A q-Analogue of Generalized

Translated Whitney Numbers: Cigler’s

Approach
Roberto B. Corcino
Research Institute for Computational Mathematics and Physics
Cebu Normal University
Cebu City, Philippines 6000
e-mail: rcorcino@yahoo.com

Jezer C. Fernandez
Department of Mathematics
Mindanao State University
Marawi City, Philippines 9700
e-mail: jezerfernandez@gmail.com

Mary Ann Ritzell P. Vega

Department of Mathematics and Statistics
College of Science and Mathematics
Mindanao State University-Iligan Institute of Technology
Iligan City 9200
e-mail: maryannritzell.vega@g.msuiit.edu.ph

October 14, 2022

Abstract

By means of a certain combinatorial interpretation in terms of set

partition, a q-analogue of generalized translated Whitney numbers of
the second kind is defined in this paper. Some properties such as the
recurrence relation, explicit formula and certain symmetric formula
are obtained. Moreover, a q-analogue of generalized translated Whit-
ney numbers of the first kind is introduced to obtain a q-analogue of
the orthogonality and inverse relations of the two kinds of generalized
translated Whitney numbers.
Keywords: Stirling numbers; r-Stirling numbers; translated Whitney
numbers; q-binomial coefficients; q-factorial

1
1 Introduction
Several generalizations of Stirling numbers have already appeared in the
literature starting from the twentieth century. Hsu and Shiue [13] have
listed almost all the generalizations of Stirling numbers which belong to
their unified generalization. One of these is the r-Stirling numbers of the
first and second kind which are defined, respectively, by Broder [3] as follows
 
n
:= number of permutations of the set {1, 2, · · · , n} into k nonempty
k r
disjoint cycles, such that the numbers 1, 2, · · · , r are in distinct
cycles.
 
n
:= number of partitions of the set {1, 2, · · · , n} into k nonempty
k r
disjoint classes (or blocks), such that the numbers 1, 2, · · · , r are in
distinct classes (or blocks).

Detailed discussions on r-Stirling numbers and some related works can be

found in the papers [3, 19, 20, 21]. Just recently, the r-Stirling numbers of
the second kind have been generalized further in [22, Mihoubi and Maamra]
by replacing the condition

the numbers 1, 2, · · · , r are in distinct classes (or blocks)

with the condition

for given subsets R1 , . . . , Rr of {1, 2, . . . , n} where |Ri | = ri and

Ri ∩ Rj = ∅, for all i, j = 1, . . . , r i 6= j, the elements of each
subsets Ri , i = 1, . . . , r are in distinct classes (or blocks).

This generalization of r-Stirling numbers of the second kind is called the

(r1 , . . . , rr )-Stirling numbers of the second kind.
Moreover, Benoumhani [2] introduced another interesting generalization
of Stirling numbers which is the pair of Whitney numbers, the Whitney
numbers of the first and second kind. They are defined, respectively, by
means of the following recurrence relations

wα (n, k) = wα (n − 1, k − 1) + (α(n − 1) + 1)wα (n − 1, k)

Wα (n, k) = Wα (n − 1, k − 1) + (αk + 1)Wα (n − 1, k).

These numbers are also known as the Whitney numbers of Dowling lat-
tices, which have their origin in a class of lattices, based on a finite group
G of order m, introduced by Dowling [11]. Numerous properties on both

2
kinds of Whitney numbers have been established [2] including the horizontal
generating functions
n
X
(x + 1|α)n = (−1)n+k wα (n, k)xk
k=0
n
X
xn = Wα (n, k)(x − 1| − α)k .
k=0

With the introduction of Whitney numbers coupled by the motivation

from the work of Broder [3], Belbachir and Bousbaa [1] defined combinatori-
ally certain generalization of Whitney numbers by introducing new parame-
ter α. These numbers are called the generalized translated Whitney numbers
 (α)  (α)
n n
of the first and second kinds and are denoted by and . For
k r k r
brevity, we use GTWN to abbreviate the phrase “generalized translated
Whitney numbers”. More precisely, these numbers are defined respectively
as follows
 (α)
n
:= number of permutations of n elements with k cycles such that the
k r
first r elements are in distinct cycles and the elements of each cycle
can mutate in α ways, except the dominant one;
 (α)
n
:= number of partitions of n elements with k parts such that the
k r
first r elements are in distinct parts and the elements of each part
can mutate in α ways, except the dominant one.

The concept of mutation and domination of the elements in the defini-

tion above were introduced by Belbachir and Bousbaa in [1]. The occurence
of more than one element in a certain part displays the situation of domina-
tion, one element dominates the other, and the dominated one can mutate
in α ways to another color, for example. Mutation of elements occurs when
it is allowed to change its color, for example, thereby contributing one per-
mutation or partition.

