����;� TeX output 1999.09.23:0824�������K����0�����q�)p�0J
�3
cmsl10�William��fA.�Stein�x$��K�`y
�3
cmr10�1�����?���r����uv���N�ffcmbx12�Researc���h�ffPlan���f����William��fA.�Stein�����E�Septem��!b�M�er��f23,�1999��"�A����q�1��**<In���tro�s3duction�����q�My�X�researc��!h�program�sits�at�the�in�tersection�of�recen�t�w�ork�of�B.�Mazur,���K.�Rib�M�et,�J-P��e.�Serre,�����qR.�T��ea��!ylor,�`�and�A.�Wiles�on�Galois�represen�tations�attac�hed�to�mo�M�dular�ab�elian�v��dDarieties�(see�����q[�21��
�4,���24����,���26��,���28��])��with�w��!ork�of�J.�Cremona,��N.�Elkies,�and�J.-F.�Mestre�on�explicit�computations�����qin��!v�olving�=�mo�M�dular�forms�(see�[�9��y�,��11��1-]).��In�1969�B.�Birc��!h�[�4��]�describ�M�ed�computations�that�led�to�the�����qmost��ffundamen��!tal�op�M�en�conjecture�in�the�theory�of�elliptic�curv�es:���f��-=uI���w��!an�t���to�describ�M�e�some�computations�undertak��!en�b�y�m�yself�and�Swinnerton-Dy�er�on����-=uEDSA��!C���b�y���whic�h�w�e�ha�v�e�calculated�the�zeta-functions�of�certain�elliptic�curv�es.�`�As����-=ua��.result�of�these�computations�w��!e�ha�v�e�found�an�analogue�for�an�elliptic�curv�e�of�the����-=uT��eamaga��!w�a�Z�n�um�b�M�er�of�an�algebraic�group;���and�conjectures�(due�to�ourselv�es,���due�to����-=uT��eate,��fand�due�to�others)�ha��!v�e��fproliferated.�����qThe�يric��!h�tap�M�estry�of�arithmetic�conjectures�and�theory�w�e�enjo�y�to�M�da�y�w�ould�not�exist�without�the�����qground-breaking���application�of�computing�b��!y�Birc�h�and�Swinnerton-Dy�er.��<Computations�in�the�����q1980s�̐b��!y�Mestre�w�ere�k�ey�in�con�vincing�Serre�that�his�conjectures�on�mo�M�dularit�y�of�o�M�dd�irreducible�����qGalois��@represen��!tations�w�ere�w�orth�y�of�serious�consideration�(see�[�24��
�4]).�lThese�conjectures�ha�v�e�����qinspired�{�m��!uc�h�recen�t�w�ork;��Ufor�example,���Rib�M�et's�pro�of�of�the���b>
�3
cmmi10��-conjecture,���whic��!h�pla�y�ed�an�essen�tial�����qrole��fin�Wiles's�pro�M�of�of�F��eermat's�Last�Theorem.����"�qMy�0�w��!ork�on�the�Birc�h�and�Swinnerton-Dy�er�conjecture�for�mo�M�dular�ab�elian�v��dDarieties�and�����qsearc��!h��pfor�new�examples�of�mo�M�dular�icosahedral�Galois�represen�tations�has�led�me�to�disco�v�er�����qand�g}implemen��!t�algorithms�for�explicitly�computing�with�mo�M�dular�forms.�!"My�researc�h�program�����qre
ects���the�essen��!tial�in�terpla�y�b�M�et�w�een�abstract�theory�and�explicit�mac�hine�computation�during�����qthe��flatter�half�of�the�t��!w�en�tieth��fcen�tury��e.��"�A����q�2��**<In���v���arian�ts�ffof�mo�s3dular�ab�elian�v���arieties�����q�No��!w�#that�the�Shim�ura-T��eaniy�ama�conjecture�has�b�M�een�pro�v�ed,�<the�main�outstanding�problem�in�����qthe��Held�is�the�Birc��!h�and�Swinnerton-Dy�er�conjecture�(BSD��=conjecture),�ށwhic�h�ties�together�the�����qarithmetic�^�in��!v��dDarian�ts�of�an�elliptic�curv�e.