<ph f="cmr"/><ph f="cmbx">Short equations for the genus 2 covers of degree 3 of an elliptic curve</ph>

### Jan Christian Rohde

Universitat Duisburg-Essen, Fachbereich Mathematik, 45117 Essen, Germany E-mail address : jan.rohde@uni-essen.de

• Abstract. E. Kani [4has shown that the Hurwitz functor ${\mathcal{ℋ}}_{E/K,3}$  , which parameterizes the (normalized) genus 2 covers of degree 3 of one elliptic curve $E$  over a field $K$  , is representable. In this paper the moduli scheme ${H}_{E/k,3}$  and the universal family are explicitly calculated over an algebraically closed field $k$  and described by short equations.

Introduction

T. Shaska [7has given a long equation, which describes the genus 2 curves with covers of degree 3 onto elliptic curves over an algebraically closed field of characteristic 0.
One can obtain simpler equations for the genus 2 covers of degree 3 of one elliptic curve.
We use the theoretical framework, which has been introduced by E. Kani [4. Let $K$  be a field with $char\left(K\right)\ne 2,3$  . We consider a $K$  -scheme $S$  , an elliptic curve $E/K$  with zero point 0 and a relative genus 2 curve $C/S$  . A normalized genus 2 cover $f:C\to E×S={E}_{S}$  of degree 3 is a morphism of $S$  -schemes of degree 3 such that ${f}_{*}\left({W}_{C/S}\right)=2\cdot \left[{0}_{E/S}\right]+{E}_{S}\left[2\right]$  , where ${W}_{C/S}$  is the divisor of Weierstrass points and $\left[{0}_{E/S}\right]$  is the zero section of $E×S\to S$  .
Two genus 2 covers ${f}_{i}:{C}_{i}\to {E}_{S}$  are isomorphic if there is an isomorphism $\phi :{C}_{1}\to {C}_{2}$  such that ${f}_{1}={f}_{2}\circ \phi$  . One denotes by ${\mathcal{ℋ}}_{E/K,3}\left(S\right)$  the set of isomorphism classes of normalized genus 2 covers of degree 3 onto ${E}_{S}$  . The assignment $S\to {\mathcal{ℋ}}_{E/K,3}\left(S\right)$  yields a Hurwitz functor ${\mathcal{ℋ}}_{E/K,3}:{\underline{Sch}}_{/K}\to \underline{Sets},$  which is represented by a smooth and geometrically connected modular curve ${H}_{E/K,3}$  ([4, Theorem 1.1).
Let $k$  be an algebraically closed field with $char\left(k\right)\ne 2,3$  . Here ${H}_{E/k,3}$  and the universal family $\mathcal{C}\to {H}_{E/k,3}$  will be explicitly calculated. We construct and parameterize genus 2 covers of elliptic curves by suitable coverings $u:{\mathbb{P}}^{1}\to {\mathbb{P}}^{1}$  (Frey-Kani coverings) in Section 1. Section 2 treats the positions of the ramification points of the Frey-Kani coverings.
The pattern of these positions allows us to determine ${H}_{E/k,3}$  .
Gerhard Frey gave some hints for the presentation of this paper. Martin Möller helped to improve this paper. I would like to thank them, and Eckart Viehweg for all his time and effort spent in guiding me for my Diplomarbeit (”master thesis”), from which this paper is originated.

