Documenti di Didattica
Documenti di Professioni
Documenti di Cultura
Preprint
AND
M UNEO C H O
(Communicated by M. S. Moslehian) Abstract. We show that if x1 , , xn are vectors in a normed linear space (X , ) and s1 , , sn belong to the interval [0, ) , then
n n
fn (s1 , , sn ) =
j =1
sjxj
s jxj
j =1
PR
where (X , ) is a normed linear space over the real or complex eld K and x j , j {1, 2, , n} are vectors in X plays a fundamental role in establishing various analytic and geometric properties of such spaces. This inequality has been studied by several authors (see e.g. [2, 9, 15]). We are interested to know under which conditions the generalized triangle inequality on X is reversed, i.e., inequalities of the following type
n n
with C 0 , which we call (additive) reverse of the triangle inequality. Kato, Saito and Tamura [8] proved the following reverse of the triangle inequality.
Mathematics subject classication (2010): 46B20, 46B99, 26D15. Keywords and phrases: triangle inequality, inner product space, strictly convex Banach space.
c , Zagreb Paper MIA-3163
EP R
xj
xj ,
j =1 j =1
j =1
xj
xj
j =1
IN
+C 1
is a non-negative valued continuous function such that fn (s1 , , sn ) fn (t1 , , tn ) for all s1 , , sn and t1 , , tn in [0, ) with s j t j (1 j n) . By using it, we prove several versions of reverse triangle inequality in inner product spaces and discuss equality attainedness of norm inequalities in strictly convex Banach spaces.
T HEOREM 1.1. ([8, Theorem 1]) Let (X , ) be a Banach space, and x j X \ {0}, j {1, , n} . Then
n n n
j =1
xj
xj
j =1
j =1
xj xj
1 j n
max x j .
(1.1)
Moreover they proved equality attainedness of (1.1) in a strictly convex Banach space. T HEOREM 1.2. ([8, Theorem 3]) Let (X , ) be a strictly convex Banach space, and x j X \{0}, j {1, , n} . Let x j0 = min1 j n x j and x j1 = max1 j n x j . Then the equality
n n n
j =1
xj
xj
j =1
j =1
xj xj
1 j n
max x j
T HEOREM 1.3. [3, Theorem 7] Let (H , , ) be an inner product space over the real or complex number eld K and e, x j H , j {1, , n} with e = 1 . If k j 0, j {1, , n} , are such that x j Re e, x j
n
PR
EP R
n n
After that, several authors improved and generalized these inequalities (cf. [5, 6, 11, 12, 13, 14]). On the other hand, Dragomir [3] established the reverse of triangle inequality in real or complex inner product spaces as follows.
j =1
xj
xj
j =1 n
IN
j =1 j =1
xj =
xj
k j.
j =1
T
x j0 , x j0 (1.2)
j =1
xj
kj
j =1 n
and
n n
xj =
j =1
j =1
x j k j e.
j =1
Ansari and Moslehian modied the above result in [1], and related results can also be found in [4, 7]. The aim of this paper is to discuss equality attainedness of norm inequalities in a strictly convex Banach space. Moreover, we investigate the relation between the inequalities (1.1) and (1.2) in inner product spaces by using a continuous monotone function. 2. Equality attainedness in strictly convex Banach spaces Let (X , ) be a normed linear space over the real or complex number eld K , and x j X , j {1, 2, , n} . Throughout this paper, for each x j X , j {1, , n} , denote by fn a function on n j =1 [0, ) as
n n n
fn (t1 , , tn ) =
j =1
t jx j
t jx j
j =1
(t1 , , tn ) [0, ) .
j =1
| fn (t1 , , tn ) fn (s1 , , sn )|
n n
EP R
n
Proof. For any (s1 , , sn ), (t1 , , tn ) n j =1 [0, ) , applying the triangle inequality, we have
= =
j =1
t jx j
t jx j
j =1
j =1
t jx j s jx j
j =1
PR
j =1 n
(t j s j )
tj sj
xj
+
n
t jx j
j =1 j =1 n j =1
xj + xj + xj .
