APPROXIMATELY BISECTRIX-ORTHOGONALITY PRESERVING

MAPPINGS

arXiv:1506.06218v1 [math.FA] 20 Jun 2015

ALI ZAMANI
Abstract. Regarding the geometry of a real normed space X , we mainly introduce
a notion of approximate bisectrix-orthogonality on vectors x, y X as follows:


1
1+


x W y if and only if 2
kxk kyk kykx + kxky 2
kxk kyk.
1+
1
We study class of linear mappings preserving the approximately bisectrix-orthogonality

W . In particular, we show that if T : X Y is an approximate linear similarity,

then
x W y = T x W T y
(x, y X )
for any [0, 1) and certain 0.

1. Introduction and preliminaries

There are several concepts of orthogonality appeared in the literature during the past
century such as BirkhoffJames, Phythagorean, isosceles, Singer, Roberts, Diminnie,
Carlsson, R
atz, -orthogonality, area orthogonality, etc, in an arbitrary real normed
space X , which can be regarded as generalizations of orthogonality in the inner product
spaces, in general [1, 2]. These are of intrinsic geometric interest and have been studied
by many mathematicians. Among them we recall the following ones:
(i) BirkhoffJames B : x B y if kxk kx + tyk for all scalars t (see [3]).
(ii) Phythagorean P : x P y if kx + yk2 = kxk2 + kyk2 (see [9]).
(iii) Isosceles I : x I y if kx + yk = kx yk (see [9, 14]).
(iv) Roberts R : x R y if kx + tyk = kx tyk for all scalars t (see [15]).
The following mapping h.|.ig : X X R was introduced by Milicic [12]:
1
hy|xig = (+ (x, y) + (x, y)),
2
where mappings + , : X X R are defined by
(x, y) = lim

t0

kx + tyk2 kxk2
.
2t

In addition the -orthogonality x y means hy|xig = 0.

Note that R , B [1] and the relations P , I , B are, however, independent. If
(H, h.|.i) is a real inner product space, then all above relations coincide with usual
orthogonality derived from h.|.i [2].
In the present note, we consider the so-called bisectrix-orthogonality and we study the
orthogonality preserving property of this kind of orthogonality.
2010 Mathematics Subject Classification. Primary 46B20; Secondary 46C50, 47B99.
Key words and phrases. Bisectrix-orthogonality, Approximate orthogonality, Isometry, Orthogonality preserving mapping.
1

A. ZAMANI

Let X be a real normed space and x, y X . The bisectrix-orthogonality relation

x W y (cf. Section 5.2 in [1]) is defined by




kykx + kxky = 2kxk kyk,
which for nonzero x and y means

x
y


kxk + kyk = 2.

For instance consider the space (R2 , k|.k|) where k|(r, s)k| = max{|r|, |s|} for (r, s) R2 .
Then (1, 0) W (r, s) if and only if either (r, s) is the zero vector or

(r, s) ({ 2 1, 2 1}) [ 2, 2] ([ 2 1, 2 1] { 2, 2}).

Now we recall some properties of bisectrix-orthogonality (the proofs can be found in

[1, Proposition 5.2.1]):
(P.1) x W 0, 0 W y for all x, y X ;
(P.2) x W y if and only if y W x;
(P.3) If x W y and x, y 6= 0, then x, y are linearly independent;
(P.4) If x W y and 0, then x W y;
(P.5) In an inner product space, x W y if and only if hx|yi = 0.
By the definition of bisectrix-orthogonality and Pythagorean orthogonality one can
easily get the following properties:
y
x
P kyk
;
(P.6) For all nonzero vectors x, y X , x W y if and only if kxk
(P.7) For all x, y SX = {z X : kzk = 1}, x W y if and only if x P y.
We state some relations between bisectrix-orthogonality and other orthogonalities. It
is known [1, 2] that each of the following properties implies that the norm k.k comes
from an inner product.
(P.8) W I over X ;
(P.9) I W over X ;
(P.10) W B over SX ;
(P.11) B W over SX ;
(P.12) If for all x, y X , x W y implies kx + yk2 + kx yk2 2kxk2 + 2kyk2
where stands either for or . An easy consequence of R , B states that if
for all x, y SX , the relation x W y implies x R y or x y, then the norm k.k
comes from an inner product.
In this paper we introduce two notions of approximate bisectrix-orthogonality W and
W in a real normed space X and study the class of linear mappings which preserve
the approximately bisectrix-orthogonality of type W . In particular, we show that if
T : X Y is an approximate linear similarity, then
x W y = T x W T y

for any [0, 1) and certain 0.

