Properties of \( \lambda \)-Terms

Computer Science Foundations: Normalisation & Church–Rosser

Definitions

1. Normalisation Concepts

Let \( r \in \{\beta, \eta, \beta\eta\} \).

Lemma. If \( A \) is strongly \( r \)-normalising, then \( A \) is weakly \( r \)-normalising and has an \( r \)-normal form.
Theorems

2. Uniqueness and Confluence

The Church–Rosser Property

If a term reduces to \( B_1 \) and \( B_2 \), then \( B_1 \equiv_\alpha B_2 \). Hence, if a normal form exists, it is unique.

Confluence: Different reduction orders lead to the same result if they terminate.
Case Studies

3. Reduction Behaviour

A. Non-terminating Terms

The term \( (\lambda x.\, x x)(\lambda x.\, x x) \) is not weakly \( \beta \)-normalising.

B. Growing Terms

\[ (\lambda x.\, x x x)(\lambda x.\, x x x) \;\to_\beta\; (\lambda x.\, x x x)(\lambda x.\, x x x)(\lambda x.\, x x x) \;\to_\beta\; \dots \]

C. Path Dependency

Appendix

4. Strategy & Notation

A safe reduction strategy guarantees finding the normal form whenever one exists.

All reduction orders of \( (\lambda x y z.\, x z (y z))(\lambda x.\, x)(\lambda x.\, x) \) converge to: \[ \lambda z.\, z z \]
Left associativity: \( x z (y z) \) means \( (x z)(y z) \).