3
$\begingroup$

I am trying to validate that the Schwartz-Christoffel mapping does indeed take the upper-half plane $\mathbb{H} = \{ z: \mathrm{Re}(z) > 0\}$ to a polygon.

This involves integrals of functions of the type and computing at every point.

$$ f(z) = \int_0^z \frac{dx}{\sqrt{(x-a)(x-b)}}$$

where $a < b \in \mathbb{R}$.

I would like the case $a = 1,b=-1$. I know this integral can be computed exactly, but not if the denominator $P(z)$ were a cubic or quartic like $x^4 - 1$.

Here is a numpy script I wrote... sorry if it's wrong.

import numpy as np
impott matplotlib.pyplot as plt

f = lambda z: np.cumsum(z, axis=1) + 1j*np.cumsum(z, axis=0)    
n = 200
dt = 1.0/n
y = (np.ones((n,2*n)))/n
y = f(y)-1
w = y[(y.imag > 0.001)] # semicircle of radius 1

g = lambda t: np.sqrt(1 - t**2)**-0.5 # function to be integrated

plt.plot((g(w)).real, g(w).imag, 'b.')

plt.axis("Equal")
plt.show()

Here's the picture I came up with... defintely not a polygon. Maybe I have to change the original mesh?

enter image description here


I don't want to compute the antiderivative exactly, since I would like it to work for more complex examples like $\int_0^z \frac{dx}{\sqrt{(x-1)(x-2)(x-3)}}$. Instead I proposed two strategy:

  • build a mesh representing values of $\frac{1}{\sqrt{P(z)}}$
  • transform the mesh into the values of $\int_0^z$ at each point
  • if the roots of $P(z)$ are real and we stick to $\mathrm{Re}(z)> 0$ the square-root is still single-valued at least
$\endgroup$
  • $\begingroup$ Alternatively can think of the differential equation $\frac{dw}{dz} = \tfrac{1}{\sqrt{P(z)}}$. $\endgroup$ – john mangual Nov 11 '14 at 20:50
  • 1
    $\begingroup$ Where exactly are you integrating the function? I can see $1/\sqrt{1-w^2}$ being plotted. $\endgroup$ – Kirill Nov 11 '14 at 23:24
  • $\begingroup$ @Kirill yeah it loos like I forget the cumulative sum. it doesn't really compute the "integral" anyway $\endgroup$ – john mangual Nov 12 '14 at 11:07
2
$\begingroup$

Note that the integral is $$ \int^z \frac{dx}{\sqrt{x-a}\sqrt{x-b}}, $$ and in general for complex variables $$ \sqrt{x-a}\sqrt{x-b} \not\equiv \sqrt{(x-a)(x-b)}. $$

I'm not sure what you were trying to do with your code, so I can't help you there. I don't see where in your code you do the integration, it seems to plot some other function entirely: the issue is not with the mesh. When I did your example in Mathematica, $a=-b=1$, and I got this:

-((I \[Pi])/2) + 2 ArcSinh[Sqrt[-1 + z]/Sqrt[2]] /. z -> x + I y;
ParametricPlot[{Re@%, Im@%}, {x, -8, 8}, {y, 0, 8}, PlotRange -> {{-1, 4}, {-2, 2}}]

enter image description here

$\endgroup$
  • $\begingroup$ Yes... in this case Mathematica computes the anti-derivative exactly. $\endgroup$ – john mangual Nov 12 '14 at 11:07

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

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