Provable by Gentzen Proof System iff Negation has Closed Tableau/Set of Formulas

From ProofWiki
Jump to navigation Jump to search

Theorem

Let $\mathscr G$ be instance 1 of a Gentzen proof system.

Let $U$ be a set of WFFs of propositional logic.

Then $U$ is a $\mathscr G$-theorem iff:

$\bar U$ has a closed semantic tableau

where $\bar U$ is the set comprising the logical complements of all WFFs in $U$.


Proof

Denote with $\bar{\mathbf A}$ and $\bar U$ the logical complement of a WFF $\mathbf A$ and set $U$, respectively.


Necessary Condition

We aim to prove this result using the Second Principle of Mathematical Induction, applied to the minimal length of a formal proof of $U$.


Suppose first that $U$ is an axiom of $\mathscr G$.

Then $U$ contains a complementary pair of literals.

Hence, so does $\bar U$, by definition of logical complement.


Therefore, the semantic tableau comprising only a root node labeled $\bar U$ is closed.


Next, suppose that the last step in proving $U$ was an instance of the $\alpha$-rule.

That is, for some $\alpha$-formula $\mathbf A$ and corresponding $\mathbf A_1, \mathbf A_2$:

$U = U_1 \cup U_2 \cup \left\{{\mathbf A}\right\}$

where $U_1 \cup \left\{{\mathbf A_1}\right\}$ and $U_2 \cup \left\{{\mathbf A_2}\right\}$ are $\mathscr G$-theorems.

By induction hypothesis, the logical complements:

$\bar U_1 \cup \left\{{\bar{\mathbf A}_1}\right\}$
$\bar U_2 \cup \left\{{\bar{\mathbf A}[email protected]}\right\}$

of these theorems have closed tableaus.

From the Soundness Theorem for Semantic Tableaus, it follows that these sets are unsatisfiable.

Now by Superset of Unsatisfiable Set is Unsatisfiable, so are:

$\bar U' := \bar U_1 \cup \bar U_2 \cup \left\{{\bar{\mathbf A}_1}\right\}$
$\bar U'' := \bar U_1 \cup \bar U_2 \cup \left\{{\bar{\mathbf A}_1}\right\}$

By the Completeness Theorem for Semantic Tableaus, these have closed tableaus.

Since $\mathbf A$ is an $\alpha$-formula, it is equivalent to $\mathbf A_1 \land \mathbf A_2$.

By De Morgan's Laws, it follows that:

$\bar{\mathbf A}$ is equivalent to $\bar{\mathbf A}_1 \lor \bar{\mathbf A}_2$.

Therefore, $\bar{\mathbf A}$ is a $\beta$-formula, and $\bar{\mathbf A}_1, \bar{\mathbf A}_2$ correspond to it as in the table of $\beta$-formulas.


This allows the Semantic Tableau Algorithm to expand a leaf labeled $\bar U$ into two leaves labeled $\bar U'$ and $\bar U''$.

Because $\bar U'$ and $\bar U''$ have closed tableaus, so does $\bar U$.


Finally, suppose that the last step in proving $U$ was an instance of the $\beta$-rule.

That is, for some $\beta$-formula $\mathbf B$ and corresponding $\mathbf B_1, \mathbf B_2$:

$U = U_1 \cup \left\{{\mathbf B}\right\}$

where $U' := U_1 \cup \left\{{\mathbf B_1, \mathbf B_2}\right\}$ is a $\mathscr G$-theorem.

By induction hypothesis, the logical complement:

$\bar U' = \bar U_1 \cup \left\{{\bar{\mathbf B}_1, \bar{\mathbf B}_2}\right\}$

of this $\mathscr G$-theorem has a closed tableau.

Since $\mathbf B$ is a $\beta$-formula, it is equivalent to $\mathbf B_1 \lor \mathbf B_2$.

By De Morgan's Laws, it follows that:

$\bar{\mathbf B}$ is equivalent to $\bar{\mathbf B}_1 \land \bar{\mathbf B}_2$.

Therefore, $\bar{\mathbf B}$ is an $\alpha$-formula, and $\bar{\mathbf B}_1, \bar{\mathbf B}_2$ correspond to it as in the table of $\alpha$-formulas.


This allows the Semantic Tableau Algorithm to expand a leaf labeled $\bar U$ into a leaf labeled $\bar U'$.

Because $\bar U'$ has a closed tableau, so does $\bar U$.


The result follows by the Second Principle of Mathematical Induction.

$\Box$


Sufficient Condition

Let $T$ be a closed tableau for $\bar U$.

We prove that $U$ is a $\mathscr G$-theorem by the Second Principle of Mathematical Induction, applied to the number of nodes of $T$.


Suppose $T$ has only one node.

Then $\bar U$ has to contain a complementary pair of literals for $T$ to be closed.

But then $U$ also contains a complementary pair.

Hence, it is an axiom of $\mathscr G$.


If $T$ has more than one node, the first iteration of the Semantic Tableau Algorithm must have selected either an $\alpha$-formula $\mathbf A$ or a $\beta$-formula $\mathbf B$ from $\bar U$.


First the case that an $\alpha$-formula $\mathbf A$ was selected.

Define, for convenience, $\bar U' = \bar U \setminus \left\{{\mathbf A}\right\}$.

Then the rest of $T$ is a semantic tableau for $\bar U' \cup \left\{{\mathbf A_1, \mathbf A_2}\right\}$.

This tableau has fewer nodes than $T$.

For $T$ to be closed, this subtree must also be a closed tableau.

By induction hypothesis, this means that $U' \cup \left\{{\bar{\mathbf A}_1, \bar{\mathbf A}_2}\right\}$ is a $\mathscr G$-theorem.


Now, since $\mathbf A$ is an $\alpha$-formula, it is equivalent to $\mathbf A_1 \land \mathbf A_2$.

Hence by De Morgan's Laws, $\bar{\mathbf A}$ is equivalent to $\bar{\mathbf A}_1 \lor \bar{\mathbf A}_2$.

That is, $\bar{\mathbf A}$ is a $\beta$-formula, so that the $\beta$-rule may be applied to it.

Hence, $U = U' \cup \left\{{\bar{\mathbf A}}\right\}$ is a $\mathscr G$-theorem, as desired.


Now the case that a $\beta$-formula $\mathbf B$ was selected.

Define, for convenience, $\bar U' = \bar U \setminus \left\{{\mathbf B}\right\}$.

Then the rest of our closed tableau $T$ comprises two semantic tableaus for $\bar U' \cup \left\{{\mathbf B_1}\right\}$ and $\bar U' \cup \left\{{\mathbf B_2}\right\}$, respectively.

Each of these has fewer nodes than $T$.


For $T$ to be closed, these subtrees must also be closed tableaus.

By induction hypothesis, this means that $U' \cup \left\{{\bar{\mathbf B}_1}\right\}$ and $U' \cup \left\{{\bar{\mathbf B}_2}\right\}$ are $\mathscr G$-theorems.


Now, since $\mathbf B$ is a $\beta$-formula, it is equivalent to $\mathbf B_1 \lor \mathbf B_2$.

Hence by De Morgan's Laws, $\bar{\mathbf B}$ is equivalent to $\bar{\mathbf B}_1 \land \bar{\mathbf B}_2$.

That is, $\bar{\mathbf B}$ is an $\alpha$-formula, so that the $\alpha$-rule may be applied to it.

Hence, $U = U' \cup \left\{{\bar{\mathbf B}}\right\}$ is a $\mathscr G$-theorem, as desired.


The result now follows from the Second Principle of Mathematical Induction.

$\blacksquare$


Sources