Chapter IV · Limits of functions of a positive integral variable
Powers of and the complex geometric series
The powers spiral into exactly when , the geometric series converges exactly then, and its real and imaginary parts sum two trigonometric series.
Why this matters
For real we know exactly what does (§72) and when converges (§78). §87 and §88 answer the same questions for complex , and the answers are cleaner: the only thing that matters is . The real line’s four cases (, , , ) merge into “inside, on, or outside the unit circle”. As a bonus, taking real and imaginary parts of the complex sum gives closed forms for and , which would be awkward to find directly.
The idea
Write in polar form, with . By De Moivre’s theorem (§45),
Each multiplication by turns the point through the angle and multiplies its distance from by . So the powers lie on a spiral: winding inwards if , round a circle if , outwards if . The angle decides how the spiral is drawn; only decides where it goes.
Try this. Drag around inside the unit circle, then onto it (it snaps to the circle), then outside. Try close to and close to . Then switch to “partial sums” and repeat.
- z (drag it)
- zⁿ, joined in order
- unit circle
What to notice
Inside the circle the powers always spiral into , whatever the angle. On the circle they run round for ever (unless , where they sit still). The partial sums lay the same steps end to end: inside the circle they wind into the hollow point , which runs off to infinity as approaches . On or outside the circle they never settle.
Precise statements
If then . If then . For every other , has no limit.
The series converges if and only if , and then its sum is .
Putting and separating real and imaginary parts gives two real series.
If , then for every
Proofs
First find what the limit would have to be: if , then also , while , so . Then use the modulus to decide when the limit really exists.
A limit, if it exists, is or
Suppose . By §86 (1), ; by §86 (4), . Limits are unique, so , that is : either or .
: the limit is
, and because (§72). By §86 (10), .
: the limit is
Every term is .
, : no limit
By the first step, the only possible limit is . But for every , so does not tend to , and by §86 (10) neither does . Hence there is no limit.
The real formula for the partial sums is pure algebra, so it holds for complex . Then the behaviour of the series is the behaviour of .
The partial sums
For , , so . For , .
Convergence when
Then , so .
No convergence when
If , has no limit. Otherwise, if converged, then would converge too, which the previous theorem rules out. (Or: the terms do not tend to , so by §86 (8) the series cannot converge.)
Sum the complex series with , rationalise the denominator of , and compare real and imaginary parts. A complex series converges to exactly when its real and imaginary parts converge to those of (§85).
The real and imaginary parts of the terms
, so the series of real parts is and of imaginary parts .
Rationalise the sum
Multiply top and bottom of by the conjugate . The denominator becomes . So
Compare parts
For we have , so the series converges to this value, and its real and imaginary parts give the two formulas.
Negative
If , write and (since ), and similarly for the sines. The formulas for and are the formulas for and .
Worked examples and exercises
Show that if then and . Which theorems of §86 do you use?
Hint
Solution
is times , so by §86 (7) it converges to .
The second is followed by (§86 (7) again), and putting the single term in front (§86 (2)) gives
For the series converge only if , since otherwise their terms do not tend to .
Show that if , then .
Hint
Solution
By the corollary, with ,
Doubling a convergent series and changing its first term are allowed by §86 (7) and (2). (This function of and is the Poisson kernel, central in the theory of Fourier series.)
Show that converges to exactly when the real part of is greater than .
Hint
Solution
For , it is the geometric series in , which converges if and only if , that is , and then its sum is .
is the distance from to and the distance from to . The points nearer to than to are those to the right of the perpendicular bisector of the segment from to , the line . So the condition is . (Algebraically: with , iff .)
Show that when , and that when the sequence oscillates finitely unless is a multiple of (when it tends to ).
Hint
Hint
Solution
. .
. If is a multiple of , every term is . Otherwise . Suppose . Letting in gives , so . But then ; and , being terms of the sequence at the even places, must also tend to (a subsequence of a convergent sequence has the same limit). Since , this is a contradiction. So has no limit. It is bounded by , so it does not tend to : it oscillates finitely.
Check yourself
For which does have a limit?
Inside the unit circle the limit is ; at it is . Every other point of the circle (such as , , or ) makes run round the circle for ever. The last option forgets the complex numbers inside the circle, such as .
If and the angle of is not a rational multiple of , the points never repeat, so must tend to a limit.
False. Never repeating is not converging. Since for every , the only possible limit, , is ruled out. In fact the powers come back arbitrarily close to every point of the circle.
Find .
With and , , so the formula gives . Directly: the series is , the real part of .
Does converge for ?
On the unit circle the terms have modulus , so they cannot tend to , and a convergent series needs terms tending to . The partial sums are bounded, but bounded partial sums imply convergence only for series of positive real terms. The formula is only valid inside the circle.
Hardy’s language
Summary
- : each power turns by and scales by .
- iff ; iff ; otherwise no limit. A limit would have to satisfy .
- converges iff , with sum .
- Taking real and imaginary parts sums and for .