Chapter 1: Foundations — Logic, Sets, Relations and Proof
University-level mathematical foundations for rigorous reasoning: formal statements, quantifiers, set operations, relations, functions and proof methods, with original worked problems.
UNDERGRADUATE UNIVERSITY MATHEMATICS — SELECT A CHAPTER
The undergraduate pathway now includes Chapters 1–22. Select a chapter to open its university-level notes and worked assessment.
Chapter 1 — Foundations: Logic, Sets, Relations and Proof
Learning outcomes
Formalise mathematical statements; distinguish implication from equivalence; manipulate predicates and quantifiers; prove set identities; classify relations; construct equivalence classes and quotient sets; analyse injective, surjective and bijective functions; and write direct, contrapositive, contradiction and induction proofs.
Study method: Reproduce definitions and proofs without looking. Attempt each problem before opening its individual worked solution. Numerical examples alone do not establish a universal claim.
1.1 Mathematical statements and propositional logic
A proposition is a declarative statement with exactly one truth value, true or false. A sentence with an unassigned variable, such as \(x^2=4\), is an open sentence rather than a proposition until a value or quantifier is supplied.
The implication \(P\Rightarrow Q\) is false only when \(P\) is true and \(Q\) is false. Its converse is \(Q\Rightarrow P\); its inverse is \(\neg P\Rightarrow\neg Q\); its contrapositive is \(\neg Q\Rightarrow\neg P\). An implication is equivalent to its contrapositive, but not generally to its converse.
For example, “if an integer \(n\) is divisible by 6, then it is divisible by 3” is true. Its converse is false, since 3 is divisible by 3 but not by 6.
1.2 Predicates, domains and quantifiers
A predicate \(P(x)\) has a truth value depending on \(x\); its domain must be stated. The universal quantifier \(\forall\) means “for every”; the existential quantifier \(\exists\) means “there exists”.
Quantifier order matters. Over \(\mathbb R\), \(\forall x\,\exists y\,(y>x)\) is true, but \(\exists y\,\forall x\,(y>x)\) is false. In the first statement, \(y\) may depend on \(x\); in the second, one fixed real number must exceed every real number.
To prove a universal statement, take an arbitrary element of the domain.
To disprove a universal statement, provide one counterexample.
To prove an existential statement, exhibit and verify a witness.
To disprove an existential statement, show every candidate fails.
1.3 Sets, operations and algebra of sets
A set is a well-defined collection of distinct objects. \(x\in A\) denotes membership; \(A\subseteq B\) means every element of \(A\) belongs to \(B\). The empty set is \(\varnothing\). The power set \(\mathcal P(A)\) consists of all subsets of \(A\); if \(|A|=n\), then \(|\mathcal P(A)|=2^n\).
Union: \(A\cup B=\{x:x\in A\text{ or }x\in B\}\). Intersection: \(A\cap B=\{x:x\in A\text{ and }x\in B\}\). Difference: \(A\setminus B=\{x\in A:x\notin B\}\). For universal set \(U\), \(A^c=U\setminus A\).
Distributivity and De Morgan: \[A\cap(B\cup C)=(A\cap B)\cup(A\cap C),\qquad(A\cup B)^c=A^c\cap B^c.\]
Proof by element method: To prove \(X=Y\), establish both inclusions. If \(x\in A\cap(B\cup C)\), then \(x\in A\) and either \(x\in B\) or \(x\in C\); hence \(x\in(A\cap B)\cup(A\cap C)\). Conversely, membership in the right-hand side implies \(x\in A\) and \(x\in B\cup C\), giving the left-hand side. Thus equality holds. For finite sets, \(|A\cup B|=|A|+|B|-|A\cap B|\).
1.4 Cartesian products, relations and equivalence classes
The Cartesian product \(A\times B=\{(a,b):a\in A,b\in B\}\). A binary relation from \(A\) to \(B\) is a subset of \(A\times B\).
Properties of a relation \(R\) on \(A\): reflexive if \(aRa\) for every \(a\); symmetric if \(aRb\Rightarrow bRa\); antisymmetric if \(aRb\) and \(bRa\Rightarrow a=b\); transitive if \(aRb\) and \(bRc\Rightarrow aRc\).
An equivalence relation is reflexive, symmetric and transitive. Its equivalence class is \([a]=\{x\in A:xRa\}\). Equivalence classes partition \(A\): each element lies in one class, and two classes are equal or disjoint. A partial order is reflexive, antisymmetric and transitive.
Example: On \(\mathbb Z\), define \(a\sim b\) iff \(5\mid(a-b)\). This is an equivalence relation with five classes, represented by \([0],[1],[2],[3],[4]\) modulo 5.
1.5 Functions, images and inverse mappings
A function \(f:A\to B\) assigns each element of \(A\) exactly one element of \(B\). The domain is \(A\), codomain \(B\), and image \(f(A)=\{f(a):a\in A\}\subseteq B\).
Injective: \(f(a_1)=f(a_2)\Rightarrow a_1=a_2\). Surjective: for every \(b\in B\), some \(a\in A\) satisfies \(f(a)=b\). Bijective: both injective and surjective.
A function has an inverse \(f^{-1}:B\to A\) exactly when it is bijective. Composition is \((g\circ f)(x)=g(f(x))\), when the domains and codomains match; composition is not generally commutative.
1.6 Proof methods and mathematical induction
Direct proof
Assume the hypotheses and derive the conclusion from definitions and established results. If \(a=2r\) and \(b=2s\), then \(a+b=2(r+s)\), so the sum of two even integers is even.
Contrapositive and contradiction
To prove \(P\Rightarrow Q\), one may prove the equivalent \(\neg Q\Rightarrow\neg P\). For contradiction, assume the desired conclusion is false and derive an impossibility. For example, assuming \(\sqrt2=p/q\) in lowest terms gives \(p^2=2q^2\); hence \(p\) is even, and substitution forces \(q\) even, a contradiction.
Mathematical induction
For \(P(n)\) on integers \(n\ge n_0\), prove (i) \(P(n_0)\), and (ii) for arbitrary \(k\ge n_0\), \(P(k)\Rightarrow P(k+1)\). Then \(P(n)\) holds for all \(n\ge n_0\).
Example: \(1+\cdots+n=n(n+1)/2\). The base case \(n=1\) holds. Assuming the identity for \(k\), add \(k+1\): \(k(k+1)/2+(k+1)=(k+1)(k+2)/2\), establishing the next case. Strong induction assumes all cases from \(n_0\) through \(k\) to prove \(k+1\); it is useful when a step depends on multiple earlier cases.
Chapter 1: University problem set
These original exercises assess proof construction, logical precision and mathematical reasoning. Attempt each question before opening its worked solution.
Question 1 — Quantifier negation and logical equivalence
Let \(x,y\in\mathbb R\). (a) Negate \(\forall x\in\mathbb R\,\exists y\in\mathbb R\;(y^2=x^2+1)\). (b) Determine whether the original is true. (c) Explain why reversing the quantifiers changes the assertion.
(b) It is true: for arbitrary real \(x\), choose \(y=\sqrt{x^2+1}\). Since \(x^2+1\ge1\), this is real and satisfies the equation.
(c) The reversed assertion requires one fixed \(y\) to work for every \(x\). Taking \(x=0\) forces \(y^2=1\), while \(x=1\) forces \(y^2=2\); impossible.
Question 2 — Set identity by double inclusion
For arbitrary \(A,B,C\subseteq U\), prove \(A\setminus(B\cap C)=(A\setminus B)\cup(A\setminus C)\). Establish both inclusions and identify the logical law used.
8 marksShow worked answer
If \(x\in A\setminus(B\cap C)\), then \(x\in A\) and \(x\notin B\cap C\). De Morgan gives \(x\notin B\) or \(x\notin C\), so \(x\in(A\setminus B)\cup(A\setminus C)\).
Conversely, if \(x\) belongs to the right-hand side, then \(x\in A\) and at least one of \(x\notin B\), \(x\notin C\) holds. Thus \(x\notin B\cap C\), proving membership in the left-hand side. Both inclusions give equality.
Question 3 — Cardinality and power sets
Let \(A\) be finite with \(|A|=n\ge1\), and fix \(a\in A\). Prove that exactly \(2^{n-1}\) subsets of \(A\) contain \(a\). Deduce the number that do not contain \(a\).
7 marksShow worked answer
Every subset containing \(a\) has a unique representation \(\{a\}\cup S\), where \(S\subseteq A\setminus\{a\}\). This is a bijection with \(\mathcal P(A\setminus\{a\})\), which has \(2^{n-1}\) members. Subsets not containing \(a\) are exactly the subsets of \(A\setminus\{a\}\), so there are also \(2^{n-1}\). These two families partition \(\mathcal P(A)\).
Question 4 — Equivalence relation and quotient set
On \(\mathbb Z\), define \(a\sim b\) iff \(a-b\in6\mathbb Z\). Prove that this is an equivalence relation, list its distinct classes and show \([14]=[2]\).
10 marksShow worked answer
Reflexive: \(a-a=0=6(0)\). Symmetric: \(a-b=6k\) implies \(b-a=6(-k)\). Transitive: \(a-b=6k\) and \(b-c=6m\) imply \(a-c=6(k+m)\). Thus all three properties hold.
The six classes are \([0],[1],[2],[3],[4],[5]\), with \([r]=\{6k+r:k\in\mathbb Z\}\). Since \(14-2=12=6(2)\), \(14\sim2\), hence \([14]=[2]\).
Question 5 — Partial order on a power set
Let \(S=\{1,2,3\}\) and order \(\mathcal P(S)\) by inclusion. Prove inclusion is a partial order; identify its least and greatest elements; give an incomparable pair and explain why this is not a total order.
9 marksShow worked answer
Reflexivity follows from \(X\subseteq X\). Antisymmetry follows because \(X\subseteq Y\) and \(Y\subseteq X\) imply the sets have identical elements. Transitivity follows because \(X\subseteq Y\subseteq Z\) implies \(X\subseteq Z\). The least element is \(\varnothing\), and the greatest is \(S\). The sets \(\{1\}\) and \(\{2\}\) are incomparable, so not every pair is comparable; the order is not total.
Question 6 — Injectivity, surjectivity and inverse
For \(f:\mathbb R\to\mathbb R\), \(f(x)=x^3-3x\), determine whether \(f\) is injective and surjective. Does \(f^{-1}:\mathbb R\to\mathbb R\) exist? Justify without relying solely on a graph.
10 marksShow worked answer
It is not injective: \(f(0)=0=f(\sqrt3)\), although \(0\ne\sqrt3\). It is surjective: for arbitrary \(y\in\mathbb R\), set \(g(x)=x^3-3x-y\). This polynomial is continuous, tends to \(-\infty\) as \(x\to-\infty\) and \(+\infty\) as \(x\to+\infty\). The intermediate value theorem gives a real root, so \(f(x)=y\). No inverse on all of \(\mathbb R\) exists because \(f\) is not injective.
Question 7 — Contrapositive proof and parity
Prove that if \(n^2\) is odd for an integer \(n\), then \(n\) is odd. Use the contrapositive and identify where integrality is used.
6 marksShow worked answer
Prove the contrapositive: if \(n\) is even, then \(n^2\) is even. Since \(n\in\mathbb Z\), write \(n=2k\) for \(k\in\mathbb Z\). Then \(n^2=4k^2=2(2k^2)\), which is even. The contrapositive is equivalent to the required implication. Integrality permits the even representation \(n=2k\).
Question 8 — Irrationality by contradiction
Prove \(\sqrt3\) is irrational. State the lowest-terms assumption and justify why \(3\mid p^2\) implies \(3\mid p\).
8 marksShow worked answer
Assume \(\sqrt3=p/q\), with integers \(p,q\), \(q\ne0\), and \(\gcd(p,q)=1\). Squaring gives \(p^2=3q^2\), so \(3\mid p^2\). Since 3 is prime, Euclid's lemma gives \(3\mid p\); write \(p=3r\). Substitution gives \(9r^2=3q^2\), hence \(q^2=3r^2\) and \(3\mid q\). This contradicts \(\gcd(p,q)=1\). Therefore \(\sqrt3\) is irrational.
Question 9 — Induction and a summation identity
Prove for every integer \(n\ge1\) that \(\sum_{k=1}^{n}k(k+1)=\frac{n(n+1)(n+2)}3\). Explicitly use the induction hypothesis.
8 marksShow worked answer
Base case \(n=1\): both sides equal 2. Assume \(\sum_{k=1}^{n}k(k+1)=n(n+1)(n+2)/3\). Then
This is the formula for \(n+1\). Induction completes the proof.
Question 10 — Characterisation of injective mappings
Let \(f:A\to B\). Prove that \(f\) is injective iff for all \(X,Y\subseteq A\), \(f(X\cap Y)=f(X)\cap f(Y)\).
12 marksShow worked answer
Forward direction: If \(z\in f(X\cap Y)\), then \(z=f(a)\) for \(a\in X\cap Y\), so \(z\in f(X)\cap f(Y)\). Conversely, if \(z\in f(X)\cap f(Y)\), then \(z=f(x)=f(y)\) for some \(x\in X,y\in Y\). Injectivity gives \(x=y\), hence \(z\in f(X\cap Y)\).
Reverse direction: Suppose the identity holds for all subsets. If \(f(a)=f(b)=z\), choose \(X=\{a\}\), \(Y=\{b\}\). Then \(z\in f(X)\cap f(Y)=f(X\cap Y)\), so \(X\cap Y\ne\varnothing\). Therefore \(a=b\), proving injectivity.
Chapter 2 — Calculus I: Limits and Single-Variable Differentiation
Undergraduate · Chapter 2 This unit develops the limit as a mathematical object, continuity, differentiability, the derivative as a linear approximation, and the principal theorems governing differentiable functions.
2.1 The limit of a function
Let f be defined on a deleted neighbourhood of a∈ℝ. We say limx→af(x)=L if for every ε>0 there exists δ>0 such that 0<|x−a|<δ implies |f(x)−L|<ε. The value f(a), if defined, is irrelevant to the limit. Limits are unique: if a function had two distinct limits L and M, disjoint sufficiently small neighbourhoods around L and M would both have to contain all sufficiently near function values, a contradiction.
One-sided limits. The right-hand limit x→a⁺ restricts x>a; the left-hand limit x→a⁻ restricts x<a. A two-sided finite limit exists precisely when both one-sided limits exist and are equal.
Limit laws (sum, product, quotient with nonzero limiting denominator) follow from the ε–δ definition. For rational functions, cancellation may simplify a punctured-neighbourhood expression but does not change the original function's domain.
2.2 Continuity
f is continuous at a if f(a) exists, limx→af(x) exists, and the limit equals f(a). Equivalently, for every ε>0 there is δ>0 such that |x−a|<δ implies |f(x)−f(a)|<ε. Sums and products of continuous functions are continuous; a quotient is continuous where its denominator is nonzero; compositions preserve continuity.
