Abstract (eng)
Gerhard Gentzen proves the consistency of first-order Peano arithmetic \textbf{PA}. His method works as follows: Define a simple part \textbf{SPA} of peano arithmetic (\textbf{SPA} does in particular not contain induction) and first show the consistency of \textbf{SPA}. Now assume towards a contradiction that \textbf{PA} deducts an contradiction. Show that this deduction can be transformed into a deduction in \textbf{SPA}, this contradicts the consistency of \textbf{SPA}. How to get a deduction in \textbf{SPA}: We assign an ordinal (more exact an ordinal term of an ordinal notation system) to each deduction in \textbf{PA}, called the rank of the deduction. Next show that for each deduction which deducts a contradiction (and is not in \textbf{SPA}) there is a deduction (also deducting and contradiction) with smaller rank.
This method requires that the ordinal notation system (which goes up to $\varepsilon_0$) is well-founded. It turns out that Gentzen's method requires only to the following properties of \textbf{PA}:
\begin{enumerate}
\item All axioms of \textbf{PA} are universal sentences or instances of the induction schema
$$ \varphi(0) \wedge \forall x [\varphi(x) \rightarrow \varphi(x+1)] \rightarrow \forall x \varphi(x).$$
\item All closed terms are provable equal to a term build up just from $0$ and the symbol of the successor function.
\end{enumerate}
This allows a slight generalisation of Gentzen's method.
In this Diploma Thesis we consider theories $T_f = (T_f)_0 \cup (T_f)_{\text{Ind}}$ with the following properties:
\begin{enumerate}
\item $(T_f)_0$ contains only universal sentences.
\item $(T_f)_{\text{Ind}}$ contains all instances of the general induction schema
$$\varphi(c_1)\wedge ... \wedge \varphi(c_m) \wedge \forall \vec{x} [\varphi(x_1)\wedge...\wedge \varphi(x_n) \rightarrow \varphi(f(x_1,...,x_n))] \rightarrow \forall x \varphi(x).$$
\item The simple part of $T_f$ proves for every closed term $t$ the equality of $t$ to a term $\bar{t}$ build up just from $c_1,...,c_m$ and $f$.
\end{enumerate}
As Gentzen did for \textbf{PA}, the consistency of $T_f$ can be shown with respect to their simple part which corresponds to the simple part of Gentzen (also without induction). As a consequence, one gets the following result for all such theories.
\begin{cor2}
Assume $\varphi(a)$ is quantifier free and $T_f$ consistent.\\
If $T_f \models \exists x \varphi(x)$, then $(T_f)_0 \models \exists x \varphi(x)$.\\
I.e., $T_f$ is $\Sigma_1^0$-conservative over $(T_f)_0$.
\end{cor2}
It seems that this method is different in an essential way to the method Kurt Sch\"{u}tte uses in his consistency proof of \textbf{PA}. Sch\"{u}tte, Tait and others uses calculi with infinite deduction rules. These methods compute, in some sense, the proof theoretical ordinal of the considered theory by embedding the deductions of the theory (in ordinary first-order logic) in an infinite system which allows cut-elimination. In contrast to Gentzen's method Sch\"{u}tte's and Tait's methods are closely related to the proof theoretical ordinals.\\
We do not provide an analysis of the disparities of both methods. Instead we present the point of view Wolfram Pohlers propose, to measure the $\Pi_1^1$-ordinal of theories $\mathsf{TA}_f$ (presented as a Tait-calculus) satisfying the following conditions:
\begin{enumerate}
\item $\mathsf{TA}_f$ includes all defining axioms for primitive recursive functions.
\item All instances of the schema $$\varphi(c_1)\wedge ... \wedge \varphi(c_m) \wedge \forall \vec{x} [\varphi(x_1)\wedge...\wedge \varphi(x_n) \rightarrow \varphi(f(x_1,...,x_n))] \rightarrow \forall x \varphi(x)$$ are included. Here $f$ is an $m$-array primitive recursive function constant and $c_1,...,c_l$ are individual constants.
\item Every $n \in \mathbb{N}$ is equal to a composition of $f^\mathbb{N}$ and the elements $c_1^\mathbb{N},...,c_l^\mathbb{N}$.
\end{enumerate}