1 Construction of genus 2 covers

Let us fix some elliptic curve $E$  , which is given by the 4 different points $0,1,\lambda ,\infty \in {\mathbb{P}}^{1}$  , over an algebraically closed field $k$  with $char\left(k\right)\ne 2,3$  . We consider a normalized genus 2 cover $f:C\to E$  of degree 3. There exist covers $h:C\to {\mathbb{P}}^{1}$  and $i:E\to {\mathbb{P}}^{1}$  of degree 2 and there is a cover $v:{\mathbb{P}}^{1}\to {\mathbb{P}}^{1}$  of degree 3, which is called ”Frey-Kani covering”, (see [2, [5and [6) such that this diagram commutes:
 $\begin{array}{c}C---\to fEh↓i↓{\mathbb{P}}^{1}---\to v{\mathbb{P}}^{1}\end{array}$ (1.0.1)
Let the zero point of $E$  lie over $\infty$  , and 0, 1 and $\infty$  be ramification points of a cover $u:{\mathbb{P}}^{1}\to {\mathbb{P}}^{1}$  of degree 3. Assume that $u\left(0\right)=0$  , $u\left(1\right)=1$  and $u\left(\infty \right)=\lambda$  . Later we will see that almost all covers $u$  are Frey-Kani coverings. Let us first study $u$  . Then the results of the following calculation will be used for a calculation and parametrization of the normalized genus 2 covers $f:C\to E$  of degree 3. We have for some ${p}_{1},c\in k$  :
 $\begin{array}{c}u\left({x}_{0}:{x}_{1}\right)=\left(g\left({x}_{0},{x}_{1}\right):c\left({x}_{1}-{p}_{1}{x}_{0}\right){x}_{1}^{2}\right)\end{array}$ (1.0.2)
with $g\left({x}_{0},{x}_{1}\right)={x}_{1}^{3}+{g}_{2}{x}_{1}^{2}{x}_{0}+{g}_{1}{x}_{1}{x}_{0}^{2}+{g}_{0}{x}_{0}^{3}$  Therefore we get by $u\left(0:1\right)=\left(1:\lambda \right)$  , which implies that $\lambda =c$  , and by $u\left(1:1\right)=\left(1:1\right)$  :
 $\begin{array}{c}{p}_{1}=1-\frac{g\left(1,1\right)}{\lambda }\end{array}$ (1.0.3)
Now we want to consider the situation over the branch points 1 and $\lambda$  of $u$  . Therefore we define $\stackrel{`}{u}:=\left({x}_{0}:{x}_{1}-{x}_{0}\right)\circ u$  and $\stackrel{^}{u}:=\left({x}_{0}:{x}_{1}-\lambda {x}_{0}\right)\circ u$  . Then one has $\stackrel{`}{u}\left(0\right)=-1$  , $\stackrel{`}{u}\left(1\right)=0$  , and $\stackrel{`}{u}\left(\infty \right)=\lambda -1$  . We obtain for some ${p}_{2},d\in k$  : $\stackrel{`}{u}\left({x}_{0}:{x}_{1}\right)=\left(g\left({x}_{0},{x}_{1}\right):d\left({x}_{1}-{p}_{2}{x}_{0}\right)\left({x}_{1}-{x}_{0}{\right)}^{2}\right)$  Thus, we get by $\stackrel{`}{u}\left(0:1\right)=\left(1:\lambda -1\right)$  , which implies $\lambda -1=d$  , and by $\stackrel{`}{u}\left(1:0\right)=\left(1:-1\right)$  :
 $\begin{array}{c}{p}_{2}=\frac{{g}_{0}}{\lambda -1}\end{array}$ (1.0.4)
One has $\stackrel{^}{u}\left(0\right)=-\lambda$  , $\stackrel{^}{u}\left(1\right)=1-\lambda$  , and $\stackrel{^}{u}\left(\infty \right)=0$  . We obtain for some $e,{p}_{3}\in k$  : $\stackrel{^}{u}\left({x}_{0}:{x}_{1}\right)=\left(g\left({x}_{0},{x}_{1}\right):e\left({p}_{3}{x}_{1}-{x}_{0}\right){x}_{0}^{2}\right)$  Therefore we get by $\stackrel{^}{u}\left(1:0\right)=\left(1:-\lambda \right)$  , which implies that $\lambda {g}_{0}=e$  , and by $\stackrel{^}{u}\left(1:1\right)=\left(1:1-\lambda \right)$  :
