Comment.Math.Univ.Carolin.

56,3 (2015) 307–317 307

Orthosymmetric bilinear map on Riesz spaces

Elmiloud Chil, Mohamed Mokaddem, Bourokba Hassen

Abstract. Let E be a Riesz space, F a Hausdorff topological vector space (t.v.s.).

We prove, under a certain separation condition, that any orthosymmetric bilinear
map T : E × E → F is automatically symmetric. This generalizes in certain way
an earlier result by F. Ben Amor [On orthosymmetric bilinear maps, Positivity
14 (2010), 123–134]. As an application, we show that under a certain separation
condition, any orthogonally additive homogeneous polynomial P : E → F is
linearly represented. This fits in the type of results by Y. Benyamini, S. Lassalle
and J.L.G. Llavona [Homogeneous orthogonally additive polynomials on Banach
lattices, Bulletin of the London Mathematical Society 38 (2006), no. 3 123–134].

Keywords: orthosymmetric multilinear map; homogeneous polynomial; Riesz

space
Classification: 06F25, 46A40

1. Introduction
One of the relevant problems in operator theory is to describe orthogonally ad-
ditive polynomials via linear operators. This problem can be treated in a different
manner, depending on domains and co-domains on which polynomials act. Inter-
est in orthogonally additive polynomials on Banach lattices originates in the work
of K. Sundaresan [19] where the space of n-homogeneous orthogonally additive
polynomials on the Banach lattices lp and Lp [0, 1] was characterized. It is only
recently that the class of such mappings has been getting more attention. We are
thinking here about works on orthogonally additive polynomials and holomorphic
functions and orthosymmetric multilinear mappings on different Banach lattices
and also C∗ -algebras, see for instance [3], [6], [7], [18] and [8], [13], [14], [17]. Proofs
of the aforementioned results are strongly based on the representation of these
spaces as vector spaces of extended continuous functions. So they are not appli-
cable to general Riesz spaces. That is why we need to develop new approaches.
Actually, the innovation of this work consists in making a relationship between
orthogonally additive homogeneous polynomials and orthosymmetric multilinear
mappings which leads to a constructive proofs of Sundaresan results [19], those of
D. Prez-Garcı́a and I. Villanueva in [18], Y. Benyamini, S. Lassalle and Llavona [3],
D. Carando, S. Lassalle and I. Zalduendo in [8], and those of A. Ibort, P. Linares
and J.G. Llavona [13].

DOI 10.14712/1213-7243.2015.132
308 Chil E., Mokaddem M., Hassen B.

A multilinear mapping T : E n −→ F is said to be orthosymmetric if

T (x1 , . . . , xn ) = 0 whenever x1 , . . . , xn ∈ E satisfy xi ⊥xj for some i 6= j. The
main purpose of this paper is to show that orthosymmetric multilinear mappings
are symmetric with general conditions on E and F . There are many results of this
kind in the literature. In this context, we mention some of them. In [7], where
orthosymmetric bilinear maps were introduced, the authors showed that any or-
thosymmetric positive bilinear map between Archimedean Riesz spaces is sym-
metric. Later, in [10] it is shown that the result remains valid for order bounded
orthosymmetric bilinear maps. Recall that a bilinear map T : E × E → F is
positive if T (x, y) ≥ 0 whenever (x, y) ∈ E + × E + , and is order bounded if
given (x, y) ∈ E + × E + there exists a ∈ F + such that |T (z, w)| ≤ a for all
(0, 0) ≤ (z, w) ≤ (x, y) ∈ E × E. Every positive bilinear map is order bounded.
Recall also that T is said to be symmetric if T (x, y) = T (y, x). In a recent paper
[2, Theorem 13], the result is proved for (r-u) continuous orthosymmetric bilinear
operators. Recall that T : E × E → F is (r-u) continuous if xn , yn −→ 0 (r-u) in
E implies that T (xn , yn ) −→ 0 (r-u) in F . Finally, in a preprint [11] of the first
and second author together with M. Meyer the result is obtained for any continu-
ous orthosymmetric bilinear map defined on a relatively uniformly complete Riesz
space with values in a Hausdorff t.v.s. The purpose of this paper is to extend the
result in [11] to a general domain E. Actually, the relative uniform completeness
condition on the domain will be removed and substituted by a separation condi-
tion. Namely, (T (E n ))′ , the topological dual of the vector space generated by the
range of T , separates points. Recall that (T (E n ))′ separates points if for every
two vectors x, y in the vector space generated by the range of T such that x 6= y
there is some f ∈ (T (E n ))′ with f (x) 6= f (y). This replaces T positive, T order
bounded and domain uniformly complete and co-domain Riesz space in [7], [10]
and [2] respectively. In the second section of this paper this result is applied to
show under the same requirements
Q that each of such mappings T is factorized
by a linear operator Q S : ni=1 E → F such that T (x1 , . . . , xn ) = S(x1 · · · xn ) for
x1 , . . . , xn ∈ E and ni=1 E = {x1 · · · xn , xi ∈ E} (the multiplication under con-
sideration is the f -algebra multiplication of E u , the universal completion of E).
The paper concludes with an application to orthogonally additive polynomials.

