De Moivre's theorem and the triangle inequality
Multiplying complex numbers multiplies moduli and adds amplitudes; hence ; and the modulus of a sum never exceeds the sum of the moduli.
Why this matters
In coordinates, a product of complex numbers is a small calculation; a tenth power is a large one. In polar form both are trivial, because multiplication becomes “multiply the lengths, add the angles”. That rule, applied repeatedly, is De Moivre’s theorem, which turns trigonometric identities into algebra and is the key to the roots of the next lesson. The section also contains the most used inequality in analysis, , which Chapter IV needs as soon as it deals with complex sequences.
The idea
Let and . Multiply out and use the addition formulas:
This is the similar-triangle construction of the first lesson, now in formulas: lengths multiply, angles add.
Try this. Put on the unit circle (the dashed circle) and drag it round: turns but keeps the length of . Then put on the positive real axis and change its length. Finally make both angles large, say and both near .
- z (drag)
- w (drag)
- zw
- unit circle
; .
What to notice
A of modulus is a pure rotation; a positive real is a pure stretch; in general it is both. In the last experiment the angles add to more than , and the read-out reports the principal amplitude of as the sum minus : the angles add, but only up to a whole turn.
Multiplying by itself times multiplies the angle by and raises the length to the -th power. For a number on the unit circle the length stays , and we get De Moivre’s theorem.
Try this. Increase step by step. Then drag just outside the unit circle, just inside it, and onto it.
- z (drag)
- z, z², …, zⁿ
- unit circle
De Moivre: with , : each multiplication by turns through and scales by . .
What to notice
The powers turn by the same angle each time. Outside the unit circle they spiral outwards, inside it they spiral in to , and on it they walk round the circle for ever. This is the picture behind the sequence in Chapter IV.
Precise statements
For complex numbers :
- and .
- .
- is equal to or differs from it by .
- (if ).
- is equal to or differs from it by .
The “or differs by ” is not pedantry. Hardy’s example: has principal amplitude , but has principal amplitude , not . With some values of the amplitudes the rule is exact; with principal values it holds only up to a whole turn.
For every integer , positive, negative or zero,
More generally for .
For any complex numbers ,
with equality only when all the nonzero have the same amplitude. Also .
Proofs
Everything follows from the product formula in polar form, proved above from the addition formulas. Positive powers come by induction, and negative powers from the reciprocal.
Product: (2) and (3)
We showed . So , and is a value of . If are principal values, lies in , so it is the principal value or differs from it by .
Reciprocal
, so . Hence : the modulus is inverted and the amplitude reversed. Combined with the product rule this gives (4) and (5).
Positive powers, by induction
The case is clear. If , multiply both sides by ; the product rule adds the angles: .
Zero and negative powers
For both sides are . For with : by the positive case.
Hardy’s proof of the triangle inequality is the picture of Fig. 23: lay the displacements nose to tail; the sum is the straight line from start to finish, and a straight line is the shortest route. Try it.
Try this. Drag the blue summands and compare the two lengths in the read-out. Then press the button that turns them all to one direction, and drag one of them so that it points the opposite way.
- the summands zₖ (drag)
- copies laid nose to tail: length Σ|zₖ|
- the sum OU: length |Σzₖ|
What to notice
The ochre broken line is never shorter than the purple straight line OU, and they are equal only when the broken line does not bend, that is, when all the summands have the same amplitude. A summand pointing backwards makes the gap largest.
Hardy also asks (Examples XXI, 1) for a proof that does not rely on geometry. Here it is.
Prove the case of two numbers by squaring, using and ; then add one number at a time.
Expand
. The bracket is a number plus its conjugate, so it equals .
Bound the cross term
For any , . So , using (2) and .
Two numbers
Hence , and taking non-negative square roots, .
Many numbers
By induction: .
The reverse form
Apply the two-number case to : , so . Swapping gives . Together, .
Worked examples and exercises
Compute and .
Hint
Solution
. , so its sixth power is .
Express and in terms of and .
Hint
Solution
Write , . Then . By De Moivre this is , and equal complex numbers have equal parts: and (using ).
Show that (all ) only when all the have the same amplitude.
Hint
Hint
Solution
Two numbers. In the algebraic proof the only inequality was . Equality there means is real and positive (it is nonzero). Now , so is a positive real number , and : the two have the same amplitude. Conversely, if with then .
Many numbers. Suppose equality holds for , and let . Then , so both inequalities are equalities. The first says and have the same amplitude (if ); the second, by induction, says have a common amplitude, which is then that of their sum . (, since .) Geometrically: the broken line of Fig. 23 has the length of its chord only if it is straight and never turns back.
Let . Find the principal amplitudes of , and , and explain the discrepancy.
Hint
Solution
. But , whose principal amplitude is . The sum is an amplitude of , but it lies outside ; subtracting gives the principal value . So (3) of the theorem cannot be sharpened to “equals” for principal values.
Show that , and interpret it geometrically.
Hint
Solution
The proof is the last step above. Geometrically, , and are the three sides of the triangle with vertices , , , and the inequality says that a side of a triangle is at least the difference of the other two. In analysis it is used to show that whenever .
Check yourself
What is ?
, and the modulus of a product is the product of the moduli, so the answer is . No expansion is needed.
What is ?
The number has modulus , so its reciprocal has modulus and the opposite amplitude: , which is also its conjugate. Taking reciprocals of the parts separately is wrong (and undefined when ); is the half turn, not the inverse turn.
For all nonzero , the principal value of equals (principal values).
False: is a counterexample (the sum is , the principal value is ). The statement is true up to a multiple of .
For which pair is ?
Equality needs the same amplitude, i.e. with . In the second pair . In the third, : the summands point in opposite directions, and while . The first pair are conjugates with different amplitudes.
Hardy’s language
Summary
- In polar form, : moduli multiply, amplitudes add.
- With principal values, amplitudes add only up to a multiple of .
- De Moivre: for every integer ; it turns multiple-angle formulas into binomial expansions.
- , with equality only when all amplitudes agree; also .