3
$\begingroup$

Suppose we have a fixed function $f: \mathbb{R}^2 \rightarrow \mathbb{R}^2$ (sufficiently regular, say $C^\infty$ ). The question is: for which $f$ there exists a scalar function $g: \mathbb{R}^2 \rightarrow \mathbb{R}$ such that the product $fg = (f_1g, f_2g)$ is holomorphic (using the standard identificacion $\mathbb{C} \sim \mathbb{R}^2$)? Are there any sufficient or necessary conditions for $f$ in order to exist such a $g$? In other words, is there any "nice" description of the product-set of the set of holomorphic functions on the plane with the set of real (sufficiently regular) functions on the plane? (Of course, one can try to solve Cauchy-Riemann but I get a pair of transport PDEs $(f_1g)_x = (f_2g)_y$ and $(f_1g)_y = -(f_2g)_x$ and is not clear that there is a common solution to both of them). Any references would be very much appreciated

EDIT: As someone commented I am looking for $g\neq 0$, this value of $g$ gives a trivial solution

$\endgroup$
1
  • $\begingroup$ $g \equiv 0$ always works. Presumably you look for a nontrivial solution. $\endgroup$ May 8, 2020 at 18:30

1 Answer 1

3
$\begingroup$

First, if $fg$ were holomorphic and nontrivial, then $\{g = 0\}$ has to be discrete. This implies that $g$ has to be either $\geq 0$ or $\leq 0$. So we can assume WLOG $g \geq 0$.

Restricting away from its zero set, we can study the function $v = \ln g$. The Cauchy-Riemann relations becomes

$$ \begin{cases} f_1 v_x - f_2 v_y = (f_2)_y - (f_1)_x \\ f_2 v_x + f_1 v_y = - (f_1)_y - (f_2)_x \end{cases} $$

Unless $f_1 = f_2 = 0$ (note also that the set on which this happens is also discrete), the matrix $A_f = \begin{pmatrix} f_1 & -f_2 \\ f_2 & f_1\end{pmatrix}$ is invertible, and hence you can explicitly solve the expression for $v_x, v_y$ in terms of $f$ and its first derivatives.

To solve the "local" problem, all you are missing is the integrability conditions, which is $(v_x)_y = (v_y)_x$. If you write the vector

$$ V = A_f^{-1} \begin{pmatrix} (f_2)_y - (f_1)_x \\ - (f_1)_y - (f_2)_x \end{pmatrix} \tag{Vdef}$$

the integrability condition is $$ (V_1)_y = (V_2)_x. \tag{INT}$$

Summary

Starting from $f$, construct $V$ by (Vdef), which is well-defined away from the set $\{f = (0,0)\}$.

  1. If (INT) fails to hold, then there exists no solution to your problem.
  2. If $U$ is a simply connected domain on which (INT) holds, then there exists a function $v:U\to \mathbb{R}$ such that $V = \nabla v$. Then on $U$ you have $g = e^v$ solves your problem.

Additional necessary conditions

By Picard's Theorem, necessarily that every direction must be represented in the image of $f$ for there to be a solution $g$.

$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

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