2. Preliminaries
We take it for granted that the reader is familiar with the notion of Riesz spaces
(or vector lattices) and operators between them. For terminology, notations and
concepts not explained or proved in this paper we refer the reader to the standard
monographs [1], [15].
A Riesz space E is called Archimedean if for each non-zero a ∈ E the set
{na, n = ±1, ±2, . . . } has no upper bound in E. In order to avoid unnecessary
repetition, we will assume throughout the paper that all Riesz spaces under con-
sideration are Archimedean. The relatively uniform topology on Riesz spaces
plays a key role in the context of this work. Let us recall the definition of the
relatively uniform convergence. Let E be an Archimedean Riesz space and an
 Orthosymmetric bilinear map on Riesz spaces 309

element u ∈ E + . A sequence (xn )n of elements of E converges u-uniformly to

an element x ∈ E whenever, for every ǫ > 0, there exists a natural number Nǫ
such that |xn − x| ≤ ǫu holds for all n ≥ Nǫ . This will be denoted xn → x(u).
The element u is called the regulator of the convergence. The sequence (xn )n
converges relatively uniformly to x ∈ E, whenever xn → x(u) for some u ∈ E + .
We shall write xn → x(r-u) if we do not want to specify the regulator. Relatively
uniform limits are unique if and only if E is Archimedean [15, Theorem 63.2]. A
nonempty subset D of E is said to be relatively uniformly closed if every relatively
uniformly convergent sequence in D, has its limit also in D. We emphasize that
the regulator does not need to be an element of D. The relatively uniformly closed
subsets are the closed sets of the relatively uniform topology in E. The notion of
relatively uniform Cauchy sequence is defined in the obvious way. A Riesz space
E is said to be relatively uniformly complete whenever every relatively uniformly
Cauchy sequence has a (unique) limit. For more details we refer the reader to
[15].
Next, we discuss linear operators on Riesz spaces. Let E and F be Riesz spaces
with positive cones E + and F + respectively, and let T be an operator from E
into F . T is said to be order bounded if for each x ∈ E + there exists y ∈ F +
such that |T (z)| ≤ y in F whenever |z| ≤ x in E. The operator T is said to be
positive if T (x) ∈ F + for all x ∈ E + . Every positive operator is of course order
bounded. The set Lb (E) of all order bounded operators on E is an ordered vector
space with respect to the pointwise operations and order. The positive cone of
Lb (E) is the subset of all positive operators. An element T in Lb (E) is referred
to as an orthomorphism if, for all x, y ∈ E, |T (x)| ∧ |y| = 0 whenever |x| ∧ |y| = 0.
Under the ordering and operations inherited from Lb (E), the set Orth(E) of all
orthomorphisms on E is an Archimedean Riesz space. The absolute value in
Orth(E) is given by |T | (x) = |T (x)| for all x ∈ E + . More details about order
bounded operators and orthomorphisms can be found in [1], [12].
The following paragraph deals with f -algebras. The Riesz space E is said to
be a Riesz algebra if there exists an associative multiplication in E with the usual
algebra properties such that xy ∈ E + for all x, y ∈ E + . We say that the Riesz
algebra is semiprime if 0 is the only nilpotent element. The Riesz algebra E is
said to be an f -algebra whenever x ∧ y = 0 implies xz ∧ y = zx ∧ y = 0 for all
z ∈ E + . It follows that multiplication in an f -algebra is an orthomorphism. If
E is a Riesz space then the Riesz space Orth(E) is an f -algebra with respect to
the composition as multiplication. Moreover the identity map on E is the multi-
plicative unit of Orth(E). In particular, the f -algebra Orth(E) is semiprime and
commutative. We end this paragraph with a remarkable result due to A.C. Za-
anen [1, Theorem 2.62] that if E is an f -algebra with unit element, then the
mapping π : x → πx from E into Orth(E) is a Riesz and algebra isomorphism,
where πx (y) = xy for all y ∈ E. For more details about f -algebras we refer the
reader to [1], [4], [12].
We end this section recalling that a Dedekind complete Riesz space E is said to
be universally complete whenever every set of pairwise disjoint positive elements
310 Chil E., Mokaddem M., Hassen B.

