Interactive · 2 questions
Like the root test, the ratio test gives us another method of seeing when a series converges or diverges. And the reason why we introduce this after the root test is that we can show (and we will at the end of this section) that the root test can be used to prove the ratio test.
But wait, why do we need another test, if the ratio test follows from the root test? The next example gives us a clue:
Does the series converge or diverge? (Recall that .)
If we try to solve this question through the root test, we have But what exactly is that denominator? One way to find out is by using Stirling's approximation, which states that for large values of , is very close to . But that's pretty complicated! Is there another way to see whether this series converges?
Thankfully, the answer is yes, and it's called the ratio test. The test states the following:
Theorem For the series , consider the number Then we have the following:
-
If , then converges absolutely.
-
If , the series diverges.
-
If , then the test is inconclusive.
Knowing this theorem, let's return to the example problem above, and see whether the series converges or diverges. Since we have So by the ratio test, the series converges absolutely.
Wow, that was a lot nicer and easier! Like this case, there are many instances in which the ratio test is more fitting to use than the root test. Usually, whenever a series involves factorials or exponential functions, the ratio test is a good idea to start with.
Now you try!
Does the series converge or diverge?
Let's work this out together: since , the ratio is Therefore, and so by the ratio test we know this series diverges.
We'll end this section with a proof of how the root test can be used to show the ratio test. Suppose we have a series and First, we'll show that if then the series converges uniformly.
So, suppose - i.e. This means that eventually, the ratio of consecutive terms of the sequence will become less than : i.e. there exist a constant with and an integer such that for any integer , we have We can rewrite this as So, repeating the process, we can write for any . Consequently, if , then so we can say that Taking the -th root of both sides, we get Next, remember that is a constant, and so if we take the limit of both sides as approaches infinity, we get which is less than by the definition of . Therefore, the limit on the left is less than , and so by the root test we get that converges.
Next, let's consider the case when . In this case, we know that the ratio of consecutive terms of the sequence will become greater than : i.e. there exist a constant with and an integer such that for any integer , we have Hence, by the same reasoning as before, we can write So again, taking the limit of both sides as approaches , we have Therefore, by the root test, diverges.
Phew, that was long! The important thing to take from this proof is that this specifically relied on the root test, showing that in a sense the root test is ``stronger'' than the ratio test. Another shocking fact is that the ratio test (by itself) isn't enough to prove the root test!
In practice, we will choose to use either the Ratio Test or the Root Test, not both! They both are automated comparisons to geometric series (so only use them if something is growing/decaying geometrically fast), and tend to work in the same cases. In particular, if one of the tests gives you , then the other one will as well. So don't waste your time trying the other! (There are series that converge by the Root test when the Ratio test is inconclusive, but that requires
The limit of the ratio in the ratio test doesn't exist.
A stronger version of the Root test Such a series is .)