Hindawi Publishing Corporation

Journal of Applied Mathematics

Volume 2014, Article ID 264652, 7 pages
http://dx.doi.org/10.1155/2014/264652

Research Article
New Inequalities for Gamma and Digamma Functions

M. R. Farhangdoost and M. Kargar Dolatabadi

Department of Mathematics, College of Science, Shiraz University, Shiraz 71475-44776, Iran

Correspondence should be addressed to M. R. Farhangdoost; [email protected]

Received 6 December 2013; Accepted 1 July 2014; Published 12 November 2014

Academic Editor: Vijay Gupta

Copyright © 2014 M. R. Farhangdoost and M. Kargar Dolatabadi. This is an open access article distributed under the Creative
Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the
original work is properly cited.

By using the mean value theorem and logarithmic convexity, we obtain some new inequalities for gamma and digamma functions.

1. Introduction In Section 2, by applying the mean value theorem on

Let Γ(𝑥), 𝜓(𝑥), 𝜓𝑛 (𝑥), and 𝜁(𝑥) denote the Euler gamma 󸀠
function, digamma function, polygamma functions, and (log Γ (𝑥)) = 𝜓 (𝑥) , for 𝑥 > 0, (5)
Riemann zeta function, respectively, which are defined by
∞ we obtain some new inequalities on gamma and digamma
Γ (𝑥) = ∫ 𝑒−𝑡 𝑡𝑥−1 𝑑𝑡, for 𝑥 > 0, functions.
0
(1) Section 3 is devoted to some new inequalities on
Γ󸀠 (𝑥)
𝜓 (𝑥) = , for 𝑥 > 0, digamma function, by using convex properties of logarithm
Γ (𝑥) of this function.
𝜓(𝑛) (𝑥) Note that in this paper by 𝛾 = lim𝑛 → ∞ (∑𝑛𝑘=1 (1/𝑘) −
(2) log(𝑛)) = 0.5772156 ⋅ ⋅ ⋅ we mean Euler’s constant [5].
∞
𝑡𝑛 𝑒−𝑥𝑡
= (−1)𝑛+1 ∫ 𝑑𝑡, for 𝑥 > 0; 𝑛 = 1, 2, 3, . . . ,
0 1 − 𝑒−𝑡
∞
2. Inequalities for Gamma and Digamma
1 Functions by the Mean Value Theorem
𝜁 (𝑥) = ∑ 𝑥 , for 𝑥 > 1. (3)
𝑛=1 𝑛
Lemma 1. For 𝑡 > 0, one has
In the past different papers appeared providing inequalities
for the gamma, digamma, and polygamma functions (see [1–
18]). −𝜓󸀠󸀠 (𝑡)
< 1. (6)
By using the mean value theorem to the function log Γ(𝑥) 𝜓󸀠 (𝑡)2
on [𝑢, 𝑢 + 1], with 𝑥 > 0 and 𝑢 > 0, Batir [19] presented the
following inequalities for the gamma and digamma functions: Proof. By [6, Proposition 1], we have
−𝛾
𝜓 (𝑥) ⩽ log (𝑥 − 1 + 𝑒 ) , for 𝑥 > 0,
2
1 𝜓󸀠 (𝑡) 𝜓󸀠󸀠󸀠 (𝑡) − 2 [𝜓󸀠󸀠 (𝑡)] < 0, for 𝑡 > 0. (7)
log (𝑥) − 𝜓 (𝑥) < 𝜓󸀠 (𝑥) , for 𝑥 > 1,
2 (4)
2
𝜋 −𝜓(𝑥) Thus the function 𝜓󸀠󸀠 (𝑡)/𝜓󸀠 (𝑡)2 is strictly decreasing on
𝜓󸀠 (𝑥) ⩾ 𝑒 , for 𝑥 ⩾ 1. (0, ∞).
6𝑒𝛾
2 Journal of Applied Mathematics

By using asymptotic expansions [20, pages 253–256 and We show that the function 𝜃(𝑢) has the following properties:
364],
(1) 𝜃(𝑢) is strictly increasing on (0, ∞);
󸀠 1 1 1 𝜃󸀠 󸀠 (2) lim𝑢 → ∞ 𝜃(𝑢) = 1/2;
𝜓 (𝑡) = + 2 + 3 + , (0 ⩽ 𝜃 ⩽ 1) , (8)
𝑡 2𝑡 6𝑡 30𝑡5
(3) 𝜃󸀠 (𝑢) is strictly decreasing on (0, ∞);
󸀠󸀠
1 1 1 1 𝜃
𝜓󸀠󸀠 (𝑡) = − − − + − , (0 ⩽ 𝜃󸀠󸀠 ⩽ 1) . (9) (4) lim𝑢 → ∞ 𝜃󸀠 (𝑢) = 0.
𝑡2 𝑡3 2𝑡4 6𝑡6 6𝑡8
To prove these four properties, since 𝜓󸀠 is a decreasing
For 𝑡 > 0, we get
function on (0, ∞), we put 𝑢 = 1/𝜓󸀠 (𝑡), where 𝑡 > 0; by
formula (13) we have
𝜓󸀠󸀠 (𝑡)
lim = −1. (10)
𝑡 → ∞ 𝜓󸀠 (𝑡)2 1 1
𝜓󸀠 ( + 𝜃 ( 󸀠 )) = 𝜓󸀠 (𝑡) . (∗)
󸀠󸀠 󸀠 2
𝜓󸀠 (𝑡) 𝜓 (𝑡)
Now, the proof follows from the monotonicity of 𝜓 (𝑡)/𝜓 (𝑡)
on (0, ∞) and Since by formula (8) we have 𝜓󸀠󸀠 (𝑡) < 0 and 𝜓󸀠 (𝑡) > 0, for
all 𝑡 > 0, then the mapping 𝑡 → 𝜓󸀠 (𝑡) from (0, ∞) into
𝜓󸀠󸀠 (𝑡) (0, ∞) is injective since also 𝜓󸀠 (𝑡) → 0 and 𝜓󸀠 (𝑡) → ∞
lim = −1. (11)
𝑡 → ∞ 𝜓󸀠 (𝑡)2 when 𝑡 → ∞ and 𝑡 → 0+ , respectively, then the mapping
𝑡 → 𝜓󸀠 (𝑡) from (0,∞) into (0,∞) is a bijective map. Clearly,
by injectivity of 𝜓󸀠 , we find that
Theorem 2. One has the following: 1 1
𝜃( )=𝑡− 󸀠 , for 𝑡 > 0. (14)
󸀠
(a) 𝑥 − (1/2) < 1/𝜓 (𝑥) ⩽ 𝑥 + (6/𝜋 ) − 1 for 𝑥 ⩾ 1;2 𝜓󸀠 (𝑡) 𝜓 (𝑡)

