + 4.1717 ) ( x 2.3535 )
the function s 1 ( x ) is positive for x > 2 11 322 5 . The proof of the case (a) is complete.
The case (b). Let 1.27 x < 5.7 . Using the following formulas
ψ ( x ) = 1 x 2 + ψ ( 1 + x ) , ψ ( x ) = 2 x 3 + ψ ( 1 + x )
for x > 0 yields
Δ ( x ) = ψ ( x ) 2 + ψ ( x ) = 1 x 4 2 x 3 + 2 x 2 ψ ( 1 + x ) + Δ ( 1 + x ) .
Making use of the inequality (see [5], p. 9)
Δ ( x ) > 1 x 4 x 2 + 12 12 ( x + 1 ) 2 .
yields
Δ ( x ) > z ( x ) = 1 x 4 2 x 3 + 2 x 2 ψ ( 1 + x ) + 1 ( x + 1 ) 4 ( x + 1 ) 2 + 12 12 ( x + 2 ) 2 .
To prove the case (b) it suffices to show
α ( x ) = z ( x ) 1 x 4 x 2 + 4 x + 12 12 ( x + 1 ) 2 6 / 5 > 0 .
It will be done if we prove
  • z ( x ) > 0 f o r 1.27 x < 5.7 ,
  • F ( 1.27 ) 0 ,
  • F ( x ) > 0 f o r 1.27 x < 5.7
where
F ( x ) = log ( z ( x ) ) + 4 log ( x ) 6 5 log x 2 + 4 x + 12 12 ( x + 1 ) 2 .
Differentiating F yields
F ( x ) = 4 x 5 + 6 x 4 4 x 3 ψ ( 1 + x ) + 2 x 2 ψ ( 1 + x ) + 2 x 3 7 x 2 44 x 63 6 ( x + 1 ) 5 ( x + 2 ) 3 1 x 4 2 x 3 + 2 x 2 ψ ( 1 + x ) + 1 ( 1 + x ) 4 ( x + 1 ) 2 + 12 12 ( x + 2 ) 2 + 4 x + 6 5 2 x + 20 ( x + 1 ) ( x 2 + 4 x + 12 ) .
It is clear that, F ( 1.27 ) 0 is equivalent to α ( 1.27 ) 0 .
In the paper [8], inequality (2.3) it was established that
ψ ( 1 + x ) > 1 x 1 2 x 2 + 1 6 x 3 1 30 x 5 + 1 42 x 7 1 30 x 9 = 210 x 8 105 x 7 + 35 x 6 7 x 4 + 5 x 2 7 210 x 9
for x > 0 .
Using (6) gives
ψ ( 2.27 ) = 1 2.27 2 + ψ ( 3.27 ) > 1 2.27 2 + 210 × 2.27 8 105 × 2.27 7 + 35 × 2.27 6 7 × 2.27 4 + 5 × 2.27 2 7 210 × 2.27 9 > 0.551 > 0.55 .
By (7), it follows that
z ( x ) > 1 x 4 2 x 3 + 2 x 2 0.55 + 1 ( x + 1 ) 4 ( x + 1 ) 2 + 12 12 ( x + 2 ) 2 = q ( x ) = 66 x 8 + 408 x 7 + 821 x 6 + 274 x 5 949 x 4 960 x 3 + 324 x 2 + 720 x + 240 60 x 4 ( x + 1 ) 4 ( x + 2 ) 2 .
Putting x = 1.27 in q ( x ) yields z ( 1.27 ) > q ( 1.27 ) = 0.0950575080326270 > 0.095 .
Because of
1 . 27 4 × 0.095 5 1 . 27 2 + 4 × 1 . 27 + 12 12 × 2 . 27 2 6 1.58 10 04 > 0
we have α ( 1.27 ) > 0 . So F ( 1.27 ) > 0 .
Next, we show z ( x ) > 0 for 1.27 < x < 5.7 . It is well known [9] that
ψ ( x ) > 1 x + 1 2 x 2 f o r x > 0 .
This implies that
z ( x ) > z 1 ( x ) = 1 x 4 2 x 3 + 2 x 2 1 x + 1 + 1 2 ( x + 1 ) 2 + 1 ( x + 1 ) 4 ( x + 1 ) 2 + 12 12 ( x + 2 ) 2 = x 6 + 2 x 5 + 25 x 4 + 72 x 3 + 156 x 2 + 144 x + 48 12 x 4 ( x + 1 ) 4 ( x + 2 ) 2 > 0 .
Finally, we show F ( x ) > 0 for for 1.27 x < 5.7 . The inequality F ( x ) > 0 is equivalent to
