ProofComplete

Proof of König's Theorem

König's theorem is the most powerful general inequality in cardinal arithmetic, implying Cantor's theorem as a special case and restricting the behavior of the exponential function.


Statement

Theorem5.4König's theorem

Let II be an index set and {κi}iI\{\kappa_i\}_{i \in I}, {λi}iI\{\lambda_i\}_{i \in I} be families of cardinals with κi<λi\kappa_i < \lambda_i for all iIi \in I. Then:

iIκi<iIλi.\sum_{i \in I} \kappa_i < \prod_{i \in I} \lambda_i.


Proof

Proof

We need to show: (a) κiλi\sum \kappa_i \leq \prod \lambda_i, and (b) κiλi\sum \kappa_i \neq \prod \lambda_i.

Part (a): For each ii, fix an injection fi:κiλif_i: \kappa_i \hookrightarrow \lambda_i (which exists since κi<λi\kappa_i < \lambda_i). We construct an injection from iIκi\bigsqcup_{i \in I} \kappa_i into iIλi\prod_{i \in I} \lambda_i.

For each (i0,α)iIκi(i_0, \alpha) \in \bigsqcup_{i \in I} \kappa_i (where α<κi0\alpha < \kappa_{i_0}), define a function gi0,α:Iλig_{i_0, \alpha}: I \to \bigcup \lambda_i by:

gi0,α(i)={fi0(α)if i=i0,0if ii0.g_{i_0, \alpha}(i) = \begin{cases} f_{i_0}(\alpha) & \text{if } i = i_0, \\ 0 & \text{if } i \neq i_0. \end{cases}

The map (i0,α)gi0,α(i_0, \alpha) \mapsto g_{i_0, \alpha} is injective: if gi0,α=gj0,βg_{i_0, \alpha} = g_{j_0, \beta}, then i0=j0i_0 = j_0 (the only coordinate that can be nonzero) and fi0(α)=fi0(β)f_{i_0}(\alpha) = f_{i_0}(\beta), so α=β\alpha = \beta.

Part (b): Suppose for contradiction that h:iIκiiIλih: \bigsqcup_{i \in I} \kappa_i \to \prod_{i \in I} \lambda_i is a surjection. For each iIi \in I, consider the "slice" Si={h(i,α)(i):α<κi}S_i = \{h(i, \alpha)(i) : \alpha < \kappa_i\}. This is the set of values in λi\lambda_i that appear as the ii-th coordinate of some h(i,α)h(i, \alpha).

Since Siκi<λi|S_i| \leq \kappa_i < \lambda_i, there exists βiλiSi\beta_i \in \lambda_i \setminus S_i for each ii.

Define a function g:Iλig^*: I \to \bigcup \lambda_i by g(i)=βig^*(i) = \beta_i. Then giIλig^* \in \prod_{i \in I} \lambda_i.

By surjectivity, g=h(j,γ)g^* = h(j, \gamma) for some jIj \in I, γ<κj\gamma < \kappa_j. Then g(j)=h(j,γ)(j)Sjg^*(j) = h(j, \gamma)(j) \in S_j. But g(j)=βjSjg^*(j) = \beta_j \notin S_j by construction. Contradiction. \blacksquare


Corollaries

ExampleImportant consequences

Cantor's theorem as a special case: Take I=AI = A, κi=1\kappa_i = 1, λi=2\lambda_i = 2 for all ii. Then 1=A\sum 1 = |A| and 2=2A\prod 2 = 2^{|A|}, so A<2A|A| < 2^{|A|}.

Cofinality restriction: Taking I=cf(κ)I = \mathrm{cf}(\kappa) and suitable sequences: cf(2κ)>κ\mathrm{cf}(2^\kappa) > \kappa. In particular:

  • cf(c)>0\mathrm{cf}(\mathfrak{c}) > \aleph_0, so cω\mathfrak{c} \neq \aleph_\omega.
  • 21ω12^{\aleph_1} \neq \aleph_{\omega_1} (since cf(ω1)=ω1=1\mathrm{cf}(\aleph_{\omega_1}) = \omega_1 = \aleph_1).
RemarkKönig's theorem and pcf theory

König's theorem gives an upper bound on κi\sum \kappa_i in terms of λi\prod \lambda_i. Shelah's pcf theory (possible cofinalities) provides much deeper results about products of cardinals, particularly the remarkable theorem that if ω\aleph_\omega is a strong limit, then 2ω<ω42^{\aleph_\omega} < \aleph_{\omega_4}. This was a breakthrough showing that the exponential function at singular cardinals is far more constrained than at regular cardinals.