Some convolution identities and an inverse

relation involving partial Bell polynomials

Daniel Birmajer
Department of Mathematics
Nazareth College
Rochester, NY 14618, U.S.A.
[email protected]
Juan B. Gil Michael D. Weiner
Penn State Altoona
3000 Ivyside Park
Altoona, PA 16601, U.S.A.
{jgil,mdw8}@psu.edu
Submitted: Jun 26, 2012; Accepted: Nov 21, 2012; Published: Dec 6, 2012
Mathematics Subject Classications: 05A19, 05A10
Abstract
We prove an inverse relation and a family of convolution formulas involving par-
tial Bell polynomials. Known and some presumably new combinatorial identities
of convolution type are discussed. Our approach relies on an interesting multino-
mial formula for the binomial coecients. The inverse relation is deduced from a
parametrization of suitable identities that facilitate dealing with nested composi-
tions of partial Bell polynomials.
Keywords: Inverse relations; convolution identities; partial Bell polynomials
1 Introduction
The main goal of this paper is to obtain the inverse relation (Theorem 17):
y
n
=
n

k=1
_
an + bk
k 1
_
(k 1)!B
n,k
(x),
x
n
=
n

k=1
an + bk
an + b
_
an b
k 1
_
(k 1)!B
n,k
(y),
(1)
where a and b are integers (not both equal to 0), x and y stand for x = (x
1
, x
2
, . . . ) and
y = (y
1
, y
2
, . . . ), and B
n,k
(z) denotes the (n, k)th partial Bell polynomial in the variables
z
1
, z
2
, . . . , z
nk+1
. There is a vast literature about Bell polynomials and their applications,
see e.g. [1, 2, 5, 8]. Recall that
B
n,k
(z) =

i(n,k)
n!
i
1
!i
2
!
_
z
1
1!
_
i
1
_
z
2
2!
_
i
2
,
the electronic journal of combinatorics 19(4) (2012), #P34 1
where (n, k) is the set of all sequences i = (i
1
, i
2
, . . . ) of nonnegative integers such that
i
1
+ i
2
+ = k and i
1
+ 2i
2
+ 3i
3
+ = n.
The above relation generalizes similar inverse relations available in the literature. For
instance, with b = 1 we recover a result by Comtet [5, Theorem F, p. 151], which he
obtained by means of the inversion formula of Lagrange. Moreover, for a = 0 and b = 1,
the inverse pair (1) ts into the Fa` a di Bruno relations given by Chou, Hsu, and Shiue
[3].
To achieve (1) we develop an alternative approach. We consider a parametrization of
suitable identities (Theorem 15) that allow us to deal with nested compositions of partial
Bell polynomials. This approach seems novel and relies on interesting (presumably new)
convolution formulas for partial Bell polynomials, see Section 3.
In particular, if (, m) is a polynomial in and m of degree at most one, then for any
sequence x = (x
1
, x
2
, . . . ) we get (Corollary 11):
1
k

=0
n

m=

_
(,m)
k
__
(,m)

__
n
m
_
(, m)
_
(, m)
__
k

_B
m,
(x)B
nm,k
(x) =
(0,0)+(k,n)
(k,n)((0,0))
_

k
_
B
n,k
(x).
A general convolution identity is presented in Theorem 9, and cases of particular interest
are given in Corollary 10. For the special case when (, m) = r is a positive integer
r k, we recover the identity
_
k
r
_
B
n,k
(x) =
nr

m=kr
_
n
m
_
B
m,kr
(x)B
nm,r
(x),
recently discovered by Cvijovic [6, Eqn. (1.4)]. Identities involving Bell polynomials have
of course the usual direct consequences for the Stirling numbers of rst and second kind.
This paper is essentially self-contained. We start with a multi-variable version of the
well known identity

n
j=0
(1)
j
_
n
j
_
P(j) = 0 for any polynomial P(x) of degree less than n,
and use it to prove Theorem 2. This theorem provides core identities for the convolution
formulas of Bell polynomials given in Section 3. In Section 4, we prove the inverse relation
(1) and discuss some special cases.
It is worth noting that the results of Section 2 can be viewed as a template for a
variety of combinatorial identities of convolution type. For instance, special choices of the
parameters in equations (3), (4), and (5) give us the famous HagenRothe identities.
2 Convolution identities
Given n N it is known that, for any polynomial P(x) of degree less than n,
n

