Q-system Cluster Algebras, Paths and Total Positivity
In the first part of this paper, we provide a concise review of our method of solution of the Ar Q-systems in terms of the partition function of paths on a weighted graph. In the second part, we show that it is possible to modify the graphs and transfer matrices so as to provide an explicit connecti...
Gespeichert in:
| Datum: | 2010 |
|---|---|
| Hauptverfasser: | , |
| Format: | Artikel |
| Sprache: | English |
| Veröffentlicht: |
Інститут математики НАН України
2010
|
| Schriftenreihe: | Symmetry, Integrability and Geometry: Methods and Applications |
| Online Zugang: | http://dspace.nbuv.gov.ua/handle/123456789/146152 |
| Tags: |
Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
|
| Назва журналу: | Digital Library of Periodicals of National Academy of Sciences of Ukraine |
| Zitieren: | Q-system Cluster Algebras, Paths and Total Positivity / P. di Francesco, R. Kedem // Symmetry, Integrability and Geometry: Methods and Applications. — 2010. — Т. 6. — Бібліогр.: 28 назв. — англ. |
Institution
Digital Library of Periodicals of National Academy of Sciences of Ukraine| id |
irk-123456789-146152 |
|---|---|
| record_format |
dspace |
| spelling |
irk-123456789-1461522019-02-08T01:23:23Z Q-system Cluster Algebras, Paths and Total Positivity di Francesko, P. Kedem, R. In the first part of this paper, we provide a concise review of our method of solution of the Ar Q-systems in terms of the partition function of paths on a weighted graph. In the second part, we show that it is possible to modify the graphs and transfer matrices so as to provide an explicit connection to the theory of planar networks introduced in the context of totally positive matrices by Fomin and Zelevinsky. As an illustration of the further generality of our method, we apply it to give a simple solution for the rank 2 affine cluster algebras studied by Caldero and Zelevinsky. 2010 Article Q-system Cluster Algebras, Paths and Total Positivity / P. di Francesco, R. Kedem // Symmetry, Integrability and Geometry: Methods and Applications. — 2010. — Т. 6. — Бібліогр.: 28 назв. — англ. 1815-0659 2010 Mathematics Subject Classification: 05E10; 13F16; 82B20 http://dspace.nbuv.gov.ua/handle/123456789/146152 en Symmetry, Integrability and Geometry: Methods and Applications Інститут математики НАН України |
| institution |
Digital Library of Periodicals of National Academy of Sciences of Ukraine |
| collection |
DSpace DC |
| language |
English |
| description |
In the first part of this paper, we provide a concise review of our method of solution of the Ar Q-systems in terms of the partition function of paths on a weighted graph. In the second part, we show that it is possible to modify the graphs and transfer matrices so as to provide an explicit connection to the theory of planar networks introduced in the context of totally positive matrices by Fomin and Zelevinsky. As an illustration of the further generality of our method, we apply it to give a simple solution for the rank 2 affine cluster algebras studied by Caldero and Zelevinsky. |
| format |
Article |
| author |
di Francesko, P. Kedem, R. |
| spellingShingle |
di Francesko, P. Kedem, R. Q-system Cluster Algebras, Paths and Total Positivity Symmetry, Integrability and Geometry: Methods and Applications |
| author_facet |
di Francesko, P. Kedem, R. |
| author_sort |
di Francesko, P. |
| title |
Q-system Cluster Algebras, Paths and Total Positivity |
| title_short |
Q-system Cluster Algebras, Paths and Total Positivity |
| title_full |
Q-system Cluster Algebras, Paths and Total Positivity |
| title_fullStr |
Q-system Cluster Algebras, Paths and Total Positivity |
| title_full_unstemmed |
Q-system Cluster Algebras, Paths and Total Positivity |
| title_sort |
q-system cluster algebras, paths and total positivity |
| publisher |
Інститут математики НАН України |
| publishDate |
2010 |
| url |
http://dspace.nbuv.gov.ua/handle/123456789/146152 |
| citation_txt |
Q-system Cluster Algebras, Paths and Total Positivity / P. di Francesco, R. Kedem // Symmetry, Integrability and Geometry: Methods and Applications. — 2010. — Т. 6. — Бібліогр.: 28 назв. — англ. |
| series |
Symmetry, Integrability and Geometry: Methods and Applications |
| work_keys_str_mv |
AT difranceskop qsystemclusteralgebraspathsandtotalpositivity AT kedemr qsystemclusteralgebraspathsandtotalpositivity |
| first_indexed |
2025-07-10T23:17:55Z |
| last_indexed |
2025-07-10T23:17:55Z |
| _version_ |
1837303842993078272 |
| fulltext |
Symmetry, Integrability and Geometry: Methods and Applications SIGMA 6 (2010), 014, 36 pages
Q-system Cluster Algebras, Paths
and Total Positivity?
Philippe DI FRANCESCO † and Rinat KEDEM ‡
† Institut de Physique Théorique du Commissariat à l’Energie Atomique, Unité de Recherche
associée du CNRS, CEA Saclay/IPhT/Bat 774, F-91191 Gif sur Yvette Cedex, France
E-mail: philippe.di-francesco@cea.fr
URL: http://ipht.cea.fr/en/Phocea/Pisp/visu.php?id=14
‡ Department of Mathematics, University of Illinois Urbana, IL 61801, USA
E-mail: rinat@illinois.edu
URL: http://www.math.uiuc.edu/∼rinat/
Received October 15, 2009, in final form January 15, 2010; Published online February 02, 2010
doi:10.3842/SIGMA.2010.014
Abstract. In the first part of this paper, we provide a concise review of our method of
solution of the Ar Q-systems in terms of the partition function of paths on a weighted
graph. In the second part, we show that it is possible to modify the graphs and transfer
matrices so as to provide an explicit connection to the theory of planar networks introduced
in the context of totally positive matrices by Fomin and Zelevinsky. As an illustration of
the further generality of our method, we apply it to give a simple solution for the rank 2
affine cluster algebras studied by Caldero and Zelevinsky.
Key words: cluster algebras; total positivity
2010 Mathematics Subject Classification: 05E10; 13F16; 82B20
Contents
1 Introduction 2
2 Partition functions 4
2.1 Hard particles on Gr . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
2.2 Transfer matrix on the dual graph . . . . . . . . . . . . . . . . . . . . . . . . . . 5
2.3 Hard particles and paths . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
3 Application to rank 2 cluster algebras of aff ine type 9
3.1 Rank two cluster algebras . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
3.2 The (2, 2) case: solution and path interpretation . . . . . . . . . . . . . . . . . . 9
3.3 The (1, 4) case: solution and relation to paths . . . . . . . . . . . . . . . . . . . . 11
4 Application to the Ar Q-system 14
4.1 Definition . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
4.2 Conserved quantities of Q-systems as hard particle partition functions . . . . . . 14
4.3 Q-system solutions and paths . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
4.4 An alternative path formulation . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16
?This paper is a contribution to the Proceedings of the Workshop “Geometric Aspects of Discrete and Ultra-
Discrete Integrable Systems” (March 30 – April 3, 2009, University of Glasgow, UK). The full collection is available
at http://www.emis.de/journals/SIGMA/GADUDIS2009.html
mailto:philippe.di-francesco@cea.fr
http://ipht.cea.fr/en/Phocea/Pisp/visu.php?id=14
mailto:rinat@illinois.edu
http://www.math.uiuc.edu/~rinat/
http://dx.doi.org/10.3842/SIGMA.2010.014
http://www.emis.de/journals/SIGMA/GADUDIS2009.html
2 P. Di Francesco and R. Kedem
5 Cluster algebra formulation: mutations and paths for the Ar Q-system 18
5.1 The Q-system as cluster algebra . . . . . . . . . . . . . . . . . . . . . . . . . . . 18
5.2 Target graphs and weights . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19
5.3 Mutations, paths and continued fraction rearrangements . . . . . . . . . . . . . . 21
5.4 Q-system solutions as strongly non-intersecting paths . . . . . . . . . . . . . . . 22
6 A new path formulation for the Ar Q-system 25
6.1 Compactified graphs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25
6.2 Examples of compactification . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26
6.3 Definition of compactified graphs . . . . . . . . . . . . . . . . . . . . . . . . . . . 27
6.4 An alternative construction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28
6.5 Equality of generating functions . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29
7 Totally positive matrices and compactif ied transfer matrices 30
8 Conclusion 34
References 35
1 Introduction
Discrete dynamical systems may take the form of recursion relations over a discrete time variable,
describing the evolution of relevant physical quantities. Within this framework, of particular
interest are the discrete integrable recursive systems, for which there exist sufficiently many
conservation laws or integrals of motion, so that their solutions can be expressed in terms
of some initial data. Interesting examples of such systems of non-linear integrable recursion
relations arise from matrix models used to generate random surfaces, in the form of discrete
Toda-type equations [1, 17, 22]. More recently, a combinatorial study of intrinsic geometry
in random surfaces has also yielded a variety of integrable recursion relations, also related to
discrete spatial branching processes [4].
We claim that other fundamental examples are provided by the so-called Q-systems for Lie
groups, introduced by Kirillov and Reshetikhin [20] as combinatorial tools for addressing the
question of completeness of the Bethe ansatz states in the diagonalization of the Heisenberg spin
chain based on an arbitrary Lie algebra. We proved integrability for these systems in the case
of Ar in [7]. In the case of other Dynkin diagrams, evidence suggests integrability still holds.
The Q-system with special (singular) initial conditions was originally introduced [20] as the
recursion relation satisfied by the characters of special finite-dimensional modules of the Yan-
gian Y (g), the so-called Kirillov–Reshetikhin modules. Remarkably, in the case g = Ar, the same
recursion relation also appears in other contexts, such as Toda flows in Poisson geometry [15],
preprojective algebras [14] and canonical bases [3].
In [7], we used methods from statistical mechanics to study the solutions of the Q-system
associated with the Lie algebra Ar, for fixed but arbitrary initial conditions. Our approach
starts with the explicit construction of the conserved quantities of the system, which appear as
coefficients in a linear recursion relation satisfied by Q-system solutions. These are finally used
to reformulate the solutions in terms of partition functions for weighted paths on graphs, the
weights being entirely expressed in terms of the initial data.
Note that there is a choice of various sets of 2r variables which constitute an initial condition
fixing the solutions of the Q-system recursion relation, a choice parametrized by Motzkin paths of
length r. This set of initial variables determines the graphs and weights which solve the problem.
This is the key point addressed in [7], and is related to the formulation as a cluster algebra.
Q-system Cluster Algebras, Paths and Total Positivity 3
Cluster algebras [9] are another form of discrete dynamical systems. They describe a specific
type of evolution, called mutation, of a set of variables or cluster seed. Mutations are rational,
subtraction-free expressions. This type of structure has proved to be very universal, and arises
in many different mathematical contexts, such as total positivity [13, 11], quiver categories [19],
Teichmüller space geometry [8], Somos-type sequences [12], etc.
Cluster algebras have the property that any cluster variable is expressible as a Laurent
polynomial of the variables in any other cluster in the algebra. It is conjectured that these
Laurent polynomials have nonnegative coefficients [9] (the positivity conjecture). This property
has only been proved in a few context-specific cases so far, such as finite type acyclic case [10],
affine type acyclic case [2], or clusters arising from surfaces [24]. The Q-system solutions for Ar
are also known to form a subset of the cluster variables in the cluster algebra introduced in [18]
(a result later generalized to all simple Lie algebras in [6]). In [7], we interpreted the solutions
of the Ar Q-system in terms of partition functions of paths on graphs with positive weights:
this proved positivity for the corresponding subset of clusters. Moreover, we obtained explicit
expressions, in the form of finite continued fractions, for these cluster variables. We review our
results in the first part of this paper.
Of particular interest to us is the connection of cluster algebras to total positivity. Fomin
and Zelevinsky [13] expressed a parametrization of totally positive matrices in terms of electrical
networks, and established total positivity criteria based on relations between matrix minors,
organized into a cluster algebra structure. In this paper we show the explicit connection of their
construction to the Q-system solutions.
In this paper, we review the methods and results of [7] in a more compact and hopefully
accessible form. We first apply this method to the case of rank 2 affine cluster algebras,
studied by Caldero and Zelevinsky [5]. These are the cluster algebras which arise from the
Cartan matrices of affine Kac–Moody algebras. We obtain a simple explicit solution for the
cluster variables in terms of initial data. We then proceed to describe the general solution of
the Ar Q-system using the same methods. In particular, we obtain an explicit formula for
the fundamental cluster variables, which generalizes the earlier results of [5] to higher rank.
Finally, we make the explicit connection between the path interpretation of the solutions of the
Q-system and a subclass of the totally positive matrices of [13] and their associated electrical
networks.
More precisely, it turns out that the generating function for the family of cluster variables R1,n
(n ∈ N) of the Ar Q-system is given by the resolvent of the transfer matrix Tm associated with
a graph Γm for some seed variables associated with the Motzkin path m. We show that it is
possible to locally modify the graph without changing the path generating function, so that we
obtain a transfer matrix of smaller size r+1, equal to the rank of the algebra. From there, there
is a straightforward identification with the networks associated with totally positive matrices of
special type, related to the Coxeter double Bruhat cells of [15].
