Professional Documents
Culture Documents
Micah J. Leamer
Thesis submitted to the faculty of
Virginia Polytechnic Institute and State University
in partial fulllment of the requirements for the degree of
Master of Science
in
Mathematics
Edward Green, Chair
Charles Parry
John Rossi
July 1, 2004
Blacksburg, Virginia
Keywords: Groebner Bases, Path Algebra
Copyright 2004, Micah J. Leamer
Gr obner Finite Path Algebras
Micah J. Leamer
Abstract
Let K be a eld and a nite directed multi-graph. In this paper I
classify all path algebras K and admissible orders with the property that
all of their nitely generated ideals have nite Gr obner bases.
Acknowledgements
First of all I would like to thank Charles Parry and John Rossi for joining my
committee on such short notice. They exemplify the supportive environment that
the VA Tech math department fosters. Joseph Ball offered up his time and his
expertise to meet with me for several weeks and teach me the nuances of opera-
tor algebras. He was a faithful committee member until the rescheduling of my
defense conicted with his prior commitments. Adrian Keister offered me help
with many latex problems. Hannah Swiger has been invaluable. She has gone
above and beyond the line of duty to help me process forms and meet all of the
administrative requirements. She has been positive, reassuring and at all times
candid. I would especially like to thank Dean Reiss for encouraging me to join
the graduate school at Tech. He also opened doors so that my application process
was smooth and simple. I would not have spent this year at Tech nor written this
thesis otherwise.
Edward Green the chair of my committee has been a core gure in my math-
ematical development, as well as the central resource for this thesis. He has con-
sidered me valuable enough to meet with on a weekly basis for the past two years.
All of our meetings for the past year have been for the purpose of this thesis. I am
a better communicator and certainly much better at explaining my mathematical
ideas because of him. He has time and again listened to my theoretical babble and
helped me nd the gems therein. I have seen how he has balanced a successful
career in mathematics with an adept social intelligence. The marriage of these
two skills is rare. He introduced me to Gr obner bases and path algebras. He has
guided me in my paper writing skills. He has reviewed multiple drafts of this
paper from my original chicken scratches to its current state. He has been both
capable at noting aws in proof and humble enough to correct my grammatical
mistakes. Most importantly he has been patient enough to put up with the eccen-
tricities of a young mathematician, who is full of himself.
There have been an endless number of other mathematicians who have aided
in my development, showed me kindness, and treated me as much like a colleague
as a pupil. Among the best of these are: Glen van Brummelen, Ezra Brown, Gail
Letzter, Neil Calkin and Peter Haskell.
iii
Contents
1 Introduction 1
2 The Fundamentals of Path Algebras 2
2.1 Examples of Path Algebras . . . . . . . . . . . . . . . . . . . . . 3
3 Path Orders 4
3.1 Examples of Admissible Orders: . . . . . . . . . . . . . . . . . . 5
4 Prerequisites for Gr obner Theory 5
5 Basics of non-commutative Gr obner theory 7
6 Induced Subgraphs 12
7 Classifying Noetherian Path Algebras 14
8 Producing Non-commutative Gr obner Bases 16
9 Classifying Gr obner Finite Path Algebras 18
iv
1 Introduction
The rst half of this paper may serve as a brief introduction to path algebras and
non-commutative Gr obner bases theory. Nothing is assumed of the reader other
than a general understanding of algebra and graph theory, at the graduate level.
For now it will sufce to know that path algebras are a type of non-commutative
algebra over a eld. In particular, given a eld K and a nite directed multi-graph
, the path algebra K is the set of all K-linear combinations of paths of nite
length on . We give a concise denition of path algebras in section 2. Sections 3-
5 introduce some key concepts in non-commutative Gr obner theory, all of which
are well known results. Given an ordering < on the set of paths of nite length on
a graph and an ideal I in the path algebra K there is a special type of generat-
ing set for I, called a Gr obner basis for I. We dene Gr obner bases in section 5.
Since Gr obner bases are order dependent, in section 3 we identify the path orders
which are relevant to the multiplicative structure of a path algebra. These relevant
orders on the set of paths will be called admissible orders. We will see that given
an ideal I K and an admissible order < there always exist Gr obner bases for
I. If every ideal in a ring R has a Gr obner basis then we say R has a Gr obner basis
theory. So any path algebra K has a Gr obner basis theory. Some of the key con-
cepts for dening Gr obner bases and understanding their uses will be introduced
in section 4. After dening Gr obner bases in section 5 we go on to further explain
some of their properties. Many of the concepts found in commutative Gr obner
theory are relevant to the non-commutative theory. For the interested reader both
[4] and [9] contain an introduction to commutative Gr obner bases. Other intro-
ductions to non-commutative Gr obner bases can be found in [5], [6], and [8].
The main thrust of this paper is to determine some conditions as to when
nitely generated ideals in path algebras have nite Gr obner bases. This study
begins in section 6, where we introduce a special type of subgraph (v
i
, v
j
) called
the induced subgraph between the vertices v
i
and v
j
of the graph . Addition-
ally, in section 6 we explore some of the useful properties of induced subgraphs.
Induced subgraphs and their properties will be instrumental in proving our main
result in section 9. Induced subgraphs are rst utilized in section 7 where we clas-
sify the Noetherian path algebras. It is known that every ideal in a Noetherian ring
has nite Gr obner bases. Although it is generally known to those experienced in
working with path algebras which ones are Noetherian, a proof of this classica-
tion, to the best of my knowledge, has not been previously spelled out. In section
8 we give an algorithm for nding Gr obner bases and prove one of the central
1
results of Gr obner bases theory: Bergmans diamond lemma [2].
Given a nitely generated ideal in a path algebra it often occurs that any
Gr obner basis of that ideal will be innite, regardless of what ordering is being
used. It is known that deciding whether an ideal I K has a nite Gr obner
basis is unsolvable in general. It is often necessary to rst nd a nite Gr obner
basis to show one exists, which is not always true. Our main result, theorem 9.9, is
a classication of all path algebra, admissible order pairs which have the property
that all of their nitely generated ideals have nite Gr obner bases. We dene a
path algebra with this property for some admissible order to be a Gr obner nite
path algebra. The paper is structured such that the concepts in each section build
upon one another leading up to the climactic main result theorem 9.9 in section 9.
