Complex analysis basics

In short

Complex analysis is the calculus of functions of a complex variable z = x + iy, built on the surprising fact that mere differentiability at every point of a region (analyticity) forces a function to be infinitely differentiable, to have a power series, and to be fully determined by its boundary values. Its central tools — the Cauchy–Riemann equations, Cauchy's integral theorem and formula, Laurent series and the residue theorem — turn hard real integrals into easy algebra and give conformal maps used in fluid flow, electrostatics and aerofoil theory. In Indian universities it is a core paper in B.Sc/M.Sc Mathematics (UGC CBCS) and in engineering mathematics; it carries steady marks in GATE MA (and GATE engineering papers), CSIR-UGC NET, IIT JAM MA, TIFR and ISI entrance. In placement and interview settings it appears through quantitative aptitude on complex roots, FFT/signal-processing questions in analytics and EDA firms, and transfer-function stability questions (poles, Nyquist contour) in control and semiconductor roles.

Explained

Complex numbers, the plane and topology

• Field C: z = x + iy, i² = −1; x = Re z, y = Im z. Conjugate z̄ = x − iy, modulus |z| = √(x²+y²), |z|² = z z̄.
• Polar/Euler form: z = r(cos θ + i sin θ) = r e^{iθ}, r = |z|, θ = arg z. Principal argument Arg z ∈ (−π, π]; arg z = Arg z + 2kπ is multi-valued.
• De Moivre: (cos θ + i sin θ)ⁿ = cos nθ + i sin nθ; n-th roots of z lie on a circle of radius r^{1/n}, equally spaced by 2π/n.
• Triangle inequalities: |z₁+z₂| ≤ |z₁|+|z₂| and ||z₁|−|z₂|| ≤ |z₁−z₂|. C is a complete metric space with d(z,w)=|z−w| but is not ordered — "z₁ < z₂" is meaningless.
• Regions: open disc |z−a| < r (neighbourhood), domain = open + connected, simply connected = no holes (every closed curve shrinks to a point).
• Extended plane C ∪ {∞} = Riemann sphere by stereographic projection; circles/lines on the plane ↔ circles on the sphere.

Analytic functions and the Cauchy–Riemann equations

• Limit and continuity: as in R², but the limit must be the same along *every* path.
• Derivative: f′(z₀) = lim_{h→0} [f(z₀+h) − f(z₀)]/h with h complex — the quotient must tend to one value from all directions. This is far stronger than real differentiability.
• Analytic (holomorphic) at z₀: differentiable in some neighbourhood of z₀. Entire: analytic on all of C (e.g. polynomials, e^z, sin z, cos z).
• Necessary condition (C–R): if f = u + iv is differentiable at z, then u_x = v_y and u_y = −v_x at that point.
  • Derivation outline: take h real → f′ = u_x + i v_x; take h purely imaginary → f′ = v_y − i u_y; equate.
• Sufficient condition: if u, v have continuous first partials in a neighbourhood and satisfy C–R there, f is analytic. (C–R alone is *not* enough — the standard counterexample is f(z) = z̄²/z (with f(0) = 0) type functions, or f = (x³−y³ + i(x³+y³))/(x²+y²) at 0, where C–R holds at 0 but f is not differentiable.)
• Polar C–R: u_r = (1/r)v_θ, v_r = −(1/r)u_θ.
• Consequences: f′ = 0 on a domain ⇒ f constant; |f| constant ⇒ f constant; f analytic and real-valued on a domain ⇒ f constant. f cannot depend on z̄ — a useful test: write in z, z̄ and check ∂f/∂z̄ = 0.
• Harmonic functions: u analytic-part ⇒ u_xx + u_yy = 0. Given harmonic u, the harmonic conjugate v is found by integrating C–R (or by Milne–Thomson: f(z) = ∫[u_x(z,0) − i u_y(z,0)]dz + C).
• Nowhere analytic examples: z̄, |z|², Re z, Im z (they fail C–R except possibly at isolated points, and a single point is not a neighbourhood).

Elementary functions and branches

• e^z = e^x(cos y + i sin y): entire, period 2πi, never zero, |e^z| = e^x.
• sin z = (e^{iz} − e^{−iz})/2i, cos z = (e^{iz}+e^{−iz})/2: entire but unbounded (sin(iy) = i sinh y) — unlike the real case; zeros of sin z are z = nπ only.
• Hyperbolic links: sin(iz) = i sinh z, cos(iz) = cosh z, cosh²z − sinh²z = 1.
• log z = ln|z| + i arg z: infinitely many values; principal Log z = ln|z| + i Arg z, analytic on C minus the branch cut along (−∞, 0]; derivative 1/z. z = 0 is a branch point.
• Powers: z^a = e^{a Log z} (principal); multi-valued in general, so iⁱ = e^{−π/2} is only the principal value.
• Branch cut discipline: pick the branch *before* computing; log(z₁z₂) = log z₁ + log z₂ holds only modulo 2πi.

