Recipe
Classify an isolated singularity
Use when: you have an isolated singularity of f at z_0 and need to know whether it is removable, a pole, or essential
Goal: a firm classification of z_0, plus, when it's a pole, its exact finite order m
Taught in Isolated singularities -- removable, pole, and essential
Quick recipe
-
1
Check that f is analytic on some punctured disk 0<|z-z_0|<r around z_0. No other bad points may crowd arbitrarily close to z_0.
Example
For f(z)=\dfrac{1-\cos z}{z^4} at z_0=0: f is analytic everywhere except at z=0, since \cos z is entire and only z^4 vanishes there. A punctured disk of any radius around 0 works.
Why: without an isolated punctured disk, a Laurent series does not even exist at z_0 to classify.
-
2
Try to simplify f, or bound |f(z)| as z\to z_0. If it settles to a finite value, the singularity is removable and you are done.
Example
Here 1-\cos z\sim\dfrac{z^2}{2} near 0, while the denominator is z^4, a higher power. The ratio behaves like \dfrac{1}{2z^2}. That blows up rather than settling.
- f simplifies or stays bounded near z_0 — removable. Assign f(z_0) that limiting value and stop here.
- |f(z)|\to\infty as z\to z_0 instead — continue to Step 3.
Why: a removable singularity is exactly the case where the Laurent series has no negative powers. Boundedness is the fastest test for that, with no need to build the series.
-
3
Test whether \phi(z)=(z-z_0)^mf(z) becomes analytic and nonzero at z_0. Start at m=1 and raise m by one each time.
Example
Testing m=1: zf(z)=\dfrac{1-\cos z}{z^3}\sim\dfrac{1}{2z} near 0. It is still unbounded, so m=1 fails. Testing m=2: z^2f(z)=\dfrac{1-\cos z}{z^2}\to\dfrac12 as z\to0. That is finite and nonzero.
- some finite m makes \phi analytic and nonzero at z_0 — pole of order exactly that m.
- no finite m ever works, and the search never ends — essential singularity.
Why: a pole of order m is defined by this property. It is the smallest power that clears every negative term in one multiplication.
-
4
State whether z_0 is removable, a pole of order m, or essential. Each answer points to a different next tool.
Why: only a true finite-order pole can use the derivative shortcut formula from the next section. An essential singularity always needs the full Laurent series.
Common mistakes
- Treating any 0/0-looking expression as removable without checking that the limit is finite.
- Stopping the search in Step 3 after one or two failed values of m, and calling the singularity essential too soon.
- Mixing up "not defined at z_0" with "badly behaved at z_0." A removable singularity is not really bad once patched.
- Using the next section's pole-order shortcut on a singularity that was never confirmed to be a finite-order pole.