Intermediate Value Theorem. If f is continuous on [a,b], then every value between f(a) and f(b) is attained at some c∈[a,b]. In particular, if f(a)f(b)<0, f has a zero in (a,b).
2.3 Derivative and differentiability
The derivative at a is the limit of the difference quotient, when it exists:
\[f'(a)=\lim_{h\to0}\frac{f(a+h)-f(a)}{h}.\]
Differentiability implies continuity, but continuity does not imply differentiability. For example, |x| is continuous at 0 but its left and right derivatives are −1 and +1.
Core differentiation rules. \((cf)'=cf'\), \((f+g)'=f'+g'\), \((fg)'=f'g+fg'\), and \((f/g)'=(f'g-fg')/g^2\) where g≠0. The chain rule gives \((f\circ g)'(x)=f'(g(x))g'(x)\).
For n∈ℤ, the power rule gives (xⁿ)'=nxⁿ⁻¹ on the domain where the expression is differentiable. Exponential and logarithmic derivatives include (eˣ)'=eˣ and (ln x)'=1/x for x>0. Trigonometric derivatives are (sin x)'=cos x and (cos x)'=−sin x.
2.4 Mean value theorems and consequences
Rolle's theorem. If f is continuous on [a,b], differentiable on (a,b), and f(a)=f(b), then some c∈(a,b) satisfies f'(c)=0.
Mean Value Theorem. If f is continuous on [a,b] and differentiable on (a,b), there exists c∈(a,b) such that \[f'(c)=\frac{f(b)-f(a)}{b-a}.\]
If f′(x)≥0 throughout an interval, the MVT implies f is nondecreasing there. If f′=0 throughout an interval, f is constant. The theorem also gives the standard uniqueness argument: if two differentiable functions have the same derivative on an interval, their difference is constant.
2.5 Taylor approximation and error
If f has sufficiently many derivatives near a, its degree-n Taylor polynomial is \(T_n(x)=\sum_{k=0}^{n}f^{(k)}(a)(x-a)^k/k!\). Under the hypotheses of Taylor's theorem, the remainder has Lagrange form \(R_n(x)=f^{(n+1)}(\xi)(x-a)^{n+1}/(n+1)!\) for some ξ between a and x. A polynomial approximation is useful only when its remainder is controlled.
Chapter 2 — University assessment problems
Show the argument, not only the final limit or derivative. State the theorem used and check its hypotheses.
Question 1 — Epsilon–delta proof of a limit
Using the formal ε–δ definition, prove that limx→2(3x−1)=5.
8 marksShow worked answer
Let ε>0. Choose δ=ε/3. If 0<|x−2|<δ, then |(3x−1)−5|=3|x−2|<3δ=ε. Thus the definition is satisfied and the limit is 5.
Question 2 — Removable discontinuity
Define f(x)=(x²−9)/(x−3) for x≠3 and f(3)=k. Determine the value of k that makes f continuous at 3, and prove continuity there using the limit definition.
8 marksShow worked answer
For x≠3, f(x)=x+3, so limx→3f(x)=6. Continuity requires k=6. With k=6, |f(x)−f(3)|=|(x+3)−6|=|x−3|. Given ε>0 choose δ=ε; then |x−3|<δ implies |f(x)−f(3)|<ε. Hence f is continuous at 3 exactly when k=6.
Question 3 — Derivative from first principles
For f(x)=x²−4x, calculate f′(a) directly from the difference quotient. Hence determine all stationary points and classify them.
10 marksShow worked answer
\[f'(a)=\lim_{h\to0}\frac{(a+h)^2-4(a+h)-(a^2-4a)}h=\lim_{h\to0}(2a+h-4)=2a-4.\] Thus f′(x)=2x−4, which vanishes at x=2. Since f(x)=(x−2)²−4, the stationary point is a global minimum at (2,−4).
Question 4 — Differentiability versus continuity
Let f(x)=|x−1|. Prove that f is continuous at x=1 but not differentiable there. Use one-sided difference quotients.
8 marksShow worked answer
Continuity follows from |f(x)−f(1)|=|x−1|→0. For h>0, [f(1+h)−f(1)]/h=h/h=1. For h<0, the quotient is (−h)/h=−1. Since the one-sided limits differ, f′(1) does not exist, despite continuity.
Question 5 — Chain rule with nested functions
Differentiate \(F(x)=\ln(1+e^{\sin(x^2)})\), stating the intermediate functions and their domains.
8 marksShow worked answer
Set u=x², v=sin u, w=eᵛ and F=ln(1+w). By the chain rule, F′(x)=[1/(1+eᵛ)]·eᵛ·cos(u)·2x. Therefore \[F'(x)=\frac{2xe^{\sin(x^2)}\cos(x^2)}{1+e^{\sin(x^2)}}.\] The logarithm's argument is strictly positive for all real x.
Question 6 — Product and quotient rules
For x>0, differentiate \(g(x)=x^2\ln x/(1+x)\) and simplify your answer.
8 marksShow worked answer
Let u=x²ln x and v=1+x. Then u′=2xln x+x and v′=1. The quotient rule gives \[g'(x)=\frac{(2x\ln x+x)(1+x)-x^2\ln x}{(1+x)^2}=\frac{x[(x+2)\ln x+x+1]}{(1+x)^2}.\]
Question 7 — Mean Value Theorem application
Verify the hypotheses of the Mean Value Theorem for f(x)=x² on [1,3], find all permissible c, and interpret the result geometrically.
7 marksShow worked answer
f is a polynomial, hence continuous on [1,3] and differentiable on (1,3). The theorem gives f′(c)=(9−1)/(3−1)=4. Since f′(c)=2c, c=2, which lies in (1,3). The tangent slope at x=2 equals the secant slope joining (1,1) and (3,9).
Question 8 — Monotonicity and uniqueness of a root
Show that the equation x³+x−1=0 has exactly one real root. You may use the Intermediate Value Theorem and a derivative argument.
10 marksShow worked answer
Let f(x)=x³+x−1. It is continuous, with f(0)=−1 and f(1)=1, so the IVT gives at least one root in (0,1). Also f′(x)=3x²+1>0 for every real x, so f is strictly increasing by the MVT. A strictly increasing function has at most one zero. Therefore exactly one real root exists.
Question 9 — Taylor polynomial and remainder bound
Find the degree-3 Maclaurin polynomial for eˣ and use Taylor's theorem to bound the error when approximating e⁰·² by this polynomial. Give a numerical upper bound.
10 marksShow worked answer
The polynomial is P₃(x)=1+x+x²/2+x³/6. Taylor's remainder satisfies |R₃(0.2)|≤e^0.2(0.2)^4/24 because the fourth derivative is eˣ and on [0,0.2] it is at most e^0.2. Thus |R₃|≤1.222·0.0016/24<0.000082.
Question 10 — A parameter-dependent differentiability problem
Let \(f(x)=\begin{cases}ax+b,&x\leq1\\x^2+1,&x>1.\end{cases}\) Determine a and b so that f is differentiable at 1. Prove the resulting function is differentiable there.
10 marksShow worked answer
Continuity at 1 requires a+b=2. Equality of one-sided derivatives requires a=2, since the left derivative is a and the right derivative is 2x evaluated at 1. Hence b=0. With (a,b)=(2,0), both pieces equal 2 at 1 and both one-sided difference quotients tend to 2, so f′(1)=2.
Chapter 3 — Calculus II: Integration and Series
Undergraduate level · Second calculus unit. This chapter develops integration as an inverse process to differentiation and as a limit of sums, then studies improper integrals, numerical approximation, sequences, infinite series and convergence tests. Students should justify convergence claims and state the hypotheses of every theorem used.
3.1 Antiderivatives and the definite integral
An antiderivative of f on an interval I is a differentiable function F satisfying F′(x)=f(x) on I. If F and G are antiderivatives of the same function on an interval, then F−G is constant. Thus an indefinite integral denotes a family: ∫f(x)dx=F(x)+C.
The definite integral is defined as a limit of Riemann sums. For a partition P of [a,b], choose sample points ξᵢ in each subinterval and form Σf(ξᵢ)Δxᵢ. If these sums approach a common limit as the mesh of the partition tends to zero, f is Riemann integrable and that limit is ∫ₐᵇf(x)dx.
Fundamental Theorem of Calculus, Part I. If f is continuous on [a,b] and F(x)=∫ₐˣ f(t)dt, then F is differentiable on (a,b) and F′(x)=f(x).
Fundamental Theorem of Calculus, Part II. If f is continuous on [a,b] and F′=f, then ∫ₐᵇ f(x)dx=F(b)−F(a).
The theorem links the limit-of-sums definition to antiderivatives. It does not mean every function has an elementary antiderivative; for example, ∫e−x²dx has no elementary closed form.
3.2 Integration techniques
Substitution. If u=g(x), then ∫f(g(x))g′(x)dx=∫f(u)du. For definite integrals, transform both the integrand and the limits, or substitute and then evaluate using the original limits consistently.
Integration by parts. From the product rule, ∫u,dv=uv−∫v,du. For definite integrals, ∫ₐᵇu v′dx=[uv]ₐᵇ−∫ₐᵇv u′dx.
Partial fractions. A proper rational function can be decomposed into simpler fractions after factoring its denominator over the relevant field. Repeated linear factors require terms A₁/(x−a)+A₂/(x−a)²+…; irreducible real quadratic factors require linear numerators.
Trigonometric substitution. Expressions involving √(a²−x²), √(a²+x²), or √(x²−a²) can be simplified using x=a sinθ, x=a tanθ, or x=a secθ, respectively, with domains and signs handled carefully.
3.3 Improper integrals
An integral is improper when the interval is unbounded or the integrand is unbounded at an endpoint or interior point. It is defined by a limit. For example, ∫₁∞f(x)dx means limR→∞∫₁ᴿf(x)dx, if the limit exists finitely. If the limit fails to exist or is infinite, the integral diverges.
p-integral tests. ∫₁∞x−pdx converges iff p>1. ∫₀¹x−pdx converges iff p<1. These are distinct endpoint behaviours and should not be confused.
For nonnegative functions, comparison is useful: if 0≤f≤g and ∫g converges, then ∫f converges; if 0≤g≤f and ∫g diverges, then ∫f diverges. Limit comparison applies when f/g tends to a positive finite constant.
3.4 Numerical integration
The composite trapezoidal rule approximates an integral by replacing the curve on each subinterval with a straight segment. With equal width h=(b−a)/n:
If f has a continuous second derivative, the composite trapezoidal error is bounded by (b−a)h² max|f″|/12. If f has a continuous fourth derivative, the composite Simpson error is bounded by (b−a)h⁴ max|f⁽⁴⁾|/180. These are error bounds, not estimates that always equal the actual error.
3.5 Sequences and limits
A sequence (aₙ) converges to L if for every ε>0 there exists N such that n≥N implies |aₙ−L|<ε. A sequence is bounded if |aₙ|≤M for some M and all n. Every convergent sequence is bounded, and every monotone bounded sequence converges. The converse of the latter statement is false in general: convergence does not require monotonicity.
Monotone convergence theorem for sequences. Every increasing sequence bounded above converges to its supremum; every decreasing sequence bounded below converges to its infimum.
3.6 Infinite series and convergence
An infinite series Σaₙ is defined through its partial sums sₙ=Σk=1ⁿaₖ. The series converges precisely when (sₙ) converges. A necessary condition is aₙ→0, but this condition alone is not sufficient: the harmonic series Σ1/n diverges.
Geometric series: Σₙ₌₀∞arⁿ converges to a/(1−r) iff |r|<1.
p-series: Σ1/nᵖ converges iff p>1.
Comparison: compare nonnegative terms with a known convergent or divergent series.
Limit comparison: if aₙ,bₙ>0 and lim aₙ/bₙ=c∈(0,∞), both series have the same convergence behaviour.
Ratio test: if L=limsup |aₙ₊₁/aₙ|, convergence is absolute for L<1 and divergence for L>1; L=1 is inconclusive.
Root test: use limsup ⁿ√|aₙ| with the same thresholds.
Alternating series test: Σ(−1)ⁿbₙ converges if bₙ decreases monotonically to zero. The error after N terms is at most bₙ₊₁.
3.7 Absolute and conditional convergence
A series Σaₙ converges absolutely if Σ|aₙ| converges. Absolute convergence implies convergence. A series is conditionally convergent if it converges but does not converge absolutely. For conditionally convergent real series, rearranging terms can change the sum; this is why absolute convergence is a stronger and structurally important property.
3.8 Power series and Taylor series
A power series centred at a is Σcₙ(x−a)ⁿ. Its radius of convergence R is determined, when the ratio limit exists, by R=lim |cₙ/cₙ₊₁|. It converges absolutely for |x−a|<R and diverges for |x−a|>R; each endpoint must be tested separately. Within the interval of convergence, a power series may be differentiated and integrated term by term on compact subintervals.
The Taylor series of f at a is Σf⁽ⁿ⁾(a)(x−a)ⁿ/n!. A function need not equal its Taylor series merely because all derivatives exist. Equality requires a suitable remainder to tend to zero. For example, eˣ has the globally convergent series Σxⁿ/n!, while smoothness alone is not enough to guarantee such representation.
Chapter 3 — Undergraduate assessment problems
Write complete solutions. For convergence problems, identify the test and verify its conditions; for improper integrals, show the limiting process explicitly.
Question 1 — Riemann integral and the Fundamental Theorem
Evaluate \(\int_0^2(3x^2-4x+1)\,dx\). Then verify the result by interpreting the antiderivative difference and state which part of the Fundamental Theorem of Calculus is being used.
7 marksShow worked answer
An antiderivative is F(x)=x³−2x²+x. By the Fundamental Theorem of Calculus, \[\int_0^2(3x^2-4x+1)dx=F(2)-F(0)=(8-8+2)-0=2.\] The integrand is continuous on [0,2], so the theorem applies. This is Part II: a definite integral is evaluated by an antiderivative at the endpoints.
Question 2 — Integration by parts with a parameter
For a>0, derive a closed expression for \(I(a)=\int_0^a xe^{2x}\,dx\) using integration by parts. Evaluate I(1) exactly.
8 marksShow worked answer
Choose u=x and dv=e²ˣdx, so du=dx and v=e²ˣ/2. Then \[I(a)=\left[\frac{xe^{2x}}2\right]_0^a-\frac12\int_0^a e^{2x}dx=\frac{ae^{2a}}2-\frac{e^{2a}-1}{4}.\] Thus I(1)=(e²+1)/4.
Question 3 — Partial fractions
Evaluate \(\int \frac{2x+5}{(x+1)(x+2)}dx\), showing the decomposition and the domain on which the antiderivative is valid.
