Miskolc Mathematical Notes HU e-ISSN 1787-2413 Vol. 20 (2019), No. 1, pp. 99–114 DOI: 10.18514/MMN.2019.2536
THE GENERALIZED t-COMTET NUMBERS AND SOME COMBINATORIAL APPLICATIONS
HAC `ENE BELBACHIR, CAROLINA FORERO, AND JOS ´E L. RAM´IREZ Received 22 February, 2018
Abstract. In the present article we use a combinatorial approach to generalize the Comtet num- bers. In particular, we establish some combinatorial identities, recurrence relations and gen- erating functions. Additionally, for some particular cases we study their relationship witht- successive associated Stirling numbers and theirq-analogue.
2010Mathematics Subject Classification: 11B83; 11B73; 05A15; 05A19
Keywords: Stirling numbers of the second kind,t-successive associated Stirling numbers, Comtet numbers, combinatorial identities
1. INTRODUCTION
It is well-known that the Stirling numbers of the second kind˚n
k count the number of partitions of a set withnelements intoknon-empty blocks. This sequence satisfies the recurrence relation
(n k )
Dk (n 1
k )
C (n 1
k 1 )
; with the initial conditions˚0
0 D1and˚n
0 D˚0
n D0.
The Stirling numbers˚n
k can be generalized to theassociated Stirling numbers of the second kind ˚n
k m (cf. [1,4,7,8,10,11,18,20])) by means of a restriction on the size of the blocks. In particular, this sequence gives the number of partitions ofnelements intokblocks, such that each block contains at leastmelements. It is clear that˚n
k 1D˚n
k . This combinatorial sequence has been applied to the study of some special polynomials such as generalized Bernoulli and Cauchy polynomials, (see, e.g., [12–16]).
Recently, Belbachir and Tebtoub [2] considered a variation for the associated Stirl- ing numbers. They introduced the2-successive associated Stirling numbers of the second kind˚n
k
Œ2. This new sequence counts the number of partitions ofnelements
The research of Jos´e L. Ram´ırez was partially supported by Universidad Nacional de Colombia, Project No. 37805.
c 2019 Miskolc University Press
intokblocks, with the additional condition that each block contains at least two con- secutive elements. Moreover, the last elementn must either form a block with its predecessor or belong to another block satisfying the previous conditions. In [2], the authors derived the recurrence
(n k
)Œ2
Dk (n 1
k )Œ2
C (n 2
k 1 )Œ2
; n2k;
with the initial conditions˚0 0
Œ2D1,˚ n n 1
Œ2D0and˚n
0 D0forn1.
Inspired by these results, in this paper we aim to investigate the sequence faŒt .n; k/gn;k0, defined by the recurrence relation
aŒt .n; k/DukaŒt .n 1; k/CaŒt .n t; k 1/; nt k; (1.1) with the initial conditionsaŒt .0; 0/D1; aŒt .n; n `/D0for`D1; 2; : : : ; t 1and aŒt .n; 0/D0, forn1. Moreover,fungis a sequence of real numbers.
We will call the sequence faŒt .n; k/gn;k0 the generalized t-Comtet numbers.
The reason for this name is that fortD1we recover the Comtet numbers (see, e.g., [9,21]). Note that if uk Dk, then aŒt .n; k/D˚n
k
Œt . This sequence is called by Belbachir and Tebtoub [3] as the t-successive associated Stirling numbers. IftD2 anduk Dk, thenaŒ2.n; k/D˚n
k
Œ2. IftD1anduk Dk, thenaŒ1.n; k/D˚n k . In this paper our goal is to give the recurrence relation, the generating function and some combinatorial identities. For some particular cases, we give combinatorial interpretations.
2. BASIC PROPERTIES
From the recurrence relation (1.1) we obtain the following generating function.
Theorem 1. Fork1, AŒt k .x/WD X
nt k
aŒt .n; k/xnD xt k
.1 u0x/.1 u1x/.1 u2x/ .1 ukx/; (2.1) withAŒt 0 .x/D1 u10x.
