Module 25 · UPSC Optional: Mathematics

Mathematics Paper I: Core Topics

Updated 18 Nov 2026 · 8,208 words · 5 sections · illustrated
Updated 17 Nov 2026
Mathematics Paper I: core syllabus areas, key concepts, standard references and answer-writing practice.
Algebra · Calculus · Real Analysis · Complex Analysis
🎯 Take Mock Test
Ecology BasicsPopulation EcologyBiodiversityConservationClimate ChangeEnvironmentalPollutionEnvironmentalPolicies & ActsInternational Orgs(UNEP, WHO)India Initiatives(Swachh, NAPCC)supportsdrivesimpactsexacerbatesregulatesguidesimplementsEcosystem = biotic + abioticKey policy hooks: NGT Act, Paris Accordvlymbooq
Comprehensive Flow of Core Ecology Topics for GS3 Environment — Shows the logical progression of key ecology concepts and their policy links for quick revision. · Open the full diagram →

Learning Objectives

  • Understand the fundamental concepts of Mathematics Paper I: Core Topics
  • Apply key formulas and techniques to solve problems
  • Practice with exam-level questions to build speed and accuracy

Vector Spaces and Linear Transformations

1.1 Vector Spaces over Real and Complex Fields

A vector space over a field F (which for the UPSC syllabus means either the real numbers R or the complex numbers C) is a set V equipped with two operations: vector addition and scalar multiplication. These operations must satisfy eight axioms that guarantee the algebraic structure behaves like the familiar Euclidean space Rn. The axioms ensure closure, associativity, commutativity of addition, existence of an additive identity (the zero vector) and additive inverses, and compatibility of scalar multiplication with field multiplication. When the field is R, we speak of a real vector space; when it is C, a complex vector space. The distinction matters because eigenvalues of a linear operator on a real vector space may not exist in R but always exist in C by the Fundamental Theorem of Algebra, which is why complexification is a standard technique in spectral theory.

1.2 Subspaces, Linear Independence, and Basis

A subset W of a vector space V is a subspace if it is non-empty and closed under addition and scalar multiplication; equivalently, W is itself a vector space under the inherited operations. The span of a set of vectors is the smallest subspace containing them. A set of vectors is linearly independent if no non-trivial linear combination yields the zero vector; otherwise it is linearly dependent. A basis of V is a linearly independent spanning set. Every vector space has a basis (assuming the Axiom of Choice), and all bases of a given vector space have the same cardinality, which defines the dimension of V. The dimension theorem states that for any subspace W of a finite-dimensional space V, dim W + dim V/W = dim V, where V/W is the quotient space whose elements are cosets v + W. This quotient construction is essential for understanding the first isomorphism theorem for vector spaces, which says that the image of a linear transformation is isomorphic to the quotient of the domain by the kernel.

1.3 Linear Transformations, Kernel, and Image

A linear transformation T : V → W between vector spaces over the same field preserves addition and scalar multiplication: T(u + v) = T(u) + T(v) and T(αv) = αT(v). The kernel (or null space) of T is the set of vectors mapped to zero; it is a subspace of V. The image (or range) is the set of all outputs; it is a subspace of W. The rank-nullity theorem, a cornerstone of linear algebra, asserts that dim ker T + dim im T = dim V when V is finite-dimensional. This theorem connects the injectivity of T (kernel is trivial) with surjectivity (image equals W) when dim V = dim W: in that case, T is injective if and only if it is surjective if and only if it is bijective. The matrix representation of T depends on the choice of bases for V and W; a change of basis corresponds to multiplying the matrix on the left and right by invertible transition matrices. If A is the matrix of T relative to bases B and C, and P, Q are the transition matrices to new bases B' and C', then the new matrix is Q-1AP.

1.4 Eigenvalues, Eigenvectors, and the Characteristic Polynomial

For a linear operator T : V → V on a finite-dimensional space, a scalar λ is an eigenvalue if there exists a non-zero vector v such that T(v) = λv; such a v is an eigenvector. The eigenvalues are precisely the roots of the characteristic polynomial p(λ) = det(T - λI), which is a monic polynomial of degree dim V. Over C, the polynomial splits completely, guaranteeing at least one eigenvalue; over R, it may not split, which is why real operators can lack eigenvalues (e.g., a rotation in R2). The algebraic multiplicity of an eigenvalue is its multiplicity as a root of the characteristic polynomial; the geometric multiplicity is the dimension of the corresponding eigenspace ker(T - λI). The geometric multiplicity never exceeds the algebraic multiplicity. The Cayley-Hamilton theorem states that every linear operator satisfies its own characteristic polynomial: p(T) = 0. This theorem is not merely a formal substitution; it implies that any polynomial in T can be reduced to a polynomial of degree less than dim V, and it provides a method to compute the inverse of an invertible operator as a polynomial in that operator.

Inner Product Spaces and Canonical Forms

2.1 Inner Products, Norms, and Orthogonality

An inner product on a vector space V over F (R or C) is a map ⟨·, ·⟩ : V × V → F that is linear in the first argument, conjugate symmetric (⟨u, v⟩ = ⟨v, u⟩), and positive definite (⟨v, v⟩ > 0 for v ≠ 0). It induces a norm ||v|| = √⟨v, v⟩ and a notion of angle via the Cauchy-Schwarz inequality |⟨u, v⟩| ≤ ||u|| ||v||. Two vectors are orthogonal if their inner product is zero. An orthonormal set is a set of mutually orthogonal unit vectors. The Gram-Schmidt orthogonalization process takes any linearly independent set and produces an orthonormal set with the same span; this algorithm is constructive and guarantees that every finite-dimensional inner product space has an orthonormal basis. Bessel's inequality states that for any orthonormal set {ei} and any vector v, the sum of squares of Fourier coefficients |⟨v, ei⟩|2 is bounded by ||v||2. When the orthonormal set is a basis, equality holds — this is Parseval's identity, which generalizes the Pythagorean theorem to infinite sums and underlies Fourier series theory.

