DIGITAL (CO)HOMOLOGY MODULES AND DIGITAL PONTRYAGIN ALGEBRAS

DAE-WOONG LEE

A BSTRACT. We study the digital homology and cohomology modules and investigate their fun-
damental properties based on digital images. We also develop digital Hopf spaces and digital
Hopf functions between the digital Hopf spaces with digital multiplications, and explore the dig-
ital primitive homology and cohomology classes, the digital Pontryagin algebras and coalgebras
on the digital Hopf spaces as digital images.

1. I NTRODUCTION
1.1. History and Hopf space. Digital geometry deals with the bounded and finite discrete sets
in the sense of classical topology which is considered to be digital images or digitalized models
of bounded and finite subsets of the lattice points of the Euclidean space. The homology and
cohomology modules, higher homotopy groups, stable homotopy groups and equivariant homo-
topy groups are useful algebraic and topological tools in a large number of problems on algebraic
geometry and algebraic topology. In the same lode, the digital counterparts of classical homol-
ogy and cohomology modules can be an important gadget to classify (pointed) digital images
from the point of view for the digital version of the homotopy type, mathematical morphology,
image synthesis and so on in computer science. In particular, the informal definitions of many
terms in elementary homotopy and simplicial homology theory based on a digital picture on
Z2 or Z3 were nicely presented in [33], [32], [5], [10] and [13]; see also [26] for digital quasi
co-Hopf space.
In the 20th century, a lot of interesting and remarkable results on Lie groups and (pointed)
Hopf spaces as the Eckmann-Hilton dual notions of (pointed) co-Hopf spaces have been widely
investigated and developed by suitable methods for CW-spaces and usual topological spaces.
The (pointed) Hopf spaces has been the direct outgrowth of compact Lie groups in classical
homotopy theory as described in [17], [1] and [2]. Indeed, a pointed Hopf space is a triple
(Y, y0 , mY ) which consists of a pointed topological space (Y, y0 ) and a base point preserving
continuous multiplication mY : Y × Y → Y such that a constant function ey0 : Y → Y at y0
plays a role of a homotopy identity, that is, mY (cy0 , y) = y = mY (y, cy0 ) for all y ∈ Y in the
pointed homotopy category. The notion of (pointed) Hopf spaces is one of the Eckmann-Hilton
dual notions of a co-Hopf space; see [15], [20], [19], [3], [21], [4], [22], [23], [24], [25], [28],
[31] and [29] for topics related to those basic notions. It can be seen that all (pointed) Lie groups

2010 Mathematics Subject Classification. Primary 16T05; Secondary 16T15, 68U10, 57T25, 55N35.
Key words and phrases. Digital homotopy, digital n-simplex, digital (co)homology module, digital primitive
(co)homology class, digital convolution, digital Pontryagin algebra.
This work was supported by the National Research Foundation of Korea (NRF) grant funded by the Korean govern-
ment (MSIT) (No. 2018R1A2B6004407).
Tel.: +82-63-270-3367.
1
2 DAE-WOONG LEE

are (pointed) Hopf spaces. In general, the Hopf spaces lack of the associative and inversive
properties, and do not have the structure of usual topological manifolds at all.
The multiplication for a Hopf space provides the homology modules of a Hopf space with
an algebraic structure which is natural with respect to Hopf functions. In fact, we can construct
the algebra structure on the homology level which is called the Pontryagin algebra of the Hopf
space. Under the suitable conditions, the diagonal function gives the non-negatively graded
homology module a coalgebra structure. Those two algebraic structures are related to each other
and make the non-negatively graded homology modules into a classical Hopf algebra.

1.2. Motivation. There are a few standard approaches for considering a digital analogue of the
well-known usual topology on Rn such as the graph-theoretic approach, the imbedding approach,
the axiomatic approach and so on; see [11]. From this point of view, we need to investigate
another approach to study digital topology out of classical algebraic topology. In this paper,
we introduce another consideration of a digital analogue as the so-called algebraic approach
from the classical homology, cohomology and Pontryagin algebra. More precisely, this paper is
concerned with setting up more algebraic invariants and their fundamental properties of digital
homology and cohomology modules over a commutative ring with identity for a digital image
with an adjacent relation which are based on the classical homology and cohomology groups of
topological spaces in algebraic topology.

1.3. Overview. We now give a short overview of the structure of this paper. In Section 2 we
introduce the general notions of digital images with kX -adjacent relations. In Section 3 we de-
fine a digital n-simplex, the digital n-chains, and the digital homology and cohomology modules
over a commutative ring R with identity of digital images. We then construct a covariant and
contravariant functor from a category of digital images and digital continuous functions to the
one of R-modules and R-module homomorphisms, and investigate some fundamental and in-
teresting properties of digital homology and cohomology modules and the primitive homology
classes of digital images. Moreover, we show that the digital version of the dimension axiom
is guaranteed as one of the Eilenberg-Steenrod axioms in algebraic topology. In Section 4 we
consider a pointed digital Hopf space together with digital multiplications, digital homotopy as-
sociative and commutative multiplications, and base pointed preserving digital Hopf functions
between the pointed digital Hopf spaces with digital multiplications based on the pointed digital
sets. We also explore the important properties of the digital primitive cohomology classes and
digital Pontryagin algebras on the digital Hopf spaces as digital images.

2. P RELIMINARIES
Let Z be the ring of integers and R the field of all real numbers. Let Zn be the set of all lattice
points in the n-dimensional Euclidean space Rn . A digital image is a pair (X, kX ), where X is a
bounded and finite subset of Zn ⊊ Rn and kX indicates some adjacent relation for the members
of X; see below.
For an integer u with 1 ≤ u ≤ n, we firstly define an adjacent relation of a digital image in
Z as follows.
n
 DIGITAL (CO)HOMOLOGY MODULES AND DIGITAL PONTRYAGIN ALGEBRAS 3

Definition 2.1. ([9, 16]) Two points p = (p1 , p2 , . . . , pn ) and q = (q1 , q2 , . . . , qn ) with p ̸= q in
Zn are k(u, n)-adjacent if
(1) there are at most u distinct indices i with the property |pi − qi | = 1; and
(2) if |pj − qj | ̸= 1, then pj = qj for all indices j.

A k(u, n)-adjacency relation on Zn may be denoted by the number of points that are k(u, n)-
adjacent to a point p ∈ Zn . Moreover,
• the k(1, 1)-adjacent points of Z are called 2-adjacent;
• the k(1, 2)-adjacent points of Z2 are called 4-adjacent, and the k(2, 2)-adjacent points
in Z2 are called 8-adjacent;
• the k(1, 3)-adjacent points of Z3 are called 6-adjacent, the k(2, 3)-adjacent points of Z3
are called 18-adjacent, and the k(3, 3)-adjacent points of Z3 are called 26-adjacent;
• the k(1, 4)-, k(2, 4)-, k(3, 4)-, and k(4, 4)-adjacent points of Z4 are called 8-adjacent,
32-adjacent, 64-adjacent, and 80-adjacent, respectively; and so forth.
We note that the above number k(u, n), 1 ≤ u ≤ n is just the cardinality of the set of lattice
points which have the k(u, n)-adjacent relations centered at p in Zn . We mostly denote the
k(u, n)-adjacent relation on a digital image X by kX -adjacent relation for short if there is no
chance of ambiguity.

Definition 2.2. ([6, 7]) A digital image (X, kX ) in Zn is said to be kX -connected if for every
pair of points {x, y} ⊂ X with x ̸= y, there exists a set P = {x0 , x1 , . . . , xs } ⊂ X of s + 1
distinct points such that x = x0 , xs = y, and xi and xi+1 are kX -adjacent for i = 0, 1, . . . , s − 1.