Proof. Multiplying both sides of (1.1) byxnand summing overnt k, we have AŒt k .x/Duk X
nt k
aŒt .n 1; k/xnC X
nt k
aŒt .n t; k 1/xn Dukx X
nt k
aŒt .n; k/xnC X
nt k t
aŒt .n; k 1/xnCt DukxAŒt k .x/CxtAŒt k 1.x/:
THE GENERALIZED -COMTET NUMBERS 101
Then
AŒt k .x/DxtAŒt k 1.x/
1 ukx :
Iterating this last recurrence, we obtain (2.1).
From the above relation, we have the following combinatorial expression.
Corollary 1. The generalizedt-Comtet numbers are given by the explicit identity aŒt .n; k/D X
i1Ci2CCikDn t k
ui11ui22 uikk; (2.2) fornt k.
Theorem 2. The generalizedt-Comtet numbers satisfy the following recurrence relation
aŒt .n; k/D
n t k
X
iD0
uikaŒt .n i t; k 1/: (2.3) Proof. Fornt k;
a.n; k/ Duka.n 1; k/ Ca.n t; k 1/;
uka.n 1; k/ Du2ka.n 2; k/ Cuka.n 1 t; k 1/;
u2ka.n 2; k/ Du3ka.n 3; k/ Cu2ka.n 2 t; k 1/;
::: ::: ::: ::: :::
un t k 1k a.t kC1; k/Dun t kk a.t k; k/ Cun t k 1k a.t kC1 t; k 1/;
un t kk a.t k; k/ Dun t kk C1a.t k 1; k/Cun t kk a.t .k 1/; k 1/;
by summing, we get the result.
Theorem 3. We have the following rational explicit formula
aŒt .nCt k; k/D
k
X
jD0
ujkCn Q
i¤j.uj ui/; (2.4) which is independent fromt.
Proof. We have
AŒt k .x/D X
nt k
aŒt .n; k/xnDxt kX
n0
aŒt .nCt k; k/xn;
then
X
n0
aŒt .nCt k; k/xnD 1
.1 u0x/.1 u1x/ .1 ukx/
D
k
X
jD0
˛j
1 ujx D
k
X
jD0
ujk Q
i¤j.uj ui/ X
n0
ujnxn
DX
n0
0
@
k
X
jD0
ujkCn Q
i¤j.uj ui/ 1 Axn;
which gives the result.
Corollary 2. The dual expression depending ont aŒt .n; k/D
k
X
jD0
ujnCk.t 1/
Q
i¤j.uj ui/: (2.5)
2.1. Exponential generating function for the t-Comtet numbers
Let u1; : : : ; uk be a sequence of complex numbers and let .Am/mD1;:::;n be the sequence of matrices such thatAmismm-matrix
AmD 2 6 6 6 4
uk m uk mC1 uk 1
uk mC1 uk
::: : :: : :: 0
uk 1 uk 0 0
3 7 7 7 5
;
with the convention thatu<0D0.
Consider also
j D. 1/j X
1k1<k2<<kjk
uk1 ukj;
(the alternate sequence of elementary symmetric function associated tou1; u2; : : : ; uk).
We have.p u1/.p u2/ .p uk/DpkC1pk 1C2pk 2C Ck. Now we can state the following lemma which will be used to establish the main result of this subsection.
Lemma 1. We have the following decomposition 1
pn.p u1/ .p uk/ D
n
X
iD0
˛n i
pi C
k
X
jD1
ˇj
p uj
; (2.6)
THE GENERALIZED -COMTET NUMBERS 103
with˛iD. 1/b.iC1/=2c
kiC1 det.Ai/,˛0D1=k, andˇj D 1 ujn
k
Y
iD1 i¤j
.uj ui/ .
Proof. We leave the proof to the reader.