(x, y X )

2. Approximately bisectrix-orthogonality preserving mappings

Let , be elements of an inner product space (H, h.|.i) and [0, 1). The approximate orthogonality defined by
or equivalently,

|h|i| kk kk

| cos(, )| (, 6= 0).

APPROXIMATE BISECTRIX-ORTHOGONALITY

So, it is natural to generalize the notion of approximate orthogonality for a real normed
space X . This fact motivated Chmieli
nski and Wojcik [5] to give for two elements
x, y X the following definition of the approximate isosceles-orthogonality x I y as
follows:




 kx + yk kx yk  (kx + yk + kx yk).
They also introduced another approximate isosceles-orthogonality x I y by



2
2
kx
+
yk

kx

yk

 4kxkkyk.

Inspired by the above approximate isosceles-orthogonality, we propose two definitions

of approximate bisectrix-orthogonality.
Let [0, 1) and x, y X , let us put x W y if











 kykx + kxky 2kxk kyk kykx + kxky + 2kxk kyk
or equivalently,

which means

1+

1


2
kxk kyk kykx + kxky 2
kxk kyk,
1+
1

1
y
x
1+
2

+
2
1+
kxk kyk
1
for nonzero vectors x and y.
Also we define x W y if
2




2
2
 kykx + kxky 2kxk kyk  2kxk2 kyk2
or equivalently,
p

which means

p


2(1 )kxk kyk kykx + kxky 2(1 + )kxk kyk,
p

x
y
p

+
2(1 )
2(1 + )
kxk kyk

for nonzero x, y X .
It is easy to check that in the case where the norm comes from a real-valued inner
product we have
4
kxk kyk
x W y |hx, yi|
(1 )2
and
x W y |hx, yi| kxk kyk x y
Thus the second approximate bisectrix-orthogonality coincides with the natural notion
of approximate orthogonality for inner product spaces.
Note that the relations W and W are symmetric and almost homogeneous in the
sense that

and

x W y = y W x and x W y for 0
x W y = y W x and x W y for 0.

A. ZAMANI

1
Remark 2.1. It is easy to see that W implies W with the same . Also if [0, 16
),

1
1+
16
then W implies W . Indeed, for x, y 6= 0, since 0 < 16 , so 1 1 + 16 and

1 16 1
1+ . Hence x W y implies


p
p
x
y
2(1 + 16)

+
2(1 16)

kxk kyk

or equivalently, x 16
W y.

Now, suppose that X and Y are real normed spaces of dimensions greater than or
equal to two and let , [0, 1). We say that a linear mapping T : X Y preserves
the approximate bisectrix-orthogonality if
x W y = T x W T y

(x, y X ).

Notice that if = = 0, we have

x W y = T x W T y

(x, y X ),

and we say that T preserves the bisectrix-orthogonality.

Koldobsky [10] (for real spaces) and Blanco and Turnsek [4] (for real and complex ones)
proved that a linear mapping T : X Y preserving the Birkhoff orthogonality has to
be a similarity, i.e., a non-zero-scalar multiple of an isometry. Further, Chmieli
nski and
Wojcik [7, 16] proved that a linear mapping T : X Y preserving the -orthogonality
has to be a similarity. Approximately orthogonality preserving mappings in the framework of normed spaces have been recently studied. In the case where = 0, Mojskerc
nski [6] verified the properties of mappings that preand Turnsek [13] and Chmieli
serve approximate Birkhoff orthogonality. Also Chmieli
nski and Wojcik [5, 7] studied
some properties of mappings that preserve approximate isosceles-orthogonality and orthogonality in the case when = 0. Recently Zamani and Moslehian [17] studied
approximate Roberts orthogonality preserving mappings.
The next lemma plays an essential role in our work. It provides indeed a reverse of the
triangle inequality; see [8].
Lemma 2.2. [11, Theorem 1] Let X be a normed space and x, y X \ {0}. Then
x
y


+
kxk + kyk + (
2) max{kxk, kyk}
kxk kyk
kx + yk
x
y


+
kxk + kyk + (
2) min{kxk, kyk}
kxk kyk

To reach our main result, we need some lemmas, which are interesting on their own
right. We state some prerequisites for the first lemma. For a bounded linear mapping
T : X Y, let kT k = sup{kT xk; kxk = 1} denote the operator norm and [T ] :=
inf{kT xk; kxk = 1}. Notice that for any x X , we have [T ]kxk kT xk kT kkxk.
Lemma 2.3. Let , [0, 1). If a nonzero bounded linear mapping T : X Y satisfies

1+
1
kxk kT xk
kxk
1+
1
h
i
1+
for all x X and all 1
[T
],
kT
k
, then
1+
1
x W y = T x W T y

(x, y X ).

