Holonomy and combinatorics - Simon Fraser Universitypeople.math.sfu.ca/~mmishna/PUB/holtalk.pdf ·...

Post on 23-Sep-2018

222 views 0 download

Transcript of Holonomy and combinatorics - Simon Fraser Universitypeople.math.sfu.ca/~mmishna/PUB/holtalk.pdf ·...

Marni Mishna, Freaky Friday the 13th, 2004 Carleton University Holonomy and Combinatorics – p. 1/26

Holonomy and combinatoricsFrom automatic identities to systematic enumeration

Marni MishnaLaBRI, Université Bordeaux I

http://www.labri.fr/~mishna

The Set Up

● A grab bag of problems

● Basic Idea

● Summary of how it works

Creative telescoping

A combinatorial framework

The scalar product

Conclusion

Marni Mishna, Freaky Friday the 13th, 2004 Carleton University Holonomy and Combinatorics – p. 2/26

A grab bag of problems

Prove Dixon’s identity:∑

k(−1)k(n+an+k

)(n+bb+k

)(a+ba+k

)= (n+a+b)!

n!a!b!

Calculate the dimensionof one representation ofthe symmetric group inanother

ζ(2, 2) = 1/2(ζ(2)2 − ζ(4)

)

Find a recursive formulafor the number of 4-

regular graphs of size n

Find a formula for standard

Young tableaux of size n

5 7

4 6 8

1 2 3

Enumerate plane parti-tions of shape 2× n

633321

522111

Prove q-Hypergeometric identities

The Set Up

● A grab bag of problems

● Basic Idea

● Summary of how it works

Creative telescoping

A combinatorial framework

The scalar product

Conclusion

Marni Mishna, Freaky Friday the 13th, 2004 Carleton University Holonomy and Combinatorics – p. 3/26

Basic Idea

Set of easy identities

A set of recurrences forp(n, k)

A recurrence forφ(p(n, k))

Systems of differentialequations satisfied by f

and g

A differential equationfor ψ(f, g)

The recurrences and DEs are useful! They give sequences,can be solved, allow asymptotic analysis, ...

The Set Up

● A grab bag of problems

● Basic Idea

● Summary of how it works

Creative telescoping

A combinatorial framework

The scalar product

Conclusion

Marni Mishna, Freaky Friday the 13th, 2004 Carleton University Holonomy and Combinatorics – p. 3/26

Basic Idea

Set of easy identities Elimination inoperator algebra

A set of recurrences forp(n, k)

A recurrence forφ(p(n, k))

Systems of differentialequations satisfied by f

and g

A differential equationfor ψ(f, g)

The recurrences and DEs are useful! They give sequences,can be solved, allow asymptotic analysis, ...

The Set Up

● A grab bag of problems

● Basic Idea

● Summary of how it works

Creative telescoping

A combinatorial framework

The scalar product

Conclusion

Marni Mishna, Freaky Friday the 13th, 2004 Carleton University Holonomy and Combinatorics – p. 3/26

Basic Idea

Set of easy identities Elimination inoperator algebra Profound identity

A set of recurrences forp(n, k)

A recurrence forφ(p(n, k))

Systems of differentialequations satisfied by f

and g

A differential equationfor ψ(f, g)

The recurrences and DEs are useful! They give sequences,can be solved, allow asymptotic analysis, ...

The Set Up

● A grab bag of problems

● Basic Idea

● Summary of how it works

Creative telescoping

A combinatorial framework

The scalar product

Conclusion

Marni Mishna, Freaky Friday the 13th, 2004 Carleton University Holonomy and Combinatorics – p. 3/26

Basic Idea

Set of easy identities Elimination inoperator algebra Profound identity

A set of recurrences forp(n, k)

A recurrence forφ(p(n, k))

Systems of differentialequations satisfied by f

and g

A differential equationfor ψ(f, g)

The recurrences and DEs are useful! They give sequences,can be solved, allow asymptotic analysis, ...

The Set Up

● A grab bag of problems

● Basic Idea

● Summary of how it works

Creative telescoping

A combinatorial framework

The scalar product

Conclusion

Marni Mishna, Freaky Friday the 13th, 2004 Carleton University Holonomy and Combinatorics – p. 3/26

Basic Idea

Set of easy identities Elimination inoperator algebra Profound identity

A set of recurrences forp(n, k)

A recurrence forφ(p(n, k))

Systems of differentialequations satisfied by f

and g

A differential equationfor ψ(f, g)

The recurrences and DEs are useful! They give sequences,can be solved, allow asymptotic analysis, ...

Marni Mishna, Freaky Friday the 13th, 2004 Carleton University Holonomy and Combinatorics – p. 4/26

Summary of how it works

Dixon’s Identity 4-Regular Graphs

k(−1)k

n+a

n+k

� �

n+b

b+k

� �

a+b

a+k

=(n+a+b)!

n!a!b!

1

2

3 4

5

6

7

Operator Algebra Shift Differential

Closure Operation +, * 〈, 〉

“Easy” Identity recurrence for n! DE for ex

Elimination Answer is in a left ideal. Find a Gröbner basis for this ideal

Output Recurrences for hypergeo-metric sums

Diff eq. satisfied by gen-erating function

The Set Up

Creative telescoping

● P-Recursive Functions

● Express recurrences as operators

● Elimination

● Zeilberger’s Creative Telescoping

● How does this work in general?

A combinatorial framework

The scalar product

Conclusion

Marni Mishna, Freaky Friday the 13th, 2004 Carleton University Holonomy and Combinatorics – p. 5/26

Creative telescoping

The Set Up

Creative telescoping

● P-Recursive Functions

● Express recurrences as operators

● Elimination

● Zeilberger’s Creative Telescoping

● How does this work in general?

A combinatorial framework

The scalar product

Conclusion

Marni Mishna, Freaky Friday the 13th, 2004 Carleton University Holonomy and Combinatorics – p. 6/26

P-Recursive Functions

Def: A function f : N → R is P-recursive if it satisfies a finitelinear recurrence with polynomial coefficients.There exists polynomials γi such that:

γd(n)f(n+ d) + . . .+ γ1(n)f(n+ 1) + γ0(n)f(n) = 0

Ex: f(n) = n! is P-recursive: 1(n+ 1)!− (n+ 1)n! = 0Ex: Rational: p(x)/q(x) =

n f(n)xn =⇒ γi = const.

The Set Up

Creative telescoping

● P-Recursive Functions

● Express recurrences as operators

● Elimination

● Zeilberger’s Creative Telescoping

● How does this work in general?

A combinatorial framework

The scalar product

Conclusion

Marni Mishna, Freaky Friday the 13th, 2004 Carleton University Holonomy and Combinatorics – p. 7/26

Express recurrences as operators

Shift operator:

SnP (n, k) = P (n+ 1, k), SkP (n, k) = P (n, k + 1)

Define a (non-commutative) algebra:

Bn,k := C〈n, k, Sn, Sk〉

Snn = (n+ 1)Sn, Skk = (k + 1)Sk, all other variables commute

0 = P (n, k) + P (n, k + 1) + (3n2 + k)P (n+ 2, k + 1)

= P (n, k) + SkP (n, k) + (3n2 + k)S2nSkP (n, k)

= (1 + Sk + (3n2 + k)S2nSk) · P (n, k)

= ψ(n, k, Sn, Sk) · P

■ If ψ, ξ ∈ IP , then A(n, k, Sn, Sk)ψ +B(n, k, Sn, Sk)ξ ∈ IP .■ ψ is in the annihilating (left) ideal IP of P .

The Set Up

Creative telescoping

● P-Recursive Functions

● Express recurrences as operators

● Elimination

● Zeilberger’s Creative Telescoping

● How does this work in general?

A combinatorial framework

The scalar product

Conclusion

Marni Mishna, Freaky Friday the 13th, 2004 Carleton University Holonomy and Combinatorics – p. 7/26

Express recurrences as operators

Shift operator:

SnP (n, k) = P (n+ 1, k), SkP (n, k) = P (n, k + 1)

Define a (non-commutative) algebra:

Bn,k := C〈n, k, Sn, Sk〉

Snn = (n+ 1)Sn, Skk = (k + 1)Sk, all other variables commute

0 = P (n, k) + P (n, k + 1) + (3n2 + k)P (n+ 2, k + 1)

= P (n, k) + SkP (n, k) + (3n2 + k)S2nSkP (n, k)

= (1 + Sk + (3n2 + k)S2nSk) · P (n, k)

= ψ(n, k, Sn, Sk) · P

■ If ψ, ξ ∈ IP , then A(n, k, Sn, Sk)ψ +B(n, k, Sn, Sk)ξ ∈ IP .■ ψ is in the annihilating (left) ideal IP of P .

