See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/224825389

Leavitt Path Algebras of Finite Gelfand-Kirillov Dimension

Article in Journal of Algebra and Its Applications · April 2012

DOI: 10.1142/S0219498812502258 · Source: arXiv

CITATIONS READS
72 411

4 authors:

A. Alahmadi Hamid Al-Sulami

King Abdulaziz University King Abdulaziz University
151 PUBLICATIONS 944 CITATIONS 92 PUBLICATIONS 1,605 CITATIONS

SEE PROFILE SEE PROFILE

S. K. Jain Efim Isaakovich Zelmanov

Ohio University University of California, San Diego
82 PUBLICATIONS 710 CITATIONS 68 PUBLICATIONS 1,281 CITATIONS

SEE PROFILE SEE PROFILE

All content following this page was uploaded by A. Alahmadi on 03 June 2014.

The user has requested enhancement of the downloaded file.

 LEAVITT PATH ALGEBRAS OF FINITE GELFAND-KIRILLOV
DIMENSION
arXiv:1204.5258v1 [math.RA] 24 Apr 2012

ADEL ALAHMADI(1) , HAMED ALSULAMI(1) , S. K. JAIN(1,2) , AND EFIM ZELMANOV(1,3)

Abstract. Groebner-Shirshov basis and Gelfand-Kirillov dimension of the

Leavitt path algebra are derived.

1. Introduction.
Leavitt path algebras were introduced in [AA] as algebraic analogs of graph
Cuntz-Kreiger C*-algebras. Since then they have received significant attention
from algebraists. In this paper we (i) find a Groebner-Shirshov basis of a Leavitt
path algebra, (ii) determine necessary and sufficient conditions for polynomially
bounded growth, and (iii) find Gelfand-Kirillov dimension.

2. Definitions and Terminologies

A (directed) graph Γ = (V, E, s, r) consists of two sets V and E, called vertices

and edges respectively, and two maps s, r : E → V . The vertices s(e) and r(e)
are referred to as the source and the range of the edge e, respectively. The graph
is called row-finite if for all vertices v ∈ V , |(s−1 (v))| < ∞. A vertex v for which
(s−1 (v)) is empty is called a sink. A path p = e1 ...en in a graph Γ is a sequence of
edges e1 ...en such that r(ei ) = s(ei+1 ) for i = 1, ..., (n − 1). In this case we say that
the path p starts at the vertex s(e1 ) and ends at the vertex r(en ). If s(e1 ) = r(en ),
then the path is closed. If p = e1 ...en is a closed path and the vertices s(e1 ), .., s(en )
are distinct, then the subgraph (s(e1 ), ..., s(en ); e1 , ..., en ) of the graph Γ is called
a cycle.
Let Γ be a row-finite graph and let F be a field. The Leavitt path F -algebra
L(Γ) is the F -algebra presented by the set of generators {v, v ∈ V }, {e, e∗ | e ∈ E}
and the set of relators (1) vi vj = δ vi ,vj vi for all vi , vj ∈ V ; (2) s(e)e = er(e) = e,
∗
r(e)eP = e∗ s(e) = e∗ , for all e ∈ E; (3) e∗ f = δ e,f r(e), for all e, f ∈ E; (4)
v = s(e)=v ee∗ , for an arbitrary vertex v ∈ V \{sinks}.
The mapping which sends v to v, for v ∈ V, e to e∗ and e∗ to e, for e ∈ E, extends
to an involution of the algebra L(Γ). If p = e1 ...en is a path, then p∗ = e∗n ...e∗1 .

3. A Basis of L(Γ)

For an arbitrary vertex v which is not a sink, choose an edge γ(v) such that
s(γ(v)) = v. We will refer to this edge as special. In other words, we fix a function
γ : V \{sinks} → E such that s(γ(v)) = v for an arbitrary v ∈ V \{sinks}.