The following is a minor modification of an earlier definition of a digital continuous function

given in [7, Definition 2.3]; see also [34].

Definition 2.3. Let (X, kX ) and (Y, kY ) be the digital images with kX -adjacent and kY -adjacent
relations, respectively. A function f : X → Y from (X, kX ) to (Y, kY ) is said to be a (kX , kY )-
continuous function if the image of every kX -connected subset of the digital image X under f
is a kY -connected subset of Y .

Let (X, kX ), (Y, kY ) and (Z, kZ ) be digital images. Then if f : X → Y is a (kX , kY )-

continuous function and g : Y → Z is a (kY , kZ )-continuous function, then it is not difficult to
show that the composite g ◦ f : X → Z of f and g is (kX , kZ )-continuous. Thus, we are able
to construct the category D of digital images and digital continuous functions; that is, the object
classes of D are digital images and the morphism classes are digital continuous functions.

Definition 2.4. ([6, 8]) Let a, b ∈ Z with a < b. A digital interval is a set of the form [a, b]Z =
{z ∈ Z | a ≤ z ≤ b} in which the 2-adjacent relation in Z is assumed.

Definition 2.5. ([18, 7, 9]) Let (X, kX ) and (Y, kY ) be digital images with kX -adjacent and
kY -adjacent relations, respectively, and let f, g : X → Y be (kX , kY )-continuous functions.
Suppose that there is a positive integer m and a function F : X × [0, m]Z → Y such that
• F (x, 0) = f (x) and F (x, m) = g(x) for all x ∈ X;
• the induced function Fx : [0, m]Z → Y, x ∈ X defined by Fx (t) = F (x, t) for all
t ∈ [0, m]Z is (2, kY )-continuous; and
4 DAE-WOONG LEE

• the induced function Ft : X → Y, t ∈ [0, m]Z defined by Ft (x) = F (x, t) for all x ∈ X
is (kX , kY )-continuous.
Then F is called a digital (kX , kY )-homotopy between f and g, written F : f ≃(kX ,kY ) g, and
f and g are called digitally (kX , kY )-homotopic in Y .

We end this section with the description of the pointed versions of digital images to develop
the pointed digital category as follows.

Definition 2.6. ([27, 30]) A pointed digital image with kX -adjacent relation is a triple (X, x0 , kX ),
where X is a digital image and x0 ∈ X. In this case, x0 is said to be a base point of (X, x0 , kX ).
A pointed digital continuous function
f : (X, x0 , kX ) → (Y, y0 , kY )
is a (kX , kY )-continuous function from (X, x0 ) to (Y, y0 ) such that
f (x0 ) = y0 .
A digital homotopy
F : X × [0, m]Z → Y
between pointed digital continuous functions f and g is said to be pointed digital (kX , kY )-
homotopy between f and g if F (x0 , t) = y0 for all t ∈ [0, m]Z .

We now construct the so-called pointed digital category D∗ of pointed digital images and
base point preserving digitally continuous functions; that is, the object classes of D∗ are pointed
digital images and the morphism classes are base point preserving digital continuous functions.

3. D IGITAL HOMOLOGY AND COHOMOLOGY MODULES

In this section, we consider the digital homology and cohomology modules over a commu-
tative ring R with identity ‘1R ’ (compare with [10] in the case of digital simplicial homology
groups). For i = 0, 1, . . . , n, we let ei be the point in Zn+1 having coordinates all zeros except
for 1 in the (1 + i)th coordinate; that is, e0 = (1, 0, 0, . . . , 0), e1 = (0, 1, 0, . . . , 0), . . ., and
en = (0, 0, . . . , 0, 1) in Zn+1 .

Definition 3.1. A digital convex combination of points e0 , e1 , . . . , en in Zn+1 is a point x with

x = r0 e0 + r1 e1 + · · · + rn en ,
∑n
where i=0 ri = 1, and ri = 0 or 1. The coefficients (r0 , r1 , . . . , rn ) of
x = r0 e0 + r1 e1 + · · · + rn en
are called the digital barycentric coordinates of x.

Unlike the classical convex combination of points, it can be easily verified that x is a digital
convex combination of e0 , e1 , . . . , en if and only if x is an element of {e0 , e1 , . . . , en }. We
let ∆n be the set of all digital convex combinations of points e0 , e1 , . . . , en in Zn+1 ; that is,
∆n = {e0 , e1 , . . . , en } which is completely different from the usual convex combinations in
algebraic topology when n ≥ 1. Considering ∆n as the digital image with k(2, n + 1)-adjacent
 DIGITAL (CO)HOMOLOGY MODULES AND DIGITAL PONTRYAGIN ALGEBRAS 5

relation, we can see that it is k(2, n + 1)-connected. We denote by k∆n the k(2, n + 1)-adjacent
relation in the digital image ∆n for our notational convenience as just mentioned earlier.

Definition 3.2. Using the k∆n -adjacent relation in the digital image ∆n = {e0 , e1 , . . . , en }, we
consider an orientation of ∆n as a linear ordering of its vertices, and call it a digital standard
n-simplex with a linear order for its orientation.

Definition 3.3. Let (X, kX ) be a digital image with kX -adjacent relation. A digital n-simplex in
(X, kX ) is a (k∆n , kX )-continuous function σ : (∆n , k∆n ) → (X, kX ), where ∆n is the digital
standard n-simplex.

A constant function is an example of a digital n-simplex in (X, kX ). Moreover, for every

{ei , ej } ⊂ ∆n such that ei and ej are k∆n -adjacent in ∆n , either σ(ei ) = σ(ej ), or σ(ei ) and
σ(ej ) are kX -adjacent in (X, kX ).

Definition 3.4. Let R be a commutative ring with identity ‘1R ’ and let (X, kX ) be a digital
image with kX -adjacent relation. For each n ≥ 0, we define dCn (X; R) to be the non-negatively
graded free R-module with basis all digital n-simplexes in (X, kX ). The elements of dCn (X; R)
are called digital n-chains in (X, kX ).

We note that the oriented boundary of a digital n-simplex σ : (∆n , k∆n ) → (X, kX ) has to
∑
be ni=0 (−1)i (σ|{e0 ,...,êi ,...,en } ), where the symbol êi means that the vertex ei is to be deleted
from the array in the digital standard n-simplex ∆n .

Definition 3.5. For each n and i, we now define the ith face function
ϵi = ϵni : ∆n−1 −→ ∆n
to be the function which sends the ordered vertices {e0 , . . . , en−1 } to the ordered vertices {e0 , . . . , eˆi , . . . , en }
{e0 , . . . , ei , . . . , en } preserving the displayed orderings as follows:
• ϵn0 : (r0 , r1 , . . . , rn−1 ) 7−→ (0, r0 , r1 , . . . , rn−1 ); and
• ϵni : (r0 , r1 , . . . , rn−1 ) 7−→ (r0 , . . . , ri−1 , 0, ri , . . . , rn−1 ) for i ≥ 1.

For example, there are three face functions ϵ2i : ∆1 → ∆2 such as ϵ20 : {e0 , e1 } → {e1 , e2 };
ϵ21 : {e0 , e1 } → {e0 , e2 }; and ϵ22 : {e0 , e1 } → {e0 , e1 }.

Definition 3.6. Let (X, kX ) be a digital image with kX -adjacent relation, and let σ : (∆n , k∆n )
→ (X, kX ) be a digital n-simplex in (X, kX ). Then a map ∂n : dCn (X; R) → dCn−1 (X; R)
defined by  n

 ∑
 (−1)i σ ◦ ϵni if n ≥ 1;
∂n σ = i=0


0 if n = 0.
is called the digital boundary operator of the digital image (X, kX ).