has a supremum. Every Archimedean Riesz space E has a unique (up to a Riesz
isomorphism) universally completion denoted E u , i.e., there exists a unique uni-
versal complete Riesz space such that E can be identified with an order dense
Riesz subspace of E u . Moreover E u is furnished with a multiplication, under
which E u is an f -algebra with unit element. For more details see [1].

3. Main results
The following definition turns out to be useful in the sequel. We shall say that
a mapping T from a Riesz space E into a Hausdorff t.v.s. F is continuous, if
xn →0 (r-u) in E implies T (xn ) → 0 in F .
The main result of this paper is strongly based on the following theorem [10].

Theorem 1. Let E be a relatively uniformly complete Riesz space, F be a Haus-

dorff t.v.s. (not necessarily Riesz space) and let ϕ : E × E → F be a continuous
orthosymmetric bilinear map. Then ϕ is symmetric.

In what follows, the relative uniform completeness condition on the domain

will be removed and substituted by a separation condition. First let us prove the
following useful proposition.

Proposition 2. Let E be a Riesz space, F be a Hausdorff t.v.s., and let T :

E n → F be a continuous orthosymmetric multilinear map such that (T (E n ))′
separates points. If σ ∈ S(n) is a permutation then

T (x1 , . . . , xn ) = T (xσ(1) , . . . , xσ(n) )

for all x1 , . . . , xn ∈ E.

Proof: We begin by proving the case where n = 2. Let x, y ∈ E + and put e = x+

y. The principal ideal generated by e is denoted Ee . It follows from the Kakutani
Theorem [16, Theorem 2.1.3] that Ee is uniformly dense in C(K)-space for a
compact Hausdorff space K. Denote Te the restriction of T to Ee × Ee . We claim
that for any f ∈ (T (E × E))′ , f ◦ Te can be extended to a uniformly continuous
orthosymmetric bilinear map from C(K) × C(K) to R. To see this let (u, v) ∈
C(K)×C(K). From the uniform density of Ee ×Ee in C(K)×C(K) there exists a
sequence ((un , vn ))n ∈ Ee × Ee such (un , vn ) → (u, v)(r-u). From bilinearity and
continuity of f ◦Te there exists c > 0 such that |f ◦ Te (um , vm ) − f ◦ Te (un , vn )| ≤
c(kum − un k + kvm − vn k). Consequently, f ◦ Te (u, v) will be defined as the limit
of the Cauchy sequence f ◦ Te (un , vn ). Denoting Ge : C(K) × C(K) → R the
extension of f ◦Te and applying the above theorem it is easy to see that Ge (x, y) =
Ge (y, x) and then f ◦ Te (x, y) = f ◦ Te (y, x) for all f ∈ (T (E × E))′ . Since
(T (E × E))′ separates points, it follows that T (x, y) = T (y, x). For the general
case where n ≥ 2, let i, j ∈ {1, . . . , n} such that i 6= j, x1 , . . . , xn ∈ E, and define
 Orthosymmetric bilinear map on Riesz spaces 311

the mapping φ as follows:

φ : E × E −→ F
x , . . . , y , . . . , xn )
(x, y) 7−→ T (x1 , . . . , |{z}
|{z}
i j

It is easily seen that φ is a continuous orthosymmetric bilinear map, then by

the preceding case φ is symmetric. So

T (x1 , . . . , xi , . . . , xj , . . . , xn ) = T (x1 , . . . , xj , . . . , xi , . . . , xn ).

Since S(n) is generated by transpositions, we obtain

T (x1 , . . . , xn ) = T (xσ(1) , . . . , xσ(n) )

for all σ ∈ S(n). 

As an immediate application to the above proposition we obtain the following
result.
Proposition 3. Let E be a Riesz space, F be a Hausdorff t.v.s., and let T :
E n → F be a continuous orthosymmetric multilinear map such that (T (E n ))′
separates points. Then

T (π1 (x1 ), . . . , πn (xn )) = T (x1 , . . . , xn−1 , π1 . . . πn (xn ))

for all x1 , . . . , xn ∈ E and π1 , . . . , πn ∈ Orth(E).

Proof: It is sufficient to prove that if i ≤ j ∈ {1, . . . , n} then

T (x1 , . . . , π(xi ), . . . , xn ) = T (x1 , . . . , π(xj ), . . . , xn )

for all x1 , . . . , xn ∈ E and π ∈ Orth(E).

Let i ≤ j ∈ {1, . . . , n}, π ∈ Orth(E) and define the mapping φ as follows

φ : E × E −→ F
(x, y) 7−→ T (x1 , . . . , π(x), . . . , y , . . . , xn ).
|{z} |{z}
i j

Since (T (E n ))′ separates points, it is not hard to verify that as a consequence

(φ(E × E))′ separates points. Now it is straightforward to show that φ(x, y) =
φ(y, x) because φ is a continuous orthosymmetric bilinear map such that (φ(E ×
E))′ separates points. Therefore

T (x1 , . . . , π(x), . . . , y , . . . , xn ) = T (x1 , . . . , π(y), . . . , |{z}

x , . . . , xn )
|{z} |{z} |{z}
i j i j
312 Chil E., Mokaddem M., Hassen B.

on the other hand T is symmetric (Proposition 2), so

T (x1 , . . . , π(y), . . . , |{z} x , . . . , π(y), . . . , xn ).

x , . . . , xn ) = T (x1 , . . . , |{z}
|{z} |{z}
i j i j

Consequently,

T (x1 , . . . , π(x), . . . , y , . . . , xn ) = T (x1 , . . . , |{z}

x , . . . , π(y), . . . , xn )
|{z} |{z} |{z}
i j i j

for all x, y ∈ E. In particular, we have

T (x1 , . . . , π(xi ), . . . , xj , . . . , xn ) = T (x1 , . . . , xi , . . . , π(xj ), . . . , xn )

and the proof is finished. 

