Properties of GKM graphs

These are some of the main properties of GKM graphs.

General Properties

The following properties make sense for any abstract GKM graph.

GKMtools.valency — Function
valency(G::AbstractGKMGraph) -> Int64

Return the valency of G, i.e. the number of flags at each vertex.

Warning

This function does not check if G is a valid GKM graph (use isvalid to check this). In particular, it does not check if every vertex has the same degree. The returned value is the degree of vertex 1.

Example:

The valency of the GKM graph of $\mathbb{P}^3$ is 3, since all of the fixed points $[1:0:0:0], \dots, [0:0:0:1]$ are connected to each other via some $T$-invariant $\mathbb{P}^1$'s. For example, $[1:0:0:0]$ and $[0:1:0:0]$ are connected by $\{[x:y:0:0] : x,y\in\mathbb{C}\}$.

julia> valency(projective_space(GKMGraph, 3))
3
julia> valency(grassmannian(GKMGraph, 2, 4)) # The Grassmannian of 2-planes in C^4
4
julia> valency(flag_variety(GKMGraph, [1, 1, 1, 1])) # The variety of full flags in C^4
6
source
GKMtools.rank_torus — Function
rank_torus(G::AbstractGKMGraph) -> Int64

Return the rank of the torus acting on G. That is, the rank of the character group.

Examples

By default, the torus acting on $\mathbb{P}^n$ is $(\mathbb{C}^\times)^{n+1}$, acting by rescaling the homogeneous coordinates.

julia> P3 = projective_space(GKMGraph, 3);

julia> rank_torus(P3)
4

Taking products adds the rank:

julia> H6 = gkm_graph_of_toric(hirzebruch_surface(NormalToricVariety, 6));

julia> rank_torus(H6)
4
julia> rank_torus(H6 * P3)
8
source
GKMtools.is_compact — Function
is_compact(G::AbstractGKMGraph) -> Bool

Return true if G is compact, i.e. all flags at all vertices are associated with edges (no standalone flags).

Example

julia> G = projective_space(GKMGraph, 2)
GKM graph with 3 nodes, valency 2 and axial function:
2 -> 1 => (1, -1, 0)
3 -> 1 => (1, 0, -1)
3 -> 2 => (0, 1, -1)
Birkhoff-Grothendieck connection for GKM graph with 3 nodes and valency 2

julia> is_compact(G)
true

julia> A3 = gkm_graph_of_toric(affine_space(NormalToricVariety, 3)) # affine space
GKM graph with 1 nodes, valency 3 and axial function:
Standalone flags:
Algorithmic connection for GKM graph with 1 nodes and valency 3

julia> is_compact(A3)
false
source
GKMtools.gkm_independence — Function
gkm_independence(G::AbstractGKMGraph) -> Int64

Return the maximum integer in i in 0,1,...,valency(G) such that the GKM graph $G$ is $i$-independent. The GKM graph is $i$-independent if for each vertex $v$, each $i$-tuple of flags at $v$ has linearly independent axial function values.

Example

julia> G = gkm_3d_twisted_flag()
GKM graph with 6 nodes, valency 3 and axial function:
2 -> 1 => (0, -1)
3 -> 2 => (1, 0)
4 -> 1 => (1, -2)
4 -> 3 => (-1, 1)
5 -> 2 => (1, -1)
5 -> 4 => (0, -1)
6 -> 1 => (1, -1)
6 -> 3 => (2, -1)
6 -> 5 => (1, 0)
Algorithmic connection for GKM graph with 6 nodes and valency 3

julia> gkm_independence(G)
2
source
GKMtools.is2_indep — Function
is2_indep(G::AbstractGKMGraph) -> Bool

Return true if G is 2-independent, i.e. the weights of every two flags at a vertex are linearly independent.

source
GKMtools.is3_indep — Function
is3_indep(G::AbstractGKMGraph) -> Bool

Return true if G is 3-independent, i.e. the weights of every three flags at a vertex are linearly independent.