It can be seen that ∂n : dCn (X; R) → dCn−1 (X; R) is an R-module homomorphism. In

∑
particular, if X = ∆n and 1∆n : ∆n → ∆n is the identity, then ∂n (1∆n ) = ni=0 (−1)i ϵni .
Moreover, if k < j, then we obtain
ϵn+1
j ◦ ϵnk = ϵn+1
k ◦ ϵnj−1 : ∆n−1 −→ ∆n+1 . (3.1)
6 DAE-WOONG LEE

Proposition 3.7. For all n ≥ 0, we have ∂n ◦ ∂n+1 = 0.

Proof. It suffices to show that the equation holds for each basis of the free R-module, i.e., digital
(n + 1)-simplex σ : (∆n+1 , k∆n+1 ) → (X, kX ) because we can extend it by linearity to the
digital n-chains. We have
∑
n+1
∂n ◦ ∂n+1 (σ) = ∂n ( (−1)j σ ◦ ϵn+1
j )
j=0
= ∂n (σ ◦ ϵn+1
0 − σ ◦ ϵn+1
1 + σ ◦ ϵn+1
2 + · · · + (−1)n+1 σ ◦ ϵn+1
n+1 )
∑
n+1 ∑n
= (−1)j ( (−1)k σ ◦ ϵn+1
j ◦ ϵnk )
j=0 k=0
∑∑
n+1 n (3.2)
= (−1)j+k σ ◦ ϵn+1
j ◦ ϵnk
∑
j=0 k=0 ∑
= (−1)j+k σ ◦ ϵn+1
j ◦ ϵnk + (−1)j+k σ ◦ ϵn+1
j ◦ ϵnk
∑
j≤k ∑
k<j
= (−1)j+k σ ◦ ϵn+1
j ◦ ϵnk + (−1)j+k σ ◦ ϵn+1
k ◦ ϵnj−1 .
j≤k k<j

The left-hand term and the right-hand term on the last two summations in (3.2) can be expressed
by the upper triangular region and the lower triangular region, respectively, in the (n + 2) × (n +
1)-matrix-like form as follows:

TABLE 1. In case j ≤ k

j ≤k k=0 k=1 ... k=n

j =0 σϵn+1
0 ϵn0 −σϵn+1
0 ϵn1 ... (−1)n σϵn+1
0 ϵnn
j =1 0 σϵn+1
1 ϵn1 ... (−1)n+1 σϵn+1
1 ϵnn
j =2 0 0 ... (−1)n+2 σϵn+2
2 ϵnn
.. .. .. .. ..
. . . . .
j=k 0 0 ... (−1)n+j σϵn+1
j ϵnn
.. .. .. .. ..
. . . . .
j=n 0 0 ... (−1)2n σϵn+1
n ϵn
n

j =n+1 0 0 ... 0

By using the equation (3.1), we can rewrite Table 2 as follows:

Those terms on the Tables 1 and 3 cancel in pairs. Thus
∂n ◦ ∂n+1 (σ) = 0
for each digital (n + 1)-simplex σ : (∆n+1 , k∆n+1 ) → (X, kX ) as required. □

Definition 3.8. The kernel of ∂n : dCn (X; R) → dCn−1 (X; R) is called the module of dig-
ital n-cycles in (X, kX ) and denoted by dZn (X; R). The image of ∂n+1 : dCn+1 (X; R) →
dCn (X; R) is called the module of digital n-boundaries in (X, kX ) and denoted by dBn (X; R).
 DIGITAL (CO)HOMOLOGY MODULES AND DIGITAL PONTRYAGIN ALGEBRAS 7

TABLE 2. In case k < j

k<j k=0 k=1 ... k=n

j=0 0 0 ... 0
j=1 −σϵ1 ϵ0
n+1 n
0 ... 0
j=2 σϵn+1
2 ϵn0 −σϵ2 ϵ1
n+1 n
... 0
.. .. .. ..
. . . .
j=k (−1)j σϵn+1
j ϵn0 (−1)j+1 σϵn+1
j ϵn1 ... 0
.. .. .. ..
. . . .
j=n (−1)n σϵn+1
n ϵ0
n (−1)n+1 σϵn+1
n ϵ1
n ... 0
n+1 n+1
j =n+1 (−1)n+1 σϵn+1 ϵn0 (−1)n+2 σϵn+1 ϵn1 ... (−1)2n+1 σϵn+1 n
n+1 ϵn

TABLE 3. In case k < j

k<j k=0 k=1 ... k=n

j=0 0 0 ... 0
j=1 −σϵ0 ϵ0
n+1 n
0 ... 0
j=2 σϵn+1
0 ϵn1 −σϵn+1
1 ϵn1 ... 0
.. .. .. ..
. . . .
j=k (−1)j σϵn+1
0 ϵnj−1 (−1)j+1 σϵn+1
1 ϵnj−1 ... 0
.. .. .. ..
. . . .
j=n (−1)n σϵn+1
0 ϵnn−1 (−1)n+1 σϵn+1
1 ϵnn−1 ... 0
j =n+1 (−1)n+1 σϵn+1
0 ϵnn (−1)n+2 σϵn+1
1 ϵnn ... (−1)2n+1 σϵn+1
n ϵn
n

By Proposition 3.7, each digital n-boundary of digital (n+1)-chains is automatically a digital

n-cycle; that is, dBn (X; R) is a submodule of dZn (X; R) for each n ≥ 0. Thus we can define
the following:
Definition 3.9. For each n ≥ 0, the nth digital homology module dHn (X; R) over R of a digital
image (X, kX ) with kX -adjacent relation is defined by
dHn (X; R) = dZn (X; R)/dBn (X; R).
The coset [zn ] = zn + Bn (X; R) is called the digital homology class of zn , where zn is a digital
n-cycle. The nth digital cohomology module dH n (X; R) over R of a digital image (X, kX ) is
defined as the corresponding cohomology of the cochain complex over R obtained by the dual
modules over R and dual R-module homomorphisms.
We note that if f : (X, kX ) → (Y, kY ) is (kX , kY )-continuous and if σ : (∆n , k∆n ) →
(X, kX ) is a digital n-simplex in (X, kX ), then
f ◦ σ : (∆n , k∆n ) → (Y, kY )
8 DAE-WOONG LEE

is a digital n-simplex in (Y, kY ). By extending by linearity, we have an R-module homomor-

phism
f♯ : dCn (X; R) → dCn (Y ; R)
defined by
f♯ (Σrσ σ) = Σrσ (f ◦ σ),
where rσ is an element of the ring R. Moreover, the following diagram
f♯
dCn (X; R) / dCn (Y ; R)
′
∂n
∂n
 f♯ 
dCn−1 (X; R) / dCn−1 (Y ; R),

commutes for every n ≥ 0, where ∂n′ is a digital boundary operator of the digital image (Y, kY ).
Let D be the category of digital images and digital continuous functions, and let M be the
category of R-modules and R-module homomorphisms. Then we have the following (see [10,
Theorem 3.15] for the case of digital simplicial homology groups).

Proposition 3.10. For each n ≥ 0, dHn : D → M is a covariant functor, and dH n : D → M

is a contravariant functor.