8 marksShow worked answer
Write (2x+5)/[(x+1)(x+2)]=A/(x+1)+B/(x+2). Equating coefficients gives A+B=2 and 2A+B=5, so A=3, B=−1. Therefore the integral is \(3\ln|x+1|-\ln|x+2|+C\), valid on any interval not containing x=−1 or x=−2.
Question 4 — Improper integral and convergence threshold
For real p, determine exactly when \(\int_1^\infty x^{-p}dx\) converges. Evaluate it when it converges and explain the borderline case p=1.
10 marksShow worked answer
For p≠1, \(\int_1^R x^{-p}dx=(R^{1-p}-1)/(1-p)\). If p>1, R^{1−p}→0 and the integral equals 1/(p−1). If p<1, R^{1−p}→∞ and it diverges. For p=1, the integral is ln R→∞. Hence convergence occurs exactly for p>1.
Question 5 — Integral comparison
Determine whether \(\int_1^\infty \frac{1}{x^2+\sqrt{x}}dx\) converges. Supply a rigorous comparison argument and do not attempt to find an elementary antiderivative.
8 marksShow worked answer
For x≥1, x²+√x≥x², so 0<1/(x²+√x)≤1/x². Since ∫₁∞x⁻²dx=1 converges, the direct comparison test proves the given improper integral converges.
Question 6 — Sequence convergence by a recurrence
Let x₁=1 and \(x_{n+1}=\sqrt{2+x_n}\). Prove that (xₙ) converges and determine its limit. Justify boundedness and monotonicity.
12 marksShow worked answer
First show 0<xₙ<2 by induction: it holds at n=1; if xₙ<2, then xₙ₊₁=√(2+xₙ)<2 and is positive. Next, xₙ₊₁>xₙ is equivalent, since both are positive, to 2+xₙ>xₙ², i.e. (2−xₙ)(xₙ+1)>0, true for 0<xₙ<2. Thus the sequence is increasing and bounded above by 2, so it converges. If L is its limit, continuity of the square root gives L=√(2+L), hence L²−L−2=0. Since L≥0, L=2.
Question 7 — Ratio test and radius of convergence
Find the radius and interval of convergence of \(\sum_{n=1}^{\infty}\frac{(x-2)^n}{n3^n}\). Test both endpoints separately.
12 marksShow worked answer
For aₙ=(x−2)ⁿ/(n3ⁿ), the ratio of successive absolute terms tends to |x−2|/3. The ratio test gives absolute convergence if |x−2|<3 and divergence if |x−2|>3, so R=3. At x=5, the series is Σ1/n, which diverges. At x=−1, it is Σ(−1)ⁿ/n, which converges by the alternating series test (but not absolutely). The interval is [−1,5).
Question 8 — Conditional versus absolute convergence
Classify \(\sum_{n=1}^{\infty}\frac{(-1)^{n-1}}{\sqrt n}\) as absolutely convergent, conditionally convergent or divergent. Give a separate argument for the original series and its absolute-value series.
9 marksShow worked answer
The positive sequence bₙ=1/√n decreases to zero, so the alternating series test proves convergence. The absolute-value series is Σ1/√n, a p-series with p=1/2≤1, hence divergent. Therefore the original series is conditionally convergent, not absolutely convergent.
Question 9 — Taylor series and remainder control
Use Taylor's theorem about 0 to obtain a degree-4 approximation for ln(1+x). For x=0.2, give the polynomial value and a rigorous bound for the absolute error.
10 marksShow worked answer
The expansion is ln(1+x)=x−x²/2+x³/3−x⁴/4+R₄. At x=0.2 the polynomial is 0.2−0.02+0.0026666667−0.0004=0.1822666667. Since the fifth derivative is 24/(1+x)⁵, Taylor's Lagrange remainder has magnitude \(|R_4|=|f^{(5)}(\xi)|0.2^5/5!\) for some ξ∈(0,0.2). On this interval, |f⁽⁵⁾(ξ)|≤24, so |R₄|≤24(0.2)⁵/120=0.000064. (The bound is conservative.)
Question 10 — Simpson's rule and an error bound
Apply the composite Simpson rule with n=2 to approximate \(\int_0^1 e^x dx\). Then use the standard fourth-derivative error bound to obtain a rigorous upper bound for the error.
10 marksShow worked answer
Here h=1/2 and the nodes are 0, 1/2, 1. Simpson gives \[S_2=\frac{1}{6}[f(0)+4f(1/2)+f(1)]=\frac{1+4e^{1/2}+e}{6}\approx1.71886.\] Since f⁽⁴⁾(x)=eˣ≤e on [0,1], the error bound is \(|E_S|≤(b-a)h^4M/180=e/(16·180)=e/2880\approx0.000944.\]
Chapter 4 — Multivariable Calculus
Undergraduate level · Several-variable differential and integral calculus. This chapter extends limits, continuity, differentiation and integration to functions of several real variables. It develops partial derivatives, differentiability, the chain rule, gradient methods, constrained extrema, multiple integrals and coordinate transformations.
4.1 Functions of several variables and domains
A scalar-valued function of two variables is a mapping f:D⊆ℝ²→ℝ, while a vector-valued mapping may take values in ℝᵐ. The domain must be determined from restrictions such as nonzero denominators, nonnegative radicands and positive logarithm arguments. A level set f(x,y)=c is a curve in the domain; in three dimensions it is a level surface.
The graph z=f(x,y) is a surface. A trace is obtained by intersecting it with a coordinate plane or another plane. Contour plots display level curves and are useful for understanding gradients and extrema.
4.2 Limits and continuity in ℝ²
We say lim(x,y)→(a,b)f(x,y)=L if for every ε>0 there is δ>0 such that 0<√((x−a)²+(y−b)²)<δ implies |f(x,y)−L|<ε. Unlike one-variable limits, checking a few paths is not enough to prove a multivariable limit exists. However, two paths giving different limiting values prove that the limit does not exist.
Useful existence methods include algebraic simplification, polar coordinates x=r cosθ, y=r sinθ, and bounds. If |f(x,y)−L|≤g(r) where g(r)→0 independently of θ, the squeeze theorem proves the limit. Continuity is defined by equality of the function value and the limit.
4.3 Partial derivatives and higher derivatives
The first partial derivatives are defined by holding the other variables fixed:
Second partial derivatives include fₓₓ, fₓᵧ, fᵧₓ and fᵧᵧ. If the relevant mixed partial derivatives are continuous in a neighbourhood, Clairaut's theorem gives fₓᵧ=fᵧₓ. The Hessian matrix of a twice differentiable scalar function is the matrix of second partial derivatives.
4.4 Differentiability and the total differential
Partial derivatives existing at a point do not, by themselves, guarantee differentiability there. Differentiability at (a,b) means there is a linear map giving a first-order approximation with an error that is small relative to the distance from (a,b):
A sufficient condition is that the first partial derivatives exist in a neighbourhood and are continuous at the point. The total differential is df=fₓdx+fᵧdy. For small increments, Δf≈df, with the approximation justified by differentiability.
4.5 The gradient and directional derivatives
The gradient is ∇f=(fₓ,fᵧ) in ℝ² and ∇f=(fₓ,fᵧ,f_z) in ℝ³. For a unit vector u, the directional derivative is Dᵤf=∇f·u. The gradient points in the direction of steepest increase and its magnitude is the maximum directional derivative. At a regular point of a level curve or surface, ∇f is normal to the level set.
4.6 The multivariable chain rule
If z=f(x,y), where x=x(t) and y=y(t), then dz/dt=fₓ dx/dt+fᵧ dy/dt. For a composition of mappings, the derivative is the product of the corresponding Jacobian matrices. This matrix form is essential in coordinate changes, optimisation and numerical algorithms.
For implicit relations F(x,y,z)=0, if F_z≠0 locally, the implicit function theorem permits z to be treated locally as a function of x and y, with zₓ=−Fₓ/F_z and zᵧ=−Fᵧ/F_z.
4.7 Critical points and the second derivative test
An interior critical point occurs where ∇f=0 or a required derivative does not exist. For a twice continuously differentiable function of two variables, define D=fₓₓfᵧᵧ−(fₓᵧ)² at a critical point. If D>0 and fₓₓ>0, there is a local minimum; if D>0 and fₓₓ<0, a local maximum; if D<0, a saddle point. If D=0, the test is inconclusive.
For global extrema on a compact domain, continuity guarantees that maximum and minimum values are attained. One must examine interior critical points and the boundary, including corners or endpoints where applicable.
4.8 Constrained optimisation and Lagrange multipliers
To optimise f(x,y) subject to g(x,y)=c, regular constrained extrema satisfy ∇f=λ∇g, provided ∇g≠0 at the candidate. Geometrically, the level curves of f and g are tangent at a regular constrained extremum. The candidates must then be evaluated and compared; the multiplier equations alone do not classify a point as a maximum or minimum.
4.9 Double and triple integrals
A double integral ∬Rf(x,y)dA accumulates a scalar density over a planar region. For a rectangular region, an iterated integral can be evaluated by Fubini's theorem when the integrand is continuous (or under suitable integrability hypotheses). For nonrectangular regions, describe the bounds carefully using vertical or horizontal slices. Triple integrals extend this construction to volume regions.
4.10 Change of variables and Jacobians
For a differentiable one-to-one coordinate transformation (u,v)↦(x(u,v),y(u,v)), area transforms according to dA=|J|dudv, where J=∂(x,y)/∂(u,v). Thus ∬Rf(x,y)dA= ∬Sf(x(u,v),y(u,v))|J|dudv. In polar coordinates x=r cosθ, y=r sinθ and |J|=r, so dA=r dr dθ. In cylindrical coordinates dV=r dz dr dθ; in spherical coordinates dV=ρ²sinφ dρ dφ dθ under the usual convention for φ.
4.11 Applications and interpretation
Multivariable calculus is used to model temperature fields, pressure, population surfaces, production functions and energy landscapes. Partial derivatives measure marginal effects; the gradient describes local sensitivity; Hessians describe local curvature; multiple integrals calculate mass, charge or accumulated quantities from densities.
Chapter 4 — Undergraduate assessment problems
Show all mathematical reasoning. For limit questions, justify path or bounding arguments; for optimisation, identify and test every candidate; for integration, state the region and Jacobian.
Question 1 — Domain and level curves
Let \(f(x,y)=\ln(9-x^2-4y^2)\). Determine the domain, describe its boundary, and derive the level curve corresponding to f(x,y)=ln 5. State whether the level curve is a circle or ellipse.
8 marksShow worked answer
The logarithm requires 9−x²−4y²>0, so the domain is the open ellipse x²+4y²<9. Its boundary is x²+4y²=9. Setting f=ln5 gives 9−x²−4y²=5, hence x²+4y²=4, an ellipse centred at the origin with semi-axes 2 and 1.
Question 2 — A two-path limit test
Investigate whether \(\lim_{(x,y)\to(0,0)}\frac{x^2-y^2}{x^2+y^2}\) exists. Your conclusion must be justified rigorously.
8 marksShow worked answer
Along y=0 the expression equals 1 for x≠0, so the path limit is 1. Along x=0 it equals −1 for y≠0, so the path limit is −1. Since two paths approaching the same point give different limits, the multivariable limit does not exist.
Question 3 — Polar-coordinate squeeze argument
Evaluate \(\lim_{(x,y)\to(0,0)}\frac{x^2y^2}{x^2+y^2}\), using a bound that is independent of the direction of approach.
8 marksShow worked answer
Write x=r cosθ and y=r sinθ. The expression becomes r²cos²θsin²θ. Since 0≤cos²θsin²θ≤1/4, its absolute value is at most r²/4, which tends to zero as r→0 independently of θ. By the squeeze theorem, the limit is 0.
Question 4 — Partial derivatives and Hessian
For \(f(x,y)=x^3y^2+e^{xy}\), calculate fₓ, fᵧ, fₓₓ, fₓᵧ and fᵧᵧ. Verify equality of the mixed partial derivatives.
12 marksShow worked answer
fₓ=3x²y²+ye^{xy}; fᵧ=2x³y+xe^{xy}. Differentiating again gives fₓₓ=6xy²+y²e^{xy}; fₓᵧ=6x²y+e^{xy}+xye^{xy}; fᵧᵧ=2x³+x²e^{xy}. The mixed derivative obtained by differentiating fᵧ with respect to x is 6x²y+e^{xy}+xye^{xy}, equal to fₓᵧ. This agrees with Clairaut's theorem because the function is smooth on ℝ².
Question 5 — Differentiability and linear approximation
Let \(f(x,y)=\sqrt{4+x+2y}\). Find the linearisation at (0,0), then use it to approximate f(0.04,−0.03). Explain why the linear approximation is valid locally.
9 marksShow worked answer
f(0,0)=2. The partial derivatives are fₓ=1/[2√(4+x+2y)] and fᵧ=1/√(4+x+2y), so at the origin they are 1/4 and 1/2. Hence L(x,y)=2+x/4+y/2. The approximation is 2+0.04/4−0.03/2=2.005. The radicand is positive in a neighbourhood of the origin and the first partial derivatives are continuous there, so f is differentiable and its linearisation gives a first-order local approximation.
Question 6 — Directional derivative and steepest ascent
For \(f(x,y,z)=x e^{yz}+y^2z\), calculate the directional derivative at P=(1,0,2) in the direction from P to Q=(3,1,2). Also determine the direction and magnitude of steepest ascent at P.
10 marksShow worked answer
∇f=(e^{yz}, xze^{yz}+2yz, xye^{yz}+y²). At P this is (1,2,0). The direction vector PQ=(2,1,0) has unit vector u=(2/√5,1/√5,0). Thus Dᵤf=∇f·u=4/√5. The steepest-ascent direction is ∇f/|∇f|=(1/√5,2/√5,0), and the maximum directional derivative is |∇f|=√5.
Question 7 — Implicit differentiation in three variables
The surface is defined by \(F(x,y,z)=x^2+y^2+z^2+xyz-7=0\). At the point (1,2,1), verify that the point lies on the surface and calculate zₓ and zᵧ, assuming z is locally a function of x and y.
10 marksShow worked answer
At (1,2,1), F=1+4+1+2−7=1, not zero. Therefore the stated point is not on the surface and implicit derivatives of this surface at that point are not defined. This is an intentional consistency check: one must verify the constraint before applying the implicit function theorem. For a valid nearby point, the formulas would be zₓ=−(2x+yz)/(2z+xy) and zᵧ=−(2y+xz)/(2z+xy), provided 2z+xy≠0.
Question 8 — Classification of critical points
Find and classify all critical points of \(f(x,y)=x^3+y^3-3xy\). Use the Hessian determinant and identify any inconclusive cases.
12 marksShow worked answer
fₓ=3x²−3y and fᵧ=3y²−3x. The critical equations give y=x² and x=y², hence x=x⁴, so x=0 or x=1; corresponding y values are 0 and 1. The points are (0,0) and (1,1). The Hessian entries are fₓₓ=6x, fᵧᵧ=6y, fₓᵧ=−3, so D=36xy−9. At (0,0), D=−9<0, a saddle. At (1,1), D=27>0 and fₓₓ=6>0, a strict local minimum.