(b) 1/𝑥2 < 𝜓󸀠 (𝑥)𝜓󸀠 (𝑥 + 1) < 2/𝑥2 for 𝑥 > 0; Differentiating between both sides of this equation, we get
󸀠 2 󸀠󸀠 4 2 󸀠
(c) [𝜓 (𝑥)] /𝜓 (𝑥) ⩾ −𝜋 /72𝜁(3) for 𝑥 ⩾ 1 and 𝑥 𝜓 (𝑥 + 2
1)𝜓󸀠 (𝑥) < 𝜋4 /72𝜁(3) for 𝑥 > 2; 1 − [(𝜓󸀠 (𝑡)) + 𝜓󸀠󸀠 (𝑡)]
󸀠 (15)
𝜃 ( 󸀠 )= .
(d) ([𝜓󸀠 (𝑥 + ℎ)]2 − 𝜓󸀠 (𝑥)𝜓󸀠 (𝑥 + ℎ))/ℎ𝜓󸀠 (𝑥) > 𝜓󸀠󸀠 (𝑥 + ℎ) 𝜓 (𝑡) 𝜓󸀠󸀠 (𝑡)
for 𝑥 > 0 and ℎ > 0;
Since by formula (8), 𝜓󸀠󸀠 (𝑡) < 0, where 𝑡 > 0, hence formula
(e) (𝜓󸀠 (𝑥 + ℎ)𝜓󸀠 (𝑥) − [𝜓󸀠 (𝑥)]2 )/ℎ𝜓󸀠 (𝑥 + ℎ) < 𝜓󸀠󸀠 (𝑥) for (15) gives 𝜃󸀠 (1/𝜓󸀠 (𝑡)) > 0, for all 𝑡 > 0. Since the mapping 𝑡 →
𝑥 > 0 and ℎ > 0; 1/𝜓󸀠 (𝑡) from (0, ∞) to (0, ∞) is also bijective, then 𝜃󸀠 (𝑡) > 0
(f) −𝑥2 𝜓󸀠󸀠 (𝑥) < 𝜓󸀠 (𝑥)/𝜓󸀠 (𝑥 + 1) and 𝜓󸀠 (𝑥 + 1)/𝜓󸀠 (𝑥) < for all 𝑡 > 0, and the proof of (1) is completed.
−𝑥2 𝜓󸀠󸀠 (𝑥 + 1) for 𝑥 > 0; From (8) we have
2
(g) ((𝜋2 𝑥/6) + 1)(𝑥+(6/𝜋 )) 𝑒−𝑥(𝛾+1) ⩽ Γ(𝑥 + 1) < (2𝑥 + lim 𝜃 (𝑢)
𝑢→∞
1)(𝑥+(1/2)) 𝑒−𝑥(1+𝛾) for 𝑥 ⩾ 1;
(h) (1/𝑥) − 𝜓󸀠 (𝑥) < (1/2)𝜓󸀠󸀠 (𝑥 + (1/2)) for 𝑥 > 0 and 1 1
= lim 𝜃 ( ) = lim (𝑡 − 󸀠 )
(1/𝑥) − 𝜓󸀠 (𝑥) > ((𝜓󸀠 )−1 (1) − 1)𝜓󸀠󸀠 (𝑥) for 𝑥 > 1; 𝑡→∞ 𝜓󸀠 (𝑡) 𝑡→∞ 𝜓 (𝑡)
−1 (16)
(i) 𝜓(𝑥 + 1) > log(𝑥 + (1/2)) + 𝜓((𝜓󸀠 ) (1)) for 𝑥 ⩾ 1/2; 1
= lim (𝑡 − )
−1 −1 𝑡→∞ (1/𝑡) + (1/2𝑡 ) + (1/6𝑡3 ) + (1/3𝑡5 )
2
(j) (𝜋4 /72𝜁(3)) log(𝑥 − (𝜓󸀠 ) (1) + 2) + 𝜓((𝜓󸀠 ) (1)) ⩾
−1
𝜓(𝑥 + 1) for 𝑥 > (𝜓󸀠 ) (1) − 1. 1
= .
2
Proof. Let 𝑢 be a positive real number and 𝜓(𝑥) defined on the
closed interval [𝑢, 𝑢+1]. By using the mean value theorem for Differentiating between both sides of (15), we obtain
the function 𝜓(𝑥) on [𝑢, 𝑢 + 1] with 𝑢 > 0 and since 𝜓󸀠 is a
decreasing function, there is a unique 𝜃 depending on 𝑢 such 1
𝜃󸀠󸀠 ( )
that 0 ⩽ 𝜃 = 𝜃(𝑢) < 1, for all 𝑢 ⩾ 0; then 𝜓󸀠 (𝑡)
3 (∗∗)
𝜓 (𝑢 + 1) − 𝜓 (𝑢) = 𝜓󸀠 (𝑢 + 𝜃 (𝑢)) , (12) [𝜓󸀠 (𝑡)] 󸀠󸀠 2 󸀠 󸀠󸀠
= [2 (𝜓 (𝑡)) − 𝜓 (𝑡) 𝜓 (𝑡)] .
󸀠 󸀠 2 𝜓󸀠󸀠 (𝑡)
Since 𝜓(𝑥 + 1) − 𝜓(𝑥) = 1/𝑥 and 𝜓 (𝑥 + 1) − 𝜓 (𝑥) = −1/𝑥 ,
we have Since 𝜓󸀠 (𝑡) > 0 and 𝜓󸀠󸀠 (𝑡) < 0, where 𝑡 > 0, then
1 𝜃󸀠󸀠 (1/𝜓󸀠 (𝑡)) < 0 for all 𝑡 > 0. Proceeding as above we conclude
𝜓󸀠 (𝑢 + 𝜃 (𝑢)) = , for 𝑢 > 0. (13) that 𝜃󸀠󸀠 (𝑡) < 0, for 𝑡 > 0. This proves (3).
𝑢
Journal of Applied Mathematics 3