G ( x ) = 4 x 5 + 6 x 4 4 x 3 ψ ( 1 + x ) + 2 x 2 ψ ( 1 + x ) 2 x 3 + 7 x 2 + 44 x + 63 6 ( x + 1 ) 5 ( x + 2 ) 3 + 1 x 4 2 x 3 + 2 x 2 ψ ( 1 + x ) + ( x + 1 ) 2 + 12 12 ( 1 + x ) 4 ( x + 2 ) 2 × 4 x + 6 5 2 x + 20 ( x + 1 ) ( x 2 + 4 x + 12 ) > 0 .
The inequality G ( x ) > 0 may be rearranged as
G ( x ) = ψ ( 1 + x ) 20 x 3 + 124 x 2 + 560 x + 240 5 x 3 ( x + 1 ) ( x 2 + 4 x + 12 ) + 2 ψ ( 1 + x ) x 2 1 30 x 3 ( x + 1 ) 5 ( x + 2 ) 3 ( x 2 + 4 x + 12 ) × 60 x 9 + 1044 x 8 + 9267 x 7 + 47554 x 6 + 147543 x 5 + 285108 x 4 + 344260 x 3 + 254352 x 2 + 105984 x + 18624 > 0 .
In the paper [10], by using asymptotic expansion, it was deduced that
ψ ( x ) = 1 x 2 1 x 3 1 2 x 4 + 1 6 x 6 θ 6 x 8 f o r x > 0 , 0 θ 1 .
This implies
ψ ( x ) 1 x 2 1 x 3 1 2 x 4 + 1 6 x 6 1 6 x 8 f o r x > 0 .
Utilizing (6), (8) we obtain that G ( x ) G 1 ( x ) , where
G 1 ( x ) = 1 x 1 2 x 2 + 1 6 x 3 1 30 x 5 + 1 42 x 7 1 30 x 9 20 x 3 + 124 x 2 + 560 x + 240 5 x 3 ( x + 1 ) ( x 2 + 4 x + 12 ) 2 x 2 1 ( 1 + x ) 2 + 1 ( 1 + x ) 3 + 1 2 ( 1 + x ) 4 1 6 ( 1 + x ) 6 + 1 6 ( 1 + x ) 8 1 30 x 3 ( x + 1 ) 5 ( x + 2 ) 3 ( x 2 + 4 x + 12 ) × 60 x 9 + 1044 x 8 + 9267 x 7 + 47554 x 6 + 147543 x 5 + 285108 x 4 + 344260 x 3 + 254352 x 2 + 105984 x + 18624 .
Rewriting G 1 ( x ) yields G 1 ( x ) = φ ( x ) / ϕ ( x ) , where
φ ( x ) = 385 x 19 + 1435 x 18 + 2100 x 17 + 18662 x 16 + 153421 x 15 + 477255 x 14 + 718276 x 13 + 568208 x 12 + 232044 x 11 + 38256 x 10 46764 x 9 314648 x 8 1233916 x 7 2806448 x 6 3866740 x 5 3395384 x 4 1933504 x 3 697344 x 2 145600 x 13440 , ϕ ( x ) = 1050 x 12 ( x + 1 ) 8 ( x + 2 ) 3 ( x 2 + 4 x + 12 ) .
To show the positivity of the function G 1 ( x ) it suffices to prove φ ( x ) > 0 . By virtue of x 1.27 we see that φ ( x ) n 1 ( x ) n 2 ( x ) n 3 ( x ) where the functions n 1 ( x ) , n 2 ( x ) , n 3 ( x ) are derived in Appendix A.
It is obvious that the proof of φ ( x ) > 0 will be done if we show n 3 ( 1.27 ) > 0 , n 3 ( 1.27 ) > 0 , n 3 ( x ) > 0 on ( 1.27 , 5.7 ) .
A direct differentiation yields
n 3 ( x ) = 14197117854302775 274877906944 x 5 + 44097108486849535 274877906944 x 4 + 1613308847079861 8589934592 x 3 + 1424065194179289 2147483648 x 2 328178783611353 536870912 x + 7951559879725953 8589934592
and
n 3 ( x ) = 70985589271513875 274877906944 x 4 + 44097108486849535 68719476736 x 3 + 4839926541239583 8589934592 x 2 + 1424065194179289 1073741824 x 328178783611353 536870912 .
