Module 25 · UPSC Optional: Mathematics

Mathematics Paper II: Core Topics

Updated 17 Nov 2026 · 3,519 words
Updated 21 Nov 2026
Mathematics Paper II: core syllabus areas, key concepts, standard references and answer-writing practice.
Differential Equations · Statistics · Numerical Analysis · Linear Programming
🎯 Take Mock Test

Learning Objectives

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

Mathematics Paper II — Algebra, Complex Analysis, ODE & PDE, Mechanics

Algebra

  • Groups: Groups (definition, examples), subgroups, Lagrange's theorem, cyclic groups, permutation groups (S_n), homomorphisms & isomorphisms, normal subgroups, quotient groups, first/second/third isomorphism theorems, automorphisms, inner automorphisms, Cayley's theorem, Cayley table, center of group, center of GL(n,R), center of SL(n,R).
  • Rings: Rings (definition, examples), ideals, quotient rings, prime ideals, maximal ideals, Euclidean domains (ED), principal ideal domains (PID), unique factorization domains (UFD), Euclidean algorithm, polynomial rings over fields, irreducibility criteria (Eisenstein's criterion, Gauss's lemma).
  • Fields: Fields (definition, examples), field extensions (algebraic & transcendental), algebraic extensions, finite fields (Galois fields GF(p^n)), splitting fields, normal extensions, separable extensions, Galois theory (Galois correspondence, fundamental theorem of Galois theory), solvability by radicals, Galois groups of polynomials (x^n - 1, x^p - p, etc.).
  • Matrices: Matrices (definition, operations), rank, inverse (if exists), elementary operations, row echelon form, canonical forms (Smith normal form, Jordan canonical form), diagonalization, eigenvalues & eigenvectors, characteristic polynomial, minimal polynomial, characteristic polynomial, trace, determinant, properties of determinants, Cayley-Hamilton theorem, bilinear forms, quadratic forms, orthogonal diagonalization, spectral theorem for symmetric matrices.

Complex Analysis

  • Complex Numbers: Complex numbers (z = x + iy), complex plane, modulus & argument, Euler's formula (e^(iθ) = cos θ + i sin θ), De Moivre's theorem, roots of unity, complex functions, limits, continuity, differentiability.
  • Complex Functions: Complex functions (f(z) = u(x,y) + i v(x,y)), Cauchy-Riemann equations (∂u/∂x = ∂v/∂y, ∂u/∂y = -∂v/∂x), analyticity (holomorphic/analytic functions), harmonic functions (∇²u = 0, ∇²v = 0), conformal mappings (conformal maps preserve angles locally).
  • Contour Integration: Contour integration (line integrals in complex plane), Cauchy's theorem (∮_C f(z) dz = 0 for analytic f in simply connected domain), Cauchy's integral formula (f(a) = (1/2πi) ∮_C f(z)/(z-a) dz), Cauchy's integral formula for derivatives (f^(n)(a) = n!/(2πi) ∮_C f(z)/(z-a)^(n+1) dz), Cauchy's estimate, Liouville's theorem (bounded entire function is constant), maximum modulus principle, minimum modulus principle.
  • Series & Residues: Taylor series (Taylor's theorem for complex functions), Laurent series (Laurent series expansion around isolated singularity), residue theorem (∮_C f(z) dz = 2πi Σ Res(f, a_k)), residue calculation (residue at simple pole: Res(f, a) = lim_{z→a} (z-a)f(z); residue at pole of order m: Res(f, a) = (1/(m-1)!) lim_{z→a} d^(m-1)/dz^(m-1)[(z-a)^m f(z)]), evaluation of real integrals using residue theorem (trigonometric integrals, rational function integrals, Fourier transforms).
  • Singularities: Singularities (isolated singularities: removable singularity, pole, essential singularity; non-isolated singularities: branch points, natural boundaries), Casorati-Weierstrass theorem (image of punctured neighborhood of essential singularity is dense in C), Picard's great theorem, Riemann sphere (extended complex plane), stereographic projection.
  • Conformal Mapping: Conformal mappings (angle-preserving maps), examples (z^n, e^z, log z, sin z, cos z, tan z, Möbius transformations (z → (az+b)/(cz+d))), mapping regions (upper half-plane to unit disk via Möbius transformation), applications to boundary value problems (Laplace's equation in simply connected domains).

