vol. 55, no. 2 (2015), 157–162
On orthogonally additive injections and surjections
Karol Baron
Summary. Let E be a real inner product space of dimension at least 2 and V a linear topological Hausdorff space. If card E ⩽ card V , then the set of all orthogonally additive injections mapping E into V is dense in the space of all orthogonally additive functions from E into V with the Tychonoff topology. If card V ⩽ card E, then the set of all orthogonally additive surjections mapping E into V is dense in the space of all orthogonally additive functions from E into V with the Tychonoff topology.
Keywords
orthogonal additivity;
inner product space;
linear topological space;
Tychonoff topology;
dense set
MSC 2010
39B55; 46C99; 46A99 Received: 2016-02-06, Accepted: 2016-04-01
Dedicated to Professor Henryk Hudzik on his 70th birthday.
1. Introduction
Let E be a real inner product space of dimension at least 2 and V a linear topological Hausdorff space.
A function f mapping E into V is called orthogonally additive, if f(x + y) = f (x) + f (y) for all x, y ∈ E with x ⊥ y.
It is well known (see [6, Corollary 10] and [4, Theorem 1]) that every orthogonally additive function f defined on E has the form
f(x) = a(∥x∥2) + b(x) for x ∈ E, (1)
Karol Baron, Institute of Mathematics University of Silesia, Bankowa 14, 40–007 Katowice, Poland (e-mail:baron@us.edu.pl)
DOI 10.14708/cm.v55i2.1109 © 2015Polish Mathematical Society
where a and b are additive functions uniquely determined by f . Consequently, we have an operator Λ which to any orthogonally additive f∶ E → V assigns a pair (a, b) of additive functions such that (1) holds, i.e.
Λ f = (a, b) (2)
where
a∶ R → V, b∶ E → V are additive and (1) holds. (3) Putting
Hom(E, V) = { f ∶ E → V ∶ f is orthogonally additive}
and
Hom(S, V) = { f ∶ S → V ∶ f is additive}
for S ∈ {R, E}, we see that Λ∶ Hom(E, V) → Hom(R, V) × Hom(E, V), given by (2) and (3), is a linear bijection.
Given a non-empty set S, consider the set VS of all functions from S into V with the usual Tychonoff topology; clearly VSis a linear topological space. In what follows, we consider Hom(E, V) and Hom(S, V) for S ∈ {R, E} with the topology induced by VE and VS, respectively. According to [3, Theorem 1]:
1. The isomorphism Λ∶ Hom(E, V) → Hom(R, V) × Hom(E, V) given by (2) and (3) is a homeomorphism.
As an immediate consequence of [2, Theorem] and [3, Corollary 1] we have the follo- wing information:
2. The set
{ f ∈ Hom(E, V)∶ f is injective and f (E) = V}
is nowhere dense in Hom(E, V).
Basing on the continuity of Λ−1and ideas from [1, Remarks 1 and 2], we are going to show that each of the sets
{ f ∈ Hom(E, V) ∶ f is injective}, (4)
{ f ∈ Hom(E, V) ∶ f (E) = V} (5)
is dense in Hom(E, V) (cardinalities of E and V permitting).
2. Results
They read as follows.
2.1. Theorem. If card E ⩽ card V , then set (4) is dense in Hom(E, V).
2.2. Theorem. If card V ⩽ card E, then set (5) is dense in Hom(E, V).
Since the intersection of a dense set with the complement of a nowhere dense set is dense, the above theorems imply the following corollaries.
2.3. Corollary. If card E ⩽ card V , then the set
{ f ∈ Hom(E, V) ∶ f is injective and f (E) /= V}
is dense in Hom(E, V).
2.4. Corollary. If card V ⩽ card E, then the set
{ f ∈ Hom(E, V) ∶ f (E) = V and f is not injective}
is dense in Hom(E, V).
3. Proofs
We start with two lemmas.
3.1. Lemma. If card E ⩽ card V , then the set
{(a, b) ∈ Hom(R, V) × Hom(E, V) ∶ a(R) ∩ b(E) = {0} = b−1({0})} (6)
is dense in Hom(R, V) × Hom(E, V).