2.2 Adjoint Operators, Self-Adjoint, Normal, and Unitary Operators

Given a linear operator T on a finite-dimensional inner product space, its adjoint T* is the unique operator satisfying ⟨Tu, v⟩ = ⟨u, T*v⟩ for all u, v. The matrix of T* relative to an orthonormal basis is the conjugate transpose of the matrix of T. An operator is self-adjoint (Hermitian) if T = T*; its eigenvalues are real, and eigenvectors corresponding to distinct eigenvalues are orthogonal. The Spectral Theorem for self-adjoint operators asserts that V has an orthonormal basis of eigenvectors of T, so T is orthogonally diagonalizable. A normal operator satisfies TT* = T*T; self-adjoint, skew-adjoint, and unitary operators are all normal. The Spectral Theorem extends to normal operators over C: a normal operator is unitarily diagonalizable. A unitary operator satisfies T*T = I; it preserves inner products and norms, and its eigenvalues lie on the unit circle in C. These classes of operators are central to quantum mechanics and numerical linear algebra, where stability and orthogonality are paramount.

2.3 Jordan and Rational Canonical Forms

Not every linear operator is diagonalizable. The Jordan canonical form provides the next best thing: over an algebraically closed field (like C), every operator can be represented by a block diagonal matrix where each block is a Jordan block — a square matrix with an eigenvalue λ on the diagonal, ones on the superdiagonal, and zeros elsewhere. The sizes of the Jordan blocks for a given eigenvalue are determined by the structure of the generalized eigenspaces ker(T - λI)k. The minimal polynomial of T is the monic polynomial of least degree that annihilates T; it divides the characteristic polynomial and has the same irreducible factors. The rational canonical form works over any field and uses companion matrices of the invariant factors (the non-constant diagonal entries of the Smith normal form of λI - A over the polynomial ring). While the Jordan form is more intuitive for analyzing powers and exponentials of matrices, the rational canonical form is field-independent and reveals the invariant factors, which are the similarity invariants of the matrix. Both forms are essential for solving systems of linear differential equations and for understanding the structure of finitely generated modules over a principal ideal domain.

Worked Example: Diagonalization and Spectral Theorem
Let A be the real symmetric matrix A = [[2, 1, 1], [1, 2, 1], [1, 1, 2]]. Find an orthogonal matrix P such that PTAP is diagonal, and write down the diagonal matrix.
Since A is real and symmetric, the Spectral Theorem guarantees an orthonormal basis of eigenvectors. The characteristic polynomial is det(A - λI) = det([[2-λ, 1, 1], [1, 2-λ, 1], [1, 1, 2-λ]]). Subtracting the first row from the second and third gives a block structure; expanding yields (2-λ)[(2-λ)2 - 1] - 1[(2-λ) - 1] + 1[1 - (2-λ)] = (2-λ)(λ2 - 4λ + 3) - (1-λ) - (1-λ) = (2-λ)(λ-1)(λ-3) + 2(λ-1) = (λ-1)[(2-λ)(λ-3) + 2] = (λ-1)[-λ2 + 5λ - 6 + 2] = (λ-1)(-λ2 + 5λ - 4) = -(λ-1)(λ-1)(λ-4) = -(λ-1)2(λ-4). The eigenvalues are λ = 1 (algebraic multiplicity 2) and λ = 4 (algebraic multiplicity 1). For λ = 4, solve (A - 4I)v = 0: [[-2, 1, 1], [1, -2, 1], [1, 1, -2]]v = 0. Adding all rows gives 0 = 0; the system reduces to v1 = v2 = v3. An eigenvector is v1 = (1, 1, 1); normalize to u1 = (1/√3, 1/√3, 1/√3). For λ = 1, solve (A - I)v = 0: [[1, 1, 1], [1, 1, 1], [1, 1, 1]]v = 0 ⇒ v1 + v2 + v3 = 0. This plane has dimension 2. Choose two orthogonal vectors in it, e.g., w2 = (1, -1, 0) and w3 = (1, 1, -2) (check: 1+1-2=0, and dot product with w2 is 1-1+0=0). Normalize: u2 = (1/√2, -1/√2, 0), u3 = (1/√6, 1/√6, -2/√6). The orthogonal matrix is P = [u1 u2 u3] (columns are eigenvectors). Then PTAP = diag(4, 1, 1).
Exam Insight: Canonical Forms in UPSC
UPSC frequently asks for the Jordan form of a specific 3×3 or 4×4 matrix with a repeated eigenvalue. Do not merely state the eigenvalues; you must determine the geometric multiplicity by computing the rank of A - λI. If the geometric multiplicity equals the algebraic multiplicity, the matrix is diagonalizable and the Jordan blocks are all 1×1. If the geometric multiplicity is one less, there is exactly one 2×2 Jordan block and the rest are 1×1. Always show the computation of the nullity explicitly — the examiner awards marks for the rank-nullity step, not just the final block diagram.

Real Numbers, Sequences, and Series

3.1 The Real Number System and Completeness

The real numbers R form a complete ordered field. The field axioms govern addition and multiplication; the order axioms introduce a total order compatible with the algebraic operations; the completeness axiom distinguishes R from Q. Several equivalent formulations of completeness exist: the least upper bound property (every non-empty subset bounded above has a supremum), the monotone convergence theorem (every bounded monotone sequence converges), the Bolzano-Weierstrass theorem (every bounded sequence has a convergent subsequence), the Cauchy criterion (every Cauchy sequence converges), and the Heine-Borel theorem (a subset of R is compact if and only if it is closed and bounded). The Archimedean property — for any real x there exists a natural n with n > x — follows from completeness and implies the density of Q and RQ in R. These properties are not merely abstract; they are the logical bedrock upon which the entire edifice of calculus rests. For instance, the Intermediate Value Theorem and the Extreme Value Theorem for continuous functions on closed intervals are direct consequences of the completeness of R via the least upper bound property.

3.2 Sequences: Convergence, Cauchy Criterion, and Limit Superior

A sequence (xn) in R converges to a limit L if for every ε > 0 there exists N such that |xn - L| < ε for all n ≥ N. The limit is unique. A sequence is Cauchy if for every ε > 0 there exists N such that |xm - xn| < ε for all m, n ≥ N. In R, a sequence converges if and only if it is Cauchy — this is the Cauchy completeness of R. The limit superior (lim sup) and limit inferior (lim inf) of a bounded sequence are defined as the greatest and least subsequential limits, respectively; they always exist in R for bounded sequences and provide a way to describe the asymptotic behavior of non-convergent sequences. The Bolzano-Weierstrass theorem guarantees that every bounded sequence has at least one convergent subsequence, so lim sup and lim inf are well-defined. A sequence converges if and only if its lim sup equals its lim inf. These concepts are indispensable for analyzing series and for the root and ratio tests.