The GTWN reduce to both kinds of r-Stirling numbers when α = 1.

Furthermore, several properties of these numbers have been established in
[1] including the recurrence relations

n + r (α) n + r − 1 (α) n + r − 1 (α)

     
= + α(n + r − 1) (1)
k+r r k+r−1 r k+r r
n + r (α) n + r − 1 (α) n + r − 1 (α)
     
= + α(k + r) , (2)
k+r r k+r−1 r k+r−1 r

3
the horizontal generating functions
n 
n + r (α)
X 
(x|α)n = (x − αr)k (3)
k+r r
k=0
n 
n + r (α)
X 
n
(x + αr) = (x| − α)k , (4)
k+r r
k=0

and certain explicit formulas

 (α)     
n X r r+α r + (n − 1)α
= ... (5)
k r i0 i1 in−1
i0 +i1 +...in−1 =n−k
ij ∈{0,1}
 (α) n−1
Y 
n X r + α(j − i0 − i1 − . . . ij−1 )
= . (6)
k r ij
i0 +i1 +...in−1 =n−k j=0
ij ∈{0,1}

This paper aims to develop further extension of GTWN using the concept
of q-analogue.

2 A q-Analogue of GTWN of the Second Kind

In [7], the classical Stirling numbers of the second kind S(n, k) were defined
as the cardinality of set B of partitions of {0, 1, 2, · · · , n−1} into k nonempty
disjoint subsets. With this definition, Cigler [5] defined a q-analogue of
S(n, k) to be the following sum
P
i
X
w(π), w(π) = q i∈B0
π∈B

where B0 is a subset in partition π which contains 0.

At this point, we can now restate the above definition of GTWN of the
 (α)
n
second kind as follows:
k r
 (α)
n
:= number of partitions π of {0, 1, ..., n − 1} into k nonempty subsets
k r
(α) (α) (α)
B0 , B1 , ..., Bk−1 such that the r first elements are in distinct
subsets and the elements of each subset can mutate in α ways,
except the dominant one.

In this section, a q-analogue of GTWN of the second kind will be defined

(α)
parallel to the work of Cigler. First, we choose B0 so that the number
(α)
0 ∈ B0 . Then, let us define the following notations:

4
 • the weight of partition π
 
(α)  
s B0 (α)
X
w(π) = q , s B0 = i.
(α)
i∈B0

• the weight of each set of partitions A

X
w(A) := w(π)
π∈A

(α)
• An,k,r := the set of all partitions of {0, 1, ..., n − 1} into k nonempty
parts such that the first r elements are in distinct partitions and the
elements of each subset can mutate in α ways, except the dominant
one.

We now have the following definition:

Definition 2.1. A q-analogue of GTWN of the second kind denoted by

 (α)
n
, is defined by
k q,r
 (α)
n (α)
:= w(An,k,r ) n, k ≥ 1, n ≥ k ≥ r
k q,r
 (α)  (α)  (α)
0 n n
where := δ0k , := δ0n , n, k ≥ 0 and := 0 when
k q,r 0 q,r k q,r
k < r.

Remark 2.1. The above weight function are chosen so that, when q = 1,
 (α)  (α)
n (α) n
= |An,k,r | = .
k 1,r k r

Moreover, the above weight function is a kind of variation of the weight

function corresponding to the q-Stirling numbers of the second kind by [4]
and [12] resulting to a new q-analogue of second kind Stirling-type numbers.
One may also try to define a q-analogue of GTWN of the second kind using
the weight function in terms of non-inversion numbers.
When n = 4, k = 3, r = 2, and α = 1, we have the following partitions
of {0, 1, 2, 3}:
(1)
A4,3,2 = {{0}{1}{2, 3}}, {{0, 2}{1}{3}}, {{0}{1, 2}{3}}, {{0, 3}{1}{2}},

{{0}{1, 3}{2}}

5
Then
 (1)
4
= q 0 + q 0+2 + q 0 + q 0+3 + q 0 = 3 + q 2 + q 3 . (7)
3 q,2

To compute quickly the first values of the q-analogue, let us consider the
following recurrence relation:
 (α)
n
Theorem 2.1. The number satisfy the following recurrence relation
k q,r

 (α)  (α)  (α)

n+1 n n n
= + α(k − 1 + q )
k q,r k−1 q,r k q,r

where n ≥ k ≥ r ≥ 0.

(α)
Proof. We write An+1,k,r = C1 ∪ C2 ∪ C3 such that

(α)
• C1 is the set of all π ∈ An+1,k,r such that {n} is one of the nonempty
parts of π.

