Wiki
Wiki

Research notes on every problem, and a library of the papers behind them. Built from the open erdos repository.

Updated


Source. Theorem 2, stated on p. 709 and again on p. 712, proof pp. 712--713, of R. D. Anderson and J. E. Keisler, An example in dimension theory, Proc. Amer. Math. Soc. 18 (1967), no. 4, 709--713, DOI 10.1090/S0002-9939-1967-0215288-0, the edition named on the source card.

Statement

Setting (p. 709). EnE^n is Euclidean nn-space; dim⁡\dim is the (inductive) topological dimension of Hurewicz and Wallman (the paper cites their Dimension theory, p. 24); KsK^s is the product of ss copies of KK and KωK^\omega the product of countably many copies. In section I (p. 709) ω\omega denotes the set of positive integers.

Theorem 2 (p. 712, quoted; the statement on p. 709 has the same content and adds the definitions above). "There exists a set K⊂EnK\subset E^n such that for each positive integer, ss, dim⁡K=dim⁡Ks=dim⁡Kω=n−1\dim K=\dim K^s=\dim K^\omega=n-1."

So for the given nn a single set KK serves every exponent at once: KK, each finite power KsK^s with s≥1s\ge1, and the countable power KωK^\omega all have dimension exactly n−1n-1. The statement leaves nn unquantified; the paper says the construction is "for arbitrary nn" (p. 709), and Theorem 1, whose proof the paper adapts for this one, takes n∈ωn\in\omega.

Context in the paper (p. 709). The paper contrasts the theorem with a result it calls known: if AA and BB are nonvoid separable metric spaces, AA compact and dim⁡B>0\dim B>0, then dim⁡(A×B)≥dim⁡A\dim(A\times B)\ge\dim A, with equality only if dim⁡A=∞\dim A=\infty. It also lists the cases with easy examples: for n=1n=1, a Cantor set or the rationals of the line; for n=2n=2 with the requirement K⊂EnK\subset E^n deleted or relaxed to K⊂En+1K\subset E^{n+1}, the rationals in Hilbert space; and it notes that for n>2n>2 the standard nn-dimensional examples (Hurewicz and Wallman, pp. 29 and 64) contain cells, so their finite powers increase in dimension.

Read depth. Claims checked: both printings of the statement (pp. 709 and 712) and the context on p. 709 were read clause by clause on the page images. The proof was read in outline only; its steps and Lemmas 1--4 were not checked. Nothing here is independently reviewed.

Proof pointer

Pages 712--713, written here in outline. The proof reruns the transfinite construction of Theorem 1 for all exponents at once: for every ss it fixes the countable family of (ns−n)(ns-n)-spheres in EnsE^{ns} of Lemma 3, and at each stage of the induction over the nondegenerate continua of EnE^n it picks the new point outside a union, over all ss, of sets of fewer than c\mathfrak c excluded points, which is still of size below c\mathfrak c. The resulting KK has dim⁡Ks=n−1\dim K^s=n-1 for every ss; Lemma 4 (p. 711: for K⊂EnK\subset E^n, if dim⁡Ks<t\dim K^s<t for all s∈ωs\in\omega then dim⁡Kω<t\dim K^\omega<t) gives dim⁡Kω<n\dim K^\omega<n, and KωK^\omega contains a homeomorphic copy of KK, so dim⁡Kω≥n−1\dim K^\omega\ge n-1.

Dependencies

Theorem 1 and its proof (p. 712), and Lemmas 1, 3 and 4 of the same paper (pp. 710--711), with Lemma 2 (pp. 710--711), which the paper uses in a weakened form "without explicit proof here" (p. 710). Lemma 4's proof cites J. Nagata, Modern dimension theory, Interscience, 1965, p. 126.

Bears on

  • Problem 909: the problem asks, for n≥2n\ge2, for a space SS of dimension nn with S2S^2 also of dimension nn. Theorem 2 applied in En+1E^{n+1} gives a set K⊂En+1K\subset E^{n+1} with dim⁡K=dim⁡K2=n\dim K=\dim K^2=n, for every n≥1n\ge1, with dim⁡\dim the inductive dimension of Hurewicz and Wallman. The paper does not mention the problem.