You are on page 1of 39

Revisiting the works of Mihail Benado

Sergiu Rudeanu, Dragos Vaida


University of Bucharest
Faculty of Mathematics and Informatics
Str. Academiei No. 14, 010014 Bucharest, Romania
srudeanu@yahoo.com,corresponding author; dvaida@rdslink.ro
May 20, 2010
Abstract
Mihail Benado was a Romanian algebraist who worked in lattice the-
ory, group theory and interconnections. He is the creator of the theory
of multilattices, which nowadays has important applications in computer
science. The present paper tries to sketch an overview of Benados works,
which comprise not only multilattices, but several other important lines of
research, which deserve to be recalled to the contemporary mathematical
community.
Key words: multilattice, regular product, Schreier renement theorem,
Dedekind quadrilateral, division semigroup, partially additive semantics,
monotone connection, semidistance, diametric space
MSC 06-02, 06A06, 06B99
Mihail Benado was a Romanian algebraist who worked in lattice theory,
group theory and interconnections. He is the creator of the theory of multi-
lattices, which nowadays has important applications in computer science. The
present paper tries to sketch an overview of Benados works, which comprise not
only multilattices, but several other important lines of research, which deserve
to be recalled to the contemporary mathematical community.
Benado
1
was born on July 5, 1920 in Bucharest. He was a scholar at the ele-
mentary school of the Jewish Community, then at high school Matei Basarab,
from which he graduated in 1940. Because of the anti-Semitic law Numerus
Clausus in vigour in Romania of those days, Benado could not become a stu-
dent of a State University, but he attended the courses of a private university.
In 1944 he became a student of the Faculty of Sciences (including divisions of
mathematics, physics and chemistry) of the University of Bucharest from which
he graduated in 1948, the very year in which the Faculty of Mathematics and
1
The biographical data of Benado are taken from the monograph by G.St. Andonie [GSA],
vol.III, 36.
1
Physics was created as a separate faculty. From 1948 to 1962 Benado held teach-
ing positions at this faculty, and from 1950 to 1962 he was also a researcher at
the Institute of Mathematics of the Romanian Academy. In January 1962 he
resigned both positions.
Unfortunately, in the sixties Benado gradually isolated himself from the
mathematical community of Romania and, except a rather short period in which
he was with the 3-year Pedagogical Institute of Bucharest, there are no infor-
mations about Benados life after 1964, except that he emigrated to Israel.
2
Mihail Benado was a pupil of Dan Barbilian, who directed Benados interest
towards group theory and lattice theory, the latter being at that time a young
branch of mathematics. The fact that Barbilian was both an algebraist and a
geometer had a strong inuence in the scientic career of Mihail Benado.
Two major research themes of Benado were the theory of regular products
of operator groups initiated by Oleg Golovin and the Schreier-Zassenhaus and
Jordan-Holder renement theorems in group theory and their generalizations to
modular lattices. Benado introduced new concepts related to regular products,
generalizing several results of Golovin and generalized this theory to the lattice-
theoretical level. The main concern (which can be detected even in his rst
paper [1] dealing with Galois theory) of Benados ample contribution to the
eld of renement theorems was the passage from modular lattices to arbitrary
lattices and even to partially ordered sets and to his multilattices.
The theory of multilattices is in fact the most important creation of Mihail
Benado. He has relaxed the concept of a lattice by requiring only the existence
of certain minimal upper bounds and of certain maximal lower bounds instead
of the least upper bound and the greatest lower bound.
The spirit of multilattices is also present in other papers of Benado. Thus, he
introduced the monotone connections of types I, II and III, which are similar to
but dierent from the conventional Galois connections, and characterized semi-
lattices and semimultilattices in terms of these monotone connections. Benado
also generalized several results of Glivenko and Barbilian on metric lattices. The
works of Mihail Benado comprise two other big constructions. One of them is
the theory of diametric spaces, an attempt to develop an axiomatic theory of the
distance between two bodies (Korper). The other construction is a general the-
ory of partially ordered sets, even more general than the theory of multilattices.
Unfortunately, these two projects have remained unnished.
Mihail Benado, like his master Dan Barbilian, had a complex and rather
contradictory human personality. To a certain extent, this had consequences
on his work. Thus e.g. he has no written course, although he was an excel-
lent teacher. In particular the mathematical orientation of the present authors
has been strongly inuenced by his teaching and research
3
Also, we have no
information about a possible PhD obtained by Benado, although as we hope
to prove in this paper and a next one he was a highly inventive author of
mathematical papers. Benado wrote no book, yet he is the author of several
2
Yet the paper [58] still mentions Benados address in Bucharest.
3
In particular Benado used his paper [16] as the support of a one-semester free course
taught to the second author.
2
comprehensive theories which deserve to be better known by the mathematical
community.
The papers of Mihail Benado are written in French (31), Romanian (17),
German (8) and Russian (3). He has no paper in collaboration, but he corre-
sponded with several mathematicians. Among them the most active were Jan
Jakubk and Milan Kolibiar, who proved the existence or non-existence of math-
ematical objects having certain properties that interested Benado and pointed
out a few minor gaps subsequently corrected by Benado; cf. footnotes in papers
[24], [32], [38], [51], [53], [54]. On the other hand, the rst works on multilattices
of Jakubk and Kolibiar appeared starting with the years 1956-1958, as shown in
this paper. A full report on the subsequent development of multilattice theory
and its applications is planned for a future paper of ours.
The present paper is structured as follows. Part A: theory of regular products
of groups. Part B: theory of Schreier-Zassenhaus and Jordan-Holder theorems.
Part C: multilattices and applications, with a link to contemporary applications
in computer science. Part D: the other papers of Benado. Part E contains
two bibliographies: the papers of Mihail Benado
4
and a selective bibliography
of the sources of Benado. Part E is followed by paragraphs Further research,
Acknowledgements and the references of the present paper.
Parts A, B and C can be read independently of each other. Part D relies on
Part C.
We refer to the papers of Benado by simple numerical codes [1], . . . ,[59].
The sources of Benado are cited in the form Author [year], e.g. O.N. Golovin
[1950], [1951a,b]. The other citations are listed as References under suggestive
acronyms; for instance, [Bir] and [Bir67] indicate a paper and a book by G.
Birkho.
PART A
The theory of regular products of groups
In this Section we survey the papers [21], [30], [31], [32], whose main results
were announced in [25], [27], [28], [29], and the papers [37], [41]. In this series
of articles M. Benado generalizes the theory of regular products of groups, in-
troduced by O.N. Golovin [1950], [1951a,b]; cf. O.N. Golovin and N.P.Goldina
[1951].
The works of Golovin and Benado refer to operator groups and permitted
subgroups; yet for the sake of convenience in the sequel we refer simply to groups
and subgroups.
We begin with some notation and other prerequisites.
4
To the best of our knowledge, the list of Benados papers is complete and he has written
no book.
3
Let G be a group. The subgroup generated by a family G
i

iI
of sub-
groups is denoted by
_
iI
G
i
. The singleton subgroup is denoted by E. If H is
a subgroup, the notation H means the normal subgroup generated by H. The
commutator of two elements x, y is the element xy = x
1
y
1
xy. The commu-
tator of two subgroups K, H of G is the subgroup K H generated by all the
commutators k h with k K and h H. The commutator K H is a normal
subgroup of KH; in particular GH is a normal subgroup of G. For any sub-
group H of G, by setting G
1
H = GH , G
n+1
H = G(G
n
H) (n = 1, 2, . . . ),
one obtains a decreasing sequence G
1
H G
n
H . . . of normal sub-
groups of G.
Now let G = G
1
G
n
be a nite direct product of groups. For each
i = 1, . . . , n, let H
i
= (g
1
, . . . , g
n
) G [ g
j
= e for j ,= i. It is easy to see that
the H
i
s are normal subgroups of G which generate G and for each i = 1, . . . , n,
the intersection between H
i
and the subgroup of G generated by the other H
j
s
reduces to E. Therefore, following A.G. Kurosh [1953], a group G is a direct
product of a family G
i

iI
of subgroups, written
G =

iI
G
i
,
provided the following conditions are satised:
each G
i
is a normal subgroup of G ;
G =
_
iI
G
i
= ;
(i I) G
i

_
jI{i}
G
j
= E .
O.N. Golovin has introduced the following more general concept. A group
G is a regular product of a family G
i

iI
of subgroups, written
G =

iI
G
i
provided the following conditions are satised;
G =
_
iI
G
i
,
(i I) G
i

_
JI{i}
G
j
= E.
Golovin has also dened the rst normalized commutator of the subgroups
G
i
as
K
1
=

i,jI;i=j
(G
i
G
j
) ,
and the n-th step normalized commutators are dened inductively:
K
n+1
= G K
n
(n = 1, 2, . . . ) ;
they form a decreasing sequence of normal divisors of G.
Following Benado, a subgroup H of a group G is called an n-step normal
subgroup if G
n
H H. The conventional normal subgroups coincide with the
1-step normal subgroups, but for n > 1 there exist n-step normal subgroups
4
that are not normal subgroups. Clearly each n-step normal subgroup is also an
(n + 1)-step normal subgroup. For each n, the n-step normal subgrups form a
complete sublattice of the lattice of all subgroups of G.
Also, a regular product

iI
G
i
is called by Benado an n-step direct product,
written
(n)

iI
G
i
,
provided its factors are n-step normal subgroups: G
n
G
i
G
i
(i I).
Here are a few sample results linking these concepts [31].
A regular product is an n-step direct product if and only if K
n
= E. The
n-step direct product is associative, to the eect that if G =

(n)
iI
G
i
and each
G
i
=

(n)
jJ
i
G
ij
(i I), then G
i
=

(n)
iI,jJ
i
G
ij
. The subgroup property says
that if G =

(n)
iI
G
i
and H
i
are subgroups of G such that H
i
G
i
(i I), then
_
iI
H
i
=

(n)
iI
H
i
. A non-trivial free group cannot be expressed as an n-step
direct product. If G = G
1
G
2
and G
1
(or G
2
) is an n-step normal subgroup,
then K
n
G
1
(or K
n
G
2
).
The above theorems (previously known for n = 1) and other similar results
generalize theorems due to Baer, Levi and Golovin, and solve a problem raised
by Malcev and Golovin.
Benado also found [21] a condition equivalent to the complete associativity
of a regular product, meaning that if G =

iI
G
i
then for any partition
I = I

, I

= we have G = (

iI

G
i
) (

iI

G
i
).
On the other hand, the concept of a direct product of subgroups can be lifted
to a lattice-theoretical level, as Kurosh rst did. He introduced the concept of a
completely modular latice, meaning a complete lattice L which satises the fol-
lowing strengthening of the modular law: for every two families x
i

iI
, y
i

iI
of elements of L, if x
i
y
j
for all i, j I with i ,= j, then
(

iI
x
i
) (

iI
y
i
) =

iI
(x
i
y
i
) .
Taking a two-element set I and y
1
= x
1
x
2
, one sees that every completely
modular lattice is indeed modular, but there exist modular complete lattices
that are not completely modular. Kurosh dened direct decompositions
5
a =
.
_
iI
a
i
in a completely modular lattice, meaning that the following conditions are ful-
lled:
a =
_
iI
a
i
,
(i I) a
i
a

i
= 0, where a

i
=
_
jI{i}
a
j
.
So a direct product of groups

iI
G
i
is a direct decomposition of the greatest
element G of the completely modular lattice of normal divisors of G.
5
which he rather improperly called direct sums
5
The endomorphisms of a direct decomposition 1 =
.
_
iI
a
i
of the unit 1 of L
are the mappings
i
: L L dened by

i
(x) = a
i
(x a

i
) (i I) .
These functions are in fact just order-preserving maps, but when applied to the
lattice of normal subgroups they yield a system of group endomorphisms.
In order to study regular products of groups, Benado uses a similar tech-
nique, but unlike Kurosh, his axiomatics has in view the lattice of all subgroups
of a group, which is not modular. So, in his lattice-theoretical approach, Benado
works with a complete lattice in which he needs a closure operator as a model
for the passage from a subgroup to the normal subgroup generated by it. Based
on works by O. Ore, V. Kornek and D. Barbilian, Benado [30] endows L with a
binary relation N subject to the following axioms. Whenever aNb one requires
that a b and
a x y imply x (b y) = (x b) y ,
a y b and a x imply y (x b) = (y x) b ,
a x implies xN(x b) ,
a x and xNy imply (b x)N(b y) :
further, 1N1, 1N0 and the set L
N
= a L [ 1Na is a complete sublattice
of L. The lattice L is then called a Lie lattice. The idea of this axiomatization
is that L should represent the lattice of subgroups of G, the relation ANB
meaning that B is a normal subgroup of the subgroup A, the element 1 L
represents the group G, so that L
N
is an abstract version of the sublattice of
normal subgroups of G. It should be noted that L
N
is a modular lattice by
the rst axiom above, but need not be a completely modular lattice. Other
instances of Lie lattices are the lattice of subalgebras of a Lie algebra and the
lattice of subrings of a ring [27]; in the latter case ANB means that B is a
two-sided ideal of the subring A.
Benado works with the closure operator associated with L
N
. So he denes
[30] a regular product
a =