 $\begin{array}{c}{p}_{3}=1+\frac{\left(1-\lambda \right)g\left(1,1\right)}{\lambda {g}_{0}}\end{array}$ (1.0.5)
By the definitions of $\stackrel{`}{u}$  and $\stackrel{^}{u}$  , and the preceding results, we get: $\left(\lambda -1\right)\left({x}_{1}-\frac{{g}_{0}}{\lambda -1}{x}_{0}\right)\left({x}_{1}-{x}_{0}{\right)}^{2}=\lambda \left({x}_{1}-\left(1-\frac{g\left(1,1\right)}{\lambda }\right){x}_{0}\right){x}_{1}^{2}-g\left({x}_{0},{x}_{1}\right),$  $\lambda {g}_{0}\left(\left(1+\frac{\left(1-\lambda \right)g\left(1,1\right)}{\lambda {g}_{0}}\right){x}_{1}-{x}_{0}\right){x}_{0}^{2}=\lambda \left({x}_{1}-\left(1-\frac{g\left(1,1\right)}{\lambda }\right){x}_{0}\right){x}_{1}^{2}-\lambda g\left({x}_{0},{x}_{1}\right)$  These equations of polynomials imply the following equations of coefficients of ${x}_{1}^{2}{x}_{0}$  :
 $\begin{array}{c}-2\left(\lambda -1\right)-{g}_{0}=g\left(1,1\right)-\lambda -{g}_{2},\end{array}$ (1.0.6)
 $\begin{array}{c}g\left(1,1\right)-\lambda -\lambda {g}_{2}=0\end{array}$ (1.0.7)
By  1.0.3  $\left(\right)$  and  1.0.4  $\left(\right)$  , one has the equations $g\left(1,1\right)=\lambda -\lambda {p}_{1}$  and ${g}_{0}=\lambda {p}_{2}-{p}_{2}$  . We substitute for $g\left(1,1\right)$  and ${g}_{0}$  in  1.0.6  $\left(\right)$  and  1.0.7  $\left(\right)$  and obtain:
 $\begin{array}{c}{p}_{1}=-{g}_{2}=-2\left(\lambda -1\right)+{p}_{2}-\lambda {p}_{2}+{p}_{1}\lambda \end{array}$ (1.0.8)
$⇒\left(\lambda -1\right)\left(-2-{p}_{2}+{p}_{1}\right)=0$  One can divide by $\lambda -1$  , because we have $\lambda \ne 1$  . Thus, we have:
 $\begin{array}{c}-2-{p}_{2}+{p}_{1}=0⇔{p}_{2}={p}_{1}-2\end{array}$ (1.0.9)
One obtains by ( 1.0.3    ), ( 1.0.4    ) and ( 1.0.5    ):
 $\begin{array}{c}{p}_{3}=1+\frac{\left(1-\lambda \right)g\left(1,1\right)}{\lambda {g}_{0}}=1-\frac{1-{p}_{1}}{{p}_{1}-2}=\frac{2{p}_{1}-3}{{p}_{1}-2}\end{array}$ (1.0.10)
The equations ${p}_{2}={p}_{1}-2$  and ${p}_{2}=\frac{{g}_{0}}{1-\lambda }⇔{p}_{2}\left(\lambda -1\right)={g}_{0}$  imply:
 $\begin{array}{c}\left(\lambda -1\right)\left({p}_{1}-2\right)={g}_{0}\end{array}$ (1.0.11)
Using the equations ( 1.0.3    ), ( 1.0.8    ) and ( 1.0.11    ) we get:
 $\begin{array}{c}{g}_{1}=g\left(1,1\right)-1-{g}_{2}-{g}_{0}=\lambda -{p}_{1}\lambda -1+{p}_{1}+\left(1-\lambda \right)\left({p}_{1}-2\right)=-2{p}_{1}\lambda +2{p}_{1}+3\lambda -3\end{array}$ (1.0.12)
Thus, by ( 1.0.8    ), ( 1.0.11    ), and ( 1.0.12    ), we obtain
 $\begin{array}{c}g\left(1,x\right)={x}^{3}-{p}_{1}{x}^{2}+\left(-2\lambda {p}_{1}+2{p}_{1}+3\lambda -3\right)x-\left(1-\lambda \right)\left({p}_{1}-2\right).\end{array}$ (1.0.13)
Hence the cover $u$  is completely determined by ( 1.0.2    ), $\lambda =c$  , and ( 1.0.13    ). Now we can apply the results of this calculation and construct normalized genus 2 covers of degree 3.
