Next Article in Journal
Sine-Cosine Algorithm to Enhance Simulated Annealing for Unrelated Parallel Machine Scheduling with Setup Times
Next Article in Special Issue
Alternating Asymmetric Iterative Algorithm Based on Domain Decomposition for 3D Poisson Problem
Previous Article in Journal
Blended Root Finding Algorithm Outperforms Bisection and Regula Falsi Algorithms
Previous Article in Special Issue
Duality for Unified Higher-Order Minimax Fractional Programming with Support Function under Type-I Assumptions
 
 
Font Type:
Arial Georgia Verdana
Font Size:
Aa Aa Aa
Line Spacing:
Column Width:
Background:
Article

Weighted Fractional Iyengar Type Inequalities in the Caputo Direction

by
George A. Anastassiou
Department of Mathematical Sciences, University of Memphis, Memphis, TN 38152, USA
Mathematics 2019, 7(11), 1119; https://doi.org/10.3390/math7111119
Submission received: 8 October 2019 / Accepted: 14 November 2019 / Published: 16 November 2019
(This article belongs to the Special Issue Multivariate Approximation for solving ODE and PDE)

Abstract

:
Here we present weighted fractional Iyengar type inequalities with respect to L p norms, with 1 p . Our employed fractional calculus is of Caputo type defined with respect to another function. Our results provide quantitative estimates for the approximation of the Lebesgue–Stieljes integral of a function, based on its values over a finite set of points including at the endpoints of its interval of definition. Our method relies on the right and left generalized fractional Taylor’s formulae. The iterated generalized fractional derivatives case is also studied. We give applications at the end.

1. Introduction

We are motivated by the following famous Iyengar inequality (1938), [1].
Theorem 1.
Let f be a differentiable function on a , b and f x M . Then
a b f x d x 1 2 b a f a + f b M b a 4 2 f b f a 2 4 M .
We need
Definition 1
([2]). Let α > 0 , α = n , · the ceiling of the number. Here g A C a , b (absolutely continuous functions) and strictly increasing. We assume that f g 1 n g L a , b . We define the left generalized g-fractional derivative of f of order α as follows:
D a + ; g α f x : = 1 Γ n α a x g x g t n α 1 g t f g 1 n g t d t ,
x a .
If α N , by [3], pp. 360–361, we have that D a + ; g α f C a , b .
We see that
I a + ; g n α f g 1 n g x = D a + ; g α f x , x a .
We set
D a + ; g n f x : = f g 1 n g x ,
D a + ; g 0 f x = f x , x a , b .
When g = i d , then
D a + ; g α f = D a + ; i d α f = D a α f ,
the usual left Caputo fractional derivative.
We mention the following g-left fractional generalized Taylor’s formula:
Theorem 2
([2]). Let g be a strictly increasing function and g A C a , b . We assume that f g 1 A C n g a , g b , i.e., f g 1 n 1 A C g a , g b , where N n = α , α > 0 . Also we assume that f g 1 n g L a , b . Then
f x = f a + k = 1 n 1 f g 1 k g a k ! g x g a k +
1 Γ α a x g x g t α 1 g t D a + ; g α f t d t , x a , b .
Calling R n a , x the remainder of (7), we find that
R n a , x = 1 Γ α g a g x g x z α 1 D a + ; g α f g 1 z d z , x a , b .
We need
Definition 2
([2]). Here g A C a , b and is strictly increasing. We assume that f g 1 n g L a , b , where N n = α , α > 0 . We define the right generalized g-fractional derivative of f of order α as follows:
D b ; g α f x : = 1 n Γ n α x b g t g x n α 1 g t f g 1 n g t d t ,
all x a , b .
If α N , by [3], p. 378, we find that D b ; g α f C a , b .
We see that
I b ; g n α 1 n f g 1 n g x = D b ; g α f x , a x b .
We set
D b ; g n f x = 1 n f g 1 n g x ,
D b ; g 0 f x = f x , x a , b .
When g = i d , then
D b ; g α f x = D b ; i d α f x = D b α f ,
the usual right Caputo fractional derivative.
We mention the g-right generalized fractional Taylor’s formula:
Theorem 3
([2]). Let g be a strictly increasing function and g A C a , b . We assume that f g 1 A C n g a , g b , where N n = α , α > 0 . Also we assume that f g 1 n g L a , b . Then
f x = f b + k = 1 n 1 f g 1 k g b k ! g x g b k +
1 Γ α x b g t g x α 1 g t D b ; g α f t d t , all a x b .
Calling R n b , x the remainder in (13), we find that
R n b , x = 1 Γ α g x g b z g x α 1 D b ; g α f g 1 z d z , x a , b .
Denote by
D b ; g n α : = D b ; g α D b ; g α D b ; g α ( n times ) , n N .
We mention the following g-right generalized modified Taylor’s formula:
Theorem 4
([2]). Suppose that F k : = D b ; g k α f , for k = 0 , 1 , , n + 1 , fulfill: F k g 1 A C c , d , where c = g a , d = g b , and F k g 1 g L a , b , where 0 < α 1 . Then
f x = i = 0 n g b g x i α Γ i α + 1 D b ; g i α f b +
1 Γ n + 1 α x b g t g x n + 1 α 1 g t D b ; g n + 1 α f t d t =
i = 0 n g b g x i α Γ i α + 1 D b ; g i α f b + D b ; g n + 1 α f ψ x Γ n + 1 α + 1 g b g x n + 1 α ,
where ψ x x , b , any x a , b .
Denote by
D a + ; g n α : = D a + ; g α D a + ; g α . . . D a + ; g α ( n times ) , n N .
We mention the following g-left generalized modified Taylor’s formula:
Theorem 5
([2]). Suppose that F k : = D a + ; g k α f , for k = 0 , 1 , , n + 1 , fulfill: F k g 1 A C c , d , where c = g a , d = g b , and F k g 1 g L a , b , where 0 < α 1 . Then
f x = i = 0 n g x g a i α Γ i α + 1 D a + ; g i α f a +
1 Γ n + 1 α a x g x g t n + 1 α 1 g t D a + ; g n + 1 α f t d t =
i = 0 n g x g a i α Γ i α + 1 D a + ; g i α f a + D a + ; g n + 1 α f ψ x Γ n + 1 α + 1 g x g a n + 1 α ,
where ψ x a , x , any x a , b .
Next we present generalized fractional Iyengar type inequalities.