��There�is�no�general�class�of�elliptic�curv�es�for�whic�h�the�����qfull���BSD��Kconjecture�is�kno��!wn.�}�Approac�hes���to�the�BSD�conjecture�that�rely�on�congruences�b�M�et��!w�een�����qmo�M�dular��forms�require�a�deep�understanding�of�the�analogous�conjecture�for�higher-dimensional�����qab�M�elian��Pv��dDarieties.���As�a�rst�step,���I��ha��!v�e��Pobtained�theorems�that�mak��!e�p�ossible�explicit�computation�����qof��fsome�of�the�arithmetic�in��!v��dDarian�ts��fof�mo�M�dular�ab�elian�v��dDarieties.��R�����q�%��N�cmbx12�2.1��0�qThe��BSD�conjecture��Aݍ��q�By��Z[�6��y�]�w��!e�no�w�kno�w�that�ev�ery�elliptic�curv�e�o�v�er��&�"V
�3
cmbx10�Q��is�a�quotien�t�of�the�curv�e��X���z�|{Ycmr8�0����(�N�1��)�whose�����qcomplex���p�M�oin��!ts�are�the�isomorphism�classes�of�pairs�consisting�of�a�(generalized)�elliptic�curv�e�����qand�"�a�cyclic�subgroup�of�order��N�1��.�S9Let��J���z�0����(�N��)�denote�the�Jacobian�of��X���z�0����(�N��);�athis�is�an�ab�M�elian�����*��K����0�����q�William��fA.�Stein�x$�2�����?���r����qv��dDariet��!y�CWof�dimension�equal�to�the�gen�us�of��X���z�0����(�N�1��)�whose�p�M�oin�ts�corresp�M�ond�to�the�degree�0�divisor���f���qclasses��fon��X���z�0����(�N�1��).����"�qAn����*�':
�3
cmti10�optimal���quotient��of��J���z�0����(�N�1��)�is�a�quotien��!t�b�y�an�ab�M�elian�sub�v��dDariet�y��e.��[Consider�an�optimal�����qquotien��!t�@�A��suc�h�that��L�(�A;��1�1)��R�!",�
�3
cmsy10�6�=�0.�	��By�@[�13��
�4],��A�(�Q�)�and����hV1
wncyr10�X��7o�(�A=�Q�)�are�b�M�oth�nite.�The�BSD�����qconjecture��fasserts�that��ى�����������M��L�(�A;��1�1)���M��㦉p "�ߟ
���	߹
������2cmmi8�A�������f�=�����}�=�#��X��
�n�(�A=�Q�)�n������ɖ���u
�3
cmex10�Q����e�<��p��K�cmsy8�j�N��V��c���z�p���=ڟ�R�p c�
���1�#�A�(�Q�)�n����#�A����0�_��*��(�Q�)�����h��:���ɍ��q�Here�the�Shafarevic��!h-T��eate�group���X����(�A=�Q�)�is�a�measure�of�the�failure�of�the�lo�M�cal-to-global�princi-�����qple;�~�the�j�T��eamaga��!w�a�n�um�b�M�ers��c���z�p��2	�are�the�orders�of�the�comp�onen��!t�groups�of��A�;�~�the�real�n�um�b�M�er�
�����A������q�is��the�v��!olume�of��A�(�R�)�with�resp�M�ect�to�a�basis�of�dieren�tials�ha�ving�ev�erywhere�nonzero�go�M�o�d�����qreduction;�and����A�����_��
ŏ�is�the�dual�of��A�.��IMy�goal�is�to�v��!erify�the�full�conjecture�for�man�y�sp�M�ecic�����qab�M�elian�Vv��dDarieties�on�a�case-b��!y-case�basis.�!�This�is�the�rst�step�in�a�program�to�v�erify�the�ab�M�o�v�e�����qconjecture��ffor�an�innite�family�quotien��!ts�of��J���z�0����(�N�1��).��~����q�2.2��0�qThe��ratio��+��g�cmmi12�L�$X�Qcmr12�(�A;����1)�=�
�����A���Aݍ��q�Generalizing��fw��!ork�of�Y.�Manin,�A.�Agash��"���De�and�I�pro�v�ed�the�follo�wing�theorem�in�[�2��y�].��H������q�Theorem�2�1.���V�4�L��p�et����m��b�e�the�lar�gest�squar�e�dividing��N�1��.��sThe�r�atio��L�(�A;��1�1)�=�
�����A��
���is�a�r�ational�numb�er�����qthat���c��p�an�b�e�explicitly�c�ompute�d,�up�to�a�unit�in��Z�[1�=�(2�m�)]�.����"�q�My��7pro�M�of�uses�mo�dular�sym��!b�ols�com��!bined�with�an�extension�of�the�argumen�t�used�b�y�Mazur�����qin��2[�17��
�4]�to�b�M�ound�the�Manin�constan��!t.��wThe�ratio��L�(�A;��1�1)�=�
�����A��	��is�expressed�as�the�lattice�index�of�t�w�o�����qmo�M�dules���o��!v�er�the�Hec�k�e�algebra.��I���exp�M�ect�the�metho�d�to�giv��!e�similar�results�for�sp�ecial�v��dDalues�of�����qt��!wists,��-and���of��L�-functions�attac�hed�to�eigenforms�of�higher�w�eigh�t.��!I���ha�v�e�computed��L�(�A;��1�1)�=�
�����A������q�for��all�optimal�quotien��!ts�of�lev�el��N�<F��
��1500;�K=this�data�con�tin�ues�to�b�M�e�of�v��dDalue�to�n�um�b�M�er�theorists.��~����q�2.3��0�qThe��torsion�subgroup��Aݍ��q�I�<�ha��!v�e�<�found�go�M�o�d�<�b�ounds�on�#�A�(�Q�)���z�tor��^C�,�Rbut�I�<�can�not�y��!et�determine�#�A�(�Q�)���z�tor���/�in�all�cases.���I�also�����qobtained��fthe�follo��!wing�in�triguing�corollary�that�suggests�cancellation�b�M�et�w�een�torsion�and��c���z�p���]�.�������q�Corollary�2�2.���Y'U�L��p�et����n��b�e�the�or�der�of�the�image�of��(0)�Q���(�1�)����in��A�(�Q�)�,��and�let��m��b�e�the�lar�gest�����qsquar��p�e���dividing��N�1��.�	vThen��n�n����L�(�A;��1�1)�=�
�����A��
���is���an�element�of��Z�[1�=�(2�m�)]�.��~����q�2.4��0�qT���amaga��w�a��n�um�b�`ers��Aݍ����q�Theorem�2�3.���V�4�When����p�����2��ʫ�!���
�3
msbm10�-�
��N�1��,�the�numb��p�er��c���z�p���G�c�an�b�e�explicitly�c�ompute�d�(up�to�a�p�ower�of��2�).����"�q�I�P�pro��!v�e�P�this�in�[�25��
�4].���Sev��!eral�related�problems�remain:�2�when��p�����2����j�&��N��^�it�should�b�M�e�p�ossible�to�����qcompute����c���z�p��V �using�the�Drinfeld-Katz-Mazur�in��!terpretation�of��X���z�0����(�N�1��);���it�should�also�b�M�e�p�ossible�to�����quse��fm��!y�metho�M�ds�to�treat�optimal�quotien�ts�of��J���z�1����(�N�1��).����"�qI�Q�w��!as�Q�surprised�to�nd�that�systematic�computations�using�this�form�ula�indicate�the�follo�wing�����qconjectural��frenemen��!t�of�a�result�of�Mazur�[�16��
�4].�������q�Conjecture�2�4.���a���Supp��p�ose����N��%�is�prime�and��A��is�an�optimal�quotient�of��J���z�0����(�N�1��)�.�oIThen��A�(�Q�)���z�tor����is�����qgener��p�ate�d���by�the�image�of��(0)�-���(�1�)����and��c���z�p����=�
�#�A�(�Q�)���z�tor��^C�.���F��)urthermor��p�e,���the�pr�o�duct�of�the��c���z�p�����over�����qal��Fl���optimal�factors�e��p�quals�the�numer�ator�of��(�N�����n�1)�=�12�.������K����0�����q�William��fA.�Stein�x$�3�����?���r���"�qI���ha��!v�e��-c�hec�k�ed�this�conjecture�for�all��N��;����631�and,��up�to�a�p�M�o�w�er�of�2,��for�all��N��;����2113.���f���qThe�2Mrst�part�is�kno��!wn�when��A��is�an�elliptic�curv�e�(see�[�20��
�4]).���Up�M�on�hearing�of�this�conjecture,�����qMazur�_Jobtained�some�partial�results.��)There�are�t��!w�o�_Jdistinct�and�promising�approac��!hes�to�nding�����qa��pro�M�of.���One�in��!v�olv�es��the�explicit�form��!ula�of�Theorem�3;��ethe�other�is�based�on�Rib�et's�lev��!el�����qlo��!w�ering��ftheorem.����"�qTheorem��3�also�suggests�a�w��!a�y��to�compute�T��eamaga��!w�a��n�um�b�M�ers�of�motiv�es�attac�hed�to�eigen-�����qforms�Aof�higher�w��!eigh�t.�	nThese�An�um�b�M�ers�app�ears�in�the�Blo�c��!h-Kato�conjecture�[�5��y�],�_wwhic�h�is�a�����qgeneralization��fof�the�BSD�conjecture.��R�����q�2.5��0�qUpp�`er��b�ounds�on��#���hV1ff
wncyr10�X���Aݍ��q�V.�-�Kolyv��dDagin�and�K.�Kato�[�12��
�4,��23��!]�obtained�upp�M�er�b�ounds�on�#��X��
�n�(�A�).�tRT��eo�v��!erify�the�full�BSD�����qconjecture�{4for�certain�ab�M�elian�v��dDarieties,��it�is�necessary�is�to�mak��!e�these�b�ounds�explicit.�z"Kolyv��dDagin's�����qb�M�ounds�_�in��!v�olv�e�computations�with�Heegner�p�M�oin�ts,���and�Kato's�in�v�olv�e�a�study�of�the�Galois�����qrepresen��!tations��fasso�M�ciated�to��A�.���I�plan�to�carry�out�b�oth�of�these�computations�in�man��!y�cases.������q�2.6��0�qLo��w�er��b�`ounds�on��#��X���Aݍ��q�One��approac��!h�to�sho�wing�that���X���z�is�as�large�as�predicted�b�y�the�BSD���conjecture�is�suggested�����qb��!y��Mazur's�notion�of�the�visible�part�of���X��Ң�(see�[�10��
�4,��18����]).��xLet��A�����_��
J�b�M�e�the�dual�of��A�.�The�visible�����qpart���of���X���k�(�A�����_��*��=�Q�)�is�the�k��!ernel�of���X���(�A�����_��*��=�Q�)�B��!���X��:�(�J���z�0����(�N�1��)).�B�Mazur���observ��!ed�that�if�an�elemen�t�����qof���order��p��in���X���9�(�A�����_��*��=�Q�)�is�visible,�Ǥthen�it�is�explained�b��!y�a�jump�in�the�rank�of�Mordell-W��eeil�����qin�<�the�sense�that�there�is�another�ab�M�elian�v��dDariet��!y��B�����W�J���z�0����(�N�1��)�suc�h�that��p�W�j��#(�A�����_�����\��3�B����)�<�and�the�����qrank��Jof��B�+�is�p�M�ositiv��!e.�ȊI��
think�that�this�observ��dDation�can�b�e�turned�around:�Ϧif�there�is�another�����qab�M�elian�s�v��dDariet��!y��B�$�of�p�ositiv��!e�rank�suc�h�that��p�
��j��#(�A�����_��3��\�	�B����),�}�then,�under�s�mild�h�yp�M�otheses,�}�there�is�����qan���elemen��!t�of���X���6�(�A�����_��*��=�Q�)�of�order��p�.��In�some�cases,��`the�theory�of�congruences�b�M�et�w�een�mo�M�dular�����qforms�6can�b�M�e�used�to�obtain�a�lo��!w�er�6b�ound�on�#��X��
�n�(�A�����_��*��=�Q�).��pI�6am�trying�to�use�the�cohomological�����qmetho�M�ds��fof�[�15��
�4]�and�suggestions�of�B.�Conrad�to�pro��!v�e��fthe�follo��!wing�conjecture.���f�����q�Conjecture�2�5.���a���L��p�et���A�����_��
���and��B�W��b�e�ab�elian�subvarieties�of��J���z�0����(�N�1��)�.���Supp�ose�that��p����j��#(�A�����_��	<�\�W�B����)�,�����qthat��V�p�w��-��N�1��,��and�that��p��do��p�es�not�divide�the�or�der�of�any�of�the�torsion�sub�gr�oups�or�c�omp�onent�����qgr��p�oups���of��A��or��B����.�	vThen��B��(�Q�)�n��
��Z�=p�Z���������
������l���
��=���������X����(�A�����_��*��=�Q�)��
��Z�=p�Z�.����"�q�Unfortunately��e,��b}�X��Y�(�A�����_��*��=�Q�)�Q�can�fail�to�b�M�e�visible�inside��J���z�0����(�N�1��).���F�or�example,�b}I�Qmfound�that���X���r�is�����qin��!visible��for�at�least�15�of�the�37�optimal�quotien�ts�at�prime�lev�el���
��2113�suc�h�that���X��Բ�should�ha�v�e�����qnon��!trivial���elemen�ts�of�o�M�dd�order.�$nF��eor�ev�ery�in�teger��M�(Rib�M�et�[�22��
�4]�tells�us�whic�h��M�to�consider),�����qw��!e�O�can�consider�the�images�of��A�����_��
z��in��J���z�0����(�N�1�M��).��cThere�O�is�not�y�et�enough�evidence�to�conjecture�����qthat�pthere�is�an�in��!teger��M����suc�h�that�all�of���X��g��(�A�����_��*��=�Q�)�is�visible�in�some��J���z�0����(�N�1�M��).���I�pam�pgathering�����qevidence��fin�order�to�determine�whether�or�not�one�exp�M�ects�the�existence�of�suc��!h��M�1��.��R�����q�2.7��0�qMotiv��@ation��for�considering�ab�`elian�v�arieties��Aݍ��q�If�kD�A��is�an�elliptic�curv��!e,�wthen�explaining���X��b��(�A=�Q�)�using�only�congruences�b�M�et�w�een�elliptic�curv�es�����qis��b�M�ound�to�fail.��lThis�is�b�ecause�one�exp�ects�the�existence�of�an�in��!teger��b��suc�h�that�if��A��and��B�����q�are���nonisogeneous�elliptic�curv��!es�o�v�er��Q�,��and��p��is�greater�than��b�,�then��A�[�p�]�is�not�isomorphic�to�����q�B����[�p�].���It��fis�crucial�to�understand�what�happ�M�ens�in�all�dimensions.�����"��K����0�����q�William��fA.�Stein�x$�4�����?���r���"�qWithin�K�the�range�accessible�b��!y�computer,�]�ab�M�elian�v��dDarieties�exhibit�more�ric�hly�textured�struc-���f���qture��	than�elliptic�curv��!es.��F��eor�example,���I���disco�v�ered�a�visible�elemen�t�of�prime�order�83341�in�����qthe�X�Shafarevic��!h-T��eate�group�of�an�ab�M�elian�v��dDariet�y�of�prime�conductor�2333;���in�con�trast,��o�v�er�all�����qoptimal�p�elliptic�curv��!es�of�conductor�up�to�5500,�{�it�app�M�ears�that�the�largest�order�of�an�elemen�t�of�����qa��fShafarevic��!h-T��eate�group�is�7.��"�A����q�3��**<Conjectures�ffof�Artin,�Merel,�and�Serre��Ō����q�3.1��0�qCyclotomic��p�`oin��ts�on�mo�dular�curv��es��Aݍ��q�If�_��E�~�is�an�elliptic�curv��!e�o�v�er��Q��and��p��is�an�o�M�dd�prime,��Nthen�the��p�-torsion�on��E�~�can�not�all�lie�����qin��7�Q�;��b�M�ecause�of�the�W��eeil�pairing�the��p�-torsion�generates�a�eld�that�con��!tains��Q�(����z�p���]�).�RQF�or�whic��!h�����qprimes��%�p��do�M�es�there�exist�an�elliptic�curv��!e��E�8��o�v�er��Q�(����z�p���]�)�with�all�of�its��p�-torsion�rational�o�v�er�����q�Q�(����z�p���]�)?�H�When��
�p��=�2�;��1�3�;��5�the�corresp�M�onding�mo�duli�space�has�gen��!us�zero�and�innitely�man�y�����qexamples�� exist.�Recen��!t�w�ork�of�L.�Merel,�)�com�bined�with�computations�he�enlisted�me�to�do,�����qsuggest��Athat�these�are�the�only�primes��p��for�whic��!h�suc�h�elliptic�curv�es�exist.��mIn�[�19��
�4],��Merel�uses�����qcyclotomic��analogues�of�the�tec��!hniques�used�in�his�pro�M�of�of�the�uniform�b�oundedness�conjecture�����q(mo�M�dular�Y|sym��!b�ols,�h�Euler�systems,�etc.)��:to�obtain�an�explicit�criterion�that�can�b�M�e�used�to�answ��!er�����qthe�1:ab�M�o��!v�e�question�for�man�y�primes��p��on�a�case-b�y-case�basis.���Computations�I�1ha�v�e�done�giv�e�the�����qfollo��!wing��fresult�(see�[�19��
�4,��x�3.2]):���f�����q�Theorem�2�6.���V�4�L��p�et����p�
����3���(�mo�M�d���14)��b��p�e�a�prime�satisfying��7����p�<��1000�.��STher��p�e�ar�e�no�el��Fliptic�curves�����qover����Q�(����z�p���]�)��al��Fl�of�whose��p�-torsion�is�r��p�ational�over��Q�(����z�p���)�.����"�q�The�Y'case�in�whic��!h��p��is�congruen�t�to�1�mo�M�dulo�4�presen�ts�additional�diculties�that�in�v�olv�e�����qsho��!wing��f�Y�n�(�p�)�has�no��Q�(����-t�p���	 ��-t�p �\�Ҍ��p�����]�)-rational�p�M�oin�ts.���W��ee�hop�M�e�to�tac�kle�these�in�the�near�future.��R�����q�3.2��0�qIcosahedral��Artin�represen��tations��Aݍ��q�Artin���conjectured�in�[�3��y�]�that�the��L�-series�asso�M�ciated�to�an��!y�con�tin�uous�irreducible�represen�tation�����q��
��:��G�����'2�@�cmbx8�Q��
��!���GL���ze��z�n��"��(�C�),�Z�with�G��n�>��1,�is�G�en��!tire.��dRecen�t�w�ork�of�T��ea�ylor�suggests�that�a�complete�pro�M�of�of�����qArtin's�cconjecture�ma��!y�b�M�e�on�the�horizon�when��n�
��=�2�cand����is�o�dd.��mThis�case�of�the�conjecture�is�����qkno��!wn��when�the�image�of����in��PGL�������z�2��X��(�C�)�is�solv��dDable�(see�[�27��
�4]),���and�in�innitely�man�y�cases�when�����qthe��fimage�of����is�not�solv��dDable�(see�[�7��y�]).����"�qIn�s_1998,���K.�Buzzard�suggested�a�w��!a�y�s_to�com��!bine�the�main�theorem�of�[�8��y�],�along�with�a�computer�����qcomputation,�δto�ƥdeduce�mo�M�dularit��!y�of�certain�icosahedral�Galois�represen�tations.�>�Buzzard�and�I�����qrecen��!tly��fobtained�the�follo�wing�theorem.�������q�Theorem�2�7.���V�4�The���ic��p�osahe�dr�al�A��\rtin�r�epr�esentations�of�c�onductor��1376�
�=�2�����5��.���n�43��ar��p�e�mo�dular.����"�q�W��ee���exp�M�ect�our�metho�d�to�yield�sev��!eral�more�examples.���These�ongoing�computations�are�la�ying�����qa�Kksmall�part�of�the�tec��!hnical�foundations�necessary�for�a�full�pro�M�of�of�the�Artin�conjecture�for�o�dd�����qt��!w�o���dimensional���,���as�w��!ell�as�stim�ulating�the�dev�elopmen�t�of�new�algorithms�for�computing�with�����qmo�M�dular��fsym��!b�ols�in�c��!haracteristic��`�.�����5��K����0�����q�William��fA.�Stein�x$�5�����?���r�����q�3.3��0�qSerre's��conjecture�mo�`dulo��pq��Aݍ��q�Let����p��and��q�!*�b�M�e�primes,��#and�consider�a�con��!tin�uous���represen�tation���/��:��G�����Q���!���GL���a(2�;��1�Z�=pq�d��Z�)�that�is���f���qirreducible��in�the�sense�that�its�reductions�mo�M�dulo��p��and�mo�dulo��q�sq�are�b�oth�irreducible.�ECall�������q�mo��p�dular�7�if�there�is�a�mo�M�dular�form��f�e��suc��!h�that�a�mo�d��p��represen��!tation�attac�hed�to��f�e��is�the�mo�M�d��p�����q�reduction��of���,�-�and�ditto�for��q�d��.�"�I��ha��!v�e��carried�out�sp�M�ecic�computations�suggested�b��!y�Mazur�in�����qhop�M�es�kof�determining�when�one�should�exp�ect�that�suc��!h�mo�d��pq���represen��!tations�are�mo�dular;�"the�����qcomputation���suggest�that�the�answ��!er�is�not�so�simple.