Proof. If f : (X, kX ) → (Y, kY ) is a (kX , kY )-continuous function, then we define

f∗ = dHn (f ) : dHn (X; R) → dHn (Y ; R)
by
dHn (f )([zn ]) = f♯ (zn ) + dBn (Y ; R),
where zn ∈ dZn (X; R). We note that if zn ∈ dZn (X; R), then ∂n (zn ) = 0 and thus
∂n′ ◦ f♯ (zn ) = f♯ ◦ ∂n (zn ) = 0;
that is,
f♯ (zn ) ∈ ker ∂n′ = dZn (Y ; R).
Moreover, if bn ∈ dBn (X; R), then
∂n+1 (bn+1 ) = bn
for some bn+1 ∈ dCn+1 (X; R), and
′
f♯ (bn ) = f♯ ◦ ∂n+1 (bn+1 ) = ∂n+1 ◦ f♯ (bn+1 ) ∈ dBn (Y ; R).
If bn ∈ dBn (X; R), then
f♯ (zn + bn ) + dBn (Y ; R) = f♯ (zn ) + dBn (Y ; R);
that is, the definition of dHn (f ) is independent of the choice of representatives.
If 1X : (X, kX ) → (X, kX ) is the identity and σ : (∆n , k∆n ) → (X, kX ) is a digital
n-simplex, then
1X ♯ (σ) = 1X ◦ σ = σ;
that is, dHn (1X ) = 1dHn (X;R) , the identity automorphism on dHn (X; R).
 DIGITAL (CO)HOMOLOGY MODULES AND DIGITAL PONTRYAGIN ALGEBRAS 9

If f : (X, kX ) → (Y, kY ) is digitally (kX , kY )-continuous and g : (Y, kY ) → (Z, kZ ) is a

digitally (kY , kZ )-continuous function, then
(g ◦ f )♯ (σ) = (g ◦ f ) ◦ (σ) = g ◦ (f ◦ σ) = g♯ (f♯ (σ)).
Therefore, we have
dHn (g ◦ f ) = dHn (g) ◦ dHn (f ).
Similarly, it can be shown that dH n : D → M is a contravariant functor in a precisely analogous
way. □
Remark 3.11. Let (X, kX ) be a digital image with kX -adjacent relation, and let X × [0, m]Z
be a digital image with the cartesian product whose adjacent relation is denoted by kC . Then the
map ψi : X → X × [0, m]Z defined by ψi (x) = (x, i) for i = 0, m is (kX , kC )-continuous.
Moreover, if
dHn (ψ0 ) = dHn (ψm ) : dHn (X; R) → dHn (X × [0, m]; R)
as R-module homomorphisms, and if
f ≃(kX ,kY ) g : (X, kX ) → (Y, kY ),
then we have
dHn (f ) = dHn (g) : dHn (X, R) → dHn (Y ; R).
Let i1 : X → X × X be the first inclusion and i2 : X → X × X the second inclusion. Then,
by Proposition 3.10, we have R-module homomorphisms
i1∗ , i2∗ : dH∗ (X; R) → dH∗ (X × X; R)
of digital homology modules induced by i1 and i2 , respectively.
Definition 3.12. An element x ∈ dH∗ (X; R) is said to be a digital primitive homology class if
∆∗ (x) = i1∗ (x) + i2∗ (x), where ∆ : X → X × X is the diagonal map.
Let P dH∗ (X; R) denote the submodule of dH∗ (X; R) with coefficients in a commutative
ring R consisting of all digital primitive homology classes. Then we have the following.
Proposition 3.13. Let f : X → Y be a (kX , kY )-continuous function. Then
f∗ (P dH∗ (X; R)) ⊆ P dH∗ (Y ; R).
Proof. If x is any digital primitive homology class of dH∗ (X; R) with coefficients in a commu-
tative ring R, then from the commutative diagram
f∗
dH∗ (X; R) / dH∗ (Y ; R)

∆∗ ∆∗
 (f ×f )∗ 
dH∗ (X × X; R) / dH∗ (Y × Y ; R),

we have
∆∗ (f∗ (x)) = (f × f )∗ ◦ ∆∗ (x)
= (f × f )∗ (i1∗ (x) + i2∗ (x))
= i1∗ (f∗ (x)) + i2∗ (f∗ (x)),
10 DAE-WOONG LEE

where i1 : W → W × W is the first inclusion and i2 : W → W × W is the second inclusion on

W = X or Y ; that is, the homomorphic image of the digital primitive homology classes is also
digital primitive as required. □

The following result is the digital counterpart of the dimension axiom as one of the Eilenberg-
Steenrod axioms in classical algebraic topology.

Proposition 3.14. If (X, kX ) is a one-point digital image with kX -adjacent relation, then

0 for all n ≥ 1;
dHn (X; R) =
R n = 0.

Proof. For each n ≥ 0, we have dCn (X; R) ∼= R ∼ = < σn > generated by the digital n-simplex
σn : (∆ , k∆n ) → (X, kX ) which is a constant function in this case. From the definition of the
n

boundary operator ∂n : dCn (X; R) → dCn−1 (X; R), we obtain


0 if n is odd;
∂n (σn ) =
σn−1 if n is even.

Thus, ∂n is the trivial homomorphism if n is odd, and it is an isomorphism if n is even.

Assume that n ≥ 1, and consider the chain complex of R-modules
∂n+1 ∂n
··· / dCn+1 (X; R) / dCn (X; R) / dCn−1 (X; R) / ··· .

(1) If n ≥ 1 is odd, then ∂n is the trivial homomorphism and ∂n+1 is an isomorphism, and
thus
dCn (X; R) = ker ∂n = dZn (X; R)
and
dCn (X; R) = im ∂n+1 = dBn (X; R).
Therefore, we have
dHn (X; R) = dZn (X; R)/dBn (X; R) = dCn (X; R)/dCn (X; R) = 0.
(2) Similarly, if n ≥ 2 is even, then ∂n is an isomorphism, and thus
Zn (X, R) = ker ∂n = 0
so that
dHn (X; R) = dZn (X; R)/dBn (X; R) = 0/dBn (X; R) = 0.
On the other hand, from the sequence
∂1 ∂0
··· / dC1 (X; R) / dC0 (X; R) / 0,

we have 
ker ∂ = dZ (X; R) ∼= R; and
0 0
im ∂1 = dB0 (X; R) ∼
=0
 DIGITAL (CO)HOMOLOGY MODULES AND DIGITAL PONTRYAGIN ALGEBRAS 11

so that
dH0 (X; R) = dZ0 (X; R)/dB0 (X; R) ∼
=R
as required. □

Definition 3.15. Given a digital image (X, kX ), we define an equivalence relation on (X, kX )
by setting x ∼ y if there is a kX -connected digital subimage containing both x and y. The
equivalence classes are called the kX -connected components of (X, kX ).

Proposition 3.16. Let {(Xλ , kX )|λ ∈ Λ} be the set of kX -connected components of a digital
image (X, kX ). Then ⊕
dHn (X; R) ∼= dHn (Xλ ; R)
λ∈Γ
for every n ≥ 0.
∑
Proof. If d = ri σi ∈ dCn (X; R), then the digital continuity of the digital n-simplex shows
that each im(σi ) is contained in a unique kX -connected component of (X, kX ). Thus we may
∑ ∑
write c = λ∈Λ cλ , where cλ is the sum of those terms; that is, cλ = rj σjλ in c for which
im(σjλ ) ⊂ Xλ for each λ ∈ Λ. Now we define an isomorphism
⊕
Ψ : dCn (X; R) → dCn (Xλ ; R)
λ∈Λ

by c 7→ (cλ )λ∈Λ . Now c is a cycle if and only if each cλ is a cycle. Since im(σjλ ) ⊂ Xλ
∑
implies im(σjλ ◦ ϵi ) ⊂ Xλ , ∂σjλ ∈ dCn−1 (Xλ ; R), and the assumption 0 = ∂c = ∂cλ implies
⊕
∂cλ = 0 for all λ ∈ Λ. Indeed, an element in the direct sum λ∈Λ dCn−1 (Xλ ; R) is zero if and
only if its coordinates are all zero. It follows that the map
⊕
ψ : dHn (X; R) → dHn (Xλ ; R)
λ∈Λ

given by [c] 7→ ([cλ ])λ∈Λ is well defined.