Question 9 — Lagrange multipliers on a circle
Find the maximum and minimum values of \(f(x,y)=3x+4y\) subject to x²+y²=25. Solve using Lagrange multipliers and interpret the result geometrically.
10 marksShow worked answer
Let g=x²+y². The equations ∇f=λ∇g give (3,4)=λ(2x,2y), so (x,y) is parallel to (3,4). Combining with x²+y²=25 gives (x,y)=(3,4) or (−3,−4). The corresponding values are 25 and −25. Geometrically, the dot product of (3,4) with a vector of length 5 is bounded by 5·5=25, with equality in the parallel and antiparallel directions.
Question 10 — Double integral and change of variables
Evaluate \(\iint_R(x+y)\,dA\), where R is the ellipse x²/4+y²=1, using the substitution x=2u, y=v and the unit disk u²+v²≤1. Explain the role of the Jacobian.
12 marksShow worked answer
The Jacobian determinant is |∂(x,y)/∂(u,v)|=|2·1−0|=2. The transformed integral is ∬u²+v²≤1(2u+v)·2 dudv. The disk is symmetric about both coordinate axes, so the integrals of u and v over it vanish. Therefore the value is 0. The factor 2 is the area-scaling factor of the transformation; omitting it would give an incorrect transformed integral, even though symmetry makes this particular final value zero.
Chapter 5 — Vector Calculus
Undergraduate level · Vector fields, integral theorems and geometric analysis. This chapter develops the differential and integral calculus of scalar and vector fields. It connects local operators—gradient, divergence and curl—to global quantities through line, surface and volume integrals. Students should be able to state theorem hypotheses, orient curves and surfaces consistently, and select an appropriate integral theorem for a given problem.
5.1 Scalar fields, vector fields and parameterised curves
A scalar field on a region D⊆ℝⁿ is a function φ:D→ℝ, such as temperature or density. A vector field is a mapping F:D→ℝⁿ. In three dimensions write F(x,y,z)=P(x,y,z)i+Q(x,y,z)j+R(x,y,z)k. A vector field assigns both a direction and magnitude to each point in its domain. Examples include velocity, force and electromagnetic fields.
A smooth parameterised curve is r(t)=(x(t),y(t),z(t)), a≤t≤b. Its velocity is r′(t), its speed is |r′(t)|, and its acceleration is r″(t). The curve is regular where r′(t)≠0. Its arc length is L=∫ₐᵇ|r′(t)|dt. Reversing the parameter changes the direction of traversal but not the geometric curve or its length.
Unit tangent. For a regular curve, T(t)=r′(t)/|r′(t)|. When the derivative exists and is nonzero, the curvature is κ=|dT/ds|=|r′×r″|/|r′|³ in ℝ³. Curvature measures the rate at which the unit tangent turns per unit arc length.
5.2 Differentiation and integration of vector-valued functions
For r(t)=(x₁(t),…,xₙ(t)), differentiation and integration are componentwise: r′(t)=(x₁′,…,xₙ′) and ∫r(t)dt=(∫x₁dt,…,∫xₙdt)+C. The fundamental theorem of calculus therefore applies componentwise. In mechanics, if r is position, v=r′ is velocity and a=v′ is acceleration. If acceleration is known, integration determines velocity and position only after initial conditions are supplied.
For a particle moving along a curve, speed is |v|, not generally the signed derivative of a single coordinate. The tangential component of acceleration is dv/dt and the normal component has magnitude κv². Thus a=(dv/dt)T+κv²N where the Frenet normal N is defined when curvature is nonzero.
5.3 The gradient of a scalar field
For a differentiable scalar field φ(x,y,z), its gradient is ∇φ=(φₓ,φᵧ,φ_z). For a unit direction u, the directional derivative is Dᵤφ=∇φ·u. The gradient is normal to a regular level surface φ=c and points in the direction of greatest local increase. The differential dφ=∇φ·dr describes the first-order change in φ.
For example, if φ(x,y,z)=x²y+z³, then ∇φ=(2xy,x²,3z²). At a point where ∇φ≠0, the tangent plane to the level surface φ=constant is ∇φ(a)·(r−a)=0. A normal vector is therefore ∇φ(a).
5.4 Divergence and its interpretation
For F=(P,Q,R), the divergence is the scalar field div F=∇·F=Pₓ+Qᵧ+R_z. It measures the local net outward flux per unit volume in the infinitesimal limit. Positive divergence indicates a local source, negative divergence a local sink, and zero divergence a solenoidal field. The physical interpretation depends on the field and its units.
Flux density interpretation. For a small volume V around a point where F is smooth, div F is the limit of the outward flux through ∂V divided by the volume, as V shrinks to the point. This is a local statement; zero divergence does not imply that F itself is zero.
Curl measures infinitesimal circulation and the axis of local rotation. A field with curl zero is called irrotational. On a simply connected open region in ℝ³, a continuously differentiable field with curl F=0 is conservative; the domain condition matters because holes can obstruct a global potential. In two dimensions, for F=(P,Q), the scalar out-of-plane curl is Qₓ−Pᵧ.
5.6 Line integrals of scalar fields and vector fields
Let C be parameterised by r(t), a≤t≤b. The scalar line integral is ∫Cg ds=∫ₐᵇg(r(t))|r′(t)|dt. It accumulates a scalar density along the curve; for example, it can calculate the mass of a thin wire with linear density g.
The work integral of a vector field is ∫CF·dr=∫ₐᵇF(r(t))·r′(t)dt. It measures work done by a force along a path. Reversing the orientation changes the sign of a vector line integral, while a scalar line integral with ds is unchanged. A valid parameterisation must trace the intended curve with the stated orientation.
5.7 Conservative fields, potential functions and path independence
A vector field F is conservative if there exists a scalar potential φ such that F=∇φ. For a piecewise smooth curve C from A to B, the fundamental theorem for line integrals gives ∫CF·dr=φ(B)−φ(A). Consequently, the integral is independent of the path and is zero around every closed curve in the domain.
To test a proposed potential in ℝ³, integrate one component and compare the remaining partial derivatives. For F=(P,Q,R), a candidate φ must satisfy φₓ=P, φᵧ=Q and φ_z=R. Equality of cross-partials is a useful necessary test. On a simply connected domain, a C¹ field with curl zero is conservative; on a domain with holes, curl zero alone need not guarantee a global potential.
5.8 Green's theorem in the plane
Let C be a positively oriented, simple, closed, piecewise smooth curve bounding a region D, and let P,Q have continuous first partial derivatives on an open region containing D. Green's theorem states
The positive orientation is counterclockwise, leaving the region on the left as the curve is traversed. The flux form is
\[\oint_C(P\,dy-Q\,dx)=\iint_D(P_x+Q_y)\,dA.\]
Green's theorem converts a boundary line integral into a double integral over the enclosed region, or conversely. It is especially useful when the boundary is complicated but the planar curl or divergence is simple. For regions with holes, all boundary components must be oriented consistently with the positive orientation of the region.
5.9 Surfaces, parametrisation and surface area
A smooth surface patch may be parameterised by r(u,v)=(x(u,v),y(u,v),z(u,v)). The tangent vectors are rᵤ and rᵥ; a normal vector is rᵤ×rᵥ. The area element is dS=|rᵤ×rᵥ|dudv. The parameterisation is regular where rᵤ×rᵥ≠0. Surface area is ∬S1 dS.
For a graph z=f(x,y), a standard upward orientation has dS=√(1+fₓ²+fᵧ²)dA and an upward normal area vector n dS=(-fₓ,-fᵧ,1)dA. The unit normal is obtained by dividing by its magnitude. Orientation is essential for flux integrals but not for scalar surface area.
5.10 Surface integrals and flux
The scalar surface integral is ∬Sg dS=∬Dg(r(u,v))|rᵤ×rᵥ|dudv. The flux of F through an oriented surface is
where the order of the cross product must agree with the chosen orientation. Reversing the normal reverses the flux sign. Flux measures the amount of a vector field crossing a surface, rather than the amount lying along it.
5.11 Stokes' theorem
Stokes' theorem. If S is an oriented, piecewise smooth surface with positively oriented boundary C=∂S, and F is continuously differentiable on an open set containing S, then
The orientation of C and the normal n are linked by the right-hand rule: when the right thumb points along n, the curled fingers indicate the positive boundary direction. Stokes' theorem equates circulation around the boundary with the flux of curl through the surface. Different surfaces with the same oriented boundary give the same value when the theorem's hypotheses apply.
5.12 The divergence theorem (Gauss' theorem)
Divergence theorem. Let V be a bounded solid region with a piecewise smooth, closed boundary surface ∂V, oriented by the outward unit normal n. If F is continuously differentiable on an open set containing V, then
The theorem converts total outward flux across a closed surface into the volume integral of divergence. The boundary must be closed and outward-oriented. It is often simpler to integrate divergence over the volume than to parameterise every face of the boundary.
5.13 Identities and theorem selection
For sufficiently smooth scalar and vector fields, important identities include ∇×(∇φ)=0, ∇·(∇×F)=0, and ∇·(φF)=∇φ·F+φ∇·F. These identities are consequences of equality of mixed partial derivatives under appropriate regularity assumptions. They provide checks on calculations and appear throughout fluid mechanics, electromagnetism and continuum physics.
Use the gradient for directional change and normals to level sets.
Use divergence for local sources and sinks, or total flux through a closed surface.
Use curl for local circulation, or Stokes' theorem on an oriented surface with boundary.
Use Green's theorem for a positively oriented planar boundary.
Use a potential function to evaluate work when the field is conservative and the endpoints are known.
5.14 Worked university examples
Example 1 — Work in a conservative field. Let F=(2xy,x²+2y). Since F=∇φ for φ=x²y+y², the work from A=(0,0) to B=(1,2) is φ(1,2)−φ(0,0)=1·2+4=6. No path parameterisation is required.
Example 2 — Divergence theorem. For F=(x,y,z) and the sphere x²+y²+z²≤a², div F=3. Hence the outward flux is 3 times the volume, namely 3(4πa³/3)=4πa³.
Example 3 — Stokes' theorem. For F=(-y/2,x/2,0) and the unit disk in the xy-plane with upward normal, curl F=(0,0,1). The surface integral is the disk area π, so the counterclockwise boundary circulation is π.
Chapter 5 — Undergraduate assessment problems
Attempt each problem before opening its worked solution. Show parameterisations, bounds, orientations and theorem hypotheses. Unsupported theorem names or unexplained sign choices receive no full credit.
Question 1 — Kinematics, arc length and curvature
A particle follows the curve \(r(t)=(3\cos t,3\sin t,4t)\), for \(0\leq t\leq2\pi\). (a) Find velocity, acceleration and speed. (b) Determine the distance travelled over the interval. (c) Calculate the curvature and explain whether it is constant.
12 marksShow worked answer
(a) v=r′=(-3sin t,3cos t,4), a=r″=(-3cos t,-3sin t,0). The speed is √(9sin²t+9cos²t+16)=5. (b) The distance is ∫₀²π5dt=10π. (c) r′×r″=(12sin t,-12cos t,9), whose magnitude is √(144+81)=15. Since |r′|=5, κ=|r′×r″|/|r′|³=15/125=3/25. The curvature is constant.
Question 2 — Gradient, tangent plane and directional derivative
Let \(\phi(x,y,z)=x^2y+yz^2\). At P=(1,2,1), (a) calculate the gradient; (b) find the tangent plane to the level surface through P; (c) calculate the directional derivative in the direction from P to Q=(3,2,2).
12 marksShow worked answer
∇φ=(2xy,x²+z²,2yz). At P, ∇φ=(4,2,4), and φ(P)=3. The tangent plane is ∇φ(P)·((x−1),(y−2),(z−1))=0, so 4(x−1)+2(y−2)+4(z−1)=0, or 2x+y+2z=6. The direction PQ=(2,0,1) has unit vector (2/√5,0,1/√5). The directional derivative is (4,2,4)·(2/√5,0,1/√5)=12/√5.
Question 3 — Divergence and curl
For \(F=(x^2y, yz^2, xz)\), calculate div F and curl F. Evaluate both at (1,2,−1), and interpret the divergence at that point.
12 marksShow worked answer
With P=x²y, Q=yz² and R=xz, div F=Pₓ+Qᵧ+R_z=2xy+z²+x. Thus at (1,2,−1), div F=4+1+1=6, indicating positive local outward source strength. Curl F=(Rᵧ−Q_z,P_z−Rₓ,Qₓ−Pᵧ)=(-2yz,-z,-x²). At the point this is (4,1,−1).
Question 4 — Work along a parameterised curve
Let \(F(x,y)=(y,x+2y)\) and let C be the straight line from A=(0,0) to B=(2,1). (a) Parameterise C with its orientation. (b) Evaluate ∫CF·dr directly. (c) Determine whether F is conservative and verify the result using a potential.
12 marksShow worked answer
Take r(t)=(2t,t), 0≤t≤1, so r′=(2,1). Along C, F=(t,4t). Hence F·r′=2t+4t=6t and the integral is ∫₀¹6t dt=3. For F=(P,Q), Pᵧ=1 and Qₓ=1 on ℝ², a simply connected domain, so F is conservative. Integrate φₓ=y to obtain φ=xy+h(y). Then φᵧ=x+h′(y)=x+2y, giving h=y² and φ=xy+y². The endpoint difference is φ(2,1)−φ(0,0)=2+1=3.
Question 5 — A non-conservative field and path dependence
Consider \(F(x,y)=(-y,x)\). Compute the line integral from (0,0) to (1,1) along (i) the two-segment path consisting of the x-axis followed by a vertical segment, and (ii) the straight line y=x. Explain why the answers differ.
12 marksShow worked answer
(i) On the x-axis, y=0 and dy=0, so F·dr=−y dx+x dy=0. On the vertical segment x=1, y=t, 0≤t≤1, dx=0 and the integral is ∫₀¹1dt=1. (ii) Set x=t,y=t, 0≤t≤1. Then dx=dy=dt and −y dx+x dy=−t dt+t dt=0. The two values differ. Indeed Qₓ−Pᵧ=1−(−1)=2, so the field is not conservative on ℝ²; the integral is path-dependent.
Question 6 — Green's theorem and area
Use Green's theorem to evaluate \(\oint_C(-y^3\,dx+x^3\,dy)\), where C is the positively oriented boundary of the rectangle 0≤x≤2, 0≤y≤1. State the orientation and show the double integral.
10 marksShow worked answer
Here P=−y³ and Q=x³. Green's theorem gives the integral as ∬D(Qₓ−Pᵧ)dA=∫₀²∫₀¹(3x²+3y²)dy dx. The inner integral is 3x²+1. Integrating from 0 to 2 gives [x³+x]₀²=8+2=10. The positive orientation is counterclockwise.
