Notes on subsemigroups of Nm
0(+)
BARBORA BATÍKOVÁ
Faculty of Engineering
Czech University of Life Sciences
Kamýcká 129 165 21 Praha 6 - Suchdol
CZECH REPUBLIC
ANTONÍN JANČAŘÍK
Faculty of Education
Charles University
M. Rettigove 4 116 39 Praha 1
CZECH REPUBLIC
PETR NĚMEC
Faculty of Engineering
Czech University of Life Sciences
Kamýcká 129 165 21 Praha 6 - Suchdol
CZECH REPUBLIC
TOMÁŠ KEPKA
Faculty of Education
Charles University
M. Rettigove 4 116 39 Praha 1
CZECH REPUBLIC
Abstract: The paper summarizes and extends the knowledge of various subsemigroups of Nm
0(+) (=N0(+)m).
It creates a theoretical basis for further study in this area and applications in other areas, such as the investigation
of context-free languages. The last chapter introduces the notion of pure subsemigroups and presents one con-
struction of a pure subsemigroup to a chosen semisubgroup of Nm
0(+).
Key-Words: integer, semigroup, pure, Nm
0(+), subsemigroups, pure subsemigroups
Introduction
This note presents some results on various subsemi-
groups of Nm
0(+) (=N0(+)m). The results show-
cased in this article may encompass both new find-
ings and components of previously published results.
However, consolidating them into a singular output is
a theoretical basis for further research.
The motivation for investigating such problems
stems from several parts of mathematics. Holes in
subsemigroups of Nm
0(+) (if Ais a subsemigroup
of Nm
0(+) then an element hof the pure subsemi-
group generated by Ais said to be a hole in Aif
h/A) were applied to transportation problems in
[1]. There also is a close connection with the the-
ory of semirings. For example, the problem whether
every commutative parasemifield (i.e., an algebraic
structure with two commutative and associative bi-
nary operations such that the multiplication is a group
and distributes over addition) finitely generated as a
semiring is additively idempotent was in [2] affirma-
tively answered in 2-generated case by transferring
the problem to subsemigroups of N0m(+) with special
properties; this method was further developed in [3],
[4]. It also seems that the investigation of finitely
generated cones could be useful in the investigation
of context-free languages (see e.g. [5], [6], [7]).
2 Preliminaries (a)
The following notation will be used throughout the
note:
Received: June 5, 2023. Revised: September 4, 2023. Accepted: October 2, 2023. Published: October 20, 2023.
WSEAS TRANSACTIONS on MATHEMATICS
DOI: 10.37394/23206.2023.22.85
Barbora Batíková, Antonín Jančařík, Petr Němec, Tomáš Kepka
E-ISSN: 2224-2880
773
Volume 22, 2023
N... the semiring of positive integers;
N0... the semiring of non-negative integers;
Z... the ring of integers;
Q+... the parasemifield of positive rationals;
Q+
0... the semifield of non-negative rationals;
R+
0... the semifield of non-negative reals.
Throughout this section, let mNand let Abe a
subset of Qm(the ring of ordered m-tuples of ratio-
nals, where the operations of addition and multipli-
cation are defined componentwise), where we denote
0 = (0, . . . , 0) Qm(no confusion can arise). We
put
conv(A) =
=Pn
i=1 qiai
nN, aiA, qiQ+,Pqi= 1
and cone(A) = Pn
i=1qiai
nN,aiA,qiQ+
0.
Lemma 1.1. (i) Aconv(A).
(ii) conv(conv(A)) = conv(A).
(iii) If A(Q+
0)mthen conv(A)(Q+
0)mand,
moreover, 0conv(A)if and only if 0A.
(iv) If A(Q+)mthen conv(A)(Q+)mand
0 /conv(A).
Proof. The assertions (i), (iii) and (iv) are quite easy.
As concerns (ii), take aconv(conv(A)). Then
a=Pn
i=1 qiai,nN,aiconv(A),qiQ+and
Piqi= 1. Furthermore, ai=Pki
j=1 qij aij ,kiN,
aij A,qij Q+and Pjqij = 1, and hence
a=PiPjqiqij aij . However, PiPjqiqij =
Pi(qiPjqij ) = Piqi= 1. Thus aconv(A).
Lemma 1.2. (i) Aconv(A)cone(A) =
conv(B), where B={qa |aA, q Q+
0}.
(ii) If A6=then 0cone(A).
(iii) cone(cone(A)) = conv(cone(A)) =
cone(conv(A)) = cone(A).
(iv) If A(Q+
0)mthen cone(A)(Q+
0)m.
Proof. It is easy.