For (4), from (8), (9), we conclude that By using this inequality and the fact that 𝜓(𝑥+1)−𝜓(𝑥) = 1/𝑥
2
and
1 [(𝜓󸀠 (𝑡)) + 𝜓󸀠󸀠 (𝑡)] 1
󸀠 󸀠
lim 𝜃 (𝑢) = lim 𝜃 ( 󸀠 ) = lim − 𝜓󸀠 (𝑥 + 1) − 𝜓󸀠 (𝑥) = − , (26)
𝑢→∞ 𝑡→∞ 𝜓 (𝑡) 𝑡→∞ 𝜓󸀠󸀠 (𝑡) 𝑥2
2 we obtain
[𝜓󸀠 (𝑡)]
= −1 − lim = 0. 1
𝑡→∞ 𝜓󸀠󸀠 (𝑡) 𝜓󸀠 (𝑡 + 1) 𝜓󸀠 (𝑡) > , 𝑡 > 0. (27)
𝑡2
(17)
Since 𝜃 is strictly increasing on (0, ∞), by (1), it is clear that
Now, we prove the theorem. To prove (a), let 1/𝜓󸀠 (1) =
1 1
6/𝜋2 ⩽ 𝑡 < ∞; then by (1) and (2) we have 𝜃( ) − 𝜃( 󸀠 )
𝜓󸀠 (𝑡 + 1) 𝜓 (𝑡)
1 (28)
𝜃( ) ⩽ 𝜃 (𝑡) < lim 𝜃 (𝑡) . (18) +1
𝜓󸀠 (1) 𝑡→∞ < lim 𝜃 (𝑡) − 𝜃 (0 ) = , 𝑡 > 0.
𝑡→∞ 2
Equation (13) and 𝜓󸀠󸀠 (𝑡) < 0 for all 𝑡 > 0 give and then it is clear that (b) holds.
For (c), since 𝑡 > 2, 𝑡 + 𝛿(𝑡) > 1 + 𝛿(1), and 𝜃󸀠 is strictly
−1 1
𝜃 (𝑡) = (𝜓󸀠 ) ( ) − 𝑡. (19) decreasing on (0, ∞) by (3), then
𝑡
2
By substituting the value of 𝜃(𝑡) into (18), we get 1 1 [𝜓󸀠 (1)]
𝜃󸀠 ( ) < 𝜃󸀠
( ) = −1 − ,
𝜓󸀠 (𝑡 + 𝛿 (𝑡)) 𝜓󸀠 (1) 𝜓󸀠󸀠 (1) (29)
1 −1 1 1
1− ⩽ (𝜓󸀠 ) ( ) − 𝑡 < lim 𝜃 (𝑡) = . (20)
𝜓󸀠 (1) 𝑡 𝑡→∞ 2 ∀𝑡 > 2.

By substituting the value 𝑡 = 1/𝜓󸀠 (𝑢) into this inequality, we Since 𝜓(𝑥 + 1) − 𝜓(𝑥) = 1/𝑥 and 𝜓󸀠 (𝑥 + 1) − 𝜓󸀠 (𝑥) = −1/𝑥2 ,
get by using (24), we obtain
1 1 6 𝜋4
𝑢− < 󸀠 ⩽ 𝑢 + 2 − 1, (21) 𝑡2 𝜓󸀠 (𝑡 + 1) 𝜓󸀠 (𝑡) < , (30)
2 𝜓 (𝑢) 𝜋 72𝜁 (3)
where 𝑢 ⩾ 1. where 𝑡 > 2.
In order to prove (b), by using the mean value theorem on Since 𝜃󸀠 is strictly decreasing on (0, ∞) by (3) and 𝜓󸀠󸀠 (𝑡) <
the interval [1/𝜓󸀠 (𝑡), 1/𝜓󸀠 (𝑡 + 1)], and since 𝜃 is a decreasing 0, for all 𝑡 > 0, we have
function, there exists a unique 𝛿 such that 1 1
𝜃󸀠 ( ) ⩽ 𝜃󸀠 ( 󸀠 ), (31)
0 < 𝛿 (𝑡) < 1, (22) 𝜓󸀠 (𝑡) 𝜓 (1)

for 𝑡 > 0 and where 𝑡 ⩾ 1.