LetCkŒt .x/WD X
nt k
aŒt .n; k/xn
nŠ, withC0Œt .x/D1. We have
@t
@xtCkŒt .x/D X
nt .k 1/
aŒt .nCt; k/xn nŠ; which gives using relation (1.1),
@t
@xtCkŒt .x/Duk @t 1
@xt 1CkŒt .x/CCk 1Œt .x/: (2.7) To solve the linear recurrence differential equation we use Laplace transform. Using the fact that
CkŒt .0/D @
@xCkŒt .y/
ˇ ˇ ˇ ˇy
D0
D D @t 1
@xt 1CkŒt .y/
ˇ ˇ ˇ ˇy
D0
D0;
we get
k
Y
iD1
.pt uipt 1/L.CkŒt .y//DL.Ck 1Œt /;
whereL.CkŒt .y//D Z 1
0
CkŒt epy@y:
Thus by recursion, we get p.t 1/k
k
Y
iD1
.p ui/L.CkŒt .y//DL.C0Œt .y//DL.u.y//;
whereu.t /is the Heaviside function. Using Lemma1, we have L.CkŒt .y///DL.u.y//
2 4
.t 1/k
X
iD0
˛i
pi C
k
X
jD1
ˇj
p uj
3 5: The inverse Laplace transform gives,
CkŒt .y/D
.t 1/k
X
iD1
˛i
yi 1 .i 1/ŠC
k
X
jD1
ˇjeujy: (2.8)
Theorem 4. The exponential generating function oft-Comtet numbers is given by X
nt k
aŒt .n; k/xn nŠ D
.t 1/k
X
iD1
˛i
xi 1 .i 1/ŠC
k
X
jD1
ˇjeujx: (2.9) 3. THE2-SUCCESSIVE ASSOCIATEDr-WHITNEY NUMBERS
In this section, we study the particular caseukDkmCr. Letn; r0be integers.
Let˘r.n; k/denote the set of partitions of the setŒnCrWD f1; : : : ; n; nC1; : : : ; nC rgintokCrblocks, such that, the firstrelements are in distinct blocks. The elements f1; 2; : : : ; rgwill be called special elements. A block of a partition of the above set is called special if it contains special element. The cardinality of ˘r.n; k/ is the r-Stirling numbers of the second kind [5].
The 2-successive associated r-Whitney numbers of the second kind, denoted Wm;rŒ2.n; k/, count the number of partitions in˘r.n; k/, such that:
theknon-special blocks contain at least two consecutive numbers,
all the elements but the last one and its predecessor in non-special blocks are coloured with one ofmcolours independently,
the elements in the special blocks are not coloured,
the last elementnCrmust either form a block with its predecessor or belong to another block (special or not-special) satisfying the previous conditions.
We denote by˘r;mŒ2.n; k/the set of partitions in˘r.n; k/that satisfying the previous conditions. It is clear that ifrD0andmD1, thenW1;0Œ2.n; k/D˚n
k
Œ2, (see [2]).
For example,W2;3Œ2.5; 2/D15with the partitions being (themD2different colours of the elements will be fixed asredandblue, and therD3special elements are1; 2 and3):
˚f1g;f2g;f3g;f4; 5; 6g;f7; 8g ; ˚
f1g;f2g;f3g;f4; 5; 6g;f7; 8g ;
˚f1; 6g;f2g;f3g;f4; 5g;f7; 8g ; ˚
f1g;f2; 6g;f3g;f4; 5g;f7; 8g ;
˚f1g;f2g;f3; 6g;f4; 5g;f7; 8g ; ˚
f1; 4g;f2g;f3g;f5; 6g;f7; 8g ;
˚f1g;f2; 4g;f3g;f5; 6g;f7; 8g ; ˚
f1g;f2g;f3; 4g;f5; 6g;f7; 8g ;
˚f1; 8g;f2g;f3g;f4; 5g;f6; 7g ; ˚
f1g;f2; 8g;f3g;f4; 5g;f6; 7g ;
˚f1g;f2g;f3; 8g;f4; 5g;f6; 7g ; ˚
f1g;f2g;f3g;f4; 5; 8g;f6; 7g ;
˚f1g;f2g;f3g;f4; 5; 8g;f6; 7g ; ˚
f1g;f2g;f3g;f4; 5g;f6; 7; 8g ;
˚f1g;f2g;f3g;f4; 5g;f6; 7; 8g : Theorem 5. Forn2k, we have
Wm;rŒ2.n; k/D.kmCr/Wm;rŒ2.n 1; k/CWm;rŒ2.n 2; k 1/: (3.1)
THE GENERALIZED -COMTET NUMBERS 105
Proof. For any set partition of ˘r;mŒ2.n; k/, there are three options: either nCr form a block with its predecessor.nCr 1/, ornCr is in a special block ornCr is in a non-special block. In the first case, there areWm;rŒ2.n 2; k 1/possibilities.
In the second case, the elementnCr can be place into one of ther special blocks and the remaining elements can be chosen inWm;rŒ2.n 1; k/. Altogether, we have rWm;rŒ2.n 1; k/possibilities. For the third case, we can follow a similar argument,
then we obtainkmWm;rŒ2.n 1; k/possibilities.