Lemma 1.3. cone(A) = Aif and only if conv(A) =
Aand qA Afor every qQ+
0.
Proof. It is easy.
Lemma 1.4. If A6=then cone(A)is just the sub-
semimodule of the Q+
0-semimodule Qmgenerated by
the set A.
Proof. It is easy.
Lemma 1.5. cone(A) = Aif and only if either A=
or Ais a subsemimodule of the Q+
0-semimodule Qm.
Proof. See 1.3 and 1.4.
Lemma 1.6. Let nN,nm+2,q1, . . . , qnQ+
0,
a1, . . . , anQmand a=Pqiai. Then there are
r1, . . . , rnQ+
0such that a=Priaiand ri1= 0
for at least one i1,1i1n.
Proof. Clearly, there are s1, . . . , sn1Qsuch that
Pn1
i=1 si(aian) = 0 and si06= 0 for at least one
i0,1i0n1. Then Pn
i=1 siai= 0, where
sn=s1 · · · sn1, and we have Psi= 0.
Consequently, the set J={j|sj>0}is a non-
empty subset of the set {1, . . . , n}and we find j0
Jsuch that t=qj0/sj0qj/sjfor every jJ.
Clearly, tQ+
0. Now, put ri=qitsifor all
i= 1, . . . , n. Then riQ+
0and rj0= 0. Finally,
Priai=a.
Lemma 1.7. cone(A) = Scone(B),BA,|B|
m+ 1.
Proof. Let a=Pn
i=1 qiai,nN,qiQ+
0,aiA.
If nis the smallest possible number with this property
then nm+ 1 by 1.6. The rest is clear.
Lemma 1.8. cone(A) = {Pm+1
i=1qiai|aiA,qi
Q+
0}.
Proof. See 1.7.
Lemma 1.9. Assume that Ais a subsemigroup of
Qm(+) that is generated by a non-empty set B. Then
cone(A) = cone(B).
Proof. Since BA, we have cone(B)cone(A).
On the other hand, cone(B)is a subsemigroup of
Qm(+) by 1.4, and hence Acone(B). Conse-
quently, cone(A)cone(cone(B)) = cone(B)by
1.2(iii).
Lemma 1.10. Assume that Ais a subsemigroup of
Qm(+). Then:
(i) For every acone(A),a6= 0, there is kNwith
ka A(conv(A)).
(ii) Either 0 /Aand cone(A) = {0} SkNA/kor
0Aand cone(A) = SkNA/k.
Proof. We have a=Pn
i=1 qiai,nN,aiA,
qiQ+. If kNis such that kqiNfor every i
then ka =PkqiaiA.
Lemma 1.11. If Ais a subsemigroup of Qm(+) then
cone(A) = {0} SkNconv(A/k).
Proof. See 1.10.
Lemma 1.12. Assume that Ais a subsemimodule of
the Q+-semimodule Qm. Then:
(i) If 0Athen conv(A) = cone(A) = A.
(ii) If 0 /Athen conv(A) = Aand cone(A) = A
{0}.
WSEAS TRANSACTIONS on MATHEMATICS
DOI: 10.37394/23206.2023.22.85
Barbora Batíková, Antonín Jančařík, Petr Němec, Tomáš Kepka
E-ISSN: 2224-2880
774
Volume 22, 2023
Proof. It is easy.
Lemma 1.13. If cone(A) = Qmthen 0conv(A).
Proof. There are nN,aiAand qi, riQ+
0
such that (1,0, . . . , 0) = Pqiai,(1,0, . . . , 0) =
Priai,q=PqiQ+and r=PriQ+. Now,
0 = P(qi+ri)ai, and so Pqi+ri
q+rai= 0 as well. But
Pqi+ri
q+r= 1 and it follows easily that 0conv(A).
Lemma 1.14. Let Abe a subsemigroup of Qm(+)
and let aA,bQmand k, l N0,k+l1,
be such that kb la A(A {0}, resp.). Then
(k1)a+kb A(A {0}, resp.).
Proof. We have (k1)a+kb = (kb la)+(k+
l1)aA.
Lemma 1.15. Let a1, . . . , anconv(A),nN.
Then (a1+· · · +an)/nconv(A).
Proof. It is easy.
3Preliminaries (b)
For a= (q1, . . . , qn)Qm, we put ||a|| =pPq2
i
R+
0.
Remark 2.1. Let ε={e1, . . . , em}be a basis of
the vector Q-space Qm. For every aQmthere
are uniquely determined πi(a)Qsuch that a=
Pπi(a)ei. Then πi:QmQare projections of
the vector Q-spaces and these projections are continu-
ous. Consequently, for every qQ+there is tQ+
such that |πi(a)|< q for all i= 1, . . . , m, whenever
aQmis such that ||a|| < t.
