A complex number extends real arithmetic and represents a point in a plane. These two interpretations reinforce each other: addition combines displacements, while multiplication combines rotations and changes of scale. The aim of this chapter is to move between these descriptions without losing a sign, a root, or a condition.

We use the real-number operations from Number Systems, Algorithms, and Recursion, elementary trigonometry, and the triangle inequality. MIT’s introductory complex-variable notes provide a companion treatment of complex algebra, geometry, and exponentials [1][1] J. Orloff, “Topic 1: Complex Algebra and the Complex Plane,” 2018. MIT OpenCourseWare, 18.04 Complex Variables with Applications, Spring 2018. https://ocw.mit.edu/courses/18-04-complex-variables-with-applications-spring-2018/resources/mit18_04s18_topic1/.

Construct complex arithmetic

No real number squares to −1-1. To extend the number system consistently, take ordered pairs of real numbers and define their operations.

DefinitionComplex numbers

The set C\mathbb C is R2\mathbb R^2 with addition and multiplication given by

(a,b)+(c,d)=(a+c,b+d),(a,b)(c,d)=(ac−bd,ad+bc).\begin{aligned} (a,b)+(c,d)&=(a+c,b+d),\\ (a,b)(c,d)&=(ac-bd,ad+bc). \end{aligned}

Identify a real number aa with (a,0)(a,0) and set i=(0,1)i=(0,1). Then i2=(−1,0)=−1i^2=(-1,0)=-1, and (a,b)=a+bi(a,b)=a+bi. Equality means equality of both coordinates.

For z=a+biz=a+bi, the real part is Re⁡z=a\operatorname{Re}z=a and the imaginary part is Im⁡z=b\operatorname{Im}z=b. Both parts are real numbers; the imaginary part is bb, not bibi. Real numbers have b=0b=0, and nonzero purely imaginary numbers have a=0a=0, b≠0b\ne0.

The pair rules yield the familiar formulas

(a+bi)+(c+di)=(a+c)+(b+d)i,(a+bi)(c+di)=(ac−bd)+(ad+bc)i.\begin{aligned} &(a+bi)+(c+di)\\ &\quad=(a+c)+(b+d)i,\\ &(a+bi)(c+di)\\ &\quad=(ac-bd)+(ad+bc)i. \end{aligned}

Thus ordinary expansion works, followed by replacing i2i^2 with −1-1. For example, (2+i)(3−2i)=8−i(2+i)(3-2i)=8-i. Powers of ii repeat every four steps:

i0=1,i1=i,i2=−1,i3=−i.i^0=1,\quad i^1=i,\quad i^2=-1,\quad i^3=-i.

This cycle also handles negative integer powers: i−1=−ii^{-1}=-i, since i(−i)=1i(-i)=1.

Conjugation, modulus, and division

DefinitionConjugate and modulus

For z=a+biz=a+bi, define

z‾=a−bi,∣z∣=a2+b2.\overline z=a-bi,\qquad |z|=\sqrt{a^2+b^2}.

The square root here is the nonnegative real square root. In particular, ∣z∣≥0|z|\ge0, and ∣z∣=0|z|=0 exactly when z=0z=0.

Multiplication gives the key identity zz‾=∣z∣2z\overline z=|z|^2. If z≠0z\ne0, this positive real number supplies an inverse:

1z=z‾∣z∣2=a−bia2+b2.\frac1z=\frac{\overline z}{|z|^2} =\frac{a-bi}{a^2+b^2}.

To divide by a complex number, multiply numerator and denominator by the conjugate of the denominator. For example,

3+4i1+2i=(3+4i)(1−2i)5=115−25i.\begin{aligned} \frac{3+4i}{1+2i} &=\frac{(3+4i)(1-2i)}{5}\\ &=\frac{11}{5}-\frac25i. \end{aligned}

Conjugation respects addition and multiplication:

z+w‾=z‾+w‾,zw‾=z‾ w‾,z‾‾=z.\begin{aligned} \overline{z+w}&=\overline z+\overline w,\\ \overline{zw}&=\overline z\,\overline w,\\ \overline{\overline z}&=z. \end{aligned}

Also, z+z‾=2Re⁡zz+\overline z=2\operatorname{Re}z and z−z‾=2iIm⁡zz-\overline z=2i\operatorname{Im}z. These identities follow by writing out the coordinates; they often avoid a much longer expansion.

Why these operations form a field
Proof

Addition inherits commutativity and associativity from real coordinate addition. Its identity is (0,0)(0,0), and the additive inverse of (a,b)(a,b) is (−a,−b)(-a,-b). The multiplication formula is symmetric in the two pairs, so multiplication is commutative, with identity (1,0)(1,0).

For z=a+biz=a+bi, w=c+diw=c+di, and u=e+fiu=e+fi, expanding either (zw)u(zw)u or z(wu)z(wu) gives real part

ace−bde−adf−bcface-bde-adf-bcf

and imaginary part

acf−bdf+ade+bce.acf-bdf+ade+bce.