A comparison of (3.1) and (1.1) shows thataŒ2.n; k/DWm;rŒ2.n; k/forukDkmC r. Therefore, from Theorem1and Corollary1we get the following corollaries.
Corollary 3. Fork1, WkŒ2.x/WD X
n2k
Wm;rŒ2.n; k/xn
D x2k
.1 rx/.1 .mCr/x/.1 .2mCr/x/ .1 .kmCr/x/; (3.2) withW0Œ2.x/D1 rx1 . Moreover, the2-successive associatedr-Whitney numbers of the second kind are given by the explicit identity
Wm;rŒ2.n; k/D X
i0Ci1Ci2CCikDn 2k
ri0.mCr/i1 .kmCr/ik; (3.3) forn2k.
In particular, for mD1 and r D0 we obtain the generating function of the 2- successive associated Stirling numbers of the second kind.
Corollary 4. (see [2, Theorem 2.3 and Corollary 2.4] and [3, Theorem 18]) For k1,
Ak.x/WD X
n2k
(n k
)Œ2
xnD x2k
.1 x/.1 2x/ .1 kx/; (3.4) withA0.x/D1. Moreover,
(n k
)Œ2
D X
i1Ci2CCikDn 2k
1i12i2 kik:
Our next identity expressesWm;rŒ2.n; k/in terms of˚i
k
Œ2forin.
Theorem 6. Letn; k0, Wm;rŒ2.n; k/D
n
X
iD2k
rn i n k n i
! mi 2k
(i k
)Œ2
: (3.5)
Proof. From (3.2) we have X
n2k
Wm;rŒ2.n; k/xnD x2k
.1 rx/.1 .mCr/x/.1 .2mCr/x/ .1 .kmCr/x/
D x2k
.1 rx/kC1 1 1 rxmx
1 1 rx2mx
1 1 rxkmx D.1 rx/k 1
m2k
mx 1 rx
2k
1 1 rxmx
1 1 rx2mx
1 1 rxkmx D.1 rx/k 1
m2k
y2k
.1 y/.1 2y/ .1 kmy/; whereyD1 rxmx .
Therefore from (3.4), we have X
n2k
Wm;rŒ2.n; k/xnD.1 rx/k 1 m2k
X
i2k
(i k
)Œ2
yi
D X
i2k
(i k
)Œ2
mi 2kxi .1 rx/i kC1 D X
i2k
X
j0
mi 2k i kCj i k
! rj
(i k
)Œ2
xiCj:
Comparing the coefficients ofxn, we obtain (3.5).
Combinatorial proof:We can construct any set partition of˘r;mŒ2.n; k/as follows:
we putn i elements in the special blocks. Then there are n ki k
rn i possibilities.
Note that we have to subtractkelements ofnbecause in the non-special blocks there are at least two consecutive numbers. The remaining i elements (i 2k) can be chosen in mi 2k˚i
k ways. The factor mi 2k accounts for thei 2k non-minimal elements within these blocks that are each to be colored in one ofmways.
From Theorem5and by induction onnwe obtain the following identity.
THE GENERALIZED -COMTET NUMBERS 107
Theorem 7. Forn2kwe have Wm;rŒ2.n; k/D 1
mkkŠ
k
X
jD0
. 1/k j k j
!
.mjCr/n k: Proof. Let,
Wm;rŒ2.n; k/D.mkCr/
mkkŠ
k
X
jD0
. 1/k j k j
!
.mjCr/n 1 k
C 1 mk 1kŠ
k
X
jD0
. 1/k 1 j k 1 j
!
.mjCr/n 1 k
D.mkCr/
mk 1kŠ
k
X
jD1
. 1/k j k 1 j 1
!
.mjCr/n 1 k
C r mkkŠ
k
X
jD0
. 1/k j k j
!
.mjCr/n 1 k
D 1 mkkŠ
n 1 k
X
iD0
rn 1 k imi 2 4
k
X
jD0
. 1/k jjiC1 k j
! m
C
k
X
jD0
. 1/k jjir k j
!3 5
D 1 mkkŠ
k
X
jD0
. 1/k j k j
!
