The first thing to do is note that, whenever a and b are positive integers and b≥a, we have:
[k=a, b]∑1/k = [a, b+1]∫1/⌊x⌋ dx
Proof: note that since whenever n is an integer ⌊x⌋ = n almost everywhere on [n, n+1]. Thus, [n, n+1]∫1/⌊x⌋ dx = [n, n+1]∫1/n dx = 1/n. Thus we have [a, b+1]∫1/⌊x⌋ dx = [k=a, b]∑[k, k+1]∫1/⌊x⌋ dx = [k=a, b]∑1/k dx, as required.
Now note that since x-1 ≤ ⌊x⌋ ≤ x, we have 1/(x-1) ≥ 1/⌊x⌋ ≥ 1/x. Thus by monotonicity, we have that:
[n+1, 2n+1]∫1/(x-1) dx ≥ [n+1, 2n+1]∫1/⌊x⌋ dx ≥ [n+1, 2n+1]∫1/x dx.
Now, resolving these integrals, we have that:
[n+1, 2n+1]∫1/(x-1) dx = ln (2n) - ln n = ln 2
[n+1, 2n+1]∫1/⌊x⌋ dx = [k=n+1, 2n]∑1/k
[n+1, 2n+1]∫1/x dx = ln (2n+1) - ln (n+1) = ln ((2n+1)/(n+1))
Thus per the above inequality, we have:
ln 2 ≥ [k=n+1, 2n]∑1/k ≥ ln ((2n+1)/(n+1))
But [n→∞]lim ln 2 = [n→∞]lim ln ((2n+1)/(n+1)) = ln 2, so by the squeeze theorem, [n→∞]lim [k=n+1, 2n]∑1/k = ln 2. And we are done.
2007-12-06 12:18:52
·
answer #1
·
answered by Pascal 7
·
0⤊
0⤋