From n 3 ( x ) > 0 , n 3 ( 1.27 ) 3.96 10 06 we deduce n 3 ( x ) is a convex function on ( 1.27 , 5.7 ) . The proof of the case (b) follows from n 3 ( 1.27 ) 2.12 10 06 > 0 , n 3 ( 1.27 ) 2.19 10 06 > 0 .
The case (c). We have 0 < x 1.27 . Using
ψ ( x ) 2 = 1 x 4 + 2 x 2 ψ ( 1 + x ) + ψ ( 1 + x ) 2 ,
ψ ( x ) = 2 x 3 + ψ ( 1 + x )
gives
s ( x ) = ϕ 6 5 , x = Δ ( x ) 1 x 4 x 2 + 4 x + 12 12 ( x + 1 ) 2 6 / 5 = ψ ( x ) 2 + ψ ( x ) 1 x 4 x 2 + 4 x + 12 12 ( x + 1 ) 2 6 / 5 = 1 x 4 2 x 3 + 2 x 2 ψ ( 1 + x ) + ψ ( 1 + x ) 2 + ψ ( 1 + x ) 1 x 4 x 2 + 4 x + 12 12 ( x + 1 ) 2 6 / 5 = 1 x 4 2 x 3 + 2 x 2 ψ ( 1 + x ) + Δ ( 1 + x ) 1 x 4 x 2 + 4 x + 12 12 ( x + 1 ) 2 6 / 5 .
Replacing x by 1 + x in (5) yields
Δ ( 1 + x ) 1 ( 1 + x ) 4 ( 1 + x ) 2 + 12 12 ( 2 + x ) 2 .
So
s ( x ) = 1 x 4 2 x 3 + 2 x 2 ψ ( 1 + x ) + Δ ( 1 + x ) 1 x 4 x 2 + 4 x + 12 12 ( x + 1 ) 2 6 / 5 t ( x ) ,
where
t ( x ) = 1 x 4 2 x 3 + 2 x 2 ψ ( 1 + x ) + ( 1 + x ) 2 + 12 12 ( 1 + x ) 4 ( 2 + x ) 2 1 x 4 x 2 + 4 x + 12 12 ( x + 1 ) 2 6 / 5 .
The inequality t ( x ) 0 is equivalent to
r ( x ) = 12 ( 1 + x ) 4 ( 2 + x ) 2 ( 1 2 x ) + x 4 ( ( 1 + x ) 2 + 12 ) + 24 x 2 ( 1 + x ) 4 ( 2 + x ) 2 ψ ( 1 + x ) 12 ( 1 + x ) 4 ( 2 + x ) 2 × ( x 2 + 4 x + 12 ) / ( 12 ( 1 + x ) 2 ) 6 / 5 0 .
Making use of the following inequality (see [11], p. 6)
ψ ( x ) > 1 x 2 + 1 x + 1 | ψ ( 1 ) |
we come to the conclusion that r ( x ) h ( x ) , where
h ( x ) = 12 ( 1 + x ) 4 ( 2 + x ) 2 ( 1 2 x ) + 24 x 2 ( 1 + x ) 4 ( 2 + x ) 2 ( 1 / ( 1 + x ) 2 + 1 / ( 2 + x ) 2 + 1 / ( 3 + x ) 2 + 1 / ( 3 + x + 1 / 1.64 ) ) + x 4 ( ( 1 + x ) 2 + 12 ) 12 ( 1 + x ) 4 ( 2 + x ) 2 ( x 2 + 4 x + 12 ) / ( 12 ( 1 + x ) 2 ) 6 / 5 .
(We used ψ ( 1 ) = π 2 / 6 > 1.64 . )
The inequality h ( x ) 0 is equivalent to
T ( x ) = log 12 ( 1 + x ) 4 ( 2 + x ) 2 ( 1 2 x ) + 24 x 2 ( 1 + x ) 4 ( 2 + x ) 2 1 / ( 1 + x ) 2 + 1 / ( 2 + x ) 2 + 1 / ( 3 + x ) 2 + 1 / ( 3 + x + 1 / 1.64 ) + x 4 ( 1 + x ) 2 + 12 log 12 ( 1 + x ) 4 ( 2 + x ) 2 log ( x 2 + 4 x + 12 ) / ( 12 ( 1 + x ) 2 ) 6 / 5 0 .
Differentiating T ( x ) yields T ( x ) = ϱ ( x ) / τ ( x ) , where
ϱ ( x ) = 190374 x 13 4286906 x 12 45015548 x 11 282667076 x 10 1146875462 x 9 3046852634 x 8 5055591704 x 7 4276245096 x 6 + 386216448 x 5 + 4183878144 x 4 + 3096950400 x 3 + 875102976 x 2 + 292571136 x
and
τ ( x ) = 5 ( 41 x + 148 ) ( x + 1 ) ( x + 2 ) ( x + 3 ) ( x 2 + 4 x + 12 ) × 67 x 9 412 x 8 + 1054 x 7 + 20728 x 6 + 98457 x 5 + 251532 x 4 + 404436 x 3 + 425376 x 2 + 252144 x + 63936 .