Then it is clear that (c) is true.
1 1 Now we prove (d) and (e) by using the mean value
𝜃( ) − 𝜃( 󸀠 )
𝜓󸀠 (𝑡 + 1) 𝜓 (𝑡) theorem on [1/𝜓󸀠 (𝑡), 1/𝜓󸀠 (𝑡 + ℎ)] (𝑡 > 0, ℎ > 0), for 𝜃, we
(23) conclude
1 1 1
=( 󸀠 − ) 𝜃󸀠 ( 󸀠 ). 1 1
𝜓 (𝑡 + 1) 𝜓󸀠 (𝑡) 𝜓 (𝑡 + 𝛿 (𝑡)) 𝜃( ) − 𝜃( 󸀠 )
𝜓󸀠 (𝑡 + ℎ) 𝜓 (𝑡)
Now, by (14), we have (32)
1 1 1
=( 󸀠 − ) 𝜃󸀠 ( 󸀠 ),
1 1 𝜓 (𝑡 + ℎ) 𝜓󸀠 (𝑡) 𝜓 (𝑡 + 𝑎)
1− +
𝜓󸀠 (𝑡 + 1) 𝜓󸀠 (𝑡)
(24) where 0 < 𝑎 < ℎ.
1 1 1 After brief computation we have
=( 󸀠 − ) 𝜃󸀠 ( 󸀠 ).
𝜓 (𝑡 + 1) 𝜓󸀠 (𝑡) 𝜓 (𝑡 + 𝛿 (𝑡)) ℎ𝜓󸀠 (𝑡 + ℎ) 𝜓󸀠 (𝑡)
1
𝜃󸀠 ( ) = − 1, 𝑡 > 0. (33)
Since 𝜃 is strictly increasing on (0, ∞), by (1), we have 𝜓󸀠 (𝑡 + 𝑎) 𝜓󸀠 (𝑡) − 𝜓󸀠 (𝑡 + ℎ)

𝜓󸀠 (𝑡 + 1) − 𝜓󸀠 (𝑡) Since 𝑡 + 𝑎 > 𝑡 for all 𝑎 > 0, 𝑡 > 0, and by the monotonicity of
1+ 𝜃󸀠 and 𝜓󸀠 we have 𝜃󸀠 (1/𝜓󸀠 (𝑡 + 𝑎)) < 𝜃󸀠 (1/𝜓󸀠 (𝑡)); then
𝜓󸀠 (𝑡 + 1) 𝜓󸀠 (𝑡)
(25) 2
1 1 𝜓󸀠 (𝑡 + ℎ) 𝜓󸀠 (𝑡) − [𝜓󸀠 (𝑡)]
= 𝜃( 󸀠 ) − 𝜃 ( 󸀠 ) > 0. < 𝜓󸀠󸀠 (𝑡) , 𝑡 > 0, ℎ > 0. (34)
𝜓 (𝑡 + 1) 𝜓 (𝑡) ℎ𝜓󸀠 (𝑡 + ℎ)
4 Journal of Applied Mathematics