We now define a map ⊕
φ: dHn (Xλ ; R) → dHn (X; R)
λ∈Λ
∑
by ([cλ ])λ∈Λ 7→ [ λ∈Λ cλ ]. Then it is easy to see that ψ and φ are the mutual inverses as
required. □

4. D IGITAL H OPF SPACES , P ONTRYAGIN ALGEBRAS AND COALGEBRAS

From many kinds of algebraic structures in algebra, we can think of a Hopf group in algebraic
topology as a generalization of a usual group in algebra (see [35] and [36]). As an application to
the topics of computer science, we now transform the notions of Hopf spaces or Hopf groups in
algebraic topology into those of digital theoretic counterparts (compare with [12] and [14]).

Definition 4.1. ([27, 30]) Let ey0 : Y → Y be the constant function at y0 and let 1Y : Y → Y
be the identity function on Y . A digital Hopf space Y = (Y, y0 , kY , mY ) consists of a pointed
12 DAE-WOONG LEE

digital image (Y, y0 ) with an adjacent relation kY and a (kY ×Y , kY )-continuous function mY :
Y × Y → Y such that the following diagram
(ey0 ,1Y ) (1Y ,ey0 )
Y MM / Y ×Y o Y
MMM
MMM m q qqqq
M Y qq
1Y MMMM  qqqqq 1Y
& xq
Y
is commutative up to pointed digital homotopy. Here, kY ×Y is an adjacent relation on Y × Y ,
and (ey0 , 1Y ) : Y → Y × Y is the composite of the diagonal function ∆ : Y → Y × Y with the
product of functions (ey0 × 1Y ) : Y × Y → Y × Y , and similarly for (1Y , ey0 ). In this case, the
(kY ×Y , kY )-continuous function mY : Y × Y → Y above is called a digital multiplication on
(Y, y0 , kY , mY ), and ey0 is called a digital homotopy identity.

As usual, we denote by [f ] the pointed digital homotopy class as the equivalence class of a
pointed digital continuous function f : (X, x0 , kX ) → (Y, y0 , kY ).

Definition 4.2. ([27, 30]) Let (Y, y0 , kY , mY ) be a digital Hopf space with a digital multiplica-
tion mY : Y × Y → Y . For pointed (kX , kY )-continuous functions f, g : (X, x0 ) → (Y, y0 ),
we define a digital convolution [f ] ⊞ [g] between pointed digital homotopy classes [f ] and [g] by
the pointed digital homotopy class of the following compositions
∆ f ×g mY
f ⊞g :X / X ×X / Y ×Y / Y;

that is, [f ] ⊞ [g] = [mY (f × g)∆] = [f ⊞ g], where ∆ is the diagonal function.

Theorem 4.3. Let (Y, y0 , kY , mY ) be a digital Hopf space with a digital multiplication mY :
Y × Y → Y . If f, g : (X, x0 ) → (Y, y0 ) are (kX , kY )-continuous functions and x ∈
P dH(X; R), then
([f ] ⊞ [g])∗ (x) = f∗ (x) + g∗ (x),
where f∗ , g∗ : dH∗ (X; R) → dH∗ (Y ; R) are homomorphisms of digital homology modules
over R induced by (kX , kY )-continuous functions f and g, respectively.

Proof. We note that f ⊞ g ≃(kX ,kY ) mY ◦ (f × g) ◦ ∆ and the following diagram

iX
1 ∗
dH∗ (X; R) / dH∗ (X × X; R)

f∗ (f ×g)∗
 iY
1 ∗

dH∗ (Y ; R) / dH∗ (Y × Y ; R),

2 ∗ : dH∗ (X; R) → dH∗ (X × X; R) together with

is strictly commutative and similarly for iX
i2 ∗ : dH∗ (Y ; R) → dH∗ (Y × Y ; R) in digital homology R-modules. Since Y has the digital
Y
 DIGITAL (CO)HOMOLOGY MODULES AND DIGITAL PONTRYAGIN ALGEBRAS 13

Hopf structure and x is a digital primitive homology class, we have

(f ⊞ g)∗ (x) = (mY ◦ (f × g) ◦ ∆)∗ (x)
= mY ∗ ◦ (f × g)∗ ◦ ∆∗ (x)
= mY ∗ ◦ (f × g)∗ (iX X
1 ∗ (x) + i2 ∗ (x))
= mY ∗ (iY1 ∗ ◦ f∗ (x) + iY2 ∗ ◦ g∗ (x))
= (mY ∗ ◦ iY1 ∗ )f∗ (x) + (mY ∗ ◦ iY2 ∗ )g∗ (x)
= 1Y ∗ ◦ f∗ (x) + 1Y ∗ ◦ g∗ (x)
= f∗ (x) + g∗ (x),
where 1Y : Y → Y is the identity map. □

The following shows that the digital Hopf spaces are closed under the digital products of
digital images.

Theorem 4.4. If (X, x0 , kX , mX ) and (Y, y0 , kY , mY ) are digital Hopf spaces, then (X ×
Y, x0 × y0 , kX×Y , mX×Y ) is a digital Hopf space.

Proof. Let mX : X × X → X and mY : Y × Y → Y be the digital multiplications on X and

Y , respectively. We define a function
mX×Y : (X × Y ) × (X × Y ) −→ X × Y
by making the following diagram commute:
mX×Y
(X × Y ) × (X × Y ) / X ×Y
ll 6
llll
l
1X ×SY ×X ×1Y lll
lllllmX ×mY
 l
(X × X) × (Y × Y ),
where SY ×X : Y × X → X × Y is the switching function, and 1X and 1Y are the identity
functions on X and Y , respectively. Then, for all (x, y) ∈ X × Y , we have
mX×Y (1X×Y , cx0 ×y0 )(x, y) = mX×Y (1X×Y × cx0 ×y0 )∆(x, y)
= mX×Y (x, y, x0 , y0 )
= (mX × mY )(1X × SY ×X × 1Y )(x, y, x0 , y0 )
= (mX × mY )(x, x0 , y, y0 )
= (mX (x, x0 ), mY (y, y0 ))
= (mX (1X × cx0 )∆(x), mY (1Y × cy0 )∆(y))
= (mX (1X , cx0 )(x), mY (1Y , cy0 )(y))
≃(kX×Y ,kX×Y ) (1X (x), 1Y (y)) = (x, y)
= 1X×Y (x, y);
that is,
mX×Y (1X×Y , cx0 ×y0 ) ≃(kX×Y ,kX×Y ) 1X×Y .
Similarly, we also obtain
mX×Y (cx0 ×y0 , 1X×Y ) ≃(kX×Y ,kX×Y ) 1X×Y .
14 DAE-WOONG LEE

Therefore,
mX×Y : (X × Y ) × (X × Y ) → X × Y
is a digital multiplication; that is, (X × Y, x0 × y0 , kX×Y , mX×Y ) is a digital Hopf space. □

Corollary 4.5. Let (Y, y0 , kY , mY ) be a digital Hopf space and let i1 , i2 : Y → Y × Y be the
first and second inclusions, respectively. Then we obtain
(i1 ⊞ i2 )∗ = ∆∗ ,
and, in particular,
(i1 ⊞ i2 )∗ (P dH∗ (Y ; R)) = ∆∗ (P dH∗ (Y ; R)).

