Questions: The Constructible Universe

Question 1

A set theorist claims that L_{ω+1} contains all subsets of ω, just as V_{ω+1} does. Why is this claim incorrect?

• A. L_{ω+1} doesn't exist because ω+1 is not a well-defined ordinal in ZF
• B. L_{ω+1} contains only the countably many subsets of ω that are first-order definable from elements of L_ω — not all subsets
• C. L_{ω+1} equals V_{ω+1} at successor stages; the difference only appears at limit ordinals
• D. Subsets of ω are not sets in L because ω itself is not constructible

Answer: (B)

This is the key distinction between L and V. V_{ω+1} = P(ω) contains all subsets of ω — an uncountable collection. L_{ω+1} = Def(L_ω) contains only those subsets of ω definable by a first-order formula with parameters from L_ω. Since there are only countably many first-order formulas and countably many elements of L_ω to use as parameters, L_{ω+1} is countable. This is where L first becomes dramatically 'thinner' than V, and it illustrates why the definability restriction has enormous consequences.

Question 2

What is the correct statement of Gödel's consistency result about the constructible universe?

• A. AC and GCH are theorems of ZF — they hold in every model of ZF, as Gödel proved
• B. If ZF is consistent, then ZFC + GCH is consistent, because L is a model of ZF in which both AC and GCH hold
• C. V = L is provably true in ZFC, so every set is automatically constructible
• D. The continuum hypothesis is false in L because L restricts the power set of ω

Answer: (B)

Gödel's 1938 result is a relative consistency proof: he showed that L is an inner model satisfying all ZFC axioms plus GCH. Therefore, if ZF has a model at all (is consistent), then ZFC + GCH also has a model (namely L). This does not say AC or GCH are theorems — they are independent of ZF. Option A is the classic misstatement. Option C is also wrong: V = L is an additional axiom, not a theorem. Cohen later proved V ≠ L is also consistent (via forcing), showing both directions are independent.

Question 3

The constructible universe L is 'thin' relative to V because at each successor stage it takes only first-order definable subsets rather than all subsets, while still containing all the ordinals.

• T. True
• F. False

Answer: True

L has exactly the same ordinal height as V — it is 'thin' in density at each stage, not in reach. L₀ = ∅, and at each stage L_{α+1} adds only those subsets of L_α that are first-order definable, whereas V_{α+1} = P(V_α) adds all subsets. For countable α, L_{α+1} is countable while V_{α+1} is uncountable. But since L contains all ordinals (every ordinal α is constructible), L is unbounded — it just has far fewer sets at each level. This is the precise meaning of 'minimal inner model.'

Question 4

V = L is a theorem of ZFC — it can be derived from the standard axioms that most set is constructible.

• T. True
• F. False

Answer: False

V = L is an independent statement that cannot be proved or disproved from ZFC. Gödel showed ZFC + (V = L) is consistent (by constructing L); Cohen showed ZFC + (V ≠ L) is also consistent (by forcing). Most set theorists do not accept V = L as true, because it is incompatible with measurable cardinals and other large cardinal axioms that are widely studied. V = L is best understood as an additional axiom that produces the minimal possible set-theoretic universe — not a consequence of ZFC.

Question 5

Why does the constructible universe L automatically satisfy the axiom of choice, even though AC is not provable from ZF alone?

Model answer: L carries a canonical definable well-ordering of all its sets: order them first by the stage L_α at which they first appear, then by the Gödel number of the formula that defines them, then lexicographically by their parameters from L_α. This construction is explicit, total, and absolute — it well-orders the entire universe L without invoking any choice principle. The axiom of choice merely asserts that well-orderings or choice functions exist; L provides a concrete, definable one. This shows AC is not contradictory with ZF — it fails only in models that lack any such canonical ordering.

The same definable order is why GCH holds in L: at each stage, the 'power set' operation in L adds only definably-many new subsets, keeping the cardinality of L_{α+1} as controlled as possible. The constructibility restriction simultaneously gives you AC (via the canonical order) and GCH (via controlled power sets) — both for the same underlying reason: L has just enough sets to be a model of ZFC, and no more.