2. Main Results

We present the following Caputo type generalized g-fractional Iyengar type inequality:
Theorem 6.
Let g be a strictly increasing function and g A C a , b . We assume that f g 1 A C n g a , g b , where N n = α , α > 0 . We also assume that f g 1 n g L a , b (clearly here it is f C a , b ). Then
(i)
a b f x d g x k = 0 n 1 1 k + 1 ! f g 1 k g a g t g a k + 1
+ 1 k f g 1 k g b g b g t k + 1
max D a + ; g α f L a , b , D b ; g α f L a , b Γ α + 2
g t g a α + 1 + g b g t α + 1 ,
t a , b ,
(ii) at g t = g a + g b 2 , the right hand side of (21) is minimized, and we have:
a b f x d g x k = 0 n 1 1 k + 1 ! g b g a k + 1 2 k + 1
f g 1 k g a + 1 k f g 1 k g b
max D a + ; g α f L a , b , D b ; g α f L a , b Γ α + 2 g b g a α + 1 2 α ,
(iii) if f g 1 k g a = f g 1 k g b = 0 , for k = 0 , 1 , , n 1 , we obtain
a b f x d g x
max D a + ; g α f L a , b , D b ; g α f L a , b g b g a α + 1 Γ α + 2 2 α ,
which is a sharp inequality,
(iv) more generally, for j = 0 , 1 , 2 , , N N , it holds
a b f x d g x k = 0 n 1 1 k + 1 ! g b g a N k + 1
j k + 1 f g 1 k g a + 1 k N j k + 1 f g 1 k g b
max D a + ; g α f L a , b , D b ; g α f L a , b Γ α + 2
g b g a N α + 1 j α + 1 + N j α + 1 ,
(v) if f g 1 k g a = f g 1 k g b = 0 , for k = 1 , , n 1 , from (24) we obtain
a b f x d g x g b g a N j f a + N j f b
max D a + ; g α f L a , b , D b ; g α f L a , b Γ α + 2
g b g a N α + 1 j α + 1 + N j α + 1 ,
j = 0 , 1 , 2 , , N ,
(vi) when N = 2 , j = 1 , (25) turns to
a b f x d g x g b g a 2 f a + f b
max D a + ; g α f L a , b , D b ; g α f L a , b Γ α + 2 g b g a α + 1 2 α .
(vii) when 0 < α 1 , inequality (26) is again valid without any boundary conditions.
Proof. 
We have by (7) that
f x k = 0 n 1 f g 1 k g a k ! g x g a k =
1 Γ α a x g x g t α 1 g t D a + ; g α f t d t ,
x a , b .
Also by (13) we obtain
f x k = 0 n 1 f g 1 k g b k ! g x g b k =
1 Γ α x b g t g x α 1 g t D b ; g α f t d t ,
x a , b .
By (27) we derive (by [4], p. 107)
f x k = 0 n 1 f g 1 k g a k ! g x g a k
D a + ; g α f L a , b Γ α + 1 g x g a α ,
and by (28) we obtain
f x k = 0 n 1 f g 1 k g b k ! g x g b k
D b ; g α f L a , b Γ α + 1 g b g x α ,
x a , b .
Call
φ 1 : = D a + ; g α f L a , b Γ α + 1 ,
and
φ 2 : = D b ; g α f L a , b Γ α + 1 .
Set
φ : = max φ 1 , φ 2 .
That is
f x k = 0 n 1 f g 1 k g a k ! g x g a k φ g x g a α ,
and
f x k = 0 n 1 f g 1 k g b k ! g x g b k φ g b g x α ,
x a , b .
Equivalently, we have
k = 0 n 1 f g 1 k g a k ! g x g a k φ g x g a α
f x k = 0 n 1 f g 1 k g a k ! g x g a k + φ g x g a α ,
and
k = 0 n 1 f g 1 k g b k ! g x g b k φ g b g x α
f x k = 0 n 1 f g 1 k g b k ! g x g b k + φ g b g x α ,
x a , b .
Let any t a , b , then by integration against g over a , t and t , b , respectively, we obtain
k = 0 n 1 f g 1 k g a k + 1 ! g t g a k + 1 φ α + 1 g t g a α + 1
a t f x d g x
k = 0 n 1 f g 1 k g a k + 1 ! g t g a k + 1 + φ α + 1 g t g a α + 1 ,
and
k = 0 n 1 f g 1 k g b k + 1 ! g t g b k + 1 φ α + 1 g b g t α + 1
t b f x d g x
k = 0 n 1 f g 1 k g b k + 1 ! g t g b k + 1 + φ α + 1 g b g t α + 1 .
Adding (37) and (38), we obtain
k = 0 n 1 1 k + 1 ! f g 1 k g a g t g a k + 1
f g 1 k g b g t g b k + 1
φ α + 1 g t g a α + 1 + g b g t α + 1
a b f x d g x
k = 0 n 1 1 k + 1 ! f g 1 k g a g t g a k + 1
f g 1 k g b g t g b k + 1 +
φ α + 1 g t g a α + 1 + g b g t α + 1 ,
t a , b .
Consequently we derive:
a b f x d g x k = 0 n 1 1 k + 1 ! f g 1 k g a g t g a k + 1
+ 1 k f g 1 k g b g b g t k + 1
φ α + 1 g t g a α + 1 + g b g t α + 1 ,
t a , b .
Let us consider
θ z : = z g a α + 1 + g b z α + 1 , z g a , g b .
That is
θ g t = g t g a α + 1 + g b g t α + 1 , t a , b .
We have that
θ z = α + 1 z g a α g b z α = 0 ,
giving z g a α = g b z α and z g a = g b z , that is z = g a + g b 2 the only critical number of θ . We have that θ g a = θ g b = g b g a α + 1 , and θ g a + g b 2 = g b g a α + 1 2 α , which is the minimum of θ over g a , g b .
Consequently the right hand side of (40) is minimized when g t = g a + g b 2 , with value φ α + 1 g b g a α + 1 2 α .
Assuming f g 1 k g a = f g 1 k g b = 0 , for k = 0 , 1 , , n 1 , then we obtain that
a b f x d g x φ α + 1 g b g a α + 1 2 α ,
which is a sharp inequality.