3.3 Infinite Series: Convergence Tests and Rearrangement

An infinite series ∑ an is the sequence of partial sums SN = ∑n=1N an. The series converges if (SN) converges. Absolute convergence (∑ |an| converges) implies convergence, but the converse fails (conditional convergence). For series with non-negative terms, the comparison test, ratio test, root test, and integral test are standard tools. The ratio test examines lim |an+1/an|; if the limit is less than 1, the series converges absolutely; if greater than 1, it diverges; if equal to 1, the test is inconclusive. The root test uses lim sup |an|1/n and is strictly stronger than the ratio test. For alternating series ∑ (-1)n bn with bn ≥ 0 decreasing to zero, the Leibniz test guarantees convergence. Riemann's rearrangement theorem states that a conditionally convergent series can be rearranged to converge to any real number, or to diverge to ±∞; this dramatic result underscores why absolute convergence is the robust notion for infinite sums. Power series ∑ an xn have a radius of convergence R = 1/lim sup |an|1/n; inside the open interval (-R, R) they converge absolutely and uniformly on compact subsets, and they define analytic functions.

3.4 Uniform Convergence of Sequences and Series of Functions

When dealing with sequences of functions fn : D → R, pointwise convergence (for each x, fn(x) converges) is too weak to preserve continuity, differentiability, or integrability. Uniform convergence — supx∈D |fn(x) - f(x)| → 0 — is the correct notion. The uniform limit of continuous functions is continuous; if fn are differentiable, fn' converges uniformly, and fn converges at one point, then fn converges uniformly to a differentiable function f with f' = lim fn'. For integration, if fn are Riemann integrable and converge uniformly to f, then f is Riemann integrable and ∫ f = lim ∫ fn. The Weierstrass M-test gives a practical criterion: if |fn(x)| ≤ Mn for all x and ∑ Mn converges, then ∑ fn converges uniformly. Abel's theorem on power series states that if ∑ an Rn converges, then the power series converges uniformly on [0, R] and the sum is continuous at R from the left. These theorems justify term-by-term differentiation and integration of power series within their interval of convergence.

Comparison of Convergence Types for Sequences of Functions
PropertyPointwise ConvergenceUniform Convergence
Definition∀x ∈ D, ∀ε > 0, ∃N(x, ε) such that n ≥ N ⇒ |fn(x) - f(x)| < ε∀ε > 0, ∃N(ε) such that n ≥ N ⇒ supx∈D |fn(x) - f(x)| < ε
Preserves ContinuityNo (counterexample: xn on [0,1])Yes
Allows Termwise IntegrationNo (requires uniform convergence or dominated convergence)Yes (for Riemann integrals)
Allows Termwise DifferentiationNoOnly if derivatives also converge uniformly
Weierstrass M-test ApplicableNoYes (for series of functions)
Circuit TheoryElectromagneticFieldsPower SystemsControl SystemsElectrical MachinesfeedsdrivescontrolsArt. 1: Core Syllabus CoverageMechanism: Energy Conversion Flowvlymbooq
Electrical Engineering Core Topics Classification Diagram for GS1 — Shows the main core topics of Electrical Engineering Paper I and their interrelations. · Open the full diagram →

Continuity, Differentiability, and Riemann Integration

4.1 Continuity, Uniform Continuity, and the Intermediate Value Property

A function f : D → R is continuous at x0 ∈ D if limx→x0 f(x) = f(x0). The ε-δ definition makes this precise: for every ε > 0 there exists δ > 0 such that |x - x0| < δ implies |f(x) - f(x0)| < ε. Uniform continuity strengthens this by requiring δ to work for all points in the domain simultaneously: for every ε > 0 there exists δ > 0 such that for all x, y ∈ D, |x - y| < δ implies |f(x) - f(y)| < ε. Every continuous function on a compact set (closed and bounded interval in R) is uniformly continuous (Heine-Cantor theorem). The Intermediate Value Theorem states that if f is continuous on [a, b] and L lies between f(a) and f(b), then there exists c ∈ (a, b) with f(c) = L. This is a topological property of connected sets: the continuous image of a connected set is connected. The Extreme Value Theorem asserts that a continuous function on a compact set attains its maximum and minimum; this follows because the image is compact, hence closed and bounded in R.

4.2 Differentiation: Mean Value Theorems and Taylor's Theorem

The derivative f'(x0) = limh→0 (f(x0+h) - f(x0))/h captures the instantaneous rate of change. Differentiability implies continuity, but not conversely. Rolle's Theorem: if f is continuous on [a, b], differentiable on (a, b), and f(a) = f(b), then there exists c ∈ (a, b) with f'(c) = 0. Lagrange's Mean Value Theorem (LMVT) generalizes this: under the same hypotheses, there exists c ∈ (a, b) such that f'(c) = (f(b) - f(a))/(b - a). Cauchy's Mean Value Theorem extends LMVT to two functions: (f(b) - f(a)) g'(c) = (g(b) - g(a)) f'(c). These theorems are the engine behind Taylor's Theorem, which approximates a function by a polynomial. Taylor's Theorem with Lagrange remainder: if f is n+1 times differentiable on an interval containing a, then for each x there exists ξ between a and x such that f(x) = ∑k=0n f(k)(a) (x-a)k/k! + f(n+1)(ξ) (x-a)n+1/(n+1)!. The Cauchy remainder form is f(n+1)(ξ) (x-ξ)n (x-a)/n!. Taylor series (the infinite sum) converges to f(x) if and only if the remainder tends to zero; functions equal to their Taylor series are called analytic. L'Hôpital's rule for indeterminate forms 0/0 or ∞/∞ is a direct corollary of Cauchy's Mean Value Theorem.

