1 Scope of this note
The Praeger–Xu graphs, introduced by Praeger and Xu in [Reference Praeger and Xu2], have exponentially large groups of automorphisms, with respect to the number of vertices. This fact causes various complications with regard to many natural questions.
In their recent work [Reference Jajcay, Potočnik and Wilson1], Jajcay et al. gave a sufficient and necessary condition for a Praeger–Xu graph to be a Cayley graph. Explicitly, [Reference Jajcay, Potočnik and Wilson1, Theorem 1.1] states that, for any positive integer
$n\geq 3$
,
$n\ne 4$
, and for any positive integer
$k\leq n-1$
, the Praeger–Xu graph
$\textrm {PX}(n,k)$
is a Cayley graph if and only if one of the following holds:
-
(i) the polynomial
$t^{n}+1$ has a divisor of degree
$n-k$ in
$\mathbb {Z}_{2}[t]$ ;
-
(ii) n is even, and there exist polynomials
$f_{1},f_{2},g_{1},g_{2},u,v\in \mathbb {Z}_{2}[t]$ such that
$u,v$ are palindromic of degree
$n-k$ , and
(1.1)$$ \begin{align} t^{n}+1 = f_{1}(t^{2})u(t) + tg_{1}(t^{2})v(t) = f_{2}(t^{2})v(t) + tg_{2}(t^{2})u(t). \end{align} $$
Our aim here is to prove that (ii) implies (i), thus obtaining the following refinement. (It can be verified that
$\textrm {PX}(4,1)$
,
$\textrm {PX}(4,2)$
and
$\textrm {PX}(4,3)$
are Cayley graphs.)
Theorem 1.1. For any positive integer
$n\geq 3$
and for any positive integer
$k\leq n-1$
, the Praeger–Xu graph
$\mathrm{PX}(n,k)$
is a Cayley graph if and only if the polynomial
$t^{n}+1$
has a divisor of degree
$n-k$
in
$\mathbb {Z}_{2}[t]$
.
Using the factorisation of
$t^{n}+1$
in
$\mathbb {Z}_{2}[t]$
, we give a purely arithmetic condition for the Cayleyness of
$\textrm {PX}(n,k)$
. Let
$\varphi $
be the Euler
$\varphi $
-function and, for every positive integer d, let
![](https://static.cambridge.org/binary/version/id/urn:cambridge.org:id:binary:20230128160005443-0196:S0004972722000454:S0004972722000454_eqnu1.png?pub-status=live)
be the multiplicative order of
$2$
modulo d.
Corollary 1.2. Let a be a nonnegative integer, let b be an odd positive integer, let
$n:=2^{a}b$
with
$n\geq 3$
and let k be a positive integer with
$k\leq n-1$
. The Praeger–Xu graph
$\mathrm{PX}(n,k)$
is a Cayley graph if and only if k can be written as
![](https://static.cambridge.org/binary/version/id/urn:cambridge.org:id:binary:20230128160005443-0196:S0004972722000454:S0004972722000454_eqn2.png?pub-status=live)
2 Proof of Theorem 1.1
Suppose (ii) holds. We aim to show that
$t^{n}+1$
is divisible by a polynomial of degree
$n-k$
in
$\mathbb {Z}_{2}[t]$
, implying (i). Working in characteristic
$2$
, (1.1) can be written as
![](https://static.cambridge.org/binary/version/id/urn:cambridge.org:id:binary:20230128160005443-0196:S0004972722000454:S0004972722000454_eqnu2.png?pub-status=live)
in short,
![](https://static.cambridge.org/binary/version/id/urn:cambridge.org:id:binary:20230128160005443-0196:S0004972722000454:S0004972722000454_eqn3.png?pub-status=live)
If
$g_{1}=0$
or if
$g_{2}=0$
, then the result follows from (2.1), and the fact that u and v have degree
$n-k$
. Therefore, for the rest of the argument, we may suppose that
$g_{1},g_{2}\ne 0$
. Moreover, observe that
$f_{1},f_{2}\ne 0$
, because t does not divide
$t^{n}+1$
.