Key words and phrases. Leavitt Path Algebra. Cuntz-Krieger C*-Algebras. Groebner-
Shirshov Basis. Polynomially Bounded Growth. Gelfand-Kirillov Dimension.
1
2 ADEL ALAHMADI(1) , HAMED ALSULAMI(1) , S. K. JAIN(1,2) , AND EFIM ZELMANOV(1,3)

Theorem 1. The following elements form a basis of the Leavitt path algebra L(Γ) :
(i) v, v ∈ V , (ii) p, p∗ ,where p is a path in Γ, (iii) pq ∗ , where p = e1 ...en ,
q = f1 ...fm , ei , fj ∈ E, are paths that end at the same vertex r(en ) = r(fm ), with
the condition that the last edges en and fm are either distinct or equal, but not
special.

Proof. Recall that a well-ordering on a set is a total order (that is, any two elements
can be ordered) such that every non-empty subset of elements has a least element.
As a first step, we will introduce a certain well-ordering on the set of generators
X = V ∪ E ∪ E ∗ . Choose an arbitrary well-ordering on the set of vertices V.
If e, f are edges and s(e) < s(f ) then e < f. It remains to order edges that
have the same source. Let v be a vertex which is not a sink. Let e1 , ..., ek be
all the edges that originate from v. Suppose ek = γ(v). We order the edges as
follows: e1 < e2 < .... < ek = γ(v). Choose an arbitrary well-ordering on the set
E ∗ . For arbitrary elements v ∈ V, e ∈ E, f ∗ ∈ E ∗ , we let v < e < f ∗ . Thus
the set X = V ∪ E ∪ E ∗ is well-ordered. Let X ∗ be the set of all words in the
alphabet X. The length-lex order (see [B, Be]) makes X ∗ a well-ordered set. For
all v ∈ V and e ∈ E, we extend the set of relators (1) - (4) by (5): ve = 0, for
v 6= s(e); ev = 0, for v 6= r(e); ve∗ = 0, for v 6= r(e); e∗ v = 0, for v 6= s(e). The
straightforward computations show that the set of relators (1) - (5) is closed with
respect to compositions (see [B, BE]). By the Composition-Diamond Lemma ([B,
BE]) the set of irreducible words (not containing the leading monomials of relators
(1) - (5) as subwords) is a basis of L(Γ). This completes the proof. 

4. Leavitt path algebras of polynomial growth

Recall some general facts on the growth of algebras. Let A be an algebra (not
necessarily unital), which is generated by a finite dimensional subspace V . Let
V k denote the span of all products v1 · ·· vk , vi ∈ V, k ≤ n. Then V = V 1 ⊂
V 2 ⊂ · · ·, A = ∪n≥1 V n and gV (n) = dim V n < ∞. Given the functions f, g
from N = {1, 2, ...} to the positive real numbers R+ , we say that f 4 g if there
exists c ∈ N such that f (n) ≤ cg(cn) for all n. If f 4 g and g 4 f then the
functions f, g are said to be asymptotically equivalent, and we write f ∼ g. If W is
another finite dimensional subspace that generates A, then gV (n) ∼ gW (n) . If gV (n)
is polynomially bounded, then we define the Gelfand-Kirillov dimension of A as
ln g
GKdim A = lim supn→∞ lnVn(n) . The definition of GK-dimension does not depend
on a choice of the generating space V as long as dim V < ∞. If the growth of A is
not polynomially bounded, then GKdim A = ∞.
We now focus on finitely generated algebras and we will assume that the graph
Γ is finite. Let C1 , C2 be distinct cycles such that V (C1 ) ∩ V (C2 ) 6= φ. Then we can
renumber the vertices so that C1 = (v1 , · · ·, vm ; e1 , · · ·, em ), C2 = (w1 , · · ·, wn ; f1 , · ·
·, fn ), v1 = w1 . Let p = e1 · · · em , and q = f1 · · · fn .
Lemma 2. The elements p, q generate a free subalgebra in L(Γ).
Proof. By Theorem 1, different paths viewed as elements of L(Γ) are linearly inde-
pendent. If u1 , u2 are different words in two variables, then u1 (p, q) and u2 (p, q) are
different paths. Indeed, cutting out a possible common beginning, we can assume
that u1 , u2 start with different letters, say, u1 (p, q) = p · ··, u2 (p, q) = q · · · . If
 LEAVITT PATH ALGEBRAS OF FINITE GELFAND-KIRILLOV DIMENSION 3