2 The Fundamentals of Path Algebras
Let = (
0
,
1
) be a nite directed graph. We allow for to have arrows from
a vertex to itself and multiple arrows between the same set of vertices. Through-
out the paper when referring to a graph we will assume it to be a nite directed
graphs of this type. We let
0
= v
1
, v
2
, . . . , v
N
be the set of vertices and
1
= A
1
, A
2
, . . . , A
M
be the set of arrows. Arbitrary arrows in
1
will be de-
noted by
i
, which need not equal A
i
. Extending the same notation, let
k
be the
set of paths of length k. Let =
i=0
i
be the set of paths of , of nite length.
We dene functions o :
0
and t :
0
, such that for any path p , o(p)
is the origin or rst vertex of the path p and t(p) is the terminus or nal vertex of
p. Note that, for any vertex v, o(v) = t(v) = v. For any path p, we dene l(p)
the length of p, to be the number of arrows that occur in p, counting multiplicities.
We will say that two paths intersect if they share a common vertex. We denote
p =
1
2
. . .
r
, when p is a path of length l(p) = r > 0. Whenever p is a vertex,
equivalently when l(p) = 0, we denote p = v
i
. Our convention is such that a path
2
n
is written from left to right, more precisely t(
i
) = o(
i+1
).
We dene multiplication of paths, such that for all p, q in , if t(p) = o(q),
then pq is the path adjoining p and q by concatenation. Otherwise, if t(p) ,= o(q)
then pq = 0. Given this denition, 0 is closed under multiplication. Let
K be an arbitrary eld. The path algebra, K, is dened to be the set of all nite
linear combinations of paths in with coefcients in K. Addition in K is the
usual Kvector space addition, where is a K-basis for K. Multiplication in
2
the path algebra, K, extends from the denition for multiplication of paths in
n
i=1
v
i
and K
n
i=1
v
i
is contained
in its center. Thus K acts centrally on K, so that k
k
i
p
i
=
(kk
i
)p
i
for any
k, k
i
K and p
i
. These operations make K a K-algebra.
2.1 Examples of Path Algebras
Let be the graph with n vertices and no arrows. Then K
n
i=1
K, with
componentwise multiplication.
Let be the graph
v
1
A
1
v
2
A
2
. . .
A
n1
v
n
. Then K is isomorphic to the
set of nn upper triangular matrices over K. Numbering the vertices from 1 to n
respectively, we may dene an isomorphism sending a path p to an n n matrix
with a 1 in the (o(p), t(p)) position and zeros in all other entries of the matrix.
We dene a loop to be an arrow from a vertex to itself. The path algebra K,
where is the graph with one vertex and m loops is the free algebra in m non-
commutative variables, K[x
1
, x
2
, . . . , x
m
].
In general,
0
is a full set of primitive orthogonal idempotents for K. Two
idempotents are orthogonal if their product in either order is 0. A primitive idem-
potent is an idempotent that cannot be written as a sumof two orthogonal non-zero
idempotents.
Let K
l
denote the set of all K-linear combinations of paths in of length
l. Let a K
i
and b K
j
. Then ab is the product of a K-linear combina-
tion of paths of length i times a K-linear combination of paths of length j. It
follows that ab is a K-linear combination of paths of length i + j. Therefore
K
i
K
j
K
i+j
. Thus K
i=0
K
l
is a grading of K. Given d N,
it follows that
d|l
K
l
is a subalgebra of K.
It is possible to represent every nite directed graph with n vertices as an nn
matrix over the non-negative integers such that, the number of arrows from vertex
i to vertex j is the (i, j) entry of the matrix. When there is any chance of confusion
about what graph is being discussed the matrix representation of the graph will be
given. Given a matrix representation for a graph, unless otherwise stated, we let
3
A
ij
(k) denote the k
th
arrow from vertex i to vertex j or, if the (i, j) entry is one,
we may simply write A
ij
.
3 Path Orders
We will now introduce certain orderings on , which will be relevant in develop-
ing a Gr obner basis theory for path algebras. A total ordering < requires that, for
distinct paths p and q either p < q or q < p. A well ordering < is a total ordering,
with the additional requirement that every nonempty set of paths has a least ele-
ment. Since not all well orderings are relevant to the multiplicative structure of ,
we will limit admissible orders to those that respect the multiplicative structure of
. More specically a well ordering < will be called admissible if it satises the
following two conditions:
For all p, q, r, s
(1) p < q = pr < qr when pr and qr are both nonzero and sp < sq if both sp
and sq are nonzero.
(2) p = qr, p ,= 0 = p q and p r.
For reference, in the commutative theory of Gr obner bases, admissible orders
are referred to as term orders or monomial orders. Also it is worth noting that,
when is the graph with one vertex and n loops then the product of paths is never
zero. In which case the conditions (1) and (2) may be simplied to not require
that the product of paths be nonzero.
Let <
lex
be the left lexicographic order, dened by the following. For paths p
and q if q = pp
2
. . .
r
and q =
2
. . .
r
, with
j
,
j
1
whenever there
exists i l(p), such that for all j < i,
j
=
j
and
i
<
i
we have p <
lex
q.
The left lexicographic order is NOT admissible in general. For example if is the
graph with one vertex and two loops
1
and
2
with
1
<
lex
2
. Then
2
>
lex
1
2
>
lex
1
2
>
lex
. So the set
2
,
2
1
,
2
1
, . . . has no
least element. Hence <
lex
is not a well ordering and consequently not admissible.
4
3.1 Examples of Admissible Orders:
The left length-lexicographic order: Order the vertices and the arrows.
v
1
< v
2
< . . . < v
N
< A
1
< . . . < A
M
, such that the vertices are less than the
arrows. If p and q are paths of length at least 1, then l(p) < l(q) implies p < q.
Otherwise, if l(p) = l(q) = r then p < q, whenever p <
lex
q. The right length
lexicographic order is dened similarly.