j=0
(1)
j
_
n
j
_
P(j) = 0.
1
Throughout this paper, the choice of parameters is restricted to those for which the expressions are
dened.
the electronic journal of combinatorics 19(4) (2012), #P34 2
For several variables, we have:
Lemma 1. Given nonnegative integers v
1
, . . . , v
d
,

i
1
,...,i
d
0
(1)
i
1
++i
d
_
v
1
i
1
_

_
v
d
i
d
_
P(i
1
, . . . , i
d
) = 0.
for any polynomial P(x
1
, . . . , x
d
) of degree less than k = v
1
+ + v
d
.
This lemma, which can be proved by reduction to the one-variable case, is the base of
our theorem below.
For a given v = (v
1
, v
2
, . . . , v
d
) with 0 v
i
Z, and such that v = 0, dene
W
m,
(v) =

i
d
(m,)
_
v
1
i
1
_

_
v
d
i
d
_
, (2)
where
d
(m, ) is the set of all i = (i
1
, i
2
, . . . , i
d
) N
d
0
such that
i
1
+ + i
d
= and i
1
+ 2i
2
+ + di
d
= m.
For those values of (m, ) for which
d
(m, ) is empty, we set W
m,
(v) = 0. In particular,
W
m,0
(v) = 0 unless m = 0.
Theorem 2. Let v = (v
1
, v
2
, . . . , v
d
) be any nite sequence of nonnegative integers with
v
1
+ + v
d
= k > 0 and v
1
+ 2v
2
+ + dv
d
= n. Let (, m) be a polynomial in and
m of degree at most one. For any C, we have
k

=0
n

m=
(k, n)
(, m)
_
(,m)
k
__
(,m)

_
_
k

_ W
m,
(v) =
_

k
_
, and (3)
k

=0
n

m=
(0, 0)
(, m)
_
(,m)
k
__
(,m)

_
_
k

_ W
m,
(v) =
_

k
_
. (4)
Proof. For simplicity, we will write = (, m). At rst, we assume to be an integer
with 0 k 1. For such we have
1

_

k
__

_
_
k

_ =
(1)

_

k
__
+1

_
_
k

_
=
(1)

!(k 1)!
k!
_
1

__
+ 1
k 1
_
,
the electronic journal of combinatorics 19(4) (2012), #P34 3
and so
k

=0
n

m=
1

_

k
__

_
_
k

_ W
m,
(v)
=
k

=0
n

m=
(1)

!(k 1)!
k!
_
1

__
+ 1
k 1
_
W
m,
(v)
=
!(k 1)!
k!
k

=0
n

m=
(1)

_
1

__
+ 1
k 1
_
W
m,
(v).
Since = (, m) is a polynomial in and m of degree at most one,
_
1

__
+1
k1
_
is a
polynomial in and m of degree at most k 1. Hence by Lemma 1,
k

=0
n

m=
(1)

_
1

__
+ 1
k 1
_
W
m,
(v) = 0,
and so
k

=0
n

m=
(k, n)
(, m)
_
(,m)
k
__
(,m)

_
_
k

_ W
m,
(v) = 0 for = 0, 1, . . . , k 1.
Now, both sides of equation (3) are polynomials in of degree k, having the same k roots
0, 1, . . . , k 1. The principal coecient of
_

k
_
is
1
k!
, and for the left hand side, the only
term invoving
k
is obtained when = k (impliying m = n), that is
_
(k,n)
k
_
. Thus its
principal coecient is also
1
k!
. This gives (3) for every C. The identity (4) can be
veried using the same argument; this is left to the reader.
Corollary 3. Let (, m) be as in Theorem 2. For any v
d
(n, k),
k

=0
n

m=

_
(,m)
k
__
(,m)

_
(, m)
_
(, m)
__
k

_W
m,
(v) =
(0, 0) + (k, n)
(k, n)
_
(0, 0)
_
_

k
_
. (5)
Proof. Replacing by in (4), we obtain
k

=0
n

m=
(0, 0)
(, m)
_
(,m)
k
__
(,m)

_
_
k

_ W
m,
(v) =
_

k
_
.
This identity together with (3) give the claimed formula.
In the case when (, m) = (), the left-hand side of (4) can be written as
k

=0
n

m=
(0, 0)
(, m)
_
(,m)
k
__
(,m)

_
_
k

_ W
m,
(v) =
k

=0
(0)
()
_
()
k
__
()

