United sight to an algebraic operations and convergence

by Gintaras VALIUKEVIČIUS

Abstract. Algebraic operations and algebraic spaces was invented to deal with various algebraic problems.
Nowadays we can say that was topologization of algebra. The theory of cathegories became an abstract the-
ory, and classical algebraic spaces became only an examples of simplest convergence spaces. In this article
we deal with convergences as instances of multivalued appointments and define the continuity as property
of commuting squares. This is beginning point for axiomatisation of multivalued mappings continuity prop-
erties.
arXiv:1004.3248v1 [math.GN] 19 Apr 2010

Keywords: convergence, continuity of multivalued mappings. Math subj. classification 2000: 46 H05, 54
B30.

1. Algebras and their homomorphisms

1.1 Algebra A will be understood as the carrying set A with some operations. The algebraic operation on
the set A defined by Birkhoff [2] defined is any function over the set of n-tuples f : An −→ A. The set of
operations names Σ is called the signature of algebra.
Example. In the space X the potential set A = 2X is defined as set of all subsets A ⊂ X. It can be
considered as Boole algebra with such operations:
The least element 0 : 1 −→ A is defined as a function from the singlepoint set 1 = {0} to the set of all
subsets 2X , appointing the wide set Ø ∈ 2X for the unique point 0 ∈ 1.
The largest element t : 1 −→ A is defined as a function appointing the whole space X ∈ 2X for the
unique point 0 ∈ 1.
The complement c : A −→ A is defined as a function, appointing the complement Ac = X \ A for every
partial set A ⊂ A.
The intersection is defined as a function ∧ : A × A −→ A, appointing the set of common points A ∩ B
for every pair of partial sets hA, Bi.
The union is defined as a function ∨ : A × A −→ A appointing the set of points belonging to either set
A ∪ B for every pair of partial sets hA, Bi.

A σ-algebra A will be an example of nonclassical algebra. In this algebra we have countable operations
AN → A. Usually such operations are called as convergence of countable sequences.
1.2 The application f : X −→ Y will be called morphism from the set X to the set Y . The first set X will
be called a source space, and the second set Y will be called a target space of morphism. For the point in the
source space x ∈ X the application appoints the point in the target space f (x) ∈ Y , sometimes we denote

y = f (x) ↑ x ∈ X .

We have the law of composition for two application with intermediant space, i.e. the target space of the
first application f : X −→ Y coincides with the source space of second application g : Y −→ Z. The result
will be f ◦ g : X −→ Z an application from the source space of the first application to the target space of
the second application. For the point x ∈ X it appoints the point z ∈ Z which is calculated

z = (f ◦ g)(x) = g(f (x)) .

For every set X we define the identity application IdX : X −→ X which for the point x ∈ X appoints the
same point. It will be neutral for the law of composition, i.e. the identity application doesn’t change such
product, if this product is possible
IdX ◦ f = f = f ◦ IdY .
Bijective application f : X −→ Y will have an inverse application g : Y −→ X for the law of composition

f ◦ g = idX , g ◦ f = IdY .

1
It will be called isomorphism of sets X and Y .
1.3 We remind the construction of set product. At first we define the two sets multiplication. The multipli-
cation of two sets X and Y is defined as the set of ordered pairs

X × Y = {ha1, a2i : a1 ∈ X, a2 ∈ Y } .

Two sets multiplication is also defined for morphisms. For any pair of sets morphisms f : X −→ Z and
g : Y −→ W we define a morphism between sets multiplications

f × g : X × Y −→ Z × W

which appoints
(f × g)(hx, yi) = hf (x), g(y)i ∈ Z × W ↑ hx, yi ∈ X × Y .
Such multiplication maintains the identity morphisms

(IdX × IdY ) = IdX×Y

and the composition of morphisms, i.e. for the pairs of composable morphisms hf1 , f2 i and hg1 , g2 i we have
the pair of composable products hf1 × g1 , f2 × g2 i and equality

(f1 ◦ g1 ) ◦ (f2 × g2 ) = (f1 × g1 ) ◦ (f2 × g2 ) .

Therefore the two sets mwltiplication can be concidered as bifunctor.

The sets product can be generated by two sets multiplication. The n-tuple hx1, x2, . . . , xni ∈ An can be
considered as vector with n components, i.e. the mapping from index set [n] to the space of values A. Any
multiplication of n sets A1 , A2 , . . . , An can be identified with the set of n-tuples with only one isomorphism
of sets. Therefore any two multiplications of the same sets are identified with distinguished isomorphisms
of sets, i.e. we indicate unique one from possible isomorphisms, which identifies taken two multiplications.
The sets product is defined as abstract notion in the category of sets, and sets multiplication will be only
concret presentation of such abstract notion.
The singlepoint set I = {∗} is neutral for set multiplication, i. e for this set we have distinguished
isomorphisms lX : I × X −→ X and rX : X × I −→ X which appoint

lX (h∗, xi) = x , rX (hx, ∗i) = x ↑ x ∈ X .

These isomorphisms are natural for morphisms of sets f : X −→ Y , i.e. we have the commuting diagrams
lX rX
I ×X /X X ×I /X

1×f f f ×1 f
   
I ×Y lY
/Y Y ×I rY
/Y

Also we have coinciding distinguished isomorphisms for one point set I × I −→ I,

lI (h∗, ∗i) = ∗ = rI (h∗, ∗i) .

Next we indicate the distinguished isomorphism aXY Z : (X × Y ) × Z −→ X × (Y × Z) defining the

associative equality
aXY Z (hhx, yi, zi) = hx, hy, zii ↑ x ∈ X, y ∈ Y, z ∈ Z .
They also are natural for morphisms of sets, i.e. for morphisms f : X −→ X ′ , g : Y −→ Y ′ , h : Z −→ Z ′ we
get commuting diagrams
aXY Z
(X × Y ) × Z / X × (Y × Z)

(f ×g) f ×(g×h)
 
(X ′ × Y ′ ) × Z ′ aX ′ Y ′ Z ′
/ X ′ × (Y ′ × Z ′ )

2
Such distinguished izomorphisms we got by identification of taken multiplications with the space of triples
X × Y × Z. Similar identifications also help to prove the pentagonal diagram identity

(X × Y ) × (Z × W )
8
a rrr
LLLa
r LLL
rr &
X × (Y × (Z × W )) ((X × Y ) × Z) × W
33 E
331×a 
3  a×1
(X × (Y × Z)) × W a
/ X × ((Y × Z) × W )

In Mac Lane book [11] the coherence theorems asserts that we can provide the distinguished isomorphisms for
arbitrary finite sets multiplications defined by different binary trees, if these trees according with properties
of associative monoid results the same word. I put so much attention to such usual product of sets, because
it will be an explicit example for another more interesting products.
If we add the distinguished isomorphisms for twisting
tXY : X × Y −→ Y × X , tX,Y (hx, yi) = hy, xi ↑ x ∈ X, y ∈ Y ,
our product becomes commuting one, and we can get all properties of commutative monoid. The realization
of product with the set of pairs remains the same, but now we identify two symmetric products with the
help of twisting isomorphisms.
Otherwise we can deal the product of sets without identifying what singlepoint set will be neutral. Then
for the product of sets we get the properties of the semigroup, without distinguished isomorphisms lX and
rX .
1.4 Next we see see another special properties of set product. Every concret realization of the set product
will be coadjoint functor to diagonal one.
Diagonal functor for every set X apoints the couple of sets < X, X >, and for ervery morphism f : X → Y
appoints the couple of morphisms < f, f >. For noncommuting product we take these couples noncommuting
< f, g >6=< g, f > .
For every set X we define the diagonal morphism
iX : X −→ X × X , iX (x) = hx, xi ↑ x ∈ X ,
and for pair of sets hY, Zi we define the projections of product
p1hY,Zi : Y × Z −→ Y , p1hY,Zi(hy, zi) = y ↑ hy, zi ∈ Y × Z ,
p2hY,Zi : Y × Z −→ Z , p2hY,Zi (hy, zi) = z ↑ hy, zi ∈ Y × Z .
Such morphisms are natural for set morphisms, i.e. for morphisms f : X ′ −→ X, g : Y −→ Y ′ , h : Z −→ Z ′
we have commuting diagrams:
iX p1hY,Zi p2hY,Zi
XO / X ×X Y ×Z /Y Y ×Z /Z
O
f f ×f g×h g g×h h
   
Y iY
/ Y ×Y Y ′ × Z ′ p1 / Y′ Y ′ × Z ′ p2 / Z′
hY ′ ,Z i
′ hY ′ ,Z i
′

The first natural morphism iX defines mapping over set of all morphism pairs
φXhY,Zi : h(X, Y ), (X, Z)i −→ (X, Y × Z) ,
which for the pair of morphism f : X −→ Y , g : X −→ Z appoints composition of morphisms iX ◦ f × g :
X −→ Y ×Z. The second natural morphism pair hp1hY,Zi , p2hY,Zi i defines mapping over the set of morphisms
ψXhY,Zi : (X, Y × Z) −→ h(X, Y ), (X, Z)i .

3
These mappings are natural for the source and target turnings of morphisms, i.e. for earlier morphisms we
have commuting diagrams of mappings
φXhY,Zi ψXhY,Zi
hhX, Y i, hX, Zii / hX, Y × Zi hX, Y × Zi / hhX, Y i, hX, Zii

hhf,gi,hf,hii hf,g×hi (f,g×h) hhf,gi,hf,hii

   
hhX ′ , Y ′ i, hX ′ , Z ′ ii / hX ′ , Y ′ × Y ′ i hX ′ , Y ′ × Y ′ i / hhX ′ , Y ′ i, hX ′ , Z ′ ii
φX ′ hY ′ ,Z ′ i ψX ′ hY ′ ,Z ′ i

The natural mappings φX,hY Zi will define adjunction, and ψX,hY Zi will define coadjunction.
The pair of identities for natural morphisms iX and hp1hY,Zi , p2hY,Zi i

iX ◦ p1hX,Xi = IdX , iX ◦ p2hX,Xi = IdX

provides the first identity of duality

φXhY,Zi ◦ ψXhY,Zi = IdhhX,Y i,hX,Zii ,

and the identity

iY ×Z ◦ p1Y Z × p2hY,Zi = IdY ×Z
provides the second identity of duality

ψXhY,Zi ◦ φXhY,Zi = IdhX,Y ×Zi .

In the case of such identities adjunction and coadjunction are reciprocal isomorphic mappings. Any
product realization with such property is isomorphic with identification of unique isomorphism. This can be
consequence of the theory for the dual functors. We conclude that the multiplication of sets is presentation
of unique abstract product. This abstract product is called Cartesian product of the sets. Every concrete
product will be an implementation of such abstract product. The abstract product is identified with possible
unique change of implementation. The commuting abstract product of the sets has more changes of imple-
mentation, therefore it is more abstract than noncommuting one. Freyd and Scedrov [5] use the commuting
Cartesian product, Mac Lane in [11] has’nt understood what abstract categorical notions mean, and he
makes no explicit differences between various notions of Cartesian products. Commuting abstract product
can be comfortably defined as limit of discrete diagram.
For abstract product we have much more isomorphic implementations. For example we can identify
arbitrary singlepoint set with I using the unique isomorphism f : {x} −→ {∗} which appoints f (x) = ∗,
therefore such singlepoint set {x} becomes neutral for such abstract set product implementation.
The abstract commuting Cartesian product extends the operator of intersection in Lattice theory and
for it the same sign X ∧ Y can be applyed.
1.5 For more complex algebraic operations many sorted algebra can be defined [9]. Let the carrying sets are
indexed by the points from sort set s ∈ S and the name of operation indicates what sort of carrying spaces
must be taken, i.e. the signature Σ is indexed over the space S ∗ × S, where S ∗ denotes the set of all finite
words of alphabet S. Therefore every operation will be a function

fs1,s2,...,sn;s : As1 × As2 × . . . × Asn −→ As .

