I apologize if I missed an existing question on this, perhaps with a different spelling of Khinchin's name. I feel like I'm missing something basic.

From Wikipedia, almost all real numbers have a continued fraction representation whose terms have a geometric mean of $K_0=2.685...$ From the definition of "almost all", I would understand that there is an at-most-countable set of counter-examples, ie real numbers with continued fraction representations whose terms have a different geometric mean.

But I also see here that continued fractions provide a homeomorphism between real numbers and and sequences of positive integers, seemingly confirming the intuition that the two sets should be isomorphic.

This seems to imply that there should be only a countable number of positive integer sequences with a geometric mean different from Khinchin's constant.

But this seems preposterous! If nothing else, we should be able to generate uncountably many sequences with a geometric mean of $2K_0$, by simply doubling the terms of any "normal" sequence with a mean of $K_0$.

Where did I go wrong?

  • 593
  • 2
  • 9

1 Answers1


Here almost all means except on a set of measure zero; see here, for instance. All countable sets have measure zero, but not all measure zero sets are countable; the middle-thirds Cantor set is a well-known example of a set of measure zero whose cardinality is equal to that of $\Bbb R$.

Brian M. Scott
  • 588,383
  • 52
  • 703
  • 1,170
  • 2
    I can see that but could you explain why his reasoning is wrong about doubling each $a_i$ and getting $2K_0$ as the geometric mean for just as many numbers? It seems reasonable. – Jam Sep 08 '16 at 00:23
  • 4
    @Jam: I’ve not worked it out in detail, but I suspect that it has to do with the rather complicated way that doubling each denominator of a continued fraction affects the real number represented by that continued fraction. – Brian M. Scott Sep 08 '16 at 00:32
  • Of course! I knew I made a simple error. – user3461142 Sep 08 '16 at 00:33
  • 3
    @Jam You will indeed get just as many numbers in the sense of cardinality, but that tells you nothing about the measure of the resulting set, which must be zero. – Jack M Sep 08 '16 at 10:19