Jacobians Alexander Mednykh Sobolev Institute of Mathematics - - PowerPoint PPT Presentation

jacobians
SMART_READER_LITE
LIVE PREVIEW

Jacobians Alexander Mednykh Sobolev Institute of Mathematics - - PowerPoint PPT Presentation

Jacobians Alexander Mednykh Sobolev Institute of Mathematics Novosibirsk State University PhD Summer School in Discrete Mathematics Rogla, Slovenia 27 June - 03 July 2015 Mednykh A. D. (Sobolev Institute of Math) Jacobians 27 June - 03 July


slide-1
SLIDE 1

Jacobians

Alexander Mednykh

Sobolev Institute of Mathematics Novosibirsk State University

PhD Summer School in Discrete Mathematics Rogla, Slovenia

27 June - 03 July 2015

Mednykh A. D. (Sobolev Institute of Math) Jacobians 27 June - 03 July 2015 1 / 20

slide-2
SLIDE 2

Jacobians

Basic definitions The notion of the Jacobian group of graph (also known as the Picard group, critical group, sandpile group, dollar group) was independently given by many authors (R. Cori and D. Rossin, M. Baker and S. Norine, N.

  • L. Biggs, R. Bacher, P. de la Harpe and T. Nagnibeda). This is a very

important algebraic invariant of a finite graph. In particular, the order of the Jacobain group coincides with the number of spanning trees for a

  • graph. Following M. Baker and S. Norine we define the Jacobian group of

a graph as follows.

Mednykh A. D. (Sobolev Institute of Math) Jacobians 27 June - 03 July 2015 2 / 20

slide-3
SLIDE 3

Jacobians

Let G be a finite, connected multigraph without loops. Let V (G) and E(G) be the sets of vertices and edges of G, respectively. Denote by Div(G) a free Abelian group on V (G). We refer to elements of Div(G) as divisors on G. Each element D ∈ Div(G) can be uniquely presented as D =

  • x∈V (G)

D(x)(x), D(x) ∈ Z. We define the degree of D by the formula deg(D) =

  • x∈V (G)

D(x). Denote by Div0(G) the subgroup of Div(G) consisting of divisors of degree zero.

Mednykh A. D. (Sobolev Institute of Math) Jacobians 27 June - 03 July 2015 3 / 20

slide-4
SLIDE 4

Jacobians

Let f be a Z-valued function on V (G). We define the divisor of f by the formula div(f ) =

  • x∈V (G)
  • xy∈E(G)

(f (x) − f (y))(x). The divisor div(f ) can be naturally identified with the graph-theoretic Laplacian ∆f of f . Divisors of the form div(f ), where f is a Z-valued function on V (G), are called principal divisors. Denote by Prin(G) the group of principal divisors of G. It is easy to see that every principal divisor has a degree zero, so that Prin(G) is a subgroup of Div0(G). The Jacobian group (or Jacobian) of G is defined to be the quotient group Jac(G) = Div0(G)/Prin(G). By making use of the Kirchhoff Matrix-Tree theorem one can show that Jac(G) is a finite Abelian group of order t(G), where t(G) is the number

  • f spanning trees of G. An arbitrary finite Abelian group is the Jacobian

group of some graph.

Mednykh A. D. (Sobolev Institute of Math) Jacobians 27 June - 03 July 2015 4 / 20

slide-5
SLIDE 5

Jacobians

Jacobians and flows By results of M. Baker and S. Norine the Jacobian Jac(G) is an Abelian group generated by flows ω(e), e ∈ E, whose defining relations are given by the two following Kirchhoff’s laws. (K1) The flow through each vertex of G is equal to zero. It means that

  • e∈

E,t(e)=x

ω(e) = 0 for all x ∈ V (G). (K2) The flow along each closed orientable walk W in G is equal to zero. That is

  • e∈W

ω(e) = 0.

Mednykh A. D. (Sobolev Institute of Math) Jacobians 27 June - 03 July 2015 5 / 20

slide-6
SLIDE 6

Jacobians

The Smith normal form Let A be a finite Abelian group generated by x1, x2, . . . , xn and satisfying the system of relations

n

  • j=1

aijxj = 0, i = 1, . . . , m, where A = {aij} is an integer m × n matrix. Set dj, j = 1, . . . , r, for the greatest common divisor of all j × j minors of A. Then, A ∼ = Zd1 ⊕ Zd2/d1 ⊕ Zd3/d2 ⊕ · · · ⊕ Zdr/dr−1.

Mednykh A. D. (Sobolev Institute of Math) Jacobians 27 June - 03 July 2015 6 / 20

slide-7
SLIDE 7

Jacobians

The Smith normal form Two integral matrices A and B are equivalent (written A ∼ B) if there are unimodular matrices P and Q such that B = PAQ. Equivalently, B is

  • btained from A by a sequence of elementary row and column operations:

(1) the interchange of two rows or columns, (2) the multiplication of any row or column by −1, (3) the addition of any integer times of one row (resp. column) to another row (resp. column). It is easy to see that A ∼ B implies that coker(A) ∼ coker(B). The Smith normal form is a diagonal canonical form for our equivalence relation: every n × n integral matrix A is equivalent to a unique diagonal matrix diag(s1(A), . . . , sn(A)), where si(A) divides si+1(A) for i = 1, 2, . . . , n − 1. The i-th diagonal entry of the Smith normal form of A is usually called the i-th invariant factor of A. We will use the fact that the values si(A) can also be interpreted as follows: for each i, the product s1(A)s2(A) · · · si(A) is the greatest common divisor of all i × i minors of A.