iI
a
i
by the following conditions:
a =
_
iI
a
i
,
(i I) a
i
(a a

i
) = 0 .
The Fitting endomorphisms associated with a regular product are the order
endomorphisms
i
: L L (i I) dened by

i
(x) = a
i
(x (a a

i
)) (i I) .
The Fitting endomorphisms characterize the regular product [30] to the eect
that they satisfy six conditions with the property that conversely, any system

iI
of order endomorphisms satisfying these conditions is the system of Fit-
ting endomorphisms of the regular product a =

iI
a
i
with a
i
=
i
(1) ( i
I).
6
The results concerning regular products of groups are generalized to the
lattice-theoretical level. Thus e.g. if 1 =

iI
a
i
and a
i
=

jJ
i
a
ij
(i I)
then 1 =

iI,jJ
i
a
ij
; if 1 =

iI
a
i
and b
i
a
i
(i I) then setting
b =
_
iI
b
i
one obtains b =

iI
b
i
; etc. Benado concentrates on the so-
called renements of two regular decompositions. Here is a sample result [30].
Consider two regular products 1 =

iI
a
i
and 1 =

jJ
b
j
with associated
Fitting endomorphisms
i

iI
and
j

jJ
, respectively. Then the following
conditions are equivalent for two renements a
i
=

jJ
a
ij
(i I) and b
j
=

iI
b
ji
(j J) : for all i I and j J,
(I)
i
(b
ji
) = a
ij
&
j
(a
ij
) = b
ji
& a
ij
= b
ji
;
(II)
i

j
(1) =
j

i
(1) .
Moreover, if these conditions hold, then for every i I and j J,
(III)
i

j
(1) = a
ij
&
j

i
(1) = b
ji
.
A variant of this result [32] says that relations (III) hold if and only if
(IV)
_
iI,jJ
(
i

j
(a

i
)
j

i
(b

j
)) = 0 ,
where a

i
=
_
kI{i}
a
k
and b

j
=
_
kJ{j}
b
k
. Besides, (IV) implies
i
(b
ji
) =
a
ij
and
j
(a
ij
) = b
ji
for all i I and j J.
Another tool used by Benado in the abstract theory of regular products
was the concept of a relative closure operator (operateur de fermeture relatif),
meaning a map which associates with every pair x, y satisfying x y, an element
y
x
subject to the following conditions:
(FR1) x
x
= x ,
(FR2) x y =(y
x
)
x
y
x
,
(FR3) x y z =y
x
z
x
z
y
.
It is easily seen that these properties imply
x y =x y
x
y ,
x y =(y
x
)
x
= y
x
.
So, for a xed x the map
x
is a closure operator on the set (x] of all lower
bounds of x.
For instance, if

is a Kuratowski closure on a distributive lattice, then
y
x
= xy is a relative closure operator. Note that the elements aa

i
, occurring
in the denitions of a regular product and of the Fitting endomorphisms and
largely used by Benado in the theory of regular products, are in fact the relative
closures (a

i
)
a
. Another relative closure operator which occurs here is H
G
=
(G
n
H) H.
7
Benado has also developed [27], [28], [29], [32], [37], [41] a more comprehen-
sive theory of Lie lattices endowed with a commutating function (commuta-
trice), that is, a function q(x, y) obeying a list of 12 axioms which abstract the
properties of the n-step commutator H
n
K of subgroups. As a matter of fact,
the ultimate goal of Benado in this series of papers refers to groups. So he pays
attention to lattices of subgroups equipped with a function q(H, K) satisfying
the axioms of a commutatrice. A subgroup H of a group G is called q-normal
provided q(G, H) H, while a regular product

iI
G
i
is said to be q-regular
provided all the factors G
i
are q-normal. We conclude this Section with the
following sample theorem [37], which generalizes several previous theorems. If
G =

iI
G
i
is a regular product for which there exists i
0
I such that all
the factors G
i
with i ,= i
0
are q-normal, then G
i
0
is q-normal too, so that the
product is q-regular.
Part B
Schreier-Zassenhaus and Jordan-Holder theorems in
arbitrary lattices
There is a huge literature around the Schreier-Zassenhaus and Jordan-Holder
theorems in group theory and their generalizations to modular lattices and be-
yond. One of the major research lines of Benado refers to this eld, which may
be characterized as dealing with renements of chains. The main concern of
Benados work in this area was the passage from modular lattices to arbitrary
lattices [5], [10], [13], [14], [20] and even to posets [13] and multilattices [17],
[23]. The results were announced in [2], [3], [6], [7], [9], 11] and, with emphasis
on groups, in [18], [19], [22]. We can only sketch his ample contributions and to
do this we begin with a short presentation of the modular lattice background.
The two theorems in question are related to another well-known result, which
states that in a modular lattice, every two intervals of the form H = [a, a b]
and K = [a b, b] are isomorphic. This is proved by using the isotone maps
+
: H K dened by x
+
= b x and

: K H dened by y

= a y,
which we call the canonical maps associated with the conjugate or transposed
intervals H and K. Modularity implies that for every x H and every y K,
x
+
= a (b x) = (a b) x = x ,
y
+
= b (a y) = (b a) y = y .
These identities show that the canonical maps establish an order isomorphism
between H and K, hence a lattice isomorphism, too.
Benado proves the following converse of the above theorem: if in a lattice L
the canonical maps satify x
+
= x (a, b L) (x H) or y
+
= y (a, b
L) (x H), then the lattice is modular.
8
Indeed, note rst that the canonical maps can be dened and are isotone in
every lattice. Now suppse e.g. the identity y
+
= y holds. Take a, b, c L
with c b. Then
b b (a c) (b a) c a b ,
hence (b a) c K, therefore
(b a) c = ((b a) c)
+
= (a c)
+
= b (a c) (b a) c ,
so that (b a) c = b (a c), proving that L is modular.
So every lattice in which the canonical maps asociated with a, b L es-
tablish an isomorphism is modular. Note that this result cannot be proved
by observing that in the pentagon non-modular lattice o, a, b, c, u we have
[a, a b] = a, c, u and [a b, b] = o, b, because the intervals H and K of the
entire lattice may be dierent from the above.
Now let us come to the problem we have announced, i.e., the renement of
chains. We deal with nite chains x
0
x
1
x
n
of a modular lattice L.
The intervals [x
i1
, x
i
] (i = 1, . . . , n 1) are called the factors of the chain (a
term reminiscent of the quotient or factor groups G
i
/G
i1
of a normal chain
G
0
G
1
G
n
dealt with in group theory). A factor [x
i1
, x
i
] is called
proper if x
1
,= x
i
. The length of a chain is the number of its proper factors.
A renement of a chain a
0
a
1
a
r
is a chain c
0
c
1
c
t
such that a
0
= c
0
, a
r
= c
t
and a
0
, a
1
, . . . , a
r
c
0
, c
1
, . . . , c
t
. If the latter
inclusion is strict, the renement is called proper. The Zassenhaus renements
of two nite chains having common extremities, that is,
(1.1) a
0
a
1
a
r
,
(1.2) b
0
b
1
b
s
,
(1.3) a
0
= c
0
, a
r
= b
s
,
are the chains a
ij
and b
ji
dened by
(2) a
ij
= a
i
(a
i1
b
j
) (i = 1, . . . , r; j = 0, 1, . . . , s) ,
(3) b
ji
= b
j
(b
j1
a
i
) (i = 0, 1, . . . , r; j = 1, . . . , s) .
This makes sense because, as can be easily checked,
(4) a
0
= a
10
a
i1
= a
i0
a
i1
a
is
= a
i
a
rs
= a
s
,
(5) b
0
= b
10
b
j1
= b
j0
b
j1
b
jr
= b
j
b
js
= b
s
.
Another property is that for every i 1, . . . , r and every j 1, . . . , s,
there are two isotone maps
: [a
i, j1
, a
ij
] [b
j, i1
, b
ji
] ,
9
: [b
j, i1
, b
ji
] [a
i, j1
, a
ij
] ,
dened by
(x) = b
j
(a
i1
b
j1
x) ,
(y) = a
i
(a
i1
b
j1
y) ,
respectively. Indeed, set a
i1
= m a
i
= M and b
j1
= n b
j
= N. Then
b
j, i1
= N (n m) N (m n x) = (x) .
But x a
ij
= M (m N), therefore
(x) N (m n (M (n N))) N (M n) = b
ji
and similarly (y) [a
i, j1
, a
ij
].
As a matter of fact, the Zassenhaus renements and the existence of the
isotone maps and are valid in arbitrary lattices. If the lattice is modular,
there exist rs isomorphisms
(6) [a
i, j1
, a
ij
] [b
j, i1
, b
ji
] (i = 1, . . . , r; j = 1, . . . , s) .
To prove this we x i and j. Using the above notation, we observe that
M (m N) x M (m N) ,
from which we infer
(x) = M (m n (N (m n x))) = M (m n N) (m n x)
= M (m N x) (m n x) = M (m n x) = (M (m n)) x = x
and similarly (y) = y.
At this point the following denition is necessary. Two chains C and C

are said to be isomorphic if they have the same length and there is a bijection
between the proper factors of C and the proper factors of C

such that the


proper factors which correspond to each other are isomorphic.
Now the above results yield the following corollary, known as the Schreier-
Zassenhaus theorem for modular lattices: every two nite chains with common
extremities have isomorhic renements.
This results from the following two remarks. The factors of the chains (4)
and (5) are the intervals ocurring in (6) plus the improper factors [a
is
, a
i+1, 0
] =
[a
i
, a
i
] and [b
jr
, b
j+1, 0
] = [b
j
, b
j
]. Each of the rs pairs of factors occurring in
(6) consists either of two improper factors or of two isomorphic proper factors.
So the number of proper factors is the same for the chains (4) and (5), and the
required bijection is the restriction of (6) to the proper factors.
The next theorem requires the following denition. A strict chain a
0
< a
1
<
< a
r
is called maximal if it has no proper renements. It is easy to see that
this happens if and only if for each i = 1, . . . r, the element a
i1
is covered by
a
i
, that is, [a
i1
, a
i
] = a
i1
, a
i
.
10
It should be mentioned that a modular lattice may have no maximal chains.
However, if maximal chains do exist then the following properties hold.
If C
0
: a
0
< a
1
< < a
r
is a maximal chain and C

0
is a Zassenhaus
renement of C
0
, then C

0
is not a proper renement, hence it must be of the
form
. . . (a
i1
) = a
i0
= a
i1
= = a
i, k(i)
< a
i, k(i)+1
= = a
is
(= a
i
) . . .
for i = 1, . . . , r. It follows that C

0
is a maximal chain isomorphic to C
0
.
We are now in a position to prove the Jordan-Holder theorem for modular
lattices: if maximal chains do exist, then every strict chain can be rened to a
maximal chain and all maximal chains have the same length.
6
Indeed, let C be a strict chain and C

, C

0
the Zassenhaus renements of C
and the maximal chain C
0
, respectively. Then the proper factors of the maximal
chain C

