2
$\begingroup$

I am trying to see if the group

$G = \langle\ a,b,c,d\ \mid\ a^3 = b^3=c^3=d^3=$

$\qquad =(ab)^4= (ab^{-1})^4= (bc)^4= (bc^{-1})^4= (cd)^4= (cd^{-1})^4= (da)^4= (da^{-1})^4\ \rangle$

is trivial or not. I have tried all methods and tools I could but have not succeeded.

Knuth -- Bendix in Magma does not finish:

F <a,b,c,d> := FreeGroup(4);

L3  := [ w^3 : w in [ a, b, c, d ] ];
L4  := [ w^4 : w in [ a*b, a*b^-1, b*c, b*c^-1, c*d, c*d^-1, d*a, d*a^-1 ] ];

G := quo< F | L3 cat L4 >;

GR := RWSGroup(G : MaxRelations := 500000, TidyInt:=1000 );
// Warning: Knuth Bendix only partly succeeded

Determining the index of a large subgroup is inconclusive:

H := sub < G | a, b, c, (c*d)^2 >; 
Index(G,H:CosetLimit:=100000000,Hard:=true);
// 0  

GAP fails as well:

gap> F:=FreeGroup("a","b","c","d");;
gap> AssignGeneratorVariables(F);
#I  Assigned the global variables [ a, b, c, d ]
gap> 
gap> L3  := List ( [ a, b, c, d ], w -> w^3 );;
gap> L4  := List ( [ a*b, a*b^-1, b*c, b*c^-1, c*d, c*d^-1, d*a, d*a^-1 ], w -> w^4 );;
gap> 
gap> R:=Concatenation(L3,L4);;
gap> G:=F/R;
<fp group on the generators [ a, b, c, d ]>
gap> 
gap> gg := GeneratorsOfGroup(G);; 
gap> 
gap> LoadPackage("kbmag");
───────────────────────────────────────────────────────────────────
Loading  kbmag 1.5.4 (Knuth-Bendix on Monoids and Automatic Groups)
by Derek Holt (http://homepages.warwick.ac.uk/staff/D.F.Holt/).
Homepage: https://gap-packages.github.io/kbmag
───────────────────────────────────────────────────────────────────
gap> R := KBMAGRewritingSystem(G);;
gap> 
gap> H := Subgroup(G,gg{[1,2,3]});; # < a, b, c >
gap> HR := SubgroupOfKBMAGRewritingSystem(R,H);;
gap> A := AutomaticStructureOnCosets(R,HR); 
#Knuth-Bendix program failed or was inconclusive. Giving up.
false

Are there any other methods I could try?

$\endgroup$
5
  • $\begingroup$ Stupid remark. First I assume the abelianization is trivial, since otherwise the answer would trivially be no and I expect you did the computation. Then writing $z=a^3$, we have $G=1$ iff $G/\langle z\rangle=1$ (since $z$ is central and $G$ perfect). So $z=1$ can be added to the input. $\endgroup$
    – YCor
    Apr 26, 2018 at 14:43
  • $\begingroup$ (Not familiar with gap but it seems that $z=1$ is actually assumed. It's missing to the presentation of $G$: you say that plenty of elements are equal, but you mean they're all equal to 1.) $\endgroup$
    – YCor
    Apr 26, 2018 at 14:44
  • 3
    $\begingroup$ It seems that the group has a lot of index 8 subgroups, at least LowIndexSubgroupsFpGroup(G,8) returns a fairly long list. (4449 elements) $\endgroup$ Apr 26, 2018 at 14:57
  • 2
    $\begingroup$ Also many of the subgroups of low index have infinite abelianization. This is a very standard method of proving computationally that finitely presented groups are infinite. $\endgroup$
    – Derek Holt
    Apr 26, 2018 at 16:33
  • $\begingroup$ @Matthias thanks. I did not realize there was a built in method for finding low index subgroups. Very convenient. $\endgroup$
    – Anvita
    Apr 26, 2018 at 23:00

1 Answer 1

7
$\begingroup$

It's nontrivial, indeed infinite.

