Also available at http://amc.imfm.si
ISSN 1855-3966 (printed edn.), ISSN 1855-3974 (electronic edn.)
ARS MATHEMATICA CONTEMPORANEA 5 (2012) 307–323
Factors of disconnected graphs and polynomials
with nonnegative integer coefficients
Christiaan E. van de Woestijne ∗
Lehrstuhl für Mathematik und Statistik, Montanuniversität Leoben
Franz-Josef-Straße 18, 8700 Leoben, Austria
Received 23 February 2011, accepted 20 May 2011, published online 10 April 2012
Abstract
We investigate the uniqueness of factorisation of possibly disconnected finite graphs
with respect to the Cartesian, the strong and the direct product. It is proved that if a graph
has n connected components, where n is prime, or n = 1, 4, 8, 9, and satisfies some addi-
tional natural conditions, it factors uniquely under the given products. If, on the contrary,
n = 6 or 10, all cases of nonunique factorisation are described precisely.
Keywords: Graphs, monoids, factorisation.
Math. Subj. Class.: 05C70,13A05,20M13
1 Introduction
The purpose of this note is to give an algebraic/combinatoric perspective on a result in [8]
about uniqueness of factorisation of certain finite or infinite graphs with respect to well-
known graph products. As already noted by the authors of [8], the structure underlying
the problem is the factorisation of multivariate polynomials with nonnegative integer co-
efficients. We expand on this theme, and provide precise classifications of all possible
non-unique factorisations for a number of small cases, which include polynomials of un-
bounded degree, however. These results immediately translate back to graphs, and as our
main result, we have the following (definitions are given below).
Theorem 1.1. LetG be a finite graph having n connected components, where either n = 1,
4, 8, or 9, or n is a prime number. Then if G is simple, it has unique Cartesian and strong
factorisation; if G has no bipartite components, it has unique direct factorisation. If, on
the contrary, n = 6, 10, 12, or 16, non-unique factorisations occur for all three products.
∗The research leading to this paper was supported by the Austrian Science Foundation FWF, project S9611,
which is part of the Austrian National Research Network “Analytic Combinatorics and Probabilistic Number
Theory.”
E-mail address: c.vandewoestijne@unileoben.ac.at (Christiaan E. van de Woestijne)
Copyright c© 2012 DMFA Slovenije
308 Ars Math. Contemp. 5 (2012) 307–323
Proof. By the Structure Theorem 2.7, together with Lemma 3.6, we reduce the problem
to the unique or non-unique factorisation of univariate polynomials P with nonnegative
integer coefficients such that P (1) = n. This problem is trivial for n = 1, easy for
n prime (see Lemma 3.2), and solvable with the aid of a computer for the cases n ∈
{4, 6, 8, 9, 10} (see Theorems 3.7 up to 3.11); for n = 12 and 16, we have the examples
(3.2) and (3.10).
The results for n prime, or n = 4, are given as Exercises 6.12 and 6.13 in [8]. As
solutions, the authors provide purely graph-theoretical proofs, so that we obtain a second
proof for unique factorisation of polynomials P with nonnegative integer coefficients and
P (1) equal to 4 or to a prime number.
The problem of unique factorisation of (disconnected or even connected) graphs with
loops with respect to the Cartesian and the strong product is still open — cf. Lemma 2.4
below.
It has already been proven [3, Theorem 2.2] that the semiring N1 of univariate polyno-
mials with nonnegative integer coefficients has full elasticity, meaning that for any rational
number p/q ≥ 1, there exists a polynomial f ∈ N1 such that the number of irreducible
factors in its longest factorisation is to the number of factors in the shortest as p is to q.
From our results, it follows that polynomials with at most 11 terms have elasticity 1; our
smallest nontrivial example (3.10) has elasticity 3/2 and 16 terms. More such examples
may exist with 12, 14, or 15 terms. Up to now, we have not found any minimal example
with more than 2 inequivalent factorisations.
2 Factorisations of graphs
We will consider the class of finite graphs (or symmetric binary relations) Γ0 and its sub-
class consisting of graphs without loops (or simple graphs) Γ. (To be precise, we consider
the category of isomorphism classes of such graphs, with graph homomorphisms as arrows.
This will be tacitly assumed in the sequel.) A graph G is given as a pair (V,E), where V is
the set of vertices and E ⊆ V × V is the set of edges. Note that (x, y) ∈ E ⇔ (y, x) ∈ E,
as our graphs are undirected.
Definition 2.1. Let G and H be graphs in Γ0.
(i) The disjoint union G ·∪ H is the graph with vertex set V (G) ·∪ V (H) and edge set
E(G) ·∪ E(H) (both again disjoint union).
(ii) The Cartesian product G  H is the graph with vertex set V (G) × V (H), such
that (u, v) and (u′, v′) are adjacent if and only if u = u′ and (v, v′) ∈ E(H) or
(u, u′) ∈ E(G) and v = v′.
(iii) The strong product G  H is the graph with vertex set V (G) × V (H), such that
distinct vertices (u, v) and (u′, v′) are adjacent if and only if u = u′ or (u, u′) ∈
E(G), and v = v′ or (v, v′) ∈ E(H), and (u, v) has a loop if and only if at least one
of u and v has a loop.
(iv) The direct product G×H is the graph with vertex set V (G)×V (H), such that (u, v)
and (u′, v′) are adjacent if and only if (u, u′) ∈ E(G) and (v, v′) ∈ E(H).
All three products are commutative and associative and distribute over the disjoint
union, and are the only products with these properties such that the projections onto the
factors are graph homomorphisms (direct product) or weak homomorphisms (the other
C. E. van de Woestijne: Factors of disconnected graphs and polynomials with. . . 309
two). Details can be found in [8]. The Cartesian and strong products in [8] are only de-
fined for simple graphs, but the extension to arbitrary graphs, using the above definitions,
is immediate. The graph K1 consisting of one vertex and no edges is a neutral element
with respect to both the Cartesian and the strong product, whereas the graph K∗1 , having
one vertex and a loop on it, is a neutral element for the direct product. Of course, the empty
graph is a neutral element for the disjoint union; by general theory, all neutral elements are
unique. Finally, note that the Cartesian and strong products of two graphs are simple if and
only if both factors are simple. It follows that both Γ and Γ0 are commutative semirings
with operations given by the disjoint union and either the Cartesian or the strong product,
whereas Γ0 is a commutative semiring with the disjoint union and the direct product.
An essential question regarding graph products is whether the possible factorisations
obtained with respect to such a product are unique. Here, we have the following results.
Definition 2.2. Let  be a graph product. A graph G that is different from the neutral
element for  is called irreducible with respect to  if in any factorisation G = H  L,
either H or L is the neutral element for .
