Research notes on every problem, and a library of the papers behind them. Built from the open erdos repository.
Updated
Source. Hall (1935), Theorem 2 and its proof, printed p. 29 (canonical PDF).
Statement. Let be a partition of a set , and let be a finite indexed family of subsets of , with . There are points belonging to pairwise distinct classes of if and only if, for every ,
There is no finiteness assumption on , on the , or on . The represented family has only indices. Repeated values among the are permitted. The resulting points themselves are distinct because their partition classes are disjoint.
Proof. Necessity follows because the points with indices in belong to distinct classes, each meeting .
For sufficiency define, for each , a subset of the set by
For every , the union is exactly the set of classes in (1): a class meets a union precisely when it meets at least one member. Thus (1) is Hall's union condition for the finite indexed family . By Theorem 1, there are pairwise distinct classes . For each of the finitely many indices define
Choose successively for . This requires only finite selection from nonempty sets. Then , and the distinct selected classes are pairwise disjoint, so the points lie in different classes and are themselves pairwise distinct. When all assignments and choices are empty and the conclusion holds.
Source precision. The source applies Theorem 1 to its class sets and then names the selected classes after relabeling. Its displayed intersection is written , but the following sentence calls it . The consistent indexed notation above repairs this mismatch. It does not change a hypothesis or conclusion and is not attributed to an author-issued erratum. The numbered source statement gives sufficiency; the reverse direction written here is the immediate necessary condition.
Used by. Theorem 3. The definitions and finite-index qualifications are recorded on the definitions page.
Bears on. No problem. The Problem 126 argument cites only Theorem 1, not this partition form.