The previous results will be used to characterize orthogonally additive homoge-
neous polynomials. Let E be a Riesz space and let F be a t.v.s. A map P : E → F
is called a homogeneous polynomial of degree n (or an n-homogeneous polyno-
mial) if P (x) = ϕ(x, . . . , x), where ϕ is an n-multilinear map from E n into F .
A homogeneous polynomial of degree n; P : E → F is said to be orthogonally ad-
ditive if P (x + y) = P (x) + P (y) where x, y ∈ E are orthogonal (i.e. |x| ∧ |y| = 0).
We denote by P0 (n E, F ) the set of n-homogeneous orthogonally additive polyno-
mials from E to F (continuous in the case that E is equipped with the relatively
uniform topology and F a Hausdorff topology). The interest in orthogonally
additive polynomials on Banach lattices originates in the work of K. Sundare-
san [19], where the space of n-homogeneous orthogonally additive polynomials
on Lp and lp was characterized. More precisely, K. Sundaresan proved that ev-
ery n-homogeneous orthogonally additive polynomial
R n P : Lp → R is determined
p
by some g ∈ L p−n via the formula P (f ) = f gdµ for all f ∈ Lp . After that,
D. Prez-Garcı́a and I. Villanueva in [18], D. Carando, S. Lassalle and I. Zalduendo
in [9] proved the following analogous result for C(X) spaces: Let Y be a Banach
space, let P : C(X) → Y be an orthogonally additive n-homogeneous polynomial
and let ϕ : (C(X))n → Y be its unique associated symmetric multilinear oper-
ator. Then there exists a linear operator S : C(X) P → Y ∗∗such that kSk = kϕk
and there exists a finitely additive measure
R µ : →P Y such that for every
f ∈ C(X), we have P (f ) = S(f n ) = X f n gdµ . Here is the Borel σ-algebra
on X. By using heavily the representation of Riesz spaces as vector spaces of
extended continuous functions, Y. Benyamini , S. Lassalle and Llavona [3] have
proven a result analogous to that of K. Sundaresan for the classes of Banach lat-
tices of functions of order continuous Köthe function space, whose dual is given by
integrals. Very recently the first and the second author together with M. Meyer
show the analogous result for uniformly complete Riesz spaces, that is, if E is a
uniformly complete Qn Riesz space, F is a Hausdorff t.v.s. and P ∈ P0 (n E, F ) then
n
there exists S : i=1 E → F such that P (x) = S(x ) for every x ∈ E, where
Q n n u
i=1 E = {πi=1 xi : xi ∈ E} is the Riesz subspace of E , the universal completion
 Orthosymmetric bilinear map on Riesz spaces 313

of E. Here the multiplication under consideration is the f -algebra multiplication

of E u , see [1] for more details. It is now relevant to investigate if the same result
still holds true in the case where the domain is just a Riesz space. Indeed, we
show in the following section that if E is a Riesz space, F is a Hausdorff t.v.s. and
P ∈ P0 (n E, F ),Q then under additional condition that (ϕ(E n ))′ separates points,
n
there exists S : i=1 E → F such that P (x) = S(xn ) for every x ∈ E.
The following theorem which is somehow an extension of the well known Kan-
torovich theorem [1, Theorem 1.10] will be of great use next.
Theorem 4. Let E be a Riesz space, F be a Hausdorff t.v.s., and let T : E + → F
be a continuous additive mapping. Then T has a unique extension to a continuous
operator from E to F . Moreover the extension, denoted by T again, is given by
T (x) = T (x+ ) − T (x− )
for all x ∈ E.
Proof: Consider the mapping S : E → F defined by S(x) = T (x+ ) − T (x− ).
Obviously S is the only possible linear extension of T to all of E. The additivity
of S can be proved as in [1, Theorem 1.10]. To show that S is homogeneous, let
λ ∈ R+ , x ∈ E and take a sequence of nonnegative rational numbers (λn ) such
that λn → λ in R, then S(λn x) = λn S(x) for all n ∈ N.
We claim that
λn S(x) −→ λS(x), and S(λn x) −→ S(λx).
From the continuity of the scalar multiplication R × F → F it is easily verified
that λn S(x) −→ λS(x).
By definition, S(λn x) = T ((λn x)+ ) − T ((λn x)− ). As (λn x)+ →(λx)+ (r-u) and
(λn x)− →(λx)− (r-u), it follows from the continuity of T that S(λn x)→S(λx).
Since F is a Hausdorff space we deduce that
λS(x) = S(λx) f or λ ∈ R+ , x ∈ E.
Finally, if λ ∈ R− and x ∈ E, since 0 = S(y − y) = S(y) + S(−y) it follows that
S(y) = −S(−y),
then

S(λx) = S(−(−λ)x)) = −S((−λ)x)) = −(−λ)S(x) = λS(x).