Lemma 2.2. Let a0, . . . , amQmbe such that
the elements a1a0, . . . , ama0are linearly Q
independent (then they form a basis of Qm). Take
q0, . . . , qmQ+such that Pm
j=0 qj= 1 and put
a=Pjqjaj. Then there is tQ+such that
a+bconv({a0, . . . , am})for every bQmwith
||b|| < t.
Proof. Put q=min({qj/m|j= 0,1, . . . , m })
Q+. By 2.1, where ε={a1a0, a2a0, . . . , am
a0}, there is tQ+such that |πi(b)|< q for all
i= 1, . . . , m, whenever bQmis such that ||b|| <
t. Moreover, if we put π0(b) = Pm
i=1 πi(b)then
b=Pm
j=0 πj(b)aj,|π0(b)|< q,Pm
j=0 πj(b)=0,
Pm
j=0(qi+πj(b)) = 1 and qj+πj(b)>0for every
j= 0,1, . . . , m. Thus a+b=Pm
j=0(qj+πj(b))ai
conv({a0, . . . , am}).
Lemma 2.3. Let a1, . . . , amQmbe linearly in-
dependent elements (then they form a basis of Qm).
Take q1, . . . , qmQ+and put a=Pm
i=1 qiai. Then
there is tQ+such that a+bcone({a1, . . . , am})
for every bQmwith ||b|| < t.
Proof. First, find rQ+such that rq < 1, where
q=Pm
i=1 qiQ+, and put a0
0= 0,a0
i=ai/rand
q0
i=rqifor i= 1, . . . , m. If q0
0= 1 rq then q0
j
Q+for every j= 0,1, . . . , m,Pm
j=0 q0
j= 1 and a=
Pm
j=0 q0
ja0
j. Now, by 2.2, there is tQ+such that
a+bconv({a0
0,a0
1, . . . , a0
m})cone({a1, . . . , am})
for every bQmwith ||b|| < t.
Lemma 2.4. Let a1, . . . , amQmbe linearly in-
dependent and let a=Pm
i=1 qiai,qiQ+. Then,
for every bQm, there is nNwith na +b
cone({a1, . . . , an}).
Proof. By 2.3, there is tQ+such that a+
ccone({a1, . . . , am})whenever ||c|| < t. Now,
||b/n|| < t for some nN, and hence a+
b/ncone({a1, . . . , am}). Then na +b
cone({a1, . . . , am}).
Lemma 2.5. Let a1, . . . , amQmbe linearly inde-
pendent and let a=Pm
i=1 qiai,qiQ+. Then, for
every bQm, there are r, s Q+with ra +sb
conv({a1, . . . , am}).
Proof. By 2.4, na+bcone({a1, . . . , am})for some
nN. Our result is clear for na +b6= 0. If
na +b= 0 6=athen 06= (n+ 1)a+b
cone({a1, . . . , am}). Finally, if a=0=na +b
then b= 0,0conv({a1, . . . , am})and we can put
r= 1 = s.
4Preliminaries (c)
In this section, the structure of these subsets is de-
scribed in more detail. Attention is paid to the case
when Ais a subsemigroup of Qm(+).
Let Abe a subset of Qm,mN. We put
α(A) = {aQm|(bQm)(r, s Q+)
ra +sb /cone(A)}, β(A) = α(A)cone(A),
γ(A) = α(A)conv(A)and δ(A) = α(A)A.
Lemma 3.1. The following conditions are equivalent
for aQm:
(i) aα(A).
(ii) There is at least one cZmsuch that ra +sc /
cone(A)for all r, s Q+.
(iii) There is at least one dQmsuch that ra +sd /
conv(A)for all r, s Q+.
(iv) There is at least one eZmsuch that ra +se /
conv(A)for all r, s Q+.
WSEAS TRANSACTIONS on MATHEMATICS
DOI: 10.37394/23206.2023.22.85
Barbora Batíková, Antonín Jančařík, Petr Němec, Tomáš Kepka
E-ISSN: 2224-2880
775
Volume 22, 2023
Proof. Clearly, the conditions (i) and (ii) are equiva-
lent, the conditions (iii) and (iv) are equivalent and (i)
implies (iii). It remains to show that (iii) implies (i).
Assume first that ra+sd cone(A)and ra+sd 6= 0.
Then there are nN,aiAand qiQ+such
that ra +sd =Pn
i=1 qiaiand Pqi=qQ+.