Ordinary Differential Equations

  • First Order ODE: First order ODE (dy/dx = f(x,y)), separable equations, homogeneous equations, exact equations (exactness condition: ∂M/∂y = ∂N/∂x), integrating factors, linear first-order ODE (y' + P(x)y = Q(x), integrating factor μ(x) = e^(∫P(x)dx)), Bernoulli equation (y' + P(x)y = Q(x)y^n), Riccati equation, Clairaut's equation, Lagrange's equation.
  • Higher Order Linear ODE: Higher order linear ODE (homogeneous & non-homogeneous), linear independence (Wronskian), fundamental set of solutions, reduction of order, variation of parameters, method of undetermined coefficients, Cauchy-Euler equation (x^n y^(n) + ... + a_1 x y' + a_0 y = f(x)), exact equations, integrating factors.
  • Systems of Linear ODE: Systems of linear ODE (X' = A(t)X + f(t)), fundamental matrix, Wronskian, Abel's formula, eigenvalue method for constant coefficient systems, complex eigenvalues, repeated eigenvalues, defective matrices.
  • Power Series Solutions: Power series solutions (ordinary points, regular singular points), Frobenius method (indicial equation, recurrence relations), special functions (Bessel functions J_v(x), Y_v(x); Legendre polynomials P_n(x); Hermite polynomials H_n(x); Laguerre polynomials L_n(x); Chebyshev polynomials T_n(x)), properties of special functions (orthogonality, recurrence relations, generating functions).
  • Sturm-Liouville Theory: Sturm-Liouville problem ((p(x)y')' + (q(x) + λw(x))y = 0), eigenvalues & eigenfunctions, orthogonality of eigenfunctions, expansion in eigenfunctions (Fourier-Bessel series, Fourier-Legendre series), Rayleigh quotient, Rayleigh-Ritz method.

Partial Differential Equations

  • First Order PDE: First order PDE (a(x,y,u) u_x + b(x,y,u) u_y = c(x,y,u)), Charpit's method, Charpit's equations, complete integral, singular integral, general integral, Clairaut's equation, singular solution vs general solution, Envelope method.
  • Second Order Linear PDE: Second order linear PDE (classification: elliptic, parabolic, hyperbolic based on discriminant B² - 4AC), canonical forms, wave equation (u_tt = c² ∇²u), heat equation (u_t = k ∇²u), Laplace equation (∇²u = 0), Poisson equation (∇²u = f(x,y,z)), separation of variables.
  • Wave Equation: Wave equation (∂²u/∂t² = c² ∇²u), D'Alembert's solution (u(x,t) = (f(x+ct) + f(x-ct))/2 + (1/(2c)) ∫_{x-ct}^{x+ct} g(s) ds for 1D wave equation), d'Alembert's solution for semi-infinite string, reflections, energy conservation.
  • Heat Equation: Heat equation (∂u/∂t = k ∇²u), separation of variables, Fourier series solutions (Fourier sine series, Fourier cosine series), steady-state solutions, maximum principle, uniqueness, Duhamel's principle (for non-homogeneous heat equation), Green's function for heat equation.
  • Laplace Equation: Laplace equation (∇²u = 0), separation of variables, Fourier series solutions, Dirichlet problem (Dirichlet problem for disk: Poisson integral formula), Neumann problem (Neumann boundary conditions), maximum principle (strong maximum principle, weak maximum principle), mean value property, Harnack's inequality, Poisson integral formula for disk.
  • Conformal Mapping: Conformal mappings (angle-preserving maps), examples (z^n, e^z, log z, sin z, cos z, tan z, Möbius transformations (z → (az+b)/(cz+d))), mapping regions (upper half-plane to unit disk via Möbius transformation w = (z - z_0)/(z - conj(z_0))), applications to boundary value problems (Laplace's equation in simply connected domains via conformal mapping).
Technique: Cauchy-Riemann equations
Question: Show that f(z) = z² is analytic everywhere.
Answer: f(z) = z² = (x+iy)² = x² - y² + 2ixy. So u(x,y) = x² - y², v(x,y) = 2xy. Compute partial derivatives: u_x = 2x, u_y = -2y, v_x = 2y, v_y = 2x. Check Cauchy-Riemann: u_x = v_y (2x = 2x) ✓, u_y = -v_x (-2y = -2y) ✓. Since u_x, u_y, v_x, v_y are continuous and C-R equations hold everywhere, f(z) = z² is analytic everywhere (entire function).