So, S is homogeneous. It remains to show the continuity of S. To this end let
x ∈ E and (xn ) be a sequence in E such that xn →x (r-u), then it follows that
x+ + − − + −
n →x (r-u) and xn →x (r-u). Writing that S(xn ) = T (xn ) − T (xn ) we see that
continuity of S follows immediately from the continuity of T , and the proof is
finished. 
Now we are able to announce the main result of this paper.
314 Chil E., Mokaddem M., Hassen B.

Theorem 5. Let E be an Archimedean Riesz space, F be a Hausdorff t.v.s.,

and let T : E n → F be a continuous orthosymmetric multilinear map Qn such that
(T (E n ))′ separates points. Then there exists a linear operator S : i=1 E → F
such that
T (x1 , . . . , xn ) = S(x1 · · · xn ).
Proof: Let 0 ≤ x, y ∈ E such that xn = y n . From the fact that E u is a unital
f -algebra, it follows that x = y [4, Proposition
Qn 2]. This implies that we can define
a mapping S on the positive cone of i=1 E by putting S(xn ) = T (x, . . . , x) for
all 0 ≤ x ∈ E. We claim that S is additive. To see this, let 0 ≤ x, y ∈ E, then
there exists a unique 0 ≤ z such that xn + y n = z n [5, Lemma 3.1]. Therefore

S(xn + y n ) = S(z n ) = T (z, . . . , z).

Now, put e = x + y + z. We denote by Ie the principal ideal generated by e.

Ie is a uniformly complete Riesz space with strong order unit, then from [12, Re-
mark 19.5] there exists πx , πy , πz ∈ Orth(Ie ) such that x = πx (e), y = πy (e), z =
πz (e). Now by Proposition 3,

S(z n ) = T (πz (e), . . . , πz (e)) = T (e, . . . , πzn (e)).

On the other hand, applying [5, Theorem 2.2] to the mapping

n
Y
φ : E × · · · × E −→ E
i=1
(x1 , . . . , xn ) 7−→ x1 x2 · · · xn

we get
φ(πxn (e), . . . , e) = φ(πx (e), πx (e), . . . , πx (e)).
So
πxn (e)en−1 = (πx (e))n = xn .
Analogously, we have

πyn (e)en−1 = (πy (e))n = y n and πzn (e)en−1 = (πz (e))n = z n .

Therefore

(πxn (e) + πyn (e))en−1 = xn + y n = z n = πzn (e)en−1 .

And thus
(πxn (e) + πyn (e) − πzn (e))en−1 = 0.
Since πxn (e) + πyn (e) − πzn (e) ∈ {e}, where {e} is the band generated by e, and
since {e} is a ring ideal of E u , we deduce that (πxn (e) + πyn (e) − πzn (e))en−2 ∈ {e} .
On the other hand, we have that (πxn (e) + πyn (e) − πzn (e))en−2 e = 0, so that
 Orthosymmetric bilinear map on Riesz spaces 315
d
(πxn (e) + πyn (e) − πzn (e))en−2 ∈ {e} (because E u is semiprime). Consequently,
(πxn (e) + πyn (e) − πzn (e))en−2 = 0 and by iteration we deduce that

πxn (e) + πyn (e) = πzn (e)

so we have that

S(z n ) = T (e, . . . e, πzn (e))

= T (e, . . . , e, πxn (e) + πyn (e))
= T (e, . . . , e, πxn (e)) + T (e, . . . , e, πyn (e))
= T (πx (e), . . . , πx (e)) + T (πy (e), . . . , πy (e))
= S(xn ) + S(y n ).
Q Q
It follows that S isQadditive on ni=1 E + . Since ni=1 E + is the positive cone
of the Riesz space ni=1 E, then by the precedingQ theorem, S can be extended
in a unique way to a continuous operator from ni=1 E to F . This extension is
also denoted by S. Now, to complete the proof it suffices to show that for all
x1 , . . . , xn ∈ E +
T (x1 , . . . , xn ) = S(x1 · · · xn ).
Put e = x1 + · · · + xn , then there exists πxi ∈ Orth(Ie ) such that xi = πxi (e) so

T (x1 , . . . , xn ) = T (πx1 (e), . . . , πxn (e)) = T (e, . . . , πx1 · · · πxn (e)).

