CoCalc -- Collaborative Calculation in the Cloud
Sharedwww / job / Prop.dviOpen in CoCalc
����;� TeX output 1999.11.12:1118������ǂ����6�;��䃍�������������	�+F
C��q
cmbxti10�Wil�8�liam��sA.�Stein�2�Detaile���Ud�R�ese�ar�ch�Plan��(�����d��k�ff���������2ō��|4V��r_���N��qcmbx12�Detailed��
Researc��Zh�Plan��(V����d���N�G�cmbx12�1��(XIn��u�tro��=duction���#���d�X�Qcmr12�My��Wresearc��rh�program�re
ects�the�essen�tial�in�terpla�y�b�S�et�w�een�abstract�theory�and�explicit������dmac��rhine��*computation�during�the�latter�half�of�the�t�w�en�tieth�cen�tury;��it�sits�at�the�in�tersec-�����dtion���of�recen��rt�w�ork�of�B.�Mazur,���K.�Rib�S�et,�J-P��V.�Serre,�R.�T��Va��rylor,�and�A.�Wiles�on�Galois�����drepresen��rtations�Y�attac�hed�to�mo�S�dular�ab�elian�v��X�arieties�(see�[�21����,��24���,��26��,��28��])�with�w��rork�of�����dJ.�KCremona,�
4N.�Elkies,�and�J.-F.�Mestre�on�explicit�computations�in��rv�olving�Kmo�S�dular�forms�����d(see��[�9����,��11����]).�����XIn�nH1969�B.�Birc��rh�[�4����]�describ�S�ed�computations�that�led�to�the�most�fundamen�tal�op�S�en�����dconjecture��in�the�theory�of�elliptic�curv��res:�����_PI�Jw��ran�t��to�describ�S�e�some�computations�undertak��ren�b�y�m�yself�and�Swinnerton-����_PDy��rer�ܠon�EDSA�C��\b�y�whic�h�w�e�ha�v�e�calculated�the�zeta-functions�of�certain�elliptic����_Pcurv��res.�	As�2�a�result�of�these�computations�w�e�ha�v�e�found�an�analogue�for�an����_Pelliptic���curv��re�of�the�T��Vamaga�w�a�n�um�b�S�er�of�an�algebraic�group;�
wand�conjectures����_P(due��to�ourselv��res,�due�to�T��Vate,�and�due�to�others)�ha�v�e�proliferated.�����dThe�]ric��rh�tap�S�estry�of�arithmetic�conjectures�and�theory�w�e�enjo�y�to�S�da�y�w�ould�not�exist�with-�����dout�sthe�ground-breaking�application�of�computing�b��ry�Birc�h�and�Swinnerton-Dy�er.�zBCom-�����dputations��fin�the�1980s�b��ry�Mestre�w�ere�k�ey�in�con�vincing�Serre�that�his�conjectures�on�����dmo�S�dularit��ry���of�o�dd�irreducible�Galois�represen��rtations�w�ere�w�orth�y�of�serious�consideration�����d(see��[�24����]).�N�These�conjectures�ha��rv�e��inspired�m��ruc�h��recen�t�w�ork;���for�example,���Rib�S�et's�pro�of�of�����dthe�����g�cmmi12��-conjecture,�whic��rh�pla�y�ed�an�essen�tial�role�in�Wiles's�pro�S�of�of�F��Vermat's�Last�Theorem.�����XMy�H|w��rork�on�the�Birc�h�and�Swinnerton-Dy�er�conjecture�for�mo�S�dular�ab�elian�v��X�arieties�and�����dsearc��rh��:for�new�examples�of�mo�S�dular�icosahedral�Galois�represen�tations�has�led�me�to�disco�v�er�����dand�/timplemen��rt�algorithms�for�explicitly�computing�with�mo�S�dular�forms.��yMy�researc�h,�T�whic�h�����din��rv�olv�es���nding�w��ra�ys���to�compute�with�mo�S�dular�forms�and�mo�dular�ab�elian�v��X�arieties,���is�driv��ren�����db��ry��outstanding�conjectures�in�n�um�b�S�er�theory��V.��(V����d�2��(XIn��u�v���arian�ts�z�of�mo��=dular�ab�elian�v���arieties���#���d�No��rw�tthat�the�Shim�ura-T��Vaniy�ama�conjecture�has�b�S�een�pro�v�ed,��`the�main�outstanding�prob-�����dlem�Lwin�the�eld�is�the�Birc��rh�and�Swinnerton-Dy�er�conjecture�(BSD�L^conjecture),�d�whic�h�ties�����dtogether�r�the�arithmetic�in��rv��X�arian�ts�r�of�an�elliptic�curv��re.�фThere�is�no�general�class�of�elliptic�����dcurv��res��for�whic�h�the�full�BSD��conjecture�is�kno�wn.