A Note on Minimal Galois Embeddings of Abelian Surfaces
HISAOYOSHIHARA
ABSTRACT- We show that the least numberNsuch that an abelian surface has a Galois embedding inPNis seven and then we give examples of such surfaces.
1. Introduction.
This is a continuation of our previous paper [7]. The least numberN suchthat an abelian surface can be embedded inPNis four, and in that case the abelian surface has a special structure, see for example [3]. Similarly it might have some interest to study the least numberNsuchthat an abelian surfaceAcan be Galois-embedded inPN. Moreover, in that case we want to know the structure of A. In this short note we give the answer to the problem. Before stating it, we recall from [7] some of the definitions and properties of Galois embeddings of algebraic varieties.
Let kbe the ground field of our discussions, which is assumed to be algebraic closed. Let V be a nonsingular projective algebraic variety of dimensionnandDa very ample divisor. We denote this by a pair (V;D).
Let f fD:V,!PNbe the embedding ofV associated withthe complete linear system jDj, wh ere N1dimH0(V;O(D)). Suppose thatW is a linear subvariety of PN suchthat dimW N n 1 and W\f(V) ;.
Then consider the projectionW,pW:PN >W0withcenterW, whereW0 is an n-dimensional linear subvariety not meeting W. The composition ppWf is a surjective morphism fromVtoW0Pn. LetK k(V) and K0k(W0) be the function fields ofV andW0 respectively. The covering map p induces a finite extension of fields p:K0,!K of degree ddegf(V)Dn, which is the self-intersection number ofD. It is easy to see that the structure of this extension does not depend on the choice ofW0
(*) Indirizzo dell'A.: Department of Mathematics, Faculty of Science, Niigata University, Niigata 950-2181, Japan.
E-mail: yosihara@math.sc.niigata-u.ac.jp
but only onW, hence we denote byKWthe Galois closure of this extension and byGW Gal(KW=K0) the Galois group of KW=K0. Note that GW is isomorphic to the monodromy group of the coveringp:V !W0.
DEFINITION1. In the above situation we callGWthe Galois group atW.
If the extensionK=K0is Galois, we callf andWa Galois embedding and a Galois subspace for the embedding, respectively.
DEFINITION2. A nonsingular projective algebraic varietyV is said to have a Galois embedding if there exist a very ample divisorDsuch that the embedding associated withthe complete linear system jDj has a Galois subspace. In this case the pair (V;D) is said to define a Galois embedding.
In this note we use the following notation.
Zm : the cyclic group of orderm Dm : the dihedral group of order 2m jGj: the order of a groupG
r: exp (2p
p 1
=6)
Aut(V) : the automorphism group of a varietyV ha1;. . .;ami: the subgroup generated bya1;. . .;am 12 : the unit matrix of degree two
We shall make use of the following criterion (cf. [7, Theorem 2.2]).
THEOREMA. Let V and D be as above. The pair(V;D)defines a Galois embedding if and onlyif the following conditions hold:
(1) There exists a subgroup G of Aut(V)such thatjGj Dn.
(2) There exists a G-invariant linear subspace L of H0(V;O(D)) of dimension n1such that, for anys2G, the restrictionsjLis a multiple of the identity.
(3) The linear system L has no base points.
The original form of the study of the Galois embedding is given in [5] or [6]. We have applied the above method to abelian surfacesAoverkC and obtained some results.
2. Statement of Theorem.
LetAbe an abelian surface defined overkCandGbe a finite sub- group ofAut(A). Fix a covering morphismC2 !A. An elementg2Ghas
a representation eg on the universal covering C2 suchthat egzM(g)zt(g), where M(g)2GL(2;C),z2C2 andt(g)2C2. We call M(g) andt(g) the matrix and translation part of the representationeg, re- spectively. PutG0 fg2GjM(g)12gandH fM(g)jg2Gg. Then, we have the following exact sequence of groups
1 !G0 !G !H !1:
ClearlyBA=G0is also an abelian surface andHG=G0is a subgroup of Aut(B).