j =1
t jx j s jx j
j =1
j =1 n
tj sj
(t j s j )x j
2 tj sj
j =1
IN
n
T HEOREM 2.1. Let (X , ) be a normed linear space, and x1 , , xn X . Then fn is a non-negative valued continuous function on n j =1 [0, ) such that f n (s1 , , sn ) [ 0 , ) with (s1 , , sn ) (t1 , , tn ) . fn (t1 , , tn ) for all (s1 , , sn ), (t1 , , tn ) n j =1
j =1
s jx j
t jx j
n j =1
j =1
s jx j
T
s jx j
j =1 n
(t1 , , tn ) if s j
t j for
s jx j
j =1
t jx j
j =1
j =1 n
(t j s j )x j + s j x j
n
j =1 n
(t j s j )x j +
(t j s j )x j +
s jx j
j =1 n
j =1 n
s jx j
j =1 n
= = =
j =1 n
(t j s j ) x j +
n j =1 n
s jx j
j =1 n
tj
j =1 n
xj sj xj +
s jx j
j =1 n
j =1
t jx j s jx j +
j =1
s jx j
j =1
min1
xj
x1
EP R
j n
This is a generalization of [8, Theorem 1]. Indeed if x j = 0 for all j {1, , n} , then we see that xj xj 1 max1
j n
xj
,,
min1
j n
xj
xn
fn (1, , 1) max1
IN
xj
j n
( j {1, , n}).
T
. xj ,, max1
j n
fn
xj
Since
PR
x1
xn
fn (1, , 1) = min1
j n
j =1
xj =
xj
j =1 n
, xj xj
fn
min1
j n
xj
x1
,,
xj
xn
j =1 n
1 j n
min x j
and fn max1
j n
xj
x1
,,
max1
j n
xj
xn
j =1
xj xj
1 j n
max x j
we obtain [8, Theorem 1]. Next, we discuss when the equality fn (s1 , , sn ) = fn (t1 , , tn ) holds in Theorem 2.1 in a strictly convex Banach space. To do it, we need the following lemma. L EMMA 2.2. [8, Lemma 1] Let (X , ) be a strictly convex Banach space. For each x j X \ {0}, j {1, 2, , n} , the following assertions are equivalent:
n n
(i)
jx j
j =1 n
j
j =1 n
(ii) (iii)
jx j
j =1
j
j =1
xj xi = xi xj
and
n n
sjx j =
j =1
t jx j
j =1
EP R
j =1 j =1 j =1 n
s jx j
t jx j
IN
n
T HEOREM 2.3. Let (X , ) be a strictly convex Banach space and (s1 , , sn ) and (t1 , , tn ) n t j for all j {1, , n} . Put J = { j j =1 [0, ) satisfy s j {1, , n} : s j < t j } . Then the following assertions are equivalent: (i) the equality fn (s1 , , sn ) = fn (t1 , , tn ) (2.1)
t jx j
(2.2)
(t j s j )x j
j =1
(2.3)
(iii) either J = or
and
PR
xj xi = xi xj
n j =1
for all i, j J xi xi
(2.4)
sjx j = sjx j
j =1
for some i J.
(2.5)
Proof. (i) (ii) We may assume that J = . If (2.1) holds, then it is clear that (2.2) is valid and we need only to prove (2.3). If n j =1 s j x j = 0 , then, by (2.1) and (2.2), we have
n n n
0=
s jx j
j =1
t jx j
j =1
(t j s j )x j
j =1
6 and
n
s jx j = 0 =
j =1
t jx j
j =1
(t j s j )x j
j =1
xi xi
(i J ).