When g t = g a + g b 2 , then (40) becomes
a b f x d g x k = 0 n 1 1 k + 1 ! g b g a k + 1 2 k + 1
f g 1 k g a + 1 k f g 1 k g b
φ α + 1 g b g a α + 1 2 α .
Next let N N , j = 0 , 1 , 2 , , N and g t j = g a + j g b g a N , that is g t 0 = g a , g t 1 = g a + g b g a N , , g t N = g b .
Hence it holds
g t j g a = j g b g a N , g b g t j = N j g b g a N ,
j = 0 , 1 , 2 , , N .
We notice
g t j g a α + 1 + g b g t j α + 1 =
g b g a N α + 1 j α + 1 + N j α + 1 ,
j = 0 , 1 , 2 , , N ,
and (for k = 0 , 1 , , n 1 )
f g 1 k g a g t j g a k + 1 +
1 k f g 1 k g b g b g t j k + 1 =
g b g a N k + 1 f g 1 k g a j k + 1 +
1 k f g 1 k g b N j k + 1 ,
j = 0 , 1 , 2 , , N .
By (40) we have
a b f x d g x k = 0 n 1 1 k + 1 ! g b g a N k + 1
f g 1 k g a j k + 1 + 1 k f g 1 k g b N j k + 1
φ α + 1 g b g a N α + 1 j α + 1 + N j α + 1 ,
j = 0 , 1 , 2 , , N .
If f g 1 k g a = f g 1 k g b = 0 , k = 1 , , n 1 , then (46) becomes
a b f x d g x g b g a N j f a + N j f b
φ α + 1 g b g a N α + 1 j α + 1 + N j α + 1 ,
j = 0 , 1 , 2 , , N .
When N = 2 and j = 1 , then (47) becomes
a b f x d g x g b g a 2 f a + f b
φ α + 1 2 g b g a α + 1 2 α + 1 = φ α + 1 g b g a α + 1 2 α .
Let 0 < α 1 , then n = α = 1 .
In that case, without any boundary conditions, we derive from (48) again that
a b f x d g x g b g a 2 f a + f b
φ α + 1 g b g a α + 1 2 α .
We have proved theorem in all possible cases. □
Next we give modified g-fractional Iyengar type inequalities:
Theorem 7.
Let g be a strictly increasing function and g A C a , b , and f C a , b . Let 0 < α 1 , and F k : = D a + ; g k α f , for k = 0 , 1 , , n + 1 ; n N . We assume that F k g 1 A C g a , g b and F k g 1 g L a , b . Also let F k ¯ : = D b ; g k α f , for k = 0 , 1 , , n + 1 , they fulfill F k ¯ g 1 A C g a , g b and F k ¯ g 1 g L a , b . Then
(i)
a b f x d g x i = 0 n 1 Γ i α + 2 D a + ; g i α f a g t g a i α + 1
+ D b ; g i α f b g b g t i α + 1
max D a + ; g n + 1 α f , a , b , D b ; g n + 1 α f , a , b Γ n + 1 α + 2
g t g a n + 1 α + 1 + g b g t n + 1 α + 1 ,
t a , b ,
(ii) at g t = g a + g b 2 , the right hand side of (50) is minimized, and we have:
a b f x d g x i = 0 n 1 Γ i α + 2 g b g a i α + 1 2 i α + 1
D a + ; g i α f a + D b ; g i α f b
max D a + ; g n + 1 α f , a , b , D b ; g n + 1 α f , a , b Γ n + 1 α + 2 g b g a n + 1 α + 1 2 n + 1 α ,
(iii) assuming D a + ; g i α f a = D b ; g i α f b = 0 , for i = 0 , 1 , , n , we obtain
a b f x d g x
max D a + ; g n + 1 α f , a , b , D b ; g n + 1 α f , a , b Γ n + 1 α + 2 g b g a n + 1 α + 1 2 n + 1 α ,
which is a sharp inequality,
(iv) more generally, for j = 0 , 1 , 2 , , N N , it holds
a b f x d g x i = 0 n 1 Γ i α + 2 g b g a N i α + 1
D a + ; g i α f a j i α + 1 + D b ; g i α f b N j i α + 1
max D a + ; g n + 1 α f , a , b , D b ; g n + 1 α f , a , b Γ n + 1 α + 2
g b g a N n + 1 α + 1 j n + 1 α + 1 + N j n + 1 α + 1 ,
(v) if D a + ; g i α f a = D b ; g i α f b = 0 , for i = 1 , , n , from (53) we obtain:
a b f x d g x g b g a N j f a + N j f b
max D a + ; g n + 1 α f , a , b , D b ; g n + 1 α f , a , b Γ n + 1 α + 2
g b g a N n + 1 α + 1 j n + 1 α + 1 + N j n + 1 α + 1 ,
for j = 0 , 1 , 2 , , N ,
(vi) when N = 2 , j = 1 , (54) becomes
a b f x d g x g b g a 2 f a + f b
max D a + ; g n + 1 α f , a , b , D b ; g n + 1 α f , a , b Γ n + 1 α + 2 g b g a n + 1 α + 1 2 n + 1 α .
Proof. 
We have by (19) that
f x = i = 0 n g x g a i α Γ i α + 1 D a + ; g i α f a +
1 Γ n + 1 α a x g x g t n + 1 α 1 g t D a + ; g n + 1 α f t d t ,
x a , b .
Also by (16) we find
f x = i = 0 n g b g x i α Γ i α + 1 D b ; g i α f b +
1 Γ n + 1 α x b g t g x n + 1 α 1 g t D b ; g n + 1 α f t d t ,
x a , b .
Clearly here it is D a + ; g n + 1 α f , D b ; g n + 1 α f C a , b .
By (56) we derive (by [4], p. 107)
f x i = 0 n g x g a i α Γ i α + 1 D a + ; g i α f a
D a + ; g n + 1 α f , a , b g x g a n + 1 α Γ n + 1 α + 1 ,
and by (57) we obtain
f x i = 0 n g b g x i α Γ i α + 1 D b ; g i α f b
D b ; g n + 1 α f , a , b g b g x n + 1 α Γ n + 1 α + 1 ,
x a , b .
Call
γ 1 : = D a + ; g n + 1 α f , a , b Γ n + 1 α + 1 ,
and
γ 2 : = D b ; g n + 1 α f , a , b Γ n + 1 α + 1 .
Set
γ : = max γ 1 , γ 2 .
That is
f x i = 0 n g x g a i α Γ i α + 1 D a + ; g i α f a γ g x g a n + 1 α ,
