Past exam of the mathematics course of the University of Cambridge / 2008 / iii / Paper 2 / 1

The identified involution is $z=g^2=h^5$. It is central: it commutes with $g$ because it is a power of $g$, and with $h$ because it is a power of $h$. We use the normal form theorem for an amalgamated free product: the two factors embed, an alternating product of elements outside the amalgamated subgroup is nonidentity, and fixed coset representatives give a unique reduced expression. Here choose representatives $1,g$ in the first cyclic group and $1,h,h^2,h^3,h^4$ in the second.

An effective expression from the normal form theorem for an amalgamated free product iswhere each $x_j$ is a nonidentity representative and successive representatives come from different factors. Reduce powers by $g^u=z^{\lfloor u/2\rfloor }{g}^{u\mathrm{mod}2}$ and $h^v=z^{\lfloor v/5\rfloor }{h}^{v\mathrm{mod}5}$, using these identities for negative integers too. Move the central $z$ factors left and reduce their total exponent modulo two. Whenever adjacent representatives come from one factor, multiply them, split off their $z$ contribution, and delete a resulting identity. Each such combination reduces the number of representatives, so the procedure terminates. By the normal form theorem for an amalgamated free product, the input is trivial exactly when $\boxed{r=0\text{ and }\epsilon =0}$. In particular $z\ne 1$ because each factor embeds. This gives the soluble word problem in finite cyclic amalgams algorithm without merely asserting that a normal form exists.

In the specified quotient, the extra relator makes $g,h$ commute. Put $t=g{h}^{-2}$. Its abelian group presentation givesConversely $g\mapsto s^5$, $h\mapsto s^2$ satisfy every quotient relator in $C_{20}=⟨s\mid s^{20}=1⟩$, and send $t$ to $s$. Thus the quotient is exactly $C_{20}$, rather than just a cyclic group of order dividing twenty. If $M,N$ are the exponent sums of $g,h$ in the input, its quotient word problem is

Use $c=[a,b]=a^{-1}{b}^{-1}ab$. The two defining relators say that $c$ commutes with $a,b$, hence is central. The given group commutator identities giveConsequently a word collection in a two-step nilpotent group algorithm maintains an integer triple $(m,n,k)$, initially $(0,0,0)$. Reading $a^{ϵ}$, with $ϵ=\pm 1$, changes it to $(m+ϵ,n,k-nϵ)$; reading $b^{ϵ}$ changes it to $(m,n+ϵ,k)$. Each step uses the displayed multiplication rule and reads one letter, so it terminates and produces the requested expression.

By the stipulated uniqueness, the word is trivial exactly when $\boxed{(m,n,k)=(0,0,0)}$, which solves its word problem for a group. Moreover the commutator subgroup is $⟨c⟩$: the quotient by $⟨c⟩$ is abelian, and $c$ itself is a commutator. It is central, so the nilpotency class is at most two. Uniqueness also implies $c\ne 1$, since $(0,0,1)$ and $(0,0,0)$ are distinct triples. Hence the exact nilpotency class is $\boxed{2}$.