(α) (α)
• C2 is the set of all π ∈ An+1,k,r such that n ∈ Bi , i 6= 0, and
(α)
Bi \{n} =
6 φ.

(α) (α)
• C3 is the set of all π ∈ An+1,k,r such that n ∈ B0 .

Then we have
 (α)  (α)  (α)
n n n
w(C1 ) = , w(C2 ) = α(k − 1) , and w(C3 ) = αq n .
k − 1 q,r k q,r k q,r

Using this recurrence relation, we can generate the first values of the q-
 (α)
n
analogue. For instance, the following tables contain the values of
k q,r
for n = 0, 1, 2, 3, 4 with r = 2 and α = 1, 2:

 (1)
4
Note that the value of as computed in (7) coincides with that in
3 q,2
Table 1.

6
  (α)
n
Table 1: Table of Values for when α = 1
k q,r

n\k 0 1 2 3 4
0 1 0
1 0 1
2 0 0 1
3 0 0 q2 +1 1
4 0 0 q5 + q3 + q2 + 1 q3 + q2 + 3 1

 (α)
n
Table 2: Table of Values for when α = 2
k q,r

n\k 0 1 2 3 4
0 1 0
1 0 1
2 0 0 1
3 0 0 2q 2 + 2 1
4 0 0 4q 5 + 4q 3 + 4q 2 + 4 2q 3 + 2q 2 + 6 1

7
 By applying the triangular recurrence relation twice, we can have

n + 1 (α)
   (α)  (α)
n n
= + α(k − 1 + q n )
k q,r k − 1 q,r k q,r
(α)
n − 1 (α)
  
n n
= + α(k − 1 + q )
k − 1 q,r k − 1 q,r
n − 2 (α)
 
2 n n−1
+ α (k − 1 + q )(k − 1 + q )
k − 1 q,r
n − 2 (α)
 
3 n n−1 n−2
+ α (k − 1 + q )(k − 1 + q )(k − 1 + q ) .
k q,r

It can be observed easily that, if we continue applying the triangular recur-

rence relation up to n − k + 1 times, we get

n + 1 (α)
(α)
n − 1 (α)
    
n n
= + α(k − 1 + q )
k q,r k − 1 q,r k − 1 q,r
n − 2 (α)
 
2 n n−1
+ α (k − 1 + q )(k − 1 + q )
k − 1 q,r
n − 2 (α)
 
3 n n−1 n−2
+ α (k − 1 + q )(k − 1 + q )(k − 1 + q )
k q,r
n−k
k − 1 (α)
Y  
n−k+1 n−i
+ ... + α (k − 1 + q )
k − 1 q,r
i=0
n−k+1
k − 1 (α)
Y  
+ αn−k+2 (k − 1 + q n−i ) .
k q,r
i=0

k − 1 (α)
 
Now, using the fact that = 0, we have the following vertical
k q,r
 (α)
n
recurrence relation for .
k q,r
 
n
Theorem 2.2. The q-analogue of GTWN of the second kind satisfies
k q,r
the following vertical recurrence relation
(α) n−k+1 j−1
n − j (α)
  
n+1 X
j
Y
n−i
= α (k − 1 + q ) . (8)
k q,r k − 1 q,r
j=0 i=0

For example, when α = 2, n = 3, k = 3, using equation (8) and Table 2, we

8
obtain
 (2)  (2)  (2)
4 3 3 2
= + 2(3 − 1 + q )
3 q,r 2 q,r 2 q,r
= (2q 2 + 2) + 2(3 − 1 + q 3 ) = 2q 3 + 2q 2 + 6,
 (2)
4
which is exactly the value of that appeared in Table 2.
3 q,r
 
n
Another form of recurrence relation can also be established for ,
k q,r
which is classified as horizontal recurrence relation. The following theorem
contains the said recurrence relation.
 
n
Theorem 2.3. The q-analogue of GTWN of the second kind satisfies
k q,r
the following horizontal recurrence relation
 (α) n−k n  (α)
n j (q − 1)k+j+1 n+1
X
= (−α) . (9)
k q,r (q n − 1)k+1 k + j + 1 q,r
j=0

Proof. Using the triangular recurrence relation in Theorem 2.1, the right-
hand side (RHS) of (9) gives
n−k (α)
(q n − 1)k+j+1

X n
RHS = (−1)j αj
(q n − 1)k+1 k + j q,r
j=0
n−k (α)
(q n − 1)k+j+1

X n
+ (−1)j αj α(k + j + q n
)
(q n − 1)k+1 k + j + 1 q,r
j=0
n−k (α)
(q n − 1)k+j+1

