Révision 179 CSL17/arithmetic.tex

arithmetic.tex (revision 179)
1 1
\section{An arithmetic for the polynomial hierarchy}
2
Our base language is $\{ 0, \succ{} , + , \times, \smsh , |\cdot| , \leq \}$.
2
Our base language is $\{ 0, \succ{} , + , \times, \smsh , |\cdot| , \leq \}$. We use classical logic connectives $\neg$, $\cand$, $\cor$, $\forall$, $\exists$. The formula $A \cimp B$ will be a notation for $\neg A \cor B$.
3 3

  
4 4
The $\basic$ axioms are as follows:
5 5
\[
......
11 11
\forall x^\safe . (x = 0 \cor \exists y^\safe.\  x = \succ{} y   )\\
12 12
\forall x^\safe, y^\safe . \safe(x+y)\\
13 13
\forall u^\normal, x^\safe . \safe(u\times x) \\
14
\forall u^\normal , v^\normal . \safe (u \smsh v)
14
\forall u^\normal , v^\normal . \safe (u \smsh v)\\
15
\patrick{\forall u^\normal \safe(u) \; ?}
15 16
\end{array}
16 17
\]
17 18
\anupam{in fact, we use essentially the same language, so just take Buss' Basic axioms after proper typing. Should also add the symbol $\hlf{\cdot}$ for binary predecessor then we have the full language of bounded arithmetic.}
18 19

  
19 20

  
20

  
21
Notation: if $\vec t=t_0,\dots, t_k$, we will denote as $\safe(\vec t)$ the sequence of formulas $\safe(t_0),\dots, \safe(t_k)$. Similarly for $\normal(\vec t)$.
22
  
21 23
\begin{definition}
22 24
[Derived functions and notations]
23 25
We write $1,2,3,\dots$ for the terms $\succ{} 0, \succ{} \succ{} 0, \succ{} \succ{} \succ{} 0 \dots$, and frequently omit the $\times$ symbol.
......
69 71
\[
70 72
 \dfrac{\forall \vec x^\normal . \exists  y^\safe .  A }{ \forall \vec x^\normal .\exists y^\normal . A}
71 73
\]
74
\patrick{For $\forall \vec x^\normal . \exists  y^\safe .  A$ closed ?} 
72 75
\end{definition}
73 76
\anupam{In induction,for inductive cases, need $u\neq 0$ for $\succ 0$ case.}
74 77

  
......
102 105

  
103 106

  
104 107
\subsection{A sequent calculus presentation}
105
\todo{Write out usual first-order sequent calculus}
106 108

  
107 109
\begin{figure}
108 110
\[
......
132 134
\vliinf{\rigrul{\cand}}{}{\Gamma \seqar \Delta, A \cand B }{\Gamma \seqar \Delta, A}{\Gamma \seqar \Delta, B}
133 135
\\
134 136
\noalign{\bigskip}
135
\vliinf{\lefrul{\cimp}}{}{\Gamma, A \cimp B \seqar \Delta}{\Gamma \seqar A, \Delta}{\Gamma, B \seqar \Delta}
136
&
137

  
137 138
\vlinf{\lefrul{\neg}}{}{\Gamma, \neg A \seqar \Delta}{\Gamma \seqar A, \Delta}
138 139
&
139 140

  
140
\vlinf{\rigrul{\cimp}}{}{\Gamma \seqar \Delta, A \cimp B}{\Gamma, A \seqar \Delta,  B}
141
\vlinf{\lefrul{\neg}}{}{\Gamma, \seqar \neg A, \Delta}{\Gamma, A \seqar  \Delta}
141 142
&
143
&
144
%\vliinf{\lefrul{\cimp}}{}{\Gamma, A \cimp B \seqar \Delta}{\Gamma \seqar A, \Delta}{\Gamma, B \seqar \Delta}
145
%&
146
%
147
%\vlinf{\rigrul{\cimp}}{}{\Gamma \seqar \Delta, A \cimp B}{\Gamma, A \seqar \Delta,  B}
142 148

  
143
\vlinf{\lefrul{\neg}}{}{\Gamma, \seqar \neg A, \Delta}{\Gamma, A \seqar  \Delta}
149

  
144 150
\\
145 151

  
146 152
\noalign{\bigskip}
......
218 224
Similarly the $\rais$ inference rule of Def. \ref{def:ariththeory} is represented by the nonlogical rule:
219 225
 \[
220 226
 \begin{array}{cc}
221
    \vlinf{\rais}{}{  \Gamma \seqar \forall \vec x^\normal . \exists  y^\normal .  A , \Pi  }{ \Gamma \seqar \forall \vec x^\normal . \exists  y^\safe .  A,  \Pi }
227
    \vlinf{\rais}{}{  \normal(t_1), \dots, \normal(t_k) \seqar  \exists  y^\normal .  A }{  \normal(t_1), \dots, \normal(t_k) \seqar \exists  y^\safe .  A}
222 228
\end{array}
223 229
\]
224
 The sequent calculus for $\arith^i$ is that of Fig. \ref{fig:sequentcalculus} extended with nonlogical rules for the axioms of $\basic$,  $\cpind{\Sigma^\safe_i } $ and $\rais$.
230