The advantage of our approach is that we have explicit expressions for the cluster variables
in terms of any mutated cluster seed parametrized by a Motzkin path m.
The paper is organized as follows. In Section 2, we explain the basic tools from statistical
mechanics which we use, the partition functions of hard particles on graphs and the equivalent
partition functions of paths on weighted dual graphs.
For illustration, in Section 3, we use this to give the explicit expression for the generating
function of the cluster variables of rank 2 cluster algebras corresponding to affine Dynkin dia-
grams, a problem extensively studied in [27, 5, 25].
In Section 4, we review our solution [7] of the Ar Q-system for the simplest choice of initial
variables. This solution uses the partition functions of Section 2 as well as the theorem of [23, 16]
relating the partition function of non-intersecting paths to determinants of partition functions
of paths. We also introduce a new notion in this section, the “compactification” of the graph,
4 P. Di Francesco and R. Kedem
which gives a new transfer matrix which is associated with the same partition function. This is
a key tool in making the connection with totally positive matrices.
Section 5 generalizes the results of Section 4, and we give expressions for the Q-system
solutions in terms of any set of cluster variables in a fundamental domain. We also introduce
the notion of “strongly non-intersecting paths” and a generalization of [23, 16]. This gives
generating functions for the cluster variables in terms of the other seeds in the cluster algebra,
and also provides a proof of the positivity conjecture [9]. This section is a quick review of the
results of [7].
In Section 6, we extend the graph compactification procedure of Section 4 to the other
graphs, corresponding to mutated cluster variables, introduced in Section 5. This yields transfer
matrices of size (r + 1) × (r + 1), the resolvent of which is an alternative expression for the
generating function for cluster variables. Finally, in Section 7, we use this to give the explicit
relation to totally positive matrices [13, 15].
2 Partition functions
2.1 Hard particles on Gr
2.1.1 Vertex-weighted graphs
Let G be a finite graph with N vertices labeled 1, 2, . . . , N , and single, non-oriented edges
connecting some vertices. The adjacency matrix AG of the graph G is the N ×N matrix with
entries AG
i,j = 1 if vertex i is connected by an edge to a vertex j, and 0 otherwise. To each
vertex i is associated a positive weight yi.
2.1.2 Configurations
A configuration C of m hard particles on G is a subset of {1, . . . , N} containing m elements,
such that AG
i,j = 0 for all i, j ∈ C.
This is called a hard particle configuration, because if we view the elements of C to be the
vertices occupied by particles on the graph, the condition AG
i,j = 0 for i, j ∈ C enforces the rule
that two neighboring sites cannot be occupied at the same time. Each vertex can be occupied
by at most one particle.
We denote by CG
m the set of all hard particle configurations on G with m particles.
2.1.3 Partition function
The weight of a configuration C is the product of all the weights associated with the elements
of C. That is,
wC =
∏
i∈C
yi.
The partition function ZG
m for m hard particles on G is the sum over all configurations CG
m
of the corresponding weights:
ZG
m(y) =
∑
C∈CG
m
wC .
Q-system Cluster Algebras, Paths and Total Positivity 5
Figure 1. The graph Gr, with 2r + 1 vertices labeled i = 1, 2, . . . , 2r + 1.
Figure 2. The graph G̃r, dual to Gr, with total of 2r + 2 vertices.
2.1.4 The graph Gr
A basic example is the graph G = Gr of Fig. 1. It has 2r + 1 vertices and 3r− 1 edges, with the
(symmetric) adjacency matrix defined by
A1,2 = A2r,2r+1 = A2i,2i+1 = A2i,2i+2 = A2i+1,2i+2 = 1, i ∈ {1, . . . , r − 1}.
When r = 0, G0 is reduced to a single vertex labeled 1, while for r = 1, G1 is a chain of three
vertices 1, 2, 3 with the two edges (1, 2) and (2, 3).
Example 2.1. The non-vanishing partition functions ZGr
m for hard particles on Gr, r = 0, 1 are
ZG0
0 = 1, ZG0
1 = y1, ZG1
0 = 1, ZG1
1 = y1 + y2 + y3, ZG1
2 = y1y3. (2.1)
2.1.5 Recursion relations for the partition function
The transfer matrices ZGr
m satisfy recursion relations in the index r. They are obtained by
considering the possible occupancies of the vertices 2r + 1 and 2r:
ZGr
m = ZGr−1
m + y2r+1Z
Gr−1
m−1 + y2rZ
Gr−2
m−1 , r ≥ 2, m ≥ 0. (2.2)
For example, ZGr
0 = 1 is the partition function of the empty configuration and ZGr
r+1 =
r∏
i=0
y2i+1
for the maximally occupied configuration.
2.2 Transfer matrix on the dual graph
Associated to the graph Gr, there is a dual graph G̃r, as in Fig. 2.
It is dual in the sense that Gr is the medial graph of G̃r: each edge of G̃r corresponds to
a vertex of Gr, and any two edges of G̃r share a vertex iff the corresponding vertices of Gr are
adjacent.
We fix the labeling so that the correspondence is between edges of G̃r and vertices of Gr is:
1) edge (k, k′) of G̃r corresponds to the vertex 2k − 1 of Gr, where k = 2, 3, . . . , r;
2) edge (k, k + 1) of G̃r corresponds to the vertex 2k of Gr, where k = 1, 2, . . . , r;
3) edge (0, 1) of G̃r corresponds to the vertex 1 of Gr;
4) edge (r + 1, r + 2) of G̃r corresponds to the vertex 2r + 1 of Gr.
6 P. Di Francesco and R. Kedem
Figure 3. The graph G̃r with oriented edges drawn in.
2.2.1 Transfer matrix on G̃r
We choose an ordering of the 2r + 2 vertices of G̃r to be 0 < 1 < 2 < 2′ < 3 < 3′ < · · · < r <
r′ < r + 1 < r + 2, and consider this to be an index set. We construct the (2r + 2) × (2r + 2)
transfer matrix Tr(ty) with these indices. Its entries are 0 except for:
(Tr)k′,k = 1, (Tr)k,k′ = ty2k−1, k = 2, 3, . . . , r;
(Tr)k+1,k = 1, (Tr)k,k+1 = ty2k, k = 1, 2, . . . , r;
(Tr)1,0 = (Tr)r+2,r+1 = 1, (Tr)0,1 = ty1, (Tr)r+1,r+2 = ty2r+1.
In matrix form,
Tr(ty) =
0 ty1 0 0 0 · · · 0 0 0 0
1 0 ty2 0 0 · · · 0 0 0 0
0 1 0 ty3 ty4 · · · 0 0 0 0
0 0 1 0 0 · · · 0 0 0 0
0 0 1 0 0 · · · 0 0 0 0
...
...
. . .
...
0 0 0 0 0 · · · 0 ty2r−1 ty2r 0
0 0 0 0 0 · · · 1 0 0 0
0 0 0 0 0 · · · 1 0 0 ty2r+1
0 0 0 0 0 · · · 0 0 1 0
. (2.3)
This matrix is a weighted adjacency matrix for the graph G̃r as drawn in Fig. 3. Each edge
of G̃r corresponds to two oriented edges pointing in opposite directions, with the weights in
the transfer matrix corresponding to those oriented edges. The element of the transfer matrix
indexed by i, j is the weight corresponding to the edge j → i.
Lemma 2.1. The generating function for hard particle partition functions on Gr with weights
tyi per particle at vertex i is
ZGr(ty) :=
r+1∑
m=0
tmZGr
m = det(I − Tr(−ty)) (2.4)
with Tr as in equation (2.3).
Proof. Expanding the determinant Dr(t) = det(I−Tr(−ty)) along the last column, one obtains
the recursion relation:
Dr(t) = Dr−1(t) + ty2r+1Dr−1(t) + ty2rDr−2(t), r ≥ 2.
Let Dr,m denote the coefficient of tm in Dr(t). It satisfies the same recursion relation as
equation (2.2) for ZGr
m , with D0(t) = 1 + ty1 and D1(t) = 1 + t(y1 + y2 + y3) + t2y1y3, in
agreement with the initial values of ZGr(ty), r = 0, 1 in equation (2.1). Thus, Dr,m = ZGr
m for
all r, m. �
Q-system Cluster Algebras, Paths and Total Positivity 7
Example 2.2. For the case r = 1, G1 is a chain of 3 vertices 1, 2, 3. The dual G̃1 is a chain of
three edges connecting four vertices 0, 1, 2, 3, and the transfer matrix is:
T (ty) =
0 ty1 0 0
1 0 ty2 0
0 1 0 ty3
0 0 1 0
.
One checks that det(I −T (−ty)) = 1 + t(y1 + y2 + y3) + t2y1y3, in agreement with the partition
functions ZG1
i of equation (2.1).
2.3 Hard particles and paths
Let G be a graph with oriented edges. For example, the graph G̃r of the previous section can
be made into an oriented graph, by taking each edge to be a doubly-oriented edge. Then each
oriented edge from i to j receives a weight w(i, j), which is the corresponding entry of the transfer
matrix T (ty) of equation (2.3). This may be interpreted as the transfer matrix for paths on G̃r
as follows.
2.3.1 The partition function of paths
A path of length n on an oriented graph G is a sequence of vertices of G, P = (v0, v1, . . . , vn),
such that there exists an arrow from vi to vi+1 for each i. Let PG
a,b(n) be the set of all distinct
paths of length n, starting at vertex a and ending at vertex b on the graph G.
A path on a graph with weighted edges has a total weight which is the product of the weights
w(vi, vi+1) associated with the edges traversed in the path.
The partition function ZG
a,b(n) for PG
a,b(n) is
ZG
a,b(n) =
∑
P∈PG
a,b(n)
n−1∏
i=0
w(vi, vi+1).
Assuming G is finite, and labeling its vertices i = 1, 2, . . . , N , let us introduce the N×N transfer
matrix T with entries Ti,j = w(j, i), we have the following simple expression for ZG
a,b(n):
ZG
a,b(n) =
(
Tn
)
b,a
.
The matrix T (ty) of (2.3) is then the transfer matrix for paths on G̃r with weights 1 for edges
pointing away from the vertex 0, and tyi for the i-th edge pointing to the origin. The partition
function for weighted paths of arbitrary length on G̃r from the vertex 0 to itself is
ZG̃r(ty) =
∑
n≥0
(
T (ty)n
)
0,0
=
(
(I − T (ty))−1
)
0,0
. (2.5)
Lemma 2.2. The partition function of paths of arbitrary length from 0 to 0 on the graph G̃r is
equal to
ZG̃r(ty) =
ZGr(0,−ty2, . . . ,−ty2r+1)
ZGr(−ty1,−ty2, . . . ,−ty2r+1)
. (2.6)
Proof. By definition, (2.5) is the ratio of the (0, 0)-minor of the matrix I−T (ty) to its determi-
nant. Using equation (2.4) and the explicit form (2.3), this immediately yields the relation. �
This relation between hard-particle partition functions and path partition functions may be
interpreted as a boson-fermion correspondence, and is a particular case of Viennot’s theory of
heaps of pieces [28, 21].
8 P. Di Francesco and R. Kedem
Figure 4. A path on G̃3 with 16 steps. The weights yi are associated to the descending steps (1,−1) and
to the second half of the horizontal steps (1, 0)+ (1, 0) of the path, i being the label of the corresponding
edge of G̃3 (see on left). Here, the path receives the weight y2
1y2y3y4y5y6y7.
2.3.2 The path partition function as a continued fraction
A direct way to compute ZG̃r(ty) in equation (2.5) is by using Gaussian elimination on I−T (ty)
to bring it to lower-triangular form. The resulting pivot in the first row is 1/
(
(I −T (ty))−1
)
0,0
.
If we do this systematically, by left-multiplication by upper-triangular elementary matrices, the
result is
Lemma 2.3.
ZG̃r(ty) =
1
1− t y1
1−t
y2
1−ty3−t
y4
1−ty5−t
y6
...
1−ty2r−1−t
y2r
1−ty2r+1
. (2.7)
2.3.3 Non-Intersecting paths: the LGV formula
We may represent paths of length 2n from vertex 0 to itself on G̃r as paths on the lattice Z2
≥0 (see
Fig. 4 for an illustration). Such paths start at (0, 0) and end at (2n, 0), and have the following
possible steps:
1) to the northeast, (j, k) → (j + 1, k + 1), corresponding to the jth step in the path going
from vertex k to vertex k + 1;
2) to the southeast, (j, k) → (j + 1, k − 1), corresponding to the jth step in the path going
from vertex k to vertex k − 1;
3) to the east, (j, k) → (j + 2, k), corresponding to the steps k → k′ → k if k ∈ {2, 3, . . . , r}.
Paths from the origin to the origin have an even number of steps, by parity.
We will need to consider the partition function of families of α non-intersecting paths on G̃r,
ZG̃r
s,e. Here, the fixed starting points are parametrized by s = (s1, . . . , sα) and the endpoints by
e = (e1, . . . , eα).
By non-intersecting paths, we mean paths that do not share any vertex. We have the cele-
brated Lindström–Gessel–Viennot formula [23, 16]
ZG̃r
s,e = det
1≤i,j≤α
ZG̃r
si,ej
. (2.8)
The determinant has the effect of subtracting the contributions from paths that do intersect.