Next, we consider the case n j =1 s j x j = 0 . As in the proof of Theorem 2.1, we see that
n n n j =1
t jx j
j =1
(t j s j )x j + s j x j
j J n
(t j s j )x j +
j J
s jx j
j =1 n
j J n
(t j s j )x j + (t j s j )x j +
n
s jx j
j =1
= =
j =1 n
s jx j
j =1
j =1
t jx j s jx j +
j =1
IN
s jx j
.
j =1 n
T
(2.6)
(t j s j )x j + s j x j
j J j =1
EP R
n
j J
(t j s j )x j +
sjx j
j =1
PR
j =1
sjx j = sjx j
=
j =1
xi xi
n j =1
t j x j (t j s j )x j
xi . xi
(ii) (i) It is clear that if J = , then (2.1) holds. Thus we assume that J = . In this case, if (2.2) and (2.3) hold, then we have
n n n
s jx j
j =1
=
n
t jx j
j =1
(t j s j )x j
j =1 n
xi xi
= =
t jx j
j =1 n
(t j s j )x j
j =1 n j =1 n j =1
t jx j
j =1
t jx j + s jx j
j J
j J
Hence, if n j =1 s j x j = 0 , then
n n
j =1
EP R
j =1 n j =1
t jx j
(t j s j )x j + s j x j
=
j J n
(t j s j )x j
j =1
(t j s j )x j =
n
Thus we have (2.1). On the other hand, if n j =1 s j x j = 0 , then, by (2.4) and (2.5), we see that, for each i, j J , n (t j s j )x j xj j =1 si xi = . = n x ( t j s j )x j j j=1 si xi Applying Lemma 2.2, we have
n n
PR
j J
j =1
t jx j s jx j +
j =1
(t j s j )x j + s j x j
j =1
j J
IN
n
(t j s j )x j
(t j s j )x j .
j J n
(t j s j )x j
n j =1
j =1
t jx j s jx j
n
s jx j
j =1
(t j s j )x j +
T
.
and the equality (2.1) holds. (iii) (i) Assume that (2.4) and (2.5) hold. Then, by (2.4), we see that, for each i, j J , (t j s j )x j (ti si )xi = . (ti si )xi (t j s j )x j
sjx j
j =1
8 and
n
t jx j
j =1
j =1
t jx j s jx j +
j =1
s jx j
j =1
(i) (iii) If (2.1) holds, then all the equalities in (2.6) hold. Thus, by Lemma 2.2, we have (2.4) and (2.5). This completes the proof. Recall that, by Theorem 2.1, if (t1 , , tn ) belongs to n j =1 [1, ) , then fn (1, , 1) fn (t1 , , tn ),
j =1
xj
xj
j =1
j =1
t jx j
t jx j
j =1
and
n j =1
xj =
EP R
j =1 j =1 j =1 n n
xj
t jx j
t jx j
j =1
(t j 1)x j
j =1
and
PR
n j =1
xj xi = xi xj
n j =1
for all i, j I
xj = xj
xi xi
xj
xj
IN
n
C OROLLARY 2.4. Let (X , ) be a strictly convex Banach space and (t1 , , tn ) n j =1 [1, ) . Put I = { j {1, , n} : 1 < t j } . Then the following assertions are equivalent: (i) the equality fn (1, , 1) = fn (t1 , , tn )
t jx j
xi xi
for some i I .
( j {1, , n}),
T
for some i I ;
C OROLLARY 2.5. (cf. [8, Theorem 3]) Let (X , ) be a strictly convex Banach space and x1 , , xn belong to X \ {0} . If we put I = { j {1, , n} : x j < max1 j n x j } , then the following assertions are equivalent: (i) the equality
n n n
j =1
xj
xj
j =1
j =1
xj xj
1 j n
max x j
j =1
xj
j =1
xj xj
1 j n
max x j
and
n n
xj =
j =1
j =1
xj xj
n 1 j n
max x j + x j
j =1
and
Throughout this section, let (H , , ) be an inner product space over K . Note that all results in 2 are valid for (H , , ) . The following result makes it possible to connect Kato, Saito and Tamuras inequality to Dragomirs one. L EMMA 3.1. Let x j H \ {0}, j {1, , n} and (s1 , , sn ) n j =1 [0, ) . For each i {1, , n} ,
n t
PR
i
EP R
n n
for all i, j I
xj = xj
j =1
j =1
xi xi
lim fn (s1 , , t , , sn ) =
j =1
s j x j Re
IN
for some i I . xi , s jx j xi .