��JOnce�w�e�kno�w�that����is�mo�M�dular,�Ǿit�is�����qthen�v�of�in��!terest�to�try�and�nd�the�minimal�w�eigh�t�and�lev�el�of�a�form�giving�rise�to���.���The�righ�t�����qconjectures��fare�elusiv��!e.��"V����q�References���W����W�[1]���(_�A.�d=Agash��"���De,����On���invisible�elements�of�the�T��)ate-Shafar��p�evich�gr�oup�,���Th��"���Deorie�d=des�nom��!bres��328����(_��(1999),��f369{374.��|{����W[2]���(_�A.���Agash��"���De�and�W.��MA.�Stein,��R�Visibility��of�Shafar��p�evich-Tate�gr�oups�of�mo�dular�ab�elian�varieties�,����(_�in��fpreparation�(1999).������W[3]���(_�E.�B�Artin,��`���;j���i�Ub��p�er�z�eine�neue�Art�von�L-r�eihen�,�i�Abh.�B�Math.�Sem.�in�Univ.�Ham��!burg�i��3��(1923),����(_�89{108.������W[4]���(_�B.��MJ.��/Birc��!h,��:�El��Fliptic���curves�over��Q�:�vA���pr��p�o�gr�ess���r�ep�ort�,�1969��/Num��!b�M�er�Theory�Institute�(Pro�c.����(_�Symp�M�os.��Pure�Math.,�2�V��eol.�XX,�State�Univ.�New�Y�ork,�2�Ston��!y�Bro�M�ok,�N.Y.,�1969),�Amer.����(_�Math.��fSo�M�c.,�Pro��!vidence,�R.I.,�1971,�pp.�396{400.������W[5]���(_�S.�SUBlo�M�c��!h�and�K.�Kato,�~��L�-functions��and�Tamagawa�numb��p�ers�of�motives�,�The�SUGrothendiec��!k����(_�F��eestsc��!hrift,��fV�ol.�I,�Birkh���fauser�Boston,�Boston,�MA,�1990,�pp.�333{400.������W[6]���(_�C.��]Breuil,��B.�Conrad,�F.�Diamond,�and�R.�T��ea��!ylor,��On���the�mo��p�dularity�of�el��Fliptic�curves�over����(_��Q�,��fin�preparation.������W[7]���(_�K.��PBuzzard,���M.�Dic��!kinson,�N.�Shepherd-Barron,�and�R.�T��ea��!ylor,��On��ic��p�osahe�dr�al�Artin�r�epr�e-����(_�sentations�,��fin�preparation.������W[8]���(_�K.�7	Buzzard�and�R.�T��ea��!ylor,�MO�Comp��p�anion��vforms�and�weight�one�forms�,�Annals�7	of�Math.�(1999).���W�����W[9]���(_�J.��ME.�1
Cremona,����A��\lgorithms�U�for�mo��p�dular�el��Fliptic�curves�,�second�1
ed.,�Cam��!bridge�Univ�ersit�y����(_�Press,��fCam��!bridge,�1997.�������q[10]���(_�J.��ME.��$Cremona�and�B.�Mazur,��d�Visualizing��elements�in�the�Shafar��p�evich-T��)ate�gr�oup�,��dPro�M�ceedings����(_�of��fthe�Arizona�Win��!ter�Sc�ho�M�ol�(1998).�������q[11]���(_�N.��MD.�ֹElkies,����El��Fliptic�_and�mo��p�dular�curves�over�nite�elds�and�r�elate�d�c�omputational�issues�,����(_�Computational�4p�M�ersp�ectiv��!es�on�n�um�b�M�er�theory�(Chicago,�J�IL,�1995),�Amer.�Math.�So�M�c.,�Pro��!v-����(_�idence,��fRI,�1998,�pp.�21{76.�������q[12]���(_�V.��MA.�_�Kolyv��dDagin,�n�On��$the�structur��p�e�of�Shafar�evich-Tate�gr�oups�,�nAlgebraic�_�geometry�(Chicago,����(_�IL,��f1989),�Springer,�Berlin,�1991,�pp.�94{121.�������q[13]���(_�V.��MA.�Kolyv��dDagin�and�D.�Y.�Logac��!hev,�,��Finiteness�M�of���X���Z�over�total��Fly�r��p�e�al�M�elds�,�Math.�USSR����(_�Izv��!estiy�a��f�39��(1992),�no.�1,�829{853.�������q[14]���(_�S.�]Lang�and�J.�T��eate,����Princip��p�al���homo�gene�ous�sp�ac�es�over�ab�elian�varieties�,���Amer.