Mednykh A. D. (Sobolev Institute of Math) Jacobians 27 June - 03 July 2015 7 / 20

slide-8
SLIDE 8

Jacobians

Embedding graphs into Jacobians. Abel-Jacobi map. For a fixed base point x0 ∈ V (G) we define the Abel-Jacobi map Sx0 : G → Jac(G) by the formula Sx0(x) = [(x) − (x0)], where [d] is the equivalence class of divisor d. The continuous version of the following theorem is well known in complex analysis.

Theorem (M. Baker and S. Norine, 2009)

If graph G is 2-edge connected (=bridgeless) then Sx0 is an imbedding.

Mednykh A. D. (Sobolev Institute of Math) Jacobians 27 June - 03 July 2015 8 / 20

slide-9
SLIDE 9

Jacobians

Jacobians and Laplacians Consider the Laplacian matrix L(G) as a homomorphism Z|V | → Z|V |, where |V | = |V (G)| is the number of vertices of G. Then coker(L(G)) = Z|V |/im(L(G)) is an abelian group. Let coker(L(G)) ∼ = Zt1 ⊕ Zt2 ⊕ · · · ⊕ Zt|V |, be its Smith normal form satisfying ti

  • ti+1, (1 ≤ i ≤ |V |). If graph G is

connected then the groups Zt1, Zt1, . . . Zt|V |−1 are finite and Zt|V | = Z. In this case, Jac(G) = Zt1 ⊕ Zt2 ⊕ · · · ⊕ Zt|V |−1 is the Jacobian group of the graph G.

Mednykh A. D. (Sobolev Institute of Math) Jacobians 27 June - 03 July 2015 9 / 20

slide-10
SLIDE 10

Jacobians

Jacobians and harmonic maps Let ϕ : G → G ′ be a harmonic morphism. We define the push-forward homomorphism ϕ∗ : Div(G) → Div(G ′) by ϕ∗(D) =

  • x∈V (G)

D(x)ϕ(x). Similarly, we define the pullback homomorphism ϕ∗ : Div(G ′) → Div(G) by ϕ∗(D′) =

  • x∈V (G ′)
  • x∈V (G), ϕ(x)=y

mϕ(x)D′(y)x =

  • x∈V (G)

mϕ(x)D′(ϕ(x))x.

Mednykh A. D. (Sobolev Institute of Math) Jacobians 27 June - 03 July 2015 10 / 20

slide-11
SLIDE 11

Jacobians

Jacobians and harmonic maps We note that if ϕ : G → G ′ is a harmonic morphism and D′ ∈ Div(G ′), then deg(ϕ∗(D′)) = deg(ϕ) · deg(D′). Since, deg(ϕ) =

  • x∈V (G), ϕ(x)=y

mϕ(x) we also have the following simple formula:

Lemma 1.

Let ϕ : G → G ′ be a harmonic morphism, and let D′ ∈ Div(G ′). Then ϕ∗(ϕ∗(D′)) = deg(ϕ)D′.

Mednykh A. D. (Sobolev Institute of Math) Jacobians 27 June - 03 July 2015 11 / 20

slide-12
SLIDE 12

Jacobians

Jacobians and harmonic maps Suppose ϕ : G → G ′ is a harmonic morphism and that f : V (G) → A and f ′ : V (G ′) → A are functions, where A is an abelian group. We define ϕ∗f : V (G ′) → A by ϕ∗f (y) := deg(ϕ) =

  • x∈V (G), ϕ(x)=y

mϕ(x)f (x) and ϕ∗f ′ : V (G) → A by ϕ∗f ′ := f ′ ◦ ϕ.

Lemma 2.

Let ϕ : G → G ′ be a harmonic morphism, and f : V (G) → Z and f ′ : V (G ′) → Z are functions. Then ϕ∗(Div(f )) = Div(ϕ∗f ), ϕ∗(Prin(G)) ⊆ Prin(G ′) and ϕ∗(Div(f )) = Div(ϕ∗f ), ϕ∗(Prin(G ′)) ⊆ Prin(G).

Mednykh A. D. (Sobolev Institute of Math) Jacobians 27 June - 03 July 2015 12 / 20

slide-13
SLIDE 13

Jacobians

Jacobians and harmonic maps As a consequence of Lemma 2, we see that ϕ induces group homomorphisms (which we continue to denote by ϕ∗, ϕ∗) ϕ∗ : Jac(G) → Jac(G ′), ϕ∗ : Jac(G ′) → Jac(G). It is straightforward to check that if ψ : G → G ′ and ϕ : G ′ → G ′′ are harmonic morphisms and D ∈ Div(G), D′′ ∈ Div(G ′′), then ϕ ◦ ψ : G → G ′′ is harmonic, and we have (ϕ ◦ ψ)∗(D) = ϕ∗(ψ∗(D)) and (ϕ ◦ ψ)∗(D′′) = ψ∗(ϕ∗(D′′)). Therefore we obtain two different functors from the category of graphs to the category of abelian groups: a covariant “Albanese” functor (G → Jac(G), ϕ → ϕ∗) and a contravariant “Picard” functor (G → Jac(G), ϕ → ϕ∗).

