SOME GRÜSS TYPE INEQUALITIES FOR FRÉCHET

DIFFERENTIABLE MAPPINGS

T. TEIMOURI-AZADBAKHT1 , A. G. GHAZANFARI2∗
arXiv:1908.05326v1 [math.FA] 14 Aug 2019

Abstract. Let X be a Hilbert C ∗ -module on C ∗ -algebra A and p ∈ A. We

denote by Dp (A, X) the set of all continuous functions f : A → X, which are
Fréchet differentiable on a open neighborhood U of p. Then, we introduce
some generalized semi-inner products on Dp (A, X), and using them some
Grüss type inequalities in semi-inner product C ∗ -module Dp (A, X) and
Dp (A, X n ) are established.

1. Introduction
Let A, X be two normed vector spaces over K(K = C, R), we recall that a
function f : A → X is Fréchet differentiable in p ∈ A, if there exists a bounded
linear mapping u : A → X such that
kf (p + h) − f (p) − u(h)kX
lim = 0, (1.1)
h→0 khkA
and in this case, we denote u by Df (p). Let Dp (A, X) denotes the set of all
continuous functions f : A → X, which are Fréchet differentiable on a open
neighborhood (say U) of p. The main purpose of differential calculus consists
in getting some information using an affine approximation to a given nonlinear
map around a given point. In many applications it is important to have Fréchet
derivatives of f , since they provide genuine local linear approximation to f .
For instance let U be an open subset of A containing the segment [x, y] =
{(1 − θ)x + θy : 0 ≤ θ ≤ 1}, and let f : A → X be Fréchet differentiable on U,
then the following mean value formula holds
kf (x) − f (y) ≤ kx − yk sup kDf ((1 − θ)x + θy)k . (1.2)
0<θ<1

For two Lebesgue integrable functions f, g : [a, b] → R, consider the Čebys̆ev

functional:

b b b
1 1 1
Z Z Z
T (f, g) := f (t)g(t)dt − f (t)dt g(t)dt.
b−a a b−a a b−a a

2010 Mathematics Subject Classification. 26D10, 46C05, 46L08.

Key words and phrases. Fréchet differentiable mappings, C ∗ -modules, Grüss inequality.
*Corresponding author.
1
2 A. G. GHAZANFARI

In 1934, G. Grüss [4] showed that

1
|T (f, g)| ≤ (M − m)(N − n), (1.3)
4
provided m, M, n, N are real numbers with the property −∞ < m ≤ f ≤ M <
∞ and −∞ < n ≤ g ≤ N < ∞ a.e. on [a, b]. The constant 14 is best possible
in the sense that it cannot be replaced by a smaller quantity and is achieved
for
 a + b
f (x) = g(x) = sgn x − .
2
The discrete version of (1.3) states that: If a ≤ ai ≤ A, b ≤ bi ≤ B, (i =
1, ..., n) where a, A, b, B, ai , bi are real numbers, then
 
1 X n n n
1 X 1 X  1
ai bi − ai . bi  ≤ (A − a)(B − b), (1.4)

n n i=1 n i=1  4

i=1

where the constant 41 is the best possible for an arbitrary n ≥ 1. Some refine-
ments of the discrete version of Grüss inequality (1.4) for inner product spaces
are available in [1, 6].
Theorem 1. Let (H; h·, ·i) and K be as above and x = (x1 , ..., xn ) ∈ H n ,
α = (α1 , ..., αn ) ∈ Kn and p = (p1 , ..., pn ) a probability vector. If x, X ∈ H are
such that
Re hX − xi , xi − xi ≥ 0 f or all i ∈ {1, ..., n},
or, equivalently,

xi − x + X ≤ 1 kX − xk f or all i ∈ {1, ..., n},

2 2
holds, then the following inequality holds
 
X n Xn Xn 1 Xn  Xn 
pi αi xi − p i αi pi xi ≤ kX − xk pi αi − p j αj 
 
2

 
i=1 i=1 i=1 i=1 j=1

n
 n 2  12
1 X
2
 X 
≤ kX − xk pi |αi | −  p i αi   .
  