Question 7 — Surface parametrisation and scalar surface integral
Let S be the portion of the plane z=2x+3y above the rectangle 0≤x≤1, 0≤y≤2. Calculate its surface area and evaluate ∬S(x+y)dS.
12 marksShow worked answer
Parameterise r(x,y)=(x,y,2x+3y). Then rₓ=(1,0,2), rᵧ=(0,1,3), and rₓ×rᵧ=(-2,-3,1), with magnitude √14. Therefore dS=√14 dxdy. The area is √14 times the rectangle area 2, namely 2√14. The scalar integral is √14∫₀¹∫₀²(x+y)dy dx. The inner integral is 2x+2, and the outer integral is [x²+2x]₀¹=3. The answer is 3√14.
Question 8 — Flux through a parametrised surface
Let S be the upper hemisphere \(x^2+y^2+z^2=a^2\), z≥0, with outward orientation, and let F=(x,y,z). Use a geometric argument or a direct surface integral to find the flux through S. Then explain why the divergence theorem cannot be applied to this hemisphere alone without adding a boundary cap.
10 marksShow worked answer
On the sphere the outward unit normal is n=(x,y,z)/a, so F·n=a. The area of the upper hemisphere is 2πa², giving flux 2πa³. The divergence theorem requires a closed boundary surface. The hemisphere has an equatorial boundary circle, so it is not closed. Adding the flat disk z=0 closes the surface; on that disk F·n=0, so the total closed-surface flux remains 2πa³, consistent with div F=3 and the half-ball volume 2πa³/3.
Question 9 — Stokes' theorem with orientation
Let \(F=(-y,x,z)\) and let S be the part of the plane z=2−x−y in the first octant, oriented upward. Use Stokes' theorem to evaluate the circulation around its positively oriented triangular boundary. Identify the boundary orientation and calculate the surface integral of curl F.
14 marksShow worked answer
curl F=(0,0,2). The upward-oriented surface is the triangle with vertices (0,0,2), (2,0,0), (0,2,0). Its projection onto the xy-plane is the triangle x≥0,y≥0,x+y≤2, of area 2. For z=2−x−y, the upward vector area element is (−zₓ,−zᵧ,1)dA=(1,1,1)dA. Thus curl F·(1,1,1)=2, and the surface integral is 2×2=4. Stokes' theorem gives boundary circulation 4. Positive orientation is counterclockwise when viewed from above; the right-hand rule links it to the upward normal.
Question 10 — Divergence theorem on a sphere
For \(F=(x^3,y^3,z^3)\), calculate the outward flux through the sphere x²+y²+z²=a² by applying the divergence theorem. Show the reduction to a radial integral and give the exact result.
14 marksShow worked answer
div F=3x²+3y²+3z²=3r². The divergence theorem gives flux=∭V3r²dV. In spherical coordinates, dV=r²sinφ drdφdθ, with 0≤r≤a, 0≤φ≤π, 0≤θ≤2π. Hence flux=3(∫₀ᵃr⁴dr)(∫₀ᵖⁱsinφdφ)(∫₀²ᵖⁱdθ)=3(a⁵/5)(2)(2π)=12πa⁵/5.
Question 11 — Potential function and a closed-curve integral
Let \(F=(2xy+z^2,x^2+2y,2xz)\) on ℝ³. (a) Find a potential function. (b) Evaluate the line integral along any piecewise smooth path from A=(0,1,0) to B=(1,2,2). (c) State the value around any closed curve.
12 marksShow worked answer
Integrate φₓ=2xy+z² with respect to x: φ=x²y+xz²+g(y,z). Then φᵧ=x²+gᵧ must equal x²+2y, so g=y²+h(z). Finally φ_z=2xz+h′(z)=2xz, hence h is constant. A potential is φ=x²y+xz²+y². At B it is 1·2+1·4+4=10; at A it is 0+0+1=1. The line integral is 10−1=9, independent of the path. Around any closed curve it is zero.
Question 12 — Choosing and justifying an integral theorem
A continuously differentiable vector field F is defined on an open region containing a solid ball V and its boundary S. Explain how you would compute the outward flux through S if div F is substantially simpler than F. State the theorem, its geometric hypotheses, and one reason a direct surface integral might be less efficient. Then explain how the conclusion changes if the surface is only an open hemisphere.
10 marksShow worked answer
Since S is the closed, piecewise smooth boundary of V and F is C¹ on a neighbourhood of V, apply the divergence theorem: ∬SF·n dS=∭Vdiv F dV, with n outward. A direct surface integral may require a complicated parameterisation and algebra, whereas divergence may be simple and the volume region symmetric. An open hemisphere is not a closed boundary: its equatorial rim must be closed by a disk, and the flux through the disk must be included. Only after summing the hemisphere and cap fluxes can the divergence theorem be applied to the half-ball.
Chapter 6 — Linear Algebra and Matrix Theory
Undergraduate level · Vector spaces, linear transformations, matrices, determinants and spectral theory. This chapter develops linear algebra from the abstract structure of vector spaces to computational matrix methods and eigenvalue theory. Emphasis is placed on the relationship between algebraic definitions, geometric interpretation and reliable computation. Unless stated otherwise, vector spaces and matrices are over the real field ℝ; the corresponding complex results are analogous, with conjugate transpose used in inner-product settings.
6.1 Linear systems and the geometry of solutions
A linear system in n unknowns can be written Ax=b, where A∈ℝᵐˣⁿ, x∈ℝⁿ and b∈ℝᵐ. Each row represents a linear constraint. The solution set is either empty or an affine translate of the null space N(A)={x:Ax=0}. Thus if x₀ is one solution, every solution has the form x=x₀+v, where v∈N(A).
A system is consistent precisely when b belongs to the column space of A. It has a unique solution precisely when it is consistent and the null space is trivial. It has infinitely many solutions when it is consistent and at least one free variable remains. Row operations preserve the solution set when applied to the augmented matrix [A|b].
6.2 Gaussian elimination and row reduction
Elementary row operations are: interchange two rows; multiply a row by a nonzero scalar; and add a scalar multiple of one row to another. They produce row-equivalent matrices. Gaussian elimination reduces a matrix to row-echelon form; Gauss–Jordan elimination continues to reduced row-echelon form. Pivot columns identify leading variables, while non-pivot columns correspond to free variables.
For numerical work, partial pivoting selects a large-magnitude pivot in the current column to reduce amplification of round-off error. A matrix is in row-echelon form if nonzero rows precede zero rows, each leading entry lies to the right of the leading entry in the row above, and entries below each leading entry are zero. Reduced row-echelon form additionally has leading entries equal to 1 and zero entries elsewhere in each pivot column.
6.3 Matrix algebra and elementary matrices
For compatible matrices, addition is entrywise and multiplication is defined by (AB)ᵢⱼ=Σₖaᵢₖbₖⱼ. Matrix multiplication is associative and distributive but generally not commutative. The identity matrix I satisfies AI=IA=A. The transpose satisfies (AB)ᵀ=BᵀAᵀ. An elementary row operation is equivalent to left multiplication by an elementary matrix.
A square matrix A is invertible if there exists A⁻¹ with AA⁻¹=A⁻¹A=I. Equivalent conditions include det(A)≠0, rank(A)=n, trivial null space, and linear independence of its columns. When A is invertible, Ax=b has the unique solution x=A⁻¹b. In computation, solving by factorisation is normally preferable to explicitly forming A⁻¹.
6.4 Determinants and their properties
The determinant det(A) is a scalar associated with a square matrix. It is multilinear and alternating in the rows (or columns), and det(I)=1. It measures the signed volume-scaling factor of the associated linear transformation. A zero determinant means that the transformation collapses dimension and the matrix is singular.
Key properties are det(AB)=det(A)det(B), det(Aᵀ)=det(A), and det(A⁻¹)=1/det(A) when A is invertible. Interchanging two rows changes the sign; multiplying a row by c multiplies the determinant by c; adding a multiple of one row to another leaves it unchanged. For triangular matrices, the determinant is the product of the diagonal entries. Cofactor expansion is useful for theoretical calculations, while elimination is generally more efficient for larger matrices.
6.5 Vector spaces and subspaces
A vector space V over a field F is a set with vector addition and scalar multiplication satisfying closure, associativity, commutativity of addition, additive identity and inverses, and the distributive and scalar identity laws. Familiar examples include ℝⁿ, polynomial spaces Pₙ, spaces of continuous functions and matrix spaces.
A nonempty subset W⊆V is a subspace if it is closed under linear combinations: for all u,v∈W and α,β∈F, αu+βv∈W. Equivalently, it contains zero and is closed under addition and scalar multiplication. The solution set of a homogeneous linear system is a subspace; the solution set of a nonhomogeneous system is generally an affine set rather than a subspace.
6.6 Span, linear independence, bases and dimension
The span of vectors v₁,…,vₖ is the set of all linear combinations Σᵢcᵢvᵢ. The vectors are linearly independent if Σᵢcᵢvᵢ=0 implies every cᵢ=0. A basis is a linearly independent spanning set. Every finite-dimensional vector space has a basis, and all bases have the same number of elements; this number is the dimension.
In ℝⁿ, a set of n vectors is a basis if and only if its column matrix is invertible. A spanning set can be reduced to a basis by removing redundant vectors; an independent set can be extended to a basis. Coordinates relative to a basis are unique, and changing basis is represented by an invertible transition matrix.
6.7 Fundamental subspaces: row space, column space and null space
For A∈ℝᵐˣⁿ, the column space Col(A) is a subspace of ℝᵐ, the row space Row(A) is a subspace of ℝⁿ, and the null space N(A) is a subspace of ℝⁿ. The rank is dim Col(A), equivalently dim Row(A). The nullity is dim N(A).
Rank–nullity theorem. For a linear map T:V→W with V finite-dimensional, dim V=dim ker(T)+dim im(T). For an m×n matrix A, rank(A)+nullity(A)=n.
The pivot columns of the original matrix form a basis for its column space. The nonzero rows of a row-echelon form form a basis for the row space. A basis for the null space is obtained by expressing pivot variables in terms of free variables in Ax=0.
6.8 Linear transformations and matrix representations
A map T:V→W is linear if T(αu+βv)=αT(u)+βT(v). Its kernel ker(T)={v:T(v)=0} is a subspace of V, and its image im(T)={T(v):v∈V} is a subspace of W. T is injective exactly when its kernel is {0}; it is surjective exactly when its image is W.
Once bases are chosen in V and W, T is represented by a matrix. The columns of the representation are the coordinates of the images of the domain basis vectors. Composition corresponds to matrix multiplication, with the order reflecting the order of composition. A change of basis changes the matrix representation but not the underlying linear map.
6.9 Eigenvalues, eigenvectors and characteristic polynomials
For a square matrix A, a nonzero vector v is an eigenvector with eigenvalue λ if Av=λv. Thus (A−λI)v=0 has a nonzero solution precisely when det(A−λI)=0. The polynomial p_A(λ)=det(λI−A) is the characteristic polynomial. Its roots are the eigenvalues, counted with algebraic multiplicity.
The eigenspace E_λ=ker(A−λI) has dimension called the geometric multiplicity. Geometric multiplicity is at least 1 and at most algebraic multiplicity. Eigenvectors belonging to distinct eigenvalues are linearly independent. The trace is the sum of eigenvalues and the determinant is their product, counting algebraic multiplicity over ℂ.
6.10 Diagonalisation and matrix powers
An n×n matrix A is diagonalizable if A=PDP⁻¹ for an invertible P and diagonal D. The columns of P are a basis of eigenvectors and the diagonal entries of D are the corresponding eigenvalues. A matrix with n distinct eigenvalues is diagonalizable. More generally, it is diagonalizable if the direct sum of its eigenspaces has dimension n.
When A=PDP⁻¹, powers are efficiently computed as Aᵏ=PDᵏP⁻¹. Diagonalisation is useful for linear recurrences, repeated applications of a linear transformation, and systems of constant-coefficient differential equations. Not every matrix is diagonalizable; repeated eigenvalues require checking the dimensions of their eigenspaces.
6.11 Inner products, orthogonality and projections
An inner product on a real vector space is a symmetric, bilinear, positive-definite map ⟨u,v⟩. In ℝⁿ the standard inner product is u·v=Σᵢuᵢvᵢ. The induced norm is ||v||=√⟨v,v⟩. Vectors are orthogonal if their inner product is zero. The Cauchy–Schwarz inequality states |⟨u,v⟩|≤||u||||v||, with equality precisely when the vectors are linearly dependent.
For a nonzero vector u, the orthogonal projection of v onto span{u} is projᵤ(v)=(⟨v,u⟩/⟨u,u⟩)u. The residual v−projᵤ(v) is orthogonal to u. For an orthonormal set q₁,…,qₖ, projection onto its span is Σᵢ⟨v,qᵢ⟩qᵢ.
6.12 Gram–Schmidt orthogonalisation and QR factorisation
Given linearly independent vectors v₁,…,vₖ in an inner-product space, Gram–Schmidt constructs an orthonormal basis q₁,…,qₖ for the same span. Set u₁=v₁ and q₁=u₁/||u₁||. Recursively, uⱼ=vⱼ−Σᵢ<ⱼ⟨vⱼ,qᵢ⟩qᵢ and qⱼ=uⱼ/||uⱼ||. In exact arithmetic, each uⱼ is nonzero when the original vectors are independent.
For a matrix with independent columns, the process yields A=QR, where Q has orthonormal columns and R is upper triangular. QR factorisation supports least-squares solutions and stable numerical algorithms. In floating-point computation, modified Gram–Schmidt or Householder transformations are often preferred for improved numerical stability.
6.13 Symmetric matrices and the spectral theorem
Real spectral theorem. Every real symmetric matrix A=Aᵀ is orthogonally diagonalizable: A=QΛQᵀ, where Q is orthogonal and Λ is real diagonal. Its eigenvalues are real, and eigenvectors corresponding to distinct eigenvalues are orthogonal.
This result underpins principal component analysis, quadratic optimisation, vibration analysis and covariance modelling. For a symmetric matrix, the quadratic form xᵀAx becomes Σᵢλᵢyᵢ² after the orthogonal coordinate change x=Qy. The signs of the eigenvalues determine whether the form is positive definite, negative definite, semidefinite or indefinite.
6.14 Quadratic forms and definiteness
A quadratic form is q(x)=xᵀAx, where A may be taken symmetric because xᵀAx depends only on (A+Aᵀ)/2. It is positive definite if q(x)>0 for all x≠0, positive semidefinite if q(x)≥0, negative definite if q(x)<0 for all x≠0, and indefinite if it takes both positive and negative values.
For real symmetric A, positive definiteness is equivalent to all eigenvalues being positive. Sylvester's criterion states that A is positive definite if and only if all leading principal minors are positive. Definiteness determines local curvature in optimisation and stability in many physical systems.