This proves associativity. Distributivity follows by distributing the real coordinates in the same way. The inverse formula above gives a multiplicative inverse for every nonzero element. These are the field axioms.

In particular, zw=0zw=0 implies z=0z=0 or w=0w=0: if z≠0z\ne0, multiply by its inverse. This justifies solving a factored equation one factor at a time.

Being a field does not make C\mathbb C an ordered field. In an ordered field every nonzero square is positive; applying that to ii would make −1-1 positive. Complex inequalities therefore compare real quantities such as moduli or real parts, rather than assigning every complex number a compatible position in an order.

RemarkAn equation has roots; a square-root symbol needs a convention

The equation z2=−5z^2=-5 has two roots, ±i5\pm i\sqrt5. If −5\sqrt{-5} denotes the principal complex square root, its value is i5i\sqrt5. Do not extend the real product rule for square roots without conditions:

−1−1=−1≠1=(−1)(−1).\sqrt{-1}\sqrt{-1}=-1\ne1=\sqrt{(-1)(-1)}.

We will obtain all roots systematically from polar form.

Read arithmetic as geometry

Represent z=x+iyz=x+iy by the point (x,y)(x,y). Its modulus is its distance from the origin; ∣z−w∣|z-w| is the distance between two points. Addition is vector addition, subtraction is displacement, and conjugation reflects a point across the real axis.

Complex conditionGeometric meaning
$z-a
$z-a
Re⁡z>c\operatorname{Re}z>c, c∈Rc\in\mathbb ROpen half-plane to the right of x=cx=c
$z-a

For radius zero, the circle equation and closed-disk inequality both reduce to the single point aa. A negative radius gives no solution.

The product zw‾z\overline w connects complex algebra to Euclidean geometry. If z=x+iyz=x+iy and w=u+ivw=u+iv, then

Re⁡(zw‾)=xu+yv.\operatorname{Re}(z\overline w)=xu+yv.

This is the ordinary dot product of the two corresponding real vectors. Consequently,

∣z+w∣2=∣z∣2+∣w∣2+2Re⁡(zw‾).|z+w|^2=|z|^2+|w|^2+2\operatorname{Re}(z\overline w).
TheoremModulus of a product and quotient

For all z,w∈Cz,w\in\mathbb C,

∣zw∣=∣z∣∣w∣.|zw|=|z||w|.

If w≠0w\ne0, then ∣z/w∣=∣z∣/∣w∣|z/w|=|z|/|w|.

Proof

Using conjugation and commutativity,

∣zw∣2=zwz‾ w‾=(zz‾)(ww‾)=∣z∣2∣w∣2.\begin{aligned} |zw|^2&=zw\overline z\,\overline w\\ &=(z\overline z)(w\overline w)\\ &=|z|^2|w|^2. \end{aligned}

Both proposed moduli are nonnegative, so equality of squares gives the result. Apply it to z=(z/w)wz=(z/w)w to obtain the quotient rule.

TheoremTriangle and reverse triangle inequalities

For all complex z,wz,w,

∣z+w∣≤∣z∣+∣w∣,|z+w|\le|z|+|w|,

and

∣∣z∣−∣w∣∣≤∣z−w∣.\bigl||z|-|w|\bigr|\le|z-w|.

In either inequality, equality holds exactly when one number is zero or the two nonzero numbers point in the same direction.

Proof

For any complex vv, Re⁡v≤∣v∣\operatorname{Re}v\le|v|, with equality exactly when vv is a nonnegative real number. Therefore

∣z+w∣2=∣z∣2+∣w∣2+2Re⁡(zw‾)≤(∣z∣+∣w∣)2.\begin{aligned} |z+w|^2 &=|z|^2+|w|^2+2\operatorname{Re}(z\overline w)\\ &\le(|z|+|w|)^2. \end{aligned}

Equality means zw‾z\overline w is nonnegative real. If w≠0w\ne0, the identity z/w=zw‾/∣w∣2z/w=z\overline w/|w|^2 shows that this means z=twz=tw for some real t≥0t\ge0.

The reverse inequality follows by applying the triangle inequality to z=(z−w)+wz=(z-w)+w and then exchanging z,wz,w. For its equality condition, observe that

∣z−w∣2−(∣z∣−∣w∣)2=2(∣z∣∣w∣−Re⁡(zw‾)).\begin{aligned} &|z-w|^2-(|z|-|w|)^2\\ &\quad=2\bigl(|z||w|-\operatorname{Re}(z\overline w)\bigr). \end{aligned}

The difference vanishes under exactly the same condition.

Polar form: magnitude and direction

For z≠0z\ne0, let r=∣z∣>0r=|z|>0. An angle θ\theta describing its direction satisfies

z=r(cos⁡θ+isin⁡θ).z=r(\cos\theta+i\sin\theta).

We abbreviate cos⁡θ+isin⁡θ\cos\theta+i\sin\theta as cis⁡θ\operatorname{cis}\theta. The angle is not unique: adding an integer multiple of 2π2\pi gives the same point.

