Root test
Let be a series.
i) if there exists some such thatthen is absolutely convergent.
ii) Conversely, ifthen is divergent.
Proof:
… by comparison test with a geometric series.
Assume there is some s.t. for .
First split the sum, then, for , we have :
is finite, is also finite, for , so is finite, which means is absolutely convergent.
Equivalenty, via limsup, the root test says:
i) absolutely convergent
ii) divergent
iii) inconclusive
Note: says nothing about the terms failing to vanish (that’d be the case if you remove the root). The root goes to for any fixed , as .
… terms vanish faster than every geometric sequence, e.g. (root: )
… terms vanish like (geometric decay)
… terms decay subexponentially (slower than every geometric sequence) or not at all
… terms grow geometrically along an infinite subsequence
EXAMPLE
Take : that means , i.e. . Infinitely many positive terms, yes, but a sum of infinitely many positive terms is finite whenever they shrink fast enough, as the geometric series shows.
The root test only detects exponential decay.
diverges.
converges.It asks “do the terms behave like for some ?” Polynomial decay like for some converges for some , but since independent of , the root test sees both as “no exponential decay” () and can’t tell them apart.
Give the definition of the root test.
Let be a series.
i) if there exists some such thatthen is absolutely convergent. Conversely, if that value is for infinitely many , the series is divergent.
Equivalently:
convergent,
divergent,
inconclusive.
What is the idea behind the root test?
The root test asks whether the terms behave like , i.e. if they decay geometrically (exponentially).
Determine whether converges. Since sin is [0, 1], it only shrinks the terms, so we can easily drop/split it for a comparison test:
Apply the root test to the latter:
The series converges absolutely, hence converges.