The orthogonal group of a Form Hilbert space

Hans A. Keller∗ Herminia Ochsenius †

Abstract
Form Hilbert spaces are constructed over fields that are complete in a non-
archimedean valuation. They share with classical Hilbert spaces the basic
property expressed by the Projection Theorem. However, there appear some
remarkable geometric features which are unknown in Euclidean geometry.
In fact, due to the so-called type condition there are only a few orthogonal
straight lines containing vectors of the same length, so these non-archimedean
spaces are utmost inhomogeneous.
In the paper we consider a typical Form Hilbert space (E, < , >) and we show
that this geometric feature has a strong impact on the group O(E) of all
isometries T : E −→ E and on the lattice L of all normal subgroups of O.
In particular, we describe some remarkable sublattices of L which have no
analogue in the classical orthogonal groups.

1 Preliminaries

1.1 Form Hilbert spaces

In order to describe a non-archimedean normed vector space E over a valued field

K we will introduce G, the range of the valuation | | of K, and X, the range of the
norm of the vectors in E.
The value group G is a multiplicative linearly ordered group. Of particular interest
is the set {Hi : i ∈ I} of its proper convex subgroups (recall that a subgroup H
is convex if for all h ∈ H, h > 1, the interval [h−1 , h] ⊆ H). This set is linearly
S
ordered by inclusion, therefore i∈I Hi is a subgroup of G. We require that there
∗
Supported by Fondecyt No 7050105
†
Supported by Fondecyt No 1050889
2000 Mathematics Subject Classification : Primary 46C99; Secondary 51F25.
Key words and phrases : Krull valuation, isometries, orthogonal group, Form Hilbert spaces,
lattices.

Bull. Belg. Math. Soc. Simon Stevin 14 (2007), 937–946

938 H. A. Keller – H. Ochsenius

exists a denumerable subset N of I such that for each n ∈ N Hn is a proper convex

S
subgroup of G and G = n∈N Hn .
The range of the norm function is a G-module, that is a linearly ordered set X
together with an action of the group G on X which preserves both the order of G
and the order of X, and such that for every x, y ∈ X there is a j ∈ G with jx < y,
see [4]. A crucial object is the orbit Gx for x ∈ X, it is coinitial, and therefore also
cofinal in X, but in general it is not evenly distributed over X.
For a vector v ∈ E the set Gkvk will be called the algebraic type of v. It is clear
that if two vectors v1 and v2 of E have different algebraic types, then v1 and v2 are
(norm-)orthogonal. Therefore a set {vi : i ∈ I} such that Gkvi k = 6 Gkvj k whenever
i 6= j is a (norm-)orthogonal set.
Consider
√ now a Banach vector space E over a valued field K with√value group G.
Let G be the subgroup of G, e the divisible hull of G, defined by G := {γ ∈ G e :
2
√ √
γ ∈ G}. Since G is a subgroup of G, under the multiplication G is a G-module.
Let < , > be an inner product in E (i.e. a symmetric bilinear form such that
∀u ∈ E < u, u >= 0 =⇒ u = 0). We shall say that E is a Form Hilbert space
(FHS) if the following conditions are true,

(i) the inner product induces a non-Archimedean norm on E by

q √
kxk := | < x, x > | ∈ G

(ii) the Projection Theorem is valid in E.

Our specific frame of reference will be the canonical FHS constructed by H. Keller
(1980).

1.2 Construction of a Form Hilbert space

In this section we shortly review the standard construction given in [3] and [2]. In
contrast to these articles we prefer, for the present purpose, to write the value group
multiplicatively.
1. The base field. For i = 1, 2, . . . let Gi =< gi > be an infinite cyclic group
ordered by gir < gis if and only if r < s. The value group G is the direct sum G =
n1 n2 nr
i∈N Gi . That is, g ∈ G if and only if g = (g1 , g2 , . . . , gr , 1, . . .) for some r ∈ N,
L

ni ∈ Z. It is ordered antilexicographically, thus g = (g1 , g2n2 , . . . , grnr , 1, . . .) > 1 if

n1

nr > 0.
Next, let F = R (Xi )i∈N be the field of all rational functions in the variables
X1 , X2 , . . . with real coefficients. There is a uniquely determined Krull valuation
| | : F −→ G ∪ {0} for which

(a) | | is the trivial valuation on R.

(b) |Xn | = (1, . . . , 1, gn , 1, . . . ) for n = 1, 2, . . .