_
the electronic journal of combinatorics 19(4) (2012), #P34 4
since

n
m=
W
m,
(v) =
_
k

_
for any v
d
(n, k). Thus (4) becomes
k

=0
(0)
()
_
()
k
__
()

_
=
_

k
_
, (6)
and similarly, (5) turns into
k

=0

_
()
k
__
()

_
()( ())
=
(0) + (k)
(k)
_
(0)
_
_

k
_
. (7)
Example 4. With appropriate choices of and (), the above equations give the well-
known HagenRothe identities. For instance, with = x+y +kz and () = y +(k )z,
the identity (7) gives
k

=0
x
x + z
_
x + z

_
y
y + (k )z
_
y + (k )z
k
_
=
x + y
x + y + kz
_
x + y + kz
k
_
,
and with = x + y + kz and () = x + z, identity (6) leads to
k

=0
x
x + z
_
x + z

__
y + (k )z
k
_
=
_
x + y + kz
k
_
.
The special case when z = 0 is known as the ChuVandermonde identity:
k

=0
_
x

__
y
k
_
=
_
x + y
k
_
.
Example 5. For = 1, identity (4) gives
k

=0
n

m=
(0, 0)
(, m)
_
1(,m)
k
__
(,m)

_
_
k

_ W
m,
(v) = (1)
k
.
Now, since
_
1 (, m)
k
__
(, m)

_
= (1)
k
_
(, m) + k
k
__
k

_
,
we then get
k

=0
n

m=
(1)

(0, 0)
(, m)
_
(, m) + k
k
_
W
m,
(v) = 1. (8)
Choosing (, m) = z k + with z > k, and since

n
m=
W
m,
(v) =
_
k

_
, we get
1 =
k

=0
(1)

z k
z k +
_
z
k
__
k

_
=
_
z
k
_
k

=0
(1)
1
_
k

_

z k +
.
the electronic journal of combinatorics 19(4) (2012), #P34 5
In other words, we recover the known formula
1
_
z
k
_ =
k

=1
(1)
1
_
k

_

z k +
.
Example 6. Let 0 z k and N. Choose v = (v
1
, v
2
, . . . ) such that v

= z,
v
+1
= k z, and v
j
= 0 for every other j. Then
W
m,
(v) =
_
z
( + 1) m
__
k z
m
_
and (8) turns into
k

=0
(+1)

m=
(1)

(0, 0)
(, m)
_
(, m) + k
k
__
z
( + 1) m
__
k z
m
_
= 1.
Other identities can be obtained by choosing a dierent v.
We nish this section with a simple and straightforward abstraction of Theorem 2.
Theorem 7. Let v = (v
1
, v
2
, . . . , v
d
) be any nite sequence of nonnegative integers with
v
1
+ +v
d
= k > 0 and v
1
+2v
2
+ +dv
d
= n. Let p
m,
() be polynomials in of degree
at most , and such that for every
0
{0, 1, . . . , k 1}, the function (m, ) p
m,
(
0
) is
a polynomial in m and of degree at most k 1. If p
n,k
() has degree k in , then
k

=0
n

m=
(1)

k!
p
m,
()W
m,
(v) =
k
_

k
_
,
where
k
is the coecient of
k
in (1)
k
p
n,k
().
For example, in equation (3) we have
p
m,
() = (1)

k!
(k, n)
(, m)
_
(,m)
k
__
(,m)

_
_
k

_ ,
which for
0
{0, 1, . . . , k 1} can be written (cf. proof of Thm. 2) as
p
m,
(
0
) = !(k
0
1)! (k, n)
_
(, m) 1

0
__
(, m)
0
+ 1
k
0
1
_
.
Recall that (, m) is a polynomial in and m of degree at most one, thus (m, ) p
m,
(
0
)
is a polynomial in and m of degree at most k 1.
the electronic journal of combinatorics 19(4) (2012), #P34 6
3 Identities for partial Bell polynomials
In this section we derive some convolution formulas for the partial Bell polynomials:
B
n,k
(x
1
, x
2
, . . . ) =

i(n,k)
n!
i
1
!i
2
!
_
x
1
1!
_
i
1
_
x
2
2!
_
i
2
. . . ,
where (n, k) is the set of all sequences i = (i
1
, i
2
, . . . ) of nonnegative integers such that
i
1
+ i
2
+ = k and i
1
+ 2i
2
+ 3i
3
+ = n.
Lemma 8. For any x = (x
1
, x
2
, . . . ) and integers 0 k and 0 m n, we have
_
n
m
_
B
m,
(x)B
nm,k
(x) =