Theorem 2.3. [8, Theorems 6.6, 7.14, 8.17]
(i) (Sabidussi, Vizing) Let G be a connected simple graph. Then any two factorisa-
tions of G into irreducibles with respect to the Cartesian product are unique up to
permutation and isomorphism of the factors.
(ii) (Dörfler and Imrich, McKenzie) The same result holds for the strong product.
(iii) (McKenzie) Let G be a connected nonbipartite graph. Then any two factorisations
ofG into irreducibles with respect to the direct product are unique up to permutation
and isomorphism of the factors.
The next result shows that Cartesian and strong factorisation for graphs with loops is in
general not unique. The proof is easy.
Lemma 2.4. Let G be a graph in which every node has a loop. Then the same holds for
G  H and G  H , where H is any graph.
Bipartite graphs, with or without loops, generally have several nonequivalent direct fac-
torisations. A classification of these phenomena is a topic of current research by Hammack
and his co-authors [1, 6, 7].
In this note, we will consider factorisations of disconnected graphs. In this case, there
are certain well-known examples of non-unique factorisations. However, we will show that
these examples are all reducible to the same phenomenon in the semiring of polynomials
in several variables with nonnegative integral coefficients.
To do this, we slightly change our perspective on the class of graphs. Instead of thinking
of the connected components of a graph as being separate graphs standing next to each
other, and possibly repeated more than once, we will think of each connected component
as occurring only once, but with some multiplicity assigned to it. Then, an easy step is to
allow also negative multiplicities.
Definition 2.5. We define the class of graphs with integer-weighted components Γ̃0 to be
the set of formal sums
∑
G aGG, whereG runs over the connected graphs in Γ0, and where
310 Ars Math. Contemp. 5 (2012) 307–323
the aG are integers, only finitely many of which are nonzero. We define
(
∑
G
aGG) + (
∑
G
bGG) =
∑
G
(aG + bG)G;
(
∑
G
aGG) (
∑
G
bGG) =
∑
G
( ∑
HL=G
aHbL
)
G.
Doing exactly the same for the subclass Γ of simple graphs yields the set Γ̃.
In the definition of multiplication, we obtain a double sum, because it is possible that
several combinations of components H on the left and L on the right have the same prod-
uct H  L = G. We combine all these and add the multiplicities together to obtain the
multiplicity of (the isomorphism class of) G in the product.
We note that we will continue to use the term graph in the usual way; the term graph
with integer-weighted components will denote a formal sum of connected graphs, with
possibly negative multiplicities.
Theorem 2.6. For any choice of graph product  =  ,  , ×, the set Γ̃0 is a com-
mutative ring with the addition and multiplication as defined above, and the semiring Γ0
embeds into it. The additive group of Γ̃0 is the free Abelian group generated by the con-
nected graphs in Γ0.
Analogously, Γ̃ is a ring, and the semiring Γ embeds into it. The connected simple
graphs freely generate its additive group.
Proof. It is clear from the definition that the set Γ̃0 is just the free Abelian group generated
by the connected graphs.
If in a formal sum all weights aG are nonnegative, then the formal sum is just a graph
as we know it. Conversely, every graph can be written as such a sum in exactly one way,
because the decomposition of graphs as a disjoint union of connected graphs is unique.
Thus, we can embed the set of graphs into the set Γ̃0.
Furthermore, it is easy to see that addition and multiplication of nonnegatively weighted
graphs agrees exactly with the operations of ·∪ and  on graphs, so that the embedding is
a homomorphism of semirings.
Now what remains is to consider the operations on the whole set Γ̃0. The operations of
addition and multiplication are clearly commutative. Their associativity and the distribu-
tivity of  over addition are easily derived from the corresponding properties of nonneg-
atively weighted graphs, using the difference representation of formal sums, as follows.
Every graph with integer-weighted components X can be written as a difference X1−X2,
where both X1 and X2 have only nonnegative multiplicities. If we also say that every con-
nected graph has nonzero multiplicity in at most one of X1 and X2, this decomposition is
uniquely determined, and we see that elements of Γ̃0 can be uniquely written as the differ-
ence G −H of two graphs G and H that have no connected component in common (i.e.,
no component of G is isomorphic to any component of H and vice versa). The details are
left to the reader.
The construction used here is known as forming the Grothendieck group of a commu-
tative semigroup with cancellation. When the semigroup is a semiring (as in our case), the
proof shows that the construction easily extends to define a compatible ring structure on
the Grothendieck group.
C. E. van de Woestijne: Factors of disconnected graphs and polynomials with. . . 311
Now we come to the main structure theorem (see also [5, 9]).
Theorem 2.7. The following three rings are each isomorphic to the ringR = Z[x1, x2, . . . ,
xn, . . .] of polynomials with integer coefficients in countably many commuting variables:
(i) the ring Γ̃ of simple graphs with the Cartesian product;
(ii) the ring Γ̃ of simple graphs with the strong product;
(iii) the subring Γ̃0,odd of Γ̃0 of graphs having only nonbipartite connected components,
with the direct product.
Note that we cannot extend the third part of the result to the set of all nonbipartite
graphs, because this set is not a free Abelian group under disjoint union: if G is connected
bipartite and H , L are connected, nonbipartite, and nonisomorphic, then we have
(G ·∪ H) ·∪ L = (G ·∪ L) ·∪ H
where none of the factors can be written as a sum of nonempty graphs within the class of
nonbipartite graphs. One might call this kind of phenomenon a “non-unique partition”.
Proof. Let  be either the Cartesian or the strong product. As unique factorisation is only
known for connected simple graphs, we restrict ourselves to the ring Γ̃.
Now letG1, G2, . . . be some (arbitrary) enumeration of the set of irreducible connected
simple graphs.
We now define a ring homomorphism φ : R → Γ̃ by sending the variable xi to Gi
(“evaluating the polynomial at the values Gi, for i ≥ 1”), the unit element 1 to the neutral
element K1, and 0 to the empty graph. By linearity, we can obviously extend the homo-
morphism to all of R. As every connected simple graph has a unique factorisation into
irreducible graphs, the images of the monomials xe1i1 · . . . · x
er
ir
contained in R are all dis-
tinct, and every connected simple graph is the image of one such monomial. Thus, because
the generators of the additive group ofR (which are the monomials) are mapped one-to-one
onto the generators of the additive group of Γ̃, and both groups are free Abelian, we find
that φ is an isomorphism.
Next, consider the direct product. It is an easy exercise to show that the direct product
of two connected graphs in Γ0 is bipartite if and only if at least one of the factors is bipartite.
This means that the subgroup Γ̃0,odd of Γ̃0 generated by the nonbipartite connected graphs
is closed under taking direct products. As it is obviously a subgroup with respect to the
disjoint union and contains the neutral element K∗1 , it is a subring.
