I want to prove the following:

Let $R,S$ be two Riemann surfaces, and suppose Green's function $g_S$ exists for $S$. Let $f: R \to S$ be a nonconstant analytic function. Prove that Green's function $g_R$ exists for $R$, and $g_R(p,q) \leq g_S(f(p),f(q))$ for all $p,q \in R$.

This is an exercise in Gamelin's book, XVI.3.2. The exercise does not forbid $f$ being constant, but it seems evidently necessary.

Some thoughts

Let $(U, z)$ be a coordinate patch of $R$ containing $q$ with $z(q) = 0$, let $v$ be a subharmonic function in $R \setminus\{q\}$ with compact support such that $$\limsup_{x \to q} v(x)+\log |z(x)|< \infty,$$ then by definition $g_R(\cdot,q)$ is the supremum of all such $v$, so it is enough to prove that $v(p) \leq g_S(f(p), f(q))$.

Letting $(V,w)$ be a coordinate chart around $f(q)$ in $S$ with $w(f(q)) = 0$, we know that $$\lim \limits_{x \to q} g_S(f(x),f(q))+\log |w(f(x))|$$ exists, as the function is even harmonic there.

Subtracting, we get that $$\limsup \limits_{x \to q} \;v(x)- g_S(f(x),f(q))+\log \lvert \frac{z(x)}{w(f(x))}\rvert < \infty.$$

If $w \circ f$ is a coordinate function, which is not always true, then $\limsup _{x \to q} \;v(x)- g_S(f(x),f(q)) < \infty$ so for every $\epsilon>0$, the function $\varphi_\epsilon(x) := v(x)- (1+\epsilon)g_S(f(x),f(q))$ extends to be subharmonic at all $R$. It is $\leq 0$ off a compact set, the support of $v$, so the maximum principle forces $\varphi_\epsilon \leq 0$. Letting $\epsilon \to 0$ we are done.

This question is a slightly more quantitative version of the fact, of which I do not know a proof, that an analytic map from a parabolic Riemann surface to a hyperbolic Riemann surface is constant. The main tool in proofs on this topic seems to be the maximum principle; how can it be leveraged here without adding an assumption on $f$?


A friend solved it as follows. The composition $w \circ f\circ z^{-1}$ is an automorphism of the unit disk fixing the origin, so by Schwarz's lemma $|w(f(x)) |\leq |z(x)|$, whence $$\log |\frac{z(x)}{w(f(x))}| \geq 0$$.

Combining this with the claim in the post that $$\limsup \limits_{x \to q} \;v(x)- g_S(f(x),f(q))+\log \lvert \frac{z(x)}{w(f(x))}\rvert < \infty,$$

we obtain

$$\limsup \limits_{x \to q} \;v(x)- g_S(f(x),f(q)) < \infty$$

and then the method of proof in the post works without adding assumptions. $\Box$

(I guess the bounty will go to whoever will prove it in another way.)


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.