Dear Mark and David,

I believe now that I can prove Mark's group automatic, by a combination
of computer calculation and reasoning. If my proof is correct, then it
quite an interesting example of such a combined approach.
I hope that at least one of you will check the theoretical part of
the argument for reasonableness!

The given presentation of the group was:

>xaxa=t
>bxbx=t
>bbtaa=t
>Abr=rAb
>yt=ty
>btay=ybta

I could not make much progress with it as it stood, but I found a few
substitutions made things easier.

Eliminating t, we get:

xaxa=bxbx
bbbxbxaa=bxbx
Abr=rAb
ybxbx=bxbxy
bbxbxay=ybbxbxa

Then putting u=xa and eliminating a (=Xu),
and v=bx and eliminating x (=Bv), we get

u^2 = v^2,
b^2v^2VbuVbu = v^2  which simplifies (using u^2=v^2) to bvbu = BuBv,
UBvbr = rUBvb,
yu^2 = u^2y
ybvbu = bvbuy.

Call this group G. I still could not do much with the complete presentation,
but leaving out the generator y and the last two relations, kbmag
completes fairly easily on the resulting group 
H = < u,b,v,r | u^2 = v^2, bvbu = BuBv, UBvbr = rUBvb>
showing that this is shortlex automatic.
(Word acceptor has 311 states and multipliers 1425 states.)

Now G is an HNN extension of the group H, where the new generator  y
centralises the subgroup  K = <u^2, bvbu> of H.
Let T be a right transversal of K in H.
Then by the theory of HNN extensions, each g in G has a unique expression

k t_1 y^(n_1) t_2 y^(n_2) ... t_r y^(n_r)

where k is in K, t_i in T and n_i are integers (with various nontriviality
conditions like all t_i except possibly t_1 are nontrivial and
all n_i except possibly n_r are nonzero).

I want to show that this normal form essentially forms the language for
an automatic structure for G.

I can show using kbmag that H is coset automatic with respect to the
subgroup K. This means that we can choose T (in fact we take the
shortlex least representatives of the cosets of K in H as members of T)
to be a regular language, such that the set of pairs (t_1,t_2) in T x T
with the property that Kt_1x = Kt_2, for some (monoid) generator x of H
is also a regular language.

In fact, by inspecting the automata involved in these calculations, it
can be seen that in all equations of the form
t_i x = k t_j, for t_i in T, k in K, and x a generator of H, or one of the
elements u^2,  bvbu, bvbU of K, the only elements  k  that arise are these
same three elements u^2,  bvbu, bvbU and their inverses.
(This property is specific to this particular example, but is essential
for the proof of automaticity of  G.)

A presentation of K can also be derived in a fairly easy way from this
automatic coset structure, and it turns out (not surprisingly) that K is
free on its two generators u^2 and bvbu. Thus, we can use the obvious
normal form based on this for the initial element  k  in the normal form

k t_1 y^(n_1) t_2 y^(n_2) ... t_r y^(n_r)

for  G.

Putting these facts together, it follows that in an equation of the form

k t_1 y^(n_1) t_2 y^(n_2) ... t_r y^(n_r) x =
l u_1 y^(m_1) u_2 y^(m_2) ... u_s y^(m_s)

in G, with x a monoid generator of G, t_i, u_i in T, and  k,l in K, the two
words 

k t_1 y^(n_1) t_2 y^(n_2) ... t_r y^(n_r) and
l u_1 y^(m_1) u_2 y^(m_2) ... u_s y^(m_s)

will fellow travel. This is obvious if x = y or Y, and otherwise  x  is
a generator of  H, and when we multiply the first word by  x, the 
t_i change into u_i, and one of the elements u^2,  bvbu, bvbU or their
inverses (which all commute with y) moves each time past the power of y 
and into the next position to the left.

Best wishes,
Derek.
From daemon Wed Jul 30 16:54:02 1997
Received: from mail.bna.bellsouth.net by ribble.maths.warwick.ac.uk; Wed, 30 Jul 1997 16:53:56 +0100 (BST)
Received: from dialin.inetnebr.com (host-32-96-72-76.bna.bellsouth.net [32.96.72.76])
	by mail.bna.bellsouth.net (8.8.5/8.8.5) with SMTP id LAA22825;
	Wed, 30 Jul 1997 11:53:48 -0400 (EDT)
Message-ID: <33DF6397.4964@bellsouth.net>
Date: Wed, 30 Jul 1997 10:53:59 -0500
From: Mark Sapir <msapir@bellsouth.net>
Reply-To: mvs@unlinfo.unl.edu
Organization: Vanderbilt
X-Mailer: Mozilla 3.01Gold (Win95; I)
MIME-Version: 1.0
To: Derek Holt <dfh@maths.warwick.ac.uk>
Subject: Re: What is this group?
References: <7799.199707291328@thames>
Content-Type: text/plain; charset=us-ascii
Content-Transfer-Encoding: 7bit
Status: RO
Content-Length: 2141

Dear Derek,

Sorry for a long delay. I was moving to Nashville (TN): I am now a
Professor at Vanderbilt. Now I am almost settled, only a few boxes are
still to be unpacked.

The group I sent to David was supposed to be a building block in the
construction of groups with given Dehn functions. In order to construct
such a group one needed to 
take a Turing machine and then using iseas similar to those of Boone,
Collins and others, interpret it in a group. The 
goal was to get a group whose Dehn function is not much bigger than the
time function of the original Turing machine. The problem with this idea
is that Boone and others used relations like x^a=x^2 where x^a=a^-1xa
(Baumslag-Solitar relations). These
relations make the Dehn function exponential, so there is no hope we'll
get a subexponential Dehn functions as a result. So at some point I
thought that it would be possible to replace the Baumslag-Solitar
relations by some other set of relations which would do the same trick
as B-S relations but would not cause exponential blow up of the Dehn
function. The main property (for my purpose) of the B-S-relation is that
x^(a^n)=x^k for some k if and only if n is positive. I tried to get
another set of relations with 
similar property. This is how I came up with the set of relations that I
sent to David. I thought that I proved that the property I needed is
true and I wanted to make sure that the Dehn function of this group is
polynomial. 
But after I sent this group to David and received your confirmation that
the group is automatic, I discovered that the group does not satisfy my
property. Later I abandoned 
my  attempts to replace B-S relations and found a completely new
approach to the problem (the paper has been submitted to a journal, you
can download it from 
http://www.math.unl.edu/~msapir, it is big, about 100 pages). Thus this
group turned out not to be very useful. But the fact that you proved
that it is automatic saved me some time: insteads of proving that the
Dehn function of this group is polynomial, I could concentrate on "my
property". This is the history of this group. 

Best regards,

Mark