0
are two-element chains, and since C

is isomorphic to C

0
, the proper
factors of C

are also two-element chains, therefore by deleting from C

the
redundant terms we obtain a strict chain C

which is still a renement of C


and whose prime factors are two-element chains, therefore C

is maximal.
For the second claim suppose that C is also a maximal chain. Then C and
C
0
are isomorphic to their Zassenhaus renements C

and C

0
, and the latter are
isomorphic. But the isomorphism of chains is clearly an equivalence relation,
therefore the strict chains C : a
0
< a
1
< < a
r
and C
0
: c
0
< c
1
< < c
s
are isomorphic, which implies r = s.
Benado proves the following Schreier-Zassenhaus theorem which holds in
arbitrary lattices. Suppose the chains (1) satisfy the isomorphisms
(7.1) [a
i
, a
i
b
j
] [a
i
b
j
, b
j
] ,
(7.2) [b
j
, a
i
b
j
] [a
i
b
j
, a
i
] ,
for i = 1, . . . , r 1 and j = 1, . . . , s1. Then their Zassenhaus renements (2),
(3) have the same length and the isomorphisms (6) hold. Besides,
(8.1) [a
i, j1
, a
i, j1
b
j, i1
] [a
i, j1
b
j, i1
, b
j, i1
] ,
(8.2) [b
j, i1
, a
i, j1
b
j, i1
] [a
i, j1
b
j, i1
, a
i, j1
] ,
for i = 1, . . . , r 1 and j = 1, . . . , s 1.
Also, Benado obtains the following Jordan-Holder theorem for arbitrry lat-
tices. Suppose the chains (1) are strict, satisfy (7) and for every i = 1, . . . , r 1
and every j = 1, . . . , s 1 there are no elements a [a
i1
, a
i
] and b [b
j1
, b
j
]
satisfying simultaneously [a, a b] [a b, b] and [b, a b] [a b, a]. Then
r = s and the intervals of the two chains are isomrphic (in a certain order).
Another major concern of Benado was the theory of normality in arbitrary
lattices, initiated by O. Ore. An element b is called -normal if
x z =x (b z) = (x b) z ;
6
Unlike what happens for groups, in a lattice the isomorphism of two maximal chains
having the same length is trivial, because the proper factors are two-element chains.
11
an element a is said to be -normal if
a z =a (y z) = (a y) z ,
and an element which is both -normal and -normal is called seminormal.
Benado studied in great detail these concepts, as well as several new related
concepts of normality introduced by himself, for instance a concept called binor-
mality such that seminormal = binormal = -normal. With each of these
concepts a corresponding concept of chain is associated, called -normal chain,
-normal chain, seminormal chain, etc. Benados endeavour was to establish
Schreier-Zassenhaus and Jordan-Holder theorems for the various types of chains.
For example, an interval [x, y] is called -normal if x is an -normal element
of the sublattice (y] = z L [ z y. A chain (1.1) or (1.2) is said to be
-normal if all of its proper intervals are -normal. Then the corresponding
Schreier-Zassenhaus theorem states the following properties of two -normal
chains with common extremities (1): the Zassenhaus renements have the same
length, the isomorphisms (6) hold and each a
ij
is -normal in [a
i, j1
, a
i
] and
each b
ji
is -normal in [b
j, i1
, b
j
]. Further, an interval [x, y], where x < y,
is called -normal-prime if among the elements -normal in [x, y] there is no
element z such that x < z < y. The corresponding Jordan-Holder theorem
states that if all the intervals of two -normal chains with common extremities
(1) are -normal prime, then r = s and the intervals of the two chains are
isomorphic (in a certain order).
The paper [13] proposes an axiomatic approach which passes from the var-
ious normality relations such as -normality, -normality, etc., to an abstract
normality N, the problem under investigation being: what conditions can be im-
posed upon N such as to force the validity of the Schreire-Zassenhaus theorem?
Also, Benado obtains generalizations to posets [14] by creating a technique based
on Dedekind quadrilaterals (, a, b, ), which mean a and b ,
thus generalizing (a b, a, b, a b). For the generalizations to multilattices [17]
see Section C1.
The paper [26] introduces several normality relations and characterizes them
by appropriate relative closure operators. For instance, the operator y
x
= yx
is associated with the normality relation
xKy x y and x y = y .
Then every two nite K-normal chains with common extremities have isomor-
phic Zassenhaus renements.
PART C
C1. Multilattices
In 1953 Mihail Benado introduced the concept of a multilattice in a paper [12]
which announces two equivalent denitions of this concept and a few examples.
12
The standard reference for the theory of multilattices is the paper [17] (with
corrections in [24]), while other papers of Benado on multilattices and their
applications are [15], [16], [33], [35], [36]. [39], [40], [51]. In this Section we
select several denitions and theorems from [17].
The basic denition is the following. Let M be a partially ordered set. If
a, b M, denote by U(a, b) and L(a, b) the (possibly empty) sets of upper
bounds of a, b and of lower bounds of a, b, respectively. The poset M is a
multilattice if for any c U(a, b), the set U(a, b) x M [ x c has a
minimal element, and dually. Note that lattices are dened by the stronger
requirement that for all a, b M, the non-empty sets U(a, b) and L(a, b) have
least element and greatest element, respectively.
It is easily seen that this denition is equivalent to the following one:
A multilattice is a poset M such that the following conditions are fullled
for any a, b M:
(MLsup) every upper bound of a, b includes a minimal upper bound,
and
(MLinf) every lower bound of a, b is included in a maximal lower bound;
if there is no common upper bound of a, b/common lower bound of a, b, then
condition (MLsup)/condition (MLinf) is vacuously satised.
The denition of multilattices can also be given in terms of the follow-
ing three well-known pre-orders (which are important in the context of non-
deterministic programming languages):
X _
S
Y y Y x X x y (Smyth ordering) ,
X _
H
Y x X y Y x y (Hoare ordering),
X _
EM
Y X _
S
Y & X _
H
Y (Egli Milner ordering) .
A remarkable theorem says that a multilattice M is a lattice if and only if
M is directed, i.e., every two elements a, b M have upper bounds and lower
bounds, and has the Riesz Interpolation Property (RIP), i.e., if a
i
, b
j
M satisfy
a
i
b
j
(i, j = 1, 2), then there exists c M such that a
i
c b
j
(i, j = 1, 2).
For every upper bound of a, b, let (ab)

denote the set of minimal upper


bounds of a, b included in , and dually, let (a b)

denote the set of maximal


lower bounds that include the lower bound . Then the set-theoretical union
a b of all (a b)

is the (possibly empty) set of all minimal upper bounds of


a, b and the set-theretical union ab of all (ab)

is the (possibly empty) set of


all maximal lower bounds of a, b. Benado proved the following properties (Mn)
and their duals (Mn

):
(M1) (commutativity) if a b ,= then a b = b a;
(M2) (partial associativity) if a b ,= and (a b) c ,= , then b c ,=
, a (b c) ,= and for every m (a b) c there is m

a (b c) such
that m m

,= and m m

= m;
(M3) (absorption) if a b ,= then a (a b) = a;
(M4) a a ,= ;
13
(M5) if the elements a, b, m, m

satisfy m, m

a b, m m

,= and
m ,= m

, then m, m

, m m

.
Conversely, if M is a set endowed with two multioperations , : M
2

T(M) which satisfy (M1)-(M5) and their duals, then a b = b a b = a


and the relation a b ab = b ab = a is a partial order on M, such
that conditions (ML) above are satised and a b/a b is the set of minimal
upper bounds of a, b/of maximal lower bounds of a, b.
Thus multilattices can also be dened as multialgebras, i.e., as sets endowed
with two multioperations ab, ab satisfying the above axioms. Jakubk [Jak1],
solved two problems raised by Benado in [17], namely he proved that (i) axiom
(M5)&(M5

) does not follow from the other axioms of the system above, and
(ii) if the multioperations and are associative and M is directed, then the
multilattice M is in fact a lattice, the latter hypothesis being essential.
Multilattices appear in various contexts, such as divisibility theory [14], [15]
(see the next Section), Jordan-Dedekind chain conditions, the Mobius func-
tion [33], [35], [40], partial dierential equations (the paper [16] solves a prob-
lem about the wave equation in two-dimensional space-time raised by G. Birk-
ho [1948, Ch.9,13]), functional analysis, theory of geometrical continua (e.g.
squares ordered by set inclusion), or topological complexes. See [17] and the
survey papers [39], [51].
Benado introduces the concepts of modular multilattice and distributive mul-
tilattice by generalizing two well-known charcterizations of modular lattices and
distributive lattices, respectively. A multilattice M is distributive (modular) if
the conditions v a b, v a b

, u a b, u a b

(and b b

) imply that
b = b

.
Lihova and Repask y [LiRe] have proved that the classes of directed modular
multilattices and directed distributive multilattices are varieties, that is, they
are closed under the constructions of submultialgebras, homomorphic algebras
and direct products.
The set of all squares with the sides parallel to two rectangular directions,
partially ordered by set inclusion, is a multilattice which is not a lattice; cf.
Benado [17], page 312. This multilattice is modular, since the map v which
assigns to any square a the length v(a) of its side, is a norm, as shown on page
336, and any multilattice pssessing a norm is modular by Theorem 5.5 on page
337. It would be interesting to check when the multilattices provided by words,
partially ordered by
pf
,
sf
or
fct
(see C3), are modular or distributive.
The following reexive relation is introduced between the quotients x/y (i.e.,
intervals [y,x]): x/y x

/y

provided x x

, y y

, x (x

y)
x
and
y

(x

y)
y
. The transitive closure of is called T-similarity. Then the
following generalization of the Schreier renement theorem holds: in a modular
multilattice, two chains with common extremities possess a system of canonical
renements having the same length and whose conjugate quotients are T-similar.
A technique introduced by Benado in [14] and largely used throughout his
14
entire work is the use of what he calls Dedekind quadrilaterals or simply quadri-
laterals. A quadrilateral is a very particular case of two chains with common
extremities, namely a quadruple (, ; a, b), where a, b . A subquadri-
lateral of (, ; a, b) is any quadrilateral (m, d; a, b) with m and d.
Given a quadrilateral (, ; , b) and two elements m (a b)

and d
(a b)

