Rational Parking Functions

Transcription

1 Rational Parking Functions Drew Armstrong et al. University of Miami armstrong December 9, 2012

2 Rational Catalan Numbers CONVENTION Given x Q \ [ 1, 0], there exist unique coprime (a, b) N 2 such that x = We will always identify x (a, b). a b a. Definition For each x Q \ [ 1, 0] we define the Catalan number: Cat(x) = Cat(a, b) := 1 ( ) a + b (a + b 1)! =. a + b a, b a!b!

3 Rational Catalan Numbers CONVENTION Given x Q \ [ 1, 0], there exist unique coprime (a, b) N 2 such that x = We will always identify x (a, b). a b a. Definition For each x Q \ [ 1, 0] we define the Catalan number: Cat(x) = Cat(a, b) := 1 ( ) a + b (a + b 1)! =. a + b a, b a!b!

4 Special cases When b = 1 mod a... Eugène Charles Catalan ( ) (a, b) = (n, n + 1) gives the good old Catalan number: ( ) n Cat(n) = Cat = 1 ( ) 2n + 1. (n + 1) n 2n + 1 n Nicolaus Fuss ( ) (a, b) = (n, kn + 1) gives the Fuss-Catalan number: ( ) ( ) n 1 (k + 1)n + 1 Cat =. (kn + 1) n (k + 1)n + 1 n

5 Special cases When b = 1 mod a... Eugène Charles Catalan ( ) (a, b) = (n, n + 1) gives the good old Catalan number: ( ) n Cat(n) = Cat = 1 ( ) 2n + 1. (n + 1) n 2n + 1 n Nicolaus Fuss ( ) (a, b) = (n, kn + 1) gives the Fuss-Catalan number: ( ) ( ) n 1 (k + 1)n + 1 Cat =. (kn + 1) n (k + 1)n + 1 n

6 Special cases When b = 1 mod a... Eugène Charles Catalan ( ) (a, b) = (n, n + 1) gives the good old Catalan number: ( ) n Cat(n) = Cat = 1 ( ) 2n + 1. (n + 1) n 2n + 1 n Nicolaus Fuss ( ) (a, b) = (n, kn + 1) gives the Fuss-Catalan number: ( ) ( ) n 1 (k + 1)n + 1 Cat =. (kn + 1) n (k + 1)n + 1 n

7 Symmetry Definition By definition we have Cat(a, b) = Cat(b, a), which translates to Cat(x) = Cat( x 1) (i.e. symmetry about x = 1/2), which implies that ( ) ( ) 1 x Cat = Cat. x 1 1 x We call this the derived Catalan number: ( ) ( ) 1 x Cat (x) := Cat = Cat. x 1 1 x

8 Symmetry Definition By definition we have Cat(a, b) = Cat(b, a), which translates to Cat(x) = Cat( x 1) (i.e. symmetry about x = 1/2), which implies that ( ) ( ) 1 x Cat = Cat. x 1 1 x We call this the derived Catalan number: ( ) ( ) 1 x Cat (x) := Cat = Cat. x 1 1 x

9 Symmetry Definition By definition we have Cat(a, b) = Cat(b, a), which translates to Cat(x) = Cat( x 1) (i.e. symmetry about x = 1/2), which implies that ( ) ( ) 1 x Cat = Cat. x 1 1 x We call this the derived Catalan number: ( ) ( ) 1 x Cat (x) := Cat = Cat. x 1 1 x

10 Symmetry Definition By definition we have Cat(a, b) = Cat(b, a), which translates to Cat(x) = Cat( x 1) (i.e. symmetry about x = 1/2), which implies that ( ) ( ) 1 x Cat = Cat. x 1 1 x We call this the derived Catalan number: ( ) ( ) 1 x Cat (x) := Cat = Cat. x 1 1 x

11 Euclidean Algorithm Observation The process Cat(x) Cat (x) Cat (x) is a categorification of the Euclidean algorithm. Example: x = 5/3 and (a, b) = (5, 8) Subtract the smaller from the larger: Cat(5, 8) = 99, Cat (5, 8) = Cat(3, 5) = 7, Cat (5, 8) = Cat (3, 5) = Cat(2, 3) = 2, Cat (5, 8) = Cat (3, 5) = Cat (2, 3) = Cat(1, 2) = 1 (STOP)