Proof. Let mY : Y × Y → Y be the digital multiplication on (Y, y0 ). Then, by Theorem 4.4,

Y × Y is also a digital Hopf space with a digital multiplication mY ×Y . Since mY ◦ i1 ≃(kY ,kY )
1Y and mY ◦ i2 ≃(kY ,kY ) 1Y , we have

[i1 ⊞ i2 ] = [mY ×Y (i1 × i2 )∆]

= [(mY × mY ) ◦ (1Y × SY ×Y × 1Y ) ◦ (i1 × i2 )∆]
= [(mY × mY ) ◦ (i1 × i2 )∆]
= [((mY ◦ i1 ) × (mY × i2 ))∆]
= [(1Y × 1Y )∆]
= [∆],
where SY ×Y : Y × Y → Y × Y is the switching function. If y is any digital primitive homology
class, then, by Theorem 4.3, we obtain
(i1 ⊞ i2 )∗ (y) = ∆∗ (y) = i1∗ (y) + i2∗ (y)
as required. □

Definition 4.6. Let (Y, y0 , kY , mY ) be a digital Hopf space with a digital multiplication mY :
Y × Y → Y . Then an element y ∈ dH ∗ (Y ; R) is said to be a digital primitive cohomology
class if
m∗Y (y) = p∗1 (y) + p∗2 (y),
where p∗1 , p∗2 : dH ∗ (Y ; R) → dH ∗ (Y × Y ; R) are the homomorphisms of cohomology R-
modules induced by p1 , p2 : Y × Y → Y , the first and second projections onto Y , respectively.

Definition 4.7. Let (X, x0 , kX , mX ) and (Y, y0 , kY , mY ) be pointed digital H-spaces. A (kX , kY )-
continuous function f : X → Y is said to be a digital Hopf function if the following diagram
f ×f
X ×X / Y ×Y
mX mY
 f 
X /Y

commutes up to the digital homotopy.

Let P dH ∗ (X; R) denote the submodule of dH ∗ (X; R) consisting of all digital primitive
cohomology classes. Then we have the dual of Proposition 3.13 as follows.
 DIGITAL (CO)HOMOLOGY MODULES AND DIGITAL PONTRYAGIN ALGEBRAS 15

Proposition 4.8. Let (X, x0 , kX , mX ) and (Y, y0 , kY , mY ) be digital Hopf spaces, and let f :
X → Y be a digital Hopf function between digital Hopf spaces. Then
f ∗ (P dH ∗ (Y ; R)) ⊆ P dH ∗ (X; R).

Proof. Since f : X → Y be a digital Hopf function, we have

f ◦ mX ≃(kX×X ,kY ) mY ◦ (f × f ).
If y is any digital primitive cohomology class of dH ∗ (Y ; R) with coefficients in a commutative
ring R, then from the commutative diagram
f∗
dH ∗ (Y ; R) / dH ∗ (X; R)

mY ∗ mX ∗
 (f ×f )∗ 
dH ∗ (Y × Y ; R) / dH ∗ (X × X; R),

we have
m∗X (f ∗ (y)) = (f × f )∗ ◦ m∗Y (y)
= (f × f )∗ (p1 ∗ (y) + p2 ∗ (y))
= p1 ∗ (f ∗ (y)) + p2 ∗ (f ∗ (y)),
where p1 , p2 : W → W × W are the first and second projections, respectively, onto W (= X
or Y ); that is, the homomorphic image of the digital primitive cohomology classes is also digital
primitive as required. □

The following is a dual of Theorem 4.3.

Theorem 4.9. Let (Y, y0 , kY , mY ) be a digital Hopf space with a digital multiplication mY :
Y × Y → Y , and let f, g : (X, x0 ) → (Y, y0 ) be (kX , kY )-continuous functions. If y ∈
dH(Y ; R) is digital primitive cohomology class, then
([f ] ⊞ [g])∗ (y) = f ∗ (y) + g ∗ (y),
where f ∗ , g ∗ : dH ∗ (Y ; R) → dH ∗ (X; R) are homomorphisms of digital cohomology R-
modules induced by the (kX , kY )-continuous functions f and g, respectively.

Proof. Let p1 W , p2 W : W × W → W be the first and second projections, respectively, onto

W (= X or Y ). Since f ⊞ g ≃(kX ,kY ) mY ◦ (f × g) ◦ ∆ and the following diagram
p∗1 Y
dH ∗ (Y ; R) / dH ∗ (Y × Y ; R)

f∗ (f ×g)∗
 p∗1 X 
dH ∗ (X; R) / dH ∗ (X × X; R),
16 DAE-WOONG LEE

is strictly commutative and similarly for p∗2X : dH ∗ (X; R) → dH ∗ (X × X; R) together with

p∗2Y : dH ∗ (Y ; R) → dH ∗ (Y × Y ; R) in digital cohomology modules over R, we have

(f ⊞ g)∗ (y) = (mY ◦ (f × g) ◦ ∆)∗ (y)

= ∆∗ ◦ (f × g)∗ ◦ m∗Y (y)
= ∆∗ ◦ (f × g)∗ (p∗1Y (y) + p∗2Y (y))
= ∆∗ ◦ (p∗1X ◦ f ∗ (y) + p∗2X ◦ g ∗ (y))
= (∆∗ ◦ p∗1X )f ∗ (y) + (∆∗ ◦ p∗2X )g ∗ (y)
= 1dH ∗ (X;R) ◦ f ∗ (y) + 1dH ∗ (X;R) ◦ g ∗ (y),
= f ∗ (y) + g ∗ (y),
where 1dH ∗ (X;R) : dH ∗ (X; R) → dH ∗ (X; R) is the identity automorphism of R-modules, and
the third equality is guaranteed by the fact that y is a digital primitive cohomology class. □

Definition 4.10. The digital homology cross product is defined by the homomorphism
µX = × : dH∗ (X; R) ⊗R dH∗ (X; R) → dH∗ (X × X; R)
of R-modules sending [x] ⊗ [y] to [x × y].

Under what conditions can we say that the submodule P dH∗ (X; R) ⊆ dH∗ (X; R) consisting
of digital primitive homology classes is equal to dH∗ (X; R)? The following gives an answer to
this query:

Theorem 4.11. Let dHs (X; R) = 0 for s ≤ n − 1. Then P dHs (X; R) is equal to dHs (X; R)
for all s ≤ 2n.

Proof. We consider the Künneth exact sequence

⊕
0 −→ s+t=m dHs (X; R) ⊗ dHt (X; R) −→ dHm (X × X; R)
⊕ (4.1)
−→ s+t=m TorR (dHs (X; R), dHt−1 (X; R)) −→ 0
of R-modules in algebraic topology; see [35, page 228]. If m ≤ 2n, then, by our assumption, the
torsion part in the above short exact sequence is trivial so that the cross product is an isomorphism
of R-modules. We now consider a homomorphism of R-modules
∆∗ : dHm (X; R) → dHm (X × X; R)
for all m ≥ 0 induced by the diagonal map
∆ : X → X × X.
We see that the target of the R-module homomorphism ∆∗ is isomorphic to the first term of the
Künneth exact sequence (4.1). Therefore, for all xm ∈ dHm (X; R), we have
∆∗ (xm ) = xm ⊗ 1 + xm−1 ⊗ x1 + xm−2 ⊗ x2 + · · · + x2 ⊗ xm−2 + x1 ⊗ xm−1 + 1 ⊗ xm
= xm ⊗ 1 + xm−1 ⊗ 0 + xm−2 ⊗ 0 + · · · + 0 ⊗ xm−2 + 0 ⊗ xm−1 + 1 ⊗ xm
= xm ⊗ 1 + 1 ⊗ xm
= i1∗ (xm ) + i2∗ (xm )
for each m ≤ 2n; that is, xm is a digital primitive homology class as required. □
 DIGITAL (CO)HOMOLOGY MODULES AND DIGITAL PONTRYAGIN ALGEBRAS 17

Remark 4.12. Let (Y, y0 , kY , mY ) be a digital Hopf space with a digital multiplication mY :
Y × Y → Y . Then the homomorphism
mY ∗ : dH∗ (Y × Y ; R) → dH∗ (Y ; R)
between digital homology modules with coefficients in a commutative ring R with identity gives
an algebra structure on the digital homology dH∗ (Y ; R) which is called the digital Pontryagin
algebra on a digital Hopf space (Y, y0 ). Moreover, if the digital homology R-module dH∗ (Y ; R)
is free, then the diagonal map ∆ : Y → Y × Y induces a coalgebra structure on the digital
homology module dH∗ (Y ; R).