Now exactly the same argument as before shows that this ring is isomorphic to R.
Of the many consequences of the above isomorphisms, we only give the following.
Recall that a set S has the cancellation property with respect to some binary operation 
on S, if for all G,H,L ∈ S, the isomorphism G H ∼= G  L implies the isomorphism
H ∼= L. It means that whenever we find that a certain element of S has a factor G (is
“divisible” by G), the “quotient” by G is uniquely determined within S. Note that the
cancellation property is inherited by any subset of S that is closed under the operation ,
whereas unique factorisation does not necessarily descend in this way.
The following result is due to Lovász [10, Theorem 9] for the case of the direct product;
see also [8, Theorem 9.10]. In fact, what Lovász proved was that quotients by nonbipartite
simple graphs and by reflexive nonbipartite graphs are well-defined within the class of (not
necessarily symmetric) binary relations. However, his argument extends almost verbatim
312 Ars Math. Contemp. 5 (2012) 307–323
to the case of nonbipartite graphs that may have some loops, and the result follows by the
inheritance property just mentioned. Of course, our proof only shows that the cancellation
law follows directly from the unique factorisation of connected graphs. As said above,
we restrict to graphs without bipartite components, to avoid problems with non-unique
partition.
For the Cartesian and strong cases, the argument appears as [8, Theorems 6.21 and 9.5].
Corollary 2.8. The set of nonempty simple graphs has the cancellation property with re-
spect to the Cartesian and the strong product. The set of graphs without bipartite compo-
nents has the cancellation property with respect to the direct product.
Proof. The given sets are embedded (with the operation being respected) in the ring R,
which is a unique factorisation domain and hence certainly has the cancellation property
with respect to multiplication of polynomials.
As any two distinct factorisations GH ∼= GL would lead to distinct factorisations
of the same polynomial, we conclude that cancellation holds.
3 Polynomials with nonnegative integer coefficients
We now proceed to the actual purpose of this note. By the above isomorphisms, it is easy to
exhibit examples of disconnected graphs that have multiple non-equivalent factorisations.
For example, consider the polynomial identity
(1 + x+ x2)(1 + x3) = (1 + x2 + x4)(1 + x). (3.1)
We have factorisations 1 + x3 = (1 + x)(1 − x + x2) and 1 + x2 + x4 = (1 − x +
x2)(1 + x+ x2), but as these contain factors with negative coefficients, they do not corre-
spond to honest graph factorisation. Restricting ourselves to the semiringN of polynomials
with nonnegative integer coefficients, it follows that all the factors occurring in (3.1) are
irreducible, and hence so are the corresponding graphs, if we let x correspond to some ir-
reducible connected simple graph of our choice (as K1 is a “unit”, we do not consider it to
be irreducible). Mapping x to K2, the graph with two nodes and one edge, we obtain
(K1 +K2 +K
,2
2 ) (K1 +K
,3
2 ) = (K1 +K
,2
2 +K
,4
2 ) (K1 +K2),
where is either the Cartesian or the strong product, andG,n means the n-fold-product
of G with itself (also called the nth -power of G).
For the direct product, we can use the same example. However, here we must use
K∗1 (the vertex with a loop) as the neutral element, and the choice for φ(x) must be some
irreducible connected nonbipartite graph, such as K∗2 (two vertices with loops and an edge
between them).
Another example, due to Nüsken [5] and containing nontrivial integer coefficients, is
(2 + x+ x3)(2 + x) = (1 + x)(4 + x2 + x3). (3.2)
Here the factor 2 − x + x2 divides both cubic factors, and it follows that the latter are
irreducible in the semiring N .
However, by the structure theorem 2.7, it follows that every instance of non-unique
factorisation existing among graphs arises from a non-unique factorisation in N . We can
C. E. van de Woestijne: Factors of disconnected graphs and polynomials with. . . 313
use this fact to limit the possible phenomena inside our sets of graphs. Every graph cor-
responds to a polynomial, with each connected component corresponding to a monomial,
whose coefficient is equal to the number of components that share the same isomorphism
class.
Hence, what we will do in the sequel is to give a precise description of all factorisations
of a given polynomial (in several variables) with nonnegative integer coefficients. It turns
out that this can be done variable by variable, and (for a univariate polynomial) that it is
useful to classify the polynomials by number of terms (i.e., evaluation of the polynomial
at 1) rather than degree. We will prove results for the cases where the number of terms is
at most 10, or equal to a prime number, with the aid of computer algebra systems such as
MAGMA [2]. All software used to produce these proofs is available from the author.
3.1 General remarks
By definition of the polynomial ring (even in infinitely many variables), every given poly-
nomial involves only finitely many variables of R; therefore, also all its possible factorisa-
tions in R, taken together, comprise only finitely many variables, if we identify isomorphic
factors that differ only in the labelling of the variables. Thus, we may assume we are in the
ring Rm = Z[X1, . . . , Xm].
Let Nm ⊆ Rm be the subsemiring of polynomials in X1, . . . , Xm with nonnegative
integer coefficients. Every polynomial P ∈ Nm has a very sparse representation given by
P =
n∑
i=1
Xαi
where we use the shortcut notation Xγ =
∏m
j=1X
γj
j for a vector γ = (γ1, . . . , γm) of
nonnegative integer exponents. The vectors αi may be repeated in order to accommodate
nontrivial coefficients. We immediately have the following results.
Lemma 3.1. Let P ∈ Nm be a sum of n monomials (possibly repeated, all with coefficient
1), and suppose it factors in Nm as P = ST , where S has s monomials and T has t. Then
n = st. If all monomials of P are distinct, then so are those of S and T .
Proof. Suppose P = ST where S, T are in Nm. We have
n = P (1, 1, . . . , 1) = S(1, 1, . . . , 1)T (1, 1, . . . , 1) = st.
This proves the first claim.
Now assume all monomials of P are distinct. Then each term of P arises in at most
one way as the product of a term of S and a term of T , and hence S and T cannot have
repeated terms.
Lemma 3.2. Let P ∈ Nm be a sum of n monomials, where n is prime. Then P has unique
factorisation inside Nm.
Proof. Obviously, in any factorisation P = ST , either S or T is a monomial. Now the
greatest monomial factor is just the greatest common divisor of all terms of P , and all
other monomial factors of P divide it. Thus P is the product of a sum of n terms, without
common factor, that is irreducible in Nm, and a monomial. But monomials clearly have
unique factorisation in Nm.
314 Ars Math. Contemp. 5 (2012) 307–323
As soon as the number of terms is not prime, the problem becomes more difficult. In
order to keep track of all possible factorisations, it is important to consider the order of
terms in a polynomial as fixed, and to fix the order in which we form the terms of the
product of two polynomials. This leads to the following definition.