12 Euclidean Algorithm Observation The process Cat(x) Cat (x) Cat (x) is a categorification of the Euclidean algorithm. Example: x = 5/3 and (a, b) = (5, 8) Subtract the smaller from the larger: Cat(5, 8) = 99, Cat (5, 8) = Cat(3, 5) = 7, Cat (5, 8) = Cat (3, 5) = Cat(2, 3) = 2, Cat (5, 8) = Cat (3, 5) = Cat (2, 3) = Cat(1, 2) = 1 (STOP)

13 Euclidean Algorithm Observation The process Cat(x) Cat (x) Cat (x) is a categorification of the Euclidean algorithm. Example: x = 5/3 and (a, b) = (5, 8) Subtract the smaller from the larger: Cat(5, 8) = 99, Cat (5, 8) = Cat(3, 5) = 7, Cat (5, 8) = Cat (3, 5) = Cat(2, 3) = 2, Cat (5, 8) = Cat (3, 5) = Cat (2, 3) = Cat(1, 2) = 1 (STOP)

14 Euclidean Algorithm Observation The process Cat(x) Cat (x) Cat (x) is a categorification of the Euclidean algorithm. Example: x = 5/3 and (a, b) = (5, 8) Subtract the smaller from the larger: Cat(5, 8) = 99, Cat (5, 8) = Cat(3, 5) = 7, Cat (5, 8) = Cat (3, 5) = Cat(2, 3) = 2, Cat (5, 8) = Cat (3, 5) = Cat (2, 3) = Cat(1, 2) = 1 (STOP)

15 How to put it in Sloane s OEIS Suggestion The Calkin-Wilf sequence is defined by q 1 = 1 and q n := 1 2 q n 1 q n Theorem: (q 1, q 2,...) = Q >0. Proof: See Proofs from THE BOOK, Chapter 17. Study the function n Cat(q n ). q Cat(q)

16 How to put it in Sloane s OEIS Suggestion The Calkin-Wilf sequence is defined by q 1 = 1 and q n := 1 2 q n 1 q n Theorem: (q 1, q 2,...) = Q >0. Proof: See Proofs from THE BOOK, Chapter 17. Study the function n Cat(q n ). q Cat(q)

17 How to put it in Sloane s OEIS Suggestion The Calkin-Wilf sequence is defined by q 1 = 1 and q n := 1 2 q n 1 q n Theorem: (q 1, q 2,...) = Q >0. Proof: See Proofs from THE BOOK, Chapter 17. Study the function n Cat(q n ). q Cat(q)

18 Pause Well, that was fun.

19 The Prototype: Actuarial Science Consider the Dyck paths in an a b rectangle. Example (a, b) = (5, 8)

20 The Prototype: Actuarial Science Again let x = a/(b a) with a, b positive and coprime. Example (a, b) = (5, 8)

21 The Prototype: Actuarial Science Let D(x) = D(a, b) denote the set of Dyck paths. Example (a, b) = (5, 8)

22 The Prototype: Actuarial Science Theorem (Grossman 1950, Bizley 1954) The number of Dyck paths is the Catalan number: D(x) = Cat(x) = 1 ( ) a + b. a + b a, b Claimed by Grossman (1950), Fun with lattice points, part 22. Proved by Bizley (1954), in Journal of the Institute of Actuaries. Proof: Break ( a+b a,b ) lattice paths into cyclic orbits of size a + b. Each orbit contains a unique Dyck path.

23 The Prototype: Actuarial Science Theorem (Grossman 1950, Bizley 1954) The number of Dyck paths is the Catalan number: D(x) = Cat(x) = 1 ( ) a + b. a + b a, b Claimed by Grossman (1950), Fun with lattice points, part 22. Proved by Bizley (1954), in Journal of the Institute of Actuaries. Proof: Break ( a+b a,b ) lattice paths into cyclic orbits of size a + b. Each orbit contains a unique Dyck path.

24 The Prototype: Actuarial Science Theorem (Grossman 1950, Bizley 1954) The number of Dyck paths is the Catalan number: D(x) = Cat(x) = 1 ( ) a + b. a + b a, b Claimed by Grossman (1950), Fun with lattice points, part 22. Proved by Bizley (1954), in Journal of the Institute of Actuaries. Proof: Break ( a+b a,b ) lattice paths into cyclic orbits of size a + b. Each orbit contains a unique Dyck path.