The left weight-lexicographic order: Let W :
1
N. Dene W : N,
such that W(
1
. . .
r
) =
r
i=1
W(
i
). Next, order the vertices and let W(v
i
) =
0 for all v
i
0
. Order the arrows so that
i
<
j
whenever W(
i
) < W(
j
).
Finally dene p < q, if W(p) < W(q), else if W(p) = W(q), then use the left
lexicographic order. The length lexicographic order is a special case of the weight
lexicographic order, where all the arrows are assigned the same weight.
The left weight-reverse-lex order: Dene the weight function W on the set of
paths as in denition of the left weight lexicographic order. For the ordering <
dene p < q if W(p) < W(q) or if W(p) = W(q) and p >
lex
q. Note that well
ordering does not fail since there are only a nite number of paths of any given
weight.
The total lexicographic order: Order the arrows arbitrarily A
1
< . . . < A
m
.
Also order the vertices. The vertices will be less than all paths of positive length.
Let p, q . Then p < q, if there exists i such that j < i, A
j
occurs in p
and q the same number of times, and A
i
occurs in p less than it occurs in q. If p
and q have the same number of each arrow then p <
lex
q = p < q. Another
way of thinking about this ordering is that it rst applies the lexicographic order
as if the arrows were commutative and in the case of equality it applies the non-
commutative lexicographic order.
4 Prerequisites for Gr obner Theory
Denition 4.1 For any x K the support of x denoted Supp(x) is the set of
all paths that occur in x. For example let x =
n
i=1
i
p
i
with
i
K0 and
p
i
. Then Supp(x) = p
1
, . . . , p
n
.
Denition 4.2 Given an admissible ordering <, for any nonzero x K, the tip
of x, denoted Tip(x), is the largest path in Supp(x). That is, Tip(x) Supp(x)
5
and for all p Supp(x), p Tip(x).
For reference, the equivalent of the tip in the commutative theory of Gr obner
bases is called the head or leading term.
Denition 4.3 Given X K the set p [ p = Tip(x) for some x X is
denoted as Tip(X).
Let I be an ideal in a path algebra K, with admissible ordering < on . If
p Tip(I) and q then qp, pq Tip(I) whenever qp, pq ,= 0 respectively.
Proposition 4.4 Tip(I) is a K-basis for the monomial ideal that it generates in
K.
Proof: Let p Tip(I), q , and qp ,= 0. Then there exists x I, such that
Tip(x) = p. Let x = p +
n
i=1
i
p
i
with p, p
i
and ,
i
K. Let q ,
such that t(q) = o(p). Then qx = qp +
n
i=1
i
qp
i
I. By the principles of
an admissible ordering p > p
i
for i ,= 1 implies that qp > qp
i
for qp
i
,= 0. Thus
qp = Tip(qx) Tip(I). So p Tip(I), q implies qp Tip(I). A similar
argument shows that pq Tip(I) for all q , such that t(p) = o(q).2
Corollary 4.5 Span(Tip(I)) is an ideal of K.
The set elements of that are not the tip of any element of I is denoted
NonTip(I) = Tip(I). Since = Tip(I) NonTip(I) is a K-basis for K
and Tip(I)NonTip(I) = , it follows that Span(Tip(I)) and Span(NonTip(I))
provide a direct sum decomposition of K as a K-vector space.
Proposition 4.6 Let be a graph, K a eld and < an admissible order on .
For each ideal I in K K Span(Tip(I)) Span(NonTip(I)) and K
I Span(NonTip(I)) as K-vector spaces.
Proof: Let x be a nonzero element in Span(NonTip(I))). Then Tip(x)
NonTip(I). Thus x / I. Thus I
i1
j=0
c
Tip(x
i
)
c
1
Tip(g
i
)
a
i
g
i
b
i
(). Thus, () = I. 2
Let X be a subset of K and y K. We say, y can be reduced by X, if for
some p in Supp(y), there exists x X, such that Tip(x) divides p. A reduction of
y by X is given by y kpxq, where x X, p, q and k K0, such that
kp(Tip(x))q is a term in y. A total reduction of y by X is an element, resulting
from a sequence of reductions, that cannot be further reduced by X. In general,
two total reductions of an element y need not be the same element. We say an
element y reduces to 0 by X if there is a total reduction of y by X which is 0. The
proof of proposition 5.2 also shows that every element of an ideal I reduces to
zero by a Gr obner basis (. A set X K is said to be reduced if for all x X,
x cant be reduced by Sx.
Algorithm 5.3 Given an element f K, an admissible ordering < and a set
f
1
, f
2
, . . . , f
n
, the following algorithm gives, as an output, a total reduction r
of f by f
1
, f
2
, . . . , f
n
.
Input: f, f
1
, f
2
, . . . , f
n
Output: r
r = f
reduced=n
WHILE (reduced,=0)
reduced=0
FOR (i=1:n)
IF (p = uTip(f
i
)v for some p Supp(r) and u, v )
Let p be the largest such path.
r = r c
1
Tip(f
i
)
c
p
uf
i
v
reduced++
We will denote the particular total reduction of an element f by a set S, that
the algorithm above produces, as Red
S
(f). The order < will be made apparent
by the context.
Algorithm 5.4 Given a nite generating set f
1
, f
2
, . . . , f
n
for an ideal I
K, and an admissible ordering < the following algorithm gives, as an output a
nite monic reduced generating set for I.
8
Input:f
1
, f
2
, . . . , f
n
Output: R
S = f
1
, f
2
, . . . , f
n
R =
WHILE (S ,= )
choose f S
S = Sf
S
=
R
=
WHILE (S ,= )
choose x S
S = Sx
S
= S
Red
{f}
(x)
WHILE (R ,= )
choose x T
R = Rx
IF (x = Red
{f}
(x))
R
= R
x
ELSE IF (Red
{f}
(x) ,= 0)
S
= S
Red
{f}
(x)
S = S
R = R
c
1
Tip(f)
f
We will denote the particular monic reduced generating set that this algorithm
produces from a generating set S, as R(S). The order < will be made apparent
by the context.
Proposition 5.5 Given x I and ( a Gr obner basis for I, every sequence of re-
ductions of x by ( terminates, with x totally reducing to zero after a nite number
of reductions.
