Chapter X · The general theory of the logarithmic, exponential and circular functions
Defining the logarithm of a complex number
Log ζ defined as the integral of dz/z along a path from 1 to ζ, and the question whether its value depends on the path.
Why this matters
Hardy built the real logarithm as an area, (§205), and derived everything else (the exponential, powers, ) from it. §230 makes the same move in the complex plane: the logarithm of is the integral of from to . The definition is easy to state, but it hides a question. There are many paths from to , so the “logarithm” might have many values. This short lesson sets up that question; the next one answers it.
The idea
To integrate from to we must choose a path. Any path will do, with one restriction: it may not pass through , where is not defined. For each admissible path the integral is a perfectly definite complex number. The question is how that number changes when we change .
Try this. The blue path runs from to the purple point . Drag the two blue way-points and itself, and read the value of the integral below the figure. First deform the path a lot while keeping it on the same side of the origin. Then drag a way-point so that the path goes round the other side of . Finally use the stepper to add whole turns round the origin.
- path C (drag the blue way-points) and its log-image
- end-point ζ (drag it)
- straight line from 1 to ζ
What to notice
The value does not change at all while the path moves without crossing the origin. It jumps when the path is dragged across , and every jump is by exactly (the imaginary part changes by , the real part not at all). So there are many values, but not one per path: whole families of paths give the same value.
The experiment already suggests the answer. But Hardy is careful to say that, from the definition alone, three things are conceivable, and that nothing yet tells us which is true:
- every path from to gives the same value;
- every different path gives a different value;
- there are several values, each shared by a whole class of paths.
The explorer points to (3). The next lesson proves it, and identifies the classes: two paths give the same value exactly when they wind round the origin the same number of times.
Precise statements
Let be a complex number other than . If is a path from to that does not pass through the origin, we write
is called the general logarithm of . Its value is defined only once the path has been chosen.
The capital L is deliberate. Hardy keeps for one particular value, which will be the principal value of the next lesson; the first piece of evidence is that the straight path reproduces the old logarithm.
If is real and positive and is the straight segment from to , then , the logarithm of Chapter IX.
The integral depends on : for , the trivial path (staying at ) gives , while the unit circle traversed once counter-clockwise gives . So alternative (1) is false, and is a many-valued function.
A useful way to compare two paths is to join them into one closed path.
If and both run from to , avoiding , then
where is the closed path “along from to , then back along from to ”.
Proofs
Parametrise the segment by the real variable itself and use the definition of §229.
Parametrise
Take for from to ; so , , and . (If the parameter runs downwards; the formula below then reads , which is the same thing.)
Compute
by the definition of §205. The segment does not meet because .
Exhibit two paths from to with different integrals. One is the constant path; the other goes once round the origin, and its integral was computed in the previous lesson.
The trivial path
A path of zero length contributes nothing: its integral is . Equivalently, the segment from to gives .
Once round the circle
On , , we have , so .
Conclude
Both paths are admissible paths from to , and . The value of depends on the path.
Integrals along paths add when paths are joined, and change sign when a path is reversed.
Join and reverse
The integral along is the integral along plus the integral along traversed backwards. Reversing a path changes the sign of the integral (substitute ; see the last quiz of the previous lesson).
Rearrange
Hence , which is the claim. So the whole question “how does the value depend on the path” reduces to: what can be round a closed path avoiding ?
Worked examples and exercises
Compute from to (a) along the quarter circle , ; (b) along the three-quarter circle , , which goes clockwise through and .
Hint
Solution
(a) , so the integral is .
(b) Now and , so the integral is .
The two values differ by . Together the two arcs make one complete counter-clockwise circle (route (a) out, route (b) back reversed), as the lemma predicts.
Let . Find along the path that goes straight from to and then once counter-clockwise round the circle .
Hint
Solution
The segment gives . On the circle , so the circle gives , independently of its radius. Total: . So is one value of , not the only one; this is why Hardy refuses to write for the general logarithm.
Which of these are admissible paths for : (a) the segment from to ; (b) the upper half of the unit circle from to ; (c) the path from straight up to , then straight to , then straight down to ? Compute the integral for each admissible path.
Hint
Solution
(a) is not admissible: it passes through , where is undefined. (b) is admissible: , , gives . (c) is admissible, and the closed path formed by (c) and the reversed semicircle (b) does not enclose ; by the next lesson its integral is , so (c) also gives . (One can check this directly by integrating along the three segments, though it is longer.)
Let be a path from to avoiding , and let be a complex number. Show that , where is the path from to .
Hint
Solution
. The path avoids because does and . Combined with a path from to , this is the germ of the functional equation proved in the next lesson.
Check yourself
Why must the path in the definition of avoid the origin?
The definition of §229 needs the integrand to be defined and continuous on the path. has a pole at . The reason is not about (the end-point is excluded separately), nor about principal values, which have not been defined yet.
Two paths from to form a closed path that goes once counter-clockwise round the origin. By how much do the imaginary parts of the two values of differ? (Give a decimal.)
By the lemma the difference of the values is round the closed path. For one counter-clockwise turn this is , so the imaginary parts differ by , and the real parts not at all.
For real positive , every value of equals .
False. The straight path gives , but a path that also goes once round the origin gives (second exercise). Only the principal value is .
Hardy’s language
Log ξ is (possibly at any rate) a many-valued function, and log ξ is only one of its values.
Summary
- along a path from to that avoids .
- For the straight path gives the real ; other paths can give other values, so is many-valued.
- Two paths give values differing by round the closed path they form.
- Experiment suggests the values change only when the path crosses the origin, by multiples of ; the next lesson proves it.