A BIJECTION FROM m-DYCK PATHS OF AREA=2 TO m-DYCK

PATHS OF DINV=2

ALLISON CRUIKSHANK AND KYUNGYONG LEE†

1. Introduction
Let m and n be positive integers. Given an mn × n grid, an m-Dyck path is a lattice
1
path from (0, 0) to (mn, n) going north or east without going below a diagonal of y = m x.
(m)
Let Dn be the set of all m-Dyck paths in the grid. For an m-Dyck path, we use the
area and dinv statistics to describe the layout of the path. It is known that there exists
a symmetry between the two statistics. It is desired to discover an explicit bijection on
m-Dyck paths interchanging the two statistics (for references, see [10]).
An m-Dyck word is a sequence r = (r0 , r1 , ..., rn−1 ) such that ri ∈ Z≥0 , r0 = 0, and
ri ≤ ri−1 + m for 1 ≤ i < n. For each m-Dyck path D, we get the corresponding m-Dyck
word by counting the numbe of unit cells in the (i + 1)-th row (from the bottom) between
D and the diagonal.
Below is an example of a 2-Dyck path and its corresponding 2-Dyck word r = (0, 1, 3, 5, 6).

6
5
3
1
0

Definition 1.1. For an m-Dyck word r = (r0 , r1 , ..., rn−1 ), define

X
area(r) = ri .
0≤i<n

Definition 1.2. For an m-Dyck word r = (r0 , r1 , ..., rn−1 ), define

X
dinv(r) = sc(ri − rj ),
0≤i<j<n

†
This work was partially supported by NSF grant DMS 1800207, the University of Nebraska–Lincoln,
and Korea Institute for Advanced Study.
1
2 ALLISON CRUIKSHANK AND KYUNGYONG LEE†

where

m + 1 − p,
 if 1 ≤ p ≤ m;
(1) sc(p) = m + p, if − m ≤ p ≤ 0;

0, otherwise.


When we consider the corresponding m-Dyck path, the above definitions can be trans-
lated as follows. Let D be an m-Dyck path in the grid. The area of D is the number of
1
unit squares between D and the diagonal line of y = m x. The arm of a box w above D in
the grid is the number of boxes to the right of w but above D. Similarly, the leg of w is
the number of boxes below w but above D. A good box is a box that meets the needs as
so:
arm(w) − mleg(w) ∈ {0, 1, .., m}.
The dinv of D is the number of good boxes. A bad box is a non-good box above D.
Haiman and Loehr [8] discovered the higher combinatorial q, t-Catalan polynomial that
is defined by X
Cn(m) (q, t) = q dinv(D) tarea(D) .
(m)
D∈Dn
For m = 1, it was proven [3] that
Cn(1) (q, t) = Cn(1) (t, q).
The joint q, t-symmetry for m ≥ 2 had been a conjecture until recently Mellit [9] proved
the following:
Theorem 1.3.
Cn(m) (q, t) = Cn(m) (t, q)
for all m, n ≥ 1.
There have been some approaches to understanding this symmetry combinatorially [1,
2, 4, 5, 6, 7]. We state our main result in this paper.
(m) (m)
Theorem 1.4. Let A2 ⊂ Dn be the set of m-Dyck words with area of 2, and D2 ⊂ Dn
the set of m-Dyck words with dinv of 2. There is an explicit combinatorial bijection between
A2 and D2 such that
X X
q dinv(D) tarea(D) = tdinv(D) q area(D) .
D∈A2 D∈D2

2. Exploring the symmetry

In Definitions 2.1 and 2.2 below, we describe all the m-Dyck words of area= 2.
Definition 2.1. For 1 ≤ i ≤ n − 1, let si = (r0 , ..., rn−1 ), where
(
2, if j = n − i;
rj =
0, otherwise.
 A BIJECTION FROM m-DYCK PATHS OF AREA=2 TO m-DYCK PATHS OF DINV=2 3

For n ≤ i ≤ 2n − 3, let si = (r0 , ..., rn−1 ), where

(
1, if j ∈ {1, 2n − i − 1};
rj =
0, otherwise.
Definition 2.2. For 2 ≤ i < j ≤ n − 1, let ti,j = (r0 , ..., rn−1 ), where
(
1, if k ∈ {i, j};
rk =
0, otherwise.
2.1. The case of m ≥ 2.
Lemma 2.3. For g ∈ {1, ..., n − 1}, we have
 