Later we will use that we get by ( 1.0.13    ):
 $\begin{array}{c}\lambda \left(x-{p}_{1}\right){x}^{2}-\lambda g\left(1,x\right)=-\lambda \left(1-\lambda \right)\left(\left(2{p}_{1}-3\right)x-{p}_{1}+2\right),\end{array}$ (1.0.14)
 $\begin{array}{c}\lambda \left(x-{p}_{1}\right){x}^{2}-g\left(1,x\right)=\left(\lambda -1\right)\left(x-{p}_{1}+2\right)\left(x-1{\right)}^{2}\end{array}$ (1.0.15)
Now we consider smooth curves $C$  of genus 2, which are given by
 $\begin{array}{c}{y}^{2}=\left(x-{p}_{1}\right)\left(x-{p}_{1}+2\right)\left(\left(2{p}_{1}-3\right)x-{p}_{1}+2\right)g\left(1,x\right)\end{array}$ (1.0.16)
for some ${p}_{1}$  .
Proposition 1.1. The curve $C$  given in ( 1.0.16    ) is smooth if $u$  is not ramified over $\infty$  . Let $h:C\to {\mathbb{P}}^{1}$  and $i:E\to {\mathbb{P}}^{1}$  be the natural projections, which are given by $\left(x,y\right)\to x$  . Then the normalized covers ${f}_{±}:C\to E$  of degree 3 with Frey-Kani covering $u$  for $h$  and $i$  are given by ${f}_{±}\left(x,y\right)=\left(g\left(1,x\right):\lambda \left(x-{p}_{1}\right){x}^{2}:±\frac{\lambda \left(\lambda -1\right)yx\left(x-1\right)}{g\left(1,x\right)}\right).$
• Proof. Using that $u$  is unramified over $\infty$  we see that $g\left(1,x\right)$  has 3 different zeros. By  1.0.9  $\left(\right)$  and  1.0.10  $\left(\right)$  , we conclude that $u$  maps the 3 other points in ${\mathbb{P}}^{1}$  , which give the Weierstrass points of $C$  , to 3 different points. Therefore the Weierstrass points of $C$  are given by 6 different points of ${\mathbb{P}}^{1}$  , and $C$  is smooth. Let ${X}_{1}=\lambda \left(x-{p}_{1}\right){x}^{2}$  and ${X}_{0}=g\left(1,x\right)$  .
Note that $E=\left\{{y}^{2}{x}_{0}-{x}_{1}\left({x}_{1}-{x}_{0}\right)\left({x}_{1}-\lambda {x}_{0}\right)=0\right\}\subset {\mathbb{P}}^{2}$  . There is a cover $f:C\to E$  with Frey-Kani covering $u$  for the chosen degree 2 covers $h$  and $i$  if and only if there is an $Y\in k\left(C\right)$  , which satisfies the equation ${Y}^{2}{X}_{0}={X}_{1}\left({X}_{1}-{X}_{0}\right)\left({X}_{1}-\lambda {X}_{0}\right)$  in $k\left(C\right)$  . By ( 1.0.14    ) and ( 1.0.15    ), one can easily check that this equation has the solutions: $Y=±\frac{\lambda \left(\lambda -1\right)yx\left(x-1\right)}{g\left(1,x\right)}$  By the definitions of $C$  and ${f}_{±}$  , one can easily see that ${f}_{+}$  and ${f}_{-}$  map the 3 different Weierstrass points of $C$  , which are given by the zeros of $g\left(1,x\right)$  , to $\left(0:0:1\right)={0}_{E}$  . Thus, ${f}_{+}$  and ${f}_{-}$  are normalized.