Example

The weights of $\mathbb{P}^3$ at the fixed point $[1:0:0:0]$ are $\{t_i-t_0:i\in\{1, 2, 3\}\}$, which are linearly independent over $\mathbb{C}$.

julia> is3_indep(projective_space(GKMGraph, 3))
true

The variety of complete flags in $\mathbb{C}^3$ is an example of a GKM graph that is not 3-independent:

julia> G = flag_variety(GKMGraph, [1, 1, 1])
GKM graph with 6 nodes, valency 3 and axial function:
13 -> 12 => (0, 1, -1)
21 -> 12 => (1, -1, 0)
31 -> 13 => (1, 0, -1)
31 -> 21 => (0, 1, -1)
23 -> 13 => (1, -1, 0)
23 -> 21 => (1, 0, -1)
32 -> 12 => (1, 0, -1)
32 -> 31 => (1, -1, 0)
32 -> 23 => (0, 1, -1)
Birkhoff-Grothendieck connection for GKM graph with 6 nodes and valency 3

julia> is3_indep(G)
false
Warning

This function throws an error if the valency of G is less than 3, since in this case it is not possible to pick three different flags at a vertex.

source

Symplectic / almost complex properties

The following properties are well-defined as soon as there is a well-defined first Chern class. This holds for GKM graphs coming from $T$-compatibly almost complex manifolds and from $T$-compatibly symplectic manifolds. In the latter case, all choices of almost complex structures compatible with the symplectic form yield the same first Chern class, making the properties below intrinsic to the symplectic structure.

GKMtools.fano_index — Function
fano_index(G::AbstractGKMGraph) -> ZZRingElem

Return the Fano index of the GKM graph, which is the greatest common divisor of the first Chern numbers of all its edges.

Examples

julia> P2 = projective_space(GKMGraph, 2);

julia> fano_index(P2)
3

julia> F3 = flag_variety(GKMGraph, [1, 1, 1]);

julia> fano_index(F3)
2
source
GKMtools.pseudo_index — Function
pseudo_index(G::AbstractGKMGraph) -> ZZRingElem

Return the pseudo index of the GKM graph, which is the minimum of the first Chern numbers of all its edges.

Examples

The examples below show that the pseudo-index differs from the Fano index in general.

julia> P2 = projective_space(GKMGraph, 2);

julia> fano_index(P2), pseudo_index(P2)
(3, 3)

julia> P3 = projective_space(GKMGraph, 3);

julia> fano_index(P3), pseudo_index(P3)
(4, 4)

julia> G = P2 * P3;

julia> fano_index(G), pseudo_index(G)
(1, 3)

julia> T = gkm_3d_twisted_flag();

julia> fano_index(T), pseudo_index(T)
(2, 0)
source
GKMtools.is_strictly_nef — Function
is_strictly_nef(G::AbstractGKMGraph) -> Bool

Return true if and only if the Chern numbers of all curve classes corresponding to edges of the GKM graph are strictly positive.

Examples

julia> F3 = flag_variety(GKMGraph, [1,1,1]);

julia> print_curve_classes(F3)
13 -> 12: (1, 0), Chern number: 2
21 -> 12: (0, 1), Chern number: 2
31 -> 13: (0, 1), Chern number: 2
31 -> 21: (1, 1), Chern number: 4
23 -> 13: (1, 1), Chern number: 4
23 -> 21: (1, 0), Chern number: 2
32 -> 12: (1, 1), Chern number: 4
32 -> 31: (1, 0), Chern number: 2
32 -> 23: (0, 1), Chern number: 2

julia> is_strictly_nef(F3)
true

julia> H5 = gkm_graph_of_toric(hirzebruch_surface(NormalToricVariety, 5))
GKM graph with 4 nodes, valency 2 and axial function:
2 -> 1 => (-1, 0, 1, 0)
3 -> 2 => (-5, -1, 0, 1)
4 -> 1 => (0, -1, -5, 1)
4 -> 3 => (1, 0, -1, 0)
Algorithmic connection for GKM graph with 4 nodes and valency 2

julia> print_curve_classes(H5)
2 -> 1: (-5, 1), Chern number: -3
3 -> 2: (1, 0), Chern number: 2
4 -> 1: (1, 0), Chern number: 2
4 -> 3: (0, 1), Chern number: 7

julia> is_strictly_nef(H5)
false
source

Compact Hamiltonian properties

The following properties have natural geometric meaning for GKM graphs of compact Hamiltonian GKM spaces. That being said, they can be defined and checked for any abstract GKM graph, although one needs to be careful about their geometric interpretation outside of the Hamiltonian setting.

