Construction of the Lindstr om valuation of an algebraic matroid - - PowerPoint PPT Presentation

construction of the lindstr om valuation of an algebraic
SMART_READER_LITE
LIVE PREVIEW

Construction of the Lindstr om valuation of an algebraic matroid - - PowerPoint PPT Presentation

Construction of the Lindstr om valuation of an algebraic matroid Dustin Cartwright University of Tennessee, Knoxville August 3, 2017 Dustin Cartwright Construction of Lindstr om valuation Algebraic matroids Given K L = K ( x 1 , . .


slide-1
SLIDE 1

Construction of the Lindstr¨

  • m valuation of an

algebraic matroid

Dustin Cartwright

University of Tennessee, Knoxville

August 3, 2017

Dustin Cartwright Construction of Lindstr¨

  • m valuation
slide-2
SLIDE 2

Algebraic matroids

Given K ⊂ L = K(x1, . . . , xn) a field extension, algebraic matroid of K ⊂ L is:

◮ Independent sets are sets I such that {xi | i ∈ I} are

algebraically independent

◮ Bases are the maximal independent sets (called transcendence

bases in field theory)

◮ Circuits are the minimal dependent sets ◮ Rank function rk(S) is the transcendence degree of the

extension K ⊂ K(xi | i ∈ S)

◮ Geometrically: If L = Frac(K[x1, . . . , xn]/I), then rk(S) is the

dimension of the projection defined by K[xi | i ∈ S] ∩ I Algebraic matroids are hard!

Dustin Cartwright Construction of Lindstr¨

  • m valuation
slide-3
SLIDE 3

Differentials: Linearizing algebraic relations

◮ If K ⊂ L = K(x1, . . . , xn) is a field extension, the vector space

  • f differentials ΩL/K is a L-vector space whose dimension is the

transcendence degree of K ⊂ L, generated by elements dx1, . . . , dxn

◮ Geometrically: If L = Frac(K[x1, . . . , xn]/I), then ΩL/K is the

dual to the tangent space of V (I) at the generic point

◮ If K has characteristic 0, the algebraic matroid of K ⊂ L is the

same as the linear matroid of the differentials ΩL/K with vectors dx1, . . . , dxn

◮ On the other hand, if K has characteristic p, and

L = Frac K[x1, x2]/xp

1 − x2, then:

◮ dx1 is non-zero, dx2 = 0, so only basis of linear matroid of

ΩL/K is {1}

◮ Bases of the algebraic matroid of K ⊂ L are {1} and {2}

Frobenius function x → xp is weird in characteristic p

Dustin Cartwright Construction of Lindstr¨

  • m valuation
slide-4
SLIDE 4

Tropicalization of a vector space

◮ Set-up: k field with valuation val: k× ։ Γ ⊂ R, V a k-vector

space, x1, . . . , xn ∈ V

◮ Given: w = (w1, . . . , wn) ∈ Γn ◮ Scale: tw1 1 x1, . . . , twn n xn, where twi ∈ k, val(twi) = wi ◮ Generate: R-submodule of V generated by tw1x1, . . . , twnxn,

where R is the valuation ring of k

◮ Reduce: tensor with R/mR to get inw(V ), where m is the

maximal ideal of R

◮ Tropicalization: Trop(V ) ∩ Γn is the set of w ∈ Γn such that

reductions tw1x1, . . . , twnxn ∈ inw(V ) are all non-zero

◮ The tropicalization is equivalent to the valuated matroid of V

Dustin Cartwright Construction of Lindstr¨

  • m valuation
slide-5
SLIDE 5

Rough idea

“Tropicalization” for fields:

◮ Scaling =

⇒ Frobenius

◮ Reduction =

⇒ Differentials

Dustin Cartwright Construction of Lindstr¨

  • m valuation
slide-6
SLIDE 6

“Tropicalization” of field extensions

◮ Set-up: K field of char. p, L = K(x1, . . . , xn) ◮ Given: w = (w1, . . . , wn) ∈ Zn ◮ Scale: F −w1x1, . . . , F −wnxn, where F is Frobenius: Fx = xp, in

˜ L =

l K(x1/pl 1

, . . . , x1/pl

n

)