X n
= (−1)j αj
(q n − 1)k+1 k + j q,r
j=0
n−k (α)
(q n − 1)k+j+2

X n
+ (−1)j αj+1
(q n − 1)k+1 k + j + 1 q,r
j=0
 (α) n−k (α)
(q n − 1)k+j+1

n X n
= + (−1)j αj
k q,r (q n − 1)k+1 k + j q,r
j=1
n−k (α)
(q n − 1)k+j+1

X n
+ (−1)j−1 αj
(q n − 1)k+1 k + j q,r
j=1
 (α)
n
= .
k q,r

9
  (α)
n
The next theorem contains an explicit formula for , which is
k q,r
analogous to certain identity of [3]. But before that, let us consider first the
following lemma from [10].
Lemma 2.4. [10, Corcino-Fernandez] The following relation holds
 
X
j1 +j2 +···+ji n−r+1 i(i−1+2r)
q = q 2 .
i q
r≤j1 <j2 <···<ji ≤n

(α)
Writing π ∈ An+1,k+1 in the form
(α) (α)
π = {0, j1 , j2 , · · · , ji }(α) /B1 / · · · /Bk

where jl 6= 1, 2, . . . , r − 1, we get therefore

 (α)
n+1 (α)
X
= w(An+1,k+1 ) = w(π)
k+1 q,r (α)
π∈An+1,k+1
n

n − i (α)
X X  
j1 j2 ji



= αq αq · · · αq
k r−1
i=0 r≤j1 <···<ji ≤n
n   (α)
X X n−i
= αi q j1 +j2 +···+ji
k r−1
i=0 r≤j1 <···<ji ≤n

Thus, using Lemma 2.4, we obtain the following explicit formula.

n + 1 (α)
 
Theorem 2.5. The explicit formula for is given by
k + 1 q,r
n 
n + 1 (α) X n − r + 1 n − i (α)
    
i(i−1+2r)
i
= αq 2 . (10)
k + 1 q,r i q k r−1
i=0

Remark 2.2. Equation (10) is a q-analogue of the following identity

(α) n
n − r + 1 n − i (α)
   
n+1 X
i
= α
k+1 r i k r−1
i=0

which, by taking α = 1, coincides to the identity by [3]

  X n − rn − p − k 
n
= pk ,
m r k m − p r−p
k

with p = 1.

10
Remark 2.3. The horizontal generating function for GTWN of the second
kind in (4) can be expressed as
n 
n + r (α) k D x E
X 
n
(x + αr) = α
k+r r α k
k=0
n 
n + r (α) k
X 
(αx + αr)n = α hxik .
k+r r
k=0

where hxik = (x)(x − 1)(x − 2) · · · (x − k + 1).

Using Newton’s interpolation formula, we obtain
k
n + r (α)
   
1 X k−j k
= k (−1) (αj + αr)n .
k+r r α k! j
j=0

This formula can also be written as

k
n + r (α) αn−k X
   
k−j k
= (−1) (j + r)n . (11)
k+r r k! j
j=0

Thus, using (11), the explicit formula in (10) can be rewritten as

i(i−1+2r)
k−r+1 n−r+1
X (−1)k−r+1−j iq − 1 + j)n−r+1−i
(α) n k−r+1




n+1 X j i q
α 2 (r
=
k+1 q,r αk−n+i (k − r + 1)!
i=0 j=0
k−r+1
αn−k
 
k−r+1−j k − r + 1
X
= (−1) (r − 1 + j)n−r+1 ×
(k − r + 1)! j
j=0
(n−r+1  i )
X n − r + 1 i(i−1)  qr
× q 2 .
i q r−1+j
i=0

Applying a q-identity in the book by [7], which is given by

n   n−1
X n i(i−1) Y
xi = 1 + xq i ,


q 2
i q
i=0 i=0

we obtain
k−r+1
n + 1 (α) αn−k
   
X k−r+1
= (−1)k−r+1−j × (12)
k + 1 q,r (k − r + 1)! j
j=0
n−r
Y
× (r − 1 + j + q r+i ). (13)
i=0

This identity is a kind of q-analogue of that identity in (11) since, when

q = 1, (12) reduces immediately to (11).

