Fractional Calculus and Applied Analysis (2022) 25:2420–2445

https://doi.org/10.1007/s13540-022-00106-3

ORIGINAL PAPER

Stability analysis of Hadamard and Caputo-Hadamard

fractional nonlinear systems without and with delay

Bin-Bin He1 · Hua-Cheng Zhou2 · Chun-Hai Kou3

Received: 12 May 2022 / Revised: 2 November 2022 / Accepted: 3 November 2022 /

Published online: 14 November 2022
© Diogenes Co. Ltd 2022

Abstract
This paper handles with the Hadamard and the Caputo-Hadamard fractional derivative
and stability of related systems without and with delay. Firstly, the derivative inequal-
ities are obtained, which is indispensable in applying the theorems derived in this
paper. Then, for systems without delay, we get the stability results by using the Lya-
punov direct method and for systems with delay, we explore two useful inequalities
to verify the stability. Examples are presented with numerical simulations to illustrate
the effectiveness of our results.

Keywords Hadamard system · Caputo-Hadamard system · Stability · Fractional

Lyapunov method · Fractional Halanay inequality

Mathematics Subject Classification 26A33 · 34A08 · 37B25 · 34K37

1 Introduction

Differential equations with fractional derivatives have achieved growing attentions

in recent years and many monographs have appeared [1–7]. The most common

B Hua-Cheng Zhou
[email protected]
Bin-Bin He
[email protected]
Chun-Hai Kou
[email protected]

1 College of Science, Zhejiang University of Technology, Hangzhou 310023, Zhejiang, People’s

Republic of China
2 School of Mathematics and Statistics, HNP-LAMA, Central South University, Changsha
410083, Hunan, People’s Republic of China
3 Department of Applied Mathematics, Donghua University, Shanghai 201620, People’s Republic of
China

123
Stability analysis of Hadamard and Caputo-Hadamard fractional nonlinear 2421

definitions of fractional calculus (integration and differentiation) are for the Riemann-
Liouville and the Caputo derivatives. Compared with these two types of definitions,
the Hadamard fractional calculus, which was first introduced in 1892 by Hadamard
[8], did not get much attention. Due to the integral kernel of form ln xt , the Hadamard
derivative is efficient in characterizing some ultra-slow diffusion processes, besides,
d α
the fractional power (t dt ) is well suited to the case of the half-axis and is invariant
relative to dilation.
Recently, the studies concerning the Hadamard equations are mainly on the fun-
damental theoretical fields. For example, the existence of the solutions are studied
in [9–11], where the strip conditions and fixed point theorems are used. In [12], the
authors present a new fractional comparison principle, and explored the stability of
Hadamard fractional systems. In [13], the Gronwall inequality and the dependence of
solution on parameters in Hadamard case are investigated. In [14], the asymptotic of
higher order Caputo-Hadamard fractional equations is investigated. The regional con-
trollability and observability of the Caputo-Hadamard fractional systems are studied
in [15, 16]. For more results on the Hadamard calculus and the properties, we refer to
[17–20].
For the application, we know that fractional calculus is widely used in anoma-
lous diffusion, see [21, 22]. For the Hadamard fractional calculus, it is convenient
to analyze the functions describing by logarithmic operator. In [23, 24], the authors
used logarithmic function to describe the ultraslow diffusion, which implies that the
Hadamard derivative is more effective in characterizing ultraslow diffusion than the
Caputo derivative or the Riemann-Liouville derivative. For example, the logarithmic
creep law is used to characterize the ultraslow diffusion of sedimentary rocks [25]
and igneous rocks [26, 27]. In [28], the authors studied the dynamic evolution of
COVID-19 caused by the Omicron variant via the Caputo-Hadamard fractional SEIR
model and the prediction of the spread is obtained. Base on the above applications,
the researches on Hadamard fractional calculus not only have theoretical significance,
but also can provide theoretical support for practical applications.
The Lyapunov direct method is a useful tool for analysing the stability of nonlinear
systems. Its fractional version is first introduced by Li et al. in [29] with the Riemann-
Liouville and the Caputo fractional derivative. In [30–32], the authors obtained very
useful fractional derivative inequalities which makes the fractional Lyapunov method
to handle the stability more implementable. Some results are obtained to deal with
the stability of fractional order systems, like fractional system with noninstantaneous
impulses [33], the fractional boundary value problems with order 2 < α < 3, [34].
Considering the systems with delay, which is a common phenomenon in practi-
cal application, Halanay [35] proposed a powerful tool to verify the stability, called
Halanay inequality. Recently, this classical inequality was extended to fractional case,
see [36, 37], then, some other consequences are obtained for different kinds of systems,
for instance, the fractional difference Halanay inequality [38], the fractional Halanay
inequality with distributed delays [39], the fractional Halanay inequality for stochastic
systems [40]. These researches are mainly on the Caputo or the Riemann-Liouville
fractional case. Very recently, the Caputo-Hadamard Halanay inequality was explored
in [41].

123
2422 B.-B. He et al.

Motivated mainly by [29, 30, 37, 41], this paper concentrates on the stability of
the Hadamard and the Caputo-Hadamard systems. The rest of this article is organized
as follows: Section 2 presents some fundamental conceptions and properties of the
Hadamard and the Caputo-Hadamard fractional calculus. In Section 3, some useful
derivative inequalities are given. Section 4 investigates the stability of fractional sys-
tems without and with delay. To support our theory, some numerical examples are
shown in Section 5.

2 Hadamard and Caputo-Hadamard fractional calculus

Now let us review some notions and properties which are necessary for this present
work. For more related works on fractional calculus, Hadamard and Caputo-Hadamard
operator, we refer to [4–7].

2.1 Definition and properties of Hadamard fractional calculus

Definition 1 [1, Page 110] For a ≥ 0 and t > a, the Hadamard integral and derivative
of order α > 0 are defined as
  
H −α
t 1 t α−1 y(η)
a Dt y(t) = ln dη
a Γ (α) η η

and
   t 
H α d m 1 t m−α−1 y(η)
a Dt y(t) = t ln dη
dt Γ (m − α) a η η

respectively, where m = [α] + 1.

Particularly, for 0 < α < 1, the Hadamard derivative becomes
 t 
H α d 1 t −α y(η)
a Dt y(t) =t ln dη.
dt Γ (1 − α) a η η

Property 1 [1, Page 112, Property 2.24] The Hadamard derivative of a constant is
generally not equal to zero. In fact, for any constant c,
 
H α c t −α
a Dt c = ln , 0 < α < 1.
Γ (1 − α) a

Property 2 [1, Page 113, Corollary 2.4] The equality aH Dtα y(t) = 0 is valid if and only
if

  
t α− j
m
y(t) = c j ln , for any ci ∈ R
a
j=1

123
Stability analysis of Hadamard and Caputo-Hadamard fractional nonlinear 2423

where m = [α] + 1.
Particularly, aH Dtα y(t) = 0 holds iff y(t) = c(ln at )α−1 for any c ∈ R and 0 < α <
1.

Property 3 [1, Page 114, Property 2.26] For α > 0, β > 0, if y(t) is integrable, then

H −α H −β −(α+β)
a Dt a Dt y(t) = aH Dt y(t).

Property 4 [1, Page 115, Property 2.27] If α > β > 0 and y(t) is integrable, then

H β H −α −α+β
a Dt a Dt y(t) = aH Dt y(t). (2.1)

Property 5 [1, Page 116, Property 2.28] If α > 0 and y(t) is integrable, then

H α H −α
a Dt a Dt y(t) = y(t). (2.2)

β−α
Property 6 Assume that y(t) is continuous and the derivative aH Dt y(t) exists when
β > α > 0, then the formula

H β H −α −α+β
a Dt a Dt y(t) = aH Dt y(t) (2.3)

holds for α, β > 0.

