Let (an)n be a bounded sequence of real numbers and x∈(0,1), we can then consider the series:
∞∑n=0anxn.
Let now d∈{2,3,…} we can then consider (∑∞n=0anxn)d. I would like to write this in the following form:
(∞∑n=0anxn)d=∞∑n=0bnxn,(1)
for certain bn.
From the Multinomial Theorem we find that it can be written as:
∑∑nkn=d(dk1,…)∞∏t=0(atxt)kt=∑∑nkn=d(dk1,…)∞∑v=0∑∑ttkt=v(∞∏t=0(at)kt)xv
we can now try to move all these sums to the left, this way we can write:
(∞∑n=0anxn)=∞∑v=0(∑∑nkn=d(dk1,…)∑∑ttkt=v(∞∏t=0aktt))xv
we have thus found the representation (1) with:
bv=∑∑nkn=d(dk1,…)∑∑ttkt=v(∞∏t=0aktt).
I was wondering if there isn't a more elegant representation of these bv?
Answer
I don't know about what you consider "elegant", but are you aware that Faà di Bruno's formula can be expressed in terms of (partial) Bell polynomials? (See also Charalambides's book.)
Applied to your problem, we have
(∑n≥0anxn)d=∑n≥0(d∑j=0d!(d−j)!ad−j0Bn,j(a1,2a2,…,n!an))xnn!
No comments:
Post a Comment