Apostol Analytic Number Theory Ch3 ex1

84 Views Asked by At

I have worked through Exercise 1 in Chapter 3. Part (a) requires you to show that

$$\sum\limits_{n \le x} {\frac{{\log n}}{n} = \frac{1}{2}{{\log }^2}x + A + O\left( {\frac{{\log x}}{x}} \right)} $$

The Euler summation formula to apply is

$$\sum\limits_{y < n \le x} {f(n) = \,\,\,\int\limits_y^x {f(t)\,\,{\rm{d}}t} } \,\,\, + \,\,\,\int\limits_y^x {\left( {t - \left[ t \right]} \right)f'(t)\,\,{\rm{d}}t} \,\, + \,\,f(x)\left( {\left[ x \right] - x} \right)\,\, + \,\,f(y)\left( {\left[ y \right] - y} \right)$$

I managed to solve it satisfactorily, but I have also looked at other solutions posted online. In particular, one by a chap called Greg Hirst who has kindly made availabe his solutions to the entire book, which can be accessed here

The extract I am referring to in this post is here: here

His approach was clear enough, except one line where he claims,

$$\left| {\int\limits_x^\infty {\left( {t - \left[ t \right]} \right)\left( {\frac{{1 - \log t}}{{{t^2}}}} \right)\,\,{\rm{d}}t} } \right|\,\, \le \,\,2\int\limits_x^\infty {\frac{{\log t}}{{{t^2}}}} \,\,{\rm{d}}t$$

I have constructed my own argument why the "2" can be incorportated as shown, but it seems rather awkward. Is there anyone who reads this post that can explain why this might be the "go to" approach favoured by some people?

Any insight much appreciated.

1

There are 1 best solutions below

1
On BEST ANSWER

Note that for $x\geq e$, $$\left| {\int_x^\infty {\left( {t - \left[ t \right]} \right)\left( {\frac{{1 - \log t}}{{{t^2}}}} \right)\,dt} } \right|\leq {\int\limits_x^\infty { \underbrace{\left({t - \left[ t \right]}\right)}_{<1} \Big( {\frac{{\overbrace{1}^{\leq \log t} + \log t}}{{{t^2}}}} \Big)\,dt} } \leq 2\int_x^\infty {\frac{{\log t}}{{{t^2}}}} \,dt.$$ A more natural upper bound would be $$\int_x^\infty \frac{1}{{t^2}} \,dt+\int_x^\infty {\frac{{\log t}}{{{t^2}}}} \,dt=\frac{1}{x}+\int_x^\infty {\frac{{\log t}}{{{t^2}}}} \,dt.$$ But in both cases, at end, we will arrive to $O\left(\frac{\log(x)}{x}\right)$.