On the other hand Orth(Ie ) is a uniformly complete semiprime f -algebra then

there exists π ∈ Orth(Ie ) such that πx1 · · · πxn = π n therefore

T (x1 , . . . , xn ) = T (e, . . . , π n (e))

= S(π(e)n )
= S(πx1 · · · πxn (e)e · · · e)
= S(πx1 (e) · · · πxn (e))
= S(x1 · · · xn )

and the proof is finished. 

In what follows, we intend to apply the above theorem to the orthogonally
additive polynomials case. To do this, we will need the following lemma.
Lemma 6. Let E be a Riesz space, F a Hausdorff t.v.s., and let P ∈ P0 (n E, F )
whose associated symmetric multilinear map ϕ satisfies that (ϕ(E n ))′ separates
points. Then, ϕ is orthosymmetric.
Proof: We begin by proving the case where n = 2. Let x, y ∈ E such that
|x|∧|y| = 0 and put e = |x|+|y|. The principal ideal generated by e is denoted Ee .
It follows from the Kakutani Theorem [16, Theorem 2.1.3] that Ee is uniformly
dense in C(K)-space for a compact Hausdorff space K. Denote ϕe the restriction
316 Chil E., Mokaddem M., Hassen B.

of ϕ to Ee × Ee . Using the same technique of Proposition 2 we show that for any

f ∈ (ϕ(E ×E))′ , f ◦ϕe can be extended to a uniformly continuous orthosymmetric
bilinear map from C(K) × C(K) to R, denote Ge this extension. Now it is not
hard to see that Pe , the homogenous polynomial associated to Ge , is orthogonally
additive since P is orthogonally additive. Applying [7, Lemma 4.1] it follows
that Ge is orthosymmetric, thus Ge (x, y) = 0 and then f ◦ ϕe (x, y) = 0 for all
f ∈ (ϕ(E × E))′ . Since (ϕ(E × E))′ separates points, it follows that ϕ(x, y) = 0.
The general case is left to the reader. 
Corollary 7. Let E be a Riesz space, F a Hausdorff t.v.s., and let P ∈ P0 (n E, F )
whose associated symmetric multilinear map ϕQsatisfies that (ϕ(E n ))′ separates
points. Then there exists a linear operator S : ni=1 E → F such that
P (x) = S(xn ).
We end this paper by giving some remarks. First of all we point out that our
approach works in the special case where ψ is any orthosymmetric multilinear
map with co-domain R[X], the Hausdorff t.v.s. of all real polynomials, while [2,
Theorem 13] and [3, Theorem 2.3] do not apply for such co-domain since R[X] is
neither a Riesz space nor a Banach lattice. Note that the result of [7] and [11]
also fails for such maps. Now let E be the Archimedean Riesz space of piecewise
linear functions on [0, 1], with only finite number of discontinuities. Then E is
not uniformly complete so the result in [9] does not apply but Theorem 2 in the
current paper holds. It is easily seen that for every multilinear map T such that
T (E n ) = R[X] we have (T (E n ))′ separates points. Finally note that our approach
fails for non continuous orthosymmetric multilinear mappings as it is shown in
the following example.
Example 8. Let E be the Riesz space of all real valued functions f on [0, 1]
satisfying that there is a finite subset {x0 , . . . , xn } of [0, 1] such that 0 = x0 < x1 <
· · · < xn = 1 and f is linear on each interval [xi−1 , xi ), i.e., f (x) = mi (f )x + bi (f )
for all x ∈ [xi−1 , xi ). Now define
ϕ : E × E −→ R
(f, g) 7−→ m0 (f )b0 (g)

It is easily checked that ϕ is well defined, orthosymmetric and bilinear map.

