GENERALIZATIONS
OF
SCHEIN
THEOREM
*KUNITAKA SHOJI DEPARTMENT
OFMATHEMATICS
SHIMANE UNIVERSITY
In the paper B.M. Schein
proved
that anyfintecyclic subsemigroup
of the fulltrans‐formation
semigroup
T(X)
on a finite set X is coveredby
an inversesubsemigroup
ofT(X)
. In this paper,westudy
ageneralization
of Schein theorem.1
Schein theorem
Let X be a finite set and
Tx
the full transformationsemigroup
on X withcomposition
being
from theright
to the left.For any
a\in Tx
, We say thata(X)
isthe rangeofa and\displaystyle \bigcap_{k=0}^{\infty}a^{k}(X)
iscalled the stablerangeofa and denoted
by
SR(a)
.(1)
Define thedepth
\mathrm{d}(x)
ofxby
aninteger
k such thatx\in a^{k}(X)
,butx\not\in(X)a^{k+1}
forx\in X-SR(a)
.\mathrm{d}(x)=\infty
ifx\in SR(a)
.(2)
Define theheight
ofxby
the leastnon‐negative integer
k such that\mathrm{d}(a^{k}(x))>k+\mathrm{d}(x)
if\mathrm{d}(x)<\infty
. Denoteitby
\mathrm{h}(x)
.Further,
\mathrm{h}(x)=\infty
ifx\in SR(a)
.Givena
partial
map b ofa(X)
toX such that forx\in a(X)
,b(x)\in a^{-1}(x)
and\mathrm{d}(b(x))\geq
\mathrm{d}(y)
forally\in a^{-1}(x)
.(3)
Define thegap\mathrm{g}(x)
asthe greatestinteger
k such thatb^{k}a^{k}(x)=x.
(4)
Define the reach\mathrm{r}(x)
by
\mathrm{r}(x)=\mathrm{g}(x)+1.
Thus,
\mathrm{h}(x)\geq \mathrm{r}(x)>\mathrm{g}(x)
. Hence(b^{h}a^{h})(x)\neq x
and(b^{r}a^{r})(x)\neq x.
Schein theorem Foran element
a\in Tx,
there exists an element
b\in Tx
such thaty=bx(x,y\in X)
if
andonly if
(1)
if
d(x)>0(x\in X)
theny\in a^{-1}(x)
has amaximaldepth.
(2) d(x)=0(x\in X)
thena^{h}(x)=a^{h+1}(y)
and(b^{r}a^{r})(ab)(x)=(ab)(b^{r}a^{r})(x)
, whereh=h(x)
,r-\neg(x),g(y)\geq r.
In this case, the
subsemigroup
<a, b>is an inversesemigroup
generated by
a and b inTx.
*
This isanabsrtact and thepaperwillappearelsewhere.
数理解析研究所講究録
Acutally,
takey=b^{h+1}a^{h}(x)
the coIt ispossible
because of definition of h(that
is,
\mathrm{d}(a^{h}(x))>h+\mathrm{d}(x))
.Remark, By
L.M. Gluskintheorem[l],
inversesemigroups
<a, b> and<a',
b'> areisomorphic
to each other if<a> and <a'>isomorphic.
In[2]\mathrm{T}.\mathrm{E}
. Hall showed thatSchein theorem is
applicable
toamalgamation problem.
2
Generalizations
of
Schein
theorem
Wepose
generalizations
of Schein theorem fromcyclic semigroups
toextensionsofcyclic
semigroups
by
groups inT(X)
and severalquestionsasfollows :Let a
semigroup
S=G\displaystyle \cup\bigcup_{i=1}^{n}Ga^{i}
, where G is thegroupof units in S and Ga=aG.Question
1Suppose
that S is asubsemigroup
ofT(X)
. The does there exist anelemenet
b\in T(X)
such that<S,
b>\mathrm{i}\mathrm{s} aninversesemigroup?
(the
firstgenralization
of Scheintheorem)
Let X be a finite set and
a\in T(X)
such that there exists distinct x,y\in X-a(X)
witha(x)=a(y)
. Letg\in T(X)
such thatg(x)=y, g(y)=x
andg(z)=z
for allz\in X-\{x, y\}.
Let
b\in T(X)
such thatbx=b^{h(x)+1}a^{h(x)}(x)
Then ga=ag =a andbg=b
(since
h(x)=h(y))
butgb\neq b
. Bothgb
and bare aninverseelement ofa.<S,
b>\mathrm{i}\mathrm{s} a left inversesemigroup.
Question
1 has anegative
answer.Leta
semigroup
S=G\displaystyle \cup\bigcup_{i=1}^{n}Ga^{i}
,where Gisthegroupofunits in S and Ga=aG.Question
2Suppose
that S is asubsemigroup
ofT(X)
. The does there exist anelemenet
b\in T(X)
such that<S,
b> isan orthodoxsemigroup?
(That
is,
isthe set ofidempotents
in aregular
semigroup <S,
b>\mathrm{a}subsemigroup?)
(the
secondgenralization
of Scheintheorem)
TheanswerofQuestion
2would benegative.
Let a
semigroupS
=G\displaystyle \cup\bigcup_{i=1}^{n}Ga^{i}
,where Gisthe groupofunits in S andga=agfor any a\in G.When
S\subseteq T(X)
, X is adirectedgraph
withedges
labelledby
a. Forx\in \mathrm{S}\mathrm{R}(X)
, thesubgraph
Gr(a;x)=\{y\in X|ya^{k}=x, \mathrm{d}(x)<\infty\}
is a tree, thatis,
a directedgraph
without
cycle.
Eachg\in G
indeucesanautomorphism
ofadirectedgraph
X.Question
3 Let asemigroupS
=G\displaystyle \cup\bigcup_{i=1}^{n}Ga^{i}
, where Gisthe group ofunits inS and ga=agfor anya\in G.Suppose
that S isasubsemigroup
ofT(X)
. The does there existanelemenetb\in T(X)
such that