4.3 Riemann Integration and the Fundamental Theorems

The Riemann integral of a bounded function f on [a, b] is defined via Darboux sums. For a partition P = {x0, ..., xn}, the lower sum L(P, f) = ∑ mi Δxi and upper sum U(P, f) = ∑ Mi Δxi, where mi = inf f and Mi = sup f on [xi-1, xi]. The lower integral is the supremum of lower sums; the upper integral is the infimum of upper sums. f is Riemann integrable if these are equal. A bounded function is Riemann integrable if and only if its set of discontinuities has measure zero (Lebesgue's criterion). The First Fundamental Theorem of Calculus: if f is integrable on [a, b] and F(x) = ∫ax f(t) dt, then F is continuous on [a, b] and differentiable at every point where f is continuous, with F'(x) = f(x). The Second Fundamental Theorem: if F' = f and f is integrable, then ∫ab f = F(b) - F(a). These theorems link differentiation and integration as inverse operations. The Mean Value Theorem for Integrals states that if f is continuous, there exists c ∈ [a, b] such that ∫ab f = f(c)(b-a).

4.4 Improper Integrals, Beta and Gamma Functions

Improper integrals extend the Riemann integral to unbounded intervals or unbounded functions. The integral ∫a∞ f is defined as limb→∞ ∫ab f if the limit exists. Convergence can be absolute (∫ |f| converges) or conditional. The comparison test: if 0 ≤ f ≤ g and ∫ g converges, then ∫ f converges. The Beta function B(p, q) = ∫01 tp-1 (1-t)q-1 dt (for p, q > 0) and the Gamma function Γ(z) = ∫0∞ tz-1 e-t dt (for Re(z) > 0) are fundamental special functions. They satisfy B(p, q) = Γ(p)Γ(q)/Γ(p+q) and Γ(z+1) = z Γ(z), with Γ(n) = (n-1)! for positive integers. The Gamma function analytically continues the factorial to complex numbers. These functions appear constantly in probability, statistics, and physics; UPSC often asks for evaluations of definite integrals that reduce to Beta or Gamma forms via substitutions like t = sin2 θ or t = x/(1+x).

Functions of Several Variables and Vector Analysis

5.1 Limits, Continuity, and Partial Derivatives in Rn

For a function f : Rn → R, the limit limx→a f(x) = L means for every ε > 0 there exists δ > 0 such that 0 < ||x - a|| < δ implies |f(x) - L| < ε. Continuity is defined analogously.

Multiple Integration and Change of Variables

The extension of the Riemann integral to functions of several variables begins with the definition of the double integral over a rectangle R = [a, b] × [c, d]. For a bounded function f : R → ℝ, the lower and upper Darboux sums are formed using partitions of R into sub‑rectangles. The function is integrable if the supremum of lower sums equals the infimum of upper sums; the common value is denoted ∬_R f(x, y) dA. Fubini’s theorem guarantees that if f is continuous on R, the iterated integrals ∫_a^b (∫_c^d f(x, y) dy) dx and ∫_c^d (∫_a^b f(x, y) dx) dy both exist and equal the double integral. For non‑rectangular regions, the domain is described by inequalities such as a ≤ x ≤ b, g₁(x) ≤ y ≤ g₂(x) (type I) or c ≤ y ≤ d, h₁(y) ≤ x ≤ h₂(y) (type II). The integral is then evaluated as an iterated integral with variable limits.

Change of variables generalises the substitution rule. Let T : (u, v) ↦ (x(u, v), y(u, v)) be a C¹ transformation mapping a region S in the uv‑plane onto R. If the Jacobian determinant J = ∂(x, y)/∂(u, v) = x_u y_v − x_v y_u is non‑zero on S, then ∬_R f(x, y) dx dy = ∬_S f(x(u, v), y(u, v)) |J| du dv. The absolute value of the Jacobian accounts for orientation reversal. Polar coordinates (x = r cos θ, y = r sin θ) give J = r, so dA = r dr dθ. Cylindrical and spherical coordinates in ℝ³ are obtained by successive applications of the same principle, with Jacobians r and ρ² sin φ respectively. The change‑of‑variables formula is essential for evaluating integrals over circles, spheres, ellipsoids, and for computing volumes, centres of mass, and moments of inertia.

Improper multiple integrals arise when the domain is unbounded or the integrand has singularities. Convergence is defined via limits of integrals over exhausting compact subsets. The comparison test extends: if 0 ≤ f ≤ g and ∬ g converges, then ∬ f converges. Absolute convergence (∬ |f| converges) implies convergence, but conditional convergence can occur. The Gamma and Beta functions appear naturally when evaluating integrals such as ∬x²+y²≤1 (1−x²−y²)α dx dy after switching to polar coordinates.

Vector Calculus: Gradient, Divergence, Curl and Integral Theorems

For a scalar field φ : ℝ³ → ℝ of class C¹, the gradient ∇φ = (∂φ/∂x, ∂φ/∂y, ∂φ/∂z) points in the direction of maximal increase and its magnitude equals the directional derivative in that direction. For a vector field F = (P, Q, R) of class C¹, the divergence div F = ∇·F = ∂P/∂x + ∂Q/∂y + ∂R/∂z measures the net outward flux per unit volume. The curl curl F = ∇×F = (∂R/∂y − ∂Q/∂z, ∂P/∂z − ∂R/∂x, ∂Q/∂x − ∂P/∂y) measures the infinitesimal rotation. Identities such as ∇·(∇×F) = 0 and ∇×(∇φ) = 0 hold for sufficiently smooth fields and are the differential forms of the integral theorems.

Green’s theorem in the plane relates a line integral around a simple closed curve C to a double integral over the region D it bounds: ∮_C (P dx + Q dy) = ∬_D (∂Q/∂x − ∂P/∂y) dA. Stokes’ theorem generalises this to surfaces in ℝ³: ∮_∂S F·dr = ∬_S (∇×F)·n dS, where ∂S is the positively oriented boundary of the oriented surface S. The divergence theorem (Gauss’s theorem) connects flux through a closed surface ∂V to the volume integral of divergence: ∬_∂V F·n dS = ∭_V (∇·F) dV. These theorems convert difficult surface or line integrals into simpler volume or double integrals and vice‑versa. They also provide physical interpretations: divergence theorem expresses conservation of mass in fluid flow; Stokes’ theorem relates circulation to vorticity.