This formula can be proved by expanding the determinant as:∑
σ∈Sα
sgn(σ)
α∏
i=1
ZG̃r
sσ(i),ei
. (2.9)
Q-system Cluster Algebras, Paths and Total Positivity 9
One then considers the involution ϕ on families of paths, defined by interchanging the beginnings
of the two first paths that share a vertex until the vertex, or by the identity if no two paths
in the family intersect. It is clear that when ϕ does not act as the identity it relates two path
configurations with opposite weights in the expansion of the determinant, as the two starting
points are switched by ϕ, hence these cancel out of the expansion (2.9). We are thus left only
with non-intersecting families, all corresponding to σ = Id, hence all with positive weights.
3 Application to rank 2 cluster algebras of affine type
3.1 Rank two cluster algebras
In this section, we use the partition functions introduced in the previous section to the prob-
lem of computing the cluster variables of rank two cluster algebras of affine type with trivial
coefficients. This allows us to give explicit, manifestly positive formulas for the variables, prov-
ing the positivity of the variables in these cases.
3.1.1 The recursion relations
The rank 2 cluster algebras of affine type [9, 27] with trivial coefficients may be reduced to the
following recursion relations for n ∈ Z:
xn+1 =
1
xn−1
(1 + xb
n) if n is odd,
1
xn−1
(1 + xc
n) if n is even,
(3.1)
where b, c are two positive integers with bc = 4, hence (b, c) = (2, 2), (4, 1) or (1, 4). The aim is
to find an expression for xn, n ∈ Z in terms of some initial data, e.g. (x0, x1).
The connection to rank 2 affine Lie algebras is via the Cartan matrix
(
2 −b
−c 2
)
.
The cases (4, 1) and (1, 4) are almost equivalent: If xn(x0, x1) is a solution of the (b, c)
equation, then x1−n(x1, x0) is the solution of the (c, b) equation. We may thus restrict ourselves
to the (1, 4) case.
However, the symmetry xn ↔ x1−n changes the parity of n. Therefore we need to also
consider the dependence of xn in the “odd” initial data (x1, x2).
It turns out that the recursion relations (3.1) are all integrable evolutions. This allows us
to compute the generating function for xn, n ≥ 0. The result is a manifestly positive (finite)
continued fraction. In the light of the results of the previous section, this allows to reinterpret xn
as the partition function for weighted paths on certain graphs. This path formulation gives yet
another direct combinatorial interpretation for the expression of xn as a positive Laurent polyno-
mial of the initial data, to be compared with the approach of [27, 5] using quiver representations
and that of [25] using matchings of different kinds of graphs.
3.2 The (2, 2) case: solution and path interpretation
Consider the recursion relation
xn+1xn−1 = x2
n + 1 (3.2)
with x0 = (x0, x1) = (x, y). This is the A1 case of the renormalized Q-system considered in [6].
Due to the symmetry n ↔ 1− n of the equation, the solution satisfies
xn(x0, x1) = x1−n(x1, x0)
so we may restrict our attention to computing xn, for n ≥ 0.
10 P. Di Francesco and R. Kedem
3.2.1 Constants of the motion
Equation (3.2) is integrable. To see this, we rewrite (3.2) as
ϕn :=
∣∣∣∣xn−1 xn
xn xn+1
∣∣∣∣ = 1.
Then
0 = ϕn − ϕn+1 =
∣∣∣∣xn−1 + xn+1 xn
xn + xn+2 xn+1
∣∣∣∣ .
We conclude that there exists a constant c, independent of n, such that xn−1 + xn+1 = cxn for
all n. Using equation (3.2),
c =
xn−1 + xn+1
xn
=
xn+1
xn
+
1
xnxn+1
+
xn
xn+1
=
x1
x0
+
1
x0x1
+
x0
x1
. (3.3)
We interpret c as an integral of motion of the three-term relation (3.2): all solutions of the latter
indeed satisfy the two-term recursion relation (3.3), for some “integration constant” c fixed by
the initial data.
Note that c coincides with the partition function ZG1
1 = y1 + y2 + y3 for one hard particle on
the graph G1, with weights
y1 =
x1
x0
, y2 =
1
x0x1
, y3 =
x0
x1
. (3.4)
Note also that the only other non-vanishing hard particle partition functions on G1 are ZG1
0 = 1
and ZG1
2 = y1y3 = 1.
3.2.2 Generating function for xn
Let X(t) =
∞∑
n=0
tnxn be the generating function for the variables xn with n ≥ 0. Using xn+1 −
cxn + xn−1 = 0, we have by direct calculation
X(t) =
x0 − (cx0 − x1)t
1− ct + t2
=
x0
1− t y1
1−t
y2
1−ty3
, (3.5)
with yi as in (3.4). This gives xn as a manifestly positive Laurent polynomial of (x0, x1). In
fact, expanding the r.h.s. of (3.5), we get
X(t) =
∑
p,q,`≥0
(
p + q − 1
q
)(
q + `− 1
`
)
tp+q+`x1+`−p−q
0 xp−q−`
1 ,
from which xn is obtained by extracting the coefficient of tn. This agrees with the result of [5].
It follows that the dependence of the variables xn on any other pair (xk, xk+1) is also as
a positive Laurent polynomial. This is clear from the translational invariance of the system:
xn(xk, xk+1) = xn−k(x0, x1)
∣∣
x0 7→xk,x1 7→xk+1
.
Q-system Cluster Algebras, Paths and Total Positivity 11
3.2.3 Relation to the partition function of paths
Upon comparing the continued fraction expression (3.5) and equation (2.7), we see that there is
a path interpretation to the variables xn as follows.
The denominator of the fraction X(t) is equal to the partition function ZG1(−ty) for hard
particles on the graph G1 introduced in Section 2, with the weights −tyi per particle at vertex i.
Therefore,
X(t) = x0
ZG1(0,−ty2,−ty3)
ZG1(−ty1,−ty2,−ty3)
,
so x−1
0 X(t) is equal the partition function for paths on the graph G̃1 of Example 2.2, beginning
and ending at the vertex 0. The weights are as follows: The weight is equal to 1 for any step
away from vertex 0, and and is equal to tyi per step from vertex i to i− 1.
3.3 The (1, 4) case: solution and relation to paths
We now consider the system:
x2n =
1 + x2n−1
x2n−2
,
x2n+1 =
1 + x4
2n
x2n−1
, n ∈ Z.
We determine the dependence of the variables xn on two different types of initial conditions:
case 0 : (x0, x1),
case 1 : (x1, x2). (3.6)
We can eliminate the odd variables and get an equation for the even variables. Let un = x2n.
Using x2n−1 = unun−1 − 1, the even variables satisfy the recursion relation
un+1 =
u3
n + un−1
unun−1 − 1
,
or, equivalently, un(un+1un−1 − u2
n) = un+1 + un−1.
3.3.1 Conserved quantities
The variable wn = un+1un−1 − u2
n satisfies
wn =
un+1 + un−1
un
=
u3
n + un−1 + (unun−1 − 1)un−1
un(unu−1 − 1)
=
u2
n + u2
n−1
unun−1 − 1
. (3.7)
Moreover,
wn+1 − wn =
u2
n+1 + u2
n
un+1un − 1
− un+1 + un−1
un
=
u3
n − un+1unun−1 + un+1 + un−1
un(unun+1 − 1)
= 0.
We conclude that wn is a conserved quantity, that is, it is independent of n.
Using (3.7), there exists a constant c such that un+1 + un−1 = cun. We may compute this
constant explicitly in terms of the initial conditions in cases 0 and 1 of (3.6), using u0 = x0,
u1 = x2 and x0x2 = 1 + x1:
case 0 : c(0) =
u2
1 + u2
0
u1u0 − 1
=
x2
0 + x2
2
x1
=
x4
0 + (1 + x1)2
x2
0x1
,
case 1 : c(1) =
x4
2 + (1 + x1)2
x2
2x1
.
12 P. Di Francesco and R. Kedem
3.3.2 Generating function
The linear recursion relation un+1 − cun + un−1 = 0 implies the following formulas for the
generating functions
U (0)(t) :=
∑
n≥0
untn =
u0 − t(c(0)u0 − u1)
1− c(0)t + t2
=
x0
1− ty
(0)
1 − t2y
(0)
2
1−ty
(0)
3
, (3.8)
U (1)(t) :=
∑
n≥0
un+1t
n =
u1 − t(c(1)u1 − u2)
1− c(1)t + t2
=
x2
1− ty
(1)
1 − t2y
(1)
2
1−ty
(1)
3
, (3.9)
where the parameters are expressed in terms of the initial data as:
y
(0)
1 =
1 + x1
x2
0
, y
(0)
2 =
x4
0 + (1 + x1)2
x4
0x1
, y
(0)
3 =
x4
0 + 1 + x1
x2
0x1
,
y
(1)
1 =
x4
2 + 1 + x1
x2
2x1
, y
(1)
2 =
x4
2 + (1 + x1)2
x4
2x1
, y
(1)
3 =
1 + x1
x2
2
. (3.10)
Note that both sets satisfy the same relation y
(i)
1 y
(i)
3 = 1 + y
(i)
2 where i = 0, 1, as they are
a related via the substitution x2 ↔ x0, which maps
(
y
(1)
1 , y
(1)
2 , y
(1)
3
)
↔
(
y
(0)
3 , y
(0)
2 , y
(0)
1
)
.
Upon expanding the right hand sides of (3.8) and (3.9) in power series of t, we get all the
un’s as explicit positive Laurent polynomials of either initial data (x0, x1) and (x1, x2). Using
1
1− ta1 − t2a2
1−ta3
=
∑
p,q,`≥0
tp+2q+`ap
1a
q
2a
`
3
(
p + q
p
)(
q + `− 1
`
)
we obtain the expressions, valid for all n ≥ 0:
case 0 : x2n(x0, x1) =
∑
q,`,r,s,m≥0
x
1+4(q+`)−2n−4(r+s)
0 xm−q−`
1
×
(
n− q − `
q − r, r
)(
q + `− 1
`− s, s
)(
n + 2r + s− 2q − `
m
)
, (3.11)
case 1 : x2n+2(x1, x2) =
∑
q,`,r,s,m≥0
x
1+2n−4(q+`+r+s)
2 xq+`+m−n
1
×
(
n− q − `
q − r, r, s
)(
q + `− 1
`
)(
2r + s + `
m
)
, (3.12)
where we have used the multinomial coefficients
(
n
m1,m2, . . . ,mk
)
:=
n!
m1! · · ·mk!(n−
∑
mi)!
,
∑
i mi ≤ n,
0 otherwise.
The expressions for the odd variables follow from x2n+1 = x2nx2n+2−1. The positivity of these
latter expressions is easily checked: Note that the product x2nx2n+2 in both equations (3.11)
and (3.12) has a constant term 1 as a Laurent polynomial of (x0, x1) and (x1, x2) respectively,
hence x2nx2n+2 − 1 is a positive Laurent polynomial.
Q-system Cluster Algebras, Paths and Total Positivity 13
Figure 5. The graph G̃(1,4) and the four corresponding path steps.
3.3.3 Path interpretation
The continued fraction expressions (3.8) and (3.9) allow for a path interpretation of x2n as
follows. Consider the graph G̃(1,4) on the left hand side of Fig. 5, with two vertices labelled 0, 1
connected by an edge, and connected to themselves via a loop. We assign weights to the oriented
edges as follows:
w(0 → 0) = ta1, w(0 → 1) = t, w(1 → 1) = ta3, w(1 → 0) = ta2.
We can also associate a path in Z2
≥0 to a path on G̃(1,4) composed of the steps shown in Fig. 5.
The corresponding path transfer matrix is: T = t
(
a1 a2
1 a3
)
. Using Gaussian elimination on
I −T as in the previous example, we find that the partition function of paths from the vertex 0
to itself is:(
(I − T )−1
)
0,0
=
1
1− ta1 − t2a2
1−ta3
.
Hence, x2n is (up to a factor of x0 in case 0 and x2 in case 1) the partition function for paths
of 2n steps on G̃(1,4), from and to the vertex 0, with weights (a1, a2, a3) = (y1, y2, y3) in case (i)
and (a1, a2, a3) = (z1, z2, z3) in case (ii), where the weights are as in (3.10).
3.3.4 The graph G̃1,4 as a compactif ication of the graph G̃1
We will see later that, quite generally, it is possible to “compactify” the graphs G̃r. The result
for r = 1 is precisely the graph G̃1,4. Thus, The case (b, c) = (2, 2) of the previous section may
also be interpreted in terms of paths on the graph G̃(1,4), but with different weights:
w(0 → 0) = ty1, w(0 → 1) = 1, w(1 → 1) = ty3, w(1 → 0) = ty2.
where yi are as in equation (3.4).
To see this, note that the continued fraction (3.5) may be rewritten as:
X(t) =
x0
1− t y1
1−t
y2
1−ty3
= x0 +
tx1
1− ty1 − ty2
1−ty3
= x0 + tx1
((
I −
(
ty1 ty2
1 ty3
))−1
)
0,0
.
Thus, for n ≥ 0, the solution xn+1 of the (2, 2) case is up to a factor x1 the partition function
for paths on G̃(1,4), from and to the origin vertex 0, and with a total of n steps of the form
0 → 0, 1 → 1 or 1 → 0, with respective weights y1, y2, y3.
