The context for this is series approximation for perturbed iterations of the Mandelbrot set with arbitrary power $P$ (aka multibrot).

Define an arbitrary precision reference orbit: $$Z_{n+1} = Z_n^P + C$$ Using lower case for low precision differences: $$Z_{n+1} + z_{n+1} = (Z_n+z_n)^P + (C + c)$$ so: $$z_{n+1} = \sum_{p = 1}^P \begin{pmatrix} P \\ p \end{pmatrix} Z_n^{P-p} z_n^p + c$$ Now approximate the difference in iterate $z_n$ by a power series in the difference in parameter $c$: $$z_n = \sum_{k=1}^\infty A_{P,k,n} c^k$$ Combining the previous two equations gives: $$\sum_{k=1}^\infty A_{P,k,n+1} c^k = \sum_{p=1}^P \begin{pmatrix} P \\ p \end{pmatrix} Z_n^{P-p} \left(\sum_{j=1}^\infty A_{P,j,n} c^j\right)^p + c$$ Equating coefficients of $c^m$ gives a collection of iteration formulas for $A_{P,m,n+1}$ in terms of the previous coefficients and $Z_n$. These iterations are defined implicitly by the previous equation. The question is:

Is there an explicit closed form expression, perhaps in terms of nested sums, for these iteration equations for arbitrary $P$?

For example, for $P=2$ a closed form equation is: $$A_{2,1,n+1} = 2 Z_n A_{2,1,n} + 1 \\ A_{2,m,n+1} = 2 Z_n A_{2,m,n} + \sum_{k=1}^{m-1} A_{2,k,n} A_{2,m-k,n}\quad (m \gt 1)$$

up vote 0 down vote accepted

I think the answer for arbitrary $P$ is "no". It is possible to generate the series recurrence for arbitrary fixed $P$, but the recurrence contains nested sums with depth $P-1$, which is hard to express in a closed form (that can be translated to computer code without too much effort). Some experiments with the Maxima computer algebra system:

$ maxima
(%i1) sumexpand: true$
(%i2) cauchysum: true$
(%i3) F(p, c, z) := z^p + c$
(%i4) D(p, c, z) := F(p, c + C, z + Z) - F(p, C, Z)$
(%i5) A(p) := niceindices(collectterms(expand(D(p,c,sum(a[i]*c^i,i,1,inf))),c))$
(%i6) tex(A(2))$

$$\sum_{j=2}^{\infty }{c^{j}\,\sum_{i=1}^{j-1}{a_{i}\,a_{j-i}}}+2\,Z \,\sum_{i=1}^{\infty }{c^{i}\,a_{i}}+c$$

(%i7) tex(A(3))$

$$\sum_{j=3}^{\infty }{c^{j}\,\sum_{{\it i_6}=1}^{j-2}{a_{{\it i_6}} \,\sum_{i=1}^{j-{\it i_6}-1}{a_{i}\,a_{j-{\it i_6}-i}}}}+3\,Z\, \sum_{j=2}^{\infty }{c^{j}\,\sum_{i=1}^{j-1}{a_{i}\,a_{j-i}}}+3\,Z^2 \,\sum_{i=1}^{\infty }{c^{i}\,a_{i}}+c$$

(%i8) tex(A(4))$

$$\sum_{k=4}^{\infty }{c^{k}\,\sum_{{\it i_{14}}=1}^{k-3}{a_{ {\it i_{14}}}\,\sum_{j=1}^{k-{\it i_{14}}-2}{a_{j}\,\sum_{i=1}^{k-j- {\it i_{14}}-1}{a_{i}\,a_{k-j-{\it i_{14}}-i}}}}}+4\,Z\,\sum_{j=3}^{ \infty }{c^{j}\,\sum_{{\it i_{18}}=1}^{j-2}{a_{{\it i_{18}}}\,\sum_{ i=1}^{j-{\it i_{18}}-1}{a_{i}\,a_{j-{\it i_{18}}-i}}}}+6\,Z^2\, \sum_{j=2}^{\infty }{c^{j}\,\sum_{i=1}^{j-1}{a_{i}\,a_{j-i}}}+4\,Z^3 \,\sum_{i=1}^{\infty }{c^{i}\,a_{i}}+c$$

These are not quite the desired form (have to collect a few things to get the coefficient of $c^j$) but are close enough for now. And it seems niceindices() fails... (Maxima 5.38.1 on Debian).

  • I implemented recursive code for arbitary $P$, but it doesn't scale efficiently - there are products that are calculated $P!$ times, corresponding to the permutations of the indices (the Maxima-generated sums have the same issue). $a_1 a_2 + a_2 a_1$ had better be calculated as $2 a_1 a_2$. – Claude Oct 10 '17 at 18:15

Your Answer

 

By clicking "Post Your Answer", you acknowledge that you have read our updated terms of service, privacy policy and cookie policy, and that your continued use of the website is subject to these policies.

Not the answer you're looking for? Browse other questions tagged or ask your own question.