25 The Prototype: Actuarial Science Theorem (Grossman 1950, Bizley 1954) The number of Dyck paths is the Catalan number: D(x) = Cat(x) = 1 ( ) a + b. a + b a, b Claimed by Grossman (1950), Fun with lattice points, part 22. Proved by Bizley (1954), in Journal of the Institute of Actuaries. Proof: Break ( a+b a,b ) lattice paths into cyclic orbits of size a + b. Each orbit contains a unique Dyck path.

26 The Prototype: Actuarial Science Theorem (Grossman 1950, Bizley 1954) The number of Dyck paths is the Catalan number: D(x) = Cat(x) = 1 ( ) a + b. a + b a, b Claimed by Grossman (1950), Fun with lattice points, part 22. Proved by Bizley (1954), in Journal of the Institute of Actuaries. Proof: Break ( a+b a,b ) lattice paths into cyclic orbits of size a + b. Each orbit contains a unique Dyck path.

27 The Prototype: Actuarial Science Theorem (Armstrong 2010, Loehr 2010) The number of Dyck paths with k vertical runs equals Nar(x; k) := 1 ( )( ) a b 1. a k k 1 Call these the Narayana numbers. And the number with r j vertical runs of length j equals Krew(x; r) := 1 ( ) b (b 1)! = b r 0, r 1,..., r a r 0!r 1! r a!. Call these the Kreweras numbers.

28 The Prototype: Actuarial Science Theorem (Armstrong 2010, Loehr 2010) The number of Dyck paths with k vertical runs equals Nar(x; k) := 1 ( )( ) a b 1. a k k 1 Call these the Narayana numbers. And the number with r j vertical runs of length j equals Krew(x; r) := 1 ( ) b (b 1)! = b r 0, r 1,..., r a r 0!r 1! r a!. Call these the Kreweras numbers.

29 The Prototype: Actuarial Science Theorem (Armstrong 2010, Loehr 2010) The number of Dyck paths with k vertical runs equals Nar(x; k) := 1 ( )( ) a b 1. a k k 1 Call these the Narayana numbers. And the number with r j vertical runs of length j equals Krew(x; r) := 1 ( ) b (b 1)! = b r 0, r 1,..., r a r 0!r 1! r a!. Call these the Kreweras numbers.

30 Motivation: Core Partitions Definition Let λ n be an integer partition of size n. Say λ is a p-core if it has no cell with hook length p. Say λ is an (a, b)-core if it has no cell with hook length a or b. Example The partition (5, 4, 2, 1, 1) 13 is a (5, 8)-core.

31 Motivation: Core Partitions Theorem (Anderson 2002) The number of (a, b)-cores (of any size) is finite if and only if (a, b) are coprime, in which case they are counted by the Catalan number Cat(a, b) = 1 ( ) a + b. a + b a, b Theorem (Olsson-Stanton 2005, Vandehey 2008) For (a, b) coprime unique largest (a, b)-core of size (a2 1)(b 2 1) 24, which contains all others as subdiagrams. Suggestion Study Young s lattice restricted to (a, b)-cores.

32 Motivation: Core Partitions Theorem (Anderson 2002) The number of (a, b)-cores (of any size) is finite if and only if (a, b) are coprime, in which case they are counted by the Catalan number Cat(a, b) = 1 ( ) a + b. a + b a, b Theorem (Olsson-Stanton 2005, Vandehey 2008) For (a, b) coprime unique largest (a, b)-core of size (a2 1)(b 2 1) 24, which contains all others as subdiagrams. Suggestion Study Young s lattice restricted to (a, b)-cores.

33 Motivation: Core Partitions Theorem (Anderson 2002) The number of (a, b)-cores (of any size) is finite if and only if (a, b) are coprime, in which case they are counted by the Catalan number Cat(a, b) = 1 ( ) a + b. a + b a, b Theorem (Olsson-Stanton 2005, Vandehey 2008) For (a, b) coprime unique largest (a, b)-core of size (a2 1)(b 2 1) 24, which contains all others as subdiagrams. Suggestion Study Young s lattice restricted to (a, b)-cores.

34 Motivation: Core Partitions Theorem (Anderson 2002) The number of (a, b)-cores (of any size) is finite if and only if (a, b) are coprime, in which case they are counted by the Catalan number Cat(a, b) = 1 ( ) a + b. a + b a, b Theorem (Olsson-Stanton 2005, Vandehey 2008) For (a, b) coprime unique largest (a, b)-core of size (a2 1)(b 2 1) 24, which contains all others as subdiagrams. Suggestion Study Young s lattice restricted to (a, b)-cores.