14 P. Di Francesco and R. Kedem
4 Application to the Ar Q-system
In [18, 6], we showed that the recursion relations (Q-systems) satisfied by the characters of the
Kirillov–Reshetikhin modules of the quantum affine algebras associated with any simple Lie
algebra can be described in terms of mutations of a cluster algebra. Solutions to the cluster
algebra recursion relations are more general, in that the initial conditions are not specialized, as
they are in the original Q-system satisfied by characters of Kirillov–Reshetikhin modules [20].
A particularly simple example of this is the case when the Lie algebra is of type A. Characters
of An KR-modules, which are just Schur functions corresponding to rectangular Young diagrams,
are given by the cluster variables upon specialization of the boundary conditions. In [7], we gave
the general solution for the cluster variables without specialization of initial conditions. For
these variables, we proved the positivity conjecture of Fomin and Zelevinsky by mapping the
problem to a partition function of weighted paths on a graph. Let us review the results obtained
in [7].
4.1 Definition
The Ar Q-system is the following system of recursion relations for a sequence Rα,n, α ∈ Ir =
{1, 2, . . . , r} and n ∈ Z:
Rα,n+1Rα,n−1 = R2
α,n + Rα+1,nRα−1,n, (4.1)
with boundary conditions
R0,n = Rr+1,n = 1 for all n ∈ Z.
We note that the case r = 1 of A1 coincides with the case (2, 2) treated in Section 3.2.
Remark 4.1. The original Q-system, which is the one satisfied by the characters of KR-modules,
differs from the system (4.1) not only in that it is specialized to the initial conditions Rα,0 = ±1
for all α ∈ Ir, but also by a minus sign in the second term on the right. In this discussion, we
choose to renormalize the variables for simplicity. It is also possible (see the appendix of [6])
to consider the recursion relation with nontrivial coefficients. In that case both (4.1) and the
original Q-system result from a specialization of the coefficients.
We wish to study the solutions of equation (4.1) for any given initial data. Our standard initial
data for the Q-system are the variables x0 = (Rα,0, Rα,1)α∈Ir . We may then view equation (4.1)
as a three-term recursion relation in n, which requires two successive values of n (for each α) as
initial data. Our first goal is to express all {Rα,n}α,n in terms of the initial data x0.
4.1.1 Symmetries of the system
We may use equation (4.1) to get both n ≥ 2 and n ≤ −1 values in terms of x0. These
are related via the manifest symmetry of the Q-system under n ↔ 1 − n, which implies
Rα,1−n
(
(Rα,0, Rα,1)α∈Ir
)
= Rα,n
(
(Rα,1, Rα,0)α∈Ir
)
. In addition, equation (4.1) is translation-
ally invariant under n → n + k: Rα,n+k
(
(Rα,k+1, Rα,k)α∈Ir
)
= Rα,n
(
(Rα,1, Rα,0)α∈Ir
)
for all
n, k ∈ Z.
4.2 Conserved quantities of Q-systems as hard particle partition functions
The Q-system turns out to be a discrete integrable system, in that it is possible to find
a sufficient number of integrals of the motion, as in the r = 1 case.
Q-system Cluster Algebras, Paths and Total Positivity 15
First, equation (4.1) may be used to express Rα,n (α ≥ 2) as polynomials in {R1,n}:
Rα,n = det
1≤i,j≤α
(R1,n+i+j−α−1). (4.2)
This is proved by use of the Desnanot–Jacobi identity for the minors of the (α + 1) × (α + 1)
matrix M with entries Mi,j = R1,n+i+j−α, i, j ∈ Iα+1.
The boundary condition Rr+1,n = 1 (for all n) together with (4.2) implies the equation of
motion which determines R1,n:
det
1≤i,j≤r+1
(R1,n+i+j−r−2) = 1.
From equation (4.1), it is clear that Rr+2,n = 0 for all n. Hence there exists a linear recursion
relation of the form
r+1∑
m=0
(−1)mcr+1−m R1,n+m = 0, (4.3)
with c0 = cr+1 = 1. The fact that cr+1−m do not depend on n follows from the fact that each
row in the matrix in the determinant for Rr+2,n is just a shift in n of any other row.
The constants cp, p = 1, 2, . . . , r are the r integrals of motion of the Ar Q-system. They can
be expressed explicitly in terms of the R1,ns as follows:
cp = det
1≤i≤r+1
1≤j≤r+2, j 6=r+2−p
R1,n+i+j−2
independently of n. These quantities are similar to those found in [26] for the so-called Coxeter–
Toda integrable systems.
By using simple determinant identities, we show in Theorem 3.5 of [7] that cp satisfy recursion
relations which allow us to identify them as the partition functions ZGr
p of p hard particles on
the graph Gr of Fig. 1, where the weights given by:
y2α−1,k =
Rα−1,kRα,k+1
Rα,kRα−1,k+1
, 1 ≤ α ≤ r + 1, y2α,k =
Rα−1,kRα+1,k+1
Rα,kRα,k+1
, 1 ≤ α ≤ r. (4.4)
This is true for any k: The functions ZGr
p are independent of the choice of k. Thus, unless
otherwise stated, yα will stand for yα,0 below.
4.3 Q-system solutions and paths
The linear recursion relation (4.3) allows to compute the generating function R(r)(t) =
∞∑
n=0
tnR1,n
explicitly. Indeed, R(r)(t)ZGr(−ty) is a polynomial of degree r, and it is easy to see that
R(r)(t)ZGr(−ty) = R1,0Z
Gr(0,−ty2, . . . ,−ty2r+1).
Using (2.6), we may interpret R(r)(t)/R1,0 as the generating function ZG̃r
0,0(ty) for paths on G̃r
with the weights (4.4), say for k = 0. In other words, R1,n/R1,0 is the partition function for
weighted paths of 2n steps on G̃r starting and ending at the origin (or starting at (0, 0) and
ending at (2n, 0) in the two-dimensional representation).
Comparing the determinant formula for Rα,n (4.2) and the LGV formula (2.8) for families of
α non-intersecting paths, we have [7]
16 P. Di Francesco and R. Kedem
Figure 6. The six pairs of non-intersecting paths on G̃2 of 8 and 4 steps, starting respectively at (0, 0)
and (2, 0) and ending at (6, 0) and (8, 0).
Figure 7. The graph G̃′
r, with r + 1 vertices. We have indicated the weights attached to each oriented
edge.
Lemma 4.1. The quantity Rα,n/(R1,0)α is the partition function ZG̃r
s,e for α non-intersecting
weighted paths on G̃r with starting and ending points si = (2i − 2, 0), ei = (2n + 2α − 2i, 0),
i = 1, 2, . . . , α.
Proof. It is clear that ZG̃r
si,ej
is the partition function for paths from (2i−2, 0) to (2n+2α−2j, 0),
which is equal to that from (0, 0) to (2(n + α + 1 − i − j), 0) by translational invariance. This
allows to identify the two determinants (4.2) and (2.8), up to an overall factor of (R1,0)α, and
the result follows. �
As an illustration, we have represented in Fig. 6 the six pairs of non-intersecting paths
contributing to R2,3, solution of the A2 Q-system.
Thus, we have proved
Theorem 4.1. The variables Rα,n which satisfy (4.1), when expressed in terms of the va-
riables x0, are equal to (R1,0)α times partition functions for paths on the graph G̃r, involving
only the weights yα of (4.4). These weights are explicit Laurent monomials of the initial data
x0 = (Rα,0, Rα,1)α∈Ir . This gives an explicit expression for Rα,n as Laurent polynomials of the
initial data, with non-negative integer coefficients.
4.4 An alternative path formulation
In the same spirit as Section 3.3.4, one can show (see below) that the solution R1,n of the Ar
Q-system may also be interpreted in terms of paths on a new (“compactified”) graph G̃′
r of
Fig. 7. This is a graph with r + 1 vertices, labelled 1, 2, . . . , r + 1, connected via oriented edges
i → i + 1, i + 1 → i, i = 1, 2, . . . , r and loops i → i, i = 1, 2, . . . , r + 1.
To each edge is attached a weight as follows:
wt(i → i + 1) = 1, wt(i + 1 → i) = ty2i, wt(i → i) = ty2i−1. (4.5)
The corresponding transfer matrix encoding these weights is an (r + 1)× (r + 1)-matrix of the
Q-system Cluster Algebras, Paths and Total Positivity 17
form
T ′ =
ty1 ty2 0 · · · · · · · · · · · · 0
1 ty3 ty4 0
...
0 1 ty5 ty6 0
...
...
. . . . . . . . . . . . . . .
...
...
. . . . . . . . . . . . . . .
...
... 0 1 ty2r−3 ty2r−2 0
... 0 1 ty2r−1 ty2r
0 · · · · · · · · · · · · 0 1 ty2r+1
. (4.6)
Then we have
R(r)(t)
R1,0
= 1 + t
y1
1− ty1 − t y2
1−ty3−t
y4
1−ty5−t
y6
...
1−ty2r−1−t
y2r
1−ty2r+1
= 1 + t
R1,1
R1,0
(
(I − T ′)−1
)
1,1
. (4.7)
This is readily proved by Gaussian elimination.
Comparing this expression to the results of the previous section, we conclude that
Lemma 4.2. For n ≥ 0, R1,n+1/R1,1 is the partition function for paths on the weighted graph G̃′
r,
with a total of n steps along the edges of type i → i or i + 1 → i in G̃′
r, starting and ending at
vertex 1.
The weights are given in equation (4.5) with yi as in (4.4) with k = 0.
In Section 7 we will relate this result to the total positivity conjecture of Fomin and Zelevinsky
and networks.
We may actually write an explicit expression for R1,n by simply expanding the continued
fraction (4.7) as:
R(r)(t) = R1,0
(
1 + ty1
∑
p1,p2,...,p2r+1≥0
p0=p2r+2=0
r∏
`=0
(ty2`+1)p2`+1
× (ty2`+2)p2`+2
(
p2` + p2`+1 + p2`+2 − 1
p2` − 1, p2`+1
))
.
Substituting the values (4.4) for the weights yα ≡ yα,0, and extracting the coefficient of tn+1,
we get for all n ≥ 0:
R1,n+1 = R1,1
∑
p1,p2,...,p2r+1≥0
p0=p2r+2=0, Σpi=n
r∏
i=1
(Ri,0)p2i+2+p2i+1−p2i−p2i−1
(Ri,1)p2i+1+p2i−p2i−1−p2i−2
r∏
`=0
(
p2` + p2`+1 + p2`+2 − 1
p2` − 1, p2`+1
)
as explicit positive Laurent polynomials of the initial data. This gives a rank-r generalization
of the formula given in [5] for r = 1.
18 P. Di Francesco and R. Kedem
5 Cluster algebra formulation: mutations and paths
for the Ar Q-system
In this section, we show that the solutions {Rα,n | α ∈ Ir, n ∈ Z} of the Q-system are positive
Laurent polynomials when expressed as functions of an arbitrary set initial conditions. This
generalizes our result for the initial condition x0 in the previous section.
The recursion relation (4.1) has a solution once a certain set of initial conditions is specified,
but this set need not necessarily be the set x0. We will explain below that the most general
possible choice of initial conditions is specified by a Motzkin path of length r.
The solutions of (4.1) can be viewed as cluster variables in the Ar Q-system cluster algebra
defined in [18]. Hence, our proof provides a general confirmation of the conjecture of [9] for this
particular cluster algebra: When the cluster variables are expressed as functions of the variables
in any other cluster (i.e. an arbitrary set of initial conditions), they are Laurent polynomials
with non-negative coefficients.
The results of this sections were explained in detailed in [7], and this section should serve as
a summary of the proofs contained therein.
5.1 The Q-system as cluster algebra
In [18], it was shown that the Ar Q-system solutions {Rα,n} may be viewed as a subset of
the cluster variables of the Ar Q-system cluster algebra. This is a cluster algebra with trivial
coefficients, which includes the the seed cluster variable (R1,0, R2,0, . . . , Rr,0, R1,1, R2,1, . . . , Rr,1),
with an associated associated 2r × 2r exchange matrix has the block form:
(
0 −C
C 0
)
, C the
Cartan matrix of Ar.
In this language, each cluster is a vector with 2r variables, and the subset of clusters relevant
to the Q-system are those which have entries made up entirely of solutions to the Q-system
(we restrict to these in the following). These clusters are all related by sequences of cluster
mutations which are one of the relations (4.1).
Because of the form of equation (4.1), it is easy to see that the restricted set of clusters
corresponding to the Q-system are characterized by a set of r integers (m1, . . . ,mr), subject to
the condition that |mα −mα+1| ≤ 1. This defines what is known as a Motzkin path.
Definition 5.1. The cluster xm corresponding to the set of integers m = (m1, . . . ,mr) is the
vector of 2r variables {Rα,mα , Rα,mα+1}α∈Ir , ordered so that all variables with an even second
index appear first.
For example, the initial cluster x0 corresponds to the Motzkin path m = (0, 0, . . . , 0), and
x0 = (R1,0, . . . , Rr,0, R1,1, . . . , Rr,1).