The Set Up

Creative telescoping

● P-Recursive Functions

● Express recurrences as operators

● Elimination

● Zeilberger’s Creative Telescoping

● How does this work in general?

A combinatorial framework

The scalar product

Conclusion

Marni Mishna, Freaky Friday the 13th, 2004 Carleton University Holonomy and Combinatorics – p. 7/26

Express recurrences as operators

Shift operator:

SnP (n, k) = P (n+ 1, k), SkP (n, k) = P (n, k + 1)

Define a (non-commutative) algebra:

Bn,k := C〈n, k, Sn, Sk〉

Snn = (n+ 1)Sn, Skk = (k + 1)Sk, all other variables commute

0 = P (n, k) + P (n, k + 1) + (3n2 + k)P (n+ 2, k + 1)

= P (n, k) + SkP (n, k) + (3n2 + k)S2nSkP (n, k)

= (1 + Sk + (3n2 + k)S2nSk) · P (n, k)

= ψ(n, k, Sn, Sk) · P

■ If ψ, ξ ∈ IP , then A(n, k, Sn, Sk)ψ +B(n, k, Sn, Sk)ξ ∈ IP .■ ψ is in the annihilating (left) ideal IP of P .

The Set Up

Creative telescoping

● P-Recursive Functions

● Express recurrences as operators

● Elimination

● Zeilberger’s Creative Telescoping

● How does this work in general?

A combinatorial framework

The scalar product

Conclusion

Marni Mishna, Freaky Friday the 13th, 2004 Carleton University Holonomy and Combinatorics – p. 8/26

Making new identities using elimination

Recall from your youth:n equations, m unknowns variables + linear algebra =⇒eliminate some unknown variables.

3x+ 4y + z = 2 (1)

5x− 2y + 2z = 3 (2)

5 ∗ (1)− 3 ∗ (2) =⇒

26y − z = 1

We can do this in the setting of non-commutative algebras.(Ore algebra)■ Left ideals will have Gröbner basis (Galligo)■ Modified traditional tools work (Buchbergers + Euclidean

algorithm)

The Set Up

Creative telescoping

● P-Recursive Functions

● Express recurrences as operators

● Elimination

● Zeilberger’s Creative Telescoping

● How does this work in general?

A combinatorial framework

The scalar product

Conclusion

Marni Mishna, Freaky Friday the 13th, 2004 Carleton University Holonomy and Combinatorics – p. 9/26

Zeilberger’s Creative Telescoping

F (n, k) =(nk

)

■ Easy identities: F (n+1,k)F (n,k) = n+1

n−k+1F (n,k+1)

F (n,k) = n−kk+1

■ Cross Multiply:

((n− k + 1)Sn − (n+ 1))F = 0 ♥

((k + 1)Sk − (n− k))F = 0 ♣

■ A linear combination: (Sk + 1)♥+ Sn♣ = 0(n+ 1) (SnSk − Sk − 1)F = 0

■ Reorganize this to have only Sn and n on the left, and afactor of (Sk − 1) on the right:

(n+ 1)(Sn − 2)F = (Sk − 1)G(n, k), for some G(n, k)

The Set Up

Creative telescoping

● P-Recursive Functions

● Express recurrences as operators

● Elimination

● Zeilberger’s Creative Telescoping

● How does this work in general?

A combinatorial framework

The scalar product

Conclusion

Marni Mishna, Freaky Friday the 13th, 2004 Carleton University Holonomy and Combinatorics – p. 9/26

Zeilberger’s Creative Telescoping

F (n, k) =(nk

)

■ Easy identities: F (n+1,k)F (n,k) = n+1

n−k+1F (n,k+1)

F (n,k) = n−kk+1

■ Cross Multiply:

((n− k + 1)Sn − (n+ 1))F = 0 ♥

((k + 1)Sk − (n− k))F = 0 ♣

■ A linear combination: (Sk + 1)♥+ Sn♣ = 0(n+ 1) (SnSk − Sk − 1)F = 0

■ Reorganize this to have only Sn and n on the left, and afactor of (Sk − 1) on the right:

(n+ 1)(Sn − 2)F = (Sk − 1)G(n, k), for some G(n, k)

The Set Up

Creative telescoping

● P-Recursive Functions

● Express recurrences as operators

● Elimination

● Zeilberger’s Creative Telescoping

● How does this work in general?

A combinatorial framework

The scalar product

Conclusion

Marni Mishna, Freaky Friday the 13th, 2004 Carleton University Holonomy and Combinatorics – p. 9/26

Zeilberger’s Creative Telescoping

F (n, k) =(nk

)

■ Easy identities: F (n+1,k)F (n,k) = n+1

n−k+1F (n,k+1)

F (n,k) = n−kk+1

■ Cross Multiply:

((n− k + 1)Sn − (n+ 1))F = 0 ♥

((k + 1)Sk − (n− k))F = 0 ♣

■ A linear combination: (Sk + 1)♥+ Sn♣ = 0(n+ 1) (SnSk − Sk − 1)F = 0

■ Reorganize this to have only Sn and n on the left, and afactor of (Sk − 1) on the right:

(n+ 1)(Sn − 2)F = (Sk − 1)G(n, k), for some G(n, k)

The Set Up

Creative telescoping

● P-Recursive Functions

● Express recurrences as operators

● Elimination

● Zeilberger’s Creative Telescoping

● How does this work in general?

A combinatorial framework

The scalar product

Conclusion

Marni Mishna, Freaky Friday the 13th, 2004 Carleton University Holonomy and Combinatorics – p. 9/26

Zeilberger’s Creative Telescoping

F (n, k) =(nk

)

■ Easy identities: F (n+1,k)F (n,k) = n+1

n−k+1F (n,k+1)

F (n,k) = n−kk+1

■ Cross Multiply:

((n− k + 1)Sn − (n+ 1))F = 0 ♥

((k + 1)Sk − (n− k))F = 0 ♣

■ A linear combination: (Sk + 1)♥+ Sn♣ = 0(n+ 1) (SnSk − Sk − 1)F = 0

■ Reorganize this to have only Sn and n on the left, and afactor of (Sk − 1) on the right:

(n+ 1)(Sn − 2)F = (Sk − 1)G(n, k), for some G(n, k)

The Set Up

Creative telescoping

● P-Recursive Functions

● Express recurrences as operators

● Elimination

● Zeilberger’s Creative Telescoping

● How does this work in general?

A combinatorial framework

The scalar product

Conclusion

Marni Mishna, Freaky Friday the 13th, 2004 Carleton University Holonomy and Combinatorics – p. 9/26

Zeilberger’s Creative Telescoping

F (n, k) =(nk

)

(n+ 1)(Sn − 2)F = (Sk − 1)G(n, k) (∗)

■ Take the sum:

(n+ 1)(Sn − 2)

a(n) =∑

k≥0

(n+ 1)(Sn − 2)

F (n, k)

=∑

k≥0

(Sk − 1)(G(n, k))

=∑

k≥0

(G(n, k + 1)−G(n, k)) = 0

■ Gives result about the sum:

a(n+ 1)− 2a(n) = 0 =⇒ a(n) = 2n =⇒∑

k

(nk

)= 2n

The Set Up

Creative telescoping

● P-Recursive Functions

● Express recurrences as operators

● Elimination

● Zeilberger’s Creative Telescoping

● How does this work in general?

A combinatorial framework

The scalar product

Conclusion

Marni Mishna, Freaky Friday the 13th, 2004 Carleton University Holonomy and Combinatorics – p. 9/26

Zeilberger’s Creative Telescoping

F (n, k) =(nk

)

(n+ 1)(Sn − 2)F = (Sk − 1)G(n, k) (∗)

■ Apply the linear combination to this sum:

(n+ 1)(Sn − 2)a(n) =∑

k≥0

(n+ 1)(Sn − 2)F (n, k)

=∑

k≥0

(Sk − 1)(G(n, k))

=∑

k≥0

(G(n, k + 1)−G(n, k)) = 0

■ Gives result about the sum:

a(n+ 1)− 2a(n) = 0 =⇒ a(n) = 2n =⇒∑

k

(nk

)= 2n

The Set Up

Creative telescoping

● P-Recursive Functions

● Express recurrences as operators

● Elimination

● Zeilberger’s Creative Telescoping

● How does this work in general?