v(n,k)
n!
v
1
!v
2
!
W
m,
(v)
_
x
1
1!
_
v
1
_
x
2
2!
_
v
2
,
where W
m,
(v) is dened as in (2).
Proof. This is a straightforward consequence of the denitions. For simplicity, we let
z
j
=
x
j
j!
and use the convenient multinomial notation
_
n
v
_
=
n!
v
1
!v
2
!
, z
v
= z
v
1
1
z
v
2
2
. . . , and so on.
Thus
_
n
m
_
B
m,
(x)B
nm,k
(x)
=
_
n
m
_
_
_

i(m,)
_
m
i
_
z
i
_
_
_
_

j(nm,k)
_
n m
j
_
z
j
_
_
=
_
n
m
_

v(n,k)

i(m,)
_
m
i
__
n m
v i
_
z
v
=

v(n,k)
n!
v
1
!v
2
!
_
_

i(m,)
_
v
1
i
1
__
v
2
i
2
_

_
_
z
v
=

v(n,k)
n!
v
1
!v
2
!
W
m,
(v)
_
x
1
1!
_
v
1
_
x
2
2!
_
v
2

As a consequence of Theorem 7 and the previous lemma, we obtain the following
convolution formula:
the electronic journal of combinatorics 19(4) (2012), #P34 7
Theorem 9. Let p
m,
() be polynomials in with deg(p
m,
) , and such that for every

0
{0, 1, . . . , k 1}, the function (m, ) p
m,
(
0
) is a polynomial in m and of degree
at most k 1. If p
n,k
() has degree k in , then for any sequence x = (x
1
, x
2
, . . . ),
k

=0
n

m=
(1)

k!
p
m,
()
_
n
m
_
B
m,
(x)B
nm,k
(x) =
k
_

k
_
B
n,k
(x),
where
k
is the coecient of
k
in (1)
k
p
n,k
().
Proof. By Lemma 8, we have
k

=0
n

m=
(1)

k!
k
p
m,
()
_
n
m
_
B
m,
(x)B
nm,k
(x)
=
k

=0
n

m=
(1)

k!
k
p
m,
()

v(n,k)
n!
v
1
!v
2
!
W
m,
(v)
_
x
1
1!
_
v
1
_
x
2
2!
_
v
2

=

v(n,k)
n!
v
1
!v
2
!
_
k

=0
n

m=
(1)

k!
k
p
m,
()W
m,
(v)
_
_
x
1
1!
_
v
1
_
x
2
2!
_
v
2

which by Theorem 7 becomes
=
_

k
_

v(n,k)
n!
v
1
!v
2
!
_
x
1
1!
_
v
1
_
x
2
2!
_
v
2
=
_

k
_
B
n,k
(x).
For the special cases in Theorem 2, we obtain:
Corollary 10. Let (, m) be a polynomial in and m of degree at most one. For any
sequence x = (x
1
, x
2
, . . . ) and any C, we have
k

=0
n

m=
(k, n)
(, m)
_
(,m)
k
__
(,m)

__
n
m
_
_
k

_ B
m,
(x)B
nm,k
(x) =
_

k
_
B
n,k
(x),
k

=0
n

m=
(0, 0)
(, m)
_
(,m)
k
__
(,m)

__
n
m
_
_
k

_ B
m,
(x)B
nm,k
(x) =
_

k
_
B
n,k
(x).
And from Corollary 3, we deduce:
Corollary 11. For any sequence x = (x
1
, x
2
, . . . ), we have
k

=0
n

m=

_
(,m)
k
__
(,m)

__
n
m
_
(, m)
_
(, m)
__
k

_B
m,
(x)B
nm,k
(x) =
(0,0)+(k,n)
(k,n)((0,0))
_

k
_
B
n,k
(x).
the electronic journal of combinatorics 19(4) (2012), #P34 8
Example 12. Let r be an integer with 0 < r k. In the special case when (, m) = r
and = k, the rst identity in Corollary 10 gives
B
n,k
(x) =
k