$G\neq 1$; moreover it is an amalgam (actually a double) over a subgroup of index $\ge 3$ (thus is infinite and contains nonabelian free subgroups).

Indeed, write $$H=H(t,u)=\langle t,u\mid t^3,u^3,(tu)^4,(t^{-1}u)^4\rangle.$$

Conversely, let us first show that $H\neq 1$ implies $G\neq 1$. Note that $H\neq 1$ means that $t,u$ have order 3. Define $$L=L(t,u,v)=\langle t,u,v\mid t^3,u^3,v^3,(tu)^4,(t^{-1}u)^4,(uv)^4,(uv^{-1})^4\rangle.$$

Then $L$ is the amalgam of $H(t,u)$ and $H(u,v)$ over the cyclic subgroup of order 3 $\langle u\rangle$ (which by the way has index $\ge 3$, since $H$ is perfect and nontrivial and hence of cardinal $\ge 60$).

Let $M$ be the subgroup of $L(t,u,v)$ generated by $t$ and $v$. Consider the double $D$ of $L=L(t,u,v)$ over $M$, that is, the amalgam $L\ast_M L$. Then it has the presentation

$$D=\langle t,u,v,t_2,u_2,v_2\mid t^3,u^3,v^3,(tu)^4,(t^{-1}u)^4,(uv)^4,(uv^{-1})^4,t_2^3,u_2^3,v_2^3,(t_2u_2)^4,(t_2^{-1}u_2)^4,(u_2v_2)^4,(u_2v_2^{-1})^4,t=t_2,v=v_2\rangle.$$ Eliminate $v_2(=v)$ and $t_2(=t)$ and write $w=u_2$. This yields $$D=\langle t,u,v,w\mid t^3,u^3,v^3,(tu)^4,(t^{-1}u)^4,(uv)^4,(uv^{-1})^4,w^3,(tw)^4,(t^{-1}w)^4,(wv)^4,(wv^{-1})^4\rangle,$$ where we precisely recognize $G$. Hence $G\neq 1$.

Now let us check that $G$ is a nontrivial amalgam- actually double- (over subgroups of index $\ge 3$). Indeed, let me discuss. If $M$ has index $\ge 3$, it's OK. $M$ can't have index 2 because $L$ is perfect. If $M=1$ (sounds unlikely but I haven't checked), then $L=M$ itself was described as an amalgam, actually a double.

Finally, to check that $H\neq 1$, just observe that in the symmetric group $\mathfrak{S}_7$, $t=(123)(456)$ and $u=(234)(576)$ satisfies the given relators of $H$.

$\endgroup$
3
  • $\begingroup$ Note: I found the proof that $H\neq 1$ after Matthias Wendt's comment that there is a subgroup of index 8 (so that I knew I should find in $\mathfrak{S}_8$ a pair of elements of order 3 with the given relations; there is not much choice). $\endgroup$
    – YCor
    Apr 26, 2018 at 15:25
  • $\begingroup$ PS: actually $M\neq L$: consider the given explicit elements $t,u$ of $\mathfrak{S}_7$; map $(t,u,v)$ to $(t,u,t)$ (sorry for the double meaning of $u$. This maps $L$ to $\mathfrak{S}_7$ and maps $M$ to a cyclic subgroup of order 3. So $M$ has index $\ge 20$ in $L$ (actually larger, since $H$ has no nontrivial homomorphism into $\mathrm{Alt}_5$). $\endgroup$
    – YCor
    Apr 26, 2018 at 15:33
  • $\begingroup$ PPS: the amalgam argument is similar to the argument of nontriviality of Higman's group, which similarly has a "cyclic" presentation on 4 generators and only relators involving two consecutive generators. $\endgroup$
    – YCor
    Apr 26, 2018 at 15:35

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.