A combinatorial framework

The scalar product

Conclusion

Marni Mishna, Freaky Friday the 13th, 2004 Carleton University Holonomy and Combinatorics – p. 9/26

Zeilberger’s Creative Telescoping

F (n, k) =(nk

)

(n+ 1)(Sn − 2)F = (Sk − 1)G(n, k) (∗)

■ Apply (∗):

(n+ 1)(Sn − 2)a(n) =∑

k≥0

(n+ 1)(Sn − 2)F (n, k)

=∑

k≥0

(Sk − 1)(G(n, k))

=∑

k≥0

(G(n, k + 1)−G(n, k)) = 0

■ Gives result about the sum:

a(n+ 1)− 2a(n) = 0 =⇒ a(n) = 2n =⇒∑

k

(nk

)= 2n

The Set Up

Creative telescoping

● P-Recursive Functions

● Express recurrences as operators

● Elimination

● Zeilberger’s Creative Telescoping

● How does this work in general?

A combinatorial framework

The scalar product

Conclusion

Marni Mishna, Freaky Friday the 13th, 2004 Carleton University Holonomy and Combinatorics – p. 9/26

Zeilberger’s Creative Telescoping

F (n, k) =(nk

)

(n+ 1)(Sn − 2)F = (Sk − 1)G(n, k) (∗)

■ Expand:

(n+ 1)(Sn − 2)a(n) =∑

k≥0

(n+ 1)(Sn − 2)F (n, k)

=∑

k≥0

(Sk − 1)(G(n, k))

=∑

k≥0

(G(n, k + 1)−G(n, k)) = 0

■ Gives result about the sum:

a(n+ 1)− 2a(n) = 0 =⇒ a(n) = 2n =⇒∑

k

(nk

)= 2n

The Set Up

Creative telescoping

● P-Recursive Functions

● Express recurrences as operators

● Elimination

● Zeilberger’s Creative Telescoping

● How does this work in general?

A combinatorial framework

The scalar product

Conclusion

Marni Mishna, Freaky Friday the 13th, 2004 Carleton University Holonomy and Combinatorics – p. 9/26

Zeilberger’s Creative Telescoping

F (n, k) =(nk

)

(n+ 1)(Sn − 2)F = (Sk − 1)G(n, k) (∗)

■ Expand:

(n+ 1)(Sn − 2)a(n) =∑

k≥0

(n+ 1)(Sn − 2)F (n, k)

=∑

k≥0

(Sk − 1)(G(n, k))

=∑

k≥0

(G(n, k + 1)−G(n, k)) = 0

■ Gives result about the sum:

a(n+ 1)− 2a(n) = 0 =⇒ a(n) = 2n =⇒∑

k

(nk

)= 2n

The Set Up

Creative telescoping

● P-Recursive Functions

● Express recurrences as operators

● Elimination

● Zeilberger’s Creative Telescoping

● How does this work in general?

A combinatorial framework

The scalar product

Conclusion

Marni Mishna, Freaky Friday the 13th, 2004 Carleton University Holonomy and Combinatorics – p. 10/26

How does this work in general?

■ Begin with a set of “independent” relationsP (n, k, Sn, Sk)F (n, k) = 0 Q(n, k, Sn, Sk)F (n, k) = 0■ Find the good linear combination which eliminates k and has

a factor of Sk − 1

R(n, k, Sn, Sk) = A(n, k, Sn, Sk)P +B(n, k, Sn, Sk)Q

R(n, k, Sn, Sk) = S(Sn, n) + (Sk − 1)R′(n, Sn, Sk)

■ Take the infinite sum over k. We have a telescoping sum:0 =

k R(n, k, Sn, Sk)F (n, k)

0 =∑

k S(Sn, n)F +∑

k(Sk − 1)R′(N,K, n)F︸ ︷︷ ︸

G(n,k)

■ ...And we get a recurrence for the sum

The Set Up

Creative telescoping

A combinatorial framework

● Generating functions

● D-finite functions

● D-finiteness

The scalar product

Conclusion

Marni Mishna, Freaky Friday the 13th, 2004 Carleton University Holonomy and Combinatorics – p. 11/26

A combinatorial framework

The Set Up

Creative telescoping

A combinatorial framework

● Generating functions

● D-finite functions

● D-finiteness

The scalar product

Conclusion

Marni Mishna, Freaky Friday the 13th, 2004 Carleton University Holonomy and Combinatorics – p. 12/26

Combinatorial generating functions

Counting information about combinatorial objects can beencoded in formal power series:

Ex. exponential generating function (egf) :

C(z) =∑

n

(number of objects in C of size n)zn

n!

Ex. C = C...

=⇒ C(z) = z + z2/2 + 2z3/3! + 3x4/4! + . . .

Nice: algebraic operations correspond with combinatorialoperations.

■ C = A ∪ B =⇒ C(z) = A(z) +B(z)

■ C = {(a, b), a ∈ A, b ∈ B} =⇒ C(z) = A(z)B(z)

The Set Up

Creative telescoping

A combinatorial framework

● Generating functions

● D-finite functions

● D-finiteness

The scalar product

Conclusion

Marni Mishna, Freaky Friday the 13th, 2004 Carleton University Holonomy and Combinatorics – p. 12/26

Combinatorial generating functions

Counting information about combinatorial objects can beencoded in formal power series:

Ex. exponential generating function (egf) :

C(z) =∑

n

(number of objects in C of size n)zn

n!

Ex. C = C...

=⇒ C(z) = z + z2/2 + 2z3/3! + 3x4/4! + . . .Nice: algebraic operations correspond with combinatorialoperations.

■ C = A ∪ B =⇒ C(z) = A(z) +B(z)

■ C = {(a, b), a ∈ A, b ∈ B} =⇒ C(z) = A(z)B(z)

The Set Up

Creative telescoping

A combinatorial framework

● Generating functions

● D-finite functions

● D-finiteness

The scalar product

Conclusion

Marni Mishna, Freaky Friday the 13th, 2004 Carleton University Holonomy and Combinatorics – p. 13/26

D-finite functions

Def. A function f(x1, x2, . . . , xn) is Differentiably finite (D-finite)with respect to X = x1, . . . , xn if■ For 1 ≤ j ≤ n, f satisfies n linear differential equations with

polynomial coefficients:

φ0(X)f(X) + φ1(X)∂f(X)∂xj

+ . . .+ φk(X)∂kf(X)

∂xkj

= 0

Examples■ f = sin(2x3 + 3yz) is D-finite with respect to {x, y, z}.xfxx − 2fx + 9x5f = 0, fyy − 9z2f = 0, fzz − 9y2f = 0

■ exp(polynomial) is D-finite■

∑f(n)zn D-finite ⇐⇒ f(n) P-recursive

The Set Up

Creative telescoping

A combinatorial framework

● Generating functions

● D-finite functions

● D-finiteness

The scalar product

Conclusion

Marni Mishna, Freaky Friday the 13th, 2004 Carleton University Holonomy and Combinatorics – p. 13/26

D-finite functions

Def. A function f(x1, x2, . . . , xn) is Differentiably finite (D-finite)with respect to X = x1, . . . , xn if■ For 1 ≤ j ≤ n, f satisfies n linear differential equations with

polynomial coefficients:

φ0(X)f(X) + φ1(X)∂f(X)∂xj

+ . . .+ φk(X)∂kf(X)

∂xkj

= 0

Examples■ f = sin(2x3 + 3yz) is D-finite with respect to {x, y, z}.

xfxx − 2fx + 9x5f = 0, fyy − 9z2f = 0, fzz − 9y2f = 0

■ exp(polynomial) is D-finite■

∑f(n)zn D-finite ⇐⇒ f(n) P-recursive

The Set Up

Creative telescoping

A combinatorial framework

● Generating functions

● D-finite functions

● D-finiteness

The scalar product

Conclusion

Marni Mishna, Freaky Friday the 13th, 2004 Carleton University Holonomy and Combinatorics – p. 13/26

D-finite functions

Def. A function f(x1, x2, . . . , xn) is Differentiably finite (D-finite)with respect to X = x1, . . . , xn if■ For 1 ≤ j ≤ n, f satisfies n linear differential equations with

polynomial coefficients:

φ0(X)f(X) + φ1(X)∂f(X)∂xj

+ . . .+ φk(X)∂kf(X)

∂xkj

= 0

Examples■ f = sin(2x3 + 3yz) is D-finite with respect to {x, y, z}.xfxx − 2fx + 9x5f = 0,

fyy − 9z2f = 0, fzz − 9y2f = 0

■ exp(polynomial) is D-finite■

∑f(n)zn D-finite ⇐⇒ f(n) P-recursive

The Set Up

Creative telescoping

A combinatorial framework

● Generating functions

● D-finite functions

● D-finiteness

The scalar product

Conclusion

Marni Mishna, Freaky Friday the 13th, 2004 Carleton University Holonomy and Combinatorics – p. 13/26

D-finite functions

Def. A function f(x1, x2, . . . , xn) is Differentiably finite (D-finite)with respect to X = x1, . . . , xn if■ For 1 ≤ j ≤ n, f satisfies n linear differential equations with

polynomial coefficients:

φ0(X)f(X) + φ1(X)∂f(X)∂xj

+ . . .+ φk(X)∂kf(X)

∂xkj

= 0

Examples■ f = sin(2x3 + 3yz) is D-finite with respect to {x, y, z}.xfxx − 2fx + 9x5f = 0, fyy − 9z2f = 0,

fzz − 9y2f = 0

■ exp(polynomial) is D-finite■

∑f(n)zn D-finite ⇐⇒ f(n) P-recursive

The Set Up

Creative telescoping

A combinatorial framework

● Generating functions

● D-finite functions

● D-finiteness

The scalar product

Conclusion

Marni Mishna, Freaky Friday the 13th, 2004 Carleton University Holonomy and Combinatorics – p. 13/26

D-finite functions

Def. A function f(x1, x2, . . . , xn) is Differentiably finite (D-finite)with respect to X = x1, . . . , xn if■ For 1 ≤ j ≤ n, f satisfies n linear differential equations with

polynomial coefficients:

φ0(X)f(X) + φ1(X)∂f(X)∂xj

+ . . .+ φk(X)∂kf(X)

∂xkj

= 0

Examples■ f = sin(2x3 + 3yz) is D-finite with respect to {x, y, z}.xfxx − 2fx + 9x5f = 0, fyy − 9z2f = 0, fzz − 9y2f = 0

■ exp(polynomial) is D-finite■

∑f(n)zn D-finite ⇐⇒ f(n) P-recursive

The Set Up

Creative telescoping

A combinatorial framework

● Generating functions

● D-finite functions

● D-finiteness

The scalar product

Conclusion

Marni Mishna, Freaky Friday the 13th, 2004 Carleton University Holonomy and Combinatorics – p. 13/26

D-finite functions

Def. A function f(x1, x2, . . . , xn) is Differentiably finite (D-finite)with respect to X = x1, . . . , xn if■ For 1 ≤ j ≤ n, f satisfies n linear differential equations with

polynomial coefficients:

φ0(X)f(X) + φ1(X)∂f(X)∂xj

+ . . .+ φk(X)∂kf(X)

∂xkj

= 0

Examples■ f = sin(2x3 + 3yz) is D-finite with respect to {x, y, z}.xfxx − 2fx + 9x5f = 0, fyy − 9z2f = 0, fzz − 9y2f = 0

■ exp(polynomial) is D-finite

■∑f(n)zn D-finite ⇐⇒ f(n) P-recursive

The Set Up

Creative telescoping

A combinatorial framework

● Generating functions

● D-finite functions

● D-finiteness

The scalar product

Conclusion

Marni Mishna, Freaky Friday the 13th, 2004 Carleton University Holonomy and Combinatorics – p. 13/26

D-finite functions

Def. A function f(x1, x2, . . . , xn) is Differentiably finite (D-finite)with respect to X = x1, . . . , xn if■ For 1 ≤ j ≤ n, f satisfies n linear differential equations with

polynomial coefficients:

φ0(X)f(X) + φ1(X)∂f(X)∂xj

+ . . .+ φk(X)∂kf(X)

∂xkj

= 0

Examples■ f = sin(2x3 + 3yz) is D-finite with respect to {x, y, z}.xfxx − 2fx + 9x5f = 0, fyy − 9z2f = 0, fzz − 9y2f = 0

■ exp(polynomial) is D-finite■

∑f(n)zn D-finite ⇐⇒ f(n) P-recursive

The Set Up

Creative telescoping

A combinatorial framework

● Generating functions

● D-finite functions

● D-finiteness

The scalar product

Conclusion

Marni Mishna, Freaky Friday the 13th, 2004 Carleton University Holonomy and Combinatorics – p. 14/26

Hierarchy of Power Series

Differentiably Algebraic partitions

D-finite k-regular graphs

Algebraic Dyck Paths

Rationalunrestricted walksin the plane

1−√

1−4x2x

Regular expressionsMotzkin Paths

ex

k × n-latin squares grid walks in the first quadrant

eex−1

Algebraic closure Hadamard product Algebraic substitutionFunctional Inverse Diagonals Hadamard product

Integrals Multiplicative inverse

The Set Up

Creative telescoping

A combinatorial framework

● Generating functions

● D-finite functions

● D-finiteness

The scalar product

Conclusion

Marni Mishna, Freaky Friday the 13th, 2004 Carleton University Holonomy and Combinatorics – p. 14/26

Hierarchy of Power Series

Differentiably Algebraic partitions

D-finite k-regular graphs

Algebraic Dyck Paths

Rationalunrestricted walksin the plane

1−√

1−4x2x

Regular expressionsMotzkin Paths

ex

k × n-latin squares grid walks in the first quadrant

eex−1

Algebraic closure Hadamard product Algebraic substitutionFunctional Inverse Diagonals Hadamard product

Integrals Multiplicative inverse

The Set Up

Creative telescoping

A combinatorial framework

● Generating functions

● D-finite functions

● D-finiteness

The scalar product

Conclusion

Marni Mishna, Freaky Friday the 13th, 2004 Carleton University Holonomy and Combinatorics – p. 14/26

Hierarchy of Power Series

Differentiably Algebraic partitions

D-finite k-regular graphs

Algebraic Dyck Paths

Rationalunrestricted walksin the plane

1−√

1−4x2x

Regular expressionsMotzkin Paths

ex

k × n-latin squares grid walks in the first quadrant

eex−1

Algebraic closure Hadamard product Algebraic substitutionFunctional Inverse Diagonals Hadamard product

Integrals Multiplicative inverse

The Set Up

Creative telescoping

A combinatorial framework

The scalar product

● Symmetric Series

● D-finite symmetric series

● The Scalar Product

● The Scalar Product

● Encoding combinatorial objects

● Weyl Algebras

● How does it work?

● Holonomic Modules

● Proof

● Other applications

Conclusion

Marni Mishna, Freaky Friday the 13th, 2004 Carleton University Holonomy and Combinatorics – p. 15/26

The scalar product

The Set Up

Creative telescoping

A combinatorial framework

The scalar product

● Symmetric Series

● D-finite symmetric series

● The Scalar Product

● The Scalar Product

● Encoding combinatorial objects

● Weyl Algebras

● How does it work?

● Holonomic Modules

● Proof

● Other applications

Conclusion

Marni Mishna, Freaky Friday the 13th, 2004 Carleton University Holonomy and Combinatorics – p. 16/26

Symmetric Series

power pk p3 = x31 + x3

2 + x33 + . . .

homogeneous hk h3 = x31 + x3

2 + . . .+ x21x2 + . . .+ x1x2x3 + . . .

elementary en e3 = x1x2x3 + x1x2x4 + x1x3x4 + . . .

monomial mλ m(3,2,1) = x31x

22x3 + x3

2x21x3 + x3

3x22x6 + . . .

The Set Up

Creative telescoping

A combinatorial framework

The scalar product

● Symmetric Series

● D-finite symmetric series

● The Scalar Product

● The Scalar Product

● Encoding combinatorial objects

● Weyl Algebras

● How does it work?

● Holonomic Modules

● Proof

● Other applications

Conclusion

Marni Mishna, Freaky Friday the 13th, 2004 Carleton University Holonomy and Combinatorics – p. 16/26

Symmetric Series

power pk p3 = x31 + x3

2 + x33 + . . .

homogeneous hk h3 = x31 + x3

2 + . . .+ x21x2 + . . .+ x1x2x3 + . . .

elementary en e3 = x1x2x3 + x1x2x4 + x1x3x4 + . . .

monomial mλ m(3,2,1) = x31x

22x3 + x3

2x21x3 + x3

3x22x6 + . . .