Applications include computing work done by a force field, flux of an electric field (Gauss’s law), and verifying whether a vector field is conservative. A field F is conservative on a simply connected domain iff ∇×F = 0, in which case a scalar potential φ exists with F = ∇φ. The line integral of a conservative field depends only on endpoints, and the integral around any closed loop is zero. The Helmholtz decomposition states that any sufficiently smooth vector field decaying at infinity can be written uniquely as the sum of a gradient field and a curl field.

Evaluating a Double Integral via Polar Coordinates
Compute ∬_D (x² + y²) dA where D is the region bounded by the circles x² + y² = 1 and x² + y² = 4 in the first quadrant.
In polar coordinates x = r cos θ, y = r sin θ, the integrand becomes r². The Jacobian is r, so dA = r dr dθ. The region D corresponds to 1 ≤ r ≤ 2 and 0 ≤ θ ≤ π/2. The integral is ∫θ=0π/2 ∫r=12 r² · r dr dθ = ∫0π/2 dθ ∫12 r³ dr = (π/2) · [r⁴/4]12 = (π/2) · (16/4 − 1/4) = (π/2) · (15/4) = 15π/8.
Exam‑Day Strategy for Vector Calculus
Memorise the three integral theorems together with their hypotheses (piecewise smooth boundaries, orientation conventions). When a problem asks for a line integral, first check if the field is conservative; if so, find a potential and evaluate at endpoints. If not, decide whether Green, Stokes, or the divergence theorem simplifies the computation. Sketch the region or surface to verify orientation; a sign error is the most common source of lost marks.

Complex Analysis: Analytic Functions and Cauchy Integral Theory

A function f : Ω → ℂ defined on an open set Ω ⊂ ℂ is analytic (holomorphic) at z₀ ∈ Ω if the limit f'(z₀) = limh→0 (f(z₀+h)−f(z₀))/h exists. Analyticity on Ω means the derivative exists at every point of Ω. The Cauchy–Riemann equations provide a necessary and sufficient condition for differentiability when the partial derivatives of u = Re f and v = Im f are continuous: ∂u/∂x = ∂v/∂y and ∂u/∂y = −∂v/∂x. An analytic function is infinitely differentiable and equals its Taylor series in a neighbourhood of each point; its power series coefficients are given by f(n)(z₀)/n! = (1/2πi) ∮_γ f(z)/(z−z₀)n+1 dz, where γ is a small circle around z₀.

Cauchy’s integral theorem states that if f is analytic on a simply connected domain D and γ is a closed contour in D, then ∮_γ f(z) dz = 0. The proof uses Green’s theorem and the Cauchy–Riemann equations. Cauchy’s integral formula follows: for any z₀ inside γ, f(z₀) = (1/2πi) ∮_γ f(z)/(z−z₀) dz. This formula yields the derivative formula f(n)(z₀) = (n!/2πi) ∮_γ f(z)/(z−z₀)n+1 dz and implies that analytic functions are determined by their boundary values. Morera’s theorem gives a converse: if f is continuous on D and ∮_γ f = 0 for every closed triangle γ ⊂ D, then f is analytic on D.

Liouville’s theorem asserts that a bounded entire function (analytic on ℂ) must be constant. The proof uses the Cauchy estimate |f(n)(0)| ≤ n! M/R^n for |f| ≤ M on |z| = R and lets R → ∞. The fundamental theorem of algebra follows: a non‑constant polynomial has at least one zero. The maximum modulus principle states that if f is analytic and non‑constant on a domain D, then |f| cannot attain a maximum in the interior of D. The open mapping theorem says that a non‑constant analytic function maps open sets to open sets. These results are the backbone of complex function theory and are frequently tested through contour integration problems.

Complex Analysis: Residue Calculus and Evaluation of Real Integrals

An isolated singularity z₀ of an analytic function f is classified by the Laurent expansion f(z) = ∑n=−∞∞ a_n (z−z₀)^n in a punctured neighbourhood. If a_n = 0 for all n < 0, z₀ is a removable singularity; if finitely many negative‑power terms exist, it is a pole of order m (the largest n with a−m ≠ 0); if infinitely many negative powers appear, it is an essential singularity. The residue Res(f, z₀) is the coefficient a−1. For a pole of order m, Res(f, z₀) = (1/(m−1)!) limz→z₀ dm−1/dzm−1 [(z−z₀)^m f(z)]. Simple poles (m = 1) give Res(f, z₀) = limz→z₀ (z−z₀) f(z).

The residue theorem: if f is analytic on a domain except for isolated singularities z₁,…,z_k inside a positively oriented simple closed contour γ, then ∮_γ f(z) dz = 2πi ∑j=1^k Res(f, z_j). This reduces contour integrals to algebraic residue computations. To evaluate real integrals of the form ∫-∞∞ R(x) dx where R is a rational function with deg(denominator) ≥ deg(numerator)+2 and no poles on the real axis, close the contour with a large semicircle in the upper half‑plane. The integral over the semicircle vanishes as its radius → ∞, leaving the real integral equal to 2πi times the sum of residues in the upper half‑plane. For integrals involving trigonometric functions, the substitution z = eiθ converts ∫02π F(cos θ, sin θ) dθ into a contour integral around the unit circle.

Jordan’s lemma handles integrals of the form ∫-∞∞ f(x) ei a x dx with a > 0 by showing the contribution from the large semicircle in the upper half‑plane tends to zero if f(z) → 0 uniformly. Branch cuts are required for multivalued functions such as z^α or log z; the contour is indented to avoid the cut and the discontinuity across the cut yields the desired real integral. Typical UPSC problems ask for ∫0^∞ xp−1/(1+x) dx (Beta function) or ∫0^∞ cos(ax)/(x²+1) dx, both solved by keyhole or semicircular contours and residue summation.