Proof: Let I be an ideal in a path algebra K, with admissible ordering < and
let ( be a Gr obner basis for I. The result of reducing x I by an element of
( is another element in I. Since the only element of I that cant be reduced by
( is zero, we must show that every series of reductions of an element of I by (
terminates in a nite number of steps. Let < be an admissible ordering on the set
of paths. Then every nonempty set of paths has a least element.
9
Let S = s
1
, . . . , s
i
and T = t
1
, . . . , t
j
be nite sets of paths, ordered from
greatest to least. We say S < T provided that there exists n N, n [T[, such
that s
k
= t
k
, for all k < n and either s
n
< t
n
or [S[ = n1. Let (S
i
)
i=1
be a non-
increasing sequence of nite ordered sets of paths. Let (s
ij
)
i=1
be the sequence
of the j
th
largest elements in each of the sets S
i
, with s
ij
= 0, whenever [S
i
[ < j.
Then (s
i1
)
i=1
is a non-increasing sequence of paths and zeros and must, therefore,
stabilize to a path t
1
or to zero. Provided that (s
ij
)
i=1
stabilizes to a path t
j
, then
(s
i(j+1)
)
i=1
must be non-increasing, after (s
ij
)
i=1
stabilizes. Hence, (s
i(j+1)
)
i=1
must also stabilize to a path t
j+1
< t
j
or to zero. The sequence t
1
, t
2
, . . . is strictly
decreasing so long as t
i
is not zero. Thus (t
i
)
i=1
must stabilize to 0. It follows
that the cardinality of the sets S
i
is uniformly bounded and that (S
i
)
i=1
must also
stabilize. Let y I.
Let x be a reduction of y by g G. Then Supp(x) < Supp(y). Hence, the
supports of the elements of a sequence of reductions, of an element y I, is a
strictly decreasing sequence and must terminate after a nite number of reduc-
tions with the support of the total reduction being . Hence every sequence of
reductions must terminate at 0 in a nite number of steps and the result follows.
2
Proposition 5.6 Given an ideal I in K and an admissible order <, there is a
unique Gr obner basis (, such that ( is a reduced set and the coefcients of the
tips of the elements of ( are all 1.
Proof: Let (
1
be a Gr obner basis of an ideal I in K, with admissible order <.
We may select a subset (
2
of (
1
, such that for all p
1
, p
2
Tip((
2
) p
1
does not
divide p
2
and for all p Tip((
1
) there is a path q Tip((
2
), such that p divides
q. Then (
2
is a Gr obner basis, since Tip((
2
) generates Span(Tip(I)) and (
2
I.
Let (
3
= g
[ g
. Let g ( and
g
). Then Supp(g g
) NonTip(I) but
(g g
) I. Thus g g
= 0 and g = g
))
NonTip(I). Hence, N(x) N(x
)) NonTip(I).
It follows that, N(x)N(x
)
11
divides p. Thus, g may be reduced by g
= x S[ x K
.
Notice that for ideals I K and subgraphs
of , I
|
is an ideal in K
.
An admissible order on a set of paths remains admissible on any of its sub-
sets. Hence for any path algebra K, with admissible order <, if
is a subgraph
12
of then < remains an admissible order on K
and (
is a uniform Gr obner
bases for I
|
.
Corollary 6.6 Given a path algebra K, an admissible order <, and an ideal I
in K, let
(1)
,
(2)
, . . . ,
(n)
be the set of maximal induced subgraphs of .
Let (
(i)
be a (reduced) Gr obner basis for I
|
(i)
. Then ( =
n
i=1
(
(i)
is a (reduced)
Gr obner basis for I.
Denition 6.7 Let x K. Then the uniform decomposition of x is the set
U
x
= v
i
xv
j
[v
i
, v
j
0
.
Proposition 6.8 Let I be an ideal in K with generating set S. Then
xS
U
x
is
a uniform generating set for I.
Proof: Let y
xS
U
x
. Then y = v
i
xv
j
for some x S and vertices v
i
, v
j
.
x S implies x I, which implies y = v
i
xv
j
I. Thus
xS
U
x
) I.
Let x S. Then, x = 1x1 = (
i
v
i
)x(
j
v
j
) =
i,j
v
i
xv
j
xS
U
x
).
Therefore, I = S)
xS
U
x
). All the elements of
xS
U
x
are uniform, by
denition, and the result follows. 2
13
Proposition 6.9 An ideal I in K is nitely generated, if and only if, I
|
(i)
is
nitely generated, for each maximal induced subgraph
(i)
of .
Proof: Suppose I is a nitely generated ideal in K. Let S be a nite generating
set for I. Then U
S
=
xS
U
x
is also nite, since [U
x
[ [
0
[
2
. Let
(i)
be
a maximal induced subgraph, of . Then x U
S
[o(x), t(x)
(i)
0
is a nite
generating set, for I
|
(i)
. Let
(1)
,
(2)
, . . . ,
(n)
be the maximal induced subgraphs
of . Suppose that I
|
(1)
, I
|
(2)
, . . . , I
|
(n)
are nitely generated. Let S
1
, S
2
, . . . , S
n
be nite generating sets, for each of the restricted ideals. Let f I. Then, y U
f
implies there exists i, such that y I
|
(i)
= S
i
). So f S
1
, S
2
, . . . , S
n
) and
consequently I S
1
, S
2
, . . . , S
n
). Since S
i
I we have I = S
1
, S
2
, . . . , S
n
).
Thus I is nitely generated. 2
7 Classifying Noetherian Path Algebras
Denition 7.1 A graph is called Noetherian (not Noetherian) if its correspond-
ing path algebra is Noetherian (not Noetherian).
Proposition 7.2 A graph containing two nonidentical cycles that intersect at a
vertex is not Noetherian.
Proof: Let be a graph containing cycles C and D that intersect at a vertex v
i
.
Let c be a path of positive length along C, from v
i
to v
i
, and let d be a path of
positive length along D, from v
i
to v
i
. Then, cdc, cd
2
c, cd
3
c, . . .) is a non-nitely
generated ideal of K. So K is not Noetherian. 2
Let
=
p
1
p
2
/
.
p
3
p
4
Let be a graph containing a graph of type
(i)
, consists of two cycles A and B, with no vertices in common and m possibly
overlapping paths from A to B, allowing for A and B to be trivial cycles of one
vertex.