Let $\mathcal{U}$  be the set of coverings ${\mathbb{P}}^{1}\to {\mathbb{P}}^{1}$  of degree 3, which satisfy our choice of coordinates of ramification points and do not have any ramification point over $\infty$  .
Proposition 1.2. Using Proposition  1.1    we have a bijection $\chi :\mathcal{U}\to {\mathcal{ℋ}}_{E/k,3}\left(k\right)$  . This map is given by $u\to \left[{f}_{±}:C\to E\right]$  .
• Proof. By the hyperelliptic involution on $C$  , we conclude that the covers ${f}_{±}:C\to E$  lie in the same isomorphism class, and the map $\chi :\mathcal{U}\to {\mathcal{ℋ}}_{E/k,3}\left(k\right)$  is well defined.
Let $f:C\to E$  be a normalized genus 2 cover of degree 3. The Frey-Kani covering of $f$  is not ramified over $\infty$  and has ramification points over 0, 1 and $\lambda$  (see [1). One can put these ramification points on 0, 1 and $\infty$  such that the Frey-Kani covering is some $u\in \mathcal{U}$  .
In [1, page 92-93 the curve $C$  is described by the Frey-Kani covering. This description of $C$  , ( 1.0.14    ) and ( 1.0.15    ) imply that $C$  is given by ( 1.0.16    ) for some ${p}_{1}$  . One can assume that $h$  and $i$  are given by $\left(x,y\right)\to x$  . Hence by Proposition  1.1    , we conclude that $\chi$  is surjective.
Let $\chi \left({u}_{1}\right)=\left[{f}_{1}:{C}_{1}\to E\right]=\left[{f}_{2}:{C}_{2}\to E\right]=\chi \left({u}_{2}\right)$  and ${h}_{i}:{C}_{i}\to {\mathbb{P}}^{1}$  be the natural degree 2 covers (for $i=1,2$  ). Then there is an isomorphism $i:{C}_{1}\to {C}_{2}$  such that ${f}_{1}={f}_{2}\circ i$  and an $a\in Aut\left({\mathbb{P}}^{1}\right)$  such that $a\circ {h}_{1}={h}_{2}\circ i$  (see [3, page 304, IV, Exercise 2.2.(a)). Therefore we conclude ${u}_{1}={u}_{2}\circ a$  . Let $x\in \left\{0,1,\infty \right\}$  . We get by our assumptions that ${u}_{1}\left(x\right)={u}_{2}\left(x\right)$  . These points are ramification points of ${u}_{1}$  and ${u}_{2}$  . Thus, ${u}_{1}={u}_{2}\circ a$  implies that $a\left(x\right)=x$  ( $\forall \text{}x\in \left\{0,1,\infty \right\}$  ). Therefore we have $a=id$  and $\chi$  is injective.

2 The reckoning of the moduli space

By $u\left(\infty \right)=\lambda \ne 0=u\left({p}_{1}\right)$  , we conclude that ${p}_{1}$  and $\infty$  can not coincide. Thus, ( 1.0.2    ), $c=\lambda$  , and ( 1.0.13    ) imply that $g\left(1,x\right)$  is determined by ${p}_{1}$  and $\lambda$  , and that:
Lemma 2.1. We have an injective map $ı:\mathcal{U}\to {\mathbb{A}}^{1}$  , which maps the morphism, which is given by $x\to \frac{\lambda \left(x-{p}_{1}\right){x}^{2}}{g\left(1,x\right)}$  , to ${p}_{1}$  .
Remark 2.2. The morphism, which is given by $x\to \frac{\lambda \left(x-{p}_{1}\right){x}^{2}}{g\left(1,x\right)}$  for some ${p}_{1}$  , has the degree 3 if and only if we have $g\left(1,{p}_{1}\right)\ne 0$  and $g\left(1,0\right)\ne 0$  . By ( 1.0.13    ), one can easily see that this is true if and only if ${p}_{1}\in k\\left\{1,2\right\}$  . Thus, by ( 1.0.14    ) and ( 1.0.15    ), we have ${p}_{1}\in ı\left(\mathcal{U}\right)$  if and only if we have ${p}_{1}\in k\\left\{1,2\right\}$  and $g\left(1,x\right)$  has 3 different zeros.