Hereafter we assume that (A;D) defines a Galois embedding inPNand let G be the Galois group. Then G is a subgroup ofAut(A) and B=His isomorphic toA=GP2. With the notation above, we have:
THEOREM B. [7, Theorem 3.7] If an abelian surface A has a Galois embedding, then H is isomorphic to D3, D4or a semidirect product Z2j K, where KD4or ZmZm (m3; 4; 6)
Now, the answer to the question in the introduction is given as follows:
THEOREMC. Suppose that(A;D)defines a Galois embedding. Then (1) The least number N is seven, i.e., there exists an abelian surface A such that it can be embedded into P7 with a Galois subspace, and no abelian surface embedded inPN(N6)has a Galois subspace.
(2) The abelian surface BA=G0 is isomorphic to the self-product EE of an elliptic curve, such that HG=G0acts on B and HD4or Z2j D4.
REMARK1. The minimal Galois embedding for an elliptic curve E is given as follows. If E can be Galois-embedded in P2, th en E must have an automorphism of order three with a fixed point. In fact, the elliptic curve is unique and is defined by Y2Z4X3Z3. Th e centers of the projections are (1:0:0);(0:
p 3
:1) and (0:
p 3
:1). How- ever, note that every elliptic curve has a Galois embedding in P3, where the group is isomorphic to Z2Z2 (further, if the j-invariant is 1728, then it has another projection center whose Galois group is isomorphic to Z4).
In what follows we shall give the proof of Theorem C and some examples.
In particular, there is an example whereBis the Jacobian of a curve. Indeed,
let J(C) be the Jacobian of the normalization C of the curve y2 x(x4ax21); a6 2, and letAbe an abelian surface which is an etale double coveringq:A !J(C). Then we shall show that (A; q(CC0)) gives a minimal Galois embedding, whereC0is a translated ofC.
3. Proof.
By Theorem B we havejHj 2a3b, wh ere (a;b)(3;0), (4;0), (5;0), (1;1), (1;2) (3;2). We consider the possible values ofjGj D22m. SinceAcan be embedded inPm 1by the complete linear systemjDj, we havem 14 by [3]. In view of Theorem B we havem65, hencem6. Suppose thatm6.
Then, from Theorem B again we infer thatHD3andjG0j 2. Every two dimensional complex crystallographic group G with X=GP2 has been classified in [4]. Referring to it, we see thatAcan be expressed asAC2=V suchthatVis the period matrix
V 1 r2 v r2v
1 r v rv
1 r2
1 r
1 0 v 0
0 1 0 v
; wherevis a complex number withIv>0.
Define four vectors as follows:
v1 1 1
; v2 r2
r ; v3 v v
vv1; v4 r2v rv
vv2: LetLAbe the lattice inC2generated byv1;v2;v3andv4. Letg1be a gen- erator ofG0whose representation iseg1zze, wherez; e2C2. Sinceg12
is the identity onA, we have 2e2 LA. LetLBbe the lattice generated byLA ande. Then BC2=LBA=hg1iis also an abelian surface on which the groupHacts. As shown in [4],His generated byg2 andg3, whose matrix parts are
M2M(g2) 0 1 1 0
andM3M(g3) r 0 0 r2
respectively. Since 2e2 LA, the vectorecan be expressed as e1
2 X4
i1
nivi;
where ni0 or 1 (1i4). Since G0 is a normal subgroup of G and jG0j 2, g1 commutes witheachelement of G. Therefore, we infer that
Mie e2 LA(i2;3). Indeed we have
M2e e f (2n1n2)v1 (2n3n4)v3g=2 2 LA: Thus we have
( n2v1 n4v3)=22 LA: (1)
Similarly, consideringM3e e, we have
f n2v1(n1 n2)v2 n4v3(n3 n4)v4g=22 LA: (2)
From (1) we getn2n40, hence from (2) we getn1n30. Since e2 L= A, this is a contradiction. Thus we havem7. From Theorem B we inferjGj 614, hence we havem8. We conclude that the least numberm is 8 by the examples in the next section.
REMARK 2. Referring to [4, Theorem 1], we see that in the case (i) HD4 the dimension of the moduli space is 1, but in the case (ii) HZ2j D4the dimension is zero.
