Algebra.

Combinatorics

Combinatorics Algebra

Geometry

Combinatorics Geometry

The closure of a linear space

in a product of lines
Federico Ardila M.
San Francisco State University, San Francisco, California.
Universidad de Los Andes, Bogot, Colombia.

SACNAS
Los Angeles, California
October 16, 2014

Algebra.

Combinatorics

Combinatorics Algebra

Geometry

Combinatorics Geometry

Joint work with:

Adam Boocher (UC Berkeley U Edinburgh)

The closure of a linear space in a product of lines.

Federico Castillo (UC Davis), Jos Samper (U Washington)

The topology of the external activity complex of a matroid.
Algebra: much more than algebra.
To understand an ideal (algebra) and a variety (geometry), we need
matroids (combinatorics) and simplicial complexes (topology).
(I have nothing arithmetic to say.)

Algebra.

Combinatorics

Combinatorics Algebra

Geometry

Combinatorics Geometry

Joint work with:

Adam Boocher (UC Berkeley U Edinburgh)

The closure of a linear space in a product of lines.

Federico Castillo (UC Davis), Jos Samper (U Washington)

The topology of the external activity complex of a matroid.
Algebra: much more than algebra.
To understand an ideal (algebra) and a variety (geometry), we need
matroids (combinatorics) and simplicial complexes (topology).
(I have nothing arithmetic to say.)

Algebra.

Combinatorics

Combinatorics Algebra

Geometry

Combinatorics Geometry

ALGEBRA
HOMOGENIZATION. To homogenize a polynomial
f (x1 , . . . , xn )

ef (x1 , . . . , xn , y1 , . . . , yn )

we substitute f ( xy11 , . . . , xynn ) and clear denominators. Example:

f = x2

x12

x2

y2

x1
y1

2
7

ef = x2 y 2 x1 2 y2 .
1

The homogenization of an ideal I is eI = {ef : f I}.

Consider a linear ideal, such as
I

= hx1 + x2 + x6 , x2 x3 + x5 , x3 + x4 i
= {(x1 + x2 + x6 )f + (x2 x3 + x5 )g + (x3 + x4 )h :
f , g, h polynomials in x1 , . . . , x6 }

OUR GOAL. Study the homogenization of a linear ideal.

Algebra.

Combinatorics

Combinatorics Algebra

Geometry

Combinatorics Geometry

ALGEBRA
HOMOGENIZATION. To homogenize a polynomial
f (x1 , . . . , xn )

ef (x1 , . . . , xn , y1 , . . . , yn )

we substitute f ( xy11 , . . . , xynn ) and clear denominators. Example:

f = x2

x12

x2

y2

x1
y1

2
7

ef = x2 y 2 x1 2 y2 .
1

The homogenization of an ideal I is eI = {ef : f I}.

Consider a linear ideal, such as
I

= hx1 + x2 + x6 , x2 x3 + x5 , x3 + x4 i
= {(x1 + x2 + x6 )f + (x2 x3 + x5 )g + (x3 + x4 )h :
f , g, h polynomials in x1 , . . . , x6 }

OUR GOAL. Study the homogenization of a linear ideal.

Algebra.

Combinatorics

Combinatorics Algebra

Geometry

Combinatorics Geometry

Suppose I is a linear ideal, such as

I = hx1 + x2 + x6 , x2 x3 + x5 , x3 + x4 i .
Then we homogenize every polynomial in I:
eI = hx1 y2 y6 + y1 x2 y6 + y1 y2 x6 , x2 y3 y5 y2 x3 y5 + y2 y3 x5 , . . .i .
BUT we cant just homogenize the generators of I:
eI 6= hx1 y2 y6 + y1 x2 y6 + y1 y2 x6 , x2 y3 y5 y2 x3 y5 + y2 y3 x5 , x3 y4 + y3 x4 i
Ex: f = x2 + x4 + x5 I, but ef = x2 y4 y5 + y2 x4 y5 + y2 y4 x5 is missing.

It is generally difficult to write down explicit equations for ideals.

Motivating Question.
What is the best generating set for eI?

Algebra.

Combinatorics

Combinatorics Algebra

Geometry

Combinatorics Geometry

Suppose I is a linear ideal, such as

I = hx1 + x2 + x6 , x2 x3 + x5 , x3 + x4 i .
Then we homogenize every polynomial in I:
eI = hx1 y2 y6 + y1 x2 y6 + y1 y2 x6 , x2 y3 y5 y2 x3 y5 + y2 y3 x5 , . . .i .
BUT we cant just homogenize the generators of I:
eI 6= hx1 y2 y6 + y1 x2 y6 + y1 y2 x6 , x2 y3 y5 y2 x3 y5 + y2 y3 x5 , x3 y4 + y3 x4 i
Ex: f = x2 + x4 + x5 I, but ef = x2 y4 y5 + y2 x4 y5 + y2 y4 x5 is missing.