2 n
dinv(sn−g ) = n − 2n − g + 1 + (m − 2).
2
Proof. Using Definition 1.2, cases can be developed in order to get all the possible values
of i and j.
Case 1. Suppose {i, j} ∩ {g} = ∅. Then ri − rj = 0, so sc(0) = (m − 2) + 2.
Case 2. Suppose i 6= g and j = g. Then ri − rj = −2, so sc(−2) = m − 2.
Case 3. Suppose i = g and j 6= g. Then ri − rj = 2, so sc(2) = (m − 2) + 1
Using the given definition, the equation for dinv is as follows,
X X X
dinv(sn−g ) = 2+ 1+ (m − 2)
0≤i<j<n−1 g<j≤n−1 0≤i<j≤n−1
n−1 n



This can be simplified into 2 · 2 + (n − g − 1) +
(m − 2) to give
2
 
n
dinv(sn−g ) = n2 − 2n − g + 1 + (m − 2).
2
The case of −n + 3 ≤ g ≤ 0 can be done by a similar calculation. 
Lemma 2.4. (i) For g ∈ {−n + 3, ..., 0}, we have
 
2 n
dinv(sn−g ) = n − 2n − g + 1 + (m − 2).
2
(ii) For 2 ≤ g1 < g2 ≤ n − 1, we have
 
2 n
dinv(tg1 ,g2 ) = n − n − g1 − g2 + 1 + (m − 2).
2
Proof. We show (ii) only, since (i) can be proved by a similar calculation. Using Defi-
nition 1.2, cases can be developed similar to the proof of Lemma 2.3 in order to get all
possible values of i and j.
Case 1. Suppose {i, j} ∩ {g1, g2} = ∅. Then ri − rj = 0, so sc(0) = (m − 2) + 2.
Case 2. Suppose i = g1 and j 6= g2 . Then ri − rj = 1, so sc(1) = (m − 2) + 2.
Case 3. Suppose i 6= g1 and j = g2 . Then ri − rj = −1, so sc(−1) = (m − 2) + 1.
Case 4. Suppose i = g1 and j = g2 . Then ri − rj = 0, so sc(0) = (m − 2) + 2.
4 ALLISON CRUIKSHANK AND KYUNGYONG LEE†

Case 5. Suppose i 6= g1 and j = g1 . Then ri − rj = −1, so sc(−1) = (m − 2) + 1.

Case 6. Suppose i = g2 and j 6= g2 . Then ri − rj = 1, so sc(1) = (m − 2) + 2.
Using the given definition, the equation for the dinv is as follows,
X X X X X X
dinv(tg1 ,g2 ) = 2+ 2+ 1+2+ 1+ 2+ (m−2).
0≤i<j<n−1 g1 < j ≤ n − 1 0 ≤ i < g2 0≤i≤g1 g2 <j≤n−1 0≤i<j≤n−1
j 6= g2 i 6= g1

This can
be simplified into (n − 2)(n − 3) + (2n − 2g1 − 4) + (g2 − 1) + 2 + g1 + 2(n − 1 −

g2 ) + n2 (m − 2) to give
 
2 n
dinv(tg1 ,g2 ) = n − n − g1 − g2 + 1 + (m − 2).
2

Definition 2.5. For 1 ≤ a ≤ b ≤ n − 1, let r(a,b) = (r0 , ..., rn−1 ), where

mi,
 if i < a;
(2) ri = mi − 1, if a ≤ i < b;

mi − 2, if b ≤ i.


Lemma 2.6. The set of {r(a,b) : 1 ≤ a ≤ b ≤ n − 1} is equal to the set of all m-Dyck
words of dinv = 2.
Proof. A corner is a horizontal edge followed by a vertical line. The basis of the proof is that
boxes on the corners of Dyck paths are always good boxes as arm − leg = 0 ∈ {0, 1, ...m}.
Also, as this is true, it is enough to consider when there are at most 2 corners.
Case 1. The first case is when there are two corners on in the Dyck path. This causes
the good boxes to be on 2 different rows. This is illustrated below where the good boxes
are denoted with G.
G

The south edge of a good box should be preceded by a vertical edge. If not, that would
allow the box next to it to be good as well as arm − leg = 1 ∈ {0, 1, ..., m}.
Case 2. The second case is when there is only one corner in the Dyck path. This causes
the good boxes to be on the same row. This is illustrated below where the good boxes are
denoted with a G.

