Proof of the Axiom of Union from Replacement
One of the elegant features of ZFC is that some axioms can be derived from others. In particular, the axiom of union can be proven from the axiom schema of replacement together with the other axioms. This demonstrates the power of replacement and shows that union is, in a sense, redundant.
Assuming the axiom schema of replacement, the axiom of pairing, and the axiom schema of comprehension, the axiom of union can be derived. Specifically, for any set , there exists a set .
Let be any set. We want to construct , the union of all elements of . We proceed in several steps.
Step 1: Construct a function on A
For each , the set itself exists. Define a function by for all . This is clearly a function (even the identity function on ), and we can express this in the language of set theory by the formula:
For each , there exists a unique (namely, itself) such that holds.
Step 2: Apply the Axiom Schema of Replacement
By the axiom schema of replacement, since defines a functional relationship on , there exists a set such that:
So far we have only recovered , but this illustrates the mechanism.
Step 3: The Correct Construction
The issue is that we want the elements of the elements of , not the elements of themselves. We use a different approach. For each , consider the ordered pair where . By the axiom of pairing and the axiom schema of comprehension, we can form:
More precisely, we first form (which we are trying to construct, so this seems circular). Let's use a more direct approach.
Step 4: Direct Construction via Comprehension
Actually, we can avoid replacement by using a simpler observation. First, note that every element must be an element of some , and hence must be a subset of some or an element of some element of , continuing downward. By the axiom of foundation, this chain terminates.
But we seek a direct proof. Here is the correct approach:
Define the formula as "". We need a set containing all elements satisfying . The problem is that comprehension requires us to separate from an existing set.
Step 5: Finding a Superset
By replacement, we know that for each , the set exists. Now, we can use the fact that we want to collect elements from these sets. Consider the formula:
By replacement applied to , there exists a set . Now form:
using comprehension on... but again we need a bounding set.
Step 6: The Correct Proof
Here is the standard proof. For each , form the singleton by pairing (taking ). By replacement, the collection forms a set .
Now, for each , extract the unique element by taking the union . By replacement again, forms a set, which is just again.
The actual construction is more subtle. We use the fact that we can form the set:
To make this rigorous without assuming union, define for each the set itself (which exists). Then form the product-like set of all pairs where and . This can be done by first forming a large enough set containing all such values.
Step 7: Using a Bounding Set
The key insight: if is a set, then every element satisfies for some . By replacement applied to the formula that maps each to the set itself, and then taking a "second-order union," we can construct .
Formally: Let (from replacement). For each , we have . The axiom of replacement with the functional relationship "map each to its elements" gives us a set containing all elements of all . By comprehension, we can then separate out exactly those elements:
using the bounding set provided by replacement.
This proof is somewhat technical because we must avoid circular reasoning: we cannot assume union exists when proving union exists. The key is to use replacement to construct a large enough bounding set, then use comprehension to extract exactly the elements we want. In practice, ZFC includes union as a separate axiom for simplicity and clarity, even though it is technically derivable from replacement.
Ernst Zermelo's original 1908 axiomatization included union as a primitive axiom. When Fraenkel and Skolem independently added the replacement schema in the 1920s, it became possible to derive union. However, standard presentations of ZFC retain union as a separate axiom because:
- It is more intuitive and easier to state
- The derivation from replacement is somewhat involved
- Union is used very frequently, and it is convenient to have it as a basic principle
- Pedagogically, it is clearer to present union explicitly
Modern set theory textbooks typically present both axioms, sometimes noting that union is redundant but keeping it for practical reasons.