Proof If α > β, (2.3) holds directly according to Property 4.

If β > α, let p = [β] + 1 and q = [β − α] + 1, it can be seen that q ≤ p holds.
Then, by the definition of Hadamard derivative and Property 3, we can obtain

H β H −α d p  H −( p−β)  H −α  d p  H β−α− p
a Dt a Dt y(t) = D D y(t) = D y(t)
dt p a t a t
dt p a t
d q  β−α−q β−α
= q aH Dt y(t) = aH Dt y(t).
dt



Property 7 [1, Page 116, Theorem 2.3] If y(t) is integrable and H m−α y(t) ∈
a Dt
AC m [a, b], then


m−1
d
m−k−1 H α−k−1 y(a)  
t α−k−1
H −α H α a Dt
a Dt a Dt y(t) = y(t) − t ln ,
dt Γ (α − k + 1) a
k=1

where m = [α] + 1. In particular, it has

 
H −α H α
H α−1 y(a)
a Dt t α−1
a Dt a Dt y(t) = y(t) − ln for 0 < α < 1.
Γ (α) a

123
2424 B.-B. He et al.

Property 8 Assume that α > 0, 0 < β < 1, then

  H β−1 y(a)
H −α H β β−α a Dt t
a Dt a Dt y(t) = aH Dt y(t) − (ln )α−1 . (2.4)
Γ (α) a

Proof To prove the equality (2.4), we use Property 3 (if β ≤ α) or Property 6 (if
α ≤ β) and then Property 7, which gives

H −α H β β−α H −β H β
a Dt a Dt y(t) = aH Dt a Dt a Dt y(t)
H β−1 y(a)
H β−α a Dt t β−1
= D
a t y(t) − (ln ) (2.5)
Γ (β) a
H β−1 y(a)
β−α a Dt t
= aH Dt y(t) − (ln )α−1 ,
Γ (α) a

where the known Hadamard integral and the derivative of the logarithmic function [1,
Page 112, Property 2.24]

H −α t p−1 Γ ( p) t
a Dt (ln ) = (ln ) p+α−1 ,
a Γ ( p + α) a
H α t Γ ( p) t
a Dt (ln ) = (ln ) p−α−1 ,
p−1
a Γ ( p − α) a

are used. 

Lemma 1 Let 0 < α < 1 and y(t), z(t) ∈ R, if H α−1 y(a) ≥ H α−1 z(a)
a Dt a Dt and
H α H α
a Dt y(t) ≥ a Dt z(t) hold. Then y(t) ≥ z(t) for t > a.

Proof Since aH Dtα y(t) ≥ aH Dtα z(t), a nonnegative function R(t) exists satisfying

H α
a Dt y(t) = R(t) + aH Dtα z(t). (2.6)

Taking the Hadamard integral of order α on both sides of (2.6) and using Property 7
yields
 α−1  
H α−1 y(a)
a Dt t H α−1 z(a)
a Dt t α−1
y(t) − ln = aH Dt−α R(t) + z(t) − ln .
Γ (α) a Γ (α) a

That is
   
1 t α−1 H α−1
y(t) = z(t) + ln a Dt y(a) − a Dt z(a) + aH Dt−α R(t).
H α−1
Γ (α) a
(2.7)

By the assumption aH Dtα−1 y(a) ≥ aH Dtα−1 z(a), then

y(t) ≥ z(t) + aH Dt−α R(t).

123
Stability analysis of Hadamard and Caputo-Hadamard fractional nonlinear 2425

From the definition of Hadamard integral, R(t) ≥ 0 implies aH Dt−α R(t) ≥ 0. Then
y(t) ≥ z(t) holds. 


2.2 Definition and properties of Caputo-Hadamard fractional calculus

Definition 2 [1, 20, 42] The Caputo-Hadamard derivative of order α > 0 is defined
by

  δ k y(a)
m−1   
CH α H α t k
a Dt y(t) = a Dt y(t) − ln , t ≥ a,
k! a
k=0

where δ = t dt
d
and m = [α] + 1.
Particularly, for 0 < α < 1,

CH α
a Dt y(t) = aH Dtα [y(t) − y(a)] .

Lemma 2 [20, Lemma 2.1] [1, Equation 2.7.35] For y(t) ∈ AC m [a, b], C H D α y(t)
a t
exist everywhere and
 t m−α−1  m
CH α 1 t d dη
a Dt y(t) = ln η y(η) , where m = [α] + 1.
Γ (m − α) a η dη η

Particularly, for 0 < α < 1,

 t 
CH α 1 t −α 
a Dt y(t) = ln y (η)dη.
Γ (1 − α) a η

Lemma 3 [20, Theorem 2.2] For any continuous function y(t) , CaH Dtα y(t) is contin-
uous and

CH α
a Dt y(a) = 0, where m = [α] + 1.

Property 9 [20, Lemma 2.4] For α > 0, it has

 
CH α H −α
a Dt a Dt y (t) = y(t).

Property 10 [20, Lemma 2.5] For y(t) ∈ AC[a, b] and 0 < α < 1, it has
 
H −α CH α
a Dt a Dt y(t) = y(t) − y(a).

Lemma 4 For 0 < α < 1, and y(t), z(t) ∈ R, if y(a) ≥ z(a) and C H D α y(t) ≥
a t
C H D α z(t)
hold. Then y(t) ≥ z(t) for t ≥ a.
a t

123
2426 B.-B. He et al.

Proof Since CaH Dtα y(t) ≥ CaH Dtα z(t), a nonnegative function R(t) exists satisfying

CH α
a Dt y(t) = R(t) + CaH Dtα z(t). (2.8)

Taking the Hadamard integral of order α on both sides of (2.8) and using Property 10,
yields

y(t) − y(a) = aH Dt−α R(t) + z(t) − z(a).

By the assumption y(a) ≥ z(a), then

y(t) ≥ aH Dt−α R(t) + z(t). (2.9)

From the definition of Hadamard integral, R(t) ≥ 0 implies aH Dt−α R(t) ≥ 0. Then
y(t) ≥ z(t) holds. 


2.3 Other definitions and properties

Definition 3 [43, Definition 5.11] A continuous function k(t) : R+ → R+ is in class-

K if k(0) = 0 and it is strictly increasing.

Definition 4 The one parameter and two parameter Mittag-Leffler functions are
defined as
∞
  ∞
yk yk
E p (y) = and E p,q (y) =
Γ ( pk + 1) Γ ( pk + q)
k=0 k=0

respectively, where p > 0, q > 0.

Lemma 5 Let 0 < α < 1 and y(a) be a nonnegative constant, then

CH α
a Dt y(t) ≤ aH Dtα y(t).

Proof By the definition of Caputo-Hadamard derivative, for 0 < α < 1, it has

CH α
a Dt y(t) =aH Dtα [y(t) − y(a)]
 
y(a) t α
=aH Dtα y(t) − ln .
Γ (1 − α) a

Since y(a) ≥ 0, we get CaH Dtα y(t) ≤ aH Dtα y(t). 