ConventionArguments and the principal argument

We use arg⁡z\arg z for the set of all arguments and Arg⁡z\operatorname{Arg}z for the unique argument in (−π,π](-\pi,\pi]. Thus

arg⁡z={Arg⁡z+2πk:k∈Z}.\arg z=\{\operatorname{Arg}z+2\pi k:k\in\mathbb Z\}.

Neither arg⁡0\arg0 nor Arg⁡0\operatorname{Arg}0 is defined. Although 0cis⁡θ=00\operatorname{cis}\theta=0 for every angle, the origin has no direction.

For z=x+iy≠0z=x+iy\ne0, find the angle using both cos⁡θ=x/r\cos\theta=x/r and sin⁡θ=y/r\sin\theta=y/r. The ratio y/xy/x alone loses quadrant information and is undefined when x=0x=0.

| zz | ∣z∣|z| | Arg⁡z\operatorname{Arg}z | | --- | --- | --- | | −1/2-1/2 | 1/21/2 | π\pi | | −3+3i-3+3i | 323\sqrt2 | 3π/43\pi/4 | | −πi-\pi i | π\pi | −π/2-\pi/2 | | −23−2i-2\sqrt3-2i | 44 | −5π/6-5\pi/6 |

Multiplication rotates and scales

The trigonometric addition formulas give

cis⁡αcis⁡β=cis⁡(α+β).\operatorname{cis}\alpha\operatorname{cis}\beta =\operatorname{cis}(\alpha+\beta).

Hence, for z=rcis⁡αz=r\operatorname{cis}\alpha and w=scis⁡βw=s\operatorname{cis}\beta with r,s>0r,s>0,

zw=rscis⁡(α+β),zw=rs\operatorname{cis}(\alpha+\beta),

and

zw=rscis⁡(α−β).\frac zw=\frac rs\operatorname{cis}(\alpha-\beta).

Multiplication by a fixed nonzero ww scales every distance by ss and rotates every vector by β\beta. In particular, multiplication by ii rotates counterclockwise through π/2\pi/2.

Complex multiplication rotates and scales. Multiplication combines magnitudes and adds angles modulo a full turn. Multiplying by the imaginary unit gives a quarter turn.

Complex multiplication rotates and scales. Multiplication combines magnitudes and adds angles modulo a full turn. Multiplying by the imaginary unit gives a quarter turn.

The principal argument of a product must be brought back into (−π,π](-\pi,\pi]:

Arg⁡(zw)≡Arg⁡z+Arg⁡w(mod2π).\operatorname{Arg}(zw) \equiv\operatorname{Arg}z+\operatorname{Arg}w\pmod{2\pi}.

This is a congruence, not an unrestricted equality of real numbers. For z=w=cis⁡(3π/4)z=w=\operatorname{cis}(3\pi/4), the sum of principal arguments is 3π/23\pi/2, whereas Arg⁡(zw)=−π/2\operatorname{Arg}(zw)=-\pi/2.

For example, division in polar form gives

1+i3−i=2cis⁡(π/4)2cis⁡(−π/6)=22cis⁡5π12.\begin{aligned} \frac{1+i}{\sqrt3-i} &=\frac{\sqrt2\operatorname{cis}(\pi/4)}{2\operatorname{cis}(-\pi/6)}\\ &=\frac{\sqrt2}{2}\operatorname{cis}\frac{5\pi}{12}. \end{aligned}

Exponentials and integer powers

We can extend the real exponential by defining, for real x,yx,y,

ex+iy=ex(cos⁡y+isin⁡y).e^{x+iy}=e^x(\cos y+i\sin y).

At x=0x=0 this is Euler’s formula eiy=cos⁡y+isin⁡ye^{iy}=\cos y+i\sin y. It makes polar form z=reiθz=re^{i\theta}. The factor ii in the exponent is essential: reθre^\theta is a positive real number when r>0r>0 and θ\theta is real.

The real exponential law and the angle-addition formulas show that, for all complex u,vu,v,

eu+v=euev.e^{u+v}=e^ue^v.

This verifies that the definition preserves multiplication of exponentials. The power-series approach gives the same extension once convergence of those series is available; it is not needed for the algebra here.

In particular,

∣ex+iy∣=ex,ez‾=ez‾.|e^{x+iy}|=e^x,\qquad \overline{e^z}=e^{\overline z}.

The exponential is never zero and is periodic in the imaginary direction:

eu=ev⟺u−v∈2πiZ.e^u=e^v\quad\Longleftrightarrow\quad u-v\in2\pi i\mathbb Z.

Indeed, equality of moduli first forces equal real parts. Equality of both sine and cosine then forces the imaginary parts to differ by an integer multiple of 2π2\pi.

TheoremDe Moivre's formula

For every real θ\theta and integer nn,

(cis⁡θ)n=cis⁡(nθ).(\operatorname{cis}\theta)^n=\operatorname{cis}(n\theta).