GG
 A BIJECTION FROM m-DYCK PATHS OF AREA=2 TO m-DYCK PATHS OF DINV=2 5

The south edge of the first good box should be preceded by a vertical edge. If not, that
would allow the box to the left on the same row to be good as well as arm − leg = 2 ∈
{0, 1, .., m}.
It remains to prove the following lemma. 
Lemma 2.7.
dinv(r(a,b) ) = 2.
Proof. Using Definitions 1.2 and 2.5, we can follow the same procedures in the proofs of
Lemmas 2.3 and 2.4 and make cases. Suppose thatP a 6= b.
Case 1: i, j < a. Then ri − rj = mi − mj, so sc(mi − mj) = 0, since mi − mj ∈
{−m, −2m, ...} P
Case 2: a ≤ i < b; a ≤ j < b. Then ri − rj = mi − mj, so sc(mi − mj) = 0, since
mi − mj ∈ {−m, −2m, ...}.
Case 3: a ≤ i < b; b ≤ j, Then ri − rj = mi − mj + 1, so sc(mi − mj + 1) = 1 if i = b − 1
and j = b, and sc(mi − mj + 1) = 0 otherwise. P
Case 4: b ≤ i; b ≤ j. Then ri − rj = mi − mj, so sc(mi − mj) = 0, since mi − mj ∈
{−m, −2m, ...}.
Case 5: i < a; a ≤ j < b. Then ri − rj = mi − mj + 1, so sc(mi − mj + 1) = 1 if i = a − 1
and j = a, and sc(mi − mj + 1) = 0 otherwise. P
Case 6: i < a; b ≤ j. Then ri − rj = mi − mj + 2, so sc(mi − mj + 2) = 0, since
mi − mj + 2 ∈ {0, −m, −2m, ..}.
Using the given definition of dinv, the sum of the last terms of cases are added together.
Only cases 3 and 5 have an impact on the equation, which gives dinv(r(a,b) ) = 2.
If a = b, then sc(ri − rj ) is nonzero only when i = a − 1 and j = a. This also gives
dinv(r(a,b) ) = 2. 
Lemma 2.8.  
(a,b) 2 n
area(r ) = n − 3n + a + b + (m − 2).
2
Proof. Using Definitions 1.1 and 2.5, an equation can be produced of the form,
X X X
area(r(a,b) ) = mi + (mi − 1) + (mi − 2).
i<a a≤i<b b≤i
n


This can be simplified into 2 m − (b − a) − (2n − 2b) to give the desired equation. 
Theorem 2.9. Let m ≥ 2. Let A2 be the set of m-Dyck words with area of 2, and D2 be
the set of m-Dyck words with dinv of 2. Define
f : A2 → D 2
by
r(i,i) ,

 s2i−1 7→ for i ≥ 1;
(3) s 7→ r (i,i+1) , for i ≥ 1;
 2i (n−j,n−i+1)
ti,j 7→ r , for 2 ≤ i < j ≤ n − 1.
6 ALLISON CRUIKSHANK AND KYUNGYONG LEE†

Then f is a bijection, and

X X X
q area(r) tdinv(r) = q dinv(f (r)) tarea(f (r)) = q dinv(r) tarea(r) .
r∈A2 r∈A2 r∈D2

Proof. It is easy to see that f is a bijection. By Lemmas 2.3, 2.4, 2.7, 2.8, and Definition 2.1,
we have
 
2 n
area(s2i−1 ) = 2 and dinv(s2i−1 ) = n − 2n − (n − 2i + 1) + 1 + (m − 2);
2

 
(i,i) 2 n
area(r ) = n − 3n + 2i + (m − 2) and dinv(r(i,i) ) = 2.
2

 
2 n
area(s2i ) = 2 and dinv(s2i ) = n − 2n − (n − 2i) + 1 + (m − 2)
2

 
(i,i+1) 2 n
area(r ) = n − 3n + 2i + 1 + (m − 2) and dinv(r(i,i+1) ) = 2.
2

 
2 n
area(ti,j ) = 2 and dinv(ti,j ) = n − n − j − i + 1 + (m − 2);
2

 
(n−j,n−i+1) 2 n
area(r ) = n − 3n + n − j + n − i + 1 + (m − 2) and dinv(r(n−j,n−i+1) ) = 2.
2