To prove T ( x ) 0 it suffices to show T ( 0 ) 0 , T ( 1.27 ) 0 , τ ( x ) > 0 , ϱ ( 0 ) = 0 , ϱ ( 0 ) > 0 , ϱ ( 1.27 ) < 0 , ϱ ( x ) = 0 has only one real root in ( 0 , 1.27 ) . It is evident that if h ( 1.27 ) > 0 then T ( 1.27 ) > 0 . Some calculations give T ( 0 ) = 0 , h ( 1.27 ) 2.70 . So T ( 1.27 ) > 0 . The inequality τ ( x ) > 0 is equivalent to
d 1 ( x ) = 67 x 9 412 x 8 + 1054 x 7 + 20728 x 6 + 98457 x 5 + 251532 x 4 + 404436 x 3 + 425376 x 2 + 252144 x + 63936 > 0 .
Because of
d 1 ( x ) > d 2 ( x ) = ( 67 × 1 . 27 2 412 × 1.27 + 1054 ) x 7 + 20728 x 6 + 98457 x 5 + 251532 x 4 + 404436 x 3 + 425376 x 2 + 252144 x + 63936 ,
and 67 × 1 . 27 2 412 × 1.27 + 1054 = 422.6957 we obtain τ ( x ) > 0 . Direct computation yields ϱ ( 0 ) = 0 , ϱ ( 1.27 ) 5.88 10 10 , ϱ ( 0 ) = 292,571,136. Further, we prove that ϱ ( x ) has only one root in ( 0 , 1.27 ) . Denote n 1 ( x ) = ϱ ( x ) / x . Because of ϱ ( x ) = 0 in ( 0 , 1.27 ) if and only if n 1 ( x ) = 0 in ( 0 , 1.27 ) .
It suffices to show that
  • ϱ ( x ) > 0 for 0 < x 0.8 ,
  • n 1 ( x ) < 0 for 0.8 < x < 1.27 ,
  • ϱ ( 1.27 ) < 0 .
Let 0 < x 0.6 . First we show ϱ ( x ) > 0 for 0 < x 0.8 . It is obvious that
n 1 ( x ) n s ( x ) = 0 . 6 6 190374 x 6 4286906 x 5 45015548 x 4 282667076 x 3 1146875462 x 2 3046852634 x 5055591704 4276245096 x 5 + 386216448 x 4 + 4183878144 x 3 + 3096950400 x 2 + 875102976 x + 292571136 .
Differentiation yields
n s ( x ) = 832695876 x 5 / 15625 66819454779474 x 4 / 3125 + 24007262662032 x 3 / 15625 + 195501095104788 x 2 / 15625 + 95107555576404 x / 15625 + 11452328429814 / 15625 ,
n s ( x ) = 832695876 x 4 / 3125 267277819117896 x 3 / 3125 + 72021787986096 x 2 / 15625 + 391002190209576 x / 15625 + 95107555576404 / 15625 .
Now recall from [12,13] that if
f ( x ) = a x 4 + b x 3 + c x 2 + d x + e
and
δ = 256 e 3 a 3 192 a 2 b d e 2 128 a 2 c 2 e 2 + 144 a 2 c d 2 e 27 a 2 d 4 + 144 a b 2 c e 2 6 b 2 a e d 2 80 a b c 2 d e + 18 a b c d 3 + 16 c 4 a e 4 a c 3 d 2 27 b 4 e 2 + 18 c d e b 3 4 d 3 b 3 4 b 2 c 3 e + b 2 c 2 d 2 < 0
then the polynomial f ( x ) has two real distinct roots and two complex but not real roots.
Recall [14] the Bolzano Theorem which states: Let a < b be two real numbers, let f ( x ) be continuous function on a closed interval [ a , b ] such that f ( a ) f ( b ) < 0 . Then there is a number x 0 ( a , b ) such that f ( x 0 ) = 0 .
