Sindbad~EG File Manager

Current Path : /usr/home/beeson/public_html/michaelbeeson/research/papers/
Upload File :
Current File : /usr/home/beeson/public_html/michaelbeeson/research/papers/Triangle.dvi

����;� TeX output 2000.04.28:2234�������n����썍��US���"V

cmbx10�TRIANGLES��WITH�VER��
�TICES�ON�LA�TTICE�POINTS�������5�e�����+�-�

cmcsc10�Michael��LJ.�Beeson����f��Dep��UTar�tment��Lof�Ma��UTthema�tics��Land�Computer�Science�������San��LJose�St��UTa�te��LUniversity������San��LJose,�CA�95192�����5��K�`y

cmr10�b�Geeson@ucscc.ucsc.edu��//%���?p�0J
cmsl10�Abstract��Bڟ�^��ٓ�Rcmr7�1����[���A�>ztriangle�>�is�called��!p�0J

cmsl10�em���b�Geddable�in��Z���^��	0e�rcmmi7�n�����if�it�is�similar�to�a�triangle�whose�v�ertices�ha�v�e�in�teger�co�Gordinates���in����b>

cmmi10�R��ǟ�^��n���E�.�x�It�w���as�already�kno�wn�that�a�triangle�is�em�b�Geddable�in��Z���^��2��)t�if�and�only�if�all�its�angles�ha�v�e�rational���tangen���ts.�.W��*�e�0�sho�w�that�a�triangle�is�em�b�Geddable�in�some��Z���^��n��	�J�if�and�only�if�it�is�em�b�Geddable�in��Z���^��5��|s�,�g�and���if��and�only�if�all�its�angles�ha���v�e��tangen�ts�with�rational�squares.�:�W��*�e�reduce�the�problem�of�em�b�Geddabilit�y���to���a�certain�Diophan���tine�equation.�<W��*�e�giv�e�a�complete�c�haracterization�of�the�triangles�em�b�Geddable�in��Z���^��n�����for�<	ev���ery��n�.�%�In�particular,�u�there�are�triangles�em�b�Geddable�in��Z���^��5���|�but�not��Z���^��4��|s�,�u�and�in��Z���^��3���but�not��Z���^��2��|s�,�u�but���surprisingly��*�,�0the�&�same�triangles�are�em���b�Geddable�in��Z���^��3���"�as�are�em�b�Geddable�in��Z���^��4��|s�.�b;A�&�triangle�is�em�b�Geddable�in��DA��Z���^��3�����if�$9and�only�if�the�tangen���ts�of�its�angles�are�all�rational�m�ultiples�of�����zP�!",�

cmsy10�p���y���zP�fe�V�����k����#�for�some�in�teger��k�tЫwhic�h�is�a�sum���of�UUthree�squares.�q�The�pro�Gofs�use�only�elemen���tary�n�um�b�Ger�theory�and�quaternions.��ߍ�These���results�can�b�Ge�equiv��q�alen���tly�form�ulated�in�terms�of�em�b�Geddable��angles��rather�than�em�b�Geddable����triangles�.��An��angle�is�em���b�Geddable�in��Z���^��n��	I>�for��n��s���5��i�its�tangen�t�is�a�rational�m�ultiple�of�����zP�p���-��zP�fe�V�����k�����-�for�some��k�P��.���It���is�em���b�Geddable�in��Z���^��n��	O�for��n�4<�=�3���or��n�4<�=�4���i��k��h�is�a�sum�of�three�squares.�6=It�is�em�b�Geddable�in��Z���^��2��D�i��k��h�is�a���square,�k�that�gFis,�the�tangen���t�is�rational.���The�n�um�b�Ger��k��ݫdep�ends�only�on�the�plane�of�the�angle,�k�so�all�angles���of�UUa�triangle�ha���v�e�UUthe�same��k�P��.���[���In���tro�UVduction�����The�Ssimplest�question�concerning�em���b�Geddabilit�y�Sis�this:�m\is�the�equilateral�triangle�em���b�eddable�in��Z���^��2��|s�?���That�7�is,�=�are�there�lattice�p�Goin���ts�in�the�plane�forming�the�v�ertices�of�an�equilateral�triangle?�g�As�it�turns�out,���there���are�not.�Y{Of�course,���the�equilateral�triangle�is�em���b�Geddable�in��Z���^��3��|s�,�with�v���ertices�at�the�p�Goin�ts�one�unit���along�xeac���h�of�the�three�axes.�W�This�illustrates�that�more�triangles�ma�y�b�Ge�em�b�Geddable�if�more�dimensions�are���allo���w�ed.��The�j�general�problem�addressed�in�this�pap�Ger�is�to�c���haracterize�the�triangles�em�b�Geddable�in��Z���^��n���<�for���eac���h�UU�n�.�q�W��*�e�giv�e�a�complete�solution�of�this�problem,�as�describ�Ged�in�the�preceding�abstract.��ߍ�The�W�problem�solv���ed�in�this�pap�Ger�has�a�surprisingly�long�history��*�,�XSand�is�connected�to�the�w�ork�of�sev�eral���other�UUauthors.�q�These�p�Goin���ts�are�discussed�in�a�separate�section�near�the�end�of�the�pap�er.���[���Dimension�Tw���o�����The��2follo���wing�prop�Gosition�is�included�as�an�in�tro�Gduction�to�the�sub��8ject.�1(In�the�prop�osition,��innit���y�coun�ts���as�UUa�rational�tangen���t.)��S���Prop�Q�osition��z1.����(J.�(McCarth���y)�A��triangle�is�em�b�Geddable�in��Z���^��2�����if�and�only�if�all�its�angles�ha�v�e�rational���tangen���ts.����$�':

cmti10�Pr��}'o�of.�p�Let�{Qtriangle��AB��qC�2m�ha���v�e�{Qits�v���ertices�on�lattice�p�Goin�ts�in��Z���^��2��|s�.��Assume�for�the�momen�t�that�neither�leg���of�J�angle��A��is�parallel�to�the��y�[٫-axis.�nSLet��AP����b�Ge�a�line�through�v���ertex��A��parallel�to�the��x�-axis.�Then�angle��A����is�O�the�dierence�of�the�angles��B��qAP��S�and��C��AP�c��.�o�These�t���w�o�O�angles�eviden���tly�ha�v�e�rational�tangen�ts.�o�But�no�w���w���e�UUma�y�use�the�form�ula��ǩ�ff<�	�������-:��Aa�cmr6�1��*��|{Ycmr8�It��w�Îas�John�McCarth�y�who�p�<roin�ted�out�the�result�on�em�b�<reddabilit�y�in���2cmmi8�R�����-:�2��@;�and�ask�ed�for�a�generalization�to��R�����-:�
;�cmmi6�n��*�.�a�Thanks��
�are��Xdue�to�R.�Alp�<rerin,�for�p�oin�Îting�out�Lemma�7.������U�T�Îyp�<reset��Xb�y���K�cmsy8�A����A���M����S�-T����A���E��MX�����1���*������2�MICHAEL���J.�BEESON�DEP��J�AR�TMENT���OF�MA��J�THEMA�TICS���AND�COMPUTER�SCIENCE�SAN�JOSE�ST��J�A�TE���UNIVERSITY�SAN�JOSE,�CA�95192�BEESON@UCSCC.UCSC.EDU����n������������x�tan���B(��BZ��8����)��=�����<$���O�tan��l������tan��UR����K�w�fe?���	(֍�1�+��tan��UR����"�tan��Д�������0���to�s�conclude�that�angle��A��also�has�a�rational�tangen���t.��nIn�case�one�leg�of�angle��A��is�parallel�to�the��y�[٫-axis,���w���e�Y�in�terc�hange�the�roles�of�the��x�-axis�and��y�[٫-axis�in�this�argumen�t.�~_This�will�b�Ge�p�ossible�unless�angle��A��is�a���righ���t�UUangle,�in�whic�h�case�the�conclusion�is�immediate.����Con���v�ersely��*�,�a,supp�Gose�^�all�the�angles�of�triangle��AB��qC��ha���v�e�^�rational�tangen���ts.��2If�one�of�the�angles�is�a�righ�t���angle,���the��em���b�Geddabilit�y�is�immediate,���so�w�e�assume�that�none�of�the�angles�is�a�righ�t�angle.�,Drop�an�altitude����AP�J�from���v���ertex��A��to�side��B��qC�Y׫(p�Gossible�extended),��so�that��P��is�on�line��B��qC���.�Y�Then�the�ratios��AP��V=B�P�J�and����AP��V=C��P���are�Owrational,�P�b�Geing�the�tangen���ts�of�angles��B���and��C���resp�ectiv���ely��*�.�o�Express�these�t�w�o�fractions�o�v�er�a���common�Pdenominator�as��AP��V=B��qP�*��=���u=��q�N�g,�and��AP�=C��P�*��=���v�[�=��q�N��.�pAssume�for�the�momen���t�that��P����lies�b�Get�w�een����B�)ܫand��k�C���.�8zThen�triangle��AB��qC�`��is�similar�to�triangle�(0�;���uv�[٫),���(��N�v�;��0),���(�N�u;��0),�since��kt���w�o�corresp�Gonding�angles���ha���v�e�UUthe�same�tangen���t.�q�The�cases�where��P���lies�to�the�left�of��A��or�the�righ�t�of��C�q�are�similar.�
�9����

msam10������Remark��D�:���The�i7criterion�in�Prop�Gosition�1�do�es�not�extend�to�higher�dimensions.�#F��*�or�example,��pthe�equilateral���triangle�UUis�em���b�Geddable�in��Z���^��3���ȫbut�not�in��Z���^��2��|s�.��pj���Em���b�UVeddabilit�y�of�Angles�and�T����riangles�Compared�����F��*�or��pthe�record,��kw���e�dene�an�angle�to�b�Ge�em�b�Geddable�in��Z���^��n����if�it�is�one�of�the�angles�of�a�triangle�em�b�Geddable���in�UU�Z���^��n��q~�.���5��Prop�Q�osition��T2.����If�UUan�angle����r�is�em���b�Geddable�in��Z���^��n�����(for�an�y��n�),�then���tan������RH�2���:���r�is�rational.����Pr��}'o�of.�p�Let��triangle��AB��qC��ҫlie�in��Z���^��n���4�with�its�v���ertices�on�lattice�p�Goin�ts.�_�Consider�sides��AB��'�and��AC��ҫas�v�ectors,���and�UUtak���e�their�dot�pro�Gduct:�qDZAB��Q��8�AC�~4�=���j�AB��q�jj�AC���j�����cos���߱��,�UUwhere����r�is�the�angle�at�v�ertex��A�.�q�Hence���絍���<�cos���ʠa����2����|��5�=�����<$��F�(�AB��Q��8�AC���)���^��2����K�w�fe3a�	(֍�j�AB��q�j���r�2��|s�j�AC���j���r�2��������:���The��expression�on�the�righ���t�hand�side�is�a�rational�function�of�the�co�Gordinates�of��A�,�=B�B��q�,�and��C���.��hSince���those��co�Gordinates�are�in���tegers,�it�follo�ws�that��cos���L���^��2��sű�08�is�rational.�-Hence��sin���09���*�2��WT��~�=��a1��c����cos�����^��2��&
���is��also�rational,���and�UUhence��tan������RH�2���:��5�=���sin���6���*�2��5Q��G=�����cos���7��^��2��5R���r�is�rational�to�o.�
���pj���The���follo���wing�lemma�connects�the�em�b�Geddabilit�y�of�a�triangle�with�the�em�b�Geddabilit�y�of�its�angles�con-��DA�sidered�UUseparately��*�.�q�(W�e�coun���t�innit�y�as�a�rational�tangen�t,�and�as�a�rational�m�ultiple�of�����zP�p�������zP�fe�V�����k����0�.)����Lemma��3.����(i)�?�If�the�square�of�the�tangen���t�of�eac�h�angle�of�a�triangle��T���is�rational,�z
then�there�exists�a���z�(square-free)�UUp�Gositiv���e�in�teger��k����suc�h�that�eac�h�tangen�t�is�a�rational�m�ultiple�of�����zP�p�������zP�fe�V�����k����0�.����(ii)�y�Moreo���v�er,����k��o�dep�Gends�only�on�the�plane�of�the�triangle,�i.e.��Ran���y�t�w�o�triangles�in�the�same�plane�ha�v�e���the�UUsame��k�P��.����(iii)�UUStill�more�generally��*�,�an���y�t�w�o�lattice�angles�in�the�same�plane�ha�v�e�the�same��k�P��.����Pr��}'o�of.�p�Ad���(i):��1Let�the�angles�of�the�triangle�b�Ge���	z�,�������,�and����
��8�.�2fIn�case�one�of�the�angles�is�a�righ���t�angle,���sa�y����
��8�,��ythen����tan�������is���the�recipro�Gcal�of��tan������,��yso�the�conclusion�is�trivial.�;�W��*�e�ma���y�assume�therefore�that�none�of���the�UUangles�is�a�righ���t�angle.�q�In�particular,�1�8����tan��UR����"�tan�����q�is�UUnot�zero.�Then��dX����U�tan���rT�
�UP�=��������<$������tan������BZ�+��8�tan��UR���33�w�fe?���	(֍�1�8����tan��UR����"�tan��Д��������8��Supp�Gose��$�tan��A��*�=� ��a������p���UW����fe�ğ`��j����*«and��$�tan�������=��b����zP�p���UW��zP�fe�V�����k�����T�where�$��k�u>�and��j��2�are�square-free�p�Gositiv���e�in�tegers,�X{and��a��and��b��are���non-zero�UUrationals.�q�Then��V�����ttan����
�UP�=��������<$��33�a������p���UW����fe�ğ`��j����>��+�8�b����zP�p���UW��zP�fe�V�����k�����33�w�fe0���	�ō����1�8���ab����s4�p���UW��s4�fe
6��̍�j���k�������������Rationalizing�UUthe�denominator�on�the�righ���t,�w�e�ha�v�e�������{
tan����|�
�UP�=��������<$��33�(1�8�+��a���^��2��|s�j����)�b����zP�p���UW��zP�fe�V�����k������+�(1�+��b���^��2���k�P��)�a������p���UW����fe�ğ`��j�����33�w�fepT�	(֍�(�M�1�8���a���r�2��|s�b���r�2���j���k��������������z�?�TRIANGLES���WITH�VER��J�TICES�ON�LA�TTICE�POINTS��z-3�����n����썫Hence����tan����_��RH�2��z�
�ͫis���a�rational�plus�a�rational�m���ultiple�of�����s4�p�����s4�fe
6��̍�j���k�����,��ewhic�h�is�irrational�unless��j��c�=�رk�P��.���This�completes���the�UUpro�Gof�of�part�(i)�of�the�Lemma.����P���art��(ii)�eviden�tly�follo�ws�from�part�(iii),��since�the�six�dieren�t�angles�of�t�w�o�triangles�are�sp�Gecial�cases���of�UUangles�in�the�plane.����No���w�W`for�part�(iii).�w�Let�t�w�o�non-righ�t�lattice�angles���`ګand����|�b�Ge�giv�en�in�the�same�plane.�w�Unless�there�are���parallel�B�sides,�Fyup�Gon�extending�the�sides�of�the�angles�t���w�o�B�triangles�will�b�e�formed,�Fywith�a�common�v���ertex��P����opp�Gosite�R�angle���\r�in�one�triangle�and�opp�osite�angle�����in�the�other�triangle.�p�(It�ma���y�b�e�necessary�to�replace���one�Ipangle�with�its�v���ertical�angle�or�a�supplemen�tal�angle,��wwhic�h�w�on't�aect�the�square�of�the�tangen�t.)���Rescaling�W�the�gure�if�necessary��*�,�Xthe�in���tersection�p�Goin�ts�of�the�sides,�Xone�of�whic�h�is��P�c��,�Xcan�b�Ge�made�lattice���p�Goin���ts.��-Assuming�qwthere�are�no�parallel�sides,�xit�will�b�e�p�ossible�to�c���ho�ose�the�v���ertex��P���so�that�the�angle�at����P�xϫis�@not�a�righ���t�angle.�\k(Otherwise�the�gure�m�ust�b�Ge�a�square�with�����and����\�diagonally�opp�osite.)