Memory Tricks

  • Cauchy-Riemann equations: u_x = v_y, u_y = -v_x (u_x = v_y, u_y = -v_x — "x with x, y with -y").
  • Classification of PDEs: Elliptic (B² - 4AC < 0, e.g., Laplace), Parabolic (B² - 4AC = 0, e.g., heat equation), Hyperbolic (B² - 4AC > 0, e.g., wave equation).
  • Series convergence: p-series (∑1/n^p converges iff p > 1), geometric series (∑ar^n converges if |r| < 1), ratio test, root test, comparison test, integral test.
  • Laplace transforms: L{f'(t)} = sF(s) - f(0), L{f''(t)} = s²F(s) - sf(0) - f'(0).
Mains strategy
Math answers: derive formulas step by step, show all assumptions, cite theorems by name (Cauchy-Schwarz, Bolzano-Weierstrass, Heine-Borel, Arzela-Ascoli, Picard's existence & uniqueness theorem, Green's theorem, Stokes' theorem, Gauss divergence theorem, fundamental theorem of calculus). Data: India's R&D spending (~0.7% of GDP), number of researchers per million, math education institutions (IITs, IISERs, IISERs, CMI, IMSc, ISI, TIFR, HRI), number of PhDs in mathematics per year.

Statistics: The Central Limit Theorem and Its Implications for Sampling

In the context of UPSC Mathematics Paper II, Statistics is not merely about calculating means or variances but understanding the theoretical underpinnings that justify inference from samples to populations. The most critical theoretical pillar here is the Central Limit Theorem (CLT). While often stated simply as "the sum of independent random variables tends toward a normal distribution," a rigorous examination requires distinguishing between the convergence in distribution and the rate of convergence.

The CLT states that if X1, X2, ..., X_n are independent and identically distributed (i.i.d.) random variables with finite mean μ and finite variance σ2, then the standardized sum converges in distribution to a standard normal variable as n → ∞: Z_n = (Σi=1^n X_i - nμ)/(σ√(n)) xrightarrow{d} N(0, 1) This theorem is the justification for using the Normal distribution to approximate sampling distributions of means, even when the underlying population distribution is highly non-normal (e.g., uniform, exponential, or skewed). However, a sophisticated answer must acknowledge the conditions: independence and identical distribution are strict requirements. In real-world survey data (often relevant for GS papers on social surveys), these assumptions may be violated due to clustering or stratification, necessitating corrections like Finite Population Correction factors.

A crucial distinction in advanced statistical theory is between Parameter Estimation and Hypothesis Testing. Estimation involves finding a statistic (like the sample mean X) that best approximates an unknown parameter (μ). We evaluate estimators based on properties like Unbiasedness (E[θ] = θ), Consistency (convergence to true value as n → ∞), and Efficiency (minimum variance among unbiased estimators). The Cramér-Rao Lower Bound provides the theoretical floor for the variance of any unbiased estimator. If an estimator achieves this bound, it is called Minimum Variance Unbiased Estimator (MVUE).

Hypothesis testing, conversely, is a decision-making framework based on probability. The Null Hypothesis (H0) assumes no effect or difference, while the Alternative Hypothesis (H1) posits an effect. The test statistic is calculated, and we compare it against a critical region determined by the significance level α (commonly 0.05). A Type I error occurs when we reject H0 when it is true (false positive), while a Type II error occurs when we fail to reject H0 when it is false (false negative). The power of a test (1 - β) measures the probability of correctly rejecting a false null hypothesis. Understanding the trade-off between α and β is vital; reducing one usually increases the other unless sample size is increased.

The concept of Confidence Intervals complements hypothesis testing. A 95% confidence interval for the mean μ is given by X ± Zα/2 (σ/√(n)) (for known σ) or using the t-distribution when σ is unknown and estimated by S. The use of the t-distribution arises because, for small samples from a normal population, the quantity (X-μ)/(S/√(n)) follows Student's t distribution with n-1 degrees of freedom. This was derived by Gosset (writing as "Student") and is essential when dealing with limited data in field studies.

Examiner's trap
A common error is assuming the Central Limit Theorem applies to dependent variables or time-series data without checking for stationarity. Another frequent mistake is confusing the Z-test (large samples, known σ) with the t-test (small samples, unknown σ). Students often write the formula for confidence interval using Z even when sample size n < 30 and population variance is unknown.
Exam-day strategy
When asked to derive a distribution or test statistic, always state the assumptions (independence, normality) explicitly at the start. In answer writing, define all symbols used in the formula before substituting values. If a question asks about "robustness," mention that t-tests are robust to slight deviations from normality, unlike variance tests which are sensitive to outliers.