and
f x i = 0 n g b g x i α Γ i α + 1 D b ; g i α f b γ g b g x n + 1 α ,
x a , b .
Equivalently, we have
i = 0 n g x g a i α Γ i α + 1 D a + ; g i α f a γ g x g a n + 1 α f x
i = 0 n g x g a i α Γ i α + 1 D a + ; g i α f a + γ g x g a n + 1 α ,
and
i = 0 n g b g x i α Γ i α + 1 D b ; g i α f b γ g b g x n + 1 α f x
i = 0 n g b g x i α Γ i α + 1 D b ; g i α f b + γ g b g x n + 1 α ,
x a , b .
Let any t a , b , then by integration against g over a , t and t , b , respectively, we obtain
i = 0 n D a + ; g i α f a g t g a i α + 1 Γ i α + 2 γ n + 1 α + 1 g t g a n + 1 α + 1
a t f x d g x
i = 0 n D a + ; g i α f a g t g a i α + 1 Γ i α + 2 + γ n + 1 α + 1 g t g a n + 1 α + 1 ,
and
i = 0 n g b g t i α + 1 Γ i α + 2 D b ; g i α f b γ n + 1 α + 1 g b g t n + 1 α + 1
t b f x d g x
i = 0 n g b g t i α + 1 Γ i α + 2 D b ; g i α f b + γ n + 1 α + 1 g b g t n + 1 α + 1 .
Adding (67) and (68), we obtain
i = 0 n 1 Γ i α + 2 D a + ; g i α f a g t g a i α + 1 + D b ; g i α f b g b g t i α + 1
γ n + 1 α + 1 g t g a n + 1 α + 1 + g b g t n + 1 α + 1
a b f x d g x
i = 0 n 1 Γ i α + 2 D a + ; g i α f a g t g a i α + 1 + D b ; g i α f b g b g t i α + 1
+ γ n + 1 α + 1 g t g a n + 1 α + 1 + g b g t n + 1 α + 1 ,
t a , b .
Consequently, we derive:
a b f x d g x i = 0 n 1 Γ i α + 2 D a + ; g i α f a g t g a i α + 1
+ D b ; g i α f b g b g t i α + 1
γ n + 1 α + 1 g t g a n + 1 α + 1 + g b g t n + 1 α + 1 ,
t a , b .
Let us consider
ϕ z : = z g a n + 1 α + 1 + g b z n + 1 α + 1 ,
z g a , g b .
That is
ϕ g t = g t g a n + 1 α + 1 + g b g t n + 1 α + 1 ,
t a , b .
We have that
ϕ z = n + 1 α + 1 z g a n + 1 α g b z n + 1 α = 0 ,
giving z g a n + 1 α = g b z n + 1 α and z g a = g b z , that is z = g a + g b 2 the only critical number of ϕ . We have that
ϕ g a = ϕ g b = g b g a n + 1 α + 1 ,
and
ϕ g a + g b 2 = g b g a n + 1 α + 1 2 n + 1 α ,
which is the minimum of ϕ over g a , g b .
Consequently, the right hand side of (70) is minimized when g t = g a + g b 2 , for some t a , b , with value γ n + 1 α + 1 g b g a n + 1 α + 1 2 n + 1 α .
Assuming D a + ; g i α f a = D b ; g i α f b = 0 , i = 0 , 1 , , n , then we obtain that
a b f x d g x γ n + 1 α + 1 g b g a n + 1 α + 1 2 n + 1 α ,
which is a sharp inequality.
When g t = g a + g b 2 , then (70) becomes
a b f x d g x i = 0 n 1 Γ i α + 2 g b g a i α + 1 2 i α + 1
D a + ; g i α f a + D b ; g i α b
γ n + 1 α + 1 g b g a n + 1 α + 1 2 n + 1 α .
Next let N N , j = 0 , 1 , 2 , , N and g t j = g a + j g b g a N , that is g t 0 = g a , g t 1 = g a + g b g a N , , g t N = g b .
Hence it holds
g t j g a = j g b g a N , g b g t j = N j g b g a N ,
j = 0 , 1 , 2 , , N .
We notice
g t j g a n + 1 α + 1 + g b g t j n + 1 α + 1 =
g b g a N n + 1 α + 1 j n + 1 α + 1 + N j n + 1 α + 1 ,
j = 0 , 1 , 2 , , N ,
and (for i = 0 , 1 , , n )
D a + ; g i α f a g t j g a i α + 1 + D b ; g i α f b g b g t j i α + 1 =
g b g a N i α + 1 D a + ; g i α f a j i α + 1 + D b ; g i α f b N j i α + 1 ,
for j = 0 , 1 , 2 , , N .
By (70) we have
a b f x d g x i = 0 n 1 Γ i α + 2 g b g a N i α + 1
D a + ; g i α f a j i α + 1 + D b ; g i α f b N j i α + 1
γ n + 1 α + 1 g b g a N n + 1 α + 1 j n + 1 α + 1 + N j n + 1 α + 1 ,
j = 0 , 1 , 2 , , N .
If D a + ; g i α f a = D b ; g i α f b = 0 , i = 1 , , n , then (76) becomes
a b f x d g x g b g a N j f a + N j f b
γ n + 1 α + 1 g b g a N n + 1 α + 1 j n + 1 α + 1 + N j n + 1 α + 1 ,
j = 0 , 1 , 2 , , N .
When N = 2 and j = 1 , then (77) becomes
a b f x d g x g b g a 2 f a + f b
γ n + 1 α + 1 2 g b g a n + 1 α + 1 2 n + 1 α + 1 =
γ n + 1 α + 1 g b g a n + 1 α + 1 2 n + 1 α .
We have proved theorem in all possible cases. □
We give L 1 variants of last theorems:
Theorem 8.
All as in Theorem 6 with α 1 . If α = n N , we assume that f g 1 n g C a , b . Then
(i)
a b f x d g x k = 0 n 1 1 k + 1 ! f g 1 k g a g t g a k + 1
+ 1 k f g 1 k g b g b g t k + 1
max D a + ; g α f L 1 a , b , g , D b ; g α f L 1 a , b , g Γ α + 1
g t g a α + g b g t α ,
t a , b ,
(ii) at g t = g a + g b 2 , the right hand side of (79) is minimized, and we find:
a b f x d g x k = 0 n 1 1 k + 1 ! g b g a k + 1 2 k + 1
f g 1 k g a + 1 k f g 1 k g b
