[seqfan] A024810 floor(2^n/pi)
Kevin Ryde
user42_kevin at yahoo.com.au
Mon Sep 30 11:06:48 CEST 2019
I was contemplating Clark Kimberling's comment in A024810 about
equivalence to floor(2^n/pi). Is there an easy argument for that
(which might appear there!) ?
The closest I got was
x = 2^(n+1)/pi = 1/pi bit-shifted upwards
A024810(n) = floor(cotan(1/x)) the %N definition
cotan(1/x) = x - (1/3)/x - (1/45)/x^3 - (2/945)/x^5 - ...
so floor(x - (1/3)/x - (1/45)/x^3 - (2/945)/x^5 - ...)
vs floor(x)
-(1/3)/x is about -1/2^n so it could only push x down to a smaller
integer part if x had a run of about n+1 many 0-bits immediately below
the binary point (a very small fractional part).
This would mean bits of 1/pi (A127266) having a run of about n many
0-bits at bit position n. The initial bits of 1/pi look like the 1s are
too common, so no such run. But is it so?
More information about the SeqFan
mailing list