Formula for A120014(n) -- Simplified
Paul D. Hanna
pauldhanna at juno.com
Sat Jun 10 03:52:54 CEST 2006
Max,
Thank you for the alternate approach. Starting with the same
formula: a(n) = [x^n] { xC/(1-nxC) - [xC/1-nxC]^2 }
(where C is Catalan function), I see that you
collected terms involving the same Catalan coefficients,
while I collected terms involving the same powers of n.
> > a(n) = n^(n-1) - Sum_{k=1..n-1}
> > n^(k-1)*k*(k-1)*(n-k-1)*(2*n-k-2)!/(n-k)!/n!
>
The goal of my approach here was to separate n from k
as much as possible, hoping it would help in isolating a g.f.
My approach leads to the formula for the
table of self-compositions of A120009:
T(n,k) = [x^k] { xC/(1-nxC) - [xC/1-nxC]^2 }
to be
T(n,k) = n^(k-1) - Sum_{j=2..k-2}
n^(j-1)*j*(j-1)*(k-j-1)*(2k-j-2)!/(k-j)!/k!
where the main diagonal is A120014(n) = T(n,n).
> I've slightly different and somewhat simpler formula:
>
> a(n) = Sum_{k=1..n} n^(k-3)*(n-k+1)*k*C(2*n-k-1,n-1)
>
I concede, your approach yields simpler formulas; it leads to
the formula for the table of self-compositions of A120009
to be:
T(n,k) = Sum_{j=1..k} n^(j-2)*(n-j+1)*j*(2k-j-1)!/(k-j)!/k!
which is preferred over the formula for T(n,k) above.
> implying that a(n) is the coefficient of x^n in the expansion of
>
> (x*(1-x+x*n)) / (n^2 * (1-n*x)^2 * (1-x)^(n+1))
>
Max, this is a nice result - good work!
> > Thus, the e.g.f. of A120014 does involve the LambertW function as
> > suspected
>
> It still does not sound to me as a strong evidence.
>
Well, consider e.g.f.:
Sum_{n>=1} a(n)*x^n/n! = Sum_{n>=1} n^(n-1)*x^n/n!
- Sum_{n>=1} Sum_{k=1..n-1}
n^(k-1)*k*(k-1)*(n-k-1)*(2*n-k-2)!/(n-k)!*x^n/(n!)^2
Not very pretty - but LambertW clearly is found in the first sum on the
right.
Yet the ugly double sum may be subtracting it back out ...
So, no, I do not think there is much hope of finding e.g.f. using this
formula!
> > (I am still working on obtaining that e.g.f.).
> Any luck so far?
>
Not yet, and it may require someone more clever than I.
But I will continue to try.
Thanks for sharing your nice results!
I will credit you in any new submissions using your formulas.
Paul
More information about the SeqFan
mailing list