# G.f. for C(q^n,n)? - Generalized

pauldhanna at juno.com pauldhanna at juno.com
Wed Jan 9 04:50:04 CET 2008

```SeqFans,
A little more on the generalization.
Introduce a function G(x) (independent of summation index n):

Let F(x) be any formal power series in x such that F(0)=1; then
(7) Sum_{n>=0} m^n * F(q^n*x)^b * log( G(x)*F(q^n*x) )^n / n!  =
Sum_{n>=0} x^n * G(x)^(m*q^n) * [y^n] F(y)^(m*q^n + b)

where [y^n] F(y) denotes the coefficient of y^n in F(y).

It seems counter-intuitive that the power of G would be m*2^n;
however, it can be deduced from (6) (in prior email below).

EXAMPLE of (7).
When applying (7) to A136554:
1,3,10,82,2304,232088,81639942,99425060368,421915147527984,
G.f.: A(x) = Sum_{n>=0} log( (1 + x)*(1 + 2^n*x) )^n / n!.

it reveals the nice formulas:

G.f.: A(x) = Sum_{n>=0} C(2^n,n) * x^n * (1+x)^(2^n).

a(n) = Sum_{k=0..n} C(2^k,k) * C(2^k,n-k).

So, from (7) we find a nice simplification for:
G.f.: A(x) = Sum_{n>=0} log( z*(1 + 2^n*x) )^n / n!
as:
G.f.: A(x) = Sum_{n>=0} C(2^n,n) * x^n * z^(2^n)

FURTHER YET ...
But here is where I get stumped.
Can the following sum be simplified into a form similar to (7)?

(8) Sum_{n>=0} log( F(p^n*x)*G(q^n*x) )^n / n!  = ?

This sum again returns an integer series in x when both
F(x) and G(x) are integer series in x for all integer p, q.

But simplifying (8) into a form like (7) when p not= q
is not straitforward (at least for me).

Perhaps someone can find a nice form for (8)?

EXAMPLE of (8).
Let p=2, q=3, then we have A136578:
G.f.: A(x) = Sum_{n>=0} log( (1 + 2^n*x)*(1 + 3^n*x) )^n / n!.
1,5,78,6527,3450452,12594729052,338284182093366,70004091118158663618,

But I have not found a nice formula for even this simple case ...
Paul

On Wed, 2 Jan 2008 21:50:02 -0500 pauldhanna at juno.com writes:
Seqfans,
The identity can be generalized further.
Let F(x) be any formal power series in x such that F(0)=1.
Then
(5) Sum_{n>=0} m^n * F(q^n*x)^b * log( F(q^n*x) )^n / n!  =
Sum_{n>=0} x^n * [y^n] F(y)^(m*q^n + b)

where [y^n] G(y) denotes the coefficient of y^n in G(y).

If we let F(x) = exp(x), then we have the nontrivial result

(6) Sum_{n>=0} m^n * q^(n^2) * exp(b*q^n*x) * x^n / n!  =
Sum_{n>=0} (m*q^n + b)^n * x^n / n!

which becomes trivial at b=0.
Example of (6): q=2, m=1, b=1:
exp(x) + 2*exp(2x) + 2^4*exp(4x)*x^2/2! + 2^9*exp(8x)*x^3/3! +...+
2^(n^2)*exp(2^n*x)/n! +...
= 1 + 3x + 5^2*x^2/2! + 9^3*x^3/3! + 17^4*x^4/4! +...+
(2^n+1)^n*x^n/n! +...

Example of (6): q=2, m=1, b=-1:
exp(-x) + 2*exp(-2x) + 2^4*exp(-4x)*x^2/2! + 2^9*exp(-8x)*x^3/3! +...+
2^(n^2)*exp(-2^n*x)/n! +...
= 1 + x + 3^2*x^2/2! + 7^3*x^3/3! + 15^4*x^4/4! +...+
(2^n-1)^n*x^n/n! +...

Example of (5): F(x) = 1+x+x^2, q=2, m=1, b=0:
Sum_{n>=0} log( 1 + 2^n*x + 2^(2n)*x^2 )^n / n!  =
Sum_{n>=0} T(2^n, n) * x^n
where T(2^n,n) = trinomial coefficient of x^n in (1+x+x^2)^(2^n).

And many other examples can be given.

So the question arises, does (5) have useful applications?
Does it offer a g.f. for some significant sequences in OEIS?
Paul

On Sun, 30 Dec 2007 21:03:39 -0500 pauldhanna at juno.com writes:
Seqfans,
Recall the identity:
(2) Sum_{n>=0} log(1 + q^n*x)^n/n!  =  Sum_{n>=0} C(q^n,n)*x^n.

From this, I found the more general statements:

(3) Sum_{n>=0} m^n*log(1 + q^n*x)^n/n!  =  Sum_{n>=0} C(m*q^n,n)*x^n.

(4) Sum_{n>=0} m^n * (1 + q^n*x)^b * log(1 + q^n*x)^n/n!  =
Sum_{n>=0} C(m*q^n + b, n)*x^n.

Identity (4) is very interesting ... I wonder if it leads to other
results?
It certainly can lead to many significant sequences!

What I would really like is for formula (4) to allow the g.f.
A(x,m,b) = Sum_{n>=0} C(m*q^n + b, n)*x^n
to be manipulated to solve some functional equation ...

Any ideas along these lines from anyone?
Paul
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://list.seqfan.eu/pipermail/seqfan/attachments/20080108/cdd1c07f/attachment.htm>
```