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/dn.dvi

����;� TeX output 2005.08.26:2252�������j֍����*�1F'����-.��src:25dn.tex�D��tG�G�cmr17�Double-Negation�7tElimination�in�Some�Prop�s�ositional������pLogics��lύ�����W<�����@cmti12�Michael�35Be��ffeson������=Qr�X�Qcmr12�San��Jose�State�Univ��rersit�y������:��Math��&�Computer�Science������L	�San��Jose,�CA�95192������������R��ffob�ert�35V���er�o�������E[�Univ��rersit�y��of�New�Mexico��������Departmen��rt��of�Computer�Science���������Albuquerque,��NM�87131������=��������C�L��ffarry�35Wos������m��Mathematics��and�Computer�Science�Division��������Argonne��National�Lab�S�oratory�������*Argonne,��IL�60439-4801������BY�����August��26,�2005��?�y��-���N�cmbx12�Abstract��q���-�src:41dn.tex�K�`y
�3
cmr10�This�0earticle�answ��!ers�t�w�o�questions�(p�M�osed�in�the�literature),�G�eac�h�concerning��
����-the�Vxguaran��!teed�existence�of�pro�M�ofs���':
�3
cmti10�fr��p�e�e���of�double�ne��p�gation�.��A�VKpro�of�Vxis�free����-of��,double�negation�if�none�of�its�deduced�steps�con��!tains�a�term�of�the�form����-��b>
�3
cmmi10�n�(�n�(�t�))���for�some�term��t�,��lwhere��n��denotes�negation.��The�rst�question�asks����-for�;�conditions�on�the�h��!yp�M�otheses�that,�P�if�satised,�guaran��!tee�the�existence�of����-a���double-negation-free�pro�M�of�when�the�conclusion�is�free�of�double�negation.����-The��second�question�asks�ab�M�out�the�existence�of�an�axiom�system�for�clas-����-sical�M�prop�M�ositional�calculus�whose�use,�w�for�theorems�with�a�conclusion�free����-of�"�double�negation,�=Dguaran��!tees�the�existence�of�a�double-negation-free�pro�M�of.����-After��sgiving�conditions�that�answ��!er�the�rst�question,�εw�e�answ�er�the�sec-����-ond��iquestion�b��!y�fo�M�cusing�on�the��W7 ����L��
�juk��dDasiewicz�three-axiom�system.���W��ee�then����-extend�x�our�studies�to�innite-v��dDalued�sen��!ten�tial�x�calculus�and�to�in��!tuitionistic����-logic�%�and�generalize�the�notion�of�b�M�eing�double-negation�free.�\YThe�double-����-negation�wpro�M�ofs�of�in��!terest�rely�exclusiv�ely�on�the�inference�rule�condensed����-detac��!hmen�t,�f�a�V�rule�that�com��!bines�mo�M�dus�p�onens�with�an�appropriately�gen-����-eral��nrule�of�substitution.�}�The�automated�reasoning�program�Otter�pla��!y�ed����-an��findisp�M�ensable�role�in�this�study��e.���"$�����31����*��j֍����*�1F'���홊���-�1��A@Origin��of�the�Study��q���-�src:62dn.tex�This�Rarticle�features�the�culmination�of�a�study�whose�origin�rests�equally��
����-with��Vt��!w�o�questions,���the�rst�p�M�osed�in��Studia��8L��p�o�gic�a��[�2��y�]�and�the�second�(closely����-related��fto�the�rst)�p�M�osed�in�the��Journal���of�A��\utomate��p�d�R�e�asoning��f�[�15��
�4].����>�src:64dn.texBoth���questions�fo�M�cus�on��double-ne��p�gation-fr�e�e�pr�o�ofs�,���pro�ofs���none�of�whose����-deduced���steps�con��!tain�a�form�ula�of�the�form��n�(�n�(�t�))�for�some�term��t��with�the����-function���n��denoting�negation.��bF��eor�example,�(rwhere��i��denotes�implication,�the����-presence��(of�the�form��!ula��i�(�i�(�n�(�x�)�;��1x�)�;�x�)��(as�a�deduced�step�do�M�es�not�preclude�a����-pro�M�of�âfrom�b�eing�double-negation�free,���whereas�the�presence�of�the�form��!ula����-�i�(�n�(�n�(�x�))�;��1x�)��do�M�es.���Note�the�distinction�b�et��!w�een��deduced�steps�and�axioms;����-in�G�particular,��/use�of�the�F��erege�system�for�t��!w�o-v��dDalued�G�sen�ten�tial�calculus,����-whic��!h��icon�tains�t�w�o�axioms�in�whic�h�double�negation�o�M�ccurs,��guaran�tees�the����-existence��fof�double-negation-free�pro�M�ofs,�as�w��!e�sho�w�in�Section�5.����>�src:68dn.texThe�
)sough��!t-after�double-negation-free�pro�M�ofs�of�in�terest�here�rely�solely����-on�ƅthe�inference�rule�condensed�detac��!hmen�t�ƅ[�10��
�4],��Ka�rule�that�com��!bines�mo�M�dus����-p�M�onens��rwith�an�appropriately�general�rule�of�substitution.��F��eormally�,�εcon-����-densed�9�detac��!hmen�t�considers�t�w�o�form�ulas,�O��i�(�A;��1B����)�(the�ma���jor�premiss)�and����-�C��q�(the���minor�premiss),��that�are�tacitly�assumed�to�ha��!v�e���no�v��dDariables�in�com-����-mon��Eand,���if��C�Iƹunies�with��A�,�yields�the�form��!ula��D�M޹,�where��D��#�is�obtained�b��!y����-applying��fto��B�3
�a�most�general�unier�of��C�n�and��A�.����>�src:71dn.texIn��
[�2��y�],��the�follo��!wing�question�is�ask�ed.�`�Where��P�Wc�and��Q��ma�y�eac�h�b�M�e����-collections�$of�form��!ulas,���if��T��r�is�a�theorem�asserting�the�deducibilit�y�of��Q����-�from�Z��P���suc��!h�that��Q��is�free�of�double�negation,���what�conditions�guaran�tee����-that��]there�exists�a�pro�M�of�of��T����(relying�solely�on�condensed�detac��!hmen�t)��]all����-of�j�whose�deduced�steps�are�free�of�double�negation?�*�Then,���in�[�15��
�4],�Dolph����-Ulric��!h��asks�ab�M�out�the�existence�of�an�axiom�system�for�t�w�o-v��dDalued�sen�ten�tial����-(or���classical�prop�M�ositional)�calculus�suc��!h�that,��for�eac�h�double-negation-free����-form��!ula����Q��pro�v��dDable�from�the�axiom�system,�Ĭthere�exists�a�double-negation-����-free��fpro�M�of�of��Q�.����>�src:75dn.texAlthough���it�is�p�M�erhaps�not�ob��!vious,���the�nature�of�the�axioms�c�hosen�for����-the��study�of�some�area�of�logic�or�mathematics�can�ha��!v�e��a�mark��!ed�impact����-on�*�the�nature�of�the�pro�M�ofs�deriv��!ed�from�them.�	kUAs�a�most�enligh�tening����-illustration���of�this�relation�and�indeed�p�M�ertinen��!t�to�the�t�w�o�cited�questions����-(eac��!h�E�of�whic�h�w�e�answ�er�in�this�article),�m~w�e�turn�to�an�example�giv�en�b�y����-Ulric��!h���that�builds�on�a�result�of�C.�A.�Meredith.��PIn�the�early�1950s,��zMeredith����-found��fthe�follo��!wing�21-letter�single�axiom�for�t�w�o-v��dDalued�logic.���;��gf"�i�(�i�(�i�(�i�(�i�(�x;��1y�d��)�;�i�(�n�(�z�{I�)�;�n�(�u�)))�;�z��)�;�v�d��)�;�i�(�i�(�v�;�x�)�;�i�(�u;�x�)))����>�src:82dn.texConsider��Athe�follo��!wing�system�with�condensed�detac�hmen�t�as�the�sole���"$�����32����	Π�j֍����*�1F'���홊��-�rule��of�inference�and�four�double-negation-free�classical�theses�(of�t��!w�o-v��dDalued��
����-logic)�V�as�axioms.��(The�notation�here�is�tak��!en�from�Ulric�h�[�15��
�4]�and�should����-�not��_�b�M�e�confused�with�that�used�for�innite-v��dDalued�logic�discussed�in�Section����-7.)����>�src:84dn.tex��.����-A1������d�i�(�x;��1x�)����������-A2�������&�i�(�i�(�x;��1x�)�;�i�(�n�(�x�)�;�i�(�n�(�x�)�;�n�(�x�))))��������-A3������ɿ�i�(�i�(�x;��1i�(�x;�x�))�;�i�(�n�(�x�)�;�i�(�n�(�x�)�;�i�(�n�(�x�)�;�n�(�x�)))))��������-A4�����D��i�(�i�(�x;��1i�(�x;�i�(�x;�x�)))�;�i�(�i�(�i�(�i�(�i�(�y�d�;�z�{I�)�;�i�(�n�(�u�)�;�n�(�v��)))�;�u�)�;�w�Kn�)�;�i�(�i�(�w�;�y�d��)�;�i�(�v�;�y��))))������>�src:91dn.texOne��^can�readily�v��!erify�that�axiom�A1�and�the�an�teceden�t�(left-hand�ar-��
����-gumen��!t)�\�of�A2�are�uniable�but�that�no�other�axiom�is�uniable�with�the����-an��!teceden�t�o|of�an��!y�axiom.�ˏIn�other�w�ords,�zwno�conclusion�can�b�M�e�dra�wn�(with����-condensed��/detac��!hmen�t)�other�than�b�y�considering�A1�and�A2.�q7Therefore,����-the��frst�step�of�an��!y�pro�M�of�in�this�system�can�only�b�e��p���-�src:97dn.tex5����i�(�n�(�x�)�;��1i�(�n�(�x�)�;�n�(�x�)))�:����>�
src:100dn.tex�Similarly��e,��5the�w�only�new�path�of�reasoning�no��!w�a�v��dDailable�is�that�of�5�with��
����-the���an��!teceden�t�of�A3.��jTherefore,�Kthe�next�step�in�an�y�pro�M�of�in�this�system����-can��fonly�b�M�e��p���-�
src:105dn.tex6����i�(�n�(�n�(�x�))�;��1i�(�n�(�n�(�x�))�;�i�(�n�(�n�(�x�))�;�n�(�n�(�x�)))))�:����>�
src:108dn.tex�Of��fcourse,�6�and�the�an��!teceden�t��fof�A4�are�uniable,�and�w��!e�ma�y�obtain����-�
src:112dn.tex7����i�(�i�(�i�(�i�(�i�(�x;��1y�d��)�;�i�(�n�(�z�{I�)�;�n�(�u�)))�;�z��)�;�v�d��)�;�i�(�i�(�v�;�x�)�;�i�(�u;�x�)))�:����>�
src:116dn.tex�But,�6since��*7�is�Meredith's�single�axiom�for�t��!w�o-v��dDalued��*sen�ten�tial�calculus,��
����-w��!e���ma�y�then�deduce�all�other�theorems�of�classical�sen�ten�tial�logic.��{The����-giv��!en�{�four-axiom�system�do�M�es,��^therefore,�pro�vide�{�a�complete�axiomatization����-for��classical�t��!w�o-v��dDalued��logic;�F�but�no�pro�M�of�of�an��!y�classical�theses�except�A1{����-A4�tkand�5�can�b�M�e�giv��!en�that�do�es�not�include�at�least�form��!ula�6,���in�whic�h����-�n�(�n�(�x�))��f(double�negation)�app�M�ears�four�times.����>�
src:119dn.texTh��!us�Aone�sees�that�some�axiom�systems�ha�v�e�so�m�uc�h�con�trol�o�v�er�pro�M�ofs����-deriv��!ed�Y�from�them�that�double�negation�is�inescapable.��]As�for�the�Meredith����-single�y%axiom�(deriv��!ed�from�the�Ulric�h�example),��fwhat�is�its�status�with�regard����-to�Gxguaran��!teed�double-negation-free�pro�M�ofs�of�theorems�that�themselv�es�are����-free���of�double�negation?�,�Of�a�sharply�dieren��!t�
a�v�or,�-what�is�the�status����-in��this�regard�of�the�F��erege�axiom�system�in�view�of�the�fact�that�t��!w�o��of�its���"$�����33�������j֍����*�1F'���홊��-�mem��!b�M�ers�ԅeac�h�con�tain�a�double�negation,� �i�(�n�(�n�(�x�))�;��1x�)�and��i�(�x;�n�(�n�(�x�)))?��
����-The��fF��erege�axiom�system�consists�of�the�follo��!wing�six�axioms.���g�������i�(�x;��1i�(�y�d�;�x�))�����������fO�i�(�x;��1n�(�n�(�x�)))���������fO�i�(�n�(�n�(�x�))�;��1x�)���������"��i�(�i�(�x;��1i�(�y�d�;�z�{I�))�;�i�(�i�(�x;�y��)�;�i�(�x;�z�{I�)))���������{��i�(�i�(�x;��1y�d��)�;�i�(�n�(�y��)�;�n�(�x�)))���������ה�i�(�i�(�x;��1i�(�y�d�;�z�{I�))�;�i�(�y�;�i�(�x;�z�{I�)))������-�
src:131dn.texThese��questions�are�also�answ��!ered�in�this�article�as�w�e�complete�our�treat-��
����-men��!t��?of�t�w�o-v��dDalued�sen�ten�tial�calculus�b�y�giving�conditions�that,��if�satised����-b��!y�]the�axioms,���guaran�tee�the�existence�of�a�double-negation-free�pro�M�of�for����-eac��!h��ftheorem�that�itself�is�double-negation�free.��
�4��>�
src:133dn.texThe�S�study�of�this�logical�prop�M�ert��!y�of�ob�viating�the�need�for�double�nega-����-tion���demands�its�examination�in�other�areas�of�logic�and�demands�a�nat-����-ural��extension.�	�Therefore,�aYw��!e�in�v�estigate�this�prop�M�ert�y�in�the�con�text�of����-innite-v��dDalued�_'sen��!ten�tial�calculus�and�in�tuitionistic�logic,�mgand�w�e�presen�t�an����-extension��of�the�prop�M�ert��!y�that�fo�cuses�on�theorems�in�whic��!h�double�negation����-app�M�ears.����>�
src:136dn.texSev��!eral���of�the�results�presen�ted�in�this�pap�M�er�w�ere�found�with�the�as-����-sistance�`�of�Otter�[�7��y�],�τan�automated�reasoning�program�that�searc��!hes�for����-pro�M�ofs.�z�Otter�ڟhas�b�een�used�eectiv��!ely�to�answ�er�n�umerous�op�M�en�questions����-in�İa�v��dDariet��!y�of�algebras�and�logics.���Its�underlying�logic�is�rst-order�predicate����-calculus��	with�equalit��!y��e.��Its�inference�rules�are�based�on�resolution�[�12��
�4]|a�gen-����-eralization��Dof�mo�M�dus�p�onens�and�syllogism�that�includes�instan��!tiation|and����-equalit��!y�@�substitution.��Otter�includes�an�extensiv�e�n�um�b�M�er�of�user-con�trolled����-strategies�/�for�directing�the�application�of�inference�rules�and�for�managing����-the��9p�M�oten��!tially�h�uge�n�um�b�M�er�of�form�ulas�(clauses)�that�can�b�M�e�deduced�in�a����-giv��!en��fstudy��e,�and�w�e�rely�hea�vily�on�these�capabilities�in�our�w�ork.����>�
src:148dn.texCondensed�e�detac��!hmen�t�problems�are�easily�represen�ted�for�Otter.��Ax-����-ioms�nDand�theorems�are�represen��!ted�as�atomic�form�ulas�in�a�predicate����<x
�3
cmtt10�P�n6�ha�v-����-ing��fthe�in��!tuitiv�e��fmeaning�\is�a�theorem".�L�F��eor�example,�Ԧthe�theorem��i�(�x;��1x�)����-w��!ould��fb�M�e�represen�ted�with�the�clause���5��-�
src:155dn.tex�>��P(i(x,x)),��Y���-�
src:158dn.tex�where�gnx�is�a�univ��!ersally�quan�tied�v��dDariable.� �The�inference�rule�condensed����-detac��!hmen�t��fis�eectiv��!ely�emplo�y�ed�b�y�including�the�clause���"$�����34����$y��j֍����*�1F'���홊��-�
src:163dn.tex�>��-P(i(x,y))���|�-P(x)�|�P(y),���n��-�
src:166dn.tex�where���\�-�"�is�the�negation�sym��!b�M�ol�and�\�|�"�represen�ts�disjunction,���and�b�y��
����-using��the�inference�rule��hyp��p�err�esolution�ӹin�Otter.���The�use�of�h��!yp�M�erresolution����-on�/`clauses�with�sym��!b�M�ol��P�/=�corresp�onds�to�the�use�of�condensed�detac��!hmen�t����-on�&�the�form��!ulas�inside��P�.�F��eor�the�reader�to�whom�this�is�new,�@w�e�sp�M�ell�it�out:����-Supp�M�ose����i�(�A;��1B����)�and��C�P'�are�t��!w�o���prop�ositional�form��!ulas�and�that����9�is�a�most����-general�#unier�of��A��and��C�ȁ�.��;Then�applying�condensed�detac��!hmen�t�#to��i�(�A;��1B����)����-and�3��C��X�w��!ould�deduce��B����d��.��1On�the�other�hand,�W3applying�h�yp�M�erresolution�to����-the��clauses��P(i(A,B))�,��P(C)��չand�the�clause�displa��!y�ed��ab�M�o�v�e�deduces�the�new����-clause��f�P(B��d��)�.������>�
src:180dn.texIn���this�study��e,�1�Otter�w��!as�used�to�pro�v�e�conditions�for�double-negation����-elimination���for�a�n��!um�b�M�er���of�axiom�systems.��YThese�conditions�are�instances�of����-theorems�ʹthat�are�not�particularly�dicult�to�pro��!v�e�ʹwith�Otter.���What�mak��!es����-these��Rtheorems�in��!teresting�and�c�hallenging�from�an�automated�theorem-����-pro��!ving�A�p�M�ersp�ectiv�e�A�is�the�requiremen�t�of�nding�deriv��dDations�of�the�exact�in-����-stances��$of�the�conditions.���Resolution-based�theorem�pro��!v�ers,��b�y��$design,�dra�w����-and��retain�most�general�conclusions�and�can�easily�b�M�ecome�o��!v�erwhelmed��if����-less�J�general�conclusions�(prop�M�er�instances)�are�retained.��>Finding�deriv��dDations����-of�,�the�desired�instances�required�careful�application�of|and�in�some�cases����-mo�M�dication�wtto|the�strategies�that�Otter�t��!ypically�relies�on�for�the�man-����-agemen��!t��fof�clauses.����>�
src:192dn.texOur�� in��!terest�in�double-negation�a�v�oidance�can�b�M�e�traced�directly�to�our����-successes�1with�Otter�in�previous�w��!ork.�	}�In�particular,���a�large�n�um�b�M�er�of����-pro�M�ofs�w�w��!ere�obtained�b�y�applying�a�strategy�that�instructs�Otter�to�a�v�oid����-reten��!tion�c6of�an�y�deduced�conclusion�if�it�con�tains�a�double-negation�term.����-Use�k�of�this�strategy�sharply�increased�the�lik��!eliho�M�o�d�k�of�success.��dBecause�the����-literature�m�strongly�suggests�that�reliance�on�double�negation�is�una��!v�oidable,����-and���b�M�ecause�our�completed�pro�ofs�suggested�the�con��!trary��e,�"the�questions�that����-are��fcen��!tral�to�this�article�w�ere�studied.��$6����-�2��A@The��In��terpla�y�of�Axioms�and�Pro�`of��bۍ�-�
src:203dn.tex�Once�hp�M�osed,�t{the�question�of�double-negation�a��!v�oidance�hseems�quite�natural,����-meshing��w��!ell�with�other�concerns�for�pro�M�of�prop�erties�as�expressed�b��!y�logi-����-cians.���F��eor�^�example,�mMeredith�and�Prior�[�6��y�],�then�Thomas�[�14��
�4]�a��!vidly�sough�t����-shorter�q'and�still�shorter�pro�M�ofs;�ևsize�of�pro�of�(total�n��!um�b�er�q'of�sym��!b�ols)�is����-of�)Fin��!terest�to�Ulric�h;�j�and�the�disp�M�ensing�with�though�t-to-b�M�e-k�ey�lemmas�is����-almost��falw��!a�ys�of�general�in�terest.���"$�����35����0X��j֍����*�1F'���홊��>�
src:208dn.tex�More�,rfamiliar�to�man��!y�are�similar�concerns�for�the�axioms�of�a�theory��e.��
����-Indeed,���in��ylogic,�merited�emphasis�is�placed�on�the�nature�and�prop�M�erties�of����-v��dDarious���axiom�systems:�b�the�n��!um�b�M�er���of�mem��!b�ers,��the�length�(individually�and����-collectiv��!ely),��&the��n�um�b�M�er�of�distinct�letters�(v��dDariables),��&the�total�n�um�b�M�er�of����-o�M�ccurrences��of�v��dDarious�function�sym��!b�ols,�-�and�other�measures�of�\simplicit��!y".����-T��eo�9�men��!tion�but�one�of�man�y�examples,��Zin�the�mid-1930s�J.���^ ����L���uk��dDasiewicz����-disco��!v�ered�R8a�23-letter�single�axiom�for�t��!w�o-v��dDalued�R8sen�ten�tial�(or�classical����-prop�M�ositional)�Tgcalculus.�	��As�cited�in�Section�1,���almost�t��!w�o�Tgdecades�later����-Meredith��dfound�a�21-letter�single�axiom.���Whether�a�still�shorter�single�axiom����-for��fthis�area�of�logic�exists�is�curren��!tly�unkno�wn.����>�
src:218dn.texAlthough�Kit�is�common�to�consider�the�prop�M�erties�of�a�pro�of�or�the�prop-����-erties�-�of�an�axiom�system,�E�less�w��!ork�connects�the�t�w�o�directly|for�example,����-considering�)relationships�of�the�form�\if�an�axiom�system�has�a�prop�M�ert��!y��P������2cmmi8�A����,����-then�n7ev��!ery�theorem�necessarily�has�a�pro�M�of�satisfying�prop�ert��!y��P����P��̹".�5OHere����-w��!e��study�suc�h�a�direct�connection�when�w�e�iden�tify�prop�M�erties�of�an�axiom����-system��fthat�guaran��!tee�the�existence�of�a�double-negation-free�pro�M�of.����>�
src:226dn.texDouble-negation-free�Upro�M�ofs,�Xin�addition�to�their�aesthetic�app�eal�and����-their��in��!terest�from�a�logical�viewp�M�oin�t,�%are�relev��dDan�t�to�the�w�ork�of�Hilb�M�ert.����-Indeed,��although��)it�w��!as�unkno�wn�un�til�recen�tly�[�20��
