5
$\begingroup$

Suppose I chose two rational functions, say,

$$u = \frac{t(4+t)^5}{(1+4t)^5}, \qquad v = \frac{t^5(4+t)}{(1+4t)}.$$

Then I know that $K(X) = \mathbf{C}(u,v)$ is the function field of the projective line (Proof: If $K(Y) = \mathbf{C}(t)$, then there is an inclusion $K(X) \subseteq K(Y)$ and hence a surjection $\mathbf{P}^1 \simeq Y \rightarrow X$, and so $X$ must have genus $0$). From this it follows that $K(X) \simeq \mathbf{C}(s)$ for some $s \in \mathbf{C}(t)$.

Is there a practical easy algorithm to explicity construct such an $s$ and, moreover, write $s$ as a rational function of $u$ and $v$? Is there an easy way at least to determine the degree of the map $Y \rightarrow X$? (EDIT: There's some ambiguity here: I mean $X$ and $Y$ to be the unique smooth curves (i.e. $\mathbf{P}^1$) with function fields $K(X)$ and $K(Y)$).

In case you were wondering, the specific choice of $u$ and $v$ where motivated by the question:

Deciding whether a given power series is modular or not

(This question was posted on math.stackexchange a week ago; I am cross posting it here because it did not receive any replies: https://math.stackexchange.com/questions/17960/function-field-of-the-projective-line)

$\endgroup$
0

3 Answers 3

3
$\begingroup$

The usual algebraic proof of L\"uroth's theorem gives the following procedure for finding a single generator of the subfield $L= K(u_1,\dotsc,u_r)$ of $K(t)$: let $a$ be any non-constant coefficient of the minimal monic polynomial of $t$ over $L$. Then $K(a) = L$.

Perhaps one can concoct a fast algorithm to compute these sorts of things using linear algebra. A lazy-person's way to do it would be to use Groebner bases. For example, in the case in question, consider the ring $R = \mathbf{C}[s,t,u,v]$ equipped with the lexicographic order where s>t>u>v. (Funny to call that "lexicographic.") Let $J$ be the ideal $$(s(1+4t)-1, u - t(4+t)^5s^5, v-t^5(4+t)s) $$ of $R$. If you compute a reduced Groebner basis for $J$, you will find elements of degrees (in the order $(s,t,u,v)$) $$ (1,1,5,23), (0,2,5,23), (0,1,5,23), (0,1,5,24), (0,0,6,6), $$ whose coefficients are too enormous for me to include here. (Despite the enormous coefficients, the calculation took only a fraction of a second on my run-of-the-mill laptop.) The element of degree $(0,0,6,6)$ is the relation between $u$ and $v$ that Joe Silverman calculated using resolvents. Each of the third and fourth elements provides a way to write $t$ as a rational function of $u$ and $v$.

In the general setting (of $L=K(u_1,\dotsc,u_r)$ in $K(t)$), if you perform a similar Groebner-basis calculation, you will find the minimal polynomial (but not in monic form) of $t$ over your subfield $L$ among the elements of your Groebner basis. Dividing that relation by the coefficient of the highest power of $t$, you get the minimal monic polynomial of $t$ over $L$. Take a non-constant coefficient $a$ of this monic polynomial, and you will have $K(a) = L$. A similar Groebner-basis calculation will provide expressions for the $u_i$ in terms of $a$. Thoughtful examination of the proof of L\"uroth's theorem may allow one to speed up the calculation (and to state things in a less roundabout way). To see how this goes in practice, you could try the simple case $u=t^4, v=t^6$.

$\endgroup$
1
  • $\begingroup$ Thank you, this worked. The equations did indeed have very large coefficients. $\endgroup$
    – Philoi
    Jan 26, 2011 at 17:52
4
$\begingroup$

Dan's right, the inclusion $K(Y)\subset K(X)$ would give a map $X\to Y$. However, I think you meant to write $K(X)\subset K(Y)$, which is easier to see is the correct formulation if you write it as $K(u,v)\subset K(t)$. Also, you meant to say $s$ as a rational function of $u$ and $v$, not of $X$ and $Y$.

You can get an equation (probably singular) for $X$ by computing the resultant of $(1+4t)^5u-t(4+t)^5$ and $(1+4t)v-t^5(4+t)$ with respect to the variable $t$. This gives an equation in $u$ and $v$. After dividing by a common factor, I get that $X$ is given by the following rather horrible affine equation in the $(u,v)$-plane: $$\eqalign{ -u^6 &+ (281474976710656 v^5 - 24739011624960 v^4 + 685517045760 v^3 \cr &- 6305218560 v^2 + 11887110 v) u^5 + (-24739011624960 v^5\cr &+ 301150356111360 v^4 - 292164004085760 v^3 - 7564564531215 v^2\cr &- 6305218560 v) u^4 + (685517045760 v^5 - 292164004085760 v^4 \cr &+ 114773655178260 v^3 - 292164004085760 v^2 + 685517045760 v) u^3 \cr &+ (-6305218560 v^5 - 7564564531215 v^4 - 292164004085760 v^3 \cr &+ 301150356111360 v^2 - 24739011624960 v) u^2 + (11887110 v^5 \cr &- 6305218560 v^4 + 685517045760 v^3 - 24739011624960 v^2 \cr &+ 281474976710656 v) u - v^6 = 0. } $$ Then $t\to(u(t),v(t))$ gives the map from $\mathbb{P}^1=Y$ to $X$. Someone who is more adept than I am using computer algebra systems can probably figure out the degree of this map, and write $k(u(t),v(t))$ as $k(s(u(t),v(t)))$ for an explicit $s(u,v)$.

$\endgroup$
0
3
$\begingroup$

In the specific example, $k(u)\subset k(u,v)\subset k(t)$ and the degree $[k(t):k(u)]=6$. Joe Silverman's equation shows that $[k(u,v):k(u)]=6$, so $k(u,v)=k(t)$ and you can take $s=t$.

$\endgroup$

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.