Complex integration: Cauchy's theorem and formula

• Contour integral: ∫_C f(z)dz = ∫_a^b f(z(t)) z′(t)dt along a piecewise-smooth path; reversing orientation changes the sign.
• ML inequality: |∫_C f dz| ≤ M·L where M = max|f| on C, L = length — the standard tool for proving a contour contribution vanishes.
• Cauchy–Goursat theorem: f analytic on and inside a simple closed contour C in a simply connected domain ⇒ ∮_C f dz = 0. Goursat's refinement needs only analyticity, not continuity of f′.
• Path independence / antiderivative: in a simply connected domain, ∫ f dz depends only on endpoints; ∫_C f = F(z₂) − F(z₁).
• Deformation / extended theorem: for a region between contours, ∮_outer = Σ ∮_inner (same orientation) — this is why ∮ dz/z = 2πi around *any* contour enclosing 0.
• Cauchy's integral formula: f(a) = (1/2πi)∮_C f(z)/(z−a) dz; derivative form f^{(n)}(a) = (n!/2πi)∮_C f(z)/(z−a)^{n+1} dz. So an analytic function is infinitely differentiable and its boundary values determine it inside.
• Standard corollaries:
  • Morera: ∮ f = 0 for every closed contour in a domain + continuity ⇒ f analytic (converse of Cauchy).
  • Cauchy's inequality: |f^{(n)}(a)| ≤ n! M / Rⁿ on |z−a| = R.
  • Liouville: a bounded entire function is constant ⇒ gives a one-line proof of the Fundamental Theorem of Algebra.
  • Maximum modulus principle: a non-constant analytic f on a domain attains no interior maximum of |f|; the maximum is on the boundary (basis of many estimate questions).
  • Open mapping, Schwarz lemma, Rouché's theorem (counts zeros), Argument principle: (1/2πi)∮ f′/f dz = Z − P.

Series, singularities and residues

• Taylor series: f analytic in |z−a| < R ⇒ f(z) = Σ a_n(z−a)ⁿ with a_n = f^{(n)}(a)/n!; radius of convergence = distance from a to the nearest singularity.
• Laurent series: f analytic in an annulus r < |z−a| < R ⇒ f(z) = Σ_{n=−∞}^{∞} c_n (z−a)ⁿ; the negative powers form the principal part.
• Classification of isolated singularities at z = a:
  • Removable: no principal part; lim f exists (e.g. sin z / z at 0).
  • Pole of order m: principal part ends at (z−a)^{−m}; |f| → ∞. m = 1 is a simple pole.
  • Essential: infinitely many negative powers (e.g. e^{1/z} at 0); by Casorati–Weierstrass the image of every punctured neighbourhood is dense in C.
• Non-isolated / branch-point singularities: z = 0 for log z or √z; cosec(1/z) has a non-isolated essential singularity at 0.
• Singularity at infinity: classify by the behaviour of f(1/w) at w = 0.
• Residue: Res_{z=a} f = c_{−1} in the Laurent expansion.
  • Simple pole: Res = lim_{z→a}(z−a)f(z); for f = p/q with q simple zero: Res = p(a)/q′(a).
  • Pole of order m: Res = (1/(m−1)!) lim_{z→a} d^{m−1}/dz^{m−1}[(z−a)^m f(z)].
  • Essential singularity: expand the Laurent series and read off c_{−1}.
• Residue theorem: ∮_C f dz = 2πi Σ residues at singularities inside C.
• Applications to real integrals:
  1. Rational on (−∞,∞): ∫ P/Q dx with deg Q ≥ deg P + 2 — semicircular contour in the upper half-plane; the arc vanishes by ML.
  2. Trigonometric over [0,2π]: put z = e^{iθ}, cos θ = (z + 1/z)/2, sin θ = (z − 1/z)/2i, dθ = dz/(iz) over |z| = 1.
  3. Fourier-type ∫ f(x)cos mx or sin mx dx via e^{imz} and Jordan's lemma.
  4. Poles on the real axis: use the Cauchy principal value and indent the contour by a small semicircle (contributes ±iπ·Res).
  5. Branch-cut integrals ∫₀^∞ x^{a−1}/(1+x)dx by keyhole contours.