It is generally difficult to write down explicit equations for ideals.

Motivating Question.
What is the best generating set for eI?

Algebra.

Combinatorics Algebra

Combinatorics

Combinatorics Geometry

Geometry

COMBINATORICS
MATROIDS. Suppose
I = hx1 + x2 + x6 , x2 x3 + x5 , x3 + x4 i C[x1 , . . . , x6 ]
Use the generators of I

1 1 0
A = 0 1 1
0 0 1

as rows of a matrix:

0 0 1
0 1 0 .
1 0 0

Treat the 6 columns as points.

2
5

16

2
5

16

16
3

16
4

2
3

16
16

2
3

The matroid M is the list of flats:

points 16,2,3,4,5
lines 1256,136,146,23,24,345
planes 123456
Key. Use the matroid (combinatorics) to study the ideal (algebra).

Algebra.

Combinatorics

Combinatorics Algebra

Combinatorics Geometry

Geometry

COMBINATORICS ALGEBRA
GENERATORS I. First lets discuss generators of linear ideals.
TAKE 1. We have

I = hx1 + x2 + x6 , x2 x3 + x5 , x3 + x4 i

First observation:
support of generators of I: 126, 235, 34
complements: 345, 146, 1256 are lines (hyperplanes!)

16
4

Why not include all hyperplanes? Each gives a unique equation in I.

TAKE 2. Use the cocircuits:
I

= hx1 + x2 + x6 , x1 + x3 x5 + x6 , x1 x4 x5 + x6 ,
x2 x3 + x5 , x2 + x4 + x5 , x3 + x4 i

support of generators of I: 126, 1356, 1456, 235, 245, 34

complements: 345, 24, 23, 146, 136, 1256 (all hyperplanes)

Algebra.

Combinatorics

Combinatorics Algebra

Combinatorics Geometry

Geometry

COMBINATORICS ALGEBRA
GENERATORS I. First lets discuss generators of linear ideals.
TAKE 1. We have

I = hx1 + x2 + x6 , x2 x3 + x5 , x3 + x4 i

First observation:
support of generators of I: 126, 235, 34
complements: 345, 146, 1256 are lines (hyperplanes!)

16
4

Why not include all hyperplanes? Each gives a unique equation in I.

TAKE 2. Use the cocircuits:
I

= hx1 + x2 + x6 , x1 + x3 x5 + x6 , x1 x4 x5 + x6 ,
x2 x3 + x5 , x2 + x4 + x5 , x3 + x4 i

support of generators of I: 126, 1356, 1456, 235, 245, 34

complements: 345, 24, 23, 146, 136, 1256 (all hyperplanes)

Algebra.

Combinatorics

Combinatorics Algebra

Combinatorics Geometry

Geometry

GENERATORS I. First lets discuss generators of a linear ideal I

TAKE 1. We have
I = hx1 + x2 + x6 , x2 x3 + x5 , x3 + x4 i

GOOD: This is a minimal set of generators.

16
4

TAKE 2. We have
I

= hx1 + x2 + x6 , x1 + x3 x5 + x6 , x1 x4 x5 + x6 ,
x2 x3 + x5 , x2 + x4 + x5 , x3 + x4 i

support of generators of I: 126, 1356, 1456, 235, 245, 34 (cocircuits)

complements: 345, 24, 23, 146, 136, 1256 (hyperplanes)
BAD: This is not a minimal set of generators.
GOOD: It is the universal Grbner basis. (Great for computations.)
(Sturmfels 96)

Algebra.

Combinatorics

Combinatorics Algebra

Combinatorics Geometry

Geometry

GENERATORS II. What are the generators of the homogenization eI?

TAKE 1. We have
eI 6= hx1 y2 y6 + y1 x2 y6 + y1 y2 x6 ,

x2 y3 y5 y2 x3 y5 + y2 y3 x5 ,

ALL BAD: This is not even a set of generators.

x3 y4 + y3 x4 i

2
5

16
4

TAKE 2. What if we try homogenizing all the cocircuits?

eI =

hx1 y2 y6 + y1 x2 y6 + y1 y2 x6 ,

x1 y3 y5 y6 + y1 x3 y5 y6 y1 y3 x5 y6 + y1 y3 y5 x6 ,

x1 y4 y5 y6 y1 x4 y5 y6 y1 y4 x5 y6 + y1 y4 y5 x6 ,
x2 y4 y5 + y2 x4 y5 + y2 y4 x5 ,

x2 y3 y5 y2 x3 y5 + y2 y3 x5 ,

x3 y4 + y3 x4 i .