T
xi xi
for some i I ;
10
Proof. Take t > 0 and put, for i {1, , n} , Xi = j=i s j x j . Then we see that
n
fn (s1 , , t , , sn ) =
i n
j =i j =i n
s j x j + txi
s j x j + txi
j =i
= s j x j + ( txi Xi + txi ) = s jx j +
j =i
= s jx j
j =i
s j x j
j =i
(as t ).
Thus
t
lim fn (s1 , , t , , sn ) = s j x j
i j =i
Re xi , Xi xi xi , Xi xi
= s j x j Re
j =i
EP R
j =i
= s j x j Re
= s j x j Re
j =i n
j =1 n
s j x j Re
j =1
s j x j Re
T HEOREM 3.2. Let x j H \ {0}, j {1, , n} and (s1 , , sn ) n j =1 [1, ) . Then, for each i {1, , n} ,
n n n
PR
xj
xj
j =1
j =1
j =1
s j x j Re
IN
xi , xi
n
xi , (Xi + si xi ) si xi xi
s jx j
j =1 n
xi , xi
s jx j
j =1
xi , s jx j xi
xi , sjx j xi
T
si xi . . (3.1) (3.2)
j =1
xj
j =1
s j x j Re
xi , sjx j xi
11
and
n j =1
xj =
j =1
Re
xi , s j x j (s j 1)x j xi
xi . xi
(3.3)
Proof. Since 1
fn (1, , 1)
lim fn (s1 , , t , , sn ).
i
Thus, by Lemma 3.1, we deduce the desired inequality (3.1). Assume that (3.2) and (3.3) hold. Since
n
j =1 n
Re
xi , s j x j (s j 1)x j xi Re xi , s j x j (s j 1) x j xi xi , s jx j s jx j xi s j x j Re
= = =
j =1 n
j =1 n
x j + Re
n
j =1
0, we have
n n
j =1
PR
fn (1, , 1)
and the equality in (3.1) holds. Conversely, if the equality holds in (3.1), then it is clear that (3.2) is valid and we need only to prove (3.3). Since fn has monotonicity property, we see that
t
EP R
Re
j =1 n
xj =
=
xi , s j x j (s j 1)x j xi
Re
j =1 n
xi , s j x j (s j 1)x j xi
n
j =1
xj
j =1
s j x j Re
lim fn (1, , t , , 1)
i
IN
xi , s jx j xi
t i
xj
j =1
xi , s jx j xi
lim fn (s1 , , t , , sn ).
T
xi xi (3.4) (3.5) . (3.6)
If the equality holds in (3.1), then both equalities in (3.4) hold, that is,
n n n
j =1
xj
xj =
j =1
j =1 n
x j Re s j x j Re
xi , xj xi xi , sjx j xi
j =1
12
xj
j =1
Re
j =1
xi , xj xi
= Re
xi , xi
xj
j =1
(3.7)
We note that in Schwarzs type inequality Re u, v u v (u, v H ) the case of equality holds iff there exists a 0 such that u = v . Consequently, the equality holds in (3.7) iff there exists a 0 such that
n
xj =
j =1
xi . xi
xi xi
xj
j =1
= Re
xi , xi
xj
j =1
xj
j =1
= = =
j =1 n
xj Re Re
j =1
s j x j Re
j =1 n
j =1
Therefore we have
n
EP R
n
xi , s j x j (s j 1)x j xi
IN
xi . xi kij ,
xi , sjx j sj x j + x j xi
xj =
j =1
Re
j =1
xi , s j x j (s j 1)x j xi
PR
Note that Theorem 3.2 is a generalization of [3, Theorem 7]. Indeed, we may state that if, for each xed i {1, , n} , kij 0, j {1, , n} , satises the condition s j x j Re xi , s jx j xi
j =1
xj
xj
j =1 n
kij .