=0
n

m=
_
r
k
__
kr

__
n
m
_
_
k

_ B
m,
(x)B
nm,k
(x),
and since
_
r
k
__
kr

_
= 0 unless = k r, we arrive at
_
k
r
_
B
n,k
(x) =
nr

m=kr
_
n
m
_
B
m,kr
(x)B
nm,r
(x). (9)
Observe that when r = 1, we get the basic recurrence formula
B
n,k
(x) =
1
k
n1

m=k1
_
n
m
_
x
nm
B
m,k1
(x), (10)
see Comtet [5] (relation [3k], p. 136). In particular, since the Stirling numbers of second
kind satisfy S(n, k) = B
n,k
(1, 1, . . . ), formula (9) gives the known recurrence
_
k
r
_
S(n, k) =
nr

m=kr
_
n
m
_
S(m, k r)S(n m, r).
The same recurrence holds for the unsigned Stirling numbers of rst kind.
The identity (9) can also be found in a recent paper by Cvijovic [6]. There the author
gives furthermore a recurrence relation for B
n,k
(see [6, equation (1.3)]) which, combined
with (10), leads to the interesting identity
n1

m=k1
_
1
k
_
n
m
_

_
n 1
m
__
x
nm
B
m,k1
(x) = 0. (11)
4 Inverse relations
For n N, b Z, C, and a sequence z = (z
1
, z
2
, . . . ), we dene
Q
n,b
(, z) =
n

k=1
_
+ bk
k 1
_
(k 1)!B
n,k
(z).
These functions appear naturally in certain compositions of formal power series. For
instance, if Z(t) = 1 +

n1
z
n
t
n
n!
, then
log(Z(t)) =

n1
L
n
t
n
n!
the electronic journal of combinatorics 19(4) (2012), #P34 9
with L
n
=

n
k=1
(1)
k1
(k 1)!B
n,k
(z) (logarithmic polynomials), see Section 3.5 in the
book by Comtet [5]. Moreover, for any complex number r,
Z(t)
r
= 1 +

n1
P
(r)
n
t
n
n!
with P
(r)
n
=

n
k=1
(r)
k
B
n,k
(z) (potential polynomials). Now, since (1)
k1
=
_
1
k1
_
and
since (r)
k
= r
_
r1
k1
_
(k 1)!, we can write
L
n
(z) = Q
n,0
(1, z) and P
(r)
n
(z) = rQ
n,0
(r 1, z). (12)
Lemma 13. For N we have
Q
n,0
(, z) = z
n
+

i=1
i
+ 1
n1

m=1
_
n
m
_
z
nm
Q
m,0
(i 1, z)
Proof. Using the identity (10), and since B
n,1
(z) = z
n
, we have
Q
n,0
(, z) z
n
=
n

k=2
_

k 1
_
(k 1)!
_
1
k
n1

m=k1
_
n
m
_
z
nm
B
m,k1
(z)
_
=
n1

k=1
n1

m=k
_

k
_
k!
k + 1
_
n
m
_
z
nm
B
m,k
(z)
=
n1

m=1
_
n
m
_
z
nm
m

k=1
k!
+ 1
_
+ 1
k + 1
_
B
m,k
(z)
=
n1

m=1
_
n
m
_
z
nm
m

k=1
k!
+ 1
_

i=1
_
i
k
__
B
m,k
(z)
=

i=1
i
+ 1
n1

m=1
_
n
m
_
z
nm
m

k=1
_
i 1
k 1
_
(k 1)!B
m,k
(z).
At last, replace the interior sum over k by Q
m,0
(i 1, z) and solve for Q
n,0
(, z).
The next lemma is straightforward.
Lemma 14. Given any sequence z = (z
1
, z
2
, . . . ), the following product formula holds:
Q
n
1
,b
1
(
1
, z) Q
n
2
,b
2
(
2
, z)
=
n
1
+n
2

k=2
n
2

=1
k!
_

1
+b
1
(k)+1
k
__

2
+b
2
+1

_
(
1
+ b
1
(k ) + 1)(
2
+ b
2
+ 1)
_
k

_B
n
1
,k
(z)B
n
2
,
(z).
the electronic journal of combinatorics 19(4) (2012), #P34 10
Theorem 15. Let a, b Z. Given x = (x
1
, x
2
, . . . ), dene y = (y
1
, y
2
, . . . ) by
y
n
= Q
n,b
(an, x) =
n