ALL GOOD: The best!

Theorem. (A. Boocher 13) The homogenized cocircuits of I
are a minimal set of generators for eI.
are the universal Grbner basis for eI.
These 2 properties rarely occur together. When they do, eI is robust.

Algebra.

Combinatorics

Combinatorics Algebra

Combinatorics Geometry

Geometry

GENERATORS II. What are the generators of the homogenization eI?

TAKE 1. We have
eI 6= hx1 y2 y6 + y1 x2 y6 + y1 y2 x6 ,

x2 y3 y5 y2 x3 y5 + y2 y3 x5 ,

ALL BAD: This is not even a set of generators.

x3 y4 + y3 x4 i

2
5

16
4

TAKE 2. What if we try homogenizing all the cocircuits?

eI =

hx1 y2 y6 + y1 x2 y6 + y1 y2 x6 ,

x1 y3 y5 y6 + y1 x3 y5 y6 y1 y3 x5 y6 + y1 y3 y5 x6 ,

x1 y4 y5 y6 y1 x4 y5 y6 y1 y4 x5 y6 + y1 y4 y5 x6 ,
x2 y4 y5 + y2 x4 y5 + y2 y4 x5 ,

x2 y3 y5 y2 x3 y5 + y2 y3 x5 ,

x3 y4 + y3 x4 i .

ALL GOOD: The best!

Theorem. (A. Boocher 13) The homogenized cocircuits of I
are a minimal set of generators for eI.
are the universal Grbner basis for eI.
These 2 properties rarely occur together. When they do, eI is robust.

Algebra.

Combinatorics

Combinatorics Algebra

Geometry

Combinatorics Geometry

SYZYGIES. We know the generators (complements hyperplanes).

eI =

hx1 y2 y6 + y1 x2 y6 + y1 y2 x6 ,

x1 y3 y5 y6 + y1 x3 y5 y6 y1 y3 x5 y6 + y1 y3 y5 x6 ,

x1 y4 y5 y6 y1 x4 y5 y6 y1 y4 x5 y6 + y1 y4 y5 x6 ,
x2 y4 y5 + y2 x4 y5 + y2 y4 x5 ,

2
5

x2 y3 y5 y2 x3 y5 + y2 y3 x5 ,

x3 y4 + y3 x4 i .

16
4

What are the linear relations (syzygies) among them? Here are two:
y2 y5 (x3 y4 + y3 x4 ) + y4 (x2 y3 y5 y2 x3 y5 + y2 y3 x5 ) y3 (x2 y4 y5 + y2 x4 y5 + y2 y4 x5 ) = 0
(x2 y5 +y2 x5 )(x3 y4 +y3 x4 )x4 (x2 y3 y5 y2 x3 y5 +y2 y3 x5 )x3 (x2 y4 y5 +y2 x4 y5 +y2 y4 x5 ) = 0

But there are seven more! How to describe them? Organize them?
Hint: these two have support 2345 whose complement is 16, a point.

Algebra.

Combinatorics

Combinatorics Algebra

Combinatorics Geometry

Geometry

SYZYGIES. We have:
6 generators. (complements: 1256, 136, 146, 23, 24, 345)
eI

hx1 y2 y6 + y1 x2 y6 + y1 y2 x6 ,

x1 y3 y5 y6 + y1 x3 y5 y6 y1 y3 x5 y6 + y1 y3 y5 x6 ,
x2 y3 y5 y2 x3 y5 + y2 y3 x5 ,

x1 y4 y5 y6 y1 x4 y5 y6 y1 y4 x5 y6 + y1 y4 y5 x6 ,
x2 y4 y5 + y2 x4 y5 + y2 y4 x5 ,

x3 y4 + y3 x4 i .

(2 of) 9 relations (syzygies). (compls: 16, 16, 2, 2, 3, 3, 4, 4, 5)

y2 y5 (x3 y4 + y3 x4 ) + y4 (x2 y3 y5 y2 x3 y5 + y2 y3 x5 ) y3 (x2 y4 y5 + y2 x4 y5 + y2 y4 x5 ) = 0
(x2 y5 + y2 x5 )(x3 y4 + y3 x4 ) x4 (x2 y3 y5 y2 x3 y5 + y2 y3 x5 ) x3 (x2 y4 y5 + y2 x4 y5 + y2 y4 x5 ) = 0.
1
5

4 rels. among the rels. (2nd syzygies).

(complement: )
-1

16

- 1 16

16

16
4

- 1 16
3

-1

Tuesday, November 26, 13

-4

-1

2
3

The picture on the right (lattice of flats)

organizes the syzygies beautifully.

