1

When am I allowed to do this:

$$\sum_{n=1}^{\infty} f_n + g_n = \sum_{n=1}^{\infty} f_n + \sum_{n=1}^{\infty} g_n $$?

I know I can do it if both $\sum_{n=1}^{\infty} f_n$ and $\sum_{n=1}^{\infty} g_n$ are convergent, but can I do it if one of them is divergent and the other convergent?

John
  • 41
  • 6

3 Answers3

3

If one is convergent and one divergent the sum will diverge. The place you have trouble is that both the sum of $f_n$ and the sum of $g_n$ can be divergent while the sum is nicely convergent. For example $$f_n=2^n\\g_n=-2^n\\f_n+g_n=0$$

Ross Millikan
  • 374,822
1

If any 2 sums exist, then the third exists and the equality holds. e.g. suppose $f_n+g_n$ and $f_n$ is summable. Then $-f_n$ is summable, and apply the known result for $F_n := f_n + g_n$ and $G_n = -f_n$ to see that $F_n + G_n$ is summable, with the equality.

Calvin Khor
  • 34,903
0

It is even worse. Consider two divergent series: $\sum_{n} f_n = \sum_{n} \frac1n$ and $\sum_{n} g_n = - M \sum_{n} \frac1{M n}$ with some natural number $M$. They both diverge, since what we have here are both harmonic series. Now add the two. What you get is

$$ \sum_{n} f_n + g_n = \sum_{n} \frac1n - M \sum_{n}\frac1{M n} \\ = 1 + \frac12 + \cdots + \frac{1}{M-1}+ \frac{1-M}{M} + \frac{1}{M+1} + \cdots + \frac{1}{2M-1}+ \frac{1-M}{2M} + \frac{1}{2M+1} + \cdots \\ = \ln(M) $$ So $\sum_{n} f_n + g_n $ is convergent. The last result may be surprising (had you expected it is zero?) but can be found here. A prominent special case is $M=2$ which is $$ \sum_{n} f_n + g_n = \sum_{n} \frac1n - 2 \sum_{n}\frac1{2 n} \\ = 1 - \frac12 + \frac13 - \frac14 + \cdots = \sum_{n} \frac{(-1)^{n}}{n} = \ln(2) $$ which is the well known alternating harmonic series.

Andreas
  • 15,175