Section 3 is devoted to additive bijective mappings preserving orthogonality. The structure of such mappings is clarified. The results are quite similar to those for algebras of bounded operators given in [5].
2 Mappings preserving zero products
For the extensive literature on this topic see for example [2,3] and the references therein. In this section we prove the following theorem.
Theorem 2.1.
Let
$\mathcal{D}\subset \mathcal{\mathscr{H}}$
be an (F)domain and let
$\mathcal{A},\mathcal{\mathcal{B}}$
be standard *operator algebras on
$\mathcal{D}$
. If
$\Phi :\mathcal{A}\to \mathcal{\mathcal{B}}$
is an additive, bijective mapping that preserves zero products, then there are
$c\in \mathbb{C}$
and
$T:\mathcal{D}\to \mathcal{D}$
bijective and either linear or conjugate linear such that
$$\Phi \left(A\right)=cTA{T}^{1}(A\in \mathcal{A}).$$
If
$T$
is linear, then
$T\in \mathcal{\mathcal{L}}+(\mathcal{D})$
The main part of the proof is contained in the following theorem from ring theory (Theorem 1 in [2]).
Theorem A Let
$\mathcal{A}$
and
$\mathcal{\mathcal{B}}$
be prime rings and
$\Phi :\mathcal{A}\to \mathcal{\mathcal{B}}$
a bijective additive mapping such that
$\Phi \left(A\right)\Phi \left(B\right)=0$
for all
$A,B\in \mathcal{A}$
with
$AB=0$
. Suppose that the maximal right quotient ring
$Q(\mathcal{A})$
of
$\mathcal{A}$
contains a nontrivial idempotent
$E$
such that
$E\mathcal{A}\cup \mathcal{A}E\subset \mathcal{A}$
.
i) If
$1\in \mathcal{A}$
, then
$\Phi \left(AB\right)=\lambda \Phi \left(A\right)\Phi \left(B\right)$
for all
$A,B\in \mathcal{A}$
where
$\lambda =\frac{1}{\Phi \left(1\right)}\in Z(\mathcal{\mathcal{B}})$
the center of
$\mathcal{\mathcal{B}}$
. In particular, if
$\Phi \left(1\right)=1$
, the
$\Phi $
is a ring isomorphism from
$\mathcal{A}$
onto
$\mathcal{\mathcal{B}}$
.
ii) If deg(
$\mathcal{\mathcal{B}})\ge 3$
, then there exists
$\lambda \in C(\mathcal{\mathcal{B}})$
, the extended centroid of
$\mathcal{\mathcal{B}}$
, such that
$\Phi \left(AB\right)=\lambda \Phi \left(A\right)\Phi \left(B\right)$
for all
$A,B\in \mathcal{A}$
The definitions and basic properties of the maximal quotient ring and the extended centroid can be found in [1]. A prime ring
$\mathcal{A}$
is called centrally closed if
$C(\mathcal{A})$
is trivial. We need the following characterization of centrally closed prime algebras to prove that every standard operator algebra on
$\mathcal{D}$
is centrally closed (thanks to M. Bres̆ar for this information). Let
$\mathcal{A}$
be a prime algebra over
$\mathbb{C}$
. Then
$\mathcal{A}$
is centrally closed if and only if the following holds:
If
$\mathcal{\mathcal{I}}\subset \mathcal{A}$
is a nonzero ideal and if there is an additive mapping
$F:\mathcal{\mathcal{I}}\to \mathcal{A}$
such that
$$\begin{array}{c}F\left(UX\right)=F\left(U\right)X\text{and}F\left(XU\right)=XF\left(U\right)\end{array}$$ 
(2)

for all
$U\in \mathcal{\mathcal{I}},X\in \mathcal{A}$
, then there is a
$\lambda \in \mathbb{C}$
such that
$F\left(U\right)=\lambda U$
for all
$U\in \mathcal{\mathcal{I}}$
.
In the next lemma we use a modification of the notion of a double centralizer. A pair of (linear or additive) mappings
$L,R:\mathcal{\mathcal{I}}\to \mathcal{A}$
is called a double centralizer if
$$L\left(XY\right)=L\left(X\right)Y,R\left(XY\right)=XR\left(Y\right)\text{and}XL\left(Y\right)=R\left(X\right)Y\text{for all}X,Y\in \mathcal{A}$$
The structure of double centralizers on standard operator algebras on
$\mathcal{D}$
was described in [10] Proposition 3.4 as follows: there is a
$T\in \mathcal{\mathcal{L}}+(\mathcal{D})$
such that
$L\left(A\right)=TA,R\left(A\right)=AT$
. The same proof is valid also for the modified situation descibed above. We use this result to prove the following lemma.
Lemma 2.2.
Every standard operator algebra on
$\mathcal{D}$
is centrally closed.