By monotonicity of 𝜃󸀠 and 𝜓󸀠 , we have or

1 1 −1
𝜃󸀠 ( ) > 𝜃󸀠 ( 󸀠 ). (35) log (𝑥) + 𝜓 ((𝜓󸀠 ) (1)) < 𝜓 (𝑥 + 𝜃 (𝑥)) . (45)
󸀠
𝜓 (𝑡 + 𝑎) 𝜓 (𝑡 + ℎ)
After some simplification of this inequality (d) is proved. Again using the monotonicity of 𝜃 and 𝜓, after some simpli-
For (f), we put ℎ = 1 in (e) and (d). fications as for 𝑥 ⩾ 1/2, we can rewrite
For (g), we integrate (a) on [1, 𝑡] for 𝑡 > 0; then we have 1 −1
log (𝑥 + ) + 𝜓 ((𝜓󸀠 ) (1)) < 𝜓 (𝑥 + 1) . (46)
(𝑡 − 1) 𝜋2 2
log ( + 1) − 𝛾
6 (36) This proves (i). By inequality (c) for 𝑥 ⩾ 1, we have
⩽ 𝜓 (𝑡) < log (2𝑡 − 1) − 𝛾, for 𝑡 ⩾ 1;
72𝜁 (3) 𝑥+𝜃(𝑥) 󸀠
log (𝑥) ⩾ ∫ 𝜓 (𝑡) 𝑑𝑡
the proof is completed when we integrate these inequalities 𝜋4 (𝜓󸀠 )−1 (1)
on [1, 𝑠], for 𝑠 > 0. (47)
By using the mean value theorem for the 𝜓󸀠 (𝑡) on [𝑡, 𝑡 + 72𝜁 (3) −1
= 4
(𝜓 (𝑥 + 𝜃 (𝑥)) − 𝜓 ((𝜓󸀠 ) (1))) ;
𝜃(𝑡)], there is a 𝛼(𝑡) depending on 𝑡 such that 0 < 𝛼(𝑡) < 𝜃(𝑡) 𝜋
for all 𝑡 > 0, and so −1
since for 𝑥 ⩾ 1, 𝜃(𝑥) ⩾ 𝜃(1) = ((𝜓󸀠 ) (1) − 1)) = (𝜓󸀠 )−1 (1) − 1,
𝜓󸀠 (𝑡 + 𝜃 (𝑡)) = 𝜃 (𝑡) 𝜓󸀠󸀠 (𝑡 + 𝛼 (𝑡)) + 𝜓󸀠 (𝑡) . (37) from this inequality we find that
By formula (13) and (2), since 𝜓󸀠󸀠 is strictly increasing on 𝜋4 −1
(0, ∞), we have log (𝑥) + 𝜓 ((𝜓󸀠 ) (1))
72𝜁 (3)
(48)
𝜓󸀠󸀠 (𝑡 + 𝛼 (𝑡)) 𝜃 (𝑡)
󸀠 −1
⩾ 𝜓 (𝑥 + (𝜓 ) (1) − 1) ;
1
= − 𝜓󸀠 (𝑡) < lim 𝜃 (𝑡) 𝜓󸀠󸀠 (𝑡 + lim 𝜃 (𝑡)) , for 𝑡 > 0,
𝑡 𝑡→∞ 𝑡→∞
replacing 𝑥 by 𝑥 − (𝜓󸀠 )−1 (1) + 2, we get for 𝑥 ⩾ (𝜓󸀠 )−1 (1) − 1
(38)
or 𝜋4 −1 −1
log (𝑥 − (𝜓󸀠 ) (1) + 2) + 𝜓 ((𝜓󸀠 ) (1))
1 1 1 72𝜁 (3)
− 𝜓󸀠 (𝑡) < 𝜓󸀠󸀠 (𝑡 + ) , for 𝑡 > 0; (39) (49)
𝑡 2 2
⩾ 𝜓 (𝑥 + 1) ,
since 𝜓󸀠󸀠 is strictly increasing on (0, ∞), by (1), we have
which proves (j). Then the proof is completed.
1
𝜃 (𝑡) 𝜓󸀠󸀠 (𝑡 + 𝛼 (𝑡)) = − 𝜓󸀠 (𝑡) > 𝜃 (1) 𝜓󸀠󸀠 (𝑡) ,
𝑡 (40) Example 3. Consider the matrix
for 𝑡 > 1,
3 1 1 ⋅⋅⋅ 1
or [1 4 1 ⋅ ⋅ ⋅ 1 ]
[ ]
1 −1 𝐴𝑛 = [ .. ]. (50)
− 𝜓󸀠 (𝑡) > ((𝜓󸀠 ) (1) − 1) 𝜓󸀠󸀠 (𝑡) , for 𝑡 > 1. (41) [ . ]
𝑡 1 1 ⋅ ⋅ ⋅ 1 𝑛 + 1
[ ]
In order to prove (i) and (j), we integrate both sides of (13)
over 1 ⩽ 𝑢 ⩽ 𝑥 to obtain By using inequalities (a), we obtain
𝑥 𝑥
1
∫ 𝜓󸀠 (𝑢 + 𝜃 (𝑢)) 𝑑𝑢 = ∫ 𝑑𝑢. (42) 𝜋2 2
1 1 𝑢 ⩽ 𝜓󸀠 (𝑥) < , 𝑥 ⩾ 1. (51)
𝜋2 𝑥 + 6 − 𝜋2 2𝑥 − 1
Making the change of variable 𝑢 = 1/𝜓󸀠 (𝑡) on the left-hand
Now, we integrate on [1, 𝑡] (for 𝑡 > 0) from both sides of (51)
side, by (14), we have
to obtain
𝑥+𝜃(𝑥) −𝜓󸀠󸀠 (𝑡)
∫ 𝜓󸀠 (𝑡) 𝑑𝑡 = log (𝑥) ; (43) (𝑡 − 1) 𝜋2
(𝜓󸀠 )−1 (1) 𝜓󸀠 (𝑡)2 log ( + 1) − 𝛾 ⩽ 𝜓 (𝑡) < log (2𝑡 − 1) − 𝛾; (52)
6
since 𝜓󸀠 (𝑡) > 0 for all 𝑡 > 0 and 𝜓󸀠 (𝑥)𝜓󸀠󸀠 (𝑥) − 2[𝜓󸀠󸀠 (𝑥)]2 < 0,
we find that, for 𝑥 > 1, replacing 𝑡 by 𝑛 + 1 (𝑛 is an integer number) and using the
identity 𝜓(𝑛 + 1) = 𝐻𝑛 − 𝛾 [6] and det 𝐴 𝑛 = 𝑛!𝐻𝑛 [21], where
𝑥+𝜃(𝑥)
𝐻𝑛 = ∑𝑛𝑘=1 (1/𝑘) is the 𝑛th harmonic number, then we have
log (𝑥) < ∫ 𝜓󸀠 (𝑡) 𝑑𝑡
(𝜓󸀠 )−1 (1)
(44) 𝑛!
𝑛𝜋2
= 𝜓 (𝑥 + 𝜃 (𝑥)) − 𝜓 ((𝜓 ) (1)) 󸀠 −1 log ( + 1) ⩽ 𝑛!𝐻𝑛 < log (2𝑛 + 1)𝑛! . (53)
6
Journal of Applied Mathematics 5

