# Newton's cubic Math Modeling

I need to find the "black points" (where the function becomes undefined) of an equation x^3 -1 using newtons method. So the function turns out to be F(x) = x - (x^3-1)/(3x^2). Is there a way to find a majority of the points (where x is undefined as in the Newton cubic fractal photo aka the black points) using Sage?

edit retag close merge delete

Can you be a little more specific? x^3-1 is defined for all complex x, so I assume you are referring to something else.

( 2011-12-13 02:35:58 -0500 )edit

Sort by » oldest newest most voted

I don't think this is a built-in behavior. I'd be interested in a reference for an algorithm, even! Basically, this would just be 0, anything whose tangent line sends it to zero, anything whose tangent line sends it to one of those, etc., right?

If you just wanted to visualize this, you could try using some of the ideas at this ticket about Newton basins, though it wouldn't give the "black points".

For a graphical way of exploring this, I would search sagenb.org for implementations of Newton's method in interactive mode.

You could use solve to try to get some of these points, though I'm not sure how useful this would be.

sage: LA(x) = x - (x^3-1)/(3*x^2)
sage: LA(0)
---------------------------------------------------------------------------
RuntimeError: power::eval(): division by zero
sage: solve(LA==0,x)
[x == 1/4*(I*(-1)^(1/3)*sqrt(3) - (-1)^(1/3))*2^(2/3), x == 1/4*(-I*(-1)^(1/3)*sqrt(3) - (-1)^(1/3))*2^(2/3), x == 1/2*(-1)^(1/3)*2^(2/3)]
sage: S = solve(LA==0,x)
sage: for s in S:
solve(LA == s.rhs(),x)
....:
[x == -1/576*(-18*I*sqrt(3) - 18)*(-I*sqrt(3) + 1)*2^(1/3)/(1/8*sqrt(6) - 5/16)^(1/3) + 1/24*(3*I*sqrt(3) - 3)*(-1)^(1/3)*2^(2/3) - 1/2*(1/8*sqrt(6) - 5/16)^(1/3)*(I*sqrt(3) + 1), x == -1/576*(-18*I*sqrt(3) - 18)*(I*sqrt(3) + 1)*2^(1/3)/(1/8*sqrt(6) - 5/16)^(1/3) + 1/24*(3*I*sqrt(3) - 3)*(-1)^(1/3)*2^(2/3) - 1/2*(1/8*sqrt(6) - 5/16)^(1/3)*(-I*sqrt(3) + 1), x == 1/24*(3*I*sqrt(3) - 3)*(-1)^(1/3)*2^(2/3) + 1/288*(-18*I*sqrt(3) - 18)*2^(1/3)/(1/8*sqrt(6) - 5/16)^(1/3) + (1/8*sqrt(6) - 5/16)^(1/3)]
[x == -1/576*(-I*sqrt(3) + 1)*(18*I*sqrt(3) - 18)*2^(1/3)/(1/8*sqrt(6) - 5/16)^(1/3) - 1/24*(3*I*sqrt(3) + 3)*(-1)^(1/3)*2^(2/3) - 1/2*(1/8*sqrt(6) - 5/16)^(1/3)*(I*sqrt(3) + 1), x == -1/576*(I*sqrt(3) + 1)*(18*I*sqrt(3) - 18)*2^(1/3)/(1/8*sqrt(6) - 5/16)^(1/3) - 1/24*(3*I*sqrt(3) + 3)*(-1)^(1/3)*2^(2/3) - 1/2*(1/8*sqrt(6) - 5/16)^(1/3)*(-I*sqrt(3) + 1), x == -1/24*(3*I*sqrt(3) + 3)*(-1)^(1/3)*2^(2/3) + 1/288*(18*I*sqrt(3) - 18)*2^(1/3)/(1/8*sqrt(6) - 5/16)^(1/3) + (1/8*sqrt(6) - 5/16)^(1/3)]
[x == -1/2*(I*sqrt(3) + 1)*(1/8*sqrt(2)*sqrt(3) - 5/16)^(1/3) - 1/16*(-I*sqrt(3) + 1)*2 ...
more

thanks. i appreciate your help. I'm still looking for an algorithm but cant seem to come up with anything!

( 2011-12-13 13:23:42 -0500 )edit
1

You may find http://en.wikipedia.org/wiki/Newton_fractal useful as well - basically, you're looking for a Julia set, and you might find algorithms for finding those out there.

( 2011-12-13 14:34:21 -0500 )edit