.mjCr/n k:
3.1. Relations with ther-Whitney numbers
Ther-Whitney numbers of the second kindWm;r.n; k/were defined by Mez˝o [17]
as the connecting coefficients between some particular polynomials.
For non-negative integersn; kandrwithnk0and for any integerm > 0 .mxCr/nD
n
X
kD0
mkWm;r.n; k/xk; (3.6) wherexnDx.x 1/ .x nC1/forn1, andx0D1.
Ther-Whitney numbers of the second kind satisfy the recurrence [17]
Wm;r.n; k/DWm;r.n 1; k 1/C.kmCr/Wm;r.n 1; k/: (3.7)
Comparing (3.7) and (3.1) we have the following relation.
Corollary 5. [2, Theorem 4.1] Forn2k,
Wm;rŒ2.n; k/DWm;r.n k; k/: (3.8) Mez˝o and Ram´ırez [19] studied the r-Whitney matrices of the second and the first kind and they derived several identities for these matrices. In particular, the r-Whitney matrix of the second kind is defined by
ŒWm;r.n; k/n;k0D 2 6 6 6 6 6 6 6 4
1 0 0 0 0
r 1 0 0 0
r2 mC2r 1 0 0
r3 m2C3rmC3r2 3mC3r 1 0
r4 m3C4rm2C6r2mC4r3 7m2C12rmC6r2 6mC4r 1
::: ::: :::
3 7 7 7 7 7 7 7 5 :
Notice that the sequence.Wm;rŒ2.n; k//k corresponds with the sequence of elements on rays in direction.1; 1/over ther-Whitney matrix of the second kind.
4. THEt-SUCCESSIVE ASSOCIATEDr-WHITNEY NUMBERS
In this section, we consider the rays in direction.s; 1/, i.e., we are going to study the sequence fWm;r.n sk; k/g. We denote by Wm;rŒt .n; k/the number Wm;r.n sk; k/, wheret DsC1. We call this new sequencethet-successive associatedr- Whitney numbers of the second kind. It is possible to show that the t-successive associatedr-Whitney numbers count the number of partitions in˘r.n; k/, such that:
theknon-special blocks contain at leastt consecutive numbers,
all the elements but the last one and its t 1 predecessors in non-special blocks are coloured with one ofmcolours independently,
the elements in the special blocks are not coloured,
the last elementnCr must either form a block with its t 1-predecessors or belong to another block (special or not-special) satisfying the previous conditions.
Reasoning in a similar manner as in Theorem5we obtain the following results.
Theorem 8. Fornt k, we have
Wm;rŒt .n; k/D.kmCr/Wm;rŒt .n 1; k/CWm;rŒt .n t; k 1/: (4.1) Fork1,
WkŒt .x/WDX
n0
Wm;rŒt .n; k/xn
D xt k
.1 rx/.1 .mCr/x/.1 .2mCr/x/ .1 .kmCr/x/; (4.2)
THE GENERALIZED -COMTET NUMBERS 109
withW0Œt .x/D1 rx1 . Moreover, fornt kwe have Wm;rŒt .n; k/D 1
mkkŠ
k
X
jD0
. 1/k j k j
!
.mjCr/n .t 1/k: (4.3) As corollary fortD2we get [3, Theorem 4, Theorem 6 and Theorem 7].
It is not difficult to generalize the relation given in Theorem6.
Theorem 9. Ifn; k0, then Wm;rŒt .n; k/D
n
X
iDt k
rn i n k n i
! mi t k
(i k
)Œt
: (4.4)
Consequence.From Equation (4.2) we deduce thatWm;rŒt .nC.t 1/k; k/are the classicalr-Whitney numbersWm;r.n; k/.
From the explicit formula given in (4.3) we get the exponential generating function of thet-successive associatedr-Whitney numbers.
Theorem 10. The exponential generating function of thet-successive associated r-Whitney numbers is
WkŒt .x/WD X
nt k
Wm;rŒt xn nŠ D
k
X
jD0
k j
!. 1/k j kŠmk
e.j mCr/x
.j mCr/.t 1/k: (4.5) Corollary 6. For the2-successive associatedr-Whitney numbers,
WkŒ2.x/D
k
X
jD0
k j
!. 1/k j kŠmk
e.j mCr/x
.j mCr/k: (4.6)
These two result are more specified expressions as relation (2.9) of Theorem4.