, Benado denes the Dedekind connections


(p) = (b p)
d
, p /a and (p

) = (a p

)
m
, p

b/,
and similarly

(q) and

(q

) by interchanging the roles of a and b. A strong


multilattice is a multilattice M such that any quadrilateral contains a subquadri-
lateral for which the corresponding Dedekind connections are functions from M
to M.
The following strengthening of the previous Schreier renement theorem
holds: in a strong modular multilattice, every two chains with common extrem-
ities have isomorphic renements.
A well-known paper of Glivenko [1937] suggested to Benado the idea of
constructing a theory of valuations for multilattices. He introduced a concept
of normed multilattice and proved that any normed multilattice is modular and
besides, if a normed multilattice is ltered, then it can be made into a metric
space.
One more theorem says that every distributive multilattice is strong.
We note that dierent generalizations of lattices have been proposed in the
literature, some of them similar to multilattices, which however are not quoted.
A poset is called a hypolattice if each of its closed intervals is a lattice and
the lattice operations coincide on overlapping intervals; cf. Draskovickova et al.
[Dra].
Mittas and Konstantinidiou [MitKo] introduced superlattices, which are
structures equipped with two multioperations a b and a b verifying some
axioms (commutativity, associativity, particular forms of idempotency and ab-
sorption, together with conditions allowing to introduce a partial order like in
the case of lattices). These authors had also introduced in 1977 the concept of
a hypolattice, where only join is a multioperation. Both multilattices and hy-
perlattices are particular cases of superlattices. A superlattice is a lattice if and
only if it satises the identity (a b) a = (a b) a = a. Strong superlattices
were investigated by Jakubk [Jak4].
Schroder [Sch] considered truncated lattices, which are structures with par-
tially dened lattice operations satifying the axioms of idempotency, commuta-
tivity and weak associativity.
A nearlattice is a join semilattice such that every principal lter [a) is a
lattice with respect to the semilattice order; cf. [ChK1],[ChK2],[ChK3].
We note that while these authors take a purely algebraic approach, i.e., based
on multioperations satisfying convenient axioms, the generalization of Benado
is mainly based on the partial order.
C2. Divisibility theory
15
One of the themes studied by Mihail Benado was divisibility theory [14],
[15], [53].
In [14] he introduces the concept of a division semigroup (in French, semi-
groupe `a division or semigroupe divisionnaire). By this term is meant a semi-
group S such that (i) S has an identity 1 and a zero 0, (ii) S0 is cancellative,
(iii) aS = Sa for any a S, and (iv) ab = 1 implies a = b = 1.
The name division semigroup is justied by the fact that the divisibility
relation a b c a = bc d a = db is a partial order such that
0 a 1 and a b =xay xby.
The following condition H plays a central role in [14]: a poset satises H
provided there exists an element
0
such that for any quadrilateral (,
0
; a, b)
there is a subquadrilateral (m, d; a, b) such that the order isomorphisms m/a
b/d and m/b a/d hold.
The following main results are proved in [14]:
Corollary 5.83. Every division semigroup satises H with
0
= 0.
Theorem 5.9. If two chains of a division semigroup have common extremities,
then they have isomorphic renements.
Theorem 5.93. Let R be an integral domain with identity and P be the lattice
of its principal ideals. Then P satises the ascending chain condition (ACC) if
and only if R satises the Satz von der eindeutigen Primfaktorzerlegung, in
which case the lattice P is distributive.
In other words, the following properties hold:
) If (P, ) is a lattice satisfying the ascending chain condition (ACC), then
R is a unique-factorization domain (UFD).
) Conversely, if R is a UFD, then (P, ) is a lattice satisfying ACC.
) If R is a UFD, then the lattice P is distributive.
A very short proof in [14] provides a reference to Birkho [1948] and a hint.
We give below a more explicit proof.
Recall rst that an integral domain is a nontrivial commutative ring R with
no zero divisors. One denes (see e.g. [And]): the divisibility relation x[y if
y = dx for some d R; x is a unit if xy = 1 for some y R, and let U(R)
be the group of units of R; x and y are associates, x y, if x = uy for some
u U(R); a nonzero nonunit x is irreducible if x = yz implies y U(R)
or z U(R), and is prime if x[yz implies x[y or x[z. Clearly every prime is
irreducible , but the converse does not hold in general. An integral domain R is
a unique factorization domain (UFD) if every nonzero nonunit of R is a product
of irreducible elements of R, and this decomposition is unique up to the order
of factors and association.
Clearly the set P of principal ideals of an integral domain P, equipped with
the product of ideals (a)(b) = (ab), is a division semigroup with identity (1) = R
and zero (0) = 0. Note that (a) (b) b[a.
) It is well known that ACC for P implies that every element of R admits
at least one representation as a product of irreducible elements (see e.g. [And],
fourth Theorem), hence it remains to prove the uniqueness of this representation.
Since P is a lattice, for any two elements a, b R there exist (d) = l.u.b.((a), (b))
in P. It follows by [Doc], Theorem 2, that d = g.c.d.(a, b) in R. Therefore,
16
by Theorem 6 in [Doc], every irreducible of R is prime, and this implies the
uniqueness of the decomposition into irreducibles by a well-known argument.
) A well-known construction from number theory shows that any two ele-
ments a, b R have l.c.m. and g.c.d., hence (a), (b) P have g.l.b. and l.u.b.
So P is a lattice. Besides, if (a
1
) (a
2
) (a
3
) . . . is an increasing sequence
in P, then a
1
, a
2
, a
3
, . . . is a decreasing sequence in R such that a
k+1
[a
k
for all
k, hence all the factors of a
k+1
are associates of the factors of a
k
. Therefore the
sets of non-unit factors of a
1
, a
2
, a
3
, . . . form a decreasing sequence, hence that
sequence stops at a certain step n; this implies a
n
= a
n+1
= . . . .
) A well-known construction from number theory shows that for any two
elements a, b of a UFD we have ab = [l.c.m.(a, b)][g.c.d.(a, b)]. If P is a lattice,
this yields (a)(b) = [(a) (b)][(a) (b)], by [Doc], Theorem 2. Therefore, if
(a)(b) = (a)(b

) and (a)(b) = (a)(b

), then (ab) = (a)(b) = (a)(b

) = (ab

),
showing that ab ab

, hence b b

, therefore (b) = (b

). But it is well knwon


that the uniqueness of the relative complements implies the distributivity of the
lattice.
In partially ordered groups we have x(x y)
1
y = x y, for any x, y, the
existence of one side implying the existence of the other side and the equality.
The meaning of the above formula allows us to consider partially dened
lattices. Moreover, similar formulas have been obtained for multilattice-ordered
groups; cf. Vaida [Va], [Vai]. A classical model for dealing with partially ordered
algebraic structures is the monograph by Fuchs [Fu], whose principal source
is the classical article of Birkho [Bir] on lattice-ordered groups; for partial
algebras see [LjEv].
In the paper Benado [15] (in Romanian and therefore with a very limited
echo), the author considers the divisibility of the algebraic number eld R().
The problem raised concerns the possible divisibility rules for these integers
rather than those for ideals.
As it is known, Dedekind and afterwards Emmy Noether gave an abstract
setting for the theory of ideals. Noetherian rings were introduced and funda-
mental factorization theorems for rings were proved. Later on, a program was
carried on by Ore, Ward and Dilworth, aiming at a lattice-theoretic setting for
the Noetherian theory of ideals, which was obtained in a paper of Dilworth
published in 1961, in the form of a version of the Krull principal ideal theorem.
The main concern in [15] is the study of multilattice/lattice division semi-
groups, meaning those division semigroups (S, , 0, 1, ) for which (S, ) is a
multilattice/lattice.
Benado observes that every Noetherian (i.e., satisfying ACC) division semi-
group is a multilattice division semigroup.
After several elementary properties, as for instance ab = ab = ,
Benado proves ([15], Lemma 3.3) that a multilattice division semigroup S is
lattice-ordered if and only if for any a, b S and m (a b)
0
, we have m ab
and m ba.
An auxiliary result is the following. Theorem ([15], Lemma 3.4). In a dis-
tributive multilattice semigroup, if (a b)
1
= 1 then ab = ba.
17
Sketch of proof. Since aS = Sa, there is b
1
S such as ab = b
1
a. Benado
proves that (a b)
1
= 1 and besides, for m = ab one has (a b)
m
= m by the
computation rules in S and also (a b)
m
= m
1
(m
1
= ba). We have m
1
= m
and therefore b
1
= b by distributivity.
This proof provides an instructive use of distributivity and suggests possible
ways to obtain sucient conditions for commutativity.
Corollary (3.4 in [15]). A Noetherian distributive multilattice semigroup S
is commutative.
Indeed, if S is Noetherian then there exist maximal elements ,= 1 and irre-
ducible in the sense of divisibility, and for any such elements p, q, p ,= q, we
have (p q)
1
= 1 and therefore pq = qp by Lemma 3.4. Since S is Noetherian,
every element ,= 0 is a product of irreducible elements.
We note that in [JK] the authors raise the problem of whether the assumption
of distributivity or commutativity of the group G considered in their Theorem
4.11 can be dropped. The Corollary quoted above gives a sucient condition for
commutativity. Moreover, in [JK] one says that a quadruple a, b, u, v is regular
if u ab, v ab and va = bu (pag. 603), i.e., a+b = u+v. Considering
the latter expression written in a convenient form for the noncommutative case,
we might hope to arrive at a sucient condition for the equality of the relative
complements of a, provided they constitute regular quadruples (by an argument
similar to that used for above).
The principal result of Benado is the following.
Theorem ([15], 4, p.267). Let S be a Noetherian division semigroup. If its
multilattice reduct is distributive, then S is a lattice.
In general, the unique factorization theorem is not satised by the integers
of a eld R() and therefore the multilattice division semigroup dened with
respect to divisibility is not lattice-ordered. It follows that this multilattice is
not necessarily distributive and thus distributivity is not a general rule for the
divisibility of integers in algebraic number elds.
As to the proof of the above Theorem, we should notice the interesting way
in which distributivity is used. For in general, in order to obtain the classical
decomposition of ideals into a product of irreducible ideals one looks for sucient
conditions to ensure that the implication irreducible = prime holds for ideals,
as shown below. The use of distributivity makes ncessary a quite dierent
approach. Indeed, in the conventional approach, if ab p with p irreducible, if
a , p then gcd(p, a) = 1 and thus gcd(pb, ab) = b. Since pb, ab p, we have
b p, showing that p is prime. The last implication is not true if we have a
multilattice instead of a lattice.
Benado [15] starts by showing that if p, q are irreducibele, p ,= q and m
(p q)
0
, then m = pq = qp ; therefore (p q)
0
is a singleton. In the next step
he proves that if c (a b), a, b < c and a ,= b, then again a b and a b are
singletons. The hypothesis that S is Noetherian is used only in the nal step,
which is an induction on the lengths of [a, 1] and [b, 1].
We conjecture that the particular use of distributivity is a promising way
for other applications to decomposition theorems.
18
There are three papers of Benado with the title Remarks on the theory of
multilattices. The series is incomplete: it comprises the papers [30] (without
series number), [40] (# IV) and [53] (# VI). Paper [53] has the secondary
title Contributions `a la theorie des structures algebriques ordonnees, which
nowadays seems too vague if we refer to the contemporary subject Ordered
structures (AMS Classication 06Fxx).
As a matter of fact, this paper originates in the courses given by Dan Bar-
bilian in the 1950s on the unique factorization theorem in rings of quadratic
integers
7
, with references to ideals, following Dedekind and Emmy Noether.
The theory presented by Benado in [53], closely related to multilattices and
having one of its main sources in the lectures of Barbilian, deals with decom-
position theorems generalizing those of Lasker-Noether and Kurosh-Ore, and
builds a general framework for divisibility and residuation.
The reader of [53] may encounter diculties in going through many deni-
tions, whose lattice- or multilattice-theoretical background is not made explicit
by Benado. An eort to reveal this background is worth doing, as it may en-
lighten a theory of divisibility far more general than the conventional one.
C3. Applications in computer science and mathematical
modelling
Stricto sensu, there are no results in the work of M. Benado which might
be seen as properly related to computer science. However some of his papers
are now quoted as an adequate framework for the algebraic basis of theoretical
computer science. In the following examples, references and our own notes are
given in this sense.
In an earlier stage of computer science, many computer scientists found a
major interest in the lattice-theoretical approach, in Scotts theory of continuous
lattices and in the ADJ work dealing with -complete sets and -continuous
algebras. However, many structures of interest in computer science are not
actually lattices, thus requiring a more general theory of order structures.
A Basic Example. Let

be the set of all strings, or words, including the


empty string e, over a nite alphabet , and let u
pf
v i v = uv
1
, u being
called a prex of v. Similarly, one denes the dual u
sf
v i u is a sux of v,
i.e., v = v
2
u, and u
fct
v i u is a factor of v, i.e., v = v
1
uv
2
(e.g., roma
pf
roman
pf
romania, man
sf
roman, man
fct
romania).
Obviously, the relations
h
, for h = pf, sf, fct, are partial orders. In general,two
dierent words u and v may not have a pf- or sf- upper bound or lower bound
dierent from e (e.g. u = ab and v = ba, where a, b ), but they always have
the fct-upper bounds u
1
vu
2
uu
3
and u
1
uu
2
vu
3
and possibly others (e.g., ab and
ba have the upper bounds aba and bab).
7
The present authors were among the students of Barbilian in 1956-1957, when he gave a
course on algebraic number theory and a course on the foundations of geometry. The second
author attended as well Barbilians lectures on the multiplicative theory of ideals, as the
unique student (!).
19
Remark. If u
h
v then [u[ [v[, where [w[ is the length of w, hence (

,
h
)
is locallynite, in the sense that every interval is nite, for h = pf, sf, fct.
Moreover, L = (

, , e,
h
) is a monoid and a multilattice, the identity e being
the rst element.
Every locally nite poset is also chain-nite and therefore L satises both
the ascending and descending chain conditions, i.e., L is an archimedean poset
in the terminology of Benado [14].
The one-sided compatibility of the concatenation with the partial orders

h
, h = pf, sf, is readily checked. For the relevance of these partial orders see
Lothaire [Lot].
Partially additive semantics. The idea of partially additive semantics is the
following. Given a program prg Prog and its set of data D = InputOutput,
a possible execution exec
i
, i I, is viewed as a partially dened function
f
i
= exec
i
(prg) : D

D, dom(f
i
) = d
i
Input .
In the model introduced by Manes and Arbib [MA], dierent executions f
i
of
prg should have disjoint domains. The semantics is dened as a sum of the
possible executions
sem(prg) = f =

(f
i
[ i I) P = Pfn(D, D) ,
where Pfn(D, D) is the set of partially dened functions D

D, the sum
being dened in a partially dened monoid, by
dom(f) =
_
(dom(f
i
) [ i I), dom(f
j
) dom(f
k
) = , j ,= k ,
f(x) ::= if ( i I) (x dom(f
i
)) then f
i
(x), else undened .
The reference structure is a partially additive naturally ordered or sum-ordered
associative semiring (Pfn(D, D),