To formulate the second lemma, denote byL the set of all one-dimensional linear subspaces of E and byV the set of all pairs(V1, V2) of linear subspaces of V over the field Q of all rationals such that V1is finite-dimensional and V1⊕ V2= V.
Observe that if L∈L, then E= L ⊕ L⊥, whence card E= c⋅ card L⊥, and this implies (see, e.g., [7, formula (2.1) on p. 414]) that
card E= card L⊥. (7)
3.2. Lemma. If card V ⩽ card E, then the set
⋃
L∈L ,(V1, V2)∈V{(a, b) ∈ Hom(R, V) × Hom(E, V) ∶ a([0, ∞)) = V1, V2⊂ b(L⊥), b(L) = {0}}
(8)
is dense in Hom(R, V) × Hom(E, V).
Proof of the Lemmas. We divide it into three parts.
Part I. It concerns both lemmas. Fix non-empty open sets U ⊂ Hom(R, V) and V ⊂ Hom(E, V). To show that U × V intersects the considered set (i.e. (6) or (8)), we may assume that V /= {0},
U = a0+ {a ∈ Hom(R, V) ∶ a(αn) ∈ U for n ∈ {1, . . . , N}}
and
V = b0+ {b ∈ Hom(E, V) ∶ b(xn) ∈ U for n ∈ {1, . . . , N}}
with a0 ∈ Hom(R, V), b0 ∈ Hom(E, V), a neighbourhood U of zero in V and some α1, . . . , αN ∈ R, x1, . . . , xN∈ E, N ∈ N.
Let HR⊂ (0, ∞) and HEbe bases of the vector spacesR and E, respectively, both of them overQ. There are finite sets HR0 ⊂ HRand H0E ⊂ HE such that αn ∈ LinQH0
Rand xn ∈ LinQHE0 for n ∈ {1, . . . , N}. For every n ∈ {1, . . . , N}, α ∈ H0Rand x ∈ H0Elet ρ(n)α and r(n)x be rationals such that
αn= ∑
α∈H0R
ρ(n)α α , xn= ∑
x∈H0E
r(n)x x . (9)
Choose now a neighbourhood U0of zero in V such that
α∈H∑0
R
ρ(n)α U0∪ ∑
x∈H0
E
r(n)x U0 ⊂ U for n∈ {1, . . . , N}, (10)
and injective functions a1∶ HR0 → V, b1∶ H0E→ V such that
a1(HR0) ∪ b1(H0E) is linearly independent over Q, a1(HR0) ∩ b1(HE0) = ∅ and
a1(α) ∈ U0+ a0(α) for α ∈ HR0, b1(x) ∈ U0+ b0(x) for x ∈ H0E. (11) Part II. It concerns Lemma3.1. Let H be a basis of V overQ containing a1(H0R) ∪ b1(H0E).
Since (see, e.g., the proof of [5, Lemma 4.2.2]) card(HE/H0E) = card HE = card Lin
Q
HE= card E
⩽ card V = card Lin
Q
H= card H = card(H/(a1(H0R) ∪ b1(H0E))), we may extend b1to an injective b2∶ HE → H/a1(H0R). Let a2∶ HR→ a1(H0R) be an exten- sion of a1and consider the additive functions a∶ R → V and b∶ E → V such that a∣HR= a2
and b∣HE = b2. Clearly, b is injective, a(R) ∩ b(E) = Lin
Q
a(HR) ∩ Lin
Q
b(HE)
= Lin
Q
a2(HR) ∩ Lin
Q
b2(HE) ⊂ Lin
Q
a1(HR0) ∩ Lin
Q (H/a1(H0R)) = {0},
and, according to (9), (11) and (10), for any n∈ {1, . . . , N} we have a(αn) − a0(αn) = ∑
α∈H0
R
ρ(n)α (a1(α) − a0(α)) ∈ ∑
α∈H0
R
ρ(n)α U0⊂ U (12)
and
b(xn) − b0(xn) = ∑
x∈H0
E
r(n)x (b1(x) − b0(x)) ∈ ∑
x∈H0
E
r(n)x U0 ⊂ U. (13) It shows that(a, b) is in the set (6) and inU × V.