�-Approac�hes�to�the�BSD��conjecture�that�����drely���on�congruences�b�S�et��rw�een���mo�dular�forms�are�lik��rely�to�require�a�deep�er�understanding�of�����dthe��analogous�conjecture�for�higher-dimensional�ab�S�elian�v��X�arieties.��As�a�rst�step,���I���ha��rv�e��ob-�����dtained��.theorems�that�mak��re�p�S�ossible�explicit�computation�of�some�of�the�arithmetic�in�v��X�arian�ts�����dof��mo�S�dular�ab�elian�v��X�arieties.��"ʫ����d�'��N�ffcmbx12�2.1��$��The�ffBSD�conjecture��U���d�By��[�6����]�w��re�no�w�kno�w�that�ev�ery�elliptic�curv�e�o�v�er��(��N�cmbx12�Q��is�a�quotien�t�of�the�curv�e��X�����|{Ycmr8�0����(�N�@�)�whose�����dcomplex�NYp�S�oin��rts�are�the�isomorphism�classes�of�pairs�consisting�of�a�(generalized)�elliptic������*�ǂ����6�;��䃍�������������	�Wil�8�liam��sA.�Stein�2�Detaile���Ud�R�ese�ar�ch�Plan��(�����d��k�ff���������2ō��y�;���d�curv��re�a�and�a�cyclic�subgroup�of�order��N�@�.���Let��J�����0����(�N��)�denote�the�Jacobian�of��X�����0����(�N��);���this�is������dan�tzab�S�elian�v��X�ariet��ry�of�dimension�equal�to�the�gen�us�of��X�����0����(�N�@�)�whose�p�S�oin�ts�corresp�S�ond�to�the�����ddegree��0�divisor�classes�on��X�����0����(�N�@�).�����XAn�$��,���@cmti12�optimal�}�quotient��of��J�����0����(�N�@�)�is�a�quotien��rt�b�y�an�ab�S�elian�sub�v��X�ariet�y��V.���Consider�an�optimal�����dquotien��rt�Y�A��suc�h�that��L�(�A;����1)�F�!",�
cmsy10�6�=�0.��&By�Y[�13����],��A�(�Q�)�and����hV1
wncyr10�X��P��(�A=�Q�)�are�b�S�oth�nite.�The�BSD�����dconjecture��asserts�that���n�������ō���g�L�(�A;����1)����g�[��z�%��
�΍�
�
������2cmmi8�A�������0r�=������덑��#��X��
�n�(�A=�Q�)�������������u
cmex10�Q���
�����p��K�cmsy8�j�N����c�����p������k��z�hK�
�΍��7�#�A�(�Q�)������#�A������_��*��(�Q�)�����n��:���>���d�Here�#the�Shafarevic��rh-T��Vate�group���X�����(�A=�Q�)�is�a�measure�of�the�failure�of�the�lo�S�cal-to-global�����dprinciple;�Athe�cT��Vamaga��rw�a�n�um�b�S�ers��c�����p�����are�the�orders�of�the�comp�onen��rt�groups�of��A�;�Athe�real�����dn��rum�b�S�er��k
�����A��
�O�is�the�v��rolume�of��A�(�R�)�with�resp�ect�to�a�basis�of�dieren��rtials�ha�ving�ev�erywhere�����dnonzero��qgo�S�o�d�reduction;��.and��A����2�_��	�!�is�the�dual�of��A�.�#My�goal�is�to�v��rerify�the�full�conjecture�for�����dman��ry�d�sp�S�ecic�ab�elian�v��X�arieties�on�a�case-b��ry-case�basis.���This�is�the�rst�step�in�a�program�����dto��v��rerify�the�ab�S�o�v�e�conjecture�for�an�innite�family�of�quotien�ts�of��J�����0����(�N�@�).��"ʫ����d�2.2��$��The�ffratio��-��g�ffcmmi12�L�&X�Qffcmr12�(�A;�fd�1)�=�
��(��
�b>

cmmi10�A���U���d�F��Vollo��rwing�|�Y.�Manin's�w�ork�on�elliptic�curv�es,���A.�Agash���s�e�and�I�|�pro�v�ed�the�follo�wing�theorem�����din��[�2����].��������d�Theorem��1.���I�B�L��ffet����m��b�e�the�lar�gest�squar�e�dividing��N�@��.���The�r�atio��L�(�A;����1)�=�
�����A���|�is�a�r�ational�����dnumb��ffer�35that�c�an�b�e�explicitly�c�ompute�d,�up�to�a�unit�(c�onje�ctur�al���ly��1�)�in��Z�[1�=�(2�m�)]�.�����X�The�b�pro�S�of�uses�mo�dular�sym��rb�ols�com��rbined�with�an�extension�of�the�argumen�t�used�b�y�����dMazur�7yin�[�17����]�to�b�S�ound�the�Manin�constan��rt.��&The�ratio��L�(�A;����1)�=�
�����A��
]�is�expressed�as�the�lattice�����dindex�s`of�t��rw�o�s`mo�S�dules�o��rv�er�s`the�Hec��rk�e�s`algebra.�I�sBexp�ect�the�metho�d�to�giv��re�similar�results�for�����dsp�S�ecial��v��X�alues�of�t��rwists,��and�of��L�-functions�attac�hed�to�eigenforms�of�higher�w�eigh�t.���I��ha�v�e�����dcomputed����L�(�A;����1)�=�
�����A��
kعfor�all�optimal�quotien��rts�of�lev�el��N��6��UR�1500;���this�table�con�tin�ues�to�b�S�e�����dof��v��X�alue�to�n��rum�b�S�er��theorists.��"ʫ����d�2.3��$��The�fftorsion�subgroup��U���d�I��dcan��compute�upp�S�er�and�lo��rw�er��b�ounds�on�#�A�(�Q�)�����tor��^C�,�"but�I��dcan�not�determine�#�A�(�Q�)�����tor������d�in�+Aall�cases.���Exp�S�erimen��rtally��V,�;gthe�deviation�b�et��rw�een�+Athe�upp�er�and�lo��rw�er�+Ab�ound�is�re
ected�����din�<�congruences�with�forms�of�lo��rw�er�<�lev�el;���I�<whop�S�e�to�exploit�this�in�a�precise�w�a�y��V.�	/RI�<walso�����dobtained��the�follo��rwing�in�triguing�corollary�that�suggests�cancellation�b�S�et�w�een�torsion�and��c�����p���]�;�����dit��generalizes�to�higher�w��reigh�t��forms,�#�th�us�suggesting�a�geometric�explanation�for�reducibilit�y�����dof��Galois�represen��rtations.�������d�Corollary��2.���LV~�L��ffet����n��b�e�the�or�der�of�the�image�of��(0)�o����(�1�)����in��A�(�Q�)�,���and�let��m��b�e�the�lar�gest�����dsquar��ffe�35dividing��N�@��.�fiThen��n������L�(�A;����1)�=�
�����A���is�35an�inte�ger,�up�to�a�unit�in��Z�[1�=�(2�m�)]�.��"ʫ����d�2.4��$��T���famaga���w�a�ffn�um�b�s3ers��U�����d�Theorem��3.���I�B�When�35�p����2�2��V�#���
msbm10�-�UR�N�@��,�the�numb��ffer��c�����p�����c�an�b�e�explicitly�c�ompute�d�(up�to�a�p�ower�of��2�).������:�ǂ����6�;��䃍�������������	�Wil�8�liam��sA.�Stein�2�Detaile���Ud�R�ese�ar�ch�Plan��(�����d��k�ff���������2ō��y�;���X�I��Bpro��rv�e��pthis�in�[�25����].�K8Sev��reral�related�problems�remain:��pwhen��p����2�2��	B=�j��9�N��T�it�ma�y�b�S�e�p�ossible������dto�(�compute��c�����p��	��using�the�Drinfeld-Katz-Mazur�in��rterpretation�of��X�����0����(�N�@�);�ǻit�should�also�b�S�e�����dp�S�ossible��to�use�m��ry�metho�ds�to�treat�optimal�quotien��rts�of��J�����1����(�N�@�).�����XI�]iw��ras�]�surprised�to�nd�that�systematic�computations�using�this�form�ula�indicate�the�����dfollo��rwing��conjectural�renemen�t�of�a�result�of�Mazur�[�16����].��VS�����d�Conjecture��4.���U���Supp��ffose�.��N�o��is�prime�and��A��is�an�optimal�quotient�of��J�����0����(�N�@�)�.�d�Then��A�(�Q�)�����tor������d�is�8gener��ffate�d�by�the�image�of��(0)�N^���(�1�)��and��c�����p��	���=��#�A�(�Q�)�����tor��^C�.��sF���urthermor��ffe,�Gythe�pr�o�duct�of�����dthe�35�c�����p�����over�al���l�optimal�factors�e��ffquals�the�numer�ator�of��(�N�������1)�=�12�.�����X�I���ha��rv�e���c�hec�k�ed�this�conjecture�for�all��N��6��UR�997�and,��{up�to�a�p�S�o�w�er�of�2,��{for�all��N��6��UR�2113.�����dThe��.rst�part�is�kno��rwn�when��A��is�an�elliptic�curv�e�(see�[�20����]).��
Up�S�on�hearing�of�this�conjecture,�����dMazur��Epro��rv�ed�it�when�all�\�q�n9�-Eisenstein�quotien�ts"�are�simple.���There�are�three�promising�����dapproac��rhes�Ѽto�nding�a�complete�pro�S�of.��One�in�v�olv�es�the�explicit�form�ula�of�Theorem�3;�����danother���is�based�on�Rib�S�et's�lev��rel�lo�w�ering�theorem,���and�a�third�mak�es�use�of�a�simplicit�y�����dresult��of�Merel.�����XTheorem�Ov3�also�suggests�a�w��ra�y�Ovto�compute�T��Vamaga��rw�a�Ovn�um�b�S�ers�of�motiv�es�attac�hed�to�����deigenforms�A�of�higher�w��reigh�t.�=�These�A�n�um�b�S�ers�app�ear�in�the�conjectures�of�Blo�c��rh�and�Kato,�����dwhic��rh��generalize�the�BSD�conjecture�to�motiv�es�(see�[�5����]).��"�g����d�2.5��$��Upp�s3er�ffb�ounds�on��#���hV1ff