More generally, for r>0r>0,

(reiθ)n=rneinθ.(re^{i\theta})^n=r^ne^{in\theta}.
Proof

For positive nn, repeatedly apply the product rule; this can also be written as induction. For n=0n=0, both sides are 11. For negative nn, use the reciprocal and the fact that eiθe−iθ=1e^{i\theta}e^{-i\theta}=1.

The restriction r>0r>0 makes negative powers valid. If the base is zero, positive integer powers are still zero; no negative power is defined, and we do not assign a value to 000^0 here. Noninteger exponents need a separate convention, discussed in the final exploration.

Find every complex root

TheoremThe nth roots of a nonzero complex number

Let n≥1n\ge1 be an integer and a=reiθ≠0a=re^{i\theta}\ne0, where r>0r>0. The equation zn=az^n=a has exactly nn distinct roots:

zk=r1/nexp⁡(iθ+2πkn),z_k=r^{1/n}\exp\left(i\frac{\theta+2\pi k}{n}\right),

where k=0,1,…,n−1k=0,1,\ldots,n-1 and r1/nr^{1/n} is the positive real root.

Proof

Write any solution as z=ρeiϕz=\rho e^{i\phi}, where ρ>0\rho>0. Comparing moduli gives ρn=r\rho^n=r, so ρ=r1/n\rho=r^{1/n}. Comparing directions gives

nϕ=θ+2πkn\phi=\theta+2\pi k

for some integer kk. Conversely, each angle of this form gives a solution. Two indices give the same point exactly when their difference is divisible by nn. The listed indices therefore give all solutions without repetition.

The roots lie on a circle of radius r1/nr^{1/n}, separated by an angle 2π/n2\pi/n. Changing the chosen argument of aa only reorders the roots. If a=0a=0, the only root is zero, with multiplicity nn in the polynomial znz^n.

For instance, 2+i2=2eiπ/4\sqrt2+i\sqrt2=2e^{i\pi/4} has cube roots

21/3ei(π/12+2πk/3),k=0,1,2.2^{1/3}e^{i(\pi/12+2\pi k/3)},\qquad k=0,1,2.

Roots of unity

Set ω=e2πi/n\omega=e^{2\pi i/n}. The nnth roots of unity are 1,ω,…,ωn−11,\omega,\ldots,\omega^{n-1}. They are closed under multiplication because exponents can be reduced modulo nn.

For n≥2n\ge2, ω≠1\omega\ne1, and the finite geometric-sum formula gives

∑k=0n−1ωk=1−ωn1−ω=0.\sum_{k=0}^{n-1}\omega^k =\frac{1-\omega^n}{1-\omega}=0.

For n=1n=1, the sum is 11. The zero sum for n≥2n\ge2 expresses the symmetry of equally spaced points around the origin.

A root of unity is primitive of order nn if its smallest positive power equal to 11 is its nnth power. The root ωk\omega^k has order n/gcd⁡(n,k)n/\gcd(n,k), so it is primitive exactly when gcd⁡(n,k)=1\gcd(n,k)=1: its jjth power is 11 precisely when nn divides kjkj.

Quadratics and polynomial roots

For Az2+Bz+C=0Az^2+Bz+C=0 with complex coefficients and A≠0A\ne0, completing the square gives

(2Az+B)2=B2−4AC.(2Az+B)^2=B^2-4AC.

Choose a square root ss of the discriminant. The solutions are

z=−B±s2A,s2=B2−4AC.z=\frac{-B\pm s}{2A},\qquad s^2=B^2-4AC.

The two choices of ss give the same solution set. A zero discriminant gives one root of multiplicity two. For example,

2z2−z+1=0⟹z=1±i74.2z^2-z+1=0 \quad\Longrightarrow\quad z=\frac{1\pm i\sqrt7}{4}.
TheoremConjugate roots require real coefficients

If a polynomial PP has real coefficients, then

P(z‾)=P(z)‾.P(\overline z)=\overline{P(z)}.

Consequently, every nonreal root occurs with its conjugate, with the same multiplicity.

Proof

Conjugation preserves sums and products and fixes every real coefficient. Applying it term by term proves the identity and the root implication. If P(t)=(t−z)mQ(t)P(t)=(t-z)^mQ(t) with Q(z)≠0Q(z)\ne0, conjugating the coefficients of this factorization gives a factor (t−z‾)m(t-\overline z)^m whose remaining factor is nonzero at z‾\overline z. Thus the multiplicities agree.

For example, if −2+i-2+i is a root of

P(z)=2z3+9z2+14z+5,P(z)=2z^3+9z^2+14z+5,

then −2−i-2-i is another. Their factors multiply to

(z+2−i)(z+2+i)=z2+4z+5.(z+2-i)(z+2+i)=z^2+4z+5.

Polynomial division, or direct multiplication, now verifies

P(z)=(z2+4z+5)(2z+1).P(z)=(z^2+4z+5)(2z+1).

The remaining root is −1/2-1/2. Checking the leading coefficient is essential: the product of three monic linear factors alone would omit the factor 22.

TheoremFundamental theorem of algebra

Every complex polynomial of degree n≥1n\ge1 factors into nn linear factors over C\mathbb C, counting multiplicity. Its leading coefficient remains as the prefactor.

