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/ConstructiveGeometry.dvi

����;� TeX output 2009.05.26:0833������header=pstricks.pro�header=pst-dots.pro�s���������G�color push gray 0�����	color pop���[(���G�����color push gray 0�4��������color push gray 0�	color pop�������4src:195 ConstructiveGeometryFinalPreprintVersion.tex�D��tG�G�cmr17�Constructiv��qe�7tGeometry����������X�Qcmr12�Mic��rhael��Beeson������fj���`�Ma��ry��26,�2009��$�����p��color push gray 0�	color pop����p��4src:202 ConstructiveGeometryFinalPreprintVersion.tex�t�:		cmbx9�Abstract���э���
��color push gray 0��	color pop���&��o���		cmr9�Euclidean���geometry��:�,��as�presen��9ted�b�y���Euclid,��consists�of�straigh��9tedge-and-compass�construc-����tions��and�rigorous�reasoning�ab�A�out�the�results�of�those�constructions.���A���consideration�of�the����relation�i�of�i�the�Euclidean�\constructions"�to�\constructiv��9e�mathematics"�leads�to�the�dev��9elop-����men��9t�[�of�[�a�rst-order�theory��ECG��of�the�\Euclidean�Constructiv��9e�Geometry",���whic�h�can�[�serv�e�as����an�V�axiomatization�V�of�Euclid�rather�close�in�spirit�to�the���j��		cmti9�Elements��of�Euclid.����ECG��is�axiomatized����in�e�a�e�quan��9tier-free,�y�disjunction-free�w�a�y��:�.�
+Unlik�e�e�previous�in�tuitionistic�e�geometries,�y�it�do�A�es�e�not����ha��9v�e���apartness.��>Unlik�e�previous�algebraic���theories�of�geometric�constructions,���it�do�A�es�not�ha��9v�e�a����test-for-equalit��9y��construction.���W��:�e�sho�w��~that��ECG��is�a�go�A�o�d��geometric�theory��:�,���in�the�sense�that����with�uqclassical�logic�upit�is�equiv��|ralen��9t�to�textb�A�o�ok�uqtheories,��xand�its�mo�A�dels�are�(in��9tuitionistically)����planes�e4o��9v�er�Euclidean�elds.�W��:�e�then�e5apply�the�metho�A�ds�of�mo�dern�metamathematics�to�this����theory��:�,��sho��9wing��lthat��kif��ECG��pro�v�es�an�existen�tial��ktheorem,��then�the�ob���ject�pro�v�ed��kto�exist�can����b�A�e��Wconstructed�from�parameters,�Řusing�the�basic�constructions��Xof��ECG��(whic��9h�corresp�ond�to����the���Euclidean���straigh��9tedge-and-compass�constructions).���In�particular,��aob���jects�pro�v�ed���to�exist����in�l��ECG��dep�A�end�l�con��9tin�uously�l�on�parameters.�#fW��:�e�also�study�the�formal�relationships�b�A�et��9w�een����sev��9eral�E�v�ersions�of�Euclid's�parallel�p�A�ostulate,���and�sho�w�E�that�eac�h�corresp�A�onds�to�a�natural����axiom�Tsystem�for�Euclidean�elds.����-=��Aa�cmr6�1������������N�ffcmbx12�1��1L�In���tro�s3duction��阍��4src:217 ConstructiveGeometryFinalPreprintVersion.tex�Euclid's��geometry��:�,�T�written�do��9wn��ab�A�out�300�BCE,�has�b�A�een�extraordinarily�in
uen��9tial�in�the����dev��9elopmen�t�zof�mathematics,��)and�prior�to�the�t��9w�en�tieth�zcen�tury�zw�as�regarded�as�a�paradigmatic����example��:of��9pure�reasoning.��!But�during�those�2300�y��9ears,� �most�p�A�eople�though�t��9that�Euclid's����theory�%�w��9as�ab�A�out�something.�NWhat�w�as�it�ab�A�out?�NSome�ma�y�ha�v�e�answ�ered�that�it�w�as�ab�A�out����p�A�oin��9ts,��lines,��and�_�planes,�and�their�relationships.��$Others�ma��9y�ha�v�e�_�said�that�it�w�as�ab�A�out����metho�A�ds�Tefor�constructing�Tfp�oin��9ts,�d)lines,�d*and�planes�with�certain�sp�ecied�Tfrelationships�to�giv��9en����p�A�oin��9ts,�lines,�and��planes,�for�example,�constructing��an�equilateral�triangle�with�a�giv��9en�side.����In�&these�'t��9w�o�answ�ers,��e�see�the�'viewp�A�oin�ts�of�pure�'(classical)�mathematics�and�of�algorithmic����mathematics�Qrepresen��9ted.���Hilb�A�ert's�1899�rew�orking�of�the�theory�Q [�11��	?�]�ga�v�e�another�answ�er,����surprising�;at�the�time:�?Euclid's�theories�w��9ere��not�6�ab��out�anything�at�6�al�x�l�.�&Instead�;of�\p�A�oin�ts,����lines,���and��Uplanes",���one�should�b�A�e�able��Tto�read�\tables,�c��9hairs,���and�b�A�eer�m�ugs."��pAll�the�reasoning����should�:�still�b�A�e�v��|ralid.�ӠThe�names�of�the�\en��9tities"�w�ere�just�place�holders.�ӠThat�w�as�the�viewp�A�oin�t����of�Tt��9w�en�tieth-cen�tury�axiomatics.����&��4src:229 ConstructiveGeometryFinalPreprintVersion.texIn��vthe�late�t��9w�en�tieth��vcen�tury��:�,�$�con�temp�A�oraneously�with�the�
o�w�ering�of�computer�science,����there��w��9as�a�new��surge�of�vigor�in�algorithmic,�&9or�constructiv�e,�&8mathematics,�&9b�A�eginning�with����Bishop's�z�b�A�o�ok�z�[�4����].�L�In�algorithmic�mathematics,��one�tries�to�reduce�ev��9ery�\existence�theorem"����to��Pan��Qassertion�that�a�certain�algorithm�has�a�certain�result.��eIn�the�terminology�of�computer��
��color push gray 0��ff�ff�r�	J=�����"5��-:�1����L��|{Ycmr8�W��J�e���w�Îould���lik�e�to���thank�Jerem�y�Avigad���and�F��J�reek�Wiedijk�for�pro�<rductiv�Îe�and�in�Îteresting�discussions,���and�for��	��careful��Xreading�and�useful�suggestions.��ٛ�	color pop����G�color push gray 0���O\�K�`y

cmr10�1������	color pop����*�s���������G�color push gray 0�����	color pop���[(��������+I�science,�|�existence�htheorems�hshould�b�A�ecome�correctness�pro�ofs�of�algorithms.��The�pro�of�theory����+Iof��darithmetic�has�pro��9vided�man�y�b�A�eautiful��etheorems�to�sho�w�that�indeed,��.existence�theorems�in����+In��9um�b�A�er�\�theory�(when�\�constructiv��9ely�pro�v�ed)�con�tain�algorithms�that�can�\�b�A�e�\extracted"�from����+Ithe��pro�A�ofs.�In�particular�w��9e�men�tion�the�tec�hniques��of�recursiv�e�realizabilit�y��:�,�TZthe�Dialectica����+Iin��9terpretation���of��G��`odel,�2�and�the�extraction�of�algorithms�from�cut-free�pro�A�ofs�as�w��9ell-kno�wn����+Iexamples�Tof�the�phenomenon.����8�C�4src:237 ConstructiveGeometryFinalPreprintVersion.texIn�(�this�pap�A�er,�-Mw��9e�re-examine�(�Euclidean�geometry�from�the�viewp�oin��9t�(�of�constructiv�e�math-����+Iematics.��The��phrase�\constructiv��9e�geometry"�suggests,��/on�the�one�hand,�that�\constructiv��9e"����+Irefers�~;to�geometrical�constructions�~:with�straigh��9tedge�and�compass.��On�the�other�hand,��sthe�w�ord����+I\constructiv��9e"��~ma�y�suggest�the�use�of�in�tuitionistic��}logic.���W��:�e�in�v�estigate�the�connections�b�A�e-����+It��9w�een��these��t�w�o�meanings��of�the�w�ord.���Our��metho�A�d�is�to�pro�A�ceed�b��9y�analogy�with�the�extensiv��9e����+Ib�A�o�dy���of���w��9ork�that�has�b�A�een�done�on�n��9um�b�A�er�theory���and�analysis,���applying�the�relev��|ran��9t�metho�A�d-����+Iologies��Gto��Hthe�w��9eak�er��Gtheories�of�geometry��:�.��JThe�basis�for�the�w��9ork�describ�A�ed�here�is�the�idea����+Ithat�D�in�geometry��:�,�P�w��9e�can�tak�e�\algorithm"�in�the�restricted�sense�of�\geometric�construction."����+IThat�Tis,�w��9e�pursue�the�analogy����8�C�4src:245 ConstructiveGeometryFinalPreprintVersion.tex��
������-ꍒ��formal�Tn��9um�b�A�er�theory��w���H�fewϞ�'m��TT��:�uring�computable�functions������Q�=�����-ꍑ�pin��9tuitionistic�Tgeometry�����H�fe_��'m�geometric�Tconstructions������
��
\���8�C�4src:248 ConstructiveGeometryFinalPreprintVersion.texT��:�o�8�carry�8�out�this�program,��mw��9e�need�a�suitable�formal�theory�for�in��9tuitionistic�geometry��:�.����+IIt��qshould�b�A�e�a�theory�with�terms�for�the��rgeometric�constructions,���so�that�there�will�b�e�terms����+Ia��9v��|railable���to�denote�the�means���of�constructing�things�that�ha�v�e�b�A�een�pro�v�ed�to���exist.�
�This�leads����+Ius���to�lo�A�ok�for�a�quan��9tier-free�axiomatization.�+5In�form�ulating�a�suitable�theory�of�geometry��:�,����+Ianother���imp�A�ortan��9t�consideration���w�as�the�principle�that�\constructiv�e�pro�A�of���implies�con�tin�uit�y�in����+Iparameters".���This�tprinciple�is�familiar�to�those�who�ha��9v�e�tstudied�constructiv��9e�mathematics;��but����+Iit��is�easy�to�understand�on�an�in��9tuitiv�e��basis.��If�w��9e�wish�to�allo�w�an�in�terpretation�of�geometry����+Iin�%swhic��9h�%rp�A�oin�ts�are�giv�en�%rb�y�appro�ximations�(as�%raccurate�as�one�ma��9y�demand),�)zfor�example,�if����+Ithey��Aare�giv��9en�as�pairs�of�real�n�um�b�A�ers�(�5��"		cmmi9�x;���y�R��),���and�w�e�(constructiv�ely)�pro�v�e������		cmsy9�8�x�9�y�<�A�(�x;���y�R��),���then����+Iw��9e���ough�t���to�b�A�e�able�to�pro�A�duce�appro��9ximations�to��y��y�when�w��9e�are�giv��9en�appro�ximations���to��x�.����+IThis�videa�v
led�us�to�lo�A�ok�at�the�con��9tin�uit�y�vof�the�Euclidean�constructions.�>�W��:�e�lo�A�ok��9ed�for�(and����+Ifound)�B�an�B�axiom�system��ECG��with�in��9tuitionistic�logic,�Nsuc�h�that�B�when�an�existen��9tial�theorem����+Iis��jpro��9v�ed��iin��ECG�,�the�ob���ject�asserted�to�exist�can�b�A�e�constructed�b��9y�straigh�tedge��iand�compass,����+Icon��9tin�uously�Tin�parameters.����8�C�4src:259 ConstructiveGeometryFinalPreprintVersion.texDecision���functions,��[suc��9h�as�whether�t�w�o�p�A�oin�ts�are�equal���or�whether�a�giv�en�p�A�oin�t�lies�on����+Ia�ulgiv��9en�ukline,��rare�discon�tin�uous.�<�P�ast�w�ork�ulon�the�ukalgebraic�approac�h�ukto�constructions�(see�for����+Iexample���[�15��	?�])�has�alw��9a�ys���assumed�decision�functions,���probably�b�A�ecause�they�seem�to�b�e�needed����+Ito��dene��pro���jection�of�a�p�A�oin��9t�on�a�line,���whic��9h�is�needed�in�order�to�in��9tro�A�duce�co�ordinates��in����+Ithe�h�Euclidean�h�plane�and�connect�geometry�to�eld�theory��:�.��But�pro���jection�itself�is�con��9tin�uous.����+IIn�;�an�early�v��9ersion�of�;�our�theory��ECG�,�w�e�to�A�ok�it�as�a�fundamen�tal�op�A�eration.���In�the�v�ersion����+Ipresen��9ted�~?here,��{that�op�A�eration�is�~@not�necessary��:�,��zb�ecause�~@pro���jection�can�b�e�~@dened�in�terms�of����+Ithe�Top�A�erations�of��ECG�,�and�its�essen��9tial�prop�erties�pro��9v�ed�Tusing�the�axioms�of��ECG�.����8�C�4src:265 ConstructiveGeometryFinalPreprintVersion.texOur��0in��9terest�in�this�sub���ject�b�A�egan�with�a�computer�animation�of�Euclid's�constructions�that����+Ip�A�ermits�L7the�user�L8to�drag�the�starting�p�oin��9ts,���and�see�ho�w�the�construction�L8dep�A�ends�on�the����+Ic��9hanged��Tstarting�p�A�oin�ts.��pThe�results��Ufor�Euclid's�Bo�A�ok�I,�Prop�osition�2,�Tw��9ere�surprising�and����+Iin��9teresting.��aIn���Bo�A�ok���I,�Prop�osition�2,�1�Euclid���attempted�to�sho��9w�that�a�rigid�compass�could����+Ib�A�e���sim��9ulated�b�y�a�collapsible���compass.�&This�famous�construction�sho�ws�ho�w���to�use�the�rigid����+Icompass�Dto�Econstruct�a�p�A�oin��9t��D�ms�=�+��e�(�A;���B�r�;�C����)�suc��9h�Dthat�whenev�er�D�B��a�6�=�+��C��-�and��A��6�=�+��B�}��then����+I�AD��=��T�B�r�C����.�u�The�3rst�3problem�with�this�construction�is�that�it�do�A�es�not�w��9ork�when��A��T�=��C����.�u�Of����+Icourse,�~~in�ivthat�case�w��9e�can�just�tak�e��D�_��=���B�r��;���but�that�case�distinction�requires�classical�logic.����+IAnd�j]the�computer�j\animation�rev��9eals�that�when�w�e�j\drag�p�A�oin�t��C�E�close�to�j\�A�,��and�then�around����+I�A�,��in�,�a�small�circle,�2�then�the�constructed�p�A�oin��9t��D�n}�mo�v�es�around�,��A��in�a�circle�of�radius�close�to����+I�B�r�C����.��.Hence�K>�D���do�A�es�not�dep�end�con��9tin�uously�K>on��C����.��.This�discon��9tin�uit�y��:�,�X�together�K>with�the�need����+Ifor�T�a�case�distinction�just�T�men��9tioned,�dymak�es�T�it�clear�that�from�the�in��9tuitionistic�p�A�oin�t�of�view,�����G�color push gray 0���O\�2������	color pop����7�s���������G�color push gray 0�����	color pop���[(��������+I�there�ware�t��9w�o�wdieren�t�wv�ersions�of�Euclid�I.2:��the�v�ersion�in�whic�h�(in�waddition�to��B��q�6�=�4��C����)�w�e����+Itak��9e�V��A��7�6�=��6�C����as�a�h�yp�A�othesis,�gior�the�\uniform"�v�ersion�V�in�whic�h�w�e�do�not�assume��A��7�6�=��C����,�ghbut����+Iassert���that��D�Ѩ�can�b�A�e�constructed,���whatev��9er�the�v��|ralues�of��A�,��B�r��,�and����C����,�as�long�as��B�{�6�=����C����.���Whic��9h����+Iv��9ersion��lof��mEuclid�I.2�is�pro��9v��|rable�with�in��9tuitionistic�logic�turns�out�to�dep�A�end�on�whether�w��9e�tak�e����+I(a�Tformalism�corresp�A�onding�to)�a�rigid�compass,�or�Euclid's�collapsible�compass.����8�C�4src:283 ConstructiveGeometryFinalPreprintVersion.texP��9ast��lw�ork�on�axiomatizations�of�constructiv�e�(in�tuitionistic)�geometry��:�,��suc�h�[�10��	?�]�and�[�17��],����+Ihas�ޣreplaced�the�la��9w�ޢof�the�excluded�middle�b�y�\apartness"�(whic�h�is�ޢexplained�in�the�b�A�o�dy����+Iof��xthe��wpap�A�er).�f�One�can�in��9tro�A�duce�an�apartness�\construction",��but�it�is�also�not�con��9tin�uous.����-=�2������+I�Therefore�Tw��9e�w�ork�in�theories�without�apartness.����8�C�4src:295 ConstructiveGeometryFinalPreprintVersion.texW��:�e��are��able�to�obtain�the�main�metatheorem�w��9e�w�an�t�using��cut-elimination;�)no�new��pro�A�of-����+Itheoretic�Ħtec��9hniques�ħare�dev�elop�A�ed�here,��{so�this�ħpap�er�should�b�e�ħaccessible�to�geometers,��{not����+Ionly�Tto�logicians.����8�C�4src:298 ConstructiveGeometryFinalPreprintVersion.texGeometry���is���an�ancien��9t�sub���ject,��but�it�is�v��9ery�m�uc�h�aliv�e.�	�There���are�a�n��9um�b�A�er���of�a�v�en�ues���of����+Icurren��9t��]researc�h�that�are��^related�to�this�w�ork,��[but��^not�dealt�with�in�this�pap�A�er;��and�rather�than����+Iundertak��9e���to���surv�ey�them,��Yw�e�prefer���to�stic�k���to�the�sp�A�ecic�aims�outlined�ab�A�o��9v�e.��Similarly��:�,��Zthere����+Iis�[�a�long�history�[�of�axiomatizations�of�geometry��:�.���F�or�a�[�review�of�some�of�these,�mpsee�[�19��	?�].���Here����+Iw��9e�j�explain�j�only�ho�w�our�axiomatization�j�is�dieren�t�from�others:�ǝ(i)�it�uses�j�constructiv�e�logic,����+Irather��than�classical,��owhic��9h�distinguishes��it�from�all�previous�axiomatizations�except�Heyting's,����+Iand�D�(ii)�it�D�do�A�es�not�use�apartness,�PKwhic��9h�enables�it�to�ha��9v�e�D�the�prop�A�ert��9y�that�p�A�oin��9ts�pro�v�ed�to����+Iexist�Tcan�b�A�e�constructed�b��9y�Euclid's�(con�tin�uous)�constructions.�������+I�2��C^Euclid's�ffconstructions�as�algorithms��阍�+I�4src:307 ConstructiveGeometryFinalPreprintVersion.tex�Euclid's�8v��9e�b�A�o�oks�8presen�t�48�t�w�o-dimensional�constructions�and�ab�A�out�a�dozen�three-dimensional����+Iconstructions.�h�(F��:�or��psimplicit��9y�,�
xw�e��pwill��qnot�discuss�the�three-dimensional�part�of�Euclid,�
wcon-����+Itained�2mostly�in�2Bo�A�ok�5).�r�W��:�e�consider�the�48�t��9w�o-dimensional�2constructions�to�b�A�e�the�w��9orld's����+Irst���systematic���collection�of�algorithms.���(W��:�e�do�not�sa��9y��:�,�͝\the�w�orld's���rst�algorithms",�͝b�A�ecause����+Ithere���certainly���w��9ere�a�few�n��9um�b�A�er-theoretic�algorithms���kno�wn�in���China�and�India�m��9uc�h���earlier.)����8�C�4src:313 ConstructiveGeometryFinalPreprintVersion.texIf��sthe�constructions��rare�considered�as�algorithms,���then�Euclid's��Elements��con��9tained�the�rst����+Ipro�A�ofs�Tof�correctness�of�algorithms.����8�C�4src:316 ConstructiveGeometryFinalPreprintVersion.texEuclid�C�presen��9ts�his�readers�with�b�A�oth�\p�ostulates"�and�\axioms".��SMo�dern�mathematicians����+Ioften��treat��these�w��9ords�as�synon�yms.�L�F��:�or�Euclid��and�his�con�temp�A�oraries,���ho�w�ev�er,���they�had����+Iquite�Tdieren��9t�meanings.�pHere�is�the�dierence,�as�explained�b�y�P�am�buccian�[�20��	?�],�p.�p12.��켍���+I�color push gray 0��	color pop���AI�4src:322 ConstructiveGeometryFinalPreprintVersion.texF��:�or��Pro�A�clus,�V�who�relates�a�view�held�b��9y�Gemin�us,�V�a�p�A�ostulate�prescrib�es�that�w��9e����AIconstruct�;�or�pro��9vide�some�simple�or�easily�;�grasp�A�ed�ob���ject�for�the�exhibition�of�a����AIc��9haracter,��nwhile�kan�axiom�kasserts�some�inheren��9t�attribute�that�is�kno��9wn�at�once�to����AIone's�L^auditors.���And�just�L_as�a�problem�diers�from�a�theorem,�Z!so�a�p�A�ostulate�diers����AIfrom�[6an�axiom,�l�ev��9en�though�b�A�oth�of�them�are�undemonstrated;�~'the�one�is�assumed����AIb�A�ecause�f�it�f�is�easy�to�construct,���the�other�accepted�b�A�ecause�it�is�easy�to�kno��9w.��NThat�is,����AIp�A�ostulates��ask�for�the�pro�duction,���the��p��o�esis���of�something�not�y��9et�giv�en����:���:�:��
�Y�,���whereas����AIaxioms�_refer�`to�the��gnosis��of�a�giv��9en,�N+to�insigh�t�`in�to�the�`v��|ralidit�y�of�`certain�relationships����AIthat�Thold�b�A�et��9w�een�Tgiv�en�notions.��콍�8�C�4src:337 ConstructiveGeometryFinalPreprintVersion.texEuclid's�@�famous�\parallel�p�A�ostulate"�states�that�if�t��9w�o�@�lines�@��L��and��M�9��are�tra��9v�ersed�@�b�y�another����+Iline��P�T�H��,��Rforming��Qadjacen��9t�in�terior��Qangles�on�one�side�of��T��4�adding�up�to�less�than�t��9w�o�righ�t��Pangles,����+Ithen����L��and����M����will�in��9tersect�on�that�side�of��T�H��.���Stated�this�w��9a�y��:�,��[the���p�A�ostulate�can�b�e���view��9ed�as�a��G�
���color push gray 0��ff�ff�r�	J=�����"5��-:�2����LܻEv�Îen���though���apartness�is�in�tuitionistically���acceptable,��an�apartness���\constructor"�m�ust���mak�e�use�of���the�idea��	��that�Hp�<roin�Îts�are�Inot�giv�en�\all�at�once"�but�Ib�y�a�sequence�of�appro�ximations.�V�The�same�p�<roin�t�can�Ib�<re�giv�en�b�y���dieren�Ît�h�sequences,�~[and�though�an�h�apartness�constructor�can�b�<re�con�Îtin�uous�h�in�the�appro�Îximating�sequences,�~[it�cannot���b�<re���con�Îtin�uous�in���the�geometric�top�<rology�on�the�p�oin�Îts.�F�Hence�an�apartness�constructor���needs�algorithms�b�ey�Îond���the�V�Euclidean�V�constructions.�KVIn�other�w�Îords,�wgeometry�with�apartness�go�<res�b�ey�Îond�V�Euclid.�KVW��J�e�ha�v�e�V�extended�our���metamathematical��Xw�Îork�to�theories�with�apartness,�but�w�e�ha�v�e�not�presen�ted�those�extensions�in�this�pap�<rer.��ٛ�	color pop����G�color push gray 0���O\�3������	color pop����-��s���������G�color push gray 0�����	color pop���[(��������+I�construction�D3metho�A�d�D4for�pro�ducing�certain�D4triangles.��No��9w�ada�ys,�O�the�D3parallel�p�ostulate�D4is�often����+Istated�qoas�an�axiom:�ԦGiv��9en�a�line��L�,��vand�a�p�A�oin�t��P��R�not�on��L�,��vthere�exists�exactly�one�parallel����+Ito�Q�L��through��P�H��.�p(P��9arallel�lines�are�b�y�denition�lines�that�Rdo�not�meet.)�oW��:�ritten�this�w�a�y��:�,�Rthe����+Iparallel�Tp�A�ostulate�do�es�not�directly�assert�the�existence�of�an��9y�sp�ecic�p�oin��9t.����-=�3����c���+I�3��C^The�ffalgebraic�approac���h�to�constructions��阍�+I�4src:351 ConstructiveGeometryFinalPreprintVersion.tex�The�7geometrical�theory�that�w��9e�shall�ev�en�tually�form�ulate�is�7	quan�tier-free,�c~with�terms�to�denote����+Ithe��geometrical��constructions.��A��mo�A�del�of�suc��9h�a�theory�can�b�A�e�regarded�as�a�man��9y-sorted����+Ialgebra�b�with�b�partial�functions�represen��9ting�the�basic�geometric�constructions.��Sp�A�ecically��:�,�vthe����+Isorts��finclude��eat�least��Point�,����Line�,�and��Cir��cle�,�and�they��fma��9y�p�A�ossibly�include��A�Îr��c�,����R�ay�,����T��J�riangle�,����+Iand�af�Squar��e�.��vW��:�e�ha��9v�e�afconstan�ts�aeand�v��|rariables�of�eac�h�sort.��vThis�collection�of�aedata�t�yp�A�es�is�almost����+Iadequate��to�co��9v�er��the�return�t��9yp�A�es�and��argumen�t�t�yp�A�es�of�the�48�plane�Euclidean�constructions.����+IThe���ob��9vious���exception�is�the�construction�of�a�regular�p�A�en��9tagon�or�hexagon.��More�generally��:�,����+Isome�f"of�the�later�constructions�f!use�the�w��9ord�\gure",�zUwhic�h�apparen�tly�means�something�lik�e����+Iwhat�4a�mo�A�dern�mathematician�w��9ould�call�a�\closed�p�olygon".�xwSome�small�xed�n��9um�b�er�4of����+Isides��xw��9ould��wsuce�for�the�Euclidean�constructions.�f�In�Euclid,��no�gure�with�more�sides�than����+Ian�t0o�A�ctagon�t/is�constructed,���and�no�gure�with�more�than�four�sides�is�an��input��to�another����+Iconstruction,�r�except�,�for�constructions�that�w��9ork�,�on�an�y�\gure."�chThe�,�general�concept�of�a����+Iclosed�^p�A�olygon�^of�an��9y�n�um�b�A�er�^of�sides�ma��9y�b�A�e�logically�problematic�as�it�drags�the�concept�of����+Iin��9teger�Tin�to�geometry��:�.����8�C�4src:375 ConstructiveGeometryFinalPreprintVersion.texOur�Talgebras�include�function�sym��9b�A�ols�for�the�basic�constructors�and�accessors:���'�����G��Line��]��(��Point��dZ�B�r�;�����Point����B��)�������*5�A�T�and��B��	�lie�on�this�line����������b%�p��ointOn1���6��(��Line��`��L�)�������*5The�Tp�A�oin��9ts�from�whic�h��L��w�as�originally�constructed����������`��p��ointOn2���6��(��Line��`��L�)���������B���Cir��cle��]���(��Point��dZ�A;�����Point����B�r��)�������*5�A�T�is�the�cen��9ter,�and�the�circle�passes�through��B����������f���c��enter������(��Cir��cle����C����)���������F�A�p��ointOnCir�cle������(��Cir��cle����C����)�������*5A�Tp�A�oin��9t�on�circle��C����,��������+I�4src:384 ConstructiveGeometryFinalPreprintVersion.texand�Tfor�the�\elemen��9tary�constructions"�(eac�h�of�whic�h�has�t�yp�A�e���Point��y��):����������Interse��ctLines���E�(��Line��`��K� Z;�����Line���r�L�)������������Interse��ctLineCir�cle1������(��Line��`��L;�����Cir��cle�����C����)������������Interse��ctLineCir�cle2������(��Line��`��L;�����Cir��cle�����C����)������������Interse��ctCir�cles1���m_�(��Cir��cle����C� Z;�����Cir��cle�����K����)������������Interse��ctCir�cles2���m_�(��Cir��cle����C� Z;�����Cir��cle�����K����)������+I�4src:392 ConstructiveGeometryFinalPreprintVersion.texEac��9h��of�these��has�sev�eral�\o�v�erloaded"�v��|rarian�ts,�#�whic�h��can�b�A�e�dened�from�these�using�con-����+Istructors�Tand�accessors.�pF��:�or�example,��������[��Interse��ctLines����'�(��Point��dZ�A;�����Point����B�r�;�����Point���C� Z;�����Point���D�A��)�����������=�����Interse��ctLines��<��(��Line��`��(�A;���B�r��)�;���Line���r�(�C� Z;�D�A��))����������[��Interse��ctLineCir�cle1���M��(��Point��dZ�A;�����Point����B�r�;�����Point���C� Z;�����Point���D�A��)�����������=�����Interse��ctLineCir�cle1��U���(��Line��`��(�A;���B�r��)�;���Cir��cle�����(�C� Z;�D�A��))����������[��Interse��ctLineCir�cle1���M��(��Point��dZ�A;�����Point����B�r�;�����Cir��cle�����C����)�����������=�����Interse��ctLineCir�cle1��U���(��Line��`��(�A;���B�r��)�;�C����)����G�	Ҕ�color push gray 0��ff�ff�r�	J=�����"5��-:�3����LܻF��J�or�~Nexample,���it�is�not�~Mimmediately�clear�whether�this�v�Îersion�implies�the�rst�v�Îersion�using�only�in�Îtuitionistic��	��(constructiv�Îe)��Xlogic,�although�of�course�it�do�<res�in�classical�logic.�� This�question�is�settled�(in�the�negativ�e)�in�[�3��@].��ٛ�	color pop����G�color push gray 0���O\�4������	color pop����G��s���������G�color push gray 0�����	color pop���[(��������+I�4src:402 ConstructiveGeometryFinalPreprintVersion.tex�As�X�these�X�three�examples�illustrate,�~Vone�can�regard�circles�and�lines�as�mere�in��9termediaries;���p�A�oin�ts����+Iare�|Gultimately�constructed�from�|Hother�p�A�oin��9ts.�QIIn�a�lemma�at�the�end�of�this�section,��w��9e�state����+Iand�Tpro��9v�e�this�principle�precisely��:�.����8�C�4src:405 ConstructiveGeometryFinalPreprintVersion.texW��:�e���can�distinguish�the���study�of�suc��9h�man�y-sorted�algebras�from�the���study�of�axiomatic�rst-����+Iorder�?�theories�con��9taining�sym�b�A�ols�for�these�algebraic�op�erations.��Some�questions�can�b�e�tak��9en����+Iup�Twithout�the�consideration�of�axioms�or�logical�inferences.�pW��:�e�shall�discuss�a�few�of�these.����8�C�4src:409 ConstructiveGeometryFinalPreprintVersion.texThere��is��a�second�constructor�for�circles,�&�whic��9h�w�e�can��describ�A�e�for�short�as�\circle�from����+Icen��9ter��and�radius",��@as��opp�A�osed�to�the�rst�constructor�ab�o��9v�e,��@\circle��from��cen�ter�and�p�A�oin�t."����+ISp�A�ecically�����Cir��cle3��!�Z�(�A;���B�r�;�C����)���constructs���a�circle�of�radius��B�C�-��and�cen��9ter��A�,���pro�vided��B��z�6�=�S��C����.����+IThese���t��9w�o���constructors�for�circles�corresp�A�ond�to�a�\collapsible�compass"�and�a�\rigid�compass"����+Iresp�A�ectiv��9ely��:�.��rThe��Ucompass�of�Euclid��Tw�as�a�collapsible�compass:��qy�ou�cannot�use��Tit�to�\hold"����+Ithe��mlength��B�r�C��U�while�y��9ou�mo�v�e�one�p�A�oin�t�of�the�compass��nto��A�.���Y��:�ou�can�only�use�it�to�hold����+Ithe�Oradius�O
�AB����while�one�p�A�oin��9t�of�the�compass�is�xed�at��A�,�]zso�in�that�sense�it�corresp�A�onds�to�����+I�Cir��cle��FC�(�A;���B�r��).�mdThe��second�constructor���Cir��cle3��%���corresp�A�onds�to�a�rigid��compass.����-=�4���
��In�the��next����+Isection�Tw��9e�will�ha�v�e�more�to�sa�y�ab�A�out�the�relationship�b�et��9w�een�Tthese�t��9w�o�Tconstructors.����8�C�4src:422 ConstructiveGeometryFinalPreprintVersion.texW��:�e���in��9tro�A�duce���here�a�rst�example�of�a�\construction"�not�considered�b��9y�Euclid,���the��test-for-����+Ie��quality����construction.��,This�\construction"���D��
g��tak��9es�four�p�A�oin�ts,��and�tests�its�rst�t�w�o�argumen�ts����+Ifor�Tequalit��9y��:�,�pro�A�ducing�the�third�or�fourth�argumen�t�dep�A�ending�on�the�outcome:�����a��D�(�a;���b;�c;�d�)���=�����&\���u

cmex10�(�����1�����c����if�T�a��=��b���������d���L�if�T�a��6�=��b����������+I�4src:428 ConstructiveGeometryFinalPreprintVersion.tex�The�ݎalgebraic�approac��9h�to�constructions�w�as�ݍpioneered�b�y�Kijne�[�15��	?�],��but�all�the�systems�he����+Iconsidered���con��9tained���\decision�functions"�suc�h�as���test-for-equalit�y�or�test-for-incidence.��In�this����+Ipap�A�er�Tw��9e�will�not�study�systems�con�taining�decision�functions.����8�C�4src:432 ConstructiveGeometryFinalPreprintVersion.texW��:�e��note��in�the�follo��9wing�lemma�that,���as�far�as�constructing�p�A�oin��9ts�go�es,���the�other��t��9yp�es�are����+Imere���con��9v�eniences.��IThe�elemen�tary�constructions���can�b�A�e�expressed,��]as�w�e�ha�v�e�noted,��]in�sev�eral����+Iw��9a�ys��using�v��|rariables��of�dieren��9t�t�yp�A�es.��F��:�or�example,��w�e�could��ha�v�e���Interse��ctLines��=+Z�(�K� Z;���L�)�where����+I�K���and�u��L�u��ha��9v�e�t�yp�A�e��u��Line��ְ�,���or��u��Interse��ctLines��<�1�(�A;���B�r�;�C� Z;�D��),���where�u��A�,��B�r��,����C����,�and�u��D����ha��9v�e�t�yp�A�e�u��Point�,����+Iand�����_�1�Interse��ctLines���*z�(�Line�(�A;���B�r��)�;�Line�(�C� Z;�D�A��)���=���Interse��ctLines��<��(�A;���B�;�C� Z;�D�A��)�:������&I�color push gray 0��Lemma���1�	color pop���['1�4src:438 ConstructiveGeometryFinalPreprintVersion.tex�L��et�,�t�,�b�e�a�,term�of�typ��e�Point,�2�whose�variables�ar��e�al�x�l�of�typ��e�Point.�4�Then�ther�e�,is�a����+Iterm���t���-=�q�%cmsy6���Ni�with��the�same�variables�as��t�,�
also�of�typ��e�Point,�
such�that�in�the�standar��d�plane��t��and��t���-=������+I�determine�C�the�same�function,�E�and��t���-=����;�c��ontains�only�function�symb�ols�of�typ�e�Point�having�Point����+Iar��guments.����+I�4src:444 ConstructiveGeometryFinalPreprintVersion.texPr��o�of�.���By�Z�induction�Z�on�the�complexit��9y�of��t�.�Supp�A�ose�that�Z��t��has�the�form��I��[nter�sectLines�(�r��9;���s�)����+Iwhere����r�Ũ�and��s��are���terms�of�t��9yp�A�e���Line����.�hNone�of�the�elemen��9tary�constructions�has�t�yp�A�e���Line����,����+Iand�
O�r�O�and��s��cannot�
Nb�A�e�v��|rariables�(since�all�v��|rariables�in��t��are�of�t��9yp�A�e���Point��q��),��so��r�O�m�ust�
Nha�v�e�the����+Iform�����Line��?��(�p;���q�R��)���for���some�terms��p��and��q�R��,���and��s��m��9ust�ha�v�e���the�form���Line��?��(�u;���v�R��)�for�some�terms��u����+I�and���v�R��.��Then��t���-=���,l�can�b�A�e��tak��9en�to�b�e����Line��7��(�p���-=���UT�;���q��R���-=������;�u���-=����;�v��R���-=������).��The��other�elemen��9tary��constructions�are����+Itreated�Tsimilarly��:�.�pThe�basis�case,�when��t��is�a�v��|rariable�or�constan��9t,�is�treated�b�y�taking��t���-=�����=����t�.�������+I�4��C^Mo�s3dels�ffof�the�elemen���tary�constructions��阍�+I�4src:453 ConstructiveGeometryFinalPreprintVersion.tex�The��	algebraic��
approac��9h�allo�ws��
us�to�consider�\mo�A�dels"�without�(y��9et)�ha�ving��
form�ulated�an�y����+Iaxioms��or�logical�theories.��There��are�sev��9eral�in�teresting�mo�A�dels,��nof�whic�h��w�e�no�w�men�tion�four.����+IT��:�o�Tdene�these�mo�A�dels,�w��9e�assume�there�are�three�constan�ts�����,���x,�,�and��
����of�t�yp�A�e���Point��y��.��G�
��color push gray 0��ff�ff�r�	J=�����"5��-:�4����LܻThere���is���a�w�Îord�in�Dutc�h,���#�f�cmti8�p���asser�,��for�this�t�yp�<re���of�compass,��whic�h�w�as���used�in�na�vigation�in���the�sev�en�teen�th��	��cen�Îtury��J�.�$�But��there�seems�to��b�<re�no�single�w�ord��in�English�that�distinguishes�either�of�the�t�Îw�o��t�yp�<res�of��compass�from���the��Xother.��ٛ�	color pop����G�color push gray 0���O\�5������	color pop����[�s���������G�color push gray 0�����	color pop���[(��������8�C�4src:457 ConstructiveGeometryFinalPreprintVersion.tex�The���standar��d��plane��is���R���-=�2��*��.��KP��9oin�ts,�#�lines,�#�and��circles�(as�w��9ell�as�segmen�ts,�#�arcs,�#�triangles,����+Isquares,�%yetc.,�%zin�"?extended�">algebras�in�whic��9h�suc�h�">ob���jects�are�considered)�are�in��9terpreted�as�the����+Iob���jects��Bthat��Ausually�b�A�ear�those�names�in�the�Euclidean�plane.��jMore�formally��:�,�ìthe�in��9terpretation����+Iof�@�the�t��9yp�A�e�sym�b�A�ol�\��Point��dZ�"�is�the�set�@�of�p�oin��9ts,���the�in�terpretation�of�\��Line��`��"�is�the�set�of����+Ilines,��<etc.���In��particular�w��9e��m�ust��c�ho�A�ose�three��sp�ecic�non-collinear��p�oin��9ts�to�serv�e��as�the����+Iin��9terpretations�=�of�����,�����x,�,���and��
����.���Let�us�c�ho�A�ose����-�=�f(0�;����1),��������=�(1�;��0),���and�=��
���=�(1�;��0).���The����+Iconstructor��[and�accessor�functions�listed�ab�A�o��9v�e��[also�ha��9v�e��[standard�and�ob��9vious�in�terpretations.����+IIt���is�when�w��9e�come���to�the�v�e�op�A�erations�for�in�tersecting�lines�and�circles���that�w�e�m�ust�b�A�e�more����+Isp�A�ecic.�pThere�Tare�three�issues�to�decide:�������+I�color push gray 0��`����	color pop���AI�4src:466 ConstructiveGeometryFinalPreprintVersion.tex�when�Tthere�are�t��9w�o�Tin�tersection�p�A�oin�ts,�whic�h�one�is�denoted�b�y�whic�h�term?�������+I�color push gray 0��`����	color pop���AI�4src:467 ConstructiveGeometryFinalPreprintVersion.tex�In�Tdegenerate�situations,�suc��9h�as���Line��v�(�P�A�;���P�H��),�what�do�w�e�do?������+I�color push gray 0��`����	color pop���AI�4src:468 ConstructiveGeometryFinalPreprintVersion.tex�When��the��indicated�lines�and/or�circles�do�not�in��9tersect,�ιwhat�do�w�e��do�ab�A�out�the�term(s)����AIfor�Ttheir�in��9tersection�p�A�oin�t(s)?������+I�4src:470 ConstructiveGeometryFinalPreprintVersion.texW��:�e�t3tak��9e�t2up�the�last�item�rst.��When,��mfor�example,�line�t3�L��do�A�es�not�meet�circle��C����,��mw��9e�sa�y�t2that�the����+Iterm����Interse��ctLineCir�cle1��Ux��(�L;���C����)��is��undene��d�.��In�other�w��9ords,��ythe��op�A�erations�of�these�\algebras"����+Ido���not�ha��9v�e���to���b�A�e�dened�on�all�v��|ralues�of�their�argumen��9ts.��\The�same�issue,��6of�course,�arises����+Iin��man��9y��other�algebraic�con�texts,��for�example,�division�is�not��dened�when�the�denominator�is����+Izero,�Tand��������p���
ʦ�����aH>8�Z��x����2�is�not�dened�(when�doing�real�arithmetic)�when��x��is�negativ��9e.����8�C�4src:475 ConstructiveGeometryFinalPreprintVersion.texRegarding���the���\whic��9h�is�whic�h"���issue,��qour�guiding�principle�is�con��9tin�uit�y��:�.��<W�e�therefore���mak��9e����+Ithe��Ffollo��9wing��Edenitions:���i�Interse��ctCir�cles1��H���(�C� Z;���K����)�is�the�in��9tersection�p�A�oin�t��E�P��)�suc�h�that��Ethe�angle����+Ifrom���`�c��enter���n�(�C����)��`to���c��enter���(�K����)��ato��P��C�mak��9es�a�\left�turn".��This�is�dened�as�in�computer�graphics,����+Iusing�;�the�sign�of�the�cross�pro�A�duct.���Sp�ecically��:�,�EJlet�;��A�ѹ�=��Ѻ�c��enter�����(�C����)�;�and��B�Do�=��Ѻ�c��enter���(�K����).���Then����+Ithe�9�sign�of�9�(�A������B�r��)����(�P�����B��)�9�determines�9�whether�angle��AB�P����is�a�\left�9�turn"�or�a�\righ��9t����+Iturn".�UiTh��9us�(Q����x,
����is�(Ra�left�turn�and��
����x,�*�is�a�righ��9t�turn.�UhIn�case�the�t��9w�o�in�tersection�(Qp�A�oin�ts�are����+Idieren��9t,��one���of�these���cases�m�ust�apply��:�.��This�explanation���has�used�a�case�distinction�as�w��9ell�as����+Ithe�&�cross�&�pro�A�duct;�/wlater�w��9e�will�sho�w�ho�w�&�to�dene�\left�turn"�and�\righ��9t�turn"�using�only�the����+Iaxioms���of�Euclidean�geometry�and���in��9tuitionistic�logic.�FF��:�or�no�w���w�e�simply�note�that�this�notion����+Iis�0�constructiv��9ely�0�app�A�ealing,�7�b�ecause�of�con��9tin�uit�y:�S�there�exists�0�a�unique�con�tin�uous�0�function�of����+I�C�g��and����K��that�satises�the�stated���handedness�condition�for���Interse��ctCir�cles1��L.