To finish the construction of the base field we define K to be the completion of

(F, v) by means of Cauchy sequences. Notice that the field K (with the extended
valuation) is far from being algebraically complete.
The orthogonal group of a Form Hilbert space 939

Remark. Let A be the valuation ring and k ∼ = R the residual field of the valuation.
We denote by Π0 : A −→ k the canonical projection.
2. The space. Let E be the space of all sequences x = (ξi )i∈N0 ∈ K N0 for which
the series ∞ 2
P
i=0 ξi Xi converges in the valuation topology, where X0 := 1. Operations
in E are of course componentwise. We define an inner product < , >: E × E −→ K
by
∞
X
< x, y >:= ξi ηi Xi for x = (ξi )i , y = (ηi )i ∈ E.
i=0

This symmetric bilinear form < , > is anisotropic. As usual we say that x, y ∈ E
are (form-)orthogonal, x ⊥ y, if < x, y >= 0, and for a subspace U ⊂ E we define
its (form-)orthogonal complement by U ⊥ := {x ∈ E : x ⊥ u for all u ∈ U }.
Next the assignment
q √
x 7−→ kxk := | < x, x > | ∈ G

is a non-Archimedean norm (see [2], [4]).

The most important properties of the space (E, < , >) thus constructed are sum-
marized in the following result.

Theorem 1. Let (E, < , >) be as above. Then

1. E is complete in the norm topology.
2. A subspace U ⊆ E is topologically closed if and only if it is orthogonally closed,
that is, U = cl(U ) if and only if U = U ⊥⊥ .
3. The Projection Theorem is valid in E : U = cl(U ) ⇒ E = U ⊕ U ⊥ .
Therefore E is a Form Hilbert space.

For a proof we refer to [3].

3. The canonical base. For n = 0, 1, 2, . . . we put en := (0, . . . , 0, 1 , 0, . . . , )
where 1 is in position n + 1. Then en ⊥ em for n 6= m, < en , en >= Xn and since
the algebraic types of the vectors ei are all different, {e0 , e1 , . . . , en , . . . } is a (norm-)
orthogonal base, called the canonical base of (E, < , >). Therefore every vector
x ∈ E can be written uniquely as
∞
X
x= ξi ei .
i=0

1.3 The orthogonal group

Definition. Let E be a K vector space, < , > be an inner product on E. A linear

operator T : E −→ E is called a (form-)isometry if < T x, T x > = < x, x > for all
x ∈ E.
Remark. In a Form Hilbert space a form-isometry preserves the norm, that is
kT xk = kxk. Hence it is also a norm-isometry.
From now on, the term isometry will be used instead of form-isometry.
940 H. A. Keller – H. Ochsenius

Definition. The orthogonal group of a space E is the group O(E) of all isometries
on E.
There is no lack of isometries. In any inner product space (E, < , > over a field K
with char K 6= 2 every vector u 6= 0 induces an isometry τu by
< x, u >
τu (x) = x − 2 u
< u, u >
called the reflection with respect to the hyperplane Hu := {w ∈ E : w ⊥ u}. It
is immediate that τu (u) = −u, τu (x) = x for every x ∈ Hu . Therefore (τu )2 = τu ,
hence τu is an involution.
The famous theorem of Cartan Dieudonné states that in an inner product space
(E, < , >) with dim E = n, every isometry is a product of at most n hyperplane
reflections. This is no longer true in infinite dimensional vector spaces. Indeed, since
τu is the identity in the hyperplane Hu , the isometry σ = −Id, with Id the identity
in E, cannot be written as a finite product of hyperplane reflections. The study of
isometries when dim E = ∞ turns out to be a difficult and challenging problem. A
variety of outstanding results were obtained by H.Gross and his school in Zürich,
above all in a purely algebraic setting of spaces of countable dimension.

1.4 Lattices, an overview

We have started an analysis of O(E) for the FHS space described above. We have
identified and described some relevant sublattices of the lattice of normal subgroups
of O(E). Therefore we give here a brief summary of the definitions and theorems
we shall use, (see [1]).
Definition. Let P be a non-empty ordered set and S ⊆ P . An element x ∈ P is
an upper (respectively lower) bound of S if s ≤ x (respectively x ≤ s) for all s ∈ S.
The least upper bound of S, if it exists, is called the supremum of S and denoted
W V
by S. Dually the infimum of S, S, is the greatest lower bound of S.
Remark. If S = {x, y} then x ∨ y denotes its supremum and x ∧ y its infimum.
Definition. Let L be a non-empty ordered set. If x ∨ y and x ∧ y exist for all x, y
in E, then L is called a lattice. If S and S exist for all S ⊆ L, then L is called
W V

a complete lattice.
Theorem 2. Let L be a non-empty ordered set that has a greatest element 1. If
V
S exists for every non-empty subset of L, then L is a complete lattice.
Corollary 3. Let O be a group and L be the set of all its normal subgroups ordered
by inclusion. Then L is a complete lattice.