4. Examples.
When we make examples, the following lemma is useful.
LEMMA3. If M(g) 12is a nonsingular matrix for g2G, then we can assume t(g)0.
PROOF. We consider tGt 1 instead of G, where t is a translation
etzz`. If M(g) 12 is nonsingular, then by putting `
(M(g) 12) 1t(g), we gett(tgt 1)0. p
In casem8 we haveHD4orZ2j D4by Theorem B. We shall give examples of both. If (i) HD4, thenjG0j 2. Since G0 is a normal sub- group of G and jG0j 2, we infer that GG0H. Suchexamples are given in Examples 4 and 5. On the other hand, if HZ2j D4, we have GH. Suchan example is given in Example 7.
EXAMPLE4. LetAbe the abelian surface with the period matrix
1 0 v 0
0 1 0 v
suchthatIv>0:
LetLAbe the lattice generated by the column vectors of the period matrix.
Let us consider the automorphismsg1,g2 andg3ofA, whose representa- tions onC2are as follows:
ge1z z12 n1n3v n2n4v
; ge2z 0 1
1 0
z a1 a2 ; ge3z 0 1
1 0
z
where (n1;n2;n3;n4)(0;0;1;1);(1;1;0;0);(1;1;1;1);
a1a2 a1a2
2 LAand 2a1 0
2 LA:
We haveg12g22g34id,g2g3g2g3 1andgig1g1gi(i2;3) onA.
Putting G hg1;g2;g3i, we have G0 hg1i and GG0H where H hM(g2);M(g3)i. ClearlyHD4. The groupGis a subgroup ofAut(A) and A=GP2. The very ample divisor D is given by p(L), where p:A !A=GP2andLis a line inP2(cf. [7, Lemma 3.5]). We infer from Theorem A that (A;D) defines a Galois embedding.
By [7, Corollary 3.8], if A has a Galois embedding, then the abelian surfaceBA=G0is isomorphic toEEfor some elliptic curveE. On th e other hand,EEcan be a Jacobian of a curve for someE(cf. [2]). So one may ask what type of genus 2 curve can give the Jacobian whose double covering has a minimal Galois embedding. Let us consider this question in the next example.
EXAMPLE 5. Let G be the curve defined by y2x(x4ax21), where we assumea6 2 (cf. [1, Theorem 4.8]). This curve has a singular point at1. LetCbe the normalization ofG. The genus ofCis two. Lets andtbe the birational transformations ofG defined by
s(x) x; s(y)iy and t(x)1=x; t(y)y=x3
respectively. Clearly we have s4t2idandtsts 1. LetHbe the group generated bysandt. Then we haveH D4. This group acts onC.
LetC(x;y) be th e function field ofC, wh erey2x(x4ax21). Clearly the invariant field of C(x;y) bys is C(x2). Let w1 andw2 be a basis of holomorphic 1-forms on C induced from dx=y and xdx=y respectively.
We have
s(w1)iw1; s(w2) iw2 and t(w1) w2; t(w2) w1: LetJ(C) be the Jacobian ofC. Taking a base pointP2C, the Abel-Jacobi mapjP is given as
jP :C !J(C) such th at jP(Q) ZQ
P
w1; ZQ
P
w2
0
@
1
A modulo the lattice;
whereQ2C. If Pis a fixed point of s, then we have Zs(Q)
P
w ZQ
P
s(w);
forQ2C, wh ereww1orw2. Similarly ifP0is a fixed point oft, th en we have
Zt(Q0)
P0
w ZQ0
P0
t(w);
whereQ02C. Note that ifjP:C !J(C) is defined witha base pointP, and jP0 :C !J(C) is defined witha base pointP0, thenjP0 tjP, wheretis a translation inJ(C). Assume thatPis a fixed point ofs. Then, lettingesandet be the representations ofsandtonC2respectively, we obtain that they can be expressed aseszM(s)zandetzM(t)zv, wherez2C2,v2C2and
M(s) i 0
0 i
andM(t) 0 1
1 0
:
PutH hM(s);M(t)i. Then we haveHD4. Note that the curvejP(C) is fixed bysandC=hsiis isomorphic to a smooth rational curve. We infer from the above arguments thatJ(C)=His isomorphic toP2andC=hsiis a line.