�when��C��and��K����+I�ha��9v�e�Tt�w�o�distinct�in�tersection�p�A�oin�ts,�and�is�dened�whenev�er��C��<�and��K��in��9tersect�(at�all).����8�C�4src:486 ConstructiveGeometryFinalPreprintVersion.texThe�8principle�of�con��9tin�uit�y�8leads�us�to�mak��9e���Interse��ctCir�cles1��F���(�C� Z;���K����)�and���Interse��ctCir�cles2���(�C� Z;���K����)����+Iundened�^�in�the�^�\degenerate�situation"�when�circles��C���and��K��coincide,��Yi.e.�ߜha��9v�e�^�the�^�same�cen��9ter����+Iand���radius.��IOtherwise,��as�the�cen��9ter�of��C�Y��passes�through�the�cen�ter�of����K����,��there�is���a�discon-����+Itin��9uit�y��:�.�It�i�mak�es�sense,�~�an�yw�a�y��:�,�to�i�ha�v�e�them�i�undened�when��C�
o�and�K�
n�coincide,�~�as�the�usual����+Iform��9ulas�-�for�computing�-�them�get�zero�denominators,�3�and�there�is�no�natural�w��9a�y�to�-�select�t��9w�o����+Iof�Tthe�innitely�man��9y�in�tersection�p�A�oin�ts.����8�C�4src:492 ConstructiveGeometryFinalPreprintVersion.texW��:�e���still���need�to�settle�the�\whic��9h�is�whic�h"�issue���for���Interse��ctLineCir�cle1��V���(��Line��`��(�A;���B�r��)�;�C����)����+Iand����Interse��ctLineCir�cle2��V
��(��Line��`��(�A;���B�r��)�;�C����).�&�Here��the��rule�is�that�these�t��9w�o��p�A�oin�ts�m�ust��o�A�ccur�in����+Ithe�Tsame�order�on���Line��v�(�A;���B�r��)�as��A��and��B��	�do.�pAgain,�w��9e�require�con�tin�uit�y�of�the�function���������o�Interse��ctLineCir�cle1����?�(�A;���B�r�;�P�A�;�Q�)����+I�4src:495 ConstructiveGeometryFinalPreprintVersion.texwith�Tfour���Point����v��|rariables,�i.e.��������Interse��ctLineCir�cle1�����(��Line��`��(�A;���B�r��)�;���Cir��cle�����(�P�A�;�Q�))�:����+I�4src:496 ConstructiveGeometryFinalPreprintVersion.tex�There�\[is�a�unique�con��9tin�uous�\[extension�from�the�\\set�of�(�A;���B�r�;�P�A�;�Q�)�\[where�there�are�t��9w�o�\[in�ter-����+Isection��p�A�oin��9ts�to�the��set�where�there�is�at�least�one�in��9tersection�p�A�oin�t;�^ that�extension�is�the����+Iin��9terpretation�Tof���Interse��ctLineCir�cle1��V$�.����8�C�4src:498 ConstructiveGeometryFinalPreprintVersion.texRegarding��xthe��ydegenerate�forms���Line��@�(�P�A�;���P�H��),��Aw��9e�sa�y��xthat�the�term�is�undened.���With�re-����+Isp�A�ect���to���degenerate�circles,��Ɗ�Cir��cle��τ�(�A;���A�),�Ƌcon��9tin�uit�y�and���computabilit�y�do���not�presen�t���the�same����+Iobstacles��as�in��the�case�of�degenerate�lines���Line��~��(�A;���A�).�6kTh��9us�w�e��ha�v�e�a�c�hoice��to�allo�w�degen-����+Ierate�n�segmen��9ts�n�and�circles,���without�destro��9ying�the�con�tin�uit�y�n�of�the�elemen��9tary�constructions.�����G�color push gray 0���O\�6������	color pop����t�s���������G�color push gray 0�����	color pop���[(��������+I�W��:�e�$�ma��9y�$�c�ho�A�ose�to�$�allo�w�them�$�or�not.�J�In�the�\standard�mo�A�del"��R��>��-=�2��_��w��9e�tak�e��$��Cir��cle��-��(�A;���A�)�to�$�b�A�e����+Idened,��oi.e.��circles�k�of�zero�radius�are�k�allo��9w�ed.��There�k�is�only�one�p�A�oin��9t�on�suc�h�a�circle.��Then�����+I�Interse��ctLineCir�cle1��~�(�L;���C���ir�A�cl�&9e�(�A;�A�))��Pis��Odened�if��A��is�on�line��L�,�ŏand�is�equal�to��A�.��cW��:�e�also����+Iconsider�إ(brie
y)�the�mo�A�del�ؤ�R��>��-=�2���ڣ�in�whic��9h�degenerate�circles�are�not�allo��9w�ed,���so��إ�Cir��cle����(�A;���A�)�إis����+Iundened.����8�C�4src:506 ConstructiveGeometryFinalPreprintVersion.texThe����r��e�cursive��plane�����R����� �E�tcmbx6�rec����R�-=�2��d�consists���of�p�A�oin��9ts�in�the�plane�whose�co�A�ordinates�are�giv��9en�b�y����+I\recursiv��9e��|reals".���W��:�e��}write��f�e�g�(�n�)�for�the�result,�Ɇif�an��9y��:�,�of��|the��e�-th�T��:�uring�mac��9hine�at�input����+I�n�.�\�Rational��hn��9um�b�A�ers��gare�co�ded��has�certain�in��9tegers,�land�mo�A�dulo�this�co�ding��hw��9e�can�sp�A�eak�of����+Irecursiv��9e��functions�from���N��to��Q�.�	A��\recursiv�e�real�n�um�b�A�er"��x���is�the�index�of�a�T��:�uring�mac��9hine����+I�e�<��suc��9h�that��jf�e�g�(�n�)�(���(��x�j�ӥ�Ӧ�1�=n��for�eac�h��n�ӥ�2�Ӧ�N�.��The�real�n�um�b�A�er�to�whic�h�the�appro�ximations����+I�f�x�g�(�n�)��con��9v�erge�is�sometimes�also�called�a�\recursiv�e�real�n�um�b�A�er",��but�w�e�call�it�the�\v��|ralue����+Iof��B�x�".��8It�is�a��Aroutine�exercise�to�sho��9w�that�the�recursiv��9e�p�A�oin�ts��Bin�the�plane�are�closed�under����+Ithe���Euclidean�constructions.�nIn�particular,��giv��9en�appro�ximations�to�t�w�o�circles�(or�to�a�circle����+Iand��a��line),�
�w��9e�can�compute�appro��9ximations�to�their�\in��9tersection�p�A�oin�ts",�
�ev�en��though�it�ma��9y����+Iturn�fout�fthat�when�w��9e�compute�b�A�etter�appro��9ximations�to�the�circles,�z>w��9e�see�that�they�do�not����+Iin��9tersect�Tat�all.����8�C�4src:515 ConstructiveGeometryFinalPreprintVersion.texIn�ӝthe�recursiv��9e�plane,�0there�is�no�computable�test-for-equalit�y�function,�0that�is,�/no�com-����+Iputable��7function��D� ��that�op�A�erates�on�t��9w�o��7T��:�uring�mac��9hine�indices��x��and��y�R��,��
and�pro�duces�0�when����+I�x���and���y�1��are�recursiv��9e�real�n�um�b�A�ers��with�the�same�v��|ralue,��and�1�when�they�are�recursiv��9e�real����+In��9um�b�A�ers��lwith�dieren��9t�v��|ralues.��xPro�of,��gif�w��9e�had�suc�h�a��D�A��,��gw�e�could�solv�e�the�halting�problem�b�y����+Iapplying���D���to�the�p�A�oin��9t��(�E��2�(�x�)�;����0),���where��f�E��(�x�)�g�(�n�)���=�1�=n���if��T��:�uring�mac��9hine��x��do�A�es�not�halt�at����+Iinput��F�x��in��Efew��9er�than��n��steps,��|and��f�E��2�(�x�)�g�(�n�)���=�1�=k���otherwise,��|where��F�x��halts�in�exactly��k���steps.����+INamely��:�,��f�x�g�(�x�)�zhalts�yif�and�only�if�the�v��|ralue�of��E��2�(�x�)�is�not�zero,�if�and�only�if��D�A��(�Z� Z;���E��2�(�x�))����6�=�0,����+Iwhere�o<�Z�#�is�an�index�o;of�the�constan��9t�function�whose�v��|ralue�is�the�(n��9um�b�A�er�o<co�ding�the)�o<rational����+In��9um�b�A�er�Tzero.����8�C�4src:523 ConstructiveGeometryFinalPreprintVersion.texReaders�9�familiar�9�with�recursion�theory�ma��9y�realize�that�there�are�sev��9eral�w�a�ys�9�to�dene����+Icomputable�<functions�<of�real�n��9um�b�A�ers.���The�mo�del�<w��9e�ha�v�e�<just�describ�A�ed�is�essen��9tially�the����+Iplane�}�v��9ersion�of�the�\eectiv�e�op�A�erations".�UHIt�is�a�w�ell-kno�wn�theorem�of�Tseitin,�׭Kreisel,����+ILaCom��9b�A�e,�y�and�e�Sho�eneld,�kno��9wn�e�traditionally�eas�KLS�ek(and�easily�adapted�to�the�plane)�that����+Ieectiv��9e� op�A�erations� are�con�tin�uous.�<�Of�course,�"�in�the� case�at�hand�w��9e�can�c�hec�k� the�con�tin�uit�y����+Iof�Tthe�elemen��9tary�constructions�directly��:�.����8�C�4src:528 ConstructiveGeometryFinalPreprintVersion.texThe����algebr��aic�ճplane��consists���of�p�A�oin��9ts�in�the�plane�whose�co�A�ordinates�are�algebraic.�֬Since����+Iin��9tersection���p�A�oin�ts�of�circles���and�lines�are�giv�en�b�y�solutions�of���algebraic�equations,��jthe�algebraic����+Iplane�M8is�also�M9closed�under�these�constructions.��Since�algebraic�n��9um�b�A�ers�M8can�b�e�M8computed,�[1this����+Iis�Ta�submo�A�del�of�the�recursiv��9e�plane.����8�C�4src:532 ConstructiveGeometryFinalPreprintVersion.texIn���the���algebraic�plane,�nthere��is��a�computable�test-for-equalit��9y�function��D�A��.��IW��:�e�assume����+Ialgebraic��n��9um�b�A�ers��are�giv�en��b�y�means��of�a�rational�in��9terv��|ral�(�a;���b�)�and�a�square-free�p�A�olynomial����+I�f��q�2����Q�[�x�]���suc��9h�that��f��P�has�only�one�ro�A�ot�in�(�a;���b�).��6T��:�o�determine�if�(�a;�b�)�and����f��O�determine�the�same����+Ior�EEa�EDdieren��9t�real�n�um�b�A�er�than�ED(�p;���q�R��)�and��g��,�QArst�EDc��9hec�k�EEif�the�t��9w�o�rational�EEin�terv��|rals�o�v�erlap.��BIf����+Inot,�h�the�Xt��9w�o�Xreals�are�dieren��9t.��If�so,�h�let�(�r�;���s�)�b�A�e�Xthe�in�tersection.��No�w�Xw�e�ha�v�e�Xto�determine����+Iif���f���and��g�n��ha��9v�e��a��common�zero�on�(�r��9;���s�).�0%There�is�a�simple�recursiv��9e�algorithm�to�do�that:�)�Sa��9y����+I�g����has�A.degree�greater�A/than�or�equal�to�that�of��f����.���Then�write��f�Ո�=����g�R�h�+t�+��r����with��r��of�A.lo��9w�er�degree����+Ithan���g�R��.�3Then��f��G�and��g�@E�ha��9v�e��a��common�zero�on�(�r��9;���s�)�if�and�only�if��f��F�and��r�/b�ha��9v�e��a�common�zero.����+IRecurse�Tun��9til�b�A�oth�p�olynomials�are�linear,�when�the�decision�is�v��9ery�easy�to�mak�e.����8�C�4src:540 ConstructiveGeometryFinalPreprintVersion.texThe�k~�T��J�arksi��jmo��del��is��K����G��K����,��where��K��J�=�!a�Q�(������p��������O\T�Q���	�Q�)�is�the�k}least�subeld�of�the�reals�con��9taining����+Ithe�4rationals�4and�closed�under�taking�the�square�ro�A�ot�of�p�A�ositiv��9e�elemen�ts.�x�This�is�4a�submo�A�del����+Iof�Tthe�algebraic�plane.����8�C�4src:543 ConstructiveGeometryFinalPreprintVersion.texIn��Gsuc��9h��Hmo�A�dels,��the�elemen�tary��Hconstructions�are�in��9terpreted�as�functions�from��M��,�to��M����,����+Iand���of�course��M����m��9ust�b�A�e�closed�under�those�functions,��
but�they�ma�y�also�ha�v�e�other�in�teresting����+Iprop�A�erties.�/�F��:�or�qexample,���in�the�examples�giv��9en�ab�o��9v�e,���the�q(in�terpretations�qof�the)�elemen�tary����+Iconstructions��are��	all�computable�functions;��windeed�they�are�algebraic�functions�of�lo��9w�degree.��In����+Iparticular,�N�they�C[are�all�CZcon��9tin�uous�C[on�their�domains.���Hence,�N�for�example,�there�is�CZno�algebraic�����G�color push gray 0���O\�7������	color pop�������s���������G�color push gray 0�����	color pop���[(��������+I�test-for-equalit��9y�_�function�_�in�the�algebraic�plane,�r}ev��9en�though�there�is�a�computable�one�(since����+Ia���test-for-equalit��9y���function�w�ould�enable�us���to�dene�a�function��f����(�x�)�suc��9h�that��f����(0)�}�=�}�0�but����+I�f����(�x�)���=�1�Twhen��x����6�=�0;�Tsince��f���is�not�con��9tin�uous,�Tit�cannot�b�A�e�algebraic).�������+I�5��H�zEuclid's�ffBo�s3ok�I,�Prop�osition�2��阍�+I�4src:552 ConstructiveGeometryFinalPreprintVersion.tex�Bo�A�ok�f�I,�f�Prop�osition�2�has�b�een�f�discussed�in�the�in��9tro�A�duction.��2The�question�it�addresses�concerns����+Ithe��constructor���Cir��cle3��!���(�A;���B�r�;�C����),��swhic��9h��constructs�a�circle�with�cen�ter��A��and�radius��B�r�C����.�}�As����+Idiscussed��9in�the��8in��9tro�A�duction,���Euclid�giv�es�a�term��8that�accomplishes�this�aim�under�the�assump-����+Itions,�\�not�Nionly�Njthat��B�c��6�=����C����,�but�Njalso�that��A����6�=��C��.�_�4F�B�c��6�=��C����.�DZW��:�e�Niconsider�the�stronger�theorem����+I\uniform��<Euclid��=I.2",��twhic��9h�asserts�that�for�ev��9ery��A��and��B�r�C�^$�with��B�{�6�=����C����,��tthere�exists�a��D���with����+I�AD��x�=�_��B�r�C����.��The���\uniformit��9y"�refers���to�the�missing�assumption��A�_��6�=�_��B�r��.��T��:�o�\realize"�uniform����+IEuclid���I.2,��w��9e�w�ould���need�a�term��e�(�A;���B�r�;�C����)���that�pro�A�duces��D���uniformly��:�,��whether��A����=��B�<��or���not.����+IIf�Tw��9e�had�suc�h�an��e�,�then�of�course�w�e�could�dene����������Cir��cle3������(�A;���B�r�;�C����)���=���Cir��cle�����(�A;���e�(�A;�B�;�C����))�:����+I�4src:561 ConstructiveGeometryFinalPreprintVersion.tex�Con��9v�ersely��:�,�Tsuc�h�an��e��can�b�A�e�dened�from���Cir��cle3��$1H�lik�e�this:�������e�(�A;���B�r�;�C����)���=���p��ointOnCir�cle��<8|�(��Cir��cle3����(�A;���B�;�C����))����+I�4src:563 ConstructiveGeometryFinalPreprintVersion.texBut�Ȩthe�mere�fact�that�Euclid's�o��9wn�construction�ȩdo�A�es�not�suce�to�dene���Cir��cle3��%���do�es�not����+Isho��9w���that�some�other�construction���w�on't�do�the�job.�"SHo�w�ev�er,��w�e�are���able�to�pro�v�e�that�no����+Iother���construction�can�dene���Cir��cle3�� ���,��tas�a�corollary�to�the�follo��9wing�theorem,�whic��9h�sho�ws�that����+Iwithout��XS�Cir��cle3�� ^��,�~!no�XStotal�function�XTis�denable,�~ at�least�if�w��9e�insist�that���Cir��cle��aN�(�x;���x�)�is�undened.������&I�color push gray 0��Theorem���1�(with�F���reek�Wiedijk)�	color pop����t��4src:570 ConstructiveGeometryFinalPreprintVersion.tex�No�\utotal�unary�p��oint-value�d�\ufunction�(other�than�the�iden-����+Itity)�Bis�denable�Ain�the�standar��d�mo�del��R��>��-=�2���?�(with�de�gener�ate�cir�cles�undene�d)�Afr�om�the�ele-����+Imentary�D�ge��ometric�al�c�onstructions�D�excluding��Cir�cle3��!Kt.�<�Mor�e�D�pr�e�cisely,�F�let��t��b�e�D�a�c�omp�ound�term����+Ic��ontaining�Fpexactly�Fqone�variable��A��of�typ��e�Point,�Hand�no�other�variables,�Hbut�p��ossibly�c�ontaining����+Isome���c��onstants����;����x,;��:�:�:��	��of���typ�e�Point.��Supp�ose�that��t��do�es���not�c�ontain�Cir�cle3.��L�et�the�c�onstants����+Ib��e��
interpr�ete�d�as�c�ertain�(xe�d)�distinct�p�oints�in�the�standar�d�plane.��mThen�for�some�value�of����+I�A�,��H�t�x��is�x�not�dene��d�in�the�standar��d�plane.��RIn�fact,��Gwe�c�an�make�x��t��undene�d�by�assigning�x��A��the����+Isame�N<value�as�any�c��onstant�o�c�curring�in��t�,�or�if��t��has�no�c�onstants,��t��is�always�undene�d.����+I�4src:584 ConstructiveGeometryFinalPreprintVersion.texPr��o�of�.�CW��:�e�"4start�b��9y�eliminating�the�\o�v�erloaded"�v�ersions�of�the�elemen�tary�constructions�from����+I�t�.��,F��:�or�\�instance,�n�if��t��con��9tains�a�subterm��I��[nter�A�sectLine�(�a;���b;�c;�d�),�w��9e�\�replace�it�b�y�a�term�using����+Ithe��\fundamen��9tal"�form�of�the�construction,��f�Interse��ctLines��=]��(��Line��`��(�a;���b�)�;���Line���r�(�c;�d�)).�#�The��result����+Iof���suc��9h�replacemen�ts�is�a���term�with�the�same�v��|ralue�as��t��under�an�y�assignmen�t���of�a�v��|ralue�to�the����+Iv��|rariable�T�A�,�and�con��9taining�no�v�ariables�of�t��9yp�A�es�other�than���Point��y��.����8�C�4src:592 ConstructiveGeometryFinalPreprintVersion.texW��:�e��*pro�A�ceed��+b��9y�induction�on�the�complexit��9y�of�terms��t��as�in�the�theorem,���but�also�con��9taining����+Ionly�Tthe�fundamen��9tal�v�ersions�of�the�elemen�tary�constructions�(no�o�v�erloaded�v�ersions).����8�C�4src:595 ConstructiveGeometryFinalPreprintVersion.texSince���the�theorem�only�applies�to�comp�A�ound�terms,���the�basis�case�o�ccurs�when��t��has�only����+Iv��|rariables�r�or�r�constan��9ts�for�argumen��9ts.�4W��:�e�note�that���Cir��cle��{��(�A;���A�),�����Line��*��(�A;�A�),�����R��ay�����(�A;�A�),�����+I�Se��gment��O�K�(�A;���A�),��@��A�Îr��c�����(�A;�A;�A�)�8	are�8
undened.���But�w��9e�also�ha�v�e�to�8
consider�the�p�A�ossibilit�y�that����+Ithe��vconstan��9ts�����or��
�&�o�A�ccup�y�one�of�the�argumen�t�places.��&F��:�or�example,���o�Cir��cle���i�(�A;����x,�)�is�undened����+Iwhen����A����tak��9es�the�v��|ralue����������`������Jw�that�is�assigned�to�the�constan��9t���x,�.��In�the�rest�of�the�pro�A�of�w��9e�shorten����+Ithis��_kind��`of�statemen��9t�to�\is�undened�when��A����=���x,�."���Similarly��:�,���*�Cir��cle���$�(��;���A�)��_is�undened�when����+I�A����=���x,�;�����Line��l�(�A;�����)�y�and��y��Line��ړ�(��;�A�)�y�are�undened�when�y��A����=���x,�,���and�y�the�same�for���R��ay��A��,�����A�Îr��c��5�,���and�����+I�Se��gment��O�K�.���Note�Uwthat�Uxthis�argumen��9t�do�A�es�not�w�ork�for���Cir��cle3��!\�,�e�since���Cir��cle3���(�A;����x,;�
����)�Uxis�alw��9a�ys����+Idened,�Tbut�b��9y�h�yp�A�othesis,��t��do�es�not�con��9tain���Cir��cle3��!��.����8�C�4src:607 ConstructiveGeometryFinalPreprintVersion.texSince�)�t�)�do�A�es�not�con��9tain�an�y�)o�v�erloaded�constructions,�.the�)basis�case�is�nished,�.as�there�is����+Ino��fundamen��9tal��construction�that�tak��9es�only�argumen�ts��of�t�yp�A�e���Point�.�{Sp�ecically��:�,��nthere��are�����G�color push gray 0���O\�8������	color pop����	�G�s���������G�color push gray 0�����	color pop���[(��������+I�just�<�v��9e�fundamen�tal�constructions�<�for�pro�A�ducing�the�in�tersections�of�lines�and�<�lines,�F�lines�and����+Icircles,�Tor�circles�and�circles,�and�they�eac��9h�need�argumen�ts�of�t�yp�A�e���Line���p�or��Cir��cle�.����8�C�4src:612 ConstructiveGeometryFinalPreprintVersion.texNo��9w��,consider�the�induction�step.��If��-the�main�sym�b�A�ol�of��t��is�a�constructor,��"suc�h�as���Line��
��,����+Ithen���t��~�has�the�form���Line��CG�(�a;���b�).�~One�of��a��or��b��m��9ust�con�tain��~a�v��|rariable,��and�hence�b�A�e�somewhere����+Iundened.�"�Hence�c�t�b�is�also�somewhere�undened,��and�indeed�the�same�assignmen��9t�of�a�v��|ralue�to����+I�A��v�that�mak��9es��a��or��b��undened�will�w�ork.��Similarly�for�the�other�constructors�(since���Cir��cle3�� ��,���to����+Iwhic��9h�Tthis�argumen�t�do�A�es�not�apply��:�,�is�not�allo�w�ed).����8�C�4src:618 ConstructiveGeometryFinalPreprintVersion.texNext��Iconsider��Jthe�case�when��t��is���Interse��ctLines��=��(�p;���q�R��).��QThen��p��and��q����ha��9v�e��It�yp�A�e��Line�.��PThe����+Ionly���terms�of�t��9yp�A�e���Line���0�are�those�of�the�form���Line���|�(�a;���b�),���or�v��|rariables�of�t�yp�A�e��Line�.���But��t��is����+Inot�Tallo��9w�ed�to�con�tain�v��|rariables�of�t�yp�A�e��Line�.�pTherefore��t��m�ust�ha�v�e�the�form,��������B�Interse��ctLines���؋�(��Line��`��(�a;���b�)�;���Line���r�(�c;�d�))����+I�4src:622 ConstructiveGeometryFinalPreprintVersion.texSince����t��con��9tains�a���v��|rariable,��lone�of��s����=���Line����(�a;���b�)���or��s����=���Line���(�c;���d�)���m��9ust�con�tain�a�v��|rariable,��kand����+Ib��9y�"%induction�"&h�yp�A�othesis�the�term��s�t��is�undened�when�substitution�"&��t��assigns�the�v��|rariable�of��s����+I�to��one��of�the�constan��9ts�in��s��(or�an��9y�constan�t��if��s��has�no�constan��9ts).��Hence��t�GD�is�also�undened.����8�C�4src:627 ConstructiveGeometryFinalPreprintVersion.texSimilarly�Tfor�the�other�elemen��9tary�constructions.�pThat�completes�the�pro�A�of.������&I�color push gray 0��Corollary���1�	color pop���d���4src:630 ConstructiveGeometryFinalPreprintVersion.tex�L��et����e�(�A;���B�r�;�C����)����b�e���any�term�built���up�fr�om�the�elementary���c�onstructions,��Znot�c�on-����+Itaining���Cir��cle3,���having�typ�e�Point,���and�c�ontaining�exactly�thr�e�e�variables��A�,����B�r��,���and����C�2y�of�typ�e����+IPoint.��KThen���it�is���not�the�c��ase�that�in��R��>��-=�2���
��,���whenever��B�i��6�=����C�)��then��e�(�A;���B�r�;�C����)����is�dene��d�and�is����+Ia�N<p��oint��D���such�that��AD�Ӎ�=����B�r�C����.����+I�4src:636 ConstructiveGeometryFinalPreprintVersion.texPr��o�of�.�C�Let�"[us�"\in��9v�en�t�t�w�o�constan�ts������and��
����,�%�and�in�terpret�them�as�t�w�o�distinct�"[p�A�oin�ts�������o��`�������`�and���w�����
������+I�(xed�d�for�the�rest�of�the�pro�A�of��q).��Then�let��f����(�A�)���=��e�(�A;����x,;�
����).��Supp�ose,��>for�d�pro�of�b��9y�con�tradiction,����+Ithat�@�e�(�A;���B�r�;�C����)�@is�dened�whenev��9er��B�K��6�=����C����and�is�a�p�A�oin��9t��D����suc�h�@that��AD���=����B�r�C����.��Then��f����(�A�)����+Iis�Tdened�for�all��A��and�is�alw��9a�ys�Tdieren�t�from��A�,�con�tradicting�the�previous�theorem.������&I�color push gray 0��Corollary���2�	color pop���d���4src:644 ConstructiveGeometryFinalPreprintVersion.tex�Cir��cle3���is���not�denable�in�the�mo��del��R��>��-=�2���
�~�fr�om���the�(other)�elementary�c��onstruc-����+Itions.����+I�4src:647 ConstructiveGeometryFinalPreprintVersion.texPr��o�of�.��p�Cir�cle3��%8d�is�Ta�term��e��fullling�the�h��9yp�A�otheses�of�the�previous�corollary��:�.����8�C�4src:651 ConstructiveGeometryFinalPreprintVersion.tex��Cir��cle3�� �3�is��in��9timately�connected�with�the��\compass"�of�\straigh�tedge�and�compass�construc-����+Itions".���Euclid's�<compass�is�supp�A�osed�to�b�e�<\collapsible",�E�so�that�y��9ou�cannot�use�it�to�measure����+I�B�r�C����and��
then��	mo��9v�e�it�to�dra�w�a��	circle�of�that�same�radius�cen��9tered�at��A�.���Therefore�Euclid����+Ipro��9v�ed���Euclid�I.2,�7�sho��9wing�ho�w���y�ou�can�accomplish�this�with�a�collapsible���compass;�rbut�for����+Ithat�G�w��9e�G�need�to�assume��A����6�=��B�r��.���In�G�eect�w��9e�need�a�test-for-equalit��9y�construction.���When�Pro-����+Iclus�ڑcriticized�Euclid�for�omitting�argumen��9ts�b�y�cases,��p�A�erhaps�this�is�what�he�had�in�mind.����+IThe��_theorem��^\for�ev��9ery��A�,��a�B�r��,�and��_�C�1F�with��B�̌�6�=�Y��C��there�exists��^a�p�A�oin��9t��D��&�with��AD����=�Y��B�r�C����"�is����+I(apparen��9tly)�Tnot�\realized"�b�y�an�elemen�tary�construction��D�Ӎ�=����f����(�A;���B�r�;�C����).����8�C�4src:658 ConstructiveGeometryFinalPreprintVersion.texThe���theorem�ab�A�o��9v�e���sho�ws���that�without���Cir��cle3�� �D�,���w��9e�cannot�ev��9en�dene�a�construction��f��O�suc��9h����+Ithat�p�for�p�eac��9h�p�A�oin�t��A�,����f����(�A�)�is�p�a�p�A�oin�t�dieren�t�p�from��A�.�.�W��:�e�also�cannot�dene�a�construction����+I�f����(�A;���L�)�Tthat�tak��9es�a�p�A�oin�t��A��and�line��L��in�to�a�p�A�oin�t�on��L�,�dieren�t�from��A�.����8�C�4src:662 ConstructiveGeometryFinalPreprintVersion.texThe��/fact�that���0�Cir��cle3��% ��is�not�denable�in��0�R��>��-=�2���
�-�means�that,��&if�w��9e�do�not�include���Cir��cle3��% ��as����+Ia��primitiv��9e�construction,�Y�w�e�shall�not��b�A�e�able�to�dene�it�in�an��9y�axiomatic�theory�that�has����+I�R��>��-=�2���
F*�for�D,a�D+mo�A�del.���It�seems�clear�that�Euclid's�b�A�o�ok�D,do�es�ha��9v�e�D,�R��>��-=�2���
F*�as�a�mo�A�del;�[�so�the�uniform����+Iv��9ersion�<�of�<�Bo�A�ok�I,�Prop�A�osition�2�is�essen��9tially�non-constructiv�e,�husing�Euclid's�<�non-rigid�compass.����+IW��:�e�E#therefore�add���Cir��cle3��$���as�a�fundamen��9tal�E$construction�(rendering�I.2�a�trivialit�y)�and�giv�e�a����+Iconstructiv��9e�Ttheory�that�w�orks�for�the�rest�of�Euclid.����8�C�4src:668 ConstructiveGeometryFinalPreprintVersion.texThere�His�Hstill�an�unresolv��9ed�tec�hnical�Hissue�here:���is���Cir��cle3��"���undenable�in��R��>��-=�2��:��?��The�ab�A�o��9v�e�Hthe-����+Iorem�8do�A�es�not�extend�to��R��>��-=�2��:��,�B�b�ecause,�for�8example,���Interse��ctLineCir�cle2��U4��(��Line��`��(�A;���B�r��)�;���Cir��cle�����(�B�;�x�))����+Iis�wa�vtotal�function�if���Cir��cle��!��is�total.�|W��:�e�conjecture�that���Cir��cle3��$��is�undenable�in��R��>��-=�2��A\�as�w��9ell�as����+Iin�T�R��>��-=�2���
��.�����G�color push gray 0���O\�9������	color pop����
Ĉ�s���������G�color push gray 0�����	color pop���[(���������+I�6��C^Circles�ffof�zero�radius�and���&���@ffcmti12�Cir��G�cle3���阍�+I�4src:675 ConstructiveGeometryFinalPreprintVersion.tex�In�wthis�xsection,�C�w��9e�address�the�t��9w�o�issues�of�xwhether�our�basic�theory�ough��9t�to�include���Cir��cle3��"%��and����+Iwhether��:w��9e�ough�t��9to�allo�w�circles�of�zero�radius.��gW��:�e�conclude��9that�b�A�oth�should�b�e�allo��9w�ed.��gOur����+Iapproac��9h��is�to�consider�what�is�required��to�ac�hiev�e�a�formalization�of�Euclid�using�in�tuitionistic����+Ilogic,�Twith�no�test-for-equalit��9y�construction�needed.����8�C�4src:679 ConstructiveGeometryFinalPreprintVersion.texHere�Tare�some�p�A�ossible�constructions�w��9e�wish�to�consider.����8�C�4src:681 ConstructiveGeometryFinalPreprintVersion.tex�Euclide��an��Extension�.��One���of�the�Euclidean���axioms�sa��9ys�that�w��9e�can�extend�a�giv��9en�segmen�t����+I�AB��=�b��9y�U�a�U�segmen�t��C���D�A��.��More�U�precisely��:�,�e�w�e�can�U�construct�a�p�A�oin��9t��P�E��=�����Extend���"�(�A;���B�r�;�C� Z;�D��)�U�suc��9h����+Ithat��I�B�r�P�ک�=����C���D�6�and��B�f��is�b�A�et��9w�een��H�A��I�and��P�H��.�lThe�assumptions�here�are�that��A����6�=��B�f��and��I�C�5��6�=��D�A��,����+Ibut�Tit�is�not�assumed�that��B�{�6�=����C��<�or��B��6�=����D�A��.����8�C�4src:685 ConstructiveGeometryFinalPreprintVersion.tex�Str��ong�\uExtension�.��ļ�ExtendA��&f��(�A;���B�r�;�C� Z;�D�A��)�8is�the�unique�con��9tin�uous�8extension�of���Extend��Ҕ�(�A;���B�;�C� Z;�D�A��)����+Ithat�Tis�dened�when��A����6�=��B�r��,�Ti.e.�without�assuming��C�5��6�=����D�A��.�pWhen��C��=����D�A��,�w��9e�ha�v�e���������ExtendA�����(�A;���B�r�;�C� Z;�D�A��)���=��B�:����+I�4src:687 ConstructiveGeometryFinalPreprintVersion.tex�W��:�e�Twill�sho��9w�b�A�elo�w�ho�w�to�dene���ExtendA��(���in�terms�of���Cir��cle3��!��.����8�C�4src:689 ConstructiveGeometryFinalPreprintVersion.tex�Pr��oje�ction�.��HThe�?�construction���pr��oje�ct���_�(�P�A�;���L�)�?�tak��9es�a�p�oin��9t�?��P����and�line��L��and�pro�A�duces�a�p�A�oin��9t��Q����+I�on�$��L�$��suc��9h�that��P�m��lies�on�the�p�A�erp�endicular�$�to��L��at��Q�.�J�The�w��9ell-kno�wn�Euclidean�$�construction����+Ifor�-�the�pro���jection�-�applies�only�if��P�v��is�kno��9wn�not�to�b�A�e�on��L�.�e�T��:�o�dene���pr��oje�ct��$��using�that����+Iconstruction,�w��9e��w�ould��require�a�test-for-incidence�that�allo��9ws�us�to�test�whether�p�A�oin��9t��P�D��is�on����+Iline��
�L��or�not.�b�But�no�suc��9h�test-for-incidence�construction�is��	computable�o�v�er�the�computable����+Iplane,���so���the��Euclidean�pro���jection�construction�do�A�es�not�lead�in�an��9y�ob�vious�w�a�y��to�a�denition����+Iof���7�pr��oje�ct��Q��.��(That��6do�A�es��7not,��=ho��9w�ev�er,�constitute��7a�pro�A�of�that���pr��oje�ct��"C��is�not�denable�in�terms�of����+Ithe�Telemen��9tary�constructions,�whic�h�w�e�giv�e�b�A�elo�w.)����8�C�4src:696 ConstructiveGeometryFinalPreprintVersion.texPro���jection��wis��xabsolutely�necessary�in�order�to�reduce�geometry�to�algebra.��'W��:�e�w��9an�t��wto�pic�k��wa����+Iline,��call���it����X��[�,�and�erect�a���p�A�erp�endicular��Y���to����X��[�,��and�pro���ject�eac��9h�p�A�oin�t����P�"��on�to�its���co�A�ordinates����+I�x����=���pr��oje�ct���H�(�P�A�;���X��[�)�Tand��y��p�=�����pr��oje�ct���(�P�A�;���Y�8�).������&I�color push gray 0��Lemma���2�	color pop���['1�4src:699 ConstructiveGeometryFinalPreprintVersion.tex�Str��ong�z�extension�and�pr�oje�ction�z�ar�e�denable�fr�om�Cir�cle3�(with�cir�cles�z�of�zer�o�r�adius����+Ial�x�lowe��d).�~uA�Îlso�
�one�
�c�an,�=�using�those�
�primitives,�=�c�onstruct�the�p�erp�endicular�
�to�line�L�
�p��assing����+Ithr��ough�N<p�oint�P,�without�c�onditions�as�to�whether�P�is�or�is�not�on�L.����+I�4src:704 ConstructiveGeometryFinalPreprintVersion.texPr��o�of�.�pFirst�Tw��9e�dene������R���ExtendA��uZ��(�A;���B�r�;�C� Z;�D�A��)���=���Interse��ctLineCir�cle2��U���(��Line��`��(�A;���B��)�;���Cir��cle3���J�(�B�;�C� Z;�D�A��))����+I�4src:707 ConstructiveGeometryFinalPreprintVersion.texSince�Tcircles�of�zero�radius�are�allo��9w�ed,�Tw�e�do�not�need�to�assume��C�5��6�=����D�A��.����8�C�4src:709 ConstructiveGeometryFinalPreprintVersion.texT��:�o���construct���the�pro���jection�of�p�A�oin��9t��P���on�line��L�,��(w��9e�just�need�some�circle�with�cen��9ter��P����+I�that���in��9tersects��L��in�t�w�o�distinct�p�A�oin�ts��Q��and��R�>�.���Then�the�pro���jection�of��P��t�on��L��is�the�midp�A�oin�t����+Iof�C
segmen��9t�C�QR�>�.���If��L��is���Line�����(�A;���B�r��),�N|then�a�suitable�radius�w��9ould�b�A�e�the�sum�of�the�lengths�of����+I�AB���and����P�H�A�.���Th��9us�the�circle�w�e���need�can�b�A�e�constructed�as���Cir��cle3�� �n�(�P�;���A;���ExtendA��$,��(�A;�B�r�;�P�;�A�)).����+INote�L�that�w��9e�need�strong�extension,�Zpnot�L�just�Euclidean�extension,�b�A�ecause�w��9e�cannot�rule�out����+I�P�ک�=����A�.�pThat�Tcompletes�the�denition�of�pro���jection.����8�C�4src:715 ConstructiveGeometryFinalPreprintVersion.texNo��9w,�~&to�i0construct�i/the�p�A�erp�endicular�to�i0�L��at��P�H��,�~&w��9e�simply�erect�the�p�A�erp�endicular�to�i0�L��at����+Ithe�$3pro���jection�of��P�m�on��L�,�'�using�the�usual�Euclidean�construction.�I
That�completes�the�pro�A�of�of����+Ithe�Tlemma.����8�C�4src:718 ConstructiveGeometryFinalPreprintVersion.texThe���follo��9wing�lemma�helps�mak�e�the�case�for���allo�wing�circles�of�zero�radius�and�for�allo�wing�����+I�Cir��cle3��I��.������&I�color push gray 0��Lemma���3�	color pop���['1�4src:720 ConstructiveGeometryFinalPreprintVersion.tex�In�Nq�R��>��-=�2���Pn�(wher��e�cir�cles�of�zer�o�Npr�adius�do�not�Npexist),��~pr�oje�ction�is�not�Npdenable�in����+Iterms�N<of�the�elementary�c��onstructions�without�Cir�cle3.����+I�4src:724 ConstructiveGeometryFinalPreprintVersion.texPr��o�of�.��Let�@0��A��and����[�b�A�e�constan��9ts,�J�whose�v��|ralues�will�@/b�e�t��9w�o�@0distinct�@/(xed)�p�oin��9ts.��Let�@/�t�(�A�)�b�e����+Ithe�0~term���pr��oje�ct����(�A;�����Line���r�(���;����x,�)).�m�Then��t�(�A�)�0is�dened�for�all�v��|ralues�of��A�.�m�By�Theorem�1,�7I�t�(�A�)�����G�color push gray 0����[�10������	color pop�����@�s���������G�color push gray 0�����	color pop���[(��������+I�is���not�denable���in�terms�of�the�elemen��9tary�constructions�without���Cir��cle3��!�(�,��in�the�mo�A�del��R��>��-=�2���
��.����+IThat�Tcompletes�the�pro�A�of.����8�C�4src:730 ConstructiveGeometryFinalPreprintVersion.texThere�7xare�also�7wsome�\constructions"�that�go�\b�A�ey��9ond�Euclid."���W��:�e�ha�v�e�7xalready�men��9tioned����+Ithe���test-for-equalit��9y���function��D�.�g�In�addition,��vthere�is�a�notion�of��ap��artness��in��9tro�A�duced�b�y����+IHeyting�&�[�10��	?�].�QThe�&�apartness�axiom�sa��9ys�that�if��B��`�6�=�Y��C����,�k?then�for�an��9y�other�p�A�oin�t�&��A�,�k?either����+I�A�/��6�=�/��B���or�tA�A��6�=��C����.�97An�tA\apartness�construction"�w��9ould�b�A�e�an�op�eration�#�suc��9h�that�if��B����6�=�/��C����+I�then�	0#(�A;���B�r�;�C����)�	/is�dened����-=�5���3��,��and�is�equal�to��B�{��or�to��C����,��and�satises��A����6�=�#(�A;���B�r�;�C����).�dIn�other����+Iw��9ords,��fwhen���B�a_�and��C����are�dieren�t,��f�A��m�ust�b�A�e��dieren�t�from�one�of�them,��fand�this�constructor����+Ipic��9ks�Tone�from�whic�h��A��is�dieren�t.����8�C�4src:740 ConstructiveGeometryFinalPreprintVersion.texIf��w��9e�ha�v�e��an�apartness�construction,��then�#(#(�A;���B�r�;�C����)�;�B�;�C����)��is��the��other��p�A�oin��9t�of��B��;�and����+I�C����,�Tso�it�w��9ould�not�b�A�e�necessary�to�add�a�second�apartness�construction.����8�C�4src:743 ConstructiveGeometryFinalPreprintVersion.texThe�|qfollo��9wing�lemma�|pjust�summarizes�what�the�recursiv��9e�mo�A�del��R��>��-=�2���~o�tells�us�ab�A�out�apartness����+Iand�Ttest-for-equalit��9y��:�.��/΍���&I�color push gray 0��Lemma���4�	color pop���['1�4src:745 ConstructiveGeometryFinalPreprintVersion.tex�(i)�f�No�f�ap��artness�c�onstruction�is�f�denable�in�terms�of�the�elementary�c��onstructions����+Iand�-Cir��cle3.�4�(ii)�T��J�est-for-e�quality�is�-not�denable�in�terms�of�the�elementary�c��onstructions�and����+ICir��cle3,�N<even�with�the�aid�of�an�ap�artness�c�onstruction.����+I�4src:750 ConstructiveGeometryFinalPreprintVersion.texPr��o�of�.��^Ad�@N(i):�rdNo�apartness�construction�@Mis�denable�in�terms�of�the�elemen��9tary�constructions����+Iand�����Cir��cle3�� �p�,���since���all���suc��9h�terms�dene�con��9tin�uous���functions,���and�#�is�not�con��9tin�uous.�DAd���(ii):����+IW��:�e��ha��9v�e�sho�wn��
in�a�previous�section�that�no�test-for-equalit��9y�function�exists�in�the�recursiv��9e����+Iplane.��&Ho��9w�ev�er,�c the�S�recursiv�e�plane�S�do�A�es�ha�v�e���Cir��cle3��$���and�S�an�apartness�construction.��&Here�is����+Iho��9w���to�compute���a�p�A�oin�t�apart�from���one�of�t�w�o�distinct���recursiv�e�p�A�oin�ts��B�?��and����C����.�JNamely��:�,��]let��n����+I�b�A�e���a�p�ositiv��9e�in�teger�suc�h�that�1�=n��is�less�than�the�length�of�segmen�t��B�r�C����,��!and�compute�rational����+Iappro��9ximations�M��a�,�[��b�,�[�and��c��to��A�,��B�r��,�[�and��C��x�within�1�=�(4�n�).��#Then��a��cannot�b�A�e�M�within�1�=�(4�n�)�of����+Ib�A�oth��i�b��j�and��c�,��oand�the�answ��9er�is��C�aQ�if��j�a�~G��~F�b�j����<��1�=�(4�n�)��iand��B�0�otherwise.��That�completes�the����+Ipro�A�of.����8�C�4src:759 ConstructiveGeometryFinalPreprintVersion.texW��:�e���in��9v�estigated�a�system�in�whic�h���Cir��cle3��%�s�is�included,�0�but�circles�of�zero�radius�are�not����+Iallo��9w�ed,��*and��`instead���_�pr��oje�ct��!`A�is��_tak�en�as��`primitiv�e.���This�is��`a�w�ork��|rable��`system,��*but�it�is�a�bit�more����+Icomplex,��(and��]there��^seems�to�b�A�e�no�in��9tuitiv�e�justication��]for�pro���jection�in�terms�of�the�compass����+Ithat�%do�A�es�$not�also�suce�to�justify�allo��9wing�circles�of�zero�radius.�`In�fact�it�can�b�A�e�argued�that����+Ithere��is��no�go�A�o�d�reason���not��to�allo��9w�the�t��9w�o�p�A�oin�ts��of�the�compass�to�coincide.�}�W��:�e�therefore����+Ic��9hose�!�to�!�allo�w���Cir��cle3��!(Y�,�$�and�to�allo�w�!�circles�of�zero�radius�b��9y�requiring�that���Cir��cle3��!(X�(�A;���B�r�;�C����)�is����+Ialw��9a�ys�Tdened.��?P���+I�7��C^Con���tin�uous�ffCo�s3ordinatization�and�Arithmetic��阍�+I�4src:769 ConstructiveGeometryFinalPreprintVersion.tex�No��9w�ada�ys���w�e���usually�think�of�analytic�geometry�as�co�A�ordinatizing�a�plane�and�translating����+Igeometrical��!relations��"b�A�et��9w�een�p�A�oin�ts��"and�lines�in�to��"algebraic�equations�and�inequalities.�@�But����+Ithe��con��9v�erse�is�also�p�A�ossible:�"�translating��algebra�in�to�geometry��:�,�Tand�this�is�imp�A�ortan�t�for�lo�w�er����+Iestimates�$ton�$uthe�p�A�o��9w�er�$tof�geometric�constructions,�(=for�example�for�sho��9wing�that�the�mo�A�dels�of����+Ithe�Tgeometry�of�constructions�are�planes�o��9v�er�TEuclidean�elds.����8�C�4src:775 ConstructiveGeometryFinalPreprintVersion.texIn��mo�A�dern��b�o�oks�(suc��9h��as�[�5����])�arithmetic�is�geometrized�as�op�A�erations�on�congruence�classes����+Iof�n�segmen��9ts.�).W��:�e�op�A�erate�instead�n�on�p�oin��9ts�on�some�xed�n�line��X��n�=��'�Line�����(0�;����1),��Nwhere�0�and�1����+Iare�O�t��9w�o�arbitrarily�xed�p�A�oin�ts.���As�far�as�I�O�can�tell,�^hpast�w�ork�on�co�A�ordinatization�has�alw�a�ys����+Iassumed���some���discon��9tin�uous�constructions,���suc�h�as���test-for-equalit�y�or���at�least�apartness.�NSince����+Ico�A�ordinatization�a�itself�a�is�paten��9tly�computable�and�con��9tin�uous,�t�it�is�a�in�some�sense�\o��9v�erkill"�a�to����+Iapp�A�eal��
to�discon��9tin�uous��
and�non-computable�\constructions"�to�ac��9hiev�e��
co�ordinatization�and����+Iarithmetization.�1�Although�q�co�A�ordinatization�q�is�standard,���old,���and�not�complicated,���w��9e�need�to����+Ic��9hec�k���that���it�can�in�fact�b�A�e�done�from�the�sp�A�ecied�primitiv��9es,��without�using�apartness�or�test-����+Ifor-equalit��9y��:�,�p3b�y�*�denitions�*�that�apply�without�(for�example)�case�distinctions�as�to�whether��G�
o��color push gray 0��ff�ff�r�	J=�����"5��-:�5����LܻStrict�$Econstructivists�$Fwill�w�Îan�t�$Eto�assume���2cmmi8�B��?�is�apart�from��C����in�the�denition�of�an�apartness�construction,�8but��	��w�Îe��Xha�v�en't�dened�that�y�et�and�the�discussion�here�mak�es�sense�with�classical�logic.��ٛ�	color pop����G�color push gray 0����[�11������	color pop�������s���������G�color push gray 0�����	color pop���[(��������+I�n��9um�b�A�ers�u�b�eing�m��9ultiplied�are�equal�to�0�or�1�or�not.�>W��:�e�note�that�it�is�crucial�that�circles�of����+Izero�Tradius�b�A�e�allo��9w�ed�T(or�else�w��9e�need�to�tak�e�pro���jection�as�another�primitiv�e).����8�C�4src:788 ConstructiveGeometryFinalPreprintVersion.texIn��this��section,��Sw��9e�giv�e�the��denitions�of�constructions�that�serv��9e�to�implemen��9t�co�A�ordinatiza-����+Ition���and���the�arithmetic�op�A�erations�in�a�con��9tin�uous���w�a�y��:�.��W�e�sho��9w���that,���in�the�mo�A�dels�discussed����+Iab�A�o��9v�e,���these��op�erations��are�dened�and�b�A�eha��9v�e��as�desired.���In�a�later�section,���w��9e�will�giv��9e�an����+Iaxiomatic�Ttheory�capable�of�formalizing�these�correctness�pro�A�ofs�without�reference�to�mo�dels.����8�C�4src:794 ConstructiveGeometryFinalPreprintVersion.texRecall��<that��=the�concepts�of�\righ��9t�turn"�and�\left�turn"�ha��9v�e��<b�A�een�in�tro�A�duced��<and�discussed����+Iin�MSection�M~4.���F��:�or�example,�uwin�Fig.���1,��AO�A�Z��f�is�Ma�\righ��9t�turn",�uwb�ecause�M~the�sign�of�the�cross�pro�A�duct����+I�O�A�A�?���?��O�B�Ғ�is�_�p�ositiv��9e.��In�tuitiv�ely��:�,�r~tra�v�eling�_�from��A��to��O����to��Z���requires�one�to�turn�righ��9t�at��O�A��.����+IThis���denition�of�\righ��9t�turn"�is�adequate�for�this�section,���since�w�e�are�only�concerned�with����+Imo�A�dels��where�cross�pro�duct�mak��9es�sense.�L�(Later,���w�e�will�exp�A�end�considerable�eort�dening����+I\righ��9t"�Tand�\left"�in�an�axiomatic�con�text.)����8�C�4src:801 ConstructiveGeometryFinalPreprintVersion.texT��:�o��=p�A�erform�addition�geometrically�w��9e�supp�ose�giv��9en�a�line��L����=���Line����(�R�>;���S���)��=and�an�\origin",����+Ia�A�p�A�oin��9t��O��I�on��L��with��S�Ĉ�b�et��9w�een�A��R�Q��and��O��.���W��:�e�need�to�dene�a�construction��Add�(�A;���B�r��),�L�whic��9h����+Ialso�~dep�A�ends,��Nof�course,�on��S���,��O�R�>�,�and��O�A��,��Osuc��9h�~that��Add�(�A;���B�r��)�is�a�p�A�oin��9t��C�"�on��L��represen��9ting����+Ithe�`(signed)�sum�of��A��and��B�r��,�#with��O�^'�considered�as�aorigin.�1�In�particular�if��A��and��B���are�on�the����+Isame���side�of��O�!��then��Add�(�A;���B�r��)���=���ExtendA��%4�(�O�A�;�A;�O�;�B�r��),��but���that�do�A�es�not�suce�to�dene��Add�.����+IIt�Tma��9y�b�A�e�instructiv�e�to�see�another�failed�attempt�to�dene��Add�(�O�A�;���A;�B�r��):�������|��Interse��ctLineCir�cle2���n��(��Line��`��(�O�A�;���B�r��)�;�C���ir�cl�&9e�3(�A;��0�;�B�r��))����+I�4src:808 ConstructiveGeometryFinalPreprintVersion.texwhic��9h�Tis�correct�when��O�Ӎ�6�=����B�r��,�but�is�undened�when��O��=����B�r��.������&I�color push gray 0��Lemma���5�	color pop���['1�4src:810 ConstructiveGeometryFinalPreprintVersion.tex�Given���line���L�Ѣ=�Line(R,S),�and�a�p��oint��O���on��L��with��S�T��b��etwe�en��R���and����O�A��,��we�c��an����+Ic��onstruct���a���p�oint��Add�(�A;���B�r��)����on��L��r��epr�esenting�the���signe�d�sum���of��A��and��B�r��,���with��O��v�c��onsider�e�d����+Ias�N<origin,�using�the�elementary�c��onstructions�and�Cir�cle3.����+I�4src:815 ConstructiveGeometryFinalPreprintVersion.texPr��o�of�.�\YW��:�e�*�rst�dene�an�*�auxiliary�construction���r��otate���v�,�/�whic��9h�requires�as�inputs�three�distinct,����+Inon-collinear�z
p�A�oin��9ts��P�H��,��7�O��,�and�z	�Q�z
�(think�of�angle��P�H�O�Q�),��7as�z	w��9ell�as�a�p�oin��9t�z	�A��on���Line�����(�O�;���P�H��).����+IThe���desired�result���of���r��otate��[t�(�P�A�;���O�;�Q;�A�)���is�a���p�A�oin��9t��Z�*��on���Line���h�(�O�;���Q�)���suc��9h�that��O�Z���=�N��O�A��and����+Iif���A��A�6�=��O����then��AO�A�Z�H��is�a�righ��9t�turn.�ˀ(See�Fig.�1.)�ˁThe�p�A�oin��9t�is�that��Z�H��is�dened�ev�en�when����'��G��NZ��color push gray 0���'���_�ٍ���|X�color push gray 0�Figure�UU1:�q��
�b>

cmmi10�Z�~4�=����(�':

cmti10�r��}'otate���I�(�P�G;���O�G;�Q;�A�)�	color pop����%'��4src:822 ConstructiveGeometryFinalPreprintVersion.tex����C��h" tx@Dict begin STP newpath 0.8 SLW 0  setgray  0.0 0.0 moveto 147.84004 0.0 L 0 setlinecap stroke  end ���v���NO���U[���NA�������NP��m" tx@Dict begin STP newpath 0.8 SLW 0  setgray  0.0 0.0 moveto 147.84004 85.35826 L 0 setlinecap stroke  end ����
���IL��m" tx@Dict begin STP newpath 0.8 SLW 0  setgray  0.0 0.0 moveto 85.35826 147.84004 L 0 setlinecap stroke  end ��DIp��wm Q��p" tx@Dict begin STP newpath 0.8 SLW 0  setgray  85.35826 0.0 moveto 85.35826 49.2774 L 0 setlinecap stroke  end ��W�=��BJB��u" tx@Dict begin STP newpath 0.8 SLW 0  setgray  85.35826 49.2774 moveto 42.67912 74.00598 L 0 setlinecap stroke  end ��"$����xZ������	color pop�����+I�A��R�=��O�p�(in�̨whic��9h�case�it�is�just��O�A��,��}of�course),�and�if��A�̩�mo��9v�es�̨along���Line��-p�(�O�A�;���P�H��)�through��O��,����+Ithen�"��Z�Ɯ�mo��9v�es�along��"��Line���|�(�O�A�;���Q�),�&
passing�through�"��O�d|�as��A��do�A�es.�D�T��:�o�construct��Z����,�&
w��9e�rst�bisect����+Ithe�/angle�/�P�H�O�A�Q��(b��9y�the�usual�Euclidean�construction,�5�whic��9h�is�not�problematic�since�the�three����+Ip�A�oin��9ts���are�not�collinear).��Let�the�angle�bisector�b�e���line��L�.��Then�let�line��K�ek�b�e�the�p�erp�endicular�����G�color push gray 0����[�12������	color pop����
7�s���������G�color push gray 0�����	color pop���[(��������+I�to����Line��c��(�O�A�;���P�H��)��at���A�,��let��B�u��b�e��the�in��9tersection�p�oin��9t�of���K����and��L�,��and�let��Z����b�A�e�the�pro���jection�of����+I�B��	�on��T�Line��v�(�O�A�;���Q�)�T(whic��9h�is�dened�no�matter�whether��O�Ӎ�=����A��or�not).����8�C�4src:833 ConstructiveGeometryFinalPreprintVersion.texNote�Gthat�there�Gare,�pHif��A����6�=��O�A��,�pGt��9w�o�Gp�oin�ts�G�Z����on��O�Q�G�suc��9h�that��O�Z�5��=����O�A�.�תThe�one�constructed����+Ib��9y��l��r��otate��A��(�P�A�;���O�;�Q;�A�)�l�is�l�suc��9h�that,���if��P�H�O�A�Q��is�a�righ��9t�turn,���then��AO�A�Z���is�a�righ��9t�turn�when��A����6�=��O�A��,����+Iregardless��Yof��Zwhether��A��is�b�A�et��9w�een��O�
 �and��Y�P�<�or�not.��Similarly��:�,��%if��P�H�O�A�Q��is�a�left�turn,��%so�is��AO�A�Z����.����8�C�4src:837 ConstructiveGeometryFinalPreprintVersion.texWith��Ǣ�r��otate�� d�in�Ǣhand,��5w��9e�ǣcan�giv�e�a�ǣconstruction�for��Add�(�A;���B�r��)�(dep�A�ending�also�on��R�>�,��6�S���,����+Iand���O�A��).�<O(The��construction�is�illustrated�in�Fig.�<O2.)�First,�"�w��9e��replace��R�01�and��S����with�new�p�A�oin��9ts����+Ion����L�K��=��K��Line�����(�R�>;���S���),���farther�a��9w�a�y���from����O�A��,�so�that����O�A��,��A�,���and��B����are�all���on�the���same�side�of��R����+I�and���S���,�یand�the��new��R��X�and��S�P�are�in�the�same�order�on�line��L��as�b�A�efore.�](This�can�b�e�done�using�����+I�Extend��Fե�.)�
CNo��9w���erect���the�p�A�erp�endicular����K����to��L��at��O�A��,��and�the�p�A�erp�endicular����H����to��L��at��B�r��.�
CIn�the����+Ipro�A�cess�93of�erecting�these�94p�erp�endiculars,�B*w��9e�will�93ha�v�e�constructed�p�A�oin�ts��C���on��K���and��D�z��on��H����+I�suc��9h�Tthat��R�>O�A�C��<�is�a�righ�t�turn.�pThen�let��ax������U�����0>�=�������b�r��otate���6��(�R�>;���O�A�;�C� Z;�A�)���������
�V�����0>�=�������b�pr��oje�ct����(�U���;���H���)��������+G�W�����0>�=�������b�r��otate���6��(�D�A�;���B�r�;�R�>;�V�8�)����aw��+I�4src:849 ConstructiveGeometryFinalPreprintVersion.texW��:�e�#�set�#��Add�(�A;���B�r��)���=����W�H��.�G�Then��Add�(�A;�B�r��)�is�#�dened�for�all��A;�B�r��.�G�Supp�A�ose��A����6�=��O��.�G�Then�#��U���V�2�is����+Ip�A�erp�endicular�?�to�?�b�oth��K��m�and��H���.��Then��U�8i�and��V�M��are�on�?�the�same�side�of��L�,�Jsince�if��U���V�M��meets����+I�L�W��at�W�a�p�A�oin��9t��X��[�,�hFthen��X�U�P��and�W��X�O��w�are�b�A�oth�W�p�erp�endicular�to�W��K����,�hFwhic��9h�implies��U��B�=�_�O�A��,�hFwhic��9h����+Iimplies��v�A����=��O�A��,��con��9tradicting��A��6�=��O�A��.�{It��uthen�follo��9ws�from�the�prop�A�ert��9y�of���r��otate��{��that��B�F*�and��W����+I�o�A�ccur��on�line��L���in�the�same�order�that��O�+��and��A��o�A�ccur.�Refer�to�Fig.�3�for�an�illustration�of�the����+Icase�Сwhen��A��is�negativ��9e.��This�implies�that��Add�(�A;���B�r��)�represen�ts�the�algebraic�Рsum�of��A��and��B�r��,����+Isince�Tin�magnitude��B�r�W�ک�=����O�A�A�.���G��G���u�color push gray 0�zQ�����<���i�i�color push gray 0�Figure�UU2:�q�Signed�addition�without�test-for-equalit���y�	color pop���i_č�4src:860 ConstructiveGeometryFinalPreprintVersion.tex���8���p" tx@Dict begin STP newpath 0.8 SLW 0  setgray  0.0 8.53635 moveto 256.07478 8.53635 L 0 setlinecap stroke  end �o" tx@Dict begin STP newpath 0.8 SLW 0  setgray  8.53635 0.0 moveto 8.53635 89.62643 L 0 setlinecap stroke  end �" tx@Dict begin STP newpath 0.8 SLW 0  setgray  /ArrowA { moveto } def /ArrowB { BeginArrow 1.  1.  scale false 0.4 1.4 1.5 2. Arrow  EndArrow  } def [ 162.18016 76.8219 8.53635 76.8219  /Lineto /lineto load def false Line  gsave 0.8 SLW 0  setgray 0 setlinecap stroke  grestore end �w" tx@Dict begin STP newpath 0.8 SLW 0  setgray  162.18016 8.53635 moveto 162.18016 85.35826 L 0 setlinecap stroke  end ��" tx@Dict begin STP newpath 0.8 SLW 0  setgray  /ArrowA { moveto } def /ArrowB { BeginArrow 1.  1.  scale false 0.4 1.4 1.5 2. Arrow  EndArrow  } def 8.53635 8.53635 /y ED /x ED /r 68.28687 def /c 57.2957 r Div def /angleA 0. 0.0 c mul 2 div add def /angleB 90. 0.0 c mul 2 div sub def x y r angleA angleB { ArrowB } r 0 gt { { sub } } { { add } } ifelse ArcArrow arc  gsave 0.8 SLW 0  setgray 0 setlinecap stroke  grestore end ��" tx@Dict begin STP newpath 0.8 SLW 0  setgray  /ArrowA { moveto } def /ArrowB { BeginArrow 1.  1.  scale false 0.4 1.4 1.5 2. Arrow  EndArrow  } def 162.18016 8.53635 /y ED /x ED /r 68.28687 def /c 57.2957 r Div def /angleA 90. 0.0 c mul 2 div sub def /angleB 0. 0.0 c mul 2 div add def x y r angleA angleB { ArrowB } r 0 gt { { add } } { { sub } } ifelse ArcArrow arcn  gsave 0.8 SLW 0  setgray 0 setlinecap stroke  grestore end ���v�O���L�hA����.B����eIW=A+B����-�U����.���IV���D�����C���W̟��YR������	color pop���3��G���u�color push gray 0�zQ�����<���u�3�color push gray 0Figure�UU3:�q�Signed�addition�when��A��is�negativ���e�	color pop���i_č�4src:866 ConstructiveGeometryFinalPreprintVersion.tex���8���p" tx@Dict begin STP newpath 0.8 SLW 0  setgray  0.0 76.8219 moveto 256.07478 76.8219 L 0 setlinecap stroke  end �o" tx@Dict begin STP newpath 0.8 SLW 0  setgray  76.8219 0.0 moveto 76.8219 89.62643 L 0 setlinecap stroke  end �" tx@Dict begin STP newpath 0.8 SLW 0  setgray  /ArrowA { moveto } def /ArrowB { BeginArrow 1.  1.  scale false 0.4 1.4 1.5 2. Arrow  EndArrow  } def [ 239.00339 8.53635 76.8219 8.53635  /Lineto /lineto load def false Line  gsave 0.8 SLW 0  setgray 0 setlinecap stroke  grestore end �w" tx@Dict begin STP newpath 0.8 SLW 0  setgray  239.00339 8.53635 moveto 239.00339 85.35826 L 0 setlinecap stroke  end ��" tx@Dict begin STP newpath 0.8 SLW 0  setgray  /ArrowA { moveto } def /ArrowB { BeginArrow 1.  1.  scale false 0.4 1.4 1.5 2. Arrow  EndArrow  } def 76.8219 76.8219 /y ED /x ED /r 68.28687 def /c 57.2957 r Div def /angleA 180. 0.0 c mul 2 div add def /angleB 270. 0.0 c mul 2 div sub def x y r angleA angleB { ArrowB } r 0 gt { { sub } } { { add } } ifelse ArcArrow arc  gsave 0.8 SLW 0  setgray 0 setlinecap stroke  grestore end ��" tx@Dict begin STP newpath 0.8 SLW 0  setgray  /ArrowA { moveto } def /ArrowB { BeginArrow 1.  1.  scale false 0.4 1.4 1.5 2. Arrow  EndArrow  } def 239.00339 76.8219 /y ED /x ED /r 68.28687 def /c 57.2957 r Div def /angleA 270. 0.0 c mul 2 div sub def /angleB 180. 0.0 c mul 2 div add def x y r angleA angleB { ArrowB } r 0 gt { { add } } { { sub } } ifelse ArcArrow arcn  gsave 0.8 SLW 0  setgray 0 setlinecap stroke  grestore end �����A���N������O����.����W=A+B����jʟ���B���N����eU����E���J�V������	color pop��0���8�C�4src:870 ConstructiveGeometryFinalPreprintVersion.tex�Ha��9ving�A�dened�A�addition,���w�e�no�w�turn�A�to�m�ultiplication,���division,���and�square�ro�A�ot.���The����+Igeometrical��denitions�of��these�op�A�erations�go�bac��9k�to�Descartes.�/�On�the�second�page�of��L��a�����G�color push gray 0����[�13������	color pop����'͠s���������G�color push gray 0�����	color pop���[(��������+I�Ge��ometrie��$�[�7����],�/�he��#giv��9es�constructions�for�m��9ultiplication�and�square�ro�A�ots.���W��:�e�repro�duce��$the����+Idra��9wings�4Ffound�4Gon�page�2�of�his�b�A�o�ok�4F[�7����]�in�Figures�4�and�5.�yGHere�is�Descartes'�explanation�of����+Ithese�Tgures:������G���u�color push gray 0�zQ�����<���jg��color push gray 0�Figure�UU4:�q��L��}'a���Multiplic�ation��according�to�Descartes�	color pop���i_č�4src:887 ConstructiveGeometryFinalPreprintVersion.tex���c���h" tx@Dict begin STP newpath 0.8 SLW 0  setgray  0.0 0.0 moveto 170.71652 0.0 L 0 setlinecap stroke  end ���NB��m" tx@Dict begin STP newpath 0.8 SLW 0  setgray  0.0 0.0 moveto 153.64513 92.18707 L 0 setlinecap stroke  end �p" tx@Dict begin STP newpath 0.8 SLW 0  setgray  76.8219 0.0 moveto 68.28687 40.97159 L 0 setlinecap stroke  end ��L�h��NA���;�"��FC��s" tx@Dict begin STP newpath 0.8 SLW 0  setgray  115.23416 0.0 moveto 102.42964 61.45804 L 0 setlinecap stroke  end ��s;��ND���]��Š�E������	color pop��@����+I�color push gray 0��	color pop���AI�4src:894 ConstructiveGeometryFinalPreprintVersion.tex������AI�color push gray 0��$��1.��
�1�	color pop���S�z�4src:897 ConstructiveGeometryFinalPreprintVersion.texL��et�N<�AB����b�e�taken�as�unity.��
����AI�color push gray 0��$�2.��
�1�	color pop���S�z�4src:898 ConstructiveGeometryFinalPreprintVersion.texL��et��Lit�b�e�r�e�quir�e�d��Kto�multiply��B�r�D���by��B�C����.�2I��:have�only�to�join��Kthe�p��oints��A��and����S�z�C����,�N<and�dr��aw��D�A�E��n�p�ar�al�x�lel�to��C���A�;�then��B�r�E��n�is�the�pr�o�duct�of��B�r�D���and��B�C����.������AI�color push gray 0��$�3.��
�1�	color pop���S�z�4src:900 ConstructiveGeometryFinalPreprintVersion.texIf�Uit�b��e�r�e�quir�e�d�to�divide��B�r�E��A�by��B�D�A��,�V�I�U
join��E��A�and��D��,�V�and�dr��aw��AC����p�ar�al�x�lel�to����S�z�D�A�E��2�;�N<then��B�r�C��$�is�the�r��esult�of�the�division.������AI�color push gray 0��$�4.��
�1�	color pop���S�z�4src:901 ConstructiveGeometryFinalPreprintVersion.texIf�RTthe�squar��e�r�o�ot�of��GH�	q�is�RSdesir�e�d,��ZI�Radd,��Yalong�the�same�str�aight�line,��Z�F�H�G����S�z�e��qual��Lto�unity;��Tthen,���bise�cting��F�H�H�Ci�at��K����,���I��<describ�e�the�cir�cle��F�H�I��[H�Ci�ab�out��K�04�as����S�za���c��enter,�%�and���dr�aw�fr�om����G��a�p��erp�endicular�and�extend���it�to��I��[�,�%�and��GI����is�the����S�zr��e�quir�e�d�N<r�o�ot.����8�C�4src:908 ConstructiveGeometryFinalPreprintVersion.tex�F��:�rom�D�the�p�A�oin��9t�D�of�view�of�constructiv�e�D�geometry��:�,�P�there�is�a�problem�with�the�construction.����+INamely��:�,���Descartes���has�only�told�us�ho��9w���to�m�ultiply�t�w�o�segmen�ts�with�non-zero�lengths,���and����+Iat�oleast�one�of�whose�lengths�is�not�1�n(the�length�of�unit��9y{he�needs�this�when�constructing��AC����+I�parallel��to��D�A�E��2�),���while�w��9e��w�an�t��to�b�e�able�to�m��9ultiply��in�general,���without�a�test-for-equalit�y����+Iconstruction.��T��:�o��solv��9e�this��problem,���w�e�recall��from�Lemma�2�that�w��9e�can�dene���p��erp���C�(�P�A�;���L�),���the����+Ip�A�erp�endicular��to���L��passing�through��P�H��,��7without�regard�to�whether��P�.��is�or�is�not�on��L�.��Then�w��9e����+Ican���dene�a�construction�����p��ar�a��?��suc��9h�that,���for�an�y�line��L��and�an�y�p�A�oin�t��P�ٔ�(whic�h�ma�y�or�ma�y����+Inot�heb�A�e�on�hf�L�),��})�p��ar�a���p�(�P�;���L�)�passes�hfthrough��P�H��,�})and�if��P��H�is�not�on��L��then���p��ar�a�����(�P�A�;���L�)�is�parallel�to����+I�L�,�Twhile�if��P�^7�is�on��L�,�then���p��ar�a��3��(�P�A�;���L�)�has�the�same�p�oin��9ts�as��L�.�pThe�denition�of���p��ar�a��H��is�������@}�p��ar�a���^��(�P�A�;���L�)���=���p��erp����(�P�;�����p��erp�����(�P�;���L�))�:����+I�4src:915 ConstructiveGeometryFinalPreprintVersion.tex�In�p'w��9ords:��First�nd�the�p�A�erp�endicular�p'to��L�p&�passing�through��P�H��.�,�Then�erect�the�p�A�erp�endicular����+Ito�Tthat�line�at��P�H��.����8�C�4src:917 ConstructiveGeometryFinalPreprintVersion.texUsing�&3the���p��ar�a��j��construction�where�Descartes�calls�for�\dra��9wing�&2�D�A�E��e�parallel�to��C���A�",�jjw�e����+Ino���longer���ha��9v�e�a�problem���m�ultiplying�n�um�b�A�ers�near���1�or�0.���W��:�e�no��9w�giv�e�a���construction�for����+Im��9ultiplication��(whic�h�of�course��could�b�A�e�written�as�a�single,��jm�uc�h�less�readable,��jterm).�=tThe����+Iconstruction�c�assumes�c�that�0�and�1�are�t��9w�o�c�distinct�p�A�oin��9ts�on�line��X��[�,�w`and��D����and��Q��are�p�A�oin��9ts����+Ion��Sline��R�X����to�b�A�e�m��9ultiplied,���and��Y���is�another�line�through��X��[�,���meeting��X����at�0�and�distinct�from����+I�X��[�.�sW��:�e�cTcould,�v�for�cUexample,�tak��9e��Y�q��to�b�A�e�cTthe�p�erp�endicular�to��X���at�0,�v�although�that�do�es�not����+Imatc��9h�Tthe�illustration�from�Descartes'�b�A�o�ok.������&I�color push gray 0�	color pop���+I�4src:936 ConstructiveGeometryFinalPreprintVersion.tex�)ߤN		cmtt9�Multiply(Point���D,�Point�Q)����+I{���I�=�IntersectLineCircle1(Y,Circle(0,1))����4�uC���=�rotate(1,0,I,Q)��X//�f�is�as�in�the�lemma,�so�0C�=�0Q�����G�color push gray 0����[�14������	color pop����C�s���������G�color push gray 0�����	color pop���[(��������4�u�//���now�we�have�to�multiply�BD�by�BC�	s,per�Descartes����4�uL���=�Line(1,C)�!//�AC�in�Descartes'�diagram.�	s,A�there�is�1�here;�B�is�0.����4�uK���=�para(1,C,D)���//�parallel�to�AC�through�D�	s,(or�AC�itself�if�D=1)����4�uE���=�IntersectLines(K,Y)�	s,//�defined�because�K�is�not�parallel�to�Y����4�uN���=�rotate(I,0,1,E)�%̰//�rotate�length�0E�back�to�line�X�from�line�Y����4�ureturn���N����+I}����8�C�4src:939 ConstructiveGeometryFinalPreprintVersion.tex�Note�|�that�since�|��I�0��is�dened�b��9y���Interse��ctLineCir�cle1��UnT�,��whic�h�side�of�|�0�it�lies�on�(on��Y�8�)�is�deter-����+Imined��iarbitrarily��:�,��1b��9y�the��hunkno�wn�order��hon��Y����of�the�t��9w�o�p�A�oin�ts���i�p��ointOn1��'�O�(�Y�8�)�and���i�p��ointOn2��)>I�(�Y��).����+IThat�m�side�m�of�0�on��Y�{��b�A�ecomes�the�\p�A�ositiv��9e"�side.��If��Q��is�p�A�ositiv��9e�(lies�on�the�same�side�of��Z���er�A�o��as����+I�O�A�ne��8�do�es)��9then��C�p �is�p�A�ositiv��9e,���i.e.�lies�on�the�same�side�of�0�as��I��[�,���and�vice-v��9ersa,�if��Q��9�is�negativ��9e����+I(lies�a�on�the�opp�A�osite�side�of�0�as�1�a�do�es),�t�then��C���lies�on�the�a�opp�osite�side�of��Z���er�o��as��U����.��This����+Imak��9es�7�m�ultiplication�of�7�signed�n�um�b�A�ers�come�out�correctly�without�7�needing�a�test-for-equalit�y����+Iconstructor.����8�C�4src:947 ConstructiveGeometryFinalPreprintVersion.texIt�7remains,�?~of�course,�to�pro��9v�e�7in�7some�geometrical�theory�that�m��9ultiplication�and�addition����+Isatisfy��;the��:eld�la��9ws.��#W��:�e�do�not�tak��9e�that�up�at�this�p�A�oin��9t�since�w��9e�ha�v�e��:not�y�et��:discussed����+Itheories�Tand�axioms.����8�C�4src:950 ConstructiveGeometryFinalPreprintVersion.texDescartes'�'xdivision�metho�A�d�'wis�handled�similarly��:�,�lusing���p��ar�a��m7�where�Descartes�constructs�a����+Iparallel.�`�T��:�urning�ֹto�Descartes'�square�ָro�A�ot�construction,�w��9e�sho�w�ֹthat�it�can�b�A�e�carried�out����+Iuniformly��:�,��without�ޡneeding�to�ޠkno��9w�that�the�p�A�oin��9t�whose�square�ro�A�ot�is�required�is�dieren��9t����+Ifrom��r�Z���er�A�o�.��$W��:�e�carry�out�Descartes's��qconstruction,��mbut�then��rw��9e�need�to��qnd�a�p�oin��9t�on��q�X�G��whose����+Idistance�Tfrom�0�is��I��[G�,�without�dividing�in��9to�cases�as�to�whether��G����=�0�Tor�not.����8�C�4src:956 ConstructiveGeometryFinalPreprintVersion.texNo��9w�V�w�e�tak�e�V�up�the�geometrical�construction�of�square�ro�A�ots.��7Fig.�5�sho��9ws�Descartes'����+Iconstruction.���Q���G���u�color push gray 0�zQ�����<���t=��color push gray 0�Figure�UU5:�q�Square�ro�Gots�according�to�Descartes�	color pop���i_č�4src:959 ConstructiveGeometryFinalPreprintVersion.tex���qϟ�t" tx@Dict begin STP newpath 0.8 SLW 0  setgray  8.53635 8.53635 moveto 179.25287 8.53635 L 0 setlinecap stroke  end �4" tx@Dict begin STP newpath 0.8 SLW 0  setgray  /ArrowA { moveto } def /ArrowB { } def 93.8946 8.53635 /y ED /x ED /r 85.35826 def /c 57.2957 r Div def /angleA 0. 0.0 c mul 2 div add def /angleB 180. 0.0 c mul 2 div sub def x y r angleA angleB arc  gsave 0.8 SLW 0  setgray 0 setlinecap stroke  grestore end �v" tx@Dict begin STP newpath 0.8 SLW 0  setgray  154.25076 8.53635 moveto 154.25076 68.8925 L 0 setlinecap stroke  end �H���U[�K�����'G����@�F����>2��q�I������	color pop������&I�color push gray 0�	color pop���+I�4src:979 ConstructiveGeometryFinalPreprintVersion.tex�SquareRoot(Point���G)����+I{���//�H�in�Descartes'�diagram�is�0����4�uF���=�Add(0,G,0,1)�	s,//�so�FG�has�unit�length����4�uK���=�Midpoint(F,0)����4�uC���=�Circle(K,F)����4�uL���=�perp(G,Line(0,F))����4�uI���=�IntersectLineCircle1(L,C)����4�uU���=�IntersectLineCircle1(L,Circle(G,F))�	s,//�rotate�unit�length�to�line�L����4�uR���=�rotate(U,G,F,I)�//�so�now�RG�=�IG�but�R�is�on�X,�on�the�same�side�of�G�as�F����4�u//���now�we�need�N�so�that�N0�=�RG����4�uMinusOne���=�IntersectLineCircle2(1,0,Circle(0,1))����4�uN���=�ExtendA(MinusOne,0,G,R)����4�ureturn���N����+I}�����G�color push gray 0����[�15������	color pop����Y!�s���������G�color push gray 0�����	color pop���[(��������+I�4src:980 ConstructiveGeometryFinalPreprintVersion.tex�The�=�p�A�oin��9t�here�is�that�w�e�do�not�need�to�assume�=��G��
�6�=�0�=�for�this�construction�to�w�ork;�Q�w�e�only����+Ineed���that���0�is�not�b�A�et��9w�een����G��and�1,��$i.e.�p�lo�A�osely�sp�eaking��G�N����0.�p�This���w��9orks�b�ecause�����p��erp��+��is����+Itotal.�������+I�8��H�zEuclidean�ffConstructiv���e�Geometry�ECG��阍�+I�4src:985 ConstructiveGeometryFinalPreprintVersion.tex�In�Jthis�Isection�w��9e�dev�elop�Ja�rst�order�axiomatic�theory�of�geometry�as�close�as�p�A�ossible�to����+IEuclid.�RW��:�e���call���it��ECG�,�for�\Elemen��9tary�Constructiv�e���Geometry".�REuclid�w��9ork�ed�with���the����+Ifollo��9wing��data��t�yp�A�es:���Point�,�=��Line�,�=��Cir��cle�,��Se�gment�,�and���A�Îr�c�.����-=�6����F��:�or�foundational�purp�A�oses,�=�it����+Iseems���simplest���to�use�only��Point�,���Line�,�and��Cir��cle�,�and���that�is�what�w��9e�do�in��ECG�.�W��:�e����+Itherefore�]Uc��9ho�A�ose�]Ta�m�ulti-sorted�theory��:�,�oUwith�]T\sorts"�corresp�A�onding�to�those�t��9yp�A�es.��sW��:�e�use�the����+Iw��9ords�"�\sort"�"�and�\t�yp�A�e"�synon�ymously�"�in�this�pap�A�er.�D�It�is,�&of�course,�&not�dicult�to�add�sorts�����+I�R��ay��=�*�,���	�A�Îr��c���$�and��t��Se��gment��'
��,��	and�t�axioms�making�the�extended�theories�t�conserv��|rativ��9e�o�v�er��ECG�,�but����+Iw��9e��,do��-not�do�so�here.��
W��:�e�tak��9e�function�sym��9b�A�ols�corresp�onding�to��-constructors�and�accessors�for����+Ithose�F}t��9yp�A�es,�R�describ�ed�F|in�detail�b�elo��9w.���The�relation�sym�b�A�ols�F|w�e�use�are�standard�F|in�axiomatic����+Igeometry��:�,�ju�B�Yn�for�Yo(strict)�b�A�et��9w�eenness�Ynand����F�for�equidistance.��W��:�e�emphasize�that��B��is�used�for����+Istrict��Sb�A�et��9w�eenness;�JSas�Hilb�A�ert�put�it,��if��C��;�is�b�et��9w�een��A��and��B�r��,��then��A�,���B��,��and��C��;�are�three����+Idistinct�Tp�A�oin��9ts.����8�C�5src:1004 ConstructiveGeometryFinalPreprintVersion.texW��:�e�Buse���on�����(�P�A�;���L�)�Cfor�the�incidence�of�p�oin��9t��P�P&�on�line��L�,�C�and���On���c�(�P�;���C����)�Cfor�the�incidence����+Iof��p�A�oin��9t��P��b�on���circle��C����.���There�is�a�complete�list�of�the�axioms�of��ECG��in�the�App�A�endix,���for����+Ireference.��In�$this�section,�J�w��9e�in�tro�A�duce�these�axioms�one�b�y�one,�J�with�discussion�and�explanation.����8�C�5src:1008 ConstructiveGeometryFinalPreprintVersion.tex�ECG�T�has�v��9e�basic�function�sym�b�A�ols,�sho�wn�here�with�argumen�ts:������������Interse��ctLines���A&�(�L;���K����)�������������Interse��ctLineCir�cle1����(�L;���C����)�������������Interse��ctLineCir�cle2����(�L;���C����)�������������Interse��ctCir�cles1���@�(�C� Z;���K����)�������������Interse��ctCir�cles2���@�(�C� Z;���K����)������+I�5src:1016 ConstructiveGeometryFinalPreprintVersion.texThe�P�in��9tuitiv�e�P�meaning�of�these�sym��9b�A�ols�has�b�A�een�discussed�ab�o��9v�e.����ECG��do�A�es�not�P�ha�v�e�\o�v�er-����+Iloaded"�Tv��9ersions�of�these�functions;�in�other�w�ords,�w�e�just�write������|'�Interse��ctLines���p�(��Line��`��(�A;���B�r��)�;���Line���r�(�P�A�;�Q�))����+I�5src:1018 ConstructiveGeometryFinalPreprintVersion.texinstead�Tof�ha��9ving�an�o�v�erloaded�v�ersion�of���Interse��ctLines��@o��that�tak�es�four�p�A�oin�ts.����8�C�5src:1021 ConstructiveGeometryFinalPreprintVersion.texOur�^underlying�logic�is�in��9tuitionistic.�
�W��:�e�]rst�giv�e�the�sp�A�ecically�in�tuitionistic�parts�of����+Iour���theory��:�,��whic��9h�are�v�ery���few�in�n�um�b�A�er.��mW��:�e�do�not���adopt�decidable�equalit�y��:�,��nor�ev�en����+Ithe�n%substitute�concept�of�\apartness"�in��9tro�A�duced�b�y�n&Heyting�(and�discussed�ab�A�o�v�e),��Yprimarily����+Ib�A�ecause���w��9e�aim�to�dev�elop�a�system�in�whic�h�denable�terms�(constructions)�denote�con�tin�uous����+Ifunctions,�|but��Aalso��@b�A�ecause�w��9e�wish�to�k��9eep�our�system�closely�related�to�Euclid's�geometry��:�,����+Iwhic��9h�}con�tains�nothing�lik�e�apartness.�%�Our�~rst�four�axioms�express�our�in�tuition�that�there�is����+Inothing��xasserting�existence�in�the�meaning�of�equalit��9y�or�incidence;��hence�assertions�of�equalit�y����+Ior�Tincidence�can�b�A�e�constructiv��9ely�pro�v�ed�b�y�con�tradiction.����8�C�5src:1031 ConstructiveGeometryFinalPreprintVersion.tex����s1����::�x����=��y��p���x��=��y���+F$�(Axiom�T1)�������::��V��(�A;���B�r�;�C� Z;�D�A��)��������(�A;���B�r�;�C� Z;�D�A��)���+F$(Axiom�T2)�������::��on��}��(�P�A�;���L�)�������on��s�(�P�;���L�)���+F$(Axiom�T3)�������::��On��
�!�(�P�A�;���C����)�������On��k��(�P�;���C����)���+F$(Axiom�T4)������G���color push gray 0��ff�ff�r�	J=�����"5��-:�6����LܻEuclid��also�w�Îork�ed��with��triangles,��tsquares,��up�<ren�tagons,�hexagons,��uand��\gures".�ƘBy�\gure"�he��seems�to�ha�Îv�e��	��mean�Ît,�\
\closed�Ap�<rolygon".�
lOne�cannot�w�ork�with�Aarbitrary�gures�without�in�tro�<rducing�v��ariables�for�in�tegers,�\
whic�h���in��the�mo�<rdern�view��tak�Îes�us�b�ey�Îond�geometry��J�.��W�e��therefore�view�those�theorems�of�Euclid��that�men�tion�\gure"�as���geometrical��Xtheorem��schemata�,�whic�Îh�result�in�a�theorem�ab�<rout�p�olygons�of��N��z�sides,�for�eac�Îh�particular�in�teger��N��"�.��ٛ�	color pop����G�color push gray 0����[�16������	color pop����i��s���������G�color push gray 0�����	color pop���[(��������8�C�5src:1040 ConstructiveGeometryFinalPreprintVersion.tex�W��:�e�Fwill�tak��9e�care�to�form�ulate�our�Eaxioms�without�quan�tiers�and�without�disjunction,����+Iwhic��9h��will��b�A�e�k�ey��to�our�applications�of�pro�A�of�theory��:�.���What�w��9e�aim�to�do�in�this�section�is����+Ito�O�form��9ulate�suc�h�a�O�theory��:�,�^ewhic�h�w�e�feel�is�quite�O�close�in�spirit�to�Euclid.���In�form�ulating�this����+Itheory��:�,���w��9e��Pmade��Quse�of�the�famous�axioms�of�Hilb�A�ert�[�11��	?�],���whic��9h�ha�v�e��Pb�A�een�giv�en��Pa�more�mo�A�dern����+Iand�z�detailed�z�form��9ulation�in�the�textb�A�o�ok�of�z�Green��9b�erg�[�9����].�L;Of�course,���w��9e�do�not�tak�e�z�the�full����+Icon��9tin�uit�y���axioms���of�Hilb�A�ert,��Vbut�only�the�line-circle�and�circle-circle�con��9tin�uit�y�axioms.�
�Where����+Ip�A�ossible,��cw��9e��-form�ulate��,our�axioms�as�correctness�statemen�ts��,ab�A�out�the�constructions;��in�that����+Iform�r�they�r�are�automatically�quan��9tier-free.�4�Some�axioms,��7whic�h�r�are�not�ab�A�out�constructions,����+Iare�n�inheren��9tly�n�quan�tier-free.�)KThe�only�n�question�of�serious�in��9terest�is�whether�disjunction�can����+Ib�A�e��rcompletely��qa��9v�oided.�/�It�can,��as�it��qturns�out.�The�details��qof�this�axiomatization�ma��9y�seem����+Isomewhat�ܧtedious,�|but�the�ܨsystem�m��9ust�of�course�b�A�e�sp�ecied�ܨin�complete�detail�in�order�to����+Iuse���it���in�metamathematical�pro�A�ofs.��Moreo��9v�er,�Ґsome�of���the�details,�as���far�as�I���can�determine,����+Iare�DHactually�new.��MIn�particular,�Pw��9e�DIsho�w�DHho�w�to�dene�DIthe�relations�\�AB�r�C��0�is�a�left�turn"�and����+I\�AB�r�C��K�is�da�crigh��9t�turn"�in�our�theory;��the�exp�A�erts�w��9e�consulted�though�t�dthis�w�as�new.��W��:�e�need����+Ithis�Tin�order�to�distinguish�the�t��9w�o�Tin�tersection�p�A�oin�ts�of�t�w�o�circles.����8�C�5src:1055 ConstructiveGeometryFinalPreprintVersion.texSince�x!man��9y�x"of�our�function�sym��9b�A�ols�denote�\partial�functions",���i.e.�D�functions�that�are�not����+Ialw��9a�ys�kdened,�Igw�e�will�use�the�\logic�of�partial�terms"��LPT�in�our�theories.��wThis�is�a�mo�A�dication����+Iof�(Drst-order�logic,�W�in�whic��9h�the�formation�(Erules�for�form�ulas�are�extended�b�y�(Eadding�the�follo�wing����+Irule:���If�I�t�I�is�a�term�then��t���#��is�a�form��9ula.���Then�in�addition�the�quan��9tier�rules�are�mo�A�died�so����+Iinstead�|�of�|��8�x�(�A�(�x�)�>x��>w�A�(�t�))�w��9e�ha�v�e��8�x�(�t�>w�#�^�A�(�x�)�>x���A�(�t�)),���and�instead�|�of��A�(�t�)���9�x���A�(�x�)�w��9e����+Iha��9v�e�T�A�(�t�)�8�^��t����#�9�x���A�(�x�).�pDetails�of�LPT�can�b�A�e�found�in�[�2����],�p.�97.����8�C�5src:1064 ConstructiveGeometryFinalPreprintVersion.texW��:�e�E�could�E�try�to�deal�with�partial�terms,�Q�suc��9h�as��������p���
�+�����aH>8�Z��x����9c�,�b��9y�simply�E�using�an�ordinary�function����+Isym��9b�A�ol��
for����	��p��p���
�[�,��Lbut�not��	sa�ying�an�ything�in��	the�axioms�ab�A�out�����p��p���
�e�of�negativ��9e�n�um�b�A�ers.�Th�us����
�����p���
�\����aH�Ÿ��1�������+Iw��9ould�{
some�real�n�um�b�A�er,���but�w�e�w�ould�not�kno�w�or�{care�whic�h�one,���and�w�e�w�ould�{not�b�A�e�able�to����+Ipro��9v�e�=�that�=�its�square�is���1.���This�approac��9h�rapidly�b�A�ecomes�a��9wkw�ard�=�when�complicated�terms����+Iin��9v�olving��`square��aro�A�ots�of�dieren��9t�quan�tities��`are�used,���and�y��9ou�m�ust��`add�extra�h��9yp�A�otheses�to����+Iev��9ery�(@theorem�asserting�(?that�what�is�under�ev�ery�square�(?ro�A�ot�is�p�ositiv��9e,�,�and�w�e�(?c�ho�A�ose�to�use����+I�LPT�T�instead.����8�C�5src:1071 ConstructiveGeometryFinalPreprintVersion.tex�LPT��T�includes�axioms��c���#��for�all�constan��9ts��c��of�an�y�theory��Uform�ulated�in��LPT�;�this�is�in����+Iaccordance�u�with�the�u�philosoph��9y�that�terms�denote�things,���and�while�terms�ma�y�u�fail�to�denote����+I(as�f�in�\the�King�f�of�F��:�rance"),�{there�is�no�suc��9h�thing�as�a�non-existen�t�thing.��Th�us�1�=�0�can�b�A�e����+Iundened,�Ti.e.�pfail�to�denote,�but�if�a�constan��9t��1��is�used�in��LPT�,�it�m�ust�denote�something.����8�C�5src:1076 ConstructiveGeometryFinalPreprintVersion.texThe��*meaning��+of��t����=��s��is��*that��t��and��s��are�b�A�oth�dened�and�they�are�equal.�

W��:�e�write��t�����=���������������=�����UP�s��to����+Iexpress�19that�1:if�one�of��t��or��s��is�dened,�83then�so�is�the�other,�83and�they�are�equal.�p!�LPT��con��9tains����+Ithe��gaxioms�of�\strictness",��0whic��9h�are�as�follo�ws�(for�eac�h�function�sym�b�A�ol��f���and�relation�sym�b�A�ol����+I�R�%��in�Tthe�language):�����������f����(�t�����1��*��;����:�:�:��
�;���t�����;�cmmi6�n��7�)����#��t�����1���m�#�^���:�:�:��c��^�8�t�����n�����#������������R�>�(�t�����1��*��;����:�:�:��
�;���t�����n��7�)������t�����1���m�#�^���:�:�:��c��^�8�t�����n�����#������8�C�5src:1084 ConstructiveGeometryFinalPreprintVersion.tex�The��7follo��9wing��6axioms�express�the�meaning�of�the�v��9e�main�function�sym��9b�A�ols�of��ECG�.�They����+Ido��8not,���ho��9w�ev�er,���mak�e��7an�y�assertions�of��7geometrical�con�ten�t,���nor�do��7they�distinguish�one�in��9ter-����+Isection�Tp�A�oin��9t�(of�a�line�and�circle,�or�of�t�w�o�circles)�from�the�other.�����G�color push gray 0����[�17������	color pop������s���������G�color push gray 0�����	color pop���[(������8�C�5src:1090 ConstructiveGeometryFinalPreprintVersion.tex����s1����P�ک�=�����Interse��ctLines��<��(�L;���K����)������on�(�P�A�;�L�)�8�^���on��
���(�P�A�;�K����)���<�(Axiom�T5)��������Interse��ctLines��:EJ�(�L;���K����)�����=����������Í�����=������UP�Interse��ctLines��F���(�K� Z;�L�)���<�(Axiom�T6)�������P�ک�=�����Interse��ctLineCir�cle1��U���(�L;���C����)�������on��s�(�P�A�;�L�)�8�^���On���Y�(�P�A�;�C����)���<�(Axiom�T7)�������P�ک�=�����Interse��ctLineCir�cle2��U���(�L;���C����)�������on��s�(�P�A�;�L�)�8�^���On���Y�(�P�A�;�C����)���<�(Axiom�T8)�������P�ک�=�����Interse��ctCir�cles1��G8)�(�C� Z;���K����)�������On��k��(�P�A�;�C����)�8�^���On���Y�(�P�A�;�K����)���<�(Axiom�T9)�������P�ک�=�����Interse��ctCir�cles2��G8)�(�C� Z;���K����)�������On��k��(�P�A�;�C����)�8�^���On���Y�(�P�A�;�K����)���7p�(Axiom�T10)��������on��}��(�P�A�;���L�)�8�^�:��on��}��(�P�;���K����)�������Interse��ctLines��<��(�L;�K����)��#���7p��(Axiom�T11)��������Interse��ctLines��:EJ�(�L;���K����)����#�^��on��}��(�P�A�;�L�)�8�^���on��
���(�P�A�;�K����)������P�ک�=���Interse��ctLines��<��(�L;�K����)���7p�(Axiom�T12)��������on��}��(�P�A�;���L�)�8�^���On���Y�(�P�;���C����)�������Interse��ctLineCir�cle1��U���(�L;�C����)��#���7p��(Axiom�T13)��������on��}��(�P�A�;���L�)�8�^���On���Y�(�P�;���C����)�������Interse��ctLineCir�cle2��U���(�L;�C����)��#���7p��(Axiom�T14)��������On��
�"�(�P�A�;���C����)�8�^���On���Y�(�P�;���K����)�������Interse��ctCir�cles1��G8)�(�C� Z;�K����)��#���7p��(Axiom�T15)��������On��
�"�(�P�A�;���C����)�8�^���On���Y�(�P�;���K����)�������Interse��ctCir�cles2��G8)�(�C� Z;�K����)��#���7p��(Axiom�T16)������J��+I�5src:1106 ConstructiveGeometryFinalPreprintVersion.texAxioms��to��distinguish�b�A�et��9w�een�the��t�w�o�in�tersection��p�A�oin�ts�in��eac�h�case��will�b�A�e�giv��9en�b�A�elo�w,��Tbut����+Ithat�Tm��9ust�a�w�ait�further�dev�elopmen�ts.����8�C�5src:1109 ConstructiveGeometryFinalPreprintVersion.texIn�(�order�to�(�rule�out�\degenerate"�lines,�-�w��9e�need�an�axiom�sa�ying�(�that�they�don't�exist;�2hbut����+Iw��9e�QS�do��allo�w�degenerate�circles.��nThe�follo�wing�axioms�also�pro�vide�for�QTlines�and�circles�to�exist����+Iwhen�Tthey�ough��9t�to.����8�C�5src:1114 ConstructiveGeometryFinalPreprintVersion.tex����s1�����Line��`��(�A;���B�r��)����#��$�����A��6�=��B���&�&�(Axiom�T17)��������Cir��cle����(�A;���B�r��)����#���&�&�(Axiom�T18)��������8�C�5src:1120 ConstructiveGeometryFinalPreprintVersion.texThere�2}are�functions�2~sym��9b�A�ols�corresp�onding�to�the�constructor�and�2~accessor�functions�for����+Ieac��9h�eoof�the�epsorts.��The�argumen�t�eoand�v��|ralue�t�yp�A�es�epof�these�sym�b�A�ols�are�epob�vious,�yvand�hence�not����+Isp�A�ecied�There.�pHere�are�the�axioms�(20�through�27)�relating�the�constructors�and�accessors.��4���8�C�5src:1125 ConstructiveGeometryFinalPreprintVersion.tex����s1�����Line��`��(��p��ointOn1��%��(�L�)�;�����p��ointOn2��(/��(�L�))���=��L���&�&�(Axiom�T19)�������A����6�=��B����^��8�p��ointOn1��' �(��Line��`��(�A;���B�r��))�=��A���&�&�(Axiom�T20)�������A����6�=��B����^��8�p��ointOn2��(��(��Line��`��(�A;���B�r��))�=��B���&�&�(Axiom�T21)��������p��ointOn1��%��(�L�)����6�=���p��ointOn2��)6��(�L�)���&�&(Axiom�T22)��������Cir��cle����(��c��enter���(�C����)�;�����p��ointOnCir�cle��;1`�(�C��))���=��C���&�&�(Axiom�T23)��������c��enter���(��Cir��cle����(�A;���B�r��))���=��A���&�&�(Axiom�T24)��������p��ointOnCir�cle��9���(��Cir��cle����(�A;���B�r��))���=��B���&�&�(Axiom�T25)��������c��enter���(�C����)����6�=���p��ointOnCir�cle��<8|�(�C��)���&�&(Axiom�T26)������7��8�C�5src:1137 ConstructiveGeometryFinalPreprintVersion.texF��:�or�Քreaders�unfamiliar�Փwith�the�logic�of�partial�terms,��w��9e�p�A�oin�t�out�that�ՓAxiom�20�could����+Iha��9v�e�Tb�A�een�written���������p��ointOn1��إ��(��Line��`��(�A;���B�r��))�����=����������Í�����=�����UP�A����+I�5src:1139 ConstructiveGeometryFinalPreprintVersion.tex�where��zthe��{relation��t�����=���Q������Í��Q��=�����
�&�s��means�that�if�either�side�is�dened,��Db�A�oth�sides�are�dened,��Dand�they����+Iare���equal.��7Also,��written���the�w��9a�y���it�is�written,�with�equalit��9y�instead�of�����=�����������=�����	�l,�Axiom�20�implies�that�����+I�Line��@r�(�A;���B�r��)�Tis�dened�when��A����6�=��B��,�Tmaking�half�of�Axiom�17�sup�A�er
uous.����8�C�5src:1143 ConstructiveGeometryFinalPreprintVersion.texThere�r�are�t��9w�o�r�incidence�relations,���for���on���7�(�P�A�;���L�)�for�p�oin��9ts�lying�on�lines,���and���On��L��(�P�;���C����)�for����+Ip�A�oin��9ts���lying�on�and�circles.��.There�are�three�constan�ts�����,�����x,�,���and��
�/�with�axioms�sa�ying�that�these����+Ithree�Tp�A�oin��9ts�are�non-collinear.�pSp�ecically�����8�C�5src:1148 ConstructiveGeometryFinalPreprintVersion.tex�����1����:��on��}��(���;�����Line���r�(��x,;���
����))���&�&(Axiom�T27)�������:��on��}��(��x,;�����Line���r�(���;���
����))���&�&(Axiom�T28)�������:��on��}��(�
���;�����Line���r�(���;����x,�))���&�&(Axiom�T30)���������8�C�5src:1155 ConstructiveGeometryFinalPreprintVersion.texW��:�e��do�not�ha��9v�e��to�sa��9y�explicitly�that������#�,��(b�A�ecause�it�is�part�of�the�logic�of�partial�terms�that����+Iev��9ery�Tconstan�t�is�dened{it�is�nothing�sp�A�ecial�to�an�y�particular�theory��:�.����+I�5src:1159 ConstructiveGeometryFinalPreprintVersion.texThe�Tother�axioms�of�incidence�are�����8�C�5src:1162 ConstructiveGeometryFinalPreprintVersion.tex�����1�����on��}��(�A;�����Line���r�(�A;���B�r��))���&�&(Axiom�T30)��������on��}��(�B�r�;�����Line���r�(�A;���B��))���&�&(Axiom�T31)��������on��}��(�P�A�;���L�)�8�^���on��
���(�Q;�L�)��^���on��
���(�R�>;���Line���r�(�P�A�;�Q�))�������on��s�(�R�>;�L�)���&�&(Axiom�T32)��������+I�5src:1169 ConstructiveGeometryFinalPreprintVersion.texW��:�e�Tdo�not�need���On���u�(�B�r�;���C���ir�A�cl�&9e�(�A;�B��))�Tb�A�ecause�that�will�follo��9w�axiom�44�b�elo��9w.�����G�color push gray 0����[�18������	color pop�������s���������G�color push gray 0�����	color pop���[(��������8�C�5src:1171 ConstructiveGeometryFinalPreprintVersion.tex�W��:�e�;&use�the�equalit��9y�;'sym�b�A�ol�;&b�et��9w�een�p�A�oin�ts�to�mean�\iden�tically�equal".���Bet�w�een�lines,����+Iequalit��9y���means�\in�tensional���equalit�y".��AIn�the�spirit���of�constructiv�e�mathematics,���lines�\come����+Iequipp�A�ed"�e�with�t��9w�o�e�asso�ciated�distinct�e�p�oin��9ts.��Th�us,��y��Line���W�(�A;���B�r��)�i=���Line��x1�(�P�;���Q�)�e�if�and�only�if����+I�A����=����B�>h�and�˴�P�
��=��Q�.�?�This�do�A�es�˳not�need�to�b�A�e�assumed,��Kas�it�follo��9ws�from�the�axioms�giv��9en����+Iab�A�o��9v�e��%for��$the�accessor�and�constructor�functions.���It�ma��9y��:�,��Yho�w�ev�er,��Xb�A�e��%confusing�to�those�not����+Iaccustomed�_to�`constructiv��9e�mathematics.��The�notion�of�\extensional�equalit��9y"�refers�to�the����+Idened��zrelation�b�A�et��9w�een��{t�w�o��zlines,��sthat�the�same�p�oin��9ts�are��{on�b�oth�lines.�
}In�practice,��sto�a��9v�oid����+Iconfusion,�{�w��9e�gqrarely�gpif�ev�er�men�tion�gpequalit�y�b�A�et�w�een�lines.��It�should�b�A�e�gpnoted,�{�ho�w�ev�er,�that����+Iour�~�theory�~�do�A�es�dep�end�on�this�~�view�of�lines,���since�the�order�of�the�t��9w�o�~�in�tersection�p�A�oin�ts�~�of�line�����+I�Line��@r�(�A;���B�r��)�,Wwith�circle��C��>�is�opp�A�osite�to�the�order�of�the�t��9w�o�,Vin�tersection�,Wp�oin�ts�,Wof���Line����(�B�r�;���A�)����+Iwith��
�C����.�k�In��other�w��9ords,���when�considering���Interse��ctLineCir�cle1��Vv��(�L;���C��),���it��is�essen��9tial�that��L��is����+Igiv��9en�Tb�y�t�w�o�p�A�oin�ts.����8�C�5src:1184 ConstructiveGeometryFinalPreprintVersion.texThe���basic�relations�of�our�theories�are�equidistance�and�b�A�et��9w�eenness,��jwhic�h���ha�v�e�b�A�een�recog-����+Inized�֩as�֪fundamen��9tal�at�least�since�Hilb�A�ert's�famous�1899�b�A�o�ok�֩[�11��	?�].��All�the�argumen��9ts�of�these����+It��9w�o�iBrelations�iCha�v�e�sort�iC�Point�.�;The�(strict)�b�A�et��9w�eenness�relation�iCis�written��B�(�a;���b;�c�).�<W��:�e�read����+Ithis���\�b��is���b�A�et��9w�een����a��and��c�".�HIt�implies�that�the�three�p�A�oin��9ts�are�collinear.�HThe�rst�b�et��9w�eenness����+Iaxiom�Tis����8�C�5src:1192 ConstructiveGeometryFinalPreprintVersion.tex����1����B�(�a;���b;�c�)������B�(�c;���b;�a�)���&�&(Axiom�T33)��������8�C�5src:1197 ConstructiveGeometryFinalPreprintVersion.texBefore�~giving�a�constructiv��9e�v�ersion�of�}the�remaining�b�A�et�w�eenness�axioms,��w�e�discuss�a����+Irelated�Tprinciple.�pBy�\Mark��9o�v's�Tprinciple�for�b�A�et��9w�eenness"�Tw�e�mean����8�C�5src:1201 ConstructiveGeometryFinalPreprintVersion.tex����1����::�B�(�A;���B�r�;�C����)������B�(�A;���B�;�C����)���&�&(Axiom�T34)��������8�C�5src:1205 ConstructiveGeometryFinalPreprintVersion.texMark��9o�v's���principle�expresses�the�idea�that�b��9y�computing�t�w�o�p�A�oin�ts��P���and��Q��etc.��to�greater����+Iand�bygreater�bzaccuracy��:�,�u�if�they�are�not�iden��9tical�w�e�will�bzev�en�tually�nd�bzthat�out.��W��:�e�w�an�t�bzit�to����+Ib�A�e�Trpro��9v��|rable�Tsin��ECG��for�sev��9eral�reasons:���it�is�used�in�Euclid�(e.g.���I.6�and�I.26,�d:as�is�discussed����+Ib�A�elo��9w);���it���is���needed�for�some�fundamen��9tal�theorems�(see�for�example�Lemma���??��
��);���and�it�mak��9es����+Ifor�@a�@smo�A�oth�metatheory��:�.��^While�some�ma��9y�consider�Mark��9o�v's�@principle�in�n��9um�b�A�er�@theory�to�b�A�e�of����+Iquestionable�
�constructivit��9y��:�,��w�e�
�consider�that�geometry�without�Mark��9o�v's�principle�
�is�a�wkw�ard.����+IAs��w��9e��shall�see�in�other�sections,�Euclid�do�A�es�use�it,�and�including�it�do�A�es�not�harm�our�abilit��9y����+Ito�ruconstruct�rvthings�that�are�pro��9v�ed�to�ruexist.�3�In�terms�of�order,���it�expresses�the�principle�that����+I�:�x������0����x�>��0.����8�C�5src:1214 ConstructiveGeometryFinalPreprintVersion.texHilb�A�ert's��9second�axiom�for�b�et��9w�eenness��9is,�r\giv�en��8three�distinct�p�A�oin�ts,�rone�and�only�one����+Iof��]the��\p�A�oin��9ts�is�b�et��9w�een��\the��]other�t��9w�o."���That�form�ulation��]is�to�A�o�strong,���constructiv��9ely��:�.���(F�or����+Iexample���its���translation�in��9to�the�recursiv��9e�plane�is�not�pro��9v��|rable�in���HA����plus�Mark��9o�v's�principle.)����+IInstead,�«w��9e���consider��the�follo�wing�v�ersion,�ªwhic�h�sa�ys�that�for���three�distinct�p�A�oin�ts,�ªif�t�w�o�of����+Ithe�Talternativ��9es�fail�then�the�third�m�ust�hold,�and�no�t�w�o�can�hold.��)��8�C�5src:1220 ConstructiveGeometryFinalPreprintVersion.tex����s1����a����6�=��b�8�^��a����6�=��c�8�^��b����6�=��c���T�()���"((Axiom�T35*)�������(�:�B�(�a;���b;�c�)�8�^�:�B�(�b;���c;�a�)������B�(�c;���a;�b�))�^��������(�:�B�(�b;���c;�a�)�8�^�:�B�(�c;���a;�b�)������B�(�a;���b;�c�))�^��������(�:�B�(�c;���a;�b�)�8�^�:�B�(�a;���b;�c�)������B�(�b;���c;�a�))�^�������:�(�B�(�a;���b;�c�)�8�^��B�(�b;���c;�a�))�8�^�:�(�B�(�a;���b;�c�)�8�^��B�(�b;���a;�c�))�^�������:�(�B�(�b;���c;�a�)�8�^��B�(�b;���a;�c�))������,���8�C�5src:1231 ConstructiveGeometryFinalPreprintVersion.texIn���the���presence�of�Mark��9o�v's���principle�for�b�A�et��9w�eenness���(Axiom�34),��w�e���can�w�eak�en���Axiom����+I35*���b��9y���replacing�the�unnegated�b�A�et��9w�eenness���form�ulas�b�y���their�double�negations.��W��:�e�obtain�the����+Ifollo��9wing:����8�C�5src:1235 ConstructiveGeometryFinalPreprintVersion.tex����s1����a����6�=��b�8�^��a����6�=��c�8�^��b����6�=��c�����&�&�(Axiom�T35)�������(�:�B�(�a;���b;�c�)�8�^�:�B�(�b;���c;�a�)�����::�B�(�c;���a;�b�))�^��������(�:�B�(�b;���c;�a�)�8�^�:�B�(�c;���a;�b�)�����::�B�(�a;���b;�c�))�^��������(�:�B�(�c;���a;�b�)�8�^�:�B�(�a;���b;�c�)�����::�B�(�b;���c;�a�))�^�������:�(�B�(�a;���b;�c�)�8�^��B�(�b;���c;�a�))�8�^�:�(�B�(�a;���b;�c�)�8�^��B�(�b;���a;�c�))�^�������:�(�B�(�b;���c;�a�)�8�^��B�(�b;���a;�c�))���������G�color push gray 0����[�19������	color pop�����ڠs���������G�color push gray 0�����	color pop���[(��������8�C�5src:1245 ConstructiveGeometryFinalPreprintVersion.tex�In���fact,�ۙAxioms�34�and�35�together�are�exactly�equiv��|ralen��9t�to�Axiom�35*,�as�the�follo��9wing����+Ilemma�X�sho��9ws.��/W��:�e�separate�Axiom�35*�in�to�X�t�w�o�axioms�to�facilitate�the�X�discussion�of�dieren�t����+Iv��9erions�Tof�Euclid's�parallel�p�A�ostulate,�whose�relations�dep�end�on�Mark��9o�v's�Tprinciple.���w����&I�color push gray 0��Lemma���6�	color pop���['1�5src:1249 ConstructiveGeometryFinalPreprintVersion.tex�Markov's�N<principle�for�b��etwe�enness�N<(Axiom�34)�is�pr��ovable�fr�om�Axiom�35*.���v��+I�5src:1252 ConstructiveGeometryFinalPreprintVersion.texPr��o�of�.���Supp�A�ose�Xs�::�B�(�a;���x;�b�).���W��:�e�w��9an�t�to�pro�v�e��B�(�a;���x;�b�).���F��:�rom��::�B�(�a;�x;�b�)�w��9e�immediately����+Iha��9v�e���a����6�=��b���and��a��6�=��x���and��x��6�=��b�.��By�Axiom��35*�it�suces�to�pro��9v�e��:�B�(�x;���a;�b�)��and��:�B�(�a;�b;�x�).����+IT��:�o��pro��9v�e��:�B�(�x;���a;�b�),���supp�A�ose��B�(�x;�a;�b�).��Then��:�B�(�a;�x;�b�),���b��9y�Axiom�35*.��But�that�con�tradicts����+I�::�B�(�a;���x;�b�).��That��cpro��9v�es��:�B�(�x;���a;�b�).��Similarly��bw��9e�ha�v�e��:�B�(�a;���b;�x�).��That�completes��bthe�pro�A�of����+Iof�Tthe�lemma.����8�C�5src:1258 ConstructiveGeometryFinalPreprintVersion.texThere�	is�
a�b�A�et��9w�eenness�	axiom�that�sa��9ys,��in�Hilb�A�ert's�form��9ulation,��that���Line��y��(�P�;���Q�)�	con��9tains����+Ia���p�A�oin��9t�b�et��9w�een����P���and��Q�.��rW��:�e�call���this�the�\densit��9y"�axiom.�Since�w��9e�w�an�t�a�quan�tier-free����+Iaxiomatization,�2�w��9e�-w�ould�-lik�e�to�sp�A�ecify�the�-p�oin��9t�asserted�to�exist.�c�The�natural�-candidate�for����+Ia��"p�A�oin��9t��#b�et�w�een��"�P�9�and��#�Q��is�the�result�of�Euclid's�segmen��9t-bisection�construction.�W��:�e�therefore����+Itak��9e�Tthe�follo�wing�axiom:��kҍ�8�C�5src:1265 ConstructiveGeometryFinalPreprintVersion.tex�����1����B�(�P�A�;�����Interse��ctLines��;���(�Line�(�P�;���Q�)�;�Line�(���&�&(Axiom�T36)�������$���Interse��ctCir�cles1��i�T�(�C���ir�A�cl�&9e�(�P�;���Q�)�;�C�ir�A�cl�&9e�(�Q;�P�H��))�;�������$���Interse��ctCir�cles2��i�T�(�C���ir�A�cl�&9e�(�P�;���Q�)�;�C�ir�A�cl�&9e�(�Q;�P�H��))))�;�Q�)������W���8�C�5src:1272 ConstructiveGeometryFinalPreprintVersion.texNote�*�that,�0Rin�view�*�of�the�strictness�axioms�of��LPT�,�this�axiom�implies�that�the�circles�and����+Iin��9tersection���p�A�oin�ts�in�v�olv�ed�are�dened.�Q�Similarly��:�,��there�is���a�b�A�et�w�eenness�axiom�that�asserts����+Ithat��l��Line��ͻ�(�A;���B�r��)�l�con��9tains�p�A�oin�ts�outside�the�segmen�t��AB�r��.��PAgain,���w�e�w�an�t�to�sp�A�ecify�suc�h�p�A�oin�ts����+Iso�Tthat�our�axiomatization�is�quan��9tier-free:���ҍ�8�C�5src:1277 ConstructiveGeometryFinalPreprintVersion.tex����s1����B�(��Interse��ctLineCir�cle1��R���(�Line�(�P�A�;���Q�)�;�C���ir�cl�&9e�(�P�;�Q�))�;�P�;�Q�)���&�&(Axiom�T37)�������B�(�P�A�;���Q;���Interse��ctLineCir�cle2��T|z�(�Line�(�P�;�Q�)�;�C���ir�cl�&9e�(�Q;�P�H��)))���&�&(Axiom�T38)��������8�C�5src:1282 ConstructiveGeometryFinalPreprintVersion.texThe�o�remaining�o�b�A�et��9w�eenness�axiom�is�o�called�the�\plane�separation�axiom".�+XT��:�o�mak��9e�its����+Istatemen��9t��Gmore��Hreadable,�Ԅw�e�in�tro�A�duce�the��Husual�denitions�of�t��9w�o�p�A�oin�ts��H�P��*�and��Q��b�A�eing�on����+Iopp�A�osite�Tsides�of,�or�on�the�same�side�of,�line��L�:����8�C�5src:1286 ConstructiveGeometryFinalPreprintVersion.tex����s1�����Opp��ositeSide��5H}�(�P�A�;���Q;�L�)���:=��B�(�P�;���Q;���Interse��ctLines��;���(��Line��`��(�P�;�Q�)�;�L�)���i�(Denition�T39)��������SameSide��(L�(�P�A�;���Q;�L�)���:=��:�B�(�P�;���Q;���Interse��ctLines��;���(��Line��`��(�P�;�Q�)�;�L�))���i�(Denition�T40)��������8�C�5src:1291 ConstructiveGeometryFinalPreprintVersion.texWhen��Vw��9e�use�the��Wsym�b�A�ol�:=,�Vw�e�mean�that��Wthe�sym�b�A�ol�on�the�left�is��Wregarded�as�an�ab-����+Ibreviation�Y�at�Y�the�meta-lev��9el,�j�rather�than�a�sym��9b�A�ol�of�the�formal�language.��}When�it�is�used�in����+Isubsequen��9t�Tform�ulas,�it�stands�for�the�formal�equiv��|ralen�t�giv�en�b�y�the�righ�t�hand�side.����8�C�5src:1295 ConstructiveGeometryFinalPreprintVersion.texNote�that�if���Line�����(�P�A�;���Q�)�do�es�not�meet��L�,���then�the�argumen��9t�of��B����is�undened,���so�b�y�the����+Istrictness�>Taxioms�>S�P��7�and��Q��are�on�the�same�side�of��L�.���This�form��9ulation,�iTho�w�ev�er,�iUdo�A�es�not�>Trequire����+Ius��]to�b�A�e�able��\to�decide�whether��L��is�or�is�not�parallel�to���Line��3%�(�P�A�;���Q�).�Using�these�denitions�w��9e����+Ican�Tgiv��9e�the�plane�separation�axiom(s):��kҍ�8�C�5src:1300 ConstructiveGeometryFinalPreprintVersion.tex����s1�����SameSide��(L�(�A;���B�r�;�L�)�8�^���SameSide��*Z8�(�B�;���C� Z;�L�)�������SameSide��*���(�A;���C�;�L�)���&�&(Axiom�T41)��������Opp��ositeSide��5H}�(�A;���B�r�;�L�)�8�^���Opp��ositeSide��7V��(�B�;���C� Z;�L�)�������SameSide��*���(�A;���C�;�L�)���&�&(Axiom�T42)�������Ӎ�+I�5src:1310 ConstructiveGeometryFinalPreprintVersion.tex�R��ays�.���Although��jw��9e�ha�v�e�not�included��ira�ys�in��ECG�,�w�e��ido�w�an�t�to�supp�A�ort�our��iclaim�that�a����+Iconserv��|rativ��9e���extension���including�ra�ys���can�easily�b�A�e�in��9tro�duced;��and���also,��%w�e�sometimes���mak�e����+Iinformal�pargumen��9ts�using�p
ra�ys�with�the�implication�that�p
they�can�b�A�e�formalized�in��ECG�.�W��:�e����+Ino��9w�|7sho�w�|8that�the�use�of�in��9tuitionistic�logic�do�A�es�not�cause�a�problem�ab�A�out�incidence�on����+Ira��9ys���or�segmen�ts.��Using�b�A�et�w�eenness,�5w�e�can�dene���incidence�for�ra�ys.��Ho�w�ev�er,�5there�is�a����+Itec��9hnicalit�y:��Cthe�{>origin��O���of��{=�R��ay��C�(�O�A�;���B�r��)�is�considered�{=to�lie�on�the�ra��9y��:�,���i.e.�N.\ra�ys�are�{>closed",����+Iwhile��	b�A�et��9w�eenness��means�\strictly�b�A�et�w�een."�
�It�is�th�us��easier�to�dene�the�\opp�A�osite�ra��9y":���P�(��is����+Ion�H�the�H�opp�A�osite�ra��9y�to���R��ay����(�O�A�;���B�r��)�if��P����is�on���Line�����(�O�;���B�r��)�H�and��O����is�b�A�et��9w�een�H��P����and��B�r��.��Then��Q����+I�is�Ton���R��ay���5�(�O�A�;���B�r��)�if�it�is�on���Line��v�(�O�;���B�r��)�but�not�on�the�opp�osite�ra��9y:��
�ҍ�8�C�5src:1320 ConstructiveGeometryFinalPreprintVersion.tex����1�����on��}��(�Q;�����R��ay��R��(�O�A�;���B�r��))���:=���on��s�(�Q;���Line���r�(�O�A�;�B�r��))�8�^�:�B�(�P�A�;�O�;�B�r��)���i�(Denition�T43)��������8�C�5src:1325 ConstructiveGeometryFinalPreprintVersion.texThis�f�denition�can�b�A�e�f�used�to�express�informal�argumen��9ts�ab�out�f�ra��9ys�in��ECG��without����+Ineeding�Tto�in��9tro�A�duce�an�explicit�sort�and�axioms�for�ra�ys.�����G�color push gray 0����[�20������	color pop�����=�s���������G�color push gray 0�����	color pop���[(��������+I�5src:1328 ConstructiveGeometryFinalPreprintVersion.tex�Se��gments�.�In��a��similar�w��9a�y�w�e��can�dene�incidence�for�segmen��9ts,���so�that�\segmen��9ts�are�closed".��f���T�v�O�A�n�(�P�;�����Se��gment��&#��(�Q;���R�>�))��$��	?��O�n�(�P�;���Line���r�(�Q;�R�>�))�8�^�:�B�(�P�A�;�Q;�R�>�)��^�:�B�(�Q;�R�>;�P�H��)����+I�5src:1330 ConstructiveGeometryFinalPreprintVersion.texRecall���that�����Se��gment��(5��(�R�>;���R��)�and���Line���W�(�R�;���R��)���are�b�A�oth�undened,��^so�the�fact�that��:�B�(�P�A�;���P�;�P�H��)����+Ido�A�es�5�not�5�mak��9e��P�~{�lie�on���Se��gment��'Κ�(�P�A�;���P�H��).�}=W��:�e�do�not�n��9um�b�A�er�5�this�denition,�=�since�it�is�not�used����+Iin�Tan��9y�further�axioms�or�pro�A�ofs.����8�C�5src:1333 ConstructiveGeometryFinalPreprintVersion.texThe�e�equidistance�e�relation�is�written���V��(�A;���B�r�;�C� Z;�D�A��).�
8W��:�e�will�e�often�express�this�using�the�in-����+Iformal���notation����AB�{�=����C���D�A��.��Axioms�for�equidistance�are�sometimes�called�\congruence�axioms"����+Isince�9�equidistance�can�9�b�A�e�though��9t�of�as�congruence�of�segmen��9ts.��JSometimes,�e�follo�wing�9�a�tradition����+Ithat�Tgo�A�es�bac��9k�to�Euclid,�w�e�write��AB�{�=����C���D�W�instead�of���V��(�A;���B�r�;�C� Z;�D�A��).����8�C�5src:1339 ConstructiveGeometryFinalPreprintVersion.texUsing�Tequidistance,�w��9e�dene�incidence�for�circles:��
����8�C�5src:1342 ConstructiveGeometryFinalPreprintVersion.tex����1�����on��}��(�P�A�;�����Cir��cle�����(�A;���Q�))��T�$��j���V��(�A;�P�;�A;�Q�)�8�^��A����6�=��Q���&�&�(Axiom�T44)������
����8�C�5src:1346 ConstructiveGeometryFinalPreprintVersion.texThe�K��A���6�=��Q�K��part�is�needed�to�a��9v�oid�con
ict�with�K�our�axiom�that���Cir��cle��T��(�A;���Q�)�is�undened����+Iif�T�A����=��Q�.����+I�5src:1348 ConstructiveGeometryFinalPreprintVersion.tex�R��emark�.��oAgain,��)w��9e���w�an�t���to�sho�w���that,��)if�desired,�arcs���can�b�A�e�correctly�handled�in�a�natural����+Iextension�Tof��ECG�:��f������1��on��=`-�(�P�A�;�����A�Îr��c��&��(�A;���C� Z;�Q�))������Ծ:=�������H�O�A�n�(�P�;�����Cir��cle�����(�C� Z;���A�))�#��^���V��(�A;�C� Z;�C�;�Q�)����������H�^��T�on����(��Interse��ctLines��:EI�(��Line��`��(�A;���Q�)�;���Line���r�(�C� Z;�P�H��))�;���Se��gment��&#��(�A;�Q�))������+I�5src:1354 ConstructiveGeometryFinalPreprintVersion.texIf��Yan��9y��Xt�w�o�of�the��Xthree�p�A�oin�ts��A�,�Y�C����,�Zand��Q��are��Xequal,�w��9e�ha�v�e��Xundened�terms�b�A�oth�on�the����+Ileft���and�the�righ��9t.��ZNote�that�this�denition�mak�es�arcs�\closed",��6in�that��A��and��Q��will�b�A�e�on�����+I�A�Îr��c��<�o�(�A;���C� Z;�Q�),�Tb�A�ecause�segmen��9ts�are�closed.�pNo�further�use�of�arcs�will�b�e�made�in�this�pap�er.����8�C�5src:1358 ConstructiveGeometryFinalPreprintVersion.texGreen��9b�A�erg's�Y;rst�Y<congruence�axiom,�j5paraphrased�from�[�9����],�j5is�closely�related�to�the�uniform����+Iv��9ersion�Tof�Euclid's�Bo�A�ok�I,�prop�osition�2:��f���=���A����6�=��B����^�8�C�5��6�=��D�Ӎ��9�R�>�(��on��}��(�R�;�����R��ay��R��(�A;���B�r��))��^���V��(�A;�R�>;�C� Z;�D�A��)���(not�Tan�axiom�of��ECG�)�����+I�5src:1361 ConstructiveGeometryFinalPreprintVersion.texThis�G8axiom�p�A�ermits�us�to�G7\la��9y�o��q"�segmen�t��C���D����along���R��ay���(�A;���B�r��).��Since�w�e�are�seeking�a����+Iquan��9tier-free�Taxiomatization,�w�e�w�an�t�to�sp�A�ecify�the�p�oin��9t��R�>�.�pThis�w�e�do�b�y�taking����mñ�R�>�(�A;���B�r�;�C� Z;�D�A��)���=���Interse��ctLineCir�cle1��U���(�A;���B�;���Cir��cle3���J�(�A;�C� Z;�D�A��))�:����+I�5src:1365 ConstructiveGeometryFinalPreprintVersion.tex�Our�Tv��9ersion�of�the�axiom�is�th�us��f���M���A����6�=��B����^�8�C�5��6�=��D�Ӎ���(��on��}��(�R�>�(�A;���B�r�;�C� Z;�D�A��)�;���R��ay��R��(�A;�B��))�8�^���V��(�A;���R�>�(�A;�B�;�C� Z;�D�A��)�;�C�;�D�A��)����+I�5src:1367 ConstructiveGeometryFinalPreprintVersion.texNote�Uthat�Vthe�strictness�axioms�then�will�imply�that��R�>�(�A;���B�r�;�C� Z;�D�A��)�Uis�dened�when��A���6�=���B��
�and����+I�C�5��6�=����D�A��.�pThe�Tocial�v��9ersion,�with��R�%��replaced�b�y�is�denition,�is�Axiom�45:������Z��A����6�=��B����^�8�C�5��6�=��D�Ӎ���������mf��(��on��}��(��Interse��ctLineCir�cle1��R���(��Line��`��(�A;���B�r��)�;���Cir��cle3���J�(�A;�C� Z;�D�A��))�;���R��ay��R��(�A;�B��))�8�^��������mf���V��(�A;�����Interse��ctLineCir�cle1��T|z�(��Line��`��(�A;���B�r��)�;���Cir��cle3���J�(�A;�C� Z;�D�A��))�;�C�;�D�A��)�����f���+I�5src:1376 ConstructiveGeometryFinalPreprintVersion.texThe�Tsecond�congruence�axiom�is��
����8�C�5src:1379 ConstructiveGeometryFinalPreprintVersion.tex����1�����V��(�A;���B�r�;�C� Z;�D�A��)�8�^����(�A;���B�r�;�E��2;�F�H��)��������(�C� Z;���D�A�;�E��2;�F�H��)���&�&(Axiom�T46)��������+I�5src:1384 ConstructiveGeometryFinalPreprintVersion.texThe�rthird�congruence�axiom�can�b�A�e�though��9t�of�ras�sa�ying�that�addition�is�w�ell-dened�on�con-����+Igruence�Tclasses�of�segmen��9ts:��
����8�C�5src:1387 ConstructiveGeometryFinalPreprintVersion.tex����1����B�(�A;���B�r�;�C����)�8�^��B�(�P�A�;���Q;�R�>�)�8�^���V��(�A;���B�;�P�A�;�Q�)�8�^���V��(�B�;���C� Z;�Q;�R�>�)�������V��(�A;���C�;�P�A�;�R�>�)���*t�(Axiom�T47)��������+I�5src:1392 ConstructiveGeometryFinalPreprintVersion.texWith�j�the�j�aid�of�the�congruence�axioms�considered�so�far,��?w��9e�can�formalize�the�notion�\�AB��	<����+IC���D�A��".�[�This�*qis�*pdened�to�mean�that�the�p�A�oin��9t��B��r���-=�0��N��on�ra�y�*p�C���D�l7�suc�h�that�*p�C���B��r���-=�0���r�=����AB��&�lies�b�A�et�w�een����+I�C��<�and�T�D�A��.�pMore�formally��:�,�w��9e�ha�v�e�to�dene��B��r���-=�0��9��b�y�a�term:�����G�color push gray 0����[�21������	color pop�����s���������G�color push gray 0�����	color pop���[(��������8�C�5src:1397 ConstructiveGeometryFinalPreprintVersion.tex����1����AB�{<���C���D�Ӎ�:=��B�(�C� Z;�����Interse��ctLineCir�cle2��T|z�(��Line��`��(�C�;���D�A��)�;���Cir��cle3���J�(�C� Z;�A;�B�r��))�;�D��)���:rl(Denition�T48)��������+I�5src:1402 ConstructiveGeometryFinalPreprintVersion.texIn��constructiv��9e��mathematics,�ׄw�e�cannot��dene��x������y���as���x�<�y��Y�_�s��x��=��y�R��.��Instead,�ׄw��9e�dene���x����y����+I�as��+�:�y��p<���x�.��
W��:�riting�out��,the�denition�of��y�<���x��to��,mak��9e�the�denition�of��x����y���directly��:�,���w��9e�ha�v�e����8�C�5src:1406 ConstructiveGeometryFinalPreprintVersion.tex����1����C���D�Ӎ�����AB�{�:=��:�B�(�C� Z;�����Interse��ctLineCir�cle2��T|z�(��Line��`��(�C�;���D�A��)�;���Cir��cle3���J�(�C� Z;�A;�B�r��))�;�D��)���:�R(Denition�T49)��������+I�5src:1412 ConstructiveGeometryFinalPreprintVersion.texand�Tas�remark��9ed�earlier�w�e�also�use�the�traditional�abbreviation����8�C�5src:1415 ConstructiveGeometryFinalPreprintVersion.tex����1����AB�{�=����C���D�Ӎ�:=���V��(�A;���B�r�;�C� Z;�D�A��)���i�(Denition�T50)��������8�C�5src:1419 ConstructiveGeometryFinalPreprintVersion.texW��:�e�Qha��9v�e�Rnot�included�angles�as�a�fundamen��9tal�data�t��9yp�A�e.�%hInstead,�statemen�ts�Rab�out�Qangles����+Ican�_b�A�e�formalized�as�statemen��9ts�ab�out�t��9w�o�`non-collinear�_ra�ys�with�the�same�origin,�H�or�ab�A�out�three����+Idistinct���p�A�oin��9ts.�
�Angles�are�th�us�alw�a�ys�less���than���R��{there�is�no�suc�h�thing���as�a�\straigh�t�angle".����+IIn��Dthis��Ew��9e�follo�w�Green�b�A�erg�[�9����].��ABut�Green�b�A�erg��Etak�es�congruence�of�angles��Eas�a�fundamen�tal����+Inotion.�&HInstead,��w��9e�m�dene�m�it,�essen�tially�m�using�the�m�principle�SAS�m�to�do�so.�&HGiv��9en�three�p�A�oin��9ts����+I(though��9t��of�as�an�angle)��AB�r�C����,��and�three�p�A�oin�ts��P�H�QR�>�,��b�y�the�rst�congruence�axiom�w�e�can����+Ind�T��P��H�-=�0��OM�on�ra��9y��QP����and��R��>��-=�0����on�ra�y��QR�d��with��QP��H�-=�0�����=��J�B�r�A��and��QR��>��-=�0���O�=��B�r�C����.��]Then�w��9e�dene�angle����+I�AB�r�C�k��to�ǧb�A�e�Ǩcongruen��9t�to�angle��P�H�QR����if�and�only�if��AC�5��=����P��H�-=�0�����Q���-=�0�����.��This�is�a�6-ary�relation�b�A�et��9w�een����+Ip�A�oin��9ts.��-Note�M>that�it�can�M=b�e�expressed�in�quan��9tier-free,�[8disjunction-free�form,�since�M=�P��H�-=�0��G��and��R��>��-=�0�����+I�are�Tgiv��9en�b�y�terms.����8�C�5src:1428 ConstructiveGeometryFinalPreprintVersion.texF��:�rom���the���denition�of�congruence�it�follo��9ws�that�if��A���-=�0��ro�distinct�from��B�3]�lies�on���R��ay�����(�B�r�;���A�)����+Iand����C����-=�0��K��distinct�from��B�h��lies�on���R��ay�����(�B�r�;���C����),��then���angle��A���-=�0�����B�C����-=�0��K��is�congruen��9t�to�angle��AB�C����.��This����+Iis�vessen��9tially�the�wre
exivit�y�of�congruence�view�ed�was�relation�b�A�et�w�een�ra�ys;�kthat�is�wthe�rst�half����+Iof�“Green��9b�A�erg's�Congruence�Axiom�”5,�� whic�h�b�A�ecomes�unnecessary��:�.��The�”same�observ��|ration�can�b�e����+Iused���to���pro��9v�e�the���symmetry�of�angle�congruence.���The�second�half�of�Congruence�Axiom�5�is�the����+Itransitivit��9y��jof�congruence��kof�angles,��0whic�h�w�e�tak�e�as��kour�fth�congruence�axiom,��0form�ulated����+Iwith�Dqnine�Drp�A�oin��9t�v��|rariables�instead�of�three�angle�v��|rariables.���It�really�comes�do��9wn�to�the�transitivit��9y����+Iof�̬the�congruence�of�triangles.�8W��:�e�w��9an�t�̬to�sa��9y�that�if�angle��AB�r�C�p��is�congruen�t�to�triangle��P�H�QR����+I�and��triangle���P�H�QR��S�is�congruen��9t�to�triangle��U���V�8W���then�triangle��AB�r�C�D��is�congruen��9t�to�triangle����+I�U���V�8W�H��.�The��congruence�of��triangle��AB�r�C����and�triangle��P�H�QR�&�is�expressed�b��9y��AB�Z�=����P�H�Q���^���B�r�C�6��=����+I�QR�v�^�8�AC�5��=����P�H�R�>�.�pHence�Tthe�transitivit��9y�axiom�w�e�need�is����8�C�5src:1440 ConstructiveGeometryFinalPreprintVersion.tex����s1����AB�{�=����P�H�Q�8�^��B�r�C�5��=��QR�v�^��AC�5��=��P�H�R�v�^��P�Q����=��U���V�p�^�8�QR���=��V�W�W�^��P�H�R���=��U���V�������C~��(Axiom�T51)��������AB�{�=����U���V�p�^�8�B�r�C�5��=��V�W�W�^��AC�5��=��U���W��������8�C�5src:1446 ConstructiveGeometryFinalPreprintVersion.tex�W��:�e�y5w��9an�t�y6to�sho�w�y6that�congruence�of�angles�really�only�dep�A�ends�on�the�four�ra��9ys�in�v�olv�ed,����+Inot�@�on�@�the�six�p�A�oin��9ts.���T��:�o�that�end�w��9e�supp�A�ose�that��A���-=�0���J�distinct�from��B��7�lies�on���R��ay��c�(�B�r�;���A�)�and����+I�C����-=�0��qF�distinct��from���B��K�lies�on���R��ay���x�(�B�r�;���C����),�'angle��AB�C���is��congruen��9t�to�angle��P�H�QR�>�,�'and��P����-=�0��A�distinct����+Ifrom�}��Q�}��lies�on���R��ay��E��(�Q;���P�H��),���and��R��>��-=�0��?��distinct�from��R����lies�on���R��ay��E��(�Q;���R�>�).�U\W��:�e�wish�to�sho��9w�that����+Iangle��a�A���-=�0�����B�r�C����-=�0��Q�is�congruen��9t��bto�angle��P��H�-=�0�����QR��>��-=�0����.�ΙLet��P�����1��&	�on���R��ay���B�(�Q;���P�H��)�ha�v�e��b�B�r�A�1�=�2�QP�����1��*��,�4�and��R�����1�����+I�on���e�R��ay��ZF�(�Q;���R�>�)��dha��9v�e��e�B�r�C��=�b7�QR�����1��*��.���Then�angle��A���-=�0�����B�r�C����-=�0����is�congruen��9t�to�angle��AB�r�C�6L�(as�observ��9ed����+Iab�A�o��9v�e),���whic�h���is�congruen��9t�to�angle��P�H�QR����b�y���h�yp�A�othesis,���whic�h�is�congruen�t�to�angle��P��H�-=�0�����QR��>��-=�0����.����+IHence�Tb��9y�the�transitivit�y�of�congruence,�w�e�are�nished.����8�C�5src:1457 ConstructiveGeometryFinalPreprintVersion.texThe�o�sixth�o�congruence�axiom�is�the�SAS�o�criterion�for�triangle�congruence.��9With�our�denition����+Iof��Ncongruence��Mfor�angles,��Othis�axiom�is�pro��9v��|rable.���In�Green�b�A�erg's��Nsystem,��Oit�simply�serv��9es�in�place����+Iof�Ta�denition�of�angle�congruence.����8�C�5src:1461 ConstructiveGeometryFinalPreprintVersion.texGreen��9b�A�erg's���fourth�congruence�axiom�states�that�for�an�y�angle��B�r�AC�Ev�and�an�y�ra�y��A���-=�0�����B��r���-=�0���
�there����+Iis�%a�%unique�ra��9y��C����-=�0��z��on�a�giv��9en�side�of��A���-=�0�����B��r���-=�0��I��suc��9h�that�angle��B��r���-=�0��$|�A���-=�0�����C����-=�0��z��is�congruen��9t�to�angle��B�r�AC����.����+IOf��course�ra��9ys�are�not�necessary�here:��this�is�a�statemen�t�that�for�an�y�v�e�p�A�oin�ts�satisfying����+Icertain���conditions,��Hthere�exists���another�p�A�oin��9t��C����-=�0��"s�satisfying�a�certain�condition.�@T��:�o�mak��9e�precise����+Ithe�	�part�	�ab�A�out�\on�a�giv��9en�side�of��q",�G%w��9e�ha�v�e�	�to�men�tion�	�another�p�A�oin�t�	��P�R��not�on��A���-=�0�����B��r���-=�0��.x�and����+Idemand���that��P�H�C����-=�0����should�not�meet���Line����(�A���-=�0�����;���B��r���-=�0��$|�).��YT��:�o�express�this�in�a�quan��9tier-free�w�a�y��:�,��`w�e����+Ineed�Tto�construct�the�p�A�oin��9t��C����-=�0��k�in�question.�pThis�w�e�do�as�follo�ws:�����G�color push gray 0����[�22������	color pop�������s���������G�color push gray 0�����	color pop���[(����ۗ��8�C�5src:1470 ConstructiveGeometryFinalPreprintVersion.tex�����1����C�5��=����C���ir�A�cl�&9e�(�A���-=�0�����;���A;�B�r��)�������B��r���-=�00���	�=�����Interse��ctLineCir�cle1��U���(��Line��`��(�A���-=�0�����;���B��r���-=�0��$|�)�;�C����)����tso�T�B��r���-=�00��k��lies�on�the�ra��9y��A���-=�0�����B��r���-=�0��������K�����1���m�=�����Cir��cle3�� �f�(�B��r���-=�00��VC�;���B�r�;�C����)�������K�����2���m�=�����Cir��cle3�� �f�(�A���-=�00����;���A;�C����)�������C����-=�0���u�=�����Interse��ctCir�cles1��G8)�(�K�����1��*��;���K�����2���)������$h��+I�5src:1479 ConstructiveGeometryFinalPreprintVersion.texNo��9w��pthe��odesired�p�A�oin�t��C����-=�0���is��oone�of�the�in��9tersection�p�A�oin�ts�of��o�K�����1����and��K�����2��*��.���There�seems�to�b�A�e����+Ino��reason,�!based�on��the�axioms�giv��9en�so�far,�!wh��9y�these�circles�in��9tersect.�8�W��:�e�tak�e��as�our�fourth����+Icongruence�:naxiom,�C�the�:massertion�that�b�A�oth�their�in��9tersection�p�A�oin�ts�:mare�dened.���W��:�riting�it�out����+Iformally�Tw��9e�ha�v�e��#���8�C�5src:1485 ConstructiveGeometryFinalPreprintVersion.tex�����1����A����6�=��B����^�8�A��6�=��C�� �^��B�{�6�=��C�� �^��A���-=�0��C��6�=��B��r���-=�0��$|�^���8��(Axiom�T52)�������B��r���-=�00���	�=�����Interse��ctLineCir�cle1��U���(��Line��`��(�A���-=�0�����;���B��r���-=�0��$|�)�;�C����)�^�������K�����1���m�=�����Cir��cle3�� �f�(�B��r���-=�00��VC�;���B�r�;�C����))�8�^��K�����2���=�����Cir��cle3�� �f�(�A���-=�00����;���A;�C����))������������Interse��ctCir�cles1��D�d�(�K�����1��*��;���K�����2���)����#�^��Interse��ctCir�cles2��D�c�(�K�����1��*��;�K�����2���)����#�^��������Opp��ositeSide��5H}�(��Interse��ctCir�cles1��D�c�(�K�����1��*��;���K�����2���)�;���Interse��ctCir�cles2��F1
�(�K�����1���;�K�����2���)�;���Line���r�(�A���-=�0�����;�B��r���-=�0��$|�))��������+I�5src:1494 ConstructiveGeometryFinalPreprintVersion.texOf��course�one�can�use��few��9er�v��|rariables�and�more�complicated�terms�to�express�this�axiom,��elim-����+Iinating��the�v��|rariables��B��r���-=�00��VC�,�Q��K�����1��*��,�and���K�����2���,�Q�at�the��cost�of�h��9uman�legibilit�y��:�.�9No�w��one�of�the�t�w�o����+Iin��9tersection�#�p�A�oin�ts�asserted�#�to�b�A�e�dened�will�b�e�#�the�p�oin��9t�needed�to�#�v�erify�Green�b�A�erg's�fourth����+Icongruence���axiom,���since���the�last�line�of�Axiom�51�sa��9ys�that�the�t��9w�o�p�A�oin�ts���are�on�opp�A�osite�sides����+Iof�T�A���-=�0�����B��r���-=�0��$|�.����-=�7���G�Hence,�b��9y�Bet�w�eenness�Axiom�3,�one�of�them�is�on�the�same�side�as��P�H��.����8�C�5src:1502 ConstructiveGeometryFinalPreprintVersion.texHa��9ving�"dened��AB��<��C���D�A��,�%Ow�e�are�in�"a�p�A�osition�to�form�ulate�the�axioms�"of�line-circle�con�ti-����+In��9uit�y��:�.�)The��rst�t��9w�o��of�these�just�tell�us�when�a�line�and�circles�in��9tersect{namely��:�,��when�there�is����+Ia�Tp�A�oin��9t�on�the�line�closer�(or�equally�close)�to�the�cen�ter�than�the�radius�of�the�circle.�����8�C�5src:1507 ConstructiveGeometryFinalPreprintVersion.tex����s1����AP�ک�����AB����^��8�on��
���(�P�A�;���L�)�����Interse��ctLineCir�cle1��U���(�L;���Cir��cle�����(�A;�B�r��))��#���&�&�(Axiom�T53)�������AP�ک�����AB����^��8�on��
���(�P�A�;���L�)�����Interse��ctLineCir�cle2��U���(�L;���Cir��cle�����(�A;�B�r��))��#���&�&�(Axiom�T54)��������+I�5src:1514 ConstructiveGeometryFinalPreprintVersion.texOur��bnext��aaxiom�sa��9ys�that�the�in��9tersection�p�A�oin�ts��adep�end��b\extensionally"�on�the�circle.��That����+Iis,�
Dif�?t��9w�o�@circles�con��9tain�the�same�p�A�oin��9ts�(whic�h�@is�guaran��9teed�if�they�ha��9v�e�@the�same�cen��9ter�and����+Iradius),��mthen��stheir��rrst�and�second�in��9tersection�p�A�oin�ts�with��ran�y�line�are��rthe�same.���Note�that�the����+Iin��9tersection��jp�A�oin�ts�dep�A�end��k\in�tensionally"�on�the�line,��b�A�ecause�the�rst��kand�second�in�tersection����+Ip�A�oin��9ts��`of���Line���(�(�A;���B�r��)��awith�circle��C�*H�are�the�second�and�rst�in��9tersection�p�A�oin�ts��`of���Line���(�(�B�r�;���A�)����+Iwith�T�C����.�pBut�the�in��9tersection�p�A�oin�ts�dep�A�end�extensionally�on�the�circle:�����8�C�5src:1523 ConstructiveGeometryFinalPreprintVersion.tex�����1����A����=���c��enter�����(�C����)�8�^��A����=���c��enter���(�K����)�8�^���On���Y�(�P�A�;���C��)��^���On���Y�(�Q;���K��)��^��AP�ک�=����AQ�����/�>�(Axiom�T55)��������Interse��ctLineCir�cle1��R���(�L;���C����)�����=����������Í�����=������UP�Interse��ctLineCir�cle1��_G �(�K� Z;�C��)�^��������Interse��ctLineCir�cle2��R���(�L;���C����)�����=����������Í�����=������UP�Interse��ctLineCir�cle2��_G �(�K� Z;�C��)���������+I�5src:1531 ConstructiveGeometryFinalPreprintVersion.texW��:�e�Tnext�giv��9e�the�basic�axioms�ab�A�out�in�tersections�of�t�w�o�circles.����8�C�5src:1534 ConstructiveGeometryFinalPreprintVersion.tex����s1�����On��
�"�(�P�A�;���C����)�8�^��AP�ک�����AB�{����3��(Axiom�T56)���������Interse��ctCir�cles1��W&\�(�C� Z;�����Cir��cle�����(�A;���B�r��))����#�^��T�Interse��ctCir�cles2��G���(�C�;�����Cir��cle�����(�A;���B�r��))��#��������+I�5src:1540 ConstructiveGeometryFinalPreprintVersion.tex�Our��next��axiom�sp�A�ecies�that�the�in��9tersection�p�A�oin�ts��of�t��9w�o�circles��dep�A�end�extensionally�on�the����+Icircles:��!���8�C�5src:1543 ConstructiveGeometryFinalPreprintVersion.tex�����1����A����=���c��enter�����(�C�����1��*��)�8�^��A����=���c��enter���(�C�����1��*��)�8�^���On���Y�(�P�A�;���C�����1���)��^���On���Y�(�Q;���C�����1���)��^��AP�ک�=����AQ�����;^�(Axiom�T57)��������Interse��ctCir�cles1��D�d�(�C�����1��*��;���K����)�����=����������Í�����=������UP�Interse��ctCir�cles1��P���(�C�����2���;�K����)�^��������Interse��ctCir�cles2��D�d�(�C�����1��*��;���K����)�����=����������Í�����=������UP�Interse��ctCir�cles2��P���(�C�����2���;�K����)�^��������Interse��ctCir�cles1��D�d�(�K� Z;���C�����1��*��)�����=����������Í�����=������UP�Interse��ctCir�cles1��P���(�K�;�C�����2��*��)�^��������Interse��ctCir�cles2��D�d�(�K� Z;���C�����1��*��)�����=����������Í�����=������UP�Interse��ctCir�cles2��P���(�K�;�C�����2��*��)�^������'��8�C�5src:1552 ConstructiveGeometryFinalPreprintVersion.tex�W��:�e��	no��9w�come�to��the�expression�of�Euclid's�parallel�p�A�ostulate.�W�e�rst�dene���Par��al�x�lel��#I��(�L;���K����)����+Ifor�Tlines��L��and��K��<�to�mean�that�the�lines�do�not�meet:����������Par��al�x�lel���{X�(�L;���K����)���:=��:��Interse��ctLines��:EI�(�K� Z;�L�)��#��:��G�@�color push gray 0��ff�ff�r�	J=�����"5��-:�7����LܻW��J�e��Xdo�not�kno�Îw�if�it�is�necessary�to�tak�e�this�assertion�as�part�of�the�axiom.�� P�erhaps�it�can�b�<re�pro�v�ed.��ٛ�	color pop����G�color push gray 0����[�23������	color pop������s���������G�color push gray 0�����	color pop���[(��������+I�5src:1556 ConstructiveGeometryFinalPreprintVersion.tex�Of�Tcourse,�in�view�of�the�other�axioms�for���Interse��ctLines��=Z��,�w��9e�ha�v�e��{������_�Par��al�x�lel���0(�(�L;���K����)��$��	?�8�x�:�(�on�(�x;�L�)�8�^��on�(�x;���K��))�;����+I�5src:1558 ConstructiveGeometryFinalPreprintVersion.tex�but��the��form�w��9e�to�A�ok�as�the�denition�has�the�adv��|ran��9tage�of�b�A�eing�quan��9tier-free.��Most�mo�A�dern����+Itreatmen��9ts��of��geometry�form�ulate�the��parallel�axiom�in�this�w��9a�y:��if��t�w�o�lines���K���and��M���are����+Iparallel�Tto��L��through�p�A�oin��9t��p�,�then��K�5��=����M����.�pIn�sym�b�A�ols:��
���8�C�5src:1564 ConstructiveGeometryFinalPreprintVersion.tex����1�����Par��al�x�lel�� ���(�K� Z;���L�)�8�^���Par��al�x�lel��"��(�M�uV;�L�)��^��on�(�p;�K����)��^��on�(�p;�M����)������K�5��=��M����
�(Pla��9yfair's�TP�ostulate)������}���8�C�5src:1569 ConstructiveGeometryFinalPreprintVersion.texW��:�e��rcall��sthis�the�\Pla��9yfair's�p�A�ostulate",���or�for�short�just�\Pla��9yfair",���after�John�Pla�yfair,���who����+Ipublished�״it�in�1795,��although�(according�to�Green��9b�A�erg�[�9����],��p.��19)�it�w�as�referred�to�b�y�Pro�A�clus.����+IEuclid's�Tp�A�ostulate�5�is��������+I�color push gray 0��	color pop���AI�5src:1573 ConstructiveGeometryFinalPreprintVersion.tex�If��Ca��Bstr��aight�line�fal�x�ling�on�two�str��aight�lines�make�the�interior�angles�on�the�same����AIside���less���than�two�right�angles,��fthe�two�str��aight�lines,��fif�pr�o�duc�e�d�indenitely,��fme�et����AIon�N<that�side�on�which�ar��e�the�angles�less�than�the�two�right�angles.��}���8�C�5src:1579 ConstructiveGeometryFinalPreprintVersion.tex�W��:�e�s�do�s�not,���ho��9w�ev�er,�tak�e�s�this�s�form�of�Euclid's�parallel�p�A�ostulate�as�an�axiom.�8^Instead�w��9e����+Itak��9e�Tthe�follo�wing�axiom:�����8�C�5src:1582 ConstructiveGeometryFinalPreprintVersion.tex����s1����:��Interse��ctLines��:EI�(�K� Z;���L�)����#�^�on�(�p;�K����)�8�^��on�(�p;�M����)��^��M����6�=����K�5�����Interse��ctLines��<��(�L;�M����)��������](Axiom�T58,�the�parallel�p�A�ostulate)������}���+I�5src:1589 ConstructiveGeometryFinalPreprintVersion.texIn�s8other�s7w��9ords,���if��K� �is�parallel�to��L��through��P�H��,���then�an��9y�other�line��M�l�through��P���m��9ust�meet����+I�L�.���This�G�diers�G�from�Euclid's�v��9ersion�in�that�w��9e�are�not�required�to�kno�w�G�in�what��dir��e�ction�G��M����+I�passes�9uthrough��P�H��;�K�but�also�the�conclusion�9tis�w��9eak�er,�B}in�9uthat�it�do�A�es�not�sp�A�ecify��wher��e��M�2Y�m��9ust����+Imeet��a�L�.�ƖThe�relationships�b�A�et��9w�een��athese��`dieren�t�parallel�axioms�are�discussed��`in�section�13����+Ib�A�elo��9w�Tand�in�[�3����].����8�C�5src:1594 ConstructiveGeometryFinalPreprintVersion.texT��:�o�illustrate�our�reasons�for�including�Mark��9o�v's�principle�in��ECG�,�w��9e�exhibit�the�follo�wing����+Ilemma.��<�����&I�color push gray 0��Lemma���7�	color pop���['1�5src:1597 ConstructiveGeometryFinalPreprintVersion.tex�(In����ECG�)���Supp��ose�neither�p�oint��A����nor�p�oint��B���lies���on�line��T�H��.�D0(i)�If��A��is�not�on����+Ithe��2same�side�of��T���as��B�r��,���then��A��is�on�the�opp��osite�side�of��T��fr��om��B�r��.�t�(ii)�If��A��is�not�on�the����+Iopp��osite�N<side�of�line��T���fr�om��B�r��,�then��A��is�on�the�same�side�of��T���as��B��.��<���+I�5src:1604 ConstructiveGeometryFinalPreprintVersion.texPr��o�of�.�[�Ad��(i).�Supp�A�ose��A��is�not��on�the�same�side�of�line��T�ȵ�as��B�r��.�[�Then�segmen��9t��AB���not�not����+Imeets���T�H��.���By��Mark��9o�v's�principle�for��line�in�tersections��(whic�h�is�a��theorem�of��ECG�),�line��AB����+I�meets����T����in�some�p�A�oin��9t����q�R��.���Since�segmen�t����AB���not�not�meets��T�H��,��,p�A�oin��9t��q����is�not�not�in�segmen��9t��AB�r��.����+ISince����A����and��B�m��do�not�lie�on��T�H��,�*�q�M��is�not�not�b�A�et��9w�een����A��and��B�r��.��Then,�*b��9y�Mark�o�v's���principle�for����+Ib�A�et��9w�eenness,��3the���in�tersection���p�oin�t���of��AB����and��T�Ӷ�lies�b�et��9w�een����A����and��B�r��.�|�That�is,��3�A��is�on�the����+Iopp�A�osite�Tside�of��T�^7�from��B�r��,�pro��9ving�(i).����8�C�5src:1612 ConstructiveGeometryFinalPreprintVersion.texAd��)(ii).���Supp�A�ose�that��(�A��is�not�on�the�opp�A�osite�side�of�line��T���as��B�r��.���T��:�o�sho��9w��A��is�on�the����+Isame�/�side�of�/��T�x��as��B�r��,�69w��9e�m�ust�/�sho�w�that�segmen�t��AB��Z�do�A�es�/�not�meet��T�H��.�kbSupp�ose�that�segmen��9t����+I�AB����do�A�es�meet��T�H��.�3�Since��A��and��B����do�not�lie�on��T�H��,�the�p�A�oin��9t�of�in��9tersection��Q��is�not�equal�to��A����+I�or�"��B�r��.�D�Hence,�&$b��9y�Mark�o�v's�"�principle�for�b�A�et�w�eenness,�&$�Q��lies�"�b�A�et�w�een��A��and��B�r��.�D�Hence��A�"��and��B����+I�are�Ton�opp�A�osite�sides�of��T�H��,�con��9tradiction.�pThat�completes�the�pro�of�of�the�lemma.����8�C�5src:1618 ConstructiveGeometryFinalPreprintVersion.texW��:�e��no��9w��ha�v�e�giv�en��a�quan�tier-free,��ldisjunction-free��axiom�system�that�enables�us�to�v��9erify����+IGreen��9b�A�erg's��%axioms,� �after�translating�angles��$as�triples�of�p�A�oin��9ts.���But�this�axiom�system�has����+Ia��Ushortcoming:�NrIt�do�A�es�not�distinguish�whic��9h�of�the�t�w�o�in�tersection�p�A�oin�ts�of�t�w�o�circles�are����+Idened���b��9y�����Interse��ctCir�cles1��J��and���Interse��ctCir�cles2��G�6�,���and�w�e���ha�v�e�not�completely���sp�A�ecied�whic�h����+Iin��9tersection���p�A�oin�t���of�a�line�and�circle�is�whic��9h,��either,�although�w��9e�ha�v�e���done�so�when�the�line�is����+Ia��wdiameter��vof�the�circle,��
b��9y�sp�A�ecifying�that���Interse��ctLineCir�cle1��U�G�(�Line�(�A;���B�r��)�;�C���ir�cl�&9e�(�A;�B�r��))��vis��won����+Ithe�nsame�nside�of��A��as��B�r��.��Hence�the�axioms�giv��9en�up�to�no��9w�do�not�suce�to�pro��9v�e�the�ncon�tin�uit�y����+Iof���w�Interse��ctCir�cles1��JQ�and���v�Interse��ctCir�cles2��GS��.���T��:�o��wput�this�matter�another�w��9a�y�,��=the��vaxioms��wgiv�en�so����+Ifar�H�ha��9v�e�mo�A�dels�in�H�whic�h���Interse��ctLineCir�cle1��Y�"�and���Interse��ctCir�cles1��K7��are�discon�tin�uous;�bTindeed����+Iarbitrarily�s�discon��9tin�uous.�6�Giv�en�one�mo�A�del,��w�e�can�s�arbitrary�switc�h�some�of�s�the�v��|ralues�of�����+I�Interse��ctLineCir�cle1��~�(�P�A�;���Q�)�Tand���Interse��ctLineCir�cle2��V$�(�P�;���Q�),�and�w��9e�still�ha�v�e�a�mo�A�del.�����G�color push gray 0����[�24������	color pop����3��s���������G�color push gray 0�����	color pop���[(��������8�C�5src:1629 ConstructiveGeometryFinalPreprintVersion.tex�W��:�e��kwish�to�add�new��laxioms�in�suc��9h�a�w�a�y�that�the�constructions��ldened�b�y�these�terms�are����+Icon��9tin�uous.��%Our�csrst�approac��9h�to�this�w�as�based�cron�the�observ��|rations�that�explicit�mo�A�duli�of�con-����+Itin��9uit�y��1can��2b�A�e�dened�for�the�in��9tersection�p�A�oin�ts��2of�circles�and�lines,��9and�the�in��9tersection�p�A�oin�ts����+Iof���circles���and�circles.��=The�latter�in��9v�olv�e������e�=����������p���:����aH�^�Z�������>�,��)so�they�are�constructible�using�straigh��9tedge����+Iand���compass.���Ho��9w�ev�er,���the�axioms�required�are�neither�short�nor�elegan�t,���and�w�e�disco�v�ered�an����+Iaxiomatization��(that,���while�still�not�v��9ery�short,�do�A�es�qualify�as�\elegan��9t".�}�The�axiomatization����+Iis���based�on�a�concept�that�is�fundamen��9tal�to�computer�graphics,���although�it�do�A�es�not�receiv�e����+Im��9uc�h�8�atten�tion�8�in�classical�geometry��:�.���Namely�,�A�the�concept�8�that�an�angle��P�H�QR�H��is�a�\left�turn"����+Ior�)
a�\righ��9t�turn".�W�Analytically��:�,�-�the�quan�tit�y�in�question�is�the�sign�of�the�v�ector�cross�pro�A�duct����+Iof��vthe��uv��9ectors��QP�*X�and��QR�>�.�&What�w��9e�will�exhibit�is�a�geometric�axiomatization�of�this�concept.����8�C�5src:1639 ConstructiveGeometryFinalPreprintVersion.texW��:�e��Ncould��Min��9tro�A�duce�a�new�predicate���L��eft���/�(�A;���B�r�;�C����),�؂but��Nthere�is�no�need�to�do�that,�؂since�w��9e����+Ican�Tjust�dene����8�C�5src:1643 ConstructiveGeometryFinalPreprintVersion.tex����s1�����L��eft�����(�A;���B�r�;�C����)���:=��C�5��=���Interse��ctCir�cles1��G8)�(��Cir��cle����(�A;���C��)�;���Cir��cle�����(�B�r�;�C��))���%�(Denition�T59)��������R�Îight����(�A;���B�r�;�C����)���:=��C�5��=���Interse��ctCir�cles2��G8)�(��Cir��cle����(�A;���C��)�;���Cir��cle�����(�B�r�;�C��))���%�(Denition�T60)��������+I�5src:1649 ConstructiveGeometryFinalPreprintVersion.texIf��5w��9e�did�in�tro�A�duce��6new�sym�b�A�ols�for���L��eft���K�and���R�Îight�� qn�then�these��6w�ould�b�A�e�axioms�instead�of����+Idenitions.�pW��:�e��Uturn�to�the��Taxioms�ab�A�out���L��eft��\��and��R�Îight�.�qRecall�that�����,��U��x,�,�and��
�M��are��U(constan��9t����+Isym��9b�A�ols�Tfor)�three�arbitrary�distinct�p�oin��9ts.�����8�C�5src:1654 ConstructiveGeometryFinalPreprintVersion.tex����s1�����L��eft�����(���;����x,;�
����)���&�&(Axiom�T61)��������R�Îight����(���;���
���;��x,�)���&�&(Axiom�T62)��������+I�5src:1660 ConstructiveGeometryFinalPreprintVersion.texThese��axioms��arbitrarily�sp�A�ecify�the�orien��9tation�of�the�plane.�9The�next�axiom�sa��9ys�that�\hand-����+Iedness"�Tis�a�prop�A�ert��9y�of�the�ra�ys�in�v�olv�ed,�not�just�the�p�A�oin�ts:����8�C�5src:1664 ConstructiveGeometryFinalPreprintVersion.tex����s1����P�ک�6�=����P��H�-=�0����^�8�R���6�=��R��>��-=�0���=�^��on�(�P��H�-=�0�����;�����R��ay��R��(�Q;���P�H��))��^���on��
���(�R��>��-=�0����;���R��ay��R��(�Q;�R�>�))�^���&�&�(Axiom�T63)���������L��eft��%G��(�P�A�;���Q;�R�>�)�������L��eft��Y��(�P��H�-=�0�����;���Q���-=�0�����;�R����-=�0����)���������+I�5src:1670 ConstructiveGeometryFinalPreprintVersion.texThe���next���axiom�sa��9ys�that�if��P�H�QR�5�is�a�left�turn,�.and�w��9e�mo�v�e����P�>��(in�an��9y�direction)�without����+Icrossing�Tline��QR�>�,�it�is�still�a�left�turn:����8�C�5src:1673 ConstructiveGeometryFinalPreprintVersion.tex����s1�����L��eft�����(�P�A�;���Q;�R�>�)�8�^�:�B�(�P�;�����Interse��ctLines��;���(��Line��`��(�Q;���R�>�)�;���Line���r�(�P�;�P��H�-=�0�����))�;�P��H�-=�0���)���&�&(Axiom�T64)�������������L��eft��Y��(�P��H�-=�0�����;���Q;�R�>�)��������+I�5src:1679 ConstructiveGeometryFinalPreprintVersion.texand�Tsimilarly�if�w��9e�mo�v�e��R�%��without�crossing��P�H�Q�:����8�C�5src:1682 ConstructiveGeometryFinalPreprintVersion.tex����s1�����L��eft�����(�P�A�;���Q;�R�>�)�8�^�:�B�(�R�;�����Interse��ctLines��;���(��Line��`��(�Q;���P�H��)�;���Line���r�(�R�;�R����-=�0����))�;�P��H�-=�0�����)���&�&(Axiom�T65)�������������L��eft��Y��(�P�A�;���Q;�R��>��-=�0����)��������+I�5src:1688 ConstructiveGeometryFinalPreprintVersion.texT��:�ogether�.�these�.�axioms�p�A�ermit�us�to�rotate�the�sides�of�a�left�turn��P�H�QR�>��as�long�as�they�do�not����+Icoincide�Tor�b�A�ecome�opp�osite,�and�it�remains�a�left�turn.����8�C�5src:1691 ConstructiveGeometryFinalPreprintVersion.texThe�Tnext�axiom�p�A�ermits�us�to�p�erform�a�translation:����8�C�5src:1694 ConstructiveGeometryFinalPreprintVersion.tex����s1�����L��eft�����(�A;���B�r�;�C����)�8�^��AB�{�=����P�H�Q��^��B�C�5��=����QR�>�^���&�&�(Axiom�T66)�������AC�5��=����P�H�R�v�^�8�AP�ک�=��B�r�Q��^��AP�ک�=��C���R������L��eft��Y��(�P�A�;���Q;�R�>�)��������+I�5src:1700 ConstructiveGeometryFinalPreprintVersion.texThose�Tare�all�the�axioms�for���L��eft���5�.�pHere�are�similar�axioms�for���R�Îight���X�:��4��8�C�5src:1703 ConstructiveGeometryFinalPreprintVersion.tex����s1����P�ک�6�=����P��H�-=�0����^�8�R���6�=��R��>��-=�0���=�^��on�(�P��H�-=�0�����;�����R��ay��R��(�Q;���P�H��))��^���on��
���(�R��>��-=�0����;���R��ay��R��(�Q;�R�>�))�^���'>��(Axiom�T67)���������R�Îight��+.��(�P�A�;���Q;�R�>�)�������R�Îight��@��(�P��H�-=�0�����;���Q���-=�0�����;�R����-=�0����)��������R�Îight����(�P�A�;���Q;�R�>�)�8�^�:�B�(�P�;�����Interse��ctLines��;���(��Line��`��(�Q;���R�>�)�;���Line���r�(�P�;�P��H�-=�0�����))�;�P��H�-=�0���)���'>�(Axiom�T68)�������������R�Îight��@��(�P��H�-=�0�����;���Q;�R�>�)��������R�Îight����(�P�A�;���Q;�R�>�)�8�^�:�B�(�R�;�����Interse��ctLines��;���(��Line��`��(�Q;���P�H��)�;���Line���r�(�R�;�R����-=�0����))�;�P��H�-=�0�����)���'>�(Axiom�T69)�������������R�Îight��@��(�P�A�;���Q;�R��>��-=�0����)��������R�Îight����(�A;���B�r�;�C����)�8�^��AB�{�=����P�H�Q��^��B�C�5��=����QR����^���'>��(Axiom�T70)��������AC�5��=����P�H�R�v�^�8�AP�ک�=��B�r�Q��^��AP�ک�=��C���R������R�Îight��@��(�P�A�;���Q;�R�>�)��������8�C�5src:1714 ConstructiveGeometryFinalPreprintVersion.texRep�A�eated��^applications��]of�Axioms�64�and�65�p�A�ermit�us�to�p�A�erform�an�arbitrary�rotation�on�a����+Ileft�.�turn��P�H�QR�>�,�5'preserving�the�fact�that�it�.�is�a�left�turn.�h�The�reader�who�wishes�to�understand����+Ithe�o<motiv��|ration�for�o;these�axioms�ab�A�out���R�Îight���{�and���L��eft���Y�should�see�Lemma�8�and�its�pro�A�of,��tb�elo��9w.����8�C�5src:1718 ConstructiveGeometryFinalPreprintVersion.texWith�h.the�aid�h-of���R�Îight��`�and���L��eft��0�,�|�w��9e�dene�the�concept�\�P���and��Q��ha��9v�e�h.the�same�order�on����+Iline��[�L��Z�as��A��and��B�r��",�ȍconstructiv��9ely�and�without�needing�case�distinctions.��rOf�course,�ȍw��9e�assume�����G�color push gray 0����[�25������	color pop����NR�s���������G�color push gray 0�����	color pop���[(��������+I�P��B�6�=�=^�Q�|I�and��A�=_�6�=��B�r��.�QPFirst,��w��9e�construct�p�A�oin�t�|J�E��{�suc�h�that��AB�r�E��{�is�a�left�turn.�QPBy�Denition����+I59,��~�E���=�����Interse��ctCir�cles1��G8)�(��Cir��cle����(�A;���B�r��)�;���Cir��cle�����(�B�;�A�))��His�suc��9h�a�p�A�oin�t.�lThen��I�P�++�and��Q��ha�v�e�the����+Isame�Torder�as��A��and��B��	�if�and�only�if���L��eft���5�(�P�A�;���Q;�E��2�).�pF��:�ormally�,����8�C�5src:1724 ConstructiveGeometryFinalPreprintVersion.tex�����1�����SameOr��der��0gI�(�A;���B�r�;�P�A�;�Q�)���:=���i�(Denition�T71)��������A����6�=��B����^�8�P�ک�6�=��Q��^���on��
���(�P�A�;�����Line���r�(�A;���B�r��))��^���on���(�Q;�����Line���r�(�A;���B�r��))�^���������L��eft��%G��(�P�A�;���Q;���Interse��ctCir�cles1��F1
�(��Cir��cle����(�A;�B�r��)�;���Cir��cle�����(�B�;�A�)))���������8�C�5src:1730 ConstructiveGeometryFinalPreprintVersion.texWith�����SameOr��der��7�.�in���hand,���it�is�easy�to�distinguish�the�t��9w�o���in�tersection���p�A�oin�ts�of�a�circle.����+IHo��9w�ev�er,�Tw�e�m�ust�b�A�e�careful�to�allo�w�for�the�case�when�the�t�w�o�in�tersection�p�A�oin�ts�coincide.����8�C�5src:1734 ConstructiveGeometryFinalPreprintVersion.tex�����1����P�ک�=�����Interse��ctLineCir�cle1��U���(��Line��`��(�A;���B�r��)�;�C����)�^���&�&�(Axiom�T72)�������Q����=���Interse��ctLineCir�cle2��U���(��Line��`��(�A;���B�r��)�;�C����)�8�^��P�ک�6�=����Q������������SameOr��der��2��(�A;���B�r�;�P�A�;�Q�)��������8�C�5src:1740 ConstructiveGeometryFinalPreprintVersion.texNext��w��9e�giv�e�the�remaining�axioms�for���Interse��ctCir�cles1��J�g�and���Interse��ctCir�cles2��G�e�.�UW��:�e�w�an�t�to����+Isa��9y�8�essen�tially�that�if��P��d�and��Q��are�the�8�t�w�o�in�tersection�p�A�oin�ts�of�circles��C��i�and��K��with�cen��9ters����+I�A���and��B�{��resp�A�ectiv��9ely��:�,�E�then��AB�r�P�Q��is�a�left�turn�and��AB�Q��is�a�righ��9t�turn.��JBut�there�is�also����+Ithe�04p�A�ossibilit��9y�that�03the�t�w�o�circles�are�03tangen�t,�6�and�the�03t�w�o�in�tersection�p�A�oin�ts�coincide.�mThen����+Ineither�PP�AB�r�P��2�nor��AB�Q�PO�is�a�left�or�righ��9t�turn.��cTherefore,�_instead�of�sa��9ying��AB�r�P��2�is�a�left�turn����+Iand��w�AB�r�Q��is��xa�righ��9t�turn,��w�e�sa�y�that��x�AB�r�P��Z�is�not�a�righ��9t�turn,��and��AB�r�Q��is�not�a�left�turn.����+IHere�Tare�the�axioms�in�question:����8�C�5src:1749 ConstructiveGeometryFinalPreprintVersion.tex����s1����R���=�����Interse��ctCir�cles1��G8)�(��Cir��cle����(�A;���P�H��)�;���Cir��cle�����(�B�r�;�Q�))�����:��R�Îight����(�A;���B�;�R�>�)���(�(Axiom�T73)�������R���=�����Interse��ctCir�cles2��G8)�(��Cir��cle����(�A;���P�H��)�;���Cir��cle�����(�B�r�;�Q�))�����:��L��eft�����(�A;���B�;�R�>�)���(�(Axiom�T74)��������+I�5src:1755 ConstructiveGeometryFinalPreprintVersion.texThis���completes�our���list�of�axioms�of��ECG�.�Note�that�these�axioms�are�all�quan��9tier-free�and����+Idisjunction-free.��W��:�e��will�consider�one�more�axiom,��whic��9h�do�A�es�con�tain�disjunction,��as�a�p�A�ossible����+Iaddition�Tto��ECG�,�in�the�next�section.������&I�color push gray 0��Lemma���8�	color pop���['1�5src:1759 ConstructiveGeometryFinalPreprintVersion.tex�(In�(��ECG�)�(�L��et�(��AB�r�C�̰�b�e�any�triangle.�ϨThen�we�c�an�determine�(�the�hande�dness�of����+Ithe���turn��AB�r�C����,��in�the�fol�x�lowing���sense.���L��et�����,���x,�,��and����
�Jm�b��e�the�thr�e�e���xe�d�non-c�ol�x�line�ar�p�oints����+Imentione��d��in��the�axioms�of���ECG��*��,�'{so�that�����x,
��Y�is�a�left�turn�by�denition.�/�L��et��L����=���Line����(���;����x,�)�.����+IThen�pwe�c��an�c�onstruct�a�p�oint�p�R��L�such�that��AB�r�C���is�a�left�turn�if�and�only�if��R��L�is�on�the�same����+Iside�N<of��L��as��
����,�and�a�right�turn�if�and�only�if��R�^z�is�on�the�opp��osite�side�of��L��fr�om��
����.����+I�5src:1765 ConstructiveGeometryFinalPreprintVersion.texPr��o�of�.���The�sidea�of�the�pro�A�of�is�this:��b��9y�a�series�of�\mo�v�es"�(applications�of�the�axioms�for�����+I�L��eft��@���and�� ��R�Îight��ϱ�,�#�whic��9h� �corresp�A�ond� �to�translations,�dilations,�and�rotations),�the� �triangle��AB�r�C����+I�is���\reduced"�to���either�triangle�����x,
�<�or�triangle����
����x,�,��preserving�the�handedness�of�the�turns;��wbut����+IAxioms��61�and��62�directly�sp�A�ecify�the�handedness�of�the�turns�����x,
��L�and���
����x,�.�>W��:�e��no��9w�giv�e�the����+Idetails.����8�C�5src:1768 ConstructiveGeometryFinalPreprintVersion.texBy�%a�\mo��9v�e",�(�applied�%to�%a�triple�of�non-collinear�p�A�oin��9ts��P�H�QR�>�,�w��9e�mean�%a�construction�of�a����+Inew��triple���U���V�8W�du�suc��9h�that,�"according�to�the�axioms�for���L��eft����and���R�Îight��ʗ�,�"if��P�H�QR�+��is�a�left�turn����+Ithen��Hso�is��G�U���V�8W�H��,��and�if��P�H�QR�ц�is�a�righ��9t�turn�then�so�is��U���V�8W�H��.�lThis�latter�condition�w��9e�describ�A�e����+Ifor��jshort��kb��9y�sa�ying�that��kthe�mo�v�e��k\preserv�es�handedness".��"The�axioms��kdescrib�A�e�sev�eral��kt�yp�A�es�of����+Imo��9v�es�_that�preserv��9e�handedness,�bsp�A�ecically��:�,�mo�ving�_�P�fB�along�`the�ra�y��QP�H��,�bmo�ving��R�-��along�the����+Ira��9y���QR�>�,�	}rotating��P�H�R���or��QR���in�suc�h�a�w�a�y�that�the�p�A�oin�ts��P�H��,�Q�,�	|and��R���nev�er�b�A�ecome�collinear,����+Iand���translating���the�whole�triple.��W��:�e�rst�sho��9w�that�the�n��9um�b�A�er���of�mo�v�es���(applications�of�these����+Iaxioms)�required��to�p�A�erform�a�giv��9en�rotation�is�b�A�ounded�b��9y�a�xed�constan��9t.�9rConsider�the����+Ifollo��9wing��Rpro�A�cedure:���First�mo�v�e��P��5�to�decrease��Qthe�angle�to�less�than�a�righ�t�angle.��oThen�w�e�can����+Irotate�)��P�H�QR�:!�b��9y�an�y�angle�)�up�to�a�righ��9t�angle,�/using�t��9w�o�)�mo�v�es�(one�)�mo�v�es��P�H�Q��and�)�one�mo�v�es����+I�QR�>�).��An��8arbitrary�rotation�can�b�A�e�p�erformed�b��9y�p�erforming�at�most�four�rotations�of�less�than����+Ininet��9y�~�degrees.�X�Th�us�in�~�four�or�few�er�~�rotations�(requiring�eigh�t�~�or�few�er�mo�v�es)�w�e�~�can�bring����+Ione�v'of�v(its�sides�on��9to�the�desired�(\target")�ra��9y��:�.�>�Then�w�e�v'can�mo�v�e�v'�P��
�b�y�the�v'same�amoun�t�v'as����+Iin�>�the�rst�step�>�resulting�in�the�desired�rotation.���Hence�ten�or�few��9er�applications�of�the�ab�A�o��9v�e����+Iaxioms�Tsuce�to�p�A�erform�an��9y�rotation.����8�C�5src:1780 ConstructiveGeometryFinalPreprintVersion.texNo��9w��consider��three�non-collinear�p�A�oin��9ts��P�H��,���Q�,��and��R�>�.�5W��:�e�giv��9e�a�pro�A�cedure�for�determining����+Iwhether�B��P�H�QR�R��is�a�left�turn�B�or�a�righ��9t�turn.��First�translate��P�H�QR�R��so�that��Q��coincides�with�the�����G�color push gray 0����[�26������	color pop����j��s���������G�color push gray 0�����	color pop���[(��������+I�p�A�oin��9t����k��(giv�en�b�y�a��constan�t�of���ECG�).�Then�rotate�it�so�that��P�<��lies�on���R��ay�����(��x,;������).�;Then�mo��9v�e����+I�P�Q�to�1����.��By�0the�axioms�ab�A�o��9v�e,�D�all�1these�steps�preserv��9e�the�handedness�of��P�H�QR�>�.��No�w,�D�if��R����+I�is���on���the�same�side�of���Line�����(���;����x,�)�as��
����,��(then�b��9y�Axioms�65�and�69,��(�P�H�QR��<�is�a�left�turn,��(since����+I����x,
��is���a�left�turn���b��9y�denition.��And�if��R����(after�the�mo�v�es���describ�A�ed)�is�on�the�opp�osite�side����+Iof���:�Line����(���;����x,�),���then��9w��9e��:claim�that��P�H�QR��x�is�a�righ��9t�turn.��!T��:�o�see�this,���let��
������-=�0��Ў�b�A�e�a�p�oin��9t��9on�the����+Isame�H�side�of��H��Line���V�(���;����x,�)�as��
����,�U\and�on�the�same�H�side�of���Line���U�(�Q;�R�>�)�as�����.��(Suc��9h�a�H�p�A�oin�t�can�b�A�e����+Iconstructed�tub��9y�bisecting�the�angle�formed�b�y��tv�R��ay��<W�(��x,;������)�and�the�opp�A�osite�ra�y�to���R��ay��<V�(�Q;���R�>�).)����+IThen�ځw��9e�ڀcan�mo�v�e��P�#c�to��
������-=�0����without�ڀc�hanging�the�handedness�ڀof��P�H�QR�>�,��Eand�then�w��9e�can�mo��9v�e�ځ�R����+I�to�bz��dB�without�c��9hanging�the�handedness�b{of��P�H�QR�>�.��But�no�w��P�H�QR�r��coincides�b{with��
����x,���,�u�whic�h�b�y����+Idenition�Tis�a�righ��9t�turn.�pThat�completes�the�pro�A�of�of�the�lemma.��
����+I�5src:1793 ConstructiveGeometryFinalPreprintVersion.tex�R��emark�.�~With��~classical�logic,���w��9e��could�pro�v�e�that��AB�r�C��f�is�either�a�righ�t�turn��or�a�left�turn.�~T��:�o����+Ireac��9h�e)that�conclusion,�yw�e�w�ould�e*need�to�kno�w�that�if�p�A�oin�t��R�ug�is�not�on��L���=���Line��w��(���;����x,�),�ythen����+Ieither�NS�R�^��is�NRon�the�same�side�of��L��as��
����or�on�the�opp�A�osite�side.��lThe�constructiv��9e�status�of�this����+Istatemen��9t�Tis�discussed�b�A�elo�w.��2�����&I�color push gray 0��Lemma���9�	color pop���['1�5src:1799 ConstructiveGeometryFinalPreprintVersion.tex�The��pr��e�dic�ates��R�Îight��0�(�A;���B�r�;�C����)���and��L��eft��H��(�A;�B�r�;�C����)��ar��e�denable��in�Gr�e�enb�er�g's�the�ory����+IG,��r��elative�to��an�arbitr�ary�choic�e�of���L�eft�����(���;����x,;�
����)���and��R�Îight�����(��;���
���;��x,�)���for�some��triple�of�non-����+Ic��ol�x�line�ar�N<p�oints�����,���x,�,�and��
����.�@This�c�an�even�b�e�done�with�intuitionistic�lo�gic.��2���+I�5src:1804 ConstructiveGeometryFinalPreprintVersion.texPr��o�of�.��It��will�suce�to�dene�the�relation��T�H��(�A;���B�r�;�C� Z;�P�A�;�Q;�R�>�)��with�the�meaning�\�AB�C����and����+I�P�H�QR�QR�ha��9v�e�Athe�same�handedness."���First�w��9e�note�that�it�is�p�A�ossible�to�dene�the�notion�of����+Ione��ztriangle�b�A�eing�a��ytranslation�of�another;��
namely��:�,����AB�r�C�Nb�is�a�translation�of��P�H�QR����if�the�t��9w�o����+Itriangles�j�are�congruen��9t�j�and��AP�iD�=� a�B�r�Q��=��C���R�>�.�!It�j�is�also�p�A�ossible�to�dene�the�notion�of��AB�r�C����+I�b�A�eing��a�rotation��of��P�H�B�r�Q��(when�the�t��9w�o��angles�share�v��9ertex��B�r��).��This�requires�t��9w�en�t�y��v��|rariables����+Ito��0express,��fso��/it�is�to�A�o�complex�to�write�do��9wn�in�telligibly��:�,��fbut��0the�denition�in�question�sa��9ys����+Ithere���exist�t��9w�en�t�y���p�A�oin�ts���represen�ting�ten�\mo�v�es"�according�to�the����ECG��axioms�for�rotations����+Igiv��9en��2ab�A�o�v�e.��dThen��AB�r�C�J�and��1�P�H�QR��p�ha�v�e�the�same��1handedness�if�there�exist��P��H�-=�0�����and��Q���-=�0��W��suc��9h�that����+I�P��H�-=�0�����B�r�R��>��-=�0�����is��a��translation�of��P�H�QR����and�there�is�a�rotation��P��H�-=�00��,q�B�r�R��>��-=�00��ً�of��P�H�QR����with��R��>��-=�00��ً�on���R��ay�����(�B�r�;���C����)����+Iand�T�P��H�-=�00��	A��is�on�the�same�side�of���Line��v�(�B�r�;���C����)�as��A�.������&I�color push gray 0��Theorem���2�	color pop���b�3�5src:1815 ConstructiveGeometryFinalPreprintVersion.texECG��f�with�classic��al��glo�gic��fis�e��quivalent�to�Gr��e�enb�er�g's��fsystem��G��with�only�line-cir��cle����+Iand�N<cir��cle-cir�cle�c�ontinuity.��2���+I�5src:1819 ConstructiveGeometryFinalPreprintVersion.texR��emark�.�qGreen��9b�A�erg's��Xsystem�is�not�completely�formal.�rBut�w�e�understand�here�a�w�eak�second-����+Iorder��xtheory��:�,���with�the�rst-order�v��|rariables�ranging�o��9v�er��yp�A�oin�ts,���and��xcircles,�lines,�segmen��9ts,�ra�ys,����+Iand��tangles�treated�as�sets�of�p�A�oin��9ts.����-=�8���	�w�Incidence��smeans�mem�b�A�ership.���There�is�a�w�eak�form�of����+Ithe�Tcomprehension�axiom,�only�for�form��9ulas�with�no�set�quan�tiers.��
����+I�5src:1827 ConstructiveGeometryFinalPreprintVersion.tex�Pr��o�of�.�ܱIt�Vsuces�to�deal�Vwith�the�p�A�oin��9t,�|Wline,�and�Vcircle�fragmen�t�of�V�ECG�,�since�the�whole�theory����+Iis��Cconserv��|rativ��9e�o�v�er�this�fragmen�t�(as�follo�ws�either�b�y�mo�A�del�theory�or�cut-elimination�from����+Ithe�h0fact�that�h/the�axioms�of�b�A�oth�theories�can�expressed�in�this�fragmen��9t).�First�w�e�h0sho�w�that����+Iset�nav��|rariables�n`are�irrelev�an��9t�n`in�Green�b�A�erg's�theory�n`G.�There�are�no�set�v��|rariables�in�the�axioms,����+Iexcept��uthose�ranging�o��9v�er��ulines,��ncircles,��osegmen�ts,�ra�ys,�and��uangles.�
{Hence,��oin�the�axioms,�those����+Iv��|rariables�q�can�b�A�e�replaced�with�v�ariables�of�q�the�corresp�A�onding�sorts�of��ECG�,�and�set�mem��9b�ership����+Ib��9y���the���appropriate�incidence�relations.��%By�Gen��9tzen's�cut-elimination�theorem,��Iif�a�form��9ula��A����+I�without�r�set�v��|rariables�has�r�a�pro�A�of�in���ECG����,��Uthen�there�is�a�cut-free�pro�of�r�of�a�sequen��9t��-��)��A�,����+Iwhere����is�a�conjunction�of�axioms.��RThis�en��9tire�pro�A�of�con��9tains�no�set�v��|rariables,��b��9y�the�cut-����+Ielimination��Dtheorem;��=hence�it�is�a��Epro�A�of�in�the�language�of��ECG�.�W��:�e�sho��9w�ed�ab�A�o�v�e��Dthat�the����+Iaxioms�_�of�_�G�_�(in��9terpreted�in��ECG��in�this�w��9a�y)�are�_�pro�v��|rable�in��ECG�.�_�That�is�one�direction�of����+Ithe�Tpro�A�of.����8�C�5src:1837 ConstructiveGeometryFinalPreprintVersion.texF��:�or��the��other�direction,���w��9e�dene�a�translation��A���-=�0��7��of��ECG��in��9to�G���as�follo�ws.��In���A���-=�0�����,���v��|rariables����+Io��9v�er���lines���and�circles�are�replaced�b��9y�set�v��|rariables,�restricted�to�appropriate�predicates�dening��G�
7
�color push gray 0��ff�ff�r�	J=�����"5��-:�8����LܻT��J�ec�Îhnically�,��in���Green�b�<rerg's���b�o�ok,��lines�are���primitiv�Îe�ob�x�jects�and�the�incidence�relation��on�(�P�.:oint;�j�Line�)�is�unde-��	��ned,�B{but�,�ra�Îys,�segmen�ts,�and�,�circles�are�sets�,�of�p�<roin�ts�and�the�incidence�relations�,�are�set�mem�b�<rership.��
See�,�p.�144���for��Xhis�explanation.��ٛ�	color pop����G�color push gray 0����[�27������	color pop�����v�s���������G�color push gray 0�����	color pop���[(��������+I�those��^concepts.��Incidence�relation�sym��9b�A�ols�are�replaced��_b�y�`��'.��F��:�or�eac�h��_term��t�(�x�����1��*��;����:�:�:��
�x�����n��7�)�of����+I�ECG�,�Fthere�is�a�form��9ula�F�G�����t��\p�(�y�R�;���x�����1��*��;��:�:�:��
�;�x�����n��7�)�Fthat�expresses��y��V�=����t�(�x�����1���;����:�:�:��
�;���x�����n��7�).���The�denition����+Iis���giv��9en���inductiv�ely�on�the���complexit�y�of��t�.�yUThe���most�dicult�part�of�this�is�to�in��9terpret����+Ithe�Nnfunction�sym��9b�A�ols�Nofor�the�t�w�o�in�tersection�p�A�oin�ts�Noof�t�w�o�circles.�ǾHo�w�ev�er,���the�dicult�y��:�,����+Iwhic��9h�`is�dening�the�concepts���L��eft���A�(�A;���B�r�;�C����)�aand��`�R�Îight���d�(�A;�B�;�C����)�`required�to�distinguish�the����+It��9w�o�Tin�tersection�p�A�oin�ts�of�t�w�o�circles,�has�b�A�een�tak�en�care�of�in�Lemma�9.����8�C�5src:1845 ConstructiveGeometryFinalPreprintVersion.texT��:�o�f+in��9terpret�f*the�function�sym�b�A�ols�f*for�the�in��9tersection�p�A�oin�ts�of�f*a�line�and�a�circle�w��9e�need����+Ito�
8dene�
9the�concept�\�P�V�and��Q��o�A�ccur�in�the�same�order�on���Line��n�(�A;���B�r��)�as��A��and��B���do."��That����+Ican���b�A�e�done�without�case�distinctions,���simply�b��9y�sa�ying���that�for�some��R�>�,����R�AB���is���a�left�turn����+Iand�T�R�>P�H�Q��is�a�left�turn.�pThat�completes�the�pro�A�of.�������+I�9��C^Euclid's�ffReasoning��阍�+I�5src:1852 ConstructiveGeometryFinalPreprintVersion.tex�Euclid's�Tpro�A�ofs�ha��9v�e�Tb�een�analyzed�in�detail�b��9y�Avigad��et.�N<al.�p�in�[�1����],�and�they�conclude:������+I�color push gray 0��	color pop���AI�5src:1855 ConstructiveGeometryFinalPreprintVersion.texEuclidean�`Vpro�A�ofs�do�little�`Umore�than�in��9tro�duce�ob���jects�satisfying�lists�`Uof�atomic�(or����AInegation��Satomic)�assertions,�ԇand�then�dra��9w�further�atomic��T(or�negation�atomic)�con-����AIclusions��xfrom�these,��qin�a�simple�linear�fashion.��There�are�t��9w�o��xminor�departures�from����AIthis���pattern.��;Sometimes�a�Euclidean�pro�A�of�in��9v�olv�es���a�case�split;��for�example,���if��ab����AI�and�[1�cd�[0�are�unequal�segmen��9ts,�l�then�one�is�longer�than�the�other,�l�and�one�can�argue����AIthat�qa�qdesired�conclusion�follo��9ws�in�either�case.�/�The�other�exception�is�that�Euclid����AIsometimes���uses���a��r��e�ductio�;���for�example,��'if���the�supp�A�osition�that��ab��and��cd��are�unequal����AIyields�Ta�con��9tradiction�then�one�can�conclude�that��ab��and��cd��are�equal.����8�C�5src:1866 ConstructiveGeometryFinalPreprintVersion.texIt���is�our���purp�A�ose�in�this�section�to�argue�that�Euclid's�reasoning�can�b�A�e�supp�orted���in��ECG�,����+Iincluding���the�t��9w�o���t�yp�A�es���of�apparen�tly�non-constructiv�e�reasoning�just�discussed.��<The�t�yp�A�e�of����+I�r��e�ductio��}�argumen��9t��~men�tioned�corresp�A�onds��~to�Mark�o�v's�principle��~�::�x�s�=��y�ŵ���x��=��y�R��,��Hwhic��9h�w�e����+Iha��9v�e���sho�wn���follo�ws�from���the�b�A�et�w�eenness���axioms�of��ECG�.�The�rst�t��9yp�A�e�(based�on�the�idea����+Ithat�M,if��ab��and��cd��are�unequal�then�M-one�of�them�is�longer)�will�b�A�e�studied�metamathematically��:�.����+IBut�Trst,�w��9e�giv�e�t�w�o�examples.����8�C�5src:1875 ConstructiveGeometryFinalPreprintVersion.texA�Tt��9ypical�example�of�suc�h�an�argumen�t�in�Euclid�is�Prop.�pI.6,�whose�pro�A�of�b�egins������+I�color push gray 0��	color pop���AI�5src:1877 ConstructiveGeometryFinalPreprintVersion.texLet�j��AB�r�C���b�A�e�a�triangle�ha��9ving�j�the�angle��AB�C���equal�to�the�angle�j��AC���B��.��I�j�sa��9y�that����AIthe��side���AB��l�is�also�equal�to�the�side��AC����.�,�F��:�or,�if��AB��l�is�unequal�to��AC����,�one�of�them����AIis�Tgreater.�pLet�AB�b�A�e�greater,���:���:�:�����+I�5src:1882 ConstructiveGeometryFinalPreprintVersion.tex�The��wsame��xpro�A�of�also�uses�an�argumen��9t�b�y��wcon�tradiction�in��xthe�form��::�x�lW�=�lV�y�����x��=��y�R��.���This����+Iprinciple,��'the��d\stabilit��9y��cof�equalit�y",��'is��cincluded�in��ECG�,�and�is�univ��9ersally�regarded�as�con-����+Istructiv��9ely��cacceptable.�i�The�conclusion��dof�I.6,��(ho�w�ev�er,��'is�negativ�e��d(has�no��9��or��_�),��'so�w��9e�can����+Isimply�fput�double�negations�fin�fron��9t�of�ev�ery�step,�zBand�apply�the�stabilit�y�fof�equalit�y�once�at����+Ithe�Tend.����8�C�5src:1888 ConstructiveGeometryFinalPreprintVersion.texProp.�wI.26�iis�janother�example�of�the�use�of�the�stabilit��9y�of�equalit��9y:��\��:���:�:��UPD�A�E����is�not�unequal����+Ito��.�AB�r��,��and�is�therefore��-equal�to�it."�a�While�w��9e�ha�v�e�pro�v�ed��-that�classical�argumen�ts��-can�b�A�e����+Ieliminated�A from�Apro�A�ofs�of�Euclid's�theorems,�Lin�fact�it�seems�that�the�only�classical�argumen��9ts����+Ithat��<�o��c�cur��in�Euclid�are��;applications�of�the�principle�\if��ab��and��cd��are�unequal�then�one�of�them����+Iis�Tlonger."����8�C�5src:1894 ConstructiveGeometryFinalPreprintVersion.texIn�G�the�G�examples�ab�A�o��9v�e,�p�this�principle�G�is�not�really�needed�to�reac��9h�Euclid's�desired�conclusion.����+ISince�-�the�conclusion�-�concerns�the�equalit��9y�of�certain�p�A�oin��9ts,�3�w�e�-�can�simply�double-negate�eac��9h����+Istep�
�of�
�the�argumen��9t,��and�then�add�one�application�of�the�stabilit��9y�of�equalit��9y�at�the�end.��The����+Idouble�_jnegation�of�\if��ab��and�_k�cd��are�unequal�then�one�is�longer"�is�pro��9v��|rable,�q�since�(in�tuitiv�ely����+Isp�A�eaking)�c��:�(�p�I<�q�R��)�is�c��q�f����p��and�c��p����q���^�Bl�q�f����p��implies�c��p��=��q�R��.�]In�fact,�w7this��had�c��to�happ�A�en:��
w��9e����+Ipro��9v�e��Ometatheorems�b�A�elo��9w�explaining�wh�y�the�principle��Pin�question,��and�indeed,��an�y�uses�of����+Iclassical��elogic��fwhatso�A�ev��9er,��bare�in�principle�eliminable�from�pro�A�ofs�of�theorems�of�the�form�found����+Iin�TEuclid.�����G�color push gray 0����[�28������	color pop�����٠s���������G�color push gray 0�����	color pop���[(��������8�C�5src:1904 ConstructiveGeometryFinalPreprintVersion.tex�In�=�order�to�arriv��9e�at�suc�h�metatheorems,�H%w�e�rst�form�ulate�the�principle�in�question�in�the����+Ilanguage�v_of�v`�ECG�,�whic��9h�do�A�es�not�con��9tain��<��as�a�primitiv��9e.�?�Our�form�ulation�v_is�as�follo��9ws:�އIf����+It��9w�o��unequal�p�A�oin��9ts���B��q�and��C�ä�are�b�oth��b�et��9w�een���A��and��D��,�"Vthen��either��B��q�is�b�et��9w�een���A���and��C�ä�or����+I�C��<�is�Tb�A�et��9w�een��A��and��B�r��.�pF��:�ormally�that�is����8�C�5src:1910 ConstructiveGeometryFinalPreprintVersion.tex����1����B�{�6�=����C�� �^�8�B�(�A;���B�r�;�D�A��)��^��B�(�A;�C� Z;�D�A��)������B�(�A;���B�r�;�C����)��_��B�(�A;�C� Z;�B�r��)���&�&(Axiom�T75)��������8�C�5src:1915 ConstructiveGeometryFinalPreprintVersion.texThe��vp�A�oin��9t��D��=�is�a��wmatter�of�con�v�enience;��the�axiom�is�really��wab�A�out��A�,��>�B�r��,��?and��C�0^�and�their����+Ip�A�ositions���on�a�ra��9y�emanating���from��A�,���but��ECG��do�es�not�ha��9v�e���ra�ys�as���primitiv�e,���so�w�e�need����+Ip�A�oin��9t�T�D�W�to�express�this�in��ECG�.����8�C�5src:1918 ConstructiveGeometryFinalPreprintVersion.texW��:�e�Tdened��AB�{<���C���D�W�in�Denition�48.�pW�e�write��AB�{>���C���D�W�to�mean��C�D�Ӎ<���AB�r��.������&I�color push gray 0��Lemma���10�	color pop���`z`�5src:1920 ConstructiveGeometryFinalPreprintVersion.tex�(in��l�ECG�)��VAxiom�75�implies�that�if��AB����6�=�,G�C���D�A��,���then�either��AB���>�,FC�D��3�or��AB���<����+IC���D�A��.����+I�5src:1923 ConstructiveGeometryFinalPreprintVersion.texPr��o�of�.���Giv��9en��'�AB�
��and��C���D�A��,���let��Q��b�e�the�in��9tersection�p�oin��9t�men�tioned�in�the�denition�of����+I�AB�X�>��C���D�A��.��'Then�G�according�G�to�Axiom�75,�T�if��Q���6�=��B����then�G�either��Q��is�b�A�et��9w�een�G��A��and��B����or��B��is����+Ib�A�et��9w�een��,�Q��-�and��A�.��cIn�the�rst�case�w��9e�ha�v�e��AB�{>���C���D�A��.��cIn�the��-second�case,����B�*��is�inside�the�circle����+Iof�radius��C���D�G��ab�A�out��A�.�\It�follo��9ws�that��D�G��is�outside�the�circle�of�radius��AB�x��and�cen��9ter��C����.�\That����+Icompletes�Tthe�pro�A�of�of�the�lemma.����8�C�5src:1930 ConstructiveGeometryFinalPreprintVersion.texThe��Ktheory��L�ECG��has�quan��9tier-free,��Idisjunction-free�axioms.��UIt�follo�ws��K(as�w��9e�will�pro��9v�e����+Iin�"KTheorem�"L8)�that�no�non-trivial�disjunction�can�b�A�e�pro��9v�ed�"Kin��ECG�.�That�is,�%�if��P�k/�is�negativ��9e����+Iand�s��ECG��pro��9v�es�s��8�x���P�H��(�x�)�.���.��A�(�x�)�M�_�M�B�r��(�x�),��then��ECG�s��pro��9v�es��8�x���P�H��(�x�)�.���.��A�(�x�)�s�or��ECG��pro�v�es����+I�8�x���P�H��(�x�)������B�r��(�x�).�pHence,�TAxiom�75�is�not�pro��9v��|rable�in��ECG�.����8�C�5src:1935 ConstructiveGeometryFinalPreprintVersion.texWhen��'w��9e�add��&Axiom�75,���w�e�will�also�in�tro�A�duce�a�new��&construction�term,���whic�h�w�e�write�����+I�if��2���(�AB�{>���C���D�A�;���P�;�Q�).��This��abbreviates�the�ocial��v��9ersion,�Ohwhic�h��is���if��	�F�(�A;�B�r�;�C� Z;�D�A�;�P�;�Q�).��Pro��9vided����+I�AB�{�6�=����C���D�A��,�n�this�E.term�E/constructs�a�p�oin��9t�whic�h�is�equal�E.either�to��P���or�to��Q�,�n�dep�A�ending�on�whether����+I�AB�{>���C���D�W�or�T�AB�<�C���D�W�.�pThe�Taxiom�expressing�this�is����8�C�5src:1940 ConstructiveGeometryFinalPreprintVersion.tex����s1���(�AB�{>���C���D�Ӎ����if��
' �(�AB�>�C���D�A�;���P�;�Q�)���=��P�H��)���&�&(Axiom�T76)�������^�(�AB�{<���C���D�Ӎ����if��
' �(�AB�>�C���D�A�;���P�;�Q�)���=��Q�)��������8�C�5src:1946 ConstructiveGeometryFinalPreprintVersion.texNote�Tthat�Axiom�76�do�A�es�not�con��9tain�disjunction,�but�that�b�y�Axiom�75,�w�e�ha�v�e��������AB�{�6�=����C���D�Ӎ���if����(�AB�>�C���D�A�;���P�;�Q�)����#��:������&I�color push gray 0��Denition���1�	color pop���g&�5src:1949 ConstructiveGeometryFinalPreprintVersion.tex�The�P�the��ory�P��ECGD��is��ECG��plus�the�new�function�symb��ol�\�if���Z",�Q�with�Axioms�75����+Iand�N<76.����+I�5src:1952 ConstructiveGeometryFinalPreprintVersion.texR��emark.�p�The�T\D"�in��ECGD��is�for�\disjunction".����8�C�5src:1954 ConstructiveGeometryFinalPreprintVersion.texThe�5%follo��9wing�lemmas�giv�e�t�w�o�app�A�ealing�theorems�5&of��ECGD��that�cannot�b�e�pro��9v�ed�5%in����+I�ECG�T�(b�A�ecause�they�are�non-trivial�disjunctions).������&I�color push gray 0��Lemma���11�	color pop���`z`�5src:1957 ConstructiveGeometryFinalPreprintVersion.tex�(in���ECGD�)��{L��et��P�=u�b�e�p�oint�not��on�line��L�.�"Then�any�p��oint��Q��is�either�on�the�same����+Iside�N<of��L��as��P�H��,�or�on�the�opp��osite�side.����+I�5src:1962 ConstructiveGeometryFinalPreprintVersion.texPr��o�of�.���Drop�E#a�E$p�A�erp�endicular��K���from�E$�P���to��L�,�Qmeeting��L��at�p�A�oin��9t��R�>�.���Pro���jecting��Q��to�p�A�oin��9t��Q���-=�0�����+I�on���K����.�6HExtend�segmen��9t��P�H�R�./�past��R�.0�b�y�the�amoun�t��R�>Q���-=�0��ϸ�t�wice,� arriving�at�p�A�oin�t��D�_��on��K����.�6HThen����+Ib�A�oth�R�R�b[�and��Q���-=�0����are�Rb�et��9w�een�R�P���and��D��,�aNso�Rb��9y�Axiom�75,�aOeither��R�bZ�is�b�A�et��9w�een�R�P���and��Q���-=�0����or��Q���-=�0���is����+Ib�A�et��9w�een����P���and��R�>�.��In���the�rst�case,��(�Q���-=�0��z��and��P���are�on�the�same�side�of��L�,��(and�in�the�second�case,����+Ithey���are�on�opp�A�osite�sides.�	PBut��Q��and��Q���-=�0�����are�on�the���same�side�of��L�.�	OHence��Q��and��P�$��are�on�the����+Isame�Tside,�or�on�opp�A�osite�sides,�of��L�.�pThat�completes�the�pro�of�of�the�lemma.������&I�color push gray 0��Lemma���12�	color pop���`z`�5src:1970 ConstructiveGeometryFinalPreprintVersion.tex�(in����ECGD�)��{In�any���triangle��AB�r�C����,�˫either��AB�C�V}�is�a���left�turn�or��AB�C�V|�is�a�right����+Iturn.�����G�color push gray 0����[�29������	color pop�����6�s���������G�color push gray 0�����	color pop���[(��������+I�5src:1974 ConstructiveGeometryFinalPreprintVersion.tex�Pr��o�of�.�6�In�szLemma�8,��w��9e�ha�v�e�already�sho�wn�syho�w�to�construct�sya�p�A�oin�t��Q��not�syon�line��L��Z�=�����+I�Line��@r�(���;����x,�),��'suc��9h��that���AB�r�C�x�is�a�left�turn�or�a�righ��9t�turn�according�as��Q��lies�on�the�same�side�of����+I�L�/�as��
����,�5�or�on�the�/opp�A�osite�side.�i�In��ECGD����b��9y�Lemma�11,�5��Q��m�ust�lie�on�one�side�or�the�other����+Iof�T�L�.�pHence��AB�r�C��<�is�either�a�left�turn�or�a�righ��9t�turn.�That�completes�the�pro�A�of.����8�C�5src:1980 ConstructiveGeometryFinalPreprintVersion.texT��:�erms��?of��@�ECGD��that�in��9v�olv�e��?the�new�sym��9b�A�ol���if�����represen�t��@geometrical�constructions�that����+Ican��pro�A�ceed�b��9y�cases,�Һwith�comparisons�b�et��9w�een��constructed��(unequal)�lengths�determining�the����+Inext�D�construction�steps.���Giv��9en�Euclid's�D�ca�v��|ralier�approac�h�to�case�splits,���the�fact�that�suc�h����+Iconstructions�ƻare�ƺnot�explicitly�men��9tioned�in�Euclid�do�A�es�not�necessarily�mean�that�they�are����+Inot��Orequired��Pto�giv��9e�a�correct�and�complete�v��9ersion�of�Euclid.��nThe�question�th��9us�arises,���whether����+IAxiom�%A75�(or�more�generally��:�,�)<disjunctiv��9e�axioms�of�an�y�%@kind)�are�required�to�formalize�Euclid.����+IBut��Vb�A�ecause�the��U�the��or�ems��V�of�Euclid�do�not�men��9tion�disjunction�in�an�y��Uessen�tial�w�a�y��:�,��w�e�can����+Isimply�!�tak��9e�the�double-negation�in�terpretation,�e'and�eliminate�!�Axioms�75�and�76,�as�will�b�A�e����+Isho��9wn��4b�A�elo�w.�Th�us�what�happ�A�ened��3in�the�example�of�Prop�A�osition�I.6�happ�A�ens�necessarily�in�all����+Iexamples�Tof�similar�logical�form.�������+I�10��Kw�Euclid's�ffparallel�p�s3ostulate�pro���v�ed�ffin�ECG��阍�+I�5src:1992 ConstructiveGeometryFinalPreprintVersion.tex�Let��]�P�>@�b�A�e�a�p�oin��9t�not�on�line��L�.���W��:�e��\consider�lines�through��P�>@�that�do�not�meet��L��(i.e.,�-_are����+Iparallel�V+to�V,�L�).��Pla��9yfair's�v�ersion�V,of�the�parallel�p�A�ostulate�sa��9ys�that�t�w�o�V,parallels�to��L��through����+I�P����are�Hequal.���Our�HAxiom�58�sa��9ys�that�if��K���is�a�parallel�to��L��through��P����and��M�@��is�another�line����+Ithrough�T�P�H��,�with��M����6�=����K����,�then��M�8�meets��L�.�pRecall�that�Euclid's�p�A�ostulate�5�is������+I�color push gray 0��	color pop���AI�5src:1998 ConstructiveGeometryFinalPreprintVersion.tex�If��Ca��Bstr��aight�line�fal�x�ling�on�two�str��aight�lines�make�the�interior�angles�on�the�same����AIside���less���than�two�right�angles,��fthe�two�str��aight�lines,��fif�pr�o�duc�e�d�indenitely,��fme�et����AIon�N<that�side�on�which�ar��e�the�angles�less�than�the�two�right�angles.����8�C�5src:2003 ConstructiveGeometryFinalPreprintVersion.tex�W��:�e�Z�consider�the�Z�formal�expression�of�Euclid's�parallel�axiom.��wSupp�A�ose��P����is�a�p�A�oin��9t�not�on����+Iline���L�,� and��K����is�parallel�to��P�f��through��L�,� �M���is�another�line�through��P�H��,�and��Q��an��9y�p�A�oin�t�on��L�.����+ILet�B��A��b�A�e�an��9y�p�oin��9t�on��M�;��not�equal�to��P����and�B�not�on��L�.��]Euclid's�condition�that��M��\mak��9e�the����+Iin��9terior�V�angles�V�less�than�t�w�o�righ�t�V�angles"�on�the�side�of��P�H�Q��where��A��lies�can�b�A�e�con��9v�enien�tly����+Iexpressed���b��9y���sa�ying�that����A��is�b�A�et��9w�een��Q��and���the�in�tersection���of��QA��with��K����.�̓Th��9us�a�formal����+Iv��9ersion�Tof�Euclid's�parallel�axiom�is�����8�C�5src:2010 ConstructiveGeometryFinalPreprintVersion.tex�����1�����Par��al�x�lel�� ���(�L;���K����)�8�^��on�(�P�A�;�K����)��^��on�(�P�A�;�M����)��^��on�(�A;�M����)�^������(Euclid's�TP��9ostulate�5)�������:�on�(�A;���L�)�8�^��A����6�=��P�W�^�8�B�(�Q;�A;���Interse��ctLines��;���(��Line��`��(�Q;�A�)�;�K����))������������B�(�P�A�;���A;���Interse��ctLines��;���(�L;�M����))���������+I�5src:2018 ConstructiveGeometryFinalPreprintVersion.texNote���that���the�logical�axioms�of�LPT��	mak��9e�it�sup�A�er
uous�to�state�in�the�conclusion�that�����+I�Interse��ctLines��eV��(�L;���M����)�Tis�dened.�pThat�follo��9ws�automatically��:�.����8�C�5src:2022 ConstructiveGeometryFinalPreprintVersion.texAxiom��#58��"has�a�w��9eak�er�conclusion��#than�Euclid's�P��9ostulate�5,��b�A�ecause�it�do�es��"not�sp�ecify����+Ion�dwhic��9h�cside�of��P�KG�the�in��9tersection�p�A�oin�t�will�clie.��On�the�dother�hand,�=�Axiom�58�also�chas�a����+Iw��9eak�er�˄h�yp�A�othesis�˃than�Euclid's�P�ostulate�5,��so�its�exactly�˃relationship�to�P�ostulate�5�˃is�not����+Iimmediately�Tclear.�pOne�direction�is�settled�b��9y�the�follo�wing�theorem:������&I�color push gray 0��Theorem���3�	color pop���b�3�5src:2026 ConstructiveGeometryFinalPreprintVersion.texECG�N<�pr��oves�Euclid's�Postulate�5.����+I�5src:2029 ConstructiveGeometryFinalPreprintVersion.texPr��o�of�.��ESupp�A�ose��FAxiom�58,���and��Elet��L��b�e�a��Eline,����P��)�a�p�oin��9t�not��Eon��L�,����K�K.�parallel�to��L��through����+I�P�H��,�}��M�a��another�h�line�h�through��P��,�}��Q��b�A�e�a�p�oin��9t�h�on��L�,�}��A��b�e�h�a�p�oin��9t�on��M�a��not�on��P�H�Q�,�}�and�h��B��y�the����+Iin��9tersection���of����QA��with��K����.��Supp�A�ose�that�the�in��9terior�angles�made�b��9y��L�,��M����,�and����P�H�Q��mak��9e�less����+Ithan�%�t��9w�o�%�righ�t�angles,�)�whic�h�formally�%�means�that��A��is�b�A�et��9w�een��Q�%��and�and��B�r��.�M�Then�b��9y�Axiom����+I58,���since����M�=��6�=�D��K����,����M�y��do�A�es���meet��L��at�some�p�A�oin��9t��R�>�.�^�It�remains�to�sho�w�that��A����is�b�A�et�w�een��P����+I�and�D��R�>�.��By�Mark��9o�v's�D�principle�D�(Axiom�34)�it�suces�to�pro��9v�e�D�that��P����is�not�b�A�et��9w�een�D��R�U�and��A�,����+Iand���R����is��not�b�A�et��9w�een���P�ȃ�and��A�.�[SSupp�ose�rst�that���P�ȃ�is�b�et��9w�een���R����and���A�.�[TThen��R����is�on�the����+Iopp�A�osite�bside�of�aline��K��J�from��A�.�.�But��R�+��is�not�not�on�the�same�side�of�line��K��I�as��Q�,��since�if��R�+��is����+Inot�6mon�6nthe�same�side�of��K��U�as��Q�,�>�then��R�>Q��(whic��9h�is��L�)�w��9ould�meet��K��U�(whic��9h�it�do�A�es�not).��By�����G�color push gray 0����[�30������	color pop�����h�s���������G�color push gray 0�����	color pop���[(��������+I�Lemma�@w7,�K?�R�P��and��Q��are�@von�the�same�side�of�line��K����.���Since��R�P��is�on�the�opp�A�osite�side�of��K��_�from����+I�A�,�[�it�M�follo��9ws�that�M��A��and��Q��are�on�opp�A�osite�sides�of�line��K����.��AHence�p�A�oin��9t��B�r��,�[�the�in�tersection�M�of����+I�AQ��N�with��K����,���m��9ust�b�A�e�b�et��9w�een��N�A��and��C����.�But�that�con��9tradicts�the�fact�that��A��is�b�et��9w�een��N�Q��and����+I�B�r��.�1Hence�0the�/assumption�that��P�e�is�b�A�et��9w�een�0�R�,m�and��A��has�led�to�a�con��9tradiction.�1No�w�supp�A�ose����+Iinstead�Tthat��R�%��is�b�A�et��9w�een�T�P�^7�and��A�.�pThen��A��and��P��are�on�opp�A�osite�sides�of��L�.����8�C�5src:2045 ConstructiveGeometryFinalPreprintVersion.texBut���B����and���P�_��are�on�the�same�side�of�line��L�,�8since��B�r�P��/�=��K�K����do�A�es�not�meet��L�.� �Hence��A��and����+I�B�)B�are���on�opp�A�osite�sides�of��L�.�But�then�the���in��9tersection�p�oin��9t�of��AB�)B�and��L�,���whic�h�is��Q�,���lies����+Ib�A�et��9w�een�<f�A�<g�and��B�r��,�F+con��9tradicting�the�fact�that��A��lies�b�A�et��9w�een��Q�<f�and��B�r��.���Hence�the�assumption����+Ithat���R�''�is��b�A�et��9w�een��P�_��and��A��has�also��led�to�a�con�tradiction.�!0Hence,�Nas�noted�already��:�,�Ob�y�Axiom����+I34,�T�A��is�b�A�et��9w�een�T�P�^7�and��R�>�.�pThat�completes�the�pro�of�of�Euclid's�P��9ostulate�5�from�Axiom�58.����8�C�5src:2050 ConstructiveGeometryFinalPreprintVersion.texNo��9w���w�e���consider�the�con��9v�erse�problem,��of�deriving���Axiom�58�from�Euclid's�P��9ostulate�5.���The����+Iob��9vious�Tpro�A�of�attempt�w�orks�only�if�w�e�assume�Axiom�75.������&I�color push gray 0��Lemma���13�	color pop���`z`�5src:2053 ConstructiveGeometryFinalPreprintVersion.tex�L��et��T�nb�e��the�the�ory��ECG���without�any�p��ar�al�x�lel�p�ostulate.�W!Then��in�T,�Axiom�75����+I(\of�N<two�une��qual�se�gments�one�is�longer")�plus�Euclid's�Postulate�5�implies�Axiom�58.����+I�5src:2057 ConstructiveGeometryFinalPreprintVersion.texPr��o�of�.���Assume�qHEuclid's�qIp�A�ostulate�and�let��p��b�A�e�a�p�A�oin��9t�not�on�line��L�,��and��K�1�parallel�to��L��through����+I�p�,���and��line��~�M��b�another�line�through��p��as�in�Axiom�58.�(Let��q�%(�b�A�e�the�p�oin��9t��on��L��at�the�fo�A�ot�of�the����+Ip�A�erp�endicular�-�to�-��L��from��p�.�f7Let��A��on�line��M�&��and��B����on�line��L��b�A�e�on�the�same�side�of��pq�R��.�f6Then����+I�K���mak��9es�d�the�d�in�terior�angles��Apq��Y�and��pq�R�B��d�equal�d�to�t�w�o�righ�t�angles.�
�Hence��M�]��do�A�es�not,�x�or�it����+Iw��9ould��*coincide��)with��K����.�bIf�angle��Apq�@��is�less�than�a�righ��9t�angle,���then�b�y��)Euclid's�P�ostulate�5,����M����+I�meets�%t�L�.�L�Similarly��:�,�)|if�%sangle��Apq�x�is�more�than�a�righ��9t�angle,�){w�e�can�%ssho�w�b�y�%sEuclid's�P�ostulate����+Ithat�-Z�M�&>�meets��L��(on�the�other�side�of��q���from��B�r��).�d�By�Axiom�75,�3\one�of�these�alternativ��9es�m�ust����+Ihold.�]�More�+(formally��:�,�0�let��S��-�b�A�e�the�+)in��9tersection�p�oin��9t�of��+)�Line�����(�A;���B�r��)�with��K����.�]�Then�angle�+)�Apq�}��is����+Iless���than���a�righ��9t�angle�if��A��is�b�A�et��9w�een����S���and��B�r��,���and�more�than�a�righ��9t�angle�if��S���is�b�A�et��9w�een����+I�A�w��and�w��B�r��.�DRBy�Axiom�75,���one�of�these�alternativ��9es�holds.�DRHence�b��9y�Euclid's�Prop�A�osition�5,����M����+I�meets�T�L�.�pThat�completes�the�pro�A�of�of�the�lemma.��"�����+I�11��Kw�Constructiv���e�ffGeometry�and�Euclidean�Fields��阍�+I�5src:2074 ConstructiveGeometryFinalPreprintVersion.tex�W��:�e��
kno��9w��(from�the�previous�section)�that��ECG��plus�classical�logic�is�not�only�\reasonable"����+Ibut�^�equiv��|ralen��9t�to�textb�A�o�ok�^�theories.��But�when�w�e�^�use�only�in�tuitionistic�logic,���is�it�still�a����+I\reasonable"�Ttheory?�pThere�are�t��9w�o�Tp�A�ossible�w��9a�ys�Tto�answ��9er�that�question:������+I�color push gray 0��`����	color pop���AI�5src:2079 ConstructiveGeometryFinalPreprintVersion.tex�Can�TEuclid�b�A�e�formalized�in��ECG�?������+I�color push gray 0��`����	color pop���AI�5src:2080 ConstructiveGeometryFinalPreprintVersion.tex�Are�Tthe�mo�A�dels�of��ECG��c��9haracterizable�in�some�elegan�t�w�a�y?����8�C�5src:2083 ConstructiveGeometryFinalPreprintVersion.texW��:�e���ha��9v�e���argued�in�the�previous�section�that�probably��ECG��suces�to�formalize�Euclid,��<and����+Icertainly�.�ECGD�.�do�A�es�suce.�f�In�this�section�w��9e�tak�e�.up�the�second�approac��9h.�f�Classically��:�,�4Mthe����+Imo�A�dels��cof��G��are�all�planes�o��9v�er��ca�Euclidean�eld,���that�is,���an�ordered�eld�in�whic��9h�ev�ery�p�A�ositiv�e����+Ielemen��9t��?has��@a�square�ro�A�ot.��2Is�that�same�thing�true�constructiv��9ely�for��ECG�?�The�main�p�A�oin��9t����+Iof�f�this�section�f�is�to�answ��9er�that�question�in�the�armativ��9e.��The�detailed�answ��9er�is�somewhat����+Isurprising,��though.��There��gturns�out��hto�b�A�e�more�than�one�natural�set�of�axioms�for�constructiv��9e����+IEuclidean���elds,�܋and���these�dieren��9t�v�ersions���corresp�A�ond�directly�to��ECG�,�to��ECGD�,�and����+Ito��b�ECG�with�our�parallel�p�A�ostulate�(Axiom�58)��areplaced�b��9y�the�(apparen�tly)�w�eak�er�v�ersion,����+IEuclid's�TP��9ostulate�5.����8�C�5src:2092 ConstructiveGeometryFinalPreprintVersion.texBefore���turning�to�the�pro�A�ofs���of�those�corresp�ondences,���w��9e�m�ust�pro�v�e�some�elemen�tary����+Ilemmas��ab�A�out��in��9tuitionistic�geometry��:�.�IaW�e��b�egin��with�a�t��9w�o-dimensional��v�ersion�of��Mark�o�v's����+Iprinciple.������&I�color push gray 0��Lemma���14�	color pop���`z`�5src:2096 ConstructiveGeometryFinalPreprintVersion.texECG�(��pr��oves�(�that,�0if�p�oint��P�qx�do�es�(�not�lie�on�line��L�,�0then�some�cir��cle�with�nonzer��o����+Ir��adius�N<and�c�enter��P���lies�entir�ely�on�the�same�side�of��L��as��P�H��.�����G�color push gray 0����[�31������	color pop���� ��s���������G�color push gray 0�����	color pop���[(��������+I�5src:2100 ConstructiveGeometryFinalPreprintVersion.tex�Pr��o�of�.�9zLet�p�A�oin��9t��P�g��not�lie�on�line��L�;���w�e�will�construct�a�circle�with�cen�ter��P�g��lying�on�the����+Isame��Kside�of��J�L��as��P�H��.��TLet�p�A�oin��9t��K�53�on��L��b�A�e�the�fo�ot�of�the��Jp�erp�endicular�from��P��.�to��L�.��TThen����+I�K����6�=��
�P�H��.��Hence���the���t��9w�o�circles����C�����1����=����Cir��cle����(�K� Z;���P�H��)�and��C�����2����=����Cir��cle����(�P�A�;���K����)�that�are�used�to����+Ibisect�Y��P�H�K��}�ha��9v�e�dieren�t�Y�cen�ters�and�eac�h�Y�con�tains�a�Y�p�A�oin�t�inside�Y�the�other.��5Hence�the�p�A�oin��9ts����+I�X����=��(��Interse��ctCir�cles1��G���(�C�����1��*��;���C�����2���)�o�and��Y�6��=��(��Interse��ctCir�cles2��G���(�C�����1��*��;�C�����2���)�o�are�o�dened.�+�If��X����=�(��Y�~�then����+I�X����is�0�b�A�et��9w�een�0��P�yf�and��K����,�wNcon�tradicting��P�H�X��=�i��P�K����.�m�Hence�the�midp�A�oin��9t�of��P�K��j�is�giv��9en�b�y����+I�M���=�����Interse��ctLines��=��(��Line��`��(�X�0�;���Y�8�)�;���Line���r�(�P�A�;�K����)).�n�Hence��0��Cir��cle��9��(�P�;�M����)�0�lies�on�the�same�side�of��L����+I�as����P�H��,��fb�A�ecause��K�d~�is�nearer�to����P�	x�than�an��9y���other�p�oin��9t���of��L�.�4That�completes�the�pro�A�of�of�the����+Ilemma.����8�C�5src:2113 ConstructiveGeometryFinalPreprintVersion.texNext��one�ma��9y�w�onder�whether�the�axioms�of�b�A�et�w�eenness�of��ECG��are�actually�sucien�t�to����+Iestablish���a�reasonable�theory�of�order�of�p�A�oin��9ts�on�a���line.�
)Instead�of�basing�our�denition�of����+Iorder�*�on�b�A�et��9w�eenness,�0=w�e�*�base�it�instead�on���R�Îight����and���L��eft����.�]Fix�*�a�line��L��and�a�p�oin��9t�0�on��L�.����+ILet����K�t��b�A�e�the�p�erp�endicular�to����L��at��K�t��and�let��I��-�b�e�a�p�oin��9t�on��K�t��that�is�not�on��L�.��Dene��x���<�y����+I�for��p�A�oin��9ts���x��and��y�r*�on��L��to�mean�that��xy�R�I����is�a�left�turn,�"and��x���<���y�r*�to�mean�that��xy�R�I����is�a�righ��9t����+Iturn.�pThen�Tw��9e�can�establish�the�fundamen�tal�facts�ab�A�out�in�tuitionistic�order�in��ECG�.������&I�color push gray 0��Lemma���15�	color pop���`z`�5src:2121 ConstructiveGeometryFinalPreprintVersion.tex�(in�N<�ECGD�)�With�notation�as�ab��ove,��ECG��pr�oves��x����6�=�0����x�>��0�8�_��x���<��0�.����+I�5src:2124 ConstructiveGeometryFinalPreprintVersion.texPr��o�of�.�]�Since�++�x�,�0�0,�0�and��I�߆�are�distinct�and�not�collinear,��xy�R�I�߆�is�either�a�left�turn�or�a�righ��9t�turn,����+Ib��9y�TLemma�12.�pThat�completes�the�pro�A�of.������&I�color push gray 0��Lemma���16�	color pop���`z`�5src:2127 ConstructiveGeometryFinalPreprintVersion.tex�(in�N<�ECG�)�With�notation�as�ab��ove,��ECG��pr�oves��0����6�<�x�8�^��x����6�<��0����x��=�0�.����+I�5src:2130 ConstructiveGeometryFinalPreprintVersion.texPr��o�of�.��kSupp�A�ose��G0����6�<�x��F�and��x����6�<��0.��lThen��F0�xI�T��is�not�a�left�turn�and��x�0�I�T��is�not�a�left�turn.��kSupp�A�ose����+I�x����6�=�0.�	�Then���x�0�I��`�is�a�triangle��and�b��9y�Lemma�8,��Inot�not�(one�of��x�0�I��`�and�0�xI��a�is�a�righ��9t�turn�and����+Ithe��other�is�a�left�turn).�.�But�neither�is�a�left�turn;��hence��::�x���=��0.�Since�in��ECG��w��9e�ha�v�e�the����+Iaxiom�T�::�x����=��y��p���x��=��y�R��,�w��9e�can�conclude��x��=�0.�pThat�completes�the�pro�A�of.����8�C�5src:2136 ConstructiveGeometryFinalPreprintVersion.texBut��cnote��bthat�neither��ECG��nor��ECGD��can�pro��9v�e��a���<�b����a�<�x��V�_��U�x�<�b�,��`since�the��cdecision����+Ias��to�whic��9h��alternativ�e��holds�cannot�b�A�e�made�con��9tin�uously�in���x�.��This�theorem�is�essen��9tially�the����+Iaxiom�Tof�apartness.����8�C�5src:2139 ConstructiveGeometryFinalPreprintVersion.texF��:�or�u�p�A�oin��9ts�on��L�,���w�e�sa�y�that��x��is�on�the�u�same�side�of��y�ȉ�as��z��l�if��y�Ȉ�do�A�es�not�lie�on���Se��gment��'��(�x;���z�c��).����+INote��that��this�is�equiv��|ralen��9t�to�sa��9ying�that��x��and��y���are�on�the�same�side�of�the�line�p�A�erp�endicular����+Ito�T�L��at��y�R��.�pThen�w��9e�ha�v�e������&I�color push gray 0��Lemma���17�	color pop���`z`�5src:2142 ConstructiveGeometryFinalPreprintVersion.tex�(in���ECG�)��With��notation�as�ab��ove,�7x�a�p��oint��1����>��0��on���L�.�+�Then��x�>��0���if�and�only����+Iif�N<�x��is�on�the�same�side�of�0�as�1.����+I�5src:2146 ConstructiveGeometryFinalPreprintVersion.texPr��o�of�.�uSSince��J0�Q`�<��1,��01�I�<��is��Ka�left�turn.�uSAssume��x��is�on�the�same�side�of�0�as�1.�uRThen,��b��9y�the����+Iaxioms��for���L��eft���~�,��0�xI���is�a�left��turn.�5KHence�0����<�x�.�5JCon��9v�ersely��:�,�assume��0����<�x�.�5JThen�0�xI���is�a�left����+Iturn.���W��:�e�<�m��9ust�sho�w�<��x��is�on�the�same�side�of�0�as�1;�P~that�is,�F�w�e�m�ust�<�sho�w�that�0�do�A�es�not�lie����+Ion��Z��Se��gment��'��(�x;����1).��Supp�A�ose�Z�0�Z�do�es�lie�Z�on���Se��gment��'��(�x;����1).��Since�0�xI��is�a�left�turn,�l0�i�6�=��x�,�since����+I00�I���is�crnot�cqa�left�turn.��Since�0���6�=�1,�v�w��9e�crha�v�e��B�(�x;����0�;��1).��But�01�I���is�cra�left�turn;���hence�b�A�oth�0�xI����+I�and�T01�I�ɯ�are�left�turns,�con��9tradicting�Lemma�8.�pThat�completes�the�pro�A�of.����8�C�5src:2154 ConstructiveGeometryFinalPreprintVersion.texWith�ethe�ebasic�prop�A�erties�of�order�established,��Vw��9e�are�ready�to�turn�to�the�c��9haracterization�of����+Imo�A�dels�~�of�~��ECG��and��ECGD��as�planes�o��9v�er�~�Euclidean�elds.��GW��:�e�rst�discuss�the�axiomatization����+Iof��Euclidean�elds�with�in��9tuitionistic�logic.�W��:�e�use�a�language��
with�sym�b�A�ols�+�for�addition�and����+I��Kx�for�m��9ultiplication,�Yand�a�unary�predicate��P�H��(�x�)�for�Kw\�x��is�p�A�ositiv�e".���W��:�e�tak�e�the�usual�axioms����+Ifor�� elds,���except�the�axiom�for�m��9ultiplicativ�e��!in�v�erse,���whic�h�sa�ys�� that�p�A�ositiv��9e�elemen�ts�ha�v�e����+Im��9ultiplicativ�e���in�v�erses.�~If���p�A�ositiv�e�elemen�ts�ha�v�e�in�v�erses,���it�is�an���easy�exercise�to�sho�w�that����+Inegativ��9e�V�elemen�ts�do�to�A�o.��RW��:�e�dene�a��Euclide��an���eld��to�b�e�V�a�comm��9utativ�e�ring�satisfying�the����+Ifollo��9wing�Tadditional�axioms:����������x����6�=�0���9�y��T�(�x�8���y��p�=���1)�������EF1���������e�a�P�H��(�x�)�8�^��P��(�y�R��)������P��(�x�8�+��y�R��)��^��P��(�x����y�R��)�������EF2���������r�E�x�8�+��y��p�=���0���:�(�P�H��(�x�)�8�^��P��(�y�R��))�������EF3��������G�color push gray 0����[�32������	color pop����!W�s���������G�color push gray 0�����	color pop���[(����������b��x�8�+��y��p�=���0��^�:�P�H��(�x�)��^�:�P��(�y�R��)������x��=�0�������EF4���������e�]�x�8�+��y��p�=���0��^�:�P�H��(�y�R��)���9����z�c��(�z�q����z��S�=��x�)�������EF5����������D�::�P�H��(�x�)������P��(�x�)������EF6,�Tor�Mark��9o�v's�Tprinciple����䍑+I�5src:2169 ConstructiveGeometryFinalPreprintVersion.texAxiom���EF5���sa��9ys�that�non-negativ��9e�elemen�ts���ha�v�e�square���ro�A�ots.��`This�is���a�stronger�axiom,����+Iin��9tuitionistically��:�,�.�than��ysimply�sp�A�ecifying��zthat�p�ositiv��9e�elemen�ts�ha�v�e�square�ro�A�ots.���W��:�e�could����+Iconserv��|rativ��9ely�Wwextend�Wxeld�theory�b��9y�a�unary�function�sym��9b�A�ol�for�negation,�h��x�,�in�whic��9h�Wwcase����+Ithe�Tlast�axiom�could�b�A�e�more�readably�written��:�P�H��(��x�)�����9�z��7�(�z��c���-=�2����=��x�).����8�C�5src:2174 ConstructiveGeometryFinalPreprintVersion.texAs���usual,��w��9e���dene��x���<�y�O��to���mean��9�z�c��(�P�H��(�z��)��\�^��]�x��+��z��S�=����y�R��,��or���informally��:�,��y�0���x����is�p�A�ositiv��9e;�and����+I�x������y�g��means�T�:�(�y��p<�x�.�pThen�Mark��9o�v's�principle�is�equiv��|ralen�t�to��:�(�x������0)����0��<�x�.����8�C�5src:2177 ConstructiveGeometryFinalPreprintVersion.texW��:�e�*Ualso�consider�*T�we��akly�a�Euclide�an�elds�,�/�whic��9h�*Uinstead�of�EF1,�are�required�*Tonly�to�satisfy����+Ithe�Tw��9eak�er�axiom��������i�P�H��(�x�)�����9�y��T�(�x�8���y��p�=���1)���EF0��������+I�5src:2182 ConstructiveGeometryFinalPreprintVersion.texEF1�"�implies�"�EF0,�&&b�A�ecause��P�H��(�x�)��4���x��5�6�=�0,�&'but�to�"�deriv��9e�EF1�from�EF0,�&&w��9e�w�ould�"�ha�v�e�to�"�kno�w����+Ithat�4	eac��9h�4