2 i=1

i=1

(1.5)
1
The constant 2
in the first and second inequalities is best possible.
In recent years several refinements and generalizations have been considered
for the Grüss inequality. We would like to refer the reader to [2–6, 8, 9] and
references therein for more information.
 SOME GRÜSS TYPE INEQUALITIES... 3

In this paper, for every Hilbert C ∗ -module X over a C ∗ -algebra A, some

Grüss type inequalities in semi-inner product C ∗ -module Dp (A, X n ) are estab-
lished. We also for two arbitrary Banach ∗-algebras, define a norm and an
involution map on Dp (A, B) and prove that Dp (A, B) is a Banach ∗-algebra.

2. Grüss type inequalities for differentiable mappings

Let A be a C ∗ -algebra. A semi-inner product module over A is a right
module X over A together with a generalized semi-inner product, that is with
a mapping h., .i on X × X, which is A-valued and has the following properties:
(i) hx, y + zi = hx, yi + hx, zi for all x, y, z ∈ X,
(ii) hx, yai = hx, yi a for x, y ∈ X, a ∈ A,
(iii) hx, yi∗ = hy, xi for all x, y ∈ X,
(iv) hx, xi ≥ 0 for x ∈ X.
We will say that X is a semi-inner product C ∗ -module. If, in addition,
(v) hx, xi = 0 implies x = 0,
then h., .i is called a generalized inner product and X is called an inner product
module over A or an inner product C ∗ -module. An inner product C ∗ -module
1
which is complete with respect to its norm kxk = khx, xik 2 , is called a Hilbert
C ∗ -module.
As we can see, an inner product module obeys the same axioms as an or-
dinary inner product space, except that the inner product takes values in a
more general structure rather than in the field of complex numbers.
If A is a C ∗ -algebra and X is a semi-inner product A-module, then the
following Schwarz inequality holds:
hx, yihy, xi ≤ khx, xikhy, yi (x, y ∈ X) (2.1)
(e.g. [7, Proposition 1.1]).
Theorem 2. [3] Let A be a C ∗ - Algebra, X a Hilbert C ∗ - module. If x, y, e ∈
X, he, ei is an idempotent in A and α, β, λ, µ are complex numbers such that


x − α + β 1
≤ |α − β|,

y − λ + µ 1
≤ |λ − µ|,
e e
2 2 2 2
hold, then one has the following inequality;
1
khx, yi − hx, eihe, yik ≤ | α − β||λ − µ | .
4
Example 1. Let A be a real C -algebra and X be a semi-inner product C ∗ -
∗

module on a C ∗ -algebra B. If functions f, g ∈ Dp (A, X), then function k :

A → B as k(a) = hf (a), g(a)i is differentiable in (p ∈ A) and derivative of
that is a linear mapping Dk(p) : A → B defined by
Dk(p)(a) = hDf (p)(a), g(p)i + hf (p), Dg(p)(a)i.
4 A. G. GHAZANFARI

Because
hf (p + h), g(p + h)i − hf (p), g(p)i − hDf (p)(h), g(p)i − hf (p), Dg(p)(h)i
= hf (p + h), g(p + h) − g(p) − Dg(p)(h))i + hf (p + h) − f (p), Dg(p)(h)i
+ hf (p + h) − f (p) − Df (p)(h), g(p)i.
Let A be a C ∗ -algebra and X a semi-inner product A-module. If f ∈
Dp (A, X) and a ∈ A, we define the function fa : A → X by fa (t) = f (t)a.
Theorem 3. Let X be a semi-inner product C ∗ -module on C ∗ -algebra A, and
p ∈ A, e ∈ X. If he, ei is an idempotent element in A, and f, g ∈ Dp (A, X),
then for every a ∈ A, the map [·, ·]a : Dp (A, X) × Dp (A, X) → A with;




[f, g]a := Df (p)(a), Dg(p)(a) 1 + f (p), g(p) 1 − D f (·), g(·) 1 (p)(a),

is a generalized semi-inner product on Dp (A, X), where

hf (a), g(a)i1 = hf (a), g(a)i − hf (a), eihe, g(a)i.
Proof. First, we show that fa ∈ Dp (A, X) and Dfa (p) = (Df (p))a. There
exists a bounded convex set V (= B(p, r)) containing p such that V ⊆ U. Let
p, h ∈ V, a ∈ A, then
kfa (p + h) − fa (p) − (Df (p)(h))ak = k[f (p + h) − f (p) − Df (p)(h)]ak
≤ kf (p + h) − f (p) − Df (p)(h)kkak.

This implies that fa ∈ Dp (A, X).