6.15 Least-squares approximation
When Ax=b is inconsistent, a least-squares solution minimises ||Ax−b||₂². Differentiating the objective yields the normal equations AᵀAx=Aᵀb. If A has full column rank, AᵀA is invertible and the solution is unique: x̂=(AᵀA)⁻¹Aᵀb. Geometrically, A x̂ is the orthogonal projection of b onto Col(A), so the residual b−Ax̂ is orthogonal to every column of A.
For numerical work, QR factorisation or singular value decomposition is generally more stable than explicitly forming and solving the normal equations, because forming AᵀA can square the condition number.
6.16 Numerical conditioning and singular value decomposition
The condition number κ₂(A)=σmax/σmin for an invertible matrix, where σmax and σmin are its largest and smallest singular values. A large condition number means small perturbations in the data can cause large relative changes in the solution. The singular value decomposition is A=UΣVᵀ, where U and V are orthogonal and Σ has nonnegative singular values on its diagonal (or rectangular diagonal).
SVD reveals rank, supports stable least-squares computation, gives the Moore–Penrose pseudoinverse, and is central to low-rank approximation. Truncating small singular values can regularise noisy or ill-conditioned problems, but the truncation threshold should be justified by the data and numerical precision.
6.17 Worked university examples
Example 1 — Null space. For A=[[1,2,−1],[2,4,−2]], the second row is twice the first. Ax=0 reduces to x+2y−z=0. Let y=s and z=t; then x=−2s+t. A null-space basis is {(-2,1,0),(1,0,1)} and the nullity is 2. Since there are 3 columns, rank–nullity gives rank 1.
Example 2 — Eigenvalues. For A=[[4,1],[2,3]], det(λI−A)=(λ−4)(λ−3)−2=λ²−7λ+10=(λ−5)(λ−2). Eigenvalues are 5 and 2. Corresponding eigenvectors can be chosen as (1,1) and (1,−2), respectively, so A is diagonalizable.
Example 3 — Orthogonal projection. Project v=(2,1) onto u=(1,1). The projection is ((v·u)/(u·u))u=(3/2)(1,1)=(3/2,3/2). The residual (1/2,−1/2) has zero dot product with u.
Chapter 6 — Undergraduate assessment problems
Give complete working. State the relevant vector-space or matrix theorem when used. For eigenvalue questions, show the characteristic polynomial and verify eigenvectors; for numerical or least-squares questions, justify uniqueness and rank assumptions.
Question 1 — Subspaces, basis and dimension
Let W={(x,y,z)∈ℝ³:x+2y−z=0}. (a) Prove that W is a subspace of ℝ³. (b) Find a basis and its dimension. (c) Determine whether (1,0,1) and (−2,1,0) form a basis for W.
10 marksShow worked answer
(a) Define T:ℝ³→ℝ by T(x,y,z)=x+2y−z. T is linear and W=ker T, so W is a subspace. (b) The constraint gives x=−2y+z. Hence (x,y,z)=y(−2,1,0)+z(1,0,1). These two vectors are independent, so they form a basis and dim W=2. (c) The proposed vectors are precisely these two basis vectors, so yes, they form a basis.
Question 2 — Linear independence and span
In ℝ³ let v₁=(1,0,1), v₂=(0,1,1), v₃=(1,1,2). (a) Determine whether the set is linearly independent. (b) Find a basis for its span. (c) State the dimension of the span and give one nonzero vector in its orthogonal complement.
10 marksShow worked answer
Since v₃=v₁+v₂, the set is linearly dependent. The vectors v₁ and v₂ are independent, so {v₁,v₂} is a basis for the span and its dimension is 2. A vector orthogonal to both solves a+b=0 and b+c=0. Taking (a,b,c)=(1,1,−1) gives a nonzero orthogonal vector.
Question 3 — Gaussian elimination and consistency
Solve the system x+y+z=6, 2x−y+z=3, x+2y−z=2 by Gaussian elimination. Display the row operations or an equivalent echelon reduction, and state whether the solution is unique.
12 marksShow worked answer
The augmented matrix is [[1,1,1|6],[2,−1,1|3],[1,2,−1|2]]. Apply R₂←R₂−2R₁ and R₃←R₃−R₁ to obtain [[1,1,1|6],[0,−3,−1|−9],[0,1,−2|−4]]. Interchange rows 2 and 3, then R₃←R₃+3R₂ to obtain [[1,1,1|6],[0,1,−2|−4],[0,0,−7|−21]]. Thus z=3, y=2, x=1. There is a pivot in every variable column, so the solution is unique.
Question 4 — Determinant and invertibility
For A=[[1,2,0],[2,1,1],[0,1,3]], calculate det(A). Determine whether A is invertible and explain what the determinant says about the associated linear transformation.
10 marksShow worked answer
Expanding along the first row, det(A)=1(3−1)−2(6−0)+0=2−12=−10. Since det(A)≠0, A is invertible. The associated transformation is one-to-one and onto ℝ³ and scales oriented volume by a factor of −10; the negative sign indicates orientation reversal.
Question 5 — Rank, nullity and the fundamental theorem
Let T:ℝ³→ℝ² be T(x,y,z)=(x+y,y+z). (a) Find ker(T). (b) Find a basis for im(T). (c) Verify rank–nullity and determine whether T is injective or surjective.
12 marksShow worked answer
For the kernel, x+y=0 and y+z=0, so (x,y,z)=y(−1,1,−1). Thus ker(T)=span{(−1,1,−1)} and nullity=1. The matrix is [[1,1,0],[0,1,1]]. Its columns (1,0) and (1,1) are independent, so they form a basis for ℝ²; rank=2. Therefore rank+nullity=2+1=3=dim ℝ³. T is not injective because its kernel is nontrivial, but it is surjective because its image is ℝ².
Question 6 — Eigenvalues and eigenvectors
For A=[[4,1],[2,3]], (a) calculate the characteristic polynomial and eigenvalues; (b) find a basis for each eigenspace; (c) construct P and D such that A=PDP⁻¹.
14 marksShow worked answer
det(λI−A)=(λ−4)(λ−3)−2=λ²−7λ+10=(λ−5)(λ−2), so λ=5,2. For λ=5, A−5I=[[-1,1],[2,−2]], giving x=y; take v₁=(1,1). For λ=2, A−2I=[[2,1],[2,1]], giving 2x+y=0; take v₂=(1,−2). Thus P=[[1,1],[1,−2]], D=diag(5,2), and A=PDP⁻¹. The distinct eigenvalues ensure diagonalizability.
Question 7 — Orthogonal diagonalisation
Let A=[[2,1],[1,2]]. Find its eigenvalues and an orthonormal eigenbasis. Write A=QΛQᵀ and verify the diagonalisation by multiplication or by its action on the eigenvectors.
12 marksShow worked answer
The characteristic polynomial is (λ−2)²−1=(λ−3)(λ−1), so eigenvalues are 3 and 1. Eigenvectors are (1,1) and (1,−1); normalising gives q₁=(1,1)/√2 and q₂=(1,−1)/√2. Set Q=(1/√2)[[1,1],[1,−1]] and Λ=diag(3,1). Since A is symmetric and its columns in Q are orthonormal eigenvectors, AQ=QΛ and QᵀQ=I, hence A=QΛQᵀ.
Question 8 — Gram–Schmidt process
Apply Gram–Schmidt to u₁=(1,1,0) and u₂=(1,0,1) in ℝ³ with the standard inner product. Obtain an orthonormal basis for their span and verify orthogonality.
12 marksShow worked answer
First q₁=u₁/||u₁||=(1,1,0)/√2. The projection coefficient is u₂·q₁=1/√2. Thus w₂=u₂−(1/√2)q₁=(1,0,1)−(1/2,1/2,0)=(1/2,−1/2,1). Its norm is √(1/4+1/4+1)=√(3/2)=√6/2. Therefore q₂=(1,−1,2)/√6. The dot product q₁·q₂=(1−1+0)/√12=0, and both vectors have unit norm.
Question 9 — Quadratic forms and definiteness
Consider q(x,y)=x²+4xy+y². (a) Write its symmetric matrix. (b) Determine its eigenvalues. (c) Classify the quadratic form and justify the classification by exhibiting vectors at which q takes opposite signs.
10 marksShow worked answer
The symmetric matrix is A=[[1,2],[2,1]], since xᵀAx=x²+4xy+y². Its characteristic polynomial is (1−λ)²−4=λ²−2λ−3=(λ−3)(λ+1), so eigenvalues are 3 and −1. The form is indefinite. For (1,1), q=6>0; for (1,−1), q=−2<0. Thus it is neither positive nor negative semidefinite.
Question 10 — Least-squares estimation
Fit the least-squares line y=α+βx to the three observations (0,1), (1,2), (2,2). (a) Form the design matrix and normal equations. (b) Find α and β. (c) Calculate the residual vector and verify its orthogonality to the columns of the design matrix.
14 marksShow worked answer
X=[[1,0],[1,1],[1,2]], b=(1,2,2)ᵀ. Then XᵀX=[[3,3],[3,5]] and Xᵀb=(5,6)ᵀ. The normal equations are 3α+3β=5 and 3α+5β=6. Subtracting gives 2β=1, so β=1/2 and α=7/6. The fitted values are (7/6,5/3,13/6), so residual b−Xθ=(−1/6,1/3,−1/6). Its entries sum to zero, and its dot product with (0,1,2) is 1/3−2/6=0. Hence Xᵀr=0, as required.
Question 11 — Inverse, composition and change of coordinates
Let S:ℝ²→ℝ² have matrix A=[[2,1],[1,1]] in the standard basis. (a) Find A⁻¹. (b) Solve S(x)=b for b=(5,3). (c) Explain why the coordinate vector of a fixed vector changes when a different basis is used, although the vector itself does not.
10 marksShow worked answer
det A=2·1−1·1=1, so A⁻¹=[[1,−1],[−1,2]]. The solution is x=A⁻¹b=(5−3,−5+6)=(2,1). A coordinate vector records coefficients relative to a chosen basis. Changing the basis changes those coefficients through an invertible transition matrix, but the underlying geometric vector and the linear map remain unchanged.
Question 12 — Singular value decomposition and rank
For A=[[3,0],[0,1],[0,0]], (a) determine the singular values; (b) give one valid SVD A=UΣVᵀ; (c) state the rank and condition number in the Euclidean 2-norm.
12 marksShow worked answer
AᵀA=diag(9,1), so its eigenvalues are 9 and 1 and the singular values are 3 and 1. One valid SVD is U=[[1,0,0],[0,1,0],[0,0,1]], Σ=[[3,0],[0,1],[0,0]], V=I₂. Direct multiplication gives A=UΣVᵀ. Both singular values are nonzero, so rank(A)=2. The 2-norm condition number is σmax/σmin=3.
Chapter 7 — Ordinary Differential Equations
Undergraduate level · First-order equations, linear theory, higher-order equations, transforms, systems and stability. An ordinary differential equation (ODE) relates an unknown function of one independent variable to one or more of its derivatives. ODEs model population dynamics, mechanical oscillations, electrical circuits, chemical kinetics, heat exchange and many other processes. This chapter develops analytical methods, existence and uniqueness principles, numerical approximations and qualitative analysis.
7.1 Definitions, order, degree and initial-value problems
An ODE has the general form F(t,y,y′,…,y⁽ⁿ⁾)=0. Its order is the highest derivative present. Its degree, when defined, is the power of the highest-order derivative after the equation has been made polynomial in derivatives. A solution is a sufficiently differentiable function satisfying the equation on an interval. A general solution typically contains arbitrary constants; initial or boundary conditions select particular solutions.
An nth-order equation can often be rewritten as a first-order system by setting x₁=y, x₂=y′,…,xₙ=y⁽ⁿ⁻¹⁾. An initial-value problem (IVP) specifies y(t₀), y′(t₀), …, y⁽ⁿ⁻¹⁾(t₀). A boundary-value problem specifies conditions at more than one point, for example y(a)=α and y(b)=β. An IVP and a boundary-value problem need not have the same existence or uniqueness properties.
7.2 Separable first-order equations
An equation y′=g(t)h(y) is separable wherever the division is valid. For h(y)≠0, write dy/h(y)=g(t)dt and integrate. Equilibrium solutions satisfying h(y)=0 must be checked separately because division by h(y) can remove them. An implicit relation may be the natural form of the solution; solving explicitly for y is not always possible or necessary.
Example. For y′=ty, separation gives dy/y=t dt, hence ln|y|=t²/2+C and y=Ce^(t²/2). The solution y=0 is included by C=0. If y(0)=2, then C=2.
7.3 First-order linear equations and integrating factors
A first-order linear equation has the standard form y′+p(t)y=q(t). An integrating factor is μ(t)=exp(∫p(t)dt). Multiplication by μ gives (μy)′=μq, so μ(t)y(t)=∫μ(t)q(t)dt+C. The method applies on intervals where p and q are continuous. The arbitrary constant is determined by an initial condition.
For y′+2y=e^(−t), choose μ=e^(2t). Then (e^(2t)y)′=e^t, giving y=e^(−t)+Ce^(−2t). This decomposition displays a particular response and a homogeneous transient.
7.4 Exact differential equations and integrating factors
An equation M(t,y)dt+N(t,y)dy=0 is exact on a suitable simply connected region if there is a potential Ψ(t,y) with Ψₜ=M and Ψᵧ=N. A necessary local condition, under continuous first partial derivatives, is Mᵧ=Nₜ. To solve, integrate M with respect to t, include an unknown function of y, and determine that function by matching Ψᵧ=N. The solution is Ψ(t,y)=C.
If Mᵧ≠Nₜ, the equation is not exact as written. An integrating factor depending only on t may exist when (Mᵧ−Nₜ)/N is a function of t alone; one depending only on y may exist when (Nₜ−Mᵧ)/M is a function of y alone. These tests are sufficient in the stated circumstances, not universal guarantees.
7.5 Homogeneous first-order equations and Bernoulli equations
A first-order homogeneous equation y′=F(y/t) can be reduced using y=vt, so y′=v+t v′. The resulting equation in v and t is often separable. The substitution must be used on intervals where t≠0, and solutions lost through division should be checked separately.
Bernoulli's equation y′+p(t)y=q(t)yⁿ, with n≠0,1, becomes linear after setting u=y^(1−n). Since u′=(1−n)y^(−n)y′, the transformed equation is u′+(1−n)p(t)u=(1−n)q(t). Solutions with y=0 may need separate consideration depending on n and the domain.
7.6 Existence and uniqueness for first-order IVPs
Picard–Lindelöf theorem (local form). For y′=f(t,y), y(t₀)=y₀, if f and ∂f/∂y are continuous on a rectangle containing (t₀,y₀), then there exists a unique local solution through that initial point.