Numerical Analysis: Error Propagation and Stability in Iterative Methods

Numerical Analysis in the UPSC syllabus focuses on the practical implementation of mathematical algorithms where exact solutions are either impossible or computationally prohibitive. The core challenge is managing Error Analysis. Errors arise from two main sources: Rounding Errors (due to finite precision of computer memory) and Truncation Errors (due to approximating infinite processes, like Taylor series expansions, with finite terms).

The concept of Conditioning is paramount. A problem is well-conditioned if small changes in input data lead to small changes in the output. Conversely, an ill-conditioned problem amplifies errors. The condition number kappa(A) of a matrix A quantifies this sensitivity: kappa(A) = \|A\| \|A-1\|. If kappa(A) is large, the system is sensitive to perturbations. In solving linear systems Ax=b, Gaussian Elimination can lead to numerical instability if the pivot element is very small. Pivoting strategies (Partial or Complete) are employed to swap rows to ensure the largest possible pivot is used, thereby minimizing error propagation.

Iterative methods, such as Jacobi and Gauss-Seidel for solving large systems of linear equations, rely on convergence criteria. The spectral radius of the iteration matrix T determines convergence: if ρ(T) < 1, the method converges. However, theoretical convergence does not guarantee numerical stability. For example, in the Newton-Raphson method for finding roots of f(x)=0, the update rule is xn+1 = x_n - f(x_n)/f'(x_n).

Linear Programming: Duality Theory and Sensitivity Analysis

Every linear programming problem (the primal) has an associated dual. If the primal is a maximisation with constraints Ax ≤ b, x ≥ 0, the dual is a minimisation with constraints Aᵀy ≥ c, y ≥ 0. The weak duality theorem states that for any feasible primal x and dual y, the primal objective value is ≤ the dual objective value. Strong duality holds when both have optimal solutions; then the optimal values are equal. Complementary slackness gives necessary and sufficient optimality conditions: for each primal constraint either the slack is zero or the corresponding dual variable is zero, and vice‑versa.

Economic interpretation: dual variables (shadow prices) measure the marginal increase in the optimal objective per unit increase in bᵢ. Sensitivity analysis examines how the optimal basis changes when b, c, or A are perturbed. If the current basis stays feasible and optimal, the new optimal value updates via the dual solution without re‑solving. The allowable range for each bᵢ is found from the simplex tableau using the minimum ratio of basic variable values to pivot column coefficients. For changes in cⱼ, reduced costs show whether a non‑basic variable becomes attractive. Adding a new constraint either leaves the current solution feasible (if satisfied) or requires the dual simplex method to restore feasibility. The dual simplex works on a tableau that is dual feasible (reduced costs optimal) but primal infeasible (some basic variables negative), iterating until primal feasibility returns.

Key computational steps: (1) Write the dual explicitly. (2) Solve either primal or dual with the simplex method. (3) Read shadow prices from the optimal dual variables. (4) Perform parametric analysis on b and c using the final tableau. (5) If a new constraint cuts off the current optimum, apply the dual simplex to find the new optimum efficiently.

Statistical Inference: Maximum Likelihood Estimation and Asymptotic Efficiency

Given a random sample X₁,…,Xₙ from a density f(x;θ), the likelihood is L(θ) = ∏ f(Xᵢ;θ). The MLE θ̂ maximises L(θ) or the log‑likelihood ℓ(θ) = Σ log f(Xᵢ;θ). Under regularity conditions (identifiability, differentiability, support independent of θ), the MLE has three large‑sample properties: consistency (θ̂ → θ in probability), asymptotic normality (√n(θ̂−θ) → N(0, I(θ)⁻¹) where I(θ) is Fisher information), and asymptotic efficiency (it attains the Cramér‑Rao lower bound). Fisher information for one observation is I(θ) = E[(∂/∂θ log f(X;θ))²] = −E[∂²/∂θ² log f(X;θ)]; for n observations it is nI(θ). The invariance principle: if θ̂ is the MLE of θ, then g(θ̂) is the MLE of g(θ) for any measurable g.

