MathB.in
New
Demo
Tutorial
About
$$ show \sum_{k=2^n}^{2*2^n-1} \frac{1}{k} >= \frac{1}{2} $$ basic induction won't work because there is no overlap in ks for n+1 $$ \sum_{k=2^{n+1}}^{2*2^{n+1}-1} \frac{1}{k} = \sum_{k=2*2^{n}}^{4*2^{n}-1} \frac{1}{k} $$ we can write dividend and divisor as separate sequences but this doesn't seem productive $$ \sum_{k=2^n}^{2*2^n-1} \frac{1}{k} = \frac{\sum_{j=2^n}^{2*2^j-1} \frac{\prod_{k=2^n}^{2*2^n-1} k}{j}} {\prod_{k=2^n}^{2*2^n-1} k} $$ $$ \ $$ my best attempt: $$ show \sum_{k=2^n}^{2*2^n-1} \frac{1}{k} >= \frac{1}{2} $$ split into odd and even. above iterates over 2^n elements. following split will iterate over 2^n/2 elements twice $$ \sum_{k=2^n}^{2*2^n-1} \frac{1}{k} = \sum_{k=2^{n-1}}^{2^n-1} \frac{1}{2k+1} + \sum_{k=2^{n-1}}^{2^n-1} \frac{1}{2k} $$ therefore $$ \sum_{k=2^{n+1}}^{2*2^{n+1}-1} \frac{1}{k} = \sum_{k=2^{n}}^{2*2^n-1} \frac{1}{2k+1} + \frac{1}{2} \sum_{k=2^{n}}^{2*2^n-1} \frac{1}{k} $$ right-hand side looks familiar. "+1" part of 2k+1 gets exceedingly irrelevant $$ \lim_{n \to \infty} (\sum_{k=2^{n}}^{2*2^n-1} \frac{1}{2k+1} - \frac{1}{2} \sum_{k=2^{n}}^{2*2^n-1} \frac{1}{k}) = 0 $$ and the following becomes true $$ \lim_{n \to \infty} (\sum_{k=2^{n+1}}^{2*2^{n+1}-1} \frac{1}{k} - (\sum_{k=2^{n}}^{2*2^n-1} \frac{1}{2k+1} + \frac{1}{2}\sum_{k=2^{n}}^{2*2^n-1} \frac{1}{k})) = 0$$ if we can substitute one for the other (2k+1 part with 2k part) $$ \lim_{n \to \infty} (\sum_{k=2^{n+1}}^{2*2^{n+1}-1} \frac{1}{k} - 2 * \frac{1}{2}\sum_{k=2^{n}}^{2*2^n-1} \frac{1}{k}) = 0$$ showing convergence n=0 into original inequation gives 1 >= 1/2 n=1 gives 1/2+1/3 >= 1/2 n=2 gives 1/4+1/5+1/6+1/7 = (1/2+1/3)/2 + 1/5+1/7 >= 1/2 we have shown convergence. still to show that its limit is >= 1/2 seems obvious that it is, but how to actually demonstrate it? n=0 1 = 1 red-factor n=1 1/2 * 1 + 2/ 3 = 5/ 6 0.83333333333333333333333333333333 n=2 1/2 * 5/ 6 + 12/ 35 = 319/ 420 0.91142857142857142857142857142857 n=3 1/2 * 319/420 + 6672/19305 = 52279/72072 0.95503503810713842061804444249585 red-factor gives the factor by which the numerical solution for n-1 must be multiplied to get the solution of n. we are converging fairly quickly and speed is bound to increase as the difference between 1/(2k+1) and 1/2k decreases $$ \ $$ seems like a decent argument, but i'm not satisfied with this
ERROR: JavaScript must be enabled to render input!
Thu, 18 Oct 2018 14:34 GMT