11
  (α)
n
The next theorem contains a symmetric formula for which is
k q,r
analogous to the horizontal generating function of Stirling numbers of the
second kind.
 
n
Theorem 2.6. A q-analogue of GTWN of the second kind satisties
k q,r
the following relation
n−1
X (α)
n
hx − α(r − 1)|αik−r+1 = (x+αq r )(x+αq r+1 ) · · · (x+αq n−1 ).
k+1 q,r
k=0

Proof. From the well-known formula

(α)
∆k

n+r
(αx + αr)n |x=0 = ,
k! k+r r

we get

(α) n−1
n − i − 1 (α)
 X   
n n−r i
i(i−1+2r)
= αq 2
k+1 q,r i q k r−1
i=0
n−1 
∆k−r+1 X n − r

i(i−1+2r)
= αi q 2 [α(x + r − 1)]n−r−i |x=0
(k − r + 1)! i q
i=0
n−r
∆k−r+1 X n − r
 
i(i−1+2r)
= αi q 2 [α(x + r − 1)]n−i−r |x=0 .
(k − r + 1)! i q
i=0

It is known that, for a positive integer n, a real number q 6= 1, and an

indeterminate z, we have
n n  
Y
i−1
X n k(k−1)
(a + q z) = q 2 z k an−k .
k q
i=1 k=0

With z = αq r and a = x + r − 1, we obtain

(α)
∆k−r+1

n
= (αq r + α(x + r − 1))(αq r+1 + α(x + r − 1)) · · ·
k+1 q,r (k − r + 1)!
(αq n−1 + α(x + r − 1))|x=0 .

The well-known formula for higher order difference operator yields

n−1
X (α)
n
hx − α(r − 1)|αik−r+1
k+1 q,r
k=0

12
  
n−1 n−1
X ∆k−r+1 Y 
= (αq j + α(x + r − 1))|x=0 ×
 (k − r + 1)! 
k=0 j=r

× hx − α(r − 1)|αik−r+1

n−1 k−r+1  
1 k−r+1−j k − r + 1
X X
= (−1) ×
(k − r + 1)!  j
k=0 j=0
n−1
)
Y
(αq l + α(r + j − 1)) hx − α(r − 1)|αik−r+1
l=r

n−1 k−r+1  
1 k−r+1−j k − r + 1
X X
= (−1) ×
(k − r + 1)!  j
k=0 j=0
 
n−r
X X 
αi q i1 +i2 +···+ii [α(r + j − 1)]n−r−i ×
 
i=0 r≤i1 <i2 <···<ii ≤n−1

× hx − α(r − 1)|αik−r+1
n−1 n−r
X 1 X X
= αi q i1 +···+ii ×
(k − r + 1)!
k=0 i=0 r≤i1 <···<ii ≤n−1
 
k−r+1  
k−r+1−j k − r + 1
X 
n−r−i
(−1) [α(r + j − 1)] ×
 j 
j=0

× hx − α(r − 1)|αik−r+1 .
Using the explicit formula for (r, β)-Stirling numbers in (11) which also
appears in the paper by [8], we have
n−1
X (α) n−r
n X X
hx − α(r − 1)|αik−r+1 = αi q i1 +i2 +···+ii
k + 1 q,r
k=0 i=0 r≤i1 <i2 <···<ii ≤n−1
(n−1   )
X n−r−i
hx − α(r − 1)|αik−r+1 .
k − r + 1 α,α(r−1)
k=0

A relation in [8] implies that

n−1
X (α)
n
hx − α(r − 1)|αik−r+1
k+1 q,r
k=0
n−r
X X
= αi q i1 +i2 +···+ii xn−r−i
i=0 r≤i1 <i2 <···<ii ≤n−1
r r+1
= (x + αq )(x + αq ) · · · (x + αq n−1 ).

13
For example, when n = 4, α = 2 and r = 2, we have
3  (2)
X 4
hx − 2|2ik−1
k + 1 q,2
k=0
 (2)  (2)  (2)
4 4 4
= + (x − 2) + (x − 2)(x − 4)
2 q,2 3 q,2 4 q,2
= (4 + 4q 2 + 4q 3 + 4q 5 ) + (6 + 2q 2 + 2q 3 )(x − 2) + (x − 2)(x − 4)
= x2 + 2q 2 x + 2q 3 x + 4q 5 = (x + 2q 2 )(x + 2q 3 ).

It is worth mentioning that [21] considered certain generalization of Bell

numbers, called r-Bell numbers, and obtained several interesting properties
of these numbers. These numbers were first defined in [8] as the sum of
r-Stirling numbers of the second kind. It will then be interesting to define a
q-analogue of r-Bell numbers in terms of the above q-analogue of r-Stirling
numbers of the second kind and establish some properties analogous to those
obtained by Mező for r-Bell numbers.