wncyr10�X���U���d�V.���Kolyv��X�agin�and�K.�Kato�[�12����,��23��L�]�obtained�upp�S�er�b�ounds�on�#��X��
�n�(�A�).��T��Vo�v��rerify�the�full�����dBSD�\8conjecture�\\for�certain�ab�S�elian�v��X�arieties,�x�it�is�necessary�is�to�mak��re�these�b�ounds�explicit.�����dKolyv��X�agin's�$b�S�ounds�in��rv�olv�e�$computations�with�Heegner�p�oin��rts,�2Xand�Kato's�in�v�olv�e�a�study�����dof��the�Galois�represen��rtations�asso�S�ciated�to��A�.��~I�Mplan�to�carry�out�suc�h�computations�in�man�y�����dsp�S�ecic��cases.������d�2.6��$��Lo���w�er�ffb�s3ounds�on��#��X���U���d�One�"�approac��rh�to�sho�wing�that���X��<̹is�as�large�as�predicted�b�y�the�BSD�"|conjecture�is�suggested�����db��ry��sMazur's�notion�of�the�visible�part�of���X��@T�(see�[�10����,��18��dk]).�[email protected]��A����2�_��
�#�b�S�e�the�dual�of��A�.�The�����dvisible���part�of���X���M�(�A����2�_��*��=�Q�)�is�the�k��rernel�of���X���(�A����2�_��*��=�Q�)���!���X���p�(�J�����0����(�N�@�)).���Mazur���observ��red�that�����dif� �an�elemen��rt�of�order��p��in���X��@�(�A����2�_��*��=�Q�)�is�visible,�.\then�it�is�explained�b�y�a�jump�in�the�rank�����dof�/�Mordell-W��Veil�in�the�sense�that�there�is�another�ab�S�elian�sub��rv��X�ariet�y�/��B�e������J�����0����(�N�@�)�suc��rh�that�����d�p�\R�j��#(�A����2�_��	>��\���B���)��+and�the�rank�of��B� 1�is�p�S�ositiv��re.�jI��think�that�this�observ��X�ation�can�b�e�turned�����daround:�if��there�is�another�ab�S�elian�v��X�ariet��ry��B�
�of�p�ositiv��re�rank�suc�h�that��p�UR�j��#(�A����2�_����\����B���),���then,�����dunder���mild�h��ryp�S�otheses,�'0there�is�an�elemen�t�of���X���O�(�A����2�_��*��=�Q�)�of�order��p�.�0�Th�us�the�theory�of�����dcongruences�j�b�S�et��rw�een�mo�S�dular�forms�can�b�e�used�to�obtain�a�lo��rw�er�b�S�ound�on�#��X��
�n�(�A����2�_��*��=�Q�).�-I�����dam�/�trying�to�use�the�cohomological�metho�S�ds�of�[�15����]�and�suggestions�of�B.�Conrad�and�Mazur�����dto��pro��rv�e�the�follo�wing�conjecture.��VS�����d�Conjecture��5.���U���L��ffet����A����2�_��	���and��B���b�e�ab�elian�subvarieties�of��J�����0����(�N�@�)�.�+�Supp�ose�that��p�UR�j��#(�A����2�_��O��\�%'�B���)�,�����dthat����p�UR�-��N�@��,��Hand�that��p��do��ffes�not�divide�the�or�der�of�any�of�the�torsion�sub�gr�oups�or�c�omp�onent�����dgr��ffoups�35of��A��or��B���.�fiThen��(�B��(�Q�)�������X����(�B�=�Q�))��
��Z�=p�Z�����P���UR����԰���n:�=�������(�A����2�_��*��(�Q�)�����X����(�A����2�_���=�Q�))��
��Z�=p�Z�.�����X�Unfortunately��V,��뺺X���(�(�A����2�_��*��=�Q�)��can�fail�to�b�S�e�visible�inside��J�����0����(�N�@�).�;rF�or�example,��I�found�that�����dthe��wBSD��7conjecture�predicts�the�existence�of�in��rvisible�elemen�ts�of�o�S�dd�order�in���X���\�for�at������!v�ǂ����6�;��䃍�������������	�Wil�8�liam��sA.�Stein�2�Detaile���Ud�R�ese�ar�ch�Plan��(�����d��k�ff���������2ō��y�;���d�least��815�of�the�37�optimal�quotien��rts�of�prime�lev�el���c�2113.�R�F��Vor�ev�ery�in�teger��M�4�(Rib�S�et�[�22����]������dtells��pus�whic��rh��M��T�to�c�ho�S�ose),���w�e�can�consider�the�images�of��A����2�_��	� �in��J�����0����(�N�@�M��).