Recall that a graded R-algebra is a graded R-module A = {Am } together with a homomor-
phism of degree 0,
mA : A ⊗ A → A,
which is called the product of the graded R-algebra. Here, mA maps Am ⊗ An into Am+n for
all m and n. The graded R-algebra A is said to be associative if the following diagram
1⊗mA
A⊗A⊗A / A⊗A

mA ⊗1A mA
 mA 
A⊗A /Y

is commutative. Let S : A ⊗ A → A ⊗ A be the switching map. Then the graded R-algebra A

is said to be commutative if the following diagram
S / A⊗A
A ⊗ AM
MMM q
MMM qqqqq
mA MMMM qq mA
& xqqq
A
commutes with sign in the sense that
mA (a ⊗ b) = (−1)s·t mA (b ⊗ a),
where a and b are homogeneous elements of degree s and t, respectively.

Definition 4.13. ([27, 30]) A digital multiplication mY : Y × Y → Y on a digital Hopf space

(Y, y0 , mY , kY ) is said to be digital homotopy associative if the following diagram
mY ×1Y
Y ×Y ×Y / Y ×Y

1Y ×mY mY
 mY 
Y ×Y /Y

is digital homotopy commutative. A digital Hopf space (Y, y0 , mY , kY ) with a digital multipli-
cation mY : Y × Y → Y is said to be a digital homotopy associative Hopf space if mY is digital
homotopy associative.

Definition 4.14. ([27, 30]) Let SY ×Y : Y × Y → Y × Y be the switching function. A digital

multiplication mY : Y × Y → Y is said to be digital homotopy commutative if the following
18 DAE-WOONG LEE

diagram
SY ×Y
Y × YL / Y ×Y
LLL rr r
LLL rrr
L rrrmY
mY LLL
L& rx rr
Y
is commutative up to digital homotopy. A pointed digital Hopf space (Y, y0 , mY , kY ) with a
digital multiplication mY : Y × Y → Y is said to be a digital homotopy commutative Hopf
space if mY is digital homotopy commutative.

For example, see [27, Example 3.8]).

We note that if (Y, y0 , kY , mY ) be a digital Hopf space with a digital multiplication mY :
Y × Y → Y , then, by Theorem 3.10, mY induces a homomorphism
mY ∗ : dH∗ (Y × Y ; R) → dH∗ (Y ; R)
of digital homology modules over a commutative ring R with identity 1R .

Theorem 4.15. Let (Y, y0 , mY , kY ) be a digital homotopy associative and commutative Hopf
space with a digital multiplication mY : Y × Y → Y . Then the digital Pontryagin algebra
dH∗ (Y ; R) becomes a non-negatively graded associative and commutative R-algebra.

Proof. Let
µY = × : dHs (Y ; R) ⊗ dHs (Y ; R) → dHs (Y × Y ; R)
be the digital homology cross product. Then we can consider the composite
µY mY ∗
dHs (Y ; R) ⊗ dHs (Y ; R) / dHs (Y × Y ; R) / dHs (Y ; R) (4.2)

of R-module homomorphisms which we denote this composition by ‘·P ’ or ‘mˆY ∗ ’. Since the
digital multiplication mY : Y × Y → Y is digital homotopy associative, we can see that the
following diagram
mˆY ∗ ⊗1dHs (Y ;R)
dHs (Y ; R) ⊗ dHs (Y ; R) ⊗ dHs (Y ; R) / dHs (Y ; R) ⊗ dHs (Y ; R)

1dHs (Y ;R) ⊗mˆY ∗ mˆY ∗

 mˆY ∗ 
dHs (Y ; R) ⊗ dHs (Y ; R) / dHs (Y ; R)

commutes on digital homology R-modules. Moreover, the digital homotopy commutative mul-
tiplication mY : Y × Y → Y induces the following commutative diagram
S∗
dHs (Y ; R) ⊗ dHs (Y ; R) / dHs (Y ; R) ⊗ dHs (Y ; R)
TTTT jj
TTTT
TTTT jjj jjjj
mˆY∗ * jt jj mˆY ∗
dHs (Y ; R)
on digital homology R-modules, where S∗ is an R-module homomorphism induced by the
switching map S : Y × Y → Y × Y . Therefore, dH∗ (Y ; R) is a graded associative and
commutative R-algebra. □
 DIGITAL (CO)HOMOLOGY MODULES AND DIGITAL PONTRYAGIN ALGEBRAS 19

Recall that a coalgebra C over a ring R is an R-module together with maps φ : C → C ⊗ C

and ϵ : C → R such that the following diagrams
φ
C / C ⊗C
φ 1⊗φ
 φ⊗1C 
C ⊗C / C ⊗ C ⊗ C,

and
φ
C TTTT / C ⊗C
TTTT 1C
TTTTT
φ TTTT 1C ⊗ϵ
 T* 
ϵ⊗1C
C ⊗C / R⊗C ∼ =C∼ = C ⊗ R,
are commutative, where 1C : C → C is the identity automorphism. The module homomorphism
φ : C → C ⊗ C above is said to be an R-algebra comultiplication on C. It can be shown that
(Y, y0 , kY , mY ) is a digital Hopf space with digital multiplication mY : Y × Y → Y , then the
digital multiplication mY provides the digital cohomology module dH ∗ (Y ; R) with the structure
of coalgebra over the commutative ring R with identity 1R .

Theorem 4.16. Let f : (X, x0 ) → (Y, y0 ) be a digital Hopf function between digital Hopf
spaces. Then f∗ : dH∗ (X; R) → dH∗ (Y ; R) is an R-module homomorphism as the Pontryagin
algebras, and f ∗ : dH ∗ (Y ; R) → dH ∗ (X; R) is an R-module homomorphism of coalgebras.

Proof. We need to show that

f∗ (r[x]) = rf∗ ([x]),
f∗ ([x1 ] + [x2 ]) = f∗ ([x1 ]) + f∗ ([x2 ])
and
f∗ ([x1 ] ·P [x2 ]) = f∗ ([x1 ]) ·P f∗ ([x2 ])
for all r in R and [x1 ], [x2 ] in dH∗ (X; R). Since it is not difficult to show that the first and second
conditions are satisfied, we only check the final condition. From the the following commutative
diagram
µX mX ∗
dHs (X; R) ⊗ dHs (X; R) / dHs (X × X; R) / dHs (X; R)

f∗ ⊗f∗ (f ×f )∗ f∗
 µY  mY ∗ 
dHs (Y ; R) ⊗ dHs (Y ; R) / dHs (Y × Y ; R) / dHs (Y ; R)

of R-modules, we have
f∗ ([x1 ] ·P [x2 ]) = f∗ (mX ∗ ◦ µX ([x1 ] ⊗ [x2 ]))
= mY ∗ ◦ µY ◦ (f∗ ⊗ f∗ )([x1 ] ⊗ [x2 ])
= mY ∗ ◦ µY ◦ (f∗ ([x1 ]) ⊗ f∗ ([x2 ])
= f∗ ([x1 ]) ·P f∗ ([x2 ]).
And similarly, for the coalgebra case. □
20 DAE-WOONG LEE

A Hopf algebra over R is a graded R-algebra A which has also the coalgebra structure whose
R-algebra comultiplication φ : A → A ⊗ A is a homomorphism of graded R-algebras. The
graded Hopf algebras are often used in algebraic topology, and they have the natural algebraic
structure on the direct sum of all homology or cohomology modules of a Hopf space.