Definition 3.3. Let X1, . . . , Xm be variables, and let c1, . . . , ct be vectors (of length m)
of nonnegative integers. A factorisation of the polynomial
∑t
i=1X
ci , of the form
(Xa1 + . . .+Xar )
(
Xb1 + . . .+Xbs
)
= Xc1 + . . .+Xct , (3.3)
is given by two sequences (a1, . . . , ar) and (b1, . . . , bs) of length-m vectors of nonnegative
integers and a bijective mapping ρ : {1, . . . , r} × {1, . . . , s} → {1, . . . , t} such that
ai + bj = cρ(i,j) whenever 1 ≤ i ≤ r and 1 ≤ j ≤ s. (3.4)
Of course, for a factorisation to exist, we must have t = rs. As remarked before, we
may assume that the polynomials to be factored are not divisible by variables, and hence we
may always assume that mini{cij} = 0 for j = 1, . . . ,m. We now show that all possible
nonunique factorisations in the semiring N1 = Z≥0[X] are parametrised by the numbers
of terms r and s and (given r, s) the bijections ρ.
Lemma 3.4. Let c1, . . . , ct ∈ Nm such that mini{cij} = 0 for 1 ≤ j ≤ m, let r, s be
positive integers such that t = rs, and let ρ be a bijection
ρ : {1, . . . , r} × {1, . . . , s} → {1, . . . , t}.
Then there exists at most one choice of sequences (a1, . . . , ar) and (b1, . . . , bs) of elements
of Nm such that ai + bj = cρ(i,j) for all i, j. If such a choice is possible, then we have
mini{aij} = mini{bij} = 0 and {aij} ∪ {bij} ⊆ {cij} for 1 ≤ j ≤ m.
Proof. We consider only the first components of all vectors. We get a system
ai1 + bj1 = cρ(i,j),1 1 ≤ i ≤ s, 1 ≤ j ≤ r
of linear equations, to be solved in nonnegative integers. Without loss of generality, we
assume c11 = 0 and ρ(1, 1) = 1. This means that a11 + b11 = c11 = 0, and hence
a11 = b11 = 0. The other sums a11 + bj1 and ai1 + b11 all occur as the left hand side
of an equation, and hence all ai1 and bj1 are determined and must occur among the ci1.
The remaining equations then define relations that the ci1 must satisfy among themselves
in order for the system to be soluble. It follows that the system has at most one solution for
a given ρ.
Finally, it is obvious that the systems for the other components are completely analo-
gous and independent from each other, and hence must all be satisfied independently. It
follows that there is at most one solution.
Corollary 3.5. Let X1, . . . , Xm be variables. Given r, s, (c1, . . . , ct) ⊆ Nm, and ρ, there
exists at most one factorisation of the form (3.3).
The last observation in this section shows that the problem of parametrising all factori-
sations can be immediately reduced to the univariate case.
C. E. van de Woestijne: Factors of disconnected graphs and polynomials with. . . 315
Lemma 3.6. Let X1, . . . , Xm be variables, and let c1, . . . , ct ∈ Nm. A quintuple
(r, s, ρ, (a1, . . . , ar), (b1, . . . , bs))
is a factorisation of
∑t
i=1X
ci in Nm if and only if (r, s, ρ, (a1j , . . . , arj), (b1j , . . . , bsj))
is a factorisation of
∑t
i=1X
cij
j inside N[Xj ], for j = 1, . . . ,m.
Proof. A polynomial in Nm in very sparse representation is determined by its exponent
vectors c1, . . . , ct. If we set all variables except X1, say, to one, this corresponds to setting
all components of the ci, except the first, to zero. If we do this for all the Xj separately,
we see that no information in the ci is lost; in fact, we map the matrix (cij) to the set
of its columns. In particular, factorisations are preserved, and factorisations with respect
to each variable separately can be pieced together to obtain a factorisation of the original
polynomial.
Note that the maps described in the proof are not the same as the semiring homo-
morphisms of evaluating all variables (except one) at 1. For example, the polynomial
XY + 1 in very sparse representation maps to (X + 1, Y + 1) as usual, but X + Y maps
to (X + 1, 1 + Y ), which is different since the order of terms is fixed. This allows us to
reconstruct polynomials in Nm from their images under these partial evaluations; in fact,
the number of terms (in the very sparse sense) does not change under our maps.
3.2 The univariate case
From now on, we restrict attention to the case of polynomials in N1. More precisely,
let t be a nonnegative integer; we consider polynomials P ∈ N1 such that P (1) = t
and P (0) > 0, so that P is not divisible by the variable. By what we just proved, it
follows that we can obtain all possible factorisations of P insideN1 by listing all nontrivial
factorisations t = rs (i.e., r > 1 and s > 1), and for every pair (r, s), listing all possible
bijections ρ : {1, . . . , r} × {1, . . . , s} → {1, . . . , t}. To complete the arguments, we will
need to prove that the factors thus found are irreducible; if they are not, we will have to
consider products of more than two factors. We will give complete descriptions of the cases
t = 1, . . . , 11; of course, the cases where t is prime were already dealt with by Lemma 3.2.
Suppose r and s are given. We will assume that factorisations have the form (3.3),
where moreover we have
0 = c1 ≤ c2 ≤ . . . ≤ ct; (3.5)
0 = a1 ≤ . . . ≤ ar and 0 = b1 ≤ . . . ≤ bs. (3.6)
As in the proof of Lemma 3.4, the exponents a1 + bj = bj correspond to cρ(1,j), and
ai + b1 = ai to cρ(i,1). Therefore, we obtain the equations
cρ(i,j) = cρ(i,1) + cρ(1,j) (2 ≤ i ≤ r, 2 ≤ j ≤ s). (3.7)
Thus, for every bijection ρ, the polynomials possessing a factorisation of the form (3.3)
with bijection ρ correspond to the solutions in the (nonnegative integer) variables ci of the
system of linear equations and inequalities made up by (3.5) together with (3.7). Such
systems are closely related to integer linear programs. The difference is that there is no
cost function to be optimised; instead, we are interested in all solutions, which hopefully
316 Ars Math. Contemp. 5 (2012) 307–323
will admit a compact description. If the polynomial P is to have two different factori-
sations, then the ci must satisfy two such linear programs simultaneously, namely those
corresponding to two different ρ.
To facilitate the analysis, we can use the fact that (3.7) also implies a partial ordering
of the ci; in fact, because all exponents are nonnegative, we have
cρ(i,j) ≥ cρ(i,1) and cρ(i,j) ≥ cρ(1,j) (2 ≤ i ≤ r, 2 ≤ j ≤ s). (3.8)
Furthermore, in the end we will drop the fixed term ordering of the factors, so that we may
assume (3.6). The implication is that
cρ(i,j) ≥ cρ(i′,j′) (1 ≤ i′ ≤ i ≤ r, 1 ≤ j′ ≤ j ≤ s). (3.9)
The number of pairs of ci whose ordering is thus determined is(
r
2
)(
s
2
)
+ r
(
s
2
)
+
(
r
2
)
s =
r2s2 + rs2 + r2s− 3rs
4
,
to be compared with the total number
(
rs
2
)
= r
2s2−rs
2 of pairs. The idea is that the partial
ordering (3.9) severely limits the possibilities for ρ.