ProofA minimum-modulus argument

Let p(z)=anzn+⋯+a0p(z)=a_nz^n+\cdots+a_0 with an≠0a_n\ne0. The triangle inequality gives

∣p(z)∣≥∣an∣∣z∣n−∑j<n∣aj∣∣z∣j⟶∞as ∣z∣→∞.|p(z)|\ge |a_n||z|^n-\sum_{j<n}|a_j||z|^j\longrightarrow\infty \quad\text{as }|z|\to\infty.

Thus the continuous function ∣p∣|p| attains a global minimum at some z0z_0: outside a sufficiently large disk it exceeds ∣p(0)∣+1|p(0)|+1, while inside the closed disk it attains a minimum. This compactness assertion follows by taking convergent subsequences of the bounded real and imaginary coordinates, using real completeness.

Suppose c=p(z0)≠0c=p(z_0)\ne0. Expand around z0z_0 and let m≥1m\ge1 be the first nonzero positive-degree coefficient:

p(z0+w)c=1+bmwm+∑j>mbjwj,bm≠0.\frac{p(z_0+w)}c=1+b_mw^m+\sum_{j>m}b_jw^j, \qquad b_m\ne0.

For 0<t<10<t<1, the already proved complex-root formula supplies ww with wm=−t/bmw^m=-t/b_m. Then ∣w∣=(t/∣bm∣)1/m|w|=(t/|b_m|)^{1/m}. For ∣w∣≤1|w|\le1, the remaining finite sum has modulus at most C∣w∣m+1C|w|^{m+1}, where C=∑j>m∣bj∣C=\sum_{j>m}|b_j|. Choose tt small enough that this is less than t/2t/2. It follows that

∣p(z0+w)c∣≤1−t+t/2<1,\left|\frac{p(z_0+w)}c\right| \le1-t+t/2<1,

contradicting minimality. Hence pp has a root. Polynomial division extracts its linear factor; induction on degree extracts all nn factors and preserves the leading coefficient.

Over R\mathbb R, conjugate factors combine into real quadratics. For example,

z4+1=(z2−2z+1)(z2+2z+1).z^4+1=(z^2-\sqrt2z+1)(z^2+\sqrt2z+1).

Both quadratics have negative discriminants, so this is a factorization into irreducible real quadratics.

Exercises

Try each problem before opening its solution. Cartesian form is usually best for linear equations, polar form for products and roots, and conjugation for modulus identities and real-coefficient polynomials.

ExerciseParts and powers of i

Show that Re⁡(iz)=−Im⁡z\operatorname{Re}(iz)=-\operatorname{Im}z. Simplify

3i11+6i3+8i20+i−1.3i^{11}+6i^3+\frac8{i^{20}}+i^{-1}.
Solution

If z=a+biz=a+bi, then iz=−b+aiiz=-b+ai. The four terms are −3i-3i, −6i-6i, 88, and −i-i, so the expression is 8−10i8-10i.

ExerciseSolve without losing a zero root

Solve each equation over C\mathbb C:

iz=4−zi,z1−z=1−5i,8z2+(2−i)z=0,z2+16=0.\begin{aligned} iz&=4-zi,\\ \frac z{1-z}&=1-5i,\\ 8z^2+(2-i)z&=0,\\ z^2+16&=0. \end{aligned}
Solution

The first gives 2iz=42iz=4, hence z=−2iz=-2i.

The second requires z≠1z\ne1. Rearranging gives (2−5i)z=1−5i(2-5i)z=1-5i, so

z=1−5i2−5i=27−5i29.z=\frac{1-5i}{2-5i}=\frac{27-5i}{29}.

This value is not 11, so it is admissible. The third factors as z(8z+2−i)=0z(8z+2-i)=0, giving z=0z=0 or z=(−2+i)/8z=(-2+i)/8. Dividing by zz at the start would lose a solution. The last equation gives z=±4iz=\pm4i.

ExerciseA complex linear system

Solve and check both equations:

(1−i)z1+3z2=2−3i,iz1+(1+2i)z2=1.\begin{aligned} (1-i)z_1+3z_2&=2-3i,\\ iz_1+(1+2i)z_2&=1. \end{aligned}
Solution

The determinant is (1−i)(1+2i)−3i=3−2i≠0(1-i)(1+2i)-3i=3-2i\ne0. Elimination gives

z1=5+i3−2i=1+i,z2=−2−3i3−2i=−i.\begin{aligned} z_1&=\frac{5+i}{3-2i}=1+i,\\ z_2&=\frac{-2-3i}{3-2i}=-i. \end{aligned}

Substitution gives (1−i)(1+i)−3i=2−3i(1-i)(1+i)-3i=2-3i and i(1+i)+(1+2i)(−i)=1i(1+i)+(1+2i)(-i)=1.

ExerciseThree real multiplications

Compute (a+bi)(c+di)(a+bi)(c+di) using three real multiplications, together with additions and subtractions. Explain why this does not by itself prove a runtime improvement on every computer.