For any m, xm is obtained from the fundamental initial seed x0 by mutations of the cluster
algebra. This is just saying that one gets xm by repeated selected applications of the recursion
relation (4.1) to x0. Each mutation changes only one of the cluster variables. That is, for some α
and n,
Rα,n 7→
Rα,n+2 =
R2
α,n+1+Rα+1,n+1Rα−1,n+1
Rα,n
(forward mutation at α),
Rα,n−2 =
R2
α,n−1+Rα+1,n−1Rα−1,n−1
Rα,n
(backward mutation at α).
Recall that we only consider here particular mutations that only involve solutions of the Q-
system. Here we have used the “time” variable n to define forward (resp. backward) mutations
according to whether the mutation increases (resp. decreases) the index n in the mutated cluster
variable.
Q-system Cluster Algebras, Paths and Total Positivity 19
Figure 8. The Motzkin path m = (2, 1, 2, 2, 2, 1, 0, 0, 1) for r = 9 (a) is decomposed into p = 6 descending
segments (12)(3)(4)(567)(8)(9) (circled, red edges). The corresponding graph pieces Γmi are indicated
in (b). They are to be glued “horizontally” for flat transitions (green edges) and “vertically” for ascending
ones (blue edges). The resulting graph Γm is represented (c) with its vertex (black) and edge (red) labels.
Alternatively, the mutation changes the Motzkin path which characterizes xm, by changing
m 7→ m ± εα, with the plus (minus) sign for a forward (backward) mutation. Here, εα is the
vector which is zero except for the entry α, which is equal to 1. We see that accordingly the
Motzkin path locally moves forward (backward).
The Laurent property of cluster algebras ensures that every cluster variable is a Laurent
polynomial of the cluster variables of any other cluster in the algebra.
The positivity property, proved only in particular cases so far, is that these Laurent polyno-
mials have non-negative integer coefficients. The property was proved for the particular clusters
considered in the present case in [7]. It may be stated as follows:
Theorem 5.1 ([7]). Each Rα,n, when expressed as a function of the seed xm for any Motzkin
path m, is a Laurent polynomial of {Rα,mα , Rα,mα+1 | α ∈ Ir}, with non-negative integer coeffi-
cients.
We outline the proof below.
For clarity let us introduce the following notation. Let F be some cluster variable. Then F
can be expressed as a function of xm for any m. The functional form is then denoted by
F = Fm(xm). Since F can also be expressed as a function of any other cluster, we can write
Fm(xm) = Fm′(xm′) for any two Motzkin paths m, m′. In particular, in the notation of the
previous section, we have R(r)(t) = R
(r)
m0(t;x0).
5.2 Target graphs and weights
Due to the reflection and translation symmetries of the Q-system, we can restrict our attention
to seeds associated with Motzkin paths in a fundamental domain Mr = {m | minα(mα) = 0}.
There are 3r−1 elements in Mr.
To each Motzkin path m ∈ Mr, we associate a pair (Γm, {ye(m)}), consisting of a rooted
graph Γm with oriented edges, and edge weights ye(m) along the edges e.
20 P. Di Francesco and R. Kedem
5.2.1 Construction of the graph Γm
The graph Γm is constructed via the following sequence of steps (see Fig. 8 for an illustration):
1. Decompose the Motzkin path m into maximal “descending segments” mi of length ki
(i = 1, . . . , p). These are segments of the form mi = (mαi ,mαi+1, . . . ,mαi+1−1) with
αi+1 = αi + ki, where mαi+j = mαi − j. Here, α1 = 1 and αp+1 − 1 = r.
2. The separation between two consecutive descending segments of the Motzkin path, mi and
mi+1 is either “flat” i.e. mαi+1 = mαi+1−1 or “ascending” i.e. mαi+1 = mαi+1−1 + 1.
3. To each descending segment mi, associate a graph Γmi , which is the graph G̃ki
with
additional, down-pointing edges a → b for all a, b such that ki + 1 ≥ a > b + 1 ≥ 2. There
are a total of ki(ki − 1)/2 extra oriented edges.
4. We glue the graphs Γmi and Γmi+1 into a graph Γmi ||Γmi+1 defined as follows (see Fig. 8
for an illustration):
(a) If the separation between mi and mi+1 is flat, we identify vertex 0 of Γmi+1 with
vertex ki + 2 of Γmi , and vertex 1 of Γmi+1 with vertex ki + 1 of Γmi , while the
connecting edges are identified.
(b) If the separation is ascending, we reverse the role of vertices 0 and 1 in the procedure
above.
The result of this procedure is the graph Γm = Γm1 ||Γm2 || · · · ||Γmp . Its root is the vertex 0
of Γm1 .
We label the vertices of the graph Γm by the integers i with i ∈ {0, . . . , r + 2 + n+(m)}
(where n+(m) is the number of mα such that mα+1 = mα + 1) and labels i′ for any univalent
vertex attached to vertex i via a horizontal edge. We do this by labeling the vertices of Γm
from bottom to top, by shifting the labels of the subgraphs Γmi so that no label is skipped nor
repeated.
The edges e pointing towards the root of Γm are of two types:
(i) the “skeleton edges” belonging to some G̃ki
in the above construction;
(ii) the extra, down-pointing edges added in the gluing procedure.
5.2.2 The weights on the graph Γm
We label the 2r+1 skeleton edges of type (i) by α = 1, 2, . . . , 2r+1 from bottom to top (see the
example in Fig. 8), and the weights are denoted by yα(m). Weights assigned to edges pointing
away from the root are all set to 1.
Alternatively, we may label the “down pointing” skeleton edges by the pairs of vertices
i+1 → i or i′ → i which they connect. The extra edges of type (ii) are also labeled by the pairs
a → b of vertices which they connect. All edge weights may be labeled by the label of the edge.
The weights of the edges of type (ii) can be expressed in terms of the skeleton weights:
ya,b(m) =
∏
b≤i<a
yi+1,i(m)∏
b<i<a
yi′,i(m)
,
so that they obey the following intertwining condition
ya,b(m)ya′,b′(m) = ya,b′(m)ya′,b(m), a > a′ > b > b′. (5.1)
Q-system Cluster Algebras, Paths and Total Positivity 21
For example, the extra weights of the example of Fig. 8 read respectively: y3,1 = y2y4/y3,
y9,7 = y14y12/y13, y8,6 = y12y10/y11, and y9,6 = y9,7y8,6/y12 = y14y12y10/(y13y11).
Finally, for a given Motzkin path m ∈ Mr, we define the skeleton weights yα(m), α =
1, 2, . . . , 2r + 1 to be:
y2α−1(m) =
λα,mα
λα−1,mα−1
, α = 1, 2, . . . , r + 1, (5.2)
y2α(m) =
µα+1,mα+1
µα,mα
(
1− δmα,mα+1+1 +
λα+1,mα+1
λα+1,mα
δmα,mα+1+1
)
×
(
1− δmα−1,mα+1 +
λα−1,mα
λα−1,mα−1
δmα−1,mα+1
)
, α = 1, 2, . . . , r, (5.3)
where
λα,n =
Rα,n+1
Rα,n
, µα,n =
Rα,n
Rα−1,n
.
Note that with these definitions the expressions (5.2), (5.3) involve only variables of the seed xm.
To each Motzkin path m ∈ Mr, we may finally associate a transfer matrix Tm ≡ Tm(ty(m)),
with entries (Tm)b,a = weight of the oriented edge a → b on Γm. Then the series in t
Zm(t;xm) :=
(
(I − Tm)−1
)
0,0
=
∑
n
tnZΓm
0,0 (n) (5.4)
is the generating function for weighted paths on Γm, with the coefficient of tn being the partition
function of walks from vertex 0 to itself on Γm which have n down-pointing steps. When
m = m0 = 0, this coincides with (2.5).
5.3 Mutations, paths and continued fraction rearrangements
Our purpose is to write an explicit expression for the functional dependence of the variables Rα,n
on the seed variable xm, that is, find (Rα,n)m(xm) for each α, n and a Motzkin path m.
To do this, we will describe how the generating function R(r)(t) = R
(r)
m (t;xm) is related to
the generating function R
(r)
m′(t;xm′), where m and m′ are related by a mutation. Then we start
from the known function R
(r)
m0(t;xm0), and apply mutations to obtain all other functions R
(r)
m
with m in the fundamental domain Mr.
One can cover the entire fundamental domain Mr starting from m0 = 0 by using only forward
mutations m 7→ m′ = m + εα of either type (i) (. . . , a, a, a + 1, . . . ) 7→ (. . . , a, a + 1, a + 1, . . . )
and (ii) (. . . , a, a, a, . . . ) 7→ (. . . , a, a + 1, a, . . . ), with the obvious truncations when α = 1 or r.
(See Remark 8.1 in [7].)
Suppose m and m′ are related by such a mutation. We compute the two generating functions
of the type (5.4), Zm(t;xm) and Zm′(t;xm′).
In fact, the two matrices, Tm and Tm′ differ only locally, so that in computing the two
generating functions by row reduction, we find that the calculation differs only in a finite
number of steps. Note that generating functions take the form of finite continued fractions with
manifestly positive series expansions of t.
We note two simple rearrangement lemmas which can be used to relate finite continued
fractions:
(R1)
1
1− a
1−b
= 1 +
a
1− a− b
,
(R2) a +
b
1− c
=
a′
1− b′
1−c′
, where a′ = a + b, b′ =
bc
a + b
, c′ =
ac
a + b
.
One checks this by explicit calculation.
22 P. Di Francesco and R. Kedem
Let α > 1. Then, using (R2) we can show that Zm(t;xm) = Zm′(t;xm′) if and only if the
weights yi ≡ yi(m) and y′i ≡ yi(m′) are related via:
(i) mα−1 = mα < mα+1 : y′β =
y2α−1 + y2α, β = 2α− 1,
y2αy2α+1/(y2α−1 + y2α), β = 2α,
y2α−1y2α+1/(y2α−1 + y2α), β = 2α + 1,
yβ, otherwise,
(ii) mα−1 = mα = mα+1 : y′β =
y2α−1 + y2α, β = 2α− 1,
y2αy2α+1/(y2α−1 + y2α), β = 2α,
y2α−1y2α+1/(y2α−1 + y2α), β = 2α + 1,
y2α+2y2α−1/(y2α−1 + y2α), β = 2α + 2,
yβ , otherwise.
One checks directly that the expressions (5.2), (5.3) indeed satisfy the above relations.
The boundary case, where α = 1, is treated analogously, but first requires a “rerooting” of
the graph to its vertex 1, which is implemented by the application of (R1): Indeed, we simply
write Zm(t;xm) =
(
(I − Tm)−1
)
0,0
= 1 + ty1(m)Z′m(t;xm) with Z′m(t;xm) =
(
(I − Tm)−1
)
1,1
.
We then rearrange Z′m(t;xm) using (R2) again, and find that Z′m(t;xm) = Zm′(t;xm′) if and
only if the weights are related via the above equations.
The net result is the following. Given a compound mutation µm which maps the fundamental
Motzkin path m0 to m = (mα)α∈Ir , then there are exactly m1 “rerootings” as described above.
This corresponds to rewriting the generating function
R
(r)
m (t;xm) =
m1−1∑
i=0
tiR1,i + tm1R1,m1Zm(t;xm)
with Zm(t;xm) as in equation (5.4).
This leads to the following main result:
Theorem 5.2. For each n ≥ 0, the function (R1,n+m1)m(xm) = R1,m1 ZΓm
0,0 (n). Thus it is
proportional to the generating function for weighted paths on the graph Γm with positive weights,
so it is a manifestly positive Laurent polynomial of the initial data xm.
We have represented in Fig. 9 the graphs Γm for the Motzkin paths m of the fundamental
domain Mr for r = 3.
5.4 Q-system solutions as strongly non-intersecting paths
To treat the case of Rα,n with α 6= 1, given a Motzkin path m ∈ Mr, we need a path interpre-
tation for the determinant formula for Rα,n+m1 :
Rα,n+m1
(R1,m1)α
=
det (R1,n+m1+i+j−α−1)1≤i,j≤α
R1,n+m1
= det
(
ZΓm
0,0 (n + i + j − α− 1)
)
1≤i,j≤α
. (5.5)
Here, we have used the result of the previous section to rewrite the formula in terms of the
partition function for paths on Γm, from and to the root, and with n+ i+ j−α− 1 down steps.
As in the standard LGV formula, we interpret this determinant as a certain partition function
for paths on Γm starting from the root at times 0, 2, 4, . . . , 2α − 2 and ending at the origin at
times 2n, 2n + 2, . . . , 2n + 2α− 2.
Q-system Cluster Algebras, Paths and Total Positivity 23
Figure 9. The Motzkin paths m of the fundamental domain M3 and the associated graphs Γm, with
their vertex and edge labels. We have also indicated the mutations by arrows, the label being α when
the mutation µα acts on variables Rα,m with an even index m and α + r for an odd index m.
5.4.1 Paths on Γm represented as paths on a square lattice
We draw paths, with allowed steps dictated by the graph Γm, on a square-lattice in two dimen-
sions. Paths start and end at y-coordinate 0. Moreover, if a path has n “down” steps (steps
towards the vertex 0), then its starting and ending point are separated by 2n horizontal steps.