Example 2.10. Shown are all of the Dyck paths for 10 × 5 grid where either area or dinv
is equal to 2. On the left are the Dyck paths that have an area of 2 and on the right are
the Dyck paths that have a dinv of 2. The symbol, A, is put in boxes accounted for area
and the symbol, G, is put in boxes accounted for dinv. Each Dyck path is labeled according
to Definitions 2.1, 2.2, and 2.5. Below each label are the values for dinv or area for each
Dyck path.
A BIJECTION FROM m-DYCK PATHS OF AREA=2 TO m-DYCK PATHS OF DINV=2 7

AA

s1 r(1,1)
dinv=12 GG area=12

AA
s2 G r(1,2)
dinv=13 G area=13

AA s3 GG r(2,2)
dinv=14 area=14

G
s4 G r(2,3)
AA dinv=15 area=15

A
GG
s5 r(3,3)
A dinv=16 area=16

G
A G
s6 r(3,4)
A dinv=17 area=17

GG

A s7 r(4,4)
A dinv=18 area=18

A
A G
t3,4 r(1,3)
dinv=14 G area=14

A G

A t2,4 r(1,4)
dinv=15 G area=15

G
A
A t2,3 G r(2,4)
dinv=16 area=16
8 ALLISON CRUIKSHANK AND KYUNGYONG LEE†

2.2. The case of m = 1.

Definition 2.11. For a and b with 1 ≤ a < b − 1 ≤ n − 2, define r(a,b) as in (2). Let
{(a, b) ∈ Z2 : n+1 n+1

Fn = 2 ≤ a = b ≤ n − 1 or 2 ≤ a = b − 1 ≤ n − 2}, if n is odd;
2 n+2 n
{(a, b) ∈ Z : 2 ≤ a = b ≤ n − 1 or 2 ≤ a = b − 1 ≤ n − 2}, if n is even.

For (a, b) ∈ Fn , define r(a,b) = (0, 1, 2, ..., n − 3, n − 2, a + b − n − 1).

This new definition allows us to transform the Dyck word from the case of m ≥ 2 to
the case of m = 1 in that it maintains the number of bad boxes, but it also puts all of the
boxes above the Dyck path in the m ≥ 2 case into the top row above the Dyck path in the
m = 1 case. An example can be seen below that illustrates this idea.
Example 2.12. Below are the 3 cases where dinv is 2 in the m = 1 case for the 5 × 5 grid.
Next to them are the corresponding Dyck paths where dinv is 2 in the m = 2 case for the
10 × 5 grid.

GG
GG

r(3,3) r(3,3)
GG G
G

r(3,4) r(3,4)
GG GG

r(4,4) r(4,4)

Theorem 2.13. Let A2 be the set of all 1-Dyck words with area of 2, D2 be the set of all
1-Dyck words with dinv of 2. Then
A2 = {sn , sn+1 , ..., s2n−3 } ∪ {ti,j : 2 ≤ i < j ≤ n − 1}
and
n+1 n+1
{r(a,b) :

2 ≤ a = b ≤ n − 1 or 2 ≤ a = b − 1 ≤ n − 2 or a < b − 1}, if n is odd;
D2 = n+2 n
{r(a,b) : 2 ≤ a = b ≤ n − 1 or 2 ≤ a = b − 1 ≤ n − 2 or a < b − 1}, if n is even.
Define
f : A2 → D 2
 A BIJECTION FROM m-DYCK PATHS OF AREA=2 TO m-DYCK PATHS OF DINV=2 9

by
r(i,i) ;

 s2i−1 →
7
(4) s 7 → r (i,i+1) ;
 2i (n−j,n−i+1)
ti,j → 7 r .
Then f is a bijection, and
X X X
q area(r) tdinv(r) = q dinv(f (r)) tarea(f (r)) = q dinv(r) tarea(r) .
r∈A2 r∈A2 r∈D2

Proof. Every 1-Dyck word of area = 2 is of the form (0...0, 1, 0...0...0, 1, 0...0) because
(0...0...0, 2, 0...0...0) violates the definition of a 1-Dyck word. Hence,
A2 = {sn , sn+1 , ..., s2n−3 } ∪ {ti,j : 2 ≤ i < j ≤ n − 1}
by Definitions 2.1 and 2.2. Mimicking an argument as in the proof of Lemma 2.6, we get
 (a,b) n+1
{r : 2 ≤ a = b ≤ n − 1 or n+1
2 ≤ a = b − 1 ≤ n − 2 or a < b − 1}, if n is odd;
D2 = (a,b) n+2 n
{r : 2 ≤ a = b ≤ n − 1 or 2 ≤ a = b − 1 ≤ n − 2 or a < b − 1}, if n is even.
For r ∈ A2 ∪ D2 , let r0 be the 2-Dyck word with the same label as r. Let k(r) (resp.
k(r0 )) be the number of bad boxes for the 1-Dyck path (resp. 2-Dyck path) corresponding
to r (resp. r0 ). Then for r ∈ A2 ,
   