■ pλ = pλ1pλ2

· · · pλk(e.g. p(3,2,2,1) = p3p

22p1)

■ Our algebra of interest: K[[p1, p2, . . . ]]

■ Ex:H =

n hn = exp(∑

npn

n)

E =∑

n en = exp(∑

n(−1)n+1 pn

n)

■ Scalar product: 〈pλ, pµ〉 = δλµzλ

λ = (1m12m2 · · · kmk ) =⇒ zλ = 1m1m1!2m2m2! · · · kmkmk!

The Set Up

Creative telescoping

A combinatorial framework

The scalar product

● Symmetric Series

● D-finite symmetric series

● The Scalar Product

● The Scalar Product

● Encoding combinatorial objects

● Weyl Algebras

● How does it work?

● Holonomic Modules

● Proof

● Other applications

Conclusion

Marni Mishna, Freaky Friday the 13th, 2004 Carleton University Holonomy and Combinatorics – p. 16/26

Symmetric Series

power pk p3 = x31 + x3

2 + x33 + . . .

homogeneous hk h3 = x31 + x3

2 + . . .+ x21x2 + . . .+ x1x2x3 + . . .

elementary en e3 = x1x2x3 + x1x2x4 + x1x3x4 + . . .

monomial mλ m(3,2,1) = x31x

22x3 + x3

2x21x3 + x3

3x22x6 + . . .

■ pλ = pλ1pλ2

· · · pλk(e.g. p(3,2,2,1) = p3p

22p1)

■ Our algebra of interest: K[[p1, p2, . . . ]]

■ Ex:H =

n hn = exp(∑

npn

n)

E =∑

n en = exp(∑

n(−1)n+1 pn

n)

■ Scalar product: 〈pλ, pµ〉 = δλµzλ

λ = (1m12m2 · · · kmk ) =⇒ zλ = 1m1m1!2m2m2! · · · kmkmk!

The Set Up

Creative telescoping

A combinatorial framework

The scalar product

● Symmetric Series

● D-finite symmetric series

● The Scalar Product

● The Scalar Product

● Encoding combinatorial objects

● Weyl Algebras

● How does it work?

● Holonomic Modules

● Proof

● Other applications

Conclusion

Marni Mishna, Freaky Friday the 13th, 2004 Carleton University Holonomy and Combinatorics – p. 16/26

Symmetric Series

power pk p3 = x31 + x3

2 + x33 + . . .

homogeneous hk h3 = x31 + x3

2 + . . .+ x21x2 + . . .+ x1x2x3 + . . .

elementary en e3 = x1x2x3 + x1x2x4 + x1x3x4 + . . .

monomial mλ m(3,2,1) = x31x

22x3 + x3

2x21x3 + x3

3x22x6 + . . .

■ pλ = pλ1pλ2

· · · pλk(e.g. p(3,2,2,1) = p3p

22p1)

■ Our algebra of interest: K[[p1, p2, . . . ]]

■ Ex:H =

n hn = exp(∑

npn

n)

E =∑

n en = exp(∑

n(−1)n+1 pn

n)

■ Scalar product: 〈pλ, pµ〉 = δλµzλ

λ = (1m12m2 · · · kmk ) =⇒ zλ = 1m1m1!2m2m2! · · · kmkmk!

The Set Up

Creative telescoping

A combinatorial framework

The scalar product

● Symmetric Series

● D-finite symmetric series

● The Scalar Product

● The Scalar Product

● Encoding combinatorial objects

● Weyl Algebras

● How does it work?

● Holonomic Modules

● Proof

● Other applications

Conclusion

Marni Mishna, Freaky Friday the 13th, 2004 Carleton University Holonomy and Combinatorics – p. 16/26

Symmetric Series

power pk p3 = x31 + x3

2 + x33 + . . .

homogeneous hk h3 = x31 + x3

2 + . . .+ x21x2 + . . .+ x1x2x3 + . . .

elementary en e3 = x1x2x3 + x1x2x4 + x1x3x4 + . . .

monomial mλ m(3,2,1) = x31x

22x3 + x3

2x21x3 + x3

3x22x6 + . . .

■ pλ = pλ1pλ2

· · · pλk(e.g. p(3,2,2,1) = p3p

22p1)

■ Our algebra of interest: K[[p1, p2, . . . ]]

■ Ex:H =

n hn = exp(∑

npn

n)

E =∑

n en = exp(∑

n(−1)n+1 pn

n)

■ Scalar product: 〈pλ, pµ〉 = δλµzλ

λ = (1m12m2 · · · kmk ) =⇒ zλ = 1m1m1!2m2m2! · · · kmkmk!

The Set Up

Creative telescoping

A combinatorial framework

The scalar product

● Symmetric Series

● D-finite symmetric series

● The Scalar Product

● The Scalar Product

● Encoding combinatorial objects

● Weyl Algebras

● How does it work?

● Holonomic Modules

● Proof

● Other applications

Conclusion

Marni Mishna, Freaky Friday the 13th, 2004 Carleton University Holonomy and Combinatorics – p. 17/26

D-finite symmetric series

Def. A symmetric series F ∈ K[[p1, p2, . . . ; t]] is D-finite if forany n,■ F (p1, ..., pn, 0, 0, . . . ; t) is D-finite wrt p1, . . . , pn, t.

■ Examples:

H =∑hnt

n = exp(∑

n pntn/n)

E =∑ent

n = exp(∑

n(−1)n+1pntn/n)

S =∑

λ sλt|λ| = exp(

n

p2

nt2n

2n+ p2n−1t2n−1

2n−1 )

λ all parts odd sλt|λ| = SE−1

The Set Up

Creative telescoping

A combinatorial framework

The scalar product

● Symmetric Series

● D-finite symmetric series

● The Scalar Product

● The Scalar Product

● Encoding combinatorial objects

● Weyl Algebras

● How does it work?

● Holonomic Modules

● Proof

● Other applications

Conclusion

Marni Mishna, Freaky Friday the 13th, 2004 Carleton University Holonomy and Combinatorics – p. 17/26

D-finite symmetric series

Def. A symmetric series F ∈ K[[p1, p2, . . . ; t]] is D-finite if forany n,■ F (p1, ..., pn, 0, 0, . . . ; t) is D-finite wrt p1, . . . , pn, t.■ Examples:

H =∑hnt

n = exp(∑

n pntn/n)

E =∑ent

n = exp(∑

n(−1)n+1pntn/n)

S =∑

λ sλt|λ| = exp(

n

p2

nt2n

2n+ p2n−1t2n−1

2n−1 )

λ all parts odd sλt|λ| = SE−1

The Set Up

Creative telescoping

A combinatorial framework

The scalar product

● Symmetric Series

● D-finite symmetric series

● The Scalar Product

● The Scalar Product

● Encoding combinatorial objects

● Weyl Algebras

● How does it work?

● Holonomic Modules

● Proof

● Other applications

Conclusion

Marni Mishna, Freaky Friday the 13th, 2004 Carleton University Holonomy and Combinatorics – p. 18/26

The Scalar Product

■ The scalar product, (symmetric, bilinear form) is defined by

〈pλ, pµ〉 = δλµzλ.

λ = (1m12m2 · · · kmk ) =⇒ zλ = 1m1m1!2m2m2! · · · kmkmk!

■ Equivalent to coefficient extraction:

[xk1

1 xk2

2 . . . xkn

n ]F = 〈F, hk1hk2

· · ·hkn〉

■ Adjunction:〈pnF,G〉 = 〈F, n ∂

∂pnG〉.

The Set Up

Creative telescoping

A combinatorial framework

The scalar product

● Symmetric Series

● D-finite symmetric series

● The Scalar Product

● The Scalar Product

● Encoding combinatorial objects

● Weyl Algebras

● How does it work?

● Holonomic Modules

● Proof

● Other applications

Conclusion

Marni Mishna, Freaky Friday the 13th, 2004 Carleton University Holonomy and Combinatorics – p. 18/26

The Scalar Product

■ The scalar product, (symmetric, bilinear form) is defined by

〈pλ, pµ〉 = δλµzλ.