35 Motivation: Core Partitions Example: The poset of (3, 4)-cores.

36 Motivation: Core Partitions Theorem (Ford-Mai-Sze 2009) For a, b coprime, the number of self-conjugate (a, b)-cores is ( a 2 + b 2 a 2, b 2 ). Note: Beautiful bijective proof! (omitted) Observation/Problem ( a 2 + b 2 ) a 2, b 2 = 1 [a + b] q [ ] a + b a, b q q= 1 Conjecture (Armstrong 2011) The average size of an (a, b)-core and the average size of a self-conjugate (a, b)-core are both equal to (a+b+1)(a 1)(b 1) 24.

37 Motivation: Core Partitions Theorem (Ford-Mai-Sze 2009) For a, b coprime, the number of self-conjugate (a, b)-cores is ( a 2 + b 2 a 2, b 2 ). Note: Beautiful bijective proof! (omitted) Observation/Problem ( a 2 + b 2 ) a 2, b 2 = 1 [a + b] q [ ] a + b a, b q q= 1 Conjecture (Armstrong 2011) The average size of an (a, b)-core and the average size of a self-conjugate (a, b)-core are both equal to (a+b+1)(a 1)(b 1) 24.

38 Motivation: Core Partitions Theorem (Ford-Mai-Sze 2009) For a, b coprime, the number of self-conjugate (a, b)-cores is ( a 2 + b 2 a 2, b 2 ). Note: Beautiful bijective proof! (omitted) Observation/Problem ( a 2 + b 2 ) a 2, b 2 = 1 [a + b] q [ ] a + b a, b q q= 1 Conjecture (Armstrong 2011) The average size of an (a, b)-core and the average size of a self-conjugate (a, b)-core are both equal to (a+b+1)(a 1)(b 1) 24.

39 Anderson s Beautiful Proof Proof. Bijection: (a, b)-cores Dyck paths in a b rectangle Example (The (5, 8)-core from earlier.)

40 Anderson s Beautiful Proof Proof. Bijection: (a, b)-cores Dyck paths in a b rectangle Example (Label the rectangle cells by height.)

41 Anderson s Beautiful Proof Proof. Bijection: (a, b)-cores Dyck paths in a b rectangle Example (Label the first column hook lengths.)

42 Anderson s Beautiful Proof Proof. Bijection: (a, b)-cores Dyck paths in a b rectangle Example (Voila!)

43 Anderson s Beautiful Proof Proof. Bijection: (a, b)-cores Dyck paths in a b rectangle Example (Observe: Conjugation is a bit strange.)

44 Rational Parking Functions Definition Label the up-steps by {1, 2,..., a}, increasing up columns. Call this a parking function. Let PF(x) = PF(a, b) denote the set of parking functions. Classical form (z 1, z 2,..., z a ) has label z i in column i. Example: (3, 1, 4, 4, 1)

45 Rational Parking Functions Definition Label the up-steps by {1, 2,..., a}, increasing up columns. Call this a parking function. Let PF(x) = PF(a, b) denote the set of parking functions. Classical form (z 1, z 2,..., z a ) has label z i in column i. Example: (3, 1, 4, 4, 1)

46 Rational Parking Functions Definition Label the up-steps by {1, 2,..., a}, increasing up columns. Call this a parking function. Let PF(x) = PF(a, b) denote the set of parking functions. Classical form (z 1, z 2,..., z a ) has label z i in column i. Example: (3, 1, 4, 4, 1)

47 Rational Parking Functions Definition Label the up-steps by {1, 2,..., a}, increasing up columns. Call this a parking function. Let PF(x) = PF(a, b) denote the set of parking functions. Classical form (z 1, z 2,..., z a ) has label z i in column i. Example: (3, 1, 4, 4, 1)

48 Rational Parking Functions Definition Label the up-steps by {1, 2,..., a}, increasing up columns. Call this a parking function. Let PF(x) = PF(a, b) denote the set of parking functions. Classical form (z 1, z 2,..., z a ) has label z i in column i. Example: (3, 1, 4, 4, 1)

49 Rational Parking Functions Definition The symmetric group S a acts on classical forms. Example: (3, 1, 4, 4, 1) versus (3, 1, 1, 4, 4) By abuse, let PF(x) = PF(a, b) denote this representation of S a. Call it the rational parking space.

50 Rational Parking Functions Definition The symmetric group S a acts on classical forms. Example: (3, 1, 4, 4, 1) versus (3, 1, 1, 4, 4) By abuse, let PF(x) = PF(a, b) denote this representation of S a. Call it the rational parking space.