On the other hand, note that ϕ(E 2 ) = R so (ϕ(E 2 ))′ separates points but ϕ
is not symmetric. Now it is a routine to verify that ϕ is neither positive nor
order bounded, so since E and R are normed Riesz spaces it follows that ϕ is
not continuous. From the preceding example we emphasize that the condition
(ϕ(E 2 ))′ separates points has no kind of relation with ϕ being positive, order
bounded or order continuous.
Acknowledgment. The authors are very grateful to the referee for providing
valuable suggestions and comments that greatly improved the quality of the paper.
 Orthosymmetric bilinear map on Riesz spaces 317

References
[1] Aliprantis C.D., Burkinshaw O., Positive Operators, Springer, Dordrecht, 2006.
[2] Ben Amor F., On orthosymmetric bilinear maps, Positivity 14 (2010), 123–134.
[3] Benyamini Y., Lassalle S., Llavona J.L.G., Homogeneous orthogonally additive polynomials
on Banach lattices, Bull. London Math. Soc. 38 (2006), no. 3, 459–469.
[4] Beukers F., Huijsmans C.B., Calculus in f-algebras, J. Austral. Math. Soc. Ser. A 37 (1984),
no. 1, 110–116.
[5] Boulabiar K., On products in lattice-ordered algebras, J. Austral. Math. Soc. 75 (2003),
no. 1, 1435–1442.
[6] Bu Q., Buskes G., Polynomials on Banach lattices and positive tensor products, J. Math.
Anal. Appl. 388 (2012), 845–862.
[7] Buskes G., van Rooij A., Almost f -algebras: commutativity and the Cauchy-Schwarz in-
equality, Positivity 4 (2000), no. 3, 227–231.
[8] Carando D., Lassalle S., Zalduendo I., Orthogonally additive polynomials over C(K) are
measures – a short proof , Integral Equations Operator Theory, 56 (2006), no. 4, 597–602.
[9] Carando D., Lassalle S., Zalduendo I., Orthogonally additive holomorphic functions of
bounded type over C(K), Proc. Edinb. Math. Soc. (2) 53 (2010), no. 3, 609–618.
[10] Chil E., Order bounded orthosymmetric bilinear operator, Czechoslovak Math. J. 61 (2011),
no. 4, 873–880.
[11] Chil E., Meyer M., Mokaddem M., On orthosymmetric multilinear maps, Positivity
(preprint).
[12] de Pagter B., f -algebras and Orthomorphisms, thesis, Leiden, 1981.
[13] Ibort A., Linares P., Llavona J.G., On the representation of orthogonally additive polyno-
mials in ℓp , Publ. RIMS Kyoto Univ. 45 (2009), 519–524.
[14] Jaramillo J.A., Prieto A., Zalduendo I., Orthogonally additive holomorphic functions on
open subsets of C(K), Rev. Mat. Complut. 25 (2012), no. 1, 31–41.
[15] Luxemburg W.A., Zaanen A.C., Riesz Spaces I , North-Holland, Amsterdam, 1971.
[16] Meyer-Nieberg P., Banach Lattices, Springer, Berlin, 1991.
[17] Palazuelos C., Peralta A.M., Villanueva I., Orthogonally additive polynomials on C ∗ -
algebras, Quart. J. Math. 59 (2008), 363–374.
[18] Perez-Garcı́a D., Villanueva I., Orthogonally additive polynomials on space of continuous
functions, J. Math. Anal. Appl. 306 (2005), no. 1, 97–105.
[19] Sundaresan K., Geometry of spaces of homogeneous polynomials on Banach lattices, Ap-
plied geometry and discrete mathematics, 571–586, DIMACS Series in Discrete Mathemat-
ics and Theoretical computer Science no. 4, Amer. Math. Soc., Providence, RI, 1991.

Chil E., Hassen B.:

Institut prparatoire aux tudes d’ingenieurs de Tunis, 2 Rue jawaher lel
Nehrou Monflery 1008 Tunisia
E-mail: [email protected]
[email protected]

Mokaddem M.:
Research Laboratory of Algebra, Topology, Arithmetic, and Order, De-
partment of Mathematics, Faculty of Mathematical, Physical Sciences of
Tunis, Tunis-El Manar University, 2092-El Manar, Tunisia
E-mail: [email protected]

(Received March 17, 2014, revised April 19, 2015)