Conformal mapping and bilinear transformations

• Conformal map: analytic with f′(z₀) ≠ 0 preserves angles in magnitude and sense; |f′| is the local scale factor. Points where f′ = 0 are critical points (angles multiplied by the order+1).
• Basic maps: translation z+b; rotation+magnification az; inversion 1/z; z² (doubles angles at 0, maps a quadrant to a half-plane); e^z (maps horizontal strips to sectors); z + 1/z (Joukowski — circle to aerofoil); Log z (sector to strip).
• Bilinear (Möbius) transformation w = (az+b)/(cz+d), ad − bc ≠ 0: a bijection of the Riemann sphere, composition of translation, rotation, inversion and scaling.
  • Circles and lines map to circles and lines (circline-preserving); cross-ratio is invariant, which lets three given points be mapped to three chosen points.
  • Fixed points solve cz² + (d−a)z − b = 0 (at most two unless the map is the identity).
  • Standard maps: w = (z−i)/(z+i) takes the upper half-plane to the unit disc; w = (z−a)/(1−āz) is a disc automorphism.
• Why it matters: Laplace's equation is preserved by conformal maps, so a hard boundary-value problem in an awkward region is solved in a disc or half-plane and mapped back — used in potential flow, electrostatics and heat conduction.
• Riemann mapping theorem: any simply connected domain (≠ C) maps conformally onto the unit disc.
• Limitations: conformal mapping needs 2-D (plane) problems and linear (Laplace-type) equations; it does not extend to 3-D or to non-linear PDEs, and finding the explicit map for a general region is usually impossible in closed form — hence numerical conformal mapping and, more often, finite-element methods in practice.

Key points

• Analytic - complex-differentiable in a neighbourhood; forces infinite differentiability and a convergent power series.
• C–R equations - u_x = v_y, u_y = −v_x; necessary always, sufficient with continuous partials.
• Harmonic - real and imaginary parts of an analytic function satisfy Laplace's equation ∇²u = 0.
• z̄ dependence - any function genuinely involving z̄, Re z, Im z or |z| is nowhere analytic.
• Cauchy–Goursat - ∮ f dz = 0 for f analytic on and inside a simple closed contour in a simply connected domain.
• Cauchy integral formula - boundary values determine interior values; gives f^{(n)} by differentiating under the integral.
• Liouville - bounded + entire ⇒ constant; yields the Fundamental Theorem of Algebra.
• Maximum modulus - |f| peaks only on the boundary for non-constant analytic f.
• Taylor radius - equals the distance from the centre to the nearest singularity.
• Laurent series - valid in an annulus; its principal part classifies the singularity.
• Singularity types - removable (no principal part), pole of order m (finite principal part), essential (infinite).
• Residue - the c_{−1} coefficient; simple pole ⇒ lim (z−a)f(z), or p(a)/q′(a).
• Residue theorem - ∮_C f = 2πi × (sum of enclosed residues); the engine for real integrals.
• Jordan's lemma / ML inequality - the standard way to kill the large semicircular arc.
• Argument principle / Rouché - counts zeros minus poles inside a contour; basis of Nyquist stability.
• Conformal - f analytic with f′ ≠ 0 preserves angles; |f′| is the local magnification.
• Bilinear map - (az+b)/(cz+d), ad−bc ≠ 0; maps circlines to circlines and preserves cross-ratio.
• sin z, cos z unbounded - true in C though bounded on R; e^z is 2πi-periodic and never zero.

Formulas and examples

(All quantities are dimensionless pure numbers; angles are in radians. In applications z carries the unit of the physical plane — e.g. metres in a flow field — and |f′| is then a dimensionless magnification.)

Example 1 — Analyticity, harmonic conjugate and f(z). Check u = x² − y² − 2x is harmonic and find the analytic f = u + iv.

• u_x = 2x − 2, u_y = −2y; u_xx = 2, u_yy = −2 ⇒ u_xx + u_yy = 0, so u is harmonic.
• C–R: v_y = u_x = 2x − 2 ⇒ v = 2xy − 2y + g(x). Then v_x = 2y + g′(x) and C–R needs v_x = −u_y = 2y ⇒ g′(x) = 0.
• So v = 2xy − 2y + c and f(z) = (x²−y²−2x) + i(2xy−2y) + ic = z² − 2z + ic.
• Check by Milne–Thomson: u_x(z,0) = 2z − 2, u_y(z,0) = 0 ⇒ f = ∫(2z−2)dz = z² − 2z + C. ✔

Example 2 — Laurent series and singularity type. Expand f(z) = 1/[z(z−2)] about z = 0 for 0 < |z| < 2, and classify the singularity.