51 Rational Parking Functions Definition The symmetric group S a acts on classical forms. Example: (3, 1, 4, 4, 1) versus (3, 1, 1, 4, 4) By abuse, let PF(x) = PF(a, b) denote this representation of S a. Call it the rational parking space.

52 Rational Parking Functions Definition The symmetric group S a acts on classical forms. Example: (3, 1, 4, 4, 1) versus (3, 1, 1, 4, 4) By abuse, let PF(x) = PF(a, b) denote this representation of S a. Call it the rational parking space.

53 Rational Parking Functions Definition The symmetric group S a acts on classical forms. Example: (3, 1, 4, 4, 1) versus (3, 1, 1, 4, 4) By abuse, let PF(x) = PF(a, b) denote this representation of S a. Call it the rational parking space.

54 A Few Facts Theorems (with N. Loehr and N. Williams) The dimension of PF(a, b) is b a 1. The complete homogeneous expansion is PF(a, b) = ( ) 1 b h r, b r 0, r 1,..., r a r a where the sum is over r = 0 r0 1 r1 a ra a with i r i = b. That is: PF(a, b) is the coefficient of t a in 1 b H(t)b, where H(t) = h 0 + h 1 t + h 2 t 2 +.

55 A Few Facts Theorems (with N. Loehr and N. Williams) The dimension of PF(a, b) is b a 1. The complete homogeneous expansion is PF(a, b) = ( ) 1 b h r, b r 0, r 1,..., r a r a where the sum is over r = 0 r0 1 r1 a ra a with i r i = b. That is: PF(a, b) is the coefficient of t a in 1 b H(t)b, where H(t) = h 0 + h 1 t + h 2 t 2 +.

56 A Few Facts Theorems (with N. Loehr and N. Williams) The dimension of PF(a, b) is b a 1. The complete homogeneous expansion is PF(a, b) = ( ) 1 b h r, b r 0, r 1,..., r a r a where the sum is over r = 0 r0 1 r1 a ra a with i r i = b. That is: PF(a, b) is the coefficient of t a in 1 b H(t)b, where H(t) = h 0 + h 1 t + h 2 t 2 +.

57 A Few Facts Theorems (with N. Loehr and N. Williams) The dimension of PF(a, b) is b a 1. The complete homogeneous expansion is PF(a, b) = ( ) 1 b h r, b r 0, r 1,..., r a r a where the sum is over r = 0 r0 1 r1 a ra a with i r i = b. That is: PF(a, b) is the coefficient of t a in 1 b H(t)b, where H(t) = h 0 + h 1 t + h 2 t 2 +.

58 A Few Facts Theorems (with N. Loehr and N. Williams) Then using the Cauchy product identity we get... The power sum expansion is PF(a, b) = r a b l(r) 1 p r z r i.e. the # of parking functions fixed by σ S a is b #cycles(σ) 1. The Schur expansion is PF(a, b) = r a 1 b s r(1 b ) s r.

59 A Few Facts Theorems (with N. Loehr and N. Williams) Then using the Cauchy product identity we get... The power sum expansion is PF(a, b) = r a b l(r) 1 p r z r i.e. the # of parking functions fixed by σ S a is b #cycles(σ) 1. The Schur expansion is PF(a, b) = r a 1 b s r(1 b ) s r.

60 A Few Facts Theorems (with N. Loehr and N. Williams) Then using the Cauchy product identity we get... The power sum expansion is PF(a, b) = r a b l(r) 1 p r z r i.e. the # of parking functions fixed by σ S a is b #cycles(σ) 1. The Schur expansion is PF(a, b) = r a 1 b s r(1 b ) s r.

61 A Few Facts Observation/Definition The multiplicities of the hook Schur functions s[k + 1, 1 a k 1 ] in PF(a, b) are given by the Schröder numbers Schrö(a, b; k) := 1 b s [k+1,1 ](1 b ) = 1 ( )( ) a 1 b + k. a k 1 b k a Special Cases: Trivial character: Schrö(a, b; a 1) = Cat(a, b). Smallest k that occurs is k = max{0, a b}, in which case Schrö(a, b; k) = Cat (a, b). Hence Schrö(x; k) interpolates between Cat(x) and Cat (x).