(i)
=
.
.
.
/
.
p
2
p
1
pm
/
.
i,j
c
ij
p
j
g
i
q
j
where g
i
(, p
j
, q
j
, and c
ij
K. Since multiples of
elements of ( sum to zero, this representation is in no way unique. Given such a
representation, for an element x, let p be the largest path in the support of any of
the elements p
j
g
i
q
j
. Let us choose a representation of x, so that p is the small-
est possible. Among the representations where p is the smallest possible, let us
choose a representation where p occurs in the least number of terms possible. It
follows that p = Tip(p
j
g
i
q
j
) for some i and j, in the representation.
16
Assume there are n > 1 pairs i, j such that p = Tip(p
j
g
i
q
j
). Let i, j and
i
, j
= p
j
, g
= g
i
and q
= q
i
. Then p = Tip(pgq) = pTip(g)q and
p = Tip(p
) = p
Tip(g
)q
).
If l(p) l(p
Tip(g
Tip(g
)s and
q
k
i
a
i
and let g
= Tip(g
) +
i
b
i
, such
that a, a
i
, b, b
i
and k, k
i
, ,
i
K.
pgq = p
Tip(g
)sgq
= p
(Tip(g
) +
b
i
)sgq
b
i
sgq
=
1
sgq
b
i
sgq
=
1
s(kTip(g) +
k
i
a
i
)q
b
i
sgq
=
k
sTip(g)q +
k
i
sa
i
q
b
i
sgq
=
k
k
i
sa
i
q
b
i
sgq
p occurs in
k
and in none of the other terms. All of the other terms only
have smaller paths occurring in them. So it follows that we may represent x so
that the largest path p occurs in less than n terms in the representation. This
contradicts the hypothesis that our representation has p occurring in a minimal
number of terms. So the assumption that p occurs in more than one term of the
representation is false. It follows that p = Tip(x) and x may be reduced.
If l(p
Tip(g
) overlap in p. So
there exists an overlap relation o(g
, g, r, s) = c
1
Tip(g
)
g
r c
1
Tip(g)
sg with r, s ,
such that p = p
s and rq = q
.
pgq = p
sgq
= c
Tip(g)
c
1
Tip(g
)
p
rq
c
Tip(g)
c
1
Tip(g
)
p
rq p
sgq
= c
Tip(g)
c
1
Tip(g
)
p
rq c
Tip(g)
p
c
1
Tip(g
)
g
r c
1
Tip(g)
sg
q
= c
Tip(g)
c
1
Tip(g
)
p
c
Tip(g)
p
o(g
, g, r, s)q
Only paths less than p occur in p
o(g
, g, r, s) may be reduced
to zero by (. So it follows that we may represent x so that the largest path p
17
occurs in less than n terms in the representation. This contradicts the hypothesis
that our representation has p occurring in a minimal number of terms. So the
assumption that p occurs in more than one term, of the representation is false. It
follows that, p = tip(x) and x may be reduced. Thus, ( reduces every element
of I to zero and ( is a Gr obner basis. 2
The Buchberger-Mora-Farkas-Green Algorithm [5] 8.2 Given a path algebra
K an admissible order <and a nite generating set f
1
, f
2
, . . . , f
m
for an ideal
I the following algorithm gives a reduced Gr obner basis for I in the limit.
Input: A generating set f
1
, f
2
, . . . , f
m
for an ideal I K
Output: A reduced Gr obner basis in the limit (
n
= g
1
, g
2
, . . .
n = 0
(
0
=
(
1
= R(f
1
, . . . , f
n
)
WHILE ((
n
,= (
n+1
)
n++
(
n
= (
n
FOR ( all pairs f, g (
n
and all overlap relations o(f, g, u, v))
(
n
= (
n
o(f, g, u, v)
(
n+1
= R((
n
)
The property that this algorithm produces a reduced Gr obner bases in the limit
will be useful in the next couple of proofs. In particular, we will use the fact that
the algorithms only performs overlap relations and reduction in order to produce a
Gr obner basis. To produce a Gr obner basis in the limit means that, if the algorithm
stops on the n
th
iteration the set (
n
is a reduced Gr obner basis. Otherwise, given a
reduced Gr obner bases (, for all x ( there exists k N, such that for all n k
we have g (
n
[ g x = g ([ g x . For a proof that the algorithm
produces a Gr obner basis in the limit, the interested reader may peruse [5].
9 Classifying Gr obner Finite Path Algebras
Denition 9.1 We will say that a path algebra is Gr obner nite if there is an
admissible order < such that all of its nitely generated ideals have nite reduced
Gr obner bases.
18
Our next goal will be to classify all Gr obner nite path algebras. Furthermore,
we will show that if a path algebra is not Gr obner nite then it contains nitely
generated ideals whose reduced Gr obner basis is not nite under any admissible
order. We show this in our main result theorem 9.9. All of the materials in this
section are either used directly in the proof of theorem 9.9 or they are implied by
the main result. Propositions 9.6 and 9.7 are direct results of theorem 9.9 and may
be omitted.
Example 9.2 Let be the graph given below.
p
_
b
/
.
q
_
/
.
b)
j
q, qc
i
bq[ j N.
Example 9.3 Let be the graph given below.
p
1
_
p
2
p
3
/
.
p
4
_
p
5
p
6
/
.
p
7
_
Let < be an admissible ordering on the paths of . If p
4
p
5
p
6
> p
3
p
2
p
4
, then
the ideal p
1
p
2
p
4
, p
4
p
5
p
6
p
3
p
2
p
4
) K has innite reduced Gr obner ba-
sis p
1
p
2
(p
3
p
2
)
i
p
4
, p
4
p
5
p
6
p
3
p
2
p
4
[ i N. Else if p
3
p
2
p
4
> p
4
p
5
p
6
, then
the ideal p
4
p
5
p
7
, p
3
p
2
p
4
p
4
p
5
p
6
) K has innite reduced Gr obner basis
p
4
(p
5
p
6
)
i
p
5
p
7
, p
3
p
2
p
4
p
4
p
5
p
6
[ i N. It follows that K has a nitely gener-
ated ideal with an innite reduced Gr obner basis, under any admissible ordering.
