26
$\begingroup$

What is the limit of the continued fraction $$\cfrac{1}{2+\cfrac{1}{3+\cfrac{1}{5+\cfrac{1}{7+\cfrac{1}{11+\cfrac{1}{13+\cdots}}}}}}\ ?$$

Is the limit algebraic, or expressible in terms of e or $\pi$? What is the fastest way to approximate the limit?

$\endgroup$
  • $\begingroup$ Does the sequence $2,3,5,7,11,13,\dots$ consist of all prime numbers? $\endgroup$ – Américo Tavares Sep 12 '11 at 11:26
  • $\begingroup$ @Angela Richardson: Are you sure that it is convergent? Also, can you really use the expression of continued fractions to express a complex number such as $i$? Maybe I am missing something here, but I am a little confused. Thanks for any explanation. $\endgroup$ – awllower Sep 12 '11 at 11:29
  • 3
    $\begingroup$ @awllower: it converges by Sleszynski-Pringsheim. $\endgroup$ – J. M. is a poor mathematician Sep 12 '11 at 11:40
  • 1
    $\begingroup$ @awllower: It is convergent by the Śleszyńsky-Pringsheim's Theorem. The continued fraction $\mathbf{K}(a_{n}/b_{n})$ converges if for all $n$ $$\left\vert b_{n}\right\vert \geq \left\vert a_{n}\right\vert +1.$$ $\endgroup$ – Américo Tavares Sep 12 '11 at 11:40
  • 2
    $\begingroup$ Reference: Chapter I, Theorem 4.1 of Continued fractions with applications by Lisa Lorentzen and Haakon Waadeland. $\endgroup$ – Américo Tavares Sep 12 '11 at 11:46
25
$\begingroup$

As I've noted in the comments, the Śleszyński-Pringsheim Theorem guarantees that this continued fraction converges to a finite value (since the prime numbers are always greater than or equal to $2$).

This arXiv preprint gives an approximate value $s$ for the continued fraction whose partial denominators are prime numbers:

$$s=0.43233208718590286890925379324199996370511089688\dots$$

which the author computed with PARI/GP. He notes that the Plouffe inverter does not at all recognize this constant.


Here's some Mathematica code for computing the constant to prec or so significant figures:

prec = 500;
y = N[2, prec + 5]; c = y; d = 0; k = 2;
While[True,
  p = Prime[k];
  c = p + 1/c; d = 1/(p + d);
  h = c*d; y *= h;
  If[Abs[h - 1] <= 10^(-prec), Break[]];
  k++];
N[1/y, prec]

which yields the result 0.43233208718590286890925379324199996370511089687765131032815206715855390511529588664247730234675307312901358874751711021925473474173059981681532525370102846860319246045704466728602248840679362020193843643798792955246786129609763893526940277522319731978458635595794036202066338633654544895108909659715862787332585763686200183679952128087865043794610126643260422526400822552675221511335417037835319471839444535578027072057047898703982387228841680143293913635134426277200532815721973910424896503203478507

$\endgroup$
  • 1
    $\begingroup$ For purposes of computation, you can use Lentz-Thompson-Barnett. $\endgroup$ – J. M. is a poor mathematician Sep 12 '11 at 11:51
  • 1
    $\begingroup$ The OEIS has it, though: oeis.org/A084255 $\endgroup$ – Charles Sep 12 '11 at 12:53
  • 1
    $\begingroup$ @J.M. can you calculate this for twin prime pairs instead of the prime (To a certain extent)...as its irrationality will prove the twin prime conjecture $\endgroup$ – Shivam Patel Feb 19 '14 at 16:23
  • 1
    $\begingroup$ @Shivam, one can certainly compute a numerical approximation of the constant you want to a pile of digits, but proving irrationality with it just seems rather unlikely. $\endgroup$ – J. M. is a poor mathematician May 1 '15 at 13:58

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.