Example. A directory C : O −→ O′ is understood as a set of arrows from some point in the first set O to
a point in the second set O′ . Points in the sets O, O′ are understood as vertexes of arrows. The arrows
between two vertexes X ∈ O, Y ∈ O′ compound an arrows set C(X; Y ). The first vertex X is called a source
of arrow, the second vertex Y is called a target of arrow. A source is an appointment from the set of arrows
to the set of vertexes s : C −→ O, for each arrow it appoints the source vertex of taken arrow. The target
is an appointment from the set of arrows to the set of vertexes t : C −→ O′ , for each arrow it appoints the
target vertex of taken arrow.
The directory is a sorted algebra with sort set 3

A1 := A , A2 := O , A3 := O ,

4
and two operations f1;2 = s : A −→ O and f1;3 = t : A −→ O.
The category C will be a dirrectory with the same source and target set O. Additionally we must choose
the unit arrows 1 ∈ (x, x) and define the arrows composition

f ◦ g ∈ (x, z) ↑ f ∈ (x, y), g ∈ (y, z) .

1.6 For more general algebras on the carrying set A we need to chose other interesting products An , and
then to define algebraic operations An −→ A as morphisms over such product.
Example. Let we have two instances of directories with some intermediant vertex space D : O1 −→ O2
and D′ : O2 −→ O3 . The product of two directories will be a product bundle for the target and source
appointments
D ⊗O2 D′ : O1 −→ O3 .
It will be a part of product of arrow sets and can be defined as equalizer for target and source appointments,
i.e. it will be the set of arrow pairs for which two maps coincide

D ×O2 D′ = (t = s′ ) = {hf, gi ∈ D × D′ : t(f ) = s′ (g)} .

This new directory product get distinguished isomorphisms from semilattice of set product. This isomor-
phisms are commuting with identity morphisms of factor spaces, therefore the equalize of target and source
appointments is maintained, and we can take the trace of distinguished isomorphisms over the set of equalizer.
There aren’t projections or diagonal morphisms for such directory product.
The singlepoint set can’t be taken as neutral element for directory product. The neutral element will be
a set of vertexes. We have the distinguished isomorphisms for equalities

lD : O1 ⊗O1 D −→ D , rD : D ⊗O2 O2 −→ D .

For the sake of clarity we express the product bundle as the sum of nonintersecting arrow sets
[ [
D = {(a, b) : a ∈ O1 , b ∈ O2 } , D′ = {(a′ , b′ ) : a′ ∈ O2 , b′ ∈ O3 } .

The product of two directories is get by the product of slices

[
D ⊗O2 D′ = {(a, b) × (b, c) : a ∈ O1 , c ∈ O2 } .

The set of vertexes O is understood as directory with identity morphism Id : O −→ O taken as source and
target appointments. There is exactly one arrow over each diagonal pair of vertexes hc, ci. The products by
such directory are defined
[
O1 ⊗O1 D = { {a} × (a, b) : a ∈ O1 , b ∈ O2 } ,
[
D ⊗O2 O2 = {(a, b) × {b} : a ∈ O1 , b ∈ O2 } .
Therefore the distinguished isomorphisms is defined over slices identifying the set product by singlepoint
set with the taken set of arrows

{a} × (a, b) −→ (a, b) , (a, b) × {b} −→ (a, b) .

1.6.1 The set product provides a new algebra with product operations.
Proposition. If the set product interchange with the product of algebra by isomorphisms

X n × Y n −→ (X × Y )n ,

then the set product of carrying spaces provides a new algebra with the set product operations.

5
Proof: It is enough to use the inverse of distinguished isomorphisms

I × I −→ I , X n × Y n −→ (X × Y )n

to define the 0-degree and n-degree operations in X × X

I ×I o I Xn × Y n o (X × Y )n
f1 ×f2 f1 ×f2 f
   
X ×Y 1
/X ×Y X ×Y / X ×Y
1

Such algebra will be called a set product algebra. In the case when the set product is associative and
commuting one, and algebra’s operations can be defined over the set product.
For any pair of directories D1 and D2 the set product D1 × D2 provides a new product directory with
product operations, i.e. the source and target appointments is defined by products

s1 × s2 : D1 × D2 −→ O1 × O2 , t1 × t2 : D1 × D2 −→ O1′ × O2′ .

Example. For the categories the set product interchanges with bundle product, used to define the operations
of category,
(A ⊗O A) × (B ⊗O′ B) −→ (A × B) ⊗O×O′ (A × B) ,
therefore we can define the appointment of unit arrow O × O′ −→ A × B and appointment of arrows
composition (A × B) ⊗O×O′ (A × B) −→ A × B. The set product of categories provides product category
with product operations.

1.7 For the algebras X and Y with 1-degree operation it is easy to define the sum of such algebras X + Y .
We take the nonintersecting union of sets X + Y and define the sticked operation as the sum

f + f ′ : X + Y −→ X + Y .

For intersecting carriers X ∩ Y 6= Ø we must check the coincidence of sticked operations over the commune
part X ∩ Y .
For 0-degree operations we need another construction of new operation over the sum of carriers. The same
problem will also arise for bigger d ≥ 1 degree operations. For the sum of groups X + Y we must demand
that the units of both groups coincided and decide what would be the composition of members from different
sets. The unique solution is to use the product group compounded of pairs X × Y = {hx, yi : x ∈ X, y ∈ Y i}
with the inclusion of summands as partial groups X −→ X × Y , Y −→ X × Y taking pairs with unit of
another sumamnd hx, ei ∈ X × Y and he, xi ∈ Y .
The sum of directories poses no problem. The sum of two directories D1 : O1 −→ O1′ and D2′ : O2 −→ O2′
will have the sum of arrow sets D1 + D2 and will have sourse set the sum O1 + O2 and target set the
sum O2 + O2′ . Therefore we can define source appointment D1 + D2 −→ O1 + O2 and target appointment
D1 + D2 −→ O1′ + O2′ .
The sum of categories A + B will have the same arrows as the sum of directories. The unit arrow can be
taken in each subcategory separately. The composition is defined only for arrows from one subcategory in
order to have common middle vertex. We shall say that taken categories are nonintermittent subcategories
A ⊂ A + B and B ⊂ A + B.
1.8 The semigroup is an example of algebra with one law of multiplication

(◦) : X × X −→ X .

Such algebra is called monoid without unit.

We ask that this multiplication would be associative, i.e. for all points we have identities

(a ◦ b) ◦ c = a ◦ (b ◦ c) .

6
 This algebra is not free, we shall say that this algebra satisfies relations.
The monoid additionally demands to take the neutral element e ∈ X with identities

x◦e = x , e◦x=x↑x∈X .

Such element is understood as operation of 0-degree

e : X 0 = I −→ X .

The group demands to have the inverse element x− ∈ X for every point x ∈ X with two identities

x ◦ x− = e = x− ◦ x ↑ x ∈ X .

The inverse element can be understood as operation of 1-degree

X −→ X , x− ↑ x ∈ X .

In the case of existence, such element is unique in associative monoid. For two points y and z wich are
inverse elements of x we get equality

y = y ◦ e = y ◦ (x ◦ z) = (y ◦ x) ◦ z = e ◦ z = z .

Therefore the inverse element x− can be defined as solution of equation in associative monoid

−◦x=e , x◦−=e .

1.9 The category C will be a directory with the same source and target vertexes set O and being associative
monoid for the directory product. The multiplication of arrows is called composition

(◦) : C ⊗O C −→ C ,

and neutral element will be appointment of unit arrows

1 : O −→ C .

Group for such directory product has its own name grupoid. The semigroup for the directory product is
not so usual.
Example. The class of all directories provides an example of category with taken composition and unit
arrows. If we take only directories with finite number of arrows, we shall not have the units in such category.
Therefore it can be called the category without units. This provides a serious example of semigroup for the
directory product.

We can define the category as an algebra with operations being partial mappings over the set product.
Such definition is given in Freyd and Scedrov [5] p. 3. Such algebras with partial operations they have called
essentially algebraic, in the sense that we can construct special products and define functional operations
over these product to get the same (isomorphic) algebra. In general case the partial operations must be deal
as arbitrary convergence.
1.10 The morphism between two algebras with the same signature for set product is defined as an application
u : A −→ B which maintains the homologic operations, i.e. for each operation name we have the commuting
diagram:
un / n
An B
f f′
 
A u
/B

Others abstract products are defined in its own categories, and application is changed by arrows between
carrying spaces.

7
Example. The functor will be morphism of categories F : A =⇒ B. At first it must be morphism of
underlying directory, i.e. we must have the application of arrows F : A −→ B and application of vertexes
F0 : A0 −→ B0 which must commute with the source and target appointments. Secondly we must demand
the maintenance of arrow operations over the directory products. It must maintain the units and the
composition of arrows
F0 F ⊗O F
O / O′ A ⊗O A / B ⊗O ′ B

1 1 (◦) (◦)
   
A F
/B A F
/B

All set mappings f : X −→ Y compounds the category of sets, we shall note it Set. Vertexes will be all
possible (small) sets. The unit arrow will be identity application IdX : X −→ X. The arrow composition is
taken the usual composition of functions f ◦ g.
The product category Set × Set has pairs of sets hX, Y i as vertexes and pairs of mappings hf : X −→
Z, g : Y −→ W i as arrows. The bifunctor will be a functor over the product category. The sets multiplication
is an example of bifunctor in the category of sets Set

(×) : Set × Set =⇒ Set .

1.11 More generally operation in the set category may be arbitrary multivalued function F : An →− A.
Such algebras are called multialgebras. When the operations are arbitrary partial functions, such algebras
are called partial algebras. If we take the operations over infinite products AN , such algebras will be better
interpreted as convergence space. It is nothing with any finite algebraic property, but traditionally it will be
called algebra. Such tradition is maintained also by categorical framework of algebras in Mac Lane’s book
[11] part VI. The united sight to an algebras calls all algebraic or nonalgebraic operations as convergences
and algebras identifies with topological spaces in generalized sense as stuctures defined by Bourbaki [3]
The category was introduced as “abstract nonsense” by S. Eilenberg and S. MacLane in [6]. Their theory
emphasizes the work with arrows. On the points of sets constructed “concrete” functions are changed by
arrows and many properties of arrows can be proved without investigation of concrete structure of these
functions. We can say that a category is one step of abstraction. We use various structure in the category’s
set of arrows, and the second step of abstraction would be changing this concrete structure in the set of arrows
by some other categorical properties. In such way we encounter the bicategories. We want emphasize that
categorical argument are not absolute fashion of modern mathematics. The “concrete” structures explored
by N. Bourbaki must also be admired by “working mathematicians”.
The algebras is an example of concrete structure. In contrast the abstract notion of Cartesian product
will be an example of abstract notion. This contrast is very clear in computer programming. Algebras will
be identifyed with working programs in concrete computer, and categorical abstract notions will be only
specification of working programs. Nevertheless the categorical approach made essential progress in todays
programming business.
The monoidal category is defined as category with some bifunctor

: C × C −→ C .