That is, a path is from (x, 0) to (x + 2n, 0) where x the starting time and n is the number of
down-steps.
Since the horizontal distance between the starting and ending points is fixed by the number
of “down” steps, a single step of the form
a → b = a− h
should have a horizontal displacement 2− h (instead of 1 as in the usual case). That is, on the
square lattice it is a segment of the form
(x, a) → (x + 2− h, a− h).
Some examples are illustrated in Fig. 10.
Thus, we identify ZΓm
0,0 (n + i + j − α− 1) = ZΓm
si,ej
with paths on the two-dimensional lattice
starting at the point si = (2i − 2, 0) and ending at the point ej = (2n + 2α − 2j, 0), with the
types of steps allowed given by the edges of Γm in the way explained in the previous paragraph.
24 P. Di Francesco and R. Kedem
Figure 10. The two-dimensional representation of a typical path on the graph Γm, m the strictly
descending Motzkin path (2, 1, 0) of the case A3. Descents of h = 2 are vertical (time displacement by
2 − h = 0), while descents of h = 3 go back one step in time (time displacement by 2 − h = −1). With
these choices, the total time distance between start and end is twice the number of descents (16 = 2× 8
here).
5.4.2 Strongly non-intersecting paths
We now look at families of α paths, corresponding to the determinant in equation (5.5). Such
paths may have crossing on the lattice. As in the case of LGV formula, the determinant cancels
out contributions from paths which share a vertex. However, other situations may occur: Two
paths may cross without sharing a vertex in our picture.
One can generalize the proof of the LGV formula to take such crossings into account. Using
the expansion (2.9), and introducing an involution ϕ on families of paths. This involution
interchanges the beginnings of the first two paths which share a vertex or which cross each
other, by transforming the crossing segments [P,Q] and [R,S] into non-crossing ones [R,Q]
and [P, S]. This effectively interchanges the two paths up to the points P and R respectively,
whichever comes first.
As in the usual case, the involution ϕ acts as the identity if no two paths cross, share a vertex,
or can be made to cross via such an exchange.
Taking into account the weights of the paths, the intertwining condition (5.1) implies that
the flip preserves the absolute value of the weight, but changes its sign, due to the transposition
of starting points. So the determinant (5.5) cancels not only the paths that share a vertex or
that cross, but also those that come “too close” to one-another, namely that can be made to
cross via a flip.
We call the families of paths which are invariant under the involution ϕ strongly non-
intersecting paths.
To summarize, we have the following theorem:
Theorem 5.3. For any Motzkin path m ∈ Mr, the variable Rα,n+m1 (with n + m1 ≥ α − 1)
is equal to (R1,m1)
α times the partition function of α strongly non-intersecting paths with steps
and weights determined by Γm. The starting points are si = (2i − 2, 0) and the end points are
ej = (2n + 2α− 2j, 0), with i, j = 1, . . . , α, and the weights are functions of the cluster xm.
Q-system Cluster Algebras, Paths and Total Positivity 25
Figure 11. The identification of horizontal (a) or vertical (b) pairs of consecutive vertices on Γm and the
result on Γ′
m. The new edge on Γ′
m with weight −1 allows to subtract the contribution from paths that
do not exist on Γm. We have also represented the situation of a longer chain (c), where the identification
now requires a network of up-pointing edges with alternating weights ±1 for the suitable subtractions.
In particular, (Rα,n+m1)m(xm) is a Laurent polynomial with non-negative coefficients of the
cluster xm.
6 A new path formulation for the Ar Q-system
In Section 5, we constructed a set of transfer matrices Tm, associated with paths on the
graphs Γm, which allowed us to interpret R1,n as generating functions of weighted paths on
a graph Γm, and hence prove their positivity as a function of the seed variables xm.
In Section 4.4, we also showed that for the special case m = m0 = 0, there is an alternative
graph Γ′m0
= G̃′
r, and that one can interpret R1,n as a generating function for paths this
alternative “compactified” graph. The graph Γ′m0
has r + 1 vertices, hence the associated
transfer matrix is of size r + 1× r + 1.
We now ask the question, is there a corresponding compactified set of graphs, Γ′m, which
give a path formulation of R1,n with weights which given by functions of the seed xm for all
Motzkin paths in Mr?
It turns out that it is always possible to find a weighted graph with r + 1 vertices which
answers this question positively for each m. This corresponds to a set of transfer matrices of
a size equal to the rank of the algebra Ar. Therefore this transfer matrix allows us to make
a direct connection between our transfer matrix approach and the totally positive matrices
of [13].
6.1 Compactified graphs
Consider the collection of graphs Γm. If we are interested in the generating function of weighted
paths on them from the vertex 0 to itself, then we can make various changes in them locally
(“compactify” them) without affecting the generating function itself.
We obtain such compactified graphs from Γm by identifying pairs of neighboring vertices,
and, when necessary, adding oriented edges to cancel unwanted terms. There are two possible
ways to make such identifications. Before presenting the general case, let us illustrate the two
situations in the following subsection.
26 P. Di Francesco and R. Kedem
6.2 Examples of compactification
In fact, the first type of compactification, applied to G̃r, leads to the alternative graph G̃′
r
obtained in Section 4.4. Recall that the generating function for paths from 1 to 1 on the
graph G̃r of Fig. 3 is related to the generating function for paths from 0 to 0 via the rerooting
procedure:
ZG̃r
0,0 = 1 + ty1 × ZG̃r
1,1.
Example 6.1. The generating function ZG̃r
1,1 is equal to the generating function of paths from 1
to 1 on the graph obtained from G̃r via the following procedure (“compactification”):
1. Identifying vertex i′ with vertex i, whenever both exist, and attaching a loop with weight
tyi′,i to the resulting vertex (see Fig. 11 (a)).
2. Identifying vertex r + 2 with vertex r + 1, and attaching a loop with weight tyr+2,r+1 to
the resulting vertex.
3. Identifying vertex 1 with vertex 0, renaming the resulting vertex 1, and attaching a loop
with weight ty1 = ty1,0 to this vertex.
This is clear, since a path from 1 to 1 through the vertex i′ must have a segment i → i′ → i with
weight tyi′,i, and if it goes through the vertex r+2 it must have a segment r+1 → r+2 → r+1
with weight tyr+2,r+1. Similarly, the loop at 1 accounts for segments of the form 1 → 0 → 1.
Note that the resulting graph is G̃′
r of Fig. 7. Thus, we have the identity of Section 4.4
ZG̃r
1,1 = Z
G̃′r
1,1.
Another example of identification of vertices is the case when the two vertices are adjacent
vertices, (i, i + 1), on the spine of Γ (see Fig. 11 (b)).
Example 6.2. Consider the graph H̃k associated to an ascending Motzkin path segment of
length k (see the example for k = 3 in the lower right hand corner of Fig. 9). This is a vertical
chain of 2k + 2 vertices, numbered from 0 to 2k + 1 from bottom to top, connected by edges
oriented in both directions. The edges i + 1 → i have weights yi+1 and the edges i → i + 1 have
weights 1.
Suppose we identify the vertices a, a + 1 in this graph for some a. A path from 0 to 0 with
a step a → a + 1 is always paired with a step a + 1 → a, for a net contribution to the weight
of the path of ya+1. We associate a loop with weight ya+1 at the newly formed vertex after the
identification of vertices a and a + 1.
However, there are “forbidden” paths on the resulting graph, paths which are not inherited
from paths on H̃r. These are paths which go from a+2 → a−1 without traversing the loop. This
would correspond to going from a+2 to a−1 on H̃r without passing through the edge a+1 → a,
which is impossible. We cancel the contribution of these paths by adding an ascending edge
a− 1 → a + 2 with weight −1. The effect is precisely to subtract the weights of the forbidden
set of paths.
More generally, a succession of identifications of the type in the example above results in the
following (see Fig. 11 (c)):
Lemma 6.1. The generating function of paths from 0 to 0 on the graph H̃k with vertices
0, . . . , 2k + 1 is equal to the generating function for paths from 0 to 0 on the following com-
pactified graph H̃ ′
k:
Q-system Cluster Algebras, Paths and Total Positivity 27
1. Identify the vertices 2i + 1 and 2i + 2; rename the resulting vertex i + 1 (i = 0, . . . , k− 1).
2. Attach a loop at the vertex i + 1 with weight wi+1 = y2i+2. Other edge weights remain
unchanged.
3. Add ascending edges j → j + 2 + a (0 ≤ a ≤ k− 1− j, 0 ≤ j ≤ k− 1) with weight (−1)a+1
to the resulting graph.
An illustration of the resulting graph H̃ ′
k is given in (c) of Fig. 11.
Proof. Consider the set of paths P of the form P1P
+P2P
−P3, where Pi are fixed paths, and P+
is a path from h0 to h1 consisting of only up steps and loop steps, and P− is a path from h1
to h0 consisting only of down steps and loop steps. We furthermore restrict ourselves to paths
with the weight(P+)× weight(P−) fixed to be ywn = wn1
1 · · ·wnk
k for some n. Here, y is the
product of the weights of the down steps in P−, and wn is the total weight coming from the
loops in the path. Let f be the weight of the remaining fixed portions of the path.
Without loss of generality, we can take h0 = 0 and h1 = k + 1.
For each such path we can decompose ni = n+
i + n−i , where n+
i is the number of times the
loop with weight wi is traversed in P+, and n−i in P−. Paths which arise from paths on H̃k must
have n−i ≥ 1 (for all i) by definition. We claim that on H̃ ′
k, paths with n−i = 0 are cancelled by
paths which pass through the new ascending edges.
The key observation is that a path which has an up step going through the ascending oriented
edge i− 1 → i + a + 1 (a ≥ 0) on H̃ ′
k has n+
i = n+
i+1 = · · · = ni+a = 0.
Then the total contribution of the paths in P to the partition function is in fact
fy
∑
n+
i
+n−
i
=ni
n−
i
>0
wn = fy
∑
n+
i
+n−
i
=ni
n+
i
>0
wn
= fy
∑
n+
i +n−i =ni
wn − fy
k∑
j=1
∑
n+
i
+n−
i
=ni
n+
j
=0
wn
+ fy
∑
j1<j2
∑
n+
j1
=n+
j2
=0
wn − fy
∑
j1<j2<j3
∑
n+
j1
=n+
j2
=n+
j3
=0
wn + · · ·
= fy
k∑
a=0
(−1)a
∑
j1<···<ja
∑
n+
j1
=···=n+
ja
=0
wn.
So the alternating sum has the effect of subtracting the terms with any n−i = 0.
A path on the graph H̃ ′
k with a spine vertices skipped (by traversing ascending edges of length
> 1) comes with a total sign (−1)x where x =
∑
(length of the ascending segments − 1) = a.
That is also the sign of the term with n+
j1
= · · · = n+
ja
= 0 in the summation.
Finally, we note that any path can be decomposed into pairs of ascending and descending
segments as above, and the proof can be applied iteratively to any path. �
6.3 Definition of compactified graphs
On a graph Γm, we call a skeleton edge horizontal if it connects (a) vertices i and i′ for some i,
(b) vertices 0 and 1, or (c) the top vertex and the one below it. We call an edge vertical otherwise.
Definition 6.1. The compactified graph Γ′m with r +1 vertices is obtained from the graph Γm
via the following compactification procedure:
28 P. Di Francesco and R. Kedem
Figure 12. (a) The graph Γm of Fig. 8, and (b) the compacted graph Γ′
m. We have circled on Γm the
pairs of vertices to be identified in the compactification procedure (in red for horizontal pairs, in blue
for vertical pairs). We also represent in (c) the Motzkin path m.
1. Introduce an order on the vertices of Γm, so that i < i + 1 and i < i′ < i + 1. Number
them from 1 to 2r + 2 accordingly.
2. Identify vertices 2j − 1, 2j (j = 1, . . . , r + 1), and rename the resulting vertex j. Double
edges connecting (2j − 1, 2j) are replaced by a loop at j with weight which is the product
of the weights on the two edges. All other edges and their weights are unchanged.
3. All maximal subgraphs of the form H̃k of Γm, consisting of vertical edges only, are replaced
by compactified weighted graphs of the form H̃ ′
k, as in Lemma 6.1 (with the obvious shift
in labels).
Example 6.3. For illustration, the identification of edges in the case of the graph of Fig. 8 (c)
are: 0 ∼ 1, 2 ∼ 2′, 3 ∼ 4, 5 ∼ 5′, 6 ∼ 6′, 7 ∼ 7′, 8 ∼ 8′, 9 ∼ 9′, 10 ∼ 11 and 12 ∼ 13. See Fig. 12.
There are two maximal subgraphs of the form H̃1 are the vertices 2, 3, 4, 5 and 9, 10, 11, 12.
Each pair now corresponds to a vertex i in Γ′m, i = 1, 2, . . . , 10, which receives a loop i → i from
the identification.
Fig. 13 shows the set of compactified graphs corresponding to Fig. 9 for the case A3.
The resulting weighted graph Γ′m has r + 1 vertices labelled 1, 2, . . . , r + 1, and hence is
associated with a transfer matrix T ′
m of size r + 1× r + 1.
6.4 An alternative construction
An alternative description of the compactified graphs Γ′m is the following.