Letp:J(C) !J(C)=HP2be the quotient morphism. Thenp(L) can be expressed asCC0, wh ereLis the line andC0is a translation ofConJ(C).
Let A be an abelian surface suchthat q:A !J(C) is an etale double covering given as follows. ExpressJ(C)C2=LandAC2=L0, whereL andL0 are lattices satisfyingjL:L0j 2. Take an element`2 L n L0 so that r(z)z` is a translation of order two on A. Then we have L hL0; `i. Since 2`2 L0, we have M(2`)2M(`)2 L0, where M2H.
Hence we infer thatsandtinduce automorphisms onA. We use the same
letterHto denote the group consisting of the elements which are induced fromH. LetGbe the automorphism group onAgenerated byHandr. Then putppq:A !A=GP2. Since deg (p)1610, we see thatp(L) is very ample (cf. [7, Lemma 3.5]). From Theorem A we infer that (A; p(L)) defines a Galois embedding inP7.
REMARK6. Note thatq(C)Ceis irreducible. Because, if not so, then q(C) can be written asC1C2. We have (q(C))24 and C12C222.
SinceC1 is ample, we have (C1;C2)1. This is a contradiction. Similarly q(C0)Ce0is also irreducible. The divisorCeCe0gives the minimal Galois embedding ofA. IfLis the image ofC, thenCeCe0p(L).
EXAMPLE7. LetAbe the abelian surface with period matrix 1 0 i (1i)=2
0 1 0 (1i)=2
:
This abelian surface has the automorphisms g1, g2 and g3, whose repre- sentations onC2are as follows:
ge1z 1 0 0 1
z e11 e12
; ge2z 0 1
1 0
z e21 e22
; ge3z i 0
0 i
z;
where the following vectors belong to the lattice generated by the column vectors of the period matrix:
2e012
; e21e22
e21e22
; e11 e12 2e21
e11e12
; (1 i)e11
(1 i)e12
; e21 ie22
e22ie21
:
We haveg12g22g34id,g1g2g1g2g23,g1g3g1g3andg2g3g2g3 1. Putting G hg1;g2;g3i, we see that G is isomorphic to the semidirect productZ2j D4 andGis a subgroup ofAut(A) andA=GP2. Th e very ample divisorDis given by p(L), wherep:A !A=GP2 andL is a line in P2. We infer from Theorem A that (A;D) defines a Galois em- bedding.
REMARK 8. The abelian surface A in Example 7 is isogenous to EiEi, whereEiC=(1; i). (In fact, we can show thatAis isomorphic to EiEi. ) Thanks to [2], this abelian surface cannot be a Jacobian of a curve.
REFERENCES
[1] A. FUJIKI,Finite automorphism groups of complex tori of dimension two, Publ. RIMS. Kyoto Univ.,24(1988), pp. 1-97.
[2] T. HAYASHIDA- M. NISHI,Existence of curves of genus two on a product of two elliptic curves, J. Math. Soc. Japan,17(1965), pp. 1-16.
[3] K. HULEK- H. LANGE,Examples of abelian surfaces inP4, Journal Reine Angew. Math.,363(1985), pp. 200-216.
[4] J. KANEKO- S. TOKUNAGA- M. YOSHIDA,Complex crystallographic groups II, J. Math. Soc. Japan,34(1982), pp. 595-605.
[5] K. MIURA- H. YOSHIHARA,Field theoryfor function fields of plane quartic curves, J. Algebra,226(2000), pp. 283-294.
[6] H. YOSHIHARA, Function field theoryof plane curves bydual curves, J.
Algebra,239(2001), pp. 340-355.
[7] H. YOSHIHARA,Galois embedding of algebraic varietyand its application to abelian surface, Rend. Sem. Mat. Universita di Padova,117(2007), pp. 69-85.
Manoscritto pervenuto in redazione il 14 aprile 2010.