��There��pis�not�y�et�����denough�Ycevidence�to�conjecture�the�existence�of�an�in��rteger��M��G�suc�h�that�all�of���X��Pѹ(�A����2�_��*��=�Q�)�is�����dvisible�W;in��J�����0����(�N�@�M��).��I�Wam�W;gathering�data�to�determine�whether�or�not�to�exp�S�ect�the�existence�����dof��suc��rh��M�@�.��"ʫ����d�2.7��$��Motiv���ation�fffor�considering�ab�s3elian�v�arieties��U���d�If�R~�A��is�an�elliptic�curv��re,�ltthen�explaining���X��I�(�A=�Q�)�using�only�congruences�b�S�et�w�een�elliptic�����dcurv��res�`is�b�S�ound�to�fail.�
�This�is�b�ecause�pairs�of�nonisogenous�elliptic�curv��res�with�isomorphic�����d�p�-torsion��are,��Xaccording�to�E.�Kani's�conjecture,�extremely�rare.�A�It�is�crucial�to�understand�����dwhat��happ�S�ens�in�all�dimensions.�����XWithin�
�the�range�accessible�b��ry�computer,��ab�S�elian�v��X�arieties�exhibit�more�ric�hly�textured�����dstructure��"than�elliptic�curv��res.��NF��Vor�example,��I���disco�v�ered�a�visible�elemen�t�of�prime�order�����d83341��in�the�Shafarevic��rh-T��Vate�group�of�an�ab�S�elian�v��X�ariet�y�of�prime�conductor�2333;��Iin�����dcon��rtrast,��]o�v�er��Jall�optimal�elliptic�curv��res�of�conductor�up�to�5500,�it�app�S�ears�that�the�largest�����dorder��of�an�elemen��rt�of�a�Shafarevic�h-T��Vate�group�is�7.��(V����d�3��(XConjectures�z�of�Artin,�Merel,�and�Serre��"#����d�3.1��$��Icosahedral�ffGalois�represen���tations�����d�E.��wArtin�conjectured�in�[�3����]�that�the��L�-series�asso�S�ciated�to�an��ry�con�tin�uous�irreducible�rep-�����dresen��rtation��R��_��:��G�����)2�@�cmbx8�Q��<q�!���GL������n����(�C�),��|with��n�>��1,��|is�en��rtire.��Recen�t�exciting�w�ork�of�T��Va�ylor�and�����dothers�R�suggests�that�a�complete�pro�S�of�of�Artin's�conjecture,�l�in�the�case�when��n�M�=�2�R�and�������d�is���o�S�dd,�A+is�on�the�horizon.�n�This�case�of�Artin's�conjecture�is�kno��rwn�when�the�image�of�������d�in��e�PGL��������2��!�"�(�C�)�e�is�solv��X�able�(see�[�27����]),���and�in�innitely�man��ry�cases�when�the�image�of����is�not�����dsolv��X�able��(see�[�7����]).�����XIn�`1998,�*K.�Buzzard�suggested�a�w��ra�y�`to�com��rbine�the�main�theorem�of�[�8����],�along�with�a�����dcomputer��computation,��kto�deduce�mo�S�dularit��ry�of�certain�icosahedral�Galois�represen�tations.�����dBuzzard��and�I�recen��rtly�obtained�the�follo�wing�theorem.��������d�Theorem��6.���I�B�The�qkic��ffosahe�dr�al�A��2rtin�r�epr�esentations�of�c�onductor��1376�UR=�2����2�5���
����43��ar��ffe�mo�dular.�����X�W��Ve���exp�S�ect�our�metho�d�to�yield�sev��reral�more�examples.� �These�ongoing�computations�are������dla��rying�)a�small�part�of�the�tec�hnical�foundations�necessary�for�a�full�pro�S�of�of�the�Artin�con-�����djecture�mkfor�o�S�dd�t��rw�o�mkdimensional���,��was�w��rell�as�stim�ulating�the�dev�elopmen�t�of�new�algorithms�����dfor��computing�with�mo�S�dular�forms�using�mo�dular�sym��rb�ols�in�c��rharacteristic��`�.��"ʫ����d�3.2��$��Cyclotomic�ffp�s3oin���ts�on�mo�dular�curv���es�����d�If�#��E�׶�is�an�elliptic�curv��re�o�v�er��Q��and��p��is�an�o�S�dd�prime,�1�then�the��p�-torsion�on��E�׶�can�not�all�����dlie��?in��Q�;�k�b�S�ecause�of�the�W��Veil�pairing�the��p�-torsion�generates�a�eld�that�con��rtains��Q�(������p���]�).�����dF��Vor�!