Example 9.4 Now let be a graph which contains two cycles P and Q, which
intersect at a vertex v. Let p be the path from v to itself along cycle P and let q be
the path from v to itself along cycle Q. If pq
2
> p
2
q, then the ideal pqpq, pq
2
p
2
q) K has Gr obner basis pqp
i
q, pq
2
p
2
q[ i N. Else if p
2
q > pq
2
,
then the ideal pqpq, p
2
q pq
2
) K has Gr obner basis pq
i
pq, p
2
q pq
2
[
i N. Therefore, any path algebra which contains two intersecting cycles, also
contains a nitely generated ideal with and innite reduced Gr obner basis under
any admissible ordering.
19
Denition 9.5 Let A be an arrow in the path algebra K. Then, for any path p,
deg
A
(p), the degree of A in p is the number of times A occurs in p. Furthermore,
for any x K, deg
A
(x) = maxdeg
A
(p)[p Supp(x)
Proposition 9.6 Let be the graph
v
1
a
v
2
b
v
3
d
0 1 0
0 1 1
0 0 1
Let < be an admissible order. Then there exists a nitely generated ideal of K
with innite reduced Gr obner basis, if and only if, there exists j > 0 such that
cd
j
> bc.
Proof: If there exists j > 0, such that cd
j
> bc, then the ideal ac, cd
j
bc) K
has innite reduced Gr obner basis ab
i
c, cd
j
bc[i N. Instead, suppose for all
j > 0 we have cd
j
< bc. Assume there exists a nitely generated ideal I K,
such that (, the reduced Gr obner basis for I is innite. By proposition 5.10, the el-
ements of ( are uniform. Let p, q be paths on such that deg
b
(p) = deg
b
(q) ,= 0.
Suppose that a occurs in neither p nor q. Then the longer of the two divides the
other. Similarly if a occurs in both p and q then the longer of the two divides the
other. Therefore, if neither p nor q divides the other, then a must occur in exactly
one of p or q. Thus every reduced set of paths, all of which have the same nonzero
degree of b, has at most 2 elements. Similarly a reduced set of paths all with de-
gree b equal to zero has at most 3 elements , a, cd
i
, d
j
, with j > i. Consequently
given a reduced set of paths S, [S[ 2sup
pS
deg
b
p + 3 (Here sup represents
the supremum not the support). It follows that the degree of b for Tip(() must
be unbounded in order for ( to be innite. Consequently, the degree of b in the
elements of ( must be unbounded.
Let x be a uniform element in K. Suppose o(x) = t(x) = v
2
. Then
x =
n
i=1
k
i
b
i
, for some k
i
K, with k
n
,= 0 and deg
b
(x) = deg
b
(Tip(x)) = n.
Suppose o(x) = v
1
and t(x) = v
2
. Then x =
n
i=1
k
i
ab
i
and deg
b
(x) =
deg
b
(Tip(x)) = n. Suppose o(x) = v
2
and t(x) = v
3
and deg
b
(x) = n, deg
d
(x) =
m. Let x =
n
i=0
m
i=0
k
ij
b
i
cd
j
, with k
ij
K. Let h be maximal such that
k
nh
,= 0. Assume b
n
cd
h
,= Tip(x) = b
i
cd
j
. Well ordering implies b
n
cd
h
does
not divide Tip(x) = b
i
cd
j
. This implies j > h. Since h was maximal i < n.
Thus by well ordering b
ni
c < cd
jh
. Furthermore bc < cd
jh
, which contradicts
20
our hypothesis. Thus b
n
cd
h
= Tip(x) and deg
b
(Tip(x)) = deg
b
(x) = n, Sup-
pose o(x) = v
1
and t(x) = v
3
. Then there exists a uniform element x
, such that
o(x
) = v
2
and t(x
) = v
3
and x = ax
. Then deg
b
(Tip(x)) = deg
b
(Tip(x
)) =
deg
b
(x
) = deg
b
(x). In all other cases deg
b
(x) = deg
b
(Tip(x)) = 0. Thus, for
all x K, deg
b
(x) = deg
b
(Tip(x)). It follows that reducing an element cant
increase the deg
b
of that element. If there exists g ( such that o(g) = t(g) = v
2
,
then deg
b
(() = deg
b
(g). So there exists no such element. If there exists g (
such that o(g) = v
1
and t(g) = v
2
, then deg
b
(v
1
() = deg
b
(g) and (v
1
( is
contained within a Noetherian subgraph and is therefore nite. Thus ( is nite, a
contradiction.
So we may assume that (v
2
= . Thus, for uniform x I either x = v
1
or t(x) = v
3
. Let (
n
be the set generated by the n
th
iteration of algorithm
8.2 starting with a nite generating set for I. Then and (
n
v
2
= . In order
for deg
b
(() to be unbounded there must exist k and g
(
k+1
such that for all
g (
k
, we have deg
b
(g
) > deg
b
(g). Since reductions do not increase the deg
b
of an element it must be that g
) deg
b
(o(f, g, u, v)) deg
b
(fv + o(f, g, u, v)) = deg
b
(fv) = deg
b
(f).
This contradicts that deg
b
(() is unbounded and the result follows. 2
Proposition 9.7 Let K be a path algebra such that every path on intersects
at most 1 cycle. Then, given any admissible order, every nitely generated ideal
I K has a nite Gr obner basis.
Proof: Let be a graph with n cycles C
i
, such that every path on intersects at
most 1 cycle, and let < be an admissible order. Note that there are graphs of this
type that are not Noetherian. Let I be a nitely generated ideal in K. Let ( be
the reduced Gr obner basis for I. Let (
0
be a nite uniform generating set for I.
Assume ( is innite. For all n N the set of paths in with deg
C
i
less than n for
all i is nite. So we may assume there exists i such that deg
C
i
of the elements of
( is unbounded.