Ordinary Differential Equations: Linear Systems and Stability

A first‑order linear system in ℝⁿ is written as x' = A(t) x + g(t), where A(t) is an n×n matrix and g(t) a vector function. For constant coefficients A, the homogeneous system x' = A x has fundamental matrix Φ(t) = eAt, where the matrix exponential is defined by the convergent series eAt = ∑k=0^∞ (A t)^k/k!. The general solution is x(t) = eAt c + ∫0^t eA(t−s) g(s) ds. Eigenvalues λ of A determine the qualitative behaviour: if all Re(λ) < 0, the origin is asymptotically stable; if any Re(λ) > 0, it is unstable; if some Re(λ) = 0 and the rest are negative, stability depends on the Jordan blocks (non‑trivial Jordan blocks for zero real part lead to polynomial growth).

For a 2×2 system, the phase portrait is classified by the trace τ = tr A and determinant Δ = det A. The characteristic polynomial λ² − τ λ + Δ = 0 gives eigenvalues λ = (τ ± √(τ²−4Δ))/2. Cases: Δ < 0 → saddle (unstable); Δ > 0, τ²−4Δ > 0 → node (stable if τ < 0, unstable if τ > 0); Δ > 0, τ²−4Δ < 0 → spiral (stable if τ < 0, unstable if τ > 0); Δ > 0, τ = 0 → centre (stable but not asymptotically). The borderline τ² = 4Δ gives degenerate nodes or improper nodes depending on eigenvectors.

Non‑homogeneous systems are solved by variation of parameters: x_p(t) = Φ(t) ∫ Φ(s)-1 g(s) ds. For periodic coefficients, Floquet theory states that the fundamental matrix can be written Φ(t) = P(t) eBt with P(t) periodic and B constant; the eigenvalues of eBT (Floquet multipliers) determine stability. The method of undetermined coefficients applies when g(t) is a combination of exponentials, polynomials, sines, and cosines. Lyapunov’s direct method provides stability criteria without solving the system: if a positive definite function V(x) has negative definite derivative V̇(x) = ∇V·(Ax) along trajectories, the origin is asymptotically stable. These tools are essential for analysing mechanical vibrations, electrical circuits, and population models.

  • Confusing the conditions for Fubini’s theorem: the function must be integrable on the product domain, not merely continuous on each slice.
  • Forgetting the absolute value of the Jacobian when changing variables; the sign of the Jacobian only matters for oriented integrals.
  • Applying Green’s theorem to a region with holes without adding the inner boundary integrals with correct orientation.
  • Misidentifying the order of a pole when computing residues, especially when the denominator has repeated factors.
  • Assuming that a zero derivative of a complex function implies a local extremum; the maximum modulus principle forbids interior maxima for non‑constant analytic functions.
  • Using the matrix exponential eAt without verifying that A is constant; for time‑varying A(t) the solution is not simply e∫A.
  • Overlooking the requirement that the contour for the residue theorem must not pass through any singularity; indentations are needed for poles on the real axis.

These advanced topics complete the core toolkit for tackling the most demanding problems in the mathematics paper.

Partial Differential Equations: First Order and Classification

First‑order partial differential equations (PDEs) arise in geometric optics, fluid dynamics, and the method of characteristics. A general first‑order PDE in two independent variables x, y and unknown u(x, y) has the form F(x, y, u, p, q) = 0 where p = ∂u/∂x and q = ∂u/∂y. The Lagrange‑Charpit equations dx/F_p = dy/F_q = du/(pF_p + qF_q) = dp/(−F_x − pF_u) = dq/(−F_y − qF_u) give the characteristic curves along which the PDE reduces to an ODE system. For a linear first‑order PDE a(x, y) u_x + b(x, y) u_y = c(x, y) u + d(x, y), the characteristic ODEs are dx/dt = a, dy/dt = b, du/dt = c u + d. Solving these yields the general solution u = Φ(ξ) where ξ is a first integral of dx/a = dy/b. The Cauchy problem prescribes u on a non‑characteristic curve; if the curve is characteristic, either no solution or infinitely many solutions exist.

Classification of Second‑Order Linear PDEs

A second‑order linear PDE in two variables takes the form A u_xx + 2B u_xy + C u_yy + D u_x + E u_y + F u = G, where coefficients may depend on x, y. The discriminant Δ = B² − AC determines the type: Δ > 0 hyperbolic (wave‑like), Δ = 0 parabolic (diffusion‑like), Δ < 0 elliptic (equilibrium). This classification is invariant under non‑singular coordinate changes. Canonical forms simplify analysis: hyperbolic → u_ξη + … = 0, parabolic → u_ηη + … = 0, elliptic → u_ξξ + u_ηη + … = 0. For equations with variable coefficients, the type may change across the domain (e.g., Tricomi equation y u_xx + u_yy = 0 is elliptic for y > 0, hyperbolic for y < 0, parabolic on y = 0). Understanding the type guides the choice of boundary conditions: hyperbolic problems need Cauchy data on a non‑characteristic curve, parabolic problems need initial data and boundary conditions, elliptic problems need boundary values on a closed curve (Dirichlet, Neumann, or Robin).

Method of Characteristics for Quasi‑Linear Equations

For a quasi‑linear PDE a(x, y, u) u_x + b(x, y, u) u_y = c(x, y, u), the characteristic system is dx/ds = a, dy/ds = b, du/ds = c. The projection of characteristics onto the (x, y) plane are called base characteristics. If two base characteristics intersect, a shock forms and the classical solution breaks down; weak solutions and entropy conditions are then required. The method extends to n dimensions: dx_i/ds = a_i, du/ds = c. The initial manifold must be non‑characteristic, i.e., the vector (a1,…,a_n) not tangent to it. This geometric viewpoint underlies the construction of solutions for conservation laws and Hamilton‑Jacobi equations.

Second Order Linear PDEs: Wave, Heat, and Laplace Equations

The three canonical linear PDEs with constant coefficients model fundamental physical phenomena. The one‑dimensional wave equation ∂²u/∂t² = c² ∂²u/∂x² is hyperbolic. Its general solution is u(x, t) = f(x − ct) + g(x + ct) (d’Alembert’s formula). For the Cauchy problem on the whole line with initial displacement φ(x) and velocity ψ(x), the solution is u(x, t) = ½[φ(x+ct)+φ(x−ct)] + (1/(2c)) ∫x−ctx+ct ψ(s) ds. On a finite interval with fixed ends, separation of variables yields a Fourier sine series: u(x, t) = ∑n=1^∞ [A_n cos(nπct/L) + B_n sin(nπct/L)] sin(nπx/L). The coefficients are determined from the initial data using orthogonality.