A simple calculation shows
D
[f, g]a = Df (p)(a) − f (p), Dg(p)(a) − g(p)i
E
− hDf (p)(a) − f (p), eihe, Dg(p)(a) − g(p)
D

= (Df (p)(a) − f (p)) − e e, (Df (p)(a) − f (p))

E
, (Dg(p)(a) − g(p)) − e e, (Dg(p)(a) − g(p)) .

Therefore,
D

[f, f ]a = (Df (p)(a) − f (p)) − e e, (Df (p)(a) − f (p))

E
, (Df (p)(a) − f (p)) − e e, (Df (p)(a) − f (p)) ≥ 0.

It is easy to show that [·, ·]a is a generalized semi-inner product on Dp (A, X).

 SOME GRÜSS TYPE INEQUALITIES... 5

Lemma 1. Let X be a semi-inner product C ∗ -module on C ∗ -algebra A, and

p, a ∈ A, e ∈ X. If he, ei is an idempotent element in A, f, g ∈ Dp (A, X), and
α, β, α′, β ′, µ, λ, µ′, λ′ are complex numbers such that

f (p) − α + β e ≤ 1 |α − β|

2 2

Df (p)(a) − α + β e ≤ 1 |α′ − β ′ |
′ ′

2 2

g(p) − λ + µ e ≤ 1 |λ − µ|

2 2

µ ′
+ λ ′
1
Dg(p)(a) − e ≤ |µ′ − λ′ |,
2 2
then the following inequality holds


khDf (p)(a), Dg(p)(a)i1 + hf (p), g(p)i1 − D f (·), g(·) 1 (p)(a)k
1
≤ (|α − β| + |α′ − β ′ |)(|λ − µ| + |λ′ − µ′ |).
2

Proof. Since [·, ·]a is a generalized semi-inner product on Dp (A, X), the Schwartz
inequality holds, i.e,

k[f, g]a k2 ≤ k[f, f ]a k k[g, g]ak.

We know that

k[f, f ]a k ≤ hDf (p)(a), Df (p)(a)i − hDf (p)(a), eihe, Df (p)(a)i

+ hf (p), f (p)i − hf (p), eihe, f (p)i

+ hDf (p)(a), f (p)i − hDf (p)(a), eihe, f (p)i

+ hf (p), Df (p)(a)i − hf (p), eihe, Df (p)(a)i .
This inequality and Theorem 2 imply that
1 1 1
k[f, f ]a k ≤ |α′ − β ′ |2 + |α − β|2 + |α′ − β ′ ||α − β|
4 4 2
1
= (|α − β| + |α′ − β ′ |)2 .
4
6 A. G. GHAZANFARI

Similarly
1 1 1
k[g, g]ak ≤ |λ′ − µ′ |2 + |λ − µ|2 + |λ′ − µ′ ||λ − µ|
4 4 2
1
= (|λ − µ| + |λ′ − µ′ |)2 .
4


Let X be a semi-inner product C ∗ -module over C ∗ -algebra A. For every

x ∈ X, we define the map x̂ : A → X n by x̂(a) = (xa, ..., xa), (a ∈ A).
Lemma 2. Let X be a semi-inner product C ∗ -module, x0 , y0 , x1 , y1 ∈ X and
(r1 , r2 , ...., rn ) ∈ Rn a probability vector. If p ∈ A and f = (f1 , ..., fn ), g =
(g1 , ..., gn ) ∈ Dp (A, X n ) such that

\

x0 + y0
x0 − y 0
Df (p) − ≤ ,

2 2

and
x\ x − y
1 + y1 1 1
Dg(p) − ≤ ,

2 2
then for all a ∈ A, we have

Xn n
DX n
X E
ri hDfi (p)(a), Dgi (p)(a)i − ri Dfi (p)(a), ri Dgi (p)(a)



i=1 i=1 i=1
1
≤ kx0 − y0 kkx1 − y1 kkak2 . (2.2)
4


Proof. For every a ∈ A, we define the map ·, · a : Dp (A, X n )×Dp (A, X n ) → A
with;
n
X D n
E DX n
X E


f, g a
= ri Dfi (p)(a), Dgi (p)(a) − ri Dfi (p)(a), ri Dgi (p)(a) .
i=1 i=1 i=1

The following Korkine type inequality for differentiable mappings holds:

n

1 X D E
f, g a
= ri rj Dfi (p)(a) − Dfj (p)(a), Dgi (p)(a) − Dgj (p)(a) ,
2 i=1,j=1



Therefore, f, f a ≥ 0. It is easy to show that ·, · a is a generalized semi-inner
product on Dp (A, X n ).
 SOME GRÜSS TYPE INEQUALITIES... 7