�4],��Hilb�M�ert�oered�a�t�w�en�t�y-����-fourth�9�problem�that�w��!as�not�included�in�the�famous�list�of�t�w�en�t�y-three����-seminal��tproblems�that�he�presen��!ted�in�P�aris�at�the�b�M�eginning�of�the�t�w�en�tieth����-cen��!tury��e.�	 QThis��t�w�en�t�y-fourth�problem�fo�M�cuses�on�the�disco�v�ery�of�simpler����-pro�M�ofs.��Hilb�ert��did�not�include�the�problem�in�his�P��!aris�talk,�zapparen�tly����-b�M�ecause��fof�the�dicult��!y�of�dening�\simpler"�precisely��e.����>�
src:235dn.texCeteris�q�paribus,��Rthe�a��!v�oidance�q�of�some�t��!yp�M�e�of�term�can�mak�e�a�pro�M�of����-simpler,�Ias��is�the�case�when�a�pro�M�of�is�free�of�doubly�negated�subform��!ulas.����-This�!�pap�M�er,���in�the�spirit�of�Hilb�ert's�t��!w�en�t�y-fourth�!�problem,���studies�this����-sp�M�ecic��!form�of�simplicit��!y��e,��Pseeking�(as�noted)�general�sucien�t�conditions����-for��an�axiom�system�of�prop�M�ositional�logic�L���that�guaran��!tees�that�doubly����-negated���form��!ulas�that�do�not�o�M�ccur�in�the�theorem�are�not�needed�in�the����-pro�M�of.���ݍ��-�3��A@F���ormalism��q���-�
src:247dn.tex�Although���prop�M�ositional�calculus�is�one�of�the�oldest�areas�of�logic,��Znot�all�of����-its��pm��!ysteries�ha�v�e�b�M�een�unlo�c��!k�ed.���The��pexistence�of�truth�tables�and�other�de-����-cision���pro�M�cedures�for�prop�ositional�logic�not��!withstanding,��@it�is�b�y�no�means����-trivial��to�pro��!v�e,�=for��example,�that�a�giv��!en�23-sym�b�M�ol�form�ula�is�in�fact�a����-single���axiom.�~�T��eruth�tables�and�decision�pro�M�cedures�can�b�e�used�to�deter-����-mine���whether�a�giv��!en�form�ula�is�a�tautology��e,��and�they�ma�y�b�M�e�helpful�in���"$�����36����=���j֍����*�1F'���홊��-�constructing��a�pro�M�of�of�a�giv��!en�form�ula�from�a�giv�en�set�of�axioms�for�prop�M�o-��
����-sitional���calculus,��gbut�generally�they�are�not�helpful�in�nding�pro�M�ofs�of�kno��!wn����-axioms�Pufrom�other�form��!ulas�(whic�h�is�what�one�m�ust�do�to�v�erify�that�a�for-����-m��!ula��is�a�single�axiom).�(�The�searc�h�for�suc�h�pro�M�ofs�has�recen�tly�b�M�ecome�a����-test��b�M�ed�in�automated�deduction.��@Not�only�do�the�theorems�w��!e�pro�v�e�here����-ab�M�out���double-negation�elimination�ha��!v�e���an�in��!trinsic,�
!aesthetic�app�eal�in�that����-they��
sho��!w�the�p�M�ossibilit�y�of�simplifying�pro�M�ofs,��wbut�they�also�are�of�in�terest����-b�M�ecause�1�they�justify�in�the�v��dDast�ma���jorit��!y�of�cases�a�shortcut�in�automated����-pro�M�of-searc��!h��+metho�ds,���namely��e,�the�automatic�discarding�of�double�negations.����>�
src:268dn.texW��ee��Vshall�w��!ork�with�logics�form�ulated�b�y�using�only�the�t�w�o�connectiv�es����-implication��and�negation.��CSev��!eral�notations�are�in�use�for�prop�M�ositional�logic����-that��#w��!e�men�tion�b�M�efore�con�tin�uing.��First,���one�can�use�inx��!",�
�3
cmsy10�!��for�implication����-and��kprex��:��for�negation.�m�F��eor�example,�"lw��!e�could�write��x�X�!��(�:�x��!��y�d��).����-Closely�)related,�Zman��!y�pap�M�ers�on�prop�ositional�logic�use�P��!olish�notation,�Zin����-whic��!h��=���"V
�3
cmbx10�C���m�is�=�used�for�implication�(conditional)�and��N��for�negation.���The�same����-form��!ula��Rw�ould�then�b�M�e�rendered�as��C�x�CN�xy�d��.�ȡFinally��e,�ݍa�notation�that�is����-appropriate��when�using�Otter�is�prex,�/$with�paren��!theses.��sW��ee�use��i�(�x;��1y�d��)����-for�]�implication�and��n�(�x�)�for�negation;�u�therefore,�lBthe�example�form��!ula�w�ould����-b�M�e�ώ�i�(�x;��1i�(�n�(�x�)�;�y�d��)).�YVIn�this�pap�M�er�w��!e�use�this�last�notation�exclusiv�ely��e.�YVIt����-p�M�ermits��us�to�cut�and�paste�mac��!hine-pro�duced�pro�ofs,�xeliminating�errors����-of��Atranscription.�XmW��ee�mak��!e�use�of�the�theorem-pro�ving�program�Otter�[�7��y�]����-to��pro�M�duce�pro�ofs�in�v��dDarious�prop�ositional�logics,�`pro�ofs�w��!e�use�to�v�erify����-that��(those�logics�satisfy�the�h��!yp�M�otheses�of�our�general�theorems�on�double-����-negation��felimination.����>�
src:291dn.texLet��jL��hb�M�e��$8 ����L��
�kuk��dDasiewicz's�form��!ulation�of�prop�ositional�calculus�in�terms�of����-implication���and�negation,���denoted�b��!y��i��and��n�,�as�giv��!en�on�page�221�of�[�19��
�4].�����-t� ����L��3�uk��dDasiewicz��fpro��!vided�the�follo�wing�axiomatization�of�L.��34�����a�L1���������i�(�i�(�x;��1y�d��)�;�i�(�i�(�y�;�z�{I�)�;�i�(�x;�z��)))�����������a�L2��������i�(�i�(�n�(�x�)�;��1x�)�;�x�)���������a�L3������K��i�(�x;��1i�(�n�(�x�)�;�y�d��))������-�
src:300dn.texThe�x�inference�rule�frequen��!tly�used�in�logic�is�kno�wn�as�condensed�detac�h-��
����-men��!t.���This��
rule�(whic�h�is�the�only�inference�rule�to�b�M�e�used�in�the�sough�t-����-after�$�double-negation-free�pro�M�ofs)�com��!bines�substitution�and�mo�dus�p�onens.����-Sp�M�ecically��e,�n1giv��!en�F<a�ma���jor�premiss��i�(�p;��1q�d��)�and�a�minor�premiss��p�,�the�con-����-clusion��of�mo�M�dus�p�onens�is��q�d��.��QThe�substitution�rule�p�ermits�the�deduction����-of���p��from��p�,���where����is�an��!y�substitution�of�terms�for�v��dDariables.�ջCondensed����-detac��!hmen�t�S�has�premisses��i�(�p;��1q�d��)�and��r����and�attempts�to�unify��p��and��r�M޹|that����-is,�I�seeks�)&a�substitution������that�mak��!es��p�I#�=���r�M��d��.�fIf�successful,�pro��!vided������is���"$�����37����K��j֍����*�1F'���홊��-�a�[�most�general�suc��!h�substitution,��$the�conclusion�of�condensed�detac�hmen�t��
����-is�š�q�d��'-�or�an�alphab�M�etic�v��dDarian��!t�of��q���.�2zThis�inference�rule�requires�renam-����-ing�Gmof�v��dDariables�in�the�premisses�b�M�efore�the�attempted�unication�to�a��!v�oid����-unin��!tended��fclashes�of�v��dDariables.������|{Ycmr8�1����3͍�>�
src:341dn.tex�A���double���negation�is�a�form��!ula��n�(�n�(�t�)),���where��t��is�an�y�term.�)�A���form�ula����-�A�]�c��p�ontains���a�double�ne�gation�]�if�it�has�a�not-necessarily-prop�M�er�subform��!ula����-that���is�a�double�negation.��HA���deriv��dDation�con��!tains�a�double�negation�if�one�of����-its��ddeduced�form��!ulas�con�tains�a�double�negation.��Supp�M�ose�that�the�form�ula����-�A��o�con��!tains�no�double�negations�and�is�deriv��dDable�in�L.�Then�(cen�tral�to�this����-pap�M�er)�k�do�es��A��ha��!v�e�a�deriv��dDation�in�L�k�that�con�tains�no�double�negation?��eW��ee����-answ��!er���this�question�in�the�armativ�e�(and�th�us�answ�er�the�cited�Ulric�h����-question),�g�not�Aonly�for���� ����L��uk��dDasiewicz's�system�L1{L3,�but�also�for�other�ax-����-iomatizations�Z�of�classical�(t��!w�o-v��dDalued)�Z�prop�M�ositional�logic,��as�w��!ell�as�other����-systems��fof�logic�suc��!h�as�innite-v��dDalued�logic.��!�o���-�4��A@Condensed��Detac��hmen�t���"��-�
src:359dn.tex�W��ee���remind�the�reader�that�the�systems�of�primary�in��!terest�in�this�pap�M�er����-use���condensed�detac��!hmen�t���as�their��sole��rule�of�inference.�ʙF��eor�example,��_if����-��,�is��a�complicated�form��!ula�and�w�e�wish�to�deduce��i�(��
`;��1��),��it��w�ould�not�b�M�e����-acceptable���to�rst�deduce��i�(�x;��1x�)�and�then�substitute����R�for��x�.�$�Rather,��it����-w��!ould��eb�M�e�necessary�to�giv�e�a�(longer)�direct�deriv��dDation�of��i�(��
`;��1��),�-�relying����-solely��fon�applications�of�condensed�detac��!hmen�t.����>�
src:370dn.texW��ee��shall�sho��!w�in�this�section�that�our�theorem�ab�M�out�the�eliminabilit�y�of����-double�jnegation�holds�for�L1{L3�with�condensed�detac��!hmen�t�jif�and�only�if�it����-holds��for�L1{L3�with�mo�M�dus�p�onens�and�substitution.��mSimilar�results�are�in����-[�3��y�,����8��	}],���but���for�other�systems:��[�3��]�treats�the�implicational�fragmen��!t,���while�w�e����-allo��!w��Enegation,��=and�[�8��y�]�treats�relev��dDance�logic.��zThe�follo�wing�three�form�ulas����-will��fpla��!y�an�imp�M�ortan�t�role.��-��B�ff�e֟
L͍����{���-=��Aa�cmr6�1�����a�"o���		cmr9�In�1the�absence�of�the�substitution�rule,�an��9y�alphab�A�etic�v��|rarian�t�of�an�axiom�is�also�ac-���cepted��fas�an�axiom.��vAn�\alphab�A�etic"�v��|rarian��9t�of��#5��"		cmmi9�A��is�a�form�ula��A�R��,��where�the�substitution�����6V�is��one-to-one�and�merely�renames�the�v��|rariables.��A��tec��9hnicalit�y��arises�as�to�whether�it�is���p�A�ermitted,��hrequired,�or��mforbidden�to�rename�the�v��|rariables�of�the�premisses�b�efore�applying���condensed�h�detac��9hmen�t.���The�denition�on�p.�212�of�[�19��	?�]�do�A�es�not�explicitly�men��9tion�renam-���ing���and,���read�literally��:�,�w��9ould�not�allo�w�it,���but�the�implemen�tation�in�Otter�requires�it,���and���[�15��	?�]�Ybexplicitly�p�A�ermits�it.��If�it�is�not�p�ermitted,�jfthen�b��9y�renaming�v��|rariables�in�the�en�tire���pro�A�of��of�the�premiss,��w��9e�obtain�a�pro�of�of�the�renamed�premiss,��using�alphab�etic�v��|rarian��9ts���of�~�the�axioms,��)so�the�same�form��9ulas�will�b�A�e�pro�v��|rable�in�either�case.�X�Similarly��:�,��)renaming���of�
iv��|rariables�in�conclusions�is�allo��9w�ed.��T��:�ec�hnically�,��w�e�
icould�w��9ait�un�til�the�conclusions�are���used��eb�A�efore�renaming�them,���but�in�practice,�Otter�renames�v��|rariables�in�eac��9h�conclusion�as���it�Tis�deriv��9ed.���"$�����3�8����	Z֠�j֍����*�1F'���홊��>�
src:382dn.tex���8����wO�D1�������i�(�x;��1x�)����������wOD2�������4�i�(�i�(�x;��1x�)�;�i�(�n�(�x�)�;�n�(�x�)))��������wOD3�������t�i�(�i�(�x;��1x�)�;�i�(�i�(�y�d�;�y��)�;�i�(�i�(�x;�y��)�;�i�(�x;�y��))))����J֍���-�Lemma�2�1���el.�
src:390dn.tex�Supp��p�ose��L��pis�any�system�of�pr�op�ositional�lo�gic�with�c�ondense�d��
����-detachment�ias�the�sole�infer��p�enc�e�irule,���and�supp��p�ose�that�ther�e�ar�e�pr�o�ofs�of����-D1{D3�1*in�L.�Then�every�formula�of�the�form��i�(��
`;��1��)�1*�is�pr��p�ovable�fr�om�L�1by����-c��p�ondense�d�j�detachment.�޶F��)urthermor�e,��Qif�ther�e�ar�e�double-ne�gation-fr�e�e�pr�o�ofs����-of� �D1{D3�in�L,�then��i�(��
`;��1��)� ��is�pr��p�ovable�without�using�double�ne�gations�exc�ept����-those���o��p�c�curring�as�subformulas�of���
`�.��-L��-�
src:403dn.texPr��p�o�of�R�.��4W��ee�Y�pro��!v�e�b�y�induction�on�the�complexit�y�of�the�prop�M�ositional�for-����-m��!ula�=4��G��that�for�eac�h���
`�,�b�the�form�ula��i�(��
`;��1��)�=4is�pro�v��dDable�in�L�=
b�y�condensed����-detac��!hmen�t.�8*The�o*base�case,��[when���y��is�a�prop�M�osition�letter,�follo��!ws�b�y�re-����-placing�L��x��b��!y���V�in�the�pro�M�of�of��i�(�x;��1x�).���An�y�line�of�the�pro�M�of�that�is�an�axiom����-b�M�ecomes��qan�alphab�etic�v��dDarian��!t�of�that�axiom,��4whic�h�is�still�considered�an����-axiom.���Actually��e,�pin�G�view�of�the�con��!v�en�tion�G�that�renaming�v��dDariables�in�the����-conclusion��nis�allo��!w�ed,��0it��nw�ould�b�M�e�enough�just�to�replace��x��b�y����ιin�the�last����-line�a�of�the�pro�M�of.���If�w��!e�ha�v�e�a�pro�M�of�of��i�(����;��1��),�ofthen�a�w�e�can�apply�condensed����-detac��!hmen�t��-and�D2�to�get�a�pro�M�of�of��i�(�n�(����)�;��1n�(���)).��1This��-could�in��!tro�duce�a����-double�rnegation�if����is�already�a�negation,���but�in�that�case�it�is�a�double����-negation�q�that�already�o�M�ccurs�in���g��=�]7�n�(����),��Sand�so�is�allo��!w�ed.�?ISimilarly��e,�if����-w��!e�<�ha�v�e�pro�M�ofs�of��i�(��
`;��1��)�and��i�(����;��1��),�bw��!e�can�apply�condensed�detac�hmen�t����-to��D3�and�get�a�pro�M�of�of��i�(�i�(��
`;��1���)�;�i�(��;����)).��That��completes�the�pro�M�of�of�the����-lemma.������-�Lemma�2�2���el.�
src:434dn.tex�If�d&�A��is�an�instanc��p�e�of��C�ȁ�,��tthen�the�r�esult�of�applying�c�ondense�d����-detachment���to��i�(�A;��1B����)��and��C��k�is��B�w��(or�an�alphab��p�etic�variant�of��B��).��-L��-�
src:440dn.texPr��p�o�of�R�:�yRename���v��dDariables�in��C��O�if�necessary�so�that��C��and��A��ha��!v�e���no�v��dDariables����-in��hcommon.���Let����b�M�e�a�most�general�unier�of��A��and��C�ȁ�.�Then�the�result�of����-applying��fcondensed�detac��!hmen�t��fto��i�(�A;��1B����)�and��C�n�is��B��d��.��
�$��>�
src:448dn.texLet�Sa�����b�M�e�a�most�general�substitution�suc��!h�that��C�ȁ��=�՜�A�;�)�since��A��is����-assumed�c!to�b�M�e�an�instance�of��C�ȁ�,�p�suc��!h�a����p�exists.��qSince�the�v��dDariables�of��C�+��do����-not���o�M�ccur�in��A��or��B����,���B��f<�=�(��B�E4�and����A��=��A�.�[Then����C�ȁ��=��A��=��A�=O�,��so������=���d�����-�for��hsome�substitution���.���Then��B�N�=����B�������=��B��d��.���Th��!us��h����is�the�iden��!tit�y�on����-�B����.��MHence�h���d���is�the�iden��!tit�y�h�on�eac��!h�v��dDariable�o�M�ccurring�in��B��.��MHence����J�and������-�do��*nothing�but�(p�M�ossibly)�rename�v��dDariables.��)Hence��B����d��,�
whic��!h�is�the�result���"$�����39����
j���j֍����*�1F'���홊��-�of��this�application�of�condensed�detac��!hmen�t,�0Fis���B��W�or�an�alphab�M�etic�v��dDarian��!t��
����-of��f�B����.���That�completes�the�pro�M�of�of�the�lemma.���͍���-�Lemma�2�3���el.�
src:469dn.tex�Supp��p�ose��sL��Eis�a�lo�gic�pr�oving�D1{D3�by�c�ondense�d�detachment.����-Then���e��p�ach�substitution�instanc�e������of�an�axiom�of�L�� is�pr�ovable�by�c�on-����-dense��p�d�C]detachment.��F��)urthermor�e,�Yzif�L�CFpr�oves�D1{D3�by�c�ondense�d�detach-����-ment�F�without�using�double�ne��p�gations,�]lthen���P��is�also�pr�ovable�without�using����-double���ne��p�gations,���exc�ept�those�double�ne�gations�o�c�curring�as�subformulas�of����-��
`�,���if�any.����-�
src:480dn.texPr��p�o�of�R�.�	Z�Let�%c��/ùb�M�e�a�substitution�instance�of�an�axiom��A�.�Renaming�the����-v��dDariables�8in�the�axiom��A��if�necessary��e,�^,w��!e�ma�y�assume�that�the�v��dDariables����-o�M�ccurring�D6in��A��do�not�o�ccur�in���
`�.�	�MBy�Lemma�1,����i�(��;��1��)�is�pro��!v��dDable�b�y����-condensed���detac��!hmen�t,��3without�using�an�y�double�negations�except�p�M�ossibly����-those�Ralready�o�M�ccurring�in���
`�.���By�Lemma�2,�4�the�result�of�applying�condensed����-detac��!hmen�t�Яto��i�(��
`;��1��)�Яand��A��is�����or�an�alphab�M�etic�v��dDarian��!t���
`�5B�of����.���If�it�is�not����-literally����
`�,�2�w��!e�can�rename�v��dDariables�in�the�conclusion�(or,�if�one�prefers�to����-a��!v�oid��Trenaming�conclusions,�Othroughout�the�en��!tire�pro�M�of��)�to�create�a�pro�of����-of��f��
`�.���This�completes�the�pro�M�of�of�the�lemma.����>�
src:502dn.texA�>�pro�M�of�?;of��B��߹in�L�from�assumptions��is�dened�as�usual:��Lines�of����-the�dpro�M�of�are�inferred�from�preceding�lines,�Ӆor�are�axioms,�or�b�M�elong�to����-.�	J�When��condensed�detac��!hmen�t��is�used�as�a�rule�of�inference,�~\ho��!w�ev�er,����-w��!e�_Qha�v�e�to�distinguish�b�M�et�w�een�(prop�M�ositional)�v��dDariables�that�o�ccur�in�the����-axioms��and�sp�M�ecic�(constan��!t)�prop�osition�letters�that�o�ccur�in�assumptions.����-F��eor�example,�0�if�w��!e�ha�v�e��i�(�n�(�n�(�x�))�;��1x�)�as�an�axiom,�0�then�w�e�can�deriv�e�an�y����-substitution�Uginstance�of�that�form��!ula,��'but�if�w�e�ha�v�e��i�(�n�(�n�(�a�))�;��1a�)�as�an����-assumption,�t�w��!e�hcannot�use�it�to�deriv�e�an�instance�with�some�other�form�ula����-substituted�B�for��a�.���W��ee�do�not�allo��!w�v��dDariables�in�assumptions,�V�only�constan�ts.����>�
src:518dn.texThe��'follo��!wing�theorem�is�the�easy�half�of�the�relation�b�M�et�w�een�condensed-����-detac��!hmen�t�+pro�M�ofs�and�mo�dus�p�onens�pro�ofs.�k�The�sense�of�the�theorem�is����-that��fsubstitutions�can�b�M�e�pushed�bac��!k�to�the�axioms.������-�Theorem�2�1�(Push��tbac�k�2�theorem)�����
src:525dn.tex�L��p�et��^L��b�e�a�system�of�pr�op�ositional�lo�gic,����-and��supp��p�ose�L��pr�oves��B����by�using�c�ondense�d�detachment�or�by�using�mo�dus����-p��p�onens��and�substitution.��Then�ther�e�exists�a�pr�o�of�of��B� ��using�mo�dus�p�onens����-fr��p�om��substitution�instanc�es�of�axioms�of�L.�Similarly,�Dif�L��Ypr�oves��B�n=�fr�om����-assumptions�9���,�M<then�ther��p�e�exists�a�pr�o�of�of��B��6�using�mo�dus�p�onens�fr�om������-�and���substitution�instanc��p�es�of�axioms�of�L.����-�
src:537dn.texR��p�emark�.�$It�hsw��!ould�not�mak�e�sense�to�sp�M�eak�of�substitution�instances�of�����-b�M�ecause��fassumptions�cannot�con��!tain�v��dDariables,�as�explained�ab�o��!v�e.���"$����f10����y&��j֍����*�1F'���홊��-�
src:542dn.tex�Pr��p�o�of�.�oFirst�]�w��!e�pro�v�e�the�theorem�for�the�case�when�the�giv�en�pro�M�of�uses��
����-mo�M�dus��p�onens�and�substitution.���W��ee�pro�ceed�b��!y�induction�on�the�length�of����-the��lgiv��!en�pro�M�of�of��B����.���If�the�length�is�zero,�then��B�m�is�an�axiom�or�assumption,����-and���there�is�nothing�to�pro��!v�e.��?If���the�last�inference�is�b��!y�mo�M�dus�p�onens,��ysa��!y����-�B�G�is���inferred�from��i�(�A;��1B����)�and��A�,��1then�b��!y�the�induction�h�yp�M�othesis�there����-exist��Ypro�M�ofs�of�these�premisses�from�substitution�instances�of�axioms,��and����-adjoining��fthe�last�inference,�w��!e�obtain�the�desired�pro�M�of�of��B����.��
�7��>�
src:555dn.texIf�%<the�last�inference�is�b��!y�substitution,�D�sa�y��B�j��=���A��Ϲis�inferred�from��A�,����-then��Db��!y�the�induction�h�yp�M�othesis�there�exists�a�pro�of���K׹of��A��using�mo�dus�p�o-����-nens��only�from�substitution�instances�of�axioms.���Apply�the�substitution���A��to����-ev��!ery�	line�of���d��;�J(the�result�is�the�desired�pro�M�of�of��B����.���If�there�are�assumptions,����-they��fare�unaected�b��!y���
��b�M�ecause�they�do�not�con�tain�v��dDariables.����>�
src:565dn.texNo��!w�hsupp�M�ose�that�the�original�pro�of�uses�condensed�detac��!hmen�t.�!�Eac�h����-condensed-detac��!hmen�t���inference�can�b�M�e�brok��!en�in�to�t�w�o�substitutions�and����-an�:�application�of�mo�M�dus�p�onens,�PBso�a�condensed-detac��!hmen�t�:�pro�of�giv��!es�rise����-to��%a�mo�M�dus�p�onens�and�substitution�pro�of,��and�w��!e�can�apply�the�preceding����-part��fof�the�pro�M�of.���That�completes�the�pro�of.����>�
src:575dn.texThe��`follo��!wing�lemma�is�not�actually�used�in�our�w�ork�but�is�of�indep�M�en-����-den��!t�V�in�terest.��MCondensed�detac�hmen�t�is�considered�as�an�inference�rule�that����-com��!bines��>mo�M�dus�p�onens�and�substitution.��%The�follo��!wing�lemma�sho�ws�that����-it��is�reasonable�to�consider�systems�whose�only�rule�of�inference�is�condensed����-detac��!hmen�t,��$b�M�ecause��Tsuc�h�systems�are�already�closed�under�the�rule�of�sub-����-stitution.�wKThis�ًis�not�ob��!vious��a��priori��since�condensed�detac�hmen�t�p�M�ermits����-only��fcertain�sp�M�ecial�substitutions.���H����-�Lemma�2�4���el.�
src:583dn.tex�Supp��p�ose�T�L�T�is�a�lo�gic�pr�oving�formulas�D1{D3�by�c�ondense�d�de-����-tachment.���If����A��is�pr��p�ovable�in�L���with�c�ondense�d�detachment�and���%��is�any����-substitution,���then��A�O}�is�pr��p�ovable�in�L�by�c�ondense�d�detachment.������-�
src:592dn.texPr��p�o�of�R�.��HBy���induction�on�the�length�of�the�pro�M�of���Vr�of��A��in�L,�w��!e�pro�v�e�that����-the�xstatemen��!t�of�the�lemma�is�true�for�all�substitutions���d��.�R�The�base�case����-o�M�ccurs�Rrwhen��A��is�an�axiom,�}uso��A���is�a�substitution�instance�of�an�axiom.����-By��fLemma�3,��A�
��is�pro��!v��dDable�in�L�b�y�condensed�detac�hmen�t.����>�
src:601dn.texF��eor�[�the�induction�step,��5supp�M�ose�the�last�inference�of�the�giv��!en�pro�of������-�has���premisses��i�(�p;��1q�d��)�and��r�M޹,�Hrwhere���2&�is�the�most�general�unier�of��p��and����-�r�M޹,�dcand�1the�conclusion�is��q�d�����=�]L�A�.�	?By�the�induction�h��!yp�othesis,�dcw�e�1ha�v�e����-condensed-detac��!hmen�t�)_deriv��dDations�of��i�(�p�=O���;��1q�d����)�)_and�of��r�M��=O��.��0Since��p�G��=�
��r�M��=O�,����-also��*�p�=O�o:�=�
��r�M���d��.��Hence�the�inference�from��i�(�p����;��1q�d����)��*and��r�M��=O�*��to��q��=O�*��is�legal����-b��!y�q�condensed�detac�hmen�t.��CHence�w�e�ha�v�e�a�condensed-detac�hmen�t�pro�M�of�of����-�q�d��=O�o:�=�
��A��.���This��fcompletes�the�pro�M�of�of�the�lemma.���"$����f11��������j֍����*�1F'���홊����-�Theorem�2�2�(D-completeness)����|�
src:619dn.tex�Supp��p�ose��_L��[is�a�lo�gic�that�pr�oves�formulas��
����-D1{D3.���If�_L�%pr��p�oves��A��by�using�mo�dus�p�onens�and�substitution,�4�then�L�%pr�oves����-�A����by�using�c��p�ondense�d���detachment.��E}��-�
src:626dn.texR��p�emark�-��.�x�W��ee��cannot�trac��!k�what�happ�M�ens�to�double�negations�in�this�pro�of.����-The�pro�M�of�do�es�not�guaran��!tee�that�passing�from�substitution�to�condensed����-detac��!hmen�t�u�will�not�in��!tro�M�duce�new�double�negations.�͗Somewhat�to�our�sur-����-prise,���w��!e���do�not�need�an�y�suc�h�result�to�pro�v�e�double-negation�elimination;����-indeed,�h�quite�Y�the�rev��!erse,�w�e�Y�shall�deriv��!e�suc�h�a�result�from�double-negation����-elimination.����-�
src:635dn.tex�Pr��p�o�of�R�.���By��tTheorem�1,��>there�exists�a�pro�M�of���!�of��A��from�substitution�instances����-of��saxioms,��6using�mo�M�dus�p�onens�as�the�only�rule�of�inference.�SBy�Lemma�3,����-there�Vexist�condensed-detac��!hmen�t�Vpro�M�ofs�of�these�substitution�instances�of����-axioms.���Since�׷mo�M�dus�p�onens�is�a�sp�ecial�case�of�condensed�detac��!hmen�t,�
if�׷w�e����-string��together�the�condensed-detac��!hmen�t��pro�M�ofs�of�the�instances�of�axioms����-required,��@follo��!w�ed���b�y�the�pro�M�of���d��,��@w�e�obtain�a�condensed-detac�hmen�t�pro�M�of����-of��f�A�.���That�completes�the�pro�M�of�of�the�theorem.���Y���-�5��A@The��Main�Theorem��q���-�
src:651dn.tex�Let�NL��b�M�e�a�system�of�prop�ositional�logic,�lgiv��!en�b�y�some�axioms�and�the����-sole��Linference�rule�of�condensed�detac��!hmen�t.���Let��LL*�b�M�e�the�system�of�logic����-whose��axioms�are�the�closure�of�(the�axioms�of��)�L��under�applications�of����-the�2�follo��!wing�syn�tactic�rule:���If��x��is�a�prop�M�osition�letter,�U�and�subterm��n�(�x�)����-app�M�ears��@in�a�form��!ula��A�,�;6then�construct�a�new�form�ula�b�y�replacing�eac�h����-o�M�ccurrence��of��x��in��A��b��!y��n�(�x�)�and�cancelling�an�y�double�negations�that�result.����-In�99other�w��!ords,�]�w�e�99c�ho�M�ose�a�set��S��˹of�prop�osition�letters�o�ccurring�negated����-in��n�A�,��pand�w��!e�replace�eac�h�o�M�ccurrence�of�a�v��dDariable��x��in��S�t�throughout��A��b�y����-�n�(�x�),���cancelling�lan��!y�doubly�negated�prop�M�ositions.�.�The�rst�description�of����-L*�calls�for�replacing�all�o�M�ccurrences�of�only��one��v��dDariable;�Ckbut�if�w��!e�rep�eat����-that��fop�M�eration,�w��!e�can�in�eect�replace�a�subset.����>�
src:673dn.texAn�܅example�will�mak��!e�the�denition�of�L*�clear.��9If�this�pro�M�cedure�is����-applied��fto�the�axiom�����"��i�(�i�(�n�(�x�)�;��1n�(�y�d��))�;�i�(�y�;�x�))�;��o���-�
src:676dn.tex�w��!e��obtain�the�follo�wing�three�new�axioms�(b�y�replacing�rst�b�M�oth��x��and��y�d��,����-then��fonly��y�d��,�then�only��x�).��E}�����V�A6���������i�(�i�(�x;��1y�d��)�;�i�(�n�(�y��)�;�n�(�x�)))�����������V�A7���������i�(�i�(�n�(�x�)�;��1y�d��)�;�i�(�n�(�y��)�;�x�))���������V�A8���������i�(�i�(�x;��1n�(�y�d��))�;�i�(�y�;�n�(�x�)))�����"$����f12����
�Ǡ�j֍����*�1F'���홊��>�
src:684dn.tex�W��ee�`(sa��!y�that�L�`admits�double-negation�elimination�if,�n4whenev�er�L�`pro�v�es��
����-a��^theorem��B����,��there�exists�a�pro�M�of��S���of��B���in�L��1suc��!h�that�an�y�double�negations����-o�M�ccurring��Has�subform��!ulas�in�the�deduced�steps�of��S�Kڹo�ccur�as�subform��!ulas�of����-�B����.������2������In��particular,��-�double-ne��p�gation-fr�e�e��Sthe�or�ems�have�double-ne�gation-fr�e�e����-pr��p�o�ofs��f�(ignoring�the�axioms).��
�K��>�
src:696dn.texSupp�M�ose���B�t��con��!tains�sev�eral�doubly�negated�subform�ulas.���W��ee�wish�to����-consider��eliminating�double�negations�on�just��some��of�those�subform��!ulas.����-Let��Ia�subset�of�the�doubly�negated�form��!ulas�in��B�\�b�M�e�selected.�[�Then�let��B���������K�cmsy8������-�b�M�e�!�the�result�of�erasing�double�negations�on��al��Fl�\3o��p�c�curr�enc�es�!��of�the�selected����-subform��!ulas�X�in��B����.��/More�precisely��e,���B��������	�)�is�obtained�from��B��%�b��!y�replacing�all����-o�M�ccurrences��vof�selected�doubly�negated�subform��!ulas��n�(�n�(�q�d��))�in��B�m�b�y��q�d��.��W��ee����-emphasize�8bthat�if�some�doubly�negated�subform��!ula�o�M�ccurs�more�than�once����-in���B����,�one�m��!ust�erase�double�negations�on�all�or�none�of�those�o�M�ccurrences.����-Generally�s�there�will�b�M�e�more�than�one�w��!a�y�s�to�select�a�set�of�doubly�negated����-subform��!ulas,�T�so����B����������
Kp�is�not�unique.��W��ee�sa�y�that�L��padmits�strong�double-����-negation�G"elimination�if,�Z/whenev��!er�L�G
pro�v�es�a�theorem��B����,�Z/and��B���������ʹis�obtained����-from��|�B� �as�describ�M�ed,��then�there�exists�a�pro�of�of��B����������	�$�in�L,�and�moreo��!v�er,����-there��fexists�a�pro�M�of�of��B����������	8�in�L��Uthat�con��!tains�only�doubly�negated�form�ulas����-o�M�ccurring��fin��B����������L��.��]�����-�Theorem�2�3���n>M�
src:722dn.tex�[Str��p�ong�5{double-ne�gation�elimination].��*Supp�ose�that�in�L�5hther�e����-exist�޺double-ne��p�gation-fr�e�e�pr�o�ofs�of�D1{D3�and�double-ne�gation-fr�e�e�pr�o�ofs����-of���al��Fl�the�axioms�of�L*.�	vThen�L�admits�str��p�ong�double-ne�gation�elimination.��=��-�
src:727dn.texR��p�emark�.��4The�wjtheorem�is�also�true�with�triple�negation,���quadruple�negation,����-and�a>so�forth�in�place�of�double�negation.���F��eor�instance,�oif��B���con��!tains�a�triple����-negation,�%9then�ةit�has�a�pro�M�of�con��!taining�no�double�negations�not�already����-con��!tained��/in��B����.���In�particular,���it�then�con�tains�no�triple�negations�not�already����-con��!tained��fin��B����,�since�ev�ery�triple�negation�is�a�double�negation.����-�
src:736dn.tex�Pr��p�o�of�.��Supp�M�ose��&�B�mʹis�pro��!v��dDable�in�L.�If��B��con��!tains�an�y�double�negations,��select����-arbitrarily�b�a�subset�of�the�doubly�negated�subform��!ula�of��B����,�piand�form��B�����������b��!y����-replacing��eac��!h�o�M�ccurrence�of�these�form�ulas��n�(�n�(�q�d��))�b�y��q�d��.���Of�course,�:��B�������������-�ma��!y�<�still�con�tain�double�negations;���if�w�e�are�pro�ving�only�double-negation����-elimination�j�and�not�strong�double-negation�elimination,���w��!e�tak�e��B����������	�2�to�b�M�e����-�B����.�/By�lTheorem�1,���there�is�a�mo�M�dus�p�onens�pro�of�of��B��ùfrom�substitution����-instances��!of�axioms.��
If�this�pro�M�of�con��!tains�an�y�double�negations�that�do����-not��o�M�ccur�in��B����������L��,���w��!e�simply�erase�them.��hThis�erasure�tak�es�a�mo�M�dus�p�onens��-�

:�ff�e֟
L͍����{���-=�2�����a�In���this�con��9text,��Ca�form�ula��t��o�A�ccurs�as�a�subform�ula�if�and�only�if��t��or�an�alphab�A�etic���v��|rarian��9t�{of��t��app�A�ears.�M�F��:�or�example,���if��n�(�n�(�i�(�u;���u�)))�o�ccurs�in��B�r��,���then��n�(�n�(�i�(�x;���x�)))�w��9ould���b�A�e�Tp�ermitted�in�the�deduced�steps�of��S��Y�but�not��n�(�n�(�i�(�x;���y�R��)))�or��n�(�n�(�i�(�i�(�z�c�;�z��)�;�i�(�z�;�z��)))).���"$����f�13�����w��j֍����*�1F'���홊��-�step�:�in��!to�another�legal�mo�M�dus�p�onens�step.���Note�that�one�cannot�\simply��
����-erase"�~�double�negations�in�a�condensed-detac��!hmen�t�~�pro�M�of;���but�no��!w�w�e�ha�v�e����-a�hmo�M�dus�p�onens�pro�of,�8and�double�negations��c��p�an��b�e�erased�in�mo�dus�p�onens����-pro�M�ofs.��dF��eor��>axioms,�C�the�pro�cess�transforms�a�substitution�instance�of�an����-axiom�iPof�L�iin��!to�a�substitution�instance�of�an�axiom�of�L*.�&�Th�us�w�e�ha�v�e����-a�:"pro�M�of�of��B����������
�ʹfrom�substitution�instances�of�axioms�of�L*�that�con��!tains����-no�.�double�negations�except�those�that�already�o�M�ccur�in��B����������L��.�	w0By�Lemma����-3,�� there�Z/exist�condensed-detac��!hmen�t�Z/pro�M�ofs�of�these�substitution�instances����-of�I�L*�(from�axioms�of�L*).��?By�h��!yp�M�othesis,�rNthe�axioms�of�L*�ha�v�e�double-����-negation-free�շpro�M�ofs�in�L.�W��ee�no��!w�construct�the�desired�pro�of�as�follo��!ws.����-First���write�the�double-negation-free�pro�M�ofs�of�the�axioms�of�L*.��;Then�write����-pro�M�ofs��of�the�substitution�instances�of�axioms�of�L*�that�are�required.��These����-actions�-Lpro��!vide�pro�M�ofs�of�all�the�substitution�instances�of�axioms�of�L*,�E�from����-L��frather��xthan�from�L*.��No��!w�write�the�pro�M�of�of��B����������	6 �from�those�substitution����-instances.��W��ee�:�ha��!v�e�the�desired�pro�M�of.��The�only�double�negations�it�con�tains����-are��fthose�con��!tained�in��B����������L��.���That�completes�the�pro�M�of�of�the�theorem.��fg��>�
src:779dn.texEsp�M�ecially�b�in�view�of�the�discussion�fo�cusing�on�the�F��erege�axiom�system,����-a��natural�question�arises�concerning�its�use�as�h��!yp�M�othesis.�(fIn�particular,�0if����-the�O�theorem�to�b�M�e�pro��!v�ed�O�is�itself�free�of�double�negation,�am��!ust�there�exist�a����-double-negation-free���pro�M�of�of�it�with�the�F��erege�system�as�h��!yp�othesis?�ݸAfter����-all,��Rthat��
system�con��!tains�t�w�o�mem�b�M�ers�exhibiting�double�negation.��Because����-w��!e�P3ha�v�e�in�hand�a�pro�M�of�that�deduces�from�the�F��erege�system�the�featured�����-t� ����L��3�uk��dDasiewicz�ɭaxiom�system�suc��!h�that�the�pro�M�of�is�free�of�double�negation,����-suc��!h��fa�pro�M�of�m�ust�exist.��6fh����-�Theorem�2�4�(Strong�D-completeness)�������
src:786dn.tex�Supp��p�ose�C�L�C�is�a�lo�gic�that�admits����-str��p�ong�3double-ne�gation�elimination.���If�L�2�pr�oves��A��using�mo�dus�p�onens�and����-substitution,�J�without�7�using�double�ne��p�gations�exc�ept�those�that�alr�e�ady�o�c�cur����-as��subformulas�of��A�,��Sthen�L��	pr��p�oves��A��using�c�ondense�d�detachment,��Swithout����-using���double�ne��p�gations�exc�ept�those�that�alr�e�ady�o�c�cur�as�subformulas�of��A�.��1����-�
src:796dn.texPr��p�o�of�R�:�
vSupp�M�ose���L��kpro��!v�es��A��using�mo�M�dus�p�onens�and�substitution.� �Then����-b��!y��TTheorem�2,�
Pthere�is�a�condensed-detac�hmen�t�pro�M�of�of��A��(p�ossibly�using����-new��rdouble�negations).��By�strong�double-negation�elimination,���there�is�a����-condensed-detac��!hmen�t�pro�M�of�of��A��in�L,�using�only�double�negations�that����-already��fo�M�ccur�as�subform��!ulas�of��A�.���"$����f14�����Ǡ�j֍����*�1F'���홊���-�6���A� ���qL��I\ruk��@asiewicz's��System�L1{L3���1��-�
src:807dn.tex�As��men��!tioned�in�Section�3,���# ����L��	�Vuk��dDasiewicz's�system�L��has�the�follo�wing�axioms.���������a�L1���������i�(�i�(�x;��1y�d��)�;�i�(�i�(�y�;�z�{I�)�;�i�(�x;�z��)))�����������a�L2��������i�(�i�(�n�(�x�)�;��1x�)�;�x�)���������a�L3������K��i�(�x;��1i�(�n�(�x�)�;�y�d��))����ߌ����-�Lemma�2�5���el.�
src:815dn.tex�F��)r��p�om�ֺL1{L3,���one�c�an�nd�double-ne�gation-fr�e�e�pr�o�ofs�of�formu-��
����-las���D1{D3.���䍑-�
src:817dn.texPr��p�o�of�R�.��UF��eorm��!ula�t�D1�is��i�(�x;��1x�).�The�follo��!wing�is�a�t�w�o-line�pro�M�of�pro�duced�b��!y����-Otter.�������:b%1.����H`�
src:831dn.tex�i�(�i�(�i�(�n�(�x�)�;��1y�d��)�;�z�{I�)�;�i�(�x;�z��))��[L1,��fL3]���ύ����:b%2.����H`�
src:832dn.tex�i�(�x;��1x�)�bK[1,��fL2]����>�
src:841dn.texThe��ffollo��!wing�is�an�Otter�pro�M�of�of�D2�from�L1{L3.�������:b%1.����H`�
src:853dn.tex�i�(�n�(�i�(�i�(�n�(�x�)�;��1x�)�;�x�))�;�y�d��)���[L3,��fL2]�������:b%2.����H`�
src:854dn.tex�i�(�i�(�i�(�i�(�x;��1y�d��)�;�i�(�z�{I;�y��))�;�u�)�;�i�(�i�(�z�{I;�x�)�;�u�))��e�[L1,��fL1]�������:b%3.����H`�
src:855dn.tex�i�(�i�(�x;��1y�d��)�;�i�(�n�(�i�(�i�(�n�(�z�{I�)�;�z��)�;�z��))�;�y�d��))����[L1,��f1]�������:b%4.����H`�
src:856dn.tex�i�(�i�(�i�(�n�(�x�)�;��1y�d��)�;�z�{I�)�;�i�(�x;�z��))��[L1,��fL3]�������:b%5.����H`�
src:857dn.tex�i�(�x;��1x�)�bK[4,��fL2]�������:b%6.����H`�
src:858dn.tex�i�(�n�(�i�(�x;��1x�))�;�y�d��)��`�[L3,��f5]�������:b%7.����H`�
src:859dn.tex�i�(�x;��1i�(�n�(�i�(�i�(�n�(�y�d��)�;�y��)�;�y��))�;�z�{I�))���c[4,��f3]�������:b%8.����H`�
src:860dn.tex�i�(�i�(�x;��1i�(�y�d�;�z�{I�))�;�i�(�i�(�u;�y��)�;�i�(�x;�i�(�u;�z�{I�))))���[2,��f2]�������:b%9.����H`�
src:861dn.tex�i�(�i�(�x;��1y�d��)�;�i�(�i�(�n�(�i�(�y�;�z�{I�))�;�i�(�y�;�z�{I�))�;�i�(�x;�z��)))�yZ�[8,��fL2]�������4�10.����H`�
src:862dn.tex�i�(�i�(�x;��1i�(�n�(�i�(�y�d�;�z�{I�))�;�i�(�y�;�z�{I�)))�;�i�(�i�(�u;�y��)�;�i�(�x;�i�(�u;�z�{I�))))�QY�[8,��f9]�������4�11.����H`�
src:863dn.tex�i�(�i�(�x;��1i�(�n�(�y�d��)�;�y��))�;�i�(�z�{I;�i�(�x;�y��)))��.�[10,��f7]�������4�12.����H`�
src:864dn.tex�i�(�i�(�n�(�x�)�;��1y�d��)�;�i�(�z�{I;�i�(�i�(�y�;�x�)�;�x�)))���|[2,��f11]�������4�13.����H`�
src:865dn.tex�i�(�i�(�x;��1i�(�y�d�;�z�{I�))�;�i�(�i�(�n�(�z��)�;�y�d��)�;�i�(�x;�z��)))��*L[10,��f12]�������4�14.����H`�
src:866dn.tex�i�(�i�(�x;��1i�(�n�(�y�d��)�;�z�{I�))�;�i�(�i�(�u;�i�(�z�;�y�d��))�;�i�(�x;�i�(�u;�y��))))�b��[8,��f13]�������4�15.����H`�
src:867dn.tex�i�(�i�(�n�(�x�)�;��1y�d��)�;�i�(�i�(�y�;�x�)�;�x�))��jD[13,��f5]�������4�16.����H`�
src:868dn.tex�i�(�i�(�n�(�x�)�;��1n�(�y�d��))�;�i�(�y�;�x�))���c[13,��fL3]�������4�17.����H`�
src:869dn.tex�i�(�x;��1i�(�y�d�;�x�))����[4,��f16]�������4�18.����H`�
src:870dn.tex�i�(�i�(�n�(�x�)�;��1y�d��)�;�i�(�x;�x�))���w[13,��f17]�������4�19.����H`�
src:871dn.tex�i�(�i�(�x;��1x�)�;�i�(�x;�x�))���[18,��f6]�������4�20.����H`�
src:872dn.tex�i�(�i�(�x;��1i�(�y�d�;�z�{I�))�;�i�(�y�;�i�(�x;�z�{I�)))����[14,��f17]���"$����f15�������j֍����*�1F'���홊�����4苹21.����H`�
src:873dn.tex�i�(�i�(�x;��1y�d��)�;�i�(�i�(�n�(�y��)�;�x�)�;�y��))��o�[20,��f15]���K�����4�22.����H`�
src:874dn.tex�i�(�n�(�x�)�;��1i�(�x;�y�d��))���
[20,��fL3]�������4�23.����H`�
src:875dn.tex�i�(�i�(�n�(�x�)�;��1y�d��)�;�i�(�n�(�y��)�;�x�))��C�[13,��f22]�������4�24.����H`�
src:876dn.tex�i�(�n�(�i�(�x;��1n�(�y�d��)))�;�y��)�ح�[23,��f17]�������4�25.����H`�
src:877dn.tex�i�(�i�(�x;��1i�(�n�(�y�d��)�;�z�{I�))�;�i�(�i�(�z�;�y�d��)�;�i�(�x;�y��)))��s�[8,��f21]�������4�26.����H`�
src:878dn.tex�i�(�i�(�n�(�x�)�;��1y�d��)�;�i�(�i�(�z�{I;�x�)�;�i�(�i�(�y�;�z�{I�)�;�x�)))���[2,��f25]�������4�27.����H`�
src:879dn.tex�i�(�i�(�x;��1i�(�y�d�;�n�(�z�{I�)))�;�i�(�i�(�z�;�x�)�;�i�(�y�d�;�n�(�z��))))�}r[26,��f24]�������4�28.����H`�
src:880dn.tex�i�(�i�(�x;��1y�d��)�;�i�(�n�(�y��)�;�n�(�x�)))���c[27,��fL3]�������4�29.����H`�
src:881dn.tex�i�(�i�(�i�(�n�(�x�)�;��1n�(�y�d��))�;�z�{I�)�;�i�(�i�(�y�;�x�)�;�z�{I�))��o�[L1,��f28]�������4�30.����H`�
src:882dn.tex�i�(�i�(�x;��1x�)�;�i�(�n�(�x�)�;�n�(�x�)))��Ep[29,��f19]��͚��>�
src:889dn.texFinally��e,�w��!e���are�ready�to�pro�v�e�D3.���A��pro�M�of�of�D3�w�as�originally�found�us-��
����-ing��$a�sp�M�ecially�compiled�v��!ersion�of�Otter.���(The�dicult�y�is�that�normal�Otter����-deriv��!es��fa�more�general�conclusion,�whic�h�subsumes�the�desired�conclusion.)�������:b%1.����H`�
src:903dn.tex�i�(�n�(�i�(�i�(�n�(�x�)�;��1x�)�;�x�))�;�y�d��)���[L3,��fL2]���K�����:b%2.����H`�
src:904dn.tex�i�(�i�(�i�(�i�(�x;��1y�d��)�;�i�(�z�{I;�y��))�;�u�)�;�i�(�i�(�z�{I;�x�)�;�u�))��e�[L1,��fL1]�������:b%3.����H`�
src:905dn.tex�i�(�i�(�x;��1y�d��)�;�i�(�n�(�i�(�i�(�n�(�z�{I�)�;�z��)�;�z��))�;�y�d��))����[L1,��f1]�������:b%4.����H`�
src:906dn.tex�i�(�i�(�i�(�n�(�x�)�;��1y�d��)�;�z�{I�)�;�i�(�x;�z��))��[L1,��fL3]�������:b%5.����H`�
src:907dn.tex�i�(�x;��1i�(�n�(�i�(�i�(�n�(�y�d��)�;�y��)�;�y��))�;�z�{I�))���c[4,��f3]�������:b%6.����H`�
src:908dn.tex�i�(�i�(�i�(�n�(�i�(�i�(�n�(�x�)�;��1x�)�;�x�))�;�y�d��)�;�z�{I�)�;�i�(�u;�z��))���[L1,��f5]�������:b%7.����H`�
src:909dn.tex�i�(�i�(�x;��1i�(�y�d�;�z�{I�))�;�i�(�i�(�u;�y��)�;�i�(�x;�i�(�u;�z�{I�))))���[2,��f2]�������:b%8.����H`�
src:910dn.tex�i�(�i�(�x;��1y�d��)�;�i�(�i�(�i�(�x;�z�{I�)�;�u�)�;�i�(�i�(�y�;�z�{I�)�;�u�)))��=�[2,��fL1]�������:b%9.����H`�
src:911dn.tex�i�(�x;��1i�(�i�(�n�(�y�d��)�;�y��)�;�y��))��[6,��fL2]�������4�10.����H`�
src:912dn.tex�i�(�i�(�x;��1i�(�n�(�y�d��)�;�y��))�;�i�(�z�{I;�i�(�x;�y��)))���C[7,��f9]�������4�11.����H`�
src:913dn.tex�i�(�i�(�n�(�x�)�;��1y�d��)�;�i�(�z�{I;�i�(�i�(�y�;�x�)�;�x�)))���|[2,��f10]�������4�12.����H`�
src:914dn.tex�i�(�x;��1i�(�y�d�;�y��))��}[10,��fL3]�������4�13.����H`�
src:915dn.tex�i�(�x;��1i�(�i�(�n�(�y�d��)�;�z�{I�)�;�i�(�i�(�z�;�y�d��)�;�y��)))��dA[10,��f11]�������4�14.����H`�
src:916dn.tex�i�(�i�(�n�(�x�)�;��1y�d��)�;�i�(�i�(�y�;�x�)�;�x�))���[13,��f13]�������4�15.����H`�
src:917dn.tex�i�(�i�(�x;��1i�(�y�d�;�z�{I�))�;�i�(�i�(�n�(�z��)�;�y�d��)�;�i�(�x;�z��)))����[7,��f14]�������4�16.����H`�
src:918dn.tex�i�(�x;��1i�(�i�(�y�d�;�x�)�;�x�))��k�[4,��f14]�������4�17.����H`�
src:919dn.tex�i�(�i�(�x;��1i�(�y�d�;�z�{I�))�;�i�(�z�;�i�(�x;�z��)))����[7,��f16]�������4�18.����H`�
src:920dn.tex�i�(�x;��1i�(�x;�x�))���[17,��f16]�������4�19.����H`�
src:921dn.tex�i�(�x;��1i�(�y�d�;�x�))����[17,��f9]�������4�20.����H`�
src:922dn.tex�i�(�i�(�x;��1i�(�y�d�;�y��))�;�i�(�x;�i�(�y�;�y��)))����[18,��f12]�������4�21.����H`�
src:923dn.tex�i�(�i�(�x;��1i�(�y�d�;�x�))�;�i�(�x;�i�(�y�;�x�)))����[18,��f19]�������4�22.����H`�
src:924dn.tex�i�(�i�(�x;��1i�(�n�(�y�d��)�;�z�{I�))�;�i�(�i�(�u;�i�(�z�;�y�d��))�;�i�(�x;�i�(�u;�y��))))�b��[7,��f15]�������4�23.����H`�
src:925dn.tex�i�(�i�(�x;��1i�(�i�(�y�d�;�z�{I�)�;�u�))�;�i�(�i�(�y�;�v��)�;�i�(�x;�i�(�i�(�v�;�z�{I�)�;�u�))))�`_�[7,��f8]���"$����f16�����g��j֍����*�1F'���홊�����4苹24.����H`�
src:926dn.tex�i�(�i�(�x;��1i�(�y�d�;�z�{I�))�;�i�(�y�;�i�(�x;�z�{I�)))����[22,��f19]���������4�25.����H`�
src:927dn.tex�i�(�i�(�i�(�x;��1i�(�y�d�;�z�{I�))�;�u�)�;�i�(�i�(�y�;�i�(�x;�z�{I�))�;�u�))���][L1,��f24]�������4�26.����H`�
src:928dn.tex�i�(�i�(�x;��1i�(�y�d�;�x�))�;�i�(�y�;�i�(�x;�x�)))����[25,��f20]�������4�27.����H`�
src:929dn.tex�i�(�i�(�x;��1x�)�;�i�(�i�(�y�d�;�x�)�;�i�(�y�;�x�)))��?#[26,��fL1]�������4�28.����H`�
src:930dn.tex�i�(�i�(�x;��1i�(�i�(�y�d�;�z�{I�)�;�u�))�;�i�(�i�(�y�;�y��)�;�i�(�x;�i�(�i�(�y�;�z�{I�)�;�u�))))�UM*[21,��f23]�������4�29.����H`�
src:931dn.tex�i�(�i�(�x;��1x�)�;�i�(�i�(�y�d�;�y��)�;�i�(�i�(�x;�y��)�;�i�(�x;�y��))))����[28,��f27]�����-�
src:939dn.texThat��fcompletes�the�pro�M�of�of�the�lemma.��,����-�Theorem�2�5���n>M�
src:941dn.tex���6� ��}lL���7ukasiewicz's��^system�L1{L3�admits�str��p�ong�double-ne�gation�elim-��
����-ination.����-�
src:945dn.texPr��p�o�of�.� �W��ee���b�M�egin�b��!y�calculating�the�form�ulas�L*�for�this�system.� �W��ee�obtain����-the��ffollo��!wing.���㍍���;�L4�����À��i�(�i�(�x;��1n�(�x�))�;�n�(�x�))�����������;�L5������K��i�(�n�(�x�)�;��1i�(�x;�y�d��))������-�
src:951dn.texBy���Theorem�3�it�suces�to�v��!erify�that�there�exist�double-negation-free�pro�M�ofs����-of�/�L4,�G�L5,�and�D1{D3.��`W��ee�ha��!v�e�already�v�eried�D1{D3�ab�M�o�v�e,�G�so�it�remains����-only�1�to�exhibit�double-negation-free�pro�M�ofs�of�L4�and�L5.��The�follo��!wing�is����-an��fOtter�pro�M�of�of�L4.��������:b%1.����H`�
src:968dn.tex�i�(�n�(�i�(�i�(�n�(�x�)�;��1x�)�;�x�))�;�y�d��)���[L3,��fL2]���������:b%2.����H`�
src:969dn.tex�i�(�i�(�i�(�i�(�x;��1y�d��)�;�i�(�z�{I;�y��))�;�u�)�;�i�(�i�(�z�{I;�x�)�;�u�))��e�[L1,��fL1]�������:b%3.����H`�
src:970dn.tex�i�(�i�(�x;��1y�d��)�;�i�(�n�(�i�(�i�(�n�(�z�{I�)�;�z��)�;�z��))�;�y�d��))����[L1,��f1]�������:b%4.����H`�
src:971dn.tex�i�(�i�(�i�(�n�(�x�)�;��1y�d��)�;�z�{I�)�;�i�(�x;�z��))��[L1,��fL3]�������:b%5.����H`�
src:972dn.tex�i�(�i�(�x;��1y�d��)�;�i�(�i�(�n�(�x�)�;�x�)�;�y��))��3�[L1,��fL2]�������:b%6.����H`�
src:973dn.tex�i�(�x;��1x�)�bK[4,��fL2]�������:b%7.����H`�
src:974dn.tex�i�(�x;��1i�(�n�(�i�(�i�(�n�(�y�d��)�;�y��)�;�y��))�;�z�{I�))���c[4,��f3]�������:b%8.����H`�
src:975dn.tex�i�(�i�(�x;��1i�(�y�d�;�z�{I�))�;�i�(�i�(�u;�y��)�;�i�(�x;�i�(�u;�z�{I�))))���[2,��f2]�������:b%9.����H`�
src:976dn.tex�i�(�i�(�x;��1y�d��)�;�i�(�i�(�n�(�i�(�y�;�z�{I�))�;�i�(�y�;�z�{I�))�;�i�(�x;�z��)))��2�[2,��f5]�������4�10.����H`�
src:977dn.tex�i�(�i�(�x;��1i�(�n�(�i�(�y�d�;�z�{I�))�;�i�(�y�;�z�{I�)))�;�i�(�i�(�u;�y��)�;�i�(�x;�i�(�u;�z�{I�))))�QY�[8,��f9]�������4�11.����H`�
src:978dn.tex�i�(�i�(�x;��1i�(�n�(�y�d��)�;�y��))�;�i�(�z�{I;�i�(�x;�y��)))��.�[10,��f7]�������4�12.����H`�
src:979dn.tex�i�(�i�(�n�(�x�)�;��1y�d��)�;�i�(�z�{I;�i�(�i�(�y�;�x�)�;�x�)))���|[2,��f11]�������4�13.����H`�
src:980dn.tex�i�(�i�(�x;��1i�(�y�d�;�z�{I�))�;�i�(�i�(�n�(�z��)�;�y�d��)�;�i�(�x;�z��)))��*L[10,��f12]�������4�14.����H`�
src:981dn.tex�i�(�i�(�x;��1i�(�n�(�y�d��)�;�z�{I�))�;�i�(�i�(�u;�i�(�z�;�y�d��))�;�i�(�x;�i�(�u;�y��))))�b��[8,��f13]�������4�15.����H`�
src:982dn.tex�i�(�i�(�n�(�x�)�;��1n�(�y�d��))�;�i�(�y�;�x�))���c[13,��fL3]�������4�16.����H`�
src:983dn.tex�i�(�x;��1i�(�y�d�;�x�))����[4,��f15]���"$����f17�������j֍����*�1F'���홊�����4苹17.����H`�
src:984dn.tex�i�(�i�(�i�(�x;��1y�d��)�;�z�{I�)�;�i�(�y�;�z�{I�))���[L1,��f16]���������4�18.����H`�
src:985dn.tex�i�(�n�(�x�)�;��1i�(�x;�y�d��))��Eq[17,��f15]�������4�19.����H`�
src:986dn.tex�i�(�i�(�n�(�x�)�;��1y�d��)�;�i�(�n�(�y��)�;�x�))��C�[13,��f18]�������4�20.����H`�
src:987dn.tex�i�(�i�(�x;��1i�(�y�d�;�z�{I�))�;�i�(�i�(�n�(�y��)�;�z�{I�)�;�i�(�x;�z��)))��*L[14,��f19]�������4�21.����H`�
src:988dn.tex�i�(�i�(�n�(�x�)�;��1y�d��)�;�i�(�i�(�x;�y��)�;�y��))���q[20,��f6]�������4�22.����H`�
src:989dn.tex�i�(�i�(�x;��1n�(�x�))�;�n�(�x�))��)[21,��f6]��位�>�
src:996dn.texThe��ffollo��!wing�is�an�Otter�pro�M�of�of�L5.�������:b%1.����H`�src:1008dn.tex�i�(�n�(�i�(�i�(�n�(�x�)�;��1x�)�;�x�))�;�y�d��)���[L3,��fL2]���������:b%2.����H`�src:1009dn.tex�i�(�i�(�i�(�i�(�x;��1y�d��)�;�i�(�z�{I;�y��))�;�u�)�;�i�(�i�(�z�{I;�x�)�;�u�))��e�[L1,��fL1]�������:b%3.����H`�src:1010dn.tex�i�(�i�(�x;��1y�d��)�;�i�(�n�(�i�(�i�(�n�(�z�{I�)�;�z��)�;�z��))�;�y�d��))����[L1,��f1]�������:b%4.����H`�src:1011dn.tex�i�(�i�(�i�(�n�(�x�)�;��1y�d��)�;�z�{I�)�;�i�(�x;�z��))��[L1,��fL3]�������:b%5.����H`�src:1012dn.tex�i�(�i�(�x;��1y�d��)�;�i�(�i�(�n�(�x�)�;�x�)�;�y��))��3�[L1,��fL2]�������:b%6.����H`�src:1013dn.tex�i�(�x;��1i�(�n�(�i�(�i�(�n�(�y�d��)�;�y��)�;�y��))�;�z�{I�))���c[4,��f3]�������:b%7.����H`�src:1014dn.tex�i�(�i�(�x;��1i�(�y�d�;�z�{I�))�;�i�(�i�(�u;�y��)�;�i�(�x;�i�(�u;�z�{I�))))���[2,��f2]�������:b%8.����H`�src:1015dn.tex�i�(�i�(�x;��1y�d��)�;�i�(�i�(�n�(�i�(�y�;�z�{I�))�;�i�(�y�;�z�{I�))�;�i�(�x;�z��)))��2�[2,��f5]�������:b%9.����H`�src:1016dn.tex�i�(�i�(�x;��1i�(�n�(�i�(�y�d�;�z�{I�))�;�i�(�y�;�z�{I�)))�;�i�(�i�(�u;�y��)�;�i�(�x;�i�(�u;�z�{I�))))�QY�[7,��f8]�������4�10.����H`�src:1017dn.tex�i�(�i�(�x;��1i�(�n�(�y�d��)�;�y��))�;�i�(�z�{I;�i�(�x;�y��)))���C[9,��f6]�������4�11.����H`�src:1018dn.tex�i�(�i�(�n�(�x�)�;��1y�d��)�;�i�(�z�{I;�i�(�i�(�y�;�x�)�;�x�)))���|[2,��f10]�������4�12.����H`�src:1019dn.tex�i�(�i�(�x;��1i�(�y�d�;�z�{I�))�;�i�(�i�(�n�(�z��)�;�y�d��)�;�i�(�x;�z��)))����[9,��f11]�������4�13.����H`�src:1020dn.tex�i�(�i�(�n�(�x�)�;��1n�(�y�d��))�;�i�(�y�;�x�))���c[12,��fL3]�������4�14.����H`�src:1021dn.tex�i�(�x;��1i�(�y�d�;�x�))����[4,��f13]�������4�15.����H`�src:1022dn.tex�i�(�i�(�i�(�x;��1y�d��)�;�z�{I�)�;�i�(�y�;�z�{I�))���[L1,��f14]�������4�16.����H`�src:1023dn.tex�i�(�n�(�x�)�;��1i�(�x;�y�d��))��Eq[15,��f13]��位�-�src:1031dn.texThat��fcompletes�the�pro�M�of�of�the�theorem.���č���-�Corollary�2�1���p�n�src:1033dn.tex�L��p�et���T����b�e�any�set�of�axioms�for�(two-value�d)�pr�op�ositional�lo�gic.��
����-Supp��p�ose�,�that�ther�e�exist�double-ne�gation-fr�e�e�c�ondense�d-detachment�pr�o�ofs�of����-L1{L3�@�fr��p�om��T��V�.�ЪThen�the�pr�e�c�e�ding�the�or�em�is�true�with��T����in�plac�e�of�L1{L3.����-�src:1041dn.texPr��p�o�of�R�.��rW��ee���m��!ust�sho�w�that��T�.C�admits�strong�double-negation�elimination.����-Let�Ǥ�A��b�M�e�pro��!v��dDable�from��T��V�,���and�let��A���b�e�obtained�from��A��b��!y�erasing�some����-of��athe�double�negations�in��A��(but�all�o�M�ccurrences�of�an��!y�giv�en�form�ula�if�there����-are�im��!ultiple�o�M�ccurrences�of�the�same�doubly�negated�subform�ula).��kW��ee�m�ust����-sho��!w��hthat��T�c��pro�v�es��A���b�y�a�pro�M�of�whose�doubly�negated�subform�ula�o�M�ccur�in����-�A��.���Since����T�ݹis�an�axiomatization�of�t��!w�o-v��dDalued���logic,����A���is�a�tautology�and����-hence�pro��!v��dDable�from�L1{L3.��hBy�the�theorem,�2there�exists�a�pro�M�of�of��A���from���"$����f18����ꌠ�j֍����*�1F'���홊��-�L1{L3�1)that�con��!tains�no�double�negations�(except�those�o�M�ccurring�in��A��,�S�if��
����-an��!y).�X�Supplying��Mthe�giv�en�pro�M�ofs�of�L1{L3�from��T��V�,�هw�e�construct�a�pro�M�of�of����-�A��ٻ�from��T�_�that�con��!tains�no�double�negations�except�those�o�M�ccurring�in��A�����-�(if��fan��!y).���That�completes�the�pro�M�of.��
�<��>�src:1061dn.tex�Example�.��W��ee��can�tak��!e��T�Hp�to�con�tain�exactly�one�form�ula,��the�single�axiom����-M���of���Meredith.��M�is�double-negation�free,��and�double-negation-free�pro�M�ofs����-of��rL1{L3�from�M��oha��!v�e��rb�M�een�found�using�Otter�[�18��
�4].�ٌTherefore,��
the�theorem����-is��ftrue�for�the�single�axiom�M.��y����-�7��A@Innite-V���alued��Logic���Q��-�src:1069dn.tex��tι ����L���uk��dDasiewicz's�w)innite-v�alued�logic�is�a�subsystem�of�classical�prop�M�ositional����-logic��that�w��!as�studied�in�the�1930s.�PThe�logic�is�of�in�terest�partly�b�M�ecause����-there��exists�a�natural�seman��!tics�for�it,��according�to�whic�h�prop�M�ositions�are����-assigned�5�truth�v��dDalues�that�are�real�(or�rational)�n��!um�b�M�ers�5�b�et��!w�een�0�and����-1,�4�with��<1�b�M�eing�true�and�0�b�eing�false.��	, ����L��r_uk��dDasiewicz's�axioms�A1{A4�are����-complete�*0for�this�seman��!tics,�K"as�w�as�pro�v�ed�(but�apparen�tly�not�published)����-b��!y��W��easjb�M�erg,�Nand�pro�v�ed�again�b�y�Chang�[�1��y�].��Axioms�A1{A4�are�form�ulated����-b��!y��<using�implication��i�(�p;��1q�d��)�and�negation��n�(�p�)�only��e.�d`The�truth�v��dDalue�of��p��is����-denoted��fb��!y��k�p�k�.���T��eruth�v��dDalues�are�giv�en�b�y�������]�k�n�(�p�)�k�
��=�1�n���k�p�k���h���6k�i�(�p;��1q�d��)�k�
��=���min���J�(1�n���k�p�k��+��k�q��k�;��1�1)�:���R��-�src:1087dn.tex�Axioms��fA1{A4�are�as�follo��!ws.������3������>�src:1093dn.tex�������C*�A1�����ӆ*�i�(�x;��1i�(�y�d�;�x�))�����������C*A2������U��i�(�i�(�x;��1y�d��)�;�i�(�i�(�y�;�z�{I�)�;�i�(�x;�z��)))���������C*A3�������`�i�(�i�(�i�(�x;��1y�d��)�;�y��)�;�i�(�i�(�y�;�x�)�;�x�))���������C*A4���������i�(�i�(�n�(�x�)�;��1n�(�y�d��))�;�i�(�y�;�x�))����H#��-�src:1101dn.texThe��fstandard�reference�for�innite-v��dDalued�logic�is�[�13��
�4].���g����-�Lemma�2�6���el.�src:1104dn.tex�A1{A4���pr��p�ove�formulas�D1{D3�without�double�ne�gation.��sō�-�src:1107dn.texPr��p�o�of�R�.���The��ffollo��!wing�is�an�Otter�pro�M�of�of�D1�from�A1{A4.��-�
6�ff�e֟
L͍����{���-=�3�����a�In��comparison�with��C� ���L����uk��|rasiewicz's�axioms�L1{L3,��|axiom�A2�is�the�same�as�L1,�and�L3���is�Tpro��9v��|rable�from�A1{A4,�but�L2�is�not�pro�v��|rable�from�A1{A4.���"$����f�19�������j֍����*�1F'���홊�����:b%�1.����H`�src:1120dn.tex�i�(�x;��1i�(�y�d�;�i�(�z�{I;�y��)))�ڧ[A1,��fA1]��dȍ����:b%2.����H`�src:1121dn.tex�i�(�i�(�i�(�x;��1y�d��)�;�z�{I�)�;�i�(�y�;�z�{I�))���N[A2,��fA1]�������:b%3.����H`�src:1122dn.tex�i�(�i�(�i�(�x;��1i�(�y�d�;�x�))�;�z�{I�)�;�z��)�ˣ�[A3,��f1]�������:b%4.����H`�src:1123dn.tex�i�(�x;��1x�)�:L[2,��f3]��W��>�src:1130dn.texThe���follo��!wing�is�an�Otter�pro�M�of�of�D2�from�A1{A4,��found�b�y�using�a��
����-sp�M�ecially��fcompiled�v��!ersion�of�Otter.�������:b%1.����H`�src:1144dn.tex�i�(�x;��1i�(�y�d�;�i�(�z�{I;�y��)))�ڧ[A1,��fA1]��dȍ����:b%2.����H`�src:1145dn.tex�i�(�i�(�i�(�i�(�x;��1y�d��)�;�i�(�z�{I;�y��))�;�u�)�;�i�(�i�(�z�{I;�x�)�;�u�))�}�*[A2,��fA2]�������:b%3.����H`�src:1146dn.tex�i�(�i�(�i�(�x;��1y�d��)�;�z�{I�)�;�i�(�y�;�z�{I�))���N[A2,��fA1]�������:b%4.����H`�src:1147dn.tex�i�(�n�(�x�)�;��1i�(�x;�y�d��))��>[3,��fA4]�������:b%5.����H`�src:1148dn.tex�i�(�i�(�i�(�x;��1y�d��)�;�z�{I�)�;�i�(�n�(�x�)�;�z��))����[A2,��f4]�������:b%6.����H`�src:1149dn.tex�i�(�x;��1i�(�i�(�x;�y�d��)�;�y��))��.K[3,��fA3]�������:b%7.����H`�src:1150dn.tex�i�(�i�(�i�(�i�(�x;��1y�d��)�;�y��)�;�z�{I�)�;�i�(�i�(�i�(�y�;�x�)�;�x�)�;�z�{I�))�~-�[A2,��fA3]�������:b%8.����H`�src:1151dn.tex�i�(�i�(�i�(�x;��1i�(�y�d�;�x�))�;�z�{I�)�;�z��)�ˣ�[A3,��f1]�������:b%9.����H`�src:1152dn.tex�i�(�x;��1x�)�:L[3,��f8]�������4�10.����H`�src:1153dn.tex�i�(�i�(�i�(�x;��1x�)�;�y�d��)�;�y��)��d�[6,��f9]�������4�11.����H`�src:1154dn.tex�i�(�i�(�x;��1i�(�y�d�;�y��))�;�i�(�y�;�y��))��I[A3,��f10]�������4�12.����H`�src:1155dn.tex�i�(�i�(�x;��1x�)�;�i�(�x;�x�))���[3,��f11]�������4�13.����H`�src:1156dn.tex�i�(�n�(�x�)�;��1i�(�x;�x�))��?�[5,��f12]�������4�14.����H`�src:1157dn.tex�i�(�i�(�x;��1i�(�y�d�;�x�))�;�i�(�x;�i�(�y�;�x�)))��$[12,��f1]�������4�15.����H`�src:1158dn.tex�i�(�i�(�x;��1y�d��)�;�i�(�y�;�i�(�x;�y��)))��K[3,��f14]�������4�16.����H`�src:1159dn.tex�i�(�i�(�x;��1x�)�;�i�(�n�(�x�)�;�i�(�x;�x�)))���P[15,��f13]�������4�17.����H`�src:1160dn.tex�i�(�i�(�x;��1y�d��)�;�i�(�x;�x�))��k�[2,��f11]�������4�18.����H`�src:1161dn.tex�i�(�x;��1i�(�i�(�y�d�;�z�{I�)�;�i�(�y�;�y��)))��y[A1,��f17]�������4�19.����H`�src:1162dn.tex�i�(�i�(�i�(�i�(�x;��1y�d��)�;�i�(�z�{I;�y��))�;�i�(�z�{I;�y��))�;�i�(�i�(�x;�z�{I�)�;�i�(�x;�y��)))�`��[7,��f2]�������4�20.����H`�src:1163dn.tex�i�(�i�(�x;��1i�(�y�d�;�z�{I�))�;�i�(�x;�i�(�y�;�y��)))����[19,��f18]�������4�21.����H`�src:1164dn.tex�i�(�i�(�x;��1x�)�;�i�(�n�(�x�)�;�n�(�x�)))��Ep[20,��f16]��W��>�src:1171dn.texThe��/follo��!wing�is�a�pro�M�of�of�D3,��found�b�y�using�a�sp�M�ecially�compiled�v�ersion����-of��fOtter.�������:b%1.����H`�src:1185dn.tex�i�(�x;��1i�(�y�d�;�i�(�z�{I;�y��)))�ڧ[A1,��fA1]��dȍ����:b%2.����H`�src:1186dn.tex�i�(�i�(�i�(�i�(�x;��1y�d��)�;�i�(�z�{I;�y��))�;�u�)�;�i�(�i�(�z�{I;�x�)�;�u�))�}�*[A2,��fA2]�������:b%3.����H`�src:1187dn.tex�i�(�i�(�i�(�x;��1y�d��)�;�z�{I�)�;�i�(�y�;�z�{I�))���N[A2,��fA1]�������:b%4.����H`�src:1188dn.tex�i�(�n�(�x�)�;��1i�(�x;�y�d��))��>[3,��fA4]�������:b%5.����H`�src:1189dn.tex�i�(�x;��1i�(�y�d�;�i�(�z�{I;�x�)))��^P[3,��fA1]�������:b%6.����H`�src:1190dn.tex�i�(�x;��1i�(�i�(�x;�y�d��)�;�y��))��.K[3,��fA3]���"$����f20������j֍����*�1F'���홊�����:b%�7.����H`�src:1191dn.tex�i�(�i�(�x;��1i�(�y�d�;�z�{I�))�;�i�(�i�(�u;�y��)�;�i�(�x;�i�(�u;�z�{I�))))���[2,��f2]��k