Oscar.betti_numbers — Function
betti_numbers(G::AbstractGKMGraph) -> Vector{Int64}

Return the array betti_numbers such that betti_numbers[i+1] is the $2i$-th combinatorial Betti number for $i$ from $0$ to the valency of G, as defined in [GZ01, section 1.3].

Note
  • i ranges from 0 to valency(G).
  • From [GZ01, Theorem 1.3.2], the combinatorial Betti numbers equal the Betti numbers of the underlying GKM space if the torus action is Hamiltonian.
Warning
  • betti_numbers[1] is the 0-th Betti number, since Julia arrays are 1-based.
  • Currently only implemented for compact GKM spaces.

Examples

julia> H6 = gkm_graph_of_toric(hirzebruch_surface(NormalToricVariety, 6));

julia> betti_numbers(H6)
3-element Vector{Int64}:
 1
 2
 1
source
GKMtools.index_periodic_betti — Function
index_periodic_betti(G::AbstractGKMGraph) -> Vector{Int64}

Return the index-periodic Betti numbers of a compact GKM graph.

For a GKM graph with Fano index $p$, this function computes the sums of Betti numbers grouped by their residue class modulo $p$.

Examples

julia> P2 = projective_space(GKMGraph, 2);

julia> index_periodic_betti(P2)
3-element Vector{Int64}:
 1
 1
 1
source

Index properties

GKMtools.is_generic — Function
is_generic(G::AbstractGKMGraph, xi) -> Bool

Return true if xi is a generic element of the weight lattice lattice(G), i.e. the Euclidean pairing of xi with every flag weight of G is nonzero.

Example

julia> P2 = projective_space(GKMGraph, 2);

julia> g1, g2, g3 = gens(lattice(P2));

julia> is_generic(P2, g1)
false

julia> is_generic(P2, g1 + g2)
false

julia> is_generic(P2, g1 + 2*g2 + 4*g3)
true
source
GKMtools.xi_index — Function
xi_index(G::AbstractGKMGraph, xi, v::Int) -> Int

Return the xi-index of vertex v of G, i.e. the number of flags at v whose weight pairs negatively with xi. Requires xi to be generic at v.

Example

julia> P2 = projective_space(GKMGraph, 2);

julia> g1, g2, g3 = gens(lattice(P2));

julia> xi = g1 + 2*g2 + 4*g3;

julia> [xi_index(P2, xi, v) for v in 1:3]
3-element Vector{Int64}:
 0
 1
 2
source
GKMtools.is_index_increasing — Function
is_index_increasing(G::AbstractGKMGraph, xi) -> Bool

Return true if xi is generic and index increasing on G. For each edge e = {v, w}, orient it as (v, w) so that the weight of Edge(v, w) pairs positively with xi; then require the xi-index at w to be strictly greater than the xi-index at v.

Example

julia> P2 = projective_space(GKMGraph, 2);

julia> g1, g2, g3 = gens(lattice(P2));

julia> is_index_increasing(P2, g1 + 2*g2 + 4*g3)
true

julia> is_index_increasing(P2, g1)
false
source
GKMtools.is_weakly_index_increasing — Function
is_weakly_index_increasing(G::AbstractGKMGraph, xi) -> Bool

Return true if xi is generic and weakly index increasing on G. For each edge e = {v, w}, orient it as (v, w) so that the weight of Edge(v, w) pairs positively with xi; then require the xi-index at w to be greater than or equal to the xi-index at v.

Example

julia> P2 = projective_space(GKMGraph, 2);

julia> g1, g2, g3 = gens(lattice(P2));

julia> is_weakly_index_increasing(P2, g1 + 2*g2 + 4*g3)
true

julia> is_weakly_index_increasing(P2, g1)
false
source
GKMtools.generic_xi_representatives — Function
generic_xi_representatives(G::AbstractGKMGraph) -> Vector

Return one representative per connected component of the set of generic directions xi in the weight lattice of G. Each representative lives in lattice(G) (as a primitive element of the lattice if lattice(G) is a free ZZ-module).

The components of the generic locus are the open chambers of the hyperplane arrangement cut out by the flag weights of G, viewed in lattice(G) tensored with QQ.