62 A Few Facts Observation/Definition The multiplicities of the hook Schur functions s[k + 1, 1 a k 1 ] in PF(a, b) are given by the Schröder numbers Schrö(a, b; k) := 1 b s [k+1,1 ](1 b ) = 1 ( )( ) a 1 b + k. a k 1 b k a Special Cases: Trivial character: Schrö(a, b; a 1) = Cat(a, b). Smallest k that occurs is k = max{0, a b}, in which case Schrö(a, b; k) = Cat (a, b). Hence Schrö(x; k) interpolates between Cat(x) and Cat (x).

63 A Few Facts Observation/Definition The multiplicities of the hook Schur functions s[k + 1, 1 a k 1 ] in PF(a, b) are given by the Schröder numbers Schrö(a, b; k) := 1 b s [k+1,1 ](1 b ) = 1 ( )( ) a 1 b + k. a k 1 b k a Special Cases: Trivial character: Schrö(a, b; a 1) = Cat(a, b). Smallest k that occurs is k = max{0, a b}, in which case Schrö(a, b; k) = Cat (a, b). Hence Schrö(x; k) interpolates between Cat(x) and Cat (x).

64 A Few Facts Observation/Definition The multiplicities of the hook Schur functions s[k + 1, 1 a k 1 ] in PF(a, b) are given by the Schröder numbers Schrö(a, b; k) := 1 b s [k+1,1 ](1 b ) = 1 ( )( ) a 1 b + k. a k 1 b k a Special Cases: Trivial character: Schrö(a, b; a 1) = Cat(a, b). Smallest k that occurs is k = max{0, a b}, in which case Schrö(a, b; k) = Cat (a, b). Hence Schrö(x; k) interpolates between Cat(x) and Cat (x).

65 A Few Facts Observation/Definition The multiplicities of the hook Schur functions s[k + 1, 1 a k 1 ] in PF(a, b) are given by the Schröder numbers Schrö(a, b; k) := 1 b s [k+1,1 ](1 b ) = 1 ( )( ) a 1 b + k. a k 1 b k a Special Cases: Trivial character: Schrö(a, b; a 1) = Cat(a, b). Smallest k that occurs is k = max{0, a b}, in which case Schrö(a, b; k) = Cat (a, b). Hence Schrö(x; k) interpolates between Cat(x) and Cat (x).

66 A Few Facts Problem Given a, b coprime we have an S a -module PF(a, b) of dimension b a 1 and an S b -module PF(b, a) of dimension a b 1. What is the relationship between PF(a, b) and PF(b, a)? Note that hook multiplicities are the same: Schrö(a, b; k) = Schrö(b, a; k + b a). See Eugene Gorsky, Arc spaces and DAHA representations, 2011.

67 A Few Facts Problem Given a, b coprime we have an S a -module PF(a, b) of dimension b a 1 and an S b -module PF(b, a) of dimension a b 1. What is the relationship between PF(a, b) and PF(b, a)? Note that hook multiplicities are the same: Schrö(a, b; k) = Schrö(b, a; k + b a). See Eugene Gorsky, Arc spaces and DAHA representations, 2011.

68 A Few Facts Problem Given a, b coprime we have an S a -module PF(a, b) of dimension b a 1 and an S b -module PF(b, a) of dimension a b 1. What is the relationship between PF(a, b) and PF(b, a)? Note that hook multiplicities are the same: Schrö(a, b; k) = Schrö(b, a; k + b a). See Eugene Gorsky, Arc spaces and DAHA representations, 2011.

69 A Few Facts Problem Given a, b coprime we have an S a -module PF(a, b) of dimension b a 1 and an S b -module PF(b, a) of dimension a b 1. What is the relationship between PF(a, b) and PF(b, a)? Note that hook multiplicities are the same: Schrö(a, b; k) = Schrö(b, a; k + b a). See Eugene Gorsky, Arc spaces and DAHA representations, 2011.

70 How about q and t? We want a Shuffle Conjecture Define a quasisymmetric function with coefficients in N[q, t] by PF q,t (a, b) := P q qstat(p) t tstat(p) F ides(p). Sum over (a, b)-parking functions P. F is a fundamental (Gessel) quasisymmetric function. natural refinement of Schur functions We require PF 1,1 (a, b) = PF(a, b). Must define qstat, tstat, ides for (a, b)-parking function P.