  
231
%\patrick{In fact, I think we rather need the following nonlogical rule, which implies the previous one but is I guess more general:
232
%\[
233
% \begin{array}{cc}
234
%    \vlinf{\rais}{}{  \normal(t_1), \dots, \normal(t_k) \seqar  \normal(t) }{  \normal(t_1), \dots, \normal(t_k) \seqar \safe(t)}
235
%\end{array}
236
%\]
237
%}
238

  
239
The $\basic$ axioms are equivalent to the following nonlogical rules, that we will also designate by $\basic$:
240
\[
241
\small
242
\begin{array}{l}
243
\begin{array}{cccc}
244
\vlinf{}{}{\seqar \safe (0)}{}&
245
\vlinf{}{}{\safe(t) \seqar \safe(\succ{} t)}{}&
246
\vlinf{}{}{ \safe (t)   \seqar 0 \neq \succ{} t}{} &
247
\vlinf{}{}{\safe (s) , \safe (t)  , \succ{} s = \succ{} t\seqar s = t }{}\\
248
\end{array}
249
\\
250
\vlinf{}{}{\safe (t) \seqar t = 0 \cor \exists y^\safe . t = \succ{} y  }{}
251
\qquad
252
\vlinf{}{}{\safe(s), \safe(t) \seqar \safe(s+t) }{}\\
253
\vlinf{}{}{\normal (s), \safe(t) \seqar \safe(s \times t)  }{}
254
\qquad
255
\vlinf{}{}{\normal (s), \normal(t) \seqar \safe(s \smsh t)  }{}\\
256
\patrick{\vlinf{}{}{\normal(t) \seqar \safe(t)  }{}}
257
\end{array}
258
\]
259

  
260
 The sequent calculus for $\arith^i$ is that of Fig. \ref{fig:sequentcalculus} extended with the $\basic$,  $\cpind{\Sigma^\safe_i } $ and $\rais$ nonlogical rules.
261
 
262
 \begin{lemma}
263
 For any term $t$, its free variables can be split in two sets $\vec{x}$ and $\vec{y}$ such  that the sequent $\normal(\vec x), \safe(\vec y) \seqar \safe(t)$ is provable. 
264
 \end{lemma}
265
 
225 266
\subsection{Free-cut free normal form of proofs}
226 267
\todo{State theorem, with references (Takeuti, Cook-Nguyen) and present the important corollaries for this work.}
227 268

  
......
243 284
   
244 285
   Now we have (see \cite{Takeuti87}): 
245 286
   \begin{theorem}
246
   [Free-cut elimination]
287
   [Free-cut elimination]\label{thm:freecutelimination}
247 288
   \label{thm:free-cut-elim}
248 289
    Given a system  $\mathcal{S}$, any  $\mathcal{S}$-proof $\pi$ can be transformed into a $\system$-proof $\pi'$ with same end sequent and without any free-cut.
249 290
   \end{theorem}
250
%   \begin{proposition}
251
%   
252
%   \end{proposition}
291
   Now we want to deduce from that theorem a normal form property for proofs of certain formulas. But before that let us define some particular classes of sequents and proofs.
292
   
293
   Say that a sequent $\Gamma \seqar \Delta$ is \textit{well-typed} if for any free variable $x$ occurring in $\Gamma$ or $\Delta$, there exists a formula $\safe(x)$ or $\normal(x)$ in $\Gamma$. A proof is well-typed if its sequence are.
294
   
295
   \begin{lemma}\label{lem:welltyped}
296
   If a well-typed sequent $\Gamma \seqar \Delta$ is provable, then it admits a well-typed proof.
297
   \end{lemma}
298
   \begin{proof}
299
   First by Thm \ref{thm:freecutelimination} we know that $\Gamma \seqar \Delta$ admits a proof $\pi$ without any free-cut. Let us then prove that $\pi$ can be transformed in a proof $\pi'$ of same conclusion and such that, for any sequent, if it is well-typed then its premises are well-typed.
300
   
301
   \patrick{to be finished}
302
     \end{proof}
303
     
304
     \patrick{ TODO: define integer-positive sequents and proofs.}
305
     
306
 \begin{theorem}
307
   Assume the $\arith^i$ sequent calculus proves a closed formula $\forall \vec u^\normal . \forall \vec x^\safe . \exists y^\safe . A(\vec u ; \vec x , y)$. Then there exists a proof $\pi$ of the sequent 
308
   $\normal(\vec u), \safe(\vec x) \seqar \exists y^\safe . A(\vec u ; \vec x , y)$ satisfying:
309
   \begin{enumerate}
310
    \item $\pi$  only contains  $\Sigma^\safe_{i}$ formulas,
311
    \item $\pi$ is a well-typed and integer-positive proof. 
312
   \end{enumerate}
313
   \end{theorem}

Formats disponibles : Unified diff