Consider a series
"Limsup" version:
Define
(RoLC): If
(RoLD): If
(RaLC): If
(RaLD): If
Note on the shortcuts: "Ro" for Root, "Ra" for Ratio, "L" for limit test, "C" for convergent part, "D" for divergent part.
Before we pass to the "inequality" versions, we address one interesting
anomaly. Note that in the case of the Root test, we only have problems with
limsup being equal to 1. On the other hand, in the Ratio test we have no
information about series for which limsup is at least 1 and liminf at most
1, which is a huge family, since for a typical sequence its limsup is larger
than liminf. Thus the "uncertain region" of the Ratio test is much larger
than for the Root test. Could we replace the "liminf" in the last statement
by "limsup"? Definitely not. It is easy to cook up a convergent series for
which limsup would be greater than 1, for instance this problem in Solved Problems - Testing convergence
has even
Now we recall the general inequality versions.
"Inequality" version:
(RoIC): If there is N and
(RoID): If
(RaIC): If there is N and
(RaID): If there is N so that
Again, there is a difference in "divergent" statements between the Root test
and the Ratio test. Above we quoted a solved problem, the same series shows
that a convergent series may have infinitely many ratios
Now we will pass to the interesting part, we will compare the "L" versions with the "I" versions to see which one is more general. To this purpose we will compare assumptions of the statements above.
1. The assumptions of (RoLC) and (RaLC) are exactly equivalent to the assumptions of (RoIC) and (RaIC), respectively. In other words, if some series satisfies assumptions of (RoLC), then it also satisfies assumptions of (RoIC), and vice versa, so these two tests give convergence for the same series; the same is true for convergence parts of (ReLC) and (ReIC). Thus for convergent cases the limit versions and the inequality versions are of the same "strength", none is better than the other, they are just two different ways of stating the same thing.
2. The assumptions of (RoLD) and (RaLD) are stronger than the assumptions of
(RoIC) and (RaIC), respectively. This means that if some series satisfies,
say, the assumption of (RoLD), then it also automatically satisfies the
assumption of (RoID), therefore we can use the latter test instead of the
former test. However, not every series that satisfies the assumption of
(RoID) also satisfies the assumption of (RoLD). Precisely, knowing that
Similarly, if a series satisfies the assumption of (RaLD), then it also must satisfy the assumption of (RaID), but it does not work the other way around, since the case of liminf equal to 1 may appear. The conclusion therefore is that when it comes to the "divergent" statements, the "inequality" versions of the two tests are stronger, more universal. In other words, every time the limit versions lead to some conclusion we could also use the inequality versions, but there are situations when only the latter can help and the limit ones fail.