CoCalc -- Collaborative Calculation in the Cloud
Sharedwww / rank4 / mestre-en.dviOpen in CoCalc
����;� TeX output 2004.02.18:2012�������\̍��i�4�x\̍���#$��O��D��t�qG�cmr17�The�B�Metho���d�of�Graphs.�m�Examples�and�Applications��#�􍍍������X�Q ff cmr12�J.-F.��/Mestre.��������<T���4r.�D>Andrei��/Jorza������.&a���\WF���4ebruary��/18,�2004��5 ��� ��N�G� cmbx12�1�� (�In��u�tro��=duction��b#��X�Q cmr12�Let�E��#��g� cmmi12�S�����$�2cmmi8�k��#��(�N���;���"�)�b�S�e�the�space�of�cusp�forms�of�w��reigh�t�E��k�g�,�\�lev�el��N��ֹand�c�haracter��"�,�\�where��k���and�����N���are�xin��rtegers��&!",�
cmsy10��F�1,��zand��"��is�a�Diric�hlet�c�haracter��N�@�.��>There�are�sev�eral�w�a�ys�to�construct���a�	basis.��F��Vor�example�one�can�use�Selb�S�erg's�trace�form��rula.�Denote�b��ry��T��Vr���(�n�)�the�trace�of��T�����n���P�,���the���n�-th�Hec��rk�e��op�S�erator.�8�The�function��"�E���)�f��Q�=����������'�K�cmsy8�1�����UR���)��u
cmex10�X���
�ҍ����n�!|{Ycmr8�=1�������T��Vr��"�.(�n�)�q��n9����n���$����is���in��S�����k��#��(�N���;���"�).�sThe�set�of��f�����i���ֹ=�e��T�����i��d��f��عgenerate�this�space,���and�this�theoretically�allo��rws�us�to���construct�,�a�basis.��pF��Vor�example,�|�if��N�mg�is�prime,��"�y(�=�1�,�and��k��E�=�y(2,�then�the�set�of�the��f�����i�����(1�UR���i����g�X�where���g��is�the�gen��rus�of��X�����0����(�N�@�))�is�a�basis�of��S�����2���(�N���;����1).������But,�3hev��ren�$�in�that�case,�whic��rh�is�the�most�fa�v�orable,�3hthe�computations�b�S�ecome�hard:��Fon���an��a��rv�erage�computer�w�e�can�only�hop�S�e�to�treat��N���of�the�size�of�5000�(alw�a�ys�with��N���prime,���w��reigh�t�~�2�and�trivial�c��rharacter);�ȑactually��V,���the�computation�of��T�r���(�n�)�requires�the�kno��rwledge���of���man��ry�class�n�um�b�S�ers�of�imaginary�quadratic�elds�of�discriminan�t�of�order�at�most��n��and���to��obtain�a�basis��f�����1����;���f�����2���;��:�:�:��ʜ;�f�����g�����one��needs�to�compute��T��Vr���.(�n�)�for��n�UR���g��n9���2�2��.=�.������In�the�follo��rwing�section�w�e�describ�S�e�the�\metho�d�of�graphs",�%�whic��rh�relies�on�the�results���of���Deuring�and�Eic��rhler,���and�dev�elop�S�ed�b�y�J.�Oestrl���s�e�and�m�yself,���whic�h�allo�ws�us�to�obtain�a���basis��for��S�����2����(�N���;����1)�more�quic��rkly�(at�least�when��N�+��is�a�prime).������In�7�the�second�section,�[aw��re�indicate�ho�w�this�metho�S�d�allo�ws�us�to�pro�v�e�that�certain�elliptic���curv��res���dened�o�v�er��5��N�cmbx12�Q��are�W��Veil�curv�es�����2�1����ڹ(whic�h,��ab�y�pro�viding�an�adequate�W��Veil�curv�e,��ayields���all��the�imaginary�quadratic�elds�of�class�n��rum�b�S�er��at�most�3,���due�to�a�result�of�Goldfeld�and���recen��rt��w�orks�of�Gross�and�Zagier).������The��-third�section�is�dedicated�to�the�v��rerication�of�a�conjecture�of�Serre�in�certain���particular�57cases;��~this�is�p�S�ossible�b�ecause�of�the�metho�d�describ�ed�in�the�rst�section.�	�It���is���kno��rwn�that�this�conjecture,���if�it�is�true,�has�n��rumerous�consequences�(e.g.,�the�Shim��rura-���T��Vaniy��rama-W�eil��conjecture,�and�th�us�F��Vermat's�Last�Theorem).��2�ff��-������
����^��ٓ�Rcmr7�1�����K�y

cmr10�It�)is�no���w�a�theorem�that�ev�ery�elliptic�curv�e�is�a�W��*�eil�curv�e,�2i.e.,�the�)Shim�ura-T��*�aniy�ama-W�eil�)conjecture���is�UUtrue.�q�{�William�Stein�������1����*��\̍��i�4�x\̍����4���2��(�The�z�metho��=d�of�graphs���#���?��N�ffcmbx12�2.1��$�cDe nitions�ffand�notations��@��In�S_the�follo��rwing��p��is�a�prime�n�um�b�S�er�and��N�����1��c�is�a�p�ositiv��re�in�teger�coprime�to��p�.�rSet��N��6�=�UR�pN�����1����.�������Let�����*\�M�����N�� n��=�UR������S�� ��Z�[�S��׹]���&�where����S�L��is�tak��ren�o�v�er�all�sup�S�ersingular�p�oin��rts�of��X�����0����(�N�����1���)���in�c�haracteristic��p�,���i.e.,�o�v�er���the�set���of�$�isomorphism�classes�of�pairs�(�E��;���C�ܞ�)�consisting�of�an�elliptic�curv��re��E��̹dened�o�v�er����K�z�UV�	���2���
msbm10�F����z����p��f�and���a�
ocyclic�group��C��
�of��E����of�order��N�����1����.��6Tw��ro�suc�h�pairs�are�iden�tied�if�they�are,�ain�the�ob�vious���sense,��ꨟ�K�z�UV�	���F����?�����p��[�-isomorphic.������Let��;���������S��	s4�=������ō����j�����Aut����(�S��׹)�j�����[��z�-���
�΍��b�2�����3^w�;�����where���Aut����(�S��׹)��is�the�group�of����K�z�UV�	���F�����r����p���Ϲ-automorphisms�of��S��.�h;W��Ve�alw��ra�ys��ha�v�e�������S��
�������12,�ӹand�if��p����do�S�es��not�divide�6�then�������S��	s4��UR�3.������Therefore�8�w��re�can�dene�a�scalar�pro�S�duct�on��M�����N��Q��b�y��h�S��;���S����i�UR�=�������S��	V��and�8��h�S�;���S���ן��2�0����i��=�0�8�if��S�)�6�=�UR�S���ן��2�0���.���Let���Eis��$�=��UR�����P����� ���������1��8O��S��� p �[�S��׹]��and�let�� ���B�M��� ���@��0���ڍ�N��� n��=���UR�� �n����� US��X���� ���x�����S�� �[�S��׹]�UR:�������X�������x�����S�� s4�=�0��� �o���� -���b�S�e��the�subspace�orthogonal�to��Eis���=.������F��Vor��all�in��rtegers��n�UR���1��coprime�with��p��w�e�de ne�an�op�S�erator��T�����n�� ���on��M�+��b�y���J���0 �T�����n���P�(�E��;���C�ܞ�)�UR=�������X�����ir�C���%;�cmmi6�n�������(�E�=C�����n���P�;����(�C��F�+����C�����n���)�=C�����n���)��$���where���C�����n��	���runs�o��rv�er��all�the�cyclic�subgroups�of�order��n��suc��rh�that��C��F�\����C�����n�����=�UR0.������F��Vor��_all��q�+��j����N�����1��	~c�and�coprime�with��q��n9���2�0���!�=��N�����1����=q�n9�,��Lw��re�dene�the�same�w�a�y�the�A�tkin-Lehner���in��rv�olutions���W�����q����b�y��������W�����q�����(�E��;���C�ܞ�)�UR=�(�E�=q��n9����0��<r�C�5�;����(�E�����q��?��+����C�ܞ�)�=q��n9����0���C��)�;����where���E�����q����is�the�group�of�p�S�oin��rts�of�order�dividing��q�X�of��E���.������Finally��_w��re�dene�an�in�v�olution��W�����p��ż�b�y��W�����p��>>�=�v�������F��Vrob����-����p��p��,�Mwhere��F��Vrob���������p��$mJ�is�the�endomorphism���of�l��M�����N�� ��that�transforms�(�E��;���C�ܞ�)�to�(�E�����2�p��{t�;���C��ܞ���2�p�����).��(The�fact�that�it�is�an�in��rv�olution�l�re ects�that�the���sup�S�ersingular��p�oin��rts�are�de ned�o�v�er��F����p������"�Aa�cmr6�2�����.)������These���op�S�erators�ha��rv�e���the�follo��rwing�prop�erties:�U^the�set�of��W�����q�� ��and��T�����n�� �7�(�n��coprime�with����N�@�)���generate�an�ab�S�elian�semigroup�of�hermitian�op�erators�with�resp�ect�to�the�scalar�pro�duct����h�;����i�.���The�&��T�����n����comm��rute�with�eac�h�other�for�all��n��coprime�to��p�.���If��q�Ë�=�UR�q�����1����q�����2��桹(�q�����1���;���q�����2���coprime)�&�and���if�&J�n�UR�=��n�����1����n�����2���N�(�n�����1���;���n�����2���coprime�&Jwith�eac��rh�other�and�with��p�)�then��W�����q���P�=�UR�W�����q�� q�1���v*�W�����q�� q�2��� �t�and��T�����n�����=��T�����n�� q�1��� ���T�����n�� q�2����.������F��Vor��+all��d�i��j��N�����1�� M/�w��re��+ha�v�e�a�homomorphism�� �����d�� I��:�i��M�����N����� �3�����،!��M�����N�$�=d���Źthat�transforms�(�E��;���C�ܞ�)�to���(�E��;���dC�ܞ�).���This��homomorphism�comm��rutes�with��T�����n��
��(�n��coprime�with��N�@�),�N�and�with��W�����q��	���(for����q�Ë�j�UR�N�F:=d�).�8�F��Vor���d��j��N�����1�����and�coprime�with��N�����1����=d��w��re�ha�v�e���X�����T�����d��ߨ������d��4��=�UR������d���(�T�����d���P�+����W�����d���)�������2����
��\̍��i�4�x\̍����4���2.2��$�cAn�ffisomorphism�with����g�ff cmmi12�S��(��2����(�N���)��@��W��Ve��consider�here�the�space��S�����2����(�N�@�)�of�cusp�forms�of�w��reigh�t��2�o��rv�er�������0���(�N�@�),��uwith�its�natural����structure��of��T�-mo�S�dule,�where��T��is�the�Hec��rk�e��algebra�[��?���].���q��Theorem�R�2.1.����@���@ cmti12�@Ther��ffe�%exists�an�isomorphism�c�omp�atible�with�the�action�of�the�He�cke�op�er�a-���tors,�%&b��ffetwe�en�!��M����2��@��0��b���N��� ��� ����C��@and�the�subsp��ffac�e�!�of��S�����2����(�N�@�)��@gener��ffate�d�!�by�the�newforms�of�level��N�b��@and���oldforms�35c��ffoming�fr�om�the�cusp�forms�of�weight�2�and�level��pd�@,��d�UR�j��N�����1����@.����Remark��2.2.�0:�Assume�I�N����(or�without�loss��N�����1����)�is�square-free.�T3W��Ve�can�determine�ecien��rtly���the��isubspace��M����2��@��0��b���N��� ���corresp�S�onding�to�newforms�in��S�����2����(�N�@�);��(it�is�the�subspace�formed�b��ry�all��x��so���that��for�all�divisors��d��of��N�����1�����w��re�ha�v�e������Nv� �����d��ߨ�(�x�)�UR=�� �����d���(�W�����d���(�x�))�=�0�:����In��Fparticular�if��N���=�S��pq�n9�,����q���prime,�it��Fis�the�subspace�of��M�����pq�� \��in��rtersection�of�the�k�ernel�of�� �����q������and��of�� �����q�����W�����q���.��"����2.3��$�cRelation�ffto�the�quaternion�algebra��@��The�:�matrices�of�the�op�S�erators��T�����n�����acting�on��M�����N��S�are�the�same�as�the�classical�Brandt�matrices���[��?���],��constructed�using�quaternion�algebras.������Let����B�����p;�1���z�b�S�e�the�quaternion�algebra�o��rv�er����Q��ramied�exactly�at��p��and�innit��ry��V,��and�let��O����is��"an�Eic��rhler�order�of�lev�el��N�����1��	�&�(dened�b�y�Eic�hler�[��?���]�in�the�case�when��N�����1��	�&�is�square-free,���and���dened�in�general�b��ry�Pizer�[��?���]),���and�let��I�����1����;���I�����2���;��:�:�:��ʜ;�I�����h��	_�b�S�e���represen��rtativ�es�of�the�left�ideal���classes��of��O�UV�.������Let� ��O�����i�����b�S�e�the�righ��rt�order�(i.e.,�.Yrigh�t�normalizers)�of�the�ideals��I�����i��dڹ,�.Yand��e�����i�����b�S�e�the�n�um�b�S�er��"!�of��units�of��O�����i��dڹ.�8�The�Brandt�matrix��B���(�n�)�UR=�(�b����ߍ�(�n�)��	5h��i;j���Dȹ)��has��i;���j��{�en��rtry�����W��b����ߍ�(�n�)��	5h��i;j�����=�UR�e�������1��
���j���$����jf� �h�:�� ��2��I���������1�� ���j��� M��I�����i��d��;�����Nor��@�(� ���)����Nor��@�(�I�����j��f �)�=�����Nor��(�I�����i���)�UR=��n�gj�����where���YNor�� �[is��Ythe�norm�o��rv�er��Y�B�����p;�1��V2�(the�norm�of�an�ideal�b�S�eing�the��gcd���of�the�norms�of�its���nonzero��elemen��rts).������In��the�language�of�sup�S�ersingular�curv��res�of�c�haracteristic��p�,���w�e�ma�y�giv�e�these�matrices���(actually��their�transp�S�oses)�the�follo��rwing�in�terpretation:������Let�T>�S��b�S�e�a�sup�ersingular�p�oin��rt�as�in��I��:�1,�rSi.e.,�a�T>sup�ersingular�elliptic�curv��re��E�U�de ned�o�v�er������K �z�UV� ���F����UV����p�����together�� with�a�cyclic�group��C�d��of�order��N�����1����.�The�ring�of�endomorphisms��O�����1�� H�of��S�:�is�an���Eic��rhler���order�of�lev�el��N�����1����.�F'T��Vo�all�the�other�sup�S�ersingular�p�oin��rts��S���ן��2�0��i�=�Y(�E������2�0���P�;���C��ܞ���2�0���׹)�w�e�asso�S�ciate���the�dset��I�����S���;S��r}�����(q�% cmsy6�0���i�of�homomorphisms�from��S��;�to��S���ן��2�0����,�c�i.e.��the�set�of�all�homomorphisms�� �+�from����E�T��to����E������2�0��"�that�send��C�}5�to��C��ܞ���2�0���׹.�Z�This�is�ob��rviously�a�left�ideal�o�v�er��O�����1����,��and�its�in�v�erse�ideal�is����I�����S��r}�����0��$D�;S�����.�kW��Ve�P�can�pro��rv�e�P�that�all�the�righ��rt�ideals�of��O�����1��	��are�obtained�in�this�w�a�y��V,�j=and�the�whole���Eic��rhler��xorder�of�lev�el��N�����1���|�if�the�rign�of�endomorphisms�of�a�sup�S�ersingular�p�oin��rt��S��׹.�-%It�is�clear���that���the�general�term��B����ߍ���(�n�)��	5h��i;j����ȹof�the��n�-th�Brandt�matrix�is�the�n��rum�b�S�er���of�isogenies�of��S�����i��Թto��S�����j�����(the��fsup�S�ersingular�p�oin��rts�b�eing�con��rv�enien�tly��findexed,)��@t�w�o�suc�h�isogenies�b�S�eing�iden�tied�is���dieren��rt�e�b�y�an�automorphism�of��S�����j��f
�.���W��Ve�can�retriev�e�the�matrix�of�the�op�S�erator��T�����n��

��acting���o��rv�er���M�����n���P�.�������3�������\̍��i�4�x\̍����4����On�c�the�other�hand�if�for�all�pairs�of�sup�S�ersingular�p�oin��rts�(�S��;���S���ן��2�0����)�w�e�asso�S�ciate�the�function�������������S���;S��r}�����0���Q�(�q�n9�)�UR=�������X���
�ҍ�	E3�������q�������deg���*�������#I���where��7���ƹgo�S�es�through�all�the�homomorphisms�of��S�p�to��S���ן��2�0����,���w��re�retriev�e�the�functions���Źclassically����asso�S�ciated��with�the�ideals�of�the�orders�of�the�quaternions,��/or,�if��one�prefers,��/asso�ciated�with���the��p�S�ositiv��re�in�teger�quadratic�forms�in�4�v��X�ariables.������It��is�therefore�easy�to�pro��rv�e��that�if�������P����D�x�����S���cx�[�S��׹]�is�an�elemen��rts�of��M�����N��
���
����C��eigen�v�ector�of�all���the��Hec��rk�e�op�S�erators�and�if��f�G��(�q�n9�)�is�the�corresp�onding�mo�dular�form,�w��re�ha�v�e,�for�all��S���ן��2�0������XR�x�����S��r}�����0���ϩ�f�G��(�q�n9�)�UR=�������X�����	1
�S������x�����S���������S���;S��r}�����0����#���whic��rh�M�allo�ws,�f�in�theory��V,�to�nd�the�co�S�ecien��rts��a�����n��	�A�of��f�G��,�using�the��x�����S���.�b�In�practice,�unfortu-���nately��V,��the��$computation�of��a�����n�� �t�demands�the�kno��rwledge�of�all�the�isogenies�of�degree��n��to��S���ן��2�0����,���and��there�do�S�esn't�seem�to�b�e�a�simple�algorithm�for�that.������Nev��rertheless,���in�}�certain�cases,�there�exists�a�di eren��rt�metho�S�d�to�calculate�the�co�ecien��rts���of��p�f�G��,�!whic��rh�is�easy�as�far�as�computation�is�concerned.��7Supp�S�ose�that��N� T�is�a�prime�(th�us���equal��ito��p�),���or��N��M�is�a�pro�S�duct�of�primes��pq����and��X�����0����(�q�n9�)�is�of�gen��rus�0�(th�us��q�Ë�=�UR2�;����3�;��5�;��7��ior�13).������In�u;the�app�S�endix,���w��re�giv�e�for�eac�h�suc�h�case�an�equation�of��X�����0����(�q�n9�)�of�the�form��xy��j�=�A1�p����2�k��#��,���th��rus��the�action�of�the�Hec�k�e�op�S�erators��T�����2�����and��T�����3���o��rv�er���X�����0����(�q�n9�),��whic�h�is�giv�en�b�y�an�equation���m��ruc�h�� simpler�than�the�equation�of�mo�S�dular�p�olynomials������2����(�j��;���j���ӟ��2�0��{ �)�;�������3���(�j��;�j���ӟ��2�0��{ �)�� (whic��rh�giv�e�the���action��of��T�����2����;���T�����3�����on��X�����0���(1),�parametrized�b��ry�the�mo�S�dular�in�v��X�arian�t��j��ӹ;�cf.�8�section�2.4).������Let���u�(Q�=��x��if��N�i5�=��pq�k=�and��u��=��j��׹if��N�i5�=��p�.�o�The�F��Vourier�expansion�of��u��at�in nit��ry�is���1�=q��ȹ+��������������.�3�Let����f�G��(�q�n9�)�UR=�������P�����a�����n���P�q�����2�n�� �z�a���normalized�newform�of�lev��rel��N�չand�w�eigh�t�2�corresp�S�onding���to��Ua�v��rector�������P����x�����S�� �[�S��׹]�of��M����2��@��0��b���N����� ����K�ܞ�,�2where��K���is�the�extension�of��Q��generated�b�y�the��a�����n���P�.���Therefore���there��exists�a�prime�ideal��}��of��K��F�o��rv�er���p��so�that�� EՍ�����Yk��� ����������X����s���x�����S��Ȋ�����u�(�S��׹)��� �����L��f�G��(�q�n9�)������ō�33�dq��33�[��z� �� �΍� �q�����z������UR���X�������x�����S�������ō�y �du��Q�[��z�-� �΍u������u�(�S��׹)�����A�)(�mo�S�d���B�}�)�:������ƽù(1)����� �0�(it��is�ab�S�out�the�congruence�b�et��rw�een��Lauren�t�series�in��q�n9�).������Supp�S�ose��for�example�that��f�K�corresp�onds�to�a�W��Veil�curv��re�of�conductor��N�@�,� 2so�that��a�����n�� �3�are���in�E#�Z�.�HPThe��x�����S�� c�are�in��Z��and�one�can�pro��rv�e�E#that�������P�����x�����S�� ��u�(�S��׹)��T�6�=�0.�Th��rus�E#w�e�kno�w��a�����n������P�mo�S�d���� -��p��for���all���n�.�8�Hasse's�inequalit��ry��j�a�����l��!��j�UR�<��2�������p��� �����z�Í� :��l�����7�for���l�.7�prime�pro�v�es�that�w�e�kno�w�the��a�����n�� ���for��n�UR<�p����2�2����=�16.��"�����2.4��$�cExplicit�ffconstruction�of�the�net��M��(��
�b>

���N�z�;�d���p����;�),��or�appro��rximately��������p���
ꪟ����z�孟z��p�����W�.������So���assume�w��re�kno�w�a�sup�S�ersingular�p�oin��rt��S�����1����.�y�Kno�wing�the�action�of��T�����2��	j��on�the�mo�S�del���giv��ren���b�y��X�����0����(�N�@�)�allo�ws�us�to�obtain�the�three�sup�S�ersingular�p�oin��rts��S�����2����;���S�����3���;�S�����4��M�(not���necessarily���distinct)�
%related�to��S�����1���)�b��ry�a�2-isogen�y��V.��VIt�comes�do�wn�to�solving�a�degree�3�p�S�olynomial�o�v�er����F����p������2�����,�-�whic��rh��Bneeds�extracting�cubic�and�square�ro�S�ots,�op�erations��Bthat�need��O�S��(�log��-I�p�)�op�erations.���Sometimes��Ww��re�ma�y�as�w�ell�exlude�this�computation.�X�Supp�S�ose�that��n�g��=��p��W�and�that�w�e�ha�v�e,���sa��ry��v�p������2(���mo�S�d�����B3).�cITh�us��v�p��is�inert�in��Q�(�������p���
���ǉz�5S�	m9��3����5U),�ѩso��j�<��=���0�is�a�sup�S�ersingular�v��X�alue,�and�w��re���kno��rw�)that�the�three�isogenies�of�degree�2�send�the�curv�e�of�the�in�v��X�arian�t�to�the�curv�e�of���complex��m��rultiplication�b�y��Z�[�������p���
���ǉz�5S�	m9��3����5U],�for�whic�h�the�in�v��X�arian�t�is��j�%�=�UR54000.������In�X4an��ry�case,�uw�e�ha�v�e�at�most�one�time�when�w�e�need�to�solv�e�a�3rd�degree�equation:��once����S�����2��j*�is��&kno��rwn,��
w�e�searc�h�from��S�����i���(�i�UR���2)�the�three�sup�S�ersingular�p�oin��rts�whic�h�are�related,��
Py�=��u3�of��T�����2��	�7�o�v�er��M�����N��
Sw�has�m�ultiplicit�y�equal�to�the�n�um�b�S�er�of�connected���comp�S�onen��rts��of�the�graph�of��T�����2����.�RBut�in��M�����N��D�,��(the�space��M����2��@��0��b���N���S�corresp�onding�to�the�cusp�forms���of��}co�S�dimension�1,�2so�3�is�a�simple�eigen��rv��X�alue�in��M�����N��
���(b�ecause�for�a�cusp�form�w��re�ha�v�e����j�a�����2����j�UR�<��2������p���
��ǉz����	�9��2������),��so�the�graph�of��T�����2�����is�connected.������In�V&conclusion,�qan�algorithm�in��O�S��(�N��@�log��n+�N�@�)�op�erations�giv��res�all�the�sup�ersingular�p�oin��rts���and��
the�Brandt�matrix��B�����2����asso�S�ciated�to�them.�5One�of�the�adv��X�an��rtages�of�this�matrix�is�that���it��is�v��rery�sparse;��|on�eac�h�line�and�column�there�are�at�most�3�nonzero�terms,�j�whic�h�are���in��rtegers��6whose�sum�is�3.�+This�allo�ws,�ɀgiv�en�an�eigen�v��X�alue,�ɀto�nd�v�ery�quic�kly��V,�ɀif��N��is�large,���the��corresp�S�onding�eigen��rv�ectors.��"ʫ���2.5��$�cExamples��@�����\h�1.���� _�T��Vak��re��for�example��N�穹=����p��=�37.��nSince��37�is�inert�in��Q�(�������p��� ���ǉz�5S� m9��2����5U),�&xone�can�tak�e�as�the� rst���� _�v��rertex��Wof�our�graph�the�curv�e��E�����1��|[�of�complex�m�ultiplications�b�y��Z�[�������p��� ���ǉz�5S� m9��2����5U],�Śfor�whic�h�the���� _�mo�S�dular�)�in��rv��X�arian�t�is��j�����1���Թ=���8000����8���mo�S�d�����B37.��UW��Ve�need�to� nd�no��rw�all�the�in�v��X�arian�ts�of���� _�curv��res���2-isogenous�to�this,��i.e.,�to���solv�e�the�equation������2����(�x;����8000)��0���0(���mo�S�d�����B37).���But������ _�����p���'_���ǉz�5S� m9��2����:Sdis��#an�endomorphism�of�degree�2�of��E�����1����,�� so��j�����1��~'�is�a�ro�S�ot�(o��rv�er��#�Q�)�of�the�p�olynomial���� _������2����(�x;����8000).�3Dividing���this�p�S�olynomial�b��ry��x������8000���w�e�get�a�second�degree�p�S�olynomials���� _�with��ro�S�ots��j�����2����;���j�����3���,�:�the��in��rv��X�arian�ts�of�the�other�t�w�o�curv�es,�:��E�����2����;���E�����3��Γ�related�to��E�����1���b��ry�a�degree���� _�2�W�isogen��ry��V.��Let��!�Ë�2�UR�F����p������2��� J�so�that��!��n9���2�2�����=���2.��One�gets�that�then��j�����2��V�=�3�+�14�!�n9;���j�����3��V�=�3����14�!�n9�.����� _�Another��5metho�S�d�to� nd��j�����2����;���j�����3��_9�consists�in�remarking�that�37�is�equally�inert�in�the� eld��� ����5����C}��\̍��i�4�x\̍����4�� _��K�1�=�UR�Q�(�������p��� ���ǉz�O� m9��15���� Q),��ufor��(whic��rh�the�class�n�um�b�S�er�is�2.�& The�second�degree�p�olynomial�giving����� _�the��v��X�alues�of�the�mo�S�dular�in��rv�arian�ts��of�2�curv��res�of�complex�m�ultiplication�b�y�the�ring���� _�of��lin��rtegers�of��K�~ �is��x����2�2����+�191025�x����121287375,��whose��lro�S�ots�generate��Q�(������p��� ��ǉz���� �9��5������),�so�mo�S�dulo���� _�37��are�conjugate�in��F����37������2��� *��.�8�W��Ve�can�th��rus� nd��j�����2����;���j�����3���.����� _�F��Vor��@�N�$�prime�congruen��rt�to�1�mo�S�d�12,��the�n�um�b�S�er�of�sup�ersingular�curv��res�mo�d��N�$�is���� _�(�N�7������1)�=�12.���F��Vor�ZO�N�TA�=�]37�w��re�get�3�sup�S�ersingular�curv�es.���It�remains�to�sho�w�that�the���� _�action�e�of��T�����2��%��on��E�����2���(b��ry�conjugations�w�e�get�the�action�on��E�����3����).� �It�is�not�p�S�ossible�to�ha�v�e���� _�2�dOisogenies�of��E�����2��$S�on��E�����1����,���b�S�ecause�then�w��re�w�ould�ha�v�e�5�isogenies�of�degree�2�starting����_�in���E�����1����.�8�Therefore�there�is�one�2-isogen��ry�of��E�����2�����o�v�er��E�����2����.����_�Actually��V,�[-if�7Othere�is�a�2-isogen��ry�of�an�elliptic�curv�e�of�in�v��X�arian�t��j��"�on�itself,�[-this�in�v��X�arian�t����_�is��the�ro�S�ot�of�the�equation������2����(�x;���x�)�}�=�0,�}a��fourth�degree�equation�that�can�b�e�written����_�as�����PV(�x������1728)(�x����8000)(�x��+�3375)�����2�����_�.�п(T��Vo�Hsee�this,�)�one�can�mak��re�the�computation�of�the�equation�of������2����(�j��;���j���ӟ��2�0��{�)�ab�S�o�v�e.�пOne����_�can�0Ealso�searc��rh�whic�h�are�the�curv�es�of�complex�m�ultiplication�that�admit�a�degree�2����_�endomorphism,��i.e.,�whic��rh���are�the�imaginary�quadratic�elds�that�con�tains�an�elemen�t����_�of���norm�2.��One�nds,� ub��ry�m�ultiplication�b�y�the�units�of�the�(\corps�pres?")��the�elemen�ts��y��_�1���+��i;����������p������ǉz�5S�	m9��2����5S�;�����؜��31�1+�����=�p���\���=�\)
�|�4Í�7�����31�dԉz��T��ꍑG(2�����%7�and�����؜����1������=p���\���=�\)
�|�4Í�7�������dԉz��T��ꍑG(2�����&���that��are�the�endomorphisms�of�degree�2�of�the�curv��res�of����_�in��rv��X�arian�t���j�%�=�UR1728�;���j��=�8000��and�for�the�last�t��rw�o,���j�%�=���3375.)����_�By��order,���mo�S�d��p�,�the�graph�of��T�����2��	Z"�cannot�con��rtain�a�lo�S�op�of�a�sup�ersingular�curv��re�on����_�itself���{�although�this�curv��re�is�dened�o�v�er��F�����p��aA�(and,��more�precisely��V,�it�is�one�of�3�curv��res����_�describ�S�ed�$�ab�o��rv�e).��Therefore,�3there�are�2�isogenies�relating��E�����2��䑹to��E�����3���and�the�graph�of���� _��T�����2�����acting��on��M�����37�� 갹is�completely�determined.���� _�T��Vo�vcompute�the�corresp�S�onding�eigen��rv�ectors,�&ione�vcan�eviden��rtly�diagonalize�the�matrix���� _�(3�;����3)��of��T�����2�����but�there�is�a�simpler�metho�S�d:���� _�the��Din��rv�olution��W�����37�� Zй=�Z�������F��Vrob����-����37��'-y�separates��M�����37�� �L�in�an�ob�vious�w�a�y�in�to�t�w�o�orthogonal���� _�prop�S�er��{subspaces,���one�generated�b��ry��u�����1��V�=�UR[�E�����2����]�@���[�E�����3���],���asso�S�ciated��{with�the�eigen��rv��X�alue�1,���� _�and� �the�other�asso�S�ciated�with�the�eigen��rv��X�alue�-1,��generates�b�y��Eis��z�=���[�E�����1����]��d+�[�E�����2���]�+�[�E�����3���]���� _�and��the�v��rector�pro�S�duct�of��u�����1�� ��and��Eis���u,�e�let�it�b�e��u�����2�� s�=�Yo2[�E�����1����]�y���[�E�����2���]����[�E�����3���].�ƈOne��can���� _�deduce,��without�� recourse�to��T�����2����,�that�there�exist�2�newforms�for�whic��rh�the��q�n9�-expansion���� _�has���rational�co�S�ecien��rt,�մand�th�us�that��J�����0����(37),�մthe�jacobian�of��X�����0���(37)�is�isogenous�to���� _�the�u�pro�S�duct�of�2�elliptic�curv��res�(whic�h�is�w�ell-kno�wn,��4see�for�example�[��?���]).��F��Vorm�ula�(1)���� _�ab�S�o��rv�e��allo�ws�us�to�obtain�the� rst�83�terms�of�their�function��L�.��������\h2.���� _��p������37�;���N��6�=�UR2����37.���� _�T��Vo� �study��X�����0����(74)�one�uses�the�homomorphism�� �����2�� ���of��M�����74�� ��to��M�����37���de ned�previously��V.���� _�The�P� bres�of�reac��rh�of�the�three�sup�S�ersingular�p�oin��rts�[�E�����1����]�;����[�E�����2���]�P�and�[�E�����3����]�of��X�����0���(1)���mo�S�d�����B37���� _�are�� formed�b��ry�three�distinct�sup�S�ersingular�p�oin��rts�of��X�����0����(2)���mo�d�����B2.�7In�a�general�w��ra�y��V,���� _�write�X that�if��S�����1����;���S�����2���;��:�:�:��ʜ;�S�����k�� {��are�X the�sup�S�ersingular�p�oin��rts�of��X�����0����(�q�n9M�@�)���mo�d�����B�p��ab�o��rv�e�X a�su-���� _�p�S�ersingular�6mp�oin��rt��S��D�of��X�����0����(�M�@�)���mo�d�����B�p��(�p;���q����coprime�and�coprime�with��M�@�),�I_one�has�the��� ����6����X���\̍��i�4�x\̍����4�� _�form��rula���p�������ō���<�q��+���1����S�[��z� ԋ� �΍�Aut�����S������Yc�=������� �5�k�����UR���X��� �ҍ� ���1��������ō�%1$�1���۟[��z� ���
�΍�Aut�����S�����i������9���:��!�3��_�The�?�equation�of��X�����0����(2)�used�here�is�that�describ�S�ed�in�the�app�endix:���uv���=���2����2�12��	�,���the�����_�in��rv�olution�X}�W�����2����switc�hing��u��and��v�n9�.�'Recall�that��W�����37��UZ�=�UR������F��Vrob����-����37��&��and�that��j�%�=�(�u���+�16)����2�3����=u����_�(where����j�Hv�is�the�in��rv��X�arian�t���of�the�curv��re��E���,���image�of�the�p�S�oin�t�(�E��;���C�ܞ�)�of��X�����0����(2)�via�the����_�homomorphism�ו\oubli�{�oblivion?"�
��of��X�����0����(2)�on��X�����0���(1).)�
��F��Vrom�the�equation��j�I*�=����_��j�����1��
��=�8��one�gets�the�v��X�alues�of�the�three�sup�S�ersingular�p�oin��rts�of��E�����1����,��of�co�ordinates����_��u�����1��	�͹=���(��1�@q+��!�n9�)�=�2�;���u�����2���=���(��1����!�n9�)�=�2�=��W�����2����(�u�����1���)�Ƨand��u�����3��	�͹=�27�=��W�����2����(�u�����2���).���(Here�Ƨagain,����_�it�EVis�p�S�ossible�to�guess�the�v��X�alue�of��u�����3����,��b�ecause�it�is�clear�b��ry�the�action�of��T��ƹ(2)�on����_��X�����0����(1)���mo�S�d�����B37�kdone�previously�that�one�of�the�ab�S�o��rv�e�k�E�����1��	+�m�ust�b�S�e�in�v��X�arian�t�relativ�e�to����_��W�����2����;��kor���the�t��rw�o���solutions�of��u����2�2��V�=�UR2����2�12���չare��u�����1���;�����u�����1���.�0BReplacing�them�in�the�equation�that����_�giv��res����j�m��one�can�see�that�it�is�ab�S�out��u�����1����.���T��Vo�get��u�����2���;���u�����3��	��it�is�enough�to�solv��re�a�second����_�degree��equation.)���㍑_�One�H�can�compute�that��u�����4��
ia�=��]�W�����2����(�u�����1���)�=�2����2�12��	�=u�����1��
ia�=���5�����5�!�n9�,��and�H�one�nds�that�the����_�corresp�S�onding�in��rv��X�arian�t��j��ӹ(�u�����4����)�is��j�����2��dP�=��L3��?+�14�!�n9�.��One�solv��res�the�second�degree�equation����_�giv��ren���2�other�p�S�oin�ts�ab�S�o�v�e�b�y��j�����2��	��and�so��u�����5��	߰�=��15�a�+�17�!�n9;���u�����6���=��16����12�!�n9�.��Note���that����_��u�����7��V�=�UR�W�����2����(�u�����2���)�=��������u�����4�����Q�;���u�����8��V�=��W�����2���(�u�����5���)�=��������u�����5������and����u�����9��V�=��W�����2���(�u�����6���)�=��������u�����6������the����x�-co�S�ordinates�of�three����_�sup�S�ersingular���p�oin��rts�o�v�er��E�����3���޹(�x����\v�����
��!��������\v�x���
���b�S�eing�the�non�trivial�automorphism�of��F����p������2�����.)�EuW��Ve����_�get��the�list�of�all�sup�S�ersingular�p�oin��rts�of��X�����0����(2)���mo�d�����B37.����_�As�=�said�ab�S�o��rv�e,�Ythe�=�space��M����2��@��new��RA��74���i�corresp�onding�to�the�newforms�is�the�in��rtersection�of�the����_�k��rernel�:�of�������2�����and�the�k�ernel�of�������2����W�����2���.��QIf�:�w�e�write�[�u�����i��dڹ],�^�i�UR�=�1�;����:�:�:��ʜ;����9)�:�the�generators�of��M�����74�����_�corresp�S�onding��to�the�sup�ersingular�p�oin��rts�of��x�-co�ordinate��u�����i��dڹ,� an�examination�of�the����_�action��>of��W�����37���F�and��W�����2��{B�pro��rv�e��>that��M����2��@��new��RA��74�����is�the�direct�sum�of�t��rw�o��>2-dimensional�subspaces,����_�one�c2�G�����1����,�~Jgenerated�b��ry��e�����1��V�=�UR[�u�����1���]������[�u�����2���]����[�u�����4���]�+�[�u�����7���]����[�u�����9���]�c2and��e�����2��V�=�UR[�u�����5���]������[�u�����6���]����[�u�����8���]�+�[�u�����9���],����_�on�N$whic��rh��W�����37�� ���=�����W�����2�����=�1�N$and�the�other,�g�G�����2����,�generated�N$b�y��e�����3�����=���[�u�����1����]��d+�[�u�����2���]����2[�u�����3���]�+���� _�[�u�����4����]������[�u�����6���]�+�[�u�����7���]����[�u�����9���],��on�whic��rh��W�����2��V�=�UR��W�����37�� UZ�=�1.���� _�Using�.nthe�equation�of��T�����3���r�acting�on��X�����0����(2)�(cf.�3app�S�endix),�?one�can�pro��rv�e�.nthat�the�ma-���� _�trix�j�of��T�����3�� *��acting�on��G�����1���(resp�S�ectiv��rely��G�����2����)�in�the�basis�(�e�����1���;���e�����2���)�(resp�S�ectiv��rely�(�e�����3���;���e�����4���))�is������� _��q������d���+5D��1���Dj�1�������/��1���Dj�0�����OJ���q����X �,��Uof��@c��rharacteristic�p�S�olynomial��x����2�2���ݹ+���x����1��@(resp�ectiv��rely�����q������d���y��3���!Y�1�������y�1���!Y�0�����,9���q����5�,��Uof�c�harac-����� _�teristic��p�S�olynomial��x����2�2��j������3�x����1).���� _�One��adeduces�that��J����2�� r�new��RA��0���Ϲ(74)�is�isogenous�to�the�pro�S�duct�of�t��rw�o��aab�elian�simple�v��X�arieties,���� _��A�����1�� ���(resp.����A�����2����),�Tof�՘real�m��rultiplication�b�y�the�ring�of�in�tegers�of��Q�(������p��� ��ǉz���� �9��5������)�;��(resp�S�ectiv�ely���� _��Q�(������p��� ��ǉz� ��� �9��13������).)��8\�� _�If��+�� �=�����CH��="��1+������p���\����\)@��o��5�����="��(�z� �T��ꍑ G(2�����$>��;�����=�����CH��="�3+������p���\����\)���o��13�����="��(�z�q��ꍑ�2�����!�5�,�+Lthen�the�v��rectors��v�����1��	��=��e�����1��Q�+�YM(���+�1)�e�����2����;���v�����2���=�	��e�����1��Q���e�����2����;���v�����3���=����_��e�����3��ι+�G��e�����4����;���v�����4��V�=�UR(3�����)�e�����3���+��e�����4��zD�corresp�S�onding��@to�the�4�newforms��f�����1����;���f�����2���;�f�����3���;�f�����4��zD�of��@w��reigh�ts�2����_�and�6)lev��rel�74.���Using�(1)�one�gets�the�rst�83�v��X�alues�of�the�co�S�ecien�ts�of�these�newforms.����_�F��Vor��example�for��f�����1�����the�list�of�the�rst�v��X�alues�of��a�����l��p�is��ǖ�����Y�������l����r�2�����G3�����5���6�7���5�r11���]��13���n����&#�a�����l�����r�1�������CH������1+������p���\����\)@��o��5���������(�z��T��ꍑG(2�����������CH��κ�1��3������p���\����\)@��o��5�����κ՟�(�z�q��ꍑ�2��������_���1���+�������p�������ǉz����	�9��5����������CH��-!D��5������p���\����\)@��o��5�����-!D��(�z��T��ꍑG(2�����������CH��VU�1+3������p���\����\)@��o��5�����VU���(�z�q��ꍑ�2��������������7����l���\̍��i�4�x\̍����4��_�and��for��f�����3�����one�gets��!�^�����Y���w�,�l����*�2����0�3����l�5����	7���8ߍ11���fTK13���n���u2��a�����l�����W��1�������CH������3+������p���\����\)���o��13�������ݟ�(�z�q��ꍑ�2�������Ȍ���1����������p�������ǉz����	�9��13����������CH���1������p���\����\)���o��13�������(�z�q��ꍑ�2�����������CH��.8]��1������p���\����\)���o��13�����.8]��(�z�!X��ꍑg*2�����������CH��[���1+������p���\����\)���o��13�����[���(�z�!X��ꍑg*2���������+�_���3��(�Application�z�to�the�study�of�W��aGeil�curv��u�es��b#��Let�r��f�8v�=���w�����P���"�a�����n���P�q��n9���2�n���$�b�S�e�a�newform�of�w��reigh�t�r�2�and�lev��rel��N���with�in�teger�co�S�ecien�ts.� йThey����corresp�S�ond��to�a�strong�W��Veil�curv��re��E��ùof�conductor��N�@�.��Unfortunately�the�co�ecien��rts��a�����n�� /P�don't���giv��re� �to�S�o�m�uc�h�information�on��E� ��and�do�not�allo�w�us�to�obtain�a�simple�equation�for��E�ù.���(In���[��?���]���there�is�a�metho�S�d�due�to�Serre�that�sometimes�allo��rws�us�to�get�suc�h�an�equation,��-but���that�f>metho�S�d�is�not�systematic.)� �Here�w��re�giv�e�a�metho�S�d�that�at�least�when��N��6�=�UR�p��is�a�prime,���one��can�solv��re�the�problem.������F��Vrom�0)no��rw�on�let��N�q �b�S�e�a�prime.� dAccording�to�the�last�section,�A�for�a�newform��f�x(�there�is���asso�S�ciated�wa�v��rector��v�����f�� e�=���F�����P��<��x�����S�� �[�S��׹],�k�x�����S�� �(�2��F�Z�,�an�weigen�v�ector�of�the�Hec�k�e�op�S�erators�de ned�in���2.1.���Theorem�sW1�do�S�esn't�describ�e�the�isomorphism�(whic��rh�is�not�canonical)�b�et��rw�een�sW�S�����2����(�N�@�)���and����M����2��@��0��b���N��� �8� ����C�.� �But�supp�S�ose�kno��rwn�the�terms��a�����n�� -0�of��f��߹(�a�����2�� D�is�sucien�t�in�general).� �The���construction�C�of�section�2.4�giv��res�us�b�S�oth�the�sup�ersingular�v��X�alues����mo�d���� ��N����and�the�graph�of����T�����2�����acting�΍on��M�����N��D�.�/�W��Ve�can�determine�the�eigenspace��V�����2���asso�S�ciated�with�the�eigen��rv��X�alue��a�����2����.�/�If���it��His�of�dimension�1,�͎w��re�also�ha�v�e��v�����f��w �,�͎or�at�least�the�space�it�generates.�,�Otherwise,�w��re�apply����T�����3�� 5��to�u��V�����2���(whic��rh�is�of�ten�tativ�ely�small�dimension|for�conductors��<�A��80000,���?dim��-��V�����2�� 5��do�S�esn't���go�S�es� �b�ey��rond�6),�F�un�til� nding�a�1-dimensional�space,�F�corresp�S�onding�to�the�same�eigen�v��X�alues�of���the��cop�S�erators��T�����l��+�as��f�G��.�\Cho�ose�in�this�space�a�v��rector��r�����f���h�=��iI�����P����x�����E��-��[�E���]�with�in�teger��x�����E�� #�coprime���in��pairs�����2�2�����;�then��r�����f�� aǹis�determined�up�to�sign.������T��Vo��bgo�further,���w��re�need�a�geometric�in�terpretation�of�the��x�����E��-��.�*�Let��UR=���N��@���2���Ќ�,���the��bdiscrim-���inan��rt�:�of�the�minimal�W��Veierstrass�mo�S�del�of��E�ù,�N�let�� ��g�:��X�����0����(�N�@�)��������� 2�!�E�M^�b�e�:�a�minimal�co��rv�er�:�of��E����of��degree��n�UR�=��deg�� � �.������According��?to�Deligne-Rap�S�op�ort��?[��?���],��Tthere�exists�a�mo�S�del��X�����0����(�N�@�)�����=�62�@�cmbx8�Z�� {n�of��X�����0���(�N�@�)�de ned�o��rv�er��?�Z����for� �whic��rh�reduction�mo�S�d��N�J��is�the�union�of�t�w�o�pro�� jectiv�e�lines,�6�one��C�����1�� ��classifying�the�elliptic���curv��res�of�c�haracteristic��N����pro�vided�with�the�group�sc�heme�k�ernel�of�the�F��Vrob�S�enius�(this���corresp�S�onding���to�inseparable�isogenies),��ythe�other�one,��C�����0����,�classifying�the�curv��res�pro�vided���with��HV��Versc��rhiebung.