The notation for families takes some getting used to, since a function is written like an element and its values like subscripts. Consider a function . In this setting we call the index set and each an index; the range of the function is called the indexed set; the function itself is called a family; and the value of at is called a term of the family, denoted . We write , or simply , for the family, and call it a family in . In particular, a family of subsets of is a function , . (Strange as this notation is at first, it soon becomes the preferable one.)
Definition 1.4.1 (Unions and intersections of families). Let be a family of sets. Its union and, when , its intersection are
When we write , and similarly for intersections, just as for sums and products.
The range of a family is the collection , and its union is . Every collection of sets is the range of some family: take and the family , . So nothing is lost by working with families, and the index set is a convenient bookkeeping device.
Example 1.4.2. For the index set we have . Another example may be given as follows: let and , so that , , , and so on. The union of the is , and their intersection is .
Example 1.4.3. Let us now consider an infinite index set in the plane: and . Each is a vertical segment. Geometrically, the union of the is the rectangle ; their intersection is , since the segments are parallel and pairwise disjoint.
Theorem 1.4.4 (Associative law of unions). Let be a family of sets with , and let be a family of sets. Then If and every are nonempty, the same holds with intersections in place of unions.
Proof. An element lies in the left side if and only if for some , that is, for some and some . This says exactly that for some , which is membership in the right side. For intersections, lies in for every if and only if, for every , it lies in for every . ∎
Proposition 1.4.5 (De Morgan’s laws and distributive laws). Let and be families of subsets of a set , with and nonempty. Then
Proof. An element of lies in no if and only if it lies in every , and it fails to lie in every if and only if it lies in some ; these are De Morgan’s laws. For the first distributive law, lies in the left side if and only if for some and for some , that is, for some pair . For the second, if lies in every or in every , then for every pair . Conversely, suppose for every pair and is not in every , say . Then forces for every . ∎
Families also let us form products of more than two sets.
Definition 1.4.6 (Products of families). Let be a family of sets. Their (Cartesian) product is the set of all families with for every ; that is, of all functions with for each . We also write its elements as . When every is the same set , the product is , the set of all functions .
This generalizes the Cartesian product of two sets. Let with , and let and . The map , , is a bijection, since a family indexed by is determined by its two terms and any choice of the two terms defines one. In the same way, an -tuple is a family indexed by , and ; an infinite sequence in is a family indexed by , an element of . There is more one can say about families of sets, sets of families, and their products, but we will not need it.