Solution

Compute

p=ac,q=bd,t=(a+b)(c+d).\begin{aligned} p&=ac,\qquad q=bd,\\ t&=(a+b)(c+d). \end{aligned}

The real part is p−qp-q, and the imaginary part is t−p−qt-p-q. This saves one multiplication but uses more additions. Runtime and floating-point error also depend on the implementation and hardware, so an operation count alone is not a benchmark.

ExerciseModuli without expansion

Find the moduli of

1+2i−2−i,(1+i)(2−3i)(4i−3),\frac{1+2i}{-2-i},\qquad (1+i)(2-3i)(4i-3),

and

i(2+i)3(1−i)2,(π+i)100(π−i)100.\frac{i(2+i)^3}{(1-i)^2},\qquad \frac{(\pi+i)^{100}}{(\pi-i)^{100}}.
Solution

Apply the product and quotient rules. The answers, in order, are

1,526,552,1.1,\qquad 5\sqrt{26},\qquad \frac{5\sqrt5}{2},\qquad1.

Every denominator is nonzero. For the last expression, numerator and denominator have equal moduli because their bases are conjugates.

ExerciseTranslate loci

Describe the sets given by ∣z−1∣=∣z+1∣|z-1|=|z+1|, ∣z−(1+i)∣≤2|z-(1+i)|\le2, and ∣z−1∣=Re⁡z+1|z-1|=\operatorname{Re}z+1.

Solution

Write z=x+iyz=x+iy. The first equation reduces to (x−1)2+y2=(x+1)2+y2(x-1)^2+y^2=(x+1)^2+y^2, hence x=0x=0, the imaginary axis. The second is the closed disk of radius 22 centred at (1,1)(1,1).

For the third, the right side must be nonnegative. Squaring gives

(x−1)2+y2=(x+1)2,(x-1)^2+y^2=(x+1)^2,

so y2=4xy^2=4x. Conversely, this equation implies x≥0x\ge0, hence x+1>0x+1>0, so taking the nonnegative square root recovers the original equation. The locus is the entire parabola y2=4xy^2=4x.

ExerciseDot products and the parallelogram law

Prove the parallelogram identity

∣z+w∣2+∣z−w∣2=2∣z∣2+2∣w∣2.|z+w|^2+|z-w|^2=2|z|^2+2|w|^2.

For nonzero z,wz,w, express perpendicularity using zw‾z\overline w.

Solution

Expand the two squared moduli. Their cross terms are 2Re⁡(zw‾)2\operatorname{Re}(z\overline w) and its negative, so they cancel. Since Re⁡(zw‾)\operatorname{Re}(z\overline w) is the real vector dot product, the vectors are perpendicular exactly when it is zero. A complex product zwzw itself is not the dot product.

ExerciseWeighted points in the unit disk

Let λ1,…,λn≥0\lambda_1,\ldots,\lambda_n\ge0, with Λ=∑k=1nλk>0\Lambda=\sum_{k=1}^n\lambda_k>0, and let ∣zk∣≤1|z_k|\le1. Prove

∣∑k=1nλkzkΛ∣≤1.\left|\frac{\sum_{k=1}^n\lambda_kz_k}{\Lambda}\right|\le1.

When does equality hold?

Solution

Repeated use of the triangle inequality gives

∣∑kλkzk∣≤∑kλk∣zk∣≤Λ.\left|\sum_k\lambda_kz_k\right| \le\sum_k\lambda_k|z_k|\le\Lambda.

Divide by the positive number Λ\Lambda. Equality holds exactly when all zkz_k with λk>0\lambda_k>0 are the same point on the unit circle. To see necessity, write the weighted average as uu with ∣u∣=1|u|=1 and take the real part after multiplying the sum by u‾\overline u. Each Re⁡(zku‾)≤1\operatorname{Re}(z_k\overline u)\le1; a positive weighted average of these numbers is 11 only if each active term equals 11. Together with ∣zk∣≤1|z_k|\le1, this forces zk=uz_k=u.

ExercisePrincipal arguments after multiplication

Find the modulus and principal argument of each number:

(1−i)(−3+i),(3−i)2,(1-i)(-\sqrt3+i),\qquad(\sqrt3-i)^2,

and

−1+3i2+2i.\frac{-1+\sqrt3i}{2+2i}.
Solution

Add or subtract the angles of the factors, then reduce into (−π,π](-\pi,\pi]. The respective pairs are

(22,7π12),(4,−π3),(12,5π12).\begin{gathered} \left(2\sqrt2,\frac{7\pi}{12}\right),\qquad \left(4,-\frac\pi3\right),\\ \left(\frac1{\sqrt2},\frac{5\pi}{12}\right). \end{gathered}

For example, the first angle is −π/4+5π/6=7π/12-\pi/4+5\pi/6=7\pi/12. All arguments of each number are its listed principal argument plus 2πk2\pi k.