Continuity of f alone is sufficient for at least one local solution under the Peano existence theorem, but does not guarantee uniqueness. A common uniqueness condition is local Lipschitz continuity in y. The theorem is local: it does not by itself establish that a solution exists for all time. Blow-up, singularities or leaving the domain may limit the interval of existence.
7.7 Autonomous equations and qualitative phase lines
An autonomous scalar equation has the form y′=f(y). Equilibria satisfy f(y*)=0. On intervals where f is positive, solutions increase; where it is negative, solutions decrease. An equilibrium is asymptotically stable if nearby solutions approach it as t increases, and unstable if nearby solutions move away. For a simple equilibrium, f′(y*)<0 implies local asymptotic stability, while f′(y*)>0 implies instability.
Qualitative analysis can reveal long-term behaviour without an explicit formula. Solutions of a well-posed autonomous scalar IVP cannot cross an equilibrium solution when uniqueness holds. This order-preserving property is useful in population and chemical concentration models.
7.8 Numerical methods: Euler and improved Euler
For y′=f(t,y), with step h and initial approximation yₙ at tₙ, forward Euler gives yₙ₊₁=yₙ+h f(tₙ,yₙ). It has local truncation error O(h²) and global error O(h) under standard smoothness assumptions. Smaller steps usually improve accuracy, but increase computational cost and do not automatically cure instability.
The explicit trapezoidal or Heun method first predicts y*=yₙ+h f(tₙ,yₙ), then corrects yₙ₊₁=yₙ+(h/2)[f(tₙ,yₙ)+f(tₙ₊₁,y*)]. It is a second-order method under suitable smoothness assumptions. Error estimates and step-size control are important in practical computation.
7.9 Fourth-order Runge–Kutta method
The classical RK4 method uses k₁=f(tₙ,yₙ), k₂=f(tₙ+h/2,yₙ+hk₁/2), k₃=f(tₙ+h/2,yₙ+hk₂/2), k₄=f(tₙ+h,yₙ+hk₃), and yₙ₊₁=yₙ+(h/6)(k₁+2k₂+2k₃+k₄). It has global error O(h⁴) for sufficiently smooth problems and suitable step sizes. RK4 is widely used, but stiff equations often require implicit or specialised solvers.
7.10 Higher-order linear equations and the superposition principle
A general nth-order linear equation is aₙ(t)y⁽ⁿ⁾+…+a₁(t)y′+a₀(t)y=g(t), with aₙ(t)≠0. The associated homogeneous equation has an n-dimensional solution space on an interval where the coefficients are continuous and the leading coefficient is nonzero. If y₁,…,yₙ form a fundamental set, the complementary solution is y_c=Σᵢcᵢyᵢ.
Superposition principle. For a linear homogeneous differential operator L, if Ly₁=0 and Ly₂=0, then L(c₁y₁+c₂y₂)=0. For a nonhomogeneous equation Ly=g, every solution has the form y=y_c+y_p, where y_p is one particular solution.
7.11 Linear equations with constant coefficients
For a homogeneous equation with constant coefficients, try y=e^(rt). The characteristic polynomial is obtained by replacing each derivative by the corresponding power of r. Distinct real roots rᵢ yield terms e^(rᵢt). A real repeated root r of multiplicity m yields e^(rt), te^(rt),…,t^(m−1)e^(rt). A complex pair α±iβ yields real solutions e^(αt)cos(βt) and e^(αt)sin(βt), with additional powers of t for repeated roots.
For example, y″−3y′+2y=0 has characteristic equation (r−1)(r−2)=0, so y=C₁eᵗ+C₂e²ᵗ. The initial conditions determine C₁ and C₂.
For constant-coefficient linear equations, undetermined coefficients construct a trial particular solution based on the forcing term. Polynomial, exponential, sine and cosine forcing terms, and their finite sums, are suitable. If a trial term overlaps with a complementary solution, multiply the trial by a sufficient power of t to obtain linear independence. The method is not generally applicable to arbitrary forcing functions or variable coefficients.
For y″−3y′+2y=e^(3t), try yₚ=Ae^(3t). Substitution gives (9−9+2)Ae^(3t)=e^(3t), hence A=1/2. The complete solution is y=C₁eᵗ+C₂e²ᵗ+(1/2)e^(3t).
7.13 Variation of parameters
For y″+p(t)y′+q(t)y=g(t), let y₁,y₂ be a fundamental pair and W=y₁y₂′−y₁′y₂ their Wronskian. A particular solution is yₚ=−y₁∫(y₂g/W)dt+y₂∫(y₁g/W)dt. The formula is valid on intervals where W≠0 and the coefficients and forcing term have the required continuity. It is more general than undetermined coefficients, though integrals may be difficult or non-elementary.
7.14 Cauchy–Euler equations
A Cauchy–Euler equation has coefficients arranged as t²y″+a t y′+b y=0. On an interval not containing t=0, try y=tʳ. This gives the indicial equation r(r−1)+ar+b=0. Distinct roots yield tʳ¹ and tʳ²; a repeated root r yields tʳ and tʳln|t|. Complex roots α±iβ yield tᵅcos(βln|t|) and tᵅsin(βln|t|). The interval and sign of t must be handled consistently.
7.15 Laplace transforms and initial conditions
The Laplace transform is ℒ{f(t)}=F(s)=∫₀∞e^(−st)f(t)dt where the integral converges. Important formulas include ℒ{1}=1/s, ℒ{e^(at)}=1/(s−a), ℒ{sin bt}=b/(s²+b²), ℒ{cos bt}=s/(s²+b²), and ℒ{f′}=sF(s)−f(0). Also ℒ{f″}=s²F(s)−sf(0)−f′(0).
Transforming a linear IVP converts derivatives into algebraic expressions in s together with the initial data. Solve for F(s), decompose into standard partial fractions, and invert. The method is particularly useful for discontinuous or piecewise forcing, impulses and circuit switching. Existence of the transform requires suitable growth and piecewise continuity conditions.
7.16 Systems of first-order linear ODEs
A linear system is x′=A(t)x+g(t). For constant A and g=0, solutions are governed by the matrix exponential e^(At)=Σₖ₌₀∞(At)ᵏ/k!. If A is diagonalizable, A=PDP⁻¹ and e^(At)=Pe^(Dt)P⁻¹, where e^(Dt) is diagonal with entries e^(λᵢt). Generalised eigenvectors handle defective matrices. For nonzero forcing, variation of constants gives x(t)=e^(A(t−t₀))x(t₀)+∫ₜ₀ᵗe^(A(t−s))g(s)ds.
For a two-dimensional autonomous system x′=Ax, eigenvalues determine the local phase portrait: negative real parts indicate asymptotic decay, any positive real part indicates instability, and purely imaginary eigenvalues require additional analysis in nonlinear systems. Repeated or defective eigenvalues can produce polynomial factors multiplying exponentials.
7.17 Modelling, dimensional consistency and interpretation
Constructing an ODE model requires defining state variables, independent variables, parameters, assumptions and initial data. Dimensional analysis checks that every term in an equation has compatible units. A model should be interpreted within its assumptions: constant coefficients, unlimited resources, linear restoring forces or instantaneous mixing may be approximations rather than universal truths.
Examples include exponential growth y′=ky, logistic growth y′=ry(1−y/K), Newton cooling T′=−k(T−Tₐ), and damped oscillation my″+cy′+ky=0. Parameter signs and magnitudes determine whether a model grows, decays, oscillates or approaches equilibrium. Model validation requires comparing predictions with observations and checking sensitivity to parameters and initial conditions.
7.18 Worked university examples
Example 1 — First-order linear IVP. Solve y′+2y=4, y(0)=1. The integrating factor is e^(2t), so (e^(2t)y)′=4e^(2t). Integration gives y=2+Ce^(−2t). The initial condition gives C=−1; hence y=2−e^(−2t).
Example 2 — Repeated characteristic root. Solve y″−4y′+4y=0. The characteristic polynomial is (r−2)², so y=(C₁+C₂t)e^(2t). The factor t is required because the root is repeated.
Example 3 — Laplace transform. Solve y′+y=1, y(0)=0. Taking transforms gives sY(s)+Y(s)=1/s, so Y=1/[s(s+1)]=1/s−1/(s+1). Therefore y(t)=1−e^(−t), t≥0.
Example 4 — Logistic equilibrium. For y′=ry(1−y/K), with r,K>0, equilibria are 0 and K. Since f′(y)=r(1−2y/K), f′(0)=r>0 and f′(K)=−r<0. Thus zero is unstable and carrying capacity K is locally asymptotically stable for positive populations.
Chapter 7 — Undergraduate assessment problems
Show all intermediate steps, identify the method used and state any interval or parameter restrictions. In numerical problems, retain sufficient precision and distinguish an approximation from an exact solution.
Question 1 — Separable IVP and equilibrium solution
Solve \(y'=t(1+y^2)\), subject to y(0)=1. State the maximal interval containing t=0 on which the solution is finite. Explain why separating variables is valid for this initial condition.
10 marksShow worked answer
Separate: dy/(1+y²)=t dt. Integrating gives arctan y=t²/2+C. Since y(0)=1, C=π/4. Thus y=tan(t²/2+π/4). The first positive singularity occurs when t²/2+π/4=π/2, so t²=π/2 and t=√(π/2). By symmetry the negative endpoint is −√(π/2). The maximal interval containing zero is (−√(π/2),√(π/2)). The denominator 1+y² is strictly positive for all real y, so division is valid throughout the real-valued solution.
Question 2 — Integrating factor and long-term behaviour
Solve \(y'-3y=6e^{2t}\), y(0)=2. Express the answer as a sum of homogeneous and particular components and determine its behaviour as t→∞.
10 marksShow worked answer
The integrating factor is μ=e^(−3t). Then (e^(−3t)y)′=6e^(−t). Integrating gives e^(−3t)y=−6e^(−t)+C, hence y=−6e^(2t)+Ce^(3t). The initial condition gives 2=−6+C, so C=8. Therefore y=8e^(3t)−6e^(2t). The homogeneous component is 8e^(3t), and a particular solution is −6e^(2t). The solution grows without bound like 8e^(3t).
Question 3 — Exact differential equation
Solve \((2xy+3)dx+(x^2+4y)dy=0\). Verify exactness, obtain an implicit solution and determine the solution curve through (1,0).
10 marksShow worked answer
Let M=2xy+3 and N=x²+4y. Mᵧ=2x=Nₓ, so the equation is exact. Integrating M with respect to x gives Ψ=x²y+3x+g(y). Differentiating with respect to y gives Ψᵧ=x²+g′(y)=x²+4y, so g(y)=2y². Thus x²y+3x+2y²=C. At (1,0), C=3, so the required curve is x²y+3x+2y²=3.
Question 4 — Bernoulli equation
Solve \(y'+y=ty^2\) for nonzero y. Use a substitution to obtain a linear equation and state whether y=0 is also a solution of the original equation.
12 marksShow worked answer
This is Bernoulli with n=2. Put u=y^(−1), so u′=−y^(−2)y′. Dividing the original equation by y² gives y′/y²+1/y=t, hence −u′+u=t, or u′−u=−t. The integrating factor is e^(−t), giving (e^(−t)u)′=−te^(−t). Integration yields e^(−t)u=(t+1)e^(−t)+C, so u=t+1+Ceᵗ. Therefore y=1/(t+1+Ceᵗ), wherever the denominator is nonzero. The original equation also admits y=0, which was excluded by the substitution.
Question 5 — Existence and uniqueness
Consider the IVP \(y'=\sqrt{|y|}\), y(0)=0. Explain why the standard local uniqueness condition is not satisfied at (0,0), and exhibit two distinct solutions satisfying the same initial condition.
10 marksShow worked answer
Here f(y)=√|y| is continuous, but its derivative with respect to y is unbounded near zero and f is not locally Lipschitz at y=0. One solution is y(t)=0. Another is y(t)=t²/4 for t≥0, because y′=t/2 and √|y|=t/2; it also satisfies y(0)=0. More generally, solutions can remain zero until an arbitrary waiting time and then follow a translated quadratic branch. Thus uniqueness fails even though existence holds.
Question 6 — Euler numerical approximation and error
Apply forward Euler with step h=0.1 to \(y'=y-t^2+1\), y(0)=0.5, to estimate y(0.2). Show both steps. The exact solution is \(y=(t+1)^2-0.5e^t\); use it to calculate the absolute error at t=0.2 to at least four significant figures.
12 marksShow worked answer
f(t,y)=y−t²+1. At t₀=0,y₀=0.5, f=1.5, so y₁=0.5+0.1(1.5)=0.65. At t₁=0.1,y₁=0.65, f=0.65−0.01+1=1.64, so y₂=0.65+0.1(1.64)=0.814. The exact value is y(0.2)=1.2²−0.5e^0.2≈1.44−0.610701379=0.829298621. The absolute error is approximately 0.015298621.
Question 7 — Constant-coefficient equation with complex roots
Solve \(y''+4y'+13y=0\), y(0)=1, y′(0)=0. Explain the role of the real and imaginary parts of the characteristic roots.
12 marksShow worked answer
The characteristic equation is r²+4r+13=0, giving r=−2±3i. Thus y=e^(−2t)(C₁cos3t+C₂sin3t). The first condition gives C₁=1. Differentiating and evaluating at zero gives y′(0)=−2C₁+3C₂=0, so C₂=2/3. Hence y=e^(−2t)(cos3t+(2/3)sin3t). The negative real part produces an exponentially decaying envelope, while the imaginary part produces oscillation of angular frequency 3.
Question 8 — Repeated roots and initial data
Solve \(y'''-3y''+3y'-y=0\) subject to y(0)=2, y′(0)=0 and y″(0)=1. Show how the multiplicity of the characteristic root determines the form of the solution.
14 marksShow worked answer
The characteristic polynomial is (r−1)³, so y=eᵗ(C₁+C₂t+C₃t²). From y(0)=2, C₁=2. Differentiating gives y′=eᵗ(C₁+C₂+ (C₂+2C₃)t+C₃t²), hence y′(0)=C₁+C₂=0 and C₂=−2. A second derivative gives y″(0)=C₁+2C₂+2C₃=1. Thus 2−4+2C₃=1, so C₃=3/2. The solution is y=eᵗ(2−2t+(3/2)t²). The repeated root of multiplicity three requires the independent factors 1,t,t² multiplying eᵗ.
Question 9 — Undetermined coefficients and resonance
Solve \(y''-2y'+y=e^t\). Explain why a trial solution proportional to eᵗ fails and construct a valid particular solution.
12 marksShow worked answer
The characteristic polynomial is (r−1)², so y_c=(C₁+C₂t)eᵗ. A trial Aeᵗ duplicates a homogeneous solution and is annihilated by the differential operator. Because the root has multiplicity two, use yₚ=At²eᵗ. Applying (D−1)² to At²eᵗ gives 2Aeᵗ, so 2A=1 and A=1/2. The general solution is y=eᵗ(C₁+C₂t+t²/2).