In the (computer-aided) proofs below, we will repeatedly compute all possibilities for
the bijection ρ that are compatible with the partial order (3.9). It is readily seen that this
problem is equivalent to the topological sorting of a given directed acyclic graph (dag) with
nodes {1, . . . , rs}. Here the graph represents the few ordering relationships that are known,
and the task is to produce a total ordering of the nodes such that node i comes before node j
whenever the graph contains an edge (i, j). Such orderings exist if and only if the graph is
acyclic, and are not unique. To find all possible such orderings, we apply Kahn’s algorithm,
with backtracking. (It is possible to use a depth-first search to find a topological ordering,
and this is the algorithm given in [4, Section 23.4], but this algorithm is unable to derive all
topological orderings — cf. Exercise 23.4-2 in [4]. In fact, on the graph with three nodes
and edges (1, 3) and (2, 3), depth-first search will never produce the ordering 123, which
is perfectly valid.)
Kahn’s algorithm works as follows: take any node with zero in-degree, write it down,
remove it and all its incident edges from the graph, and continue recursively with the re-
mainder until the graph is trivial. By backtracking on every choice that we make, we
smoothly produce all possible topological orderings.
3.3 Results for polynomials with few terms
Using the above framework, we arrive at the following results.
Theorem 3.7. Let P ∈ N1 be a sum of 4 monomials, not necessarily distinct. Then P has
unique factorisation inside N1.
Proof. We may assume the variable does not divide P . Putting t = 4, the only nontrivial
factorisation is one having r = s = 2.
Let us see which bijections ρ can occur in the factorisation
(1 +Xa2)(1 +Xb2) = 1 +Xc2 +Xc3 +Xc4 .
C. E. van de Woestijne: Factors of disconnected graphs and polynomials with. . . 317
We may put ρ(1, 1) = 1 and ρ(2, 2) = 4, because cρ(2,2) = a2 + b2 must be maximal
among the ci. The only remaining choice for ρ is whether ρ(1, 2) = 2 or 3 (i.e., whether
c2 = a2 or c3 = a2). But this choice corresponds to exchanging the factors on the left,
and this leaves the factorisation is essentially unchanged. Thus ρ is fixed, and P is either
irreducible (if c4 6= c2 + c3) or breaks up uniquely into two factors of 2 terms each, by
Corollary 3.5.
Theorem 3.8. Let P ∈ N1 be a sum of 6 monomials, not necessarily distinct. If P has
non-unique factorisation, then it is of the form
Xa(1 +Xb +X2b +X3b +X4b +X5b)
for nonnegative integers a and b ≥ 1.
Note that 1+X+X2+X3+X4+X5 = (1+X2+X4)(1+X) = (1+X+X2)(1+X3).
All factors are irreducible in N1 by Lemma 3.2.
Proof. We may assume the variable does not divide P . Any nontrivial factorisation is of
the form
(1 +Xa2 +Xa3)(1 +Xb2).
We use the topological sorting algorithm described above to find all bijections ρ that are
compatible with the partial ordering (3.9), and find that there are the following 5 possibili-
ties (to be read such that the first pair corresponds to c1, the next to c2, and so on):
Case: c1 c2 c3 c4 c5 c6
1 (1, 1) (1, 2) (2, 1) (2, 2) (3, 1) (3, 2)
2 (1, 1) (1, 2) (2, 1) (3, 1) (2, 2) (3, 2)
3 (1, 1) (2, 1) (1, 2) (2, 2) (3, 1) (3, 2)
4 (1, 1) (2, 1) (1, 2) (3, 1) (2, 2) (3, 2)
5 (1, 1) (2, 1) (3, 1) (1, 2) (2, 2) (3, 2)
We now check whether any combination of two ρ admits a simultaneous solution of the
respective sets of equations (3.7). For example, the first line implies the equations
c4 = c2 + c3 and c6 = c2 + c5,
whereas the fifth line gives
c5 = c2 + c4 and c6 = c3 + c4.
The general solution to these four equations combined is
c2 free, c3 = 2c2, c4 = 3c2, c5 = 4c2, c6 = 5c2,
and this corresponds to the factorisation given in the theorem. To see this, note that the
bijection ρ from the first line results in a2 = c3, a3 = c5, and b2 = c2, whereas the fifth
line gives a2 = c2, a3 = c3, and b2 = c4.
It is easily (but tediously) verified that in all the factorisations given by combining two
other lines of the table, the left and right hand sides coincide. Hence the only nonunique
factorisation is the one given in the Theorem, and a fortiori, polynomials that possess three
inequivalent factorisations do not exist.
318 Ars Math. Contemp. 5 (2012) 307–323
Theorem 3.9. Let P ∈ N1 be a sum of 8 monomials, not necessarily distinct. Then P has
unique factorisation inside N1.
Proof. We may assume the variable does not divide P . If P is reducible in N1, then it can
be written as a product of factors with 2 and 4 terms, where the latter possibly again breaks
up as a product of factors with 2 terms each. We will first establish all possible nonunique
factorisations of type
(1 +Xa2)(1 +Xb2 +Xb3 +Xb4).
All bijections ρ : {1, 2} × {1, 2, 3, 4} → {1, . . . , 8} that are compatible with the partial
ordering (3.9) are enumerated by MAGMA as
Case: c1 c2 c3 c4 c5 c6 c7 c8
1 (1, 1) (1, 2) (1, 3) (1, 4) (2, 1) (2, 2) (2, 3) (2, 4)
2 (1, 1) (1, 2) (1, 3) (2, 1) (1, 4) (2, 2) (2, 3) (2, 4)
3 (1, 1) (1, 2) (1, 3) (2, 1) (2, 2) (1, 4) (2, 3) (2, 4)
4 (1, 1) (1, 2) (1, 3) (2, 1) (2, 2) (2, 3) (1, 4) (2, 4)
5 (1, 1) (1, 2) (2, 1) (1, 3) (1, 4) (2, 2) (2, 3) (2, 4)
6 (1, 1) (1, 2) (2, 1) (1, 3) (2, 2) (1, 4) (2, 3) (2, 4)
7 (1, 1) (1, 2) (2, 1) (1, 3) (2, 2) (2, 3) (1, 4) (2, 4)
8 (1, 1) (1, 2) (2, 1) (2, 2) (1, 3) (1, 4) (2, 3) (2, 4)
9 (1, 1) (1, 2) (2, 1) (2, 2) (1, 3) (2, 3) (1, 4) (2, 4)
10 (1, 1) (2, 1) (1, 2) (1, 3) (1, 4) (2, 2) (2, 3) (2, 4)
11 (1, 1) (2, 1) (1, 2) (1, 3) (2, 2) (1, 4) (2, 3) (2, 4)
12 (1, 1) (2, 1) (1, 2) (1, 3) (2, 2) (2, 3) (1, 4) (2, 4)
13 (1, 1) (2, 1) (1, 2) (2, 2) (1, 3) (1, 4) (2, 3) (2, 4)
14 (1, 1) (2, 1) (1, 2) (2, 2) (1, 3) (2, 3) (1, 4) (2, 4)
The equations (3.7) are
cρ(2,2) = cρ(2,1) + cρ(1,2) = a2 + b2;
cρ(2,3) = cρ(2,1) + cρ(1,3) = a2 + b3;
cρ(2,4) = cρ(2,1) + cρ(1,4) = a2 + b4.