• 1/[z(z−2)] = (1/2)[1/(z−2)] − (1/2)(1/z)·… — do it by partial fractions: 1/[z(z−2)] = (1/2)/(z−2) − (1/2)/z.
• For |z| < 2: 1/(z−2) = −(1/2)·1/(1 − z/2) = −(1/2)Σ_{n≥0}(z/2)^n.
• So f(z) = −(1/2z) − (1/4)Σ_{n≥0} zⁿ/2ⁿ = −1/(2z) − 1/4 − z/8 − z²/16 − …
• Principal part is the single term −1/(2z) ⇒ z = 0 is a simple pole with Res = −1/2 (confirming lim_{z→0} z f(z) = 1/(0−2) = −1/2). ✔

Example 3 — Residue theorem on a contour integral. Evaluate ∮_C (z² + 1)/[(z−1)(z−3)] dz where C is |z| = 2 (positively oriented).

• Poles at z = 1 (inside, |1| < 2) and z = 3 (outside). Only z = 1 counts.
• Res_{z=1} = (1²+1)/(1−3) = 2/(−2) = −1.
• ∮ = 2πi(−1) = −2πi.

Example 4 — A real integral by contour integration. Evaluate I = ∫_{−∞}^{∞} dx/(x⁴ + 1).

• f(z) = 1/(z⁴+1); poles where z⁴ = −1 ⇒ z = e^{iπ/4}, e^{i3π/4}, e^{i5π/4}, e^{i7π/4}. Upper-half poles: z₁ = e^{iπ/4}, z₂ = e^{i3π/4}.
• Simple poles of p/q with q = z⁴+1, q′ = 4z³ ⇒ Res = 1/(4z³) = z/(4z⁴) = z/(4·(−1)) = −z/4 at each pole.
• Sum = −(z₁+z₂)/4 = −(e^{iπ/4} + e^{i3π/4})/4 = −[(√2/2 + i√2/2) + (−√2/2 + i√2/2)]/4 = −(i√2)/4.
• Semicircular arc of radius R: |f| ≤ 1/(R⁴−1), length πR ⇒ bound πR/(R⁴−1) → 0. So I = 2πi × (−i√2/4) = 2π√2/4 = π/√2 ≈ 2.2214.

Example 5 — Trigonometric integral by the unit circle. Evaluate I = ∫₀^{2π} dθ/(5 + 4cos θ).

• Put z = e^{iθ}: cos θ = (z + z^{−1})/2, dθ = dz/(iz), over |z| = 1.
• I = ∮ [1/(5 + 2(z + 1/z))] dz/(iz) = ∮ dz/[i(5z + 2z² + 2)] = (1/i)∮ dz/(2z² + 5z + 2).
• 2z² + 5z + 2 = (2z + 1)(z + 2) ⇒ poles z = −1/2 (inside) and z = −2 (outside).
• Res at z = −1/2 of 1/[(2z+1)(z+2)] = 1/[2(z+2)]|_{z=−1/2} = 1/[2(1.5)] = 1/3.
• I = (1/i)·2πi·(1/3) = 2π/3 ≈ 2.0944. (Check against the standard result 2π/√(a²−b²) with a = 5, b = 4: 2π/3. ✔)

Example 6 — n-th roots. Find all cube roots of 8i.

• 8i = 8 e^{iπ/2} ⇒ roots = 2 e^{i(π/2 + 2kπ)/3}, k = 0,1,2 ⇒ 2e^{iπ/6}, 2e^{i5π/6}, 2e^{i3π/2}.
• In rectangular form: √3 + i, −√3 + i, −2i. (Check: (−2i)³ = −8i³ = 8i ✔)

Where it is used

• Signal processing and communications (Qualcomm, Samsung R&D Bengaluru, Tejas Networks, ISRO): the z-transform and Fourier transform live in C; the region of convergence is an annulus and system stability is read off pole locations inside |z| = 1.
• Control systems and power engineering (BHEL, L&T, Siemens, ABB, NTPC design teams): the Nyquist criterion is the argument principle applied to a contour in the s-plane; root-locus work is pole–zero geometry.
• Aerospace and fluid dynamics (HAL, NAL Bengaluru, ISRO, aerodynamics groups in IITs): complex potential w = φ + iψ and the Joukowski transform z + 1/z for aerofoil lift; conformal mapping for potential flow.
• Electrostatics, heat conduction and EM simulation: conformal mapping solves 2-D Laplace problems for capacitor edges, microstrip lines and waveguide cross-sections (used in VLSI/EDA work at Cadence, Synopsys, Intel India).
• Quantitative finance and analytics: characteristic functions, Fourier-transform option pricing (Carr–Madan) and the Laplace inversion used by quant desks and risk teams.
• Academia, exams and teaching roles: a compulsory unit in B.Sc/M.Sc Mathematics (UGC CBCS), engineering mathematics under the AICTE model curriculum, and a scoring section in GATE MA, CSIR-UGC NET, IIT JAM MA and state SET/assistant-professor recruitment.