Now, (r/q)a+ (s/q)d=P(qi/q)ai,Pqi/q= 1
and (r/q)a+(s/q)dconv(A), a contradiction with
(iii). Consequently, if ra +sd 6= 0 for all r, s Q+
then ra +sb /cone(A)and aα(A).
Next, assume that r1a+s1d= 0 for some r1, s1
Q+. Then 0 /conv(A),d= (r1/s1)aand ra +
sd = (r((r1s/s1))a/conv(A)for all r, s Q+.
If acone(A),a6= 0, then r2aconv(A)for suit-
able r2Q+and, setting r=r2+ (r1s/s1), we get
ra +sd =r2aconv(A), a contradiction. It follows
that a(Qm\cone(A)) {0}.
If a/cone(A)then, setting b= 0, we get ra +
sb =ra /cone(A)for all r, s Q+and aα(A).
Finally, if a= 0 then, choosing bQm\cone(A)
(see 1.13), we get ra +sb =sb /cone(A)for all
r, s Q+. Again, we obtain aα(A).
Lemma 3.2. (i) qα(A)α(A)and qβ(A)β(A)
for every qQ+.
(ii) For every aβ(A),a6= 0, there is rQ+with
ra γ(A).
Proof. It is easy.
Lemma 3.3. The following conditions are equiva-
lent:
(i) 0α(A).
(ii) 0β(A).
(iii) β(A)6=.
(iv) α(A)6=.
(v) cone(A)6=Qm.
Proof. Clearly, (i) implies (ii), (ii) implies (iii), (iii)
implies (iv) and (iv) implies (v). If b/cone(A)then
r0 + sb =sb /cone(A)for all r, s Q+. Thus
0α(A)and (v) implies (i).
Lemma 3.4. If A(Q+)mthen 0β(A).
Proof. We have cone(A)(Q+
0)mand 0β(A)by
3.3.
Lemma 3.5. α(A) = (Qm\cone(A)) β(A).
Proof. If a/cone(A)then ra /cone(A)for every
rQ+, and hence aα(A).
Lemma 3.6. Let a1, . . . , amQmbe linearly in-
dependent and let a=Pm
i=1 qiai,qiQ+. Then
a/α(A).
Proof. See 2.4 (or 2.5).
Lemma 3.7. (i) α(cone(A)) = α(A).
(ii) β(cone(A)) = γ(cone(A)) = δ(cone(A)) =
β(A).
Proof. The equalities follow easily from the defini-
tions of the sets involved and from 1.2(iii).
Lemma 3.8. Assume that Ais a subsemigroup of
Qm(+). Then for every aβ(A),a6= 0, there is
kNwith ka δ(A).
Proof. We have a=Piqiaifor some aiA,qi
Q+,i= 1, . . . , n,nN. Choosing kNsuch that
kqiNfor every i, we get ka A. But ka β(A)
by 3.2(i), and hence ka δ(A).
Corollary 3.9. If Ais a subsemigroup of Qm(+)
then β(A) = SkNδ(A)/kor β(A) = {0}
SkNδ(A)/k.
Lemma 3.10. Assume that Ais a subset of (Q+
0)m.
Then Qm\(Q+
0)mα(A).
Proof. We have cone(A)(Q+
0)m, and hence Qm\
(Q+
0)mα(A)by 3.5. On the other hand, if a=
(q1. . . , qm)(Q+
0)m\(Q+)mthen qi= 0 for at least
one iand, setting b= (0, . . . , 0,1,0, . . . , 0) where
1is the i-th coordinate, we have ra +sb /(Q+
0)m
for all r, s Q+. Thus aα(A).
Lemma 3.11. Let a1, . . . , amQmbe linearly in-
dependent and let a=Pm
i=1 qiai,qiQ+. Then
a/α({a1, . . . , am}).
Proof. See 2.5.
Lemma 3.12. The following conditions are equiva-
lent for aQm:
(i) a/α(A).
(ii) For every bQmthere are k, l Nwith ka +
lb cone(A).
Moreover, if Ais a subsemigroup of Qm(+) with 0
Athen these two conditions are equivalent to
(iii) For every bQmthere are k1, l1Nwith
k1a+l1bA.
Proof. Clearly, (iii) implies (ii) and (ii) implies (i). If
a/α(A)then ra+sb cone(A)for some r, s Q+
and we have tra +tsb cone(A), where tNis
such that tr, ts N. Finally, if Ais a subsemigroup
of Qm(+) and 0Athen (ii) implies (iii) by 1.10(i).
Lemma 3.13. If bQm\α(A),aAand qQ+
0
then qa +b/α(A).