�m�These�t�w�o�lines�in�tersect�at�sup�S�ersingular�p�oin��rts.�m�As�far�as�the�curv�e��E����is��Mconcerned,��7reduction�mo�S�d��N�/1�of�its�Neron�mo�del�has�iden��rtit�y��Mcomp�onen�t��M�E����2����0��뀍�=�3p�p�msbm8�F��X.�N������isomorphic�����to�Oc�F����N���"�����2����N�of�the�m��rultiplicativ�e�Ocgroup��G�����m��Ĺ.� One�can�pro��rv�e�Octhat�the�co��rv�er�Ocextends�to��X�����0����(�N�@�)�����=�Z�� *���m�x����where����x��is�the�set�of�all�sup�S�ersingular�p�oin��rts�of�c�haracteristic��N�@�,���and�de ne�b�y�restriction���a��wregular�\application�?"�oNof��C�����1�� ��on��E����2����0�� ;%��=���h��\)㍟�S��F����㍟X.�N�����ù,�֫of�a�rational�function�� ��o��rv�er��w�C�����1�� �,�for�whic��rh�the���\�p�S�oles�@fand�zeros�are�in��E�ù.�:The�divisor�������P���������E��-��[�E���]�of�� �,�U��E��}�going�through�all�the�sup�ersingular���curv��res�����mo�S�d���� ���N�@�,�$Band��th�us�an�elemen�t�of��M����2��@��0��b���N���D�,�$Bde ned�up�to�sign�(dep�S�ending�on�the�c�hoice�of���isomorphism��of��E����2����0��뀍�=�F��X.�N�����k�o��rv�er���G�����m��Ĺ.)������Prop�osition��3.1.��@In�35the�ab��ffove�notation�the�divisor��()�UR=�������P����������E��-��[�E���]�35�@is�e�qual�to���r�����f��w �@.�� �8�ff��-� ����� ����^��2�����That�UUseems�to�strong�to�me;�do�w���e�just�mean�that�the�gcd�of�co�G ecien�ts�is�1?��� ����8���� ����\̍��i�4�x\̍����4����It�� is�not�dicult�to�see�that�()�is�prop�S�ortional�to��r�����f��w �.�8By�con��rtradiction,���the�fact�that����the��u�l�����E�� ���are�coprime�with�one�another�is�obtained�from�the�result�of�Rib�S�et�whic��rh�sa�ys�that�if��l����is���a�prime�di eren��rt�from�2,���3�then�all�cusp�forms�mo�S�d��l��$�of�w�eigh�t�2�and�lev�el��N�@�p��(where��N�p����is��square-free)�for�whic��rh�the�asso�S�ciated�represen�tation�mo�S�d��l�+5�is�irreducible�and�not�ramied���at��>�p�,���comes�from�a�cusp�form�mo�S�d��l�͹of�w��reigh�t��>2�and�lev��rel��N��"�(this�result�w�as�conjectured���b��ry���Serre�in�1985.��gThis�also�sho�ws�that�the�T��Vaniy�ama-W�eil���conjecture�implies�the�F��Vermat���theorem.)������T��Vo�tpro��rv�e�the�previous�theorem,�D�one�pro�v�es�rst�that������is�related�to�������E��
H��b�y���Ȅ�=��URgcd���F(������E��-��!�����E��0�����������F��O�!�����F���)��Ewhere��!�����E��!͹is�the�n��rum�b�S�er��Eof�automorphisms�of��E���.�U�Supp�ose�that�a�prime�n��rum�b�er��E�l����divides��Qthe�gcd�is�������E��-��.�L�It�also�divides���s2�,���and�one�deduces�from�here�that��p��is�not�ramied�in���the�i�eld�of�p�S�oin��rts�of�order��l��X�or��E�ù.��BIf��l��is�coprime�with�6�Rib�S�et's������2�3���
���theorem�sho��rws�that�the���mo�S�dular��form��f��asso�ciated�to��E��ڹis�congruen��rt�mo�d��l���to�a�mo�dular�form�of�w��reigh�t��2�and�lev��rel���1,�͜whic��rh��Ycannot�b�S�e�but�the�Eisenstein�series.�,�The�curv�e��E���is�semi-stable,�͜whic�h�implies�([��?���],���p.306)���that��E�Ƌ�or�a�curv��re��Q�-isogenous�to�it�has�a�p�S�oin�t�of�nite�order��l�C��.�&�If��l���=�UR2�;����3�w�e�get�the���same��_result�due�to�[��?���],���App�S�endix.�No��rw,�w�e��_kno�w�explicitly�the�curv�es�of�prime�conductor���with��Ftorsion�[��?���]�namely�the�curv��res�11A��2and�11B�of�[��?���],��Ywhic��rh�ha�v�e�a�p�S�oin�t�of�order�5,��Ycurv�es���17A,17B,17C�ܒ(p�S�oin��rt���of�order�4),�[17D�(p�S�oin��rt�of�order�2),�[19A�and�19B�(p�S�oin��rt�of�order�3),���37B,���37C��f(p�S�oin��rt�of�order�3)�and�the�curv�es�of�Setzer-Neumann�[��?���],��whic�h�ha�v�e�a�p�S�oin�t�of���order�ʤ2.���In�eac��rh�of�these�cases,��w�e�kno�w���s2�,��whic�h�is�equal�to�the�n�um�b�S�er�of�nite�p�oin��rts���rational�_�o��rv�er��Q��of�the�considered�curv�es,�}@and�one�can�v�erify�that�the�������E���w�are�coprime�with���one��Hanother.�kThis�pro��rv�es��Hthe�prop�S�osition.�Note�that�along�the�pro�S�of�w��re�sho�w�ed�that�Rib�S�et's���theorem��implies�the�follo��rwing�����Theorem�A�3.2.�V �@L��ffet��;�E��R�@b�e�a�str�ong�Weil�curve�of�prime�c�onductor��N�@��@.�7|The�valuation�of�its���discriminant�35in��N�t�@is�e��ffqual�to�the�numb�er�of�torsion�p�oints�of��E���(�Q�)�@.������W��Ve��vstate�without�pro�S�of�the�theorem�that�allo��rws�us�to�get�an�explicit�equation�for��E��9�once���w��re��kno�w�the�������E��-��.����Theorem�J�3.3.�1��@L��ffet�;��E�N��@b�e�a�str�ong�Weil�curve�of�prime�c�onductor��N�@��@,�mGand�������P���w������E��-��[�E���]��@the�element���of�35�M����2��@��0��b���N���Ly�@asso��ffciate�d�to��E�E��@via�the�c�onstructions�ab�ove.�fiTher�e�exists�an�e�quation�of��E��Q������y��n9����2�����=�UR�x�����3��j��������ō�֎�c�����4����۟[��z����
�΍�48�������x����������ō�ƌ�c�����6����۟[��z���
�΍�864������ s���@with�35�c�����4����;���c�����6��V�2�UR�Z��@so�that,�if��H�B��=��max��3|(�����l�p������l�z�y?�
B���j�c�����4����j����yA�;���UP��v��3�����US���l�p���UU���l�z�y?�
B���j�c�����6����j������Δ�)��@we�have:�������
�Q1.����_��H�B�������Fu�����8�n��������z����s�����8��p���\��8��\)�D��A��N��������2�����FԹ(�log��-K(�H���V���2�6���Z�=�1728)���+��b�)�@,�35wher��ffe��b�UR�=�((1�=�3)�=�(2�=�3))����2�3��V�=�7�:�74316962�����:�:�:��ʞ�@.��������
�Q2.����_�L��ffet�������2�0����=�)�(�c����2��3��RA�4����������c����2��2��RA�6�����)�=�1728�@.���Then������2�0���=�)���@if��E����@is�sup��ffersingular�in�char�acteristic�2,�and����_�����2�0��#��=�UR�35�@or��2����2�12��	���@otherwise.�������
�Q3.����_��c�����4��V��UR�(������P����������E��-��j�����E���)����2�4�������@mo��ffd������N�@��@.�������
�Q4.����_��c�����6��V��UR�(������P����������E��-��j�����E���)����2�6�������@mo��ffd������N�@��@.��38�ff��-������
����^��3�����K.Rib�Get,�UU�A�':

cmti10�AL��}'e�ctur�es���on�Serr��}'e's�c�onje�ctur�es�,�UUMSRI,�F��*�all�1986�������9����
���\̍��i�4�x\̍����4�����
�Q�@5.����_��n�Ȅ�=�UR�����2��2��b���E���-��!�����E���@.���B����If���the�������E��
�1�are�kno��rwn�then�5�allo�ws�us�to�get��n��and�1�allo�ws�us�to�nd�a�b�S�ound�on��H��V�,��Bth�us����on���c�����4����;���c�����6���.��By�2�w��re�ha�v�e��c����2��3��RA�4�������ƚ�c����2��2��RA�6���[4�=��01728����2�0���9�;��whic�h�allo�ws�us�to�nd��c�����4����;���c�����6���.��The�congruences�3���and�
n4�allo��rw�us�to�reduce�the�n�um�b�S�er�of�computations�signican�tly��V.��1Th�us�w�e�ha�v�e�found�an���equation��of�a�strong�W��Veil�curv��re�corresp�S�onding�to�the�initial�newform��f�G��.������This���metho�S�d�also�allo��rws�us�to�pro�v�e�that�an�elliptic�curv�e�of�small�prime�conductor�is�a���W��Veil�ƣcurv��re.�,�Supp�S�ose�that�w�e�are�giv�en�suc�h�a�curv�e�b�y�its�equation.�,�Then�w�e�ma�y�compute���the���n��rum�b�S�er�of�its�p�oin��rts��N�����l���~�mo�d��l��E�for��l���=�UR2�;����3�;��:�:�:��ʞ�.��Next�w��re�searc�h,��Mb�y�the�metho�S�d�of�graphs,���whether�1z�a�����2����=���3������N�����2���~�is�the�eigen��rv��X�alue�of��T�����2���acting�on��M�����N��D�.�
UIf�not�then�the�T��Vaniy��rama-W�eil���conjecture�J�is�false.��If�y��res,�j�then�con�tin�ue�with��T�����3��
�acting�on�the�found�eigenspace,�j�if�it�is�not�of���dimension�@�1,�b�un��rtil�w�e�get�an�eigenspace�of�dimension�1�for�the�Hec�k�e�op�S�erators,�b�with�in�tegers���eigen��rv��X�alues.�If�1there�is�no�suc�h�thing,�B�then�w�e�get�a�coun�terexample�to�the�T��Vaniy�ama-W�eil���conjecture.�YIf�.�there�is�one,�?�w��re�compute�the�equation�of�a�corresp�S�onding�W��Veil�curv�e.�YIf�this���curv��re�iis�isogenous�to�the�initial�curv�e,���w�e�are�done.�
�Otherwise,���the�initial�curv�e�is�not�a�W��Veil���curv��re.������In��particular,�this�allo��rws�us�to�pro�v�e�that�the�elliptic�equation��ڢ����1�y��n9����2����+����y�Ë�=�UR�x�����3��j����7�x��+�6���of��conductor�5077,�is�a�W��Veil�curv��re.�������This��icurv��re�seems�to�b�S�e�the�smallest�curv�e�(ordering�the�curv�es�b�y�their�conductors)���ha��rving��a�Mordell-W��Veil�rank���UR�3�[��?���].�8�The�in�terest�in�it�is�the�follo�wing:������Let���f�G��(�z���)��(=�������P��/��a�����n���P�q��n9���2�n��
K�(�q��a�=��e����2�2��I{iz����),�
�a�newform�of�w��reigh�t�2�and�conductor��N�@�,�
�and�let��L�(�s�)��(=���������P�����a�����n���P�n����2��s��
��,��the��<asso�S�ciated��L��function.�7gIf�the�order�of��L��in�1�is���UR�3�then�Goldfeld�pro��rv�ed��<that���there��exists�a�computable�constan��rt��C�����f��	aǹso�that��ڢ����M�log���z��p�UR<�C�����f��w�h�(��p�)�;����where��S�p�UR���3(���mo�S�d�����B4)�is�a�prime�n��rum�b�S�er�coprime�with��N�%7�and��h�(��p�)�is�the�n�um�b�S�er�of�classes����of��#imaginary�quadratic�elds�of�discriminan��rt���p�.�RW��Ve�ha�v�e�other�form�ulas,���but�more�com-���plicated,�ڸin�w�the�case�of�imaginary�quadratic�elds�of�non-prime�discriminan��rt�(see�[��?���]�for���example).������If��,the�Birc��rh�and�Swinnerton-Dy�er�conjecture�is�true,���all�the�W��Veil�curv�es�for�whic�h�the���Mordell-W��Veil��group�o��rv�er���Q��is�of�rank���^��3�ha��rv�e��to�b�S�e�giv��ren�b�y�suc�h�mo�S�dular�forms,�i�but���un��rtil�a+the�w�ork�of�Gross�and�Zagier�[��?���],�~�there�w�as�no�w�a�y�to�v�erify�that�the�deriv��X�ativ�e�at�1���of��ithe��L��function�of�a�W��Veil�curv��re�is�indeed�0.�'vThe�results�of�Gross�and�Zagier�allo�w�to�write����L����2�0���9�(1)���as�the�pro�S�duct�of�a�non-zero�factor�easily�computable�and�the�N��r��s�eron-T��Vate�heigh�t�of�a���Heegner��Op�S�oin��rt�(cf.�v�[��?���]�for�more�details.)�It�is�therefore�p�S�ossible,�xb��ry�decreasing�the�heigh�t�of���rational��p�S�oin��rts�on�the�curv�e�and�increasing��L����2�0���9�(1)�b�y�a�careful�computation,��ito�pro�v�e�that��L����is��wof�order���s��3�at��s��=�1.�nN(In�all�the�previous,��w��re�considered�o�S�dd�W��Veil�curv�es,��i.e.,�for��wwhic�h���the���L��function�has�an�o�S�dd�order�at�1�{�or�if�one�prefers�for�whic��rh�the�sign�of�the�functional���equation��is�-1.)������One��has�sev��reral�metho�S�d�to�construct�W��Veil�curv�es�for�whic�h�the�Mordell-W��Veil�group�is���of��rank�����3�(and�whic��rh�are�go�S�o�d��candidates�for�the�preceding�question:�%fb�y�the�metho�S�d�of�������10��������\̍��i�4�x\̍����4��Gross-Zagier,���one�o�ma��ry�compute��L����2�0���9�(1).���If�it�is�zero,�one�has�an��L��function�whic��rh�allo�ws�to����obtain��man�increase�of�the�absolute�v��X�alue�of�the�discriminan��rt�of�imaginary�quadratic�elds�of���giv��ren��class�n�um�b�S�ers;��Kif�it�is�non-zero,��8the�conjecture�of�Birc�h�and�Swinnerton-Dy�er�is�false.)���One�0can,�AWfor�example,�searc��rh�for�curv�es�of�complex�m�ultiplication�of�rank�3�(w�e�kno�w�that���they���are�W��Veil�curv��res),��nbut�the�constan�t��C�����f��	E��is�v�ery�large.�/�One�can�deform�����2�4�����a�W��Veil�curv�e�(for���example�zthe�curv��re�37C�y�of�[��?���]�un�til�getting�a�rank�3�curv�e�(for�the�curv�e�37C,�one�can�deform���b��ry�G�Q�(�������p���
���ǉz��K�	m9��139����$�M),�^$as�sho�wn�b�y�Gross�and�Zagier�[��?���].)�N
This�leads�to�a�constan�t��C�����f��	�*�of�order�of���7000������One�I1ma��ry�c�ho�S�ose�some�elliptic�curv�e�dened�o�v�er��Q�,��or�rank�3,�and�try�to�pro��rv�e�I1that�it���is�+�a�W��Veil�curv��re.��{This�w�as�done�in�[��?���]�for�the�men�tioned�curv�e�of�conductor�5077,�;�using�the���trace�9�form��rula.�&�But�the�computation�is�v�ery�long.�&�The�metho�S�d�of�graphs�allo�ws�us�to�do�it���in��ab�S�out�5�seconds�an�a�computer�that�needed�5�hours�with�the�men��rtioned�metho�d.������F��Vor�Tthis�curv��re,�n�one�has��C�����f��	�d�<�	E�50:�Qall�imaginary�quadratic�curv�es�of�discriminan�t��d��with����j�d�j���>�e����2�150���F�therefore�B:has�a�class�n��rum�b�S�er�B:����4.�	?�On�the�other�hand,��there�is�no�imaginary���quadratic���eld�of�discriminan��rt��d��and�class�n�um�b�S�er�3�for�907�z�<��j�d�j��<��10����2�2500����[��?���].�<�Therefore���(after��an�examination�of�a�table�of�class�n��rum�b�S�ers��of�the�rst�quadratic�elds):�����Theorem�a&3.4.�@��@The�M�imaginary�quadr��ffatic�elds�of�class�numb�er�3�ar�e�the�16�elds�of�discrimi-���nant:�����23�;�����31�;���59�:��83�;���107�;���139�;���211�;���283�;���307�;���331�;���379�;���499�;���547�;���643�;���883�;���907�@.��(V���4��(�Application�z�to�a�conjecture�of�Serre��b#��Let�z���b�S�e�a�con��rtin�uous�zrepresen�tation�of��Gal�C��(���S��z�
#��	�\��Q���
#��=�Q�)�in��GL�����2����(�V��p�)�where��V���is�a�dimension�2���v��rector�space�o�v�er�a�nite�eld��F�����q���^�of�c�haracteristic��p�.��Assume�this�is�an�o�S�dd�represen�tation,���i.e.,��8that�zN��(�c�)�the�image�of�the�complex�conjugation,�seen�as�an�elemen��rt�of��Gal�C��(���S��z�
#��	�\��Q���
#��=�Q�)�has���eigen��rv��X�alues��1�and�-1.�8�In�that�case�put��G�UR�=��I��m�.������In��[��?���]�Serre�denes�the�lev��rel,�the�c�haracter�and�the�w�eigh�t�of�suc�h�a�represen�tation:��������\h1.����_�The��lev��rel.�����_�Let��m�l���b�S�e�a�prime�n��rum�b�er��mdieren�t�from��p�.�+wW��Vrite��G�����i��'G�(�i�UR�=�0�;����:�:�:��ʞ�)�the�groups�of�ramica-����_�tions��of����at��l�C��.�8�Let���R������n�(�l�C��)�UR=����������1��������X���
㇍�S�i�=0��������ō��p�g�����i����۟[��z�
[�
�΍�g�����0��������#qi�co�S�dim���Bs�V���p����G��8:�i����z�;������_�where���g�����i���,�=�UR�j�G�����i��d��j�.����_�The��conductor�of�the�represen��rtation����is�dened�as����س9�N��6�=����UR���Y���'؍�=0�l�K�6�=�p������l��C�����n�(�l�K�)��*�:��+>������\h�2.����_�The��c��rharacter.��33�ff��-������
����^��4�����Twist?��������11�����+��\̍��i�4�x\̍����4��_�The�E�determinan��rt�of����yields�a�c�haracter�of��Gal�C��(���S��z�
#��	�\��Q���
#��=�Q�)�in��F����2����RA��q�����,�\hfor�whic�h�the�conductor�����_�divides���pN�@�.�8�Therefore,�one�can�write��M]�����idet�������UR�=��"�����k�6���1���;����_�where�D����is�the�cyclotomic�c��rharacter�of�conductor��p��and��"��is�the�c�haracter�(�Z�=��X�N�@��Z�)����2������V�����j�!����_��F����2����RA��q�����:�M��The�in��rteger��k��¹is�dened�mo�S�d�(�p�����1),�feand�M�the�fact�that�the�represen�tation�is�o�S�dd����_�implies��that��"�(��1)�UR=�(��1)����2�k��#��.��T2��_�By��denition,��"��is�the�c��rharacter�of�the�represen�tation���.��(d�����\h3.����_�The��w��reigh�t.����_�The�Q�in��rteger��k��͹ab�S�o�v�e�is�dened�mo�S�d�(�p������1).�m�Read�Q�Serre's�article�for�the�denition�of����_�the���w��reigh�t��k��>�2��!�Z��of�the�represen�tation���.�Z�As�the�conductor��N�ᐹdep�S�ends�only�on�the����_�b�S�eha��rvior�(of����ar�places�coprime�with��p�,�c�the�denition�of�w�eigh�t�only�uses�the�lo�S�cal����_�prop�S�erties��at��p��of�the�represen��rtation���.��!�����Then��Serre's�conjecture�is:��y+��Conjecture���4.1.�2�@L��ffet�g����@b�e�a�r�epr�esentation�as�ab�ove,�t�of�weight��k�g�@,�level��N����@and�char��ffacter��"�@.���Assume�Jthis�r��ffepr�esentation�Jis�irr��ffe�ducible.��Then�Jit�c��ffomes�fr�om�a�cusp�form����mo�d����J#�p��@of�weight����k�g�@,�35level��N�t�@and�char��ffacter��"�@.������This���conjectures,���if�true,�has�n��rumerous�consequences:�iit�implies�the�T��Vaniy�ama-W�eil���con-���jecture��and�F��Vermat's�theorem.������Man��ry��6suc�h�represen�tations����are�mo�S�dular,�
Meither�b�y�construction,�
Mor�b�S�ecause�they�are�part���of��8classical�conjectures�(Langlands,��Artin,���:���:�:��ȼ�)��8that�carry�on�the�conjecture�(but�sometimes���in��a�w��reak�form,�i.e.,�with�a�w�eigh�t�or�conductor�bigger�than�those�dened�in�[��?���].)������In��@order�to�v��rerify�(or�con�tradict)�Serre's�conjecture,���w�e�need�to�nd�the�extensions��K�5�=�Q����of���Galois�group�subgroup�of��GL�����2����(�F�����q�����)�of�o�S�dd�determinan��rt�and��p�ی�6�=�2.��It���is�in�general�not���dicult��jto�calculate,�Ūfor��l����prime�and�not�to�S�o�large,�the�trace��a�����l���2�of��F��Vrob���e�����l��"C˹in��GL�����2����(�F�����q�����):�!�if��P��ƹ(�x�)���is��a�p�S�olynomial�whose�ro�ots�generate��K��F�the�decomp�osition�of��P�����ƹmo�d����'�l�.7�usually�will�suce.������It�:�is�ho��rw�ev�er,�N�m�uc�h�:�harder�to�nd�mo�S�dular�forms����mo�d������p�,�N�if�they�exist,�that�corresp�S�ond���to��sthe�represen��rtation����giv�en�b�y�the�eld��K�ܞ�:�8�the�discriminan�t�of��K���is�usually�large,��~th�us�so���is��the�conductor�of���,�whic��rh�is�related�to�it,�so�it�is�not�easy�to�mak�e�the�computations.��"�g���4.1��$�cThe�ffcase��S�֜L��(��2����(� ���ff msbm10�F��(��4���)��@��A���troubling���case�is�that�of��p�UR�=�2,��Kb�S�ecause,�since�����1�UR���1(���mo�d�����B2)���all�represen��rtations�are�o�dd.