Proof. (Theorem10) We use the derivation.t 1/ktimes according toxand using the consequence property, we get
@.t 1/k
@.t 1/kxWk.x/D X
nt k .t 1/k
Wm;rŒt .nCk.t 1/; k/xn nŠ D 1
kŠmkerx.emx 1/k D 1
kŠmkerx
k
X
jD0
k j
!
. 1/k jej mx
D
k
X
jD0
k j
!. 1/k j
kŠmk e.j mCr/x:
Theorem 11. Fornk, we have
Wm;rŒt .nC.t 1/k; k/D X
i1CCinDn k i1;;in2f0;1g
n 1
Y
jD0
.rCm.j
j
X
`D1
i`//ijC1: (4.7)
Proof. By induction overn, we suppose that the identity is true untiln 1 X
i1CCinDn k i1;;in2f0;1g
n 1
Y
jD0
.rCm.j
j
X
lD1
il//ijC1
D X
i1CCin 1D.n 1/ .k 1/
i1;;in 12f0;1g
n 2
Y
jD0
.rCm.j
j
X
lD1
il//ijC1
C 0 B B
@
X
i1CCin 1D.n 1/ k i1;;in 12f0;1g
n 2
Y
jD0
.rCm.j
j
X
lD1
il//ijC1 1 C C A
.rCmk/:
We have
Wm;rŒt .nC.t 1/k; k/
D.mkCr/Wm;rŒt .n 1C.t 1/k; k/CWm;rŒt .n 1C.t 1/.k 1/; k/;
which gives the desired result.
Corollary 7. Fornt k, we have Wm;rŒt .n; k/D X
i1CCin .t 1/kDn t k i1;;in .t 1/k2f0;1g
n .t 1/k 1
Y
jD0
.rCm.j
j
X
`D1
i`//ijC1; (4.8)
with empty sum equal zero.
Example1. ForkDtD2we have the following formula Wm;rŒ2.n; 2/D X
i1CCin 2Dn 4
ri1.rCm.1 i1//i2.rCm.2 i1 i2//i3 .rCm.n 3 i1 i2 in 3//in 2;
Wm;rŒ2.4; 2/D X
iCjD0
ri.rCm.1 i //j D1;
THE GENERALIZED -COMTET NUMBERS 111
Wm;rŒ2.5; 2/D X
iCjCkD1
ri.rCm.1 i //j.rCm.2 i j //kD3.rCm/;
Wm;rŒ2.6; 2/D X
iCjCkC`D2
ri.rCm.1 i //j.rCm.2 i j //k.rCm.3 i j k//` D6r2C12mrC7m2:
Theorem 12. We have the following explicit formula
Wm;rŒt .nCt k; k/D 1 mkkŠ
k j
X
jD0
. 1/k j k j
!
.mjCr/nCk; (4.9) and thus
Wm;rŒt .n; k/D 1 mkkŠ
k
X
jD0
. 1/k j k j
!
.mjCr/n .t 1/k: (4.10) Proof. It suffices to setukDmkCrthen
uj ui Dm.j i /
in Theorem3.
FortD2we get Theorem7.
Theorem 13. Expression oft-successiver-Whitney numbers in terms of binomials and Stirling numbers.
Wm;rŒt .nCt k; k/Dr m
knCk
X
iD0
nCk i
! rn imi
(i k
)
: (4.11)
Proof.
Wm;rŒt .nCt k; k/D 1 mkkŠ
k
X
jD0 nCk
X
iD0
. 1/k j k j
! nCk i
!
.mj /irnCk i
D 1 mkkŠ
nCk
X
iD0
mirnCk i nCk i
! k X
jD0
. 1/k j k j
! ji
D 1 mk
nCk
X
iD0
mirnCk i nCk i
!(i k
) : Notice thatPk
jD0. 1/k j kj
ji D0fori < k.
5. Aq-ANALOGUE OF THEt-SUCCESSIVE ASSOCIATEDSTIRLING NUMBERS
Finally, we considerer aq-analogue of thet-successive associated Stirling num- bers of the second kind. For this purpose, we use a similar statistic studied by Carlitz [6], see also [21].