#whic��rh�primes��p��do�S�es�there�exist�an�elliptic�curv�e��E��:�o�v�er��Q�(������p���]�)�with�all�of�its��p�-torsion�����drational�<0o��rv�er��Q�(������p���]�)?�-wWhen��p���=�2�;����3�;��5�the�corresp�S�onding�mo�duli�space�has�gen��rus�zero�and�����dinnitely�sTman��ry�examples�exist.���Recen�t�w�ork�of�L.�Merel,��com�bined�with�computations�he�����denlisted��wme�to�do,��ksuggest�that�these�are�the�only�primes��p��for�whic��rh�suc�h�elliptic�curv�es������4��ǂ����6�;��䃍�������������	�Wil�8�liam��sA.�Stein�2�Detaile���Ud�R�ese�ar�ch�Plan��(�����d��k�ff���������2ō��y�;���d�exist.�$�In���[�19����],���Merel�exploits�cyclotomic�analogues�of�the�tec��rhniques�used�in�his�pro�S�of�of�the������duniform�1�b�S�oundedness�conjecture�to�obtain�an�explicit�criterion�that�can�b�e�used�to�answ��rer�����dthe�5�ab�S�o��rv�e�question�for�man�y�primes��p�,�H�on�a�case-b�y-case�basis.��Theoretical�w�ork�of�Merel,�����dcom��rbined�-pwith�m�y�computations�of�t�wisted��L�-v��X�alues�and�c�haracter�groups�of�tori,�>"giv�e�the�����dfollo��rwing��result�(see�[�19����,��x�3.2]):��<�����d�Theorem��7.���I�B�L��ffet�[��p������3�UP(�mo�S�d���B4)��b��ffe�a�prime�satisfying��7����p�<��1000�.��lTher��ffe�ar�e�no�el���liptic�����dcurves�35over��Q�(������p���]�)��al���l�of�whose��p�-torsion�is�r��ffational�over��Q�(������p���)�.�����X�The��Lcase�in�whic��rh��p��is�congruen�t�to�1�mo�S�dulo�4�presen�ts�additional�diculties�that�����din��rv�olv�e��sho�wing�that��Y��p�(�p�)�has�no��Q�(�������p���
�����z�孟z��p����对)-rational�p�S�oin�ts.���Merel�and�I��hop�S�e�to�tac�kle�these�����ddiculties��in�the�near�future.��"2L����d�3.3��$��Serre's�ffconjecture�mo�s3dulo��pq��U���d�Let����p��and��q��b�S�e�primes,��Jand�consider�a�con��rtin�uous���represen�tation���UR�:��G�����Q��1��!���GL����(2�;����Z�=pq�n9�Z�)�that�����dis�0zirreducible�in�the�sense�that�its�reductions�mo�S�dulo��p��and�mo�dulo��q����are�b�oth�irreducible.�����dCall�A0���mo��ffdular��if�there�is�a�mo�S�dular�form��f��/�suc��rh�that�a�mo�d��p��represen��rtation�attac�hed�����dto�;��f����is�the�mo�S�d��p��reduction�of���,�PPand�ditto�for��q�n9�.�,�I�;�ha��rv�e�;�carried�out�sp�ecic�computations�����dsuggested�Tb��ry�Mazur�in�hop�S�es�of�determining�when�one�should�exp�ect�that�suc��rh�mo�d��pq�����d�represen��rtations�rare�mo�S�dular;�P�the�computation�suggests�that�the�righ�t�conjectures�are�elusiv�e.�����dRib�S�et's�NStheorem�(see�[�22����])�pro�duces�innitely�man��ry�lev�els��pq�n9`��at�whic�h�there�is�a�form�giving�����drise��Vto����mo�S�d��p��and�another�giving�rise�to����mo�d��q�n9�;��qw��re�hop�e�to�determine�if�for�some��`��there�����dis��a�single�form�giving�rise�to�b�S�oth�reductions.��'|�����d�4��(XGen��u�us�z�one�curv�es���#���d�The���index�of�an�algebraic�curv��re��C��2�o�v�er��Q��is�the�order�of�the�cok�ernel�of�the�degree�map������dDiv���o�����Q��L)�(�C�ܞ�)���!��Z�;�_*rationalit��ry���of�the�canonical�divisor�implies�that�the�index�divides�2�g�����S��2,�����dwhere�PR�g����is�the�gen��rus�of��C�ܞ�.�i�When��g�p��=�]1�this�is�no�condition�at�all;��&Artin�conjectured,�i�and�����dLang�3�and�T��Vate�[�14����]�pro��rv�ed,�FJthat�3�for�ev��rery�in�teger��m��there�is�a�gen�us�one�curv�e�of�index��m�����d�o��rv�er�i�some�n��rum�b�S�er�i�eld.