It was first explicitly mentioned in 1963 by Bénabou [1]. The name ”monoidal” is due to Eilenberg.
This bifunctor is called a multiplication and will generalize the set multiplication of the set category. One
demands that this multiplication would be associative after identification with a natural transformation of
functors
α : (− −) − −→ − (− −) .
For the coherence one needs that the pentagon diagram with wedges get from natural transformation α
would be commuting.
At this time in physics is introduced the premonoidal categories, see W. Joyce [10]. It is interesting that
in his work noncommuting pentagon diagram are completed to commuting one with a deformation arrow.
The calculating of this arrows provides nonassociative statistics in quantum physics.

8
1.12 A multivalued mapping between points of two sets is identified with partial set of set multiplication
R ⊂ X × Y . Again the first space is called a source space and the second is called a target space. We shall
also say that we have a reform from the source space to the target space.
The pair of points hx, yi ∈ R is called related. More exactly we shall say that the first point x ∈ X
reforms himself to the second point y ∈ Y and the second point is reformed from the first one. We can also
say that second point is related to the first one, and the first point is corelated to the second one. It is more
usual to say that the second point y ∈ Y is related to the first point x ∈ X by the relation R, or the first
point x ∈ X has related point in the target space. But relation has no source or target space, therefore such
language can’t be correct for educated mathematician.
The reform A ⊂ O × O′ will be an example of directory which has no more than one arrow for every pair
of vertexes.
The reform is uniquely defined by its direct appointment

x 7−→ x ◦ R , X −→ 2Y ,

or by its opposite appointment

y 7−→ R ◦ y , Y −→ 2X .
The first appointment to a point x ∈ X appoints all related points in the target space x ◦ R. The second
appointment to a point y ∈ Y appoints all the points the source space x ∈ X to which the taken point is
related.
Two reforms F : X −→ Y and G : Y −→ Z can be composed

F ◦ G := {hx, zi ∈ X × Z : (∃y ∈ Y )(hx, yi ∈ F, hy, zi ∈ G)} .

The identity reform is defined by diagonal in set product

IdX = △ ⊂ X × X .

We can check the associativity for composition of reforms, and all reforms between arbitrary small sets
compound a new category which will be called the allegory of sets. It extends the category of sets compounded
by all mappings between arbitrary small sets.
The reform F ⊂ X × Y is called entire if every point in the source space has related point in target space.
The reform F ⊂ X × Y is called accurate if every point in the source space has no more than one related
point in the target space.
Also we rename these properties of opposite reform.
The reform F ⊂ X × Y is called covering if every point in the target space is related with some point in
the source space.
The reform F ⊂ X × Y is called exact if every point in the target space is related with no more than one
point in the source space.
The entire accurate reform is a function, the covering function is surjection, the exact function is injection.
For a reform F : X −→ Y between two spaces we construct a function of direct image between potential
spaces F P : 2X −→ 2Y defining image of arbitrary set in a source space A ⊂ X

F P (A) := {y ∈ Y : (∃x ∈ A)(hx, yi ∈ F )} ,

and a function of opposite image F ∗P : 2Y −→ 2X defining opposite image of arbitrary set in a target space
B⊂Y
F ∗P (B) := {x ∈ X : (∃y ∈ B)(hx, yi ∈ F )} .
For the mappings f : X −→ Y the opposite image is called an inverse image and is denoted more simply

f − (B) = f ∗P (B) .

1.13 The reforms can be abstractly defined in arbitrary regular category, cl. Freyd and Scedrov [5]. Cartesian
category has the beginning of every finite diagrams. It is enough to have Cartesian product, terminal object
and equalizer of every morphism pair.

9
 All these constructions are abstract, every object is identified only with uniquely existent isomorphism.
Regular category additionally demands the existence of image for every arrow and the property of product
bundle to maintain covers, i.e. for covering base mapping we get covering induced mapping of bunde. The
reform R : X →− Y in Cartesian category is defined as monopair of tabulating mappings f : Q −→ X and
g : Q −→ Y . The relation is understood as part of Cartesian product defined by this monopair. It can be
named as graphic of reform
G(R) >−→ X ∧ Y .
The reform R can be restore from it’s graphic G(R) >−→ X ∧ Y by choosing what the space is a source and
what the space is a target, i.e. the reform can be denoted by the triple

R = hX, G(R), Y i .

In regular category we have possibility to define the composition of two reforms with an intermediate space.
The mapping f : X −→ Y are identifyed with tabulation 1 : X −→ X, f : X −→ Y and arbitrary reform
is equal to fraction of its tabulation R = f − ◦ g. All reforms of regular category C ⊂ R(C) compound an
allegory. We shall call it a fractional extention. Functor of Cartesian category will be called representation
of Cartesian category if it maintains the beginnings of finite diagrams. Functor of regular category will
be called representation of regular category if it additionally maintains the covers. Every representation of
regular category F : C =⇒ C ′ has fractional extension which becomes a representation of unitary allegories
R(F ) : R(C) =⇒ R(C ′ ).
In Cartesian category we can freely define the composition of function with arbitrary relation f ◦ R. A
topos will be Cartesian category C having potential objects. For any object C we define potential [C] with
distinguished reform ∋: [C] →− C wich must be initial among reforms to this object by transformations
defined with arrows of category, i.e. for every reform R : X →− C we can find a unique representing arrow
f : X −→ C which transforms the distinguished reform to the taken one

f ◦ ∋= R .

The set category is an example of topos. The potential of the set C will be the set of all subsets [C] := 2C
and distinguished reform every partial set A ⊂ C relates with its points x ∈ C, we shall write A ∋ x.
For the reform R : X →− C the unique representing mapping f : X −→ [C] is a direct appointment of
this relation. For any reform R : X →− Y in topos we can define the unique mapping RP : [X] −→ [Y ].
In the Set category this mapping coincides with direct image appointment. The direct image appointment
defines endofunctor of set category F : Set =⇒ Set. Its fractional extention is representation of unitary
allegories R(F ) : R(Set) =⇒ R(Set). We remark that it differs from earlier direct image appointment. For
the relation R : X →− Y tabulated with the pair of mappings hf, gi : Q −→ hX, Y i the set A ⊂ X has
related set B ⊂ Y exactly when there is a set C ⊂ Q with needed direct images for tabulating mappings

f P (C) = A , g P (C) = B .

We get maintenance for composition F ◦ G and inverce F ∗ of relations

R(F ◦ G) = R(F ) ◦ R(G) , R(F ∗ ) = R(F )∗ .

The direct image f P (A) ⊂ Y are related to taken set A ⊂ X by the trace of set product (A × f P (A)) ∩ R ⊂
X × Y . It is the biggest of related sets to taken one.

2. Reform’s continuity in convergence spaces

2.1 The operations in multialgebras better to interpret as convergences. Then morphisms of multialgebras
will be defined as mappings with various continuity properties.
The convergence in the world space X is understood as a reform ρ ⊂ X Φ × X from sequences F ∈ X Φ to
the points x ∈ X of taken set X, see Gähler [8]. The direct appointment transforms the sequence F ∈ X Φ
to a partial set F ◦ ρ ⊂ X which is called a limit of the taken sequence. We shall use the notation

limX F := F ◦ ρ ⊂ X .

10
 The points of the limit will be called limit points. The opposite appointment transforms the point x ∈ X
to a set of sequences ρ ◦ x ⊂ X Φ . The sequences from this set will be called converging to the taken point.
A countable sequence in the space X will be a point from the countable set product of the world space
F ∈ X N , i. e. it is a function over the natural numbers to the world space

Fn ∈ X ↑ n ∈ N .

A natural number n ∈ N will be called an index and a value Fn ∈ X will be called a n-member of the
sequence.
The algebraic operations traditionally deals with the convergence of finite sequences taken from the finite
set product F ∈ X n , i.e. a sequence in this case coincides with a n-tuple of space points xi ∈ X.
The filter is a collection of partial sets F ⊂ X P . The space of all filters will be the second potential of
the space X P P . Filters commonly are considered as generalized sequences.
For us any set may also be considered as a sequence. The space of such sequences coincides with potential
of the space X P . The limit point of such convergence may be interpreted as integral of the taken set, for
example a geometric center point could be set’s limit point
Z
x= A.

For any reform to another world space R : X →− Y we get the reform between the spaces of sequences
RN : X N →− Y N using fractional extension of mappings. Every mapping f : X −→ Y provides the
mapping for the sequences f N : X N −→ Y N coinciding with set product. For reform we take the fractional
extension of set product. It can be expressed in such way: For every sequence with members xn ∈ X a
sequence with members yn ∈ Y will be related exactly then, when these members are related

hxn , yn i ∈ R ↑ n ∈ N .

The direct appointment appoints to the sequence x· ∈ X N the set coinciding with the product of sets
apointed to the members of the sequence
Y
hxn : n ∈ N i ◦ RN = {xn ◦ R : n ∈ N } .

We describe needed transformations also for another generalized sequences. For the partial sets A ∈ X P
we take fractional extension of mapping’s direct image appointment. Therefore for a partial set A ⊂ X
we get a related partial set B ⊂ Y in another world space exactly then, when we can find the partial set
C ⊂ Q ⊂ X × Y in the graphic set of the reform R which for projections has direct images coinciding with
the taken partial sets
pP
1 (C) = A , pP
2 (B) = B .

For the filters F ⊂ X P P we take fractional extension of mapping’s secondary direct image appointment.
This secondary image appointment is used to define the image of the filter for a mapping of the world space.
For a filter F ∈ X P P we have a related filter in another world space G ∈ Y P P exactly then, when we can
find a filter on the graphic of the reform H ∈ QP P which has the images coinciding with the taken filters

pP P
1 (H) = F , pP P
2 (H) = G .

The continuous function between two convergence spaces u : X −→ Y is any function which maintains
the limits
uP (limX F ) ⊂ limY uΦ (F ) .
For an algebraic operation, as for function over set product, the limit of (finite) sequence has exactly
one limit point. Therefore the continuity for such convergence coincides with equality of limit points. For
example, in additive monoid we demand the property of additivity

f (x + y) = f (x) + f (y) .

A map between algebras is called morphism of algebras if it is continuous for all algebra’s operations.

11
 For example a morphism between semigroups u : X −→ Y is defined demanding the equality of unique
limit points
u(F ◦ ρX ) = uΦ (F ) ◦ ρY .
This property is drawn as commuting square

uΦ
XΦ / YΦ
ρX ρY
 
X u
/Y

We shall say that commuting diagram defines predicate. In this case a predicate will be equality between
two terms get with composition of arrows

ρX ◦ u = u Φ ◦ ρY .

For more general convergence the continuous mappings are defined by by inclusion of limit sets

uP (F ◦ ρX ) ⊂ uΦ (F ) ◦ ρY .

Such inclusion can be drawn as directed square. It can be named as allegory’s commuting square

⊙ Φ uΦ / YΦ
X
ρX ρY
 
2X / 2Y
ր
uP

We can see that this inclusion is equivalent with another inclusion for opposite reforms

u− (y) ◦ ρX ⊂ (uΦ )− (y ◦ ρY ) .