max D a + ; g α f L 1 a , b , g , D b ; g α f L 1 a , b , g Γ α + 1 g b g a α 2 α 1 ,
(iii) if f g 1 k g a = f g 1 k g b = 0 , for k = 0 , 1 , , n 1 , we obtain
a b f x d g x
max D a + ; g α f L 1 a , b , g , D b ; g α f L 1 a , b , g Γ α + 1 g b g a α 2 α 1 ,
which is a sharp inequality,
(iv) more generally, for j = 0 , 1 , 2 , , N N , it holds that
a b f x d g x k = 0 n 1 1 k + 1 ! g b g a N k + 1
j k + 1 f g 1 k g a + 1 k N j k + 1 f g 1 k g b
max D a + ; g α f L 1 a , b , g , D b ; g α f L 1 a , b , g Γ α + 1
g b g a N α j α + N j α ,
(v) if f g 1 k g a = f g 1 k g b = 0 , for k = 1 , , n 1 , from (82) we obtain
a b f x d g x g b g a N j f a + N j f b
max D a + ; g α f L 1 a , b , g , D b ; g α f L 1 a , b , g Γ α + 1
g b g a N α j α + N j α ,
j = 0 , 1 , 2 , , N ,
(vi) when N = 2 , j = 1 , (83) turns to
a b f x d g x g b g a 2 f a + f b
max D a + ; g α f L 1 a , b , g , D b ; g α f L 1 a , b , g Γ α + 1 g b g a α 2 α 1 .
Proof. 
From (27) we have
f x k = 0 n 1 f g 1 k g a k ! g x g a k
1 Γ α a x g x g t α 1 g t D a + ; g α f t d t
g x g a α 1 Γ α a x g t D a + ; g α f t d t
g x g a α 1 Γ α a b g t D a + ; g α f t d t =
g x g a α 1 Γ α a b D a + ; g α f t d g t =
D a + ; g α f L 1 a , b , g Γ α g x g a α 1 ,
x a , b .
Similarly, from (28) we obtain
f x k = 0 n 1 f g 1 k g b k ! g x g b k
1 Γ α x b g t g x α 1 g t D b ; g α f t d t
g b g x α 1 Γ α x b D b ; g α f t d g t
D b ; g α f L 1 a , b , g Γ α g b g x α 1 ,
x a , b .
Call
δ : = max D a + ; g α f L 1 a , b , g , D b ; g α f L 1 a , b , g .
We have proved that
f x k = 0 n 1 f g 1 k g a k ! g x g a k
δ Γ α g x g a α 1 ,
and
f x k = 0 n 1 f g 1 k g b k ! g x g b k
δ Γ α g b g x α 1 ,
x a , b .
The rest of the proof is as in Theorem 6. □
It follows
Theorem 9.
All as in Theorem 7, with 1 n + 1 α 1 . Call
ρ : = max D a + ; g n + 1 α f L 1 a , b , g , D b ; g n + 1 α f L 1 a , b , g .
Then
(i)
a b f x d g x i = 0 n 1 Γ i α + 2 D a + ; g i α f a g t g a i α + 1
+ D b ; g i α f b g b g t i α + 1
ρ Γ n + 1 α + 1 g t g a n + 1 α + g b g t n + 1 α ,
t a , b ,
(ii) at g t = g a + g b 2 , the right hand side of (91) is minimized, and we find:
a b f x d g x i = 0 n 1 Γ i α + 2 g b g a i α + 1 2 i α + 1
D a + ; g i α f a + D b ; g i α f b
ρ Γ n + 1 α + 1 g b g a n + 1 α 2 n + 1 α 1 ,
(iii) assuming D a + ; g i α f a = D b ; g i α f b = 0 , i = 0 , 1 , , n , we obtain
a b f x d g x
ρ Γ n + 1 α + 1 g b g a n + 1 α 2 n + 1 α 1 ,
which is a sharp inequality,
(iv) more generally, for j = 0 , 1 , 2 , , N N , it holds that
a b f x d g x i = 0 n 1 Γ i α + 2 g b g a N i α + 1
D a + ; g i α f a j i α + 1 + D b ; g i α f b N j i α + 1
ρ Γ n + 1 α + 1 g b g a N n + 1 α j n + 1 α + N j n + 1 α ,
(v) if D a + ; g i α f a = D b ; g i α f b = 0 , i = 1 , , n , from (94) we find:
a b f x d g x g b g a N j f a + N j f b
ρ Γ n + 1 α + 1 g b g a N n + 1 α j n + 1 α + N j n + 1 α ,
for j = 0 , 1 , 2 , , N ,
(vi) when N = 2 and j = 1 , (95) becomes
a b f x d g x g b g a 2 f a + f b
ρ Γ n + 1 α + 1 g b g a n + 1 α 2 n + 1 α 1 .
Proof. 
By (56) we obtain
f x i = 0 n g x g a i α Γ i α + 1 D a + ; g i α f a
1 Γ n + 1 α a x g x g t n + 1 α 1 g t D a + ; g n + 1 α f t d t
g x g a n + 1 α 1 Γ n + 1 α a x g t D a + ; g n + 1 α f t d t
g x g a n + 1 α 1 Γ n + 1 α a b g t D a + ; g n + 1 α f t d t =
g x g a n + 1 α 1 Γ n + 1 α a b D a + ; g n + 1 α f t d g t =
D a + ; g n + 1 α f L 1 a , b , g Γ n + 1 α g x g a n + 1 α 1 ,
x a , b .
Similarly, from (57) we derive
f x i = 0 n g b g x i α Γ i α + 1 D b ; g i α f b
1 Γ n + 1 α x b g t g x n + 1 α 1 g t D b ; g n + 1 α f t d t
g b g x n + 1 α 1 Γ n + 1 α x b D b ; g n + 1 α f t d g t
D b ; g n + 1 α f L 1 a , b , g Γ n + 1 α g b g x n + 1 α 1 ,
x a , b .
We have proved that
f x i = 0 n g x g a i α Γ i α + 1 D a + ; g i α f a
ρ Γ n + 1 α g x g a n + 1 α 1 ,
and
f x i = 0 n g b g x i α Γ i α + 1 D b ; g i α f b
ρ Γ n + 1 α g b g x n + 1 α 1 ,
x a , b .
The rest of the proof is as in Theorem 7. □
Next follow L p variants of Theorems 6 and 7.
Theorem 10.
All as in Theorem 6 with α 1 , and p , q > 1 : 1 p + 1 q = 1 . If α = n N , we assume that f g 1 n g C a , b . Set
μ : = max D a + ; g α f L q a , b , g , D b ; g α f L q a , b , g .
Then
(i)