�����:b%8.����H`�src:1192dn.tex�i�(�i�(�x;��1y�d��)�;�i�(�z�{I;�i�(�x;�z��)))����[2,��f3]�������:b%9.����H`�src:1193dn.tex�i�(�i�(�x;��1y�d��)�;�i�(�i�(�i�(�x;�z�{I�)�;�u�)�;�i�(�i�(�y�;�z�{I�)�;�u�)))��ߑ[2,��fA2]�������4�10.����H`�src:1194dn.tex�i�(�x;��1i�(�n�(�y�d��)�;�x�))��8�[8,��f4]�������4�11.����H`�src:1195dn.tex�i�(�n�(�x�)�;��1i�(�y�d�;�i�(�z�{I;�i�(�u;�z��))))��F�[10,��f1]�������4�12.����H`�src:1196dn.tex�i�(�i�(�x;��1i�(�y�d�;�z�{I�))�;�i�(�y�;�i�(�x;�z�{I�)))���"[7,��f6]�������4�13.����H`�src:1197dn.tex�i�(�i�(�i�(�x;��1i�(�y�d�;�z�{I�))�;�u�)�;�i�(�i�(�y�;�i�(�x;�z�{I�))�;�u�))��e�[A2,��f12]�������4�14.����H`�src:1198dn.tex�i�(�x;��1i�(�y�d�;�y��))���[12,��fA1]�������4�15.����H`�src:1199dn.tex�i�(�x;��1i�(�y�d�;�i�(�z�{I;�i�(�u;�u�))))�ͫ�[5,��f14]�������4�16.����H`�src:1200dn.tex�i�(�i�(�i�(�x;��1x�)�;�y�d��)�;�y��)�ܴ�[A3,��f14]�������4�17.����H`�src:1201dn.tex�i�(�i�(�x;��1i�(�y�d�;�y��))�;�i�(�y�;�y��))��I[A3,��f16]�������4�18.����H`�src:1202dn.tex�i�(�i�(�x;��1i�(�y�d�;�y��))�;�i�(�x;�i�(�y�;�y��)))����[17,��f15]�������4�19.����H`�src:1203dn.tex�i�(�i�(�x;��1i�(�y�d�;�x�))�;�i�(�x;�i�(�y�;�x�)))����[17,��f11]�������4�20.����H`�src:1204dn.tex�i�(�i�(�x;��1i�(�i�(�y�d�;�z�{I�)�;�u�))�;�i�(�i�(�y�;�v��)�;�i�(�x;�i�(�i�(�v�;�z�{I�)�;�u�))))�`_�[7,��f9]�������4�21.����H`�src:1205dn.tex�i�(�i�(�x;��1i�(�y�d�;�x�))�;�i�(�y�;�i�(�x;�x�)))����[13,��f18]�������4�22.����H`�src:1206dn.tex�i�(�i�(�x;��1x�)�;�i�(�i�(�y�d�;�x�)�;�i�(�y�;�x�)))���[21,��fA2]�������4�23.����H`�src:1207dn.tex�i�(�i�(�x;��1i�(�i�(�y�d�;�z�{I�)�;�u�))�;�i�(�i�(�y�;�y��)�;�i�(�x;�i�(�i�(�y�;�z�{I�)�;�u�))))�UM*[19,��f20]�������4�24.����H`�src:1208dn.tex�i�(�i�(�x;��1x�)�;�i�(�i�(�y�d�;�y��)�;�i�(�i�(�x;�y��)�;�i�(�x;�y��))))����[23,��f22]��0⍍��-�Theorem�2�6���n>M�src:1215dn.tex�The�jgsystem�A1{A4�admits�str��p�ong�double-ne�gation�elimination.����-�src:1218dn.texPr��p�o�of�.�SvW��ee�"�b�M�egin�b��!y�calculating�the�form�ulas�L*�for�this�system.�SvThe�only��
����-axiom��con��!taining�negations�is�A4,�6�but�there�are�three�p�M�ossible�replacemen�ts,����-so��fw��!e�get�three�new�axioms�A6{A8�as�follo�ws.������4����0⍍���V��A6���������i�(�i�(�x;��1y�d��)�;�i�(�n�(�y��)�;�n�(�x�)))�����������V�A7���������i�(�i�(�n�(�x�)�;��1y�d��)�;�i�(�n�(�y��)�;�x�))���������V�A8���������i�(�i�(�x;��1n�(�y�d��))�;�i�(�y�;�n�(�x�)))������-�src:1232dn.texBy���Theorem�3�it�suces�to�v��!erify�that�there�exist�double-negation-free�pro�M�ofs��
����-of�k@A6,��vA7,�A8,�and�D1{D3.�,jW��ee�ha��!v�e�already�v�eried�D1{D3�ab�M�o�v�e,��vso�it����-remains��fonly�to�pro�M�duce�double-negation-free�pro�ofs�of�A6{A8.����>�src:1238dn.texThe��ffollo��!wing�is�an�Otter�pro�M�of�of�A6.��0⍍���:b%1.����H`�src:1251dn.tex�i�(�x;��1i�(�y�d�;�i�(�z�{I;�y��)))�ڧ[A1,��fA1]��k
�����:b%2.����H`�src:1252dn.tex�i�(�i�(�i�(�i�(�x;��1y�d��)�;�i�(�z�{I;�y��))�;�u�)�;�i�(�i�(�z�{I;�x�)�;�u�))�}�*[A2,��fA2]�������:b%3.����H`�src:1253dn.tex�i�(�i�(�i�(�x;��1y�d��)�;�z�{I�)�;�i�(�i�(�n�(�y��)�;�n�(�x�))�;�z�{I�))��T�[A2,��fA4]��-�
J�ff�e֟
L͍����{���-=�4�����a�The��rname�A5�is�already�in�use�for�another�form��9ula,��loriginally�used�as�an�axiom�along���with�TA1{A4,�but�later�sho��9wn�to�b�A�e�pro�v��|rable�from�A1{A4.���"$����f�21����R��j֍����*�1F'���홊�����:b%�4.����H`�src:1254dn.tex�i�(�i�(�i�(�x;��1y�d��)�;�z�{I�)�;�i�(�y�;�z�{I�))���N[A2,��fA1]���������:b%5.����H`�src:1255dn.tex�i�(�n�(�x�)�;��1i�(�x;�y�d��))��>[4,��fA4]�������:b%6.����H`�src:1256dn.tex�i�(�i�(�x;��1i�(�y�d�;�z�{I�))�;�i�(�i�(�u;�y��)�;�i�(�x;�i�(�u;�z�{I�))))���[2,��f2]�������:b%7.����H`�src:1257dn.tex�i�(�i�(�x;��1y�d��)�;�i�(�i�(�i�(�x;�z�{I�)�;�u�)�;�i�(�i�(�y�;�z�{I�)�;�u�)))��ߑ[2,��fA2]�������:b%8.����H`�src:1258dn.tex�i�(�i�(�x;��1y�d��)�;�i�(�n�(�y��)�;�i�(�x;�z�{I�)))�ē[6,��f5]�������:b%9.����H`�src:1259dn.tex�i�(�i�(�i�(�n�(�x�)�;��1y�d��)�;�z�{I�)�;�i�(�i�(�i�(�x;�u�)�;�y��)�;�z�{I�))��k�[7,��f5]�������4�10.����H`�src:1260dn.tex�i�(�x;��1i�(�i�(�x;�y�d��)�;�y��))��.K[4,��fA3]�������4�11.����H`�src:1261dn.tex�i�(�i�(�i�(�x;��1i�(�y�d�;�x�))�;�z�{I�)�;�z��)�ˣ�[A3,��f1]�������4�12.����H`�src:1262dn.tex�i�(�i�(�x;��1i�(�y�d�;�z�{I�))�;�i�(�y�;�i�(�x;�z�{I�)))��u�[6,��f10]�������4�13.����H`�src:1263dn.tex�i�(�i�(�n�(�x�)�;��1n�(�i�(�y�d�;�i�(�z�{I;�y��))))�;�x�)���[3,��f11]�������4�14.����H`�src:1264dn.tex�i�(�i�(�i�(�x;��1y�d��)�;�z�{I�)�;�i�(�i�(�x;�u�)�;�i�(�i�(�u;�y��)�;�z�{I�)))���^[12,��f7]�������4�15.����H`�src:1265dn.tex�i�(�i�(�x;��1y�d��)�;�i�(�i�(�z�{I;�x�)�;�i�(�z�;�y�d��)))��?![12,��fA2]�������4�16.����H`�src:1266dn.tex�i�(�i�(�x;��1i�(�y�d�;�z�{I�))�;�i�(�i�(�z�;�u�)�;�i�(�x;�i�(�y�d�;�u�))))���^[6,��f15]�������4�17.����H`�src:1267dn.tex�i�(�i�(�i�(�x;��1y�d��)�;�z�{I�)�;�i�(�i�(�x;�u�)�;�i�(�n�(�u�)�;�z��)))��pQ[16,��f8]�������4�18.����H`�src:1268dn.tex�i�(�i�(�n�(�x�)�;��1y�d��)�;�i�(�n�(�y��)�;�x�))��C�[17,��f13]�������4�19.����H`�src:1269dn.tex�i�(�i�(�i�(�x;��1y�d��)�;�z�{I�)�;�i�(�n�(�z��)�;�x�))��I{[9,��f18]�������4�20.����H`�src:1270dn.tex�i�(�i�(�i�(�x;��1y�d��)�;�z�{I�)�;�i�(�i�(�z�;�u�)�;�i�(�n�(�u�)�;�x�)))����[14,��f19]�������4�21.����H`�src:1271dn.tex�i�(�i�(�x;��1y�d��)�;�i�(�n�(�y��)�;�n�(�x�)))��C�[20,��f13]���ҍ�>�src:1278dn.texThe��ffollo��!wing�is�an�Otter�pro�M�of�of�A7.���Ӎ����:b%1.����H`�src:1291dn.tex�i�(�x;��1i�(�y�d�;�i�(�z�{I;�y��)))�ڧ[A1,��fA1]�������:b%2.����H`�src:1292dn.tex�i�(�i�(�i�(�i�(�x;��1y�d��)�;�i�(�z�{I;�y��))�;�u�)�;�i�(�i�(�z�{I;�x�)�;�u�))�}�*[A2,��fA2]�������:b%3.����H`�src:1293dn.tex�i�(�i�(�i�(�x;��1y�d��)�;�z�{I�)�;�i�(�i�(�n�(�y��)�;�n�(�x�))�;�z�{I�))��T�[A2,��fA4]�������:b%4.����H`�src:1294dn.tex�i�(�i�(�i�(�x;��1y�d��)�;�z�{I�)�;�i�(�y�;�z�{I�))���N[A2,��fA1]�������:b%5.����H`�src:1295dn.tex�i�(�n�(�x�)�;��1i�(�x;�y�d��))��>[4,��fA4]�������:b%6.����H`�src:1296dn.tex�i�(�i�(�x;��1i�(�y�d�;�z�{I�))�;�i�(�i�(�u;�y��)�;�i�(�x;�i�(�u;�z�{I�))))���[2,��f2]�������:b%7.����H`�src:1297dn.tex�i�(�i�(�x;��1y�d��)�;�i�(�n�(�y��)�;�i�(�x;�z�{I�)))�ē[6,��f5]�������:b%8.����H`�src:1298dn.tex�i�(�x;��1i�(�i�(�x;�y�d��)�;�y��))��.K[4,��fA3]�������:b%9.����H`�src:1299dn.tex�i�(�i�(�i�(�x;��1i�(�y�d�;�x�))�;�z�{I�)�;�z��)�ˣ�[A3,��f1]�������4�10.����H`�src:1300dn.tex�i�(�i�(�x;��1i�(�y�d�;�z�{I�))�;�i�(�y�;�i�(�x;�z�{I�)))���"[6,��f8]�������4�11.����H`�src:1301dn.tex�i�(�i�(�n�(�x�)�;��1n�(�i�(�y�d�;�i�(�z�{I;�y��))))�;�x�)��|6[3,��f9]�������4�12.����H`�src:1302dn.tex�i�(�i�(�x;��1y�d��)�;�i�(�i�(�z�{I;�x�)�;�i�(�z�;�y�d��)))��?![10,��fA2]�������4�13.����H`�src:1303dn.tex�i�(�i�(�x;��1i�(�y�d�;�z�{I�))�;�i�(�i�(�z�;�u�)�;�i�(�x;�i�(�y�d�;�u�))))���^[6,��f12]�������4�14.����H`�src:1304dn.tex�i�(�i�(�i�(�x;��1y�d��)�;�z�{I�)�;�i�(�i�(�x;�u�)�;�i�(�n�(�u�)�;�z��)))��pQ[13,��f7]�������4�15.����H`�src:1305dn.tex�i�(�i�(�n�(�x�)�;��1y�d��)�;�i�(�n�(�y��)�;�x�))��C�[14,��f11]����>�src:1312dn.texThe��ffollo��!wing�is�an�Otter�pro�M�of�of�A8.���"$����f22����g��j֍����*�1F'���홊�����:b%�1.����H`�src:1325dn.tex�i�(�x;��1i�(�y�d�;�i�(�z�{I;�y��)))�ڧ[A1,��fA1]��T�����:b%2.����H`�src:1326dn.tex�i�(�i�(�i�(�i�(�x;��1y�d��)�;�i�(�z�{I;�y��))�;�u�)�;�i�(�i�(�z�{I;�x�)�;�u�))�}�*[A2,��fA2]�������:b%3.����H`�src:1327dn.tex�i�(�i�(�i�(�x;��1y�d��)�;�z�{I�)�;�i�(�i�(�n�(�y��)�;�n�(�x�))�;�z�{I�))��T�[A2,��fA4]�������:b%4.����H`�src:1328dn.tex�i�(�i�(�i�(�x;��1y�d��)�;�z�{I�)�;�i�(�y�;�z�{I�))���N[A2,��fA1]�������:b%5.����H`�src:1329dn.tex�i�(�n�(�x�)�;��1i�(�x;�y�d��))��>[4,��fA4]�������:b%6.����H`�src:1330dn.tex�i�(�i�(�x;��1i�(�y�d�;�z�{I�))�;�i�(�i�(�u;�y��)�;�i�(�x;�i�(�u;�z�{I�))))���[2,��f2]�������:b%7.����H`�src:1331dn.tex�i�(�i�(�x;��1y�d��)�;�i�(�i�(�i�(�x;�z�{I�)�;�u�)�;�i�(�i�(�y�;�z�{I�)�;�u�)))��ߑ[2,��fA2]�������:b%8.����H`�src:1332dn.tex�i�(�i�(�x;��1y�d��)�;�i�(�n�(�y��)�;�i�(�x;�z�{I�)))�ē[6,��f5]�������:b%9.����H`�src:1333dn.tex�i�(�i�(�i�(�n�(�x�)�;��1y�d��)�;�z�{I�)�;�i�(�i�(�i�(�x;�u�)�;�y��)�;�z�{I�))��k�[7,��f5]�������4�10.����H`�src:1334dn.tex�i�(�x;��1i�(�i�(�x;�y�d��)�;�y��))��.K[4,��fA3]�������4�11.����H`�src:1335dn.tex�i�(�i�(�i�(�x;��1i�(�y�d�;�x�))�;�z�{I�)�;�z��)�ˣ�[A3,��f1]�������4�12.����H`�src:1336dn.tex�i�(�i�(�x;��1i�(�y�d�;�z�{I�))�;�i�(�y�;�i�(�x;�z�{I�)))��u�[6,��f10]�������4�13.����H`�src:1337dn.tex�i�(�i�(�n�(�x�)�;��1n�(�i�(�y�d�;�i�(�z�{I;�y��))))�;�x�)���[3,��f11]�������4�14.����H`�src:1338dn.tex�i�(�i�(�i�(�x;��1y�d��)�;�z�{I�)�;�i�(�i�(�x;�u�)�;�i�(�i�(�u;�y��)�;�z�{I�)))���^[12,��f7]�������4�15.����H`�src:1339dn.tex�i�(�i�(�x;��1y�d��)�;�i�(�i�(�z�{I;�x�)�;�i�(�z�;�y�d��)))��?![12,��fA2]�������4�16.����H`�src:1340dn.tex�i�(�i�(�x;��1i�(�y�d�;�z�{I�))�;�i�(�i�(�z�;�u�)�;�i�(�x;�i�(�y�d�;�u�))))���^[6,��f15]�������4�17.����H`�src:1341dn.tex�i�(�i�(�x;��1y�d��)�;�i�(�i�(�z�{I;�u�)�;�i�(�i�(�y�;�z�{I�)�;�i�(�x;�u�))))���^[2,��f16]�������4�18.����H`�src:1342dn.tex�i�(�i�(�i�(�x;��1y�d��)�;�z�{I�)�;�i�(�i�(�x;�u�)�;�i�(�n�(�u�)�;�z��)))��pQ[16,��f8]�������4�19.����H`�src:1343dn.tex�i�(�i�(�n�(�x�)�;��1y�d��)�;�i�(�n�(�y��)�;�x�))��C�[18,��f13]�������4�20.����H`�src:1344dn.tex�i�(�i�(�i�(�x;��1y�d��)�;�z�{I�)�;�i�(�n�(�z��)�;�x�))��I{[9,��f19]�������4�21.����H`�src:1345dn.tex�i�(�i�(�i�(�x;��1y�d��)�;�z�{I�)�;�i�(�i�(�z�;�u�)�;�i�(�n�(�u�)�;�x�)))����[14,��f20]�������4�22.����H`�src:1346dn.tex�i�(�i�(�x;��1y�d��)�;�i�(�n�(�y��)�;�n�(�x�)))��C�[21,��f13]�������4�23.����H`�src:1347dn.tex�i�(�i�(�x;��1y�d��)�;�i�(�i�(�i�(�n�(�z�{I�)�;�n�(�u�))�;�x�)�;�i�(�i�(�u;�z��)�;�y�d��)))�d�[17,��f22]�������4�24.����H`�src:1348dn.tex�i�(�i�(�i�(�n�(�x�)�;��1n�(�y�d��))�;�i�(�n�(�z�{I�)�;�n�(�u�)))�;�i�(�i�(�y�;�x�)�;�i�(�u;�z�{I�)))�D
�[23,��fA4]�������4�25.����H`�src:1349dn.tex�i�(�i�(�x;��1n�(�y�d��))�;�i�(�y�;�n�(�x�)))��C�[24,��f19]��)0��-�src:1357dn.texThis��fcompletes�the�pro�M�of�of�the�theorem.�� H򍍑-�8��A@An��In��triguing�Example��Ѝ�-�src:1360dn.tex�One�ہof�the�motiv��dDations�for�this�w��!ork�w�as�the�existence�of�a�form�ula�that�is��
����-double-negation�hrfree�and�pro��!v��dDable�from�A1{A4�but�for�whic�h�W��eos�had�b�M�een����-unable��fto�nd�a�double-negation-free�pro�M�of.���The�form��!ula�in�question�is��
配�>�src:1367dn.tex���Z����WG�DN1������	q�i�(�i�(�n�(�x�)�;��1n�(�i�(�i�(�n�(�y�d��)�;�n�(�z�{I�))�;�n�(�z��))))�;����������wUn�(�i�(�i�(�n�(�i�(�n�(�x�)�;��1y�d��))�;�n�(�i�(�n�(�x�)�;�z�{I�)))�;�n�(�i�(�n�(�x�)�;�z��)))))�:�����"$����f�23����.~��j֍����*�1F'���홊��-�src:1373dn.tex�W��eos�w�pro��!vided�a�pro�M�of�of�45�condensed-detac�hmen�t�steps�of�this�theorem,��016��
����-of��whose�lines�in��!v�olv�ed��a�double�negation.��EBeeson�used�this�pro�M�of�as�input����-to�6/a�computer�program�implemen��!ting�the�algorithms�implicit�in�the�pro�M�of�of����-our��kmain�theorem.���The�output�of�this�program�w��!as�a�double-negation-free����-pro�M�of���b��!y�mo�dus�p�onens�of�the�example,��}from�substitution�instances�of�A1{����-A4.��0The�J`pro�M�of��'s�length�and�size�w��!ere�surprising.�It�w��!as�796�lines,�\�and�man�y����-of���its�lines�in��!v�olv�ed���thousands�of�sym��!b�M�ols.�&jThe�input�pro�of�tak��!es�ab�out�3.5����-kilob��!ytes,�	�the�‡output�pro�M�of�ab�out�200�kilob��!ytes.�2@No�w�‡w�e�kno�w�what�the����-�c��p�ondense�d��@�means�in�\condensed�detac��!hmen�t"!�[kThe��@expansion�in�size�is�due����-to�Umaking�the�substitutions�in��!tro�M�duced�b�y�condensed�detac�hmen�t�explicit.����-The���expansion�in�length�is�due�to�duplications�of�m��!ultiply�referenced�lines,����-whic��!h��m�ust�b�M�e�done�b�efore�the�substitutions�are�\pushed�up��!w�ard"�in�the����-pro�M�of.��{In��?other�w��!ords,���one�line�of�the�pro�of�can�b�e�referenced�sev��!eral�times,����-and�	iwhen�the�pro�M�of�is�con��!v�erted�	ito�tree�form,�"*eac��!h�reference�will�require�a����-separate��cop��!y�of�the�referenced�line.��This�796-line�pro�M�of,��considered�as�a����-tree,��Uhas���substitution�instances�of�the�axioms�at�the�lea��!v�es.��AAfter���obtaining����-this��9pro�M�of,�,-w��!e�could�ha�v�e�con�tin�ued�with�the�algorithm,�,-pro�viding�pro�M�ofs����-of�j�the�substitution�instances�of�the�axioms.�
+�That�approac��!h�w�ould�ha�v�e����-substan��!tially��increased�the�length.���Instead,�9�McCune�put�the�lines�of�the����-796-line�|pro�M�of�in��!to�an�Otter�input�le�as�\hin�ts"�[�16��
�4],���and�Otter�pro�M�duced�a����-27-line��tdouble-negation-free�condensed-detac��!hmen�t��tpro�M�of�of�DN1�from�A1{����-A4���and�A6{A8.�ܻThis�run�generates�some�6,000�form��!ulas�and�tak�es�b�M�et�w�een����-one-half��and�t��!w�o��hours,��!dep�M�ending�on�what�mac��!hine�is�used.�W�If�the�lines�of����-this���pro�M�of,�Ktogether�with�the�pro�ofs�of�A6{A8,�Kare�supplied�as�resonators�[�17��
�4]����-in�F�a�new�input�le,�Y�Otter�can�then�nd�a�37-step�pro�M�of�of�DN1�from�A1{A4.��$����-�9��A@D-Completeness��of�In��tuitionistic�Logic��q���-�src:1420dn.tex�Let�\�H�\�b�M�e�the�follo��!wing�form�ulation�of�in�tuitionistic�prop�M�ositional�calculus�in����-terms��fof�implication�and�negation,�denoted�b��!y��i��and��n�.������5����T�������ɹH1�����ӆ+�i�(�x;��1i�(�y�d�;�x�))�������������H2���������i�(�i�(�x;��1i�(�y�d�;�z�{I�))�;�i�(�i�(�x;�y��)�;�i�(�x;�z�{I�)))�����������H3������/��i�(�i�(�x;��1n�(�x�))�;�n�(�x�))�����������H4��������i�(�x;��1i�(�n�(�x�)�;�y�d��))����-�	a��ff�e֟
L͍����{���-=�5�����a�These�iaxioms�can�b�A�e�found�in�App�endix�I�h�of�[�11��	?�],�~as�the��ː ���L��	0�uk��|rasiewicz�2-basis�in�12.1���plus�o�the�t��9w�o�o�axioms�lab�A�eled�(4)�of�3.2,��pas�sp�ecied�in�12.5.�+�According�to�[�11��	?�],��pif�w��9e�also���add��Z(2)�and�(3)�of�3.2,��w��9e�get�the�full�in�tuitionistic�prop�A�ositional�calculus;�0�but�(2)�and���(3)�ٌof�3.2�concern�disjunction�and�conjunction.��If�they�are�omitted,��the�four�axioms�listed���form�Ta�4-basis�for�the�implication-negation�fragmen��9t.�pThis�will�b�A�e�pro�v�ed�in�Corollary�2.���"$����f�24����<���j֍����*�1F'���홊��-�src:1435dn.tex�The�^�inference�rules�of�H�^4are�mo�M�dus�p�onens�and�substitution.�
�It�is�also��
����-p�M�ossible�S�to�consider�H1{H4�with�condensed�detac��!hmen�t.��KThese�S�t�w�o�systems����-ha��!v�e��fthe�same�theorems,�as�will�b�M�e�sho��!wn�in�detail�b�elo��!w.����>�src:1440dn.texW��ee�Z�note�that�H�Z�do�M�es�not�satisfy�strong�double�negation�elimination.����-Substituting����n�(�y�d��)�for��x��in�axiom�H3�pro�M�duces��i�(�i�(�n�(�y��)�;��1n�(�n�(�y��)))�;�n�(�n�(�y��))).����-Cancelling�xythe�double�negations�pro�M�duces��i�(�i�(�n�(�y�d��)�;��1y��)�;�y��),���whic��!h�xyis�not�pro�v-����-able�q1in�in��!tuitionistic�logic.�>>This�same�example�demonstrates�directly�that����-H���do�M�es���not�satisfy�the�h��!yp�othesis�of�Theorem�3.��Nev��!ertheless,��and�p�erhaps����-surprising,���H��do�M�es��satisfy�double�negation�elimination|but�w��!e�shall�need�a����-dieren��!t��fpro�M�of�to�sho�w�that.���9����-�Lemma�2�7���el.�src:1449dn.tex�D1{D3�WIhave�double-ne��p�gation-fr�e�e�WIc�ondense�d-detachment�pr�o�ofs����-fr��p�om���H1{H4.����-�src:1454dn.texPr��p�o�of�R�:��
The���follo��!wing�is�a�double-negation-free�condensed-detac�hmen�t�pro�M�of����-of��fD1�from�H1{H4.���8�����:b%1.����H`�src:1467dn.tex�i�(�i�(�x;��1y�d��)�;�i�(�x;�x�))��x�[H2,��fH1]��Z�����:b%2.����H`�src:1468dn.tex�i�(�x;��1x�)��[1,��fH1]����>�src:1476dn.texD2�P]is��i�(�i�(�x;��1x�)�;�i�(�n�(�x�)�;�n�(�x�))).�	��The�P]follo��!wing�is�a�double-negation-free����-condensed-detac��!hmen�t��Zpro�M�of�of�D2�from�H1{H4,��found�b��!y�using�a�sp�ecially����-compiled��fv��!ersion�of�Otter.���Curiously��e,�H3�is�not�used.���8�����:b%1.����H`�src:1492dn.tex�i�(�x;��1i�(�y�d�;�i�(�z�{I;�y��)))�ڧ[H1,��fH1]��Z�����:b%2.����H`�src:1493dn.tex�i�(�x;��1i�(�y�d�;�i�(�z�{I;�i�(�u;�z��))))�ˠ�[H1,��f1]�������:b%3.����H`�src:1494dn.tex�i�(�i�(�i�(�x;��1i�(�y�d�;�z�{I�))�;�i�(�x;�y��))�;�i�(�i�(�x;�i�(�y�;�z�{I�))�;�i�(�x;�z��)))�P�[H2,��fH2]�������:b%4.����H`�src:1495dn.tex�i�(�i�(�x;��1n�(�x�))�;�i�(�x;�y�d��))��a�[H2,��fH4]�������:b%5.����H`�src:1496dn.tex�i�(�i�(�x;��1y�d��)�;�i�(�x;�x�))��x�[H2,��fH1]�������:b%6.����H`�src:1497dn.tex�i�(�x;��1i�(�i�(�y�d�;�z�{I�)�;�i�(�y�;�y��)))���[H1,��f5]�������:b%7.����H`�src:1498dn.tex�i�(�x;��1i�(�i�(�y�d�;�n�(�y��))�;�i�(�y�;�z�{I�)))����[H1,��f4]�������:b%8.����H`�src:1499dn.tex�i�(�i�(�x;��1i�(�y�d�;�z�{I�))�;�i�(�x;�i�(�y�;�y��)))����[H2,��f6]�������:b%9.����H`�src:1500dn.tex�i�(�i�(�x;��1i�(�x;�y�d��))�;�i�(�x;�y��))���[3,��f5]�������4�10.����H`�src:1501dn.tex�i�(�i�(�x;��1i�(�i�(�y�d�;�x�)�;�z�{I�))�;�i�(�x;�z��))��o�[3,��f1]�������4�11.����H`�src:1502dn.tex�i�(�i�(�x;��1i�(�x;�x�))�;�i�(�x;�x�))��^[3,��f9]�������4�12.����H`�src:1503dn.tex�i�(�i�(�x;��1i�(�y�d�;�x�))�;�i�(�x;�i�(�y�;�x�)))��$[11,��f2]�������4�13.����H`�src:1504dn.tex�i�(�n�(�x�)�;��1i�(�x;�y�d��))��[10,��f7]�������4�14.����H`�src:1505dn.tex�i�(�n�(�x�)�;��1i�(�x;�x�))��?�[8,��f13]�������4�15.����H`�src:1506dn.tex�i�(�i�(�x;��1y�d��)�;�i�(�y�;�i�(�x;�y��)))��K[12,��f1]�������4�16.����H`�src:1507dn.tex�i�(�i�(�x;��1x�)�;�i�(�n�(�x�)�;�i�(�x;�x�)))���P[15,��f14]���"$����f25����J��j֍����*�1F'���홊�����4苹17.����H`�src:1508dn.tex�i�(�i�(�x;��1x�)�;�i�(�n�(�x�)�;�n�(�x�)))�ſ