It can be drawn by another directed square

ր u∗Φ
XO Φ o Y OΦ
ρ∗
X ρ∗
Y

2X o 2Y ⊙
u∗P

We shall say that these two directed squares define two different reversed predicates.
2.2 For diagrams of reforms the predicate is defined by choosing the birth vertex, the death vertex, the two
paths from the birth vertex to the death vertex and the arrow of inclusion between such paths. The path is
cyclic if it has some repeated vertex. The path is simple if it hasn’t cyclic parts. It is a unique possibility
for simple path to go from the birth vertex to the death vertex. Therefore the predicates with simple paths
are determined by choosing the birth and the death vertexes and the direction of inclusion.
If we denote the vertexes of commuting square

1 /2

 
4 /3

and choose clockwise the positive direction of inclusion, then the former first directed square will be noted
(−13) and the former second directed square will be noted (+31). There are 24 predicates with simple paths
get from such square diagram. Every of such predicates can be viewed as a definition of some continuity

12
property. Therefore clever mathematician would wish to choose the 24 different names for every of such
property. The linguistic problems usually are not interesting for mathematicians, but there is no other way
to understand this new field of multifunctional algebra. For new theories we need a new efforts of richer
language. My attempt to propose a vocabulary for englishmen can be corrected by others more prudent
wizards of English language.
2.3 At first we shall give only four names initiated by two different definitions of continuous multivalued
mappings. Our language will use the notion of sequences and their limit points.
2.3.1 A reform F : X →− Y will be called respecting if for every sequence in the source space F ∈ X Φ we
have in the target space the inclusion of points related to the limit points in the source space

F ◦ ρX ◦ F ⊂ F ◦ F Φ ◦ ρY ,

we shall write
Φ
F P (limX F ) ⊂ limP
Y (F (F )) ,

i.e. every point y ∈ Y , related to the limit point x ∈ limX F , must be a limit point for some related
sequence F ′ . The limP Y for some set of filters notes the direct image appointment. Such continuity property
is noted by predicate (−13). This property expresses the regularity of limit points in the source space. The
respective reform can be also understood as distributive operation. In earlier example the additive morphism
is distributive for a law of addition.
Such property is useful when we want to construct the algebras having two operations. For example
multiplication can be distributive for the operation of sum. Taking all possible products of generaters, we
get the set of members which sums maintains the products. Every product of sums will be expressed as a
sum of products
(x + y) · (z + w) = x · z + y · z + x · w + y · w .

2.3.2 A reform F : X →− Y will be called creating if for every point in the source space x ∈ X we have in
the target space the inclusion of sequences F ′ related to convergent sequences F in the source space

x ◦ ρ∗X F Φ ⊂ x ◦ F ◦ ρ∗Y ,

we shall write
F ΦP (lim∗X (x)) ⊂ lim∗P
Y (F (x)) ,

i.e. every sequence related to a x-convergent sequence must be y-convergent for some point y ∈ Y related
to taken point x. Such property is noted by predicate (+42). This property expresses the existence of limit
points in the target space.
2.3.3 These two properties for mappings coincide with usual continuity property: Every limit point in the
source space x ∈ limX F must have image f (x) ∈ Y which is a limit point of the image sequence f (F ).
Reverse predicate will say that the x-convergent sequence F has the image f (F ) convergent to the image
f (x) ∈ Y of the taken point.
2.4 The opposite reform F ∗ : Y →− X provides additional two predicates. We shall give them new names.
2.4.1 A reform F : X →− Y will be called reversely cautious if for every sequence in the target space F ′ ∈ Y Φ
we have the inclusion of points x ∈ X having related limits in the target space

F ′ ◦ ρY ◦ F ∗ ⊂ F ′ ◦ F Φ∗ ◦ ρX ,

we shal write
(limY F ′ ) ◦ F ∗ ⊂ limP ′
X (F ◦ F
Φ∗
),
i.e. every point in the source space x ∈ X, related to the limit point y ∈ limY F ′ of the taken sequence, must
be a limit point for some sequence F in the source space X for which the taken sequence is related. Such
property is noted by predicate (+24). Philologically the convergence in the source space must be cautious
as the convergence in the target space could be respecting.

13
2.4.2 A reform F : X →− Y will be called reversely wasting if for every point in the target space y ∈ Y we
have inclusion of sequences having related sequences with taken point in the target space

y ◦ ρ∗Y ◦ F Φ∗ ◦ ρ∗X ,

we shall write
(lim∗Y (y)) ◦ F Φ∗ ⊂ lim∗P ∗
X (y ◦ F ) ,

i.e. every sequence F in the source space having related sequence F ′ in the target space with taken limit
point y ∈ limY F ′ , must be convergent to some point x ∈ X for which taken point y ∈ Y is related. Such
property is noted by predicate (−31). Philologically the convergence in the source space must be wasting,
as the convergence in the target space could be creating.
2.5 The opposition maintains the inclusion of the graphics, therefore we can define additionally 4 predicates
with known properties. The predicate (+31) notes the reversely respecting reform F , the predicate (−24)
notes the reversely creating reform F , the predicate (−42) notes the cautious reform F , and the predicate
(+13) notes the wasting reform F . These predicates will define different properties for pairs of conjugate
functors. For reforms such functors are given by direct image appointment and opposite image appointment.
Therefore we wished to have different names for all predicates. At this moment I choose the predicate with
direct reform as principal.
2.6 We have defined the names for 8 predicates. We shall say that these predicates are from the first octet.
When the reform F :→− Y is a mapping, we can drive along this mapping the death or the birth vertexes
for the first two predicates (−13), (−24). We get the new predicates (−14) and (−23), which defines the
same continuity properties for mappings. These predicates begins the second octet.
2.6.1 A reform F : X →− Y will be called lavishing if for every sequence F in a source space X every limit
point has related point in the target space y ∈ Y which is a limit point for some related sequence F ′

F ◦ ρX ⊂ F ◦ F Φ ◦ ρY ◦ F ∗ ,

we can write
Φ ∗
limX F ⊂ (limP
Y (F ◦ F )) ◦ F .

It is more convenient to express such inclusion as large opposite image of convergence in the target space

ρX ⊂ (F Φ × F )∗P (ρY ) .

For the reforms there we apply the fractional extension of set product. Such property is noted by predicate
(−14).
2.6.2 A reform F : X →− Y will be called pressing if for every sequence F ′ in a target space Y the limit
point x ∈ X of sequences F in a source space, which has related taken sequence F ′ , has related some limit
point y ∈ limY F ′
F ′ ◦ F Φ∗ ◦ ρX ◦ F ⊂ F ′ ◦ ρY ,
we shall write
′
(limP
X (F ◦ F
Φ∗
)) ◦ F ⊂ limY F ′ .
It can also be expressed that a convergence in a source space has small direct image

(F Φ × F )P (ρX ) ⊂ ρY .

Such property is noted by predicate (−23).

2.7 The opposite reform F ∗ : Y →− X provides a new names.
2.7.1 A reform F : X →− Y will be called hiding if for every sequence F ′ in the target space Y every limit
point y ∈ Y is related to some point in the source space x ∈ X, which is a limit point for some sequence F
in a source space, which has the taken sequence F ′ as related

F ′ ◦ ρY ⊂ F ′ ◦ F Φ∗ ◦ ρX ◦ F ,

14
we shall write
limY F ′ ⊂ (limP ′
X (F ◦ F
Φ∗
)) ◦ F .
It can be expressed that the convergence in a source space has large direct image

ρY ⊂ (F Φ × F )P (ρX ) .

Such property is noted by predicate (+23). Philologically the convergence in a source space must be hiding,
as the convergence in a target space could be lavishing.
2.7.2 A reform F : X →− Y will be called reflecting if for every sequence F in the source space X every
point x ∈ X, which has related limit point of some related sequence F ′ in a target space, is a limit point of
taken sequence F
F ◦ F Φ ◦ ρY ◦ F ∗ ⊂ F ◦ ρX ,
we shall write
Φ ∗
(limP
Y (F ◦ F )) ◦ F ⊂ limX F .

It can be expressed that the convergence in a target space has a small opposite image

(F φ × F )∗P (ρY ) ⊂ ρX .

Tis property is noted by predicate (+14).

2.8 The reversing will give the same names for additional 4 predicates. At this moment the principal
predicate deals with the limit points of convergence, and reversed predicate deals with convergent sequences
of convergence. The predicate (+41) notes the reversely lavishing reform F , the predicate (+32) notes the
reversely pressing reform F , the predicate (−32) notes the reversely hiding reform F , the predicate (−41)
notes the opposite reflecting reform F .
2.9 We have just defined the 8 predicates from the second octet. The rest predicates compound the third
octet. For functional convergence we can get the earlier continuity properties. This helps us to choose the
preference order in description of last predicates.
2.9.1 A reform F : X →− Y will be called thin if for every sequence F in a source space X every sequence
converging to a point, related to some limit point x ∈ X of taken sequence F , is related with taken sequence
F
F ◦ ρX ◦ F ◦ ρ∗Y ⊂ F ◦ F Φ ,
we shall write
lim∗P P
Y (F (limX F )) ⊂ F
ΦP
(F ) .
It is better to say that the inverse image of the taken reform is smaller than the reform for the sequences

(ρX × ρY )∗P (F ) ⊂ F Φ .

This property is noted by predicate (−12).

2.9.2 A reform F : X →− Y will be called thick if for every point x ∈ X in a source space the limit points
in target space y ∈ limY F ′ , of some sequence related to some sequence F in source space converging to the
taken point x, is related to the taken point x

x ◦ ρ∗X ◦ F Φ ◦ ρY ⊂ x ◦ F ,

we shall write
limP
Y (F
ΦP
(lim∗X (x)) ⊂ F (x) .
It is better to say that the direct image of the reform for sequence is smaller than the taken reform itself

(ρX × ρY )P (F Φ ) ⊂ F .

15
This property is noted by predicate (+43). This property speaks about regularity of existing limit points.
Such reforms were often encountered in the theory of differential operators, they are called operator with
closed graphic.
2.10 The reversing provides additional two names. The predicate (+21) notes the reversely thin reform F ,
and the predicate (−34) notes the reversely thick reform F . These predicates deal with graphic of opposite
reform R∗ .
2.11 Rest 4 last predicates.
2.11.1 A reform F : X →− Y will be called binding if for every sequence F in the source space X every
related sequence F in the target space Y has limit point y ∈ Y related to some limit point x ∈ X of taken
sequence F
F ◦ F Φ ⊂ F ◦ ρX ◦ F ◦ ρ∗Y ,
we shall write
F Φ (F ) ⊂ lim∗P P
Y (F (limX F )) .

It is better to say that the inverse image of the taken reform is larger than the reform for sequences

F Φ ⊂ (ρX × ρY )∗P (F ) .

This property is noted by the predicate (+12).

2.11.2 A reform F : X →− Y will be called parting if for every point x ∈ X in a source space X the related
point y ∈ Y in the target space Y must be a limit point of some sequence F which is related to some sequence
F in a source space X converging to taken point x

x ◦ F ⊂ x ◦ ρ∗X ◦ F Φ ◦ ρY ,

we shall write
Φ ∗
F (x) ⊂ limP
Y F (limX (x)) .

It is better to say that direct image of the sequence reform is larger than the taken reform itself

F ⊂ (ρX × ρY )P (F Φ ) .