Lemma 6 [1, Page 235, Example 4.13] The solution to the Cauchy problem with
Hadamard derivative
H α
a Dt y(t) = λy(t) + f (t), 0 < α < 1,
H −(1−α) y(a) = b, λ, b ∈ R,
a Dt

123
Stability analysis of Hadamard and Caputo-Hadamard fractional nonlinear 2427

can be expressed as
   t   f (ξ )
t α−1 t α t α−1 t α
y(t) = b ln E α,α λ ln + ln E α,α λ ln dξ.
a a a ξ ξ ξ

Lemma 7 [41] The solution to the Cauchy problem with Caputo-Hadamard derivative
CH
a Dtα y(t) = λy(t) + f (t), 0 < α < 1,
y(a) = b, λ, b ∈ R,

can be expressed as
   t   f (ξ )
t α t α−1 t α
y(t) = bE α λ ln + ln E α,α λ ln dξ.
a a ξ ξ ξ

3 Several derivative inequaities for Hadamard and

Caputo-Hadamard derivative

This section presents several key lemmas on the inequalities of the Hadamard and the
Caputo-Hadamard derivative, which makes the Lyapunov method and the Halanay
inequality applicable to the Hadamard and the Caputo-Hadamard fractional systems.

3.1 Derivative inequaities for Hadamard fractional derivative

Theorem 1 Let α ∈ (0, 1) and y(t) ∈ Rn be differentiable, V (y) : Rn → R be

continuous and differentiable, suppose that V (0) = 0 and V (y) is convex over Rn ,
then
  ∂V
H α H α
a Dt V y(t) ≤ a Dt y(t) . (3.1)
∂y

Proof Clearly, inequality (3.1) is equivalent to

 
H α H α ∂V
a Dt V y(t) − a Dt y(t) ≤ 0. (3.2)
∂y

It suffices to prove that (3.2) holds. Due to the Newton-Leibniz formula, y(t) can be
expressed as
t
y(t) = y(a + ) + a ẏ(ξ )dξ
t (3.3)
= y(a + ) + a ξ ẏ(ξ ) + H −1 t ẏ(t) .
ξ dξ = y(a ) + a Dt

derivative on both sides of (3.3) and using (2.1), we have

123
2428 B.-B. He et al.

H α
a Dt y(t) = aH Dtα y(a + ) + aH Dtα−1 t ẏ(t)
   t 
y(a + ) t −α 1 t −α
= ln + ln ẏ(ξ )dξ
Γ (1 − α) a Γ (1 − α) a ξ
    t  
1 + t −α t −α
= y(a ) ln + ln ẏ(ξ )dξ . (3.4)
Γ (1 − α) a a ξ

Similar to (3.3) and (3.4), the Hadamard derivative of V (y(t)) is


H α 1 t −α
a Dt V y(t) = V y(a + ) ln
Γ (1 − α) a
 t −α   
t ∂ V y(ξ )
+ ln ẏ(ξ ) dξ . (3.5)
a ξ ∂y

Combing (3.4) and (3.5), we derive

  ∂V
H α H α
a Dt V y(t) − a Dt y(t)
∂y
     ∂ V y(t) 
1 t −α + +
= ln V y(a ) − y(a ) (3.6)
Γ (1 − α) a ∂y
 t  −α   
1 t ∂ V y(ξ ) ∂ V y(t) 
+ ln ẏ(ξ ) − dξ.
a Γ (1 − α) ξ ∂y ∂y

Calculating the integral term of (3.6) leads to

 t    ∂ V y(ξ ) 
t −α  ∂ V y(t)
ln ẏ(ξ ) − dξ
a ξ ∂y ∂y
 t   
t −α ∂ V y(ξ ) ∂ V y(t)
= ln − dy(ξ )
a ξ ∂y ∂y
 t     ∂ V y(t)
t −α
= ln d V y(ξ ) − y(ξ )
a ξ ∂y
  ∂ V y(t) 
+ y(t) − V y(t)
∂y
     ∂ V y(t)   ∂ V y(t)
t −α
= ln V y(ξ ) − y(ξ ) + y(t)
ξ ∂y ∂y
t  t  −α−1 
 1 t
− V y(t)  − α ln V y(ξ )
a a ξ ξ
  ∂ V y(t)   ∂ V y(t) 
− y(ξ ) + y(t) − V y(t) dξ
∂y ∂y
     ∂ V y(t)   ∂ V y(t)
t −α
= lim ln − y(ξ ) + y(t)
ξ →t − ξ ∂y ∂y

123
Stability analysis of Hadamard and Caputo-Hadamard fractional nonlinear 2429

 
    ∂ V y(t)
t −α
× V y(ξ ) − V y(t) V y(a) − y(a)
− ln
a ∂y
  ∂ V y(t)   t  −α−1
1 t
+ y(t) − V y(t) − α ln
∂y a ξ ξ
   ∂ V y(t)   
× y(t) − y(ξ ) − V y(t) − V y(ξ ) dξ. (3.7)
∂y

By the L’Hopital’s rule, one gets

 −α 
  ∂ V y(t)   ∂ V y(t)
t
lim V y(ξ ) − y(ξ )
ln + y(t)
ξ →t − ξ ∂y ∂y
     (3.8)
 ξ ẏ(ξ ) ∂ V (y(ξ
∂y
))
− ẏ(ξ ) ∂ V ∂(y(t))
y
− V y(t) = lim α−1
= 0.
ξ →t − −α ln t − ln ξ

Combing (3.6), (3.7) and (3.8), it is seen that (3.2) can be reduced to
  ∂ V (y(t))
H α H α
a Dt V y(t) − a Dt y(t)
∂y
 −α    ∂ V y(t) 
1 t
= ln V y(t) − y(t)
Γ (1 − α) a ∂y
 t  −α−1    ∂ V y(t)  (3.9)
α 1 t
− ln y(t) − y(ξ )
Γ (1 − α) a ξ ξ ∂y
 
− V y(t) − V y(ξ ) dξ ≤ 0.

Noting
  ∂ V (y(t))
V (y(ξ )) = V (y(t)) + y(ξ ) − y(t)
∂y
1   ∂ V (y(ζ ))
2
+ y(ξ ) − y(t) (y(ξ ) − y(t))
2 ∂ y2

and
 −α−1   
limξ →t − ln ξt y(t) − y(ξ ) ∂ V ∂(y(t))
y − V (y(t)) − V (y(ξ ))
 −α−1
y(ξ ) − y(t) ∂ V∂(y(ζ ))
2
= limξ →t − 21 ln ξt y2
y(ξ ) − y(t)
= 0,

we know
 t    ∂ V (y(t))   
1 t −α−1
ln y(t) − y(ξ ) − V y(t) − V y(ξ ) dξ
a ξ ξ ∂y

123
2430 B.-B. He et al.

makes sense. From the property of convex function and V (y(t)) being convex, it is
seen that
   ∂ V y(t)   
y(t) − y(ξ ) − V y(t) − V y(ξ ) ≥ 0
∂y

∂ V (y(t))
holds. This, together with V (0) = 0, yields V y(t) ≤ y(t) ∂y , which implies
∂V
that (3.9) is always true, hence aH Dtα V y(t) ≤ ∂y
H α
a Dt y(t) holds. 


If we take V (y) = y 2 (t) for y(t) ∈ R or V (y) = y (t)y(t) for y(t) ∈ Rn in

Theorem 1, it is clear that V (y) is convex and V (0) = 0. Then we can obtain the
following corollaries directly.
Corollary 1 Let 0 < α < 1 and y(t) be differentiable, then

1H α 2
D y (t) ≤ y(t)aH Dtα y(t), for y(t) ∈ R,
2a t
and
1H α  
H α
D y (t)y(t) ≤ D y(t) y(t), for y(t) ∈ Rn
2a t a t

hold.
Corollary 2 Let 0 < α < 1 and y(t) ∈ Rn be differentiable vector function, then for
constant matrix S ∈ Rn×n , it has

1H α  
H α
a Dt (y (t)Sy(t)) ≤ a Dt y(t) Sy(t),
2
where S is symmetric and positive definite.
Proof From the property of matrix S, one can see that V (y) = y (t)Sy(t) is convex
and satisfies V (0) = 0. The result can be derived directly according to Theorem 1. 