3. New Inequalities for Digamma Function Let V = 3 and 𝑢 = 𝑎𝑥 + 3. Note that 𝜓(3) = (3/2) − 𝛾 and
by Properties of Strictly Logarithmically (1/𝑎)𝑢 + (1 − (1/𝑎))V = 𝑥 + 3; also we obtain
Convex Functions 𝑎 𝑎−1
[𝜓 (𝑥 + 3)] 3 𝑢−3 3
> ( − 𝛾) for 𝑥 = >− . (58)
Definition 4. A positive function 𝑓 is said to be logarithmi- 𝜓 (𝑎𝑥 + 3) 2 𝑎 𝑎
cally convex on an interval 𝐼 if 𝑓 has derivative of order two
on 𝐼 and In order to prove (b), let
󸀠󸀠
(log 𝑓 (𝑥)) ⩾ 0 (54) 𝑓 (𝑥) = log 𝜓 (𝑎𝑥 + 3) − log 𝜓 (3 + 𝑎)
(59)
for all 𝑥 ∈ 𝐼. − 𝑎 log 𝜓 (𝑥 + 3) ;
If inequality (54) is strict, for all 𝑥 ∈ 𝐼, then 𝑓 is said to be
strictly logarithmically convex [22]. since 𝜓(4) = (11/6) − 𝛾, we have 𝑓(1) = log((11/6) − 𝛾)−𝑎 .
Also
Lemma 5. The function Γ is increasing on [𝑐, ∞), where 𝑐 =
1/46163 ⋅ ⋅ ⋅ is the only positive zero of 𝜓 [1, 19]. 𝜓󸀠 (𝑎𝑥 + 3) 𝜓󸀠 (𝑥 + 3)
𝑓󸀠 (𝑥) = 𝑎 [ − ]. (60)
𝜓 (𝑎𝑥 + 3) 𝜓 (𝑥 + 3)
Lemma 6. If 𝑥 ⩾ 𝑐 and 𝑘(𝑥) = 1/𝜓(𝑥), then 𝑘 is strictly
logarithmically convex on [𝑐, ∞). By Lemma 6, log(1/𝜓(𝑡)) is strictly convex on [𝑐, ∞); then
(log 𝜓(𝑡))󸀠󸀠 < 0 and so (𝜓󸀠 (𝑡)/𝜓(𝑡))󸀠 < 0; this implies that
Proof. By differentiation we have (𝜓󸀠 (𝑡)/𝜓(𝑡)) is strictly decreasing on [𝑐, ∞). Since 𝑎 > 1 and
𝑥 ∈ (0, 1), we have 𝑎𝑥 + 3 > 𝑥 + 3. Then
󸀠 2
󸀠󸀠 −𝜓󸀠 (𝑥) −𝜓󸀠󸀠 (𝑥) 𝜓 (𝑥) + [𝜓󸀠 (𝑥)]
[log 𝑘 (𝑥)] = [ ] = 2
; 𝜓󸀠 (𝑎𝑥 + 3) 𝜓󸀠 (𝑥 + 3)
𝜓 (𝑥) [𝜓 (𝑥)] < . (61)
𝜓 (𝑎𝑥 + 3) 𝜓 (𝑥 + 3)
(55)
󸀠
And then 𝑓󸀠 (𝑥) < 0; also 𝑓(1) = log((11/6) − 𝛾)−𝑎 . Then
by Lemma 5, we obtain 𝜓(𝑥) = Γ (𝑥)/Γ(𝑥) > 0, for every
𝑥 ∈ [𝑐, ∞) and since 𝜓󸀠󸀠 (𝑥) < 0 on (0, ∞), then we have 11 −𝑎
(log 𝑘(𝑥))󸀠󸀠 > 0, for 𝑥 ⩾ 𝑐. 𝑓 (𝑥) > 𝑓 (1) = log ( − 𝛾) (62)
6
This implies that 1/𝜓(𝑥) is strictly logarithmically convex
on [𝑐, ∞). for 𝑎 > 1 and 𝑥 ∈ (0, 1) or
𝑎 𝑎
Theorem 7. One has the following: [𝜓 (𝑥 + 3)] ((11/6) − 𝛾)
< . (63)
𝑎 𝑎−1 𝜓 (𝑎𝑥 + 3) 𝜓 (3 + 𝑎)
(a) [𝜓(𝑥 + 3)] /𝜓(𝑎𝑥 + 3) > ((3/2) − 𝛾) , for 𝑎 > 1 and
𝑥 > −3/𝑎; So (b) is proved.
(b) [𝜓(𝑥 + 3)]𝑎 /𝜓(𝑎𝑥 + 3) < ((11/6) − 𝛾)𝑎 /𝜓(3 + 𝑎), for By
𝑎 > 1 and 𝑥 ∈ (0, 1);
𝑎𝑥 + 3 > 𝑥 + 3, for 𝑎 > 1, 𝑥 > 1,
(c) [𝜓(𝑥 + 3)]𝑎 /𝜓(𝑎𝑥 + 3) > ((11/6) − 𝛾)𝑎 /𝜓(3 + 𝑎), for
𝑎 > 1 and 𝑥 > 1; 𝑎𝑥 + 3 < 𝑥 + 3, for 𝑎 ∈ (0, 1) , 𝑥 ∈ (0, 1) , (64)
𝑎 𝑎
(d) [𝜓(𝑥 + 3)] /𝜓(𝑎𝑥 + 3) > ((11/6) − 𝛾) /𝜓(3 + 𝑎), for 𝑎𝑥 + 3 < 𝑥 + 3, for 𝑎 ∈ (0, 1) , 𝑥 > 1,
𝑎 ∈ (0, 1) and 𝑥 ∈ (0, 1);
(e) [𝜓(𝑥 + 3)]𝑎 /𝜓(𝑎𝑥 + 3) < ((11/6) − 𝛾)𝑎 /𝜓(3 + 𝑎), for (c), (d), and (e) are clear.
𝑎 ∈ (0, 1) and 𝑥 > 1.
Corollary 8. For all 𝑥 ∈ (0, 1) and all integers 𝑛 > 1, one has
Proof. By Lemma 6 we have, for 𝑎 > 1, 𝑛 𝑛
𝑛−1 [𝜓 (𝑥 + 3)] ((11/6) − 𝛾)
3
( − 𝛾) < < , (65)
𝑢 V 1/𝑝 1/𝑞 2 𝜓 (𝑛𝑥 + 3) 𝐻𝑛+2 − 𝛾
𝜓[ + ] > [𝜓 (𝑢)] [𝜓 (V)] , (56)
𝑝 𝑞
where 𝐻𝑛 = ∑𝑛𝑘=1 (1/𝑘) is the 𝑛th harmonic number.
where 𝑝 > 1, 𝑞 > 1, (1/𝑝) + (1/𝑞) = 1, 𝑢 ⩾ 𝑐, and V ⩾ 𝑐.
If 𝑝 = 𝑎 and 𝑞 = 𝑎/(𝑎 − 1), then Proof. By [6], for all integers 𝑛 ⩾ 1, we have

1 1 1/𝑎 1−(1/𝑎) 𝜓 (𝑛 + 1) = 𝐻𝑛 − 𝛾, (66)

𝜓 [ 𝑢 + (1 − ) V] > [𝜓 (𝑢)] [𝜓 (V)] (57)
𝑎 𝑎
and replacing 𝑎 by 𝑛 in Theorem 7, the proof is completed.
for 𝑢 ⩾ 𝑐 and V ⩾ 𝑐.
6 Journal of Applied Mathematics