Remark 4.17. The relationship between the Pontryagin algebras and coalgebras of a digital
Hopf space is that of dual Hopf algebras. Indeed, if (Y, y0 , kY , mY ) is a digital Hopf space, and
dH(Y ; R) is a free graded R-module, then it can be shown that dH∗ (Y ; R) and dH ∗ (Y ; R) are
dual Hopf algebras.

5. C ONCLUSIONS
Hopf spaces and co-Hopf spaces play a pivotal role in algebraic topology especially in (equi-
variant) homotopy theory. Digital topology deals with the so-called discrete sets with a topology
in the sense of Euclidean topology considered to be digital sets or images of finite and bounded
subsets of the n-dimensional Euclidean space, and the digital process substitutes a bounded and
finite discrete set for a suitable object in some category. In this paper, we develop a consid-
eration of a digital counterpart of classical notions in algebraic topology as the so-called alge-
braic approach from the classical homology, cohomology and Pontryagin algebra. This paper is
concerned with setting up more algebraic invariants and their fundamental properties of digital
homology and cohomology modules over a commutative ring with identity for a digital image
with an adjacent relation which are based on the classical homology and cohomology groups of
topological spaces in algebraic topology.

R EFERENCES
[1] Arkowitz, M.; Curjel, C. R. On maps of H-spaces. Topology 1967, 6, 137–148.
[2] Arkowitz, M.; Lupton, G. Loop-theoretic properties of H-spaces. Math. Proc. Cambridge Philos. Soc. 1991,
110, 121–136.
[3] Arkowitz, M.; Lee, D.-W. Properties of comultiplications on a wedge of spheres. Topology Appl. 2010, 157,
1607–1621.
[4] Arkowitz, M.; Lee, D.-W. Comultiplications on a wedge of two spheres. Sci. China Math. 2011, 54, 9–22.
[5] Arslan, H.; Karaca, I,; Öztel, A. Homology groups of n-dimensinal images. Proceeding of XXI. Turkish Na-
tional Mathematics Symposium 2008, B1–13.
[6] Boxer, L. Digitally continuous functions. Pattern Recognition Letters 1994, 15, 833–839.
[7] Boxer, L. A classical construction for the digital fundamental group. J. Math. Imaging Vis. 1999, 10, 51–62.
[8] Boxer, L. Properties of digital homotopy. J. Math. Imaging Vis. 2005, 22, 19–26.
[9] Boxer, L. Homotopy properties of sphere-like digital images. J. Math. Imaging Vis. 2006, 24, 167–175.
[10] Boxer, L; Karaca, I.; Öztel A. Topological invariants in digital images. J. Math. Sci. Adv. Appl. 2011, 11 (2),
109–140.
[11] Eckhardt, U; Latecki, L. Digital topology. In Current Topics in Pattern Recognition Research, (Research
Trends), Council of Scientific Information, Vilayil Gardens, Kaithamukku, Trivandrum, India 1995.
[12] Ege, O.; Karaca, I. Digital H-spaces. 3rd International Symposium on Computing in Science and Engineering,
ISCSE 2013, 133–138.
[13] Ege, O.; Karaca, I. Digital homotopy fixed point theory. C. R. Acad. Sci. Paris, Ser. I 2015, 353, 1029–1033.
 DIGITAL (CO)HOMOLOGY MODULES AND DIGITAL PONTRYAGIN ALGEBRAS 21

[14] Ege, O.; Karaca, I. Some properties of digital H-spaces. Turkish Journal of Electrical Engineering and Com-
puter Sciences 2016, 24, 1930–1941.
[15] Ganea, T. Cogroups and suspensions. Invent. Math. 1970, 9, 185–197.
[16] Han, S.-E. Multiplicative property of the digital fundamental group. Acta Appl. Math. 2010, 110, 921–944.
[17] James, I. M. On H-spaces and their homotopy groups. Quart. J. Math. 1960, 11, 161–179.
[18] Khalimsky, E. Motion, deformation, and homotopy in finite spaces. Proceedings IEEE International Confer-
ence on Systems, Man, and Cybernetics 1987, 227–234.
[19] Lee, D.-W. Phantom maps and the Gray index. Topology Appl. 2004, 138, 265–275.
[20] Lee, D.-W. On the same n-type conjecture for the suspension of the infinite complex projective space. Proc.
Amer. Math. Soc. 2009, 137 (3), 1161–1168.
[21] Lee, D.-W. On the same n-type structure for the suspension of the Eilenberg-MacLane spaces. J. Pure Appl.
Algebra 2010, 214, 2027–2032.
[22] Lee, D.-W. On the same n-type of the suspension of the infinite quaternionic projective space. J. Pure Appl.
Algebra 2013, 217, 1325–1334.
[23] Lee, D.-W. On the generalized same N -type conjecture. Math. Proc. Camb. Phil. Soc. 2014, 157, 329–344.
[24] Lee, D.-W. On the same N -types for the wedges of the Eilenberg-MacLane spaces. Chin. Ann. Math. Ser. B
2016, 37 (6), 951–962.
[25] Lee, D.-W. Comultiplication structures for a wedge of spheres. Filomat 2016, 30 (13), 3525–3546.
[26] Lee, D.-W. On the digitally quasi comultiplications of digital images. Filomat 2017, 31(7), 1875–1892.
[27] Lee, D.-W. Near-rings on digital Hopf groups. Appl. Algebra Engrg. Comm. Comput. 2018, 29, 261–282.
[28] Lee, D.-W. Algebraic loop structures on algebra comultiplications. Open Math. 2019, 17 (1), 742–757.
[29] Lee, D.-W. Comultiplications on the localized spheres and Moore spaces. Mathematics 2020, 8 (1), 1–19.
[30] Lee, D.-W. Digital H-spaces and actions in the pointed digital homotopy category. To appear in Appl. Algebra
Engrg. Comm. Comput., https://doi.org/10.1007/s00200-019-00398-8.
[31] Lee, D.-W.; Lee, S. Homotopy comultiplications on the k-fold wedge of spheres. Topology Appl. 2019, 254,
145–170.
[32] McAndrew, A.; Osborne, C. Algebraic methods for multidimensional digital topology. Proc. SPIE 1993, 2060,
14–25.
[33] McAndrew, A.; Osborne, C. A survey of algebraic methods in digital topology, J. Math. Imaging Vis. 1996, 6,
139–159.
[34] Rosenfeld, A. Continuous functions on digital pictures. Pattern Recognition Letters 1986, 4, 177–184.
[35] Spanier, E. Algebraic Topology McGraw-Hill, New York, 1996.
[36] Whitehead, G. W. Elements of Homotopy Theory, Graduate Texts in Math 61, Springer-Verlag, New York,
Heidelberg, Berlin, 1978.

D EPARTMENT OF M ATHEMATICS , AND I NSTITUTE OF P URE AND A PPLIED M ATHEMATICS , C HONBUK NA -

TIONAL U NIVERSITY, 567 BAEKJE - DAERO , D EOKJIN - GU , J EONJU - SI , J EOLLABUK - DO 54896, R EPUBLIC OF
KOREA
E-mail address: dwlee@jbnu.ac.kr