k=1
_
an + bk
k 1
_
(k 1)!B
n,k
(x)
for every n N. Then, for any C, we have
n

k=1
_

k 1
_
(k 1)!B
n,k
(y) =
n

k=1
_
+ an + bk
k 1
_
(k 1)!B
n,k
(x). (13)
In other words, for any ,
Q
n,0
(, y) = Q
n,b
( + an, x).
Proof. At rst, we assume to be a positive integer. As both sides of (13) are polynomials
in , the statement will then be valid for any C.
We proceed by induction in n. For n = 1,
Q
1,0
(, y) = B
1,1
(y) = y
1
= B
1,1
(x) = Q
1,b
( + a, x) for every .
Assume Q
m,0
(, y) = Q
m,b
( + am, x) for every 1 m < n and any N. By
Lemma 13,
Q
n,0
(, y) = y
n
+

i=1
i
+ 1
n1

m=1
_
n
m
_
y
nm
Q
m,0
(i 1, y).
Since m < n, we use the induction hypothesis on Q
m,0
(i 1, y) to obtain
Q
n,0
(, y) y
n
=

i=1
i
+ 1
n1

m=1
_
n
m
_
y
nm
Q
m,b
(i 1 + am, x)
=

i=1
i
+ 1
n1

m=1
_
n
m
_
Q
nm,b
(a(n m), x) Q
m,b
(i 1 + am, x),
which by Lemma 14 becomes
=

i=1
i
+ 1
n1

m=1
_
n
m
_
n

k=2
m

=1
k!
_
+1
k
__
+i

_
( + 1)( + i)
_
k

_B
nm,k
(x)B
m,
(x)
=

i=1
n

k=2
ik!
+ 1
k1

=1
n1

m=
_
+1
k
__
+i

__
n
m
_
( + 1)( + i)
_
k

_B
nm,k
(x)B
m,
(x)
with = an + bk and = am + b. Now, with = + 1 and = + i + 1, the
identity from Corollary 11 becomes
k

=0
n

m=
( + i + 1)
_
+1
k
__
+i

__
n
m
_
( + 1)( + i)
_
k

_ B
m,
B
nm,k
=
i + 1
i
_
+ i + 1
k
_
B
n,k
.
the electronic journal of combinatorics 19(4) (2012), #P34 11
Hence
k

=0
n

m=
_
+1
k
__
+i

__
n
m
_
( + 1)( + i)
_
k

_B
m,
B
nm,k
=
i + 1
( + i + 1)i
_
+ i + 1
k
_
B
n,k
,
and so the sum from = 1 to = k 1 (as needed above) is equal to
_
i + 1
( + i + 1)i
_
+ i + 1
k
_

1
+ i
_
+ i
k
_

1
( + 1)i
_
+ 1
k
__
B
n,k
(x).
Therefore,
Q
n,0
(, y) y
n
=

i=1
n

k=2
ik!
+ 1
_
(i + 1)
_
+i+1
k
_
( + 1 + i)i

_
+i
k
_
+ i

_
+1
k
_
( + 1)i
_
B
n,k
(x)
=
n

k=2
k!
+ 1

i=1
_
(i + 1)
_
+i+1
k
_
( + i + 1)

i
_
+i
k
_
+ i

_
+1
k
_
+ 1
_
B
n,k
(x)
=
n

k=2
k!
_
_
++1
k
_
( + + 1)

_
+1
k
_
+ 1
_
B
n,k
(x)
=
n

k=1
k!
_
1
k
_
+
k 1
_

1
k
_

k 1
__
B
n,k
(x)
=
n

k=1
(k 1)!
__
+ an + bk
k 1
_

_
an + bk
k 1
__
B
n,k
(x)
= Q
n,b
( + an, x) Q
n,b
(an, x).
Since y
n
= Q
n,b
(an, x), we get Q
n,0
(, y) = Q
n,b
( + an, x) as desired.
Remark 16. The identity (13) is also valid in the form
n

k=k
0
_

k k
0
_
(k 1)!B
n,k
(y) =
n

k=k
0
_
+ an + bk
k k
0
_
(k 1)!B
n,k
(x).
For instance, for k
0
= 2, this follows from the identity
_

k2
_
=
_
+1
k1
_

_

k1
_
together with
the fact that (13) holds for every .
Theorem 17. Let a, b Z, not both equal to 0. Given any sequence x = (x
1
, x
2
, . . . ), we
consider the sequence y = (y
1
, y
2
, . . . ) dened by
y
n
=
n

k=1
_
an + bk
k 1
_
(k 1)!B
n,k
(x). (14)
the electronic journal of combinatorics 19(4) (2012), #P34 12
Then, for every n N, we have
x
n
=
n