Theorem 9. Let 𝑓 be a function defined by Corollary 13. For all 𝑥 ∈ (0, 1) and all 𝑎 > 0 and 𝑏 < 0, one
has
𝑎
[𝜓 (3 + 𝑏𝑥)] 𝑎 𝑎
𝑓 (𝑥) = ; ∀𝑥 > 0, (67) [𝜓 (3 + 𝑏𝑥)] [𝜓 (3 + 𝑏𝑦)]
[𝜓 (3 + 𝑎𝑥)]
𝑏
𝑏
< 𝑏
, (74)
[𝜓 (3 + 𝑎𝑥)] [𝜓 (3 + 𝑎𝑦)]
where 3 + 𝑎𝑥 ⩾ 𝑐 and 3 + 𝑏𝑥 ⩾ 𝑐; then for all 𝑎 > 𝑏 > 0 or where 3 + 𝑎𝑥 ⩾ 𝑐, 3 + 𝑏𝑥 ⩾ 𝑐, 3 + 𝑎𝑦 ⩾ 𝑐, 3 + 𝑏𝑦 ⩾ 𝑐, and
0 > 𝑎 > 𝑏 (𝑎 > 0 and 𝑏 < 0), 𝑓 is strictly increasing (strictly 0 < 𝑦 < 𝑥 < 1.
decreasing) on (0, ∞).
Remark 14. Taking 𝑎 = 𝑛 and 𝑏 = 1 in Corollary 10, we obtain
Proof. Let 𝑔 be a function defined by inequalities of Corollary 8.
𝑔 (𝑥) = log 𝑓 (𝑥) = 𝑎 log 𝜓 (3 + 𝑏𝑥) − 𝑏 log 𝜓 (3 + 𝑎𝑥) ; (68)
Conflict of Interests
then
The authors declare that there is no conflict of interests
󸀠 󸀠 regarding the publication of this paper.
𝜓 (3 + 𝑏𝑥) 𝜓 (3 + 𝑎𝑥)
𝑔󸀠 (𝑥) = 𝑎𝑏 [ − ]. (69)
𝜓 (3 + 𝑏𝑥) 𝜓 (3 + 𝑎𝑥)
References
By proof of Theorem 7, we have
[1] H. Alzer and S. Ruscheweyh, “A subadditive property of
󸀠󸀠 the gamma function,” Journal of Mathematical Analysis and
(log 𝜓 (𝑥)) < 0, for 𝑥 ∈ [𝑐, ∞) ; (70)
Applications, vol. 285, no. 2, pp. 564–577, 2003.
[2] H. Alzer, “Some gamma function inequalities,” Mathematics of
this implies that 𝑔󸀠 (𝑥) > 0 if 𝑎 > 𝑏 > 0 or 0 > 𝑎 > 𝑏 (𝑔󸀠 (𝑥) < 0 Computation, vol. 60, no. 201, pp. 337–346, 1993.
if 𝑎 > 0 and 𝑏 < 0); that is, 𝑔 is strictly increasing on (0, ∞) [3] A. V. Boyd, “Gurland’s inequality for the gamma function,”
(strictly decreasing on (0, ∞)). Hence 𝑓 is strictly increasing Scandinavian Actuarial Journal, vol. 1960, pp. 134–135, 1960.
on (0, ∞), if 𝑎 > 𝑏 > 0 or 0 > 𝑎 > 𝑏 (strictly decreasing if [4] J. Bustoz and M. E. H. Ismail, “On gamma function inequalities,”
𝑎 > 0 and 𝑏 < 0). Mathematics of Computation, vol. 47, no. 176, pp. 659–667, 1986.
[5] C. P. Chen, “Asymptotic expansions of the logarithm of the
Corollary 10. For all 𝑥 ∈ (0, 1) and all 𝑎 > 𝑏 > 0 or 0 > 𝑎 > 𝑏, gamma function in the terms of the polygamma functions,”
one has Mathematical Inequalities and Applications, vol. 2, pp. 521–530,
𝑎−𝑏 𝑎 𝑎 2014.
3 [𝜓 (3 + 𝑏𝑥)] [𝜓 (3 + 𝑏)]
( − 𝛾) < 𝑏
< 𝑏
, (71) [6] C.-P. Chen, “Some properties of functions related to the gamma,
2 [𝜓 (3 + 𝑎𝑥)] [𝜓 (3 + 𝑎)] psi and tetragamma functions,” Computers & Mathematics with
Applications, vol. 62, no. 9, pp. 3389–3395, 2011.
where 3 + 𝑏𝑥 ⩾ 𝑐, 3 + 𝑎𝑥 ⩾ 𝑐, 3 + 𝑏 ⩾ 𝑐, and 3 + 𝑎 ⩾ 𝑐. [7] C. P. Chen, “Unified treatment of several asymptotic formulas
for the gamma function,” Numerical Algorithms, vol. 64, no. 2,
Proof. To prove (71), applying Theorem 9 and taking account pp. 311–319, 2013.
of 𝜓(3) = (3/2) − 𝛾, we get 𝑓(0) < 𝑓(𝑥) < 𝑓(1) for all 𝑥 ∈ [8] T. Erber, “The gamma function inequalities of Gurland and
(0, 1), and we obtain (71). Gautschi,” vol. 1960, no. 1-2, pp. 27–28, 1960.
[9] J. D. Keckic and P. M. Vasic, “Some inequalities for the gamma
Corollary 11. For all 𝑥 ∈ (0, 1) and all 𝑎 > 0 and 𝑏 < 0, one function,” Publications de l’Institut Mathématique, vol. 11, pp.
has 107–114, 1971.
𝑎 𝑎 [10] A. Laforgia, “Further inequalities for the gamma function,”
[𝜓 (3 + 𝑏)] [𝜓 (3 + 𝑏𝑥)] 𝑎−𝑏
3 Mathematics of Computation, vol. 42, no. 166, pp. 597–600, 1984.
𝑏
< 𝑏
< ( − 𝛾) , (72)
[𝜓 (3 + 𝑎)] [𝜓 (3 + 𝑎𝑥)] 2 [11] D. Lu and X. Wang, “A new asymptotic expansion and some
inequalities for the gamma function,” Journal of Number Theory,
where 3 + 𝑎𝑥 ⩾ 𝑐, 3 + 𝑏𝑥 ⩾ 𝑐, 3 + 𝑏 ⩾ 𝑐, and 3 + 𝑎 ⩾ 𝑐. vol. 140, pp. 314–323, 2014.
[12] C. Mortici, “A continued fraction approximation of the gamma
Proof. Applying Theorem 9, we get 𝑓(1) < 𝑓(𝑥) < 𝑓(0) for function,” Journal of Mathematical Analysis and Applications,
all 𝑥 ∈ (0, 1), and we obtain (72). vol. 402, no. 2, pp. 405–410, 2013.
[13] C. Mortici, “Best estimates of the generalized Stirling formula,”
Corollary 12. For all 𝑥 ∈ (0, 1) and all 𝑎 > 𝑏 > 0 or 0 > 𝑎 > 𝑏, Applied Mathematics and Computation, vol. 215, no. 11, pp.
one has 4044–4048, 2010.
[14] C. Mortici, V. G. Cristea, and D. Lu, “Completely monotonic
𝑎 𝑎
[𝜓 (3 + 𝑏𝑦)] [𝜓 (3 + 𝑏𝑥)] functions and inequalities associated to some ratio of gamma
𝑏
< 𝑏
, (73) function,” Applied Mathematics and Computation, vol. 240, pp.
[𝜓 (3 + 𝑎𝑦)] [𝜓 (3 + 𝑎𝑥)] 168–174, 2014.
[15] C. Mortici, “Estimating gamma function by digamma function,”
where 3 + 𝑎𝑥 ⩾ 𝑐, 3 + 𝑏𝑥 ⩾ 𝑐, 3 + 𝑎𝑦 ⩾ 𝑐, 3 + 𝑏𝑦 ⩾ 𝑐, and Mathematical and Computer Modelling, vol. 52, no. 5-6, pp. 942–
0 < 𝑦 < 𝑥 < 1. 946, 2010.
Journal of Applied Mathematics 7