As a typical example, we will consider polynomials admitting factorisations under both
bijections 1 and 8. Besides c1 = 0, which always holds, we find
c6 = c2 + c5; c7 = c3 + c5; c8 = c4 + c5;
c4 = c2 + c3; c7 = c3 + c5; c8 = c3 + c6,
which is obviously solvable taking c2, c3, and c5 as free (nonnegative integer) parameters.
The corresponding inequivalent factorisations are
(1 +Xc5)(1 +Xc2 +Xc3 +Xc2+c3) = (1 +Xc3)(1 +Xc2 +Xc5 +Xc2+c5).
However, in fact both can be further reduced to
(1 +Xc2)(1 +Xc3)(1 +Xc5).
Automated verification by MAGMA shows that likewise all pairs of inequivalent 2 × 4-
factorisations that arise by combining two cases from the table can be further reduced to
equivalent 2× 2× 2-factorisations. This proves the theorem.
C. E. van de Woestijne: Factors of disconnected graphs and polynomials with. . . 319
Theorem 3.10. Let P ∈ N1 be a sum of 9 monomials, not necessarily distinct. Then P
has unique factorisation inside N1.
Proof. We may assume the variable does not divide P . Putting t = 9, the only nontrivial
factorisation is one having r = s = 3.
We establish all factorisations of the form
(1+Xa2 +Xa3)(1+Xb2 +Xb3) = 1+Xc2 +Xc3 +Xc4 +Xc5 +Xc6 +Xc7 +Xc8 +Xc9 .
Exchanging the factors does not change the factorisation, so in addition to the partial or-
dering (3.9), we also require that a2 ≥ b2, hence cρ(2,1) ≥ cρ(1,2). All bijections ρ that are
compatible with these orderings are enumerated by MAGMA to be
Case: c1 c2 c3 c4 c5 c6 c7 c8 c9
1 (1, 1) (1, 2) (1, 3) (2, 1) (2, 2) (2, 3) (3, 1) (3, 2) (3, 3)
2 (1, 1) (1, 2) (1, 3) (2, 1) (2, 2) (3, 1) (2, 3) (3, 2) (3, 3)
3 (1, 1) (1, 2) (1, 3) (2, 1) (2, 2) (3, 1) (3, 2) (2, 3) (3, 3)
4 (1, 1) (1, 2) (1, 3) (2, 1) (3, 1) (2, 2) (2, 3) (3, 2) (3, 3)
5 (1, 1) (1, 2) (1, 3) (2, 1) (3, 1) (2, 2) (3, 2) (2, 3) (3, 3)
6 (1, 1) (1, 2) (2, 1) (1, 3) (2, 2) (2, 3) (3, 1) (3, 2) (3, 3)
7 (1, 1) (1, 2) (2, 1) (1, 3) (2, 2) (3, 1) (2, 3) (3, 2) (3, 3)
8 (1, 1) (1, 2) (2, 1) (1, 3) (2, 2) (3, 1) (3, 2) (2, 3) (3, 3)
9 (1, 1) (1, 2) (2, 1) (1, 3) (3, 1) (2, 2) (2, 3) (3, 2) (3, 3)
10 (1, 1) (1, 2) (2, 1) (1, 3) (3, 1) (2, 2) (3, 2) (2, 3) (3, 3)
11 (1, 1) (1, 2) (2, 1) (2, 2) (1, 3) (2, 3) (3, 1) (3, 2) (3, 3)
12 (1, 1) (1, 2) (2, 1) (2, 2) (1, 3) (3, 1) (2, 3) (3, 2) (3, 3)
13 (1, 1) (1, 2) (2, 1) (2, 2) (1, 3) (3, 1) (3, 2) (2, 3) (3, 3)
14 (1, 1) (1, 2) (2, 1) (2, 2) (3, 1) (1, 3) (2, 3) (3, 2) (3, 3)
15 (1, 1) (1, 2) (2, 1) (2, 2) (3, 1) (1, 3) (3, 2) (2, 3) (3, 3)
16 (1, 1) (1, 2) (2, 1) (2, 2) (3, 1) (3, 2) (1, 3) (2, 3) (3, 3)
17 (1, 1) (1, 2) (2, 1) (3, 1) (1, 3) (2, 2) (2, 3) (3, 2) (3, 3)
18 (1, 1) (1, 2) (2, 1) (3, 1) (1, 3) (2, 2) (3, 2) (2, 3) (3, 3)
19 (1, 1) (1, 2) (2, 1) (3, 1) (2, 2) (1, 3) (2, 3) (3, 2) (3, 3)
20 (1, 1) (1, 2) (2, 1) (3, 1) (2, 2) (1, 3) (3, 2) (2, 3) (3, 3)
21 (1, 1) (1, 2) (2, 1) (3, 1) (2, 2) (3, 2) (1, 3) (2, 3) (3, 3)
As before, we consider the systems of 8 equations in 8 variables (putting the trivial c1 = 0
apart) that arise by combining two possibilities for ρ as given in the table. Among the
8 variables, at most four are conceptually suited as free parameters, namely those corre-
sponding to one of the variables a2, a3, b2, and b3 under both bijections simultaneously.
It is interesting to note that it is always possible to choose the free parameters among this
small set of “interesting” variables (it would be even more interesting to prove this before-
hand), even if some of the occurring systems have rank as small as 5 (out of 8).
We let MAGMA run through all combinations of two bijections, and the result is that
in all cases, the obtained factorisations are equivalent.