This property is noted by the predicate (−43).

2.12 The reversing provides additional two names. The predicate (−21) notes the reversely binding reform
F , and the predicate (+34) notes the reversely parting reform F . These predicates deals with graphic of
opposite reform R∗ .
2.13 The reform composition maintains the inclusion

F ⊂ G =⇒ R ◦ F ⊂ R ◦ G, F ◦ H ⊂ G ◦ H .

This property gives a possibility to apply the composition of directed squarees. This is straightforward
applied for the predicates of first octet.
2.13.1
Proposition. The composition of reform maintains the property of respecting reforms (−13), creating
reforms (+42), cautious reforms (+24), reversely wasting reforms (+31) and their reversed counterparts.
Proof: It is enough to apply composition of neighbouring squares.

2.13.2 The predicates of second octet needs the composition of embracing squares.
Proposition. The composition of reform maintains the property of lavishing reforms (−14), pressing reforms
(−23), hiding reforms (+23), reflecting reforms (+14) and their reversed counterparts.
Proof: We construct the embracing squarees.

16
2.13.3 For the predicates of third octet we must pose restrictions for the convergence in the mediating space.
At first we explain the terminology.
For every reform R : X →− Y we have defined two appointments: the direct image RP : X P −→ Y P
and inverse image R∗P : Y P −→ X P . They are monotonic mappings of Boole algebras, therefore we have
instances of functors.
We shall call the pair of opposite functors F and G between two categories an adjunction with the same
notation as for reform R : A →− B. Therefore for this adjunction we have the source category A and the
target category B. The direct functor F : A =⇒ B is called adjoint functor, and opposite functor G : B =⇒ A
is called coadjoint functor. The unit of adjunction is a transformation of Identity functor i : Id −→ F G
and counit of adjunction is a transformation of functor composition e : GF −→ Id. The naturality and
triangular identities is needed for the usual true adjunction, but we can deal with “generalized” adjunctions
without such properties.
The entire reform R provides a unit of its adjunction

A ⊂ R∗P (RP (A)) ↑ A ⊂ X ,

the accurate reform R provides a counit of its adjunction

RP (R∗P (A′ )) ⊂ A′ ↑ A′ ⊂ Y ,

the covering reform R provides a unit of opposite adjunction

A′ ⊂ RP (R∗P (A′ )) ↑ A′ ⊂ Y ,

and the exact reform R provides counit of opposite adjunction

R∗P (RP (A)) ⊂ A ↑ A ⊂ X .

The mapping R := f provides a unit and counit together, therefore such adjunction for Boole algebras is
true. The inverse of mapping R := f ∗ provides a unit and counit for opposite adjunction, therefore in this
case the opposite adjunction is true. For bijective map R both adjunctions are true.
Proposition. The composition maintains the property of thin reforms (−12) if the convergence in the middle
space is simple. The composition maintains the property of thick reforms (+43) if the convergence in the
middle space is entire. The composition maintains the property of binding reforms (+12) if the convergence
in the middle space is covering. The composition maintains the property of parting reforms (−43) if the
convergence in the middle space is exact. The reversed counterparts needs the same conditions for the
convergence in the middle space.
Proof: These conditions provide needed unit and co-unit arrows of direct and opposite adjunctions.

For the countable sequence xn ∈ X ↑ n ∈ X an accurate convergence is known as Hausdorff. Usual

sequence covergence is covering, as it appoints the point x ∈ X for constant sequence xn = x. The usual
convergence is not exact, because different sequences can converge to the same limit point. Entire convergence
may be only for some algebraic operation over the finite part of sequences.
2.14 The identity mapping has continuity property of every kind, therefore the respective reforms compound
the suballegory. Sometimes it is beneficial to use the decomposition with tabulating mappings R = f − ◦ g
in this suballegory, because the continuity properties for mappings is much simple.
The death vertex can be pushed forward along adjoint functor with counit, and backword along adjoint
functor with unit. The direction is defined with the arrow of predicate. In the case of set reforms we get
that death vertex can be pushed forward along the accurate reform, and can be pushed backward along the
entire reform.
The birth vertex can be pushed forward along coadjoint functor with unit, and can be pushed backward
along coadjoint functor with counit. In the case of set reforms we get that the birth vertex can be pushed
forward along covering reform, and can be pushed backward along exact reform.
After pushing we get the continuity for the new predicate. In the case of mappings we get the equivalent
continuity properties. The death vertex can be pushed forward and backward along mappings, and birth
vertex can be pushed backward and forward along inverse mappings.

17
 Therefore we can formulate the proposition about continuity properties of mappings.
Proposition. For mappings 4 continuity properties are equivalent: respecting (−13), pressing (−23), lav-
ishing (−14) and reversely creating (−24).
Proof: Such predicates allow to push the death and birth vertexes along the mappings or inverse mappings.

This proposition also is valid for 4 rversed predicates.

3. Reform’s continuity in adherence spaces

3.1 The adherence in the space X is defined as generalized convergence Adh : 2X →− X. A sequence now
will be an arbitrary partial set A ⊂ X. The direct appointment appoints the set of adherent points

[A] := A ◦ Adh

and will be called an adherence operator. The opposite appointment appoints the filter of sets having taken
adherent point
Fx := Adh ◦ x .
We shall call such filter an adherence filter over the point x ∈ X.
We remember that any of these appointments restore the adherence. We shall say the point x ∈ X is
adherent to the set A ⊂ X and the set adherent point. All adherent points for the set A ⊂ X compound the
adherence set [A] ⊂ A.
The conjugate operator is called inside operator, it is defined with the Boole completement of the sets

hAi := A ◦ Ins = (Ac ◦ Adh)c .

The sets with the adherent point now will be changed by the neighbourhoods of taken point. The set V ⊂ X
is a neighbourhood of the point x ∈ X and this point x ∈ X is inside point of this set exactly then, when
the completment of this set Ac = X \ A hasn’t taken point as adherent point. We shall write

Vx := Ins ◦ x .

We shall say that the inside points x ∈ A compound the inside set hAi ⊂ X. The inside operator or
neighbourhood filters define inside as adherence conjugate to taken one.
It is obvious that secondly conjugate operator coincides with the taken one. The names adherence or
inside are only relative consideration. At this moment every adherence can be inside of the dual adherence.
Next we can ask some asymmetric properties for the adherence.
3.1.1 An increasing adherence appoints the bigger adherence set A ⊂ [A]. An isotonic adherence for the
bigger set apoints a bigger set
A ⊂ B =⇒ [A] ⊂ [B] .
The isotonic increasing adherence is called a closure. The set is called closed if it coinsides with it’s adherence
set
A = [A] .
For such adherence the inside will be diminishing A ⊃ hAi and isotonic

A ⊂ B =⇒ hAi ⊂ hBi .

The isotonic diminishing inside will be called interior. The set is open if it coincides with it’s interior

A = hAi .

3.2 We shall look the continuity properties for the reforms of adherence or inside spaces. We shall content
himself with the case of mappings f : X −→ Y . Then the mapping of partial sets is provided by direct
image f P : X P −→ Y P . We shall use the predicates from the first and second octets only. Such predicates

18
are good for the sequential adherences, but we haven’t opportunity to speak there about this subject now.
The adherence operator and the adherence filters helps us to deal continuity properties more easy.
At first we shall check the equivalent predicates which coincide with property of traditional continuous
mapping.
3.2.1 The respecting (−13) mapping of adherence spaces is defined by directed squaree:
P
f
⊙ /·
AdhX  (−13)  AdhY

· /·
ր
f

The corresponding inclusion of reforms

A ◦ AdhX ◦ f ⊂ A ◦ f P ◦ AdhY

is easily defined with adherence operators i.e. we get usual condition for continuous mappings in topological
spaces
f P ([A]) ⊂ [f P (A)] ↑ A ⊂ X .
The reversely respecting (+31) mapping is easily defined with adherence filters:
P∗
ր f
· o ·
∗ O
AdhX (+31) O Adh∗Y
· o
∗
⊙
f

The inclusion of reforms

y ◦ f ∗ ◦ Adh∗X ⊂ y ◦ Adh∗Y ◦ f P ∗
can be written as properties of adherence filters

Ff −1 (y) ≤ (f P )− (Fy ) .

For every point in a target space y ∈ Y every point of inverse image x ∈ f − (y) is adherent only for such
set A ⊂ X which has direct image f P (A) ⊂ Y for which taken point is adherent y ∈ [f P (A)].
3.2.2 The reversely creating (−24) mapping of adherence spaces
P∗
f
· o ⊙
AdhX  (−24)  AdhY
ց
· o
∗
·
f

is easily defined with adherence operators. The inclusion of reforms

B ◦ f P ∗ ◦ AdhX ⊂ B ◦ AdhY ◦ f ∗

we want to change with inclusion of adherence sets. For every partial set in the source space A ⊂ X having
the direct image coinciding with taken partial set in the target space B = f P (A), the adherence set is smaller
than inverse image of taken partial set
A ⊂ f − (B) .
For isotonic adherences this property can be get more easily.
Proposition. For reversely creating mapping f : X −→ Y every covered partial set B ⊂ f P (X) has inclusion
for adherence set of inverse image
[f − (B) ⊂ f − ([B]) .

19
If the adherence in the source space is isotone, then this property is also sufficient for mapping to be reversely
creating.
If the adherence in the target space is isotone, then this property is equivalent to the same requirement
for every partial set in the target space B ⊂ Y .
Proof: For covered set B ⊂ f P (X) the inverse image A = f − (B) will have the image coinciding with the
taken set f P (A) = B, therefore its adherence set is included in the inverse image of adherence set for the
taken set
[f − (B)] ⊂ f − ([B]) .
For every partial set in the target space B ⊂ Y we can take smaller set B ′ = f P (f − (B)), and for isotone
adherence in the target space

[f − (B)] = [f − (B ′ )] ⊂ f − ([B ′ ]) ⊂ f − ([B]) .

For every partial set A ⊂ X with image f P (A) = B we have A ⊂ f − (A), therefore for isotone adherence
in the source space we shall have
[A] ⊂ [f − (B)] ⊂ f − ([B]) .

For isotonic adherences this property is equivalent with the inclusion of adherence for the inverse image

[f − (B)] ⊂ f − ([B]) .

The creating (+42) mapping

P
f ց
·o ·
Adh∗
X
O (+42) O Adh∗Y
⊙o ·
f

is defined by inclusion of reforms

x ◦ Adh∗X ◦ f P ⊂ x ◦ f ◦ Adh∗Y .
This can be written with inclusion of image for adherence filter

f P P (Fx ) ≤ Ff (x) .

Every partial set A ⊂ X adherent by the taken point x ∈ X has image f P (A) ⊂ Y adherent by the image
of the taken point y = f (x).
3.2.3 The pressing (−23) mapping
P∗
f
· o ⊙
AdhX  (−23)  AdhY
· /·
ր
f

is defined by inclusion of reforms

B ◦ f P ∗ ◦ AdhX ◦ f ⊂ B ◦ AdhY .

It can be expressed as small image of adherence AdhX ⊂ X P × X

(f P × f )P (AdhX ) ⊂ AdhY .

For the adherence operator we shall get small images

f P ([A]) ⊂ [B]

20
for the adherent set of sets A ⊂ X which has direct image coinciding with taken set f P (A) = B. This can
be written as inclusion of the filters

f P P ([(f P )− (B)]) ≤ {[B]} .