Consider the equation n s ( x ) = 0 . Direct computation gives
δ 2.21 10 64 < 0
and
This implies that there are only two real roots of n s ( x ) = 0 . Table 1 implies the first root of n s ( x ) = 0 is in (−330,000, −320,000) and the second root of n s ( x ) = 0 is in ( 0.6 , 0.7 ) .
Because of n s ( 0 ) > 0 , n s ( 0 ) > 0 , n s ( 0 ) > 0 we obtain ϱ ( x ) > 0 for 0 < x 0.6 .
Similarly, for 0.6 < x 0.8 we deduce
n 1 ( x ) n r ( x ) = 0.8 . 6 190374 × 0 . 8 6 4286906 × 0 . 8 5 45015548 × 0 . 8 4 282667076 × 0 . 8 3 1146875462 × 0 . 8 2 3046852634 × 0.8 5055591704 4276245096 x 5 + 386216448 x 4 + 4183878144 x 3 + 3096950400 x 2 + 875102976 x + 292571136 .
Differentiation yields
n r ( x ) = 21381225480 x 4 + 1544865792 x 3 + 12551634432 x 2 + 6193900800 x + 875102976 .
Consider the equation n r ( x ) = 0 . Straightforwardly computing acquires
δ 3.27 10 59 < 0
and
This implies that there are only two real roots of n r ( x ) = 0 . Table 2 implies that the first root of n r ( x ) = 0 is in ( 0.2 , 0.3 ) and the second root of n r ( x ) = 0 is in ( 0.9 , 1 ) .
It brings n r ( x ) > 0 on 0.6 < x 0.8 (evidently n r ( 0 ) > 0 ). Because of n r ( 0.6 ) 3.53 10 08 > 0 we obtain n r ( x ) > 0 on 0.6 < x 0.8 . So ρ ( x ) > 0 for 0.6 < x 0.8 . Now we prove n 1 ( x ) < 0 for 0.8 < x 1.27 .
n 1 ( x ) = 2284488 x 11 47155966 x 10 450155480 x 9 2544003684 x 8 9175003696 x 7 21327968438 x 6 30333550224 x 5 21381225480 x 4 + 1544865792 x 3 + 12551634432 x 2 + 6193900800 x + 875102976 .
We first show n 1 ( x ) < 0 for 0.85 x 1.27 . It is obvious that
n 1 ( x ) < v ( x ) = 2284488 x 11 47155966 x 10 450155480 x 9 2544003684 x 8 9175003696 x 7 21327968438 x 6 30333550224 x 5 21381225480 x 4 + 1544865792 × 1 . 27 3 + 12551634432 × 1 . 27 2 + 6193900800 × 1.27 + 875102976 .
Because of v ( x ) < 0 , v ( 0.85 ) 4.26 10 09 we obtain n 1 ( x ) < 0 for 0.85 x 1.27 .
Next, we show n 1 ( x ) < 0 for 0.8 x 0.85 . Easy to see that
n 1 ( x ) < v ( x ) = 2284488 x 11 47155966 x 10 450155480 x 9 2544003684 x 8 9175003696 x 7 21327968438 x 6 30333550224 x 5 21381225480 x 4 + 1544865792 × 0 . 85 3 + 12551634432 × 0 . 85 2 + 6193900800 × 0.85 + 875102976 .
Because of v ( x ) < 0 , v ( 0.8 ) 1.05 10 10 we get n 1 ( x ) < 0 for 0.8 x 0.85 .
From n 1 ( 0.8 ) 1.67 10 09 and n 1 ( 1.27 ) 4.63 10 10 we can conclude that ϱ ( x ) has only one root in ( 0 , 1.27 ) which completes the proof of Lemma 2. □