, , 0, 1, ) (see Rudeanu and Vaida [RV],


Vaida [Vai1], [Vai2] and further partial semirings). The multiplication is con-
catenation, a posible interpretation being the composition of functions, corre-
sponding to the sequential execution, dom(0) = , 1 = 1
D
and f g means
that g is an extension of f (the approximation ordering). P is a (additive)-
cancellative, i.e., f +f
1
= f +f
2
implies f
1
= f
2
.
The above structure of Pfn(D, D) is further explained.
Denition. A naturally ordered partial semiring, also called a sum-ordered
partial semiring, is a partial algebra (S, +, , 0, 1, ) with the following prop-
erties: (S, +, 0) is a partial commutative monoid with a neutral element called
zero, the addition + being partially dened; this monoid is partially associative,
in the sense that if one side of the associative rule exists then the other exists as
well and they are equal; (S, +, ) is partially ordered by a b i c S b = a+c
(in Pfn(D, D) this is equivalent to the approximation ordering previously de-
ned); this order is compatible with addition, meaning that if a b then the
existence of b +d implies that a+d exists as well and a+d b +d; (S, , 1, 0, )
20
is a partially ordered monoid with identity element 1 and annihilator element 0;
the following rules of partial distributivity are satised: if b +c S exists then
a(b +c) = ab +ac and (b +c)a = ba +ca a S ,
with the meaning that the sums of the right sides exist as well and the equalities
hold.
The structure P = (Pfn(D, D), 0, ) is not a lattice; as a matter of fact it
is not even directed, because f, g h is possible only if f(x) = h(x) = g(x) for
any x dom(f) dom(g). Yet it is a meet semilattice with rst element 0 by
taking
dom(f g) = x D [ x dom(f) dom(g) and f(x) = g(x) ,
inf(f, g) = f g with (f g)(x) = f(x) = g(x) x dom(f g) .
Consequently, one has the Riesz interpolation property (RIP): if x
1
, x
2
y
1
, y
2
then z P such that x
1
, x
2
z y
1
, y
2
.
Moreover, P is a multilattice. It remains to show that if for f and g there
is h Pfn(D, D) such that f, g h, then there exists sup(f, g) = f g
Pfn(D, D). Indeed, we take
dom(f g) = dom(f) dom(g) ,
(f g)(x) ::= if x dom(f) then f(x) else g(x) x dom(f g) .
Readily, dom(f), dom(g) dom(f g) and f, g f g. From f, g h we
obtain dom(f g) dom(h) and for x dom(f g), if x dom(f) then
(f g)(x) = f(x) since f h and similarly for g.
In fact P is a particular case of a multilattice, since it is a dual nearlattice.
The term nearlattice was introduced by Cornish and Noor [NC] for a concept
previously studied by Sholander under a dierent name. It designates a join
semilattice in which every principal lter is a lattice with respect to the induced
order. Alternatively, a nearlattice was described by Chajda and Kolark [ChK1,
ChK2, ChK3] as an algebra with a ternary operation satisfying eight simple
identities. Hence the class of nearlattices is a variety.
The multilattice join f g can be seen as a partially dened algebraic sum
if we extend the denition of summability from families of disjoint functions to
families with overlapping functions, i.e., functions which coincide on the overlaps
of their domains. The support structure would be a partially additive naturally
ordered associative semiring P
1
= (Pfn(D, D),

1
, , 0, 1, ). In P
1
if the sum
f +
1
g exists then it coincides with f g. This sum exists i f and g are
compatible, in the sense that there exists h Pfn(D, D) such that f, g h.
Three more properties of Pfn(D, D) related to its structure can be provided
merely for motivating a further study.The proofs should remain for another
paper.
The sum decomposition property (SDP), saying that
0 x a +b =( s, t) 0 s a &0 t b &x = s +t ,
21
is equivalent to (RIP) under rather general conditions, but this equivalence is
nor proved for P and therefore an independent proof is needed.
Property 1. If f, g, h Pfn(D, D) and f g +
1
h then g and h exist such
that g
1
g, h
1
h and f = g
1
+
1
h x dom(f).
Property 2. Every directed family of functions has a sup.
Property 3. In P
1
the relative complements, whenever they exist, are unique.
As a matter of fact, it turns out that mulilattices and several more general
structures have applications in theoretical computer science and in logic. We
plan to come back in a future paper on the literature which quotes Benado in
this respect. Right now we illustrate the intended applications only by a few
references which are not coming from a rigorous selection, being rather a result
of our limited knowledge.
Thus the papers [Cab], [Cord] and [Mart] are examples of the systematic and
constant attention devoted to multilattices by the Spanish school. The lecture
[Mark] states that lattices are not enough for the specic needs of computer
science. Moreover, the basic examples given therein concern syntax word
ordering and semantics partially dened functions yet without a specic
development for these examples. The book [DP] does not quote Benado, but the
interesting and new examples of order structures proposed to the reader suggest
possible links for a future investigation, e.g., dierence posets or interpolation
conditions. The study [Maru] motivates an interest for many-valued logics based
on lattice-ordered algebraic structures and thus possibly based on multilattice-
ordered monoids.
Part D
D1. The general theory of partially ordered sets
Benado aimed at a general theory of partially ordered sets, which he cre-
ated under this very name. The results were announced in [34], [38], [42]-[46],
[49], [50], and full proofs were given in [47], [48], [52]. As curious as it may
seem nowadays, Benado regarded his theory of multilattices as being just an
intermediate step towards the general theory of partially ordered sets.
Roughly speaking, the idea is to endow a partially ordered set (poset) with
a supplementary structure in such a way that the most important classes of
posets and lattices be obtained as specializations of this general structure.
Technically speaking, Benado considers a poset (P, ) endowed with two
correspondences , between P and the set of subsets of P, subject to the
conditions
dA =d is an u.b. of A and mB =m is a l.b. of B,
where u.b. and l.b. stand for upper bound and lower bound, respectively;
besides, it is assumed that tuples (a, b, d, m) such that da, b and ma, b
do exist. We refer to all this as a (, ) structure.
22
A few particular (, ) structures play an important role:
1) the Dedekind structure (
0
, M
0
), dened by
d
0
A d is the l.u.b. of A and mM
0
B m is the g.l.b. of B;
2) the Hausdor structure (
H
, M
H
), dened by
d
H
A d is a minimal u.b. of A, and
mM
H
B m is a maximal l.b. of B;
3) the Riesz structure (
R
, M
R
), dened by
d
R
A d is an u.b. of A and for any u.b. u of A there is an u.b. t of A
such that t d, u, and
mM
R
B m is a l.b. of B and for any l.b. v of B there is a l.b. w of B
such that w m, v;
4) the ltering structure (structure ltrante) (
F
, M
F
), dened by
d
F
A d is an u.b. of A and for every u.b. u of A there is an u.b. t of
A such that t d, u, and
dM
F
B m is a l.b. of B and for every l.b. v of B there is a l.b. w of B
such that w m, v;
5) the discrete structure (, M) dened by
dA d is an u.b. of A, and mMB m is a l.b. of B.
Having in view a triple (a, a, a, a), we see that the discrete structure is a
paraphrase of the concept of poset itself. A multilattice is a Hausdor structure
satisfying the following conditions:
u a, b = d u d & d
H
a, b,
v a, b = m v m & mM
H
a, b.
The general theory of posets consists in the study of numerous conditions
that a (, ) structure may full, meaning the discovery of relationships that
exist between these conditions, called elementary incidence properties. Here
are a few samples. A (, ) structure is said to be: natural, if a b =
aa, b & ba, b; saturated, if d, d

a, b &m, m

a, b &d d

&m
m

= d = d

&m = m

; with similarity (`a similitude), if conditions da, b


and ma, b imply d = a b = m; relatively complemented, if u a
v = b ua, b &va, b; complemented, if the poset P has 0 and 1 and
a b 1a, b &0a, b; rening (ranante), if u a v = d, m u
d &v m&da, b &ma, b. Other elementary incidence properties have
longer denitions.
The study of relationships between elementary incidence properties was ex-
tended to include several generalizations of the concepts of modular and dis-
tributive lattices. For instance, a (, ) structure is called:
K-modular, if da, b &da, b

&ma, b &ma, b

&b b

= b =
b

;
O-modular, if the conditions da, b &ma, b &d a

a &b b

m
imply a

a, b

, b;
W-modular, if da, b &ma, b =[a, d] [m, b] (lattice isomorphism);
G-distributive, if da, b &da, b

&ma, b &a, b

=b = b

;
23
G

-distributive, if the hypothesis of G-distributivity and d d

and m m

imply b b

.
The above types of modularity are not equivalent, nor are the variants of
distributivity. Also, distributivity does not imply modularity, unless certain
supplementary hypotheses are assumed.
Among the numerous theorems obtained by Benado in this eld we mention
a Schreier-like renement theorem for W-modular rening structures, and a
generalization of the Glivenko theorem which makes every normed lattice into
a metric space. Another theorem characterizes the Dedekind structure of a
Boolean algebra viewed as a poset.
Benado acknowledges the valuable observations made by M. Kolibiar and J.
Jakubk in their correspondence with him during the preparation of the general
theory of partially ordered sets.
D2. Monotone connections
Another research project of Benado concerns the monotone conenctions in-
troduced in [5]. Certain peculiarities of the Zassenhaus renements of -normal
chains and
1
-normal chains have led thim to the introduction of the concepts
of monotone connections of type I and II, respectively, while the monotone con-
nections of type III have been suggested by the well-known Galois connections.
Let P and Q be two posets, and let (+, ) be a pair of isotone (also called
increasing) functions + : P Q and : Q P. Then (+, ) is called a
monotone connction of type I or II or III between P and Q according as
x
+
x x P & y
+
y y Q or
x x
+
x P & y y
+
y Q or
x
+
x x P & y y
+
y Q .
Recall also that a covariant or isotone Galois connection between P and Q is a
pair (+, ) of isotone functions + : P Q and : Q P such that
x x
+
x P & y
+
y y Q ,
while a contravariant or antitone Galois connection between P and Q is a pair
(+, ) of antitone (also called decreasing) functions + : P Q and : Q
P such that
x x
+
x P & y y
+
y Q .
The paper [59] continues the study of monotone connections of type I. In
the introduction to [59] Benado states that he has introduced the monotone
connections of type I rather in order to exhaust the types of monotone con-
nections (plutot en exhaustion des esp`eces de connexions monotones), among
which those of type III, introduced by J.R. B uchi under the name Paarungen,
have been independently rediscovered by myself and by R. Croisot.
However the aforementioned ve concepts do not exhaust the eight possibil-
ities of combining +, isotone or +, antitone with x P x x
+

24
or x P x
+
x and with y Q y y
+
or y Q y
+
y .
We want to note here that exhaustion is however realized, to the eect that the
study of these eight concepts reduces to the study of Benados monotone connec-
tions of type I and of type III. Indeed, denote by X
op
the dual of a poset X. It
was remarked by Benado himself that the monotone connections of type I and
II are dual to each other, to the eect that if + : P Q and : Q P form
a connection of type I (of type II), then + : P
op
Q
op
and : Q
op
P
op
form a connection of type II (of type I). Therefore the properties of monotone
connections of type II are obtained by duality from the properties of monotone
connections of type I. It is also immediately seen that the monotone connections
of type III and the covariant Galois connections are dual to each other, therefore
the study of the latter reduce to the study of the former. We have thus proved
that the study of the four isotone connections (meaning that +, are isotone)
reduces to the study of the monotone connections of types I and III, and it
remains to show that the same reduction is valid for the four antitone connec-
tions (meaning that +, are antitone). This follows immediately by observing
that if we associate with each monotone connection + : P Q, : Q P
of type I or III the two antitone connections + : P
op
Q,
:
Q P
op
and
+ : P Q
op
, : Q
op
P, then we obtain the four antitone connections.
The paper [59] is devoted to monotone connections of type I. It studies
involutive connections, meaning that x P x
++
= x
+
and y Q y
+
=
y

. It also provides characterizations of semilattices and semimultilattices in


terms of monotone connections. Here is the former theorem.
It is easy to see that if (S, ) is a meet semilattice, then for every a, b S
the mappings
(S0) + : (a] (b], x
+
= b x and : (b] (a], y