APPROXIMATE BISECTRIX-ORTHOGONALITY

Proof. Let x, y X \ {0} and x W y. Then

for some s R r {0}, then for =

1
1+ [T ]

1
2 1+

we have

x
y
1+
kxk + kyk 2 1
. If x = sy

 x
Tx
Ty
y 



+
+

= T

kT xk kT yk
kT xk kT yk
 sy

y


= T
+

ksT yk kT yk
 sy
kyk
y 


=
+

T
kT yk
ksyk kyk
 x
kyk
y 


=
+

T
kT yk
kxk kyk
kyk 1 +
y
x

+

kT yk 1 kxk kyk
y
kyk 1 + 1
x

[T ]
+
=

kT yk 1 1 +
kxk kyk
1 + 1 1 +

2
11+
1
1+
= 2
,
1



1
Tx
Tx
Ty
Ty
1+
whence kT xk + kT yk 2 1 . Similarly, 2 1+ kT xk + kT yk . Thus T x W T y.
Assume that x, y are linearly independent. Set 0 :=

that

kxk
kT xk

kyk
kT yk .

2
.

x
y
kT xk + kT yk

We may assume

By Lemma 2.2 we have

x
x
y
kyk
y
kxk
kxk kyk




+
+
+ (
+
,
}


2) min{
kT xk kT yk
kT xk kT yk
kxk kyk
kT xk kT yk
1+
kyk
kxk

+( 2
1)
kT yk
1
kT xk
1+
1
1
+( 2
1)

[T ]
1
[T ]
1+ 1
.
= 2
1 [T ]
So that 0

1+ 1
2 1
[T ]

Similarly we get 0
that

1
1+ [T ].

1+
1 kT k.

Thus 0

1+
1
1+ [T ], 1 kT k

. Our hypothesis implies

1+
1
0 kzk kT zk
0 kzk
1+
1

(z X )





kT zk 0 kzk (kT zk + 0 kzk)

(z X )

or equivalently,

A. ZAMANI

Putting kT ykx + kT xky instead of z in the above inequality we get




2




kT
ykx
+
kT
xky
 kT ykT x+kT xkT y


y
x
kT xk + kT yk



2




kT ykT x + kT xkT y +
kT
ykx
+
kT
xky

.
y
x
kT xk + kT yk

Thus











 kT ykT x + kT xkT y 2kT xk kT yk  kT ykT x + kT xkT y + 2kT xk kT yk ,
whence T x W T y.

Lemma 2.4. Let , [0, 1). If a nonzero bounded linear mapping T : X Y satisfies
1+
1+
1 kT zk kuk 1 kT uk kzk for all z, u X , then
x W y = T x W T y

(x, y X ).

1+
kT zk 1+
Proof. By our assumption we have, 1
1 kT uk for all z, u with kzk = kuk = 1.
Passing to the infimum over kuk = 1, we get
1+
1+
kT zk
[T ] (kzk = 1),
1
1
and passing to the supremum over kzk = 1 we obtain
1+
1+
kT k
[T ].
1
1
h
i
1+
Now, let 1
[T
],
kT
k
and x X . Therefore we have
1+
1

1 1+
1
kxk

kT k kxk
1+
1+ 1
1 1+ 1+ 1

[T ] kxk
1+ 1 1 1+
kT xk

Thus

kT k kxk
1+ 1

[T ] kxk

1 1+
1+ 1 1+

kxk
1 1+ 1
1+
=
kxk.
1

1
1+
kxk kT xk
kxk
1+
1
Making a use of Lemma 2.3 just completes the proof.

We are now in position to establish the main result. Following [13], we say that a
linear mapping U : X Y is an approximate linear isometry if
(1 1 ())kzk kU zk (1 + 2 ())kzk

(z X ),

where 1 () 0 and 2 () 0 as 0. Notice that if 1 () = 2 () = 0, then U is

an isometry.

APPROXIMATE BISECTRIX-ORTHOGONALITY

A linear mapping U : X Y is said to be an approximate similarity if it is a

non-zero-scalar multiple of an approximate linear isometry, or equivalently it satisfies
(1 1 ())kwk kU wk (1 + 2 ())kwk
for some unitary U , some > 0 and for all w X , where 1 () 0 and 2 () 0 as
0.
Theorem 2.5. Let U : X Y be an approximate linear similarity and [0, 1). If a
nonzero bounded linear mapping T : X Y satisfies kT U k kU k, then
x W y = T x W T y
where =

(x, y X ),

2+2+(1)1 ()+(1++2)2 ()
2+2(1)1 ()+(1++2)2 () .