Let ${p}_{1}\ne 1,2,\infty$  , and $u$  be the morphism, which is given by $x\to \frac{\lambda \left(x-{p}_{1}\right){x}^{2}}{g\left(1,x\right)}$  . The preceding remark, ( 1.0.14    ) and ( 1.0.15    ) imply that $u$  has the degree 3 and satisfies our choice of coordinates of ramification points, and that we have ${p}_{1}\in ı\left(\mathcal{U}\right)$  if and only if $u$  does not have a ramification point over $\infty$  . By the Hurwitz formula, $u$  has a ramification divisor of degree 4. But we do not know the position of a fourth branch point. Let $u\left(\mu \right)=u\left(\delta \right)=\zeta$  and $\delta$  be a ramification point of $u$  . We have $\left({x}_{0}:{x}_{1}-\zeta {x}_{0}\right)\circ u=\left(g:F\left({x}_{1}-\mu {x}_{0}\right)\left({x}_{1}-\delta {x}_{0}{\right)}^{2}\right)$  for some $F\in k$  resp., $F\left({x}_{1}-\mu {x}_{0}\right)\left({x}_{1}-\delta {x}_{0}{\right)}^{2}=\left(\lambda -\zeta \right){x}_{1}^{3}+\left(g\left(1,1\right)-\lambda -\zeta {g}_{2}\right){x}_{1}^{2}{x}_{0}-\zeta {g}_{1}{x}_{1}{x}_{0}^{2}-\zeta {g}_{0}{x}_{0}^{3}.$  This equation of polynomials implies the following equations of coefficients:
 $\begin{array}{ccc}F=\lambda -\zeta & & \end{array}$
 $\begin{array}{ccc}-F\mu -2F\delta =g\left(1,1\right)-\lambda -\zeta {g}_{2}& & \end{array}$
 $\begin{array}{ccc}F{\delta }^{2}+2F\delta \mu =-\zeta {g}_{1}& & \end{array}$
 $\begin{array}{ccc}-\mu {\delta }^{2}F=-\zeta {g}_{0}& & \end{array}$
Using $F=\lambda -\zeta$  , ${p}_{1}=-{g}_{2}$  , $-{p}_{1}\lambda =g\left(1,1\right)-\lambda$  , ( 1.0.8    ), and ( 1.0.3    ) we substitute in the second equation: $\left(-\lambda +\zeta \right)\mu +2\left(-\lambda +\zeta \right)\delta =-{p}_{1}\lambda +\zeta {p}_{1}$  Let $\lambda \ne \zeta$  . Then we have:
 $\begin{array}{c}\mu ={p}_{1}-2\delta \end{array}$ (2.2.1)
By $\mu ={p}_{4}-2\delta$  and and the same substitution as above, we obtain:
 $\begin{array}{c}\left(\lambda -\zeta \right)\left(2\delta {p}_{1}-3{\delta }^{2}\right)=-\zeta \left(1-\lambda \right)\left(2{p}_{1}-3\right),\end{array}$ (2.2.2)
 $\begin{array}{c}\left(2{\delta }^{3}-{p}_{1}{\delta }^{2}\right)\left(\lambda -\zeta \right)=\zeta \left(1-\lambda \right)\left({p}_{1}-2\right)\end{array}$ (2.2.3)