= a y ,
form a monotone connection of type I such that for every x (a] and every
y (b] the following properties hold:
(S1) x
+
x and y

y ,
(S2) x
+
= x x a &x b and y
+
= y y a &y b .
Conversely, if P is a poset such that for every a, b P there is a monotone
connection of type I + : (a] (b], : (b] (a] satisfying (S1) and (S2),
then P is a meet semilattice.
To prove the second statement, take a, b P and use the monotone con-
nection associated with these elements. Then a
+
(b], that is a
+
b, while
a
+
a by (S1). Now take an arbitrary lower bound v of a and b. Then v (a]
and v
+
v (type I). On the other hand v
+
= v by (S2) and v a implies
v
+
a
+
a
+
by (S1), hence v a
+
. 2.
Benado intended to continue this study by obtaining characterizations of
other ordered structures, such as lattices and multilattices, or Kurepas ramied
sets and trees. Unfortunately, this research project has been interrupted, the
paper [59] being apparently the last paper of Benado.
25
D3. Metric lattices
The paper [8], whose results were announced in [4], studies lattices endowed
with a metric. While K. Menger, V. Glivenko and other mathematicians in-
vestigate lattices equipped with a valuation which plays the role of a norm in
analysis and which in particular induces a distance, Benado takes the opposite
way. He denes a distance-like function which he studies in great detail; in
particular this function induces a valuation of the intervals of the lattice.
Following M. Frechet, Benado denes a semidistance on a lattice L as a
function d : L
3
R satisfying, for all x, y, z L,
d(x, y) 0 ,
d(x, y) = 0 x = y ,
d(x, y) = d(y, x) ;
if, moreover, d fulls the triangle inequality
d(x, z) d(x, y) +d(y, z) ,
then d is called a distance or a metric. We refer to a lattice equipped with a
semidistance or a distance as an sd-lattice or a d-lattice, respectively.
8
Thus
every d-lattice is a metric space. A semidistance d on L and the lattice L
itself are called regular or convex or concave according as the identity d(x, y) =
d(x y, x y) or d(x, y) d(x y, x y) or d(x, y) d(x y, x y) holds; note
that a semidistance may have none of these properties. The semidistance and
the lattice are said to be additive provided
x y z =d(x, z) = d(x, y) +d(y, z) .
Taking x := x y and z := x y, we see that every additive sd-lattice satises
the weak additivity property
9
d(x y, x y) = d(x y, x) +d(x, x y) = d(x y, y) +d(y, x y) .
The functions
H
1
(x, y) = d(x y, x) d(y, x y) ,
H
2
(x, y) = d(x y, y) d(x, x y) ,
called torsions, are essentially due to D. Barbilian [1964]. Note that H
2
(x, y) =
H
1
(y, x). Let us say that an sd-lattice is torsion free if H
1
= 0 identically,
or equivalently, H
2
= 0 identically. This amounts to the equivalent identities
d(x y, x) = d(y, x y and d(x y, y) = d(x, x y).
Benado studies the relationships between the above and other properties.
Here are a few typical results.
If an sd-lattice satises
8
Benado uses the terms metrisable lattice and metric lattice, respectively. Yet the estab-
lished terms metrisable topological space and metric lattice have dierent meanings.
9
Called elementary additivity by Benado.
26
(0) x y a x y =d(a, x) +d(a, y) = d(a, x y) +d(a, x y) ,
then it is regular, torsion free, weakly additive and distributive.
To prove this we write down the hypothesis on a, x, y taking in turn a :=
x, a := y, a := x y, a := x y :
(1) d(x, y) = d(x, x y) +d(x, x y) ,
(2) d(y, x) = d(y, x y) +d(y, x y) ,
(3) d(x y, x) +d(x y, y) = d(x y, x y) ,
(4) d(x y, x) +d(x y, y) = d(x y, x y) .
The sums (1) + (2) and (3) + (4) yield
2d(x, y) = d(x y, x) +d(x y, y) +d(x, x y) +d(y, x y) = 2d(x y, x y) ,
hence the lattice is regular. Now (1) together with the regularity condition and
(4) yield
d(x, x y) +d(x, x y) = d(x, y) = d(x y, x y) = d(x, x y) +d(y, x y) ,
hence d(x, xy) = d(y, xy), that is, the lattice is torsion free. This transforms
(3) into d(y, xy)+d(xy, y) = d(xy, xy), showing that d is weakly additive.
It is well known that a lattice is distributive if and only if it satises
a x = a y & a x = a y =x = y .
So we assume the elements a, x, y full ax = ay and ax = ay and must
prove that x = y. But these elements satify a y x a y, so that taking
a := x and x := a in the hypothesis (0) we get
d(x, a) +d(x, y) = d(x, a y) +d(x, a y) .
Using again the hypothesis (0) for a x x a x and the hypothesis on
a, x, y, we obtain
d(x, a) = d(x, a x) +d(x, a x) = d(x, a y) +d(x, a y) .
By comparing the results we obtain d(x, a) + d(x, y) = d(x, a). So d(x, y) = 0,
therefore x = y, completing the proof.
Any additive distance is convex.
For the hypotheses imply
d(x y, x y) = d(x y, x) +d(x, x y) ,
d(x y, x y) = d(x y, y) +d(y, x y) ,
d(x, y) d(x, x y) +d(x y, y) ,
27
d(x, y) d(x, x y) +d(x y, y) ,
hence 2d(x, y) 2d(x y, x y), as was claimed.
Any additive regular distance is torsion free.
For the hypotheses imply
d(x y, y) +d(y, x y) = d(x y, x y) = d(x, y) d(x, x y) +d(x y, y) ,
d(x y, x) +d(x, x y) = d(x y, x y) = d(x, y) d(x, x y) +d(x y, y) ,
therefore d(y, xy) d(x, xy) and d(xy, x) d(xy, y), that is, H
1
(x, y) =
0.
The following converse holds: any torsion free additive regular semidistance
is a distance. Therefore an additive regular semidistance lattice is a distance
lattice if and only if it is torsion free. A large part of the paper is devoted to
these lattices, called elementary d-lattices. Other classes of sd-lattices are also
introduced and studied, many results being expressed in terms of valuation. This
means that if x y, then d(x, y) is interpreted as a valuation of the interval
[x, y] = z L [ x z y and this is written in the form
d(x, y) = v(
y
x
) .
The research, suggested by previous works of Glivenko [1936] and Barbilian
[1946], generalizes several results of these authors.
Benado has gone farther in [17], where he introduced the concept of a normal-
ized multilattice (multistructure normee). This means a multilattice M equipped
with a map v : M M (valuation) which satises the conditions
a < b =v(a) < v(b) ,
a b ,= ,= a b =v(a) +v(b) = v(m) +v(d) ( m a b) ( d a b) .
The following theorems generalize well-known results of Glivenko for lattices:
1. Every normalized multilattice is modular.
2. Every directed normalized multilattice is a metric space with respect to
the distance (a, b) = v(m) v(d).
Another approach was taken by Kolibiar [Kol1], who dened a metric mul-
tilattice as a multilattice which is also a metric space with respect to a distance
which satises the following supplementary conditions:
a b c =abc, x a b =axb, x a b =axb ,
where xyz denotes metric betweenness, i.e., (x, y) + (y, z) = (x, z). The
directed metric multilattices coincide with Benados normalized multilattices.
Two metric lattices M and M

are said to be m-equivalent if there is a bijec-


tion : M M

such that abc (a)(b)(c). In [Kol1], m-equivalence


is studied in some detail. Sample result: two directed distributive metric mul-
tilattices M and M

are m-equivalent if and only if there are two multilattices


28
A
1
and A
2
such that M A
1
A
2
and M

A
1


A
2
, where denotes iso-
morphism and

A is the dual of A. In [Kol2], necessary and sucient conditions
are given for a metric space to be partially orderable so as to become a metric
multilattice.
A related concept was introduced by Jakubk [Jak4]. A graph isomorphism
between two multilattices M, M

is a bijection : M M

such that two


elements a, b M are neighbouring (benachtbar) if and only if (a), (b) are
neighbouring. The meaning of this term is that a b or b a, where x y
means that x < y and there is no z satisfying x < z < y. Besides, a multilattice
M is said to be of nite length (langenendlich) if every bounded chian in M is
nite. Jakubk has proved a theorem which can be paraphrased as follows: two
directed distributive multilattices of nite length are graph-isomorphic if and
only if they are m-equivalent.
D4. Diametric spaces
The theory of diametric spaces is the last eld initiated and explored by
Benado. The papers devoted to this subject are [54]-[58] and a few manuscripts
which are announced in the references of the printed papers and which appar-
ently have remained unpublished.
A diametric space is a set E endowed with a function : EE R which
satises the conditions
(E1) (a, b) = (b, a) ,
(E2) (a, b) +(c, c) (a, c) +(b, c) .
The elements of E are called bodies (Korper), while (a, b) represents the spatial
dimension (Maraum) or diameter of a and b.
Several examples of diametric spaes are listed in [54], but unfortunately the
reader is referred to an unpublished manuscript for details. However the review
MR38#3774 to [55] makes explicit the following diametric space. Let E be the
set of all spheres of a metric space and take (a, b) = (a, b) +r(a) +r(b), where
(a, b) is the distance between the centers of a and b, while r(c) is the radius of
c. Having this example in mind, the motivation of most of Benados denitions
becomes clear.
In every diametric space E the following function is also introduced:
(E3) 2(a, b) = 2(a, b) (a, a) (b, b) .
It is easy to see that is a quasimetric, that is, (a, a) = 0, (a, b) = (b, a)
and (a, b) (a, c) + (c, b). Conversely, if is a quasimetric on a set E and
: E R, then E becomes a diametric space with
(E4) (a, b) = (a, b) +(a) +(b) .
Having in view that (a, b) = 0 &(a, a) = (b, b) (a, b) = (a, a) =
(b, b), the following equivalence relation on E is taken in the role of equality
on E:
29
(E5) a = b (a, b) = (a, a) = (b, b) .
To prove the transitivity of this relation, suppose
(a, b) = (a, a) = (b, b) = (c, c) = (b, c) .
Then (a, b) + (c, c) (a, c) + (c, b) implies (c, c) (a, c), while (a, c) +
(b, b) (a, b) + (b, c) implies (a, c) (a, b) = (c, c). Therefore (a, c) =
(c, c) and we have also (a, a) = (c, c).
The following property is important:
(E6) (b, a) (a, a) =(b, c) (a, c) .
Indeed, from (b, a) (a, a) we infer
(b, c) +(b, a) (b, c) +(a, a) (b, a) +(a, c) ,
hence (b, c) (a, c).
Having in view that (a, b) +
1
2
(b, b)
1
2
(a, a) (a, b) (a, a), the
relation b is interior to A, written b a, is dened by
(E7) b a (b, a) (a, a) .
This relation is a partial order on E. Indeed, reexivity is trivial, transitivity
follows immediately by (E6), and if (b, a) (a, a) and (a, b) (b, b), then
(b, b) (a, b) and (a, a) (b, a), therefore (b, a) = (a, a) and (a, b) =
(b, b), hence a = b. The order (E6) is called the intrinsic or natural order of
(E, ).
The relative positions b is internally tangent to a, a and b are concentric,
and c is between a and b, are dened by b IT a (b, a) = (a, a), a CC
b (a, b) = 0 and c ab (a, b) = (a, c) +(c, b), respectively.
Besides the above concepts there are many other operations and relations
introduced on a diametric space and whose relationships have been studied in
great detail. Morphisms between diametric spaces have been also introduced.
If (E, ) and E