���Their�construction�yields�gen��rus�one�curv�es�o�v�er��Q��only�for�a�few�����dv��X�alues�?kof��m�,�T�and�they�ask�whether�one�can�nd�gen��rus�one�curv�es�o�v�er��Q��of�ev�ery�index.�7*I�����dha��rv�e��answ�ered�this�question�for�o�S�dd��m�.�������d�Theorem��8.���I�B�L��ffet�Q��K�.A�b�e�any�numb�er�eld.���Ther�e�ar�e�genus�one�curves�over��K�.A�of�every�o�dd�����dindex.�����X�The�~�pro�S�of�in��rv�olv�es�~�sho�wing�that�enough�cohomology�classes�in�Kolyv��X�agin's�Euler�system�����dof�vzHeegner�p�S�oin��rts�do�not�v��X�anish�com�bined�with�explicit�Heegner�p�S�oin�t�computations.�	�WI�����dhop�S�e���to�sho��rw�that�curv�es�of�ev�ery�index�o�S�ccur,���and�to�determine�the�consequences�of�m�y�����dnon��rv��X�anishing��2result�for�Selmer�groups.�cThis�can�b�S�e�view�ed�as�a�con�tribution�to�the�problem�����dof��understanding��H���V���2�1���Z�(�Q�;���E���).��'|����d�References��b#�����`�[1]���&pA.���Agash��r��s�e,��n�On�Iinvisible�elements�of�the�Tate-Shafar��ffevich�gr�oup�,��nC.���R.�Acad.�Sci.�P��raris����&pS��r��s�er.��I�Math.��328��(1999),�no.�5,�369{374.������Ea�ǂ����6�;��䃍�������������	�Wil�8�liam��sA.�Stein�2�Detaile���Ud�R�ese�ar�ch�Plan��(�����d��k�ff���������2ō��y�;�����`�[2]���&pA.��-Agash��r��s�e�and�W.��oA.�Stein,�@��Visibility�.�of�Shafar��ffevich-Tate�gr�oups�of�mo�dular�ab�elian�����&pvarieties�,��in�preparation�(1999).��������`[3]���&pE.�B�Artin,���֟��x���[email protected]��ffer���eine�neue�Art�von�L-Reihen�,�Abh.�B�Math.�Sem.�Univ.�Ham��rburg�lm�3��(1923),����&p89{108.�������`[4]���&pB.��oJ.�u�Birc��rh,��j�El���liptic���curves�over��Q�:�e�A���pr��ffo�gr�ess���r�ep�ort�,�1969�u�Num��rb�S�er�Theory�Institute����&p(Pro�S�c.�&8Symp�os.�Pure�Math.,�M�V��Vol.�XX,�State�Univ.�New�Y�ork,�M�Ston��ry�Bro�S�ok,�N.Y.,�1969),����&pAmer.��Math.�So�S�c.,�Pro��rvidence,�R.I.,�1971,�pp.�396{400.�������`[5]���&pS.��
Blo�S�c��rh�and�K.�Kato,�
f�L�-functions���and�T���amagawa�numb��ffers�of�motives�,�The����&pGrothendiec��rk��F��Vestsc�hrift,�V��Vol.�I,�Birkh�� auser�Boston,�Boston,�MA,�1990,�pp.�333{400.�������`[6]���&pC.�ԼBreuil,��B.�Conrad,�F.�Diamond,�and�R.�T��Va��rylor,��On�the�mo��ffdularity�of�el���liptic�curves����&pover�35�Q�,��in�preparation.�������`[7]���&pK.�t�Buzzard,��M.�Dic��rkinson,�N.�Shepherd-Barron,�and�R.�T��Va��rylor,��On���ic��ffosahe�dr�al�Artin����&pr��ffepr�esentations�,��a��rv��X�ailable�at��9߆�Tcmtt12�http://www.math.harvard.edu/~rtaylor/�.�������`[8]���&pK.�XOBuzzard�and�R.�T��Va��rylor,�s��Comp��ffanion���forms�and�weight�one�forms�,�Annals�XOof�Math.����&p(1999).�������`[9]���&pJ.��oE.���Cremona,��~�A��2lgorithms��for�mo��ffdular�el���liptic�curves�,�second���ed.,�Cam��rbridge�Univ�ersit�y����&pPress,��Cam��rbridge,�1997.�������d[10]���&pJ.��oE.��Cremona�and�B.�Mazur,�Ј�Visualizing�1elements�in�the�Shafar��ffevich-T���ate�gr�oup�,�ЈPro-����&pceedings��of�the�Arizona�Win��rter�Sc�ho�S�ol�(1998).�������d[11]���&pN.��oD.�RElkies,�I|�El���liptic�5/and�mo��ffdular�curves�over�nite�elds�and�r�elate�d�c�omputational����&pissues�,��Computational��'p�S�ersp�ectiv��res�on�n�um�b�S�er�theory�(Chicago,��IL,�1995),�Amer.�Math.����&pSo�S�c.,��Pro��rvidence,�RI,�1998,�pp.�21{76.�������d[12]���&pV.��oA.�YyKolyv��X�agin,��.�On�o�the�structur��ffe�of�Shafar�evich-Tate�gr�oups�,��.Algebraic�Yygeometry����&p(Chicago,��IL,�1989),�Springer,�Berlin,�1991,�pp.