m > n then the path u2 (p, q) returns to the vertex v1 at the n-th step, whereas
u1 (p, q) does not. If m = n, then the left segments of length m of u1 (p, q), u2 (p, q)
are different. This proves the lemma. 
Corollary 3. If two distinct cycles have a common vertex, then L(Γ) has exponen-
tial growth.
From now on we will assume that any two distinct cycles of the graph Γ do not
have a common vertex.
′′ ′′
For two cycles C ′ , C , we write C ′ =⇒ C , if there exists a path that starts in
′′
C ′ and ends in C .
′′ ′′ ′′
Lemma 4. If C ′ , C are two cycles such that C ′ =⇒ C , and C =⇒ C ′ , then
′′
C′ = C .
′′
Proof. Choose a path p that starts in C ′ and ends in C . Similarly, choose a
′′ ′′
path q that starts in C and finishes in C ′ . There exists also a path p′ on C ,
′ ′
which connects r(p) with s(q) and a path q on C , which connects r(q) with s(p).
′′
Now, pp′ qq ′ is a closed path, which visits both C ′ and C . Let t be a closed path
′′
with this property (visiting both C ′ and C ) having a minimal length. Write t =
e1 · · · en , ei ∈ E. We claim that the vertices s(e1 ), · · ·, s(en ) are all distinct, thus
t = (s(e1 ), · · ·, s(en ); e1 , · · ·, en ) is a cycle. Assuming the contrary, let s(ei ) = s(ej ),
1 ≤ i < j ≤ n, and j − i is minimal with this property. Then t′ = (s(ei ), s(ei+1 ), · ·
·, s(ej ); ei , ei+1 , · · ·, ej−1 ) is a cycle. Let us ”cut it out”, that is, consider the path
′′ ′′
t = e1 · ·· ei−1 ej · ·· en . This path is shorter than t. Hence t can not visit both
′′ ′′
C ′ and C . Suppose that t does not visit C ′ . Then at least one of the vertices
s(ei ), · · ·, s(ej−1 ) lies in C ′ . Since two intersecting cycles coincide, it implies that
′′
t′ = C ′ , hence s(ej ) lies in C ′ . This contradicts our assumption that t does not
′′
visit C ′ . Hence t = C ′ = C . This proves the lemma. 
A sequence of distinct cycles C1 , · · ·, Ck is a chain of length k if C1 =⇒ · · · =⇒
Ck . The chain is said to have an exit if the cycle Ck has an exit (see [AA]), that is,
if there exists an edge e such that s(e) ∈ V (Ck ), but e does not belong to Ck .Let d1
be the maximal length of a chain of cycles in Γ, and let d2 be the maximal length
of chain of cycles with an exit. Clearly, d2 ≤ d1 .
Theorem 5. Let Γ be a finite graph.
(1) The Leavitt path algebra L(Γ) has polynomially bounded growth if and only
if any two distinct cycles of Γ do not have a common vertex;
(2) If d1 is the maximal length of a chain of cycles in Γ, and d2 is the maximal
length of chain of cycles with an exit, then GK dim L(Γ) = max(2d1 − 1, 2d2 ).
Proof. As in the proof of Theorem 1 we consider the generating set X = V ∪ E ∪ E ∗
of L(Γ). Let E ′ be the set of edges that do not belong to any cycle. Let P ′ be the
′
set of all paths that are composed from edges from E . Then an arbitrary path
from P ′ never arrives to the same vertex twice. Hence, |P ′ | < ∞.
By Theorem 1 the space Span(X n ) is spanned by elements of the following types:
(1) a vertex,
(2) a path p = p′1 p1 p′2 p2 · · · pk p′k+1 , where pi is a path on a cycle Ci , 1 ≤ i ≤ k,
C1 =⇒ · · · =⇒ Ck is a chain, p′i ∈ P ′ , length(p) ≤ n,
(3) p∗ , where p is a path of the type (2),
4 ADEL ALAHMADI(1) , HAMED ALSULAMI(1) , S. K. JAIN(1,2) , AND EFIM ZELMANOV(1,3)