Examples: For Xᵢ ∼ N(μ,σ²) with both unknown, MLEs are μ̂ = X̄ and σ̂² = (1/n)Σ(Xᵢ−X̄)² (divisor n, not n−1). For Xᵢ ∼ Exp(λ), λ̂ = 1/X̄. For binomial(n,p), p̂ = X/n. Asymptotic variances are the inverse Fisher information: Var(μ̂) ≈ σ²/n, Var(λ̂) ≈ λ²/n, Var(p̂) ≈ p(1−p)/n. The likelihood ratio, Wald, and score (Lagrange multiplier) tests are three asymptotically equivalent methods for hypotheses on θ; all rely on the quadratic approximation of the log‑likelihood near the MLE.

Worked Example: MLE for Exponential Distribution
Let X₁,…,Xₙ be i.i.d. Exp(λ) with density f(x;λ)=λe−λx, x>0. Derive the MLE of λ, its asymptotic distribution, and a 95% confidence interval for λ based on n=25, ΣXᵢ=12.5.
Log‑likelihood: ℓ(λ)=n log λ − λ ΣXᵢ. Derivative: ∂ℓ/∂λ = n/λ − ΣXᵢ. Set to zero ⇒ λ̂ = n/ΣXᵢ = 25/12.5 = 2. Fisher information per observation: I(λ)=1/λ²; sample information: n/λ². Asymptotic variance of λ̂ is λ²/n. Estimated SE = λ̂/√n = 2/5 = 0.4. Approximate 95% CI: λ̂ ± 1.96 × 0.4 = 2 ± 0.784 ⇒ (1.216, 2.784). Exact interval uses 2λ ΣXᵢ ∼ χ²2n.
Exam‑Day Strategy for MLE Questions
Write the likelihood, take logs, differentiate, and solve. State regularity conditions (support not depending on θ, differentiability). For asymptotic results, quote Fisher information and asymptotic normality. If a confidence interval is requested, give both the asymptotic normal approximation and, when feasible, the exact pivot (e.g., chi‑square for exponential, t for normal mean). Show numerical substitution clearly; marks are awarded for correct formula and arithmetic.

Numerical Solution of Ordinary Differential Equations: Runge–Kutta Methods and Stability Regions

Runge–Kutta (RK) methods are one‑step schemes for y' = f(t,y). An s‑stage explicit RK method computes slopes kᵢ = f(tₙ + cᵢh, yₙ + h Σ aᵢⱼ kⱼ) and updates yₙ₊₁ = yₙ + h Σ bᵢ kᵢ. Coefficients are chosen to achieve order p. The classic RK4 uses four stages with c = (0, ½, ½, 1), a₂₁=½, a₃₂=½, a₄₃=1, and b = (1/6, 1/3, 1/3, 1/6). Its local truncation error is O(h⁵), global error O(h⁴). Embedded RK pairs (e.g., Fehlberg 4(5), Dormand–Prince 5(4)) yield two estimates of different orders from the same stages, enabling adaptive step‑size control: if the error estimate exceeds tolerance, the step is rejected and h reduced; otherwise h may increase.

Stability analysis uses the test equation y' = λy (Re(λ) < 0). The numerical solution satisfies yₙ₊₁ = R(z) yₙ with z = λh; R(z) is the stability function (polynomial for explicit RK, rational for implicit). The region of absolute stability is {z ∈ ℂ : |R(z)| ≤ 1}. For explicit RK4 this region is bounded; the method is not A‑stable. A‑stability requires the region to contain the entire left half‑plane; only implicit methods (e.g., Gauss–Legendre RK of order 2s) achieve this. L‑stability further requires R(z) → 0 as z → −∞, strongly damping stiff components. For stiff problems, implicit RK or BDF are preferred despite higher per‑step cost because they allow much larger steps.

Practical implementation: (1) Choose an embedded pair (e.g., Dormand–Prince 5(4)). (2) Set initial h, tolerances atol, rtol. (3) Compute two solutions, estimate error e = |y⁽⁵⁾ − y⁽⁴⁾|. (4) If e ≤ atol + rtol·|y|, accept step and update h ← h × min(5, max(0.2, 0.9 (tol/e)¹ᐟ⁵)). (5) Else reject step and reduce h. (6) Monitor stiffness: if h becomes extremely small, switch to an implicit solver.