We introduce four polynomials
$u_{e},u_{o},v_{e},v_{o} \in \mathbb {Z}_{2}[t]$
such that
![](https://static.cambridge.org/binary/version/id/urn:cambridge.org:id:binary:20230128160005443-0196:S0004972722000454:S0004972722000454_eqnu3.png?pub-status=live)
Substituting these expansions for u and v in (2.1),
![](https://static.cambridge.org/binary/version/id/urn:cambridge.org:id:binary:20230128160005443-0196:S0004972722000454:S0004972722000454_eqnu4.png?pub-status=live)
Recall that n is even. By splitting the equations into even and odd degree terms, we obtain
![](https://static.cambridge.org/binary/version/id/urn:cambridge.org:id:binary:20230128160005443-0196:S0004972722000454:S0004972722000454_eqnu5.png?pub-status=live)
Set
$m:=n/2$
. Since we are working in characteristic
$2$
,
![](https://static.cambridge.org/binary/version/id/urn:cambridge.org:id:binary:20230128160005443-0196:S0004972722000454:S0004972722000454_eqn4.png?pub-status=live)
![](https://static.cambridge.org/binary/version/id/urn:cambridge.org:id:binary:20230128160005443-0196:S0004972722000454:S0004972722000454_eqn5.png?pub-status=live)
Since u and v are palindromic by assumption, we get
$1=u(0) =u_{e}(0)$
and
${1=v(0) =v_{e}(0)}$
. In particular, both
$u_{e}$
and
$v_{e}$
are not zero. From (2.2) and (2.3),
![](https://static.cambridge.org/binary/version/id/urn:cambridge.org:id:binary:20230128160005443-0196:S0004972722000454:S0004972722000454_eqn6.png?pub-status=live)
Our candidate for the desired divisor of
$t^{n} + 1$
is
$s:=u_{e}v_{e}+tu_{o}v_{o}$
. Let us show first that
$\deg (s)=n - k$
. Since
$u_{e}v_{e}$
and
$u_{o}v_{o}$
have even degree, we deduce
![](https://static.cambridge.org/binary/version/id/urn:cambridge.org:id:binary:20230128160005443-0196:S0004972722000454:S0004972722000454_eqnu6.png?pub-status=live)
Recall
$u=u_{e}^{2}+tu_{o}^{2}$
and
$v=v_{e}^{2}+tv_{o}^{2}$
. If
$n-k$
is even, then
![](https://static.cambridge.org/binary/version/id/urn:cambridge.org:id:binary:20230128160005443-0196:S0004972722000454:S0004972722000454_eqnu7.png?pub-status=live)
However, if
$n-k$
is odd, then
![](https://static.cambridge.org/binary/version/id/urn:cambridge.org:id:binary:20230128160005443-0196:S0004972722000454:S0004972722000454_eqnu8.png?pub-status=live)
Therefore, in both cases,
$\deg (s) = n-k$
.
It remains to prove that s divides
$t^{n}+1$
. Since
$f_{1},g_{1},f_{2},g_{2}$
are polynomials, by (2.4), s divides
![](https://static.cambridge.org/binary/version/id/urn:cambridge.org:id:binary:20230128160005443-0196:S0004972722000454:S0004972722000454_eqnu9.png?pub-status=live)
Observe that
$\gcd (v_{e},v_{o},u_{e},u_{o})$
divides
$f_{1}u_{e}+tg_{1}v_{o}$
, and hence, in view of the first equation in (2.2),
$\gcd (v_{e},v_{o},u_{e},u_{o})$
divides
$t^{m}+1$
. Therefore, s divides
${(t^{m}+1)^{2}=t^{n}+1}$
.
3 Proof of Corollary 1.2
By Theorem 1.1, deciding if a Praeger–Xu graph
$\textrm {PX}(n,k)$
is a Cayley graph is tantamount to deciding if
$t^{n}+1$
admits a divisor of order k in
$\mathbb {Z}_{2}[t]$
. An immediate way to proceed is to study how
$t^{n}+1$
can be factorised in irreducible polynomials.
Let
$n=2^{a}b$
, with
$\gcd (2,b)=1$
. Since we are in characteristic
$2$
,
![](https://static.cambridge.org/binary/version/id/urn:cambridge.org:id:binary:20230128160005443-0196:S0004972722000454:S0004972722000454_eqnu10.png?pub-status=live)
Furthermore, if
$\lambda _{d}(t)\in \mathbb {Z}[t]$
denotes the dth cyclotomic polynomial, then
![](https://static.cambridge.org/binary/version/id/urn:cambridge.org:id:binary:20230128160005443-0196:S0004972722000454:S0004972722000454_eqnu11.png?pub-status=live)
is the factorisation of
$t^{b}+1$
in irreducible polynomials over
$\mathbb {Q}[t]$
, by Gauss’ theorem. Since the Galois group of any field extension of
$\mathbb {Z}_{2}$
is a cyclic group generated by the Frobenius automorphism, the degree of an irreducible factor of
$\lambda _{d}(t)$
in
$\mathbb {Z}_{2}[t]$
is the smallest c such that a dth primitive root
$\zeta $
raised to the power
$2^{c}$
is
$\zeta $
, that is,
$\omega (d)$
. Hence,
$\lambda _{d}(t)$
in
$\mathbb {Z}_{2}[t]$
factorises into
$\varphi (d)/\omega (d)$
irreducible polynomials, each having degree
$\omega (d)$
.
Therefore,
$t^{n}+1\in \mathbb {Z}_{2}[t]$
has a divisor of degree k if and only if k can be written as the sum of some
$\omega (d)$
terms, each summand repeated at most
$2^{a}\varphi (d)/\omega (d)$
times, which is exactly (1.2).