a b f x d g x k = 0 n 1 1 k + 1 ! f g 1 k g a g t g a k + 1
+ 1 k f g 1 k g b g b g t k + 1
μ Γ α α + 1 p p α 1 + 1 1 p
g t g a α + 1 p + g b g t α + 1 p ,
t a , b ,
(ii) at g t = g a + g b 2 , the right hand side of (102) is minimized, and we have:
a b f x d g x k = 0 n 1 1 k + 1 ! g b g a k + 1 2 k + 1
f g 1 k g a + 1 k f g 1 k g b
μ Γ α α + 1 p p α 1 + 1 1 p g b g a α + 1 p 2 α 1 q ,
(iii) if f g 1 k g a = f g 1 k g b = 0 , for k = 0 , 1 , , n 1 , we obtain
a b f x d g x
μ Γ α α + 1 p p α 1 + 1 1 p g b g a α + 1 p 2 α 1 q ,
which is a sharp inequality,
(iv) more generally, for j = 0 , 1 , 2 , , N N , it holds
a b f x d g x k = 0 n 1 1 k + 1 ! g b g a N k + 1
j k + 1 f g 1 k g a + 1 k N j k + 1 f g 1 k g b
μ Γ α α + 1 p p α 1 + 1 1 p g b g a N α + 1 p j α + 1 p + N j α + 1 p ,
(v) if f g 1 k g a = f g 1 k g b = 0 , for k = 1 , , n 1 , from (105) we obtain
a b f x d g x g b g a N j f a + N j f b
μ Γ α α + 1 p p α 1 + 1 1 p g b g a N α + 1 p j α + 1 p + N j α + 1 p ,
j = 0 , 1 , 2 , , N ,
(vi) when N = 2 , j = 1 , (106) turns to
a b f x d g x g b g a 2 f a + f b
μ Γ α α + 1 p p α 1 + 1 1 p g b g a α + 1 p 2 α 1 q .
Proof. 
From (27) we find
f x k = 0 n 1 f g 1 k g a k ! g x g a k
1 Γ α a x g x g t α 1 g t D a + ; g α f t d t =
(by [5], p. 439)
1 Γ α a x g x g t α 1 D a + ; g α f t d g t
(by [6])
1 Γ α a x g x g t p α 1 d g t 1 p a x D a + ; g α f t q d g t 1 q
1 Γ α g x g a α 1 q p α 1 + 1 1 p D a + ; g α f L q a , b , g .
That is
f x k = 0 n 1 f g 1 k g a k ! g x g a k
D a + ; g α f L q a , b , g Γ α p α 1 + 1 1 p g x g a α 1 q ,
x a , b .
Similarly, from (28) we obtain
f x k = 0 n 1 f g 1 k g b k ! g x g b k
1 Γ α x b g t g x α 1 g t D b ; g α f t d t =
(by [5], p. 439)
1 Γ α x b g t g x α 1 D b ; g α f t d g t
(by [6])
1 Γ α x b g t g x p α 1 d g t 1 p x b D b ; g α f t q d g t 1 q
1 Γ α g b g x α 1 q p α 1 + 1 1 p D b ; g α f L q a , b , g .
That is
f x k = 0 n 1 f g 1 k g b k ! g x g b k
D b ; g α f L q a , b , g Γ α p α 1 + 1 1 p g b g x α 1 q ,
x a , b .
We have proved that
f x k = 0 n 1 f g 1 k g a k ! g x g a k
μ Γ α p α 1 + 1 1 p g x g a α 1 q ,
and
f x k = 0 n 1 f g 1 k g b k ! g x g b k
μ Γ α p α 1 + 1 1 p g b g x α 1 q ,
x a , b .
The rest of the proof is as in Theorem 6. □
We continue with
Theorem 11.
All as in Theorem 7, with 1 n + 1 α 1 , and p , q > 1 : 1 p + 1 q = 1 . Set
θ : = max D a + ; g n + 1 α f L q a , b , g , D b ; g n + 1 α f L q a , b , g .
Then
(i)
a b f x d g x i = 0 n 1 Γ i α + 2 D a + ; g i α f a g t g a i α + 1
+ D b ; g i α f b g b g t i α + 1
θ Γ n + 1 α n + 1 α + 1 p p n + 1 α 1 + 1 1 p
g t g a n + 1 α + 1 p + g b g t n + 1 α + 1 p ,
t a , b ,
(ii) at g t = g a + g b 2 , the right hand side of (115) is minimized, and we have:
a b f x d g x i = 0 n 1 Γ i α + 2 g b g a i α + 1 2 i α + 1
D a + ; g i α f a + D b ; g i α f b
θ Γ n + 1 α n + 1 α + 1 p p n + 1 α 1 + 1 1 p g b g a n + 1 α + 1 p 2 n + 1 α 1 q ,
(iii) assuming D a + ; g i α f a = D b ; g i α f b = 0 , i = 0 , 1 , , n , we obtain
a b f x d g x
θ Γ n + 1 α n + 1 α + 1 p p n + 1 α 1 + 1 1 p g b g a n + 1 α + 1 p 2 n + 1 α 1 q ,
which is a sharp inequality,
(iv) more generally, for j = 0 , 1 , 2 , , N N , it holds that
a b f x d g x i = 0 n 1 Γ i α + 2 g b g a N i α + 1
D a + ; g i α f a j i α + 1 + D b ; g i α f b N j i α + 1
θ Γ n + 1 α n + 1 α + 1 p p n + 1 α 1 + 1 1 p
g b g a N n + 1 α + 1 p j n + 1 α + 1 p + N j n + 1 α + 1 p ,
(v) if D a + ; g i α f a = D b ; g i α f b = 0 , i = 1 , , n , from (118) we obtain:
a b f x d g x g b g a N j f a + N j f b
θ Γ n + 1 α n + 1 α + 1 p p n + 1 α 1 + 1 1 p
g b g a N n + 1 α + 1 p j n + 1 α + 1 p + N j n + 1 α + 1 p ,
j = 0 , 1 , 2 , , N ,
(vi) when N = 2 , j = 1 , (119) turns to
a b f x d g x g b g a 2 f a + f b
θ Γ n + 1 α n + 1 α + 1 p p n + 1 α 1 + 1 1 p g b g a n + 1 α + 1 p 2 n + 1 α 1 q .
Proof. 
By (56) we find
f x i = 0 n g x g a i α Γ i α + 1 D a + ; g i α f a
1 Γ n + 1 α a x g x g t n + 1 α 1 g t D a + ; g n + 1 α f t d t =
(by [5])
1 Γ n + 1 α a x g x g t n + 1 α 1 D a + ; g n + 1 α f t d g t
(by [6])
1 Γ n + 1 α a x g x g t p n + 1 α 1 d g t 1 p
a x D a + ; g n + 1 α f t q d g t 1 q
1 Γ n + 1 α g x g a p n + 1 α 1 + 1 p p n + 1 α 1 + 1 1 p D a + ; g n + 1 α f L q a , b , g .