Theorem 3.11. Let P ∈ N1 be a sum of 10 monomials, not necessarily distinct. If P has
320 Ars Math. Contemp. 5 (2012) 307–323
non-unique factorisation, then it is of the form
(1 +X)(1 +X2 +X4 +X6 +X8) = (1 +X5)(1 +X +X2 +X3 +X4);
(1 +X)(1 +X4 +X6 +X8 +X12) = (1 +X5)(1 +X +X4 +X7 +X8);
(1 +X3)(1 +X2 +X4 +X6 +X8) = (1 +X5)(1 +X2 +X3 +X4 +X6),
up to multiplication by a power of X , and up to replacing X by a power, or it has one of
the forms
(1 +X3b)(1 +Xa +Xb +Xa+b+Xa+2b) =
(1 +Xb)(1 +Xa +Xa+2b +X3b +Xa+4b);
(1 +X3b)(1 +Xa +Xb +Xa+b +X2b) =
(1 +Xb)(1 +Xa +X2b +Xa+3b +X4b)
for integers a ≥ 0 and b ≥ 1, up to multiplication by a power of X .
The products in the theorem evaluate to
1 +X +X2 +X3 +X4 +X5 +X6 +X7 +X8 +X9;
1 +X +X4 +X5 +X6 +X7 +X8 +X9 +X12 +X13;
1 +X2 +X3 +X4 +X5 +X6 +X7 +X8 +X9 +X11;
1 +Xa +Xb +Xa+b +Xa+2b +X3b +Xa+3b +X4b +Xa+4b +X5b;
1 +Xa +Xb +Xa+b +X2b +X3b +Xa+3b +X4b +Xa+4b +X5b,
respectively. All factors are irreducible in N1 by Lemma 3.2. Note that the first three cases
are all self-reciprocal; the reciprocals of all polynomials in the two 2-parameter families
again belong to one of these two families.
Proof. We may assume the variable does not divide P . Any nontrivial factorisation is of
the form
(1 +Xa2)(1 +Xb2 +Xb3 +Xb4 +Xb5).
We use topological sorting and backtracking to find all bijections ρ that are compatible with
the partial ordering (3.9), and find the cases listed in Figure 1. Solving the linear systems
obtained by every combination of two entries in the table, this time we find 102 pairs with
nonequivalent factorisations. A good way to classify the cases is to look at the two values
for the variable a2, i.e., the nontrivial exponent of X in the left-hand factor.
In some cases, such as the combination Cases 5 and 38, the two choices for the left-
hand factor are (1 + Xa) and (1 + X−a), for a free parameter a. As the parameter must
be integer and nonnegative, it is forced to be 0, and the corresponding factorisations are
equivalent after all.
In the combinations 1 and 42, as well as 10 and 42, the left-hand factors are (1 + Xa)
and (1 +X5a), giving the first and second case in the theorem.
In the combinations (1, 9), (1, 17), (1, 30), (2, 12), (2, 20), (2, 33), (3, 14), (3, 22),
(6, 25), (6, 38), and (7, 27), the left-hand factors are (1 +X3a) and (1 +X5a), giving the
third case in the theorem (all combinations yield the same right-hand factors).
C. E. van de Woestijne: Factors of disconnected graphs and polynomials with. . . 321
Case: c1 c2 c3 c4 c5 c6 c7 c8 c9 c10
1 (1, 1) (1, 2) (1, 3) (1, 4) (1, 5) (2, 1) (2, 2) (2, 3) (2, 4) (2, 5)
2 (1, 1) (1, 2) (1, 3) (1, 4) (2, 1) (1, 5) (2, 2) (2, 3) (2, 4) (2, 5)
3 (1, 1) (1, 2) (1, 3) (1, 4) (2, 1) (2, 2) (1, 5) (2, 3) (2, 4) (2, 5)
4 (1, 1) (1, 2) (1, 3) (1, 4) (2, 1) (2, 2) (2, 3) (1, 5) (2, 4) (2, 5)
5 (1, 1) (1, 2) (1, 3) (1, 4) (2, 1) (2, 2) (2, 3) (2, 4) (1, 5) (2, 5)
6 (1, 1) (1, 2) (1, 3) (2, 1) (1, 4) (1, 5) (2, 2) (2, 3) (2, 4) (2, 5)
7 (1, 1) (1, 2) (1, 3) (2, 1) (1, 4) (2, 2) (1, 5) (2, 3) (2, 4) (2, 5)
8 (1, 1) (1, 2) (1, 3) (2, 1) (1, 4) (2, 2) (2, 3) (1, 5) (2, 4) (2, 5)
9 (1, 1) (1, 2) (1, 3) (2, 1) (1, 4) (2, 2) (2, 3) (2, 4) (1, 5) (2, 5)
10 (1, 1) (1, 2) (1, 3) (2, 1) (2, 2) (1, 4) (1, 5) (2, 3) (2, 4) (2, 5)
11 (1, 1) (1, 2) (1, 3) (2, 1) (2, 2) (1, 4) (2, 3) (1, 5) (2, 4) (2, 5)
12 (1, 1) (1, 2) (1, 3) (2, 1) (2, 2) (1, 4) (2, 3) (2, 4) (1, 5) (2, 5)
13 (1, 1) (1, 2) (1, 3) (2, 1) (2, 2) (2, 3) (1, 4) (1, 5) (2, 4) (2, 5)
14 (1, 1) (1, 2) (1, 3) (2, 1) (2, 2) (2, 3) (1, 4) (2, 4) (1, 5) (2, 5)
15 (1, 1) (1, 2) (2, 1) (1, 3) (1, 4) (1, 5) (2, 2) (2, 3) (2, 4) (2, 5)
16 (1, 1) (1, 2) (2, 1) (1, 3) (1, 4) (2, 2) (1, 5) (2, 3) (2, 4) (2, 5)
17 (1, 1) (1, 2) (2, 1) (1, 3) (1, 4) (2, 2) (2, 3) (1, 5) (2, 4) (2, 5)
18 (1, 1) (1, 2) (2, 1) (1, 3) (1, 4) (2, 2) (2, 3) (2, 4) (1, 5) (2, 5)
19 (1, 1) (1, 2) (2, 1) (1, 3) (2, 2) (1, 4) (1, 5) (2, 3) (2, 4) (2, 5)
20 (1, 1) (1, 2) (2, 1) (1, 3) (2, 2) (1, 4) (2, 3) (1, 5) (2, 4) (2, 5)
21 (1, 1) (1, 2) (2, 1) (1, 3) (2, 2) (1, 4) (2, 3) (2, 4) (1, 5) (2, 5)
22 (1, 1) (1, 2) (2, 1) (1, 3) (2, 2) (2, 3) (1, 4) (1, 5) (2, 4) (2, 5)
23 (1, 1) (1, 2) (2, 1) (1, 3) (2, 2) (2, 3) (1, 4) (2, 4) (1, 5) (2, 5)
24 (1, 1) (1, 2) (2, 1) (2, 2) (1, 3) (1, 4) (1, 5) (2, 3) (2, 4) (2, 5)
25 (1, 1) (1, 2) (2, 1) (2, 2) (1, 3) (1, 4) (2, 3) (1, 5) (2, 4) (2, 5)
26 (1, 1) (1, 2) (2, 1) (2, 2) (1, 3) (1, 4) (2, 3) (2, 4) (1, 5) (2, 5)
27 (1, 1) (1, 2) (2, 1) (2, 2) (1, 3) (2, 3) (1, 4) (1, 5) (2, 4) (2, 5)
28 (1, 1) (1, 2) (2, 1) (2, 2) (1, 3) (2, 3) (1, 4) (2, 4) (1, 5) (2, 5)
29 (1, 1) (2, 1) (1, 2) (1, 3) (1, 4) (1, 5) (2, 2) (2, 3) (2, 4) (2, 5)