������The��lrepresen��rtations�of��Gal�C��(���S��z� #�� �\��Q��� #��=�Q�)�in��GL�����2����(�F�����2���)��=��S�����3�� �p�(although��laltogether�real,�6�cf.�V-[��?���])���come���from�w��reigh�t���1�mo�S�dular�forms;��mthe�group��S�����3���ӹcan�b�e�realized�as�a�subgroup�of��GL�����2����(�C�).���One�2bcan�hop�S�e�that�b��ry�m�ultiplication�with�con�v�enien�t�Eisenstein�series,��Qone�can�obtain�a���mo�S�dular��form�of�w��reigh�t��and�lev��rel�predicted�b�y�the�Serre�conjecture�(cf.�8�[��?���]�for�examples.)������In���order�to�obtain�the�most�in��rteresting�case�for�c�haracteristic�2,���one�considers�the�rep-���resen��rtations���with�v��X�alues�in��GL�����2����(�F�����4���).�u�The���isomorphism��A�����5�� ZU�'��Q�S���L�����2����(�F�����4���)���allo�ws�us�to�obtain���sev��reral�F1examples.�KzLet��K�"Ϲb�S�e�an�extension�of��Q��of�Galois�group��A�����5����.�Since��A�����5�� 5�\immerses�?"��� ����12���� ����\̍��i�4�x\̍����4��in��rto�d��P���GL�����2����(�C�),�cif�the� eld�is�not�completely�real,�the�asso�S�ciated�represen��rtation����comes�from����a�/�w��reigh�t�1�mo�S�dular�form�(mo�dule�Artin's�conjecture,�@�cf.��[��?���]).�Supp�S�ose�no��rw�that��K� 0�is�real.���None�H[of�the�classical�conjectures�allo��rw�us�to�susp�S�ect�that����comes�from�a�mo�dular�form,�h�ev��ren���if�F�of�higher�w��reigh�t�F�or�lev��rel.�M|It�is�this�case�that�w�e�will�study�in�what�follo�ws.�M|The�metho�S�d���of�8�graphs�here�is�indisp�S�ensable,�LYthe�mo�dular�forms�that�w��re�lo�ok�at�ha��rving�a�conductor�to�o���large��to�b�S�e�studied�with�the�Eic��rhler-Selb�erg�trace�form��rula.������Let�LQ�P��ƹ(�x�)�UR=��x����2�5��'B�+�g>�a�����1����x����2�4���+��a�����2����x����2�3���+��a�����3����x����2�2���+��a�����4����x��+��a�����5�� U�b�S�e�LQa�rational�p�olynomial�of�discriminan��rt��D��.���In�h+order�that�the� eld�of�ro�S�ots�of��P� �b�e��A�����5��(/�it�is�sucien��rt�and�necessary�that��P� �b�e�irreducible,���that�u�D�ȣ�b�S�e�square-free,���and�that�there�exist�a�prime�n��rum�b�er�u�l����not�dividing��D�ȣ�so�that��P�����ƹmo�d���� '�l����ha��rving��exactly�t�w�o�ro�S�ots�in��F�����l�� p�(this�last�condition�assuring�that�the�group�is�all�of��A�����5����).������It��is�clear�that��"�l'�=�1.�a If���p��j��D�S��,��l�p��coprime�with�30,��n�(�p�)�l'=�1�if�it�\seulmen��rt�si�l'inertie�en���p�J�==?"�Y�is�of�order�2,�cand�th��rus�the�p�S�olynomial��P��¹has�at�most�double�ro�ots�mo�d��p�.�Y�As�far���as��the�w��reigh�t���k�|��is�concerned,� Zit�is�either�2�or�4�according�to�the�rami cation�of��K��:�at�2.���T��Vo���simplify��the�computation,���w��re�ha�v�e�limited�to�searc�hing�examples�among�the�represen�tations���of��prime�lev��rel�and�w�eigh�t�2.������On�l�the�other�hand,���since�it�is�ab�S�out�represen��rtations�in��S���L�����2����(�F�����4���)m�l�the�co�ecien��rt��a�����2��,��of�the���sough��rt���mo�S�dular�form,���if�it�exists,�cannot�b�S�e�in��F�����4����,�but�in��F�����16�� �.� �This�comes�from�the�fact�that���the�cco�S�ecien��rt��a�����l��*+�of�a�mo�dular�form����mo�d���� ���l�K�is�equal�to�an�eigen��rv��X�alue�of��F��Vrob���������l�� �Z�,��and�not�to�its���trace.�8�No��rw,��if�a�matrix�in��S���L�����2����(�F�����4���)��is�of�order�5,�its�eigen�v��X�alues�are�in��F�����16�� 갹not�in��F�����4����.������The�Y�examples�treated�ab�S�o��rv�e�Y�w�ere�obtained�b�y�making�a�systematic�searc�h�on�a�computer���of�Q�con��rv�enien�t�p�S�olynomials�(totally�real,�pdof�t�yp�S�e��A�����5����,�pdfor�whic�h�the�conductor�of�the�asso�S�ciated���represen��rtation��is�a�prime��N�@�,�and�for�whic�h�the�w�eigh�t�is�2).������Thereafter,�Bfor��ieac��rh�suc�h�p�S�olynomial��P��ƹ,�Bone�computes�the�corresp�onding�eigen��rv��X�alue��a�����2���m�(in����F�����16�� �),���and���one�tries�to� nd�whether�there�exists�a�mo�S�dular�form�mo�d�2�of�lev��rel��N�ۆ�and�w�eigh�t���2�[cso�that��T�����2�� g�has��a�����2���as�an�eigen��rv��X�alue.��In�all�the�cases�considered,�w�w�e�ha�v�e�thereafter�found���an���eigenspace�of�dimension�1�or�2.�"�Using�the�op�S�erators��T�����3����;���T�����5���,���one���calculates�the�co�ecien��rts����a�����3����;���a�����5���,�DJand�2]v��reri es�that�they�corresp�S�ond�to�the�v��X�alues�predicted�b�y�the�decomp�S�osition�of��P����in��3�and�5.������Clearly��V,��Athis�}�do�S�esn't�really�pro��rv�e�}�that�the�represen��rtation����asso�ciated�to��P� O�is�mo�dular:���w��re�O%ha�v�e�only�exhibited�a�mo�S�dular�form�mo�d�2�of�prop�er�lev��rel�and�w�eigh�t�for�whic�h�the���terms�5��a�����2����;���a�����3���;�a�����5�� ���are�5�con��rv�enien�t.� �But�there�is�a�go�S�o�d�indication�of�the�truthfulness�of�the���conjecture�|wof�Serre�in�the�considered�cases:��an�exhaustiv��re�searc�h�o�v�er�n�umerous�primes��N��[�of���the��{co�S�ecien��rts��a�����2����of�mo�dular�forms�of�w��reigh�t��{2�and�lev��rel��N�_�pro�v�es�that�it�is�rare�that�there���are�_� elds�of�small�degree.� �(Actually��V,�{�is�seems�that�2,�and�in�general�the�small�primes,�are�the���most�Ò\inert"�p�S�ossible�in�the� elds�that�app�ear�in�the�Hec��rk�e�Òalgebra�of�mo�dular�forms,��d elds���whic��rh��>themselv�es�in�general�app�S�ear�to�ha�v�e�the�largest�degree�p�S�ossible,�ctaking�in�to�accoun�t���constrain��rts��suc�h�as�the�A�tkin-Lehner�in�v�olutions,���primes�of�Eisenstein,�etc.�One�gets�that���one���has�small�factors,���{�corresp�S�onding�for�example�to�elliptic�curv��res�with�prime�conductor�{���but��this�is�apparen��rtly�rare.)��"ʫ���4.2��$�cA�fffew�examples��@�����\h�1.����_��P��ƹ(�x�)�UR=��x����2�5��j������10�x����2�3���+�2�x����2�2���+�19�x����6.�������13�����M��\̍��i�4�x\̍����4��_�The��7discriminan��rt�is�(2����2�3����887)����2�2���.�M�This��7p�S�olynomial�is�irreducible�mo�d�5,�țth��rus�irreducible�����_�o��rv�er���Q�.�8�Its�ro�S�ots�are�all�real�(apply�Sturm's�algorithm).�One�has�that��ው��4>�P��ƹ(�x�)�UR���x�(�x������1)(�x�����3��j��+��x�����2�����1)���mo�S�d�����B3�;����_�whic��rh�Y�giv�es�a�cycle�of�order�3;��the�Galois�group�of��K�ܞ�,�uGthe�eld�of�ro�S�ots�of��P��ƹ,�is�th��rus����_��A�����5����.��Dˍ�_�F��Vrom����P��ƹ(�x�)�UR���(�x�K<���462)(�x����755)����2�2����(�x����788)����2�2������mo�S�d����EF887���one�gets�that�the�conductor��N��Թof����_�the�Ujasso�S�ciated�represen��rtation�is��N��6�=�UR887.�!One�can�also�pro�v�e�that�2�is�\little�ramied"����_�in�P�the�sense�of�[��?���],�i�th��rus����has�w�eigh�t�2.�jpExamining�the�reduction�mo�S�d�2�of��P��I�pro�v�es����_�that�@]the�co�S�ecien��rts��a�����2����;���a�����3���;�a�����5��
a�of�@]the�mo�S�dular�form�mo�d�2�of�lev��rel�887�(whic�h�m�ust����_�corresp�S�ond��Fto����via�the�Serre�conjecture)�are�1,���1,�j��F(where��j�"��2�v�F�����4��	TJ�has�the�prop�ert��ry����_�that���j���ӟ��2�2���+����j�W{�+�1�UR=�0).����_�One�%.therefore�applies�the�metho�S�d�of�graphs:���the�space�of�mo�dular�forms�mo�d�2�of����_�w��reigh�t�w2�and�lev��rel�887�has�dimension�73,�#�and�computation�sho�ws�that�the�eigenspace����_��G�����1��	���of��~�T�����2���corresp�S�onding�to�the�eigen��rv��X�alue�1�has�dimension�2;�Fi�T�����3���acts�as�the�iden��rtit�y����_�on�>v�G�����1����,�Siand��j��;���j���ӟ��2�2��	�M�are�the�eigen��rv��X�alues�of��T�����5���z�acting�on��G�����1���,�Sifrom�where�get�a�basis�of��G�����1�����_�formed���b��ry��f�����1��	,�=�_(�q��6�+���q��n9���2�2��C:�+��q��n9���2�3���+��q��n9���2�4���+��j���q��n9���2�5���+����������Ϲand����f�����2��	,�=�_(�q��6�+��q��n9���2�2���+��q��n9���2�3���+��q��n9���2�4���+��j���ӟ��2�2��l��q��n9���2�5���+�����������,����_�eigen��rv�ectors��of�Hec��rk�e��op�S�erators.�8�These�corrob�orate�the�conjecture.��	������\h2.����_��P��ƹ(�x�)�UR=��x����2�5��j������23�x����2�3���+�55�x����2�2�����33�x����1.����_�Then��o�D���=�UR13613����2�2����;���P��ƹ(�x�)����(�x�3����6308)(�x����2211)����2�2���(�x����8248)����2�2������mo�S�d����EF13613,���N��6�=�UR13613;����P����_�b�S�eing��irreducible�mo�d�2,���kF��Vrob���F�����2��#���is�a�cycle�of�order�5,��kand��a�����2��V�=�UR������5��J �is�a�fth�ro�ot�of�unit��ry��V,����_�view��red���as�an�elemen�t�of��F�����16��	�.���Computation�also�sho�ws�that�in�the�space�of�mo�S�dular����_�forms���mo�S�d�2�of�lev��rel�13613�and�w�eigh�t�2,�'whic�h�has�dimension�1134,�'������5��	�ιis�a�simple����_�eigen��rv��X�alue�Bof��T�����2����.�?	The�co�S�ecien�ts��a�����3����;���a�����5��	�are�resp�S�ectiv�ely�equal�to�1��'+������2����2��RA�5����,�+������2����3��RA�5���	��=���j��޹and����_������2����2��RA�5���R��+��������2����3��RA�5����W�=�UR�j���ӟ��2�2��l׹,��whic��rh�are�the�traces�of��F��Vrob����ן���3�� S��;�����F��Vrob����-����5��"Sٹin��S���L�����2����(�F�����4���).��	������\h3.����_�W��Ve�
�write�the�other�found�p�S�olynomials;���in�eac��rh�case�there�exists�a�mo�dular�form�of����_�w��reigh�t��O2�and�appropriate�lev��rel,��afor�whic�h�the�rst�terms��a�����n��	V��corresp�S�ond�to�those�v��X�alues����_�predicted��b��ry�the�Serre�conjecture.��ው�\f��P��ƹ(�x�)�UR=��x�����5��j��+����x�����4�����16�x�����3�����7�x�����2���+�57�x����35�;���N��6�=�UR8311�;�����A�p�����A�z�
�
���D����jY�=��N��9��W/�P��ƹ(�x�)�UR=��x�����5��j��+���2�x�����4�����43�x�����3���+�29�x�����2���+�2�x����3�;���N��6�=�UR8447�;�����A�p�����A�z�
�
���D����jY�=�2�����2����N�����S��P��ƹ(�x�)�UR=��x�����5��j��+����x�����4�����13�x�����3�����14�x�����2���+�18�x��+�14�;���N��6�=�UR15233�;�����A�p�����A�z�
�
���D����jY�=�2�N����TX�P��ƹ(�x�)�UR=��x�����5��j��+����x�����4�����37�x�����3���+�67�x�����2���+�21�x��+�1�;���N��6�=�UR24077�;�����A�p�����A�z�
�
���D����jY�=�2�����2����N��'�(���5��(�App��=endix:���The�z�curv��u�es��D��g�G�cmmi12�DX�����0��_��D��tG�G�cmr17�(�Dp�)��of�gen�us�0��b#��In��[��?���],�N�it�is�pro��rv�en��that�if��p��is�a�prime�n��rum�b�S�er��then�the�curv��re��X�����0����(�p�)�o�v�er��Z�����p��	��is�formally���isomorphic���to�the�curv��re�of�equation��xy�6��=����p����2�k��#��,��nin�the�neigh�b�S�orho�o�d���of�eac�h�p�S�oin�t�reducing���mo�S�d���p��to�a�sup�ersingular�p�oin��rt��S��׹,��k�QŹb�eing�one�half�the�n��rum�b�er��of�automorphisms�of��S��׹.�������14�����k��\̍��i�4�x\̍����4����If��L�X�����0����(�p�)�has�gen��rus�0�(i.e.,����p�p��=�2�;����3�;��5�;��7�;��13)��Lone�has�suc�h�a�mo�S�del�o�v�er��Z�,���giv�en�b�y�the����function��"V'�����>;�x�UR�=�����q��������ō�P���n9�(�z���)��
]ݟ[��z�A��
�΍��n9�(�pz���)������)�Ɵ�q�������~�6���24��3�Q�s^�\)
L�����p��1����D��;������ƽù(2)�����"�z�where����n9�(�z���)�UR=��q�����2�1�=�24����G�����Q����*���C��1��	U_��C��i�=1���-��(1������q�����2�n����)��and��q�Ë�=�UR�e����2�2��I{iz����.������This�	�results�from�F��Vric��rk�e�	�[��?���],�6�who�giv��res�for�eac�h�of�the�ab�S�o�v�e��p�'s�an�expression�of�the�\oubli���?"���homomorphism���j�F��:����X�����0����(�p�)���������
�%!��X�����0���(1),��whic��rh��asso�S�ciates�to�eac�h�p�S�oin�t�(�E��;���C�ܞ�)�of��X�����0����(�p�)�the���p�S�oin��rt��(�E���)�of��X�����0����(1),�parametrized�b�y�the�mo�S�dular�in�v��X�arian�t��j��ӹ.������In�B�the�follo��rwing�w�e�recall�these�equations�and�giv�e�the�expressions�of�the�corresp�S�ondences����T�����2����;���T�����3���|�o��rv�er��xthese�curv��res.�VPThe�v��X�ariable��x��is�the�one�giv�en�b�y�equation�(2),���the�in�v�olution��W�����p�����switc��rhes���x��and��y�Oȹand�the�divisor�of��x��is�(0)�����(�1�),��awhere��0�and��1��are�t�w�o�p�S�oin�ts�of��X�����0����(�p�).��u������\h1.����_��p�UR�=�2��The�equations�giv��ren�b�y�F��Vric�k�e�(mo�S�died�to�giv�e�the�mo�S�del�of��X�����0����(2)�o�v�er��Z�)�are:��w<����i�xy�Ë�=�UR2�����12��� �l���L��j�%�=������ō���(�x����+�16)����2�3������[��z�.ɮ�
�΍�.�x���������_�T�����2�����is��giv��ren�b�y��������y��n9����2���������y�n9�(�x�����2��j��+�2�����4����3�x�)����2�����12��	�x�UR�=�0���>��_�(to���eac��rh�p�S�oin�t��x��is�asso�S�ciated�b�y��T�����2���̹the�formal�sum�of�p�S�oin�ts�of�co�S�ordinate��y�i�that�are����_�ro�S�ots��of�this�p�olynomial.)��~���_��T�����3�����is��giv��ren�b�y����_��x�����4����+�y��n9����4��.=��x�����3���y��n9����3����2�����3���3�����2���x�����2���y��n9����2���(�x�+�y�n9�)��2�����2���3�����2���5�����2���xy��(�x�����2���+�y������2��.=�)+2��3�����2���1579�x�����2���y������2��.=��2�����15��	�3�����2���xy��(�x�+�y��)��2�����24��	�xy�Ë�=�UR0��u������\h2.����_��p�UR�=�3.������k�xy�Ë�=�UR3�����6���rn���8�j�%�=������ō���(�x����+�27)(�x��+�3)����2�3������[��z�R�\�
�΍�&#�x������Ӷ�����T�����2��V�:�UR�x�����3��j��+����y��n9����3�������2�����3����3�xy�n9�(�x��+��y��)����x�����2����y������2�������3�����6����xy�Ë�=�UR0���>��W��T�����3��V�:�UR�y��n9����3���������y��n9����2��.=�(�x�����3��j��+�2�����2����3�����2���x�����2���+���2����3�����2����5�y�n9�)����3�����6���y�n9x�(�x��+�2�����2���3�����2���)����3�����12��	�x�UR�=�0���������\h3.����_��p�UR�=�5.������k�xy�Ë�=�UR5�����3������i��j�%�=������ō���(�x����2�2��j��+���10�x��+�5)����2�3������[��z�N�؟
�΍�#�C�x������Ӷ�����T�����2��V�:�UR�x�����3��j��+����y��n9����3�������x�����2����y��n9����2�����2�����3����xy�n9�(�x��+��y��)����7�����2����xy�Ë�=�UR0���>��_��T�����3��V�:�UR�x�����4�����+�6��y��n9����4��d����x�����3����y��n9����3�����2����3�����2����x�����2���y��n9����2��.=�(�x�6��+��y�n9�)����3�����4����xy��(�x�����2�����+��y������2��.=�)����2����3�����2����23�x�����2���y������2��d���6��2250�xy��(�x��+��y��)����5�����6����xy�Ë�=�UR0�������15����t��\̍��i�4�x\̍����4�����\h�4.����_��p�UR�=�7.�������k�xy�Ë�=�UR7�����2���v0������j�%�=������ō���(�x����2�2��j��+���13�x��+�49)(�x����2�2���+�5�x��+�1)����2�3������[��z��_��
�΍�H�-�x�������x�����T�����2��V�:�UR�x�����3��j��+����y��n9����3�������x�����2����y��n9����2�����2�����3����xy�n9�(�x��+��y��)����7�����2����xy�Ë�=�UR0����_��T�����3��V�:�UR�x�����4����+�y��n9����4��.=��x�����3���y��n9����3����2�����2���3�x�����2���y��n9����2���(�x�+�y�n9�)��2��3��7�xy��(�x�����2����+�y������2���)��3��53�x�����2����y������2����2�����2����3��7�����2���xy�n9�(�x�+�y��)��7�����4����xy�Ë�=�UR0�������\h5.����_��p�UR�=�13.�����^o�xy�Ë�=�UR13����~˱�j�%�=������ō���(�x����2�2��j��+���5�x��+�13)(�x����2�4���+�7�x����2�3���+�20�x����2�2���+�19�x��+�1)����2�3������[��z�����
�΍�k�U�x�������x����T�����2��V�:�UR�x�����3��j��+����y��n9����3�������x�����2����y��n9����2�����2�����2����xy�n9�(�x��+��y��)����13�xy�Ë�=�UR0����_��T�����3��V�:�UR�x�����4����+���y��n9����4��L!���x�����3����y��n9����3�����2����3�x�����2����y��n9����2��.=�(�x��+��y�n9�)����3����5�xy��(�x�����2����+��y������2��.=�)����3����11�x�����2����y������2��L!���2����3����13�xy��(�x��+��y��)����13�����2����xy�Ë�=�UR0�������The���p�S�olynomials�ab�o��rv�e���that�giv��re��T�����2����;���T�����3��	k��are�of�simpler�form�than�the�classical�mo�dular���equations�]�����2����(�j��;���j���ӟ��2�0��{�)�and������3���(�j��;���j���ӟ��2�0��{�)�(that�corresp�S�ond�to�the�action�of��T�����2���a�and��T�����3���on��X�����0����(1)).���F��Vor���comparison,��w��re�recall�their�expressions:��������@�8�����2����(�j��;���j���ӟ���0��{�)�����sf]=���������j���ӟ���3���+����j���ӟ���0�3��
e����j���ӟ���2��l��j���ӟ���0�2���+�2�����4����3����31�j���j������0��{�(�j�W{�+����j������0���)������2�����4����3�����4���5�����3���(�j���ӟ���2���+����j���ӟ���0�2����)������������+3�����4����5�����3���4027�j���j������0��%��+���2�����8����3�����7���5�����6���(�j�W{�+��j���ӟ���0��{�)����2�����12��	�3�����9���5�����9������������������3����(�j��;���j���ӟ���0��{�)�����Df�=�����W�_�j���ӟ���4���+����j���ӟ���0�4��
e����j���ӟ���3��l��j���ӟ���0�3�����2�����2����3�����3���9907�j���j������0��{�(�j������2���+����j������0�2����)�+�2�����3����3�����2���31�j���ӟ���2��l��j���ӟ���0�2����(�j�W{�+��j���ӟ���0��{�)��������W�_��2�����16��	�5�����3����3�����5���17������263�j���j������0��{�(�j�W{�+����j������0���)���+�2�����15��	�3�����2����5�����3���(�j���ӟ���3���+����j���ӟ���0�3����)�+�2����3�����4����13����193����6367�j���ӟ���2��l��j���ӟ���0�2���������W�_��2�����31��	�5�����6����22973�j���j������0��%��+���2�����30���3�����3����5�����6���(�j���ӟ���2���+����j���ӟ���0�2����)�+�2�����45���3�����3����5�����9���(�j�W{�+����j���ӟ���0��{�)���������16��������;��\���G
�D��g�G�cmmi12�A�':

cmti10�@���@cmti12�?��N�ffcmbx12�62�@�cmbx8�5��N�cmbx12�3p�p�msbm8�2���
msbm10�)��u
cmex10�(q�%cmsy6�'�K�cmsy8�&!",�
cmsy10�%;�cmmi6�\$�2cmmi8�#��g�cmmi12�"�Aa�cmr6�!|{Ycmr8� ��N�G�cmbx12�D��tG�G�cmr17����ff
msbm10���g�ffcmmi12�X�Qffcmr12�D��t�qG�cmr17�X�Qcmr12�
�b>

cmmi10�K�y

cmr10�ٓ�Rcmr7�	�������