Note that performing overlap relations between uniform elements and reduc-
ing uniform elements by uniform elements, produces uniform elements. Conse-
quently, at any stage in the algorithm, (
n
is a uniform set. Let S
i
be the nite set
21
of paths that either begin or end on C
i
but do not traverse C
i
at all. Let s S
i
with o(s) C
i
and let p = as and q = bs with a and b on C
i
. Without loss of
generality we may assume a < b and consequently a divides b and p divides q. It
follows that every reduced set of paths that all begin or end on C
i
is nite. Thus
v
i
(v
j
is nite whenever v
i
or v
j
is on one of the cycles.
Consequently, there exists v
i
and v
j
not on any cycle, such that v
i
(v
j
is in-
nite. We say a vertex v
k
is between v
i
and v
j
if there is a path p, such that o(p) = v
i
and t(p) = v
j
with v
k
a vertex on p, v
k
,= v
i
and v
k
,= v
j
. There must exist at least
one pair v
i
, v
j
not on any cycle with v
i
(v
j
innite, such that, for all v
k
between
v
i
and v
j
, both v
i
(v
k
and v
k
(v
i
are nite. Starting with a nite generating set for
I we may reach a point in algorithm 8.2 such that v
i
(v
k
(
n
and v
k
(v
i
(
n
for all v
k
between v
i
and v
j
, then there are only a nite number of element in (
n
whose overlap relations are in v
i
Kv
j
. After these overlap relations have been
reduced no other elements in v
i
Kv
j
will be produced by algorithm 8.2. Thus
v
i
(v
j
must be nite. This contradicts the assumption that ( was innite. 2
Proposition 9.8 Suppose is a graph with no intersecting cycles and does not
contain a graph of the form
=
p
1
_
p
2
p
3
/
.
p
4
_
p
5
p
6
/
.
p
7
_
as a subgraph, such that p
1
, p
2
, p
4
, p
5
, p
7
are paths of positive length and p
3
, p
6
are paths of possibly length zero. Then a nonempty uniform subgraph (v
i
, v
j
),
with v
i
,= v
j
consists of a cycle A, n
i
N paths p
ij
from A to a cycle B
i
, m
i
N
paths q
ij
from cycle B
i
to a cycle C, for i, t N such that j < t and s N 0
paths h
i
from A to C. It is possible for either cycle A or C to be trivial cycles of
one vertex. The p
ij
may intersect and overlap one another, the q
ij
may intersect
and overlap one another and the h
i
may intersect one another and as a group they
may rst intersect the p
ij
and then the q
ij
. These restrictions on the intersections
of the p
ij
, q
ij
and h
i
entail that no path on (v
i
, v
j
) may overlap more than one of
the B
i
.
Proof: Supposing a graph doesnt contain any intersecting cycles then every
path p on must be contained within a subgraph
C
1
/
.
q
1
_
C
2
/
.
q
2
_
q
n1
_
Cn
/
.
22
Since does not contain the graph
=
p
1
_
p
2
p
3
/
.
p
4
_
p
5
p
6
/
.
p
7
_
p can travel on at most 2 of the paths q
i
, it follows that either p is contained within
the subgraph
p
, such that either
p
=
/
.
p
ij
_
B
i
/
.
q
ik
_
/
.
or
p
=
/
.
h
i
_
/
.
pv
i
v
j
p
. The
cycles B
i
and B
j
do not intersect for i ,= j since contained no intersecting
cycles. Furthermore, the paths p
ij
may overlap and the paths q
ij
may overlap. If
the p
ij
, B
i
or q
ij
shared any vertices with another not of their type then the graph
would have to contain a graph of the form
/
.
q
_
/
.
bq < qc
i
for some i.
(3) contains a subgraph of the form
/
.
p
_
b
/
.
q
_
23
where l(b) may be zero, with pbb
< a
i
p for some i.
Proof: Let K be a path algebra and < an admissible order. We have already
shown in examples 9.2 and 9.4 that if and < satisfy any of (1), (2), or (3) then
K contains a nitely generated ideal with an innite reduced Gr obner basis. So
we may assume that none of conditions (1), (2), or (3) are satised. Example 9.3
shows that if were to have a graph of the form
=
p
1
_
p
2
p
3
/
.
p
4
_
p
5
p
6
/
.
p
7
_
as a subgraph then it would have to satisfy either (2) or (3). So does not have
any graph of the form
ijk
be the path along B
i
such that b
ijk
b
ijk
completes one cycle around B
i
. Since
(2) and (3) are not satised we have p
ij
b
ijk
b
ijk
> a
h
ij
p
ij
and b
ijk
b
ijk
q
ik
> q
ik
c
h
ik
for all i, j, k and h.
We may choose (v
i
, v
j
) to be minimal in the sense that for all (v
h
, v
k
)
strictly contained in (v
i
, v
j
), we have (
|
(v
i
,v
j
)
is innite but (
|
(v
h
,v
k
)
is nite.
Thus v
h
(v
k
is nite whenever v
h
and v
k
are not both on A and C respectively and
(
v
i
A
v
i
)((
v
j
C
v
j
) is innite. Let
be the subgraph of (v
i
, v
j
) containing
cycles A, C and only the paths h
i
from A to C that do not intersect any of the B
i
.
Since
v
i
A
v
i
)Tip(()(
v
j
C
v
j
) containing only those paths which do not
intersect some cycle B
i
is nite.
We may partition the paths starting on A and ending on C that intersect some
cycle B
i
into sets, S
pq
= ap(bb
)
h
bqc = p(bb
)
h
b
1
qc[ h N 0, a and c
are any path on A and C respectively with t(a) = o(p) and o(c) = t(q) where
for some i, j, k we have p = p
ij
, q = q
ik
, b = b
ijk
, and b
= b
ijk
. There are only
nitely many such sets. Therefore, there exists S
pq
, such that Tip(() S
pq
is
innite.
Let 1 be a relation on the sets S
pq
, such that S
pq
1S
p
q
whenever for all
x S
pq
there exist y S
p
q
, such that x < y. Suppose we dont have S
pq
1S
p
then exists x S
pq
such that x > y for all y S
p
q
, which implies S
p
q
1S
pq
.