nonzero�elemen�t�4
is�either�p�A�ositiv�e�4
or�negativ�e.�x�In�the�language�4
without�a�sym��9b�A�ol�for����+Iadditiv��9e�Tin�v�erse,�this�can�expressed�as������E�x�8�+��y��p�=���0��^��x��6�=�0����P�H��(�x�)��_��P��(�y�R��)���EF7�����+I�5src:2186 ConstructiveGeometryFinalPreprintVersion.texFields�N�that�satisfy�N�EF0�and�EF2�through�EF7�then�automatically�satisfy�EF1�as�w��9ell.��bW��:�e����+Icall���these�elds��str��ongly��Euclide�an�.��5T��:�o���recap:�'�In�a�Euclidean�eld,��Pall�nonzero�elemen��9ts�ha�v�e����+Im��9ultiplicativ�e�c�in�v�erses,��cwhile�in�c�a�w�eakly�Euclidean�eld,��cit�is�only�c�required�that�elemen�ts�kno�wn����+Ito��qb�A�e��rp�ositiv��9e�or��rnegativ�e�ha�v�e��rm�ultiplicativ�e�in�v�erses.���In�a�strongly��rEuclidean�eld,��9nonzero����+Ielemen��9ts�Tare�either�p�A�ositiv�e�or�negativ�e,�so�the�distinction�do�A�esn't�matter.��Ir��+I�5src:2194 ConstructiveGeometryFinalPreprintVersion.tex�R��emark�.�J�Since�z(the�z)double�negation�of�EF7�follo��9ws�from�EF4,��]ev��9ery�Euclidean�eld�is�not�not����+Istrongly��Euclidean;�
in�other�w��9ords,�]one�cannot�giv�e�an��example�of�a�Euclidean�eld�that�is�not����+Istrongly�)NEuclidean.�X]The�standard�plane�is�strongly�Euclidean�if�and�)Monly�if�Mark��9o�v's�)Nprinciple����+I(in�Tthe�form��9ulation��::�x���>��0����x�>��0)�Tholds�in�the�reals.���+����&I�color push gray 0��Theorem���4�	color pop���b�3�5src:2199 ConstructiveGeometryFinalPreprintVersion.tex�The��$mo��dels��#of��ECG��ar�e��#al�x�l�isomorphic�to�planes��F��H�-=�2��s��,��wher��e��F��is�a�Euclide��an�eld.����+Iand���the�r��elations�and���function�symb�ols�of��ECG��ar�e�interpr�ete�d�as���usual�in�r�e�ducing�ge�ometry����+Ito��eld�the��ory.�h�The�mo�dels��of��ECGD��ar�e�al�x�l��isomorphic�to�planes��F��H�-=�2��s��,���wher�e��F����is�a�str�ongly����+IEuclide��an�i�eld.��)The�mo�dels�of��ECG��with�the�p�ar�al�x�lel�Axiom�i�58�r�eplac�e�d�by�Euclid's�Postulate����+I5���ar��e�al�x�l�isomorphic�to���planes��F��H�-=�2��s��,�˳wher�e��F��~�is�a�we�akly�Euclide�an�eld.�m!In�fact,�˳e�ach�of�these����+Ige��ometric�al�N<the�ories�interpr�ets,�and�is�interpr�etable�in,�the�c�orr�esp�onding�eld�the�ory.���䍑+I�5src:2207 ConstructiveGeometryFinalPreprintVersion.texR��emark�.��Theory����A����is�in��9terpretable�in�theory��B��Y�if�there�is�a�map�from�the�syn��9tax�of��A��(v��|rariables,����+Iterms,��$form��9ulas)���to���that�of��B�	��that�preserv�es�pro�v��|rabilit�y��:�.��GTh�us�the�theorem���implies,��$for�example,����+Ithat��
Euclid's�P��9ostulate�5�implies��	Axiom�58�if�and�only�if�axiom�EF0�implies�EF1�with�the�help����+Iof�TEF2�through�EF6.����+I�5src:2212 ConstructiveGeometryFinalPreprintVersion.tex�Pr��o�of�.��W��:�e�F�sho��9w�F�ho�w�to�F�in�terpret�eld�F�theory�in�our�geometrical�theories.��W��:�e�ha��9v�e�three�F�xed����+Ip�A�oin��9ts�A�����,�M��x,�,�and��
����,�pairwise�A�unequal.��MLet�0�b�A�e�another�name�for���C��and�let�1�b�A�e�another�name����+Ifor�F���x,�.���Let��L�F��b�A�e���Line���x�(���;�����),�Sand�F�let��K���b�A�e�the�p�A�erp�endicular�F�to��L��at�0.���Let��I���b�A�e�a�p�A�oin��9t�on��K����+I�but��not�on��L��suc��9h�that�01�I���is�a�left�turn,�
�and�dene��P�H��(�x�)�to�mean�that�0�xI��is�a�left�turn.�?The����+Ifunction���sym��9b�A�ols��+�and����of�eld�theory�are�in��9terpreted�b�y��terms���A��dd�����and���Multiply��(3M�of��ECG�,����+Idened�Tas�sho��9wn�earlier.����8�C�5src:2220 ConstructiveGeometryFinalPreprintVersion.texT��:�ec��9hnically�,�uw�e�bBough�t�bCto�exhibit�formal�pro�A�ofs�in��ECG��of�bBthe�(in�terpretations�of��q)�the�ring����+Iaxioms,��corresp�A�onding��to��the�informal�pro�ofs�in�the��section�on�arithmetization,��and�it�ma��9y�w�ell����+Ib�A�e�
p�ossible��to�exhibit�suc��9h�pro�A�ofs�using�a�theorem-pro��9v�er��or�pro�A�of-c�hec�k�er,��but��here�w�e��rely�on����+Ithe��treader��uto�b�A�e�con��9vinced�that�suc�h��upro�A�ofs�exist�based�on�an�examination�of�the�pro�A�ofs�and����+Ithe�Taxioms�of��ECG�.�����G�color push gray 0����[�33������	color pop����" A�s���������G�color push gray 0�����	color pop���[(��������8�C�5src:2225 ConstructiveGeometryFinalPreprintVersion.tex�No��9w��Xw�e��Wwill�c�hec�k�axiom�EF3.��{Supp�A�ose��W�x�b��+�b��y��!�=�ew0�holds�in�this�mo�A�del.��|That�means�that����+I�Add�(�x;���y�R��)�Օ=�0.�cJSupp�A�ose�םalso�that��:�P�H��(�x�)�and��:�P��(�y�R��),�/i.e.�cKneither�0�xI����nor�0�y�I����is�a�left�turn.����+IThen���x���is�not�b�A�et��9w�een��0�and�1,�>since�01�I���is�a�left�turn,�>and�if��x��is�b�A�et��9w�een�0��and�1,�=0�xI���w��9ould����+Ib�A�e�0a�left�turn�to�o.�l�Similarly��y����is�not�b�et��9w�een�00�and�1.�l�No��9w�assume��x��l<��0.�l�then�0�Add�(�x;���y�R��)�=�0����+Iimplies��Wthat��X�x��and��y���are�on�opp�A�osite�sides�of�0.��zSince��x�0�I�Y��is�a�left�turn,��Xb��9y�Lemma�8�0�xI�Y��is����+Ia�t�righ��9t�turn.�:�Hence�0�y�R�I�)+�is�a�left�turn,���so�0�0��<�0�y��,���con��9tradiction.�:�Hence��:�x�0�<��0.�:�Since�t�w�e�ha�v�e����+Ipro��9v�ed��%�x��$�is�neither�negativ��9e�nor�p�A�ositiv��9e,���w�e��%ha�v�e��x��"�=�0��%b��9y�Lemma�16.�(�That�completes�the����+Iv��9erication�Tof�axiom�EF3.����8�C�5src:2234 ConstructiveGeometryFinalPreprintVersion.texNo��9w��w�e��turn�to�axiom�EF5.��uSupp�A�ose��x����+����y�R��=���0.�Supp�A�ose�also���:�P�H��(�y�R��);�^�that�is,�'�it��is�not����+Ithat��case���that��y�T&�6�=�}0�and�0�y�R�I��P�is�a�left�turn.��RV��:�erifying�this�axiom�amoun��9ts�to�c��9hec�king��that����+Iin�C,�ECG�,�Descartes'�C+square�ro�A�ot�construction�can�b�A�e�extended�to�a�function�dened�for��x��-���.�0,����+Iwithout�<requiring�<a�case�distinction�as�to�whether�or�not�the�argumen��9t�is�zero.���W��:�e�ha��9v�e�<giv�en����+Ijust�+gsuc��9h�an�extension�+hof�Descartes'�construction�in�the�section�on�arithmetization,�0�and�no��9w�it����+Ionly�Tremains�to�remark�that�the�argumen��9t�giv�en�there�can�b�A�e�carried�out�in��ECG�.����8�C�5src:2240 ConstructiveGeometryFinalPreprintVersion.texNext�[Aw��9e�[@c�hec�k�axiom�[@EF7�in��ECGD�.�Let��x�<��+��y�X��=�P0�and�[@�x�Q�6�=�0.��6Then��x�[A�and��y����are�on�line����+I�L�,�6_and�/�w��9e�/�can�nd�a�p�A�oin�t��D�q��on�the�same�/�side�of�0�as�1�and�a�p�A�oin�t��A�/��on�the�opp�A�osite�side�of����+I0�|�from�|�1,����x��b�A�et��9w�een��A�|��and��D�A��.�SBThen�either�0�is�b�A�et��9w�een�|��A��and��x�,���in�whic��9h�case��x��is�p�A�ositiv��9e����+I(that��Ois�0�xI����is�a�left�turn),���or��x��is�b�A�et��9w�een��O�A��and�0,���in�whic��9h��Pcase��y�C��is�on�the�opp�osite�side�of�0����+Ifrom�T�x�,�i.e.�pthe�same�side�as�1,�and�hence��y�g��is�p�A�ositiv��9e.����8�C�5src:2247 ConstructiveGeometryFinalPreprintVersion.texNext���w��9e�v�erify���Mark�o�v's�principle��::�P�H��(�x�)��|���{�P��(�x�).�	Supp�A�ose����::�P��(�x�),��i.e.�	�::�x�>��0.�Let����+I�y��p�=�����A��dd��h��(�x;����1),�_�so�2�that�2��y�>���x�.���Then��x�>��0�2�is�equiv��|ralen��9t�to��B�(0�;���x;�y�R��).���Hence�w��9e�ha�v�e��::�B�r��(0�;���x;�y�R��).����+IThen�Tb��9y�Mark�o�v's�principle�in��ECG�,�w�e�ha�v�e��B�r��(0�;���x;�y�R��),�Ti.e.�p�x���>��0,�i.e.�p�P�H��(�x�),�as�desired.����8�C�5src:2251 ConstructiveGeometryFinalPreprintVersion.texW��:�e��$no��9w�turn�to�the�v�erications��%of�the�parallel�Axiom�58�and�Euclid's�P�ostulate�5.�In�order����+Ito���dene���the�recipro�A�cal�1�=x��in�geometry��:�,���w��9e�use�Descartes's�metho�A�d.��That�is,���w��9e�x�a�line��L��to����+Iserv��9e���as���the��x�-axis,���and�a�p�A�oin�t���0�on��L��and�a�p�A�oin��9t�1����6�=�0�on����L�.���Then�let��X�\Y�b�A�e�a�p�A�oin��9t�on��L�,���and����+Isupp�A�ose���X�F!�6�=���0.�W��:�e��wish�to�dene�a�p�oin��9t�1�=X��[�.�Erect��the�p�erp�endicular��K�o��to��L��at��X��[�,�ڻand�nd����+Ia�:�p�A�oin��9t�on��K�ސ�at�distance�1�from��X��[�,�C�for�example��Q���=����Interse��ctLineCir�cle1��U���(�K� Z;�����Cir��cle�����(�X�0�;����0�;��1)).����+IThen�x�Q���6�=��0,� �since��Q�y�is�not�on��L�,�so�yw��9e�can�form��M����=����Line����(0�;���Q�).�7�Erect�the�p�A�erp�endicular�x�H����+I�to�y��L�y��at�1.�I�Then��K���and��H�0��are�parallel,���since�they�are�b�A�oth�p�erp�endicular�y�to��L�.�I�Line��M�r��do�es����+Inot��coincide��with��K����,���since�0�lies�on��M����but�not�on��K����(since��X�F!�6�=���0).��Hence��M���is�a�line�through����+I�Q��@�that��?is�not�parallel�to��H���.�`3Then,��:b��9y�Axiom�58,��:�M�z$�meets��H�8]�in�a�p�A�oin��9t��R�>�.�`3The�segmen�t��@�R�>�1����+Ihas��the��desired�length.���The�desired�p�A�oin��9t�1�=X��t�on�line��L��is�one�of�the�in��9tersection�p�A�oin�ts��of�����+I�Cir��cle��FC�(0�;����1�;�R�>�)�P�with��L�.��Whic��9h�one�it�is�dep�A�ends�P�on�the�sign�of��X��[�,�_Uwhic�h�w�e�do�not�kno�w;�n"but����+Ithe�Tselection�is�made�automatically�b��9y�the�denition�of���Interse��ctLineCir�cle2��V$�:�����x�f1�=X�F!�=�����Interse��ctLineCir�cle2��U���(��Line��`��(0�;���X��[�)�;���Cir��cle�����(0�;��1�;�R�>�))����+I�5src:2262 ConstructiveGeometryFinalPreprintVersion.texsince���the�t��9w�o���in�tersection���p�A�oin�ts�are�n�um�b�A�ered�in�the�same�order�as���0�and��X�p�o�ccur�on��L�.���Hence����+Iarithmetic��Don��L��satises�the��Caxioms�of�a�Euclidean�eld.��@Similarly��:�,��if�w��9e�only�ha�v�e�Euclid's����+IP��9ostulate��C5,��Gw�e�can�still�construct�1�=X�~��if�w�e�kno�w��Dthat��X�F!>����0,��Fas�follo�ws.�kConsider�the�in�terior����+Iangles�d�made�b��9y��M�]w�and��H���with�the�p�A�erp�endicular�d�dropp�ed�d�from��Q��to��H���.�
/They�will�mak��9e�less����+Ithan���t��9w�o�righ�t���angles�if��X�>�Z��0,���and�more�than�t��9w�o�righ�t�angles���if��X�<�Z��0.��	Hence�b�y�Euclid's����+IP��9ostulate���5,��jif��X�F!>����0,��M����meets��H�s��as�sho��9wn�in�the�gure,�while�if��X�F!<����0,��M����meets��H�s��at�a�p�A�oin��9t����+Isouth��of��L�.�XOne�can��
then�v��9erify�that�arithmetic�on��L��satises�the�w��9eak�Euclidean�eld�axioms.����+IFinally��:�,���Axiom���75�erases���the�distinction�in�question,���and�enables�us�to�v��9erify�axiom�EF7�as�w��9ell.����8�C�5src:2270 ConstructiveGeometryFinalPreprintVersion.texCon��9v�ersely��:�,��kassume��3�F���is��4a�Euclidean�eld.��W�e�will�sho��9w�ho�w��4to�turn��F��H�-=�2��	)��in�to��4a�mo�A�del�of�����+I�ECG��D�c�(or,��,to��bdescrib�A�e�the��aconstruction�more�formally��:�,�w��9e�will�sho�w�ho�w�to��ain�terpret�geometry�in����+Ield��<theory).���As�usual��=in�the�corresp�A�onding�classical�theories,���w��9e�tak�e�the�p�A�oin�ts�to�b�A�e�elemen�ts����+Iof�n��F��H�-=�2��s��,��#and�n�let�lines,�circles,�arcs,��"and�segmen��9ts�ha�v�e�their�n�usual�analytic�denitions;��Uin�particular����+Iw��9e��dene�circles�so�that�circles�of�zero�radius��are�allo�w�ed.���Hence���Cir��cle3��%D��can�b�A�e�in�terpreted.����+IMark��9o�v's�ojprinciple�in�ok�F��M�allo��9ws�us�to�v�erify�that�okaxiom�of��ECG�.�The�in�tersection�okp�A�oin�ts�of�circles����+Iand�€lines,���and��the�in��9tersection�p�A�oin�ts��of�circles�and�circles,���can�b�A�e�dened�b��9y�the�solution�of����+Iquadratic��Mequations;�5Iand�whic��9h�one��Lis�whic�h�(i.e.�\[the��Lconcept�\�AB�r�C�y5�is�a�left�turn")�can�b�A�e�����G�color push gray 0����[�34������	color pop����#8�s���������G�color push gray 0�����	color pop���[(��������+I�dened��as��usual�in�computer�graphics,���b��9y�the�cross�pro�A�duct,���whic��9h�can�b�A�e�dened�in�Euclidean����+Ield���theory���(note�that�division�is�not�required).��9Then�one�has�to�v��9erify�that�the�axioms�of��ECG����+I�ab�A�out���handedness���are�v��|ralid.�y�The�details�follo��9w�the�sk�etc�h���giv�en�in�an���earlier�section,���and�are����+Iomitted.�AThe�f�v��9erication�f�of�the�parallel�Axiom�58�requires�that�the�recipro�A�cal�1�=x��b�A�e�dened����+Iwhen����x����6�=�0;��the�v��9erication���of�Euclid's�P�ostulate���5�only�requires�that�the�recipro�A�cal�b�e���dened����+Iwhen�`��x�=>��0.��GIf��F���is�`�strongly�Euclidean�(satises�EF7)�then�w��9e�can�also�v��9erify�Axiom�75,�smand����+Iin�qthat�case�there�qis�on�ob��9vious�in�terpretation�of�the�function�sym�b�A�ol��q�if��wq�of��ECGD��v��|ralidating����+IAxiom�T76.�pThat�completes�the�pro�A�of�of�the�theorem.����8�C�5src:2286 ConstructiveGeometryFinalPreprintVersion.texIn��^[�3����]�it��_is�sho��9wn�that�the�dieren��9t�constructiv�e�v�ersion�of��_Euclidean�eld�theory�are�not�only����+Iapparen��9tly���dieren�t,��Kbut�really�are�not�equiv��|ralen�t�with�constructiv�e�logic;��and�as�a�corollary��:�,����+Ithe�Tdieren��9t�v�ersions�of�the�parallel�p�A�ostulate�are�also�not�equiv��|ralen�t.�������+I�12��Kw�Classical�fflogic�not�needed�for�negativ���e�theorems��阍�+I�5src:2292 ConstructiveGeometryFinalPreprintVersion.tex�Our��Tplan��Sin�this�section�is�to�in��9v�estigate��Tthe�double-negation�in��9terpretation�for�geometric�theo-����+Iries.�VSince��the��double-negation�in��9terpretation�applies�a�double-negation�to�atomic�form��9ulas,���w�e����+Ineed�L�to�ha��9v�e�L��::�A������A��for�L�eac��9h�atomic�form�ula��A�.��lW��:�e�rst�consider�the�L�case�when��A��has�the����+Iform�T�t����#�.����8�C�5src:2297 ConstructiveGeometryFinalPreprintVersion.texThe�L�follo��9wing�sc�hema�L�seems�initially�to�ha�v�e�the�c�haracter�L�of�Mark�o�v's�principle,�Z�since��t��5�#�,����+Iin�Tn��9um�b�A�er�theory�or�ev�en�in�the�recursiv�e�mo�A�del��R���-=�2��*��,�in�v�olv�es�an�existen�tial�quan�tier.��������)��::�t����#��t��#��'�for�Tall�terms��t��������+I�5src:2302 ConstructiveGeometryFinalPreprintVersion.tex�As���it�turns�out,���ho��9w�ev�er,�this���sc�hema���is�pro��9v��|rable�in��ECG�;�ho�w�ev�er,���w�e�require�Axiom�58�for����+Ithe�Tpro�A�of,�and�Euclid's�P��9ostulate�5�apparen�tly�do�A�es�not�suce.������&I�color push gray 0��Lemma���18�	color pop���`z`�5src:2306 ConstructiveGeometryFinalPreprintVersion.tex�L��et��?�t��>�b�e�any�term��>of��ECG�.�Then��ECG��pr��oves��::�t����#��t��#�.��Mor��e�over,��?Axiom�34����+I(Markov's�N<principle�for�b��etwe�enness)�N<is�not�ne��e�de�d�N<in�the�pr��o�of.����+I�5src:2310 ConstructiveGeometryFinalPreprintVersion.texPr��o�of�.���W��:�e�C�pro�A�ceed�C�b��9y�induction�on�the�complexit��9y�of�the�term��t�.���If��t��is�a�v��|rariable�or�constan��9t����+Ithen���t����#���is�an�axiom�of��LPT�,�so�w��9e�need�consider�only�comp�A�ound�terms.��Consider�the�induction����+Istep��xwhen��t��is�a�term��f����(�q�R�;���r�A��),�
for�terms��q�.!�and��r��.�n�By�the�induction�h��9yp�othesis,�
�ECG��pro��9v�es����+I�::�q�
��#�� �q��#�.�3�Argue�ǽin��ECG��as�Ǿfollo��9ws:��BSupp�A�ose��::�f����(�q�R�;���r��)�� �#�Ǿ�W��:�e�claim��::�q�
��#�.�3�F�or�supp�A�ose����+I�:�q���#�.�08Then�Ɩb��9y�Ɨthe�strictness�axioms�of��LPT�,�w��9e�ha�v�e��:�f����(�q�R�;���r�A��)��5�#�,���con�tradicting�Ɩ�::�f��(�q�R�;���r�A��)��5�#�.����+IThat�
�con��9tradiction�
�completes�the�pro�A�of�that��::�q��p�#�.��Using�the�pro�of�
�in��ECG��that��::�q��p�#����q��#�,����+Iw��9e�k�conclude��q�t �#�.�Similarly��:�,��w�e�ha�v�e��r�c=�#�.�Hence,��to�complete�the�pro�A�of,��it�k�suces�to�pro�v�e�for����+Ieac��9h�Tfunction�sym�b�A�ol��f���of��ECG��that������S{�s�����1���m�#���^�����:�:�:��c��^�8�s�����n�����#�^::�f����(�s�����1��*��;����:�:�:��
�;���s�����n��7�)��#��f��(�s�����1��*��;����:�:�:��
�;���s�����n��7�)��#����+I�5src:2322 ConstructiveGeometryFinalPreprintVersion.tex�W��:�e���will�pro��9v�e���this���b�y�an�exhaustiv�e�consideration�of�eac�h�function���sym�b�A�ol�of��ECG�.�First����+Iconsider���the�case�when��f��J�is���Interse��ctLines��<���.���Supp�A�ose�that���Interse��ctLines���(�K� Z;���L�)�is�not�undened.����+IBy��Axiom��12,��yif��K����=���L��then���Interse��ctLines��>�"�(�K� Z;���L�)�is�undened;�5hence��K����6�=���L�.�[�Then�b��9y����+IAxiom�3�58�3�(the�parallel�axiom),�;�K��k�meets��L�.�v�Note�that�neither�Pla��9yfair's�v�ersion�3�of�the�parallel����+Iaxiom���nor���Euclid's�w��9ould�seem�to�suce�here:�@�Pla��9yfair's�do�A�es�not�guaran��9tee�the�existence�of����+Ithe��in��9tersection��p�A�oin�t,�and�Euclid's��guaran�tees�it��only�when�w�e�kno�w��on�whic�h�side��of�some����+Itransv��9ersal�T�K��<�mak�es�in�terior�angles�less�than�t�w�o�righ�t�angles.����8�C�5src:2330 ConstructiveGeometryFinalPreprintVersion.texConsider�J/the�case�when�J0�f�D��is���Interse��ctLineCir�cle1��Y�.�or���Interse��ctLineCir�cle2��V;��.��Then�w��9e�ha�v�e�a����+Iline����L����and�a�circle��C�Wz�that�do�not�fail�to�meet,��!and�w��9e�m�ust���sho�w�that���they�do�indeed�meet.����+ILet�-R�P�v5�b�A�e�-Qthe�cen��9ter�of��C����,�3Qand�let��K��:�b�e�the�-Qline�through��P�v5�p�erp�endicular�to��L�-Q�(whic��9h�w�e�ha�v�e����+Isho��9wn�$�ho�w�$�to�construct�without�kno��9wing�whether��P�m��is�on��L��or�not).�J�Let��F�m��b�A�e�the�fo�A�ot�of�this����+Ip�A�erp�endicular,��i.e.�@the���in��9tersection���p�oin�t���of��L����and��K����.�@Then��L��meets��C�a��if�and�only�if��P�H�F�|�is�����G�color push gray 0����[�35������	color pop����$UY�s���������G�color push gray 0�����	color pop���[(��������+I�less�A0than�A/or�equal�to�the�radius��r����of��C����.��So�w��9e�ha�v�e��::�P�H�F�#������r�A��.��Then�(ev�en�A/without�Mark�o�v's����+Iprinciple)�Tw��9e�ha�v�e��P�H�F�ک�����r�A��,�so��C��<�do�es�meet��L��as�desired.����8�C�5src:2337 ConstructiveGeometryFinalPreprintVersion.texNext�1�consider�the�case�1�when��f�,��is���Interse��ctCir�cles1��K
!�or���Interse��ctCir�cles2��G�B�.�rThen�w��9e�ha�v�e�t�w�o����+Icircles���C�M�and��K��that�do�not�fail�� to�in��9tersect,��and�w�e�m�ust�� sho�w�that�they�do�in�tersect.���The����+Irelev��|ran��9t��geometrical��fact�is�that�t��9w�o��circles�in��9tersect�if�and�only�if�the�distance��d��b�A�et��9w�een��their����+Icen��9ters��ais�less�than��bor�equal�to�the�sum�of�their�radii��r�����1�����+����r�����2��*��.�h�So�if��C�}I�and��K�}J�do�not�fail�to����+Iin��9tersect�implies��::�d��������r�����1��9R�+���r�����2��*��.�uEv�en�without�Mark�o�v's�principle�w�e�then�ha�v�e��d��������r�����1��9R�+���r�����2��*��,�,so����+I�C��<�and�T�K��do�in��9tersect.����8�C�5src:2343 ConstructiveGeometryFinalPreprintVersion.texNext�l�w��9e�l�consider�the�constructors.���N�Cir��cle���H�(�P�A�;���Q�)�is�alw�a�ys�l�dened,���since�w�e�l�allo�w�zero-radius����+Icircles;��similarly���for�����Cir��cle3��!�1�(�P�A�;���Q;�R�>�).���)�Line��!��(�P�;�Q�)���is���dened�if�and�only�if��P��e�6�=�{��Q�.��(Hence�w��9e����+Ineed�޲�::�P�*E�6�=��b�Q����c�P��6�=��Q�.�x�This�޲follo��9ws�from�ޱthe�general�in�tuitionistic�logical�ޱprincipal�that����+Itriple�*3negation�is�*4equiv��|ralen��9t�to�single�negation.�[The�constructors�for�segmen�ts�and�*4arcs�can�b�A�e����+Itreated�Tsimilarly��:�.����8�C�5src:2348 ConstructiveGeometryFinalPreprintVersion.texFinally��Mw��9e��Nconsider�the�accessors,���suc��9h�as���c��enter���[�,�����p��ointOn1��'͜�,���etc.��These�are�all�total,���so�there����+Iis�Tnothing�to�pro��9v�e.�pThat�Tcompletes�the�pro�A�of�of�the�lemma.����8�C�5src:2354 ConstructiveGeometryFinalPreprintVersion.texLet�N��A���-=���	���b�A�e�N�the�G��`odel�double-negation�in��9terpretation�of��A�,�]Jobtained�b��9y�replacing��9��b��9y��:8:����+I�and����A�'s�_��B���b��9y����:�(�:�A�'t�^�:�B�r��).���W��:�e�do���not�replace��A��b��9y��::�A��for�atomic��A��since�these�are�equiv��|ralen��9t����+Iin�Tin��9tuitionistic��ECG�.��tՍ���&I�color push gray 0��Theorem���5�(Double�negation�in��Cterpretation)�	color pop����;��5src:2360 ConstructiveGeometryFinalPreprintVersion.tex�Supp��ose�c�ECG��with�classic�al�clo�gic�pr�oves��A�.����+IThen�N<�ECG��with�intuitionistic�lo��gic�pr�oves��A���-=���G�.����+I�5src:2364 ConstructiveGeometryFinalPreprintVersion.texPr��o�of�.�^�First�� w��9e�observ�e��!that��::�A��is�equiv��|ralen�t��!to��A��for�atomic��A�.�^�This�is�an�axiom�for�all����+Iatomic�^�form��9ulas�^�not�of�the�form��t����#�,��'and�also�for�those�of�that�form�when��t��has�the�form�����+I�Interse��ctLines��eV��(�u;���v�R��).��F��:�or�other�form��9ulas�of�the�form��t����#�,�	�w��9e�ha�v�e�pro�v�ed�it�in�Lemma�18.��Since����+Ithe�3axioms�3
of��ECG��are�quan��9tier-free�and�disjunction-free,�:xit�follo�ws�3
that�so��A���-=���	z#�is�equiv��|ralen�t����+Ito��4�A��for�axioms��A��5�of�the��ECG�.�No��9w�the�theorem�follo�ws�as�so�A�on��5as�w�e�c�hec�k�the�soundness����+Iof���the���double-negation�in��9terpretation�in�a�m��9ulti-sorted�logic�with�partial�terms.��5But�that�is����+Istraigh��9tforw�ard;��bsorts���and���partial�terms�oer�no�complications�o��9v�er�the���usual�rst-order�case.����+IThat�Tcompletes�the�pro�A�of.������&I�color push gray 0��Corollary���3�	color pop���d���5src:2374 ConstructiveGeometryFinalPreprintVersion.texECG���with�classic��al��lo�gic��is�c��onservative�over��ECG��with�intuitionistic�lo��gic�for����+Ine��gative�N<formulae.����+I�5src:2377 ConstructiveGeometryFinalPreprintVersion.texPr��o�of�.�pF��:�or�Tnegativ��9e��A�,��A���-=���	\m�is�iden�tical�to��A�.����8�C�5src:2379 ConstructiveGeometryFinalPreprintVersion.texA�*�t��9ypical�+theorem�+of�Euclid�has�the�form��H����`��A�,�pywhere��A��will�b�A�e�quan��9tier-free�when����+Iform��9ulated��hin��i�ECG�,�and��H����is�a�collection�of�h��9yp�A�otheses�that�certain�p�A�oin��9ts�are�distinct,�4�or����+Icertain��incidence�relations��hold�or�do�not�hold.��As�Pro�A�clus�p�oin��9ted�out,�sometimes�this�implies����+Ia���theorem���form��9ulated�b�y���cases.��F��:�or�example,��Euclid�I.2���has�one�pro�A�of�if��A�u^�=��C�A��and�another����+Ipro�A�of�hif�i�A����6�=��C����.��Using�the�hdouble-negation�in��9terpretation,�ew�e�hnd�a�pro�A�of�that��A����=��C��I�_�a�A��6�=��C����+I�implies�=the�<conclusion�of�Euclid�1.2,�H�but�without�the�la��9w�of�the�excluded�middle�w��9e�cannot����+Iconclude�Tthe�\uniform�v��9ersion"�of�Euclid�I.2.����8�C�5src:2386 ConstructiveGeometryFinalPreprintVersion.texRecall��"the�example�giv��9en�ab�A�o�v�e�of�Euclid's�Prop.�
I.6,��where�Axiom�75�is�used,��but�the�same����+Iconclusion��can��b�A�e�reac��9hed�without�it.�I�Since�Axiom�75�is�classically�tautological,��~the�double-����+Inegation���in��9terpretation���sho�ws�that�it���is��always��eliminable�from�pro�A�ofs�of�negativ��9e�theorems.��3But����+Iall�?kthe�theorems�in�Euclid�?jare�either�already�negativ��9e,�j4or�assert�the�existence�of�some�ob���jects�that����+Ican��b�A�e��constructed�using�the�terms�of��ECG�;�when�form��9ulated�more�explicitly��:�,���they�are�negativ��9e����+Iin�0pthe�0qsense�that�they�sa��9y�that�the�result�of�a�certain�construction�has�certain�(quan��9tier-free)����+Iprop�A�erties.��l	���+I�13��Kw�F���from�ffpro�s3ofs�to�geometric�algorithms��阍�+I�5src:2396 ConstructiveGeometryFinalPreprintVersion.tex�In���this���section�w��9e�tak�e���up�our�plan�of�doing�for��ECG��what�cut-elimination�and�recursiv��9e����+Irealizabilit��9y��Gdid�for�in�tuitionistic�arithmetic�and�analysis,�ױnamely��:�,�װto�sho�w�that�existence�pro�A�ofs�����G�color push gray 0����[�36������	color pop����%l�s���������G�color push gray 0�����	color pop���[(��������+I�lead�pto�programs�(or�terms)�pro�A�ducing�the�ob���ject�whose�existence�is�pro��9v�ed.��[In�pthe�case�of��ECG����+I�w��9e��w�an�t�to��pro�A�duce�geometrical�constructions,�B�not�just�recursiv�e��constructions�(whic�h�could����+Ialready�5Wb�A�e�pro�duced�5Vb��9y�kno�wn�tec�hniques,�=Wsince��ECG��is�in�terpretable�5Vin�Heyting's�arithmetic����+Iof�inite�t��9yp�A�es).�tT��:�erms�of��ECG��corresp�ond�in�a�natural�w��9a�y�ito�straigh��9tedge�and�compass����+Iconstructions.������&I�color push gray 0��Theorem���6�(Geometric�constructions�extracted�from�in��Ctuitionistic�pro�K�ofs)�	color pop����L�5src:2405 ConstructiveGeometryFinalPreprintVersion.tex�(i)Supp��ose����+I�ECG�T��pr��oves�T��P�H��(�x�)�����9�y��T�(�x;���y�R��)��wher��e��P����is�T�ne�gative�(do�es�not�T�c�ontain��9��or��_�).�S�Then�ther�e�T�is�a����+Iterm�]��t�(�x�)��of��ECG��(r��epr�esenting�]�a�ge��ometric�]�c�onstruction)�]�such�that��P�H��(�x�)��������(��(�x;���t�(�x�))��is�also����+Ipr��ovable�N<in��ECG�.����8�C�5src:2410 ConstructiveGeometryFinalPreprintVersion.tex(ii)�N<Same�as�(i)�but�with��ECGD��in�plac��e�of��ECG�.����8�C�5src:2413 ConstructiveGeometryFinalPreprintVersion.tex(iii)�^�L��et�^�ECG��+��DE��b��e��ECG�,�augmente��d�with�a�c��onstant��D��and�the�axiom�saying��D��is�a����+Itest-for-e��quality�N<function.�@Then�the�analo�gue�of�(i)�holds�for��ECG��+��DE�.����+I�5src:2417 ConstructiveGeometryFinalPreprintVersion.texPr��o�of�.�=W��:�e�ɺuse�cut-elimination.����-=�9���-��Since�ɻour�axiomatization�is�quan��9tier-free,���if�� ��p�!���9�y��T��is�pro�v-����+Iable�߂in�߁constructiv��9e��ECG�,�then�there�is�a�list��of�quan��9tier-free�axioms�suc�h�߁that��;��� ��p�)���9�y��T����+I�is�Ckpro��9v��|rable�Cjb�y�a�cut-free�Cj(hence�quan�tier-free)�pro�A�of.���Since�our�axiomatization�Cjis�disjunction-����+Ifree,�r�b��9y�`&[�14��	?�]�`'(or�rather,�b��9y�`'its�adaptation�to�m��9ulti-sorted�logic�with�the�logic�of�partial�terms)����+Iw��9e�r�can�p�A�erm�ute�the�inferences�so�that�the�existen�tial�quan�tier�is�in�tro�A�duced�at�the�last�step.����+IThen�v�w��9e�obtain�the�desired�v�pro�A�of�just�b�y�omitting�the�last�step�of�the�v�pro�A�of.�@{That�completes����+Ithe��pro�A�of��of�part�(i).�$3All�the�w��9ork�w�as�in��arranging�the�axiom�system�to�b�A�e�quan��9tier-free�and����+Idisjunction-free.�\\P��9art��N(iii)��Mis�pro�v�ed��Min�the�same�w��9a�y��:�,�Lnoting�that��Nthe�axioms�for��D��are�also����+Idisjunction-free.����8�C�5src:2431 ConstructiveGeometryFinalPreprintVersion.texP��9art�x�(ii)�x�requires�a�bit�more�w��9ork.�F Since�Axiom�75�con��9tains�a�disjunction,��]there�is�an����+Iissue�eHab�A�out�p�erm��9uting�eGthe�inferences�in�a�cut-free�pro�of.�KSupp�ose�that�w��9e�ha�v�e�eGa�cut-pro�A�of�of����+I���)��� ���!�9�y��T�,��Owhere����is���a�list�of�axioms�(or�subform��9ulas�of�axioms)�of��ECGD�.�Among�����+Ithere�ϐma��9y�Ϗb�A�e�o�ccurrences�of�ϏAxiom�75,�݄whic��9h�con�tains�Ϗa�disjunction.�.W��:�e�pro�v�e�b�y�Ϗinduction�on����+Ithe���n��9um�b�A�er�of�disjunctions�in��that�there�exists�a�term��t��of���ECGD��)�j�and�a�list��of�axioms����+Iof�e��ECGD��suc��9h�that�e��;�������)����(�t�)�is�pro�v��|rable.�
-The�basis�e�case,�y�when�there�e�are�no�disjunctions,����+Iis�>5part�>4(i)�of�the�theorem.��No��9w�for�the�induction�step.��If�the��9��on�the�righ��9t�is�in�tro�A�duced�>4at�a����+Ilo��9w�er��>lev�el�(nearer�the�end-sequen�t)�than�the�lo�w�est�in�tro�A�duction�of�disjunction�on�the�left,��Cthen����+Iw��9e�]can�complete�the�pro�A�of�as�ab�o��9v�e,�n�since�]the�]line�just�b�efore�the��9��is�in��9tro�duced�will�con��9tain����+Ithe� desired�term,�"�and� the��9�-in��9tro�A�duction�can�just�b�e�p�ostp�oned�un��9til�the� end.�<tOtherwise�there����+Iis�Ta�part�of�the�pro�A�of�that�lo�oks�lik��9e�this:�����8�C�5src:2450 ConstructiveGeometryFinalPreprintVersion.tex����?��AB�{>���AC� Z;���������1���m�)�9�y��T��AB�<�AC� Z;���������2���m�)�9�y��T��%�������p�IR���e�����AB�{>���AC�� �_�8�AB�<�AC� Z;���������1��*��;�������2���m�)�9�y��T���������8�C�5src:2453 ConstructiveGeometryFinalPreprintVersion.tex�By�H8induction�H9h��9yp�A�othesis,���there�are�terms��t�����1��r��and��t�����2��r��suc�h�H9that��AB��>��CAC� Z;���������1�����)���(�t�����1��*��)�H8is����+Ipro��9v��|rable�Z�and��AB�x
<�WAC� Z;���������2��/��)���(�t�����2��*��)�Z�is�Z�pro�v��|rable.��uLet��t�Z��b�A�e�the�term���if��
��(�AB�x>�XAC� Z;���t�����1��*��;�t�����2���).��uThen����+I�ECGD����pro��9v�es����AB��(>��sAC�(Z���t��r�=��t�����1��і�(b��9y�Axiom���76),��Uand��ECGD��pro��9v�es����AB��'<��sAC�(Z���t��r�=��t�����2��*��.����+IHence,��Zfor���some�list����of�axioms�of��ECGD�,�there�is�a�cut-free�pro�A�of�of��AB�{>���AC� Z;�����;�������1���m�)���(�t�),����+Iand���a�cut-free���pro�A�of�of��AB��<���AC� Z;�����;�������2��̆�)�����(�t�).�6These�t��9w�o���pro�A�ofs�can�then�b�A�e�com��9bined�as����+Ifollo��9ws:��U���8�C�5src:2465 ConstructiveGeometryFinalPreprintVersion.tex���j?��AB�{>���AC� Z;�����;�������1���m�)���(�t�)���AB�{<�AC� Z;���;�������2���m�)���(�t�)����������p��h����e����s�AB�{>���AC�� �_�8�AB�<�AC� Z;�����;�������1��*��;�������2���m�)���(�t�)��������8�C�5src:2468 ConstructiveGeometryFinalPreprintVersion.texThat�Tcompletes�the�induction�step,�and�with�it,�the�pro�A�of�of�the�lemma.����8�C�5src:2473 ConstructiveGeometryFinalPreprintVersion.tex�Example���1.��The�\other�interse��ction���p�oint".����Man��9y�V|Euclidean�V}constructions�in�v�olv�e�construct-����+Iing��Mone��Lin��9tersection�p�A�oin�t��P�0�of��La�line��L��d�=���e�Line��-�(�A;���B�r��)�and�a�circle��C����,��and�then�w��9e�sa�y��M\Let��G�@�color push gray 0��ff�ff�r�	J=�����"5��-:�9����LܻIn��fact,�
Iw�Îe�use�cut-elimination�for�man�y-sorted�logic�with�the�logic�of�partial�terms.���The�details�of�the�cut-��	��elimination��theorem��for�suc�Îh�logics�ha�Îv�e�not��b�<reen�published,�)ybut�they�are�not�signican�Îtly�dieren�t��from�Gen�tzen's���form�Îulation��Xfor�rst-order�logic.��ٛ�	color pop����G�color push gray 0����[�37������	color pop����&���s���������G�color push gray 0�����	color pop���[(��������+I�Q����b�A�e���the�other�in��9tersection�p�oin��9t���of��L��and��C����".�g$Of�course�w��9e�can�pro�v�e���\if��P�!��lies�on��L��and����+I�C�$o�and���the���t��9w�o�in�tersection���p�A�oin�ts�of��L����and��C�$o�are�not�equal,��Tthen�there�exists�an��x��suc��9h�that����+I�x��g�6�=��P�l��^��$�on��
���(�x;���L�)�$�^���On���3�(�x;�C����)."�~�Then�6b��9y�the�theorem,�>Mthere�m�ust�b�A�e�a�term��t�(�A;���B�r�;�C� Z;�P�H��)�6suc�h����+Ithat,�P�under�D�the�stated�conditions,��t�(�A;���B�r�;�C� Z;�P�H��)�D�is�an�in��9tersection�p�A�oin�t�of��L�D��and��C����that�is�not����+Iequal���to��P�H��.��*It�is�not�immediately�ob��9vious�what���this�term��t��migh�t�b�A�e,���and�it�w�ould�b�A�e�in�teresting����+Ito��extract��it�b��9y�computer�from�a�pro�A�of.��But�w��9e�should�b�A�e�able�to�see�directly�ho��9w�suc�h��a�term����+Icould�Tb�A�e�constructed.����8�C�5src:2481 ConstructiveGeometryFinalPreprintVersion.texHere��is��a�sk��9etc�h��of�suc��9h�a�construction.��First,���pro���ject�the�cen��9ter�of��C��n�on�to�line��L�,���obtaining����+Ip�A�oin��9t�J��R�Z��on��L�.���Since�the�t�w�o�in�tersection�p�A�oin�ts�are�distinct,�s,�R���6�=����P�H��.���Then�w�e�ask�whether�(�A;���B�r��)����+Iis���in���the�same�order�on��L��as�(�R�>;���P�H��)�or�not.�
�No��9w�Denition�71�sho��9ws�ho�w���to�construct�a�certain����+Ip�A�oin��9t�'?�E��r�(giv�en�b�y�'@a�term�in�v�olving�'@�A��and��B�r��)�suc��9h�that�(�A;���B�r��)�is�in�the�same�order�as�(�R�>;���P�H��)�if�and����+Ionly��if��E��2R�>P�_��is�a�left�turn.�!And�the�pro�A�of�of�Lemma�8�sho��9ws�ho�w�to�construct�a�(complicated)����+Iterm�B��`�(�A;���B�r�;�P�A�;�C����)�B�that�will�b�A�e�equal�to��
��Z�if�and�only�if��E��2R�>P����is�a�left�turn.��BIf��E��2R�>P����is�a�righ��9t�turn,����+Ithen�
�`�(�A;���B�r�;�P�A�;�C����)�can�b�A�e�arranged�to�b�A�e�another�sp�A�ecic�p�oin��9t����ǟ�-=�0�����on�the�other�side�of�����{6�from����+I�
����,��vthe�l�same�l�distance�from�line��������along�line���x,
��L�as�����,��vbut�on�the�other�side.��>Com��9bining�this�term����+I�`�,%�with�,&terms�represen��9ting�an�appropriate�dilation,�1�translation,�1�and�rotations,�w��9e�,&can�construct����+Ia�&term��f����(�A;���B�r�;�P�A�;�C����)�&suc��9h�that�if�(�A;���B��)�has�the�%same�order�as�(�R�>;���P�H��)�then�(�f����(�A;�B�r�;�P�A�;�C����)��$=��A�,����+Iand�~if�(�A;���B�r��)�~
has�the�opp�A�osite�order�as�(�R�>;���P�H��),��9it�is�another�p�A�oin��9t��A���-=�0��/��with��B�(�A;���B�r�;�A���-=�0�����.�V�Hence����+I(�f����(�A;���B�r�;�P�A�;�C����)�;�B��)�V�has�the�same�or�opp�A�osite�order�V�as�(�A;���B��),�gCdep�A�ending�on�whether�(�A;���B��)�has����+Ithe��`same��_or�opp�A�osite�order�as�(�R�>;���P�H��).�r�Note�that�if�(�A;���B�r��)�has�the�same�order�as�(�R�>;���P�H��),���then����+I�P�� �=���=�Interse��ctLineCir�cle2��V�
�(�L;���C����),�ۭwhile��if��not,��P��=���=�Interse��ctLineCir�cle1��V�
�(�L;���C����).��yHence��w��9e��can����+Itak��9e����i��t�(�A;���B�r�;�C� Z;�P�H��)���=���Interse��ctLineCir�cle1��U���(��Line��`��(�f����(�A;���B�;�P�A�;�C����)�;�B��)�;�C����)����+I�5src:2497 ConstructiveGeometryFinalPreprintVersion.texas�Tthe�\other�in��9tersection�p�A�oin�t"�constructor.����8�C�5src:2499 ConstructiveGeometryFinalPreprintVersion.texIn�m;case�m<one�thinks,��4\this�is�not�what�Euclid�had�in�mind!",��5that�is�of�course�true;�/but����+IEuclid��lnev��9er�tried�to�giv�e�uniform�constructions�of�this��kt�yp�A�e.�޸P�erhaps�the�complexit�y�of�this����+Iconstruction���is�one���reason�wh��9y�not.��7This�example�sho�ws���that�one�could�conserv��|rativ�ely���add�a����+Ifunction�Tsym��9b�A�ol�to��ECG��for�\the�other�in�tersection�p�A�oin�t".������&I�color push gray 0��Theorem���7�(Geometric�constructions�extracted�from�classical�pro�K�ofs)�	color pop���r=D�5src:2506 ConstructiveGeometryFinalPreprintVersion.tex�Supp��ose�\u�ECG����+I�with�f�classic��al�lo�gic�f�pr�oves��P�H��(�x�)�����9�y��T�(�x;���y�R��)�f��wher�e��P��c�is�f�quantier-fr�e�e�and�disjunction-fr�e�e.���Then����+Ither��e�P�ar�e�terms�P��t�����1��*��(�x�)�;����:�:�:��
�;���t�����n��7�(�x�)��of��ECG��such�that��P�H��(�x�)��@���A��(�x;���t�����1��*��(�x�))��_���:�:�:����_���(�x;�t�����n��7�(�x�))�P��is�also����+Ipr��ovable�N<in��ECG��with�classic�al�lo�gic.����+I�5src:2512 ConstructiveGeometryFinalPreprintVersion.texPr��o�of�.�pThis�Tis�a�sp�A�ecial�case�of�Herbrand's�theorem.����+I�5src:2515 ConstructiveGeometryFinalPreprintVersion.tex�Example��2�.��Euclid's��bpro�A�of��aof�Bo�ok��aI,�Prop�osition��a2�pro��9vides�us�with�t��9w�o��bsuc�h�constructions,����+I�t�����1��*��(�A;���B�r�;�C����)���=��C�q�and�o��t�����2���(�A;���B�r�;�C����)�o�the�o�result�of�Euclid's�construction�of�a�p�A�oin��9t��D��Q�with��AD�Ӎ�=����B�r�C����,����+Iv��|ralid�8)if��A����6�=����B�r��.���Classically�w��9e�ha�v�e��8�A;���B�r�;�C�.��9�D�A��(�AD�
��=����B�C����),�@�but�8)w��9e�need�t�w�o�8(terms��t�����1��b��and��t�����2�����+I�to�Tco��9v�er�all�cases.����8�C�5src:2519 ConstructiveGeometryFinalPreprintVersion.tex�Example�mn3�.��ILet�7F�P��*�and�7G�Q��b�A�e�distinct�p�A�oin��9ts�and��L��a�giv��9en�line,�?�and��A�,��B�r��,�?�and�7F�C��/�p�A�oin��9ts�on����+I�L�,�>with�5��A�5��and��B����on�the�same�side�of��C����.�~.Then�there�exists�a�p�A�oin��9t��D�w��whic�h�is�equal�5�to��P�~��if��B����+I�is��b�A�et��9w�een���A��and��C����and�equal�to��Q��if��A��is�b�A�et��9w�een��B��_�and��C����.�7The�t�w�o��terms��t�����1��<Q�and��t�����2���for�this����+Iexample�Y�can�Y�b�A�e�tak��9en�to�b�e�Y�the�v��|rariables��P����and��Q�.��~One�term�will�not�suce,�j�since��D��u�cannot����+Idep�A�end�z�con��9tin�uously�z�on��A��and��B�r��,��<but�all�constructed�p�A�oin��9ts�do�dep�A�end�con��9tin�uously�z�on�their����+Iparameters.�pThis�Tclassical�theorem�is�therefore�not�constructiv��9ely�pro�v��|rable.����8�C�5src:2527 ConstructiveGeometryFinalPreprintVersion.texW��:�e�F�men��9tioned�F�ab�A�o�v�e�that��ECG�F��cannot�pro�v�e�F�an�y�non-trivial�disjunctiv�e�F�theorem.��WThat�is����+Ia�%simple�consequence�of�the�fact�that�its�&axioms�con��9tain�no�disjunction.�W��:�e�no�w�sp�A�ell�this�out:������&I�color push gray 0��Theorem���8�(�ECG�cannot�pro��Cv�e���a�non��Ctrivial�disjunctiv�e�theorem)�	color pop���b`�5src:2532 ConstructiveGeometryFinalPreprintVersion.tex�Supp��ose�\u�ECG��pr�oves����+I�H���(�x�)��l���m�P�H��(�x�)�n�_��Q�(�x�)�,��gwher��e��^�H��{�is��_ne�gative.��jThen�either��ECG��pr�oves��_�H���(�x�)��l���P�H��(�x�)��or��^�ECG����+I�pr��oves�F��H���(�x�)�T���T��Q�(�x�)�.�*(This�F�r�esult�dep�ends�F�only�on�the�lack�of�disjunction�in�the�axioms�of����+I�ECG�.)�����G�color push gray 0����[�38������	color pop����'�3�s���������G�color push gray 0�����	color pop���[(��������+I�5src:2537 ConstructiveGeometryFinalPreprintVersion.tex�Pr��o�of�.�b�Consider�,�a�cut-free�pro�A�of�of�,��;���H���(�x�)������P�H��(�x�)���_���Q�(�x�),�2�where��,�is�a�list�of�some�axioms�of����+I�ECG�.�AOT��:�racing�the�disjunction�up��9w�ards�AOin�the�pro�A�of,�LMif�w��9e�reac�h�a�place�where�the�disjunction����+Iw��9as���in�tro�A�duced�on���the�righ�t�b�A�efore�reac�hing�a�leaf���of�the�pro�A�of�tree,��mthen�w�e�can���erase�the�other����+Idisjunct��{b�A�elo��9w�that�in�tro�A�duction,��obtaining�a�pro�of��|of�one�disjunct�as�required.�r�If�w��9e�reac�h�a����+Ileaf��7of�the��8pro�A�of�tree�with��P�H��(�x�)��%�_��Q�(�x�)��7still�presen��9t�on�the�righ��9t,�*�then�it�o�A�ccurs�on�the�left,����+Iwhere�7�it�app�A�ears�7�p�ositiv��9ely��:�.��Its�7�descendan�ts�will�7�also�b�A�e�p�ositiv��9e,�@�so�it�cannot�7�participate�in�in����+Iapplication���of���the�rule�for�pro�A�of�b��9y�cases�(whic��9h�in�tro�A�duces����_��in�the�left�side�of�a�sequen��9t);��and����+Iit�|Qcannot�reac��9h�|Rleft�side�of�the�b�A�ottom�sequen��9t,��namely��;���H���(�x�),�as�|Qthese�form��9ulas�con�tain�no����+Idisjunction.�[[But��a��glance�at�the�rules�of�cut-free�pro�A�of,��6e.g.�[[on�p.�442�of��[�13��	?�],��6will�sho��9w�that����+Ithese�Tare�the�only�p�A�ossibilities.�pThat�completes�the�pro�of.��;?����&I�color push gray 0��Theorem���9�(Disjunction�Prop�K�erties�for�ECG�and�ECGD)�	color pop���:���5src:2551 ConstructiveGeometryFinalPreprintVersion.tex�Supp��ose�N<�ECGD��pr�oves��y����}P�H���(�x�)������P�H��(�x�)�8�_��Q�(�x�)��y���+I�5src:2552 ConstructiveGeometryFinalPreprintVersion.tex�wher��e�N<�H�Y�is�ne�gative.�@Then�ther�e�is�a�term��t�(�x�)��of��ECGD��such�that��ECG��pr�oves�����@�H���(�x�)������t�(�x�)�=�����_�8�t�(�x�)�=������+I�5src:2553 ConstructiveGeometryFinalPreprintVersion.tex�and�N<�ECGD��pr��oves������F�t�(�x�)���=��������P�H��(�x�)�8�^��t�(�x�)���=���	����Q�(�x�)�:��|ԍ�+I�5src:2554 ConstructiveGeometryFinalPreprintVersion.tex�(Her��e�N<��P�and����h�ar�e�two�c�onstants�of��ECGD�,�with������6�=������h�an�axiom.)�����+I�5src:2559 ConstructiveGeometryFinalPreprintVersion.texPr��o�of�.�pSupp�A�ose�T�ECGD��pro��9v�es�T�H���(�x�)������P�H��(�x�)�8�_��Q�(�x�).�Then�Talso��ECGD��pro��9v�es�����G��H���(�x�)�����9�y��T�((�y��p�=��������P�H��(�x�))�8�^��(�y��=�����	����Q�(�x�)))�:����+I�5src:2562 ConstructiveGeometryFinalPreprintVersion.tex�The�Tform��9ula�on�the�righ�t�is�disjunction-free,�so�b�y�Theorem�6,�there�is�a�term��t��as�required.��])���+I�14��Kw�Conclusions��阍�+I�5src:2566 ConstructiveGeometryFinalPreprintVersion.tex�W��:�e�f�ha��9v�e�f�giv�en�a�quan�tier-free,���disjunction-free�f�axiomatization��ECG��of�Euclidean�Constructiv��9e����+IGeometry��:�,�>Xmaking�6$use�of�the�Logic�of�P��9artial�6%T�erms�(�LPT�).�W�e�ha��9v�e�6$v�eried�that�this�theory����+Iis��ea�reasonable�v��9ersion��dof�in�tuitionistic�geometry��:�,��ib�y�c�hec�king�that�its��dmo�A�dels�are�planes�o�v�er����+IEuclidean�|�elds.�SmP��9ast�|�v�ersions�of�in�tuitionistic�geometry�|�ha�v�e�included�either�|�apartness�or����+Idecidable�Չequalit��9y��:�.�]Both�of�these�destro�y�the�prop�A�ert�y�of�con�tin�uit�y�that�the�terms�of��ECG����+I�p�A�ossess.��cThe��Oterms��Pof�this�theory�corresp�A�ond�in�a�natural�w��9a�y�to��OEuclidean�straigh�tedge-and����+Icompass���constructions.���Making�use���of�more-or-less�standard�pro�A�of-theoretical�to�ols,��w��9e���ha�v�e����+Isho��9wn��ethat�pro�A�ofs�of�existen�tial�theorems�con�tain��fEuclidean�constructions�of�the�ob���jects�pro�v�ed����+Ito���exist,��Wand���that�these�constructions�can�b�A�e�automatically�extracted�from�suc��9h�pro�A�ofs.���As�a����+Icorollary��:�,�Tob���jects�pro��9v�ed�Tto�exist�in��ECG��dep�A�end�con��9tin�uously�Ton�parameters.����8�C�5src:2575 ConstructiveGeometryFinalPreprintVersion.texW��:�e�Tset�out�to�pursue�the�analogy��K�������-ꍒ��formal�Tn��9um�b�A�er�theory��w���H�fewϞ�'m��TT��:�uring�computable�functions������Q�=�����-ꍑ�pin��9tuitionistic�Tgeometry�����H�fe_��'m�geometric�Tconstructions�������I��+Iand��w��9e�think��that�w�e�ha�v�e�found�the�correct��theory��ECG��to�place�on�the�righ��9t�side�of�this��
�j��+Iequation.����8�C�5src:2580 ConstructiveGeometryFinalPreprintVersion.texW��:�e�pXha��9v�e�sho�wn�that�pWEuclid�is�essen�tially�constructiv�e,��in�the�pWpro�A�cess�exp�osing�these�in��9ter-����+Iesting�Tfacts:��|Ս���+I�color push gray 0��`����	color pop���AI�5src:2583 ConstructiveGeometryFinalPreprintVersion.tex�in�Tconstructiv��9e�geometry��:�,�w�e�need�a�rigid�compass�(for�the�uniform�v�ersion�of�Prop.�pI.2).��|ԍ���+I�color push gray 0��`����	color pop���AI�5src:2584 ConstructiveGeometryFinalPreprintVersion.tex�W��:�e�>�need�Mark��9o�v's�>�principle�to�pro��9v�e�>�the�fundamen��9tal�prop�A�erties�of�same-side�and�opp�osite-����AIside�Wof�a�line.��W��:�e�therefore�conclude�that�Mark��9o�v's�Wprinciple�Xis�fundamen��9tal�to�geometry;����AItheories�Twithout�it�do�not�corresp�A�ond�to�our�geometrical�in��9tuition.�����G�color push gray 0����[�39������	color pop����(��s���������G�color push gray 0�����	color pop���[(����������+I�color push gray 0��`����	color pop���AI�5src:2587 ConstructiveGeometryFinalPreprintVersion.tex�Dieren��9t�SBv�ersions�of�SCthe�parallel�p�A�ostulate�corresp�ond�to�dieren��9t�SCaxiom�systems�for�Eu-����AIclidean�}helds:��zEuclid's�}gp�A�ostulate�5�amoun��9ts�to�assuming�1�=x��is�dened�when��x���>��0��`�_��_�x�<��0,����AIwhile���Axiom�58�amoun��9ts�to�assuming�that��x����6�=�0���implies�1�=x��is�dened.�BIt�ma�y�b�A�e�simpler����AIto�'�determine�the�logical�'�relations�b�A�et��9w�een�'�these�eld�theories�than�to�w��9ork�directly�in����AIgeometrical�Ttheories.������+I�color push gray 0��`����	color pop���AI�5src:2590 ConstructiveGeometryFinalPreprintVersion.tex�Axiom���58���implies�Euclid's�P��9ostulate�5,��Pand�con��9v�ersely�with���the�aid�of�Axiom�75�(corre-����AIsp�A�onding�"to��x���6�=�0����0��<�x���_��x��<���0).�B�That�"the�rev��9erse�implications�"are�not�constructiv�ely����AIv��|ralid�Tis�sho��9wn�in�[�3����].������+I�color push gray 0��`����	color pop���AI�5src:2592 ConstructiveGeometryFinalPreprintVersion.tex�Since�c=�ECG��and��ECGD��corresp�A�ond�w��9ell�to�Euclid�c>and�to�constructiv�e�eld�theories,�v�w�e����AIconclude�Tthat�apartness�is�not�necessary�for�constructiv��9e�geometry��:�.�������+I�15��Kw�App�s3endix:�32List�ffof�axioms�of�ECG��阍�+I�5src:2663 ConstructiveGeometryFinalPreprintVersion.tex�In��$this�section�w��9e��#list�the�axioms�of��ECG��giv�en��#ab�A�o�v�e�without�commen�t,��Xfor��#reference.���The����+Iunderlying�;�logic�;�is�three-sorted�in��9tuitionistic�logic�(sorts�for�p�A�oin��9ts,�Elines,�and�;�circles)�with�the����+Ilogic�Tof�partial�terms��LPT��[�2����],�p.�p97.����8�C�5src:2667 ConstructiveGeometryFinalPreprintVersion.texIn�ގthe�ޏfollo��9wing�list,��the�sym��9b�A�ol�:=�is�used�for�a�denition�(macro).�
.The�sym��9b�A�ols�on�the�left����+Iside��are�to�b�A�e�replaced�(with�argumen��9t�substitution)��in�their�subsequen�t�uses�b�y�the�righ�t�hand����+Iside�Tof�the�denition.�p����,���x,�,�and��
����are�the�only�constan��9t�sym�b�A�ols�in��ECG�.����8�C�5src:2671 ConstructiveGeometryFinalPreprintVersion.texThe�Tunderlying�logic�of�LPT�pro��9vides�the�follo�wing:�����������8�x�(�t����#�^�A�(�x�)����A�(�t�))������������A�(�t�)�8�^��t����#�9�x���A�(�x�)������������f����(�t�����1��*��;����:�:�:��
�;���t�����n��7�)����#��t�����1���m�#�^���:�:�:��c��^�8�t�����n�����#������������R�>�(�t�����1��*��;����:�:�:��
�;���t�����n��7�)������t�����1���m�#�^���:�:�:��c��^�8�t�����n�����#������������t�����=����������Í�����=�����UP�s����:=�(�t��#��s��#�^�t��=��s�)�8�^��(�s����#��t��#�^�t��=��s�)������������x����#��'�for�Tev��9ery�v��|rariable��x������������c����#��'�for�Tev��9ery�constan�t��c�������8�C�5src:2682 ConstructiveGeometryFinalPreprintVersion.tex�In�Polisting�Ppthe�axioms�of��ECG�,�it�is�not�necessary�to�explicitly�indicate�the�t��9yp�A�es�(or�sorts)����+Iof��Cthe�v��|rariables,�&�as�this�can�b�A�e�mec��9hanically�deduced�from��Bthe�signature�of�the�relation�and����+Ifunction��;sym��9b�A�ols.��%(That��<is�one�reason�for�distinguishing���on���#�and���On���\�.)��&The�signatures�of�the����+Ifunction�ݾsym��9b�A�ols�ha�v�e�also�not�ݿb�A�een�explicitly�giv�en�here,���as�the�ݿnames�c�hosen�for�them�con�v�ey����+Ithat�Lainformation.��tThe�follo��9wing�are�the�L`axioms�of��ECG�,�including�the�denitions�used�in�stating����+Ithe�Taxioms.������Op�::�x����=��y��p���x��=��y�������;�color push gray 0�(1)�	color pop��������Op�::��V��(�A;���B�r�;�C� Z;�D�A��)��������(�A;���B�r�;�C� Z;�D�A��)�������;�color push gray 0(2)�	color pop��������Op�::��on��}��(�P�A�;���L�)�������on��s�(�P�;���L�)�������;�color push gray 0(3)�	color pop��������Op�::��On��
�!�(�P�A�;���C����)�������On��k��(�P�;���C����)�������;�color push gray 0(4)�	color pop��������Op�P�ک�=�����Interse��ctLines��<��(�L;���K����)������on�(�P�A�;�L�)�8�^���on��
���(�P�A�;�K����)�������;�color push gray 0(5)�	color pop���������Op�Interse��ctLines����K�(�L;���K����)�����=����������Í�����=������UP�Interse��ctLines��F���(�K� Z;�L�)�������;�color push gray 0(6)�	color pop��������Op�P�ک�=�����Interse��ctLineCir�cle1��U���(�L;���C����)�������on��s�(�P�A�;�L�)�8�^���On���Y�(�P�A�;�C����)�������;�color push gray 0(7)�	color pop��������Op�P�ک�=�����Interse��ctLineCir�cle2��U���(�L;���C����)�������on��s�(�P�A�;�L�)�8�^���On���Y�(�P�A�;�C����)�������;�color push gray 0(8)�	color pop��������Op�P�ک�=�����Interse��ctCir�cles1��G8)�(�C� Z;���K����)�������On��k��(�P�A�;�C����)�8�^���On���Y�(�P�A�;�K����)�������;�color push gray 0(9)�	color pop��������Op�P�ک�=�����Interse��ctCir�cles2��G8)�(�C� Z;���K����)�������On��k��(�P�A�;�C����)�8�^���On���Y�(�P�A�;�K����)������>=�color push gray 0(10)�	color pop���������Op�on��Z��(�P�A�;���L�)�8�^�:��on��}��(�P�;���K����)�������Interse��ctLines��<��(�L;�K����)��#������>=�color push gray 0�(11)�	color pop�������G�color push gray 0����[�40������	color pop����)�e�s���������G�color push gray 0�����	color pop���[(�����������Op�Interse��ctLines����K�(�L;���K����)����#�^��on��}��(�P�A�;�L�)�8�^���on��
���(�P�A�;�K����)������P�ک�=���Interse��ctLines��<��(�L;�K����)������>=�color push gray 0(12)�	color pop���������Op�on��Z��(�P�A�;���L�)�8�^���On���Y�(�P�;���C����)�������Interse��ctLineCir�cle1��U���(�L;�C����)��#������>=�color push gray 0�(13)�	color pop���������Op�on��Z��(�P�A�;���L�)�8�^���On���Y�(�P�;���C����)�������Interse��ctLineCir�cle2��U���(�L;�C����)��#������>=�color push gray 0�(14)�	color pop���������Op�On��]J#�(�P�A�;���C����)�8�^���On���Y�(�P�;���K����)�������Interse��ctCir�cles1��G8)�(�C� Z;�K����)��#������>=�color push gray 0�(15)�	color pop���������Op�On��]J#�(�P�A�;���C����)�8�^���On���Y�(�P�;���K����)�������Interse��ctCir�cles2��G8)�(�C� Z;�K����)��#������>=�color push gray 0�(16)�	color pop���������Op�Line��d���(�A;���B�r��)����#��$�����A��6�=��B������>=�color push gray 0�(17)�	color pop���������Op�Cir��cle��jx��(�A;���B�r��)����#������>=�color push gray 0�(18)�	color pop���������Op�Line��d���(��p��ointOn1��%��(�L�)�;�����p��ointOn2��(/��(�L�))���=��L������>=�color push gray 0�(19)�	color pop���������Op�p��ointOn1��t���(��Line��`��(�A;���B�r��))���=��A������>=�color push gray 0�(20)�	color pop���������Op�p��ointOn2��v��(��Line��`��(�A;���B�r��))���=��B������>=�color push gray 0�(21)�	color pop���������Op�p��ointOn1��t���(�L�)����6�=���p��ointOn2��)6��(�L�)������>=�color push gray 0(22)�	color pop���������Op�Cir��cle��jx��(��c��enter���(�C����)�;�����p��ointOnCir�cle��;1`�(�C��))���=��C������>=�color push gray 0�(23)�	color pop���������Op�c��enter��is�(��Cir��cle����(�A;���B�r��))���=��A������>=�color push gray 0�(24)�	color pop���������Op�p��ointOnCir�cle�����(��Cir��cle����(�A;���B�r��))���=��B������>=�color push gray 0�(25)�	color pop���������Op�c��enter��is�(�C����)����6�=���p��ointOnCir�cle��<8|�(�C��)������>=�color push gray 0(26)�	color pop��������Op�:��on��}��(���;�����Line���r�(��x,;���
����)������>=�color push gray 0(27)�	color pop��������Op�:��on��}��(��x,;�����Line���r�(���;���
����)������>=�color push gray 0(28)�	color pop��������Op�:��on��}��(�
���;�����Line���r�(���;����x,�)������>=�color push gray 0(29)�	color pop���������Op�on��Z��(�A;�����Line���r�(�A;���B�r��))������>=�color push gray 0(30)�	color pop���������Op�on��Z��(�B�r�;�����Line���r�(�A;���B��))������>=�color push gray 0(31)�	color pop���������Op�on��Z��(�P�A�;���L�)�8�^���on��
���(�Q;�L�)��^���on��
���(�R�>;���Line���r�(�P�A�;�Q�))�������on��s�(�R�>;�L�)������>=�color push gray 0(32)�	color pop��������Op�B�(�a;���b;�c�)������B�(�c;���b;�a�)������>=�color push gray 0(33)�	color pop��������Op�::�B�(�a;���b;�c�)������B�(�a;���b;�c�)������>=�color push gray 0(34)�	color pop��������Op�a����6�=��b�8�^��a����6�=��c�8�^��b����6�=��c��������>=�color push gray 0�(35)�	color pop��������a�(�:�B�(�a;���b;�c�)�8�^�:�B�(�b;���c;�a�)�����::�B�(�c;���a;�b�))��������a�(�:�B�(�b;���c;�a�)�8�^�:�B�(�c;���a;�b�)�����::�B�(�a;���b;�c�))��������a�(�:�B�(�c;���a;�b�)�8�^�:�B�(�a;���b;�c�)�����::�B�(�b;���c;�a�))��������a��:�(�B�(�a;���b;�c�)�8�^��B�(�b;���c;�a�))�8�^�:�(�B�(�a;���b;�c�)�8�^��B�(�b;���a;�c�))��������a��:�(�B�(�b;���c;�a�)�8�^��B�(�b;���a;�c�))��������Op�B�(�P�A�;�����Interse��ctLines��;���(�Line�(�P�;���Q�)�;�Line�(������>=�color push gray 0(36)�	color pop���������to��Interse��ctCir�cles1���U�(�C���ir�A�cl�&9e�(�P�;���Q�)�;�C�ir�A�cl�&9e�(�Q;�P�H��))�;���������to��Interse��ctCir�cles2���U�(�C���ir�A�cl�&9e�(�P�;���Q�)�;�C�ir�A�cl�&9e�(�Q;�P�H��)))�;�Q�)��������Op�B�(��Interse��ctLineCir�cle1��R���(�Line�(�P�A�;���Q�)�;�C���ir�cl�&9e�(�P�;�Q�))�;�P�;�Q�)������>=�color push gray 0(37)�	color pop��������Op�B�(�P�A�;���Q;���Interse��ctLineCir�cle2��T|z�(�Line�(�P�;�Q�)�;�C���ir�cl�&9e�(�Q;�P�H��)))������>=�color push gray 0(38)�	color pop���������Op�Opp��ositeSide����~�(�P�A�;���Q;�L�)���:=��B�(�P�;���Q;���Interse��ctLines��;���(��Line��`��(�P�;�Q�)�;�L�)������>=�color push gray 0(39)�	color pop���������Op�SameSide��w��(�P�A�;���Q;�L�)���:=��:�B�(�P�;���Q;���Interse��ctLines��;���(��Line��`��(�P�;�Q�)�;�L�))������>=�color push gray 0(40)�	color pop���������Op�SameSide��w��(�A;���B�r�;�L�)�8�^���SameSide��*Z8�(�B�;���C� Z;�L�)�������SameSide��*���(�A;���C�;�L�)������>=�color push gray 0(41)�	color pop���������Op�Opp��ositeSide����~�(�A;���B�r�;�L�)�8�^���Opp��ositeSide��7V��(�B�;���C� Z;�L�)�������SameSide��*���(�A;���C�;�L�)������>=�color push gray 0(42)�	color pop���������Op�on��Z��(�P�A�;�����R��ay��R��(�O�;���B�r��))���:=���on��s�(�P�A�;���Line���r�(�O�;�B�r��))�8�^�:�B�(�P�A�;�O�;�B�r��)������>=�color push gray 0(43)�	color pop���������Op�On��]J#�(�P�A�;�����Cir��cle�����(�A;���Q�))��$��	?���V��(�A;�P�;�A;�Q�)�8�^��A����6�=��Q������>=�color push gray 0�(44)�	color pop��������Op�A����6�=��B����^�8�C�5��6�=��D�Ӎ�������>=�color push gray 0�(45)�	color pop�������G�color push gray 0����[�41������	color pop����*�I�s���������G�color push gray 0�����	color pop���[(����������a��(��on��}��(��Interse��ctLineCir�cle1��R���(��Line��`��(�A;���B�r��)�;���Cir��cle3���J�(�A;�C� Z;�D�A��))�;���R��ay��R��(�A;�B��))�8�^��������a���V��(�A;�����Interse��ctLineCir�cle1��T|z�(��Line��`��(�A;���B�r��)�;���Cir��cle3���J�(�A;�C� Z;�D�A��))�;�C�;�D�A��)��������Op��V��(�A;���B�r�;�C� Z;�D�A��)�8�^����(�A;���B�r�;�E��2;�F�H��)��������(�C� Z;���D�A�;�E��2;�F�H��)������>=�color push gray 0(46)�	color pop��������Op�B�(�A;���B�r�;�C����)�8�^��B�(�P�A�;���Q;�R�>�)�8�^���V��(�A;���B�;�P�A�;�Q�)�8�^���V��(�B�;���C� Z;�Q;�R�>�)�������V��(�A;���C�;�P�A�;�R�>�)������>=�color push gray 0(47)�	color pop��������Op�AB�{<���C���D�Ӎ�:=��B�(�C� Z;�����Interse��ctLineCir�cle2��T|z�(��Line��`��(�C�;���D�A��)�;���Cir��cle3���J�(�C� Z;�A;�B�r��))�;�D��)������>=�color push gray 0(48)�	color pop��������Op�C���D�Ӎ�����AB�{�:=��:�B�(�C� Z;�����Interse��ctLineCir�cle2��T|z�(��Line��`��(�C�;���D�A��)�;���Cir��cle3���J�(�C� Z;�A;�B�r��))�;�D��)������>=�color push gray 0(49)�	color pop��������Op�AB�{�=����C���D�Ӎ�:=���V��(�A;���B�r�;�C� Z;�D�A��)������>=�color push gray 0(50)�	color pop��������Op�AB�{�=����P�H�Q�8�^��B�r�C�5��=��QR�v�^��AC�5��=��P�H�R�v�^��P�Q����=��U���V�p�^�8�QR���=��V�W�W�^��P�H�R���=��U���V����������>=�color push gray 0�(51)�	color pop��������a��AB�{�=����U���V�p�^�8�B�r�C�5��=��V�W�W�^��AC�5��=��U���W��������OpA����6�=��B����^�8�A��6�=��C�� �^��B�{�6�=��C�� �^��A����=�0��C��6�=��B��r����=�0��2��^������>=�color push gray 0�(52)�	color pop��������a��B��r����=�00���	�=�����Interse��ctLineCir�cle1��U���(��Line��`��(�A����=�0�����;���B��r����=�0��$|�)�;�C����)�8�^��������a��K�����1���m�=�����Cir��cle3�� �f�(�B��r����=�00��VC�;���B�r�;�C����)�8�^��K�����2���=�����Cir��cle3�� �f�(�A����=�00����;���A;�C����)�������������a��Interse��ctCir�cles1����]�(�K�����1��*��;���K�����2���)����#�^��Interse��ctCir�cles2��D�c�(�K�����1��*��;�K�����2���)����#�^���������a��Opp��ositeSide���8v�(��Interse��ctCir�cles1��D�c�(�K�����1��*��;���K�����2���)�;����������o��Interse��ctCir�cles2���E�(�K�����1��*��;���K�����2���)�;���Line���r�(�A����=�0�����;�B��r����=�0��$|�))��������Op�AP�ک�����AB����^��8�on��
���(�P�A�;���L�)�����Interse��ctLineCir�cle1��U���(�L;���Cir��cle�����(�A;�B�r��))��#������>=�color push gray 0�(53)�	color pop��������Op�AP�ک�����AB����^��8�on��
���(�P�A�;���L�)�����Interse��ctLineCir�cle2��U���(�L;���Cir��cle�����(�A;�B�r��))��#������>=�color push gray 0�(54)�	color pop��������Op�A����=���c��enter�����(�C����)�8�^��A����=���c��enter���(�K����)�8�^���On���Y�(�P�A�;���C��)��^���On���Y�(�Q;���K��)��^��AP�ک�=����AQ��������>=�color push gray 0�(55)�	color pop���������a��Interse��ctLineCir�cle1������(�L;���C����)�����=����������Í�����=������UP�Interse��ctLineCir�cle1��_G �(�K� Z;�C��)�8�^���������a��Interse��ctLineCir�cle2������(�L;���C����)�����=����������Í�����=������UP�Interse��ctLineCir�cle2��_G �(�K� Z;�C��)���������Op�On��]J#�(�P�A�;���C����)�8�^��AP�ک�����AB�{�������>=�color push gray 0�(56)�	color pop���������a��Interse��ctCir�cles1����]�(�C� Z;�����Cir��cle�����(�A;���B�r��))����#�^��T�Interse��ctCir�cles2��G���(�C�;�����Cir��cle�����(�A;���B�r��))��#��������Op�A����=���c��enter�����(�C�����1��*��)�8�^��A����=���c��enter���(�C�����1��*��)�8�^���On���Y�(�P�A�;���C�����1���)��^���On���Y�(�Q;���C�����1���)��^��AP�ک�=����AQ��������>=�color push gray 0�(57)�	color pop���������a��Interse��ctCir�cles1����]�(�C�����1��*��;���K����)�����=����������Í�����=������UP�Interse��ctCir�cles1��P���(�C�����2���;�K����)�8�^���������a��Interse��ctCir�cles2����]�(�C�����1��*��;���K����)�����=����������Í�����=������UP�Interse��ctCir�cles2��P���(�C�����2���;�K����)�8�^���������a��Interse��ctCir�cles1����]�(�K� Z;���C�����1��*��)�����=����������Í�����=������UP�Interse��ctCir�cles1��P���(�K�;�C�����2��*��)�8�^���������a��Interse��ctCir�cles2����]�(�K� Z;���C�����1��*��)�����=����������Í�����=������UP�Interse��ctCir�cles2��P���(�K�;�C�����2��*��)�8�^��������Op:��Interse��ctLines��:EI�(�K� Z;���L�)����#�^�on�(�p;�K����)�8�^��on�(�p;�M����)��^��M����6�=����K�5�����Interse��ctLines��<��(�L;�M����)������>=�color push gray 0(58)�	color pop���������Op�L��eft��b7��(�A;���B�r�;�C����)���:=��C�5��=���Interse��ctCir�cles1��G8)�(��Cir��cle����(�A;���C��)�;���Cir��cle�����(�B�r�;�C��))������>=�color push gray 0(59)�	color pop���������Op�R�Îight��h�(�A;���B�r�;�C����)���:=��C�5��=���Interse��ctCir�cles2��G8)�(��Cir��cle����(�A;���C��)�;���Cir��cle�����(�B�r�;�C��))������>=�color push gray 0(60)�	color pop���������Op�L��eft��b7��(���;����x,;�
����)������>=�color push gray 0(61)�	color pop���������Op�R�Îight��h�(���;���
���;��x,�)������>=�color push gray 0(62)�	color pop���������Op�L��eft��b7��(�P�A�;���Q;�R�>�)�8�^��P�ک�6�=����P��H��=�0����^��R���6�=��R��>���=�0���=�^��on�(�P��H��=�0�����;�����R��ay��R��(�Q;���P�H��))��^���on��
���(�R��>���=�0����;���R��ay��R��(�Q;�R�>�))������>=�color push gray 0(63)�	color pop��������d��������L��eft��Y��(�P��H��=�0�����;���Q����=�0�����;�R��>���=�0����)���������Op�L��eft��b7��(�P�A�;���Q;�R�>�)�8�^�:�B�(�P�;�����Interse��ctLines��;���(��Line��`��(�Q;���R�>�)�;���Line���r�(�P�;�P��H��=�0�����))�;�P��H��=�0���)������>=�color push gray 0(64)�	color pop��������d��������L��eft��Y��(�P��H��=�0�����;���Q;�R�>�)���������Op�L��eft��b7��(�P�A�;���Q;�R�>�)�8�^�:�B�(�R�;�����Interse��ctLines��;���(��Line��`��(�Q;���P�H��)�;���Line���r�(�R�;�R�����=�0����))�;�P��H��=�0�����)������>=�color push gray 0(65)�	color pop��������d��������L��eft��Y��(�P�A�;���Q;�R��>���=�0����)���������Op�L��eft��b7��(�A;���B�r�;�C����)�8�^��AB�{�=����P�H�Q��^��B�C�5��=����QR�v�^������>=�color push gray 0�(66)�	color pop��������a��AC�5��=����P�H�R�v�^�8�AP�ک�=��B�r�Q��^��AP�ک�=��C���R������L��eft��Y��(�P�A�;���Q;�R�>�)��������Op�P�ک�6�=����P��H��=�0����^�8�R���6�=��R��>���=�0���=�^��on�(�P��H��=�0�����;�����R��ay��R��(�Q;���P�H��))��^���on��
���(�R��>���=�0����;���R��ay��R��(�Q;�R�>�))��^������>=�color push gray 0�(67)�	color pop�������G�color push gray 0����[�42������	color pop����+�J�s���������G�color push gray 0�����	color pop���[(�����������a��R�Îight��z���(�P�A�;���Q;�R�>�)�������R�Îight��@��(�P��H��=�0�����;���Q����=�0�����;�R�����=�0����)���������Op�R�Îight��h�(�P�A�;���Q;�R�>�)�8�^�:�B�(�P�;�����Interse��ctLines��;���(��Line��`��(�Q;���R�>�)�;���Line���r�(�P�;�P��H��=�0�����))�;�P��H��=�0���)������>=�color push gray 0(68)�	color pop��������d��������R�Îight��@��(�P��H��=�0�����;���Q;�R�>�)���������Op�R�Îight��h�(�P�A�;���Q;�R�>�)�8�^�:�B�(�R�;�����Interse��ctLines��;���(��Line��`��(�Q;���P�H��)�;���Line���r�(�R�;�R�����=�0����))�;�P��H��=�0�����)������>=�color push gray 0(69)�	color pop��������d��������R�Îight��@��(�P�A�;���Q;�R��>���=�0����)���������Op�R�Îight��h�(�A;���B�r�;�C����)�8�^��AB�{�=����P�H�Q��^��B�C�5��=����QR�� �^������>=�color push gray 0�(70)�	color pop��������a��AC�5��=����P�H�R�v�^�8�AP�ک�=��B�r�Q��^��AP�ک�=��C���R������R�Îight��@��(�P�A�;���Q;�R�>�)���������Op�SameOr��der���J�(�A;���B�r�;�P�A�;�Q�)���:=������>=�color push gray 0(71)�	color pop��������a��A����6�=��B����^�8�P�ک�6�=��Q��^���on��
���(�P�A�;�����Line���r�(�A;���B�r��))��^���on���(�Q;�����Line���r�(�A;���B�r��))��^���������a��L��eft��t���(�P�A�;���Q;���Interse��ctCir�cles1��F1
�(��Cir��cle����(�A;�B�r��)�;���Cir��cle�����(�B�;�A�))��������Op�P�ک�=�����Interse��ctLineCir�cle1��U���(��Line��`��(�A;���B�r��)�;�C����)�8�^������>=�color push gray 0�(72)�	color pop��������a��Q����=���Interse��ctLineCir�cle2��U���(��Line��`��(�A;���B�r��)�;�C����)�8�^��P�ک�6�=����Q��������d��������SameOr��der��2��(�A;���B�r�;�P�A�;�Q�)��������Op�R���=�����Interse��ctCir�cles1��G8)�(��Cir��cle����(�A;���P�H��)�;���Cir��cle�����(�B�r�;�Q�))�����:��R�Îight����(�A;���B�;�R�>�)������>=�color push gray 0(73)�	color pop��������Op�R���=�����Interse��ctCir�cles2��G8)�(��Cir��cle����(�A;���P�H��)�;���Cir��cle�����(�B�r�;�Q�))�����:��L��eft�����(�A;���B�;�R�>�)������>=�color push gray 0(74)�	color pop��������Op�B�{�6�=����C�� �^�8�B�(�A;���B�r�;�D�A��)��^��B�(�A;�C� Z;�D�A��)������B�(�A;���B�r�;�C����)��_��B�(�A;�C� Z;�B�r��)������>=�color push gray 0(75)�	color pop��������Op(�AB�{>���C���D�Ӎ���if����(�AB�>�C���D�A�;���P�;�Q�)���=��P�H��)�8�^��(�AB�{<���C���D�Ӎ���if����(�AB�>�C���D�A�;���P�;�Q�)���=��Q�)������>=�color push gray 0(76)�	color pop��������+I�References��阍���+I�color push gray 0����[1]�	color pop���>t��5src:2807 ConstructiveGeometryFinalPreprintVersion.texAvigad,�;nJ.,�Dean,�Edw��9ard,�and��Mumma,�John,�A�`formal�system�for��Euclid's��Elements�,����>t��R��eview�N<of�Symb�olic�L�o�gic�T�(to�app�A�ear�in�2009).������+I�color push gray 0���[2]�	color pop���>t��5src:2811 ConstructiveGeometryFinalPreprintVersion.texBeeson,���M.,��F��J�oundations�Zof�Constructive�Mathematics�,���Springer-V��:�erlag,�Berlin/�8�Heidel-����>t�b�A�erg/�TNew�Y��:�ork�(1985).������+I�color push gray 0���[3]�	color pop���>t��5src:2814 ConstructiveGeometryFinalPreprintVersion.texBeeson,��IM.,�Constructiv��9e�`geometry�`and�the�parallel�p�A�ostulate,��Ito�app�ear.�`Preprin��9t�a�v��|railable����>t�on�Tthe�author's�w��9ebsite.������+I�color push gray 0���[4]�	color pop���>t��5src:2818 ConstructiveGeometryFinalPreprintVersion.texBishop,�TE.,��F��J�oundations�N<of�Constructive�A�Înalysis�,�TMcGra��9w-Hill,�New�Y��:�ork�(1967).������+I�color push gray 0���[5]�	color pop���>t��5src:2821 ConstructiveGeometryFinalPreprintVersion.texBorsuk,�J�K.��{and�Szmielew,�J�W.,��F��J�oundations��of�Ge��ometry�,�J�Amsterdam,�North��{Holland����>t�(1960).������+I�color push gray 0���[6]�	color pop���>t��5src:2824 ConstructiveGeometryFinalPreprintVersion.texDehn,���Max,���Die�l�Grundlegung�der�l�Geometrie�in�historisc��9her�En�t�wic�klung,���an�l�app�A�endix�in����>t�[�16��	?�].������+I�color push gray 0���[7]�	color pop���>t��5src:2827 ConstructiveGeometryFinalPreprintVersion.texDescartes,���R.,��L��a���Ge�ometrie�.�[�F��:�acsimile�[�with�English�translation�published�as��The���Ge��ometry����>t�of�N<R��ene�Desc�artes�,�TDo��9v�er,�New�Y��:�ork�(1954).������+I�color push gray 0���[8]�	color pop���>t��5src:2830 ConstructiveGeometryFinalPreprintVersion.texEuclid,�q,�The���Thirte��en���Bo�oks�of���the�Elements�,�q,second�H!edition�H"of�the�Heath�translation,�q,Do��9v�er,����>t�New�TY��:�ork�(1956).������+I�color push gray 0���[9]�	color pop���>t��5src:2833 ConstructiveGeometryFinalPreprintVersion.texGreen��9b�A�erg,��NM.��KJ.,��Euclide��an��Oand��PNon-Euclide�an�Ge�ometries�,��Mfourth��Ledition,�F��:�reeman,��NNew����>t�Y��:�ork�T(2008).������+I�color push gray 0[10]�	color pop���>t��5src:2835 ConstructiveGeometryFinalPreprintVersion.texHeyting,�TA.,��Intuitionism,�N<A�În�Intr��o�duction�,�TNorth-Holland,�Amsterdam�(1956).������+I�color push gray 0[11]�	color pop���>t��5src:2838 ConstructiveGeometryFinalPreprintVersion.texHilb�A�ert,�…D.,��F��J�oundations�͎of�͏Ge��ometry�,�„translated���from�the���ten��9th�German�edition�b��9y�Leo����>t�Unger,�TOp�A�en�Court,�La�Salle,�Illinois�(1971).������+I�color push gray 0[12]�	color pop���>t��5src:2840 ConstructiveGeometryFinalPreprintVersion.texHilb�A�ert,��D.,�and��Berna��9ys,�P��:�.,��Grund�x�lagen�2bder�2aMathematik�,�second�edition,�Springer-V��:�erlag����>t�(1970).������+I�color push gray 0[13]�	color pop���>t��5src:2842 ConstructiveGeometryFinalPreprintVersion.texKleene,�TS.�C.,��Intr��o�duction�N<to�Metamathematics�,�Tv��|ran�Nostrand,�Princeton�(1952).�����G�color push gray 0����[�43������	color pop����,��s���������G�color push gray 0�����	color pop���[(����������+I�color push gray 0�[14]�	color pop���>t��5src:2845 ConstructiveGeometryFinalPreprintVersion.texKleene,��S.��C.,��P��9erm�utabilit�y�of��inferences�in�Gen�tzen's�calculi�LK��and�LJ,�in:�	�,�q[�		cmsl9�Tw�o�P�ap�A�ers����>t�on�Tthe�Predicate�Calculus�,�A.M.S.�Memoirs��10��(1952),�A.�M.�S.,�Pro��9vidence,�R.�I.������+I�color push gray 0[15]�	color pop���>t��5src:2851 ConstructiveGeometryFinalPreprintVersion.texKijne,���D.,����Plane�`Construction�`Field�The��ory�,�Ph.�?RD.�?Qthesis,�Rijksuniv��9ersiteit�te�?QUtrec�h�t����>t�(1956).������+I�color push gray 0[16]�	color pop���>t��5src:2854 ConstructiveGeometryFinalPreprintVersion.texP��9asc�h,�gM.,�gand�V�Dehn,�M.,��V��J�orlesungen��.a�������IUb��er��HNeuer�e�Ge�ometrie�,�gzw��9eite�V�Au
age,�Springer,����>t�Berlin�#=(1926).�#<Note,�&�the�rst�edition�(1882),�&�whic��9h�is�the�one�digitized�b��9y�Go�A�ogle�Sc��9holar,����>t�do�A�es�Tnot�con��9tain�the�app�endix�b��9y�Dehn.������+I�color push gray 0[17]�	color pop���>t��5src:2858 ConstructiveGeometryFinalPreprintVersion.texv��9on��LPlato,�ԊJan,�The�axioms��Mof�constructiv��9e�geometry��:�,��A�Înnals���of�Pur��e���and�Applie�d�L�o�gic����>t��76�,�Tpp.�169{200�(1995)�.������+I�color push gray 0[18]�	color pop���>t��5src:2861 ConstructiveGeometryFinalPreprintVersion.texN.�5�Moler�5�and�P��:�.�Supp�A�es,�>Quan��9tier-free�axioms�for�constructiv��9e�plane�geometry��:�,�>�Comp��os.����>t�Math.�T�20�,�pp.�143-152�(1968).������+I�color push gray 0[19]�	color pop���>t��5src:2864 ConstructiveGeometryFinalPreprintVersion.texP��9am�buccian,�ZV.,�ZAxiomatizing�+.geometric�+-constructions,��J.�wApplie��d�wL�o�gic��6�,�Zpp.�24{46,�2008.������+I�color push gray 0[20]�	color pop���>t��5src:2867 ConstructiveGeometryFinalPreprintVersion.texP��9am�buccian,�ɾV.,�ɿAxiomatizations���of���h�yp�A�erb�olic�and�absolute�geometries,�ɾin��Non-Euclide��an����>t�ge��ometries,���J��Ganos��4Bolyai��3memorial�volume.�Pap�ers�fr�om�the��3International�Confer�enc�e�on����>t�Hyp��erb�olic��Ge�ometry�held�in��Budap�est,�tJuly�6{12,�s2002�,���pp.�ɺ119{153.�ɹEdited�b��9y�Andr��`as����>t�Pr������Xek��9opa�4�and�Emil�4�Moln��`ar.�Mathematics�and�Its�Applications�(New�Y��:�ork),�<��581�.�Springer,����>t�New�TY��:�ork�(2006).�����G�color push gray 0����[�44������	color pop����+����;�s�ï�,�,�q[�		cmsl9�)ߤN		cmtt9�(�':

cmti10�&���@ffcmti12� �E�tcmbx6�#�f�cmti8���N�ffcmbx12�����		cmsy9�5��"		cmmi9��j��		cmti9�t�:		cmbx9�o���		cmr9�q�%cmsy6�;�cmmi6��2cmmi8��Aa�cmr6�|{Ycmr8�X�Qcmr12�D��tG�G�cmr17�
�b>

cmmi10�K�`y

cmr10���u

cmex10�5e����

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