Definition. A lattice L is called distributive if for all x, y, z ∈ L,

x ∧ (y ∨ z) = (x ∧ y) ∨ (x ∧ z).
Equivalently, x ∨ (y ∧ z) = (x ∨ y) ∧ (x ∨ z) for all x, y, z ∈ L.
Definition. A lattice L is called modular if for all x, y, z ∈ L,
z ≤ x =⇒ x ∧ (y ∨ z) = (x ∧ y) ∨ z.
The orthogonal group of a Form Hilbert space 941

Theorem 4. If L is a distributive (modular) lattice then every sublattice of L is

also distributive (modular).

Remark. The lattice of normal subgroups of a group O is modular, but in general

not distributive.
Definition. A Boolean algebra is a distributive lattice B with a least element
0B and a greatest element 1B together with a map B −→ B , a 7−→ a0 such that
a ∨ a0 = 1B and a ∧ a0 = 0B for all a ∈ B.
Remark. a0 is called the complement of a. It is clear that (a0 )0 = a.
Definition. Let L1 and L2 be two lattices. A map ϕ : L1 −→ L2 is said to be
order preserving if for all x, y in L1 we have that x ≤ y =⇒ ϕ(x) ≤ ϕ(y). If ϕ
is an order preserving bijection and ϕ−1 is also an order preserving map, then ϕ is
called an isomorphism of lattices.

2 The orthogonal group O of E

Let E be the FHS described in 1.2 and O its orthogonal group. We begin with a
crucial result.
∞
X
Theorem 5. Let T : E −→ E be an isometry and T (ek ) = ξik ei where ξik =
i=0
< ei , T (ek ) >< ei , ei >−1 . Then Π0 (ξkk ) = ±1.

Proof. First we notice that Xk =< ek , ek >=< T (ek ), T (ek ) >. Since the types of the
P 2 2
vectors {ei : i ∈ N0 } are all different, | < T ek , T ek > | = | ξik Xi | = max{|ξik Xi | :
2 2 2
 i | forall i 6= k. In addition |ξkk | = 1, hence
i ∈ N0 }, thus |Xk | = |ξkk | |Xk | > |ξik X
2 Xi 2 Xi
1 > |ξik Xk | whenever i 6= k, hence Π0 ξik Xk
= 0.
2 P 2 P 2 Xi
But Xk =< T (ek ), T (ek ) >= ξkk Xk + i6=k ξik Xi , so that 1 = ξkk + i6=k ξik Xk
.
2
Therefore Π0 (ξkk ) = 1 and it follows that Π0 (ξkk ) = ±1 as claimed.

Notation. From now on, if T ∈ O and k ∈ N0 we will write T (k) for

!
< ek , T (ek ) >
T (k) := Π0 .
< ek , ek >

Definition. For all k ∈ N0 we define Nk := {T ∈ O : T (k) = 1}.

Lemma 6. Let S, T ∈ O, k ∈ N0 , then (ST )(k) = S(k)T (k).
P P P
Proof. Let S(ek ) = i ηik ei , T (ek ) = i ξik ei and (ST )(ek ) = i αik ei . Write
P P
S( i6=k ξik ei ) = δk ek + i6=k δi ei , and we have that
X X X
αkk ek + αik ei = ξkk ηkk ek + ξkk ηik ei + δk ek + δi ei ,
i6=k i6=k i6=k

so
αkk = ξkk ηkk + δk . (1)
942 H. A. Keller – H. Ochsenius

Now

|δk2 Xk | ≤ max{|δk2 Xk |, | δi2 Xi |} = |δk2 Xk + δi2 Xi | = | < S(

X X X X
ξik ei ), S( ξik ei ) > |
i6=k i6=k i6=k i6=k

but S is an isometry, hence |δk2 Xk | ≤ | < i6=k ξik ei , i6=k ξik ei > | = | i6=k ξik
2
Xi |
P P P
2
and by the proof of Theorem 5, | i6=k ξik Xi | < |Xk |. Hence |δk | < 1 and |δk | < 1,
P

therefore Π0 (δk ) = 0.
From (1) we obtain now that Π0 (αkk ) = Π0 (ξkk ηkk + δk ) = Π0 (ξkk )Π0 (ηkk ) and
(ST )(k) = S(k)T (k).