3 A q-Analogue of GTWN of the First Kind

The inverse relation for generalized translated Whitney numbers has not
been established in [1]. However, one can derive this relation by making use
of the horizontal generating function of GTWN. For example, the classical
Stirling numbers satisfy the following inverse relation
n   n  
X n X
n−k n
fn = gk ⇐⇒ gn = (−1) fk , (14)
k k
k=0 k=0

which can be obtained using the following horizontal generating functions

n  
X n
hxin = xk
k
k=0
n  
X n
xn = hxik .
k
k=0

This motivates the authors to define a q-analogue of GTWN of the first kind
to possibly establish the inverse relation for the q-analogue of GTWN.

Definition 3.1. A q-analogue of GTWN of the first kind is defined by

n  (α)
X n
hx − α(r − 1)|αin−r = (−1)n−k (x+αq r )(x+αq r+1 ) . . . (x+αq k−1 )
k q,r
k=0
(15)

14
  (α)  (α)
n n
with r ≤ k−1. By convention, = 1 when r = k and n ≥ k, =1
k q,r 0 q,r
 (α)  (α)
n n
when n = 0, = 0 when n > 0 and = 0 when n < k or n, k < 0.
0 q,r k q,r

Using the relation in Theorem 2.6, we have

n  (α) k  (α)
X n n−k
X k
hx − α(r − 1)|αin−r = (−1) hx − α(r − 1)|αim−r
k q,r m q,r
k=0 m=1
n n
(  (α)  (α) )
X X
n−k n k
= (−1) hx − α(r − 1)|αim−r .
k q,r m q,r
m=1 k=m

Comparing the coefficients of hx − α(r − 1)|αin−r , we obtain

n  (α)  (α)
X n k
(−1)n−k = δmn
k q,r m q,r
k=m

where δmn is the Kronecker delta. On the other hand, the relation in The-
orem 2.6 can be written as

(x + αq r ) · · · (x + αq n−1 )
n  (α) X k  (α)
X n k
= (−1)k−m (x + αq r ) . . . (x + αq m−1 )
k q,r m q,r
k=1 m=0
k
( n    (α) )
X X n k
= (−1)k−m (x + αq r ) . . . (x + αq m−1 ).
k q,r m q,r
m=1 k=m

Thus, we can state formally these results in the following theorem.

Theorem 3.1. The q-analogue of GTWN of the first kind satisfies the fol-
lowing orthogonality relations
n  (α)  (α)
X n
n−k k
(−1) = δmn
k q,r m q,r
k=m
n  (α)  (α)
k−m n k
X
(−1) = δmn .
k q,r m q,r
k=m

Remark 3.1. This theorem immediately implies that

 (α) !  (α) !T
i i
(−1)i−j = In+1
j q,r j q,r
0≤i,j≤n 0≤i,j≤n

15
where In+1 is the identity matrix of order n + 1. That is,
 (α) !−1  (α) !T
i i
(−1)i−j =
j q,r j q,r
0≤i,j≤n 0≤i,j≤n

and  
 (α) !  (α) !T
i−j i i
det  (−1)  = 1.
j q,r j q,r
0≤i,j≤n 0≤i,j≤n

As a direct consequence of this theorem, we have the following inverse

relations of q-analogue of GTWN.
Theorem 3.2. The q-analogue of GTWN of the first kind satisfies the fol-
lowing inverse relations
n  (α) n  (α)
X n X n
fn = (−1)n−k gk ⇐⇒ gn = fk
k q,r k q,r
k=0 k=0
∞  (α) ∞  (α)
n−k n n
X X
fk = (−1) gn ⇐⇒ gk = fn .
k q,r k q,r
n=0 n=0

Remark 3.2. When q → 1, the orthogonality and inverse relations in Theo-

rem 3.1 and 3.2 will reduce to the following orthogonality relation
n  (α)  (α)
n−k n k
X
(−1) = δmn
k r m r
k=m
n  (α)  (α)
k−m n k
X
(−1) = δmn .
k r m r
k=m

and inverse relation

n  (α) n  (α)
n−k n n
X X
fn = (−1) gk ⇐⇒ gn = fk
k r k r
k=0 k=0
∞  (α) ∞  (α)
n−k n n
X X
fk = (−1) gn ⇐⇒ gk = fn .
k r k r
n=0 n=0

for GTWN.
 (α)
n
To compute quickly the first values of , we need the following
k q,r
triangular recurrence relation.
 