WSEAS TRANSACTIONS on MATHEMATICS
DOI: 10.37394/23206.2023.22.85
Barbora Batíková, Antonín Jančařík, Petr Němec, Tomáš Kepka
E-ISSN: 2224-2880
776
Volume 22, 2023
Proof. The result follows easily from 3.12(ii).
Lemma 3.14. Assume that Ais a subsemigroup of
Qm(+). Then either δ(A) = Aor A\α(A)is a
subsemigroup of Qm(+).
Proof. If δ(A)6=Athen A*α(A)and B=A\
α(A)6=. Now, the fact that Bis a subsemigroup of
Qm(+) follows from 3.13.
Lemma 3.15. Assume that Ais a subsemigroup of
Qm(+). Then, for all aAand bQm\α(A),
there is nNwith (n1)a+nb A.
Proof. By 3.12(iii), k1bl1aA {0}for some
k1, l1N(notice that cone(A) = cone(A{0})) and
α(A) = α(A {0}). Now, by 1.14, (n1)a+nb
A {0}for some nN. If (n1)a+nb = 0 then
(2n1)a+ 2nb =aA.
5Subsemigroups of Nm
0(+) (a)
In this section, let Abe a subsemigroup of Nm
0(+).
We denote by ε(A)the set of the elements aAsuch
that baA, whenever bAand baNm
0.
Equivalently, ε(A) = {aA|A+a= (Nm
0+a)
A}.
Lemma 4.1. The following conditions are equiva-
lent:
(i) 0A.
(ii) 0ε(A).
(iii) ε(A)6=.
Proof. Clearly, (i) implies (ii) and (ii) implies (iii). It
remains to show that (iii) implies (i). If aε(A)then
a=a+ 0, and so a(Nm
0+a)A=A+a. Then
a=b+afor some bAand, of course, b= 0.
Lemma 4.2. Either 0 /Aand ε(A) = or 0
ε(A)Aand ε(A)is a subsemigroup of A.
Proof. It is easy (use 4.1).
In the rest of the section, we will assume that 0
A(see 4.1 and 4.2).
Lemma 4.3. (i) If b, a ε(A)are such that ba
Nm
0then baε(A).
(ii) (ε(A)ε(A)) Nm
0=ε(A).
(iii) ε(ε(A)) = ε(A).
Proof. (i) Since aε(A)and bA, we have ba
A. Now, if cAis such that c(ba) = (c+a)b
Nm
0then c+aAand c(ba)A, since bε(A).
Thus baε(A).
(ii) and (iii). Use (i).
Lemma 4.4. ε(A)ε(A)Zmand ε(A)ε(A)is
just the difference (sub)group of the (sub)semigroup
ε(A).
Proof. It is easy.
Lemma 4.5. If A+aε(A)for at least one aNm
0
then ε(A) = A.
Proof. First, a= 0 + aa+Aε(A), and so
aε(A). Now, A= ((A+a)a)Nm
0(ε(A)
ε(A)) Nm
0=ε(A)by 4.3(ii). Thus ε(A) = A.
Corollary 4.6. If ε(A)6=Athen for every aNm
0
there is at least one bAsuch that a+b/ε(A).
For every aA, put ϕ(A, a) = {bNm
0|a+
bA},ψ(A, a) = {cNm
0|(n1)a+nc
Afor some nN}and ξ(A, a) = ϕ(A, a)
ψ(A, a).
Lemma 4.7. (i) A {0} ξ(A, a).
(ii) Nm
0\α(A)ψ(A, a).
(iii) Nm
0Qaψ(A, a).
Proof. (i) It is obvious.
(ii) This is ensured by 3.15.
(iii) We can assume that a6= 0. Take bNm
0Qa,
b6= 0. We have b=ra/bfor some rZ,sN,
and hence sb =ra and we conclude easily that r
N. Now, sb ra = 0 and (s1)a+sb A(see
1.14).
Lemma 4.8. If aε(A)then both ϕ(A, a)and
ξ(A, a)are subsemigroups of Nm
0(+).
Proof. First, if a+b1Aand a+b2Afor some
b1, b2Nm
0then 2a+b1+b2Aand a+b1+
b2Nm
0. Since aε(A), we get a+b1+b2A.
Similarly, if c1, c2ξ(A, a)then (n1)a+nciA
for some nN,n3,(2n2)a+n(c1+c2)A
and, since aε(A), we conclude that (n1)a+
n(c1+c2)A.
6Subsemigroups of Nm
0(+) (b)
Define a relation on Nm
0by abif and only if
baNm
0. Clearly, the relation is reflexive, anti-
symmetric and transitive, and hence it is an ordering.