j =1
T
xi , s jx j xi . (3.8)
j =1
xj
kij
j =1
13
and
n n
xj =
j =1
j =1
x j kij
j =1
xi . xi
As a corollary, we obtain the following result similar to [3, Theorem 7]. C OROLLARY 3.3. Let x j H \{0}, j {1, , n} . If, for each xed i {1, , n} , 0, j {1, , n} , are such that x j Re then
n n n
kij
xi ,xj xi
kij ,
j =1
xj
xj
j =1
j =1 n
x j Re
xi ,xj xi
kij .
j =1
(3.9)
j =1
xj
kij
j =1 n
and
n j =1 n j =1
xj =
that fn and max1
x j kij
j =1
EP R
xj ,, max1
j n
We are interested in the relation between Theorem 1.1 and Corollary 3.3. Recall xj
n
j n
IN
xi . xi = n
x1
xn
j =1
lim fn (1, , t , , 1) =
i
j =1
x j Re
xi ,xj xi
PR
g( ) = f3 g( ) =
x3 x3 x3 , , x1 x2 x3
lim f3 (t , 1, 1).
t
T
xj xj
1 j n
max x j
14 In the case of =
4
, we have
3 1 1 + 3 2 2 2 2 3 1 = 3 < 0. 2 2 2
Thus we get f3 x3 x3 x3 , , x1 x2 x3
t
It is possible to compare this with Theorem 1.1 and Corollary 1.3 in the special case.
EP R
1 j n n n
PR
n
j =1
xj
xj
j =1
j =1
x j Re
(ii) if I = {i0 } for some i0 {1, , n} , then the following inequalities hold
n n
j =1
xj
xj
j =1
n
n
j =1
x j Re
IN
.
j =1
xj xj
1 j n
max x j ;
xi ,xj xi
j =1
xj xj
1 j n
max x j .
15
xn xn , , x1 xn
n
n
n
j =1
xj xj
1 j n
max x j
lim fn (1, , 1, t , 1, , , 1) =
i0
j =1
x j Re
Thus we have
n
j =1
xj xj
1 j n
max x j = fn 1, , 1,
t n
xn , 1, , 1 xi0
i0
lim fn (1, , 1, t , 1, , , 1)
j =1
x j Re
EP R
x1 + x2 Re
n
Finally, as an analogue of Theorem 1.1, we may state the following result as well: T HEOREM 3.7. Let (H , , ) be an inner product space, and x j H \ {0}, j {1, , n} . Then, for each xed i {1, , n} ,
PR
xj
xj
j =1
j =1
n Re
xi , xi
IN
x2 x1 + x1 x2 x2
n
j =1
xj xj
T
. x2 . Then the following x1 , x1 + x2 . x1
1 j n
max x j .
(3.10)
xj
j =1
n Re
xi , xi
j =1 n
xj xj
1 j n
max x j
and
n n
xj =
j =1
j =1
x j n Re
xi , xi
j =1
xj xj
1 j n
max x j
xi . xi
16
j 1 j n ( j {1, , n}) , then s j Proof. If we put s j = xj Thus, by Theorem 2.1 and Lemma 3.1, we have
1 ( j {1, , n}) .
fn (1, , 1) = = = =
t n
lim fn (s1 , , t , , sn )
i
j =1 n
s j x j Re max1
j n
xi , s jx j xi xj x j Re xi max1 j n x j , xj xi xj
j =1 n
xj 1 Re
j =1
xj xi , xi xj xi , xi
n
1 j n
max x j max x j .