3. Proof of the Main Result

In this section, we prove Qi’s Conjecture.
Theorem 1.
Let Δ ( x ) = ψ ( x ) 2 + ψ ( x ) for x > 0 . Then
1 x 4 x 2 + 4 x + 12 12 ( 1 + x ) 2 α < Δ ( x ) < 1 x 4 x 2 + 4 x + 12 12 ( 1 + x ) 2 β
holds on ( 0 , ) if and only if α 6 / 5 and β 1 .
Proof of Theorem 1. 
The upper bound of (9) follows from the following conclusion. Le x be a fixed positive real number. Denote
F ( β , x ) = 1 x 4 x 2 + 4 x + 12 12 ( 1 + x ) 2 β .
Then F β ( β , x ) < 0 . So, F ( β , x ) is a decreasing function in β for each fixed x > 0 . In the paper [5], it was obtained that
1 x 4 x 2 + 12 12 ( 1 + x ) 2 < Δ ( x ) < 1 x 4 x 2 + 4 x + 12 12 ( 1 + x ) 2
for x > 0 . From F ( 1 ) Δ ( x ) and
lim x + x 4 Δ ( x ) x 2 + 4 x + 12 12 ( 1 + x ) 2 β = 1 12 1 12 β
we can derive that β = 1 is an optimal constant.
Now we show the lower bound of (9). Lemma 2 implies
Δ ( x ) 1 x 4 x 2 + 4 x + 12 12 ( 1 + x ) 2 α > 0
on ( 0 , ) for α 6 / 5 . The cases (2), (3) of Lemma 2 imply that α = 6 / 5 is the best constant.
This completes the proof. □

4. Materials and Methods

In this paper, MATLAB software and methods of mathematical analysis were used.

5. Conclusions

The main result of this paper is the Theorem 1. The Theorem 1 says that the double inequality
1 x 4 x 2 + 4 x + 12 12 ( 1 + x ) 2 α < ψ ( x ) 2 + ψ ( x ) < 1 x 4 x 2 + 4 x + 12 12 ( 1 + x ) 2 β
holds on ( 0 , ) if and only if α 6 / 5 and β 1 where ψ ( x ) and ψ ( x ) are the tri- and tetra-gamma functions respectively. The double inequality was posed by F. Qi and R. P. Agarwal as the seventh open problem in Remark 6 of the paper [5].

Funding

The work was supported by VEGA grants No. 1/0589/17, No. 1/0649/17, No. 1/0185/19, No. 1/0348/20, No. 1/0026/20 and by Kega grant No. 007 TnUAD-4/2017.