◮ Generate: K(F −wx) := K(F −w1x1, . . . , F −wnxn) ◮ Reduce: Vector space of differentials ΩK(F −wx)/K generated by

differentials dF −w1x1, . . . , dF −wnxn

◮ Tropicalization: Trop(L/K) ∩ Zn is the set of w ∈ Zn such

that differentials dF −w1x1, . . . , dF −wnxn are all non-zero

◮ Trop(L/K) is the tropicalization of a unique valuated matroid,

called the Lindstr¨

  • m valuated matroid of K ⊂ L

(Bollen-Draisma-Pendavingh)

Dustin Cartwright Construction of Lindstr¨

  • m valuation
slide-7
SLIDE 7

Local and global structure

If V is a k-vector space, x1, . . . , xn ∈ V , then Trop(V ) is

◮ Globally: The recession fan of Trop(V ) is equivalent to the

linear (non-valuated) matroid of x1, . . . , xn ∈ V

◮ Locally: At w ∈ Γn, the link of Trop(V ) is equivalent to the

linear matroid of inw(V ) If K ⊂ L = K(x1, . . . , xn) is a field extension, then Trop(L/K) is

◮ Globally: The recession fan of Trop(L/K) is equivalent to the

algebraic matroid of K ⊂ L

◮ Locally: The link of w ∈ Zn is equivalent to the linear matroid

  • f ΩK(F −wx)/K

Dustin Cartwright Construction of Lindstr¨

  • m valuation
slide-8
SLIDE 8

A bridge example: monomials

◮ A: a d × n integer matrix ◮ L = K(z1, . . . , zd), xi = zA1i 1

· · · zAdi

d ◮ The Lindstr¨

  • m valuated matroid of K ⊂ L is the same as the

valuated matroid of the columns of A in Qd with the p-adic valuation

◮ Prior example of K[x1, x2]/xp 1 − x2 is a monomial example

with A =

  • 1

p

  • Dustin Cartwright

Construction of Lindstr¨

  • m valuation
slide-9
SLIDE 9

Circuits of the Lindstr¨

  • m matroid

As before: K ⊂ L = K(x1, . . . , xn)

◮ If C is a circuit of the algebraic matroid of K ⊂ L, then

{xi | i ∈ C} is a minimal dependent set, and there exists a polynomial relation fC ∈ K[xi | i ∈ C], unique up to scaling

◮ Write

fC =

  • u∈JC

cuxu1

1 · · · xun n

where JC ⊂ Zn

≥0, ui = 0 if i /

∈ C, and cu = 0

◮ Define:

C(fC) =

  • . . . , min{valp ui | u ∈ JC}, . . .
  • ∈ (Z ∪ ∞)n

where valp is the p-adic valuation

◮ The valuated circuits of the Lindstr¨

  • m valuation are the vectors

C(fC) + λ1 as C ranges over circuits of the algebraic matroid, and λ ∈ Z

Dustin Cartwright Construction of Lindstr¨

  • m valuation
slide-10
SLIDE 10

Valuation of the Lindstr¨

  • m matroid

As before: K ⊂ L = K(x1, . . . , xn)

◮ Let B be a basis of the algebraic matroid, meaning a maximal

independent set of variables

◮ The extension K(xi | i ∈ B) ⊂ L can be uniquely factored as

K(xi | i ∈ B) ⊂ K(xi | i ∈ B)sep ⊂ L where the first extension is separable (roughly: like characteristic 0) and the second is purely inseparable (defined by taking pth roots)

◮ The Lindstr¨

  • m valuation is:

v(B) = logp[L : K(xi | i ∈ B)sep] ∈ Z≥0

Dustin Cartwright Construction of Lindstr¨

  • m valuation
slide-11
SLIDE 11

Cocircuits of the Lindstr¨

  • m valuation

As before: K ⊂ L = K(x1, . . . , xn)

◮ Let H be a hyperplane of the algebraic matroid, meaning a

maximal set with rk(H) = rk({1, . . . , n}) − 1

◮ Define:

Cco(H) =

  • . . . , logp[L : K(xi | i ∈ H ∪ {i})sep], . . .
  • ◮ The valuated cocircuits of the Lindstr¨
  • m valuation are:

Cco(H) + λ1 as H ranges over the hyperplanes, and λ ∈ Z

Dustin Cartwright Construction of Lindstr¨

  • m valuation