We start from the graph Γ′m0
≡ G̃′
r of Fig. 7. The loop at vertex i has weight ty2i−1 and the
edge i + 1 → i has weight ty2i, where yj = yj(m0) are as in equations (5.2), (5.3).
Decompose m into maximal segments of the form:
1) descending segments, Dα,i = ((α, m), (α + 1,m− 1), . . . , (α + i− 1,m− i + 1));
2) ascending segments, Aα,i = ((α, m), (α + 1,m + 1), . . . , (α + i− 1,m + i− 1));
3) flat segments ((α, m), (α + 1,m), . . . , (α + k − 1,m)).
Here, i ≥ 2 and k ≥ 1.
Q-system Cluster Algebras, Paths and Total Positivity 29
Figure 13. The Motzkin paths m of the fundamental domain M3 and the associated compact graphs Γ′
m,
with their vertex and edge labels. Note the up-step weights: y1,3 = y2,4 = −1 and y1,4 = 1. Mutations
are indicated by arrows.
Definition 6.2. The graph Γ′′m is the graph obtained from Γ′m0
= G̃′
r via the following steps:
1. For each descending sequence Dα,i we add descending edges α + p → α + q (0 < q + 1 <
p ≤ i) to Γ′m0
, with weights tyα+p,α+q(m), where
yα+p,α+q(m) =
α+p−1∏
j=α+q
y2j(m)
α+p−1∏
j=α+q+1
y2j−1(m)
. (6.1)
2. For each ascending sequence Aα,i, we add ascending edges α+q → α+p (0 < q+1 < p ≤ i)
with weights yα+q,α+p(m) = (−1)p−q−1.
Lemma 6.2. The weighted graph Γ′′m is identical to the weighted graph Γ′m.
This is just the result of the definition of Γm using the decomposition of m, as in Fig. 8.
Maximal subgraphs the form H̃k correspond to the maximal ascending segments of the Motzkin
path. All other segments correspond to subgraphs with horizontal edges.
6.5 Equality of generating functions
To summarize, the compactified graph Γ′m is such that
Theorem 6.1.
(1− Tm)−1
1,1 = (1− T ′
m)−1
1,1.
In other words: the partition function for weighted paths from vertex 1 to vertex 1 in Γm is
identical to that for weighted paths from vertex 1 to vertex 1 in the compact graph Γ′m.
30 P. Di Francesco and R. Kedem
7 Totally positive matrices and compactified transfer matrices
We now establish the connection between the transfer matrices T ′
m for paths on Γ′m and the
totally positive matrices of [13] corresponding to double Bruhat cells for pairs of Coxeter ele-
ments.
We may express the compact transfer matrices T ′
m of the previous section in terms of the
elementary matrices fi, ei, di for GLr+1, defined as follows. Let Eij denote the standard
elementary matrix of size (r + 1)× (r + 1), with entries (Ei,j)k,` = δk,iδ`,j .
Definition 7.1. The elementary matrices {ei, fi, di} are defined by
fi = I + λiEi+1,i, ei = I + νiEi,i+1, i ∈ {1, . . . , r},
di = I + (µi − 1)Ei,i, i ∈ {1, . . . , r + 1}. (7.1)
for some real parameters λi, µi, νi.
In [13], Fomin and Zelevinsky introduced a parametrization of totally positive matrices as
products of the form
∏
i∈I
fi
r+1∏
i=1
di
∏
j∈J
ej for I, J two suitable sets of indices, and λi, µi, νi some
positive parameters. This expression allowed to rephrase total positivity in terms of networks.
Here we interpret our compact transfer matrices in terms of some of these products.
Recall that each Motzkin path can be decomposed into descending, ascending and flat pieces,
as in Section 6.4. We introduce the increasing sequence of integers (a1, . . . , a2k), such that the jth
ascending piece of m, Aαj ,ij of m starts at mαj = a2j−1 and ends at mαj+ij−1 = a2j . Similarly
for the sequence of increasing integers (b1, . . . , b2p), which mark the starting and ending points
of the descending sequences Dαj ,i.
For i < j, let ω[i, j] denote the permutation which reverses the order of all consecutive
elements between i and j in a given sequence. That is, ω[i, j] = (i, j)(i+1, j−1)(i+2, j−2) · · · .
For example, ω[i, j]: (j, j − 1, . . . , i) 7→ (i, i + 1, . . . , j),
σm =
(
k∏
i=1
ω[a2i−1, a2i]
)
◦ (r, r − 1, . . . , 1),
τm =
(
p∏
i=1
ω[b2i−1, b2i]
)
◦ (r, r − 1, . . . , 1). (7.2)
Example 7.1. For the Motzkin path m = (2, 1, 2, 2, 2, 1, 0, 0, 1) of Fig. 8, we have the ascending
segments [2, 3] and [8, 9], while the descending segments are [1, 2] and [5, 7]. The rearranged
sequences read σm = (8, 9, 7, 6, 5, 4, 2, 3, 1) and τm = (9, 8, 5, 6, 7, 4, 3, 1, 2).
Note that the sequences σm and τm consist of increasing and decreasing subsequences of
consecutive integers, and that these subsequences and their order are unique.
One can define the decomposition of the transfer matrix T ′
m into a strictly lower-triangular
part Nm and an upper triangular part Bm, so that
T ′
m = Nm + Bm.
Lemma 7.1. The matrices Nm and Bm can be expressed as
Nm = I − (fi1fi2 · · · fir)
−1, (7.3)
Bm = t(d1d2 · · · dr+1) (ej1ej2 · · · ejr), (7.4)
where the parameters in equation (7.1) are
λi = 1, µi = y2i−1, νi =
y2i
y2i−1
. (7.5)
Q-system Cluster Algebras, Paths and Total Positivity 31
Figure 14. Pictorial representation of the action of (a) f−1
j , (b) an increasing product f−1
a f−1
a+1 · · · f
−1
b−1,
and (c) a decreasing product f−1
b−1f
−1
b−2 · · · f−1
a for some a < b. These correspond to adding (a) an
ascending edge j → j + 1 with weight −1, (b) ascending edges a + i → a + i + 1, i = 0, 1, . . . , b − a − 1
with weights −1, and (c) a web of ascending edges a + i → a + k, 0 ≤ i ≤ k− 1 ≤ b− a− 1 with weights
(−1)i+k.
Proof. We give a pictorial proof. It is possible to describe multiplication by an elementary
matrix as the addition of an arrow to a graph. In our context, Nm encodes the ascending arrows
in the graph Γ′m, and Bm the descending arrows.
First, consider the product f−1
ir
· · · f−1
i1
in Nm of equation (7.3), where (fi)−1 = I − Ei+1,i.
The sequence (ir, ir−1, . . . , i1) =
(∏
i ω[a2i−1, a2i]
)
◦ (1, . . . , r), which is simply σm written in
reverse order, consists of alternating increasing and decreasing sequences of consecutive integers.
The products of matrices corresponding to increasing subsequences are
P+
j = f−1
a2j+1f
−1
a2j+2 · · · f
−1
a2j+1−1 = I −
a2j+1−a2j−1∑
i=1
Ea2j+i+1,a2j+i
and products corresponding to decreasing sequences are
P−
j = f−1
a2j
f−1
a2j−1 · · · f
−1
a2j−1
= I +
a2j−a2j−1∑
i=1
i−1∑
k=0
(−1)i+kEa2j−1+i,a2j−1+k.
We start with the graph corresponding to the identity matrix, which is the transfer matrix
of the graph consisting of r + 1 disconnected vertices labelled 1, 2, . . . , r + 1, each with a loop of
weight 1. Multiplying on the left by f−1
j creates an ascending edge j → j + 1 with weight −1.
More generally, left multiplication by P+
j creates a succession of ascending edges a2j+1 → a2j+2,
a2j + 2 → a2j + 3, . . . , a2j+1− 1 → a2j+1. Left multiplication by P−
j creates a web of ascending
edges a2j−1 + k → a2j−1 + i, 0 ≤ k ≤ i− 1 ≤ a2j − a2j−1 − 1, with alternating weights (−1)i+k.
We illustrate the resulting actions on the graphs in Fig. 14.
Recall that the segments [a2i−1, a2i] correspond to the ascending segments of m, themselves
associated to the vertical chain-like pieces of Γm (see Fig. 8). Recall that in the identification
procedure leading to Γ′m (Definition 6.2 and Lemma 6.2), we showed that such chains must
receive a web of ascending edges with alternating weights ±1 (see Fig. 11 (c)), while all the
vertices are connected via ascending edges i → i + 1 with weight 1.
Finally, comparing this with the graph associated to (fi1fi2 · · · fik)−1 as described above,
we find that the contribution of ascending edges to T ′
m (or equivalently, Γ′m) is identical to
I − (fi1fi2 · · · fik)−1.
32 P. Di Francesco and R. Kedem
The proof of equation (7.4) is similar, but now concerns the descending edges and the loops
of the graph Γ′m.
The product td1d2 · · · dr+1 is the transfer matrix of a chain of r + 1 disconnected vertices
i = 1, 2, . . . , r + 1, each with a loop with weight ty2i−1. Multiplication on the right by ei creates
a descending edge i + 1 → i, with weight ty2i−1
y2i
y2i−1
= ty2i.
Again we divide the sequence τm into increasing subsequences of consecutive integers, (b2j−1,
b2j−1 + 1, . . . , b2j) and decreasing subsequences (b2j+1 − 1, . . . , b2j + 1). Therefore the product
ej1ej2 · · · ejr consists of “ascending” factors
Q+
j = eb2j−1
eb2j
· · · eb2j
= I +
b2j−b2j−1∑
i=1
i−1∑
k=0
yb2j−1+i,b2j−1+kEb2j−1+k,b2j−1+i
and “descending” factors
Q−
j = eb2j+1−1eb2j+1−2 · · · eb2j+1 = I +
b2j+1−b2j−1∑
i=1
y2b2j+2i
y2b2j+2i−1
Eb2j+i,b2j+i+1,
where yb+i,b+j are the weights of equation (6.1).
Recall that the segments [b2i−1, b2i] correspond to the descending segments of m, which
correspond to networks of descending edges on Γm with weights (6.1) (see Fig. 8). In our
construction 6.2 of Γ′m, these descending edges have remained unchanged, while each vertex i
received a loop with weight ty2i−1, i = 1, 2, . . . , r+1. This is nothing but the graph associated to
t(d1d2 · · · dr+1)(ej1ej2 · · · ejr), which therefore encodes the contribution of loops and descending
edges to T ′
m, and equation (7.4) follows. �
We can now make a direct connection with totally positive matrices encoding the networks
associated to the Coxeter double Bruhat cells considered in [15]. Each Motzkin path m ∈ Mr
corresponds to such an element, the (r + 1)× (r + 1) matrix Pm:
Definition 7.2. Given a Motzkin path m ∈ Mr, define
Pm = (fi1fi2 · · · fir) (d1d2 · · · dr+1) (ej1ej2 · · · ejr),
where (i1i2 · · · ir) = σm and (j1j2 · · · jr) = τm are the two sequences of (7.2). The parameters
λj , µj , νj are as in equation (7.5).
As a consequence of equations (7.3), (7.4), we have
Theorem 7.1.
(I − T ′
m)−1 = (I − tPm)−1 (fi1fi2 · · · fir),
which allows to rewrite the generating function of Theorem 6.1
(
(I − T ′
m)−1
)
1,1
=
(
(I − tPm)−1
)
1,1
+
(
(I − tPm)−1
)
1,2
if a1 > 1,
a2+1∑
a=a1
(
(I − tPm)−1
)
1,a
if a1 = 1.
This yields an interpretation of the solution R1,n+m1+1 to the Ar Q-system with initial
data xm in terms of the network associated to the totally positive matrix Pm, for all n ≥ 0.
Q-system Cluster Algebras, Paths and Total Positivity 33
Figure 15. The network corresponding to the totally positive matrix Pm for the Motzkin path m of
Example 7.1, depicted in Fig. 12 (c). We have indicated in medallions the three network representations
(in red, black, blue) for the elementary matrices fi, di and ei. The network for P ′
m corresponds to the
same picture, but with the f part (red descending elements) to the right of the e part (ascending blue
elements).
Example 7.2. For the fundamental Motzkin path m = m0 with mα = 0 for all α, we have
σm0 = τm0 = (r, r − 1, . . . , 1), and therefore Pm0 = (frfr−1 · · · f1)(d1d2 · · · dr+1)(erer−1 · · · e1).
Note that the matrix F = frfr−1 · · · f1 has entries Fi,j = 1 if i ≥ j, and 0 otherwise. One can
check directly that I − tPm0 = F (I − T ′
m0
), with T ′
m0
given by equation (4.6).
An equivalent formulation uses the explicit decomposition Pm = FDE, where F = fi1 · · · fir ,
D = d1 · · · dr+1 and E = ej1 · · · ejr :
(I − tPm)−1F =
∞∑
n=0
tn(FDE)nF =
∞∑
n=0
tnF (DEF )n = F (I − tP ′
m)−1,
where P ′
m = DEF = F−1PmF , and the fact that F is a lower uni-triangular matrix, which
implies:
(
(I − tT ′
m)−1
)
1,1
=
(
(I − tPm)−1F
)
1,1
=
(
(I − tP ′
m)−1
)
1,1
.