Theorem 7. Nk is a normal subgroup of O such that (O : Nk ) = 2.

Proof. Define ϕ : O −→ {1, −1} by ϕ(T ) = T (k). Lemma 6 ensures that ϕ is a

group homomorphism and clearly ker ϕ = Nk .

2.1 The sublattice generated by the subgroups Nk

For each A ⊆ N0 we define \

NA := Nk
k∈A

and N := {NA : A ⊆ N0 }.

Lemma 8. NN0 6= {Id}, where Id is the identity mapping on E.

Proof. For each k ∈ N0 we have that O/Nk is an abelian group, hence Nk contains
Ω, the commutator subgroup of O. Therefore Ω ⊆ k∈N0 Nk = NN0 . But clearly
T

Ω 6= {Id} since O is not an abelian group.

Theorem 9. N is a sublattice of the lattice of normal subgroups of O. It is a

Boolean algebra with 0N = NN0 and 1N = O.

Proof. For every A ⊆ N0 we have that NA is a normal subgroup of O, since it is the

intersection of a family of normal subgroups. We shall prove that for A, B subsets
of N0 the subgroups NA ∧ NB = NA ∩ NB and NA ∨ NB = NA NB belong to N .
First we have
NA ∧ NB = {T ∈ O : T ∈ NA and T ∈ NB }
= {T ∈ O : ∀i ∈ A (T (i) = 1) and ∀j ∈ B (T (j) = 1)}
= {T ∈ O : ∀k ∈ A ∪ B (T (k) = 1)}
= NA∪B ∈ N .

Next we prove that NA NB = NA∩B . Let S ∈ NA , T ∈ NB , then S(k) = 1 = T (k)

for all k ∈ A ∩ B. By Lemma 6 (ST )(k) = 1. Therefore NA NB ⊆ NA∩B .
Now let R ∈ NA∩B and put C := {k ∈ N0 : R ∈ Nk }. We define P ∈ O by
(
−ei if i ∈ B \ C
P (ei ) =
ei if i ∈
/ B\C
The orthogonal group of a Form Hilbert space 943

It is readily seen that P = P −1 , moreover P ∈ NA . In fact, notice that A ∩ B ⊆ C,

from which it follows that A ∩ (B \ C) = ∅, hence if k ∈ A then k ∈ / B\C
and therefore P (ek ) = ek . On the other hand P R ∈ NB , since by construction
P (k) = R(k) for all k ∈ B. Thus R = P −1 (P R) ∈ NA NB and NA∩B = NA NB ∈ N .
Therefore N is a sublattice of the lattice of normal subgroups of O as claimed.
Clearly the least element of N is NN0 and the greatest one is O = N∅ . It is also
clear that N is distributive.
Now we define for NA ∈ N its complement NA 0 by NA 0 = NN0 \A . It is obvious
that (NA 0 )0 = NA , NA ∨ NA 0 = 1N and NA ∧ NA 0 = 0N . Therefore N is a Boolean
algebra.

Lemma 10. N is a complete lattice.

Proof. Since N has a greatest element N∅ = O, it is enough to prove that if S =

{NSi : Si ⊆ N0 , i ∈ I} is a non-empty collection of elements in N then S ∈ N .
V

We contend that S = NZ with Z = i∈I Si .

V S

In fact for every T ∈ O we have

T ∈ NZ ⇔ T (m) = 1 for all m ∈ Si ⇔ T ∈ NSi for all i ∈ I ⇔ T ∈ N Si .

S T
i∈I i∈I

Corollary 11. N is isomorphic to the Boolean algebra (P(N), ⊆op ) = P where

A ⊆op B if and only if A ⊇ B.

Proof. The map ϕ : P −→ N defined by ϕ(A) = NA is clearly surjective. It is also

injective, because if A, B ∈ P with A 6= B we may assume that A * B and so there
exists m ∈ N0 with m ∈ A and m ∈ / B. The isometry Im defined by Im (ek ) = ek if
k 6= m, Im (em ) = −em belongs to NB but not to NA . It is a direct verification that
ϕ as well as ϕ−1 are lattice homomorphisms.
Remark. By the previous result the lattice of normal subgroups of the orthogonal
group contains an isomorphic copy of P(N). The question as to whether it contains
another another copy of this Boolean algebra is still open.