LetDB1=B2= =Bkbe any block representation of a set partition in˘0;1Œt .n; k/WD
˘Œt .n; k/, with min.B1/ <min.B2/ < <min.Bk/. We define the following stat- istic on the set˘Œt .n; k/.
wŒt ./WD
k
X
iD1
.i 1/.jBij tC1/:
We now define theq-analogue of thet-successive associated Stirling numbers of the second kind.
Definition 1. Define˚n k
Œt
q as the distribution polynomial for thewŒt statistic on the set˘Œt .n; k/, that is,
(n k
)Œt
q
D X
2˘Œt .n;k/
qwŒt ./; n; k0;
whereqis an indeterminate.
It is clear that˚n k
Œt 1 D˚n
k Œt .
For example, in the set˘Œ2.7; 3/we have the following partitions:
ff1; 2g;f3; 4; 5g;f6; 7gg; ff1; 2; 3g;f4; 5g;f6; 7gg; ff1; 2; 5g;f3; 4g;f6; 7gg; ff1; 2g;f3; 4g;f5; 6; 7gg; ff1; 2g;f3; 4; 7g;f5; 6gg; ff1; 2; 7g;f3; 4g;f5; 6gg: Therefore,
(7 3
)Œ2
q
Dq4Cq3Cq3Cq5Cq4Cq3D3q3C2q4Cq5: Let us introduce the following notations.
ŒnqD1CqC Cqn 1; ŒnqŠDŒ1qŒ2q Œnq and
"
n k
#
q
D ŒnqŠ ŒkqŠŒn kqŠ: The last coefficient is calledq-binomial coefficient. IfqD1, thenn
k
1D nk . Theorem 14. Fornt k, we have
(n k
)Œt
q
DŒkq (n 1
k )Œt
q
Cqk 1 (n t
k 1 )Œt
q
: (5.1)
Proof. For any set partition of ˘Œt .n; k/, there are two options: either n form a block with its t 1-predecessors, or n is in a block that satisfies the conditions.
In the first case, there are qk 1˚n t
k 1 Œt
q possibilities. In this case, the size of the last block Bk is t, then this block contributes a factor qk 1. In the second case,
THE GENERALIZED -COMTET NUMBERS 113
the element n can be place into one of the k blocks and thus contributes a factor 1CqCq2C Cqk 1DŒkq. Moreover, the remaining elements can be chosen in
˚n 1 k
Œt
q ways. Altogether, we haveŒkq˚n 1 k
Œt
q possibilities.
From above theorem, we obtain the following corollaries.
Corollary 8. Fork1, X
nt k
(n k
)Œt
q
xnD xt kq.k2/
.1 x/.1 Œ2qx/.1 Œ3qx/ .1 Œkqx/: (5.2) Moreover, theq-analogue of thet-successive associatedr-Stirling numbers are given by the explicit identity
(n k
)Œt
q
Dq.k2/ X
i1Ci2CCikDn t k
Œ1iq1Œ2iq2 Œkiqk; (5.3) fornt k.
Corollary 9. Fornt kwe have (n
k )Œt
q
D 1 ŒkqŠ
k
X
jD1
. 1/k j
"
k j
#
q
q.k2j/.Œj q/n .t 1/k:
ACKNOWLEDGEMENT
The authors would like to thank the anonymous referee for some useful comments.
REFERENCES
[1] H. Belbachir and I. E. Bousbaa, “Associated Lah numbers and r-Stirling numbers.”
arXiv:1404.5573, pp. 1–24, 2014.
[2] H. Belbachir and A. F. Tebtoub, “Les nombres de Stirling associ´es avec succession d’ordre 2, nombres de Fibonacci-Stirling et unimodalit´e.”C. R., Math., Acad. Sci. Paris, vol. 353, no. 9, pp.
767–771, 2015, doi:doi.org/10.1016/j.crma.2015.06.008.
[3] H. Belbachir and A. F. Tebtoub, “The t-successive associated Stirling numbers, t-Fibonacci–
Stirling numbers, and unimodality.”Turkish J. Math., vol. 41, no. 5, pp. 1279–1291, 2017, doi:
doi.org/10.3906/mat-1506-83.
[4] M. B´ona and I. Mez˝o, “Real zeros and partitions without singleton blocks.”Eur. J. Comb., vol. 51, pp. 500–510, 2016, doi:doi.org/10.1016/j.ejc.2015.07.021.