Heat Equation

The heat equation ∂u/∂t = α ∂²u/∂x² is parabolic. It satisfies the maximum principle: the maximum of u on a closed space‑time cylinder occurs on the parabolic boundary (initial line or spatial boundaries). This implies uniqueness and continuous dependence on data. The fundamental solution (heat kernel) on ℝ is Φ(x, t) = (1/√(4παt)) exp(−x²/(4αt)) for t > 0. Convolution with initial data gives the solution on the whole line. On a bounded interval with Dirichlet conditions, separation of variables produces a Fourier sine series with exponential decay: u(x, t) = ∑n=1^∞ B_n exp(−α n²π²t/L²) sin(nπx/L). The series converges rapidly for t > 0, illustrating smoothing.

Laplace and Poisson Equations

Laplace’s equation ∇²u = 0 and Poisson’s equation ∇²u = f are elliptic. Solutions (harmonic functions) satisfy the mean‑value property: u(x₀) = (1/(2π)) ∫02π u(x₀ + r cosθ, y₀ + r sinθ) dθ for any circle inside the domain. The maximum principle states that a non‑constant harmonic function cannot attain an interior maximum or minimum. Green’s functions provide integral representations: u(x) = ∫_∂Ω u(y) ∂G/∂n_y dS_y − ∫_Ω G(x, y) f(y) dy for Poisson’s equation with Dirichlet data. In a disk, the Poisson kernel gives an explicit formula. Separation of variables in polar coordinates yields Fourier‑Bessel series for problems with circular symmetry. The method of images constructs Green’s functions for half‑planes, spheres, and other simple geometries.

Worked Example: Classification and Canonical Form
Classify the PDE u_xx − 4 u_xy + 4 u_yy = 0 and reduce it to canonical form.
Here A = 1, B = −2, C = 4. The discriminant Δ = B² − AC = 4 − 4 = 0, so the equation is parabolic everywhere. The characteristic equation is dy/dx = B/A = −2, giving ξ = y + 2x = constant. Choose ξ = y + 2x and any independent η, e.g., η = x. Compute derivatives: u_x = u_ξ ∂ξ/∂x + u_η ∂η/∂x = 2 u_ξ + u_η; u_y = u_ξ. Then u_xx = 4 u_ξξ + 4 u_ξη + u_ηη; u_xy = 2 u_ξξ + u_ξη; u_yy = u_ξξ. Substitute: (4 u_ξξ + 4 u_ξη + u_ηη) − 4(2 u_ξξ + u_ξη) + 4 u_ξξ = u_ηη = 0. The canonical form is u_ηη = 0, whose general solution is u = f(ξ) + η g(ξ) = f(y+2x) + x g(y+2x).
Exam‑Day Strategy for PDEs
Memorise the discriminant classification and the three canonical forms. For separation of variables, write the product ansatz, derive the ODEs, and state the eigenvalue problem. When asked for a solution on a bounded domain, always check boundary conditions to select the correct eigenfunctions (sine for Dirichlet, cosine for Neumann). For the heat equation, remember the smoothing property: the solution is C^∞ for t > 0 even if initial data is discontinuous. For the wave equation, note finite propagation speed: the domain of dependence is the interval [x−ct, x+ct]. In Laplace problems, use the maximum principle to argue uniqueness without solving. Practice converting a given PDE to canonical form by finding characteristics; this is a frequent short‑answer question.

Fourier Series and Fourier Transforms

Fourier series represent periodic functions as sums of sines and cosines (or complex exponentials). For a function f defined on [−π, π] and extended periodically, the Fourier coefficients are a₀ = (1/π) ∫−ππ f(x) dx, a_n = (1/π) ∫−ππ f(x) cos(nx) dx, b_n = (1/π) ∫−ππ f(x) sin(nx) dx. The series is f(x) ∼ a₀/2 + ∑n=1^∞ [a_n cos(nx) + b_n sin(nx)]. If f is piecewise C¹, the series converges to f(x) at points of continuity and to the average of left and right limits at jumps (Dirichlet’s theorem). Parseval’s identity states (1/π) ∫−ππ |f(x)|² dx = a₀²/2 + ∑n=1^∞ (a_n² + b_n²), linking energy in time and frequency domains.

Convergence and Gibbs Phenomenon

Uniform convergence holds if f is continuous and periodic with piecewise continuous derivative. At a jump discontinuity, the partial sums overshoot by approximately 9% of the jump height (Gibbs phenomenon); this does not vanish as N → ∞ but the overshoot region narrows. Fejér summation (Cesàro means) eliminates Gibbs phenomenon and converges uniformly for continuous periodic functions. The Fourier series of an even function contains only cosines; an odd function contains only sines. Half‑range expansions on [0, L] use sine or cosine series depending on boundary conditions.

Fourier Transform on ℝ

For non‑periodic functions, the Fourier transform pair is ̂f(ω) = ∫−∞∞ f(x) e−iωx dx, f(x) = (1/2π) ∫−∞∞ ̂f(ω) eiωx dω. The transform exists for absolutely integrable functions (L¹) and extends to L² via Plancherel’s theorem: ∫ |f|² = (1/2π) ∫ |̂f|². Key properties: linearity, shift (f(x−a) → e−iωa ̂f), scaling (f(ax) → (1/|a|) ̂f(ω/a)), derivative (f' → iω ̂f), convolution (f * g → ̂f ̂g). The transform of a Gaussian is a Gaussian: e−αx² → √(π/α) e−ω²/(4α). The Fourier transform converts constant‑coefficient PDEs into algebraic equations: the heat equation becomes ∂̂u/∂t = −α ω² ̂u, solved as ̂u(ω, t) = ̂u₀(ω) e−α ω² t, then inverted via convolution with the heat kernel.

Discrete Fourier Transform and FFT

For numerical work, the discrete Fourier transform (DFT) of a sequence x₀,…,xN−1 is X_k = ∑n=0N−1 x_n e−2πi kn/N. The fast Fourier transform (FFT) computes the DFT in O(N log N) operations. The DFT approximates the continuous Fourier transform when the signal is band‑limited and sampled above the Nyquist rate. Aliasing occurs if the sampling frequency is too low. Windowing (e.g., Hamming, Hanning) reduces spectral leakage when analysing finite records.

