Chapter X · The general theory of the logarithmic, exponential and circular functions
Sine, cosine and the hyperbolic functions of a complex variable
cos ζ and sin ζ defined through the exponential, why every trigonometric identity survives, and the surprising new facts — they are unbounded, and cos iy = cosh y.
Why this matters
The sine and cosine of school trigonometry are defined by triangles and angles, which make sense only for real numbers. The formula of the last lesson lets us turn this round: express sine and cosine through the exponential, and use that expression as the definition for complex arguments. §§237–240 carries this out. All the familiar identities survive, but some familiar facts do not: is false, and the circular and hyperbolic functions turn out to be the same functions seen along different axes.
The idea
Put in : , and with , . Adding and subtracting,
for real . The right-hand sides make sense for every complex number, so we use them to define and for all . On the real axis nothing changes. Off the real axis, one of , has modulus , which grows exponentially, so the new functions are large there. The level curves show this at a glance; this is Hardy’s Fig. 56, drawn live.
Try this. Move the purple level from up to , and drag the ochre point off the real axis to read the values. Then switch to and to with and .
- Hardy’s levels k = 0.35, 0.5, 0.71, 1, 1.4, 2, 2.8, 4
- your level k
- zeros (|f| = 0)
- a point z (drag it)
What to notice
For small the curve is a ring of small ovals round the zeros , all on the real axis. At neighbouring ovals touch, and for the curves are wavy lines running along the whole axis, moving outwards as grows: exceeds every bound once you go far enough from the real axis. For the zeros have left the real axis (Hardy's Fig. 59).
Precise statements
For every complex ,
and , , , . For real these agree with the elementary definitions (§237).
and are even, the other four odd. For all complex , :
and , . Every identity of elementary trigonometry is an algebraic consequence of these, and so holds for complex values too.
and for every complex . Then
For all , (the interesting case being real , ):
Consequently and .
The zeros of are exactly and those of exactly , an integer: there are no new, non-real zeros. Both functions have period . On each horizontal line constant they are bounded, but and , so both tend to infinity as , uniformly in . In particular neither is bounded in the plane.
Proofs
Put . Then , and , are rational functions of . Identities between them become identities between rational functions, checkable by algebra. The addition theorem for handles sums of arguments.
Express in terms of
and . Evenness and oddness: replacing by replaces by , which fixes and changes the sign of .
.
The addition formula
With , the addition theorem gives , so . On the other hand . Multiplying out, the terms and cancel and what is left is . The sine formula is proved in the same way.
Quarter periods
, so corresponds to : . Similarly for .
Everything else
Every identity of elementary trigonometry is derived from these by algebra alone (no appeal to triangles or to the size of angles), so the same derivations work for complex arguments.
Use the addition formula with , compute and from the definitions, then take moduli.
and
From the definitions, and .
Addition formula
, and likewise .
Moduli
, using . The cosine is the same with in place of .
Zeros
exactly when , i.e. and , i.e. and . The cosine is similar.
Growth
, and as ; the same holds for .
The formula also explains the picture of as a mapping: on a horizontal line constant the image point runs round an ellipse with semi-axes and , and on a vertical line it runs along a hyperbola. All these conics have foci .
Try this. Look at the images of the blue horizontal lines and the ochre vertical lines. Then drag the purple circle so that it surrounds , a zero of .
- horizontal lines and their images
- vertical lines and their images
- circle (drag centre and rim) and its image
- w = 0
What to notice
Horizontal lines go to ellipses and vertical lines to hyperbolas, all with foci at (compare Misc. Examples X, 16). A circle round a single zero of has an image winding once round ; a circle that surrounds no zero has an image that does not wind round .
Worked examples and exercises
Find all for which is (i) real, (ii) purely imaginary.
Hint
Solution
(i) is real when : either (the real axis) or (vertical lines through the multiples of , where ). (ii) Purely imaginary (or zero) when , i.e. , , since . On those vertical lines .
Find all solutions of for real . In particular solve .
Hint
Solution
From : either or .
If : , possible only for , giving with .
If : , so . Since this needs with even or with odd; if then and we are back in the first case. For , gives . So for , and for .
For : , so . Check: . ✓
Show that if is real, every solution of is real.
Hint
Solution
Let with . Then means (multiply by , both non-zero at a solution), i.e. . By the corollary the zeros of are real, so and , all real.
Let along a ray from the origin that is not part of the real axis. Show that and , and that if the ray lies in the upper half-plane and if it lies in the lower half-plane.
Hint
Solution
On a non-horizontal ray, with , so and .
For : . Now . In the upper half-plane , so and . In the lower half-plane , so the fraction tends to and .
(Hardy’s printed statement of Ex. 17 has the two signs the other way round; the computation above, which a numerical check confirms, e.g. , gives above the axis.)
Use and to turn and into identities for and .
Hint
Solution
becomes . And becomes , i.e. . The sign change wherever a product of two sines occurs is the old “Osborn’s rule”; here it is simply .
Check yourself
for every complex .
False. , which is unbounded. For example , of modulus greater than . The bound holds only on the real axis.
Compute (it is real).
. Cosine of an imaginary number is the hyperbolic cosine, and it is greater than .
Which statement about the zeros of in the complex plane is correct?
vanishes only when and . never vanishes, but is a difference of exponentials, which vanishes where , i.e. . The points are zeros of , not of .
What is , for real , ?
By the addition formula, , and , . So the real part is and the imaginary part . The last option is ‘s real part.
Hardy’s language
Summary
- , define the circular functions for all .
- Every identity of elementary trigonometry remains true, because it is algebra in .
- , : circular and hyperbolic functions are one family.
- ; .
- No new zeros, but no bound either: and grow like away from the real axis.