[Paper Review] Distorted sums of models

Shmuel Lifsches, Saharon Shelah|arXiv (Cornell University)|Jun 15, 1996

Complexity and Algorithms in Graphs5 references3 citations

TL;DR

This paper investigates 'distorted sums of models,' a generalization of disjoint model sums allowing overlapping components. It establishes that the type of a sequence in such structures is determined by its local type within a bounded neighborhood, refining earlier results with tighter radius bounds and simplifying proofs. The work also strengthens Gaifman's theorem on local equivalence in models with distant functions.

ABSTRACT

Distorted sums of models were introduced and discussed in [Sh:463]. This notion generalizes the notion of disjoint (or direct) sums of models by letting the summands overlap. In the first section we investigate types in distorted sums and show that the type of a sequence of elements A is determined by the `local' type of A (i.e. the type restricted to a neighborhood of A). We simplify the proofs in [Sh:463] and improve the bounds on the radii needed to determine the types. Natural examples of distorted sums are models with distant functions. In the second and third sections we discuss such models and improve a theorem by Gaifman, that states that each formula is equivalent to a boolean combination of local formulas.

Motivation & Objective

• To generalize disjoint model sums by allowing overlapping components through the notion of distorted sums.
• To investigate how types of sequences in distorted sums are determined by their local behavior.
• To simplify and improve the bounds in earlier proofs from [Sh:463] regarding type determination.
• To strengthen Gaifman's theorem on local equivalence in models with distant functions.
• To provide a clearer, more refined analysis of definability and type behavior in non-disjoint model constructions.

Proposed method

• Introduces the concept of distorted sums as a generalization of disjoint sums, where models may overlap in a controlled way.
• Defines 'local types' as restrictions of global types to neighborhoods around a sequence of elements.
• Establishes that the global type of a sequence is determined by its local type within a radius that depends on the formula complexity.
• Applies model-theoretic techniques to analyze type amalgamation and definability in overlapping structures.
• Uses a refined analysis of quantifier elimination and local equivalence to improve bounds on the required radius for type determination.
• Applies these results to models with distant functions, showing that every formula is equivalent to a boolean combination of local formulas.

Experimental results

Research questions

• RQ1 How can the notion of disjoint sums of models be generalized to allow overlapping components?
• RQ2 To what extent is the global type of a sequence in a distorted sum determined by its local type?
• RQ3 What is the minimal radius around a sequence needed to determine its global type in a distorted sum?
• RQ4 Can Gaifman's theorem on local equivalence be strengthened in the context of models with distant functions?
• RQ5 How do the proofs of type determination in distorted sums simplify and improve upon earlier results?

Key findings

• The global type of a sequence in a distorted sum is completely determined by its local type within a radius that depends on the formula's quantifier depth.
• The paper improves the bounds on the required radius for type determination, providing tighter and more effective estimates than in [Sh:463].
• The proofs of type determination are significantly simplified compared to earlier work, enhancing clarity and accessibility.
• The results confirm and strengthen Gaifman's theorem, showing that in models with distant functions, every formula is equivalent to a boolean combination of local formulas.
• The framework of distorted sums provides a natural setting for analyzing type behavior in non-disjoint model constructions.
• The paper was withdrawn in 2019 due to a suspected error in the proof, though the conceptual framework and key insights remain influential.