The matrix P ′
m is another way to write a totally positive matrix, and the network graph
corresponding to it has a slightly modified form from that of Pm. Both of these correspond to
electrical networks [13]. For illustration, we represent in Fig. 15 the network corresponding to the
matrix Pm for the Motzkin path m of Example 7.1, represented in Fig. 12 (c). The medallions
show the three elementary circuit representations for the three types of elementary matrices fi,
di, ei, each receiving the associated weight. The Lindström lemma [23] of network theory states
that the minor |P |c1,...,ck
r1,...,rk of the matrix P of the network, corresponding to a specific choice rows
r1, . . . , rk and columns c1, . . . , ck, is the partition function of k non-intersecting (vertex-disjoint)
paths starting at points r1, . . . , rk and ending at points c1, . . . , ck, and with steps taken only on
horizontal lines or along f , d or e type elements. Here we have only considered circuits with
entry and exit point 1, after possibly several iterations of the same network (each receiving the
weight t), and whose generating function is precisely the resolvent
(
(I − tP )−1
)
1,1
.
34 P. Di Francesco and R. Kedem
Figure 16. The concatenation of n copies of the network coded by P ′
m. The quantity (P ′
m)n
1,1 is the
partition function of electrical wires starting and ending at the two indicated arrows. We also represent
below a cylinder formulation à la [15]: the wire must wind n times around the cylinder before exiting.
8 Conclusion
In this paper, we have made the contact between our earlier study of the solutions of the Ar
Q-system, expressed in terms of initial data coded via Motzkin paths, and the totally positive
matrices for Coxeter double Bruhat cells. We showed in particular how the relevant pairs of
Coxeter elements were encoded in the Motzkin paths as well.
One would expect the total positivity of the transfer matrices Pm or P ′
m to be directly related
to the proof of the positivity conjecture in the case of the Ar Q-system. Our proof presented
in [7] relies on the path formulation of R1,n and on the formulation of Rα,n as the partition
function of families of strongly non-intersecting paths. The total positivity of the compactified
formulation should provide an alternative proof, using networks rather than paths.
The precise connection between paths on graphs and networks, as illustrated in Section 7
above is subtle. Indeed, the identity between resolvents implies that the partition function for
weighted paths from 1 to 1 on Γm with n descents, (Tn
m)1,1, is identical to the generating function
for circuits on a network made of n identical concatenated networks, each corresponding to the
totally positive matrix P ′
m, from connector 1 to connector 1 (see the top of Fig. 16). In [15],
this concatenation is realized by putting the network on a cylinder and allowing for the circuit
to wind n times around it before exiting (see the bottom of Fig. 16). Note that we could also
work with Pm instead, as it is related to P ′
m via cyclic symmetry.
More generally, it should be possible to relate our non-intersecting path families to networks
with multiple entries and exits, as in the setting of the Lindström lemma.
Another question concerns the cluster algebra attached to the Ar Q-system. As stressed in [7],
we have only considered a subset of the clusters which arise in the full Q-system cluster algebra,
namely those which consist of solutions of the Q-system. There are other cluster mutations,
however, which are not recursion relations of the form (4.1). One may ask about the other cluster
variables in the algebra. The positivity conjecture should hold for them as well. Preliminary
investigations show that the corresponding mutations can still be understood in terms of (finite)
continued fraction rearrangements, hence we expect them to also have a network counterpart.
These clearly can no longer correspond to Coxeter double Bruhat cells, as those are exhausted
by the solutions of the Q-system.
Finally, the connection to total positivity should be generalizable to the case of other simple
Lie algebras as well. Indeed, on the one hand the Q-systems based on other Lie algebras also
Q-system Cluster Algebras, Paths and Total Positivity 35
have cluster algebra formulations [6], while on the other hand the notion of total positivity
has been extended to arbitrary Lie groups [11]. We have evidence that hard particle and path
interpretations exist for all Q-systems, and it would be interesting to investigate their relation
to the corresponding generalized networks. The integrability of these systems is presumably
related to that of the Coxeter–Toda systems of [26]. This will be the subject of forthcoming
work.
Acknowledgements
We thank M. Gekhtman, S. Fomin, A. Postnikov, N. Reshetikhin and A. Vainshtein for use-
ful discussions. RK’s research is funded in part by NSF grant DMS-0802511. RK thanks
CEA/Saclay IPhT for their hospitality. PDF’s research is partly supported by the European
network grant ENIGMA and the ANR grants GIMP and GranMa. PDF thanks the department
of Mathematics of the University of Illinois at Urbana-Champaign for hospitality and support,
and the department of Mathematics of the University of California Berkeley for hospitality.
References
[1] Adler M., van Moerbeke P., Integrals over classical groups, random permutations, Toda and Toeplitz lattices,
Comm. Pure Appl. Math. 54 (2001), 153–205, math.CO/9912143.
[2] Assem I., Reutenauer C., Smith D., Frises, arXiv:0906.2026.
[3] Berenstein A., Private communication.
[4] Bouttier J., Di Francesco P., Guitter E., Geodesic distance in planar graphs, Nuclear Phys. B 663 (2003),
535–567, cond-mat/0303272.
Bouttier J., Di Francesco P., Guitter E., Random trees between two walls: exact partition function,
J. Phys. A: Math. Gen. 36 (2003), 12349–12366, cond-mat/0306602.
[5] Caldero P., Zelevinsky A., Laurent expansions in cluster algebras via quiver representations, Mosc. Math. J.
6 (2006), 411–429, math.RT/0604054.
[6] Di Francesco P., Kedem R., Q-systems as cluster algebras. II. Cartan matrix of finite type and the polynomial
property, Lett. Math. Phys. 89 (2009), 183–216, arXiv:0803.0362.
[7] Di Francesco P., Kedem R., Q-systems, heaps, paths and cluster positivity, Comm. Math. Phys. 293 (2010),
727–802, arXiv:0811.3027.
[8] Fomin S., Shapiro M., Thurston D., Cluster algebras and triangulated surfaces. I. Cluster complexes, Acta
Math. 201 (2008), 83–146, math.RA/0608367.
[9] Fomin S., Zelevinsky A., Cluster algebras. I. Foundations, J. Amer. Math. Soc. 15 (2002), 497–529,
math.RT/0104151.
[10] Fomin S., Zelevinsky A., Cluster algebras. IV. Coefficients, Compos. Math. 143 (2007), 112–164,
math.RA/0602259.
[11] Fomin S., Zelevinsky A., Double Bruhat cells and total positivity, J. Amer. Math. Soc. 12 (1999), 335–380,
math.RT/9802056.
[12] Fomin S., Zelevinsky A., The Laurent phenomenon, Adv. in Appl. Math. 28 (2002), 119–144,
math.CO/0104241.
[13] Fomin S., Zelevinsky A., Total positivity: tests and parameterizations, Math. Intelligencer 22 (2000), 23–33,
math.RA/9912128.
[14] Geiss C., Leclerc B., Schröer J., Preprojective algebras and cluster algebras, in Trends in Representation
Theory of Algebras and Related Topics, EMS Ser. Congr. Rep., Eur. Math. Soc., Zürich, 2008, 253–283,
arXiv:0804.3168.
[15] Gekhtman M., Shapiro M., Vainshtein A., Generalized Bäcklund–Darboux transformations for Coxeter–
Toda flows from cluster algebra perspective, arXiv:0906.1364.
[16] Gessel I.M., Viennot X., Binomial determinants, paths and hook formulae, Adv. in Math. 58 (1985), 300–321.
[17] Kazakov V., Kostov I., Nekrasov N., D-particles, matrix integrals and KP hierachy, Nuclear Phys. B 557
(1999), 413–442, hep-th/9810035.
http://dx.doi.org/10.1002/1097-0312(200102)54:2<153::AID-CPA2>3.0.CO;2-5
http://arxiv.org/abs/math.CO/9912143
http://arxiv.org/abs/0906.2026
http://dx.doi.org/10.1016/S0550-3213(03)00355-9
http://arxiv.org/abs/cond-mat/0303272
http://dx.doi.org/10.1088/0305-4470/36/50/001
http://arxiv.org/abs/cond-mat/0306602
http://arxiv.org/abs/math.RT/0604054
http://dx.doi.org/10.1007/s11005-009-0354-z
http://arxiv.org/abs/0803.0362
http://dx.doi.org/10.1007/s00220-009-0947-5
http://arxiv.org/abs/0811.3027
http://dx.doi.org/10.1007/s11511-008-0030-7
http://dx.doi.org/10.1007/s11511-008-0030-7
http://arxiv.org/abs/math.RA/0608367
http://dx.doi.org/10.1090/S0894-0347-01-00385-X
http://arxiv.org/abs/math.RT/0104151
http://dx.doi.org/10.1112/S0010437X06002521
http://arxiv.org/abs/math.RA/0602259
http://dx.doi.org/10.1090/S0894-0347-99-00295-7
http://arxiv.org/abs/math.RT/9802056
http://dx.doi.org/10.1006/aama.2001.0770
http://arxiv.org/abs/math.CO/0104241
http://arxiv.org/abs/math.RA/9912128
http://arxiv.org/abs/0804.3168
http://arxiv.org/abs/0906.1364
http://dx.doi.org/10.1016/0001-8708(85)90121-5
http://dx.doi.org/10.1016/S0550-3213(99)00393-4
http://arxiv.org/abs/hep-th/9810035
36 P. Di Francesco and R. Kedem
[18] Kedem R., Q-systems as cluster algebras, J. Phys. A: Math. Theor. 41 (2008), 194011, 14 pages,
arXiv:0712.2695.
[19] Keller B., Cluster algebras, quiver representations and triangulated categories, arXiv:0807.1960.
[20] Kirillov A.N., Reshetikhin N.Yu., Representations of Yangians and multiplicity of occurrence of the irre-
ducible components of the tensor product of representations of simple Lie algebras, J. Soviet Math. 52
(1990), 3156–3164.
[21] Krattenthaler C., The theory of heaps and the Cartier–Foata monoid, Appendix to the electronic repub-
lication of Cartier P., Foata D., Problèmes combinatoires de commutation et réarrangements, available at
http://www.mat.univie.ac.at/∼slc/books/cartfoa.html.
[22] Krichever I., Lipan O., Wiegmann P., Zabrodin A., Quantum integrable systems and elliptic solutions of
classical discrete nonlinear equations, Comm. Math. Phys. 188 (1997), 267–304, hep-th/9604080.
[23] Lindström B., On the vector representations of induced matroids, Bull. London Math. Soc. 5 (1973), 85–90.
[24] Musiker G., Schiffler R., Williams L., Positivity for cluster algebras from surfaces, arXiv:0906.0748.
[25] Propp J., Musiker G., Combinatorial interpretations for rank-two cluster algebras of affine type, Electron.
J. Combin. 14 (2007), no. 1, R15, 23 pages, math.CO/0602408.
[26] Reshetikhin N., Characteristic systems on Poisson Lie groups and their quantization, in Integrable Systems:
from Classical to Quantum (Montreal, QC, 1999), CRM Proc. Lecture Notes, Vol. 26, Amer. Math. Soc.,
Providence, RI, 2000, 165–188, math.CO/0402452.
[27] Sherman P., Zelevinsky A., Positivity and canonical bases in rank 2 cluster algebras of finite and affine
types, Mosc. Math. J. 4 (2004), 947–974, math.RT/0307082.
[28] Viennot X., Heaps of pieces. I. Basic definitions and combinatorial lemmas, in Combinatoire Énumérative,
Editors G. Labelle and P. Leroux, Lecture Notes in Math., Vol. 1234, Springer, Berlin, 1986, 321–350.
http://dx.doi.org/10.1088/1751-8113/41/19/194011
http://arxiv.org/abs/0712.2695
http://arxiv.org/abs/0807.1960
http://www.mat.univie.ac.at/~slc/books/cartfoa.html
http://dx.doi.org/10.1007/s002200050165
http://arxiv.org/abs/hep-th/9604080
http://dx.doi.org/10.1112/blms/5.1.85
http://arxiv.org/abs/0906.0748
http://arxiv.org/abs/math.CO/0602408
http://arxiv.org/abs/math.CO/0402452
http://arxiv.org/abs/math.RT/0307082
1 Introduction
2 Partition functions
2.1 Hard particles on Gr
2.2 Transfer matrix on the dual graph
2.3 Hard particles and paths
3 Application to rank 2 cluster algebras of affine type
3.1 Rank two cluster algebras
3.2 The (2,2) case: solution and path interpretation
3.3 The (1,4) case: solution and relation to paths
4 Application to the Ar Q-system
4.1 Definition
4.2 Conserved quantities of Q-systems as hard particle partition functions
4.3 Q-system solutions and paths
4.4 An alternative path formulation
5 Cluster algebra formulation: mutations and paths for the Ar Q-system
5.1 The Q-system as cluster algebra
5.2 Target graphs and weights
5.3 Mutations, paths and continued fraction rearrangements
5.4 Q-system solutions as strongly non-intersecting paths
6 A new path formulation for the Ar Q-system
6.1 Compactified graphs
6.2 Examples of compactification
6.3 Definition of compactified graphs
6.4 An alternative construction
6.5 Equality of generating functions
7 Totally positive matrices and compactified transfer matrices
8 Conclusion
References
|