For isotone adherences this property is expressed more easily.

Proposition. For the isotone adherence in the source space the mapping is pressing if every covered partial
set B ⊂ Y has small image of adherence set for the inverse image of the taken set

f P ([f − (B)]) ⊂ [B] .

For the isotone adherence in the target space this requirement is equivalent to such inclusion for every partial
set B ⊂ Y .
Proof: For isotone adherence in the source space a smaller partial set will have a smaller adherence set

A ⊂ f − (B) =⇒ [A] ⊂ [f − (B)] ,

therefore we shall have the inclusion for the image of adherence set

f P ([A]) ⊂ f P ([f − (B)]) .

We have shown that for such adherence every partial set A ⊂ X will have a small image of adherence set

f P ([A]) ⊂ [B] .

If the adherence in the target space is isotone, then for every partial set B ⊂ Y we have a smaller covered
set f P (f − (B)) ⊂ B, and we can check the needed inclusion

f P ([f − (B)]) ⊂ [f P (f − (B))] ⊂ [B] .

We have shown that for isotone adherences the property of pressing mapping is expressed with inverse
image of taken set in the target space

f P ([f − (B)]) ⊂ [B] ↑ B ⊂ Y .

The reversely pressing (+32) mapping

fP ց
· /·
Adh∗
X
O (+32) O Adh∗Y
· o
∗
⊙
f

is defined by inclusion of opposite reforms

y ◦ f ∗ ◦ Adh∗X ◦ f P ⊂ y ◦ Adh∗Y ,

which also express small image of adherence

(f × f P )P (Adh∗X ) ≤ Adh∗Y .

For adherence filters we get that for any point in a target space y ∈ Y the points of inverse image x ∈ f − (y)
have adherence filter Fx with small image

f P P (Ff − (y) ) ≤ Fy .

21
3.2.4 The lavishing (−14) mapping
P
f
⊙o ·
AdhX  (−14)  AdhY
ց
· o
∗
·
f

demands an inclusion of reforms

A ◦ AdhX ⊂ A ◦ f P ◦ AdhY ◦ f ∗ ,

i.e. we ask a large inverce image of the adherence

AdhX ⊂ (f P × f )− (AdhY ) .

For adherence operator it will be an inclusion of adherence set

[A] ⊂ f − ([f P (A)]) ↑ A ⊂ X .

The reversely lavishing (+41) mapping

P∗
ր f
· o ·
∗ O
AdhX (+41) O Adh∗Y
⊙ /·
f

asks another inclusion

Adh∗X ⊂ f ◦ Adh∗Y ◦ f P ∗ ,
which again is equivalent of asking the large inverse image of adherence

Adh∗X ⊂ (f × f P )− (Adh∗Y ) .

It can be expressed with large inverse image of adherence filter

Fx ≤ (f P )− (Ff (x) ) .

3.3 Remain the predicates without equivalent counterparts.

3.3.1 Wasting (+13) mapping
P
f
⊙ /·
AdhX  (+13)  AdhY

· /·
ւ
f

asks an inclusion of reforms

A ◦ f P ◦ AdhY ⊂ A ◦ AdhX ◦ f .
It can be expressed with adherence operators

[f P (A)] ⊂ f P ([A]) ↑ A ⊂ X .

The reversely wasting (−31) mapping

P∗
ւ f
· o ·
∗ O
AdhX (−31) O Adh∗Y
· o
∗
⊙
f

22
asks an inclusion of reforms
y ◦ Adh∗Y ◦ f P ∗ ⊂ y ◦ f ∗ ◦ Adh∗X .
This can be expressed with small inverse image of adherence filter

(f P )− (Fy ) ⊂ Ff − (y) .

3.3.2 The reversely cautious (+24) mapping

P∗
f
· o ⊙
AdhX  (+24)  AdhY

ց
· o
∗
·
f

is asking the inclusion of reforms

B ◦ AdhY ◦ f ∗ ⊂ B ◦ f P ∗ ◦ AdhX .

This property can be expressed with adherence operator: Every point x ∈ f − ([B]) in the inverse image of
the adherence set for taken set in the target space B ⊂ Y is adherent x ∈ [A] for some partial set in the
source space A ⊂ X with image coinsiding with taken set f P (A) = B.
Proposition. For the isotone adherence in a source space the mapping f : X −→ Y will be reversely cautious
exactly then, when for every covered partial set B ⊂ f P (X) the inverse image of the adherence set would
contained in the adherence set of the inverse image of taken set

f − ([B]) ⊂ [f − (B)] ,

and for other partial sets in a target space B 6⊂ f P (X) the adherence set would contained in completment
of mapping’s whole image
[B] ⊂ Y \ f P (X) .
Proof: This condition is obviously sufficient.
And it is necessary: If a point x ∈ X is in inverse image of the adherence set x ∈ f − ([B]) for a covered
set B ⊂ f P (X), then any set A ⊂ X with the image f P (A) = B and adherent by the taken point x ∈ [A]
will have A ⊂ f − (B), therefore for isotone adherence

x ∈ [A] ⊂ [f − (B)] .

For other partial sets B 6⊂ f P (X) the inverse image of the adherence set will be empty f − ([B]) = Ø, as
we can’t find any partial set A ⊂ X having image equal to the taken set, i.e.

f P (A) 6= B ↑ A ⊂ X .

The cautious (−42) mapping

P
f տ
·o ·
Adh∗
X
O (−42) O Adh∗Y
⊙ /·
f

asks an inclusion of reforms

x ◦ f ◦ Adh∗Y ⊂ x ◦ Adh∗X ◦ f P .
This is expressed with big image of the adherence filter

f P P (Fx ) ≥ Ff (x) .

23
It is very special property for the mapping: the point’s image f (x) ∈ Y can be adherent only such partial
sets B ⊂ Y which coincide with direct image of any partial set in the source space B = f P (A).
3.3.3 The hiding (+23) mapping
P∗
f
· o ⊙
AdhX  (+23)  AdhY

· /·
ւ
f

demands an inclusion of relations

B ◦ AdhY ⊂ B ◦ f P ∗ ◦ AdhX ◦ f ,

which equivalent to the large image of adherence

AdhY ⊂ (f P × f )P (AdhX ) .

This can be expressed with adherence operator. For the adherent point y ∈ [B] we can find partial set in
a source space A ⊂ X which image coincides with the taken set f P (A) = B and we have adherent point
x ∈ [A] covering the earlier point in the target space f (x) = y
[
[B] ⊂ {f P ([A]) : f P (A) = B} .

For the isotone adherence in the source space the hiding mapping can be defined by stronger requirement.
Proposition. If the whole target set has any adherent point [Y ] 6= Ø, then hiding mapping f : X −→ Y
will cover the whole target space.
For the isotone adherence in the source space AdhX the hiding mapping f : X −→ Y will provide every
covered partial set B ⊂ f P (X) with adherence set contained in the image of adherence set for inverse image
of taken set
[B] = Ø ↑ B 6⊂ f P (X) .
Proof: For not covered set B ⊂ Y we cant find any partial set in the source space A ⊂ X with coinciding
image f P (A) = B, therefore for hiding mapping such set cannot have any adherent point [B] = Ø.
For the covered set B ⊂ Y every adherent point y ∈ [B] must be covered by adherent point x ∈ [A] for
some partial set in the source space A ⊂ X with coinciding image f P (A) = B. If the adherence in the source
space is isotone, then
y ∈ f P ([A]) ⊂ f P ([f − (B)]) ↑ B ⊂ Y .

The reversely hiding (−32) mapping

fP տ
· /·
Adh∗
X
O (−32) O Adh∗Y
· o
∗
⊙
f

demands an inclusion of reforms

y ◦ Adh∗Y ⊂ y ◦ f ∗ ◦ Adh∗X ◦ f P ,
which is equivalent for large image of adherence’s opposite reform

Adh∗Y ≤ (f × f P )P (Adh∗X ) .

This will be expressed as large image of adherence filter appointment

[
Fy ≤ {f P P (Fx ) : x ∈ f − (x)} .

24
 We have seen that such mapping must be very special. For the target space with adherent point [Y ] 6= Ø
such mapping must be covering.
3.3.4 The reflecting (+14) mapping
P
f
⊙ /·
AdhX  (+14)  AdhY

տ
· o
∗
·
f

is defined by inclusion of reforms

A ◦ f P ◦ AdhY ◦ f ∗ ⊂ A ◦ AdhX ,

which is equivalent to the small inverse image of the adherence’s graphic

(f P × f )− (AdhY ) ⊂ AdhX .

This can be expressed with the adherence operator as large adherence set in the source space

f − ([f P (A)]) ⊂ [A] ↑ A ⊂ X .

The adherent point of some set x ∈ [A] is a property which is reflected by such mapping. If this property
is checked in the target space for image point and the set f (x) ∈ [f P (A)], then we can conclude that this
property is also content for initial point and partial set in the source space.
The reversely reflecting (−41) mapping
P∗
ւ f
· o ·
∗ O
AdhX (−41) O Adh∗Y
⊙ /·
f

is defined by inclusion of reforms

x ◦ f ◦ Adh∗Y ◦ f P + ⊂ x ◦ Adh∗X ,

which is equivalent to the small inverse image of adherence’s opposite reform

(f × f P )− (Adh∗Y ) ≤ Adh∗X .

This property can be expressed for the appointment of adherence filters

(f P )− (Ff (x) ) ≤ Fx .

3.4 Finely we formulate the predicates from the third octet. These properties is not useful for the traditional
topological adherences. But they may be interesting for some integral representation of set’s middle point.
The adherence point x ∈ [X] can be understood as middle point of this set. Such point usually exists and is
defined uniquely, but we can questionize such existence or may exist many of needed middle points for some
extraordinary sets.
3.4.1 The binding (+12) mapping
P
f
⊙ /·ց
AdhX  (+12) O Adh∗
Y

· / ·
f

is defined by inclusion of reforms

A ◦ f P ⊂ A ◦ AdhX ◦ f ◦ Adh∗Y ,

25
which is equivalent to the large inverse image of the taken mapping

f P ⊂ (AdhX × AdjY )− (f ) .

It may be defined with the union of adherence filters

f P (A) ∈ Ff P ([A]) ,

the image of the partial set A ⊂ X has the image f (y) ∈ Y of the middle point x ∈ [A] as a middle point.
For functional convergence in the source space this property would be the same as for the creating
mapping (+42), and for functional convergence in the target space this property would be the same as for
the wasting mapping (+13), therefore it is a same as usual continuity property for mappings defined with
reversely respecting (+31), reversely pressing (+32), reversely lavishing (+14) mappings.
The reversely binding (−21) mapping
P∗
ւ f
· o ⊙
Adh∗
X
O (−21)  AdhY
· o
∗
·
f

is defined by inclusion of reforms

B ◦ f P ∗ ⊂ B ◦ AdhY ◦ f ∗ ◦ AdhX ,

which is equivalent to the large inverse image of the inverse napping

f P ∗ ⊂ (AdhY × AdhX )− (f ∗ ) .