71 How about q and t? We want a Shuffle Conjecture Define a quasisymmetric function with coefficients in N[q, t] by PF q,t (a, b) := P q qstat(p) t tstat(p) F ides(p). Sum over (a, b)-parking functions P. F is a fundamental (Gessel) quasisymmetric function. natural refinement of Schur functions We require PF 1,1 (a, b) = PF(a, b). Must define qstat, tstat, ides for (a, b)-parking function P.

72 How about q and t? We want a Shuffle Conjecture Define a quasisymmetric function with coefficients in N[q, t] by PF q,t (a, b) := P q qstat(p) t tstat(p) F ides(p). Sum over (a, b)-parking functions P. F is a fundamental (Gessel) quasisymmetric function. natural refinement of Schur functions We require PF 1,1 (a, b) = PF(a, b). Must define qstat, tstat, ides for (a, b)-parking function P.

73 How about q and t? We want a Shuffle Conjecture Define a quasisymmetric function with coefficients in N[q, t] by PF q,t (a, b) := P q qstat(p) t tstat(p) F ides(p). Sum over (a, b)-parking functions P. F is a fundamental (Gessel) quasisymmetric function. natural refinement of Schur functions We require PF 1,1 (a, b) = PF(a, b). Must define qstat, tstat, ides for (a, b)-parking function P.

74 How about q and t? We want a Shuffle Conjecture Define a quasisymmetric function with coefficients in N[q, t] by PF q,t (a, b) := P q qstat(p) t tstat(p) F ides(p). Sum over (a, b)-parking functions P. F is a fundamental (Gessel) quasisymmetric function. natural refinement of Schur functions We require PF 1,1 (a, b) = PF(a, b). Must define qstat, tstat, ides for (a, b)-parking function P.

75 How about q and t? We want a Shuffle Conjecture Define a quasisymmetric function with coefficients in N[q, t] by PF q,t (a, b) := P q qstat(p) t tstat(p) F ides(p). Sum over (a, b)-parking functions P. F is a fundamental (Gessel) quasisymmetric function. natural refinement of Schur functions We require PF 1,1 (a, b) = PF(a, b). Must define qstat, tstat, ides for (a, b)-parking function P.

76 qstat is easy Definition Let qstat := area := # boxes between the path and diagonal. Note: Maximum value of area is (a 1)(b 1)/2. (Frobenius) see Beck and Robins, Chapter 1 Example This (5, 8)-parking function has area = 6.

77 ides is reasonable Definition Read labels by increasing height to get permutation σ S a. ides := the descent set of σ 1. Example Remember the height? ides = {1, 4}

78 ides is reasonable Definition Read labels by increasing height to get permutation σ S a. ides := the descent set of σ 1. Example Look at the heights of the vertical step boxes. ides = {1, 4}

79 ides is reasonable Definition Read labels by increasing height to get permutation σ S a. ides := the descent set of σ 1. Example Remember the labels we had before. ides = {1, 4}

80 ides is reasonable Definition Read labels by increasing height to get permutation σ S a. ides := the descent set of σ 1. Example Read them by increasing height to get σ = S 5. ides = {1, 4}

81 tstat is hard (as usual) Definition Blow up the (a, b)-parking function. Compute dinv of the blowup. Example Recall our favorite the (5, 8)-parking function.

82 tstat is hard (as usual) Definition Blow up the (a, b)-parking function. Compute dinv of the blowup. Example Since = 1 we blow up by 2 horiz. and 3 vert...

83 tstat is hard (as usual) Example To get this!

84 tstat is hard (as usual) Example To get this! Now compute dinv = 7.

85 tstat is hard (as usual) Example (There s a scaling factor depending on the path, so tstat = 3.)

86 All Together Example So our favorite (5, 8)-parking function contributes q 6 t 3 F {1,4}. Proof of Concept: The coefficient of s[2, 2, 1] in PF q,t (5, 8) is

87 All Together Example So our favorite (5, 8)-parking function contributes q 6 t 3 F {1,4}. Proof of Concept: The coefficient of s[2, 2, 1] in PF q,t (5, 8) is

88 A Few Facts Facts PF 1,1 (a, b) = PF(a, b). PF q,t (a, b) is symmetric and Schur-positive with coeffs N[q, t]. via LLT polynomials (HHLRU Lemma 6.4.1) Experimentally: PF q,t (a, b) = PF t,q (a, b). this will be impossible to prove (see Loehr s Maxim) Definition: The coefficient of the hook s[k + 1, 1 a k 1 ] is the q, t-schröder number Schrö q,t (a, b; k). Experimentally: Specialization t = 1/q gives [ ] [ ] 1 a 1 b + k Schrö q, 1 (a, b; k) = q [b] q k a q q (centered)