(4) pq ∗ , where p = p′1 p1 p′2 p2 · · · pk p′k+1 , q = q1′ q1 q2′ · · · qs qs+1 ′

; pi , qj are paths on
cycles Ci , Dj respectively and C1 =⇒ · · · =⇒ Ck , D1 =⇒ · · · =⇒ Ds are chains ;
p′i , qj′ ∈ P ′ , length(p) + length(q) ≤ n with r(p) = r(q). We will further subdivide
this case into two subcases:
(4.1) r(p) ∈ / V (Ck ) ∪ V (Ds );
(4.2) r(p) ∈ V (Ck ) ∪ V (Ds ).
We will estimate the number of products of length ≤ n in each of the above
cases and then use the following elementary fact:
Let (an )n∈N be the sum of s sequences (ain )n∈N , 1 ≤ i ≤ s, ain > 0. Then
lim supn→∞ lnln ann = max(lim supn→∞ lnln anin , 1 ≤ i ≤ s)
Let us estimate the number of paths of the type (2). Fix a chain C1 =⇒ C2
=⇒ ... =⇒ Ck . If C = (v1 , ...., vm ; e1 , ..., em ) is a cycle, let PC = e1 ...em . For a
given cycle there are m such paths depending upon the choice of the starting point
v1 .
Let |(V (Ci )| = mi and let PCi be any one of the mi paths described above. Then
an arbitrary path on Ci can be represented as u′ Pil u′′ , where length (u′ ), length
′′
(u ) ≤ mi − 1. Hence every path of the type (2) which corresponds to the chain C1
=⇒ C2 =⇒ ... =⇒ Ck can be represented as p′1 u′1 PCl11 u′′1 ...p′k u′k PCk lk ′′ ′
uk pk+1 , where
′ ′ ′ ′′
pi ∈ Pi and length (ui ), length (ui ) ≤ mi − 1. Clearly, m1 l1 + ... + mk lk ≤ n. This
implies that the number of such paths 4 nk ≤ nd1 . On the other hand, choosing
a chain C1 =⇒ C2 =⇒ ... =⇒ Cd1 of length d1 , we can construct ∼ nd1 paths of
length ≤ n. The case (3) is similar to the case (2).
Consider now the elements of length ≤ n of the type pq ∗ , r(p) = r(q); the path
p passes through the cycles of the chain C1 =⇒ C2 =⇒ ... =⇒ Ck on the way, the
path q passes through the cycles of the chain D1 =⇒ D2 =⇒ ... =⇒ Ds on the
way and so p = p′1 p1 p′2 ...pk p′k+1 , where p′i ∈ Pi′ , each pi is a path on the cycle Ci .
Similarly, q = q1′ q1 q2′ ...qs qs+1 ′
. Arguing as above, we see that for fixed chains C1
=⇒ C2 =⇒ ... =⇒ Ck and D1 =⇒ D2 =⇒ ... =⇒ Ds , the number of such paths
4 nk+s .
Suppose that the vertex v = r(p) = r(q) does not lie in V (Ck ) ∪ V (Ds ) . Then
both cycles Ck and Ds have exits. Hence the number of paths of type (4.1) is
≤ n2d2 . On the other hand, let C1 =⇒ C2 =⇒ ... =⇒ Cd2 be a chain and let e
be an exit of the cycle Cd2 . Select paths p′2 , ...,p′d2 , where p′i connects Ci−1 to Ci ,
p′i ∈ P ′ .
Select a path u′′1 on the cycle C1 which connects r(PC1 ) to s(p′2 ) , a path u′2
in C2 which connects r(p′2 ) to s(PC2 ), a path u′′2 on C2 which connects r(PC2 ) to
s(P3′ ), and so on. The path u′′d2 connects r(PCd2 ) to s(e).
Among the edges from s−1 (s(e)) choose a special one γ(s(e)) different from e.
Then by Theorem 1, the elements
l
PCl11 u′′1 p′2 u′2 PC2
l2 ′′ ′
u2 p3 ...PCdd2 u′′d3 ee∗ (u′′2 )∗ (PCd∗ )ld2 +1 ...(u′′1 )∗ (PC1 )l2d2 , li ≥ 1, (A),
2 2
are linearly independent. Let m be the total length of all elements other than
li
PC i , (PCi∗ )l2d2 −i+1 . The number of elements in the inequality (A) above is the
number of nonnegative integral solutions of the inequality
Pd2 2d2
i=1 mi (li + l2d2 −i+1 ) ≤ n − m, which is ∼ n .