A simple calculation shows that

n

X D x0 + y0 x1 + y1 E
f, g a
= ri Dfi (p)(a) − a, Dgi (p)(a) − a
i=1
2 2
n n
DX x0 + y0 X x1 + y1 E
− ri Dfi (p)(a) − a, ri Dgi (p)(a) − a .
i=1
2 i=1
2

From Schwartz inequality, we have

n 2 n 2

2
X x0 + y 0 X x1 + y 1
k f, g a k ≤ ri Dfi (p)(a) − a r i
Dgi (p)(a) − a
i=1
2
i=1
2
2 2
x \
0 + y
0 x\
1 + y 1

≤ Df (p) − Dg(p) − kak4

2 2
1
≤ kx0 − y0 k2 kx1 − y1 k2 kak4
16

Corollary 1. Let X be a semi-inner product C ∗ -module, x0 , y0 ∈ X, (α1 , ..., αn ) ∈

Cn and (r1 , r2 , ...., rn ) ∈ Rn a probability vector. If p ∈ A and f = (f1 , ..., fn ) ∈
Dp (A, X n ) such that

x\
0 + y x − y
0 0 0
Df (p) − ≤ ,

2 2

then for all a ∈ A, we have

n n n

X X X
ri αi Dfi (p)(a) − r i αi ri Dfi (p)(a)



i=1 i=1 i=1


n  n 2  12
x0 − y0 X
2
X 
≤ kak r |α | − r α . (2.3)
 
i i i i

2 
 
i=1 i=1
8 A. G. GHAZANFARI

Proof.

X n X n Xn
ri αi Dfi (p)(a) − r i αi ri Dfi (p)(a)



i=1 i=1
  i=1
n n
X  X  x0 + y0
= r i αi − rj αj  Dfi (p)(a) − .a
 

i=1 j=1
 2
n
 n

X  X  \
x0 + y0
≤ ri αi − rj αj  Df (p) − kak
 
i=1

j=1
 2

n  2  12
n
x0 − y0 X 2
X 
≤ kak r |α | − r α .
 
i i i i

2 



i=1 i=1


Corollary 2. Let X be a semi-inner product C ∗ -module, x0 , y0 ∈ X. If p ∈ A
and f = (f1 , ..., fn ) ∈ Dp (A, X n ) such that

x\ x − y
0 + y0 0 0
Df (p) − ≤ ,

2 2
then for all a ∈ A, we have
kakkx − y kn  (n − 1)(n + 1)  12

n n
X n + 1 X 0 0
kDfk (p)(a) − . Dfk (p)(a) ≤ ,

2 4 3


k=1 k=1
(2.4)
and

n n
X
2 (n + 1)(2n + 1) X
k Dfk (p)(a) − . Dfk (p)(a)

6


k=1 k=1
kakkx0 − y0 kn p
≤ √ (n − 1)(n + 1)(2n + 1)(8n + 11). (2.5)
12 5
Proof. If we put ri = n1 , αi = k in inequality (2.3), then we get (2.4), and if
ri = n1 , αi = k 2 in inequality (2.3), then we get (2.5). 

3. Differential mapping on Banach *-algebras

Theorem 4. Let A, B be two Banach ∗-algebras and p be a self adjoint element
in A. Then Dp (A, B) is a Banach ∗-algebra with the point-wise operations and
the involution f ∗ (a) = (f (a∗ ))∗ (a ∈ A), and the norm
 
kf k := max sup kDf (x)k, sup kf (a)k < ∞. (3.1)
x∈U a∈A
 SOME GRÜSS TYPE INEQUALITIES... 9

Proof. First we show that the involution f → f ∗ is differentiable and Df ∗ (x)(h) =

(Df (x∗ )(h∗ ))∗ (x, h ∈ U = B(p, r)). It is trivial that if x ∈ U, then x∗ ∈ U
because kx − pk = k(x − p)∗ k = kx∗ − pk. It can be shown easily that Df ∗ (x)
is a bounded linear map with kDf ∗(x)k = kDf (x∗ )k. Therefore

∗
kf ∗ (x + h) − f ∗ (x) − Df ∗ (x)(h)k = k f (x∗ + h∗ ) − f (x∗ ) − Df (x∗ )(h∗ ) k
= kf (x∗ + h∗ ) − f (x∗ ) − Df (x∗ )(h∗ )k
≤ ǫkh∗ k = ǫkhk.
From kDf ∗ (x)k = kDf (x∗ )k and kf ∗ (a)k = kf (a∗ )k, we obtain
 