Acknowledgments

The author thanks to anonymous reviewers whose comments and suggestions have greatly improved this work. The author also thanks to Ondrušová, dean of FPT TnUAD, Slovakia, for his kind grant support.

Conflicts of Interest

The author declares that he has no competing interests.

Appendix A

Deriving the polynomials n 1 ( x ) , n 2 ( x ) , n 3 ( x ) . It is obvious that
n 1 ( x ) = 1 . 27 7 ( 385 x 12 + 1435 x 11 + 2100 x 10 + 18662 x 9 + 153421 x 8 + 477255 x 7 + 718276 x 6 + 568208 x 5 + 232044 x 4 + 38256 x 3 ) 46764 x 9 314648 x 8 1233916 x 7 2806448 x 6 3866740 x 5 3395384 x 4 1933504 x 3 697344 x 2 145600 x 13440 = 1154932396539119275 562949953421312 x 12 + 4304748023463990025 562949953421312 x 11 + 1574907813462435375 140737488355328 x 10 + 14828465727041900881 281474976710656 x 9 + 283104983340120160239 562949953421312 x 8 + 737048242443739991533 562949953421312 x 7 + 143703026148044650491 140737488355328 x 6 29516054779357928685 35184372088832 x 5 303835002510877789687 140737488355328 x 4 60856544015076107363 35184372088832 x 3 697344 x 2 145600 x 13440 .
By virtue of x 1.27 one may deduce that n 1 ( x ) n 2 ( x ) where
n 2 ( x ) = 1.27 3 1154932396539119275 562949953421312 x 9 + 4304748023463990025 562949953421312 x 8 + 1574907813462435375 140737488355328 x 7 + 14828465727041900881 281474976710656 x 6 + 283104983340120160239 562949953421312 x 5 + 737048242443739991533 562949953421312 x 4 + 143703026148044650491 140737488355328 x 3 29516054779357928685 35184372088832 x 5 303835002510877789687 140737488355328 x 4 60856544015076107363 35184372088832 x 3 697344 x 2 145600 x 13440 = 10404651649353497429418550644025 2475880078570760549798248448 x 9 + 9695243582352122447439497494795 618970019642690137449562112 x 8 + 7094080670013747513516302614285 309485009821345068724781056 x 7 + 8349245205097078426781024008335 77371252455336267181195264 x 6 + 924711257333410630048926774241 4835703278458516698824704 x 5 + 1264511729803961840461767951097 2417851639229258349412352 x 4 + 3500065726786252849629593023555 9671406556917033397649408 x 3 697344 x 2 145600 x 13440 .
Similarly, we have n 2 ( x ) n 3 ( x ) where
n 3 ( x ) = 1.27 3 10404651649353497429418550644025 2475880078570760549798248448 x 6 + 9695243582352122447439497494795 618970019642690137449562112 x 5 + 7094080670013747513516302614285 309485009821345068724781056 x 4 + 8349245205097078426781024008335 77371252455336267181195264 x 3 + 924711257333410630048926774241 4835703278458516698824704 x 2 + 1264511729803961840461767951097 2417851639229258349412352 x + 3500065726786252849629593023555 9671406556917033397649408 697344 x 2 145600 x 13440 = 10656355779728833048985112408175 1237940039285380274899124224 x 6 + 19859572134949190226816117461335 618970019642690137449562112 x 5 + 227053035079449571081038231535 4835703278458516698824704 x 4 + 17102451940952370969088397817565 77371252455336267181195264 x 3 2955971695165983842306493551217 9671406556917033397649408 x 2 + 17905321055673993813657464422955 19342813113834066795298816 x + 14078982859013282843744752202495 19342813113834066795298816 .