2.2 The lattice N ∗

The subgroup J = {Id, −Id} is normal in O but does not coincide with any NA
when A ⊆ N0 . We will study here the sublattice N ∗ generated by N ∪ {J} in L.
We denote by NA∗ the element NA ∨ J. Explicitly

NA∗ = {T ∈ O : ∀m ∈ A T (m) = 1} ∪ {S ∈ O : ∀m ∈ A S(m) = −1}

.
The lattice N ∗ is modular since it is a sublattice of L, but it is not a distributive
lattice. In fact, for every k ∈ N Nk ∧J = Nk ∩J = {Id}. Hence (N1 ∧J)∨(N2 ∧J) =
{Id} but (N1 ∨ N2 ) ∧ J = Nφ ∧ J = O ∧ J = J.

Lemma 12. For all A, B ⊆ N0 we have NA∗ ∨ NB = NA∗ ∨ NB∗ = NA∩B

∗
.
944 H. A. Keller – H. Ochsenius

Proof.
NA∗ ∨ NB = J ∨ NA ∨ NB = J ∨ NA∩B ∗
= NA∩B
∗
NA∗ ∨ NB∗ = J ∨ NA ∨ J ∨ NB = J ∨ NA ∨ NB = NA∩B

Remark. If A ∩ B = ∅, N∅∗ = N∅ = O.
The suprema are shown in the following diagram.
∗
NA∩B
q
H
NA∗  ∗
 HH
H NB
q
 Hq
q
HH
 N A∩B HH
q 
 Hq
NA NB

The computation of infima is more complex.

Let A, B ⊆ N0 , r, s ∈ {0, 1, −1}. We introduce the following notation:
(A = r ∧ B = s) := {T ∈ O : ∀m ∈ A T (m) = r and ∀m ∈ B T (m) = s}
where T (m) = 0 means that no restriction is placed on T (m). Then,
NA = (A = 1 ∧ B = 0)
NB = (A = 0 ∧ B = 1)
NA∪B = (A = 1 ∧ B = 1) = NA ∩ NB
NA∗ = (A = 1 ∧ B = 0) ∪ (A = −1 ∧ B = 0)
NB∗ = (A = 0 ∧ B = 1) ∪ (A = 0 ∧ B = −1)
We obtain by direct computation
NA∗ ∩ NB∗
= [(A = 1 ∧ B = 0) ∪ (A = −1 ∧ B = 0)] ∩ [(A = 0 ∧ B = 1) ∪ (A = 0 ∧ B = −1)]
= [(A = 1 ∧ B = 0) ∩ (A = 0 ∧ B = 1)] ∪ [(A = 1 ∧ B = 0) ∩ (A = 0 ∧ B = −1)]∪
[(A = −1 ∧ B = 0) ∩ (A = 0 ∧ B = 1)] ∪ [(A = −1 ∧ B = 0) ∩ (A = 0 ∧ B = −1)].
Therefore
NA∗ ∩ NB∗ =
(A = 1 ∧ B = 1)∪(A = 1 ∧ B = −1)∪(A = −1 ∧ B = 1)∪(A = −1 ∧ B = −1).
In the same way we compute
NA∗ ∩ NB
= [(A = 1 ∧ B = 0) ∪ (A = −1 ∧ B = 0)] ∩ (A = 0 ∧ B = 1)
= [(A = 1 ∧ B = 0) ∩ (A = 0 ∧ B = 1)] ∪ [(A = −1 ∧ B = 0) ∩ (A = 0 ∧ B = 1)].
Therefore,
NA∗ ∩ NB = (A = 1 ∧ B = 1) ∪ (A = −1 ∧ B = 1).
Remark. If A ∩ B 6= ∅ then the values of A and B must be the same. Therefore,
NA∗ ∩ NB∗ = (A = 1 ∧ B = 1) ∪ (A = −1 ∧ B = −1) = NA∪B
∗
.
In addition,
NA∗ ∩ NB = NA∪B = NA ∩ NB = NA ∩ NB∗ .
The following diagrams depict the situation.
The orthogonal group of a Form Hilbert space 945

1. If A ∩ B = ∅,
NA∗ N∗
q ∗ ∗ qB
P PPNA ∧ NB 
q 
PP 

NAq qN
@
@ B
qN ∗ @
A∪B@
q q
@
NA ∧ NB∗ PPP N ∗ ∧ NB
P P
q  A
NA∪B
Note that there are elements in L∗ that do not belong to {NC∗ : C ⊆ N0 }.
They are NA ∧ NB∗ , NA∗ ∧ NB and NA∗ ∧ NB∗ .