This property can also be expressed with union of adherence filters

(f P )− (B) ≤ Ff − ([B]) ,

every partial set A ⊂ X which has the image f P (A) = B coinciding with the taken partial set in the target
space B ⊂ Y , has some adherent point in the inverse image x ∈ f − (B) of taken set B ⊂ Y .
3.4.2 The thick (+43) mapping
P
f
· /·
Adh∗
X
O (+43)  AdhY
⊙ /·
f ւ

is defined by an inclusion of reforms

x ◦ Adh∗X ◦ f P ◦ AdhY ⊂ x ◦ f ,

which is equivalent for small image of mapping’s direct image appointment

(AdhX × AdhY )P (f P ) ⊂ f .

This can be expressed with adherence operator’s union over all sets from some filter
[
{[B] : B ∈ f P P (Fx )} ⊂ {f (x)} ,

if the taken point x ∈ X adheres the partial set A ⊂ X, then the image point f (x) ∈ Y is a unique adherent
point for the image set f P (A) ⊂ Y .
For the functional convergence AdhX in the source space we get the equivalent property of wasting
mapping (+13). For the functional convergence AdhY in the target space we get the equivalent property of

26
creating mapping (+42), therefore for mappings this is equivalent with reversely respecting (+31), reversely
pressing (+32), reversely lavishing (+41) mappings also.
3.5 Now rest more special properties, without any equivalent counterparts.
3.5.1 The thin (−12) mapping
P
f
⊙ /·տ
AdhX  (−12) O Adh∗
Y

· /·
f

is defined by an inclusion of reforms

A ◦ AdhX ◦ f ◦ Adh∗Y ⊂ A ◦ f P ,

which is eqiuvalent to the small inverse image of mapping

(AdhX × AdhY )∗P (f ) ⊂ f P .

This can be expressed with the union of adherence filters

Ff P ([A]) ⊂ {f P (A)} ,

the image f P (A) ⊂ Y of every partial set A ⊂ X in the source space is unique partial set B ⊂ Y in the
target space which has an adherent point y ∈ [B] related to some adherent point of the taken partial set
x ∈ [A], f (x) = y. This means that the related adherence points must also relate the partial sets.
The reversely thin (+21) mapping
P∗
ր f
· o ⊙
∗ O
AdhX (+21)  Adhy
· o
∗
·
f

is defined by inclusion of reforms

B ◦ AdhY ◦ f ∗ ◦ Adh∗X ⊂ B ◦ f P ∗ ,

which is equivalent to the small inverse image of inverse mapping’s graphic

(AdhY × AdhY )∗P (f ∗ ) ⊂ f P ∗ .

This can be expressed with the union of adherence filters

Ff − ([B]) ⊂ (f P )− (B) ,

the partial set A ⊂ X, with the adherent points x ∈ [A] having related points y ∈ Y , f (x) = y adherent to
the taken partial set y ∈ [B], must be related to the taken set in the target space f P (A) = B.
3.5.2 The parting (−43) mapping
P
f
· /·
∗ O
AdhX (−43)  AdhY
⊙ /·
f ր

is defined by inclusion of reforms

x ◦ f ⊂ x ◦ AdhX ◦ f P ◦ AdhY ,

27
which is equivalent to a large image of the mapping’s direct image appointment

f ⊂ (AdhX × AdhY )P (f P ) .

The adherence expands set image so much that it include the mapping itself f : X −→ Y . This can be
expressed with upper covering of adherence operator over the adherence filter
[
{f (x)} ⊂ {[B] : B ∈ f P P (Fx )} ,

i. e. the image f (x) ∈ Y of a taken point x ∈ X is adherent to the image f P (A) ⊂ Y of any partial set in
the source space A ⊂ X for which the taken point is adherent x ∈ [A].
If this adherent point is defined as set’s middle point
Z
x= A,

then the parting mapping asks that the image f (x) ∈ Y would be a middle point of an direct image
Z
f (x) = f P (A)

for some set A ⊂ X for which taken point is middle point.

This property is very week. For expanding adherences such property has every mapping f : X −→ Y .
We can take A = {x}, then x ∈ [A] and f (x) ∈ [A].
The reversely parting (+34) mapping
P∗
f
· o ·
AdhX  (+34) O Adh∗Y
տ
· o
∗
⊙
f

is defined by inclusion of reforms

y ◦ f ∗ ⊂ y ◦ Adh∗Y ◦ f P ∗ ◦ AdhX ,

which is equivalent to large image of inverse of mapping’s direct image appointment

f ∗ ⊂ (AdhY × AdhX )P (f P ∗ ) .

This can be expressed with upper covering of adherence operator over the filter
[
f − (y) ⊂ {[A] : f P (A) ∈ Fy } ,

i.e. for the point in the target space y ∈ Y every point of inverce image x ∈ f − (y) is adherent some partial
set A ⊂ X which direct image f P (A) ⊂ Y is adherent by the taken point y ∈ [f P (A)].
3.6 We can define weakly cautious mappings of adherence spaces, and to show that such property is equivalent
to respecting mappings of inside spaces. This provides the fact, that some continuity properties may be
expressed both in adherence and inside spaces. More details is presented in my recent monograph [13].

4. Mapping’s global continuity

4.1 The continuity properties in adherence spaces provide local continuity properties in generated topological
spaces. The same situation also arise for bornological or measurable spaces. Therefore it is usefull to look at
general continuity properties for topological spaces. At this moment we content himeself to deal only with
mappings. These continuity properties will be called global.

28
4.2 The part in the potential set T ⊂ X P will be called a tribe and it will define a tribe space X. For
any mapping f : X → Y with set’s direct image appointment f P : X P → Y P we can define 3 continuity
properties: a carrying mapping f P P (TX ) ≤ TY , a hiding mapping f P P (TX ) ≥ TY , and an image reflecting
mapping (f P )− (TY ) ≤ TX .
For topological spaces we can investigate the topology T and cotopology T c together, therefore we get
different continuity properties: For topology compounded by closed sets a closed mapping f P P (TX ) ≤ TY ,
a closed set hiding mapping f P P (TX ) ≥ TY , and an closed sets image reflecting mapping (f P )− (TY ) ≤ TX ,
and for cotopology compounded by open sets a open mapping f P P (TXc ) ≤ TYc , a open sets hiding mapping
f P P (TXc ) ≥ TYc , and an open set image reflecting mapping (f P )− (TYc ) ≤ TXc .
With set’s inverse image appointment f − : Y P → X P we define another 3 continuity properties: a
turning mapping (f − )P (TY ) ≤ TX , a exhausting mapping (f − )P (TY ) ≥ TX , and an inverse image reflecting
mapping (f − )− (TY ) ≤ TX .
For topological spaces only the turning mappings are concidered as continuous ones.
4.2.1 We defined the closure as isotonic and expanding adherence. The closed sets is defined as fixed points
for closure operator, i.e. the closed sets are defined by the property

A ⊂ [A] ⊂ A .

The tribe compounded by all closed sets is called a topology. It is characterized by the property of freely
intersection \
E∈T ↑E⊂T ,
every tribe, which maintains the arbitrary intersection of its members, can be get as topology of closed sets
for some closure. Such topologies are used in logic calculations [12]. Bourbaki for more traditional topologies
asks additional property that wide set is closed.
Otherwise the closure was called a multistep (germ. mehrstufig) topology by Gähler [7].
The respecting (−13) mapping asks inclusion for direct image

f P ([A]) ⊂ [f P (A)] ↑ A ⊂ X ,

and reversely creating mapping (−24) asks equivalent (for mappings of closure spaces) opposite inclusion for
inverse image
[f − (B)] ⊂ f − ([B]) ↑ B ⊂ Y .
This can be also expressed by equivalent (for mappings) properties with direct or inverse images of closure
reform. The pressing (−23) mapping asks an inclusion of direct image

(f P × f )P (AdhX ) ⊂ AdhY ,

and lavishing (−14) mapping asks the same opposite inclusion of inverse image

AdhX ⊂ (f P × f )− (AdhY ) .

These all equivalent continuity properties coincide with usual continuity of mappings between closure spaces.
Such mapping is turning for topological spaces, i.e. the inverse image of closed set remains closed

[B] ⊂ B =⇒ [f − (B)] ⊂ f − ([B]) ⊂ f − (B) .

The wasting (−31) mapping asks opposite inclusion for direct image

[f P (A)] ⊂ f P ([A]) ↑ A ⊂ X .

It defines the closed mapping between closure spaces: The direct image of closed set remains closed

[A] ⊂ A =⇒ [f P (A)] ⊂ f P ([A]) ⊂ f P (A) .

Such mappings are carrying for topological spaces.

29
 Bourbaki [4] concidered the property of closed mappings only together with continuity.
The cautious (−42) mappings asks big image of the adherence filter

f P P (Fx ) ≥ Ff (x) .

For covering mapping between closure spaces it asks an inclusion of inverse image

f − ([B]) ⊂ [f − (B)] ↑ B ⊂ Y .

The more useful property of weakly cautious mapping. It demands that image of adherence filter would
generate a big hereditary filter
[f P P (Fx )] ≥ Ff (x) .
Such mapping will be equivalent to respecting mapping of conjugate interior spaces. It will be open for
generated topology.
Let 0 ⊂ X is open set of closure in a source space. It is easy to check that the direct image f P (O) ⊂ Y
will be open set of closure in a target space. It is enough to show that any set A′ ⊂ Y \ f P (O) has no
adherent point from image of taken set y ∈ f P (O). Otherwise we should have a point x ∈ O with f (x) = y,
and for weakly cautious mapping this point will be an adhered by the set A ⊂ X with smaller direct image
f P (A) ⊂ A′ . Therefore A ⊂ X \ O, and such set can’t adhere the point from the taken open set x ∈ O. So
we get a contradiction, and there is no point from the taken open set adhered by the set A′ ⊂ Y \ f P (O).
The hiding (−32) mapping asks the opposite inclusion of direct image of closure reform

(f P × f )P (AdhX ) ⊃ AdhY ,

and reflecting (+14) mapping asks the inclusion of inverse image of closure reform

(f P × f )− (AdhY ) ⊂ AdhX .

Such local properties don’t coincide with corresponding global properties in topological spaces.
4.2.2 We define the interior as isotonic and decreasing adherence. The open sets is defined as fixed points
for interior operator, i.e. the open sets are defined by the property

A ⊃ hAi ⊃ A .

The tribe compounded by all open sets we shall call a cotopology. It is characterized by the property of
freely union [
E∈K↑E⊂K,
every tribe, which maintains the arbitrary union of its members, can be get as cotopology of open sets
for some interior. Cotopologies are dual for topologies, however the continuity properties of mappings are
different.
From continuous mappings of closure sets with inverse image condition

[f − (B)] ⊂ f − ([B]) ↑ B ⊂ Y ,

we get the dual inverse image condition for mappings of interior spaces

f − (hBi) ⊂ hf − (B)i ↑ B ⊂ Y .

Such mappings are turning between the cotopological spaces

B ⊂ hBi =⇒ f − (B) ⊂ f − (hBi) ⊂ hf − (B)i .

Nevertheless this condition cannot be get as earlier continuity property for adherence spaces. The cautious
(−42) mappings between interior spaces is defined by the same inclusion for covered subsets

f − (hBi) ⊂ hf − (B)i ↑ B ⊂ f P (X) ,

30
and in the target space we demand that

hBi ⊂ Y \ f P (X) ↑ B 6⊂ f P (X) .

Therefore in most cases the cautious mappings must be surjective.