References

  1. Yang, Z.-H.; Qian, W.-M.; Chu, Y.-M.; Zhang, W. On rational bounds for the gamma function. J. Inequal. Appl. 2017, 2017. [Google Scholar] [CrossRef]
  2. Zhao, T.-H.; Chu, Y.M.; Jiang, Y.P. Monotonic and logarithmically convex properties of a function involving gamma functions. J. Inequal. Appl. 2009, 2009. [Google Scholar] [CrossRef]
  3. Yang, Z.-H.; Chu, Y.-M.; Zhang, X.-H. Sharp bounds for psi function. Appl. Math. Comput. 2015, 268, 1055–1063. [Google Scholar] [CrossRef]
  4. Andrews, G.-E.; Askey, R.; Roy, R. Special Functions; Cambridge University Press: Cambridge, UK, 1999. [Google Scholar]
  5. Qi, F.; Agarwal, R.P. On complete monotonicity for several classes of functions related to ratios of gamma functions. J. Inequal. Appl. 2019, 2019. [Google Scholar] [CrossRef]
  6. Qi, F. Complete monotonicity of a function involving the tri- and tetra-gamma functions. Proc. Jangjeon Math. Soc. 2015, 2, 235–264. [Google Scholar] [CrossRef]
  7. Cartier, P. An Introduction to Zeta Functions; Waldschmidt, M., Moussa, P., Luck, J.M., Itzykson, C., Eds.; From Number Theory to Physics; Springer: Berlin/Heidelberg, Germany, 1992. [Google Scholar] [CrossRef]
  8. Qi, F.; Mortici, C. Some inequalities for the trigamma function in terms of the digamma function. Appl. Math. Comput. 2015, 271, 502–511. [Google Scholar] [CrossRef]
  9. Guo, B.-N.; Chen, R.J.; Qi, F. A class of completely monotonic functions involving the polygamma functions. J. Math. Anal. Approx. Theory 2006, 1, 123–134. [Google Scholar] [CrossRef]
  10. English, B.J.; Rousseau, G. Bound for certain harmonic sums. J. Math. Anal. Appl. 1997, 206, 428–441. [Google Scholar] [CrossRef]
  11. Qi, F.; Guo, B.-N. Sharp inequalities for polygamma functions. Math. Slovaca 2015, 65, 103–120. [Google Scholar] [CrossRef]
  12. Janson, S. Roots of polynomials of degree 3 and 4. arXiv 2010, arXiv:1009.2373v1. [Google Scholar]
  13. Rees, E.L. Graphical discussion of the roots of a quartic equation. Am. Math. Mon. 1922, 29, 51–55. Available online: https://www.jstor.org/stable/pdf/2972804.pdf (accessed on 10 September 2019). [CrossRef]
  14. Jarník, V. Diferenciální Počet I; Academi: San Francisco, CA, USA, 1974; Available online: http://matematika.cuni.cz/jarnik-all.html (accessed on 10 September 2019).
Table 1. Values of n s ( x ) .
Table 1. Values of n s ( x ) .
PointsValues of n s ( x )
−330,000 8.63 10 25
−320,000 8.54 10 24
0.6 4.28 10 09
0.7 3.47 10 09
Table 2. Values of n r ( x ) .
Table 2. Values of n r ( x ) .
PointsValues of n r ( x )
0.2 9.18 10 07
0.3 6.83 10 07
1−215,721,480
0.9 3.71 10 09

Share and Cite

MDPI and ACS Style

Matejíčka, L. A Solution to Qi’s Conjecture on a Double Inequality for a Function Involving the Tri- and Tetra-Gamma Functions. Mathematics 2019, 7, 1098. https://doi.org/10.3390/math7111098

AMA Style

Matejíčka L. A Solution to Qi’s Conjecture on a Double Inequality for a Function Involving the Tri- and Tetra-Gamma Functions. Mathematics. 2019; 7(11):1098. https://doi.org/10.3390/math7111098

Chicago/Turabian Style

Matejíčka, Ladislav. 2019. "A Solution to Qi’s Conjecture on a Double Inequality for a Function Involving the Tri- and Tetra-Gamma Functions" Mathematics 7, no. 11: 1098. https://doi.org/10.3390/math7111098

Note that from the first issue of 2016, this journal uses article numbers instead of page numbers. See further details here.

Article Metrics

Back to TopTop