[5] A. Z. Broder, “The r-Stirling numbers.” Discrete Math., vol. 49, pp. 241–259, 1984, doi:
doi.org/10.1016/0012-365X(84)90161-4.
[6] L. Carlitz,Generalized Stirling numbers. Combinatorial Analysis Notes, Duke University, 1968.
[7] J. Y. Choi, L. Long, S.-H. Ng, and J. Smith, “Reciprocity for multirestricted Stirl- ing numbers.” J. Comb. Theory, Ser. A, vol. 113, no. 6, pp. 1050–1060, 2006, doi:
doi.org/10.1016/j.jcta.2005.10.001.
[8] J. Y. Choi and J. D. H. Smith, “On the combinatorics of multi-restricted numbers.”Ars Comb., vol. 75, pp. 45–63, 2005.
[9] L. Comtet, “Nombres de Stirling g´en´eraux et fonctions sym´etriques.”C. R. Acad. Sci., Paris, S´er.
A, vol. 275, pp. 747–750, 1972.
[10] L. Comtet,Advanced combinatorics. D. Reidel Publishing Co. (Dordrecht, Holland), 1974.
[11] F. T. Howard, “Associated Stirling numbers.”Fibonacci Quart., vol. 18, pp. 303–315, 1980.
[12] T. Komatsu, I. Mez˝o, and L. Szalay, “Incomplete Cauchy numbers.”Acta Math. Hung., vol. 149, no. 2, pp. 306–323, 2016, doi:10.1007/s10474-016-0616-z.
[13] T. Komatsu, “Incomplete poly-Cauchy numbers.”Monatsh. Math., vol. 180, no. 2, pp. 271–288, 2016, doi:10.1007/s00605-015-0810-z.
[14] T. Komatsu, K. Liptai, and I. Mez˝o, “Incomplete poly-Bernoulli numbers associated with incom- plete Stirling numbers.”Publ. Math., vol. 88, no. 3-4, pp. 357–368, 2016.
[15] T. Komatsu and J. L. Ram´ırez, “Generalized poly-Cauchy and poly-Bernoulli numbers by using incompleter-Stirling numbers.”Aequationes Math., vol. 91, no. 6, pp. 1055–1071, 2017, doi:
10.1007/s00010-017-0509-4.
[16] T. Komatsu and J. L. Ram´ırez, “Incomplete poly-Bernoulli numbers and incomplete poly-Cauchy numbers associated to theq-Hurwitz-lerch zeta function.”Mediterr. J. Math., vol. 14, no. 3, p. 19, 2017, doi:10.1007/s00009-017-0935-5.
[17] I. Mez˝o, “A new formula for the Bernoulli polynomials.” Result. Math., vol. 58, no. 3-4, pp.
329–335, 2010, doi:10.1007/s00025-010-0039-z.
[18] I. Mez˝o, “Periodicity of the last digits of some combinatorial sequences.”J. Integer Seq., vol. 17, no. 1, pp. article 14.1.1, 18, 2014.
[19] I. Mez˝o and J. L. Ram´ırez, “The linear algebra of ther-Whitney matrices.”Integral Transforms Spec. Funct., vol. 26, no. 3, pp. 213–225, 2015, doi:10.1080/10652469.2014.984180.
[20] V. H. Moll, J. L. Ram´ırez, and D. Villamizar, “Combinatorial and arithmetical properties of the restricted and associated Bell and factorial numbers.”J. Comb., vol. 9, no. 4, pp. 693–720, 2018, doi:10.4310/JOC.2018.v9.n4.a7.
[21] C. G. Wagner, “Partition statistics andq-Bell numbers.qD 1/.”J. Integer Seq., vol. 7, no. 1, pp. art. 04.1.1, 12, 2004.
Authors’ addresses
Hac`ene Belbachir
USTHB, Faculty of Mathematics, RECITS Laboratory. P.B. 32, El Alia, 16111, Bab Ezzouar, AL- GERIA
E-mail address:hbelbachir@usthb.dz; hacenebelbachir@gmail.com
Carolina Forero
Universidad Sergio Arboleda, Departamento de Matem´aticas, Bogot´a, COLOMBIA E-mail address:tcarolinafc@gmail.com
Jos´e L. Ram´ırez
Universidad Nacional de Colombia, Departamento de Matem´aticas, Bogot´a, COLOMBIA E-mail address:jlramirezr@unal.edu.co