home *** CD-ROM | disk | FTP | other *** search
- Newsgroups: sci.math
- Path: sparky!uunet!noc.near.net!black.clarku.edu!black.clarku.edu!djoyce
- From: djoyce@black.clarku.edu (Dave Joyce)
- Subject: Re: A word problem
- Message-ID: <djoyce.724714724@black.clarku.edu>
- Organization: Clark University (Worcester, MA)
- References: <1992Dec12.162349.29729@dcs.qmw.ac.uk> <djoyce.724444576@black.clarku.edu> <1992Dec15.221213.6274@dcs.qmw.ac.uk>
- Date: 18 Dec 92 21:38:44 GMT
- Lines: 54
-
- In <1992Dec15.221213.6274@dcs.qmw.ac.uk> arodgers@dcs.qmw.ac.uk (Angus H Rodgers) writes:
- >
- >I had some amazingly prompt responses to my article, on the day it was
- >posted, at the weekend. My thanks go especially to Mark "V." Sapir and
- >Nicholas Reingold, whose virtually instantaneous and simultaneous replies
- >supplied all the information I needed, and convinced me that sci.math
- >has some previously unsuspected oracular qualities. I'll summarise the
- >replies when I have time (probably after Christmas ...). Meanwhile, I'll
- >just say that the answer to my question is known, and can be found in:
- >
- > Lallement, Gerard
- > Semigroups and combinatorial applications / [by] Gerard Lallement.
- > New York : Wiley, 1979. - xiii,376p : ill ; 24cm. - (Pure and
- > applied mathematics)
- > 'A Wiley-Interscience publication'. - Text on lining paper. - Bibl.:
- > p.357-368. - Index. - 0-471-04379-6
- >
- >(At least, I'm told it can; but a computer error has prevented me from
- >borrowing the book and making sure. Aargh! I hate computers!)
- >
- >I had also better admit, with some embarrassment, that my presentation
- >of the problem was confused, in that I used the term "reduced words" to
- >refer to strings over the alphabet {a,b,c} which are (in the correct
- >terminology) "squarefree", i.e. have no nonempty substrings of the form
- >xx; *but* distinct squarefree strings may nevertheless denote equal elements
- >in all idempotent semigroups -- because the reduction rule xx = x can be
- >applied in both directions. (It never even entered my head that the
- >backwards application could be needed.) There are, indeed, recipes for
- >constructing infinite squarefree strings over a 3-character alphabet, but
- >it does not follow from this that infinite 3-generated idempotent
- >semigroups exist, and I'm sorry that I gave the misleading impression that
- >there was such an implication.
-
- I haven't yet seen the book, but after a bit of work, I found that the free
- idempotent semigroup on three generators {a,b,c} has 159 elements. I was
- surprized that it wasn't infinite. The number 159 may look strange, but it is
- divisible by 3. Every element in the semigroup starts with one of the three
- generators (no matter which string is used to represent it), so there are 53
- elements starting with any particular generator, say a. One of those elements
- is a itself. That leaves 52, which is divisible by 2. 26 of those begin with
- ab, the other 26 begin with ac.
-
- There are some interesting identities which hold in idempotent semigroups. My
- colleague down the hall, John Kennison, discovered the identity xyz=xyzyxyz.
- There are other identities similar to that. If u and v are any three-letter
- permutations of abc, and if x is any string over {a,b,c}, then uv=uxv. For
- instance, (acb)(cab)=(acb)(abcb)(cab).
-
- I don't know whether the idempotent semigroup on 4 generators is finite.
-
- --
- David E. Joyce Dept. Math. & Comp. Sci.
- Internet: djoyce@black.clarku.edu Clark University
- BITnet: djoyce@clarku Worcester, MA 01610-1477
-