Finite Difference Methods for Parabolic and Hyperbolic Partial Differential Equations

For the heat equation uₜ = α uₓₓ on 0<x<L, t>0, a uniform grid gives the explicit FTCS scheme: uᵢⁿ⁺¹ = uᵢⁿ + r (uᵢ₊₁ⁿ − 2uᵢⁿ + uᵢ₋₁ⁿ) with r = αΔt/Δx². Von Neumann analysis substitutes uᵢⁿ = ξⁿ eik iΔx, yielding amplification factor ξ = 1 − 4r sin²(kΔx/2). Stability requires |ξ| ≤ 1 for all k, which holds iff r ≤ ½ (the CFL condition for parabolic equations). The implicit Crank–Nicolson scheme averages the spatial derivative at levels n and n+1: uᵢⁿ⁺¹ − uᵢⁿ = (r/2)[(uᵢ₊₁ⁿ⁺¹ − 2uᵢⁿ⁺¹ + uᵢ₋₁ⁿ⁺¹) + (uᵢ₊₁ⁿ − 2uᵢⁿ + uᵢ₋₁ⁿ)]. It is unconditionally stable (|ξ| ≤ 1 for all r) and second‑order accurate in Δt and Δx; the tridiagonal system is solved by the Thomas algorithm.

For the wave equation uₜₜ = c² uₓₓ, the leapfrog scheme uses central differences: (uᵢⁿ⁺¹ − 2uᵢⁿ + uᵢⁿ⁻¹)/Δt² = c² (uᵢ₊₁ⁿ − 2uᵢⁿ + uᵢ₋₁ⁿ)/Δx². The CFL condition is cΔt/Δx ≤ 1. The Lax–Wendroff method, derived from a Taylor expansion replacing time derivatives by space derivatives via the PDE, gives a second‑order explicit scheme with the same CFL limit. For first‑order hyperbolic systems uₜ + A uₓ = 0, upwind schemes use the sign of eigenvalues of A to choose the spatial difference direction, ensuring stability under max|λ| Δt/Δx ≤ 1. The Lax equivalence theorem states that for a consistent linear finite‑difference scheme, stability is equivalent to convergence; thus verifying consistency (local truncation error → 0) and stability (von Neumann or energy methods) guarantees convergence.

Conformal Mapping and Its Applications in Complex Analysis

A conformal map is an analytic function with non‑zero derivative that preserves angles locally. The Riemann mapping theorem guarantees that any simply connected proper subdomain of ℂ can be mapped conformally onto the unit disc. Möbius transformations f(z) = (az+b)/(cz+d) (ad−bc ≠ 0) map circles and lines to circles and lines and form the automorphism group of the Riemann sphere. The Schwarz–Christoffel formula gives an explicit conformal map from the upper half‑plane to a polygon: f(z) = A ∫ ∏ (ζ−zₖ)αₖ/π−1 dζ + B, where αₖ are interior angles; this integral is evaluated numerically in practice.

Applications: In two‑dimensional potential flow, the complex potential Φ(z) = φ + iψ is analytic; φ is the velocity potential, ψ the stream function. Conformal maps transform complicated boundaries (e.g., airfoils) to simple ones (circle, line) where the flow is known (uniform flow, source, vortex). The Joukowski transform z = ζ + 1/ζ maps a circle to an airfoil‑like shape. In electrostatics, the complex potential solves Laplace's equation with Dirichlet or Neumann conditions; mapping the domain to a strip or disc reduces the problem to a Fourier series. Harmonic conjugates: if u is harmonic, there exists a harmonic conjugate v such that f = u+iv is analytic; v is determined up to a constant by the Cauchy–Riemann equations. The Dirichlet problem on a disc is solved by the Poisson integral formula, derivable via conformal mapping from the half‑plane.

  • Confusing primal and dual forms when the primal is a minimisation with ≥ constraints.
  • Using divisor n−1 instead of n for the MLE of σ² in a normal sample, then applying the wrong asymptotic variance.
  • Applying explicit Runge–Kutta methods to stiff problems without checking the stability region, causing severe step‑size restrictions.
  • Using FTCS for the heat equation with r > ½ and claiming stability because the solution looks smooth initially.
  • Assuming every analytic function is conformal everywhere, ignoring points where f'(z)=0.
  • Neglecting the CFL condition for hyperbolic schemes, producing non‑physical oscillations.
  • Mastering these concepts and avoiding these traps will strengthen your answer writing.

Practice Questions

← Mathematics Paper I: Core Topics ← Course Home