That is
f x i = 0 n g x g a i α Γ i α + 1 D a + ; g i α f a
D a + ; g n + 1 α f L q a , b , g Γ n + 1 α p n + 1 α 1 + 1 1 p g x g a n + 1 α 1 q ,
x a , b .
Similarly, from (57) we derive
f x i = 0 n g b g x i α Γ i α + 1 D b ; g i α f b
1 Γ n + 1 α x b g t g x n + 1 α 1 g t D b ; g n + 1 α f t d t =
(by [5])
1 Γ n + 1 α x b g t g x n + 1 α 1 D b ; g n + 1 α f t d g t
(by [6])
1 Γ n + 1 α x b g t g x p n + 1 α 1 d g t 1 p
x b D b ; g n + 1 α f t q d g t 1 q
1 Γ n + 1 α g b g x p n + 1 α 1 + 1 p p n + 1 α 1 + 1 1 p D b ; g n + 1 α f L q a , b , g .
That is
f x i = 0 n g b g x i α Γ i α + 1 D b ; g i α f b
D b ; g n + 1 α f L q a , b , g Γ n + 1 α p n + 1 α 1 + 1 1 p g b g x n + 1 α 1 q ,
x a , b .
We have proved that
f x i = 0 n g x g a i α Γ i α + 1 D a + ; g i α f a
θ Γ n + 1 α p n + 1 α 1 + 1 1 p g x g a n + 1 α 1 q ,
and
f x i = 0 n g b g x i α Γ i α + 1 D b ; g i α f b
θ Γ n + 1 α p n + 1 α 1 + 1 1 p g b g x n + 1 α 1 q ,
x a , b .
The rest of the proof is as in Theorem 7. □
Applications follow:
Proposition 1.
We assume that f ln x A C n e a , e b , where N n = α , α > 0 . We also assume that f ln x n e x L a , b , f C a , b . Set
T 1 : = max D a + ; e x α f L a , b , D b ; e x α f L a , b .
Then
(i)
a b f x e x d x k = 0 n 1 1 k + 1 ! f ln x k e a e t e a k + 1
1 k f ln x k e b e b e t k + 1
T 1 Γ α + 2 e t e a α + 1 + e b e t α + 1 ,
t a , b ,
(ii) at t = ln e a + e b 2 , the right hand side of (128) is minimized, and we find:
a b f x e x d x k = 0 n 1 1 k + 1 ! e b e a k + 1 2 k + 1
f ln x k e a + 1 k f ln x k e b
T 1 Γ α + 2 e b e a α + 1 2 α ,
(iii) if f ln x k e a = f ln x k e b = 0 , for k = 0 , 1 , , n 1 , we obtain
a b f x e x d x T 1 e b e a α + 1 Γ α + 2 2 α ,
which is a sharp inequality,
(iv) more generally, for j = 0 , 1 , 2 , , N N , it holds
a b f x e x d x k = 0 n 1 1 k + 1 ! e b e a N k + 1
j k + 1 f ln x k e a + 1 k N j k + 1 f ln x k e b
T 1 Γ α + 2 e b e a N α + 1 j α + 1 + N j α + 1 ,
(v) if f ln x k e a = f ln x k e b = 0 , for k = 1 , , n 1 , from (131) we obtain
a b f x e x d x e b e a N j f a + N j f b
T 1 Γ α + 2 e b e a N α + 1 j α + 1 + N j α + 1 ,
j = 0 , 1 , 2 , , N ,
(vi) when N = 2 , j = 1 , (132) turns to
a b f x e x d x e b e a 2 f a + f b
T 1 Γ α + 2 e b e a α + 1 2 α ,
(vii) when 0 < α 1 , inequality (133) is again valid without any boundary conditions.
Proof. 
By Theorem 6, for g x = e x .
We continue with
Proposition 2.
Here f C a , b , where a , b 0 , + . Let 0 < α 1 , and G k : = D a + ; ln x k α f , for k = 0 , 1 , , n + 1 ; n N . We assume that G k e x A C ln a , ln b and G k e x ln x L a , b . Also let G k ¯ : = D b ; ln x k α f , for k = 0 , 1 , , n + 1 , they fulfill G k ¯ e x A C ln a , ln b and G k ¯ e x ln x L a , b . Set
T 2 : = max D a + ; ln x n + 1 α f , a , b , D b ; ln x n + 1 α f , a , b .
Then
(i)
a b f x x d x i = 0 n 1 Γ i α + 2 D a + ; ln x i α f a ln t a i α + 1
+ D b ; ln x i α f b ln b t i α + 1
T 2 Γ n + 1 α + 2 ln t a n + 1 α + 1 + ln b t n + 1 α + 1 ,
t a , b ,
(ii) at t = e ln a b 2 , the right hand side of (135) is minimized, and we have:
a b f x x d x i = 0 n 1 Γ i α + 2 ln b a i α + 1 2 i α + 1
D a + ; ln x i α f a + D b ; ln x i α f b
T 2 Γ n + 1 α + 2 ln b a n + 1 α + 1 2 n + 1 α ,
(iii) assuming D a + ; ln x i α f a = D b ; ln x i α f b = 0 , i = 0 , 1 , , n , we obtain
a b f x x d x T 2 Γ n + 1 α + 2 ln b a n + 1 α + 1 2 n + 1 α ,
which is a sharp inequality,
(iv) more generally, for j = 0 , 1 , 2 , , N N , it holds
a b f x x d x i = 0 n 1 Γ i α + 2 ln b a N i α + 1
D a + ; ln x i α f a j i α + 1 + D b ; ln x i α f b N j i α + 1
T 2 Γ n + 1 α + 2 ln b a N n + 1 α + 1 j n + 1 α + 1 + N j n + 1 α + 1 ,
(v) if D a + ; ln x i α f a = D b ; ln x i α f b = 0 , i = 1 , , n , from (138) we find:
a b f x x d x ln b a N j f a + N j f b
T 2 Γ n + 1 α + 2 ln b a N n + 1 α + 1 j n + 1 α + 1 + N j n + 1 α + 1 ,
for j = 0 , 1 , 2 , , N ,
(vi) if N = 2 and j = 1 , (139) becomes
a b f x x d x ln b a 2 f a + f b
T 2 Γ n + 1 α + 2 ln b a n + 1 α + 1 2 n + 1 α .
Proof. 
By Theorem 7, for g x = ln x .
We could give many other interesting applications that are based in our other theorems, due to lack of space we skip this task.