Example

On P^1 there is a single hyperplane so two chambers:

julia> P1 = projective_space(GKMGraph, 1);

julia> length(generic_xi_representatives(P1))
2

On P^2 there are 6 chambers corresponding to the 6 orderings of the three coordinates of xi:

julia> P2 = projective_space(GKMGraph, 2);

julia> reps = generic_xi_representatives(P2);

julia> length(reps)
6

julia> all(xi -> is_generic(P2, xi), reps)
true

Non-compact example: the total space of O(1) + O(-1) on P^1:

julia> T = total_space(vector_bundle_O(1, [1, -1]));

julia> length(generic_xi_representatives(T))
8
source
GKMtools.index_increasing_xi_representatives — Function
index_increasing_xi_representatives(G::AbstractGKMGraph) -> Vector

Return one representative per connected component of the generic locus on which xi is index increasing. Each representative lives in lattice(G).

Example

Every chamber of P^2 is index-increasing:

julia> P2 = projective_space(GKMGraph, 2);

julia> reps = index_increasing_xi_representatives(P2);

julia> length(reps)
6

julia> all(xi -> is_index_increasing(P2, xi), reps)
true

For the non-compact total space of O(1) + O(-1) on P^1, only some chambers are strictly index-increasing (compare the count with the 8 generic chambers):

julia> T = total_space(vector_bundle_O(1, [1, -1]));

julia> length(index_increasing_xi_representatives(T))
6
source
GKMtools.weakly_index_increasing_xi_representatives — Function
weakly_index_increasing_xi_representatives(G::AbstractGKMGraph) -> Vector

Return one representative per connected component of the generic locus on which xi is weakly index increasing. Each representative lives in lattice(G).

Example

julia> P2 = projective_space(GKMGraph, 2);

julia> reps = weakly_index_increasing_xi_representatives(P2);

julia> length(reps)
6

julia> all(xi -> is_weakly_index_increasing(P2, xi), reps)
true

On the non-compact total space of O(1) + O(-1) on P^1, every generic chamber turns out to be weakly index-increasing even though only 6 of them are strictly index-increasing:

julia> T = total_space(vector_bundle_O(1, [1, -1]));

julia> length(weakly_index_increasing_xi_representatives(T))
8
source
GKMtools.admits_index_increasing_xi — Function
admits_index_increasing_xi(G::AbstractGKMGraph) -> Tuple{Bool, FreeModuleElem}

Return (true, xi) if G admits some generic index-increasing direction xi in lattice(G), where xi is the first such direction found during chamber enumeration. Return (false, zero(lattice(G))) otherwise.

Example

julia> P2 = projective_space(GKMGraph, 2);

julia> ok, xi = admits_index_increasing_xi(P2);

julia> ok
true

julia> is_index_increasing(P2, xi)
true

julia> G = gkm_2d([1 0; 1 1; 0 1; -1 0; -1 -1; 0 -1]);

julia> admits_index_increasing_xi(G)
(false, (0, 0))

The same is true on the non-compact total space of O(1) + O(-1) on P^1:

julia> T = total_space(vector_bundle_O(1, [1, -1]));

julia> ok, xi = admits_index_increasing_xi(T);

julia> ok
true

julia> is_index_increasing(T, xi)
true
source
GKMtools.admits_weakly_index_increasing_xi — Function
admits_weakly_index_increasing_xi(G::AbstractGKMGraph) -> Tuple{Bool, FreeModuleElem}

Return (true, xi) if G admits some generic weakly-index-increasing direction xi in lattice(G), where xi is the first such direction found during chamber enumeration. Return (false, zero(lattice(G))) otherwise.

Example

julia> P2 = projective_space(GKMGraph, 2);

julia> ok, xi = admits_weakly_index_increasing_xi(P2);

julia> ok
true

julia> is_weakly_index_increasing(P2, xi)
true

julia> G = gkm_2d([1 0; 1 1; 0 1; -1 0; -1 -1; 0 -1]);

julia> ok, xi = admits_weakly_index_increasing_xi(G);

julia> ok && is_weakly_index_increasing(G, xi)
true

Non-compact total space of O(1) + O(-1) on P^1:

julia> T = total_space(vector_bundle_O(1, [1, -1]));

julia> ok, xi = admits_weakly_index_increasing_xi(T);

julia> ok
true

julia> is_weakly_index_increasing(T, xi)
true
source