) are diametric spaces, a map : E E

is said to be a
diametric isomorphism if the equalty (a, b) =

(a, b) holds for every a, b


E; this implies that the associated posets E, ) and (E

) are isomorphic.
There are several variants of this concept, studied in some detail.
Another favourite research theme refers to interpolation properties. To state
many of these properties one needs the following notation. If is a binary
relation on E and A, B E, then we dene AB ( a A) ( b B) ab;
as usual, we write aB and Ab instead of aB and Ab. The interpolation
properties are certain implications which a diametric space may or may not
satisfy. For instance, A B = c A c B and A IT B = c A c B,
where is the natural order and IT is the relation internally tangent, are
two interpolation properties. The relationships between many interpolation
properties have been studied, with the aim of establishing sucient conditions
for their validity.
30
Another concern of Benado was the construction of diametric spaces from
given partially ordered sets. Here is a sample theorem. Recall that the concept
of a ramied set, introduced and studied by D. Kurepa, means any poset whose
order ideals are chains. Suppose (K, ) is a dually directed ramied set for
which there is a strictly increasing function f : K R. Then K can be
made into a diametric space (K, ) which satises (a, b) = max((a, a), (b, b))
whenever a and b are comparable. Besides, the natural order of (K, ) is ner
than the original order of K, i.e., a b =(a, b) (b, b).
Denote by H the hypothesis of the above theorem. A certain condition (
has been found such that H&( = coincides with the natural order = (.
Part E
Papers of Mihail Benado
1. Series canoniques principales du probl`eme de Galois. Bull. Math. Soc.
Roumaine Sci. 47(1946), 49-61.
2. Nouveaux theor`emes de decomposition et dintercalation attaches `a la
normalite . C.R. Acad. Sci. Paris 228(1949), 529-531. MR 10, p.502.
3. Le fondement axiomatique du theor`eme Jordan-Holder relatif aux series
principales. C.R. Acad. Sci. Paris 229(1949), 332-334. MR 11, p.309.
4. Teoria structurilor metrizabile. Bul. Sti. Acad. R.P. Romane A1(1949),
353-359. MR 12, p.237.
5. Not iunea de normalitate si teoremele de descompunere ale algebrei. Stud.
Cerc. Mat. 1(1950), 282-317. MR 16, p.212.
6. Normalitatea si teorema Jordan-Holder. Bul. Sti. Ser. Mat. Fiz.
Chim. 2(1950), 557-560.
7. Normalitatea Uzkov si teorema Jordan-Holder. Com. Acad. R.P.Romane
1(1951), 7-11. MR 17, p.7.
8. Teoria structurilor metrizabile. Stud. Cerc. Mat. 2(1951), 45-106;
erratum 285. MR 16, 275i.
9. Asupra teoremei de ranare a lui O. Schreier. Com. Acad. R.P. Romane
1(1951), 1021-1023. MR 17, p.7.
10. Asupra teoremelor de descompunere ale algebrei. Stud. Cerc. Mat.
3(1952), 263-288. MR 16, p.212.
11. Mult imile part ial ordonate si teorema de ranare a lui Schreier. Bul.
Sti. Sect . Sti. Mat. Fiz. 4(1952), 585-591. MR 15, p.595.
12. Asupra unei generalizari a not iunii de structura. Bul. Sti. Sect . Sti.
Mat. Fiz. 5(1953), 41-48. MR 16, p.668.
13. Teoria abstracta a relat iilor de normalitate. Stud. Cerc. Mat. 4(1953),
69-120. MR 16, p.212.
14. Les ensembles partiellement ordonnes et le theor`eme de ranement de
Schreier. I. Czechoslovak Math. J. 4(79)(1954), 105-129. MR 16, p.668.
15. Asupra teoriei divizibilitat ii. Bul. Sti. Sect . Sti. Mat. Fiz. 6(1954),
263-270. MR 16, p.668.
31
16. Asupra unei probleme a lui Garrett Birkho. Bul. Sti. Sect . Sti. Mat.
Fiz. 6(1954), 703-739. MR 17, p.341.
17. Les ensembles partiellement ordonnes et le theor`eme de ranement de
Schreier. II. Czechoslovak Math. J. 5(80)(1955), 308-344. MR 17, p.937.
18. Asupra descompunerii unui grup n produs direct. I. Bul. Sti. Sect . Sti.
Mat. Fiz. 7(1955), 241-248. MR 17, p.1050.
19. Asupra descompunerii unui grup n produs direct. II. Bul. Sti. Sect .
Sti. Mat. Fiz. 7(1955), 249-254. MR 17, p.1050.
20. Asupra unei probleme din teoria condit iilor de normalitate Ore. Com.
Acad. R.P. Romane 5(1955), 1241-1243. MR 17, p.937.
21.

Uber die allgemeine Theorie der regularen Produkte von Herrn O.N.
Golovin. I. Math. Nachr. 14(1955), 213-234 (1956). MR 18, p.871.
22. Asupra descompunerii unui grup n produs direct. III. Bul. Sti. Sect .
Sti. Mat. Fiz. 8(1956), 5-10. MR 19, p.248.
23. Remarks on a paper by O.Ore. (Russian). Rev. Math. Pures Appl.
1(1956), no.2, 5-12. MR 18, p.275.
24. Rectication `a mon travail Les ensembles partiellement ordonnes et
le theor`eme de ranement de Schreier. II. (Theorie des multistructures).
Czechoslovak Math. J. 6(81)(1956), 287-288. MR 19, p.243.
25. Sur la theorie generale des produits reguliers. C.R. Acad, Sci. Paris
243(1956), 1092-1093. MR 18, p.279.
26. Sur une interpretation topologique de la notion de normalite unitaire.
Bull. Sci. Math. 81(1957), Avril-Juin, 87-112. MR 20, 1643.
27. Sur la theorie generale des produits reguliers. C.R. Acad. Sci. Paris
244(1957), 1595-1597. MR 19, p.385.
28. Sur la theorie generale des produits reguliers. C.R. Acad. Sci. Paris
244(1957), 1702-1704. MR 19, p.528.
29. Sur la theorie generale des produits reguliers. C.R. Acad. Sci. Paris
245(1957), 267-270. MR 19, p.528.
30.

Uber die allgemeine Theorie der regularen Produkte von Herrn O.N.
Golovin. II. Math. Nachr. 16(1957), 137-194. MR 20, 1642.
31.

Uber die allgemeine Theorie der regularen Produkte von Herrn O.N.
Golovin. III. Math. Nachr. 21(1960), 1-36. MR 22A, 12135.
32. Sur la theorie generale des produits reguliers de Monsieur O.N. Golovine.
V.
10
Publ. Sci. Univ. Alger Ser. A 4(1957), no.2, 111-143. MR 20, 6470.
33. Sur la fonction de Mobius. C.R. Acad. Sci. Paris 246(1958), 863-865.
MR 20, 6370.
34. Sur la theorie generale des ensembles partiellement ordonnes. C.R.
Acad. Sci. Paris 247(1958), 2265-2268. MR 20, 5745.
35. Sur la fonction de Mobius. C.R. Acad. Sci. Paris 246(1958), 2553-2555.
MR 20, 6371.
36. Bemerkungen zur Theorie der Vielverbande. Math. Nachr. 20(1959),
1-16. MR 22A, 19.
10
It seems that No.IV of this series has never been published.
32
37. Remarques sur un theor`eme de Monsieur Oleg N. Golovine. Czechoslo-
vak Math. J. 9(84)(1959), 475-484. MR 23A, 193.
38. Sur la theorie generale des ensembles partiellement ordonnes. Publ. Sci.
Univ. Alger Ser. A 7(1960), 5-39. MR 26, 2370.
39. La theorie des multitreillis et son role en alg`ebre et en geometrie. Publ.
Sci. Univ. Alger Ser. A 7(1960), 41-58. MR 27, 2447.
40. Bemerkungen zur Theorie der Vielverbande. IV.

Uber die Mobiussche
Funktion. Proc. Cambridge Philos. Soc. 56(1960), 291-317. MR 22A, 10929.
41.

Uber den Kommutatrixbegri. Proc. London Math. J. (3) 10(1960),
514-530. MR 22A, 11034.
42. Sur la theorie generale des ensembles partiellement ordonnes. I. Proc.
Japan Acad. 36(1960), 590-594. MR 24A, 3089.
43. Sur la theorie generale des ensembles partiellement ordonnes. II. Proc.
Japan Acad. 36(1960), 595-597. MR 24A, 3089.
44. Sur la theorie generale des ensembles partiellement ordonnes. III. Proc.
Japan Acad. 36(1960), 636-638. MR 24A, 3089.
45. Sur une caracterisation abstraite des alg`ebres de Boole. I. C.R. Acad.
Sci. Paris 251(1960), 622-623. MR 22A, 7962.
46. Sur une caracterisation abstraite des alg`ebres de Boole. II. C.R. Acad.
Sci. Paris 251(1960), 835-836. MR 22A, 7963.
47. Zur abstrakten Begr undung der F uhrertheorie. Math. Japon. 6(1960/61),
1-25. MR 28, 5014.
48. Sur une propriete dinterpolation remarquable dans la theorie des en-
sembles partiellement ordonnes. Acta Sci. Math. (Szeged) 22(1961), 1-5. MR
24A, 3091.
49. Sur la theorie generale des ensembles partiellement ordonnes. IV. Proc.
Japan Acad. 37(1961), 83-84. MR 24A, 3090.
50. On the general theory of partially ordered sets. (Russian). Acta Fac.
Rer. Nat. Univ. Comenian. 5(1961), 397-429. MR 27, 61.
51. The theory of multilattices and its signicance in algebra and geometry.
(Russian). Acta Fac. Rer. Nat. Univ. Comenian. 5(1961), 431-448. MR 24A,
1228.
52. Sur la theorie generale des ensembles partiellement ordonnes. I-III.
Publ. Sci. Univ. Alger Ser. A 8(1961), 90 pp. MR 27, 5701.
53. Remarques sur la theorie des multitreillis. VI. Mat.-Fyz.

Casopis Sloven.
Akad. Vied 14(1964), 163-207. MR 31, 94.
54. Geometrische Kontinua und diametrische Raume. Spisy Prrod. Fak.
Univ. Purkyne (Brno) Ser. A 29(1965), 469-473. MR 33, 4744.
55. Proprietes dinterpolation des espaces diametriques. Math. Japon.
13(1968), 5-19. MR 38, 3774.
56. Remarques sur les subisometries dun espace diametrique. Mem. Fac.
Ci. Habana Ser. Mat. 1(1968), no.7, fasc.6, 11-18. MR 47, 5840.
57. A propos du second theor`eme disomorphisme pour les espaces diametri-
ques. Math. Japon. 13(1968), 21-31. MR 41, 6730.
58. Ensembles ordonnes, fonctions reelles, espaces diametriques. Publ. Inst.
Math. (Beograd) (NS) 9(23) (1969), 143-152. MR 40, 63.
33
59. Ensembles ordonnes et fonctions monotones. Univ. Lisboa Revista Fac.
Ci. A(2) 14(1972/73), 165-194. MR 48, 8313.
Sources of Mihail Benado
R. Baer
1945. Representations of groups as quotient groups. I. Trans. Amer. Math.
Soc. 58, 295-346.
D. Barbilian
1946. Metrisch-konkave Verbande. Disquisitiones Math. Phys. 5, 3-63.
1951. Curs de teoria grupurilor si structurilor. (lito) Univ. Bucuresti.
1953. Normalitat i locale si integral-involutive. Stud. Cerc. Mat. 4, 29-62.
1956. Teoria Aritmetica a Idealelor (n Inele Necomutative). Ed. Academiei
Romane.
1957. Argumentul lui Euclid pentru innitatea numerelor prime. Stud. Cerc.
Mat. 7, 7-72.
G. Birkhoff
1948. Lattice Theory. Amer. Math. Soc.
M.L. Dubreil-Jacotin, R. Croisot, L. Lesieur
1953. Lecons sur la Theorie des Treillis,des Structures Algebriques Or-
donnees et des Treillis Geometriques. Gauthier-Villars, Paris.
H. Fitting
1936.