3.2 Derivative inequaities for Caputo-Hadamard derivative

Theorem 2 Let y(t) ∈ Rn be differentiable, V (y) : Rn → R be continuous and

differentiable, suppose V (y) is convex over Rn , then for any t ≥ a,
  ∂V
CH α CH α
a Dt V y(t) ≤ a Dt y(t) , α ∈ (0, 1). (3.10)
∂y

Proof Clearly, inequality (3.10) means

 
CH α ∂V CH α
a Dt V y(t) − a Dt y(t) ≤ 0.
∂y

123
Stability analysis of Hadamard and Caputo-Hadamard fractional nonlinear 2431

Due to the Newton-Leibniz formula, one has y(t) = y(a) + aH Dt−1 (t ẏ(t)). Then by
Lemma 3 and Property 6, we have

  
CH α H −(1−α)
t 1 t −α
a Dt y(t) = a Dt (t ẏ(t)) = ln ẏ(ξ )dξ. (3.11)
a Γ (1 − α) ξ

Similarly, for V (y(t)), it has

   ∂ V (y(t)) 
V (y(t)) = V y(a) + aH Dt−1 ẏ(t) t , (3.12)
∂y

and

  ∂ V (y(t)) 
CH α
a Dt V y(t) =CaH Dtα−1 t ẏ(t)
∂y
 t −α    ∂ V y(ξ )  (3.13)
1 t
= ln ẏ(ξ ) dξ.
Γ (1 − α) a ξ ∂y

Combing (3.11) and (3.13), we have

 
C HDα V
a t y(t) − CaH Dtα y(t) ∂∂Vy
 −α    ∂ V y(ξ ) 
t ∂ V y(t)
= Γ (1−α)
1
a ln t
ξ ẏ(ξ ) ∂y − ∂y dξ
 −α  
t ∂ V y(ξ ) ∂ V y(t)
= Γ (1−α)
1
a ln ξt ∂y − ∂y dy(ξ )
 −α   
t ∂ V y(t)
= Γ (1−α)
1
a ln ξt d V y(ξ ) − y(ξ ) ∂y
  ∂ V y(t)  (3.14)
+ y(t) ∂y − V y(t)
 −α    ∂ V y(t)
= Γ (1−α)
1
ln ξt V y(ξ ) − V y(t) − y(ξ ) ∂y
  ∂ V y(t) t 
 −α−1 
+ y(t)  − α t 1 ln t V y(ξ )
∂y  a ξ ξ
  ∂ V y(t) a
  ∂ V y(t) 
− y(ξ ) ∂y + y(t) ∂y − V y(t) dξ
 −α    ∂ V y(t) 
= − Γ (1−α)
1
ln y(t) − y(a)
t
a ∂y
  
 −α−1
α t 1
− V y(t) − V y(a) − Γ (1−α) a ξ ln t
ξ
(3.15)
   ∂ V y(t)   
× y(t) − y(ξ ) ∂y − V y(t) − V y(ξ ) dξ.

123
2432 B.-B. He et al.

Since V (y(t)) is convex over Rn , we know from the property of convex function
that
  
∂ V y(t)   
y(t) − y(a) − V y(t) − V y(a) ≥ 0
∂y

and
  
∂ V y(t)   
y(t) − y(ξ ) − V y(t) − V y(ξ ) ≥ 0,
∂y

which, jointly with (3.14) and (3.15), gives

 
CH α CH α ∂ V (y(t))
a Dt V y(t) − a Dt y(t) ≤ 0.
∂y

The proof is completed. 



Analogous to Corollary 1 and Corollary 2, we can obtain the following two corol-
laries.

Corollary 3 Let 0 < α < 1 and function y(t) be differentiable, then

1CH α 2
D y (t) ≤ y(t)CaH Dtα y(t), for y(t) ∈ R,
2a t

and

1CH α  
CH α
D y (t)y(t) ≤ D y(t) y(t), for y(t) ∈ Rn
2a t a t

hold.

Corollary 4 Let 0 < α < 1 and function y(t) ∈ Rn be differentiable, then for constant
matrix S ∈ Rn×n , it has

1CH α  
CH α
a Dt (y (t)Sy(t)) ≤ a Dt y(t) Sy(t),
2

where S is symmetric and positive definite.

4 Fractional nonlinear systems and Hadamard Mittag-Leffler stability

Consider the following fractional nonlinear system

α
a Dt y(t) = f (t, y), (4.1)

123
Stability analysis of Hadamard and Caputo-Hadamard fractional nonlinear 2433

where 0 < α < 1, a Dtα represents the Hadamard derivative or the Caputo-Hadamard
derivative. y ∈ Rn , f : (a, ∞) (for the Hadamard case while [a, ∞) for the Caputo-
Hadamard case) ×Rn → Rn is a nonlinear function.
The initial condition of system (4.1) corresponding to the Hadamard derivative and
the Caputo-Hadamard derivative are, respectively,

H −(1−α)
a Dt y(a) = ya ,

and

y(a) = ya .

Definition 5 y0 is said to be an equilibrium point of the Hadamard or the Caputo-

Hadamard fractional system (4.1) if and only if a Dtα y0 = f (t, y0 ).
For the initial value and equilibrium point of the Hadamard fractional systems and
the Caputo-Hadamard fractional systems, one can see the explanation in [27].
In the follows, we suppose that y = 0 is the equilibrium point of system (4.1). Now
let us introduce the Lyapunov stability of the zero solution to system (4.1).
Definition 6 [27, Definition 2.8,2.9] The solution of system (4.1) is said to be:
(1) stable, if for any initial value ya , there exists an ε > 0 such that y(t) < ε for all
t > a > 0 (or t ≥ a > 0 for Caputo-Hadamard case).
(2) asymptotically stable, iff lim y(t) = 0.
t→+∞

4.1 Lyapunov method for fractional nonlinear systems without delay

Now we present the Lyapunov method for the Hadamard and the Caputo-Hadamard
fractional systems, combining the inequalities of the Hadamard and Caputo-Hadamard
derivative, these theorems can be used to check the stability for some specific examples.
Theorem 3 Let V (y) : Rn → R be a continuously differentiable function satisfying

c1 y u ≤ V (y) ≤ c2 y uv , and (4.2)

CH α
a Dt V (y) ≤ −c3 y ,
uv
(4.3)

where α ∈ (0, 1), c1 , c2 , c3 , u, v are arbitrary positive constants. Then system (4.1)
with Caputo-Hadamard derivative is asymptotically stable.
Proof Combining (4.2) and (4.3), we have

CH α c3
a Dt V (y) ≤− V (y),
c2

which implies that a nonnegative function R(t) exists satisfying

CH α c3
a Dt V (y) =− V (y) − R(t). (4.4)
c2

123
2434 B.-B. He et al.

By Lemma 7, the solution of system (4.4) with initial V (y(a)) = Va is

   
c3 t α
V (y) = Va E α − ln
c2 a
 t α−1    
t c3 t α dξ
− ln E α,α − ln R(ξ ) . (4.5)
a ξ c2 ξ ξ

Since R(t) ≥ 0, the integral term in (4.5) is non-negative. Then,

   
c3 t α
V (y) ≤ Va E α − ln . (4.6)
c2 a

Substituting (4.6) into (4.2) yields

     1
Va c3 t α u
y(t) ≤ Eα − ln .
c1 c2 a

By the property of the Mittag-Leffler function, for α ∈ (0, 1), E α (−t) tends to zero
as t → ∞, so y(t) → 0 as t → ∞. This implies the asymptotically stable of system
(4.1). The proof is completed. 