[8,��f16]���@��>�src:1515dn.texThe�r�follo��!wing�is�a�double-negation-free�condensed-detac�hmen�t�pro�M�of�of��
����-D3��ffrom�H1{H4.���H3�and�H4�are�not�used.���A�����:b%1.����H`�src:1528dn.tex�i�(�x;��1i�(�y�d�;�i�(�z�{I;�y��)))�ڧ[H1,��fH1]��􍍍��:b%2.����H`�src:1529dn.tex�i�(�i�(�i�(�x;��1i�(�y�d�;�z�{I�))�;�i�(�x;�y��))�;�i�(�i�(�x;�i�(�y�;�z�{I�))�;�i�(�x;�z��)))�P�[H2,��fH2]�������:b%3.����H`�src:1530dn.tex�i�(�x;��1i�(�i�(�y�d�;�i�(�z�{I;�u�))�;�i�(�i�(�y�;�z�{I�)�;�i�(�y�;�u�))))�~(W[H1,��fH2]�������:b%4.����H`�src:1531dn.tex�i�(�i�(�x;��1y�d��)�;�i�(�x;�x�))��x�[H2,��fH1]�������:b%5.����H`�src:1532dn.tex�i�(�i�(�x;��1i�(�x;�y�d��))�;�i�(�x;�y��))���[2,��f4]�������:b%6.����H`�src:1533dn.tex�i�(�i�(�x;��1i�(�i�(�y�d�;�x�)�;�z�{I�))�;�i�(�x;�z��))��o�[2,��f1]�������:b%7.����H`�src:1534dn.tex�i�(�x;��1i�(�i�(�y�d�;�i�(�i�(�z�{I;�y��)�;�u�))�;�i�(�y�;�u�)))���[H1,��f6]�������:b%8.����H`�src:1535dn.tex�i�(�i�(�x;��1y�d��)�;�i�(�i�(�z�{I;�x�)�;�i�(�z�;�y�d��)))���"[6,��f3]�������:b%9.����H`�src:1536dn.tex�i�(�i�(�x;��1x�)�;�i�(�x;�x�))��fX[5,��f8]�������4�10.����H`�src:1537dn.tex�i�(�i�(�i�(�x;��1y�d��)�;�i�(�z�{I;�x�))�;�i�(�i�(�x;�y��)�;�i�(�z�{I;�y��)))��c�[H2,��f8]�������4�11.����H`�src:1538dn.tex�i�(�i�(�x;��1i�(�y�d�;�y��))�;�i�(�x;�i�(�y�;�y��)))���[8,��f9]�������4�12.����H`�src:1539dn.tex�i�(�i�(�x;��1i�(�y�d�;�x�))�;�i�(�x;�i�(�y�;�x�)))����[9,��f1]�������4�13.����H`�src:1540dn.tex�i�(�i�(�x;��1x�)�;�i�(�i�(�y�d�;�x�)�;�i�(�y�;�x�)))��$[11,��f8]�������4�14.����H`�src:1541dn.tex�i�(�i�(�i�(�x;��1y�d��)�;�z�{I�)�;�i�(�y�;�z�{I�))��Y[6,��f7]�������4�15.����H`�src:1542dn.tex�i�(�i�(�x;��1y�d��)�;�i�(�i�(�y�;�z�{I�)�;�i�(�x;�z��)))����[14,��f10]�������4�16.����H`�src:1543dn.tex�i�(�i�(�i�(�i�(�x;��1y�d��)�;�i�(�z�{I;�y��))�;�u�)�;�i�(�i�(�z�{I;�x�)�;�u�))��"�[15,��f15]�������4�17.����H`�src:1544dn.tex�i�(�i�(�x;��1i�(�y�d�;�z�{I�))�;�i�(�i�(�u;�y��)�;�i�(�x;�i�(�u;�z�{I�))))��"�[16,��f16]�������4�18.����H`�src:1545dn.tex�i�(�i�(�x;��1y�d��)�;�i�(�i�(�i�(�x;�z�{I�)�;�u�)�;�i�(�i�(�y�;�z�{I�)�;�u�)))��"�[16,��f15]�������4�19.����H`�src:1546dn.tex�i�(�i�(�x;��1i�(�i�(�y�d�;�z�{I�)�;�u�))�;�i�(�i�(�y�;�v��)�;�i�(�x;�i�(�i�(�v�;�z�{I�)�;�u�))))�UlR[17,��f18]�������4�20.����H`�src:1547dn.tex�i�(�i�(�x;��1i�(�i�(�y�d�;�z�{I�)�;�u�))�;�i�(�i�(�y�;�y��)�;�i�(�x;�i�(�i�(�y�;�z�{I�)�;�u�))))�UM*[12,��f19]�������4�21.����H`�src:1548dn.tex�i�(�i�(�x;��1x�)�;�i�(�i�(�y�d�;�y��)�;�i�(�i�(�x;�y��)�;�i�(�x;�y��))))����[20,��f13]������-�Theorem�2�7���n>M�src:1555dn.tex�The�Gsame�the��p�or�ems�Gar�e�pr�ovable�fr�om�H1{H4�by�using�c�on-��
����-dense��p�d�^�detachment�as�the�sole�rule�of�infer�enc�e�as�when�we�use�mo�dus�p�o-����-nens��and�substitution�as�rules�of�infer��p�enc�e.���Mor�e�over,�5if���b��is�pr��p�ovable�without����-double�/�ne��p�gation�by�mo�dus�p�onens�fr�om�substitution�instanc�es�of�axioms,�UEthen����-ther��p�e���is�a�double-ne�gation-fr�e�e�c�ondense�d-detachment�pr�o�of�of��b�.����-�src:1566dn.texR��p�emark�.�tDThe�-�presen��!t�pro�M�of�giv�es�no�assurance�that�a�general�H�-�pro�M�of,�O�us-����-ing���substitution�arbitrarily�and�not�just�in�axioms,���can�b�M�e�con��!v�erted���to�a����-condensed-detac��!hmen�t�1�pro�M�of�without�in��!tro�ducing�additional�double�nega-����-tions.�UJThat�x�stronger�claim�is�also�true:���it�is�a�consequence�of�Theorem�8����-b�M�elo��!w.���"$����f26����X��j֍����*�1F'���홊��-�src:1574dn.tex�Pr��p�o�of�R�:�گThe���rst�claim�is�an�immediate�consequence�of�Theorem�2�and��
����-Lemma��7.�	GT��eo�pro��!v�e��the�second�claim,�f|supp�M�ose��b��has�a�double-negation-����-free���mo�M�dus�p�onens�pro�of�from�substitution�instances�of�axioms.�ضBy�Lemma����-3,�c�w��!e�=�can�supply�double-negation-free�condensed-detac�hmen�t�pro�M�ofs�of�the����-substitution�dsinstances�of�axioms�that�are�used�in�the�pro�M�of.���Adjoining�these����-pro�M�ofs,�#nw��!e��obtain�a�double-negation-free�condensed-detac�hmen�t�pro�M�of�of��b��as����-required.��e����-�10��HH��and�Sequen��t�Calculus��q���-�src:1588dn.tex�Let��-G1�b�M�e�the�in��!tuitionistic�Gen�tzen�calculus�as�giv�en�b�y�Kleene�[�5��y�].��2Let�G����-b�M�e��lG1�(min��!us�cut),��8restricted�to�implication�and�negation;���that�is,�form��!ulas����-con��!taining��{other�connectiv�es�are�not�allo�w�ed.��Th�us�the�rules�of�inference�of����-G���are���the�four�rules�in��!v�olving���implication�and�negation,��plus�the�structural����-rules.�|�The��drules�of�G1�are�listed�on�pp.�442{443�of�[�5��y�].�They�will�also�b�M�e�giv��!en����-in��,the�course�of�the�pro�M�of�of�Lemma�12.��0W��ee�shall�use�the�notation�����)������-for�1�a�sequen��!t.�dW��ee�remind�the�reader�that�what�distinguishes�in�tuitionistic����-from�=�classical�sequen��!t�calculus�is�that�the�consequen�t��in�a�sequen�t��
��)������-in��fthe�in��!tuitionistic�calculus�is�restricted�to�con�tain�at�most�one�form�ula.������6������>�src:1619dn.tex�W��ee��giv��!e�a�translation�of�H��in�to�G:�If��A��is�a�form�ula�of�H,�then��A�����0���Թis�a����-form��!ula��fof�G,�obtained�b�y�the�follo�wing�rules.��������~7�i�(�a;��1b�)���z��0��ʫ�=�
��a���z��0���!��b���z��0������������q	�n�(�a�)���z��0��ʫ�=�
��:�a���z��0�������-�src:1626dn.tex�Of��course,��when��a��is�a�prop�M�osition�letter�(v��dDariable),�then��a�����0��	��is�just��a�.�9�If��
����-�
�=��A���z�0����;���1:�:�:��l�;��1A���z�n��	N��is��fa�list�of�form��!ulas�of�L,�then������0��fj�is�the�list��A������0���g�0����;���1:�:�:��l�;��1A������0���A��n����P�.����>�src:1632dn.texW��ee��translate�G��in��!to�H�in�the�follo��!wing�manner.�=First�w�e�assign�to�eac�h����-form��!ula��f�A��of�G�a�corresp�M�onding�form�ula��A�����0��t��of�H,�giv�en�b�y���������B(�A�
��!��B����)���z��0����=��i�(�A���z��0���9�;��1B����z��0��Zݹ)����������R(�:�A�)���z��0����=�
��n�(�A���z��0���9�)�;������-�src:1639dn.tex�where�}�again��A�����0��?��=�qK�A��for�prop�M�osition�letters��A�.�cmW��ee�need�to�dene������0��KϹalso,��
����-where��1�is�a�list�of�form��!ulas;�<�since�w�e�are�treating�only�the�in�tuitionistic��-�	N��ff�e֟
L͍����{���-=�6�����a�The���translations�giv��9en�here�can�also�b�A�e�giv�en�for��! ���L���suk��|rasiewicz's�logic�L1{L3,���but�man�y���additional�$complications�are�in��9tro�A�duced�b�y�the�necessit�y�of�translating�a�sequen�t�con�taining���more��than�one�form��9ula�on�the�righ�t,��and�in�view�of�the�simpler�pro�A�ofs�of�double-negation���elimination�LMgiv��9en�ab�A�o�v�e,�t�w�e�treat�the�Gen�tzen�translation�only�for�in�tuitionistic�logic.��nNote���that���w��9e�used�Otter�only�for�the�H-pro�A�ofs�of�D2�and�D3;��ebut�if�w�e�treat�L���this�w�a�y�instead���of�TH,�w��9e�need�Otter�for�t�w�en�t�y-one�additional�lemmas.���"$����f�27����fj��j֍����*�1F'���홊��-�calculus,���w��!e��\need�this�denition�only�for�lists�o�M�ccurring�on�the�left�of��)�.��If��
����-�,�=��A���z�1����;���1:�:�:��l�;��1A���z�n��	�ʹis�Tza�list�of�form��!ulas�o�M�ccurring�on�the�left�of��)�,��then������0��"��is����-�A������0���g��1�����;���1:�:�:��l�;��1A������0���A��n����P�.������7�������>�src:1662dn.tex�These��ft��!w�o�translations�are�in�v�erse.��������-�Lemma�2�8���el.�src:1666dn.tex�L��p�et����A��b�e�a�formula�of�H.�Then���A�����0�����U��	��0���5�=�
��A�.��#���-�src:1669dn.texPr��p�o�of�.��	By�L�induction�on�the�complexit��!y�of��A�.�If��A��is�a�v��dDariable,�^�then��A�����0��ʫ�=�
��A����-�and���f�A�����0�����џ�	��0��u��=�
��A�.���W��ee��fha��!v�e����������A�i�(�x;��1y�d��)���z��0������-��}m�0������֫f�=������/�(�x���z��0��ʫ�!�
��y��d���z��0��$��)���z��0������'����֫f�=������/��i�(��x���z��0������}m�0��
�X�;���1y��d���z��0����V#��}m�0��$\�)����������֫f=������/��i�(�x;��1y�d��)�;������-�src:1677dn.tex�and��fw��!e�ha�v�e���������k�n�(�x�)���z��0������d��}m�0������؞��=������"�(�:�(�x���z��0����))���z��0���������؞��=������"��n�(��x���z��0������}m�0��
�X�)����������؞�=������"��n�(�x�)�:��������>�src:1685dn.tex�Henceforth��w��!e�simplify�our�notation�b�y�using�lo�w�er-case�letters�for�for-��
����-m��!ulas�>of�H�=�and�upp�M�er-case�letters�for�form�ulas�of�G.�Then�w�e�can�write��a����-�instead��Oof��A�����0��Y��and��A��instead�of��a�����0����.���By�the�preceding�lemma,���this�con��!v�enien�t����-notation�	mpresen��!ts�no�am�biguit�y��e.�	�Th�us,�b.for�example,�(�A�ZZ�!��B����)�����0��צ�is�	m�i�(�a;��1b�).����-Greek��fletters�are�used�for�lists�of�form��!ulas.�����>�src:1695dn.texThe�S�follo��!wing�lemma�giv�es�sev�eral�v��dDariations�of�the�deduction�theorem����-for��fH.��������-�Lemma�2�9�(Deduction�theorem�for�H)����N��src:1698dn.tex�(i)�^6If�H�^pr��p�oves��a��fr�om�assump-����-tions�v���[;��1b�,��cthen��i�(�b;�a�)��is�a�the��p�or�em�vpr�ove�d�in�H�u�fr�om�assumptions���j�,��cpr�ovide�d����-the���assumptions�c��p�ontain�only�c�onstant�pr�op�osition�letters.����>�src:1704dn.tex(ii)���If��a��is�pr��p�ovable�fr�om�assumptions����[;��1b��by�c�ondense�d�detachment�fr�om����-H1{H4,���then�j��i�(�b;��1a�)��is�derivable�by�c��p�ondense�d�j�detachment�fr��p�om���j�,�pr��p�ovide�d����-the���assumptions�c��p�ontain�only�c�onstant�pr�op�osition�letters.��-��h�ff�e֟
L͍����{���-=�7�����a�A� �similar�!#translation�has�b�A�een�giv��9en�in�[�9����]�in�connection�with���� ���L����uk��|rasiewicz's�m�ultiv��|ralued���logics���(whic��9h�include�the�innite-v��|ralued�logic�discussed�in�Section�7�of�this�pap�A�er).�d�It�is���the��ob��9vious�translation�of�Gen�tzen�calculus�in�to�the�implication-and-negation�fragmen�t�of���prop�A�ositional�ndcalculus.���W��:�e�cannot�app�eal�to�an��9y�of�the�results�of�[�9����]�b�ecause�w��9e�are�dealing���with�Tdieren��9t�logics,�and�b�A�esides�w�e�need�to�pa�y�atten�tion�to�double�negations.���"$����f�28����t7��j֍����*�1F'���홊��>�src:1710dn.tex�(iii)�	�If�ther��p�e�exists�a�pr�o�of�of��a��by�mo�dus�p�onens�fr�om����[;��1b��and�substitution��
����-instanc��p�es�4�of�H1{H4,�Y0then�ther�e�exists�a�pr�o�of�of��i�(�a;��1b�)��by�mo�dus�p�onens�fr�om����-��U�and���substitution�instanc��p�es�of�H1{H4.����>�src:1716dn.tex(iv)���In�p��p�art�(i),��Vif�the�given�pr�o�of�of��a��has�no�double�ne�gations,��Vthen�the����-pr��p�o�of���of��i�(�b;��1a�)��fr��p�om���U�has�no�double�ne�gations.����>�src:1721dn.tex(v)��In�p��p�art�(iii),��<if�the�given�pr�o�of�of��a��has�no�double�ne�gations,��<then�the����-pr��p�o�of���of��i�(�b;��1a�)��fr��p�om���U�has�no�double�ne�gations.��L͍�-�src:1728dn.texR��p�emarks���:���W��ee��do�not�pro��!v�e��a�claim�ab�M�out�double�negations�for�condensed-����-detac��!hmen�t�pro�M�ofs,�4~only�for�mo�dus�p�onens�pro�ofs.�2�That�is,�4~for�condensed-����-detac��!hmen�t��Wpro�M�ofs,�8�there�is�no�part�(vi)�analogous�to�parts�(iv)�and�(v).����-The�ܘreason�for�the�restriction�to�constan��!t�assumptions�in�(i)�and�(ii)�is�the����-follo��!wing.��XF��erom���the�assumption��i�(�n�(�n�(�x�))�;��1x�),�˃w�e�can�deriv�e�an�y�theorem����-of�-�classical�logic,���for�instance��i�(�n�(�n�(�a�))�;��1a�),�b��!y�substitution�or�condensed����-detac��!hmen�t.��But���w�e�cannot�deriv�e�the�prop�M�osition�that�the�rst�of�these����-implies���the�second,��A�i�(�i�(�n�(�n�(�x�))�;��1x�)�;�i�(�n�(�n�(�a�))�;�a�)).���Therefore���the�deduction����-theorem���is�false�without�the�restriction.���Pro�M�ofs�b��!y�mo�dus�p�onens�from�sub-����-stitution��yinstances�of�axioms�do�not�suer�from�this�dicult��!y��e,��whic�h��yis�one����-reason��fthey�are�so�tec��!hnically�useful�in�this�pap�M�er.�����-�src:1749dn.tex�Pr��p�o�of�.���First��w��!e�sho�w�that�(ii)�follo�ws�from�(iii).���If�w�e�are�giv�en�a�condensed-����-detac��!hmen�t��pro�M�of�of��a��from�assumptions����[;��1b��using�H1{H4,�@w��!e�can�nd,�b��!y����-Theorem��g1,��'a�mo�M�dus�p�onens�pro�of�of��a��from����[;��1b��and�substitution�instances����-of��1H1{H4.��>Applying�(iii),�Pcw��!e�ha�v�e�a�mo�M�dus�p�onens�pro�of�of��i�(�b;��1a�)�from����-���ҹand���substitution�instances�of�H1{H4.���By�Theorem�7,���this�pro�M�of�can�b�e����-con��!v�erted�?�to�a�condensed-detac��!hmen�t�?�pro�M�of�of��i�(�b;��1a�)�from���j�,�Tmcompleting�the����-deriv��dDation��fof�(ii)�from�(iii).����>�src:1764dn.texNext�(zw��!e�sho�w�that�(i)�follo�ws�from�(iii).�dSupp�M�ose�w�e�are�giv�en�a�pro�M�of����-of����a��from����[;��1b��in�H.�By�Theorem�1,��w��!e�can�nd�a�mo�M�dus�p�onens�pro�of�of��a����-�from�S�assumptions����[;��1b��and�substitution�instances�of�H1{H4.��SBy�(iii)�w��!e�then����-can��nd�a�mo�M�dus�p�onens�pro�of�of��i�(�b;��1a�)�from������and�substitution�instances�of����-H1{H4.���Adding��one�substitution�step�ab�M�o��!v�e��eac�h�suc�h�substitution�instance,����-w��!e�Z3ha�v�e�a�pro�M�of�in��H�=�of��i�(�b;��1a�)�from���j�.��DThat�completes�the�pro�of�that�(i)����-follo��!ws��ffrom�(iii).����>�src:1777dn.texW��ee�͎no��!w�sho�w�that�(v)�implies�(iv).�SVT��eo�do�so�requires�going�o�v�er�the����-preceding���paragraph�with�atten��!tion�to�double�negations.�Supp�M�ose�w�e�are����-giv��!en���a�double-negation-free�pro�M�of�of��a��from����[;��1b��in�H.�By�Theorem�1,��]w�e����-can��pnd�a�mo�M�dus�p�onens�pro�of�of��a��from�assumptions����[;��1b��and�substitution����-instances�L�of�H1{H4,��^whic��!h�is�also�double-negation-free.�	�By�(v)�w�e�then����-can���nd�a�double-negation-free�mo�M�dus�p�onens�pro�of�of��i�(�b;��1a�)�from���2
�and����-substitution��instances�of�H1{H4.�;)Adding�one�substitution�step�ab�M�o��!v�e��eac�h���"$����f29��������j֍����*�1F'���홊��-�suc��!h�substitution�instance,�n0w�e�ha�v�e�a�double-negation-free�pro�M�of�in��H��ɹof��
����-�i�(�b;��1a�)��ffrom���j�.���That�completes�the�pro�M�of�that�(iv)�follo��!ws�from�(v).��
���>�src:1796dn.texNo��!w�	w�e�pro�v�e�(iii)�and�(v)�sim�ultaneously�b�y�induction�on�the�n�um�b�M�er����-of�3steps�in�a�pure�mo�M�dus�p�onens�pro�of�of��a��from���rJ�and�substitution�instances����-of��fH1{H4.����>�src:1801dn.texBase��case:�1�a��is��b�,��or�a�mem��!b�M�er�of���j�,�or�a�substitution�instance�of�one�of����-H1{H4.����>�src:1804dn.texIf���a��is�a�substitution�instance�of�an�axiom�of�H1{H4,��*then�b��!y�Lemma�3����-there�G�exists�a�condensed-detac��!hmen�t�G�pro�M�of�of��a��from�H1{H4�that�con��!tains����-only��fdouble�negations�already�o�M�ccurring�in��a�.����>�src:1810dn.texIf����a��is��b�,��then�w��!e�use�the�fact�that��i�(�b;��1b�)�is�a�theorem�of�H,�pro�v��dDable����-without���double�negations�(except�those�o�M�ccurring�in��b�)�b��!y�Lemma�1.��Hence����-b��!y�STheorem�1,� �it�is�pro�v��dDable�b�y�mo�M�dus�p�onens�from�substitution�instances����-of��fH1{H4.����>�src:1817dn.texIf���a��is�a�mem��!b�M�er�of���j�,��then�consider�the�form�ula��i�(�a;��1i�(�b;�a�)),��whic�h��is�a����-substitution��.instance�of�axiom�H1.�4W��ee�can�deduce��i�(�b;��1a�)�b��!y�mo�M�dus�p�onens����-from���this�form��!ula�and��a�;��adjoining�this�step�to�a�one-step�pro�M�of�of��a��from������-�\b��!y��fassumption",�w�e�ha�v�e�a�pro�M�of�of��i�(�b;��1a�)�from���j�.����>�src:1825dn.texT��eurning�#�to�the�induction�step,�Cbsupp�M�ose�the�last�step�in�the�giv��!en�pro�of����-infers���a��from��i�(�p;��1a�)�and��p�.���By�the�induction�h��!yp�M�othesis,�Xw�e��ha�v�e�pro�M�ofs�of����-�i�(�b;��1p�)��and��i�(�b;�i�(�p;�a�))�from���j�.���By�axiom�H2�and�mo�M�dus�p�onens�(whic��!h�is�a����-sp�M�ecial��5case�of�condensed�detac��!hmen�t)��5w�e�ha�v�e��i�(�i�(�b;��1p�)�;�i�(�b;�a�)).��KApplying����-mo�M�dus�J&p�onens�once�more,�sw��!e�ha�v�e��i�(�b;��1a�)�as�desired.��Note�that�no�double����-negations��fare�in��!tro�M�duced.���That�completes�the�pro�of�of�the�lemma.����>�src:1837dn.texW��ee���shall�call�a�sequen��!t��F��)�����double-negation-free�if�it�con�tains�no����-double��fnegation.����>�src:1841dn.texW��ee���shall�refer�to�pro�M�ofs�b��!y�mo�dus�p�onens�from�substitution�instances����-of���H1{H4�as�M-pro�M�ofs�for�short.��M-pro�ofs�use�mo�dus�p�onens�only�but�can����-use�xysubstitution�instances�of�axioms,���as�opp�M�osed�to�H-pro�ofs,���whic��!h�can�use����-substitution��an��!ywhere�as�w�ell�as�mo�M�dus�p�onens.�N�W��ee�ha��!v�e��already�sho��!wn����-ho��!w�Iyto�con�v�ert�condensed-detac�hmen�t�pro�M�ofs�to�M-pro�ofs�(in�Theorem�1),����-and��dvice�v��!ersa�(since�ev�ery�substitution�instance�of�the�axioms�is�deriv��dDable����-b��!y��fcondensed�detac�hmen�t).��D�����-�Lemma�2�10���k��src:1856dn.tex�If���the�nal�se��p�quent���
��)������of�a�G-pr�o�of�is�double-ne�gation�fr�e�e,����-then���the�entir��p�e�G-pr�o�of�is�double-ne�gation�fr�e�e.��(��-�src:1861dn.texPr��p�o�of�.��By��the�subform��!ula�prop�M�ert�y�of�cut-free�pro�M�ofs:��Ev�ery�form�ula�in�the����-pro�M�of��fis�a�subform��!ula�of�the�nal�sequen�t.���"$����f30�������j֍����*�1F'���홊����-�Lemma�2�11���k��src:1865dn.tex�The��ftr��p�anslation�fr�om�H��to�G�is�sound.�F�That�is,�D�if�H�pr��p�oves��
����-�a��_�fr��p�om�assumptions���j�,��|then�G��2pr�oves�the�se�quent���JR�)��A��_�(wher�e��A��is�the����-tr��p�anslation����a�����0����,�and����is����j����0��*�).���ʍ�-�src:1872dn.texPr��p�o�of�.��W��ee��pro�M�ceed�b��!y�induction�on�the�length�of�pro�ofs.��When�the�length����-is��<zero,��1w��!e�m�ust�exhibit�a�pro�M�of�in�G��3of�eac�h�of�the�axioms�H1{H4.�I^This�is����-a�ةroutine�exercise�in�the�Gen��!tzen�sequen�t�calculus,��:whic�h�w�e�omit.�t�F��eor�the����-induction�[�step,��supp�M�ose�w��!e�ha�v�e�pro�M�ofs�in�H�[�from�assumptions����׹of��a��and����-�i�(�a;��1b�).��dThen�E�b��!y�the�induction�h�yp�M�othesis,�m�w�e�ha�v�e�pro�M�ofs�in�G�E�of���)��A����-�and�Id�
��)��A��!��B����.���W��ee�require�a�pro�M�of�in�G�ILof���)��B����.���It�is�easy�to�exhibit�a����-pro�M�of��in�G��of��A�
��!��((�A��!��B����)��!��B��).��Applying��the�cut�rule�t��!wice,��Sw�e��obtain����-a�υpro�M�of�in�G��9plus�the�cut�rule�of�����)��B����.�Y:By�υGen��!tzen's�cut-elimination����-theorem,� _there��exists�a�pro�M�of�in�G��of���G�)��B����.��This��completes�the�pro�of�of����-the��flemma.�������-�Lemma�2�12���k��src:1889dn.tex�(i)��7Supp��p�ose�G�pr�oves�the�se�quent���
��)��A�.�	:Then��7ther�e�is�an�M-����-pr��p�o�of�M(of��a��fr��p�om�assumptions��
����.���If�G�Mpr�oves���
��)�"��[]�,�l�wher�e��[]��is�the�empty�list,����-then�f�ther��p�e�is�an�M-pr�o�of�of��p��fr�om�assumptions��
����,���wher�e��p��is�any�formula����-of���H.��
�׍�>�src:1896dn.tex(ii)�
�If�any�double�ne��p�gations�o�c�cur�in�subformulas�of�the�given�se�quent����-��H��)���'��(wher��p�e�her�e����c�an�b�e�empty�or�not),�w.then�a�pr�o�of�as�in�(i)�c�an����-b��p�e�Wfound�that�c�ontains�no�double�ne�gations�exc�ept�those�arising�fr�om�the����-H-tr��p�anslations���of�double-ne�gate�d�subformulas�of���
��)���.����>�src:1904dn.tex(iii)���If�in�p��p�art�(i)�the�H-tr�anslation�of�the�given�se�quent���
��)������do�es�not����-c��p�ontain��nany�double�ne�gations,���then�the�M-pr�o�of�that�is�asserte�d�to�exist�c�an����-also���b��p�e�found�without�double�ne�gations.���ʍ�-�src:1913dn.texPr��p�o�of�.���W��ee��fpro�M�ceed�b��!y�induction�on�the�length�of�pro�of�of��
��)��A��f�in�G.����>�src:1916dn.texBase��6case:��}the�sequen��!t�has�the�form��;��1A�!��)��A�.�MW��ee��6m�ust�sho�w�that��a��is����-deriv��dDable��fin�H�from�premisses��
���;��1a�,�whic��!h�is�clear.����>�src:1921dn.texNo��!w��ffor�the�induction�step.���W��ee�consider�one�case�for�eac�h�rule�of�G.����>�src:1925dn.texCase��f1,�the�last�inference�in�the�G-pro�M�of�is�b��!y�rule��!)�:��5��>�src:1935dn.tex����K��
��)��A��jB���;��1���)����?_�����U��Bfbߎ��l卍��G�A�
��!��B���;��1��;���
��)���������>�src:1938dn.texBy�r�the�induction�h��!yp�M�othesis,��w�e�r�ha�v�e�an�M-pro�M�of�of��a��from���j�,��and�an��
����-M-pro�M�of�D�of������from��b��and��
����.���W��ee�m��!ust�giv�e�an�M-pro�M�of�of������from��i�(�a;��1b�),�lZ��j�,����-and��f�
����.����>�src:1945dn.texApplying��amo�M�dus�p�onens�to��i�(�a;��1b�)�(whic��!h�is�(�A����!��B����)�����0���9�)��aand�the�giv�en����-pro�M�of��Hof��a��from���j�,��w��!e�deriv�e��b�.��Cop�ying�the�steps�of�the�pro�M�of�of����&�from���"$����f31���� �^��j֍����*�1F'���홊��-�assumptions����b;��1
�#N�(but�c��!hanging�the�justication�of�the�step(s)��b��from�\as-��
����-sumption"��pto�the�line�n��!um�b�M�er��pwhere��b��has�b�een�deriv��!ed),��w�e��pderiv�e���>N�from����-assumptions�V\(�A�
��!��B����)�����0���9�,���[;��1
����,�f^completing�the�pro�M�of�of�case�1.��/No�double�nega-����-tions��fare�in��!tro�M�duced�b�y�this�step.��
�Ӎ�>�src:1959dn.texCase��f2,�the�last�inference�in�the�G-pro�M�of�is�b��!y�rule��)!�:���l��>�src:1968dn.tex����ԍ�T�A;��1��
��)��B��?_�����U��B9�S���l卍��
��)��A��!��B������m��>�src:1971dn.tex�By�z=the�induction�h��!yp�M�othesis,��w�e�z=ha�v�e�an�M-pro�M�of�from�H1{H4�of��b��from����-�
�}��and����a�.��DApplying�the�deduction�theorem�for�H1{H4�with�M-pro�M�ofs,�0�w��!e����-ha��!v�e���an�M-pro�M�of�in�H���of��i�(�a;��1b�)�from��
����.��But�(�A�'f�!��B����)�����0�����=��i�(�a;�b�),���completing����-this��case.��DNote�that�double�negations�are�not�in��!tro�M�duced�b�y�the�deduction����-theorem���if�they�are�not�already�presen��!t,��b�y���part�(v)�of�the�deduction�theorem.����-(One��fsees�wh��!y�w�e�m�ust�use�M-pro�M�ofs�instead�of�condensed�detac�hmen�t.)����>�src:1983dn.texCase��N3,���the�last�inference�in�the�G-pro�M�of�in��!tro�duces�negation�on�the�righ��!t:������>�src:1992dn.tex����g��A;��1��
��)��[]��������U��B+
���l卍��2�
��)�:�A������w��>�src:1995dn.tex�By�v�the�induction�h��!yp�M�othesis,��Sthere�is�an�M-pro�of�of��n�(�a�)�from��a��and��
����.�x�By����-the��>deduction�theorem�for�H1{H4�with�M-pro�M�ofs,���there�is�a�pro�of�of��i�(�a;��1n�(�a�))����-from���
����.���Hence,�<%it�suces�to�sho��!w�that��n�(�a�)�is�deriv��dDable�from��i�(�a;��1n�(�a�)).����-This���follo��!ws�from�a�substitution�instance�of�H3,��#whic�h�is��i�(�i�(�x;��1n�(�x�))�;�n�(�x�)),����-substituting��f�a��for��x�.����>�src:2006dn.texCase�pW4,�{'the�last�inference�in�the�G-pro�M�of�in��!tro�duces�negation�on�the�left:������>�src:2015dn.tex�������	 �
��)��A��N�����U��B2V뎎�	'Ǎ��:�A;��1��
��)��[]�����3���>�src:2018dn.texBy��the�induction�h��!yp�M�othesis,���w�e��ha�v�e�an�M-pro�M�of�of��a��from��
����.��mW��ee�m�ust����-sho��!w�~�that�from��n�(�a�)�and��
����,���w�e�can�deduce��b��in�L,�where��b��is�an�y�form�ula�of����-�H��.���W��ee��\ha��!v�e��i�(�a;��1i�(�n�(�a�)�;�b�))�as�a�substitution�instance�of�axiom�H4.���Applying����-mo�M�dus�"p�onens�t��!wice,�Aw�e�ha�v�e�the�desired�M-pro�M�of�of��b��from��
����,�Acompleting����-case��f4.����>�src:2027dn.texCase��f5,�the�last�inference�is�b��!y�con�traction�in�the�an�teceden�t:������>�src:2036dn.tex����K��C�,�;��1C�;���
��)����?_�����U��B:!���l卍�o�C�,�;��1��
��)���������>�src:2039dn.texBy��the�induction�h��!yp�M�othesis�w�e�ha�v�e�an�M-pro�M�of�of�����from�assumptions��
����-�c;��1c�1ȹand��
����,�Iwhic��!h�also�qualies�as�a�pro�M�of�from�assumptions��c��and��
��,�Iso�there����-is��fnothing�more�to�pro��!v�e.����>�src:2045dn.texCase��f6,�the�last�inference�is�b��!y�thinning�in�the�an�teceden�t:���"$����f32����!�t��j֍����*�1F'����j:��>�src:2054dn.tex����ԍ�o��
��)����N�����U��B-B쎎�l卍�C�,�;��1��
��)��������t��>�src:2057dn.texBy�jthe�induction�h��!yp�M�othesis,�v#w�e�jha�v�e�an�M-pro�M�of�from�H1{H4�of�����from��
����-assumptions��g�����0���9�.�p�That�coun��!ts�as�an�M-pro�M�of�from�assumptions��C�,�;��1
�s"�as�w�ell.����-That��fcompletes�case�6.��
�+��>�src:2063dn.texCase���7,�ͺthe�last�inference�is�b��!y�in�terc�hange�in�the�an�teceden�t.�<AThis�just����-means���the�order�of�form��!ulas�in�the�assumption�list�has�c�hanged,��so�there�is����-nothing��fto�pro��!v�e.����>�src:2068dn.texThat�#�completes�the�pro�M�of�of�part�(i)�of�the�lemma.�	U�Regarding�parts����-(ii)�ʜand�(iii),�Ӫb��!y�the�preceding�lemma,�an��!y�double�negations�o�M�ccurring�an�y-����-where��rin�the�G-pro�M�of�m��!ust�o�ccur�in�the�nal�sequen��!t.�یNo�new�double�nega-����-tions���are�in��!tro�M�duced�in�the�translation�to�H,�and�all�the�theorems�of�H���that����-w��!e�zqused�ha�v�e�b�M�een�giv�en�double-negation-free�condensed-detac�hmen�t�pro�M�ofs����-from���H1{H4.��
By�Theorem�1,��lthey�ha��!v�e���double-negation-free�M-pro�M�ofs,�to�o.����-Although��w��!e�ma�y�not�ha�v�e�p�M�oin�ted�it�out�in�eac�h�case,��(the�argumen�t�giv�en����-pro�M�duces�ͬan�M-pro�of�in�whic��!h�an�y�double�negations�arise�from�the�transla-����-tions�vdin��!to�H�vof�doubly�negated�subform�ulas�of�the�nal�sequen�t.�x�In�particular,����-if���the�nal�sequen��!t�con�tains�no�double�negations,�?then�the�M-pro�M�of�pro�duced����-also��fcon��!tains�no�double�negations.��(�����-�Corollary�2�2���p�n�src:2088dn.tex�H�1tis�1�a�b��p�asis�for�the�implic�ation-ne�gation�fr�agment�of�intuition-����-istic��^lo��p�gic.�"�That�is,��{every�intuitionistic�al��Fly�valid�formula�in�this�fr�agment�is����-pr��p�ovable���in�H.���n��-�src:2093dn.texR��p�emark�.�	�In�X[�4��y�],�j�this�lemma�w��!as�pro�v�ed�for�a�dieren�t�axiomatization�of����-the�6implication-negation�fragmen��!t�of�in�tuitionistic�calculus,�L�so�this�corollary����-could�Jalso�b�M�e�pro��!v�ed�Jb�y�demonstrating�the�equiv��dDalence�of�the�t�w�o�fragmen�ts����-directly��e.����-�src:2101dn.tex�Pr��p�o�of�.�T^Supp�M�ose�x��A��is�an�in��!tuitionistically�v��dDalid�form�ula�con�taining�no�con-����-nectiv��!es�m
other�than�implication�and�negation.�ʾBy�Gen�tzen's�cut-elimination����-theorem,��there��|is�a�cut-free�pro�M�of�of�the�sequen��!t�[]�� �)��A��|�(with�empt�y�an-����-teceden��!t).�,5By���Lemma�12,���A��has�an��M�1��-pro�M�of,�whic��!h�in�particular�is�a�pro�M�of����-in��fH.����-�src:2108dn.tex�R��p�emark�.�8�The���main�idea�of�the�corollary�is�that�b��!y�the�subform�ula�prop-����-ert��!y�MXof�cut-free�pro�M�ofs,�_'the�cut-free�pro�of�con��!tains�no�connectiv�es�other�than����-implication��fand�negation.��(�����-�Theorem�2�8���n>M�src:2112dn.tex�Supp��p�ose�ˏH��Upr�oves��b��fr�om�assumptions���5��and�neither����nor��b����-�c��p�ontains�4)double�ne�gation.��3Then�ther�e�is�a�c�ondense�d-detachment�pr�o�of�of��b����-�fr��p�om���H1{H4�and�assumptions���U�that�do�es�not�c�ontain�double�ne�gation.���"$����f�33����"�0��j֍����*�1F'���홊��>�src:2119dn.tex�Mor��p�e�X%gener�al��Fly,�ssif����<�and��b��ar�e�al��Flowe�d�to�c�ontain�double�ne�gation,�ssthen��
����-ther��p�e�5is�a�c�ondense�d-detachment�pr�o�of�of��b��fr�om�H1{H4�and�assumptions������-�that�ˌc��p�ontains�no�new�double�ne�gations.��That�is,���al��Fl�doubly�ne�gate�d�formulas����-o��p�c�curring���in�the�pr��p�o�of���ar�e�subformulas�of���U�or�of��b�.���Ǎ�-�src:2127dn.texPr��p�o�of�.��4Let����b�����0��	N�=���B�|�b�M�e�the�translation�of��b��in��!to�G���dened�earlier.�Double����-negations��'in��B�)˹arise�only�from�double�negations�in��b�.���Supp�M�ose��b��is�pro��!v��dDable����-in�3H�from�assumptions���j�.�<CBy�Theorem�1,�8fthere�is�an�M-pro�M�of�of��b��from����.����-By�dLemma�11,�<dthe�sequen��!t��ң�)��B���is�dpro�v��dDable�in��G�.�E�Hence,�<db�y�Gen�tzen's����-cut-elimination��theorem,���there�is�a�pro�M�of�in�G��of��y��)��b�.���By��the�preceding����-lemma,��there�ͩis�an�M-pro�M�of�of��B��������0��(��from�assumptions�������0��������	��0��I��that�con��!tains�no�new����-double�gnegations.���But�b��!y�Lemma�8,�g�B��������0���ڹ=����b��and������0��r6�=���j�.���Th��!us�w�e�ha�v�e�an����-M-pro�M�of��Pof��b��from���j�.���By�the�D-completeness�of��H��,��UTheorem�7,�there�is�also����-a��condensed-detac��!hmen�t�pro�M�of�of��b��from���j�.�/�The�second�part�of�Theorem�7����-sa��!ys�X�that�w�e�can�nd�a�double-negation-free�condensed-detac�hmen�t�pro�M�of�of����-�b�Dž�from���j�.���It�is�imp�M�ortan��!t�that�w�e�are�w�orking�with�M-pro�M�ofs�here�b�ecause�the����-second�(�part�of�the�D-completeness�theorem,�Bab�M�out�double�negations,�applies����-only��fto�M-pro�M�ofs.���That�completes�the�pro�of.��E����-�Theorem�2�9���n>M�src:2155dn.tex�Supp��p�ose�n��A��is�pr�ovable�fr�om�H1{H4�by�using�c�ondense�d�detach-����-ment�4.as�the�only�rule�of�infer��p�enc�e.��BThen�4.�A��has�a�pr��p�o�of�4.fr�om�H1{H4�by����-using�c��p�ondense�d�detachment�in�which�no�doubly�ne�gate�d�formulas�o�c�cur�ex-����-c��p�ept���those�that�alr�e�ady�o�c�cur�as�subformulas�of��A�.����-�src:2163dn.texPr��p�o�of�R�.��@Supp�M�ose��2�A��is�pro��!v��dDable�from�H1{H4�using�condensed�detac�hmen�t.����-Eac��!h�3�condensed-detac�hmen�t�step�can�b�M�e�con�v�erted�to�three�steps�b�y�using����-substitution���and�mo�M�dus�p�onens,�mso��A��is�pro��!v��dDable�in�H.�By�the�preceding����-theorem,���A�IY�has�a�condensed-detac��!hmen�t�IYpro�M�of�from�H1{H4�in�whic��!h�no����-doubly�֛negated�form��!ulas�o�M�ccur�except�those�that�already�o�ccur�in��A�.�n{That����-completes��fthe�pro�M�of.��
�֍�-�src:2173dn.tex�R��p�emark�-��.�	szSince�-�the�translation�bac��!k�from�Gen�tzen�calculus�pro�M�duces�M-����-pro�M�ofs,��6w��!e�Ȧdo�not�need�to�app�eal�to�D-completeness�for�arbitrary�H-pro�ofs.����-This�jis�fortunate�b�M�ecause�w��!e�do�not�kno�w�a�pro�M�of�of�D-completeness�that����-a��!v�oids�[�the�p�M�ossible�in��!tro�duction�of�double�negations,�j�except�when�restricted����-to��fM-pro�M�ofs.��E����-�Corollary�2�3���p�n�src:2182dn.tex�L��p�et��U�T�^��b�e�any�set�of�axioms�for�intuitionistic�pr�op�ositional�lo�gic.����-Supp��p�ose�r�that�ther�e�exist�c�ondense�d-detachment�pr�o�ofs�of�H1{H4�fr�om��T����in����-which�Ono�double�ne��p�gations�o�c�cur�(exc�ept�those�that�o�c�cur�in��T��V�,�nCif�any).�ՆThen����-the���pr��p�e�c�e�ding�the�or�em�is�true�with��T�p@�in�plac�e�of�H1{H4.���"$����f�34����#���j֍����*�1F'���홊��-�src:2191dn.tex�Pr��p�o�of�R�.��Let��q�b��b�M�e�pro��!v��dDable�from��T��V�.�Then��b��is�pro��!v��dDable�from�H1{H4�b�M�ecause��
����-�T����is�tRa�set�of�axioms�for�in��!tuitionistic�logic.��,By�the�theorem,�~Vthere�is�a�pro�M�of����-of����b��from�H1{H4�that�con��!tains�no�double�negations�(except�those�o�M�ccurring����-in���b�,�&-if�an��!y).��Supplying�the�giv�en�pro�M�ofs�of�H1{H4�from��T��V�,�&-w�e�construct�a����-pro�M�of��:of��b��from��T�4��that�con��!tains�no�double�negations�except�those�o�ccurring����-in��f�T�+��or�in��b��(if�an��!y).���That�completes�the�pro�M�of.��!����-�11��HDouble-Negation��P��ostp�`onemen�t���卑-�src:2206dn.tex�In�t�preceding�sections�w��!e�considered�attempts�to�deriv�e�a�double-negation-free����-theorem�j�without�in��!tro�M�ducing�double�negations�in�the�deduced�steps.��In�this����-section���w��!e�consider�the�question�of�deducing�a�theorem�that�do�M�es�con�tain�a����-double�Rnegation.���In�that�case,�b�of�course,�there�cannot�b�M�e�a�double-negation-����-free�Ppro�M�of,�Kbut�w��!e�sho�w�that�w�e�can�do�the�next�b�M�est�thing:���w�e�can�nd�a����-pro�M�of�v�that�is�double-negation�free�un��!til�the��very��last�step�|all�v�the�double-����-negations�\�are�in��!tro�M�duced�at�the�last�application�of�condensed�detac�hmen�t.����-This��Oma��!y�seem�surprising�at�rst,��!but�the�pro�M�of�sho�ws�that�for�man�y�logics,��!it����-really�۔is�just�a�simple�corollary�of�our�results�on�double-negation�elimination.����-The��result�applies�to�classical�logic�as�axiomatized�b��!y�L1{L3,�%for�example.��If����-a���prop�M�ositional�logic��T�-�has�this�prop�ert��!y��e,��w�e���sa�y��T�-�satises��double-ne��p�gation����-p��p�ostp�onement�.��Note���that�double-negation�p�M�ostp�onemen��!t���is�a�stronger�prop-����-ert��!y�ߪthan�double-negation�elimination:�Pedouble-negation�p�M�ostp�onemen�t�ߪsa�ys����-there��is�a�pro�M�of�with�no�double�negations�except�p�ossibly�in�the�axioms�or����-the���nal�conclusion,���and�if�the�nal�conclusion�has�no�double�negation,�this����-is�djust�the�statemen��!t�of�double-negation�elimination.��Note�also�that�double-����-negation��5p�M�ostp�onemen��!t�implies�triple-negation�p�ostp�onemen��!t,���and�so�forth,����-since��fa�triple�negation�is�also�a�double�negation.��'׍�>�src:2221dn.texW��ee�9�sa��!y�that��T����denes�]�e��p�quivalenc�e�9��if�there�is�a�prop�M�ositional�form�ula����-�E����(�x;��1y�d��)��fsuc��!h�that��DR���<�i�(��(�y�d��)�;��1i�(�E����(�x;�y��)�;��(�x�)))��p��-�src:2223dn.texcan��b�M�e�pro��!v�ed��in��T�4u�for�eac��!h�form�ula���.��W��ee�sa�y�that��T�4u�denes���ve��p�ctor�e�quiv-����-alenc��p�e��˹if�for�eac��!h�p�M�ositiv�e�in�teger��n��there�is�a�form�ula��E���z�n���P�(�x;��1y�d��),��dwith�free����-v��dDariables��f�x�
��=��x���z�1����;���1:�:�:��l�;��1x���z�n��	N��and��y�o:�=��y���z�1����;���1:�:�:��l�;��1y���z�n���P�,�suc��!h�that��DR������i�(��(�y�d��)�;��1i�(�E���z�n���P�(�x;�y��)�;��(�x�)))����-�src:2228dn.texcan�5�b�M�e�pro��!v�ed�5�in��T���for�eac��!h�form�ula���.���Clearly��e,�Ythe�same�form�ulas�satisfy��
����-the���conditions�for��E����(�x;��1y�d��)�and��E���z�1����(�x;�y�d��),��and�w��!e�tak�e�these�t�w�o�forms�to�b�M�e����-synon��!ymous.�qWhen��z�x��and��y�<
�are�lists�(v�ectors)�of�sev�eral�v��dDariables,��b�y��n�(�y�d��)���"$����f35����$�@��j֍����*�1F'���홊��-�w��!e�5�understand�the�v�ector��n�(�y���z�1����)�;���1:�:�:��l�;��1n�(�y���z�n���P�).��_W��ee�ma�y�drop�the�subscript�on��
����-�E���z�n��	N��for��fnotational�con��!v�enience.��oҍ���-�Lemma�2�13���k��src:2234dn.tex�Pr��p�op�ositional��lo�gic�L1{L3�c�an�dene�b�oth�e�quivalenc�e�and�ve�c-����-tor���e��p�quivalenc�e�in�such�a�way�that�the�dening�formula��E����(�x;��1n�(�y�d��))��is�double-����-ne��p�gation���fr�e�e.����-�src:2240dn.texPr��p�o�of�.��hDene�u�x�-�^��y�ٙ�to�b�M�e��n�(�i�(�x;��1n�(�y�d��))).��hDene��E����(�x;�y�d��)�to�b�M�e��i�(�x;�y�d��)�-�^��i�(�y�;��1x�),����-whic��!h��fis��n�(�i�(�i�(�x;��1y�d��)�;�n�(�i�(�y�;�x�)))).���Then��f�E����(�x;�n�(�y��))��fis��oҍ���n�(�i�(�i�(�x;��1n�(�y�d��))�;�n�(�i�(�n�(�y��)�;�x�))))�;����-�src:2243dn.tex�whic��!h��Wis�double-negation�free.�j�W��ee�no�w�pro�M�ceed�b�y�recursion�on��n��to�dene��
����-the��'form��!ula��E���z�n���P�(�x;��1y�d��)�when��x��and��y�*��are�v�ectors�of�length��n�.�=The�natural����-thing��fis�to�dene��E���z�n�+1���̹((�x;��1x���z�n�+1���)�;��(�y�d�;�y���z�n�+1���))��fto�b�M�e��oҍ���j�E���z�n���P�(�x;��1y�d��)�n��^��E����(�x���z�n�+1�����;�y���z�n�+1���)�;����-�src:2247dn.tex�where���x�
��=��x���z�1����;���1:�:�:��l�;��1x���z�n���E�and��y�o:�=��y���z�1����;���1:�:�:��l�;��1y���z�n���P�.��bUsing�the�denition�of�conjunction��
����-(�^�),��fw��!e�obtain�����[��n�(�i�(�E���z�n���P�(�x;��1y�d��)�;�n�(�E����(�x���z�n�+1�����;�y���z�n�+1���))))�;�����-�src:2250dn.tex�and��fusing�the�denition�of�equiv��dDalence�on�the�righ��!t,�w�e�get��oҍ�_@��n�(�i�(�E���z�n���P�(�x;��1y�d��)�;�n�(�n�(�i�(�i�(�x���z�n�+1�����;�y���z�n�+1���)�;�n�(�i�(�y���z�n�+1���;�x���z�n�+1���)))))))�:����-�src:2252dn.tex�A��#double��&negation�has�b�M�een�in��!tro�duced,��whic�h��&is�undesirable.�W��ee�therefore��
����-cancel��fthat�double�negation�and�dene�instead:�����c3��E���z�n�+1���̹((�x;��1x���z�n�+1���)�;��(�y�d�;�y���z�n�+1���))���6����uE:=�
��n�(�i�(�E���z�n���P�(�x;��1y�d��)�;�i�(�i�(�x���z�n�+1�����;�y���z�n�+1���)�;�n�(�i�(�y���z�n�+1���;�x���z�n�+1���)))))�:�����>�src:2262dn.tex�No��!w��w�e�can�pro�v�e�b�y�induction�on��n��that��E���z�n���P�(�x;��1y�d��)�is�double-negation�free����-for�7�eac��!h��n�.���W��ee�ha�v�e�already�done�the�base�case,�[�and�the�denition�mak�es����-the��finduction�step�apparen��!t.����>�src:2265dn.texIt��remains�to�v��!erify�that�L1{L3�pro�v�es��i�(��(�y�d��)�;��1i�(�E����(�x;�y��)�;��(�x�)))��for�eac��!h����-form��!ula��|���and�eac�h��n�,���where��x�
��=��x���z�1����;���1:�:�:��l�;��1x���z�n���P�.���Since��|L1{L3�is�an�axiomatiza-����-tion�.$of�classical�logic,�F1w��!e�can�app�M�eal�to�the�completeness�theorem:���it�suces����-to�5.sho��!w�that��i�(��(�y�d��)�;��1i�(�E����(�x;�y��)�;��(�x�)))�5.is�v��dDalid�under�eac��!h�assignmen�t�of�truth����-v��dDalues�yto�the�v�ariables��x��and��y�d��.�νF��eor�this�it�suces�to�sho��!w�that�if��E����(�x;��1y��)�is����-satised��under�a�giv��!en�truth�assignmen�t,��then��x���z�i��.�gets�the�same�truth�v��dDalue����-as���y���z�i��f��for�eac��!h��i��޹=�1�;���1:�:�:��l�;��1n�.���This��w�e�pro�v�e�b�y�induction�on��n�.���F��eor�the�base���"$����f36����%�>��j֍����*�1F'���홊��-�case�n�w��!e�ha�v�e��E����(�x;��1y�d��)�X�=��n�(�i�(�i�(�x;�y�d��)�;�n�(�i�(�y�;�x�)))),��and�n�a�four-line�truth�table��
����-can��]b�M�e�used�to�v��!erify�the�claim;���alternativ�ely�one�can�v�erify�the�truth�table����-of���x�U��^��y�e^�as�dened�ab�M�o��!v�e�and�the�truth-table�equiv��dDalence�of��E����(�x;��1y�d��)�and����-�i�(�x;��1y�d��)�x��^��i�(�y�;��1x�).�
$The��(induction�step�is�apparen��!t�from�the�denition�of��E���z�n�+1�����-�in��fterms�of��E�G��and��E���z�n���P�.���That�completes�the�pro�M�of�of�the�lemma.���g����-�Lemma�2�14���k��src:2277dn.tex�L1{L3���pr��p�oves�every�instanc�e�of��E����(�x;��1x�)�,�wher�e��x�
��=��x���z�1����;���1:�:�:��l�;��1x���z�n���P�.��i���-�src:2280dn.texPr��p�o�of�.�	W�It�$qprobably�is�p�M�ossible�to�sho��!w�ho�w�to�actually�construct�suc�h�a����-pro�M�of,�Ƣpro�ceeding��0b��!y�induction�on��n�,�but�it�is�not�necessary�b�M�ecause�L1{L3����-is��an�axiomatization�of�classical�logic,���and�all�these�statemen��!ts�are�v��dDalid,�as����-is��Ueasily�pro��!v�ed��Ub�y�induction�on��n�.���By�the�completeness�theorem�and�D-����-completeness,�%"v��dDalid��statemen��!ts�are�pro�v��dDable�in�L1{L3.�That�completes�the����-pro�M�of.������-�Theorem�2�10�(P��tostp�Y�onemen�t�2�of�double�negation)���;S��src:2286dn.tex�L��p�et�؈�T�]��b�e�any�pr�op�o-����-sitional�(�lo��p�gic�such�that�(1)��T����satises�double-ne�gation�elimination,�x
(2)��T����-�c��p�an��dene�e�quivalenc�e�and�ve�ctor�e�quivalenc�e�in�such�a�way�that��E����(�x;��1n�(�y�d��))����-�is��double-ne��p�gation�fr�e�e,��and�(3)��T�\�c�an�pr�ove�every�instanc�e�of��E���z�n���P�(�x;��1x�)�.����-Then�
��T��O�satises�double-ne��p�gation�p�ostp�onement.�i�In�p�articular,��L1{L3�satis-����-es���double-ne��p�gation�p�ostp�onement.��i���-�src:2295dn.texPr��p�o�of�.���W��ee���v��!eried�earlier�that�L1{L3�satises�h�yp�M�otheses�(1),��and�in�the����-lemmas��Yw��!e�v�eried�h�yp�M�otheses�(2)�and�(3).�
k�Hence,���it�suces�to�deriv�e����-double-negation��fp�M�ostp�onemen��!t�from�those�h�yp�M�otheses.��
�g��>�src:2298dn.texF��eor��simplicit��!y�,�1w�e��rst�consider�the�case�of�a�theorem��A��of��T��^�that�con-����-tains��=a�single�double�negation,��2but�no�triple�negation,�and�no�other�double����-negation���(though�it�ma��!y�con�tain�m�ultiple�copies�of�that�same�doubly�negated����-form��!ula).���Let��Vthe�doubly�negated�form�ula�b�M�e��n�(�n�(�P��V�)),��Ra�subform�ula�of�the����-theorem����A�.�ݨLet���(�x�)�b�M�e�the�form��!ula�resulting�from��A��b�y�replacing��n�(�P��V�)�b�y����-a��"new�v��dDariable��x�,�not�o�M�ccurring�in��A�.��Th��!us��A��is���[�x��5�:=��n�(�P��V�)],�or��"for�short����-��(�n�(�P��V�)),�!and�����is�double-negation�free.�dW��ee�m��!ust�sho�w�that��A��has�a�pro�M�of����-in�{�whic��!h�double�negation�en�ters�only�at�the�last�step.�^Since��T�%�can�dene����-equiv��dDalence,��fthere�is�a�pro�M�of�of���������i�(��(�n�(�P��V�))�;��1i�(�E����(�x;�n�(�P��))�;��(�x�)))�:����-�src:2307dn.tex�By���h��!yp�M�othesis,�
there�is�some�pro�of�of��A��(whic��!h�is���(�n�(�P��V�)))�in��T��.�'oHence��
����-there��fis�a�pro�M�of�of���������i�(�E����(�x;��1n�(�P��V�))�;��(�x�))�:���"$����f�37����&�Š�j֍����*�1F'���홊��-�src:2310dn.tex�Since��J�E����(�x;��1n�(�y�d��))�is�double-negation�free�and��P���and����are�double-negation��
����-free,���the�w�displa��!y�ed�form�ula�is�double-negation�free.��DSince��T���satises�double-����-negation���elimination,�<there�is�a�double-negation-free�pro�M�of�of�that�form��!ula.����-By���h��!yp�M�othesis�(3),���there�also�is�a�pro�of�in��T�,'�of��E����(�n�(�P��V�)�;��1n�(�P��)).��Since���w��!e����-assumed�Џthat��A��has�no�triple�negation,�the�form��!ula��P�U�is�not�a�negation,����-and��/since�w��!e�assumed�that��n�(�n�(�P��V�))�is�the�only�double�negation�in��A�,�>��P����-�is�`�double-negation�free.�
�Applying�double-negation�elimination�again,��Vw��!e����-ha��!v�e��a�double-negation-free�pro�M�of�of��E����(�n�(�P��V�)�;��1n�(�P��)).��xApplying��condensed����-detac��!hmen�t�?eto�this�form��!ula�and��i�(�E����(�x;��1n�(�P��V�))�;��(�x�)),�e�w�e�?eobtain�the�conclu-����-sion�)��(�n�(�P��V�))�as�desired.�e�That�completes�the�pro�M�of�in�case��A��has�no�triple����-negations��(and�only�one�double�negation.��#Note�that�only�the�denabilit��!y�of����-equiv��dDalence,��fnot�v��!ector�equiv�alence,�w��!as�used�so�far.����>�src:2319dn.texNext�V
w��!e�consider�ho�w�to�relax�those�assumptions�b�y�in�tro�M�ducing�more����-than���one�extra�v��dDariable��x�.��vW��ee�do�so�recursiv��!ely�.��vNamely�,��&w�e���nd�a�form��!ula����-�P��that�_�is�double-negation�free,���is�not�itself�a�negation,�and�o�M�ccurs�doubly����-negated���in��A�.�r/W��ee�in��!tro�M�duce�a�new�v��dDariable��x��in�place�of��n�(�P��V�)�as�ab�o��!v�e,��3so����-that�t��A��is�� �d��[�x�b͹:=��n�(�P��V�)]�t�for�some�� ��.�IWIf�there�is�an��!y�double�negation�in��A�,����-then�)�suc��!h�a�form�ula��P��عcan�b�M�e�found.�g0After�replacing��n�(�P��V�)�b�y��x��to�obtain����-� �d��,��w��!e��*pro�M�ceed�recursiv�ely�to�mak�e�similar�replacemen�ts�in�� �d��.�(Since�eac�h����-new�+�v��dDariable�reduces�the�n��!um�b�M�er�+�of�doubly�negated�subform��!ulas,�M	this�pro-����-cess��7will�terminate�after�in��!tro�M�ducing�v��dDariables��x���z�1����;���1:�:�:��l�;��1x���z�n��	`��for�some��n�,���and��A����-�will��b�M�e�of�the�form���[�x���z�1��	��:=���n�(�P���z�1����)�;���1:�:�:��l�x���z�n��
yS�:=��n�(�P���z�n���P�)].��If��w��!e�write��n�(�P��V�)�for����-�n�(�P���z�1����)�;���1:�:�:��l�;��1n�(�P���z�n���P�)��*and��x��for�the�v��!ector��x���z�1���;���1:�:�:��l�;��1x���z�n���P�,��[w��!e�can�use�the�more�com-����-pact��notation���[�x��D�:=��n�(�P��V�)].���W��ee��can�no��!w�rep�M�eat�the�pro�of�in�the�preceding����-paragraph�{�v��!erbatim,��but�no�w�it�applies�to�the�general�case.�]�W��ee�need�the����-denabilit��!y�C�of�v�ector�equiv��dDalence�to�sho�w�that��E����(�x;��1n�(�y�d��))�is�double-negation����-free�twhen��x��and��y�ة�are�v��!ectors�and�to�sho�w�that��E����(�n�(�P��V�)�;��1n�(�P��))�tis�double-����-negation��ffree.���That�completes�the�pro�M�of.����>�src:2331dn.tex�Example.��G�F��erege's���prop�M�ositional�logic�includes�the�axioms��i�(�n�(�n�(�x�))�;��1x�)����-and����i�(�x;��1n�(�n�(�x�))).��3By�our�theorem,�s4there�m��!ust�exist�pro�M�ofs�of�eac�h�of����-these��$axioms�from�L1{L3�that�in��!v�olv�e��$double�negation�only�in�the�con-����-clusion.��Indeed,��it��zis�straigh��!tforw�ard��zto�nd�suc��!h�pro�M�ofs�with�Otter.�W��ee����-use�b�the�second�example,����i�(�x;��1n�(�n�(�x�))),�to�b�illustrate�the�metho�M�d�implicit�in����-the�gOpro�M�of�of�the�double-negation�p�ostp�onemen��!t�theorem.���That�metho�d�tells����-us��to�in��!tro�M�duce�a�new�v��dDariable��y�sT�and�seek�double-negation-free�pro�ofs�of����-�i�(�E����(�y�d�;��1n�(�x�))�;�i�(�x;�n�(�y��)))��Iand��E����(�n�(�x�)�;��1n�(�x�)).��If�suc��!h�pro�M�ofs�are�found,���then����-the���desired�double-negation-p�M�ostp�oned���pro�of�of��i�(�x;��1n�(�n�(�x�)))�requires�only����-one��more�step.�	C�W��eriting�out�the�denition�of��E����,�{�the�t��!w�o��form�ulas�to�b�M�e����-pro��!v�ed��fare����~ۤ�i�(�n�(�i�(�i�(�y�d�;��1n�(�x�))�;�n�(�i�(�n�(�x�)�;�y��))))�;�i�(�x;�n�(�y��)))���"$����f38����'�*��j֍����*�1F'���홊��-�src:2340dn.tex�and���%���v�n�(�i�(�i�(�n�(�x�)�;��1n�(�x�))�;�n�(�i�(�n�(�x�)�;�n�(�x�)))))�:��|g��-�src:2342dn.tex�In�/principle,�F�w��!e�could�use�the�pro�M�of�of�the�double-negation�elimination�theo-��
����-rem��'to�nd�double-negation�free�pro�M�ofs�of�these�theorems�and�then�construct����-the�egdouble-negation-p�M�ostp�oned�pro�of�of��i�(�x;��1n�(�n�(�x�))).��3This�is�not�necessary��e,����-ho��!w�ev�er,�1�since��w�e�found�the�desired�pro�M�of�in�this�case�quite�easily�without����-relying�z�on�the�metho�M�d.�Z�After�all,���the�theorems�pro��!v�ed�z�in�this�pap�er�w��!ere����-designed���not�so�m��!uc�h���to�supplemen��!t�our�theorem-pro�ving�strategies�as�to����-justify��fthem.��H��>�src:2349dn.tex�R��p�emarks.����Besides��]L1{L3,�?the�other�examples�of�systems�discussed�in����-this�vpap�M�er�w��!ere�innite-v��dDalued�logic�A1{A4�and�the�in�tuitionistic�system�H,����-b�M�oth�T�of�whic��!h�are�form�ulated�with�only�implication�and�negation�as�connec-����-tiv��!es.�4nWhether�m�double-negation�p�M�ostp�onemen�t�m�is�true�for�A1{A4�or�for�H����-itself,�U=w��!e�2Edo�not�kno�w.��{If�H�2!admits�double-negation�p�M�ostp�onemen�t,�U=it�2Ewill����-require�z�a�dieren��!t�pro�M�of.��dIt�is�kno�wn�(see�[�11��
�4],���p.��d253)�that�H�z�cannot�dene����-conjunction,�pgso�Hthe�denabilit��!y�of�equiv��dDalence�cannot�b�M�e�carried�out�for�H����-in���the�same�w��!a�y���as�for�L1{L3.��Indeed,�ǧAlb�M�ert�Visser�has�pro��!v�ed���(priv��dDate����-corresp�M�ondence)��'that�H���cannot�dene�equiv��dDalence�(and�his�pro�of�repro��!v�es����-the��fresult�for�conjunction�to�M�o).��"e���-�Ac��kno�wledgmen�ts��ě��-�src:2355dn.tex�W��ee��thank�Kenneth�Harris,�f�Branden�Fitelson,�and�Dolph�Ulric��!h�for�their����-atten��!tion���to�early�drafts�of�this�pap�M�er.�3VIn�addition,��Ulric�h�con�tributed�part����-of��fthe�pro�M�of�of�D1{D3�from�L1{L3.����>�src:2362dn.texThe���w��!ork�of�Mic�hael�Beeson�w�as�supp�M�orted�in�part�b�y�National�Science����-F��eoundation��fgran��!t�no.���CCR-0204362.����>�src:2365dn.texThe���w��!ork�of�Rob�M�ert�V��eero�w�as�supp�M�orted�in�part�b�y�National�Science����-F��eoundation��fgran��!t�no.���CD�A-9503064.����>�src:2368dn.texThe��3w��!ork�of�Larry�W��eos�w�as�supp�M�orted�b�y�the�Mathematical,���Information,����-and�K�Computational�Sciences�Division�subprogram�of�the�Oce�of�Adv��dDanced����-Scien��!tic�q.Computing�Researc�h,��Oce�of�Science,�U.S.�Departmen��!t�of�Energy��e,����-under��fCon��!tract�W-31-109-Eng-38.�� �f��-�References��\����2y��[1]���C�$�src:2377dn.texC.�fyC.�Chang.�Algebraic�analysis�of�man��!y�v��dDalued�logics.��T��)r��p�ans.���A��\mer.����C�$Math.���So��p�c.�,��f88:467{490,�1958.���"$����f39����(
��j֍����*�1F'���홊����2y��[2]���C�$�src:2382dn.texB.�'Fitelson�and�L.�W��eos.�Finding�missing�pro�M�ofs�with�automated�reason-��
����C�$ing.��f�Studia���L��p�o�gic�a�,�68(3):329{356,�2001.��_����2y�[3]���C�$�src:2387dn.texJ.��R.�Hindley�and�D.�Meredith.�Principal�t��!yp�M�e-sc�hemes��and�condensed����C�$detac��!hmen�t.��f�J.���Symb��p�olic�L�o�gic�,��f55(1):90{105,�1990.������2y�[4]���C�$�src:2392dn.texA.���Horn.�The�separation�theorem�of�in��!tuitionist�prop�M�ositional�calculus.����C�$�J.���Symb��p�olic�L�o�gic�,��f27:391{399,�1962.������2y�[5]���C�$�src:2403dn.texS.��:C.�Kleene.��Intr��p�o�duction��Qto�Metamathematics�.��:v��dDan�Nostrand,���Prince-����C�$ton,��fNew�Jersey��e,�1952.������2y�[6]���C�$�src:2408dn.texC.�{A.�Meredith�and�A.�Prior.�Notes�on�the�axiomatics�of�the�prop�M�osi-����C�$tional��fcalculus.��Notr��p�e���Dame�J.�F��)ormal�L�o�gic��f�4:171{187,�1963.������2y�[7]���C�$�src:2413dn.texW.��OMcCune.�Otter�3.3�Reference�Man��!ual.�T��eec�h.�Memo�ANL/MCS-TM-����C�$263,��Mathematics��Vand�Computer�Science�Division,�Argonne�National����C�$Lab�M�oratory��e,��fArgonne,�IL,�August�2003.������2y�[8]���C�$�src:2419dn.texG.�j�Min��!ts�and�T.�T��eammet.�Condensed�detac�hmen�t�is�complete�for����C�$relev��dDance�F�logic:�
BA�E�computer-aided�pro�M�of.��J.�UUA��\utomate��p�d�R�e�asoning�,����C�$7(4):587{596,��f1991.������2y�[9]���C�$�src:2424dn.texA.�-Prijatelj.�Bounded�con��!traction�and�Gen�tzen-st�yle�form�ulation�of�����C�� ����L��JZ%uk��dDasiewicz��flogics.��Studia���L��p�o�gic�a�,��f57(2{3):437{456,�1996.������-[10]���C�$�src:2429dn.texA.�^N.�Prior.�Logicians�at�pla��!y:��YOr�Syll,�/-Simp,�and�^Hilb�M�ert.��A��\ustr��p�alasian����C�$J.���Phil.�,��f34:182{192,�1956.������-[11]���C�$�src:2434dn.texA.�[�N.�Prior.��F��)ormal��/L��p�o�gic�,�j�2nd�[�edition,�Clarendon�Press,�Oxford,�1962.������-[12]���C�$�src:2438dn.texJ.��TRobinson.�A��mac��!hine-orien�ted��Tlogic�based�on�the�resolution�principle.����C�$�J.���A��\CM�,��f12:23{41,�1965.������-[13]���C�$�src:2443dn.texA.�Q�Rose�and�J.�B.�Rosser.�F��eragmen��!ts�of�man�y-v��dDalued�statemen�t�calculi.����C�$�T��)r��p�ans.���A��\mer.�Math.�So�c.�,��f87:1{53,�1958.������-[14]���C�$�src:2448dn.texI.�uThomas.�Shorter�dev��!elopmen�t�uof�an�axiom.��Notr��p�e��Dame�J.�F��)ormal����C�$L��p�o�gic�,��f16:378,�1975.������-[15]���C�$�src:2453dn.texD.��BUlric��!h.�A��legacy�recalled�and�a�tradition�con�tin�ued.��J.��A��\utomate��p�d����C�$R��p�e�asoning�,��f27(2):97{122,�2001.������-[16]���C�$�src:2458dn.texR.�L	V��eero.�Using�hin��!ts�to�increase�the�eectiv�eness�of�an�automated����C�$reasoning���program:���Case�studies.��J.�9xA��\utomate��p�d�R�e�asoning�,�"16(3):223{����C�$239,��f1996.���"$����f40����)v��j֍����*�1F'���홊����-�[17]���C�$�src:2464dn.texL.�r W��eos.�The�resonance�strategy�.��Computers���and�Mathematics�with�Ap-��
����C�$plic��p�ations�,��f29(2):133{178,�1996.��������-[18]���C�$�src:2469dn.texL.���W��eos.�Conquering�the�Meredith�single�axiom.��J.�A��\utomate��p�d�R�e�ason-����C�$ing�,��f27(2):175{199,�2001.������-[19]���C�$�src:2474dn.texL.��W��eos�and�G.�Piep�M�er.��A�n�F��)ascinating�n�Country�in�the�World�of�Comput-����C�$ing:�
�Y��)our��Guide�to�A��\utomate��p�d�R�e�asoning�.��$W��eorld�Scien��!tic,��SSingap�M�ore,����C�$1999.������-[20]���C�$�src:2480dn.texL.�<5W��eos�and�R.�Thiele.�Hilb�M�ert's�new�problem.��Bul��Fletin��8of�the�Se��p�ction�of����C�$L��p�o�gic�,��f30(3):165{175,�2001.���"$����f41����#����;��j��)�#5��"		cmmi9�"o���		cmr9��"V
�3
cmbx10���<x
�3
cmtt10�!",�
�3
cmsy10��b>
�3
cmmi10��':
�3
cmti10���N�cmbx12����@cmti12��K�cmsy8��2cmmi8��Aa�cmr6�|{Ycmr8�X�Qcmr12�D��tG�G�cmr17�K�`y
�3
cmr10�&�������

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