�\kW��*�e�apply���part��$(i)�of�the�lemma�successiv���ely�to�the�t�w�o�triangles�con�taining���Ğ�and���B@�to�sho�w�that���Ğ�and���B@�ha�v�e�the���same�UU�k�P��.����The�\gpro�Gof�is�not�quite�nished,�^,b�ecause�w���e�still�m�ust�consider�the�case�in�whic�h�no�triangles�are�formed,���i.e.�ESt���w�o�F�angles���O��and���͟�in�the�same�plane�with�one�side�of����parallel�to�one�side�of�����.�ESIn�this�case�w���e���can�A_translate�the�angles�un���til�they�ha�v�e�a�common�v�ertex�and�a�common�side.�k Assume�for�deniteness�that���the��common�side�is�b�Get���w�een��the�t���w�o��angles.�!�Then�the�sum���i�+�`��(�is�em���b�eddable.�!�Therefore��tan���֟�RH�2��~I�(��i�+�`����)�is���rational.�&@Using��}the�same�argumen���t�as�w�e�used�to�pro�v�e�part�(i),���i.e.�&@the�form�ula�for�the�tangen�t�of�a�sum,���w���e�d�can�sho�w�that���nK�and�����m�ust�ha�v�e�the�same��k�P��.��=Similarly��*�,�h�using�the�form�ula�for��tan��֛(��L���C3����),�h�w�e�can�treat���the�UUcase�of�parallel�sides�in�whic���h�one�angle�lies�inside�the�other.�
������One��of�the�referees�p�Goin���ted�out�that�the�in�teger��k�B+�in�Lemma�3�is�related�to�the�area�of�the�lattice�triangle.��DA�Putting��Hthe�matter�simply��*�,����D��zP�p���,���zP�fe�V�����k����o9�is�a�rational�m���ultiple�of�the�area.���This�observ��q�ation�yields�another�pro�Gof�of���part�Ϫ(i)�of�Lemma�3.�E9The�computation�is�a�little�simpler,��fbut�w���e�can't�get�part�(iii)�without�the�computation���giv���en�y�ab�Go�v�e.�ߨSince�the�observ��q�ation�ab�Gout�the�area�is�in�teresting�in�its�o�wn�righ�t,��w�e�giv�e�that�computation���to�Go:���Let�hf�a��and��b��b�e�lattice�p�oin���ts�dening�t�w�o�adjacen�t�sides�of�a�triangle�with�angle������at�origin.���Then�the���p�Gerp�endicular�UUfrom��a��to��b��is�giv���en�b�y��5����ıu���=��a�8������<$��l�(�a����b�)�b��l�w�fe��	(֍�D��j�b�j���r�2�������z}��The�UUarea��S���is�th���us�giv�en�b�y�4�S������^��2����=���j�u�j���^��2��|s�j�b�j���^��2��C��=��j�a�j���^��2���j�b�j���^��2���S��8�(�a����b�)���^��2���.�q�W��*�e�UUha���v�e��w�����A8tan���]���5�=�����<$���K�j�u�jj�b�j���K�w�fe ��	(֍�)�a�8���b�����2�=�����<$���_2�S���K�w�fe�}�	(֍acdotb����������The�'Jform���ula�sho�ws�that�4�S������^��2��7J�is�an�in�teger.�bnW��*�rite�4�S������^��2����=���m���^��2��|s�k�w�where��k��is�square-free.�bnThen��tan��C���ng�is�a�rational��DA�m���ultiple�UUof�����zP�p�������zP�fe�V�����k����0�.�������The�T����riangle�Equations����Denition.����The�UU�triangle��Tequations��E����(�k�P�;���n�)��are��״����4�k�P��(�a������2����1����S�+�8�a������2����2����+���:���:�:��g�+��a������2���፴n���q~�)��=��u������2����1����S�+��u������2����2����+���:���:�:���/u������2���፴n�����ۍ����a����1��|s�u����1���S�+�8�a����2���u����2���+��8�:���:�:���/a����n��q~�u����n��8��=��0������Throughout�UUthe�pap�Ger,�w���e�consider�only�non-trivial�solutions�in�in�tegers��a����i�����and��u����i���of�these�equations.������Prop�Q�osition�� 4.����If��a�triangle�with�an�angle�with�tangen���t������zP�p���UW��zP�fe�V�����k������is�em�b�Geddable�in��Z���^��n��q~�,�?�where����is�rational,���then�>the�triangle�equations��E����(�n;���k�P��)��ha���v�e�>a�non-zero�solution,�*in�whic���h�the�v��q�ariables�ha�v�e�no�common�factor.���z��Con���v�ersely��*�,��Cif���E����(�n;���k�P��)��has�a�non-zero�solution,�and�if�triangle��AB��qC�=��has�all�its�tangen���ts�of�the�form������zP�p���UW��zP�fe�V�����k�������for�UUrational���,�then�triangle��AB��qC�q�is�em���b�Geddable�in��Z���^��n��q~�.��}N��Remark��D�:��VOf���course�there�are�man���y�non-em�b�Geddable�triangles�with��one��tangen�t�of�the�sp�Gecied�form,���as���y���ou�Fcan�x�t�w�o�v�ertices�and�let�the�third�mo�v�e�along�one�side�of�the�triangle.�^mHence�the�second�condition�in���the�UUtheorem�is�needed.����Pr��}'o�of.�p�First��supp�Gose�the�triangle��AB��qC��ޫhas�its�v���ertices�on�lattice�p�oin���ts�in��Z���^��n��q~�.�tAs�in�the�previous�pro�of,���w���e�Ȟdrop�the�altitude�from�v�ertex��B�I�to�p�Goin�t��P�,-�on�side��AC���.�ˤAs�in�that�pro�Gof,��q�P��has�rational�co�Gordinates,����)�������4�MICHAEL���J.�BEESON�DEP��J�AR�TMENT���OF�MA��J�THEMA�TICS���AND�COMPUTER�SCIENCE�SAN�JOSE�ST��J�A�TE���UNIVERSITY�SAN�JOSE,�CA�95192�BEESON@UCSCC.UCSC.EDU����n����썫and�$�enlarging�the�triangle�if�necessary��*�,�.Mw���e�ma�y�assume��P���has�in�teger�co�Gordinates.�a�P�erforming�a�translation,���w���e�}�ma�y�assume��P��T�is�the�origin.��W��*�e�no�w�ha�v�e�v�ector��A��of�magnitude��AP�c��,���and�v�ector��B��6�of�magnitude��B��qP�c��,��DA�whic���h�UUare�orthogonal.�q�The�ratio��B��qP��V=��q�AP�*��=���tan��㊱A���=������zP�p���UW��zP�fe�V�����k����0�b�y�UUh�yp�Gothesis.�q�W��*�e�th�us�ha�v�e�����Aڍ���(�tan���qʟ���2����A�)�j�A�j�����2��C��=���k�P��j�(�A�)�j�����2���=��j�B��q�j�����2��|s�:��x���Th���us�UU(�A;���B��q�)�solv�es�the�triangle�equations��E����(�n;���k�P��).����Con���v�ersely��*�,�>�supp�Gose���giv�en�a�solution�(�a;���u�)�of�the�triangle�equations,�>�and�a�triangle��AB��qC���suc�h�that���z��tan��r�A���=������zP�p���UW��zP�fe�V�����k����#�with�I@���rational.�m�As�b�Gefore�drop�an�altitude��B��qP��ϫfrom��B�ɱ�to�side��AC���.�Consider�rst�the�case�in���whic���h��ϱP��^�lies�b�Get�w�een��A��and��C���.�$7Then�the�triangle�with�t�w�o�v�ertices�at������^��O!�cmsy7��1��
�t�a�,���and��u��will�ha�v�e�the�correct��DA�tangen���t���stan����A��s�at�v�ertex��a�.��!By�h�yp�Gothesis,��zthe�tangen�t�at�v�ertex��C�l��has�the�form������zP�p���UW��zP�fe�V�����k����� �for�some�rational���.���T��*�aking�9�the�third�v���ertex�to�b�Ge�����^���1��
�t�a��yields�the�correct�tangen�t��tan��V�B���at�this�v�ertex.�h�Therefore�the�triangle�is���similar�UUto�the�giv���en�one.�q�The�cases�in�whic�h��P���do�Ges�not�lie�b�et���w�een�UU�A��and��C�q�are�treated�similarly��*�.�
�������Dimension�Fiv���e�or�More����Lemma��T5.����If�UU�n�����5��then�the�triangle�equations�ha���v�e�a�non-zero�solution�for�an�y��k�P��.��Aڍ�Pr��}'o�of.�p�It�n�suces�to�consider��n��g�=�5,�usince�n�w���e�can�alw�a�ys�let�the�v��q�ariables��u����i����and��a����i���for��i��g>��5�n�b�Ge�zero.���Let��k����b�Ge�UUgiv���en.�q�Then��k���can�b�e�written�as�the�sum�of�four�squares�(see�e.g.�q�Hardy�and�W��*�righ���t,�p.�302):�����Aڍ���#�k���=���u������2����1����S�+�8�u������2����2����+��u������2����3����+��u������2����4����Aڍ�Let�UU�u����5��C��=��0,�and�let��a����1���=��0�=��a����2���=��a����3���=��a����4��|s�,�UUand��a����5���=��1.�
������Remark��D�:�-Similarly��*�,��`if����k���is�a�sum�of��n�wL���1���squares,�then�the�triangle�equations��E����(�n;���k�P��)�ha���v�e���a�non���trivial���solution.��Aڍ�Theorem��T6.����The�UUfollo���wing�are�all�equiv��q�alen�t:������UU�T��*�riangle��T����is�em���b�Geddable�in��Z���^��n�����for�some��n�.������UU�All�the�tangen���ts�of�the�angles�of�triangle��T����ha�v�e�rational�squares.��
ߍ���UU�F��*�or�some��k�P��,�all�the�tangen���ts�of�the�angles�of�triangle��T����are�of�the�form������zP�p���UW��zP�fe�V�����k����0�for�rational���.��������The�triangle�equations��E����(�n;���k�P��)��ha���v�e��a�non-zero�solution�and�all�the�tangen���ts�of�the�angles�are�of�the���z�form�UU�����zP�p���UW��zP�fe�V�����k����0�for�rational���.������UU�T��*�riangle��T����is�em���b�Geddable�in��Z���^��5��|s�.��Aڍ�Pr��}'o�of.�p�W��*�e���sho���w�that�eac�h�claim�in�the�theorem�implies�the�next;���since�the�last�one�is�a�sp�Gecial�case�of�the���rst,���that��will�suce.���Supp�Gose�triangle��T�<��is�em���b�eddable�in��Z���^��n��q~�.���By�Prop�osition�2,���the�tangen���ts�of�all�the���angles�UUof��T���ha���v�e�UUrational�squares.������No���w�u5supp�Gose�all�the�tangen�ts�of�angles�of��T��īha�v�e�rational�squares.��iBy�Lemma�3,��-there�is�a�p�Gositiv�e��DA�square-free�UUin���teger��k���suc�h�that�all�the�tangen�ts�of�angles�of��T���are�rational�m�ultiples�of�����zP�p�������zP�fe�V�����k����0�.��
ߍ�No���w��supp�Gose�all�the�tangen�ts�are�rational�m�ultiples�of�����zP�p����k��zP�fe�V�����k����^��.��By�Prop�Gosition�4,���the�triangle�equations����E����(�n;���k�P��)�UUare�solv��q�able.������No���w��supp�Gose��E����(�n;���k�P��)�is�solv��q�able�and�the�angles�of�a�triangle�ha�v�e�tangen�ts�whic�h�are�rational�m�ultiples���z�of����۟�zP�p����2��zP�fe�V�����k����|��.�W[By���Lemma�5,���the�equations�are�solv��q�able�already�when��n�F��=�5.�By���the�second�half�of�Prop�Gosition�4,���the�UUtriangle�is�em���b�Geddable�in��Z���^��5��|s�.�
�������Quaternions�and�Orthogonal�T����ransformations�of��R��ǟ�^��4������Bac���kground���information�on�quaternions�can�b�Ge�found�in�Hardy�and�W��*�righ�t,��p.�:B303.�W��*�e���assume�the�reader���kno���ws�(the�basic�prop�Gerties�of�quaternions.��A�'�four-v�ector�(�x����1��|s�;���x����2���;�x����3���;�x����4���)�(can�b�Ge�regarded�as�a�quaternion����x����1��h��+���x����2��|s�i��+��x����3���j��+��x����4���k�.�d�T��*�o�.�x�some�notation:�^�If��x���=��x����1��h��+���x����2���i��+��x����3���j��+��x����4���k�,�6�then�.�the�conjugate��x���is�dened�b���y����x����=��x����1������z@�x����2��|s�i����x����3���j����x����4���k�,�	the��norm�is�dened�b���y��j�x�j���^��2��C��=���x���^���2��l�1������+�z@�x���^���2��l�2����+��x���^���2��l�3����+��x���^���2��l�4���|s�.�RW��*�e��ha���v�e��xx����=��j�x�j���^��2�����+�z@0�i��+�0�j��+�0�k����whic���h�UUw�e�shall�iden�tify�with�the�scalar��j�x�j���^��2��|s�.�q�The�m�ultiplicativ�e�in�v�erse�of��x��is��x���^���1��
���=���x�8���=�j�x�j���^��2��|s�.����@������z�?�TRIANGLES���WITH�VER��J�TICES�ON�LA�TTICE�POINTS��z-5�����n������Lemma�n77.����Giv���en��Ga�xed�quaternion���	z�,���an�orthogonal�transformation��T������	�)�on��R��ǟ�^��4��j��is�dened�b�y��T��������x����=��x�	z�,���where���on�the�righ���t�w�e�mean�quaternion�m�ultiplication.���That�is,���T������	m��preserv�es�orthogonalit�y�and�m�ultiplies���lengths�UUb���y�a�constan�t�factor.��
A��Pr��}'o�of.�p�A�lPsimple�l�calculation.�$/Let��x���=��x����1���+�gN�x����2��|s�i��+��x����3���j��+��x����4���k�,���y�"�=���y����1���+��y����2���i��+��y����3���j��+��y����4���k�,��and�l���В�=�������1���+������2���i��+������3���j��+������4���k�.���Note��[that�the�dot�pro�Gduct�of�t���w�o��[v�ectors��x�����y��4�is�the�real�part�of�the�quaternion�pro�Gduct��xy�[ٷ�.�4�Hence�(�x�	z�)����(�y�[���)���is��the�real�part�of�(�x�	z�)(�y�[���)����=��x�	z������6�y����=��x�j��	z�j���^��2��|s�y����=��j��	z�j���^��2��|s�xy���,��Dwhose��real�part�is��j��	z�j���^��2��|s�x��6���y��.�1+Hence��orthogonalit���y���is�UUpreserv���ed.�q�T��*�aking��x���=��y��.�w�e�UUsee�that�lengths�are�m�ultiplied�b�y��j��	z�j���^��2��|s�.�
������Dimension�4�����W��*�e�,Krst�c���haracterized�the�triangles�em�b�Geddable�in��Z���^��4�����b�y�using�a�computer�to�sho�w�that�certain�triangles���are��Tnot�em���b�Geddable�in��Z���^��4��|s�.�[�This�pro�of�sho���w�ed��Tb�y�direct�searc�h�that�the�triangle�equations��E����(4�;���k�P��)�ha�v�e�no���solutions��mo�Gd�32�in�whic���h�the�v��q�ariables�ha�v�e�no�common�factor,�O�when��k�eī=�-7�;����15�;��23�;��31.��#It��is�p�Gossible�to���pro���v�e��6Theorem�8�b�Gelo���w�from�this�result.�dkThe�program�w�e�used,��nwritten�in�the�C��!language,�ran�for�sev���eral���hours�UUon�an�IBM�PC/A��*�T.�Later�w���e�found�the�more�insigh�tful�pro�Gof�giv�en�here.��
A��Theorem�\8.����The�gFtriangle�equations��E����(4�;���k�P��)��are�solv��q�able�i��k����is�a�sum�of�three�squares.���Geometrically���z�stated:�ĴA�~�triangle�~�is�em���b�Geddable�in��Z���^��4���>�if�and�only�if�all�of�its�tangen�ts�are�rational�m�ultiples�of�����zP�p����"��zP�fe�V�����k����Yx�,��)where����k����is�UUa�sum�of�three�squares.����Pr��}'o�of.�p�If�A�a�triangle�is�em���b�Geddable�in��Z���^��n��q~�,�E�for�an�y��n�,�E�then�there�is�a��k��K�suc�h�that�the�tangen�ts�of�its�angles�all��DA�lie�fxin��Q�(����zP�p���UW��zP�fe�V�����k����
ڭ�),�j�as�has�b�Geen�pro���v�ed�fxab�o�v�e.��0Hence�fxthe�main�claim�of�the�theorem�follo���ws�from�the�equiv��q�alence���of�UUthe�rst�t���w�o�UUprop�Gositions.��
k��If���k�٩�is�a�sum�of�three�squares,��then��E����(4�;���k�P��)�is�automatically�solv��q�able,�as�remark���ed�after�Lemma�5.��Th�us���it��will�suce�to�sho���w�that�if��E����(4�;���k�P��)�is�solv��q�able,�-�then��k�S)�is�a�sum�of�three�squares.�y�Supp�Gose�that��a��and��u����are�Qfour-v���ectors�solving��E����(4�;���k�P��),�Q�that�is��k��j�a�j���^��2��C��=���j�u�j���^��2��͏�and��a�0m���u��=�0.�p_Consider�Qthe�four-v���ectors�as�quaternions.���Let�ī�b����=��aa���(considered�as�a�four-v���ector�or�quaternion,��not�as�a�scalar),�and�let��v�܀�=����ua��,�where�w���e�mean���quaternion�mm���ultiplication�on�the�righ�t.���Since�quaternion�m�ultiplication�preserv�es�orthogonalit�y��*�,���w�e�ha�v�e����b�8���v�"�=��0.�q�W��*�e�UUha���v�e��
B������������
�k�P��j�b�j�����2��������ޮܫ=���k�P��j�a�j�����4���������������������ޮܫ=���j�a�j�����2��|s�k�P��j�a�j�����2�������������������ޮܫ=���j�a�j�����2��|s�j�u�j�����2�������������������ޮܫ=���j�a�j�����2��|s�uu������#�������������ޮܫ=���uaa�8���u�������#�������������ޮܫ=���ua�(�ua�)�������������������ޮܫ=���v�[�v����������������������ޮܫ=���j�v�[ٷj�����2�������B��Hence��y�b��and��v�XR�are�a�new�solution�to�the�triangle�equations��E����(4�;���k�P��).�T)But��b���=��aa���y�has�only�its�rst�comp�Gonen���t���non-zero.��(Since��u�v�N�is�orthogonal�to��b��b���y�Lemma�7,���v��lies�in�the�three-dimensional�subspace�of�v���ectors�with���zero�UUrst�comp�Gonen���t.�q�Hence�w�e�ha�v�e��k�P��j�b�j���^��2��C��=���v���^���[ٮ2��l�2���,�+�8�v���^���[ٮ2��l�2����+��v���^���[ٮ2��l�3����L�.�q�Hence�UU�k�P��j�b�j���^��2���ȫis�a�sum�of�three�squares.��
k��Note�Nthat�since��b���=��aa��,�*w���e�Nha�v�e��j�b�j���^��2��C��=���j�a�j���^��4��|s�,�*so��j�b�j��=��j�a�j���^��2�����is�an�in���teger.�_�It�is�w�ell-kno�wn�(see�e.g.�_�LeV��*�eque���p.��187)��athat�a�n���um�b�Ger��afails�to�b�e�a�sum�of�three�squares�if�and�only�if�it�is�a�p�o���w�er��aof�4�times�a�n���um�b�er���congruen���t�~rto�7�mo�Gd�8.��If��k��	�w�ere�of�this�form,���then��k�P��j�b�j���^��2����w�ould�also�b�Ge�of�this�form,���since�ev�ery�o�Gdd�square���is���congruen���t�to�1�mo�Gd�8.��3Hence�it�follo�ws�from�the�facts�that��k�P��j�b�j���^��2��AA�is�a�sum�of�three�squares�and��j�b�j��is�an���in���teger�UUthat��k���is�also�a�sum�of�three�squares.�
���
k���Another��Opro�Gof�of�Theorem�8���:�3��J.�McCarth���y�has�p�oin���ted�out�that�the�fact�that�the�tangen�ts�are�rational��DA�m���ultiples��of�����zP�p���
k��zP�fe�V�����k����Dիwhere��k���is�a�sum�of�three�squares�can�b�Ge�pro�v�ed�without�use�of�the�triangle�equations,��as���follo���ws:�;�b�y��Lemma�3,���it�suces�to�consider�only�one�angle���G�,�with�v���ertex�at�origin�and�sides�giv�en�b�y�v�ectors����Wy������6�MICHAEL���J.�BEESON�DEP��J�AR�TMENT���OF�MA��J�THEMA�TICS���AND�COMPUTER�SCIENCE�SAN�JOSE�ST��J�A�TE���UNIVERSITY�SAN�JOSE,�CA�95192�BEESON@UCSCC.UCSC.EDU����n����썱x�UU�and��y�[٫.�q�W��*�e�ha���v�e��ְ�������������Z~tan�����H����2����c��������ͳ_�=���sec����m����2��È�����8�1������؍������������ͳ_=�����<$���K�j�x�j���^��2��|s�j�y�[ٷj���^��2����K�w�fen�	(֍�Lv�(�x�8���y�[٫)���r�2������&u̷�8�1�����Dፍ�����������ͳ_=�����<$���K�j�x�j���^��2��|s�j�y�[ٷj���^��2���S��8�(�x����y��)���^��2����K�w�feI�Ο	(֍��&�(�x�8���y�[٫)���r�2����������$���Therefore��UUtan��qDZ��r�is�UUa�rational�m���ultiple�of��ְ��������������k�������H-�=���j�x�j�����2��|s�j�y�[ٷj�����2���S��8�(�x����y��)�����2����������������������H-�=��(�x������2���፴i���|s�)(�y������[ٮ2���፴i����L�)�8���(�x����i��TL�y����i���)�����2��������The�lNpro�Gof�will�b�e�completed�b���y�observing�an�iden�tit�y�whic�h�expresses�the�last�expression�as�a�sum�of�three���squares:���������������T��(�x������2���፴i���|s�)(�y������[ٮ2���፴i����L�)�8���(�x����i��TL�y����i���)�����2�����������������L㢫=�������T��(�x������2����1����S�+�8�x������2����2����+��x������2����3����+��x������2����4���|s�)(�y������[ٮ2����1���,�+��y������[ٮ2����2����+��y������[ٮ2����3����+��y������[ٮ2����4����L�)����(�x����1��|s�y����1���S�+��x����2���y����2���S�+��x����3���y����3���S�+��x����4���y����4���)�����2�����������������L㢫=�������T��(�x����1��|s�y����2���S��8�x����2���y����1���+�8�x����3���y����4����8�x����4���y����3���)�����2�������������������V㠫+�8�(�x����1��|s�y����3���S���x����3���y����1���S���x����2���y����4���S�+��x����4���y����2���)�����2�������������������V㠫+�8�(�x����1��|s�y����4���S���x����4���y����1���S�+��x����2���y����3���S���x����3���y����2���)�����2��������McCarth���y�UUfound�this�iden�tit�y�b�y�generalizing�the�corresp�Gonding�three-dimensional�iden�tit�y������Ʒj�x�j�����2��|s�j�y�[ٷj�����2��C��=��(�x�8���y��)�����2���S�+��j�x����y��j�����2�����It���is�really�just�the�iden���tit�y���expressing�the�m���ultiplicativit�y���of�the�quaternion�norm,�%�applied�to�the�t���w�o���quaternions�UU�x��and��y�[ٷ�.����This�_�alternate�pro�Gof�is�in���teresting�b�ecause�it�sho���ws�a�uniformit�y�in�the�deriv��q�ation�of�the�necessary�condition���on�UU�k���for�dieren���t�dimensions;�the�t�w�o-dimensional�case�of�this�iden�tit�y�is���㍍�������������<$����ۙtan����Mc��RH�2���t~����i=�w�fev��	(֍�(�x�8���y�[٫)���r�2��������
�=���������((�x������2����1����S�+�8�x������2����2���|s�)(�y������[ٮ2����1���,�+��y����2���)�����2���S���(�x����1���y����1���S�+��x����2���y����2���)�����2�������������������
�=���������((�x����1��|s�y����2���S��8�x����2���y����1���)�����2������ְ��whic���h�UUexplains�wh�y�the�tangen�t�is�rational�in�t�w�o�dimensions.�����Corollary���9.����There�5mare�triangles�em���b�Geddable�in��Z���^��5�����but�not��Z���^��4��|s�.�g%F��*�or�example,�;�the�isosceles�triangle�of�base��A�2�UUand�heigh���t������P�p��������P�fe�E���7�������.��9̍��Dimension�3��{ݍ��The�UUfollo���wing�v�ery�short�pro�Gof�to�ok�a�long�time�to�nd;�see�the�P���ostscript.����Theorem���10.����If�\��k��M�is�a�sum�of�three�squares,�^�then��E����(3�;���k�P��)��is�solv��q�able�in�in���tegers.���Hence�the�same�triangles���are�UUem���b�Geddable�in��Z���^��4�����as�in��Z���^��3��|s�,�and��E����(3�;���k�P��)��is�solv��q�able�if�and�only�if��k����is�a�sum�of�three�squares.����Pr��}'o�of.�p�By�B�our�results�on�em���b�Geddabilit�y�B�in��Z���^��4��|s�,�Fmit�suces�to�pro���v�e�B�the�rst�claim.�k�Supp�ose��k���=���x���^��2����+���y��[ٟ�^��2����+��z��p���^��2���
�.���Dene����ħ�a���=�(�z�p�;���z�;���y�����8�x�)��{ݍ.������Zu�u���=�(�y��[ٟ���2��,�+�8�xy����+��z��p�����2���
�;���x�����2���S�+��xy��+��z��p�����2���
�;���xz��w���y�[�z�p��)����m������z�?�TRIANGLES���WITH�VER��J�TICES�ON�LA�TTICE�POINTS��z-7�����n����썑�One�`Ncan�easily�c���hec�k�`Nthat��a��and��u��(regarded�as�quaternions�with�zero�real�part)�are�obtained�b���y�m�ultiplying���the��#kno���wn�solution�(1�;����0�;��0�;��0)��#and�(0�;���x;�y�[�;�z�p��)��#of��E����(4�;���k�P��)�on�the�righ�t�b�y�the�quaternion�0�>~+��z�p��i��+��z��j��+�(��y��W���x�)�k�.���z�Hence,���b���y��*Lemma�7,�the�transformed�v���ectors�are�still�orthogonal�and�ha�v�e�the�same�ratio�����zP�p���܁��zP�fe�V�����k������of�length,���so����a�UU�and��u��solv���e�the�triangle�equations��E����(3�;���k�P��).�
���}���Remark��D�:�]One�+�can�pro�Gduce��a��and��u��deus�ex�mac���hina��and�v�erify�b�y�a�simple�direct�computation�that�they���do��solv���e�the�triangle�equations,�%without�ev�er�men�tioning�quaternions.�O�F��*�or�example,�%t�yp�Ge�the�three�equations���for�UU�k�P��,��a�,�and��u��in���to��Mathematica��and�then�ask����<x