Mednykh A. D. (Sobolev Institute of Math) Jacobians 27 June - 03 July 2015 13 / 20

slide-14
SLIDE 14

Jacobians

Jacobians and harmonic maps As a result we the following two important theorems.

Theorem

Let ϕ : G → G ′ be a nonconstant harmonic morphism. Then ϕ∗ : Jac(G) → Jac(G ′) is surjective.

Theorem

Let ϕ : G → G ′ be a nonconstant harmonic morphism. Then ϕ∗ : Jac(G ′) → Jac(G) is injective.

Mednykh A. D. (Sobolev Institute of Math) Jacobians 27 June - 03 July 2015 14 / 20

slide-15
SLIDE 15

Jacobians

Jacobians and harmonic maps Let t(G) = |Jac(G)| denote the number of spanning trees in a graph G. From each of the two above theorems we immediately deduce the following corollary, a special case of which is due to K. A. Berman and D. Treumann.

Corollary

If there exists a nonconstant harmonic morphism from G to G ′, then t(G ′) divides t(G).

Mednykh A. D. (Sobolev Institute of Math) Jacobians 27 June - 03 July 2015 15 / 20

slide-16
SLIDE 16

Jacobians Exercises

Exercise 5.1. Let G be a tree. Show that Jac(G) = 0. Exercise 5.2. Let Cn be a cyclic graph on n vertices. Show that Jac(Cn) = Zn.

Mednykh A. D. (Sobolev Institute of Math) Jacobians 27 June - 03 July 2015 16 / 20

slide-17
SLIDE 17

Jacobians

Exercise 5.3. Let Kn be the complete graph on n vertices. Prove that Jac(Kn) = Zn−2

n

. Exercise 5.4. Let Km,n be the complete bipartite graph. Prove that Jac(Km,n) = Zn−2

m

⊕ Zm−2

n

⊕ Zmn. (See: D.J. Lorenzini, A finite group attached to the Laplacian of a graph. Discrete Math. 91 (1991), 277–282.) Exercise 5.5. Let X be a finite connected graph. Denote by X the graph obtained from X by collapsing all bridges of X to vertices. Prove Jac(X) = Jac(X).

Mednykh A. D. (Sobolev Institute of Math) Jacobians 27 June - 03 July 2015 17 / 20

slide-18
SLIDE 18

Jacobians

Exercise 5.6. Let X1 and X2 be connected graphs sharing a common vertex. Show that Jac(X1 + X2) = Jac(X1) ⊕ Jac(X2). Exercise 5.7. Let A = Zn1 ⊕ Zn2 ⊕ . . . ⊕ Znr be a finite Abelian group and X = Cn1 + Cn2 + . . . + Cnr . Show that Jac(X) ∼ = A. Exercise 5.8. Let e be an edge of graph X such that X \ e = X1 ∪ X2 is a disjoint union

  • f two connected graphs X1 and X2. Prove that

Jac(X) = Jac(X1) ⊕ Jac(X2).

Mednykh A. D. (Sobolev Institute of Math) Jacobians 27 June - 03 July 2015 18 / 20

slide-19
SLIDE 19

Jacobians

Exercise 5.9. Describe all bridgeless graphs X with Jac(X) = G, where (i) G = Z2, (ii) G = Z3, (iii) G = Z2 ⊕ Z2, (iv) G = Z4, (v) G = Z5. Exercise 5.10. Let X and X ∗ be dual planar graphs. Prove that Jac(X) = Jac(X ∗).

Mednykh A. D. (Sobolev Institute of Math) Jacobians 27 June - 03 July 2015 19 / 20

slide-20
SLIDE 20

Jacobians

Exercise 5.11. Find the Jacobian J(Wn) of the wheel graph on n + 1 vertices. Answer: Jac(Wn) = Zℓn ⊕ Zℓn, if n is odd and Jac(Wn) = Zfn ⊕ Z5fn, if n is even. Here ℓj is j-th Lukas number and fk is k-th Fibonacci number. ℓ1 = 1, ℓ2 = 3, ℓk+2 = ℓk+1 + ℓk, k ≥ 1. f1 = 1, f2 = 1, fk+2 = fk+1 + fk, k ≥ 1. In particular, Jac(W2) = Z1 ⊕ Z5, Jac(W3) = Z4 ⊕ Z4, Jac(W4) = Z3 ⊕ Z15, Jac(W5) = Z11 ⊕ Z11, Jac(W6) = Z8 ⊕ Z40.

Mednykh A. D. (Sobolev Institute of Math) Jacobians 27 June - 03 July 2015 20 / 20