λ = (1m12m2 · · · kmk ) =⇒ zλ = 1m1m1!2m2m2! · · · kmkmk!

■ Equivalent to coefficient extraction:

[xk1

1 xk2

2 . . . xkn

n ]F = 〈F, hk1hk2

· · ·hkn〉

■ Adjunction:〈pnF,G〉 = 〈F, n ∂

∂pnG〉.

The Set Up

Creative telescoping

A combinatorial framework

The scalar product

● Symmetric Series

● D-finite symmetric series

● The Scalar Product

● The Scalar Product

● Encoding combinatorial objects

● Weyl Algebras

● How does it work?

● Holonomic Modules

● Proof

● Other applications

Conclusion

Marni Mishna, Freaky Friday the 13th, 2004 Carleton University Holonomy and Combinatorics – p. 18/26

The Scalar Product

■ The scalar product, (symmetric, bilinear form) is defined by

〈pλ, pµ〉 = δλµzλ.

λ = (1m12m2 · · · kmk ) =⇒ zλ = 1m1m1!2m2m2! · · · kmkmk!

■ Equivalent to coefficient extraction:

[xk1

1 xk2

2 . . . xkn

n ]F = 〈F, hk1hk2

· · ·hkn〉

■ Adjunction:〈pnF,G〉 = 〈F, n ∂

∂pnG〉.

The Set Up

Creative telescoping

A combinatorial framework

The scalar product

● Symmetric Series

● D-finite symmetric series

● The Scalar Product

● The Scalar Product

● Encoding combinatorial objects

● Weyl Algebras

● How does it work?

● Holonomic Modules

● Proof

● Other applications

Conclusion

Marni Mishna, Freaky Friday the 13th, 2004 Carleton University Holonomy and Combinatorics – p. 19/26

The Scalar Product

Theorem (Gessel, 90): If F ∈ K[[p1, p2, . . . , pn; t]] and G ∈ K[[p1, p2, . . . ; t]]are both D-finite symmetric series then the scalar product

〈F,G〉 = φ(t)

is a D-finite function with respect to t.

That is, φ(t) satisfies a linear DE with polynomial coefficients.Infact, this DE can be calculated (Chyzak, M., Salvy 02)

Termination guaranteed by D-finiteness.

Ex:> scalar de(exp(p1+p12/2+p2+p22),exp((p22+p2/2)*t),[t],f);

[(−126 t+ 13− 16 t2

)f (t) +

(−1 + 32 t− 256 t2

)ddtf (t)]

The Set Up

Creative telescoping

A combinatorial framework

The scalar product

● Symmetric Series

● D-finite symmetric series

● The Scalar Product

● The Scalar Product

● Encoding combinatorial objects

● Weyl Algebras

● How does it work?

● Holonomic Modules

● Proof

● Other applications

Conclusion

Marni Mishna, Freaky Friday the 13th, 2004 Carleton University Holonomy and Combinatorics – p. 19/26

The Scalar Product

Theorem (Gessel, 90): If F ∈ K[[p1, p2, . . . , pn; t]] and G ∈ K[[p1, p2, . . . ; t]]are both D-finite symmetric series then the scalar product

〈F,G〉 = φ(t)

is a D-finite function with respect to t.

That is, φ(t) satisfies a linear DE with polynomial coefficients.Infact, this DE can be calculated (Chyzak, M., Salvy 02)

Termination guaranteed by D-finiteness.

Ex:> scalar de(exp(p1+p12/2+p2+p22),exp((p22+p2/2)*t),[t],f);

[(−126 t+ 13− 16 t2

)f (t) +

(−1 + 32 t− 256 t2

)ddtf (t)]

Marni Mishna, Freaky Friday the 13th, 2004 Carleton University Holonomy and Combinatorics – p. 20/26

Encoding combinatorial objects

Object Encoding π Sum over objects in the class

MAKE SPACE

t

,,

eee

ll

%%%

,,

t t

tt

t

t2 3

4

6 7

5

1

x31x

42x

23x

54x

55x

26x

27

G =∑

g a graph π(g)

= exp(∑

n

p2

n

2n− p2n

2n)

Set(C) x21x

22x

23x4x5 H[ZC], E[ZC]

C a labelled,finite comb. class

xk

i=⇒ k occurrences

of label i

H =

hn, E =

en,Z = Polya cycle index,[]= plethystic substitution

Marni Mishna, Freaky Friday the 13th, 2004 Carleton University Holonomy and Combinatorics – p. 20/26

Encoding combinatorial objects

Object Encoding π Generating function of regular subclass

MAKE SPACE

t

t

t

t2

4 3

1

x21x

22x

23x

24 Gk(t) = 〈G,

∑hn

2 tn〉

■ Idea: Encode all objects, use scalar product to extract the number with acertain regularity

■ Scalar product with a series gives the egf of a “regular” subclass.

■ Other examples: Young tableaux, k × n-latin squares, hypergraphs

The Set Up

Creative telescoping

A combinatorial framework

The scalar product

● Symmetric Series

● D-finite symmetric series

● The Scalar Product

● The Scalar Product

● Encoding combinatorial objects

● Weyl Algebras

● How does it work?

● Holonomic Modules

● Proof

● Other applications

Conclusion

Marni Mishna, Freaky Friday the 13th, 2004 Carleton University Holonomy and Combinatorics – p. 21/26

Weyl Algebras

Ap := K〈p1, . . . , pn, ∂1, . . . , ∂n〉

∂ipi = pi∂i + 1, all other variables commute

f(p1, . . . , pn)

Left annihilating ideal: If

= {φ ∈ Ap|φ · f = 0}

Left module: Apf

= {ψ · f |ψ ∈ Ap} ' Ap/If

The Set Up

Creative telescoping

A combinatorial framework

The scalar product

● Symmetric Series

● D-finite symmetric series

● The Scalar Product

● The Scalar Product

● Encoding combinatorial objects

● Weyl Algebras

● How does it work?

● Holonomic Modules

● Proof

● Other applications

Conclusion

Marni Mishna, Freaky Friday the 13th, 2004 Carleton University Holonomy and Combinatorics – p. 21/26

Weyl Algebras

Ap := K〈p1, . . . , pn, ∂1, . . . , ∂n〉

∂ipi = pi∂i + 1, all other variables commute

f(p1, . . . , pn)

Left annihilating ideal: If

= {φ ∈ Ap|φ · f = 0}

Left module: Apf

= {ψ · f |ψ ∈ Ap} ' Ap/If

The Set Up

Creative telescoping

A combinatorial framework

The scalar product

● Symmetric Series

● D-finite symmetric series

● The Scalar Product

● The Scalar Product

● Encoding combinatorial objects

● Weyl Algebras

● How does it work?

● Holonomic Modules

● Proof

● Other applications

Conclusion

Marni Mishna, Freaky Friday the 13th, 2004 Carleton University Holonomy and Combinatorics – p. 21/26

Weyl Algebras

Ap := K〈p1, . . . , pn, ∂1, . . . , ∂n〉

∂ipi = pi∂i + 1, all other variables commute

f(p1, . . . , pn) f = sin(p1 + p2/2)

Left annihilating ideal: If

= {φ ∈ Ap|φ · f = 0} If = Ap〈∂21 + 1, 4∂2

2 + 1, 2∂1 − ∂2〉

Left module: Apf

= {ψ · f |ψ ∈ Ap} ' Ap/If Apf = K[p1, p2]sin(p1 + p2/2)

⊕K[p1, p2]cos(p1 + p2/2)

The Set Up

Creative telescoping

A combinatorial framework

The scalar product

● Symmetric Series

● D-finite symmetric series

● The Scalar Product

● The Scalar Product

● Encoding combinatorial objects

● Weyl Algebras

● How does it work?

● Holonomic Modules

● Proof

● Other applications

Conclusion

Marni Mishna, Freaky Friday the 13th, 2004 Carleton University Holonomy and Combinatorics – p. 22/26

How does it work?

Like with creative telescoping, we take linear combinations ofequations to eliminate variables and generate new equations.0− 0 = 0〈0, G〉 − 〈F, 0〉 = 0

The Set Up

Creative telescoping

A combinatorial framework

The scalar product

● Symmetric Series

● D-finite symmetric series

● The Scalar Product

● The Scalar Product

● Encoding combinatorial objects

● Weyl Algebras

● How does it work?

● Holonomic Modules

● Proof

● Other applications

Conclusion

Marni Mishna, Freaky Friday the 13th, 2004 Carleton University Holonomy and Combinatorics – p. 22/26

How does it work?