�]J.�Math.����(_��80��f�(1958),�659{684.�����E��K����0�����q�William��fA.�Stein�x$�6�����?���r������q[15]���(_�B.�v�Mazur,��n�R��p�ational��?p�oints�of�ab�elian�varieties�with�values�in�towers�of�numb�er�elds�,��nIn��!v�en�t.���f��(_�Math.��f�18��(1972),�183{266.��������q[16]����(_��ff ٟ��K�,��;�Mo��p�dular�� curves�and�the�Eisenstein�ide�al�,��;Inst.�u1Hautes��rB��;j��Etudes�Sci.�Publ.�Math.�(1977),����(_�no.��f47,�33{186�(1978).�������q[17]����(_��ff ٟ��K�,�i=�R��p�ational���iso�genies�of�prime�de�gr�e�e�(with�an�app�endix�by�D.�Goldfeld)�,�i=In��!v�en�t.�Y�Math.����(_��44��f�(1978),�no.�2,�129{162.�������q[18]����(_��ff ٟ��K�,��f�Visualizing���elements�of�or��p�der�thr�e�e�in�the�Shafar�evich-Tate�gr�oup�,��fpreprin��!t�(1999).�������q[19]���(_�L.���Merel,���Sur��la�natur��p�e�non-cyclotomique�des�p�oints�d'or�dr�e�ni�des�c�ourb�es�el��Fliptiques�,����(_�preprin��!t��f(1999).�������q[20]���(_�J.-F.��pMestre�and�J.�Oesterl��"���De,����Courb��p�es���de�Weil�semi-stables�de�discriminant�une�puissanc�e����(_��m�-i��\����eme�,��fJ.�Reine�Angew.�Math.��400��(1989),�173{184.�������q[21]���(_�K.��MA.��Rib�M�et,���On�1�mo��p�dular�r�epr�esentations�of���Gal��M�(���,R�p 	u��Ӯ��Q���	u��=�Q�)��arising�fr�om�mo�dular�forms�,��In��!v�en�t.����(_�Math.��f�100��(1990),�no.�2,�431{476.�������q[22]����(_��ff ٟ��K�,����R��p�aising��7the�levels�of�mo�dular�r�epr�esentations�,���S��"���Deminaire�q�de�Th����Deorie�des�Nom��!bres,����(_�P��!aris��f1987{88,�Birkh���fauser�Boston,�Boston,�MA,�1990,�pp.�259{271.�������q[23]���(_�A.��MJ.���Sc��!holl,��S�A��\n�intr��p�o�duction�to�Kato's�Euler�systems�,��SGalois���Represen��!tations�in�Arithmetic����(_�Algebraic��fGeometry��e,�Cam��!bridge�Univ�ersit�y�Press,�1998,�pp.�379{460.�������q[24]���(_�J-P��e.�G�Serre,�p0�Sur�nles�r��p�epr��\����esentations�mo�dulair�es�de�de�gr��\����e��2��de���Gal���p(���,R�p 	u��Ӯ��Q���	u��=�Q�),�p0Duk��!e�G�Math.�J.��54����(_��(1987),��fno.�1,�179{230.�������q[25]���(_�W.��MA.��fStein,��Comp��p�onent���gr�oups�of�optimal�quotients�of�Jac�obians�,��fpreprin��!t�(1999).�������q[26]���(_�R.�R
T��ea��!ylor�and�A.��MJ.�Wiles,�b��R��\ing-the��p�or�etic��Npr�op�erties�of�c�ertain�He�cke�algebr�as�,�b�Ann.�R
of�Math.����(_�(2)��f�141��(1995),�no.�3,�553{572.�������q[27]���(_�J.��5T��eunnell,��i�A��\rtin���s's�	c��p�onje�ctur�e�for�r�epr�esentations�of�o�ctahe�dr�al�typ�e�,��iBull.��5Amer.�Math.�So�M�c.����(_�(N.S.)��f�5��(1981),�no.�2,�173{175.�������q[28]���(_�A.��MJ.��Wiles,���Mo��p�dular�4sel��Fliptic�curves�and�Fermat's�last�the�or�em�,��Ann.��of�Math.�(2)��141��(1995),����(_�no.��f3,�443{551.�����TF���;��K��d�+��g�cmmi12�*�':
�3
cmti10�)p�0J
�3
cmsl10�'2�@�cmbx8�&�"V
�3
cmbx10�%��N�cmbx12�$X�Qcmr12�!���
�3
msbm10���u
�3
cmex10��K�cmsy8�!",�
�3
cmsy10��2cmmi8��b>
�3
cmmi10�|{Ycmr8���N�ffcmbx12��hV1ff
wncyr10��hV1
wncyr10�K�`y
�3
cmr10�_������