Interview questions

• Q: What does it mean for a function to be analytic, and why is it so much stronger than real differentiability? A: It is complex-differentiable in a whole neighbourhood, so the difference quotient has the same limit from every direction; that single condition forces infinite differentiability and a convergent power series.
• Q: Are the Cauchy–Riemann equations sufficient for differentiability? A: No — they are necessary; they become sufficient only when the first partial derivatives of u and v are continuous in a neighbourhood. There are standard functions satisfying C–R at a point yet not differentiable there.
• Q: Why is f(z) = z̄ nowhere analytic? A: Here u = x, v = −y, so u_x = 1 but v_y = −1; C–R fails everywhere. Equivalently ∂f/∂z̄ ≠ 0.
• Q: Can sin z exceed 1 in magnitude? A: Yes — sin(iy) = i sinh y grows without bound, so sin z and cos z are unbounded on C even though they are bounded on R.
• Q: Why is ∮ dz/z = 2πi around the unit circle when the integrand is "nice" everywhere on the path? A: Because 1/z is not analytic *inside* the contour — there is a pole at 0 — so Cauchy's theorem does not apply; the residue there is 1.
• Q: State Liouville's theorem and one consequence. A: A bounded entire function is constant; it gives an immediate proof of the Fundamental Theorem of Algebra, since 1/p(z) would otherwise be bounded and entire for a non-constant polynomial with no root.
• Q: How do you find the radius of convergence of a Taylor series without computing coefficients? A: It is the distance from the centre of expansion to the nearest singularity of the function.
• Q: What is the difference between a pole and an essential singularity? A: A pole has a finite principal part and |f| → ∞ as z → a; an essential singularity has infinitely many negative-power terms and no limit at all, with f taking values dense in C nearby.
• Q: Why does ∫_{−∞}^{∞} P(x)/Q(x) dx need deg Q ≥ deg P + 2 for the semicircle method? A: Otherwise the ML bound on the large arc, roughly R·R^{degP−degQ}, does not tend to zero, so the arc contribution cannot be discarded.
• Q: What makes a map conformal, and why do engineers care? A: Analyticity with f′ ≠ 0, which preserves angles; because conformal maps preserve harmonicity, a Laplace boundary-value problem on an awkward region can be transported to a disc or half-plane and solved there.
• Q: How many fixed points can a bilinear transformation have? A: At most two, from the quadratic cz² + (d−a)z − b = 0, unless it is the identity, which fixes everything.
• Q: Connect the argument principle to control engineering. A: Encircling a contour with f = 1 + G(s)H(s) counts Z − P, zeros minus poles inside; that count is exactly the Nyquist stability criterion for closed-loop poles in the right half-plane.
• Q: Why does |f| constant on a domain force f constant? A: |f|² = u² + v² constant plus the C–R equations makes all partials vanish (unless f ≡ 0), so f′ = 0 on a connected domain and f is constant.

Common mistakes

• Concluding a function is analytic from the C–R equations alone, without checking continuity of the partials — and testing C–R at a single point rather than on a neighbourhood.
• Applying Cauchy's theorem when a singularity lies inside the contour, or when the domain is not simply connected (annulus problems need the deformation theorem).
• Treating log z, √z and z^a as single-valued: ignoring the branch cut gives answers off by 2πi factors, and log(z₁z₂) = log z₁ + log z₂ is only true modulo 2πi.
• Using the simple-pole formula lim (z−a)f(z) at a double or triple pole instead of the derivative formula with (m−1)!.
• Forgetting that only singularities inside the contour contribute, and missing one that lies exactly on it (a real-axis pole needs an indented contour and a principal value, contributing ±iπ·Res).
• Dropping the factor 2πi, or the orientation sign — a clockwise contour gives the negative of the residue sum.
• Omitting the proof that the large arc vanishes (ML inequality or Jordan's lemma) while claiming the real integral equals the residue sum, and assuming |z₁ + z₂| = |z₁| + |z₂| or trying to order complex numbers.

Related topics

Vector calculus · Numerical methods · Fourier and Laplace transforms · Real analysis: sequences, series and continuity · Partial differential equations and Laplace's equation · Linear algebra: eigenvalues and matrices