Proof.
Let
$\mathcal{\mathcal{I}}\subset \mathcal{A}$
be an ideal and
$F:\mathcal{\mathcal{I}}\to \mathcal{A}$
an additive mapping such that (2) is satisfied. Note that
$\mathcal{\mathcal{F}}(\mathcal{D})\subset \mathcal{\mathcal{I}}\subset \mathcal{A}$
. The pair
$(F,F)$
is a double centralizer in the sense described above. Consequently there is a
$T\in \mathcal{\mathcal{L}}+(\mathcal{D})$
such that
$F\left(A\right)=TA,F\left(A\right)=AT$
for all
$A\in \mathcal{\mathcal{I}}$
. So,
$T$
commutes with all operators from
$\mathcal{\mathcal{I}}$
, in particular, with all operators from
$\mathcal{\mathcal{F}}(\mathcal{D})$
. But this implies
$T=\lambda I$
for some
$\lambda \in \mathbb{C}$
. □
Proof of Theorem 2.1 : Apply Theorem A, ii) to get
$\Phi =\frac{1}{\lambda}\Psi $
with a ring isomorphism
$\Psi $
. The structure of ring isomorphisms between standard *operator algebras on (F)domains follows directly from Theorem 3.1 in [
10]
. Namely, there exists a bijective either linear or conjugate linear
$T:\mathcal{D}\to \mathcal{D}$
such that
$\Psi \left(A\right)=TA{T}^{1}(A\in \mathcal{A})$
. This concludes the proof.
3 Mappings preserving orthogonality
In this section we describe the structure of orthogonality preserving mappings on several standard operator algebras. As a corollary we obtain an unbounded version of a result of L. Molnár [7].
Theorem 3.1.
Let
$\mathcal{D}\subset \mathcal{\mathscr{H}}$
be an (F)domain and let
$\mathcal{A}\subset \mathcal{\mathcal{L}}+(\mathcal{D})$
be one of the following standard operator algebras:
a)
$\mathcal{A}=\mathcal{\mathcal{F}}(\mathcal{D})$
, b)
$\mathcal{A}$
is a unital standard *operator algebra, c)
$\mathcal{A}\ne \mathcal{\mathcal{F}}(\mathcal{D})$
,
$\mathcal{A}$
a *ideal of
$\mathcal{\mathcal{L}}+(\mathcal{D})$
.
Assume that
$\Phi :\mathcal{A}\to \mathcal{A}$
is an additive bijection preserving orthogonality in both directions. Then
$\Phi $
has one of the following forms:
i) There exist a nonzero constant
$c$
and operators
$U,V:\mathcal{D}\to \mathcal{D}$
, both either unitary or antiunitary such that
$$\begin{array}{c}\Phi \left(T\right)=cUTV(T\in \mathcal{A})\end{array}$$ 
(3)

or ii) There exist a nonzero constant
$c$
and operators
$U,V:\mathcal{D}\to \mathcal{D}$
, both either unitary or antiunitary such that
$$\begin{array}{c}\Phi \left(T\right)=cU{T}^{+}V(T\in \mathcal{A})\end{array}$$ 
(4)


