-
Notifications
You must be signed in to change notification settings - Fork 0
/
Copy pathinterpolation.tex
111 lines (96 loc) · 5.64 KB
/
interpolation.tex
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
Existence of \emph{interpolant} formulas was proved by W. Craig \cite{CraigA} for classical predicate logic. In this section, we prove \emph{Craig's interpolation} theorem for $\ldl$ using a method similar to Maehara \cite{maehara1960interpolation}.
\input{thm/craig}
We can not prove the same result for the fragment $\LDL_*$, because as we observed, the constructed interpolant uses $\supset$ in a case that we would eventually face, even if we are limited to the fragment without $\supset$. However, we can prove a weaker form of the interpolation theorem, called \emph{deductive interpolation}, in which $\vdash$ plays the role that $\Rightarrow$ has in interpolation.
\begin{nota}
For the rest if this section, $\Gamma' \Rightarrow \Delta' \vdash \Gamma \Rightarrow \Delta$ means that there is a proof-tree in $\LDL_*$ for $\Gamma \Rightarrow \Delta$ with assumption $\Gamma' \Rightarrow \Delta'$.
\end{nota}
We state the deductive interpolation now, but we postpone its proof until after developing some more tools that will be used in the proof.
\begin{thm}[Deductie Interpolation for $\ldl_*$]\label{thm:deductive}
For any formulas $A$ and $B$, if $\Rightarrow A \vdash\ \Rightarrow B$ then there exists a formula $C$ such that
\begin{enumerate}[label=(\arabic*)]
\item $\Rightarrow A \vdash\ \Rightarrow C$,
\item $\Rightarrow C \vdash\ \Rightarrow B$ and
\item $V(C) \subseteq V(A) \cap V(B)$.
\end{enumerate}
\end{thm}
As we have seen, induction on the height of proof-trees is our main proof-theoretic tool here in this paper. So we must first make it clear what is the relation between $\vdash$ and $\Rightarrow$. Recall that $\Box A$ is a shorthand for $\top \rightarrow A$. We first define a \emph{variant} of a formula to be the formula with an arbitrary number of $\nabla$ and $\Box$.
\begin{dfn}
The set of \emph{variants} of a formula $A$, is defined inductively as follows:
\begin{enumerate}
\item $A$ is a variant of $A$, and
\item if $B$ is a variant of $A$, then $\nabla B$ and $\Box B$ are also variants of $A$.
\end{enumerate}
We say that a set of formulas $\Gamma$ is a variant of a set of formulas $\Sigma$, if any formula in $\Gamma$ is a variant of some formula in $\Sigma$.
\end{dfn}
To put it simply, variants of a formula $A$ are all possible combinations of $\nabla$ and $\Box$ applied on $A$. One can observe some trivial properties of sets of variants of a formula, like closure under union, or invariance under $V$. We will widely use these properties in the rest of this section.
In order to facilitate working with variants, we can define variations for the rules $R \rightarrow$ and $N$, which use $\Box$ instead of $\nabla$. The following lemma shows the admissibility of these variations.
\begin{lem}\label{lem:box-rules} For any $\Gamma$, $\Delta$, $A$, $B$ and $C$:
\begin{enumerate}
\item If $\LDL_* \vdash \Gamma, A \Rightarrow \Delta$ then $\LDL_* \vdash \Gamma, \nabla \Box A \Rightarrow \Delta$.
\item If $\LDL_* \vdash \Gamma \Rightarrow \Delta$ then $\LDL_* \vdash \Box \Gamma \Rightarrow \Box \Delta$.
\item If $\LDL_* \vdash \nabla \Gamma, \Sigma, A \Rightarrow B$ then $\LDL_* \vdash \Gamma, \Box \Sigma \Rightarrow A \rightarrow B$.
\end{enumerate}
\end{lem}
Before proving the Lemma, we set a convention to use these admissible rules in our proof-trees.
\begin{nota}
We will use the results from parts \1, \2 and \3 of Lemma \ref{lem:box-rules} as admissible rules $L \nabla \Box$, $\Box$ and $R \rightarrow'$, respectively.
\begin{multicols}{3}
\begin{prooftree}
\AXC{$\Gamma, A \Rightarrow \Delta$}
\RightLabel{$L \nabla \Box$}
\UIC{$\Gamma, \nabla \Box A \Rightarrow \Delta$}
\end{prooftree}
\columnbreak
\begin{prooftree}
\AXC{$\Gamma \Rightarrow \Delta$}
\RightLabel{$\Box$}
\UIC{$\Box \Gamma \Rightarrow \Box \Delta$}
\end{prooftree}
\columnbreak
\begin{prooftree}
\AXC{$\nabla \Gamma, \Sigma, A \Rightarrow B$}
\RightLabel{$R \rightarrow'$}
\UIC{$\Gamma, \Box \Sigma \Rightarrow A \rightarrow B$}
\end{prooftree}
\end{multicols}
\end{nota}
Now we prove Lemma \ref{lem:box-rules}
\begin{proof} \quad\\
\1
\begin{prooftree}
\AXC{$\Rightarrow \top$}
\RightLabel{$LW$}
\UIC{$\Gamma, \nabla \Box A \Rightarrow \top$}
\AXC{$\Gamma, A \Rightarrow \Delta$}
\RightLabel{$LW$}
\UIC{$\Gamma, \nabla \Box A, A \Rightarrow \Delta$}
\RightLabel{$L \rightarrow ^n$}
\BIC{$\Gamma, \nabla \Box A \Rightarrow \Delta$}
\end{prooftree}
\begin{multicols}{2}
\2
\begin{prooftree}
\AXC{$\Gamma \Rightarrow \Delta$}
\RightLabel{$L \nabla \Box^{(*)}$} \doubleLine
\UIC{$\nabla \Box \Gamma \Rightarrow \Delta$}
\RightLabel{$LW$}
\UIC{$\nabla \Box \Gamma, \top \Rightarrow \Delta$}
\RightLabel{$R \rightarrow$}
\UIC{$\Box \Gamma \Rightarrow \Box \Delta$}
\end{prooftree}
\columnbreak
\3
\begin{prooftree}
\AXC{$\nabla \Gamma, \Sigma, A \Rightarrow B$}
\RightLabel{$L \nabla \Box^{(*)}$} \doubleLine
\UIC{$\nabla \Gamma, \nabla \Box \Sigma, A \Rightarrow B$}
\RightLabel{$R \rightarrow$}
\UIC{$\Gamma, \Box \Sigma \Rightarrow A \rightarrow B$}
\end{prooftree}
\end{multicols}{3}
\end{proof}
\begin{rem}\label{rem:var-val}
Notice that any variant of a valid formula is also a valid formula. More precisely, we can prove any combination of $\nabla$ and $\Box$ applied on $A$, using the rules $N$ and $\Box$ applied to $\Rightarrow A$.
\end{rem}
The next theorem shows that every sequent provable using a formula as assumption, is also provable using some variant of that formula in the antecedent.
\input{thm/vdash}