There was a question about a particular case of this, Quadratic differentials; seemingly it contained too little information, so let me try again. This will be also a second take on my previous question Getting rid of discontinuities in plots caused by square roots, logarithms, `Arg`, etc focussed on the particular case of square roots.

A quadratic differential is simply an expression of the form $f(z)dz^2$, where $f(z)$ is (for my purposes) some meromorphic function. For example, it can be a rational function, although I in fact need things like products of the Weierstrass function or some other elliptic functions.

A trajectory of such differential would be a standard solution of a differential equation if there would be $dz$ instead of $dz^2$: a (piecewise) smooth parametric curve $(x(t),y(t))$ that at every point $z$ it passes forms the same constant angle (counted always in the same, say counterclockwise direction) with the vector $(\operatorname{Re}(f(z)),\operatorname{Im}(f(z)))$. We could then obtain it by solving, for any given angle $\alpha$, the system \begin{align*} x'(t)&=\operatorname{Re}(e^{i\alpha}f(x+iy))\\ y'(t)&=\operatorname{Im}(e^{i\alpha}f(x+iy)) \end{align*} with, say, NDSolve, or visualize it using StreamPlot.

However the term $dz^2$ means that it must form constant angle with the vector $\pm\sqrt{f(z)}$ rather than $f(z)$. And since this vector now has two possible (opposite) directions, the condition now becomes that the curve forms either angle $\alpha$ or $\alpha+\pi$ with ReIm of $\sqrt{f(z)}$. This means we now have instead of a vector field, a field of tangent lines with unspecified direction.

There are easy examples showing that the resulting ambiguity cannot be avoided: with $f(z)=-z$ and $\alpha=\pi/2$, the stream plot for ReIm[I Sqrt[-(x+I y)]] is enter image description here

so that the positive horizontal half-axis gets conflicting directions. For more complicated $f(z)$ I failed to find any general consistent way to compute with Sqrt[f[z]].

Is there some alternative way? More specifically, given a smoothly varying family of lines $\ell_{x,y}$ passing through each point $(x,y)$, like this -enter image description here

how to calculate and draw curves tangent to $\ell_{x,y}$ at each point $(x,y)$ they pass through?


Your Answer

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

Browse other questions tagged or ask your own question.