n Re
j =1
xj xj
1 j n
[1] A. H. A NSARI AND M. S. M OSLEHIAN, More on reverse triangle inequality in inner products spaces, Int. J. Math. Math. Sci. (2005), no. 18, 28832893. [2] F. D ADIPOUR , M. S. M OSLEHIAN , J. M. R ASSIAS AND S.-E. TAKAHASI , Characterization of a generalized triangle inequality in normed spaces, Nonlinear Anal. 75 no. 2 (2012), 735741. [3] S. S. D RAGOMIR, Reverses of the triangle inequality in inner product spaces, Aust. J. Math. Anal. Appl. 1 no. 2 (2004), 114. [4] S. S. D RAGOMIR, Reverses of the triangle inequality in Banach spaces, JIPAM. J. Inequal. Pure Appl. Math. 6(5) (2005), Art. 129, pp. 46. [5] M. F UJII , M. K ATO , K.-S. S AITO AND T. TAMURA, Sharp mean triangle inequality, Math. Inequal. Appl. 13 (2010), 743752. [6] C.-Y. H SU , S.-Y. S HAW AND H.-J. W ONG, Renements of generalized triangle inequalities, J. Math. Anal. Appl. 344 (2008), 1731. [7] M. K HOSRAVI , H. M AHYAR AND M. S. M OSLEHIAN, Reverse triangle inequality in Hilbert C modules, JIPAM. J. Inequal. Pure Appl. Math. 10 no. 4 (2009), Article 110, 11 pp. [8] M. K ATO , K.-S. S AITO AND T. TAMURA, Sharp triangle Inequality and its reverse in Banach spaces, Math. Inequal. Appl. 10 no. 2 (2007), 451460. [9] L. M ALIGRANDA, Some remarks on the triangle inequality for norms, Banach J. Math. Anal. 2 no. 2 (2008), 3141. [10] M. S. M ARTIROSYAN AND S. V. S AMARCHYAN, Inversion of the triangle inequality in Rn , 38 (2003), no. 4, 6572. [11] K. M INENO , Y. N AKAMURA AND T. O HWADA, Characterization of the intermediate values of the triangle inequality, to appear in Math. Ineq. Appl. [12] K. M INENO , Y. N AKAMURA , S. N AKAMURA , C. TAMIYA AND T. O HWADA, Characterization of the intermediate values of the triangle inequality II, in preparation.
PR
EP R
REFERENCES
IN
Acknowledgement. The authors wish to thank the referee for his/her comments and suggestions on the manuscript. The rst author was supported in part by the National Science Foundation of China (61074048). The second and the third author were supported in part by the Grant-in-Aid for Scientic Research(C), Japan Society for the Promotion of Science (22540184) and (20540192), respectively.
17
[13] K.-I. M ITANI , K.-S. S AITO , M. K ATO AND T. TAMURA, On sharp triangle Inequalities in Banach spaces, J. Math. Anal. Appl. 10 no. 2 (2007), 451460. [14] K.-I. M ITANI , K.-S. S AITO, On sharp triangle Inequalities in Banach spaces II, J. Inequal. Appl. (2010), Art. ID 323609, 17 pp. [15] S. S AITOH, Generalizations of the triangle inequality, JIPAM. J. Inequal. Pure Appl. Math. 4 (2003), no. 3, Article 62, 5 pp.
Lingling Zhang Department of Mathematics Taiyuan University of Technology Taiyuan, 030024 P. R. China e-mail: zllww@126.com Tomoyoshi Ohwada Faculty of Education Shizuoka University Shizuoka 422-8529, Japan e-mail: etoowad@ipc.shizuoka.ac.jp Muneo Ch o Department of Mathematics Faculty of Science Kanagawa University Hiratsuka 259-1293, Japan
e-mail: chiyom01@kanagawa-u.ac.jp
www.ele-math.com mia@ele-math.com
PR
EP R
IN