Proof.
The main step consists in proving that
$\Phi $
preserves rankone operators in both directions. For this we treat cases a) c) separately.
Case a): This can be done as in [5] using the polar decomposition of
$A\in \mathcal{\mathcal{F}}(\mathcal{D})$
.
Remark that in the unbounded case the situation is more complicated. For
$\mathcal{A}\ne \mathcal{\mathcal{F}}(\mathcal{D})$
one can not be sure that the polar decomposition can be performed within
$\mathcal{A}$
(even not within
$\mathcal{\mathcal{L}}+(\mathcal{D})$
).
Case b): Let
$F=\psi \otimes \phi $
be a rankone operator and let
${\mathcal{\mathscr{H}}}_{0}:=lin\{\phi ,\psi \}$
. Now we define an operator
$Q\in \mathcal{A}$
, which has corank one as follows.
If dim
${\mathcal{\mathscr{H}}}_{0}=1$
put
$$Q\chi =\{\begin{array}{cc}\chi & \text{for all}\chi \in {\mathcal{\mathscr{H}}}_{0}^{\perp}\cap \mathcal{D}\\ 0& \text{for all}\chi \in {\mathcal{\mathscr{H}}}_{0}.\end{array}$$
If dim
${\mathcal{\mathscr{H}}}_{0}=2$
put
$$Q\chi =\{\begin{array}{cc}\chi & \text{for all}\chi \in {\mathcal{\mathscr{H}}}_{0}^{\perp}\cap \mathcal{D}\\ \langle {\phi}_{1},\chi \rangle {\psi}_{1}& \text{for all}\chi \in {\mathcal{\mathscr{H}}}_{0}.\end{array}$$
where
${\phi}_{1},{\psi}_{1}\in {\mathcal{\mathscr{H}}}_{0}$
such that
${\phi}_{1}\perp \phi ,{\psi}_{1}\perp \psi $
. To get
$Q\in \mathcal{A}$
it is used that
$I\in \mathcal{A}$
.
Then
${F}^{+}Q=F{Q}^{+}=0$
and therefore
$\Phi (F{)}^{+}\Phi (Q)=\Phi (F)\Phi (Q{)}^{+}=0$
.
That means ran
$\Phi \left(F\right)\perp $
ran
$\Phi \left(Q\right)$
and ran
$\Phi (F{)}^{+}\perp $
ran
$\Phi (Q{)}^{+}$
.
Suppose that
$\Phi \left(F\right)$
has rank larger than one. Then also
$\Phi (F{)}^{+}$
has rank larger than one. Let
${\rho}_{1},{\rho}_{2}\in \text{ran}\Phi \left(F\right),{\rho}_{1}\perp {\rho}_{2}$
and let
${\chi}_{1},{\chi}_{2}\in \text{ran}\Phi (F{)}^{+},{\chi}_{1}\perp {\chi}_{2}$
. Put
${S}_{i}:={\rho}_{i}\otimes {\chi}_{i}$
. Then
${S}_{i}:\mathcal{D}\to \text{ran}\Phi \left(F\right),{S}_{i}^{+}:\mathcal{D}\to \text{ran}\Phi (F{)}^{+}$
and
${S}_{1},{S}_{2}$
are orthogonal.
Obviously
${S}_{i}^{+}\Phi \left(Q\right)={S}_{i}\Phi (Q{)}^{+}=0$
.
Now let
${T}_{i}\in \mathcal{A}$
such that
$\Phi \left({T}_{i}\right)={S}_{i}$
Then
${T}_{1}{T}_{2}^{+}={T}_{1}^{+}{T}_{2}=0$
and
${T}_{i}^{+}Q={T}_{i}{Q}^{+}=0$
, i.e. ran
${T}_{1}\perp $
ran
${T}_{2}$
and ran
${T}_{i}\perp $
ran
$Q$
. But this is a contradiction, because ran
$Q$
has codimension one.
Case c): Here we can argue similar to case b).
Let
$F=\psi \otimes \phi \in \mathcal{\mathcal{F}}(\mathcal{D})$
,
${\mathcal{\mathscr{H}}}_{0}=$
lin
$\{\phi ,\psi \},{\mathcal{D}}_{0}={\mathcal{\mathscr{H}}}_{0}^{\perp}\cap \mathcal{D}.$
Fix an operator
$A\in \mathcal{A}$
with infinitedimensional range. Let
$\left({\phi}_{n}\right)\subset $
ran
$A$
be an orthonormal basis from
$\overline{\text{ran}A}$
and let
$\left({\psi}_{n}\right)\subset {\mathcal{D}}_{0}$
be an orthonormal basis from
${\mathcal{\mathscr{H}}}_{0}^{\perp}$
.
Now we use property (1) of the operators defining the topology
$t$
in
$\mathcal{D}$
. One can find a positive, bounded sequence
$\left({\alpha}_{n}\right)$
such that the operator
$U$
defined by
$U=0$
on (ran
$A{)}^{\perp}$