�94{121.�������d[13]���&pV.��oA.�Kolyv��X�agin�and�D.�Y.�Logac��rhev,�Jj�Finiteness�5�of���X��c*�over�total���ly�r��ffe�al�5�elds�,�Math.����&pUSSR��Izv��restiy�a��39��(1992),�no.�1,�829{853.�������d[14]���&pS.�P�Lang�and�J.�T��Vate,��A�Princip��ffal�|Xhomo�gene�ous�sp�ac�es�over�ab�elian�varieties�,��AAmer.�P�J.����&pMath.���80��(1958),�659{684.�������d[15]���&pB.���Mazur,��N�R��ffational��p�oints�of�ab�elian�varieties�with�values�in�towers�of�numb�er�elds�,����&pIn��rv�en�t.��Math.��18��(1972),�183{266.�������d[16]����&p�ff#?莑=[�,��`�Mo��ffdular��Ocurves�and�the�Eisenstein�ide�al�,��`Inst.���Hautes���ٟ��x��Etudes�Sci.�Publ.�Math.����&p(1977),��no.�47,�33{186�(1978).�������d[17]����&p�ff#?莑=[�,�B7�R��ffational�s�iso�genies�of�prime�de�gr�e�e�(with�an�app�endix�by�D.�Goldfeld)�,�B7In��rv�en�t.����&pMath.���44��(1978),�no.�2,�129{162.�������d[18]����&p�ff#?莑=[�,��+�Visualizing���elements�of�or��ffder�thr�e�e�in�the�Shafar�evich-Tate�gr�oup�,��+preprin��rt����&p(1999).�������d[19]���&pL.��hMerel,����Sur��Ula�natur��ffe�non-cyclotomique�des�p�oints�d'or�dr�e�ni�des�c�ourb�es�el���liptiques�,����&ppreprin��rt��(1999).�������d[20]���&pJ.-F.�xcMestre�and�J.�Oesterl��r��s�e,����Courb��ffes��{de�Weil�semi-stables�de�discriminant�une�puis-����&psanc��ffe�35�m�-i��3��L�eme�,��J.�Reine�Angew.�Math.��400��(1989),�173{184.�������d[21]���&pK.��oA.�zoRib�S�et,��a�On���mo��ffdular�r�epr�esentations�of���Gal���[(���S��z�
#��	�\��Q���
#��=�Q�)��arising�fr�om�mo�dular�forms�,����&pIn��rv�en�t.��Math.��100��(1990),�no.�2,�431{476.������UW�ǂ����6�;��䃍�������������	�Wil�8�liam��sA.�Stein�2�Detaile���Ud�R�ese�ar�ch�Plan��(�����d��k�ff���������2ō��y�;�����d�[22]����&p�ff#?莑=[�,�<�R��ffaising�Q�the�levels�of�mo�dular�r�epr�esentations�,�<S��r��s�eminaire��de�Th���s�eorie�des�Nom-�����&pbres,��P��raris�1987{88,�Birkh�� auser�Boston,�Boston,�MA,�1990,�pp.�259{271.��������d[23]���&pA.��oJ.���Sc��rholl,��,�A��2n��3intr��ffo�duction�to�Kato's�Euler�systems�,��,Galois���Represen��rtations�in�Arith-����&pmetic��Algebraic�Geometry��V,�Cam��rbridge�Univ�ersit�y�Press,�1998,�pp.�379{460.�������d[24]���&pJ-P��V.��HSerre,��\�Sur��les�r��ffepr��3��L�esentations�mo�dulair�es�de�de�gr��3��L�e��2��de���Gal��w�(���S��z�
#��	�\��Q���
#��=�Q�),��\Duk��re��HMath.�J.����&p�54�꨹(1987),�no.�1,�179{230.�������d[25]���&pW.��oA.��Stein,��Comp��ffonent�35gr�oups�of�optimal�quotients�of�Jac�obians�,��preprin��rt�(1999).�������d[26]���&pR.�!CT��Va��rylor�and�A.��oJ.�Wiles,�.��R��2ing-the��ffor�etic�efpr�op�erties�of�c�ertain�He�cke�algebr�as�,�.�Ann.�!Cof����&pMath.��(2)��141��(1995),�no.�3,�553{572.�������d[27]���&pJ.�˚T��Vunnell,����A��2rtin����'s��c��ffonje�ctur�e�for�r�epr�esentations�of�o�ctahe�dr�al�typ�e�,���Bull.�˚Amer.�Math.����&pSo�S�c.��(N.S.)��5��(1981),�no.�2,�173{175.�������d[28]���&pA.��oJ.���Wiles,�
�Mo��ffdular�B�el���liptic�curves�and�Fermat's�last�the�or�em�,�
Ann.���of�Math.�(2)��141����&p�(1995),��no.�3,�443{551.������eV���;��ܙ��	�9߆�Tcmtt12�-��g�ffcmmi12�,���@cmti12�+F
C��q
cmbxti10�)2�@�cmbx8�(��N�cmbx12�'��N�ffcmbx12�&X�Qffcmr12�#���
msbm10���u
cmex10��K�cmsy8�!",�
cmsy10��2cmmi8���g�cmmi12�|{Ycmr8���N�G�cmbx12���N��qcmbx12��hV1ff
wncyr10��hV1
wncyr10�X�Qcmr12�
�b>

cmmi10�km�����