This post explains how the infimum distributes over unions of sets, and how this leads to clean formulas for minima when additional assumptions are present. We also highlight a standard counterexample showing where the minimum identity fails for infinite unions.
1. Main Result
Given a family of nonempty subsets {Ai}i∈I of R, the following distribution identity holds:
inf(i∈I⋃Ai)=i∈Iinf(infAi).
Proof. Let x=inf(∪i∈IAi),y=i∈IinfinfAi.
Since Ai⊂∪i∈IAi for all i∈I, thus, infAi≥x for all i∈I. Therefore y≥x.
For any a∈∪i∈IAi, there exists Ai containing a, thus infAi≤a. Thus, y≤a for any a, i.e. y is a lower bound for ∪i∈IAi, hence y≤x.
Thus x=y.
2. Consequences
(i) Finite index set
When I is finite, an infimum of a finite set of real numbers is actually a minimum. Therefore:
inf(i∈I⋃Ai)=i∈Imininf(Ai).
(ii) Closed and lower-bounded sets
If I is finite and each Ai is closed and bounded below, then each infAi is attained, so:
infAi=minAi
and thus (⋃i∈IAi is also closed and bounded below, thus inf(i∈I⋃Ai)=min(i∈I⋃Ai).
Hence:
min(i∈I⋃Ai)=i∈IminminAi.
3. Counterexample for Infinite Index Sets
For infinite I, the following identity does not hold
min(i∈I⋃Ai)=i∈IinfminAi.
The problem is that the minima may fail to exist even if each Ai has one. Indeed, let us consider:
Ai={1/i}.
Each Ai has a minimum equal 1/n, thus the right-hand-side is 0.
However, the union is:
i≥1⋃Ai={1,1/2,1/3,…}.
Its infimum is 0, but the min does not exist.
Conclusion
The identity
inf(i∈I⋃Ai)=i∈Iinf(infAi)
is always valid. However, replacing infimum by minimum requires additional assumptions—most notably finiteness of the index set and closedness of the sets involved. Infinite unions can easily destroy the existence of a minimum, even when every individual set has one.