The situation is rescued with the notion of weakly cautious mapping between interior spaces. This property
is equivalent to the property of respecting mapping for conjugate closure spaces.
The respecting (−13) mapping between interior spaces is defined by inclusion of direct image

f P (hAi) ⊂ hf P (A)i ↑ A ⊂ X .

It defines open mapping between topological spaces. Such mapping is carrying for cotopological spaces, i.e.
the direct image of open set remains open

A ⊂ hAi =⇒ f P (A) ⊂ f P (hAi) ⊂ hf P (A)i .

Bourbaki in [4] uses such notion together with continuity, ant doesn’t remark any equivalent formulation of
open mappings.
The reversely creating (−24) mapping between interior spaces is defined by opposite inclusion of inverse
image
hf − (B)i ⊂ f − (hBi) ↑ B ⊂ Y .
The pressing (−23) mapping between interior spaces demands the inclusion of direct image of interior reform

(f P × f )P (IntX ) ⊂ IntY .

The lavishing (−14) mapping between interior spaces demands the opposite inclusion of inverse image of
interior reform
IntX ⊂ (f P × f )− (IntY ) .
All such mappings define an open mapping of topological spaces.
The wasting (−31) mapping between interior spaces demands the opposite inclusion of direct image

f P (hAi) ⊃ hf P (A)i ↑ A ⊂ X .

This property wasn’t remarked for classical topological spaces, nevertheless it is meaningful for interesting
topologies.
The hiding (−32) mapping between interior spaces demands the opposite inclusion of direct image of
interior reform
(f P × f )P (IntX ) ⊃ IntY .
The reflecting (+14) mapping between interior spaces demands the the inclusion of inverse image of interior
reform
(f P × f )− (IntY ) ⊂ IntX .
Both last local properties for mappings of cotopological spaces differ from coresponding global properties.
4.2.3 Any tribe K ⊂ X P can be declared as bornology, and its member B ⊂ X as bounded set. The carrying
mapping between bornological spaces f : X −→ Y is called bounded, i.e. the image of bounded set must be
bounded
f P (A) ∈ KY ↑ A ∈ KX .
The turning mapping between bornological spaces f : X −→ Y we shall call perfect, i.e. the inverse image
of bounded set must be bounded
f − (B) ∈ KX ↑ B ∈ KY .
Such mappings are useful for bornologies compounded of compact parts in some topological space.
4.2.3.1 It is interesting to consider adherence space as example of ”local” bornology. The reform SidX :
X P →− X will be called proximity. We say that the couple hA, xi ∈ X P × X belongs to proximity if the set

31
A is proximal to the point x ∈ X. The direct appointment for each set A ⊂ X appoints the set of points for
which taken set is proximal, we shall call it a siding set with sign hAi ⊂ X. The opposite appointment for
every point x ∈ X appoints a proximal filter Fx of proximal sets
4.2.3.2 In proximity space we define bornology declaring all sets with biger siding set

A ⊂ hAi ↑ A ⊂ X

as bounded sets.
In bornology space we define proximity, declaring the set A ⊂ X proximal to the point x ∈ X, if we can
find a biger bounded set A ⊂ B ⊂ X with taken point x ∈ B.
We shall check continuity properties for mappings between proximity spaces. Usually we work with
proximal cofilters, demanding that smaller set remainded bounded and proximal.
4.2.3.3 The respecting (−13) mapping demands inclusion of siding sets

f P (hAi) ⊂ hf P (A)i .

Such mappings will be bounded for generated bornology spaces:

Let the set A ⊂ X is near of every it’s point

A ⊂ hAi .

Then the direct image of taken set f P (A) ⊂ Y has the same property

f P (A) ⊂ f P (hAi) ⊂ hf P (A)i .

And in opposite direction. The bounded mappings of bornological spaces are respecting for generated
proximity spaces:
Let the set A ⊂ X is near the point x ∈ X, i. e. we find a bigger bounded set A ⊂ B having the taken
point x ∈ B. We shall check that the direct image of taken set f P (A) ⊂ Y remains near the point f (x) ∈ Y .
It is enough to take bounded set f P (B) ⊂ Y , for which we have f P (A) ⊂ f P (B) and f (x) ∈ f P (B).
The reversely respecting (+31) mapping demands inclusion of proximal filters in source space

Ff − (y) ≤ (f P )− (Fy ) .

It is equivalent for creating (+42) mapping, which demands inclusion of proximal filters in target space

f P P (Fx ) ≤ Ff (x) .

Therefore this property is similar to local boundedness.

There are additional two equivalent properties defined with direct or inverse images of the proximity
reform.
The reversely pressing (+32) mapping between proximity spaces demands a small direct image

(f P × f )P (Sid∗X ) ⊂ SidY ,

and reversely lavishing (+41) mapping between proximity spaces demands a big inverse image

Sid∗X ⊂ (f P × f ()− (Sid∗Y ) .

4.2.3.4 Rest 4 more special continuity properties.

The cautious (−42) mapping between proximity spaces demands big direct image of proximity filter

f P P (Fx ) ≥ Ff (x) .

We can show that such mappings are perfect for generated bornologies: The inverse image f − (B ′ ) ⊂ X of
bounded set B ′ ⊂ Y remains bounded.

32
 Let the set B ′ ⊂ Y has a bigger siding set B ′ ⊂ hB ′ i. We shall check that the same is also an inverse
image
f − (B ′ ) ⊂ hf − (B ′ )i .
The smaller set B ′′ ⊂ B ′ remains siding, therefore a smaller set remains bounded

B ′′ ⊂ B ′ ⊂ hB ′ i ⊂ hB ′′ i .

We can take a smaller bounded set B ′′ := f P (f − (B ′ )) ⊂ B ′ covered by mapping, therefore for cautious
mapping we get
f − (hB ′′ i) ⊂ hf − (B ′′ )i ,
f − (B ′ ) = f − (B ′′ ) ⊂ f − (hB ′′ i ⊂ hf − (B ′′ )i = hf − (B ′ )i .
And in opposite direction. For perfect covering mapping we get reversely cautious mapping (+24) of
generated proximity spaces, i. e. for the set A′ ⊂ Y the inverse image of siding set f − (hA′ i) ⊂ X must be
included in union of siding sets
[
f − (hA′ i) ⊂ {hAi : f P (A) = A′ } .

Every siding point y ∈ hA′ i is included in bigger bounded set y ∈ B ′ , A′ ⊂ B ′ . We can take the inverse
image of this bounded set f − (y) ⊂ f − (B ′ ) and it will be bigger for any set A ⊂ X with f P (A) = A′ . For
covering mapping we can take A := f − (A′ ), therefore

f − (hA′ i) ⊂ hf − (A′ )i .

The reflecting mappings (+14) demmands small inverse image of siding reform

(f P × f )− (SidY ) ⊂ SidX .

We shall check that it reflects bounded sets for generated bornologies.

Let we have the set A ⊂ X with bounded direct image

f P (A) ⊂ hf P (A)i .

We shall show that taken set is also bounded

A ⊂ hAi .

Every point x ∈ A has image f (x) ∈ Y proximal to the direct image of taken set f P (A) ⊂ Y , therefore for
reflecting mapping it will be proximal to the taken set x ∈ hAi.
And in oposite direction. The mapping f : X → Y which reflects bounded sets will be also reflecting for
generated proximities.
Let f (x) ∈ Y is siding to the set f P (A) ⊂ Y , i. e. we can find a bigger bounded set f P (A) ⊂ B ′ with
f (x) ∈ B ′ . We can take B := f − (B ′ ), it’s image remains bounded f P (A) ⊂ f P (B) ⊂ B ′ , therefore for
reflecting mapping it itself is bounded. We have find a bigger bounded set A ⊂ B with x ∈ B.
The hiding (+23) mapping between siding spaces demands large direct image of siding reform

SidY ⊂ (f P × f )P (SidX ) .

We shall check that all mappings hiding the bounded sets of bornology spaces are also hiding between siding
spaces.
Let we have the point y ∈ Y and set A′ ⊂ Y siding in the bornology space Y , i. e. we can find a bigger
bounded set A′ ⊂ B ′ with taken point y ∈ B ′ . For hiding mapping we can find a bounded set B ⊂ X with
direct image f P (B) = B ′ . Therefore we also can find a covering point x ∈ X with f (x) = y. Taking the set
A := f − (A′ ) ∩ B we get needed siding set f P (A) = A′ .
In general we can’t check that hiding mapping between siding spaces is hiding the bounded sets.

33
 At first we notice that every bounded point y ∈ Y must be covered. Such point has proximal set A′ ⊂ Y ,
therefore for hiding mapping will be the point x ∈ X with f (x) = y and the set A ⊂ X with f P (A) = B.
Let the set B ′ ⊂ Y is bounded for the siding space, i. e. it has a bigger siding set B ′ ⊂ hB ′ i. Every point
y ∈ B ′ will have x ∈ X with f (x) = y and Bx ⊂ X with f P (Bx ) = B ′ , therefore these points compounds
the set A ⊂ X with direct image f P (A) = B ′ . We want to check that the new set has bigger siding set

A ⊂ hAi

But it can’t be calculated without some asumption of compactness for the union of proximal sets
[
{Bx : x ∈ X} .

The reversely wasting (+13) mapping between siding spaces demands small inverse image of siding filter

(f P )− (Fy ) ≤ Ff − (y) .

I don’t think that these property may have global counterpart for bornology spaces.
At this moment I don’t see any application of such continuity properties for bornological spaces. May be
they become interesting for various final or initial constructions of siding spaces.

References
[1] Bénabou J. Categories avec multiplication, 1887–1890, Comptes Rendue Acad. Sci. Paris 256 (1963).

[2] Birkhoff G. On the structure of abstract algebras, 433–454, Proc. Cambridge Phil. Soc. 31 (1935).

[3] Bourbaki N. Théorie des ensembles, Hermann: Paris 1963, Mir: Moskva 1965.

[4] Bourbaki N. Topologie générale Chapitre 1. Structures topologiques, Hermann: Paris 1961, Nauka:
Moskva 1968.

[5] Freyd P. Scedrov A. Categories and allegories, North Holland: Amsterdam – New York – Oxford – Tokyo
1990.

[6] Eilenberg S. MacLane S. General theory of natural equivalencees, 231-294, Trans. AMS 58 (1945).

[7] Gähler W. Grundstrukturen der Analysis I, II, Akademie Verlag: Berlin 1977, 1978.

[8] Gähler W. A unified general theory of convergence, 81–92, Topology and measure 3,1 (1982), Greifswald
GDR.

[9] Goguen J. Thatcher J. Wagner E. An initial algebra approach to the specification, correctness, and
implementation of abstract data types, 80–149, Current trends in programming methodology vol. 4 Data
structuring, ed. Yeh R., Prentice Hall: Englewood Cliffs, New Jersey in USA 1978.

[10] Joyse W. Braided premonoidal Mac Lane coherence, 155–176, J. of Pure and Applied Algebra 190
(2004).

[11] MacLane S. Categories for the working mathematician, Springer: New York – Heidelberg – Berlin 1971.

34
[12] Martin N. Pollard St. Closure spaces and logic, Kluwer: Dordrecht – Boston – London 1996.

[13] Valiukevičius G. The continuity property of multivalued functions (in lithuanian), Veja: Vilnius 2009.

Institute of Mathematics and Informatics

Akademijos 4,
LT-08663 Vilnius,
Lietuva
geva @ ktl.mii.lt

2010.04.19

35