This ordering is stable under both addition and multi-
plication. Notice that it satisfies the descending chain
condition, but not the ascending one.
Lemma 5.1. There is no infinite set of pair-wise in-
comparable m-tuples in Nm
0.
Proof. The assertion is clear for m= 1 and we pro-
ceed by induction for m2.
Let Abe a set of pair-wise incomparable elements
from Nm
0. For kN0, put Bk={bNm1
0|a
WSEAS TRANSACTIONS on MATHEMATICS
DOI: 10.37394/23206.2023.22.85
Barbora Batíková, Antonín Jančařík, Petr Němec, Tomáš Kepka
E-ISSN: 2224-2880
777
Volume 22, 2023
(b, k)for at least one aA}. Then B0B1
B2. . . and all the sets Bkare filters of the ordered
set Nm1
0. Then B=SBkis a filter and we denote
by Cthe set of minimal elements of B. The elements
of Care pair-wise incomparable, and hence Cis a
finite set by induction. Consequently, CBk0for
some k0and it follows that Bk0=Bk0+1 =· · · =
B. Now, if a= (b1, k1)A,b1Nm1
0,k1
N0, then b1Bk1B=Bk0, and therefore a1
(b1, k0)for some a1A. If k0k1then a1
a, and hence a1=aand k1=k0. It means that
k1k0anyway. Furthermore, if Cidenotes the set
of minimal elements from Bi,i= 0,1, . . . , k0, then
Ciis a finite set (and Ck0=C). Since b1Bk1, there
is c1Ck1with c1b1and (c1, k1)a. According
to the definition of Bk1, we can find a2Asuch that
a2(c1, k1). Then a2a, and therefore a2=aand
b1=c1. We have proved that ASk0
i=0(Ci× {i})
and it follows immediately that Ais finite.
Example 5.2. Put m= 2. Then An=
{(0, n),(1, n 1), . . . , (n1,1),(n, 0)}(N2
0),
nN0, is a set of incomparable elements and |An|=
n+ 1.
Proposition 5.3. Let A1be a non-empty finite subset
of Nm
0and let Abe a subsemigroup of Nm
0(+) such
that A1Acone(A1). Then Ais a finitely gener-
ated semigroup.
Proof. Let A1={a1, . . . , an},nN, and B=
{Pn
i=1 qiai|qiQ+
0, qi1} Nm
0. Clearly, A1
Acone(A1)Nm
0and Bis a finite subset of Nm
0.
If aAthen a=Priai,riQ+
0, and we have
ri=li+sifor suitable liN0and siQ+
0,si1.
From this, a=b1+Pliai, where b1=Psiai.
Since aNm
0and PliaiNm
0, we get b1B.
Moreover, either Pliai= 0 or PliaiA.
For every bB, put Nb={(k1, . . . , kn)
Nn
0|b+PkiaiA}. Then Nbis a filter of the or-
dered set Nm
0(see 5.1 and its proof) and the set Mbof
minimal elements from Nbis finite. Of course, Nb=
{dNm
0|cdfor some cMb}and the set
C={a1, . . . , an} SbB{b+kiai|(k1, . . . , kn)
Mb}is a finite subset of A. Taking into account the
preceding steps, one sees easily that the semigroup A
is generated by the finite subset C.
Theorem 5.4. The following conditions are equiva-
lent for a subsemigroup Aof Nm
0(+):
(i) Ais a finitely generated semigroup.
(ii) cone(A) = cone(A1)for a non-empty finite sub-
set A1of A.
(iii) cone(A) = cone(A2)for a non-empty finite sub-
set A2of (Q+
0)m.
Proof. (i) implies (ii) by 1.9, (ii) implies (iii) trivially
and (ii) implies (i) by 5.3. It remains to show that (iii)
implies (ii).
Since A2is a finite subset of cone(A), there is
a finite subset A1of Asuch that A2cone(A1).
Then cone(A) = cone(A2)cone(cone(A1)) =
cone(A1)cone(A). Thus cone(A) =
cone(A1).
Corollary 5.5. Let Abe a finitely generated subsemi-
group of Nm
0(+). Then every subsemigroup A0of
Nm
0(+) such that AA0cone(A)is finitely gen-
erated. In particular, cone(A)Nm
0is a finitely gen-
erated subsemigroup of Nm
0(+).