and
$U{\phi}_{n}={\alpha}_{n}{\psi}_{n}$
(and linear, bounded extension to
$\overline{\text{ran}A}$
) has the following properties.
$U\mathcal{\mathscr{H}}\subset \mathcal{D},{U}^{*}\mathcal{\mathscr{H}}\subset \mathcal{D}$
, so
$U\mathcal{D}\subset \mathcal{D},{U}^{*}\mathcal{D}\subset \mathcal{D}$
and consequently the restriction of
$U$
to
$\mathcal{D}$
(also denoted by
$U$
) belongs to
$\mathcal{\mathcal{L}}+(\mathcal{D})$
. Moreover by construction,
$U\mathcal{D}\subset {\mathcal{D}}_{0}$
and ran
$U$
is dense in
${\mathcal{\mathscr{H}}}_{0}^{\perp}$
. This implies that
$UA\in \mathcal{\mathcal{L}}+(\mathcal{D})$
and ran
$UA$
is dense in
${\mathcal{\mathscr{H}}}_{0}^{\perp}$
. Analogously to case b) we define an operator
$Q$
as follows.
If dim
${\mathcal{\mathscr{H}}}_{0}=1$
put
$$Q\chi =\{\begin{array}{cc}UA\chi & \text{for all}\chi \in {\mathcal{\mathscr{H}}}_{0}^{\perp}\cap \mathcal{D}\\ 0& \text{for all}\chi \in {\mathcal{\mathscr{H}}}_{0}.\end{array}$$
If dim
${\mathcal{\mathscr{H}}}_{0}=2$
put
$$Q\chi =\{\begin{array}{cc}UA\chi & \text{for all}\chi \in {\mathcal{\mathscr{H}}}_{0}^{\perp}\cap \mathcal{D}\\ \langle {\phi}_{1},\chi \rangle {\psi}_{1}& \text{for all}\chi \in {\mathcal{\mathscr{H}}}_{0}.\end{array}$$
where
${\phi}_{1},{\psi}_{1}\in {\mathcal{\mathscr{H}}}_{0}$
such that
${\phi}_{1}\perp \phi ,{\psi}_{1}\perp \psi $
.
As in case b) this operator
$Q$
can be used to prove that
$\Phi \left(F\right)$
is a rankone operator.
So, in any of the cases a) c)
$\Phi $
restricted to
$\mathcal{\mathcal{F}}(\mathcal{D})$
is a bijective additive mapping which preserves rankone operators in both directions. According to [12] we get the following structure of
$\Phi $
on
$\mathcal{\mathcal{F}}(\mathcal{D})$
. Either there exist a ring automorphism
$h:\mathbb{C}\to \mathbb{C}$
and bijective additive mappings
$B,C:\mathcal{D}\to \mathcal{D}$
with the properties
$B\left(\lambda \phi \right)=h\left(\lambda \right)B\phi ,C\left(\lambda \phi \right)=h\left(\lambda \right)C\phi $
such that
$$\Phi (\psi \otimes \phi )=B\psi \otimes C\phi (\phi ,\psi \in \mathcal{D})$$
or there exist a ring automorphism
$k:\mathbb{C}\to \mathbb{C}$
and bijective additive mappings
$B,C:\mathcal{D}\to \mathcal{D}$
with the properties
$B\left(\lambda \phi \right)=k\left(\lambda \right)B\phi ,C\left(\lambda \phi \right)=k\left(\lambda \right)C\phi $
such that
$$\Phi (\psi \otimes \phi )=C\phi \otimes B\psi (\phi ,\psi \in \mathcal{D}).$$
Now the proof can be completed as in the last part of the proof of Theorem 4 in [5]. □
As a corollary we get an unbounded version of Theorem 2 in [7]
Corollary 3.2.
Let
$\mathcal{D}\subset \mathcal{\mathscr{H}}$
be an (F)domain and let
$\mathcal{A}(\mathcal{D})$
be a unital standard operator *algebra. Suppose that
$\Phi :\mathcal{A}\to \mathcal{A}$
is an additive bijection such that i)
${A}^{+}B=0$
if and only if
$\Phi (A{)}^{+}\Phi (B)=0$
and ii)
$A{B}^{+}=0$
if and only if
$\Phi \left(A\right)\Phi (B{)}^{+}=0$
.
Then there exist a nonzero constant
$c$
and operators
$U,V:\mathcal{D}\to \mathcal{D}$
, both either unitary or antiunitary such that
$$\begin{array}{c}\Phi \left(T\right)=cUTV(T\in \mathcal{A}(\mathcal{D}\left)\right).\end{array}$$ 
(5)