∗ ∗ ∗
kf k = max sup kDf (x)k, sup kf (a)k
x∈U a∈A
 
∗ ∗
= max sup kDf (x )k, sup kf (a )k
x∈U a∈A
 
∗ ∗
= max sup kDf (x )k, sup kf (a )k = kf k.
x∗ ∈U a∗ ∈A

Now, we show that Dp (A, B) is complete. There exists a bounded convex set
V (= B(p, r)) containing p such that V ⊆ U. Suppose that (fn ) is a Cauchy
sequence in Dp (A, B), i.e.,
kfn (a) − fm (a)k → 0 (a ∈ A), and kDfn (x) − Dfm (x)k → 0 (x ∈ V ).
Since B is complete, therefore L(A, B) the space of all bounded linear
maps from A into B, is complete. So there are functions f, g such that
supa∈A kfn (a) − f (a)k → 0 and supx∈V kDfn (x) − g(x)k → 0. Given ε > 0, we
can find N ∈ N such that for m > n ≥ N one has
ε
kDfm − Dfn k∞ = sup kDfm (x) − Dfn (x)k < (3.2)
x∈V 3
ε
kg − Dfn k∞ = sup kg(x) − Dfn (x)k < . (3.3)
x∈V 3
We may suppose that there exist a ∈ A such that, p + a ∈ V . Using
Lipschitzian functions fm − fn , we obtain that

kfm (p + a) − fm (p) − (fn (p + a) − fn (p))k

ε
≤ sup kDfm (p + θa) − Dfn (p + θa)kkak ≤ kak.
0<θ<1 3
passing to the limit on m, we get
ε
kf (p + a) − f (p) − (fn (p + a) − fn (p))k ≤ kak. (3.4)
3
10 A. G. GHAZANFARI

Utilizing differentiability fN and (3.3), we have

kfN (p + a) − fN (p) − g(p)(a)k ≤ kfN (p + a) − fN (p) − DfN (p)(a)k
ε ε
+ kDfN (p)(a) − g(p)(a)k ≤ kak + kak.
3 3
(3.5)
From (3.4) and (3.5), we obtain
kf (p + a) − f (p) − g(p)(a)k ≤ εkak.
Therefore Dp (A, B) is a Banach ∗-algebra. 
References
[1] S. S. Dragomir, Advances in Inequalities of the Schwarz, Grüss and Bessel Type in
Inner Product Spaces, Nova Science puplishers Inc., New York, 2005.
[2] S. S. Dragomir, A Grüss type discrete inequality in inner product spaces and applica-
tions, J. Math. Anal. Appl., 250 (2000), 494-511.
[3] A.G. Ghazanfari, S.S. Dragomir, Bessel and Grüss type inequalities in inner product
modules over Banach ∗-algebra, Linear Algebra Appl. 434 (2011), 944-956.
1
Rb
[4] G. Grüss, Über das Maximum des absoluten Betrages von b−a a
f (x)g(x)dx −
1
Rb Rb
(b−a)2 a f (x)dx a g(x)dx, Math. Z. 39(1934), 215-226.
[5] D. Ilišević and S. Varošanec, Grüss type inequalities in inner product modules, Proc.
Amer. Math. Soc. 133 (2005), 3271-3280.
[6] A. I. Kechriniotis and K. K. Delibasis, On generalizations of Grüss inequality in inner
Product Spaces and applications, J. Inequal. Appl. Vol(2010), Article ID 167091.
[7] E.C. Lance, Hilbert C ∗ -Modules, London Math. Soc. Lecture Note Series 210, Cam-
bridge Univ. Press, 1995.
[8] X. Li, R. N. Mohapatra and R. S. Rodriguez, Grüss-type inequalities, J. Math. Anal.
Appl. 267 (2002), no. 2, 434-443.
[9] D. S. Mitrinović, J. E. Pečarić, and A. M. Fink, Classical and New Inequalities in
Analysis, Kluwer Academic, Dordrecht, 1993.
1,2
Department of Mathematics, Lorestan University, P.O. Box 465, Kho-
ramabad, Iran.
E-mail address: 1 t.azadbakhat88@gmail.com, 2 ghazanfari.a@lu.ac.ir