By  2.2.3  $\left(\right)$  , we get the equation $\zeta =\frac{\left(2{\delta }^{3}-{p}_{1}{\delta }^{2}\right)\left(\lambda -\zeta \right)}{\left(1-\lambda \right)\left({p}_{1}-2\right)}$  . Thus, one has by ( 1.0.10    ) and ( 2.2.2    ):
$\left(\lambda -\zeta \right)\left(2\delta {p}_{1}-3{\delta }^{2}\right)=\left(2{\delta }^{3}-{p}_{1}{\delta }^{2}\right)\left(\lambda -\zeta \right)\left(-{p}_{3}\right)$  The solution $\zeta =\lambda$  gives the ramification point $\infty$  and $\delta =0$  gives the ramification point 0. Therefore we can divide by $\lambda -\zeta$  and $\delta$  , and get $2{p}_{1}-3\delta =\left(2{\delta }^{2}-{p}_{1}\delta \right)\left(-{p}_{3}\right)⇔0={\delta }^{2}-\frac{{p}_{1}{p}_{3}+3}{2{p}_{3}}\delta +\frac{{p}_{1}}{{p}_{3}}=\left(\delta -1\right)\left(\delta -\frac{{p}_{1}}{{p}_{3}}\right),$  because we get by ( 1.0.10    ): $\frac{{p}_{1}{p}_{3}+3}{2{p}_{3}}=\frac{{p}_{1}\left(2{p}_{1}-3\right)+3\left({p}_{1}-2\right)}{2\left(2{p}_{1}-3\right)}=\frac{\left(2{p}_{1}^{2}-4{p}_{1}\right)+\left(4{p}_{1}-6\right)}{2\left(2{p}_{1}-3\right)}=\frac{{p}_{1}}{{p}_{3}}+1$  Thus, a ramification point $\delta$  of $u$  is given by $\delta =\frac{{p}_{1}}{{p}_{3}}=\frac{{p}_{1}\left({p}_{1}-2\right)}{2{p}_{1}-3}.$  We substitute $\delta =\frac{{p}_{1}\left({p}_{1}-2\right)}{2{p}_{1}-3}$  in ( 2.2.2    ) and get: $\left(\lambda -\zeta \right)\left(2\frac{{p}_{1}\left({p}_{1}-2\right)}{2{p}_{1}-3}{p}_{1}-3\left(\frac{{p}_{1}\left({p}_{1}-2\right)}{2{p}_{1}-3}{\right)}^{2}\right)=-\zeta \left(1-\lambda \right)\left(2{p}_{1}-3\right)$  $⇔\left(\lambda -\zeta \right){p}_{1}^{3}\left({p}_{1}-2\right)=-\zeta \left(1-\lambda \right)\left(2{p}_{1}-3{\right)}^{3}⇔\lambda {p}_{1}^{3}\left({p}_{1}-2\right)=\zeta \left(\left(\lambda -1\right)\left(2{p}_{1}-3{\right)}^{3}+{p}_{1}^{3}\left({p}_{1}-2\right)\right)$
 $\begin{array}{c}⇔\zeta =\frac{\lambda {p}_{1}^{3}\left({p}_{1}-2\right)}{\left(\lambda -1\right)\left(2{p}_{1}-3{\right)}^{3}+{p}_{1}^{3}\left({p}_{1}-2\right)}\end{array}$ (2.2.4)
Recall that $E$  is determined by $\lambda$  , and that $E$  resp., $\lambda$  is fixed. Hence by  2.2.4    , the set of values of ${p}_{1}$  , which induce a 4th ramification point over $\infty$  , is given by $Z=\left\{\left(\lambda -1\right)\left(2\cdot {p}_{1}-3{\right)}^{3}+{p}_{1}^{3}\left({p}_{1}-2\right)=0\right\}\subset {\mathbb{P}}^{1}$  .
Remark 2.3. Four different points lie in $Z$  .