Funding

This research received no external funding.

Conflicts of Interest

The authors declare no conflict of interest.

References

  1. Iyengar, K.S.K. Note on an inequality. Math. Stud. 1938, 6, 75–76. [Google Scholar]
  2. Anastassiou, G. Advanced Fractional Taylor’s formulae. J. Comput. Anal. Appl. 2016, 21, 1185–1204. [Google Scholar]
  3. Anastassiou, G.; Argyros, I. Intelligent Numerical Methods: Applications to Fractional Calculus; Springer: Heidelberg, Germany; New York, NY, USA, 2016. [Google Scholar]
  4. Royden, H.L. Real Analysis, 2nd ed.; MacMillan Publishing Co., Inc.: New York, NY, USA, 1968. [Google Scholar]
  5. Royden, H.L.; Fitzpatrick, P.M. Real Analysis, 4th ed.; Pearson: New York, NY, USA, 2010. [Google Scholar]
  6. Dragomir, S.S. Inequalities for the Riemann-Stieljes integral of p,q-H-Dominated integrators with applications. Appl. Math. E-Notes 2015, 15, 243–260. [Google Scholar]

Share and Cite

MDPI and ACS Style

Anastassiou, G.A. Weighted Fractional Iyengar Type Inequalities in the Caputo Direction. Mathematics 2019, 7, 1119. https://doi.org/10.3390/math7111119

AMA Style

Anastassiou GA. Weighted Fractional Iyengar Type Inequalities in the Caputo Direction. Mathematics. 2019; 7(11):1119. https://doi.org/10.3390/math7111119

Chicago/Turabian Style

Anastassiou, George A. 2019. "Weighted Fractional Iyengar Type Inequalities in the Caputo Direction" Mathematics 7, no. 11: 1119. https://doi.org/10.3390/math7111119

APA Style

Anastassiou, G. A. (2019). Weighted Fractional Iyengar Type Inequalities in the Caputo Direction. Mathematics, 7(11), 1119. https://doi.org/10.3390/math7111119

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