Corollary 5 Let V (y) : Rn → R be a continuously differentiable function convex over

Rn and satisfies

c1 y u ≤ V (y) ≤ c2 y uv , and
 
∂V CH α
a Dt y(t) ≤ −c3 y
uv
∂y

where α ∈ (0, 1), c1 , c2 , c3 , u, v are arbitrary positive constants. Then system (4.1)
with Caputo-Hadamard derivative is asymptotically stable.

Proof The result of this corollary can be obtained directly according to Theorem 2
and Theorem 3. 


Theorem 4 Let V (y) : Rn → R be a continuously differentiable function satisfying

c1 y u
≤ V (y) ≤ c2 y uv
, and (4.7)
H α
a Dt V (y) ≤ −c3 y ,
uv
(4.8)

where α ∈ (0, 1), c1 , c2 , c3 , u, v are arbitrary positive constants. Then system (4.1)
with Hadamard derivative is asymptotically stable.

Proof By Lemma 5 and the nonnegativeness of V (y(t)), we can obtain

CH α
a Dt V (y) ≤ aH Dtα V (t) ≤ −c3 y uv
,

123
Stability analysis of Hadamard and Caputo-Hadamard fractional nonlinear 2435

which means that condition (4.3) holds. Then by Theorem 3, we can get that

     1
Va c3 t α u
y(t) ≤ Eα − ln ,
c1 c2 a

By the property of the Mittag-Leffler function, for α ∈ (0, 1), E α (−t) tends to zero
as t → ∞, so y(t) → 0 as t → ∞. This implies the system (4.1) with Hadamard
derivative is asymptotically stable. 


Similar to Corollary 5, we have:

Corollary 6 Let V (y) : Rn → R be a continuously differentiable convex function and

satisfy

c1 y u ≤ V (y) ≤ c2 y uv , and
 
∂V H α
a Dt y(t) ≤ −c3 y
uv
∂y

where α ∈ (0, 1), c1 , c2 , c3 , u, v are arbitrary positive constants. Then system (4.1)
with Hadamard derivative is asymptotically stable.

Remark 1 In Theorem 3, Theorem 4, Corollary 5, and Corollary 6, we obtain the

stability in the form of
    
t α c
y(t) ≤ m(ya )E α −d ln ,
a

where c, d > 0, and m(ya ) is a function with respect to the initial condition. This
indicates the state decays towards zero being a composition of the Mittag-Leffler func-
tion and ln at . From the property of the Mittag-Leffler function [44, Theorem 4], it is
seen that the state decays towards zero like [1+(ln1 t )α ]c . In [29], the authors introduce
a
a similar expression and call it Mittag-Leffler stable, which is slightly different from
the asymptotical stable because the Mittag-Leffler stability implies the asymptotical
stability and gives the decay expression while the asymptotical stability cannot gen-
erally tell us the speed of the decay or the decay expression. Following the manner of
[29], here, we can call it Hadamard Mittag-Leffler stable.

Theorem 5 Let V (y) be a Lyapunov function and let γ1 (·) be a class-K function. If

γ1 ( y(t) ) ≤ V (y) (4.9)

and
H α
a Dt V (y) ≤ 0. (4.10)

Then system (4.1) with Hadamard derivative is asymptotically stable.

123
2436 B.-B. He et al.

Proof By (4.10), a nonnegative function R(t) exists satisfying aH Dtα V (y) + R(t) = 0.
Calculating the Hadamard integral of order α on both side of this equation yields
  
Va t t 1 t α−1 R(ξ )
V (t) − (ln )α−1 + ln dξ = 0,
Γ (α) a a Γ (α) ξ ξ

where Va = aH Dtα−1 V (a). Since R(ξ ) ≥ 0, one has

 
Va t α−1
V (t) ≤ ln .
Γ (α) a

Considering (4.9), we have

 
Va t α−1
γ1 ( y ) ≤ V (y) ≤ ln .
Γ (α) a
 
α−1
Then y ≤ γ1−1 Va
Γ (α) ln at . Hence, the asymptotical stability of system (4.1)
follows from the fact that γ1 is in class-K and the fact that as t → ∞, (ln(t/a))α−1 →
0. This ends the proof. 


Theorem 6 Let V (y) be a Lyapunov function and let γ1 (·) be a class-K function. If

γ1 ( y(t) ) ≤ V (y(t)) (4.11)

and

CH α
a Dt V (y(t)) ≤ 0. (4.12)

Then system (4.1) with Caputo-Hadamard derivative is stable.

Proof By (4.12), we can see that a nonnegative function R(t) ≥ 0 exists such that
C H D α V (y(t))
+ R(t) = 0. Calculating the Hadamard integral of order α on this
a t
equation, yields
 t 1 t α−1 R(ξ )
V (y(t)) = V (y(a)) − ln dξ,
a Γ (α) ξ ξ

which, together with R(ξ ) ≥ 0 implies V (y(t)) ≤ V (y(a)). This, combing with
(4.11), gives y(t) ≤ γ1−1 (v(y)) ≤ γ1−1 (V (a)). Hence system (4.1) with Caputo-
Hadamard derivative is stable. 


Remark 2 For the Caputo fractional differential systems, the integer order differential
systems and the Caputo-Hadamard fractional differential systems, corresponding con-
ditions (4.11) and (4.12) can only obtain the stability results, see the Caputo fractional
case in [31, Theorem 3], and the integer order case in [45, Page 62, Theorem 3.2],

123
Stability analysis of Hadamard and Caputo-Hadamard fractional nonlinear 2437

but for the Hadamard fractional differential system, the asymptotical stability can be
realised, the item (ln at )α−1 makes the solution go to zero. This is a difference between
the Caputo type system and the Hadamard system.

According to Theorem 5 and Theorem 6, we have the following corollaries.

Corollary 7 For the Hadamard fractional system

H α
a Dt y(t) = f (y(t)), (4.13)

if y(t) f (y(t)) ≤ 0 holds for ∀y(t) ∈ R, then system (4.13) is asymptotically stable.

Proof Choosing V (y(t)) = 21 y 2 (t), one has aH Dtα V (y(t)) ≤ y(t)aH Dtα y(t) =
y(t) f (y(t)) ≤ 0, from which we see that V (y(t)) satisfies (4.9) and (4.10). Hence,
by Theorem 5, system (4.13) is asymptotically stable. 


Corollary 8 For the Caputo-Hadamard fractional system

CH α
a Dt y(t) = f (y(t)), (4.14)

if y(t) f (y(t)) ≤ 0 holds for ∀y(t) ∈ R, then system (4.14) is stable.

Proof Taking V (y(t)) = 21 y 2 (t), we get CaH Dtα V (y(t)) ≤ y(t)CaH Dtα y(t) =
y(t) f (y(t)) ≤ 0. It follows from Theorem 6 that system (4.14) is stable. 


4.2 Fractional Halanay inequalities for Hadamard and Caputo-Hadamard systems

with time delay

Theorem 7 Suppose that W : (a, +∞) → R+ and τ : (a, +∞) → R+ are con-
tinuous and (ln t)1−α W (t) is continuous, and there exists a monotonically increasing
function g(t) satisfying limt→+∞ g(t) = +∞ such that t − τ (t) ≥ g(t) > a for all
t > a, and g(t)α−1 is integrable on (a, T ] for some T > a. Let W (t) satisfy
H α
a Dt W (t) ≤ −λW (t) + μ sup W (t + σ ), α ∈ (0, 1) (4.15)
−τ (t)≤σ ≤0

for some constants λ > μ > 0. Then lim W (t) = 0.

t→+∞

Proof According to (4.15), we know that there exist a nonnegative function R(t) such
that
H α
a Dt W (t) = −λW (t) − R(t) + μ sup W (t + σ ).
−τ (t)≤σ ≤0

This equation has solution

123
2438 B.-B. He et al.

