enter image description here

The idea is through conformal transformations satisfying the conditions requested of the problem make this an easier problem to deal,but i don't know which be this transformation. I would appreciate any hint how to solve this. thank you very much


Looking at my copy of Churchill, Complex Variables with Applications, the appendix on conformal transformations suggests that applying

$$ w = f(z) = \log \left(\frac{z-1}{z+1} \right) $$

will transform the region shown (for the case $a = 1$) into a strip in the $w$-plane, where the $V = 1$ rays are mapped to $\Im(w) = \pi$ and the $V = 0$ segment is mapped to $\Im(w) = 0$.

If you take $g(z) = \frac{1}{\pi} f(z)$, you map this region to one in which the $V = 1$ segment maps to $\Im(w) = 1$, and the $V = 0$ segment maps to $\Im(w) = 0$.

I seem to recall that this is what's needed in this kind of problem, but it's as far as my shoddy memory takes me. (The inverse conformal map for $f$, by the way, is $z = -\coth \frac{w}{2}$.)

OK. I went and read the relevant chapter of Churchill ... and found this problem in the exercises!

I think that having come up with the transformation, and the potential in the $w$-plane, I'm going to leave the conversion back to a potential in the $z$-plane up to you, although I confess I don't see offhand how to get from the $\log$-expression back to something involving $\arctan$.

In comments, I've been asked to add some pics to show off what the log-transform is doing. Let's look at this diagram:The point $z$ in the upper half plane, relative to $-1$ and $+1$

Now the "log" that I've used above is really $$ w = \log |\frac{z-1}{z+1}| + i \arg \frac{z-1}{z+1}. $$

Let's look at how that gets computed. First, the first ratio is just the ratio of distances to the two point $+1$ and $-1$; I'll come back to that. The second term -- the imaginary part, is the difference of the args of $z-1$ and $z - (-1)$, i.e., the angle labeled $\theta_1 - \theta_2$ in the diagram.

What's that look like on (or very near) the real axis? Well, for $z > 1$, $\theta_1$ and $\theta_2$ will both be nearly zero; for $z < -1$, they'll both be nearly $\pi$. For $-1 < z < 1$, we have $\theta_2 \sim 0$ and $\theta_1 \sim \pi$, so the difference is approximately $\pi$. In other words, the red segments of the real axis will be sent to $v = 0$ in the $w$-plane, but the blue segment will be sent to $v = \pi$. Note that the remote ends of the red segments (near $\pm \infty$) will be sent to $u = 0$, while the near ends (near $\pm 1$) are sent off to infinity.

What about the ratio of lengths? Well, along the $x = 0$ line, we'll have $r_1 = r_2$, so that will map to $u = 0$. The other $u = constant$ lines will become hyperbolic-like curves in the $xy$-plane.

Finally, the transform we really want is $$ w = \frac{1}{\pi}\log |\frac{z-1}{z+1}| + \frac{i}{\pi} \arg \frac{z-1}{z+1}, $$ which sends the blue segment to the $v = 1$ line in the $w$-plane.

Now since the imaginary part of $w$ is $v$, and we can substitute $x+iy$ for $z$, we get \begin{align} v &= \frac{1}{\pi} \arg \frac{z-1}{z+1} \\ &= \frac{1}{\pi} \arg \frac{(x-1) + iy}{(x+1) + iy} \\ &= \frac{1}{\pi} \arg \frac{x^2 + y^2 - 1 + i 2y}{(x+1)^2 + y^2} \\ &= \frac{1}{\pi} \arg {x^2 + y^2 - 1 + i 2y}\\ &= \frac{1}{\pi} \arctan \frac{2y}{x^2 + y^2 - 1}, \end{align} which is an expression for the potential at point $(x, y)$.

Finally, I'm going to show off the effect of the conformal transformation by showing an image in the $z$-plane, and the transformed image in the $w$-plane, courtesy of Matlab. The problem is that the points near $z = \pm 1$ get sent to infinity, so parts of the transformed image will be missing.

The starting image (which extends from $-2.5$ to $2.5$ in $x$, and from $0.05$ to $2.0$ in $y$ (with $y = 0$ near the top, because of Matlab's idiosyncracies!) shows a bunch of peppers: enter image description here

The transformed image, extending from -10 to 10 in $u$ and $0$ to $1$ in $v$, is this: enter image description here

If you can make sense of those, then bless you. I just followed the description given here: http://www.mathworks.com/help/images/examples/exploring-a-conformal-mapping.html to make the images. Here's my code:

function confImage

close all
A = imread('peppers.png');
A = A(31:330, 1:500, :);
figure, imshow(A)
title ('Original Image', 'FontSize', 14);

conformal = maketform('custom', 2, 2, [], @conformalInverse, []);

uData = [ -10   10];  % Bounds for REAL(w)
vData = [  0  1.0];  % Bounds for IMAG(w)
xData = [ -2.5    2.5 ];  % Bounds for REAL(z)
yData = [  0.05   2.0 ];  % Bounds for IMAG(z)

B = imtransform( A, conformal, 'cubic', ...
                'UData', uData,'VData', vData,...
                'XData', xData,'YData', yData,...
                'Size', [300 360], 'FillValues', 255 );
figure, imshow(B)
title('Transformed Image','FontSize',14)

function U = conformalInverse(X, t)

Z = complex(X(:,1),X(:,2));
W = (Z + 1./Z)/2;
U(:,2) = (1/pi) * angle((Z - 1) ./ (Z + 1))
U(:,1) = (1/pi)* log( abs((Z - 1) ./ (Z + 1)));  %real(W);
  • 1
    $\begingroup$ Should be $$\log \left(\frac{z-a}{z+a}\right),$$ I think. $\endgroup$ – Daniel Fischer Feb 10 '14 at 21:55
  • $\begingroup$ Yes. (I specified that I was handling the case $a = 1$, but this is the more general case.) $\endgroup$ – John Hughes Feb 10 '14 at 22:10
  • $\begingroup$ Ah, I overlooked that parenthesis. $\endgroup$ – Daniel Fischer Feb 10 '14 at 22:11
  • $\begingroup$ Hmm, a few pictures to see what's happening would be nice .. $\endgroup$ – Han de Bruijn Feb 11 '14 at 11:21
  • $\begingroup$ Done. By the way, the explanation is copied almost verbatim from section 82 of Churchill (the OP's textbook), "Steady Temperatures in a Half Plane", so if s/he wants to read it in more detail, it's there. $\endgroup$ – John Hughes Feb 11 '14 at 21:37

For completeness of the record, I've decided to formulate my previously given comment as an answer. Yes, the mapping in John's answer is a Schwarz-Christoffel mapping :

enter image description here
enter image description here
In general we have: $$\frac{dw}{dz}=K(z-a)^{\alpha/\pi-1}(z-b)^{\beta/\pi-1}$$ With $a=-1$ , $b=+1$ and $\alpha=\beta=0$ , giving: $$dw\;=\;K(z-1)^{-1}(z+1)^{-1}dz \;=\;\frac{K}{2}\left(\frac{dz}{z-1}-\frac{dz}{z+1}\right)$$ So indeed, integration gives apart from constants that can be determined (later): $$w=\log\left(\frac{z-1}{z+1}\right)$$


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.