Try constructing an ordinary power series at w=0 for g(w) = f(1/w), and then revert back to z via w = 1/z.
What you're doing here is creating a power series for f "about the point infinity" (or, equivalently, a Laurent series for f about z=0.)
Try constructing an ordinary power series at w=0 for g(w) = f(1/w), and then revert back to z via w = 1/z.
What you're doing here is creating a power series for f "about the point infinity" (or, equivalently, a Laurent series for f about z=0.)
Instead of trying to directly express it in a (1/z) power series, which is kind of unusual/unnatural compared to our usual efforts of trying to express something in an ordinary power series, first substitute z = 1/w, and then try to find the "ordinary" power series in w. Once you've done that, switch back, using w = 1/z, and you'll have your answer.
You can think of that process in two ways. One way is just an algebraic book keeping exercise. Instead of "confusing" 1/z terms everywhere, we track them as "nice" w terms, and then at the end reveal that those nice w's were actually ugly 1/z's all along. You don't need to do this - FernadoRevilla's post shows how the calculation is doable directly. But it's a suggestion for helping you do that calculation.
The other perspective on this algebraic process I'll put in a separate post, so as not to scare the crap out of you. You can safely ignore it and don't need to know any of it to do this problem.
. Let .
Then
.
Thus .
Now, as is *very* common, try to make it look like the geometric series .
(You could always start taking derivatives and using Taylor Series. But the geometric series trick is much easier when it's possible. It also gives you the radius of convergence of the series - for the above, it's .)
converging on .
Thus
Thus converging on .
Now w = 1/z, z = 1/w, so get
Thus converging on .
Thus converging on (Same as FernandoRevilla's result).