n n
dinv(r) = − area(r) − k(r) = − 2 − k(r0 )
2 2
     
n n 0 0
= −2− 2 − area(r ) − dinv(r )
2 2
     
T heorem 2.9 n n 0 0
= −2− 2 − dinv(f (r )) − area(f (r ))
2 2
 
n
= − 2 − k(f (r0 ))
2
 
n
= − dinv(f (r)) − k(f (r)) = area(f (r)).
2

Example 2.14. Below are the 6 cases where either dinv or area equals 2 in the m = 1
case for the 5 × 5 grid. Next to them are the corresponding Dyck paths in the m = 2 case
for the 10 × 5 grid. As one can see, the m = 1 case only applies to the last 6 cases of both
when the area is 2 and the dinv is 2 for 10 × 5 grid. The Dyck paths for 10 × 5 grid are on
the right and the Dyck paths for 5 × 5 grid are on the left. A symbol, G, denotes a good box
and a symbol, A, denotes a box accounted for area. The Dyck paths are labeled according
to Definitions 2.1, 2.2, 2.5, and 2.11. Below the label of each is the value for dinv or area
is shown.
10 ALLISON CRUIKSHANK AND KYUNGYONG LEE†

A s5 GG r(3,3) A s5 r(3,3)
GG
dinv=6 area=6 dinv=16 area=16
A A

s6 GG r(3,4) s6 G r(3,4)
A A G
dinv=7 area=7 dinv=17 area=17
A A

s7 GG r(4,4) s7 GG r(4,4)
A dinv=8 area=8 A dinv=18 area=18
A A
A t3,4 r(1,3) A t3,4 r(1,3)
A G A G
dinv=4 area=4 dinv=14 area=14
G G
A t2,4 G r(1,4) A t2,4 G r(1,4)
A dinv=5 area=5 A dinv=15 area=15
G G

t2,3 G r(2,4) t2,3 G r(2,4)

A A
A dinv=6 G area=6 A dinv=16 G area=16

References
[1] E. Gorsky, G. Hawkes, A. Schilling, and J. Rainbolt, Generalized q, t-Catalan numbers,
arXiv:1905.10973.
[2] E. Gorsky and M. Mazin, Compactified Jacobians and q, t-Catalan numbers, II, J. Algebraic Combin.
39 (2014), 153–186.
[3] M. Haiman, Vanishing theorems and character formulas for the Hilbert scheme of points in the plane,
Invent. Math. 149 (2002), 371–407.
[4] R. Kaliszewski and D. Karmakar, A Rational Catalan Formula for (m, 3)-Hikita Polynomials, Electronic
J. Comb. 25(1) (2018), #P1.21.
[5] R. Kaliszewski and H. Li, The (m, n)-rational q, t-Catalan polynomials for m = 3 and their q, t-
symmetry, FPSAC 2015, Daejeon, South Korea. 769–780.
[6] K. Lee, L. Li, and N. A. Loehr, Combinatorics of certain higher q, t-Catalan polynomials: chains, joint
symmetry, and the Garsia–Haiman formula, J. Algebraic Combin. 39 (2014), 749–781.
[7] K. Lee, L. Li, and N. A. Loehr, A combinatorial Approach to the Symmetry of q, t-Catalan Numbers,
SIAM J. Discrete Math (SIDMA) 32 (2018), 191–232.
[8] N. A. Loehr, Conjectured Statistics for the Higher q, t-Catalan Sequences, Electronic J. Comb. 12
(2005), R9.
[9] A. Mellit, Toric braids and (m, n)-parking functions, arXiv:1604.07456.
[10] J. Haglund, The q, t-Catalan Numbers and the Space of Diagonal Harmonics: with an Appendix on
the Combinatorics of Macdonald Polynomials, American Mathematical Society (2008).

University of Nebraska–Lincoln, Lincoln, NE 68588, U.S.A.

Department of Mathematics, University of Alabama, Tuscaloosa, AL 35487, U.S.A., and

Korea Institute for Advanced Study, Seoul 02455, Republic of Korea
Email address: [email protected]; [email protected]