n
Theorem 3.3. The q-analogue of GTWN of the first kind satisfies
k q,r

n + 1 (α) n (α)
     (α)
n
= + α(n − 1 + q k ) . (16)
k q,r k − 1 q,r k q,r

16
Proof. Equation (15) implies that
n+1
X (α)
n+1
(−1)n+1−k (x + αq r )(x + αq r+1 ) . . . (x + αq k−1 )
k q,r
k=0
= (x − α(r − 1) − α(n − r)) hx − α(r − 1)|αin−r
n  (α)
X n
= (−1)n−k (x + αq k − αq k + α − αn)(x + αq r )(x + αq r+1 ) . . . (x + αq k−1 )
k q,r
k=0
n  (α)
X n
= (−1)n−k (x + αq r )(x + αq r+1 ) . . . (x + αq k−1 )(x + αq k )
k q,r
k=0
n  (α)
X n
+ (−1)n−k (−αq k − α(n − 1))(x + q r )(x + q r+1 ) . . . (x + q k−1 )
k q,r
k=0
n+1
X n 
= (−1)n+1−k (x + q r )(x + q r+1 ) . . . (x + q k−1 )
k − 1 q,r
k=0
n  
X
k n
+ (α(q + n − 1)) (−1)n+1−k (x + q r )(x + q r+1 ) . . . (x + q k−1 )
k q,r
k=0

By comparing the coefficients, we obtain the desired recurrence relation.

 ∗
n
We observe that the q-Stirling numbers in [5] satisfy the relation
k q
 ∗  ∗  ∗
n+1 n k n
= + (n − 1 + q )
k q k−1 q k q

which is analogous to the recurrence relation in Theorem 3.3. This recur-

n ∗

rence relation has been used to give combinatorial interpretation of in
k q
terms of the weight of permutations in {1, 2, . . . , n} with k nonempty cy-
cles. Hence, we can also use the recurrence
  relation in Theorem 3.3 to give
n
combinatorial interpretation for by following the same argument in
k q,r
 ∗
n
constructing the combinatorial interpretation of .
k q
To sketch the construction, first, we let Pn be the set of all permuta-
(α)
tions of {1, 2, . . . , n}, Pn,r be the set of all permutations of {1, 2, . . . , n}
such that elements 1, 2, . . . , r are in different cycles and the elements of
each cycle can mutate in α ways, except the dominant one. Also, we let
w(π) be the weight of π ∈ Pn . As defined in [5], the decomposition into

17
nonempty cycles C0 , C1 , . . . , Ck−1 of a permutation π ∈ Pn is called a nat-
ural decomposition if the ordering is according to decreasing largest ele-
ments of the cycles, the natural ordering. Since max(C0 ) = n, the natural
decomposition of C0 is given by {n}, C01 , C02 , . . . , C0i . Also, by [5], for
π = [C01 |C02 | . . . |C0i |n]C1 |C2 | . . . |Ck−1 ∈ Pn , we define
w(π) := q j1 +j2 +...+ji
where jl = m if C0l lies between Cm−1 and Cm in the natural ordering of
 (α)
n
cycles and jl = k if max(C0l ) < max(Ck−1 ). Then the q-analogue of
k q,r
GTWN of the first kind can be interpreted as the sum of the weights of all
(α)
permutations π ∈ Pn,r such that the natural decomposition has exactly k
cycles.
One can easily observe that, by repeated application of the triangular
recurrence relation in Theorem 3.3, we can have
n + 1 (α) n (α)
     (α)
k n
= + α(n − 1 + q )
k q,r k − 1 q,r k q,r
(α)
n − 1 (α) n − 2 (α)
    
n k 2 k
= + α(n − 1 + q )1 + α (n − 1 + q )2
k − 1 q,r k − 1 q,r k − 1 q,r
 (α)
n−2
+ α3 (n − 1 + q k )3
k q,r
..
.
n (α) n − 1 (α) n − 2 (α)
     
k 2 k
= + α(n − 1 + q )1 + α (n − 1 + q )2
k − 1 q,r k − 1 q,r k − 1 q,r
(α)
k − 1 (α)
  
3 k n−3 n−k+1 k
+ α (n − 1 + q )3 + ... + α (n − 1 + q )n−k+1
k − 1 q,r k − 1 q,r
 (α)
k−1
+ αn−k+2 (n − 1 + q k )n−k+2
k q,r
 (α)
k−1
Since = 0, we have the following vertical recurrence relation for
k q,r
 (α)
n
.
k q,r
 
n
Theorem 3.4. The q-analogue of GTWN of the first kind satisfies
k q,r
the following vertical recurrence relation
n−k+1
n + 1 (α) n − j (α)
  X  
j k
= α (n − 1 + q )j . (17)
k q,r k − 1 q,r
j=0

