Sie sind auf Seite 1von 2

Analysis G5085

Convergence of
n=1

an bn
n=1 bn .

Theorem. Let (an ) and (bn ) be two sequences with associated sums A = n=1 an and B = If n=1 an converges absolutely and (bn ) is bounded, then n=1 an bn converges.

Proof. First the proof strategy. To prove convergence of n=1 an bn we need to say something about the sequence of partial sums. Thats complicated because we need to know about the behaviour of an bn as a function of n, and we only know about an and bn separately. That doesnt give us enough control. Instead, the plan is to use that (bn ) is a bounded sequence to try and take it out of the game. Well replace bn by its bound, which means we only need to know about the behaviour of an on its own. Then, the hypothesis of the theorem will guarantee for us that it converges. How should we say that in mathematical notation? First, we write down the information we have. Since (bn ) is bounded, we know that there exists a real number K such that |bn | < K for all n N. Also, because n=1 an converges absolutely, we know that for any > 0, there exists an N( ) such that
m m

|ak | =
k=n+1 k=n+1

|ak | <

n, m

N( ).

(1)

This is just Cauchys convergence criterion for sums. If you dont see why this must be true, you should go back and check your lecture notes. Now we must decide how to use this information. Our job is to show that for any > 0 we can nd an integer, N ( ), such that
m

an bn <
k=n+1

n, m

N ( ).

(2)

This is Cauchys criterion again, now applied to n=1 an bn . How do we show this? First we need to a b | . We can use the triangle inequality to deduce estimate a bound on | m k=n+1 n n
m m

an bn
k=n+1 k=n+1

|an bn |

(3)

and next we can use the property |AB | = |A||B | to nd


m m

an bn
k=n+1 k=n+1

|an ||bn |.

(4)

We know that |bn | < K for any n. So, we replace bn with K everywhere and take it out of the sum. Then things are looking a lot better for us:
m m

an bn
k=n+1

K
k=n+1

|an |.

(5)

To show that this becomes small at large n, we have to argue that the remaining sum over |an | becomes small at large n. Thats exactly what the assumption that n=1 an converges absolutely guarantees for us, from Eq. (1).

Suppose you pick a small number > 0 and ask me to guarantee that the right-hand side of (1) has to be smaller than . How large must I pick n and m? Large enough that
m

|an | <
k=n+1

(6)

or in other words, n, m

N( /K ). For n and m satisfying this condition,


m

an bn < ,
k=n+1

(7)

so we can conclude that this sum converges.

Das könnte Ihnen auch gefallen