−(1−α) t α−1  t α
W (t) = aH Dt W (a) ln E α,α − λ ln
a a
 t t α−1  t α   ds
+ ln E α,α − λ ln − R(s) + sup W (s + σ ) .
a s s −τ (s)≤σ ≤0 s

Considering the fact that (ln t)α−1 , E α,α [−λ(ln t)α ], R(t) are nonnegative, we have
the following estimate

−(1−α) t α−1  t α
W (t) ≤ aH Dt W (a) ln E α,α − λ ln
a a
 t
t α−1  t α ds
+ ln E α,α − λ ln sup W (s + σ ) . (4.16)
a s s −τ (s)≤σ ≤0 s

Using the notations u(t) and K (h(t)) with

t α−1
−(1−α)  t α
u(t) = aH Dt E α,α − λ ln
W (a) ln
a a
 α
and K (h(t)) = (h(t))α−1 E α,α −λ h(t) , then the inequality (4.16) can be rewritten
as
 t
W (t) ≤ u(t) + K (h(t) − h(s)) sup W (η)dh(s),
a s−τ (s)≤η≤s

where h(t) = ln t, it can be established that lim u(t) = lim K (h(t)) = 0.

t→+∞ t→+∞
 h(t)
Now we prove μ K L 1 (C ) = μ 0 K (h(s))dh(s) < 1. In fact, for the Mittag-
Leffler function, it has

d α
[t E α,α+1 (−λt α )] = t α−1 E α,α (−λt α ),
dt

if we take h(t) as an independent variable, then we have

d
[h α (t)E α,α+1 (−λh α (t))] = h α−1 (t)E α,α (−λh α (t)),
dh(t)

which implies
 h(t)
h α−1 (s)E α,α (−λh α (s))dh(s) = h α (t)E α,α+1 (−λh α (t)),
0

together with
 
1 1
E α,α+1 (−λh α (t)) = +O ,
λh α (t) λ2 h 2α (t)

123
Stability analysis of Hadamard and Caputo-Hadamard fractional nonlinear 2439

which yields

1
lim h α (t)E α,α+1 (−λh α (t)) = .
h(t)→+∞ λ

By the fact that h(t) = ln t is increasing with respect to t and the fact that
 h(t) α−1
0 h (s) E α,α (−λh α (s))dh(s) is increasing with respect to h(t), we can see
h(t) α−1
that 0 h (s)E α,α (−λh α (s))dh(s) is increasing with respect to t, and so does for
h (t)E α,α+1 (−λh α (t)). Thus
α

 h(t) 1
h α−1 (s)E α,α (−λh α (s))dh(s) ≤ , ∀t > a,
0 λ

combining λ > μ, which implies

 h(t) μ
μ K L 1 (C ) =μ K (h(s)) d(h(s)) ≤ < 1.
0 λ

By [37, Theorem 2], we know that lim W (t) = 0. 


t→+∞

Similar to Theorem 7, we have integral inequality for Caputo-Hadamard fractional

case.
Theorem 8 Suppose that W , τ : [a, +∞) → R+ are continuous. Let τ satisfy t ≥
a + τ (t) for t ∈ [a, +∞) and limt→∞ t − τ (t) = ∞. Let W (t) satisfy
CH α
a Dt W (t) ≤ −λW (t) + μ sup W (t + σ ), α ∈ (0, 1), (4.17)
−τ (t)≤σ ≤0

for some constants λ > μ > 0. Then lim W (t) = 0.

t→+∞

Remark 3 Inequalities (4.15) and (4.17) can be considered as the generalization of the
classic Halanay inequality [35], which are useful in verifying the stability of delay
system. Furthermore, the delay can be bounded, unbounded or distributed.
Remark 4 From Theorem 3, Theorem 4, Corollary 5, and Corollary 6, it is seen that
the state of system without delay under certain conditions can decay towards zero
like [1+(ln1 t )α ]c . An interesting and nontrivial problem arises naturally: Can we get
a
the decay speed of state of system with delay under under certain conditions? More
precisely, we leave an open question here: Do (4.15) and (4.17) imply that W (t) decays
towards zero like [1+(ln1 t )α ]c ?
a

5 Examples

In this section, some illustrative examples are provided to show the usefulness of the
Lyapunov theorem and the Halanay inequality.

123
2440 B.-B. He et al.

Example 1 Consider the following fractional order system with Hadamard derivative

1 Dt y1 (t) = −y1 (t) − 2y2 (t) cos t,

H 0.8
(5.1)
1 Dt y2 (t) = 2y1 (t) cos t − y2 (t).
H 0.8

Letting V (y) = 1
2 y12 (t) + y22 (t) , we derive

1 Dt V (y) ≤ y1 1H Dt0.8 y1 (t) + y2 1H Dt0.8 y2 (t)

H 0.8

= −y12 (t) − 2y1 (t)y2 (t) cos t + 2y1 (t)y2 (t) cos t − y22 (t)
= −y12 (t) − y22 (t)
≤ −2V (y).

By Theorem 3, system (5.1) is asymptotically stable.

Example 2 Consider the following fractional order system with Caputo-Hadamard

derivative
C H D 0.98 y (t) = −4y1 (t) + y2 (t) + 2y2 (t)y3 (t),
1 t 1
C H D 0.98 y (t) = −y1 (t) − 4y2 (t) + 5y1 (t)y3 (t), (5.2)
1 t 2
C H D 0.98 y (t) = −4y3 (t) − 7y1 (t)y2 (t),
1 t 3

with initial value (y1 (1), y2 (1), y3 (1)) = (0.8, 0.6, 0.4) .
Letting V (y) = 21 (y12 (t) + y22 (t) + y32 (t)), it has

C H 0.98
1 Dt V (y)
≤ y1 (t)C1H Dt0.98 y1 (t) + y2 (t)C1H Dt0.98 y2 (t) + y3 (t)C1H Dt0.98 y3 (t)
= −4y12 (t) + y1 (t)y2 (t) + 2y1 (t)y2 (t)y3 (t) − y1 (t)y2 (t) − 4y22 (t)
+5y1 (t)y2 (t)y3 (t) − 4y32 (t) − 7y1 (t)y2 (t)y3 (t)
= −4(y12 (t) + y22 (t) + y32 (t)) = −8V (y)

By Theorem 3, system (5.2) is asymptotically stable. The evolution for the system can
be seen in Fig. 1.