[16] C. Mortici, “New approximation formulas for evaluating the

ratio of gamma functions,” Mathematical and Computer Mod-
elling, vol. 52, no. 1-2, pp. 425–433, 2010.
[17] C. Mortici, “New improvements of the Stirling formula,” Applied
Mathematics and Computation, vol. 217, no. 2, pp. 699–704,
2010.
[18] C. Mortici, “Ramanujan formula for the generalized Stirling
approximation,” Applied Mathematics and Computation, vol.
217, no. 6, pp. 2579–2585, 2010.
[19] N. Batir, “Some new inequalities for gamma and polygamma
functions,” Journal of Inequalities in Pure and Applied Mathe-
matics, vol. 6, no. 4, article 103, 2005.
[20] B. J. English and G. Rousseau, “Bounds for certain harmonic
sums,” Journal of Mathematical Analysis and Applications, vol.
206, no. 2, pp. 428–441, 1997.
[21] T. P. Dence and J. B. Dence, “A survey of Euler’s constant,”
American Mathematical Monthly, vol. 82, pp. 225–265, 2009.
[22] E. Artin, The Gamma Function (Trans M. Butter), Holt, Rinehart
and Winston, San Francisco, Calif, USA, 1964.
Advances in Advances in Journal of Journal of
Operations Research
Hindawi Publishing Corporation
Decision Sciences
Hindawi Publishing Corporation
Applied Mathematics
Hindawi Publishing Corporation
Algebra
Hindawi Publishing Corporation
Probability and Statistics
Hindawi Publishing Corporation
http://www.hindawi.com Volume 2014 http://www.hindawi.com Volume 2014 http://www.hindawi.com Volume 2014 http://www.hindawi.com Volume 2014 http://www.hindawi.com Volume 2014

The Scientific International Journal of

World Journal
Hindawi Publishing Corporation
Differential Equations
Hindawi Publishing Corporation
http://www.hindawi.com Volume 2014 http://www.hindawi.com Volume 2014

Submit your manuscripts at

http://www.hindawi.com

International Journal of Advances in

Combinatorics
Hindawi Publishing Corporation
Mathematical Physics
Hindawi Publishing Corporation
http://www.hindawi.com Volume 2014 http://www.hindawi.com Volume 2014

Journal of Journal of Mathematical Problems Abstract and Discrete Dynamics in

Complex Analysis
Hindawi Publishing Corporation
Mathematics
Hindawi Publishing Corporation
in Engineering
Hindawi Publishing Corporation
Applied Analysis
Hindawi Publishing Corporation
Nature and Society
Hindawi Publishing Corporation
http://www.hindawi.com Volume 2014 http://www.hindawi.com Volume 2014 http://www.hindawi.com Volume 2014 http://www.hindawi.com Volume 2014 http://www.hindawi.com Volume 2014

International
Journal of Journal of
Mathematics and
Mathematical
Discrete Mathematics
Sciences

Journal of International Journal of Journal of

Hindawi Publishing Corporation Hindawi Publishing Corporation Volume 2014

Function Spaces
Hindawi Publishing Corporation
Stochastic Analysis
Hindawi Publishing Corporation
Optimization
Hindawi Publishing Corporation
http://www.hindawi.com Volume 2014 http://www.hindawi.com http://www.hindawi.com Volume 2014 http://www.hindawi.com Volume 2014 http://www.hindawi.com Volume 2014