2. If A ∩ B 6= ∅ the elements mentioned above vanish and the diagram collapses

to the following one:
NA∗ N∗
qP ∗ qB
PP NA∪B 
PPq 

qP q
NAPPPPq NB
NA∪B

2.3 A characterization

Theorem 13. For any i ∈ N0 let Ci = {1, −1} be the multiplicative group of two
elements and i∈N0 Ci their direct product. Then i∈N0 Ci ∼
= O/NN0 .
Q Q

Proof. Let ϕ : O −→ i∈N0 Ci be the map defined by T 7−→ (T (i))i∈N0 . By Lemma

Q

6 it is a group homomorphism. It is surjective, since for ε = (εi ) ∈ Ci the

Q

involution Tε defined by Tε (ej ) = εj is an isometry. Clearly ker ϕ = NN0 so the First

Isomorphism Theorem gives the result.
Definition. L0 := [NN0 , O] is the interval of L of all the normal subgroups of O
which contain NN0 .
Q
Corollary 14. The (complete) lattice of all subgroups of i∈N0 Ci is isomorphic to
the (complete) lattice L0 .

Proof. The map B 7−→ ϕ−1 (B) = {T ∈ O : ϕ(T ) ∈ B} is a bijection between

the subgroups of Ci and the subgroups of O which contain NN0 . This bijection
Q

is an order isomorphism, that is, if B1 , B2 are subgroups of Ci , A1 := ϕ−1 (B1 ),

Q

A2 := ϕ−1 (B2 ) then A1 ⊆ A2 ⇔ B1 ⊆ B2 . Therefore ϕ is a lattice isomorphism.

Corollary 15. Every subgroup of O which contains NN0 is a normal subgroup of O.
From the lattice N ∗ we remove the subgroup J = {1, −1} as well as the subgroup
{Id}. There remains then the interval [NN0 , O] of L∗ . We shall call it the lattice
N ∗∗ .
N ∗∗ is a sublattice of L0 . In fact, the latter is the completion of the former.
946 H. A. Keller – H. Ochsenius

Theorem 16. Let A be a subgroup of O such that NN0 ⊆ A. Then A is the supre-
mum of a family of elements of N ∗∗ .

Proof. Let ε = (εi )i∈N0 ∈ i∈N0 Ci ; set Aε = {i : εi = 1}, Bε = {i : εi = −1}. If

Q

Hε = {ε, 1} ≤ Ci then ϕ−1 (Hε ) = NAε ∧ NB∗ ε . In fact

Q

T ∈ NAε ∩ NB∗ ε ⇔ T (i) = 1 for all i ∈ Aε and T (i) = T (k) if j, k ∈ Bε

⇔ (T (i) = 1 for all i ∈ N0 ) or (T (i) = 1 if i ∈ Aε
and T (i) = −1 if i ∈ Bε )
⇔ ϕ(T ) = 1 or ϕ(T ) = ε
⇔ ϕ(T ) ∈ Hε .

Now, let A ≤ O with NN0 ⊆ A and H = ϕ(A) ∈ Ci . Assume that {αk :∈ k}

Q

Hk with Hk = {1, αk }. By Corollary 15,

W
is a set of generators of H, then H =
k∈K
A = ϕ−1 ( ϕ−1 (Hk ) = (NAαk ∧ NB∗ α ) is the supremum of a family
W W W
Hk ) =
k
k∈K k∈K k∈K
∗∗
of elements of N .

References
[1] G. Birkhoff, Lattice theory, Coll. Publ., XXV, American Mathematical Society,
Providence, R.I. (1967).

[2] H. Gross and U.M. Künzi, On a class of orthomodular quadratic spaces.

L’Enseignement Math. 31 (1985), 187-212.

[3] H. Keller, Ein nicht-klassischer Hilbertscher Raum. Math. Z. 172 (1980), 41-49.

[4] H. Ochsenius and W. Schikhof, Banach spaces over fields with an infinite rank
valuation. In p-Adic Functional Analysis, Lecture Notes in pure and applied
mathematics 207, edited by J. Kakol, N. De Grande-De Kimpe and C. Perez-
Garcia. Marcel Dekker (1999), 233-293.

H. Keller : Hochschule Technik+Architektur Luzern,

CH-6048 Horw, Switzerland.
email : [email protected]

H. Ochsenius : Facultad de Matemáticas,

Pontificia Universidad Católica de Chile,
Casilla 306 - Correo 22, Santiago, Chile.
email : [email protected]