Remark 5.6. It is easy to see that 5.4 remains true for
subsemigroups of (Q+
0)m(+). Indeed, let Bbe such a
subsemigroup and assume that cone(B) = cone(B1)
for a non-empty finite subset B1of B. We have B1=
{k1/l1, . . . , kn/ln}for some nN,kiN0and
liN. If l=l1· · · lnthen A=lB is a subsemigroup
of Nm
0(+) and cone(A) = cone(B) = cone(B1) =
cone(A1), where A1=lB1. By 5.4, Ais a finitely
generated semigroup and the same is true for B, since
the mapping a7→ a/l,aA, is an isomorphism of
Aonto B.
7Pure subsemigroups of Nm
0(+)
In this section, let Abe a subsemigroup of Nm
0(+).
Lemma 6.1. The following conditions are equiva-
lent:
(i) nA =AnNm
0for every nN.
(ii) If aNm
0and nNare such that na Athen
aA.
(iii) If aNm
0and qQ+are such that qa A
then aA.
Proof. It is easy.
If these equivalent conditions are satisfied then A
is called pure subsemigroup of Nm
0(+). In the re-
maining part of this section(except for 6.12), we will
assume that Ais a pure subsemigroup.
Lemma 6.2. A {0}is a pure subsemigroup of
Nm
0(+).
Proof. It is easy.
Lemma 6.3. If 0Athen ε(A)is a pure subsemi-
group of Nm
0(+).
WSEAS TRANSACTIONS on MATHEMATICS
DOI: 10.37394/23206.2023.22.85
Barbora Batíková, Antonín Jančařík, Petr Němec, Tomáš Kepka
E-ISSN: 2224-2880
778
Volume 22, 2023
Proof. By 4.2, ε(A)is a subsemigroup of A. Let a=
nb, where nN,aε(A)and bNm
0. Since Ais
pure, we have bA. Moreover, if cAis such that
cbNm
0then nc a=n(cb)Nm
0, and so
n(cb)A, since aε(A). Using again the fact
that Ais pure, we get cbA. Thus bε(A)and
ε(A)is pure.
Lemma 6.4. Let nN,a1, . . . , anAand
q1, . . . , qnQ+be such that a=PqiaiNm
0.
Then aA.
Proof. We have qi=ri/sifor some ri, siN. If
s=s1· · · snthen sqiN,bi=sqiaiAand
sa =PbiA. Now, aAby 6.1.
Lemma 6.5. If aNm
0and a/α(A)then aA.
Proof. By 3.3, a6= 0. If bAthen there are
r, s Q+with ra sb cone(A)and we get ra
cone(A) + sb cone(A). Thus acone(A)Nm
0
and aAby 6.4.
Proposition 6.6. (i) cone(A)Nm
0=A {0}.
(ii) If 0Athen A=cone(A)Nm
0.
(iii) If cone(A) = (Q+
0)mthen A {0}=Nm
0.
(iv) α(A)Nm
0= (Nm
0\A)δ(A).
(v) Nm
0= (α(A)Nm
0)A.
(vi) ξ(A, a) = ψ(A, a)for every aA.
Proof. (i), (ii) and (iii). Use 6.4.
(iv) and (v). Use 6.5.
(vi) If (n1)a+nc Athen n(a+c)Aand
a+cA. Thus ψ(A, a)ϕ(A, a)and ξ(A, a) =
ψ(A, a).
Remark 6.7. Let kNand a, b1, . . . , bkNm
0be
such that a+biAfor every i= 1, . . . , k (e.g.,
aAand biϕ(A, a)). Furthermore, assume that
there are niNsuch that (ni1)a+nibiAfor
every i(e.g., aAand biψ(A, a)). If nN,
nmax(ni)then (n1)a+nbiAfor all i.
In particular, if tiN0are such that t=Pti
max(ni)then (t1)a+tbiAand (t1)a+
Ptibi=P((ti/t)((t1)a+tbi),(t1)a+Ptibi
Nm
0. Now, by 6.4, we get (t1)a+PtibiA.
Lemma 6.8. If aAand bNm
0are such that
a+b/α(A)then a+bAand bξ(A, a).
Proof. Since a+b/α(A),a+bAby 6.6(v) and
there are r, s Q+such that r(a+b)sa cone(A).
We have r=k/n,s=l/nfor suitable k, l, n N
and c/n=r(a+b)sa cone(A), where c= (k
l)a+kb. Then cZmcone(A) = Zm(Q+
0)m
cone(A) = Nm
0cone(A) = A {0}(see 6.6(i)), so
that cA∪{0}and d= (k1)a+kb =c+(l1)a
A {0}. If d6= 0 then dAand bξ(A, a)(see
6.6(vi)). If d= 0 then b= 0 ξ(A, a).
Lemma 6.9. Let aAbe such that bψ(A, a)for
every bNm
0with a+bδ(A)(then bα(A)).