It follows that for all pairs of sets S
pq
and S
p
q
at least one of S
pq
1S
p
q
and
S
p
q
1S
pq
is true. Suppose that S
pq
1S
p
q
and S
p
q
1S
p
q
. Then for all x S
pq
there exist y S
p
q
, such that x < y and there exists z S
p
q
with x < y < z.
Thus S
pq
1S
p
q
. These properties of the relation 1 allow us to re-label the sets
S
pq
as the sets P
hk
with h, k N, in such a way that P
hk
1P
h
k
if and only if
h h
. Let T
h
=
k
P
hk
. Then there exists h such that Tip(() T
h
is innite.
Let be largest integer, such that Tip(() T
is innite.
Let H be the nite subset of (
v
i
A
v
i
)((
v
j
C
v
j
) consisting of those
elements whose tips dont intersect any of the cycles B
j
. Starting with a nite
generating set let (
n
be the set dened in the n
th
iteration of algorithm 8.2.
We may reach a the point in algorithm 8.2, such that for all v
h
and v
k
not
both on A and C we have v
h
(v
k
(
n
, H (
n
, and g ([ Tip(g) T
h
for h > (
n
. Furthermore let us choose n large enough so that all the
overlap relations o(f, g, u, v) with f, g H (
v
i
/ A or v
k
/ C
v
i
(v
k
) g ([
Tip(g) T
h
for h > reduce to zero by (
n
.
It follows that if an overlap relation o(f, g, u, v) (
v
i
A
v
i
)K(
v
j
C
v
j
)
does not reduce to 0, with f, g (
n
, then either Tip(g) T
h
or Tip(f) T
h
for
some h . Let x be the largest path in (
n
(
h
T
h
). Since for all h
we have P
hj
1P
1
there exists x
n
P
1
such that x
n
> x. Let x
n
= a
n
p(bb
bqc
n
.
For f, g (
n
let y ,= 0 be the element that o(f, g, u, v)
(
v
i
A
v
i
)K(
v
j
C
v
j
) reduces to when the set (
n
, of elements from (
n
25
union their overlaps, is made into the reduced set (
n+1
at the end of the n
th
iteration of algorithm 8.2. Specically, we are referring to the line in algorithm
8.2, (
n+1
= R((
n
), where algorithm 8.2 calls algorithm 5.4. In general for a gen-
erating set S every element in the reduced set R(S) is the reduction of an element
in S. If Tip(g) P
hj
for some j and h then Tip(y) Tip(o(f, g, u, v)) <
vTip(g) < va
a
n
p(bb
bqc
n
= a
p(bb
bqc
n
, where a
bqc
n
c
u = a
n
p(bb
bqc
,
where c
s and c
bqc
n+1
. By the hypothesis
on our order a
n+1
p(bb
< p(bb
)
+1
and bqc
n+1
< (bb
bqc
n+1
< p(bb
)
+2
bq.
We conclude, by induction, that for all g ( with Tip(g) T
k
for k h, we
have Tip(g) < p(bb
)
+2
bq.
We see that for all paths p
Tip(() T
h
, with h we have
p
< p(bb
)
+2
bq T
h
. Let T
h
=
s
j=1
P
hj
=
s
j=1
a
j
p
j
(b
j
b
j
)
k
b
j
q
j
c
j
, Then
for all j there exists k
j
, such that p
j
(b
j
b
j
)
k
j
b
j
q
j
> p(bb
)
+2
bq. Any reduced set
of paths in the set a
j
p
j
(b
j
b
j
)
t
b
j
q
j
c
j
[ p
j
(b
j1
b
j2
)
t
b
j1
q
j
is a xed path is nite. It
follows that any reduced set of paths in T
h
all of which are less than p(bb
)
+2
bq
is a nite union of reduced set of paths from the sets a
j
p
j
(b
j
b
j
)
t
b
j
q
j
c
j
[
p
j
(b
j
b
j2
)
t
b
j1
q
j
is a xed path. It follows that Tip(() T
is nite. This
contradicts the hypothesis that (
v
i
A
v
i
)((
v
j
C
v
j
) is innite. Thus (
|
(v
i
,v
j
)
is nite, and by corollary 6.6, ( is nite.2
We will now show an example of an admissible order on the path algebras
dened in proposition 9.8, which meets the criteria of the ordering described in
the theorem 9.9. Since such an order exists this implies the path algebras dened
in proposition 9.8 are all Gr obner nite.
The left dual-weighted-lexicographic order: Let be a graph of the form
described in proposition 9.8. Let
1
be the set of arrows which are contained
within cycles in the graph which do not have both an arrow entering them
and an arrow coming out of them. Fix a set of positive integers n
[ N
Let W :
1
N N, such that for
1
we have W() = (0, n
) and
26
for
1
we have W() = (n
r
i=1
W(
i
), with componentwise addition. Order the set NN
so that (n, m) < (n
, m
) whenever n < n
or when n = n
and m < m
. Next,
order the vertices and let W(v
i
) = 0 for all v
i
0
. Order the arrows so that
i
<
j
whenever W(
i
) < W(
j
). Finally dene p < q, if W(p) < W(q), else
if W(p) = W(q), then use the left lexicographic order.
In our example above, the set N N is given a well order and there are
only a nite number of paths with given weight (n, n
k
, 2
(, 7
c
Tip(x)
, 7
c
px
, 7
deg
A
(), 20
l(), 2
o(), 2, 11
o(, , , ), 16
t(), 2, 11
degree, 20
divides, 7
Gr obner basis, 7
reduced, 11
theory, 1
Gr obner nite, 18
graph, 2
induced subgraph, 12
maximal, 12
intersect, 2
length, 2
loop, 3
matrix representation, 3
Noetherian graph, 14
normal form, 7
order
admissible, 4
left lexicographic, 4
on a subgraph, 12
well, 4
origin
path, 2
uniform element, 11
overlap relation, 16
path, 2
path algebra, 2
reduced, 8
reduction, 8
restriction, 12
support, 5
terminus
path, 2
uniform element, 11
tip, 5
total reduction, 8
uniform
decomposition, 13
element, 11
29