Proof. For any w X we have





 kT wk kU wk  kT w U wk kT U k kwk kU k kwk (1 + 2 ())kwk,

whence

(1 + 2 ())kwk kT wk kU wk (1 + 2 ())kwk.
Since
(1 1 ())kwk kU wk (1 + 2 ())kwk,
therefore we get
h
i
(1 1 ()) (1 + 2 ()) kwk kT wk (1 + )(1 + 2 ())kwk.
Thus for any z, u X , we have
1+
kT uk
1+
i
kT zk kuk
(1 + )(1 + 2 ())kzk h
1
1
(1 1 ()) (1 + 2 ())
=
=
=
Therefore

1+
1 kT zk kuk

(1 + )(1 + 2 ())(1 + )
kT ukkzk
[(1 1 ()) (1 + 2 ())](1 )

1+

2+2+(1)1 ()+(1++2)2 ()
2+2(1)1 ()+(1++2)2 ()
kT ukkzk
2+2+(1)1 ()+(1++2)2 ()
2+2(1)1 ()+(1++2)2 ()

1+
kT ukkzk.
1

1+
1 kT uk kzk.

Now the assertion follows from Lemma 2.4. 

As a consequence, with = 0 and T = U , we have

Corollary 2.6. Let T : X Y be an approximate linear similarity. Then
x W y = T x W T y
for any [0, 1), where =

(x, y X )

2+(1)1 ()+(1+)2 ()
2(1)1 ()+(1+)2 () .

Acknowledgement. The author would like to thank his PhD supervisor, Prof. M.
S. Moslehian, for his useful comments improving the paper.

A. ZAMANI

References
1. C. Alsina, J. Sikorska and M. Santos Tom
as, Norm Derivatives and Characterizations of Inner
Product Spaces, World Scientific, Hackensack, NJ, 2009.
2. D. Amir, Characterization of Inner Product Spaces, Birkh
auser Verlag, Basel-Boston-Stuttgart,
1986.
3. G. Birkhoff, Orthogonality in linear metric spaces, Duke Math. J. 1 (1935), 169172.
4. A. Blanco and A. Turnsek, On maps that preserves orthogonality in normed spaces, Proc. Roy. Soc.
Edinburgh Sect. A 136 (2006), 709716.
5. J. Chmieli
nski and P. W
ojcik, Isosceles-orthogonality preserving property and its stability, Nonlinear
Anal. 72 (2010), 14451453.
6. J. Chmieli
nski, Remarks on orthogonality pereserving mappings in normed spaces and some stability
problems, Banach J. Math. Anal. Bold (2007), no. 1, 117124.
7. J. Chmieli
nski and P. W
ojcik, On a -orthogonality, Aequationes Math. 80 (2010), 4555.
8. F. Dadipour, M. S. Moslehian, J. M. Rassias and S.-E. Takahasi, Characterization of a generalized
triangle inequality in normed spaces, Nonlinear Anal-TMA 75 (2012), no. 2, 735741.
9. R. C. James, Orthogonality in normed linear spaces, Duke Math. J. 12 (1945), 291301.
10. D. Koldobsky, Operators preserving orthogonality are isometries, Proc. Roy. Soc. Edinburgh Sect.
A 123 (1993), 835837.
11. L. Maligranda, Simple norm inequalities, Amer. Math. Monthly 113 (2006), 256260.
12. P. M. Milicic, Sur la G-orthogonalite dans les espeaceesnormes, Math. Vesnik. 39 (1987), 325334.
13. B. Mojskerc and A. Turnsek, Mappings approximately preserving orthogonality in normed spaces,
Nonlinear Anal. 73 (2010), 38213831.
14. M. Mirzavaziri and M. S. Moslehian, Orthogonal constant mappings in isoceles orthogonal spaces,
Kragujevac J. Math. 29 (2006), 133140.
15. B. D. Roberts, On the geometry of abstract vector spaces, T
ohoku Math. J. 39 (1934), 4259.
16. P. W
ojcik, Linear mappings preserving -orthogonality, J. Math. Anal. Appl. 386 (2012), 171176.
17. A. Zamani and M.S. Moslehian, Approximate Roberts orthogonality, Aequat. Math. 89 (2015),
529-541.
Department of Pure Mathematics, Center of Excellence in Analysis on Algebraic
Structures (CEAAS), Ferdowsi University of Mashhad, P.O. Box 1159, Mashhad 91775,
Iran.
E-mail address: zamani.ali85@yahoo.com.