Now suppose that the vertex v = r(p) = r(q) lies in Ck . Assume at first that
Ck 6= Ds . Then the chain D1 =⇒ D2 =⇒ ... =⇒ Ds has an exit. If k ≤ s, then the
number of the paths of this type is ≤ nk+s ≤ n2d2 .
 LEAVITT PATH ALGEBRAS OF FINITE GELFAND-KIRILLOV DIMENSION 5

If s < k, then nk+s ≤ n2k−1 ≤ n2d1 − 1.

Next, let Ck = Ds . It means that the paths p′k+1 , qs+1 ′
are empty; pk and qs are
both paths on the cycle Ck and in this case we have,
(i) pk qs∗ = u, if pk = uqs , is a path on Ck ,
(ii)pk qs∗ = u∗ , if qs = upk , is a path on Ck ,and
(iii) pk qs∗ = 0, otherwise.
The number of such elements pq ∗ is 4 nk+s−1 ≤ n2d1 − 1.
On the other hand, let C1 =⇒ C2 =⇒ ... =⇒ Cd2 be a chain of cycles. Se-
lect paths p′2 , ..., p′d1 ∈ P ′ , p′i connects Ci−1 to Ci ; u′i , u′′i are paths on the cy-
cle Ci such that PC1 u′′1 p′2 u′2 PC2 u′′2 p′3 ...PCd1 6= 0. By Theorem 1, the elements
l
PCl11 u′′1 p′2 u′2 PCl22 u′′2 p′3 ...u′d1 PCdd1 (u′d1 )∗ (PC∗d −1 )ld1 +1 ...(PC∗1 )l2d1 −1 are linearly indepen-
1 1
dent provided that li ≥ 1, 1 ≤ i ≤ 2d1 − 1. The number of these elements is ∼
n2d1 −1. This proves Theorem 2. 
References.
[AA] G. Abrams, G. Aranda Pino, The Leavitt path algebra of a graph, J.
Algebra 293 (2005), 319-334
[B] L.A. Bokut, Imbeddings in simple associative algebras, Algebra i Logica, 15
(1976), 117-142.
[Be] G.M. Bergman, The Diamond Lemma for ring theory, Adv. in Math. 29
(1978), 178-218.
[KL] G.R.Krause, T.H. Lenagan, Growth of Algebras and Gelfand-Kinillow Di-
mension, revised edition. Graduate studies in Mathematics, 22, AMS, Providence,
RI, 2000
1. Department of Mathematics, King Abdulaziz University, Jeddah, SA, 2.Depart-
ment of Mathematics, Ohio University, USA, 3 Department of Mathematicsw, Univer-
sity of California, San Diego, Partially supported by the NSF

View publication stats