Like with creative telescoping, we take linear combinations ofequations to eliminate variables and generate new equations.0− 0 = 0〈0, G〉 − 〈F, 0〉 = 0〈φ · F,G〉 − 〈F, ψ ·G〉 = 0 φ ∈ IF , ψ ∈ IG

The Set Up

Creative telescoping

A combinatorial framework

The scalar product

● Symmetric Series

● D-finite symmetric series

● The Scalar Product

● The Scalar Product

● Encoding combinatorial objects

● Weyl Algebras

● How does it work?

● Holonomic Modules

● Proof

● Other applications

Conclusion

Marni Mishna, Freaky Friday the 13th, 2004 Carleton University Holonomy and Combinatorics – p. 22/26

How does it work?

Like with creative telescoping, we take linear combinations ofequations to eliminate variables and generate new equations.0− 0 = 0〈0, G〉 − 〈F, 0〉 = 0〈φ · F,G〉 − 〈F, ψ ·G〉 = 0 φ ∈ IF , ψ ∈ IG〈F, φ⊥ ·G〉 − 〈F, ψ ·G〉 = 0 p⊥n = n∂n, ∂

⊥n = pn/n

The Set Up

Creative telescoping

A combinatorial framework

The scalar product

● Symmetric Series

● D-finite symmetric series

● The Scalar Product

● The Scalar Product

● Encoding combinatorial objects

● Weyl Algebras

● How does it work?

● Holonomic Modules

● Proof

● Other applications

Conclusion

Marni Mishna, Freaky Friday the 13th, 2004 Carleton University Holonomy and Combinatorics – p. 22/26

How does it work?

Like with creative telescoping, we take linear combinations ofequations to eliminate variables and generate new equations.0− 0 = 0〈0, G〉 − 〈F, 0〉 = 0〈φ · F,G〉 − 〈F, ψ ·G〉 = 0 φ ∈ IF , ψ ∈ IG〈F, φ⊥ ·G〉 − 〈F, ψ ·G〉 = 0 p⊥n = n∂n, ∂

⊥n = pn/n

〈F,(φ⊥ − ψ

)·G〉 = 0 by linearity

The Set Up

Creative telescoping

A combinatorial framework

The scalar product

● Symmetric Series

● D-finite symmetric series

● The Scalar Product

● The Scalar Product

● Encoding combinatorial objects

● Weyl Algebras

● How does it work?

● Holonomic Modules

● Proof

● Other applications

Conclusion

Marni Mishna, Freaky Friday the 13th, 2004 Carleton University Holonomy and Combinatorics – p. 22/26

How does it work?

Like with creative telescoping, we take linear combinations ofequations to eliminate variables and generate new equations.0− 0 = 0〈0, G〉 − 〈F, 0〉 = 0〈φ · F,G〉 − 〈F, ψ ·G〉 = 0 φ ∈ IF , ψ ∈ IG〈F, φ⊥ ·G〉 − 〈F, ψ ·G〉 = 0 p⊥n = n∂n, ∂

⊥n = pn/n

〈F,(φ⊥ − ψ

)·G〉 = 0 by linearity

If,(φ⊥ − ψ

)= γ(t, ∂t) ∈ C[t, ∂t], then...

〈F, γ(t, ∂t) ·G〉 = 0

=⇒ γ(t, ∂t) · 〈F,G〉 = 0

The Set Up

Creative telescoping

A combinatorial framework

The scalar product

● Symmetric Series

● D-finite symmetric series

● The Scalar Product

● The Scalar Product

● Encoding combinatorial objects

● Weyl Algebras

● How does it work?

● Holonomic Modules

● Proof

● Other applications

Conclusion

Marni Mishna, Freaky Friday the 13th, 2004 Carleton University Holonomy and Combinatorics – p. 22/26

How does it work?

Like with creative telescoping, we take linear combinations ofequations to eliminate variables and generate new equations.0− 0 = 0〈0, G〉 − 〈F, 0〉 = 0〈φ · F,G〉 − 〈F, ψ ·G〉 = 0 φ ∈ IF , ψ ∈ IG〈F, φ⊥ ·G〉 − 〈F, ψ ·G〉 = 0 p⊥n = n∂n, ∂

⊥n = pn/n

〈F,(φ⊥ − ψ

)·G〉 = 0 by linearity

If,(φ⊥ − ψ

)= γ(t, ∂t) ∈ C[t, ∂t], then...

〈F, γ(t, ∂t) ·G〉 = 0 =⇒ γ(t, ∂t) · 〈F,G〉 = 0

This gives a differential equation satisfied by 〈F,G〉.

The Set Up

Creative telescoping

A combinatorial framework

The scalar product

● Symmetric Series

● D-finite symmetric series

● The Scalar Product

● The Scalar Product

● Encoding combinatorial objects

● Weyl Algebras

● How does it work?

● Holonomic Modules

● Proof

● Other applications

Conclusion

Marni Mishna, Freaky Friday the 13th, 2004 Carleton University Holonomy and Combinatorics – p. 22/26

How does it work?

Like with creative telescoping, we take linear combinations ofequations to eliminate variables and generate new equations.0− 0 = 0〈0, G〉 − 〈F, 0〉 = 0〈φ · F,G〉 − 〈F, ψ ·G〉 = 0 φ ∈ IF , ψ ∈ IG〈F, φ⊥ ·G〉 − 〈F, ψ ·G〉 = 0 p⊥n = n∂n, ∂

⊥n = pn/n

〈F,(φ⊥ − ψ

)·G〉 = 0 by linearity

If,(φ⊥ − ψ

)= γ(t, ∂t) ∈ C[t, ∂t], then...

〈F, γ(t, ∂t) ·G〉 = 0 =⇒ γ(t, ∂t) · 〈F,G〉 = 0

This gives a differential equation satisfied by 〈F,G〉.

Any non-trivial element γ of(

I⊥f + Ig

)⋂C[t, ∂t] provides a

differential equation satisfied by 〈f, g〉.

The Set Up

Creative telescoping

A combinatorial framework

The scalar product

● Symmetric Series

● D-finite symmetric series

● The Scalar Product

● The Scalar Product

● Encoding combinatorial objects

● Weyl Algebras

● How does it work?

● Holonomic Modules

● Proof

● Other applications

Conclusion

Marni Mishna, Freaky Friday the 13th, 2004 Carleton University Holonomy and Combinatorics – p. 23/26

Holonomic Modules

Ax := K〈x1, . . . , xn, ∂1, . . . , ∂n〉∂ixi = xi∂i + 1, all other variables commute

Filtration: A(d)x = {ψ ∈ Ax| degψ ≤ d}

Def. A module Axf is holonomic if

dimA(d)x · f = O(dn).

✔ K[x1, ..., xn] = Ax • 1: dimA(d)x · 1 = O(dn)

✘ Ax: dimA(d)x = O(d2n)

✔ Ax · sin(x) ' K[x] sin(x)⊕K[x] cos(x)

dimA(d)x · sin(x) = 2 dimK[x](d) = O(d1)

Theorem: (Kashiwara) f(x) D-finite ⇐⇒ Axf holonomic

The Set Up

Creative telescoping

A combinatorial framework

The scalar product

● Symmetric Series

● D-finite symmetric series

● The Scalar Product

● The Scalar Product

● Encoding combinatorial objects

● Weyl Algebras

● How does it work?

● Holonomic Modules

● Proof

● Other applications

Conclusion

Marni Mishna, Freaky Friday the 13th, 2004 Carleton University Holonomy and Combinatorics – p. 23/26

Holonomic Modules

Ax := K〈x1, . . . , xn, ∂1, . . . , ∂n〉∂ixi = xi∂i + 1, all other variables commute

Filtration: A(d)x = {ψ ∈ Ax| degψ ≤ d}

Def. A module Axf is holonomic if

dimA(d)x · f = O(dn).

✔ K[x1, ..., xn] = Ax • 1: dimA(d)x · 1 = O(dn)

✘ Ax: dimA(d)x = O(d2n)

✔ Ax · sin(x) ' K[x] sin(x)⊕K[x] cos(x)

dimA(d)x · sin(x) = 2 dimK[x](d) = O(d1)

Theorem: (Kashiwara) f(x) D-finite ⇐⇒ Axf holonomic

The Set Up

Creative telescoping

A combinatorial framework

The scalar product

● Symmetric Series

● D-finite symmetric series

● The Scalar Product

● The Scalar Product

● Encoding combinatorial objects

● Weyl Algebras

● How does it work?

