login

Year-end appeal: Please make a donation to the OEIS Foundation to support ongoing development and maintenance of the OEIS. We are now in our 61st year, we have over 378,000 sequences, and we’ve reached 11,000 citations (which often say “discovered thanks to the OEIS”).

a(n) = 2^(2^(n-1))*b(n) where b(1) = 1/2 and b(n+1) = b(n) - b(n)^2.
11

%I #33 Sep 08 2022 08:45:07

%S 1,1,3,39,8463,483008799,1841209495473815103,

%T 30574267942866411492610617989187955839,

%U 9469098402231365955845075782127013884605836706970528486326625808482327697663

%N a(n) = 2^(2^(n-1))*b(n) where b(1) = 1/2 and b(n+1) = b(n) - b(n)^2.

%C Michael Somos found that b(n) has some nice properties (see link titled "A Somos-Rusin recursion").

%C From Shai Covo (green355(AT)netvision.net.il), Mar 17 2010: (Start)

%C Define a sequence of fractions by b'(1) = 1/2 and b'(n+1) = b'(n) - b'(n)^2/2. Then, b'(n) = 2*b(n+1), n >= 1. The sequence b'(n) is the complement to 1 of the sequence f(n) defined in A167424 by f(1) = 1/2 and f(n+1) = (f(n)^2 + 1)/2, i.e., b'(n) + f(n) = 1. The numerator of b'(n) is given by a'(n) = 2^(2^n-1)*b'(n).

%C Thus a'(n) = a(n+1), n >= 1. The sequences a'(n) and A167424(n) are related by a'(n) + A167424(n) = 2^(2^n-1). (End)

%C From Shai Covo (green355(AT)netvision.net.il), Mar 25 2010: (Start)

%C The following is the counterpart of a comment in A167424. Suppose that U_1, U_2, ... is a sequence of independent uniform (0,1) random variables, and define random variables X_1, X_2, ... as follows: X_1 = U_1, and, for n >= 1, X_{n+1} = X_n or U_{n+1} according as U_{n+1} > E(X_n) or U_{n+1} < E(X_n), respectively, where E() denotes expectation. Then, the sequence E(X_n) is identical to the sequence b'(n) introduced in the comment dated Mar 17 2010. Sketch of proof. E(X_1) = 1/2; for n >= 1, by the law of total expectation, we have E(X_{n+1}) = (1 - E(X_n))*E(X_n) + E(X_n)*E(X_n)/2. Hence E(X_{n+1}) = E(X_n) - E(X_n)^2/2. (End)

%C b(n) = sqrt(p(n-1)/2) with p(n) as defined in A187131, so Sum_n b(n)^2 = 1/2 - _Henry Bottomley_, Mar 05 2011

%H Michael Somos, <a href="http://www.math.niu.edu/Papers/Rusin/known-math/99/somos">A Somos-Rusin recursion</a>

%F a(n) = 2^(2^(n-1))*[1/(n + log(n) + (1-e(oo)) + (log(n)/n)*(1+o(1)))], where 1 - e(oo) is a constant equal to about 1 - 0.232006 = 0.767994. The expression in the square brackets is Dave Rusin's "best estimate" for b(n), taken from the link "Michael Somos, A Somos-Rusin recursion" (where b(n) is denoted a(n)). - Shai Covo (green355(AT)netvision.net.il), Mar 15 2010

%F a(n+1) = a(n) * 2^(2^(n-1)) - a(n)^2 starting from a(1) = 1, also a(n) = sqrt(A187131(n-1)) - _Henry Bottomley_, Mar 05 2011

%F It appears that a(n) = Product_{k = 1..n} A100441(k). - _Peter Bala_, Feb 11 2015

%t a[ n_] := If[ n < 2, Boole[n == 1], a[n - 1] (2^(2^(n - 2)) - a[n - 1])]; (* _Michael Somos_, Jul 24 2018 *)

%o (PARI) b(n)=if(n<2,1/2,b(n-1)-b(n-1)^2); a(n)=numerator(b(n))

%o (Magma) [1] cat [n eq 1 select 1 else Self(n-1)*2^(2^(n-1))-Self(n-1)^2: n in [1..8]]; // _Vincenzo Librandi_, Jun 17 2015

%Y Cf. A100441, A167424, A187131.

%K nonn,easy

%O 1,3

%A _Benoit Cloitre_, Oct 22 2002

%E Typo in definition corrected by Shai Covo (green355(AT)netvision.net.il), Mar 14 2010