Uber die Existenz gemeinsamer Verfeinerungen bei direkten Produkt-
zerlegungen einer Gruppe. Math. Z. 39, 385-400.
1938. Beitrage zur Theorie der Gruppen endlicher Ordnung. Jahresber.
Deutscher Math. Verein 48, 77-141.
V. Glivenko
1936. Geometrie des syst`emes de choses normees. Amer. J. Math. 58,
799-828; 59(1937), 941.
O.N. Golovin
1950. Nilpotent products of groups (Russian). Mat. Sb. (NS) 27, 427-454.
1951a. Meta-abelian products of groups (Russian). Mat. Sb. (NS) 28, 431-
444.
1951b. On isomorphism problems of nilpotent decompositions of a group.
(Russian). Mat. Sb. (NS) 28, 445-452.
O.N. Golovin, N.P. Goldina
1951. Subgroups of free meta-abelian groups (Russian). Mat. Sb. 37(79),
323-336.
V. Ko rnek
1941. Das Schreiersche Satz und das Zassenhaussche Verfahren in Verbanden.
Vestnik

Cesk. Spol. Nauk Trida Mat.-Prrod. Rocnik 1, 29.
W. Krull
34
1924. Axiomatische Begr undung der allgemeinen Idealtheorie. Sitz. Phys.-
Math. Soz. Erlangen 56, 47-63.
A.G. Kurosh
1940. The Jordan-Holder theorem in arbitrary structures. (Russian). Col-
lection of Papers Deducated to Acad. D.A. Grav`e. (Russian). 110-116.
1953. Theory of Groups. Second ed. (Russian). Moscow.
O. Ore
1935. On the foundations of abstract algebra. I. Ann. Math. 26, 406-437.
1937. On the theorem of Jordan-Holder. Trans. Amer. Math. Soc. 41,
266-275.
A.I. Uzkov
1938. On the theorem of Jordan-Holder. Recueil Math. (Moscou) NS 14,
31-43.
M. Ward, R.P. Dilworth
1939. Residuated lattices. Trans. Amer. Math. Soc. 45, 335-354.
H. Zassenhaus
1937. Lehrbuch der Gruppentheorie. Leipzig and Berlin.
Further research
We are planning a possible sequel of this paper, including references to sev-
eral articles that cite Benado, including more remarks like those in Section C3,
and with substantial references to articles that develop the theory of multilat-
tices and/or use it as a tool in computer science. We would also refer to some
papers that do not cite Benado, but nevertheless contain ideas in the same direc-
tion. Such a study is appropriate in the present context since it will show that
his work, far from remaining isolated, provides remarkable ideas, some of them
not yet explored, for (i) the theory of partially ordered algebraic systems more
general than the lattice ordereded algebraic systems, (ii) a more comprehensive
theory of posets suitable for applications in the algebraic theoretical conputer
science, and (iii) a theory of general order structures. Obviously, only a few
examples can be shortly indicated here, the substantial part remaining for the
projected future paper.
As for (i), the starting point consists of papers [14], [15], already quoted here.
We are compiling a bibliography which so far contains about 15 entries referring
to multilattices, and among them 9 mention multilattices in their titles, e.g.
[Kol3], [Kol4], and [McA1], [McA2], [McA3]. The recent paper [GaP] considers
multilattice groups, quoting [Vai] and [Jak3]. The latter solves a problem raised
in [Vai]; namely, Jakubk shows that every -group can be embedded into a
J-group (as introduced in [Vai]) which is not a multilattice group.
For (ii) consider again an example, the Plotkin order; cf. [Plot]. Given a
poset A and a nite subset U of A, a set V if minimal upper bounds for U
is said to be complete, if for every upper bound x of U, there is y V with
y x. A characteristic feature for a Plotkin order is that every nite subset has
35
a complete set of minimal upper bounds, i.e., the characteristic requirement in
order to hve an upper semimultilattice. The paper [GuS] studies the theory of
Plotkin orders for semantic domains.
In the last decade, an important school of computer scientists at the Uni-
versity of Malaga is very active in focussing a great deal of research on mul-
tilattices as introduced by Benado in [12] and [14]-[17]. Their works recover
multilattices for use in very diversied contexts, both theoretical and applied
[Mart], [Med]. Since the operations of multi-suprema and multi-inma are no
longer single-valued, their research leads to the theory of hyperstructures as in
[Cord2], [Cord3].
As for (iii), the theory of multilattices developed to some extent according
to [Bir67] is one of the achievements of the school of Kolibiar [Kol1]-[Kol4]
and Jakubk [Jak1], [Jak2], [Jak4]. Their joint paper [JK] on isometries of
multilattice groups has been very inuential; see e.g. [Jas]. One should also
note the connections with universal algebra [Li], [HaLi] and the treatment of
valuations and distances in directed multilattices [Li2].
Acknowledgements
This paper is our tribute to Mihail Benado. We were his students, his courses
have determined our interest in the subject and have remained a reference to
us.
We thank Dragic Bankovic, Gheorghe Costovici, Ioana and Laurent iu Leustean,
who provided several papers of Benado not available in our libraries. Thanks
are also due to Traian Serbanut a and Andrei Popescu for references to Math.
Sci. Net.
References
[And] D.F. Anderson, Factoring in integral domains. Prep Lectures, Trinity
Univ. San Antonio, TX. May 21-25, 2007.
[GSA] G.St. Andonie, Istoria Matematicii n Romania. Ed. Stiint ica, Bu-
curesti 1967.
[Bir] G. Birkho, Lattice-orered groups. Ann. Math. 43(1942), 298-331.
[Bir67] G. Birkho, Lattice Theory. Amer. Math. Soc., Providence, RI 1967.
[Cab] I.P. Cabrera, P. Cordero, G. Gutierez, J. Martinez, M. Ojeda-Aciego,
Fuzzy congruence relations on nd-groupoids. Intern. J. Comput. Math.
86(2009), 1684-1695.
[ChK1] I. Chajda, M. Kolark, A decomposition of homomorphic images of near-
lattices. Acta Univ. Palacki. Olomouc. Fac. Rer. Nat. Math. 45(2006), 43-
52.
36
[ChK2] I. Chajda, M. Kolark, Ideals, congruences and annihilatora on nearlat-
tices. Acta Univ. Palacki. Olomouc, Fac. Rer. Nat. Math. 46(2007), 25-33.
[ChK3] I. Chajda, M. Kolark, Nearlattices. Discrete Math. 308(2008), 4906-
4913.
[Cord] P. Cordero, G. Gutierez, J. Martinez, I.P. de Guzman, A new alge-
braic tool for automatic theorem provers. Multisemilattice: A structure to
improve the eciency of powers in temporal logics. Ann. Math. Articial
Intelligence, 42(2004), 369-398.
[Cord2] P. Cordero, G. Gutierez, J. Martnez, M. Ojeda-Aciego, I. de las Pe nas,
Congruence relations on hypergroupoids and nd-groupoids. In: 14th Spanish
Conference on Fuzzy Logic and Technology, 2008 (to appear).
[Cord3] P. Cordero, C. Gutierez, J. Martnez, M. Ojeda-Aciego, I. de las Pe nas,
Congruence relations on multilattices. In: 8th International FLINS Con-
ference on Computational Intelligence in Decision and Control, 2008 (to
appear).
[DND] A. Derghan Nezhad, B. Davvaz, An introduction to the theory of Hv-
semilattices. Bull. Malays. Math. Sci. Soc. (2) 32(3)(2009), 375-390.
[Doc] Algebra 1 (MATH 6130), Handout 7 (November 30-December 2, 2009).
[Dra] H. Draskovicova, M. Kolibiar, L.A. Skornyakov, Ordered Sets and Lat-
tices. Vol.2 . AMS Translations 1992.
[DP] A. Dvurecenskij, S. Pulmanova, New Trends in Quantum Structures
(Mathematics and Its Applications). Kluwer Acad. Publ., Dordrecht 2000.
[Fu] L. Fuchs, Partially Ordered Algebraic Systems. Int. Ser. Monographs Pure
Appl. Math. vol. 28, Pergamon Press, Oxford 1963.
[GaP] B.J. Gardner, M.M. Parmenter, Some classes of directed groupoids. Al-
gebra Discrete Math. 1(2009), 44-58.
[GuS] C.A. Gunter, D.S. Scott, Semantic domains. In: J. van Leeuwen, ed.,
Handbook of Theoretical Computer Science, vol.B: Formal Methods and
Semantics. MIT Press, 1991.
[HaLi] A. Haviar, J. Lihova, Varieties of posets. Order 22(2005), 343-356.
[Jak1] J. Jakubk, On the axioms of multistructures. (Russian). Czechoslovak
Math. J. 6(81) (1956),426-430. MR 20, 6369.
[Jak2] J. Jakubk, The isomorphism of graphs of multilattices. (Slovak). Acta
Fac. Rer. Nat. Univ. Comen. Math. 1(1956), 256-264.
[Jak3] J. Jakubk,

Uber halbgeordneten Gruppen mit verallgemeinerter Jordan-
sche Zerlegung. Rev. Roumaine Math. Pures Appl. 9(1964), 187-190.
37
[Jak4] J. Jakubk, On strong superlattices. Math. Slovaca 44(1994), No.2, 131-
138.
[JK] J. Jakubk, M. Kolibiar, Isometries of multilattice groups. Czechoslovak
Math. J. 33(108)(1983), 602-612.
[Jas] M. Jasem, Isometries in non-abelian multilattice groups. Math. Slovaca
46(1990), 491-496.
[IJJ1] I.J. Johnston, Modularity and distributivity in directed multilattices. PhD
Thesis, Queens Univ. Belfast, 1985.
[IJJ2] I.J. Johnston, Some results involving multilattice ideals and distributivity.
Discrete Math. 83(1990), 27-35.
[Kol1] M. Kolibiar,

Uber metrische Vielverbande. I.. Acta Fac. Rer. Nat. Univ.
Comen. Math. 4(1959), 187-203. MR 24, A3100.
[Kol2] M. Kolibiar,

Uber metrische Vielverbande. II.. Acta Fac. Rer. Nat. Univ.
Comen. Math. 7(1963), 629-637. MR 31, 98.
[Kol3] M. Kolibiar, Direct factors of multilattice groups. Arch. Math. (Brno)
26(1990), 121-127.
[Kol4] M. Kolibiar, Direct factors of multilattice groups. II. Arch. Math. (Brno)
28(1992), 83-84.
[Li] J. Lihova, Varieties of directed multilattices. Math. Slovaca 40(1990), 233-
244.
[Li2] J. Lihova, Valuations and distance functions on directed multilattices.
Math. Slovaca 46(1996), 143-155.
[LiRe] J. Lihova, K. Repask y, Congruence relations on and varieties of directed
multilattices. Math. Slovaca 38(1988), 105-122.
[LjEv] E.S. Ljapin, A.E. Evseev, The Theory of Partial Algebraic Operations.
Springer-Verlag new York, 2009.
[Lot] M. Lothaire, Applied Combinatorics on Words. Cambridge Univ. Press,
2005.
[MA] E.G. Manes, M.A. Arbib, Algebraic Approach to Program Semantics.
Springer-Verlag, Berlin.
[Mark] G. Markowski, Posets, lattices and computer science. DIMACS Work-
shop on Applications of Lattices and Ordered Sets to Computer Science,
DIMACS Center, July 8-10, 2003.
[Mart] J. Martinez, I.P. de Guzman, P.Cordero, Generalizations of lattices via
non-deterministic operators. Discrete Math. 295(2005), 107-141.
38
[Maru] Y. Maruyama, Algebraic study of lattice-valued logic and lattice-
valued modal logic. Lecure Notes Comput. Sci. vol. 5378/2009, Springer,
Berlin/Heidelberg, 2009.
[McA1] D.B. McAllister, On multilattice groups. Proc. Cambridge Philos. Soc.
61(1965), 621-638.
[McA2] D.B. McAllister, On multilattice groups. II. Proc. Cambrifge Philos.
Soc. 62(1966), 149-164.
[McA3] D.B. McAllister, Multilattice groups with a lexico basis. Proc. Roy. Irish
Acad. Sect. A 71(1971), 53-72.
[Med] J. Medina, M. Ojeda-Aciego, J. Ruiz-Calvi no, Fuzzy logic programming
via multilattices. Fuzzy Sets Systems 158(6) (2007), 674-688.
[MitKo] J. Mittas, M. Konstantinidou, Sur une nouvelle generalisation de la no-
tion de treillis. Les supertreillis et certaines de leurs proprietes generales.
Ann. Sci. Univ. Clermont-Ferrand 2(94), Ser. Math. no.25(1989), 61-83.
[NC] A.S. Noor, W.H. Cornish, Multipliers on nearlattices. Comment. Math.
Univ. Carol. 27(1986), 815-827.
[Plot] G.D. Plotkin, A powerdomain construction. SIAM J. Computing 5(1976),
452-287.
[RV] S. Rudeanu, D. Vaida, Semirings in operations research and computer
science: more algebra. Fund. Inform. 61(2004), 61-85.
[Sch] B.S.W. Schroder, Ordered Sets: an Introduction. Birkhauser.2002.
[Va] D. Vaida, Groupes ordonnes dont ls elements admettent une decomposition
jordanienne generalisee. Publ. Fac. Sci. Univ. J.E. Purkyne, vol A28(1964)
No.9 (Tagung uber geordnete Mengen, Brno, November 1963), p.494.
[Vai] D.Vaida,Groupes ordonnes dont les elements admettent une decomposition
jordanienne generalisee. Rev. Roumaine Math. Pures Appl. 10(1964), 930-
948.
[Vai1] D. Vaida, On partially additive semirings and applications. Mult.-Val.
Logic 6(2001), 251-256.
[Vai2] D. Vaida, Note on a more general setting for complemented elements. I.
Remarks from semiring theory. J. Mult.-Val. Logic Soft Comput. 12(2006),
275-284.
39

You might also like