cmtt10�Simplify[a.u]��and��Simplify[k(a.a)�?�-�u.u]�.��s���Em���b�UVeddabilit�y�of�regular�p�olygons�in�plane�lattices�����W��*�e�٧sho���w�that�an�old�result�is�a�corollary�of�our�main�theorem.���The�original�pro�Gofs�(there�are�t�w�o���indep�Genden���t�y�ones�in�the�literature)�are�m�uc�h�easier�than�the�pro�Gof�of�our�main�theorem,���so�the�fact�that�it���is�<a�corollary�of�our�theorem�is�only�of�in���terest�for�the�connection,�A)and�not�for�the�result�itself.�i`The�original���pro�Gofs�UUare�discussed�in�the�next�section.��x��Theorem.����(Sc���ho�Gen�b�erg��N[1937],��MSc�herrer�[1946])�Supp�Gose�a�regular��n�-gon�is�em�b�Geddable�in���Z���^��f$�cmbx7�k����-�for�some��k�P��.���Then�UU�n��=�3,4,�or�6.����Pr��}'o�of.�p�If�P�w���e�ha�v�e�an�em�b�Gedded��n�-gon,�Q�then�there�is�an�em�b�Gedded�isosceles�triangle�with�one�angle�of�2��[�=n�.���The��Yother�t���w�o��Yangles�are�eac���h���[�=�2�������=n�.���Their��Ytangen���ts�are�th�us��cot���X��[�=n�.���The�non-em�b�Geddabilit�y�of�an����n�-gon�UUin�an���y���Z���^��k����;�will�then�follo�w�from�our�theorem�when��cot�������RH�2��'�(��[�=n�)�is�irrational.�q�Since��=�����!cot����x����2���$���5�=�����<$���K1�8�+��cos��G2����K�w�fe*?��	(֍�1�8����cos��G2�����������w���e�UUha�v�e���䍍���!cos��жX2��5�=�����<$����Kcot���O���RH�2��v������8�1���K�w�fe)�5�	c���cot���
UW��RH�2��|r����+�8�1��������so��W�cos��e�2���̫is�W�rational�if�and�only�if��cot���^��2��|s���is�rational.�x�Hence�an�em���b�Gedded��n�-gon�is�p�ossible�if�and�only�if����cos��
c�(2��[�=n�)��Bis�rational.���T��*�o�complete�the�pro�Gof,��=w���e�ha�v�e�to�sho�w�that��cos����(2��[�=n�)�is�rational�exactly�when����n���=�3,�UU4,�or�6.��}��Let�<*����=�Gбe���^��2��@Li=n��H��.�&HThen�the�minimal�p�Golynomial�of����
�has�degree���(�n�),�u�where����is�the�Euler���-function.���(See��
for�example�Borevic���h�and�Shafarevic�h�[1966],���p.��326.)�Since��
2����cos��7(2��[�=n�)�#�=����+�^1�=���,�w���e��
ha�v�e��f���(����)�#�=�0���for�v�a�quadratic�p�Golynomial��f��T�with�co�ecien���ts�in��Q�(�cos��
c�(2��[�=n�)).��Hence�the�degree�of�the�eld�extension���[�Q�(����)�͋:��Q�(�cos��
c�(2��[�=n�)]���is�at�most�2.�J3On�the�other�hand�it�is�at�least�2,�,since��cos��V](2��=n�)�is�real.�J3Hence�the���degree�A�of��cos���3(2�pi=n�)�o���v�er�A�the�rationals�is���(�n�)�=�2.�k7This�can�b�Ge�one�if�and�only���(�n�)��=�2,�E�that�A�is,��n���=�3,�E�4,�or���6.�
���s���History�and�Related�W����ork�����The��Zrst�pro�Gof�that�the�equilateral�triangle�is�not�em���b�eddable�in���Z���^��2���Y��w���as�giv�en�(so�far�as�I��Fkno�w)�b�y�E.���Lucas��.[1878].��SLucas'�pro�Gof�is�p�erhaps�more�accessible�in�P����oly���a�and�Szeg�o�[1954],��dpage�376�(problem�238).���Since�UUit�is�only�a�few�lines,�and�not�published�elsewhere�in�English,�it�seems�w���orth�reprin�ting:��}��Put��one�corner�of�the�h���yp�Gothetical�equilateral�triangle�at�origin,��the�other�corners�at�(�a;���b�)�and�(�x;�y�[٫),��and���supp�Gosing�UUthat��x;���y�[�;�a;�b�UU�ha���v�e�no�common�factor.�q�Then�w�e�ha�v�e��x���2Ʊx�����2���S�+�8�y��[ٟ���2���d�=���a�����2���+��b�����2��C��=��(�x����a�)�����2���+�(�y������b�)�����2�����and�UUhence��x���K�2(�xa�8�+��by�[٫)��=��x�����2���S�+�8�y������2���d�=��a�����2���S�+�8�b�����2���2ɍ���H�x�����2���S�+�8�y��[ٟ���2��,�+��x�����2���+��b�����2��C��=��4(�xa��+��y�[�b�)����T͍������+3����=�����
UN0��	��mo�Gd�� �4��x�Since���w���e�ha�v�e�excluded�the�case�of��x;���y�[�;�a;�b���all�divisible�b�y�2,��Othey�m�ust�all�b�Ge�o�dd.��In�that�case,��Oho���w�ev�er,���the�UUequation������D�x�����2���S�+�8�y��[ٟ���2���d�=��(�x����a�)�����2���+�(�y������b�)�����2���|c�mo�Gd��$�b4����~]������8�MICHAEL���J.�BEESON�DEP��J�AR�TMENT���OF�MA��J�THEMA�TICS���AND�COMPUTER�SCIENCE�SAN�JOSE�ST��J�A�TE���UNIVERSITY�SAN�JOSE,�CA�95192�BEESON@UCSCC.UCSC.EDU����n����썫is�UUimp�Gossible,�completing�the�pro�of.����So�Q;far�as�I�Q:can�determine,�R
John�McCarth���y�w�as�the�rst�to�state�and�pro�v�e�(although�he�did�not�publish)���the���generalization�of�Lucas'�theorem�to�planar�p�Golygons�(Prop�osition�1�of�this�pap�er).��#One�of�the�referees���suggested�cLthat�this�theorem�w���as�part�of�the�\folklore"�of�the�sub��8ject,���and�should�not�b�Ge�credited�to�McCarth�y;���but�5�Lucas'�pro�Gof�is�v���ery�sp�ecial,�<and�when�P����oly���a�and�Szeg�o�giv���e�it,�<as�recen�tly�as�1954,�<there�is�no�hin�t�of�a���generalization,�y�nor�rMis�this�generalization�men���tioned�in�an�y�of�the�other�related�pap�Gers�discussed�b�elo���w,�y�and���these�UUare�all�the�pap�Gers�I�could�nd�on�the�sub��8ject.����Rather�İthan�ask�ab�Gout�arbitrary�planar�triangles,��p�eople�seemed�to�ha���v�e�İgeneralized�Lucas'�theorem�in���another�UUdirection,�asking�ab�Gout�arbitrary�regular�p�olygons.����Sc���ho�Gen�b�erg�+I[1937]�pro���v�ed�+Ithat�a�regular��n�-gon�with��n��dieren���t�from�3,4,�3�and�6,�is�not�em���b�Geddable�in���Z���^��k���yګ,���or��indeed�an���y�(p�Gossibly�oblique)�rational�lattice�in��k�P��-space�for�an�y��k�P��.���Although�it�refers�to��k�`E�dimensions���instead�Aof�a�plane,�*it�actually�suces�to�consider�only�planar�lattices,�since�if�a�p�Golygon�w���ere�em�b�Geddable�in�����Z���^��k���N��,�A�then�=the�in���tersection�of�the�plane�of�the�p�Golygon�with��Z�����^��k��߷�w�ould�b�Ge�a�planar�lattice.�i�Sc�ho�Gen�b�erg's�=pro�of���is���short:�:,Let��A�,�B��q�,and��C�p��b�Ge�three�consecutiv���e�v�ertices�of�a�regular�lattice��n�-gon�with�cen�ter�at�origin.��_Let����P��;�=�_��A�u�+��C���.��nThen���j�P�c��j��=�2�j�B��q�j�����cos��7(2��[�=n�),���so��cos���q��^��2����(2��=n�)�is�rational.��nThen�w���e�can�nish�the�pro�Gof�as�in�the���previous��Dsection,��except�that�the�cases��n��P�=�8��Dand��n��P�=�12��Dstill�need�atten���tion.�'�(Sc�ho�Gen�b�erg��D[1937],��p.�'�50,���jumps�UUto�Go�quic���kly�for�me�to�follo�w�to�the�conclusion�that��cos����(2��[�=n�)�is�rational.)����Sc���herrer�u�[1946],�}�apparen�tly�una�w�are�of�Sc�ho�Gen�b�erg�u�[1937],�}�ga�v�e�another�pro�Gof�of�this�theorem.���His�pro�of���is�,Za�gem:�]JSupp�Gose�w���e�had�an�em�b�Gedded��n�-gon�(for��n��>��6).�dConsider�,Zthe�lattice�v�ectors�formed�b�y�the�sides.���T��*�ranslate���them,�ѩputting�their�tails�all�at�origin.�:�Then�their�heads�form�a��smaller��lattice��n�-gon,�in�fact�smaller���b���y��pat�least�a�certain�factor,���namely�2����sin��
��(��[�=n�).�5Iterating�this�construction�leads�to�arbitrarily�small�lattice����n�-gons,�-\a�#^con���tradiction.�a This�pro�Gof�w�orks�ev�en�for�non-square�lattices,�-\whic�h�w�e�ha�v�e�not�considered�in�this���pap�Ger.��Sc���herrer�~�also�sho�w�ed�the�case��n�{�=�5�~�is�imp�Gossible,��_b�y�a�similar�construction:��Num�b�Ger�the�sides�of�a���p�Gen���tagon,�55considered�--as�v�ectors,�55b�y�1,2,3,4,5.�deThen�taking�them�in�the�order�1,3,5,2,4,�55place�the�tail�of�eac�h���at��nthe�head�of�the�previous�one.�Y��*�ou�will�get�a�v���e-p�Goin�ted��nstar.�Connecting�the�p�Goin���ts,��4y�ou��nget�a�smaller���lattice�oMp�Gen���tagon�than�y�ou�started�with.���F��*�or�square�lattices,�u�Sc�herrer�could�ha�v�e�ruled�out��n��_�=�3�oMand��n��_�=�6���b���y�UULucas'�theorem.����The��?main�p�Goin���t�of�Sc�ho�Gen�b�erg��?[1937]�is�not�p�Golygons,��ybut�rather�necessary�and�sucien�t�conditions�for���the�em���b�Geddabilit�y�of�a�regular��n�-simplex�in���Z���^��n����t�(it�is�alw�a�ys�em�b�Geddable�in���Z���^��n�+�1���K��,�ARfor�example�taking�all���the��p�Goin���ts�with�one�co�ordinate�1�and�the�rest�0).���Although�the�equilateral�triangle�is�not�em���b�eddable�in�����Z���^��2����,���the��#tetrahedron�is�em���b�Geddable�in�the�unit�cub�e,���for�example�at�(1�;����0�;��0),�(0�;��1�;��0),�(0�;��0�;��1),�and��#(1�;��1�;��1).���Sc���ho�Gen�b�erg���sho�w�ed�that,��for��n��ev�en,��the�em�b�Gedding�is�p�ossible�if�and�only�if��n�y«+�1�is�a�p�Gerfect�square;��Sfor����n����T͍�������+3�����=�����
UN3����mo�Gd��*�4,�UUit�is�alw���a�ys�UUp�Gossible�and�for��n����T͍�������+3�����=�����1����mo�Gd��*�4,�if�and�only�if��n�8�+�1�UUis�a�sum�of�t���w�o�UUsquares.����The��fact�that�the�4-simplex�is�not�em���b�Geddable�in���Z���^��4�����refutes�the�idea�that�p�erhaps�a�p�olyhedron�is���em���b�Geddable�UUif�all�of�its�triangles�are�em�b�Geddable.����Nob�Go�dy��!seems�to�ha���v�e��!considered�the�question�of�the�em���b�Geddabilit�y��!of�arbitrary�triangles�un���til�the�1980's.���Landau��xand�Cremona�[1987]�consider�the�follo���wing�question:�"Ygiv�en�that�a�triangle�is�em�b�Geddable�in��Z���^��n��q~�,��>what���is�m=the�smallest�em���b�Gedding?�$jThat�is,���nd�the�smallest�triangle�similar�to�the�giv�en�one�whic�h�has�its�v�ertices�on���lattice��zp�Goin���ts�in��n�-space.�BThey�answ�er�the�question�in�dimensions�3�and�4�using�the�greatest-common-divisor���algorithm��in�the�quaternions.�Z�Since�no���w�w�e�kno�w�that�triangles�em�b�Geddable�in��Z���^��4��x�are�also�em�b�Geddable�in����Z���^��3��|s�,�E�w���e�A�migh�t�w�onder�if�a�smallest�em�b�Gedding�can�alw�a�ys�b�Ge�found�in��Z���^��3��|s�.�kDThe�answ�er�(according�to�a�letter���from�O�Landau)�is�no:�n�although�a�lattice�triangle�in��Z���^��4����can�alw���a�ys�O�b�Ge�rotated�and�dilated�in���to��Z���^��3��|s�,�P�sometimes���a�UUdilation�is�really�required.��4ꍑ�P���ostscript�on�the�Disapp�UVearing�Computer�����All��the�pro�Gofs�ab�o���v�e��use�only�elemen���tary�n�um�b�Ger�theory��*�.�ZSThis�is�in�teresting,�considering�that�a�computer��A�w���as�Oin�v�olv�ed�throughout�this�researc�h.�x�First�I�#used�it�to�disco�v�er�that�the�isosceles�triangle�of�heigh�t������P�p���W����P�fe�E���7������and���base�2�is�not�em���b�Geddable�in��Z���^��3��|s�;�N�then�that�the�same�triangle�is�not�em�b�Geddable�in��Z���^��4��|s�;�N�then�to�settle���the��tquestion�of�non-solv��q�abilit���y�of��E����(4�;���k�P��)�if��k��is�not�a�sum�of�four�squares.��&A�t�rst�I��Yexp�Gected�to�use�it�to���nd��{an�example�of�a�triangle�em���b�Geddable�in��Z���^��4��#�but�not�in��Z���^��3��|s�.�h9Only�after�using�it�to�nd�actual�solutions���of�I�E����(3�;���k�P��)�for��k����a�sum�of�three�squares�up�to�128�did�I�I�giv���e�up�m�y�preconceptions�and�try�to�pro�v�e�that�the���same��Ptriangles�are�em���b�Geddable�in�three-space�as�in�four-space.�ָWhen�I��1learned�ho�w�to�use�quaternions�to���describ�Ge��zorthogonal�transformations�of�four-space,�̈́all�m���y�programs�w�ere�displaced�b�y�the�concise�elegance���of�UU\real�mathematics".����	�썟���z�?�TRIANGLES���WITH�VER��J�TICES�ON�LA�TTICE�POINTS��z-9�����n����썒��b�References�����Borevic���h,�UUZ.�I.,�and�Shafarevic�h,�I.�R.,��Num�b�Ger�Theory�,�Academic�Press,�New�Y��*�ork{London�(1966).����Erd����os,�UUP��*�.,�Grub�Ger,�P�.�M.,�and�Hammer,�J.,��Lattice�P���oin�ts�,�UUWiley�,�New�Y�ork�(1989).����Hardy��*�,�UUG.�H.�and�W�righ���t,�E.�M.,��The�Theory�of�Num�b�Gers�,�Clarendon�Press,�Oxford�(1968).����Landau,�F�S.,�and�Cremona,�J.,�Shrinking�Lattice�P���olyhedra,�T��*�ec�hnical�Rep�Gort�87-05,�W��*�esley�an�Univ�ersit�y���Departmen���t�UUof�Mathematics,�Middleto�wn,�CT�06457�(1987).����LeV��*�eque,�UUW.�J.��The�Elemen���tary�Theory�of�Num�b�Gers�,�Addison-W��*�esley�,�UUReading,�Massac�h�usetts�(1962).����Lucas,�UUE.�Th�����Geor��eme�UUsur�la�G����eom��etrie�UUdes�Quinconces,��Bull�So�Gc.�Math.�F��*�rance��6��(1878)�9-10.����P���atruno,�UUG.�N.,�The�lattice�p�Golytop�e�UUproblem,��Elemen�te�der�Mathematik��38��(1983)�69-71.����P����oly���a,� �G.,�and�Szeg�o,�G.,��Aufgab�Gen�und�Lehrs�atze�aus�der�Analysis,�Band�I�GI�,�Springer-V��*�erlag,�Berlin-���G����ottingen-Heidelb�Gerg�UU(1954).����Sc���herrer,�"rW.,�Die�Einlagerung�eines�regul����aren�Vielec�ks�in�ein�Gitter,��Elemen�te�der�Mathematik�D��1��(1946)���97-98.����Sc���ho�Gen�b�erg,�UUI.�J.,�Regular�simplices�and�quadratic�forms,��J.�London�Math.�So�c.����12��(1937)�48-55.���������;��n�G
	�?p�0J
cmsl10�9����

msam10�+�-�

cmcsc10�$�':

cmti10�!p�0J

cmsl10���<x

cmtt10�f$�cmbx7��"V

cmbx10�O!�cmsy7��K�cmsy8�!",�

cmsy10�
;�cmmi6�	0e�rcmmi7��2cmmi8��b>

cmmi10��Aa�cmr6�ٓ�Rcmr7�|{Ycmr8�K�`y

cmr10��a������

Sindbad File Manager Version 1.0, Coded By Sindbad EG ~ The Terrorists