Example 3 Consider the the following Hadamard system with time-varying delay

1 Dt y1 (t) = −4y1 (t) − 2y1 (t − 21 t) + y1 (t) sin(y2 (t)),

H 0.7
(5.3)
1 Dt y2 (t) = − 27 y2 (t) − y2 (t − 21 t) + y2 (t) sin(y1 (t − 21 t)).
H 0.7

Letting W (t) = y12 (t) + y22 (t),we can verify that H1Dt0.7 W (t) ≤ −λW (t) +
μ sup−τ (t)≤η≤0 W (t + η) holds for λ = 4, μ = 1. In fact,

123
Stability analysis of Hadamard and Caputo-Hadamard fractional nonlinear 2441

0.8

0.6 0.8

0.6

0.4

0.4

1 2 3

0.2

-0.1
0 5 10 15 20

Fig. 1 The evolution of System (5.2)

1Dt W (t) + 4W (t) − W (t + η)

H 0.7
sup
−τ (t)≤η≤0
1
≤ 2y1 (t)(−4y1 (t) − 2y1 (t − t) + y1 (t) sin(y2 (t)))
2
7 1 1
+2y2 (t)(− y2 (t) − y2 (t − t) + y2 (t) sin(y1 (t − t)))
2 2 2
1 1
+4(y1 (t) + y2 (t)) − y1 (t − t) − y2 (t − t)
2 2 2 2
2 2
1 1
≤ −2y1 (t) − 4y1 (t)y1 (t − t) − y1 (t − t) − y22 (t)
2 2
2 2
1 1
−2y2 (t)y2 (t − t) − y2 (t − t)
2
2 2
1 2 1
= −(2y1 (t) + y1 (t − t)) − (y2 (t) + y2 (t − t))2 ≤ 0.
2 2
By Theorem 7, system (5.3) is asymptotically stable.
Example 4 Consider the the following Caputo-Hadamard system with time-varying
delay
C H D 0.95 y (t) = −3y1 (t) + 2y1 (t) sin(y2 (t − 41 t)),
1 t 1
C H D 0.95 y (t) (5.4)
1 t 2 = −2y2 (t) + y2 (t) cos(y1 (t − 14 t)).

The initial value is (y1 (1), y2 (1)) = (0.25, 0.2) .

To verify the stability of system (5.4), we check the conditions of Theorem 8. To
this end, let W (t) = y12 (t) + y22 (t). Since

C H D 0.95 W (t) + λW (t) − μ sup W (t + η)

1 t −τ (t)≤η≤0
≤ 2y1 (t) − 3y1 (t) + 2y1 (t) sin(y2 (t − 41 t))
+2y2 (t) − 2y2 (t) + 2y2 (t) cos(y1 (t − 41 t))
+λy12 (t) + λy22 (t) − μ sup− 1 t≤τ (t)≤0 (y12 (t + η) + y22 (t + η))
4

123
2442 B.-B. He et al.

0.25

0.2

0.25

0.15 0.2

0.1 0.1

0
0.05 0 5 10 15 20

0
0 100 200 300 400 500 600 700 800 900 1000

Fig. 2 The evolution of System (5.4)

≤ −(6 − λ)y12 (t) − (4 − λ)y22 (t) + 4y12 (t) sin(y2 (t)(t − 41 t))
+2y22 (t) cos(y1 (t − 41 t)) − μy12 (t − 41 t) − μy22 (t − 41 t)
≤ −(2 − λ)y12 (t) − (2 − λ)y22 (t) − μy12 (t − 14 t) − μy22 (t − 41 t),

the inequality C1H Dt0.95 W (t) ≤ −λW (t) + μ sup−τ (t)≤η≤0 W (t + η) holds for λ =
1, μ = 0.5. Therefore, it is seen that the system (5.4) is asymptotically stable. The
evolution of the system is shown in Fig. 2.

Example 5 Consider the following fractional partial differential system with delay
⎧CH
⎨ 1 Dt0.8 z(x, t) = z x x (x, t) − 2z(x, t) + z(x, t − sin2 t), t > 1,
z(0, t) = z(1, t) = 0, t ≥ 1, (5.5)
⎩
z 0 (x, t) = sin 3π x − 3 sin 2π x, 0 ≤ x ≤ 1, 0 ≤ t ≤ 1.
1
To verify its stability, let W (t) = 0 z 2 (x, t)d x. Then

C H D 0.8 W (t)
1 t
1
≤ 2 0 z(x, t)C1H Dt0.8 z(x, t)d x
1
= 2 0 z(x, t) z x x (x, t) − 2z(x, t) + z(x, t − sin2 t) d x (5.6)
1
= 0 2z(x, t)z x x (x, t) − 4z 2 (x, t) + 2z(x, t)z(x, t − sin2 t) dx
1
= 0 −2z 2x (x, t) − 4z 2 (x, t) + 2z(x, t)z(x, t − sin2 t) d x.

So we have

1 Dt W (t) + λW (t) − μ W (t + η)
C H 0.8
sup
− sin2 t≤η≤0
 1 
≤ − 2z 2x (x, t) + (λ − 4)z 2 (x, t) + 2z(x, t)z(x, t − sin2 t)
0

−μz 2 (x, t − sin2 t) d x

123
Stability analysis of Hadamard and Caputo-Hadamard fractional nonlinear 2443

Fig. 3 The evolution of system (5.5)

 1 
≤ (−2π 2 + λ − 4)z 2 (x, t) + 2z(x, t)z(x, t − sin2 t) − μz 2 (x, t − sin2 t) d x
0
 1   
−2π 2 − 4 + λ 1 z(x, t)
= z(x, t) z(x, t − sin t)
2 d x, (5.7)
0 1 −μ z(x, t − sin2 t)

1 1
where the Wirtinger’s inequality 0 z 2x (x, t)d x ≥ π 2 0 z 2 (x, t)d x is used. It can
be seen that when μ = 1, λ = 2π 2 + 3.5, inequality C1H Dt0.8 W (t) + λW (t) −
μ sup− sin2 t≤η≤0 W (t + η) ≤ is always true. Hence, by Theorem 8, system (5.5)
is asymptotically stable. The evolution of the system is shown in Fig. 3.

6 Concluding remarks

This paper focuses on the stability of the Hadamard and the Caputo-Hadamard
fractional systems. We have derived several useful inequalities of the Hadamard frac-
tional derivative and explored the Lyapunov method to cope with the stability of the
Hadamard and the Caputo-Hadamard fractional nonlinear systems and the Halanay
inequality to deal with the stability of delay systems. There are some other interesting
problems to be studied for the Hadamard fractional calculus such as controllability,
observability and detectability.
Acknowledgements This work is partially supported by Hunan Provincial Natural Science Foundation
of China (No. 2021JJ20081), the science and technology innovation Program of Hunan Province (No.
2022RC1188) and the National Natural Science Foundation of China (No. 62173348).

Declarations
Conflict of interest The authors declare that they have no conflict of interest.

123
2444 B.-B. He et al.

