ums, Approximations, and Asymptotics Il The area of the bars is at least the area under this curve so we have a lower bound on the n-th harmonic sum 111 (n+1) Remember that n blocks can overhang the edge of a table by Hn block wi we had, say, a million blocks, then this lower bound implies that we could achieve an overhang of at least 1m(.00069 block widths! In fact, since the lower bound of In(n+ 1) grows arbitrarily large, there is no limit on how far the stack can overhang. Of course, this assumes no breeze, defor- mation of the blocks, or gravitational variation as our stack grows thousands of miles high We can get an upper bound on the n-th harmonic number by playing essentially the same game. Now we need a curve that skims just above the bar graph. The curve defined by y=1/a fits the bill 1/2 01234 1n2 The area under this curve is an upper bound on the area of the bar graph and thus on the n-th harmonic sum. But there's a problem: the area under the curve is infinite because y=1/ has a bad asymptote at =0. This is a common problem when bounding sums with integrals and there's a simple solution: take the exact value of the first term(1/1)and then bound the remaining terms(1/2+1/3+..+1/n) with an integral. In this case, weSums, Approximations, and Asymptotics II 5 The area of the bars is at least the area under this curve, so we have a lower bound on the nth harmonic sum: 1 1 1 1 Hn = + + + . . . + � 1 2 3 n n 1 ≥ dx 0 x + 1 = ln(n + 1) Remember that n blocks can overhang the edge of a table by 1 2Hn block widths. So if we had, say, a million blocks, then this lower bound implies that we could achieve an overhang of at least ln(1, 000, 000 + 1) = 6.907 . . . 2 block widths! In fact, since the lower bound of 1 2 ln(n + 1) grows arbitrarily large, there is no limit on how far the stack can overhang. Of course, this assumes no breeze, deformation of the blocks, or gravitational variation as our stack grows thousands of miles high. We can get an upper bound on the nth harmonic number by playing essentially the same game. Now we need a curve that skims just above the bar graph. The curve defined by y = 1/x fits the bill. - 6 1 1 1 2 1 3 1 4 1 1/2 y = 1/x 0 1 2 3 4 n − 1 n The area under this curve is an upper bound on the area of the bar graph and thus on the nth harmonic sum. But there’s a problem: the area under the curve is infinite because y = 1/x has a bad asymptote at x = 0. This is a common problem when bounding sums with integrals and there’s a simple solution: take the exact value of the first term (1/1) and then bound the remaining terms (1/2 + 1/3 + . . . + 1/n) with an integral. In this case, we