k=1
an + bk
an + b
_
an b
k 1
_
(k 1)!B
n,k
(y). (15)
Proof. For k > 1 we rewrite
an + bk
an + b
_
an b
k 1
_
=
_
an b
k 1
_
b
_
an b 1
k 2
_
and split the right-hand side of (15) as
n

k=1
_
an b
k 1
_
(k 1)!B
n,k
(y) b
n

k=2
_
an b 1
k 2
_
(k 1)!B
n,k
(y).
Using (13) with = an b we obtain
n

k=1
_
an b
k 1
_
(k 1)!B
n,k
(y) =
n

k=1
_
b(k 1)
k 1
_
(k 1)!B
n,k
(x)
= x
n
+
n

k=2
_
b(k 1)
k 1
_
(k 1)!B
n,k
(x).
Using again the identity (13) now with = an b 1, and starting at k = 2, we get
b
n

k=2
_
an b 1
k 2
_
(k 1)!B
n,k
(y) =
n

k=2
_
b(k 1)
k 1
_
(k 1)!B
n,k
(x).
The statement then follows by taking the dierence.
Remark 18. Note that equations (14) and (15) give an inverse relation: It is also true
that for any given sequence y = (y
1
, y
2
, . . . ), the sequence x = (x
1
, x
2
, . . . ) dened by (15)
satises the relation (14).
Example 19. For b = 1, Theorem 17 gives the inverse relation
y
n
=
n

k=1
_
an + k
k 1
_
(k 1)!B
n,k
(x),
x
n
=
n

k=1
_
an 2
k 1
_
(k 1)!B
n,k
(y),
which can be found in [5, Theorem F, p. 151]. In particular, if x is replaced by x, we
arrive at the symmetric relation (cf. [7, Theorem 10]):
y
n
=
n

k=1
(1)
k
_
an + k
k 1
_
(k 1)!B
n,k
(x),
x
n
=
n

k=1
(1)
k
_
an + k
k 1
_
(k 1)!B
n,k
(y).
the electronic journal of combinatorics 19(4) (2012), #P34 13
Example 20 (Generating function). Consider the formal power series
Y (t) = 1 +

n1
y
n
t
n
n!
with y
n
as in (14) for any given sequence x = (x
1
, x
2
, . . . ).
As mentioned at the beginning of the section, the logarithmic polynomials associated
with Y (t) are given by Q
n,0
(1, y), see (12). By means of Theorem 15, these polynomials
can then be written in terms of x, namely L
n
(y) = Q
n,b
(1 + an, x), and so
log
_
Y (t)
_
=

n1
Q
n,b
(1 + an, x)
t
n
n!
.
Similarly, for any r, P
(r)
n
(y) = rQ
n,0
(r 1, y) = rQ
n,b
(r 1 + an, x), thus
Y (t)
r
= 1 +

n1
rQ
n,b
(r 1 + an, x)
t
n
n!
.
In particular, for any polynomial F(z) =

m
=0
c

, we have
F(Y (t)) = F(1) +

n1
_
m

=1
c

Q
n,b
( 1 + an, x)
_
t
n
n!
.
References
[1] E.T. Bell. Exponential polynomials. Ann. of Math., 35:258277, 1934.
[2] C.A. Charalambides. Enumerative Combinatorics. Chapman and Hall/CRC, Boca
Raton, 2002.
[3] W.S. Chou, L.C. Hsu, P.J.S. Shiue. Application of Fa`a di Brunos formula in char-
acterization of inverse relations. J. Comput. Appl. Math., 190:151169, 2006.
[4] W. Chu. Elementary proofs for convolution identities of Abel and HagenRothe.
Electron. J. Combin., 17(1): Note 24, 2010.
[5] L. Comtet. Advanced Combinatorics: The Art of Finite and Innite Expansions. D.
Reidel Publishing Co., Dordrecht, 1974.
[6] D. Cvijovic. New identities for the partial Bell polynomials. Appl. Math. Lett.,
24(9):15441547, 2011.
[7] M. Mihoubi. Partial Bell polynomials and inverse relations. J. Integer Seq., 13(4):
Article 10.4.5, 2010.
[8] W. Wang and T. Wang. General identities on Bell polynomials. Comput. Math.
Appl., 58(1):104118, 2009.
the electronic journal of combinatorics 19(4) (2012), #P34 14