References
1. Kilbas, A.A., Srivastava, H.M., Trujillo, J.J.: Theory and Applications of Fractional Differential Equa-
tions. Elsevier, Amsterdam (2006)
2. Mainardi, F.: Fractional Calculus and Wave in Linear Viscoelasticity: An Introduction to Mathematical
Models. World Scientific, Singapore (2010)
3. Monje, C.A., Chen, Y.Q., Vinagre, et al.: Fractional-order Systems and Controls: Fundamentals and
Applications. Springer, London (2010)
4. Hilfer, R.: Applications of Fractional Calculus in Physics. World Scientific, Singapore (2000)
5. Miller, K.S., Ross, B.: An Introduction to the Fractional Calculus and Fractional Differential Equation.
Wiley, New York (1993)
6. Podlubny, I.: Fractional Differential Equations. Academic Press, New York (1998)
7. Samko, S.G., Kilbas, A.A., Marichev, O.I.: Fractional Integrals and Derivatives: Theory and Applica-
tions. Gordon and Breach Science Publishers, Amsterdam (1993)
8. Hadamard, J.: Essai sur l’étude des fonctionsdonnées par leur développement de Taylor. Journal de
Mathématiques Pures et Appliquées 8, 101–186 (1892)
9. Wang, G.T., Wang, T.L.: On a nonlinear Hadamard type fractional differential equation with p-
Laplacian operator and strip condition. J. Nonlinear Sci. Appl. 9, 5073–5081 (2016)
10. Abbas, S., Benchohra, M., Hamidi, N., Henderson, J.: Caputo-Hadamard fractional differential equa-
tions in Banach spaces. Fract. Calc. Appl. Anal. 21(4), 1027–1045 (2018). https://doi.org/10.1515/
fca-2018-0056
11. Abbas, S., Benchohra, M., Lazreg, J.E., Zhou, Z.: A survey on Hadamard and Hilfer fractional differ-
ential equations: Analysis and stability. Chaos Soliton. Fract. 102, 47–71 (2017)
12. Wang, G., Pei, K., Chen, Y.Q.: Stability analysis of nonlinear Hadamard fractional differential system.
J Franklin I. 356(12), 6538–6546 (2019)
13. Gong, Z., Qian, D., Li, C., et al.: On the Hadamard type fractional differential system. In: Baleanu,
D., Machado, J., Luo, A. (eds.) Fractional Dynamics and Control, pp. 159–171. Springer, New York
(2012)
14. Graef, J.R., Grace, S.R., Tunc, E.: Asymptotic behavior of solutions of nonlinear fractional differential
equations with Caputo-Type Hadamard derivatives. Fract. Calc. Appl. Anal. 20(1), 71–87 (2017).
https://doi.org/10.1515/fca-2017-0004
15. Cai, R., Ge, F., Chen, Y.Q., Kou, C.H.: Regional observability for Hadamard-Caputo time fractional
distributed parameter systems. Appl. Math. Comput. 360, 190–202 (2019)
16. Cai, R., Ge, F., Chen, Y.Q., Kou, C.H.: Regional gradient controllability of ultra-slow diffusions
involving the Hadamard-Caputo time fractional derivative. Math. Control Relat. F. 10(1), 141–156
(2020)
17. Kilbas, A.A.: Hadamard-type fractional calculus. J. Korean Math. Soc. 38(6), 1191–1204 (2001)
18. Bardaro, C., Butzer, P.L., Mantellini, I.: The foundations of fractional calculus in the Mellin transform
setting with applications. J. Fourier Anal. Appl. 21, 961–1017 (2015)
19. Pooseh, S., Almeida, R., Torres, D.F.M.: Expansion formulas in terms of integer-order derivatives for
the Hadamard fractional integral and derivative. Numer. Func. Anal. Opt. 33(3), 301–319 (2012)
20. Jarad, F., Abdeljawad, T., Baleanu, D.: Caputo-type modification of the Hadamard fractional deriva-
tives. Adv. Differ. Equ. 142, 1–8 (2012)
21. Metzler, R., Klafter, J.: The random walk’s guide to anomalous diffusion: a fractional dynamics
approach. Phys. Rep. 339(1), 1–77 (2000)
22. Zeng, C., Chen, Y.Q.: Optimal random search, fractional dynamics and fractional calculus. Fract. Calc.
Appl. Anal. 17(2), 321–332 (2014). https://doi.org/10.2478/s13540-014-0171-7
23. Meerschaert, M., Scheffler, H.: Stochastic model for ultraslow diffusion. Stoch. Proc. Appl. 116(9),
1215–1235 (2006)
24. Kochubei, A.: Distributed order calculus and equations of ultraslow diffusion. J. Math. Anal. Appl.
240(1), 252–281 (2008)
25. Griggs, D.T.: Creep of rocks. J. Geol. 47, 225–251 (1939)
26. Lomnitz, C.: Application of the logarithmic creep law to stress wave attenuation in the solid earth. J.
Geophys. Res. 67(1), 365–368 (1962)
27. Li, C., Li, Z.: Stability and logarithmic decay of the solution to Hadamard-type fractional differential
equation. J. Nonlinear Sci. 31 (2021)

123
Stability analysis of Hadamard and Caputo-Hadamard fractional nonlinear 2445

28. Cai, M., Karniadakis, G.E., Li, C.: Fractional SEIR Model and Data-Driven Predictions of COVID-19
Dynamics of Omicron Variant. Chaos 32(7), 071101 (2022)
29. Li, Y., Chen, Y.Q., Podlubny, I.: Mittag-Leffler stability of fractional order nonlinear dynamic systems.
Automatica 45(8), 1965–1969 (2009)
30. Aguila-Camacho, N., Duarte-Mermoud, M., Gallegos, J.A.: Lyapunov functions for fractional order
systems. Commun. Nonlinear Sci. 19(9), 2951–2957 (2014)
31. Duarte-Mermoud, M., Aguila-Camacho, N., Gallegos, J.A., Castro-Linares, R.: Using general
quadratic Lyapunov functions to prove Lyapunov uniform stability for fractional order systems. Com-
mun. Nonlinear Sci. 22(1–3), 650–659 (2015)
32. Liu, S., Wu, X., Zhou, X.F., Jiang, W.: Asymptotically stability of Riemann-Liouville fractional non-
linear systems. Nonlinear Dynam. 86, 65–71 (2016)
33. Agarwal, R., Hristova, S., O’Regan, D.: Noninstantaneous impulses in Caputo fractional differential
equations and practical stability via Lyapunov functions. J. Franklin Inst. 354(7), 3097–3119 (2017)
34. Dhar, H., Kong, Q.: Lyapunov-type inequalities for α-th order fractional differential equations with
2 < α ≤ 3 and fractional boundary conditions. Electron. J. Differ. Eq. 203, 1–15 (2017)
35. Halanay, A.: Differential Equations: Stability, Oscillations, Time Lags. Academic Press, New York
(1966)
36. He, B.B., Zhou, H.C., Chen, Y.Q., Kou, C.H.: Asymptotical stability of fractional order systems with
time delay via an integral inequality. IET Control Theory A. 12(12), 1748–1754 (2018)
37. He, B.B., Zhou, H.C., Kou, C.H., Chen, Y.Q.: New integral inequalities and asymptotic stability of
fractional order systems with unbounded time delay. Nonlinear Dynam. 94, 1523–1534 (2018)
38. Wang, M., Jia, B.G., Du, F.F., Liu, X.: Asymptotic stability of fractional difference equations with
bounded time delays. Fract. Calc. Appl. Anal. 23(2), 571–590 (2020). https://doi.org/10.1515/fca-
2020-0028
39. Kassim, M., Tatar, N.: A neutral fractional Halanay inequality and application to a Cohen-Grossberg
neural network system. Math. Methods Appl. Sci. 44(13), 10460–10476 (2021)
40. Zhao, X., Deng, F.: Time-varying Halanay inequalities with application to stability and control of
delayed stochastic systems. IEEE T. Automat. Contr. 67(3), 1226–1240 (2022)
41. He, B.B., Zhou, H.C.: Caputo-Hadamard fractional Halanay inequality. Appl. Math. Lett. 125, 107723
(2022)
42. Li, C, Li, Z.: The blow-up and global existence of solution to Caputo-Hadamard fractional partial
differential equation with fractional Laplacian. J. Nonlinear Sci. 80 (2021)
43. Sastry, S.: Nonlinear Systems: Analysis, Stability and Control. Springer, New York (1999)
44. Simon, T.: Comparing Fréchet and positive stable laws. Electron. J. Probab. 19, 1–25 (2014)
45. Slotine, J.J.E., Li, W.: Applied Nonlinear Control. Prentice Hall, Englewood Cliffs, New Jersey (1991)

Publisher’s Note Springer Nature remains neutral with regard to jurisdictional claims in published maps
and institutional affiliations.

Springer Nature or its licensor (e.g. a society or other partner) holds exclusive rights to this article under
a publishing agreement with the author(s) or other rightsholder(s); author self-archiving of the accepted
manuscript version of this article is solely governed by the terms of such publishing agreement and applicable
law.

123