ExerciseExponentials in Cartesian form

Convert the following to a+bia+bi:

e−iπ/4,e1+3πie−1+πi/2,2e3+iπ/6.e^{-i\pi/4},\qquad \frac{e^{1+3\pi i}}{e^{-1+\pi i/2}},\qquad 2e^{3+i\pi/6}.

Also simplify (e3i−e−3i)/(2i)(e^{3i}-e^{-3i})/(2i) and find ∣ez∣|e^z| for z=4eiπ/3z=4e^{i\pi/3}.

Solution

The first three values are

1−i2,ie2,3e3+ie3.\frac{1-i}{\sqrt2},\qquad ie^2,\qquad \sqrt3e^3+ie^3.

Euler’s formula reduces the difference quotient to the real number sin⁡3\sin3. Finally, z=2+23iz=2+2\sqrt3i, so ∣ez∣=e2|e^z|=e^2; its modulus depends on the real part of zz, not on ∣z∣|z|.

ExercisePowers and a geometric sum

Write an=(1+i)na_n=(1+i)^n in polar form for integers n≥0n\ge0. Compute its first NN terms’ sum, indexed from 00 to N−1N-1, for N≥0N\ge0.

Solution

De Moivre’s formula gives

an=2n/2einπ/4.a_n=2^{n/2}e^{in\pi/4}.

For N≥1N\ge1, the geometric-sum identity gives

∑n=0N−1(1+i)n=(1+i)N−1i.\sum_{n=0}^{N-1}(1+i)^n=\frac{(1+i)^N-1}{i}.

For N=0N=0, the empty sum is zero, and the displayed right side is also zero. The points rotate by π/4\pi/4 and increase their distance from the origin by a factor 2\sqrt2 at each step.

ExerciseEnumerate roots without duplicates

Find all fourth roots of −16-16, all fifth roots of 11, and all square roots of i−1i-1.

Solution

The three solution sets are, respectively,

{2ei(π/4+kπ/2):k=0,1,2,3},\left\{2e^{i(\pi/4+k\pi/2)}:k=0,1,2,3\right\},{e2πik/5:k=0,1,2,3,4},\left\{e^{2\pi ik/5}:k=0,1,2,3,4\right\},

and

{21/4ei(3π/8+kπ):k=0,1}.\left\{2^{1/4}e^{i(3\pi/8+k\pi)}:k=0,1\right\}.

The first set can also be written {2+i2,−2+i2,−2−i2,2−i2}\{\sqrt2+i\sqrt2,-\sqrt2+i\sqrt2,-\sqrt2-i\sqrt2,\sqrt2-i\sqrt2\}.

ExerciseA roots-of-unity filter

Let n≥1n\ge1, ω=e2πi/n\omega=e^{2\pi i/n}, and m∈Zm\in\mathbb Z. Evaluate

∑k=0n−1ωmk.\sum_{k=0}^{n-1}\omega^{mk}.
Solution

If nn divides mm, every term is 11, so the sum is nn. Otherwise ωm≠1\omega^m\ne1, and the geometric-sum formula gives zero because (ωm)n=1(\omega^m)^n=1. This includes negative mm and the case n=1n=1. Thus equally spaced complex phases distinguish divisibility by nn, an identity that later appears in the discrete Fourier transform.

ExerciseProducts of roots of unity

Suppose um=1u^m=1 and vn=1v^n=1, where m,n≥1m,n\ge1. Prove (uv)mn=1(uv)^{mn}=1. Does this require m,nm,n to be coprime? If u,vu,v are primitive of orders m,nm,n and gcd⁡(m,n)=1\gcd(m,n)=1, show that uvuv is primitive of order mnmn.

Solution

Commutativity gives (uv)mn=(um)n(vn)m=1(uv)^{mn}=(u^m)^n(v^n)^m=1, without a coprimality assumption.

For the stronger claim, suppose (uv)k=1(uv)^k=1 for k>0k>0. Raising to the nnth power yields ukn=1u^{kn}=1. Thus mm divides knkn, and coprimality implies mm divides kk. Similarly, nn divides kk, hence mnmn divides kk. Since the mnmnth power is already 11, the order is exactly mnmn.

ExerciseFactor over the complex numbers

Factor each polynomial completely:

z2+z+3,z3+3z2+7z+5,z^2+z+3,\qquad z^3+3z^2+7z+5,

and

z3−iz2−4z+4i.z^3-iz^2-4z+4i.
Solution

The quadratic has roots r±=(−1±i11)/2r_\pm=(-1\pm i\sqrt{11})/2. Its factorization is

z2+z+3=(z−r+)(z−r−).z^2+z+3=(z-r_+)(z-r_-).

For the real cubic, −1-1 is a root, and division gives

z3+3z2+7z+5=(z+1)(z2+2z+5).\begin{aligned} &z^3+3z^2+7z+5\\ &\quad=(z+1)(z^2+2z+5). \end{aligned}

Its complete factorization is (z+1)(z+1−2i)(z+1+2i)(z+1)(z+1-2i)(z+1+2i).