Numerical Analysis: Interpolation, Quadrature, and ODE Solvers

Numerical analysis provides algorithms for approximating solutions to mathematical problems that cannot be solved analytically. Polynomial interpolation constructs a polynomial p_n of degree ≤ n that matches given data points (x_i, y_i). The Lagrange form p_n(x) = ∑i=0^n y_i ℓ_i(x) with ℓ_i(x) = ∏j≠i (x−x_j)/(x_i−x_j) is explicit but costly to evaluate. Newton’s divided‑difference form p_n(x) = a₀ + a₁(x−x₀) + a₂(x−x₀)(x−x₁) + … allows incremental addition of nodes. The error for a function f ∈ Cn+1 is f(x) − p_n(x) = f(n+1)(ξ)/(n+1)! ∏i=0^n (x−x_i). Runge’s phenomenon shows that equidistant nodes can cause large oscillations near the interval ends for high‑degree interpolation; Chebyshev nodes x_i = cos((2i+1)π/(2n+2)) minimise the maximum of the nodal polynomial and are preferred.

Numerical Quadrature

Quadrature rules approximate ∫_a^b f(x) dx by a weighted sum ∑ w_i f(x_i). The Newton‑Cotes family uses equally spaced nodes: trapezoidal rule (degree 1), Simpson’s rule (degree 2), Simpson’s 3/8 rule, Boole’s rule. Composite versions divide [a, b] into subintervals and apply the simple rule on each. Error estimates: composite trapezoidal error = −(b−a)h²/12 f''(ξ), composite Simpson error = −(b−a)h⁴/180 f(4)(ξ). Gaussian quadrature chooses nodes and weights to integrate polynomials of degree up to 2n−1 exactly; Gauss‑Legendre quadrature on [−1, 1] is optimal for smooth functions. For singular integrals or infinite intervals, adaptive quadrature or transformation to a standard interval is used.

Numerical Solution of ODEs

Initial value problems y' = f(t, y), y(t₀) = y₀ are solved by step‑by‑step methods. Euler’s method yn+1 = y_n + h f(t_n, y_n) is first order. The classical Runge‑Kutta method of order 4 (RK4) uses four function evaluations per step and achieves local truncation error O(h⁵). Multistep methods (Adams‑Bashforth explicit, Adams‑Moulton implicit) reuse past values for efficiency. Stiff problems, where explicit methods require impractically small steps, are handled by implicit methods such as backward Euler or BDF (backward differentiation formulas). Stability regions are analysed via the test equation y' = λ y; a method is A‑stable if its stability region contains the left half‑plane. Convergence requires consistency (local truncation error → 0) and zero‑stability (bounded error propagation). For boundary value problems, shooting methods convert to IVPs, while finite‑difference discretisation yields a linear/nonlinear system.

Metric Spaces: Completeness, Compactness, and Connectedness

A metric space (X, d) is a set X with a distance function d: X × X → ℝ satisfying positivity, symmetry, and the triangle inequality. Convergence of sequences, Cauchy sequences, and continuity are defined analogously to ℝ. A metric space is complete if every Cauchy sequence converges in X. ℝⁿ with the Euclidean metric is complete; C[a, b] with the sup norm is complete (Banach space), but C[a, b] with the L¹ metric is not. The completion of a metric space is a complete space containing the original as a dense subset; it is unique up to isometry.

Compactness

A subset K of a metric space is compact if every open cover has a finite subcover. In metric spaces, compactness is equivalent to sequential compactness (every sequence has a convergent subsequence) and to being complete and totally bounded (for every ε > 0, K can be covered by finitely many ε‑balls). The Heine‑Borel theorem characterises compact subsets of ℝⁿ as those that are closed and bounded. Continuous images of compact sets are compact; a continuous function on a compact set attains its maximum and minimum (extreme value theorem) and is uniformly continuous. Arzelà‑Ascoli theorem gives a compactness criterion in C[a, b]: a set is relatively compact iff it is uniformly bounded and equicontinuous.

Connectedness

A metric space is connected if it cannot be written as the union of two disjoint non‑empty open sets. Equivalently, the only subsets that are both open and closed are ∅ and X. Path‑connectedness (any two points joined by a continuous path) implies connectedness, but the converse fails (e.g., the topologist’s sine curve). Connected components are maximal connected subsets; they are closed. In ℝ, connected sets are exactly intervals. The intermediate value theorem holds for continuous functions on connected domains. For product spaces, X × Y is connected iff both X and Y are connected.

Baire Category Theorem

In a complete metric space, the intersection of countably many dense open sets is dense. Equivalently, a non‑empty complete metric space is not a countable union of nowhere‑dense sets. This theorem underpins the uniform boundedness principle, open mapping theorem, and closed graph theorem in functional analysis. It also shows that ℝ is uncountable and that the set of points where a function is continuous is a G_δ set.

  • Confusing the type of a PDE when coefficients depend on the solution (non‑linear) – classification applies only to linear or quasi‑linear equations at a given state.
  • Forgetting that the Fourier series of a function with a jump converges to the midpoint at the discontinuity, not to the function value.
  • Applying the trapezoidal rule to a periodic integrand over a full period and expecting the error formula with endpoint derivatives; the composite trapezoidal rule is actually spectrally accurate for smooth periodic functions.
  • Using an explicit Runge‑Kutta method on a stiff ODE without checking the stability region, leading to step‑size restrictions that defeat the purpose of higher order.
  • Assuming that a bounded sequence in a metric space has a convergent subsequence; this holds only in compact (or sequentially compact) spaces.
  • Misapplying the Heine‑Borel theorem in infinite‑dimensional normed spaces where closed and bounded does not imply compactness.
  • Neglecting the requirement that the initial curve for a first‑order PDE be non‑characteristic; otherwise the Cauchy problem is ill‑posed.

These advanced topics complete the core toolkit for tackling the most demanding problems in the mathematics paper.

Practice Questions

← Mathematics Optional: Syllabus, Sources & Strategy Mathematics Paper II: Core Topics →