30 (1, 1) (2, 1) (1, 2) (1, 3) (1, 4) (2, 2) (1, 5) (2, 3) (2, 4) (2, 5)
31 (1, 1) (2, 1) (1, 2) (1, 3) (1, 4) (2, 2) (2, 3) (1, 5) (2, 4) (2, 5)
32 (1, 1) (2, 1) (1, 2) (1, 3) (1, 4) (2, 2) (2, 3) (2, 4) (1, 5) (2, 5)
33 (1, 1) (2, 1) (1, 2) (1, 3) (2, 2) (1, 4) (1, 5) (2, 3) (2, 4) (2, 5)
34 (1, 1) (2, 1) (1, 2) (1, 3) (2, 2) (1, 4) (2, 3) (1, 5) (2, 4) (2, 5)
35 (1, 1) (2, 1) (1, 2) (1, 3) (2, 2) (1, 4) (2, 3) (2, 4) (1, 5) (2, 5)
36 (1, 1) (2, 1) (1, 2) (1, 3) (2, 2) (2, 3) (1, 4) (1, 5) (2, 4) (2, 5)
37 (1, 1) (2, 1) (1, 2) (1, 3) (2, 2) (2, 3) (1, 4) (2, 4) (1, 5) (2, 5)
38 (1, 1) (2, 1) (1, 2) (2, 2) (1, 3) (1, 4) (1, 5) (2, 3) (2, 4) (2, 5)
39 (1, 1) (2, 1) (1, 2) (2, 2) (1, 3) (1, 4) (2, 3) (1, 5) (2, 4) (2, 5)
40 (1, 1) (2, 1) (1, 2) (2, 2) (1, 3) (1, 4) (2, 3) (2, 4) (1, 5) (2, 5)
41 (1, 1) (2, 1) (1, 2) (2, 2) (1, 3) (2, 3) (1, 4) (1, 5) (2, 4) (2, 5)
42 (1, 1) (2, 1) (1, 2) (2, 2) (1, 3) (2, 3) (1, 4) (2, 4) (1, 5) (2, 5)
Figure 1: Bijections for factorisations into 2× 5 terms
322 Ars Math. Contemp. 5 (2012) 307–323
In all other combinations, the ratio of the exponents of X in the left-hand factors is
1 : 3, and we find that all cases are covered by the two-parameter families given in the
theorem. Here, some combinations give one of the two-parameter families, while others
give a specialisation of either family, obtained by putting a = kb for k ∈ {0, 1, 2, 3, 4, 5}.
In fact (an observation of Wilfried Imrich), each of the theorems given above yields a
slightly more general result. Namely, if instead of considering just terms with coefficient
1, we give each term of our polynomials a positive integer coefficient, then the same proofs
show that there is at most one factorisation of P of the given form (for example, a fac-
torisation of a 10-term polynomial into a product of a 2-term and a 5-term polynomial).
For it to hold, now also the coefficients must satisfy some easy relations. However, if we
have nontrivial coefficients, the number of terms a factor has need not divide the number
of terms of the polynomial itself, and therefore (unfortunately) this is not enough to show
that all quadrinomials have unique factorisation in N1.
The method above clearly reduces the classification of all non-unique factorisations of
univariate polynomials with a given number of terms n to a finite computation. However,
it is also clear that the cost of these computations increases exponentially with n. Hence
we refrain from attacking the complicated case n = 12. We only note that factorisation
of 12-term polynomials in N1 is non-unique, as witnessed by Nüsken’s example (3.2).
Also, although cases where n is a prime power apparently tend to possess few nonunique
factorisations or none at all, there do exist examples, such as
(1 +X3 +X5 +X6)(1 +X +X2 +X4) = (1 +X)(1 +X2)(1 + 2X4 +X7) (3.10)
for n = 16, which also shows that the monoid N1 is not half-factorial.
Acknowledgement
I would like to thank Richard Hammack and Wilfried Imrich for introducing me to the
subject in such an agreeable way, and Alfred Geroldinger for his interest and for pointing
me to the recent paper [3].
References
[1] G. Abay-Asmerom, R. H. Hammack, C. E. Larson and D. T. Taylor, Direct product factor-
ization of bipartite graphs with bipartition-reversing involutions, SIAM J. Discrete Math. 23
(2009/10), 2042–2052.
[2] W. Bosma, J. Cannon and C. Playoust, The Magma algebra system, I., The user language, J.
Symbolic Comput. 24 (1997), 235–265.
[3] P. Cesarz, S. T. Chapman, S. McAdam and G. J. Schaeffer, Elastic properties of some semirings
defined by positive systems, in: Commutative algebra and its applications, Walter de Gruyter,
Berlin, 2009, 89–101.
[4] T. H. Cormen, C. E. Leiserson and R. L. Rivest, Introduction to algorithms, The MIT Electrical
Engineering and Computer Science Series, MIT Press, Cambridge, MA, 1990.
[5] A. Fernández, T. Leighton and J. L. López-Presa, Containment properties of product and power
graphs, Discrete Appl. Math. 155 (2007), 300–311.
[6] R. H. Hammack, Isomorphic components of direct products of bipartite graphs, Discuss. Math.
Graph Theory 26 (2006), 231–248.
C. E. van de Woestijne: Factors of disconnected graphs and polynomials with. . . 323
[7] R. H. Hammack, Proof of a conjecture concerning the direct product of bipartite graphs, Euro-
pean J. Combin. 30 (2009), 1114–1118.
[8] R. H. Hammack, W. Imrich and S. Klavžar, Handbook of Product Graphs, CRC Press, sec-
ond ed., 2011.
[9] W. Imrich, S. Klavžar and D. F. Rall, Cancellation properties of products of graphs, Discrete
Appl. Math. 155 (2007), 2362–2364.
[10] L. Lovász, On the cancellation law among finite relational structures, Period. Math. Hungar. 1
(1971), 145–156.