For the complex-coefficient cubic, group the terms:

z3−iz2−4z+4i=(z−i)(z2−4)=(z−i)(z−2)(z+2).\begin{aligned} &z^3-iz^2-4z+4i\\ &\quad=(z-i)(z^2-4)\\ &\quad=(z-i)(z-2)(z+2). \end{aligned}

Its root ii does not imply a root −i-i, since its coefficients are not all real.

ExerciseRecover a real cubic

A real-coefficient cubic PP has roots 2+i2+i and 11, and P(0)=10P(0)=10. Find PP.

Solution

The conjugate root is 2−i2-i. Therefore

P(z)=A(z2−4z+5)(z−1)P(z)=A(z^2-4z+5)(z-1)

with nonzero real AA. At zero, −5A=10-5A=10, so A=−2A=-2. Expanding gives

P(z)=−2z3+10z2−18z+10.P(z)=-2z^3+10z^2-18z+10.
ExerciseCheck the hypothesis before pairing roots

Suppose 1+i1+i is a root of

P(z)=z3+az2+bz+10−6i,P(z)=z^3+az^2+bz+10-6i,

where a,b∈Ra,b\in\mathbb R. Find a,ba,b. Is 1−i1-i also a root?

Solution

Using (1+i)2=2i(1+i)^2=2i and (1+i)3=−2+2i(1+i)^3=-2+2i gives

P(1+i)=(8+b)+i(2a+b−4).P(1+i)=(8+b)+i(2a+b-4).

Equating the real and imaginary parts to zero gives b=−8b=-8 and a=6a=6. The constant coefficient is complex, so the conjugate-root theorem does not apply. Direct substitution gives P(1−i)=−12i≠0P(1-i)=-12i\ne0.

ExerciseRecover every exponent

For a=reiθ≠0a=re^{i\theta}\ne0, solve ew=ae^w=a. Explain what happens if a=0a=0.

Solution

Write w=x+iyw=x+iy. Comparing moduli gives ex=re^x=r, hence x=ln⁡rx=\ln r. Comparing directions gives y=θ+2πky=\theta+2\pi k. Therefore all solutions are

w=ln⁡r+i(θ+2πk),k∈Z.w=\ln r+i(\theta+2\pi k),\qquad k\in\mathbb Z.

There is no solution when a=0a=0, since a complex exponential is never zero.

Exploration: what does a fractional power mean?

Integer powers are unambiguous because they use repeated multiplication and inverses. A fractional power can mean a chosen principal value or a set of algebraic values. State the convention before using exponent laws.

Principal values, all values, and a failed exponent law

For z≠0z\ne0, define the pointwise principal logarithm by

Log⁡z=ln⁡∣z∣+iArg⁡z.\operatorname{Log}z=\ln|z|+i\operatorname{Arg}z.

A principal value of a real power is

zprincipalα=exp⁡(αLog⁡z).z^\alpha_{\mathrm{principal}} =\exp\bigl(\alpha\operatorname{Log}z\bigr).

This selects one value using our argument convention. It does not make every real exponent law valid. For instance,

((−1)2)principal1/2=1,\bigl((-1)^2\bigr)^{1/2}_{\mathrm{principal}}=1,

whereas (−1)principal2(1/2)=−1(-1)^{2(1/2)}_{\mathrm{principal}}=-1. Taking an intermediate principal argument discarded a full turn. This pointwise logarithm is also discontinuous across the negative real axis; choosing a continuous logarithm on a suitable domain belongs to complex analysis.

For a positive rational exponent m/nm/n in lowest terms, another convention is the set

{um:un=z}.\{u^m:u^n=z\}.

Writing z=reiθz=re^{i\theta} yields the nn distinct values

rm/nexp⁡(im(θ+2πk)n).r^{m/n}\exp\left(i\frac{m(\theta+2\pi k)}n\right).

Here k=0,1,…,n−1k=0,1,\ldots,n-1.

These are exactly the roots of vn=zmv^n=z^m. Indeed, those roots have angles (mθ+2πj)/n(m\theta+2\pi j)/n, and multiplication by mm permutes the residue classes modulo nn because gcd⁡(m,n)=1\gcd(m,n)=1.

The lowest-terms condition matters. For m=n=2m=n=2 and z=1z=1, squaring the square roots of 11 gives only {1}\{1\}, while v2=12v^2=1^2 has roots {1,−1}\{1,-1\}. Reduce the exponent before using this description.

For example, the two algebraic values of (1−i)3/2(1-i)^{3/2} are

±23/4e−3πi/8.\pm2^{3/4}e^{-3\pi i/8}.

The principal value is the one with the plus sign, obtained from Arg⁡(1−i)=−π/4\operatorname{Arg}(1-i)=-\pi/4. Both values square to (1−i)3=−2−2i(1-i)^3=-2-2i.

References

  1. [1] J. Orloff, “Topic 1: Complex Algebra and the Complex Plane,” 2018. MIT OpenCourseWare, 18.04 Complex Variables with Applications, Spring 2018. https://ocw.mit.edu/courses/18-04-complex-variables-with-applications-spring-2018/resources/mit18_04s18_topic1/ ↩