● Holonomic Modules

● Proof

● Other applications

Conclusion

Marni Mishna, Freaky Friday the 13th, 2004 Carleton University Holonomy and Combinatorics – p. 23/26

Holonomic Modules

Ax := K〈x1, . . . , xn, ∂1, . . . , ∂n〉∂ixi = xi∂i + 1, all other variables commute

Filtration: A(d)x = {ψ ∈ Ax| degψ ≤ d}

Def. A module Axf is holonomic if

dimA(d)x · f = O(dn).

✔ K[x1, ..., xn] = Ax • 1: dimA(d)x · 1 = O(dn)

✘ Ax: dimA(d)x = O(d2n)

✔ Ax · sin(x) ' K[x] sin(x)⊕K[x] cos(x)

dimA(d)x · sin(x) = 2 dimK[x](d) = O(d1)

Theorem: (Kashiwara) f(x) D-finite ⇐⇒ Axf holonomic

The Set Up

Creative telescoping

A combinatorial framework

The scalar product

● Symmetric Series

● D-finite symmetric series

● The Scalar Product

● The Scalar Product

● Encoding combinatorial objects

● Weyl Algebras

● How does it work?

● Holonomic Modules

● Proof

● Other applications

Conclusion

Marni Mishna, Freaky Friday the 13th, 2004 Carleton University Holonomy and Combinatorics – p. 23/26

Holonomic Modules

Ax := K〈x1, . . . , xn, ∂1, . . . , ∂n〉∂ixi = xi∂i + 1, all other variables commute

Filtration: A(d)x = {ψ ∈ Ax| degψ ≤ d}

Def. A module Axf is holonomic if

dimA(d)x · f = O(dn).

✔ K[x1, ..., xn] = Ax • 1: dimA(d)x · 1 = O(dn)

✘ Ax: dimA(d)x = O(d2n)

✔ Ax · sin(x) ' K[x] sin(x)⊕K[x] cos(x)

dimA(d)x · sin(x) = 2 dimK[x](d) = O(d1)

Theorem: (Kashiwara) f(x) D-finite ⇐⇒ Axf holonomic

The Set Up

Creative telescoping

A combinatorial framework

The scalar product

● Symmetric Series

● D-finite symmetric series

● The Scalar Product

● The Scalar Product

● Encoding combinatorial objects

● Weyl Algebras

● How does it work?

● Holonomic Modules

● Proof

● Other applications

Conclusion

Marni Mishna, Freaky Friday the 13th, 2004 Carleton University Holonomy and Combinatorics – p. 23/26

Holonomic Modules

Ax := K〈x1, . . . , xn, ∂1, . . . , ∂n〉∂ixi = xi∂i + 1, all other variables commute

Filtration: A(d)x = {ψ ∈ Ax| degψ ≤ d}

Def. A module Axf is holonomic if

dimA(d)x · f = O(dn).

✔ K[x1, ..., xn] = Ax • 1: dimA(d)x · 1 = O(dn)

✘ Ax: dimA(d)x = O(d2n)

✔ Ax · sin(x) ' K[x] sin(x)⊕K[x] cos(x)

dimA(d)x · sin(x) = 2 dimK[x](d) = O(d1)

Theorem: (Kashiwara) f(x) D-finite ⇐⇒ Axf holonomic

The Set Up

Creative telescoping

A combinatorial framework

The scalar product

● Symmetric Series

● D-finite symmetric series

● The Scalar Product

● The Scalar Product

● Encoding combinatorial objects

● Weyl Algebras

● How does it work?

● Holonomic Modules

● Proof

● Other applications

Conclusion

Marni Mishna, Freaky Friday the 13th, 2004 Carleton University Holonomy and Combinatorics – p. 24/26

(I⊥f + Ig

)∩ At is not empty!

Idea: model adjunction p⊥n = n∂pnby a tensor product

■ Define surjection:Ap,tf ⊗

⊥C[t] Ap,tg −→ Ap,t〈f, g〉

a⊗⊥C[t] b 7→ 〈a, b〉

Prop: I⊥f + Ig contains AnnAt(f ⊗⊥C[t] g)

■ Thm: Ap,tf ⊗⊥C[t] Ap,tg is a holonomic Ap,t-module

(

Ap,tf ⊗⊥C[t] Ap,tg

)

∩ “At” is a holonomic At-module

isomorphic to At/AnnAt(f ⊗⊥

C[t] g) as At-modules

Cor: AnnAt(f ⊗⊥

C[t] g) is non-trivial.

The Set Up

Creative telescoping

A combinatorial framework

The scalar product

● Symmetric Series

● D-finite symmetric series

● The Scalar Product

● The Scalar Product

● Encoding combinatorial objects

● Weyl Algebras

● How does it work?

● Holonomic Modules

● Proof

● Other applications

Conclusion

Marni Mishna, Freaky Friday the 13th, 2004 Carleton University Holonomy and Combinatorics – p. 24/26

(I⊥f + Ig

)∩ At is not empty!

Idea: model adjunction p⊥n = n∂pnby a tensor product

■ Define surjection:Ap,tf ⊗

⊥C[t] Ap,tg −→ Ap,t〈f, g〉

a⊗⊥C[t] b 7→ 〈a, b〉

Prop: I⊥f + Ig contains AnnAt(f ⊗⊥C[t] g)

■ Thm: Ap,tf ⊗⊥C[t] Ap,tg is a holonomic Ap,t-module

key fact: ⊥ is degree preserving

(

Ap,tf ⊗⊥C[t] Ap,tg

)

∩ “At” is a holonomic At-module

isomorphic to At/AnnAt(f ⊗⊥

C[t] g) as At-modules

Cor: AnnAt(f ⊗⊥

C[t] g) is non-trivial.

The Set Up

Creative telescoping

A combinatorial framework

The scalar product

● Symmetric Series

● D-finite symmetric series

● The Scalar Product

● The Scalar Product

● Encoding combinatorial objects

● Weyl Algebras

● How does it work?

● Holonomic Modules

● Proof

● Other applications

Conclusion

Marni Mishna, Freaky Friday the 13th, 2004 Carleton University Holonomy and Combinatorics – p. 24/26

(I⊥f + Ig

)∩ At is not empty!

Idea: model adjunction p⊥n = n∂pnby a tensor product

■ Define surjection:Ap,tf ⊗

⊥C[t] Ap,tg −→ Ap,t〈f, g〉

a⊗⊥C[t] b 7→ 〈a, b〉

Prop: I⊥f + Ig contains AnnAt(f ⊗⊥C[t] g)

■ Thm: Ap,tf ⊗⊥C[t] Ap,tg is a holonomic Ap,t-module

(

Ap,tf ⊗⊥C[t] Ap,tg

)

∩ “At” is a holonomic At-module

isomorphic to At/AnnAt(f ⊗⊥

C[t] g) as At-modules

Cor: AnnAt(f ⊗⊥

C[t] g) is non-trivial.

The Set Up

Creative telescoping

A combinatorial framework

The scalar product

● Symmetric Series

● D-finite symmetric series

● The Scalar Product

● The Scalar Product

● Encoding combinatorial objects

● Weyl Algebras

● How does it work?

● Holonomic Modules

● Proof

● Other applications

Conclusion

Marni Mishna, Freaky Friday the 13th, 2004 Carleton University Holonomy and Combinatorics – p. 25/26

Other applications

Algorithm computes scalar products whose adjoint operation isa degree preserving operation:➢Related to Macdonald Polynomials

〈pλ, pµ〉q,t = zλδµλ

λi∈λ

1− qλi

1− tλi

➢MacMahon symmetric functions

〈p(a,b)2(c,d), p(a,b)2(c,d)〉 = 2

(a!b!

(a+ b− 1)!

)2 (c!d!

(c+ d− 1)!

)

The Set Up

Creative telescoping

A combinatorial framework

The scalar product

Conclusion

● Lots to do...

Marni Mishna, Freaky Friday the 13th, 2004 Carleton University Holonomy and Combinatorics – p. 26/26

Lots to do...

■ Other D-finite closure properties?■ Effective closure properties??■ Classifying families of combinatorial objects e.g. walks in the

plane, families of generalized partitions■ Improving the Gröbner basis calculations