Part III. It concerns Lemma3.2. Let L be a one-dimensional linear subspace of E such that L∩ Lin
Q
H0E = {0}.
Let HLbe a basis of L overQ, and HL⊥a basis of L⊥overQ. Put V1= Lin
Q
a1(HR0)
and let a∶ R → V1be an additive extension of a1such that a([0, ∞)) = V1. Consider now a basis H of V overQ containing a1(HR0). Then, by (7),
card(H/a1(H0R)) = card H = card V ⩽ card E = card L⊥= card HL⊥= card(HL⊥/HE0), and so there is an additive extension b∶ E → V of b1such that
b(HL) = {0} and b(HL⊥/HE0) = H/a1(HR0).
Putting
V2= Lin
Q (H/a1(HR0)),
we see that(V1, V2) ∈V. It shows that(a, b) is in the set (8). Moreover, by (9), (11) and (10), for every n∈ {1, . . . , N} we have (12) and (13) and so(a, b) is also in U × V.
Proof of Theorem2.1. Denote the set (6) byD. Since Λ−1 is continuous, it follows from Lemma3.1that Λ−1(D) is dense in Hom(E, V), and it is enough to show that any f in Λ−1(D) is injective. Indeed, let f ∈ Λ−1(D). Then
(a, b) ∶= Λ f ∈ D, (14)
and if f(x) = f (y) for some x, y ∈ E, then
a(∥x∥2) − a(∥y∥2) = b(y) − b(x).
The left-hand side belongs to a(R) and the right-hand side is in b(E), whence b(x) = b(y) and x = y.
Proof of Theorem2.2. Denoting now byD the set (8), it is enough to show that any f in Λ−1(D) maps E onto V. Indeed, let f ∈ Λ−1(D). Keeping (14), we can find L∈ L and (V1, V2) ∈V such that
a([0, ∞)) = V1, V2⊂ b(L⊥) and b(L) = {0}.
Fix arbitrary v∈ V. Then v = v1+v2with v1∈ V1, v2∈ V2, and v2= b(x2) for some x2∈ L⊥. Choose α∈ [0, ∞) with
a(α) = v1− a(∥x2∥2) and then x1∈ L with ∥x1∥2= α. We have
f(x1+ x2) = a(∥x1∥2+ ∥x2∥2) + b(x1+ x2) = a(α) + a(∥x2∥2) + b(x2) = v1+ v2= v, which ends the proof.
The reader interested in further problems connected with orthogonal additivity is referred to a survey paper [8] by Justyna Sikorska.
Acknowledgement The research was supported by the Silesian University Mathematics
Department (Iterative Functional Equations and Real Analysis program).
References
[1] K. Baron, On some orthogonally additive functions on inner product spaces, Ann. Univ. Sci. Budapest. Sect.
Comput. 40 (2013), 123–127.
[2] K. Baron, Orthogonally additive bijections are additive, Aequationes Math. 89 (2015), 297–299, DOI 10.1007/s00010-013-0242-6.
[3] K. Baron, On the continuous dependence of solutions to orthogonal additivity problem on given functions, Ann. Math. Sil. 29 (2015), 19–23,DOI 10.1515/amsil-2015-0002.
[4] K. Baron and J. Rätz, On orthogonally additive mappings on inner product spaces, Bull. Polish Acad. Sci.
Math. 43 (1995), 187–189.
[5] M. Kuczma, An introduction to the theory of functional equations and inequalities. Cauchy’s equation and Jensen’s inequality, 2nd ed., Birkhäuser Verlag, Basel 2009.
[6] J. Rätz, On orthogonally additive mappings, Aequationes Math. 28 (1985), 35–49,DOI 10.1007/bf02189390.
[7] W. Sierpiński, Cardinal and ordinal numbers, Monografie Matematyczne, vol. 34, Państwowe Wydawnic- two Naukowe, Warszawa 1958.
[8] J. Sikorska, Orthogonalities and functional equations, Aequationes Math. 89 (2015), 215–277, DOI 10.1007/s00010-014-0288-0.
© 2015Polish Mathematical Society