Sums of sets inherit their suprema term by term, but products and intersections need more care. Multiplication preserves order only for nonnegative factors, and an intersection must satisfy two membership conditions at once. This note proves the product rule under a sign hypothesis, shows what goes wrong without it, and compares unions, intersections and sums on one concrete pair of sets. It builds on the companion note on suprema and set operations.
Criteria to keep in view
- A supremum proof needs an upper bound and arbitrary-accuracy witnesses below it; an infimum proof needs the reversed pair of conditions.
- Nonempty bounded sets have finite real suprema and infima.
- Multiplication is order-preserving for nonnegative factors. A negative factor reverses the inequality.
- A union's supremum is the larger of the separate suprema. An intersection is a subset of each source set and can have a strictly smaller supremum.
- The supremum of an independent set sum is the sum of the separate suprema; attainment needs both source suprema to be attained.
Products of nonnegative sets
For sets write .
Theorem 8.3.1
(Bounds of a product of nonnegative sets)
Let be nonempty bounded subsets of , with , , and . Then
Proof
(Product estimates with an explicit tolerance)
Nonnegativity gives and , so for every product. Fix and put
Choose with and with . Since ,
So products approach from below, which proves leastness. Neither endpoint had to belong to its source set. The added keeps positive even when ; in that case every element of and of is zero and the formula reads .
For the infimum, every product is at least , since all factors are nonnegative. Given , put and select with and with . Then
This also works when a lower endpoint is zero, and it does not assume that minima exist.
Proof X-ray
Why the tolerance is shared out this way
The non-obvious move is the rewriting . It splits the error of a product into two terms, each a fixed bound times one small gap: multiplies the gap in , and the element multiplies the gap in . Nonnegativity is used twice: it gives the upper bound , and it makes and nonnegative multipliers, so the inequalities and survive multiplication. The bound , which holds because , then replaces the second coefficient by the fixed number . The tolerance is chosen so that the two coefficients and together cost less than ; the only guarantees a positive denominator. The argument closes without any limit theorem, because both witnesses come from the approximation criterion.
If negative elements are allowed, the coefficients and can be negative, the multiplications reverse the inequalities, and the bound itself fails.
The sign hypothesis cannot simply be dropped. The next example shows the formula failing and then fixes the correct values.
Counterexample mode
The product rule needs nonnegative sets
Take and the claim . The right side is . But and , and , so is not even an upper bound of . The nonnegative formula fails because order reverses on negative numbers. The true values are and , as the worked example below shows: the roles of the two endpoint products are interchanged.
Unions, intersections and sums
Theorem 8.3.2
(Union and intersection bounds)
Let be nonempty and bounded. Then
If moreover , then
Proof
(Union by the larger set, intersection by inclusion)
The number bounds every member of the union. Choose a source set whose supremum is this larger value and use its approximation witnesses; they also belong to the union. Reversing inequalities proves the infimum formula.
For a nonempty intersection, every upper bound of bounds , and likewise for , so is at most each separate supremum. The same argument with lower bounds gives the infimum inequality. Nonempty boundedness guarantees that these real bounds exist. An empty intersection has no real supremum or infimum in this convention, so nonemptiness must be checked first.
The inequalities for intersections cannot be sharpened without extra information about common points, as the following pair shows.
Counterexample mode
Equal separate bounds do not fix an intersection
Let and . Both sets have supremum and infimum . The claim that is determined by these numbers fails. A positive common element would need for positive integers , which parity forbids, so and . Replace by itself: all separate bounds are unchanged, yet has supremum . The inequality is valid; equality needs common points near the endpoint.
Worked examples
Worked example 8.3.3
(A product set with both endpoints missing)
Let . Every product lies strictly between and , since and the product of two negatives is positive.
For the supremum, write the two values as with . Their product is , which lies below by . Choosing also gives a product within of . For the infimum, write the values as ; the product exceeds by , and gives a product within of . So and , neither attained. The separate suprema multiply to and the separate infima to , so the nonnegative formulas would interchange the actual answers.
Worked example 8.3.4
(Bounds and attainment of two parity-defined sets)
For and as in the counterexample above, zero is the minimum and infimum of each set, because every other element is positive. Each element is less than . Given , choose an integer ; then both and are less than , so the displayed members approach from below. Hence both suprema equal , and neither set has a maximum.
Worked example 8.3.5
(Intersection, union and sum of those sets)
With from the counterexample, the intersection has supremum, infimum, maximum and minimum all equal to . The union has supremum , not attained, and minimum and infimum .
For the sum, the set-sum rule gives supremum and infimum . The infimum is attained by , and since both sets are nonnegative and only has value , this is the only attaining pair. There is no maximum: every sum is strictly below because each summand is strictly below . Sums arbitrarily close to exist by choosing each summand within of .
Common mistakes
Common mistake
(Writing intersection bounds before checking nonemptiness)
For and the intersection is empty, so and do not exist as real numbers, and the inequalities with say nothing. Check that the intersection is nonempty before naming its bounds.
Common mistake
(Treating the union as a sum)
The supremum of is the larger of the two suprema, not their sum. For and the union is with supremum , while the sum set has supremum . Choosing a point from either set and choosing a point from each set are different operations.
Practice
Quick check
(Let and . Find , and whether each is attained.)
Both sets are nonnegative.
Solution · Answer
The product rule gives and . Both are attained, by and , since , , , . So and .
Quick check
(Let . Find and , and say whether has a maximum or minimum.)
Every equals .
Solution · Answer
The product rule gives and . Indeed , since every is with , and products of numbers in stay in . The endpoints are not in the set, so there is no maximum and no minimum.
Quick check
(Let and . Compare with .)
Find first.
Solution · Answer
Here and , so . But , so and . The inequality is strict.