89 A Few Facts Facts PF 1,1 (a, b) = PF(a, b). PF q,t (a, b) is symmetric and Schur-positive with coeffs N[q, t]. via LLT polynomials (HHLRU Lemma 6.4.1) Experimentally: PF q,t (a, b) = PF t,q (a, b). this will be impossible to prove (see Loehr s Maxim) Definition: The coefficient of the hook s[k + 1, 1 a k 1 ] is the q, t-schröder number Schrö q,t (a, b; k). Experimentally: Specialization t = 1/q gives [ ] [ ] 1 a 1 b + k Schrö q, 1 (a, b; k) = q [b] q k a q q (centered)

90 A Few Facts Facts PF 1,1 (a, b) = PF(a, b). PF q,t (a, b) is symmetric and Schur-positive with coeffs N[q, t]. via LLT polynomials (HHLRU Lemma 6.4.1) Experimentally: PF q,t (a, b) = PF t,q (a, b). this will be impossible to prove (see Loehr s Maxim) Definition: The coefficient of the hook s[k + 1, 1 a k 1 ] is the q, t-schröder number Schrö q,t (a, b; k). Experimentally: Specialization t = 1/q gives [ ] [ ] 1 a 1 b + k Schrö q, 1 (a, b; k) = q [b] q k a q q (centered)

91 A Few Facts Facts PF 1,1 (a, b) = PF(a, b). PF q,t (a, b) is symmetric and Schur-positive with coeffs N[q, t]. via LLT polynomials (HHLRU Lemma 6.4.1) Experimentally: PF q,t (a, b) = PF t,q (a, b). this will be impossible to prove (see Loehr s Maxim) Definition: The coefficient of the hook s[k + 1, 1 a k 1 ] is the q, t-schröder number Schrö q,t (a, b; k). Experimentally: Specialization t = 1/q gives [ ] [ ] 1 a 1 b + k Schrö q, 1 (a, b; k) = q [b] q k a q q (centered)

92 A Few Facts Facts PF 1,1 (a, b) = PF(a, b). PF q,t (a, b) is symmetric and Schur-positive with coeffs N[q, t]. via LLT polynomials (HHLRU Lemma 6.4.1) Experimentally: PF q,t (a, b) = PF t,q (a, b). this will be impossible to prove (see Loehr s Maxim) Definition: The coefficient of the hook s[k + 1, 1 a k 1 ] is the q, t-schröder number Schrö q,t (a, b; k). Experimentally: Specialization t = 1/q gives [ ] [ ] 1 a 1 b + k Schrö q, 1 (a, b; k) = q [b] q k a q q (centered)

93 A Few Facts Facts PF 1,1 (a, b) = PF(a, b). PF q,t (a, b) is symmetric and Schur-positive with coeffs N[q, t]. via LLT polynomials (HHLRU Lemma 6.4.1) Experimentally: PF q,t (a, b) = PF t,q (a, b). this will be impossible to prove (see Loehr s Maxim) Definition: The coefficient of the hook s[k + 1, 1 a k 1 ] is the q, t-schröder number Schrö q,t (a, b; k). Experimentally: Specialization t = 1/q gives [ ] [ ] 1 a 1 b + k Schrö q, 1 (a, b; k) = q [b] q k a q q (centered)

94 Motivation: Lie Theory The James-Kerber Bijection between a-cores and the root lattice of the Weyl group S a

95 Here s The Picture These are the root and weight lattices Q Λ of S a.

96 Here s The Picture Here is a fundamental parallelepiped for Λ/bΛ.

97 Here s The Picture It contains b a 1 elements (these are the parking functions ).

98 Here s The Picture But they look better as a simplex...

99 Here s The Picture...which is congruent to a nicer simplex.

100 Here s The Picture ( There are Cat(a, b) = 1 a+b ) a+b a,b elements of the root lattice inside.

101 Here s The Picture These are the (a, b)-dyck paths (via Anderson, James-Kerber).

102 Other Weyl Groups? Definition Consider a Weyl group W with Coxeter number h and let p N be coprime to h. We define the Catalan number Cat q (W, p) := j [p + m j ] q [1 + m j ] q where e 2πim j /h are the eigenvalues of a Coxeter element. Observation [ ] 1 a + b Cat q (S a, b) = [a + b] q a, b q

103 Thank You