18
  
n
Theorem 3.5. The q-analogue of GTWN of the first kind satisfies
k q,r
the following horizontal recurrence relation
 (α) X n
( j )
n + 1 (α)

n j−k
Y
i
= (−α) (n − 1 + q ) . (18)
k q,r j + 1 q,r
j=k i=k+1

Proof. By making use of (16), the right-hand side (RHS) of (18) gives
n
( j )
n + 1 (α)
X Y 
j−k i
RHS = (−α) (n − 1 + q )
j + 1 q,r
j=k i=k+1
n
( j ) (   )
n (α) n (α)
X Y 
j−k i j+1
= (−α) (n − 1 + q ) + α(n − 1 + q )
j q,r j + 1 q,r
j=k i=k+1
n
( j ) 
X Y n (α)
= (−α)j−k (n − 1 + q i )
j q,r
j=k i=k+1
n j+1
( )
n (α)
X Y 
j−k j−k+1 i
+ (−1) α (n − 1 + q )
j + 1 q,r
j=k i=k+1
n j
 (α) ( )  (α)
n X
j−k
Y
i n
= + (−α) (n − 1 + q )
k q,r j q,r
j=k+1 i=k+1
n j
( )  (α)
X
j−1−k j−k
Y
i n
+ (−1) α (n − 1 + q )
j q,r
j=k+1 i=k+1
 (α)
n
= ,
k q,r
which is exactly the left-hand side of (18).

References
[1] H. Belbachir, I. E. Bousbaa, Translated Whitney and r-Whitney Num-
bers: A Combinatorial Approach, J. Integer Seq. 16(2013), Article
13.8.6
[2] M. Benoumhani, On Whitney numbers of Dowling lattices, Discrete
Math. 159(1996), 13–33.
[3] A. Z. Broder, The r-Stirling numbers, Discrete Math. 49(1984), 241-
259.
[4] L. Carlitz, q-Bernoulli Numbers and Polynomials, Duke Math. J.
15(1948), 987-1000.

19
 [5] J. Cigler, A new q-Analogue of Stirling Numbers, Sitzunber. Abt. II
201(1992), 97-109.
[6] J. Cigler, Operatormethoden für q-identitäten, Monatshefte. für Math-
ematik 88(1979), 87-105.
[7] L. Comtet. Advanced Combinatorics, D. Reidel Publishing Company,
Dordrecht, Holland, (1974).
[8] C. B. Corcino, An asymptotic formula for the r-Bell numbers, Matimy-
ás Mat. 24(2001), 9–18.
[9] R. B. Corcino, C. B. Corcino, R. Aldema, Asymptotic Normality of the
(r, β)-Stirling Numbers, Ars Combin. 81(2006), 81-96.
[10] R. B. Corcino, J. Fernandez, A Combinatorial Approach for q-Analogue
of r-Stirling Numbers, Br. J. Math. Comput. Sci. 4(9)(2014), 1268–
1279.
[11] M. Dowling, A Class of Geometric Lattices Based on Finite Groups, J.
Comb. Theory (B) 14(1973), 61–86.
[12] H. W. Gould, The q-Stirling Numbers of the First and the Second Kinds,
Duke Math. J. 28(1968), 281-289.
[13] L. C. Hsu, P.J-S. Shiue, A unified approach to generalized Stirling num-
bers, Advances in Appl. Math. 20(1998), 366-384
[14] F. H. Jackson, Certain q-identities, Quart. J. Math. 12(1941), 167-172.
[15] T. Mansour, Combinatorics of Set Partitions, Chapman & Hall/CRC
an imprint of Taylor & Francis LLC, Discrete Mathematics and Its
Applications Series, CRC Press, Boca Raton, FL, 2012.
[16] T. Mansour, M. Schork, M. Shattuck, On a new family of general-
ized Stirling and Bell numbers, Electronic Journal of Combinatorics
18(2011), #P77 (33pp.).
[17] T. Mansour, M. Schork, M. Shattuck, On the Stirling numbers associ-
ated with the meromorphic Weyl algebra, Applied Mathematics Letters
25(2012), 1767–1771.
[18] T. Mansour, M. Schork, M. Shattuck, The generalized Stirling and Bell
numbers revisited, J. Integer Seq. 15(2012), Article 12.8.3.
[19] I. Mező, On the Maximum of r-Stirling numbers, Advances in Appl.
Math. 41 (3)(2008), 293-306.
[20] I. Mező, New Properties of r-Stirling Series, Acta Mathematica Hun-
garica 119 (4)(2008), 341-358.

20
[21] I. Mező, The r-Bell Numbers, J. Integer Seq. 14(2011), Article 11.1.1.

[22] M. Mihoubi, M. S. Maamra, The (r1 , . . . , rp )-Stirling Numbers of the

Second Kind, Integers 12(5)(2012), 787–1089.

[23] C. G. Wagner, Partition Statistics and q-Bell Numbers (q = −1), J.

Integer Seq. 7(2004), Article 04.1.1.

21