If in addition
$\Phi \left(I\right)=I$
then there is a unitary or antiunitary operator
$U:\mathcal{D}\to \mathcal{D}$
such that
$$\begin{array}{c}\Phi \left(T\right)=UT{U}^{+}(T\in \mathcal{A}(\mathcal{D}\left)\right)\end{array}$$ 
(6)


Proof.
If
$\Phi $
satisfies i) and ii) then
$\Phi $
preserves orthogonality in both directions.
Consequently
$\Phi $
has one of the representations (3),(4) of Theorem 3.1.
First we exclude representation (4). Suppose
${A}^{+}B=0$
. Then
$$\Phi (A{)}^{+}\Phi (B)=\overline{c}{V}^{+}A{U}^{+}cU{B}^{+}V=c{}^{2}{V}^{+}A{B}^{+}V.$$
But this expression is not necessarily zero because
$A{B}^{+}$
must not be zero if
${A}^{+}B=0$
. So it remains the form (3).
Now let
$\Phi \left(I\right)=I$
. Then
$\Phi \left(I\right)=cUV=I$
. So
$V={c}^{1}{U}^{1}={c}^{1}{U}^{+}$
Therefore
$$\Phi \left(T\right)=UT{U}^{+}.$$
□
References

K. I. Beidar, W. S. Martindale and A. V. Mikhalev, Rings with generalized identities, Marcel Dekker, Inc., New York 1996.

M. A. Chebotar, W.F. Ke, P.H. Lee and N.C. Wong, Mappings preserving zero products. Studia Math., 155 (2003), 77 – 94.

M. A. Chebotar,W.F. Ke and P.H. Lee, Maps characterized by action on zero products, Pacific J. Math., 216, (2004), 217 – 228.

G. Frobenius, Über die Darstellung der endlichen Gruppen durch lineare Substitutionen, I., Sitzungsber. Preuss. Akad. Wiss. Berlin, (1897), 994 – 1015.

M. Györy and L. Molnár and P. S̆emrl, Linear rank and corank preserving maps on
$B\left(H\right)$
and an application to *semigroup isomorphisms of operator ideals . Linear Algebra Appl., 280 (1998), 253 – 266.

I. Kaplansky, Ring isomorphisms of Banach algebras , Canad. Math. J. 6 (1954), 374 – 381.

L. Molnár, Characterization of additive *homomorphisms and Jordan *homomorphisms on operator ideals, Aequationes Math., 55 (1998), 259 – 272.

K. Schmüdgen, Unbounded Operator Algebras and Representation Theory. Akademie Verlag 1990.

E. Scholz and W. Timmermann, Local derivations, automorphisms and commutativity preserving maps in
${L}^{+}\left(D\right)$
. Publ. RIMS, Kyoto, 29 (1993), 977 – 995.

W. Timmermann, Approximate derivations and isomorphisms in algebras of unbounded operators. Publ. Math. (Debrecen), 63 (2003), 667 – 676.

W. Timmermann, Additive derivations and Jordan derivations on algebras of unbounded operators. Publ. Math. Debrecen, 58 (2001), 717 – 731.

W. Timmermann, Additive mappings in algebras of unbounded operators preserving operators of rank one, Preprint TU Dresden, MATHAN012004 (2004) and arXiv math.OA/4040545.
Institut fur Analysis, Technische Universitat Dresden, D01062 Dresden, Germany Email address : timmerma@math.tudresden.de