• Proof. We consider the morphism $\stackrel{~}{\zeta }:{\mathbb{P}}^{1}\to {\mathbb{P}}^{1}$  , which is given by $\stackrel{~}{\zeta }\left({p}_{1}\right)=\frac{\lambda {p}_{1}^{3}\left({p}_{1}-2\right)}{\left(\lambda -1\right)\left(2{p}_{1}-3{\right)}^{3}+{p}_{1}^{3}\left({p}_{1}-2\right)}=\left(1+\frac{\lambda -1}{\lambda }\frac{\left(2{p}_{1}-3{\right)}^{3}}{{p}_{1}^{3}\left({p}_{1}-2\right)}{\right)}^{-1}.$  Therefore the ramification points of $\stackrel{~}{\zeta }$  are the ramification points of the morphism $\sigma$  , which is given by ${p}_{1}\to \frac{\left(2{p}_{1}-3{\right)}^{3}}{{p}_{1}^{3}\left({p}_{1}-2\right)}$  . The derivative of $\sigma$  is ${\sigma }^{\prime }\left({p}_{1}\right)=\frac{6\left(2{p}_{1}-3{\right)}^{2}{p}_{1}^{3}\left({p}_{1}-2\right)-\left(2{p}_{1}-3{\right)}^{3}\left(3{p}_{1}^{2}\left({p}_{1}-2\right)+{p}_{1}^{3}\right)}{{p}_{1}^{6}\left({p}_{1}-2{\right)}^{2}}=\frac{-2\left(2{p}_{1}-3{\right)}^{2}{p}_{1}^{2}\left({p}_{1}-3{\right)}^{2}}{{p}_{1}^{6}\left({p}_{1}-2{\right)}^{2}}.$  Thus, the ramification points of $\stackrel{~}{\zeta }$  are $0$  , $\frac{3}{2}$  and $3$  . The statement follows by the fact that all $x\in \left\{0,\frac{3}{2},3\right\}$  fulfil $\left(\lambda -1\right)\left(2\cdot x-3{\right)}^{3}+{x}^{3}\left(x-2\right)\ne 0$  .
By Lemma  2.1    and Remark  2.2    , we have a bijection between $\mathcal{U}$  and the set of closed points of ${\mathbb{P}}^{1}\\left(\left\{1,2,\infty \right\}\cup Z\right)$  , where ${\mathbb{P}}^{1}\\left(\left\{1,2,\infty \right\}\cup Z\right)$  is an (Zariski) open subset of ${\mathbb{P}}^{1}$  . Recall that ${\mathcal{ℋ}}_{E/k,3}$  is represented by a smooth modular curve ${H}_{E/k,3}$  (see [4, Theorem 1.1). Therefore, we conclude by Proposition  1.1    and Proposition  1.2    :
Theorem 2.4. We have ${H}_{E/k,3}\sim ={\mathbb{P}}^{1}\\left(\left\{1,2,\infty \right\}\cup Z\right)$  . The fiber ${\mathcal{C}}_{{p}_{1}}$  of the universal family $\mathcal{C}\to {\mathbb{P}}^{1}\\left(\left\{1,2,\infty \right\}\cup Z\right)$  is given by ( 1.0.16    ) for all ${p}_{1}\in {\mathbb{P}}^{1}\\left(\left\{1,2,\infty \right\}\cup Z\right)$  .
References

1. G. Frey, On elliptic curves with isomorphic torsion structures and corresponding curves of genus 2. Elliptic Curves, Modular Forms, and Fermat's Last Theorem (Hong Kong, 1993), Ser. Number Theory, I, 79-98.
2. G. Frey, E. Kani, Curves of Genus 2 covering elliptic curves and an arithmetical application. In: Arithmetic Algebraic Geometry. Progress in Math. vol. 89, Birkhäuser, Boston, 1991, 153-176.
3. R. Hartshorne, Algebraic Geometry. Graduate Texts in Mathematics 52. Springer Verlag, New York, 1977.
4. E. Kani, Hurwitz Spaces of genus 2 covers of an elliptic curve. Collect. Math. 54(2003), 1-51.
5. R. Kuhn, Curves of genus 2 with split Jacobian. Transac. AMS 307(1988), 41-49.
6. T. Shaska, Curves of genus 2 with $\left(n,n\right)$  -split Jacobians. J. Symbolic Computation 2001(31), 603-617.
7. T. Shaska, Genus 2 fields with degree 3 elliptic subfields. Forum Math. (16)2004, 263-280.

Universitat Duisburg-Essen, Fachbereich Mathematik, 45117 Essen, Germany E-mail address : jan.rohde@uni-essen.de