-1

1
4

Algebra.

Combinatorics Algebra

Combinatorics

Geometry

Combinatorics Geometry

Theorem. (A. Boocher 13)

The syzygies of eI are supported on [n] F where F is a flat (spanned
subspace) of M. There are | (F )| ith syzygies on [n] F , where
(F ) =

([n]) = 1,

(G)

GF

is the dual Mbius function of M.

1
5

-1

16

- 1 16

16

16
4

- 1 16
3

-1

-4

-1

2
3

Tuesday, November 26, 13

-1

1
4

Algebra.

Combinatorics Algebra

Combinatorics

Geometry

Combinatorics Geometry

The same theorem, in fancier language:

Theorem. (A. Boocher 13) The Betti numbers of eI are the coefficients
of the cocharacteristic polynomial of the matroid of I.
1
5

-1

16

- 1 16

16

16
4

- 1 16
3

-1

-1

-1

2
3

1
4

-4

Tuesday, November 26, 13

The free resolution of eI as an S-module is:

S 4 S 9 S 6 S S/eI 0.
Here S = C[x1 , . . . , xn , y1 , . . . , yn ].

Algebra.

Combinatorics Algebra

Combinatorics

Geometry

Combinatorics Geometry

GEOMETRY
Algebra

Geometry

Polynomial equations

Set of solutions

Ideal I C[x1 , . . . , xn ]

Variety in Cn

Linear Ideal I C[x1 , . . . , xn ]

Subspace of Cn

Embed the line C , P1 into the projective line P1 = C . Then

Cn , (P1 )n

gives

e
L , L

e is the closure of L in (P1 )n .

where L
Example: x2 = 10x1

L=
Our algebraic goal.
Study the homogenized ideal eI.

e ="
L
Our geometric goal.
e
Study the closure variety L.

Algebra.

Combinatorics Algebra

Combinatorics

Geometry

Combinatorics Geometry

GEOMETRY
Algebra

Geometry

Polynomial equations

Set of solutions

Ideal I C[x1 , . . . , xn ]

Variety in Cn

Linear Ideal I C[x1 , . . . , xn ]

Subspace of Cn

Embed the line C , P1 into the projective line P1 = C . Then

Cn , (P1 )n

gives

e
L , L

e is the closure of L in (P1 )n .

where L
Example: x2 = 10x1

L=
Our algebraic goal.
Study the homogenized ideal eI.

e ="
L
Our geometric goal.
e
Study the closure variety L.

Algebra.

Combinatorics

Combinatorics Algebra

Combinatorics Geometry

Geometry

DEGREE. (To find the degree of a d-dim variety V in Cn , choose a

generic (n d)-plane P, count intersection points: deg V = |V P|)


Ex. I = x 2 + y 2 z 2 , V = ice cream cone deg V = 2.

Ex. I = linear ideal V = linear space deg V = 1.

e
Question. What is the degree of our closure variety L?
eI = hx1 y2 y6 + y1 x2 y6 + y1 y2 x6 , x2 y3 y5 y2 x3 y5 + y2 y3 x5 , . . .i .
e
Theorem. (A. Boocher 13) The degree of L
equals the number of bases of the matroid M.

2
5

16
4

In our example we have 13 bases (triples of points not on a line)

123, 124, 134, 135, 145, 234, 235, 236, 245, 246, 346, 356, 456.
e = 13.
So deg L

Algebra.

Combinatorics

Combinatorics Algebra

Combinatorics Geometry

Geometry

DEGREE. (To find the degree of a d-dim variety V in Cn , choose a

generic (n d)-plane P, count intersection points: deg V = |V P|)


Ex. I = x 2 + y 2 z 2 , V = ice cream cone deg V = 2.

Ex. I = linear ideal V = linear space deg V = 1.

e
Question. What is the degree of our closure variety L?
eI = hx1 y2 y6 + y1 x2 y6 + y1 y2 x6 , x2 y3 y5 y2 x3 y5 + y2 y3 x5 , . . .i .
e
Theorem. (A. Boocher 13) The degree of L
equals the number of bases of the matroid M.

2
5

16
4

In our example we have 13 bases (triples of points not on a line)

123, 124, 134, 135, 145, 234, 235, 236, 245, 246, 346, 356, 456.
e = 13.
So deg L

Algebra.

Combinatorics Algebra

Combinatorics

Geometry

Combinatorics Geometry

muchas gracias
1
5

-1

16

- 1 16

16

16
4

- 1 16
3

-1

-1

-1

2
3

1
4

-4

Tuesday, November 26, 13

The articles are available at:

http://math.sfsu.edu/federico
http://arxiv.org/