Then ϕ(A, a) = ψ(A, a) = ξ(A, a).
Proof. We have ψ(A, a) = ξ(A, a)ϕ(A, a)by
6.6(vi). If bϕ(A, a)then a+bAand if a+b/
α(A)then bξ(A, a)by 6.8. On the other hand, if
a+bα(A)then a+bα(A)A=δ(A)and
bψ(A, a)by assumption.
Lemma 6.10. Let aAbe such that δ(A)Qa.
Then ϕ(A, a) = ψ(A, a) = ξ(A, a).
Proof. In view of 6.9, if bNm
0is such that a+b
δ(A)then a+bQa,bQaand bψ(A, a)by
4.7(iii).
Lemma 6.11. The following conditions are equiva-
lent for all aAand bNm
0:
(i) bψ(A, a)(bξ(A, a), resp.).
(ii) r(a+b)sa =Pk
i=1 qiaifor some r, s Q,
kN,aiAand qiQ+.
Proof. See the proof of 6.8.
Construction 6.12. Let Abe a subsemigroup of
Nm
0(+). Put p(A) = Nm
0SnNA/n. It is
easy to check that p(A)is a pure subsemigroup of
Nm
0(+). It is the smallest pure subsemigroup con-
taining A. Clearly, Ap(A)cone(A), and so
cone(p(A)) = cone(A). Now, according to 5.4, p(A)
is finitely generated if and only if Ais so. Finally, no-
tice that Nm
0SnNε(A)/nε(p(A)). In partic-
ular, ε(A)ε(p(A)), and so ε(p(A)) 6=, provided
that ε(A)6=.
8Conclusion
In this paper, the properties of subsemigroups and
pure subsemigroups of Nm
0(+) (=N0(+)m) are in-
vestigated. A theoretical and reference basis for fur-
ther research in this area has been established. It
can be used for further research of finitely gener-
ated cones, e.g. in connection with the investigation
of context-free languages. Our further research will
be directed to a deeper description of pure subsemi-
groups of Nm
0for a finite m, in particular m= 2.
References:
[1] R. Hemmecke, A. Takemura, R. Yoshida, Com-
puting holes in semigroups and its application to
transportation problems, Contribution to Discrete
Mathematics America Vol. 1, No. 4, 2009
[2] J. Ježek, V. Kala, T. Kepka, Finitely generated
algebraic structures with various divisibility con-
ditions, Forum Mathematicum, Vol. 24, 2012, pp.
379–397.
WSEAS TRANSACTIONS on MATHEMATICS
DOI: 10.37394/23206.2023.22.85
Barbora Batíková, Antonín Jančařík, Petr Němec, Tomáš Kepka
E-ISSN: 2224-2880
779
Volume 22, 2023
[3] J. Ježek, T. Kepka, Finitely generated commu-
tative division semirings, Acta Univ. Carolinae
Math. Phys., Vol. 51, 2010, pp. 3–27.
[4] V. Kala, T. Kepka, M. Korbelář, Notes on com-
mutative parasemifields, Comment. Math. Univ.
Carolinae. Vol. 50, 2009, pp. 321–533.
[5] J. Beauquier, M. Blattner, M. Latteux, On com-
mutative context/free languages, J. Comput. Sys-
tems Sci. 35 (1987), pp. 311–320.
[6] M. Latteux, Cônes rationnels commutatifs, J.
Comput. Systems Sci.. Vol. 18, 1979, pp.
307–333.
[7] T. Kepka, J. Kortelainen, M. Latteux, On
face-structure of convex polyhedrons induced
by simple commutative context-free languages,
(preprint)
Contribution of individual authors to
the creation of a scientific article
(ghostwriting policy)
B. Batíková formal analysis and investigation.
A. Jančařík validation and writing - review &
editing.
P. Němec investigation, writing - original draft,
validation.
T. Kepka supervision, validation and methodology.
Sources of Funding for Research Presented in a
Scientific Article or Scientific Article Itself
No funding was received for conducting this study.
Conflict of Interest
The authors have no conflicts of interest to declare
that are relevant to the content of this article.
Creative Commons Attribution License 4.0
(Attribution 4.0 International, CC BY 4.0)
This article is published under the terms of the
Creative Commons Attribution License 4.0
https://creativecommons.org/licenses/by/4.0/deed.en
_US
WSEAS TRANSACTIONS on MATHEMATICS
DOI: 10.37394/23206.2023.22.85
Barbora Batíková, Antonín Jančařík, Petr Němec, Tomáš Kepka
E-ISSN: 2224-2880
780
Volume 22, 2023