Question 10 — Laplace transform with initial conditions
Use Laplace transforms to solve \(y''+3y'+2y=0\), y(0)=1, y′(0)=0. Show the algebraic equation for Y(s), perform partial fractions and invert.
12 marksShow worked answer
Taking transforms gives [s²Y−s]+3[sY−1]+2Y=0. Thus (s²+3s+2)Y=s+3, so Y=(s+3)/[(s+1)(s+2)]. Partial fractions give Y=2/(s+1)−1/(s+2). Therefore y(t)=2e^(−t)−e^(−2t). It satisfies y(0)=1 and y′(0)=−2+2=0.
Question 11 — Coupled linear system and eigenvalues
Solve the system \(x'=3x+y,\;y'=x+3y\), with x(0)=2 and y(0)=0. Use eigenvalues and eigenvectors of the coefficient matrix and interpret the resulting modes.
14 marksShow worked answer
A=[[3,1],[1,3]]. Its eigenvalues are 4 and 2, with eigenvectors (1,1) and (1,−1), respectively. Decompose (2,0)=(1,1)+(1,−1). Therefore (x,y)=e^(4t)(1,1)+e^(2t)(1,−1). Hence x=e^(4t)+e^(2t), y=e^(4t)−e^(2t). The two independent eigenmodes grow at rates 4 and 2; the faster mode dominates as t increases.
Question 12 — Damped oscillator and classification
Consider \(m y''+c y'+k y=0\), where m,c,k>0. (a) Derive the characteristic equation. (b) State the conditions for underdamping, critical damping and overdamping. (c) For m=1, c=4, k=3, solve the IVP y(0)=1, y′(0)=0 and classify the motion.
14 marksShow worked answer
The characteristic equation is mr²+cr+k=0, with discriminant Δ=c²−4mk. Underdamping occurs for Δ<0, critical damping for Δ=0, and overdamping for Δ>0. For m=1,c=4,k=3, the polynomial is r²+4r+3=(r+1)(r+3), so y=C₁e^(−t)+C₂e^(−3t). The initial conditions give C₁+C₂=1 and −C₁−3C₂=0. Solving gives C₁=3/2 and C₂=−1/2. Thus y=(3/2)e^(−t)−(1/2)e^(−3t). Since Δ=16−12=4>0, the motion is overdamped: it returns to equilibrium without oscillation.
Classification, first-order PDEs, separation of variables, Fourier methods and classical boundary-value problems. A partial differential equation (PDE) relates an unknown function of two or more independent variables to its partial derivatives. PDEs describe diffusion, waves, electrostatics, fluid flow and quantum systems. The central tasks are classification, selecting suitable coordinates, constructing solutions and interpreting initial and boundary data.
Learning outcomes
By the end of this chapter, students should be able to define the central objects precisely, state and apply key theorems with their hypotheses, construct rigorous derivations, solve unfamiliar problems and interpret mathematical results in context.
8.1 Definitions, order and linearity
For u=u(x₁,…,xₙ), a PDE is F(x,u,Du,D²u,…)=0. Its order is the highest derivative order. A linear PDE is linear in u and all its derivatives; coefficients may depend on independent variables. A homogeneous linear PDE has zero forcing term. A quasilinear PDE is linear in its highest-order derivatives but may be nonlinear in lower-order terms.
8.2 Classification of second-order equations
For A uₓₓ+2B uₓᵧ+C uᵧᵧ+ lower-order terms = f, classify using Δ=B²−AC: hyperbolic if Δ>0, parabolic if Δ=0, elliptic if Δ<0. Examples: wave equation uₜₜ=c²uₓₓ is hyperbolic; heat equation uₜ=κuₓₓ is parabolic; Laplace equation uₓₓ+uᵧᵧ=0 is elliptic. Classification is local when coefficients vary.
8.3 First-order PDEs and characteristics
For a uₓ+b uᵧ=c, characteristic curves satisfy dx/ds=a, dy/ds=b, du/ds=c. Along these curves the PDE becomes an ODE. For uₓ+uᵧ=0, characteristics have y−x constant and u is constant along each line, so u=F(y−x). Initial data must be prescribed on a non-characteristic curve for a locally determined solution.
8.4 Separation of variables
For a product ansatz u(x,t)=X(x)T(t), substitution into a linear PDE often separates independent-variable expressions and sets each equal to a constant. Boundary conditions determine admissible eigenvalues and eigenfunctions; initial data determine expansion coefficients. Separation is a construction method, not a claim that every solution is a single product.
8.5 Fourier series and eigenfunction expansions
On (0,L), the sine series of f is Σ bₙ sin(nπx/L), with bₙ=(2/L)∫₀ᴸ f(x)sin(nπx/L)dx. Sine functions satisfy homogeneous Dirichlet conditions. Cosine functions naturally satisfy homogeneous Neumann conditions. Orthogonality makes coefficient extraction possible; convergence at discontinuities is interpreted using one-sided limits.
8.6 Heat equation
For uₜ=κuₓₓ on 0
8.7 Wave equation
For uₜₜ=c²uₓₓ on a fixed string with zero endpoint displacement, modes have angular frequencies ωₙ=cnπ/L. A general finite-energy solution is represented by a Fourier sine series with time factors cos(ωₙt) and sin(ωₙt). Initial displacement and velocity determine the two sets of coefficients.
8.8 Laplace and Poisson equations
Laplace's equation Δu=0 describes source-free steady potentials; Poisson's equation Δu=g includes a source. Harmonic functions satisfy the mean-value property and, under standard assumptions, attain extrema on the boundary. Uniqueness for the Dirichlet problem follows from the maximum principle.
8.9 Boundary and initial conditions
Dirichlet data prescribe u; Neumann data prescribe a normal derivative; Robin data prescribe a linear combination. Parabolic problems typically require initial data in time and boundary data in space; hyperbolic problems require initial displacement and velocity; elliptic problems require boundary data. Compatibility and regularity affect classical solvability.
8.10 Weak solutions and energy viewpoint
When classical derivatives do not exist, integrate against smooth test functions and transfer derivatives by integration by parts. This leads to weak formulations and Sobolev spaces. Energy estimates bound solution norms and are central to uniqueness and stability.
8.11 University assessment — original problems
Attempt each problem before revealing its worked solution. Marks reward mathematical reasoning, correct notation, justified steps and interpretation.
Question 8.1 — Classify 3uₓₓ+4uₓᵧ+uᵧᵧ=0.
Classify 3uₓₓ+4uₓᵧ+uᵧᵧ=0.
8 marksShow worked answer
Here A=3, 2B=4 so B=2, and C=1. Δ=B²−AC=4−3=1>0. The equation is hyperbolic.
Question 8.2 — Solve uₓ+2uᵧ=0.
Solve uₓ+2uᵧ=0.
10 marksShow worked answer
Characteristics satisfy dx/ds=1, dy/ds=2, du/ds=0. Thus y−2x is constant and u=F(y−2x), for arbitrary differentiable F.
Question 8.3 — Find the separated heat modes on 0
Find the separated heat modes on 08 marksShow worked answer
Set u=XT. XT′=κX″T gives T′/(κT)=X″/X=−λ. Dirichlet conditions give λₙ=(nπ/L)² and Xₙ=sin(nπx/L). Tₙ=e^(−κλₙt).
Question 8.4 — Expand f(x)=x on (0,L) in a sine series.
Expand f(x)=x on (0,L) in a sine series.
10 marksShow worked answer
bₙ=(2/L)∫₀ᴸx sin(nπx/L)dx=2L(−1)^(n+1)/(nπ). Hence x~(2L/π)Σ(−1)^(n+1)sin(nπx/L)/n for 0
Question 8.5 — For uₜ=4uₓₓ on (0,π) with zero endpoints and u(x,0)=sin(3x), find u.
For uₜ=4uₓₓ on (0,π) with zero endpoints and u(x,0)=sin(3x), find u.
8 marksShow worked answer
The nth mode decays as e^(−4n²t). Only n=3 is present, so u(x,t)=e^(−36t)sin(3x).
Question 8.6 — Explain why Laplace's equation with zero Dirichlet boundary data has only the zero solution on a bounded domain.
Explain why Laplace's equation with zero Dirichlet boundary data has only the zero solution on a bounded domain.
10 marksShow worked answer
If Δu=0 and u=0 on the boundary, the maximum principle implies the maximum and minimum occur on the boundary. Both are zero; therefore u≡0.
Undergraduate · Chapter 9
Chapter 9 — Real Analysis
Sequences, limits, continuity, differentiation, integration, metric spaces and rigorous convergence. Real analysis supplies the rigorous foundations of calculus. It studies completeness of the real numbers, convergence, continuity, differentiation, integration and function spaces. Proofs depend on precise quantifiers and estimates rather than graphical intuition.
Learning outcomes
By the end of this chapter, students should be able to define the central objects precisely, state and apply key theorems with their hypotheses, construct rigorous derivations, solve unfamiliar problems and interpret mathematical results in context.
9.1 Ordered fields and completeness
The real numbers form a complete ordered field. Completeness may be expressed as the least-upper-bound property: every nonempty set bounded above has a supremum. The Archimedean property states that for every real x there is n∈N with n>x. Completeness distinguishes ℝ from ℚ.
9.2 Sequences and convergence
A sequence (aₙ) converges to L if for every ε>0 there exists N such that n≥N implies |aₙ−L|<ε. Limits are unique. Every convergent sequence is bounded. The algebraic limit laws hold when the relevant limits exist.
9.3 Cauchy sequences and subsequences
A sequence is Cauchy if for every ε>0 there is N such that m,n≥N implies |aₘ−aₙ|<ε. ℝ is complete: every Cauchy sequence converges. Bolzano–Weierstrass states every bounded real sequence has a convergent subsequence.
9.4 Series and convergence
Σaₙ converges when its partial sums converge. The Cauchy criterion says tails of partial sums become arbitrarily small. Absolute convergence implies convergence. Comparison, limit comparison, ratio, root and alternating-series tests are tools whose hypotheses must be checked.
9.5 Limits and continuity
The ε–δ definition: limₓ→a f(x)=L if for each ε>0 there is δ>0 such that 0<|x−a|<δ implies |f(x)−L|<ε. Continuity at a includes f(a)=L. Sequential continuity is equivalent in metric spaces.
9.6 Compactness and uniform continuity
In ℝ, a set is compact exactly when it is closed and bounded. Continuous functions on compact sets attain extrema and are uniformly continuous. Uniform continuity chooses δ independently of the point x; ordinary continuity may choose δ depending on x.
9.7 Differentiation and the mean value theorem
Differentiability at a means f(a+h)=f(a)+Lh+o(h), where L=f′(a). Differentiability implies continuity. Rolle's theorem and the mean value theorem connect local derivatives to global change: f(b)−f(a)=f′(c)(b−a).
9.8 Riemann integration
A bounded function on [a,b] is Riemann integrable when its upper and lower Darboux integrals agree. Every continuous function on a compact interval is integrable. The fundamental theorem links differentiation and integration.
9.9 Metric spaces and completeness
A metric d satisfies non-negativity, identity, symmetry and the triangle inequality. Open balls define topology. A metric space is complete if every Cauchy sequence converges in the space. Closed subsets of complete metric spaces are complete.
9.10 Pointwise and uniform convergence
fₙ→f uniformly if for every ε>0 there is N such that n≥N implies |fₙ(x)−f(x)|<ε for every x in the domain. Uniform limits of continuous functions are continuous. Uniform convergence permits interchange with integration under standard hypotheses; differentiation needs stronger conditions.
9.11 University assessment — original problems
Attempt each problem before revealing its worked solution. Marks reward mathematical reasoning, correct notation, justified steps and interpretation.
Question 9.1 — Prove that a convergent sequence has a unique limit.
Prove that a convergent sequence has a unique limit.
8 marksShow worked answer
Suppose aₙ→L and aₙ→M. If L≠M, let ε=|L−M|/3. For sufficiently large n, |aₙ−L|<ε and |aₙ−M|<ε. Then |L−M|≤|L−aₙ|+|aₙ−M|<2ε<|L−M|, contradiction. Thus L=M.
Question 9.2 — Show that aₙ=1/n converges to zero.
Show that aₙ=1/n converges to zero.
10 marksShow worked answer
Given ε>0 choose N>1/ε. For n≥N, |1/n−0|=1/n≤1/N<ε. Hence aₙ→0.
The p-series test gives convergence for p=2>1. Its value is π²/6, though convergence follows without evaluating the sum.
Question 9.4 — Prove f(x)=x² is uniformly continuous on [0,2].
Prove f(x)=x² is uniformly continuous on [0,2].
10 marksShow worked answer
For x,y∈[0,2], |x²−y²|=|x−y||x+y|≤4|x−y|. Given ε choose δ=ε/4. Then |x−y|<δ implies |f(x)−f(y)|<ε.
Question 9.5 — Does fₙ(x)=xⁿ converge uniformly to zero on [0,1]?
Does fₙ(x)=xⁿ converge uniformly to zero on [0,1]?
8 marksShow worked answer
Pointwise, limit is 0 for 0≤x<1 and 1 at x=1. Thus it does not converge pointwise to zero on the whole interval. Even on [0,1), convergence to zero is not uniform because supₓ∈[0,1) xⁿ=1 for every n.
Question 9.6 — Show that every Cauchy sequence in ℝ is bounded.
Show that every Cauchy sequence in ℝ is bounded.
10 marksShow worked answer
Choose N such that m,n≥N implies |aₘ−aₙ|<1. For n≥N, |aₙ|≤|a_N|+1. The finitely many terms before N are bounded; taking the maximum gives a bound for the whole sequence.
Undergraduate · Chapter 10
Chapter 10 — Complex Analysis
Complex differentiation, analytic functions, contour integration, residues and conformal maps. Complex analysis studies functions of a complex variable. Complex differentiability is substantially more restrictive than real differentiability and yields powerful integral theorems, series representations and residue techniques.
Learning outcomes
By the end of this chapter, students should be able to define the central objects precisely, state and apply key theorems with their hypotheses, construct rigorous derivations, solve unfamiliar problems and interpret mathematical results in context.
10.1 Complex numbers and geometry
Write z=x+iy=re^(iθ), with modulus |z|=√(x²+y²) and argument θ modulo 2π. The conjugate is z̄=x−iy. The identities z z̄=|z|² and e^(iθ)=cosθ+i sinθ support polar calculations and roots.
10.2 Complex limits and continuity
A limit f(z)→L as z→z₀ must be independent of the path of approach in the plane. To disprove a proposed limit, two paths giving different limits suffice. Continuity is defined using the complex modulus.
10.3 Complex differentiability and Cauchy–Riemann equations
For f=u+iv, complex differentiability implies uₓ=vᵧ and uᵧ=−vₓ. If first partial derivatives are continuous near a point and satisfy these equations, f is analytic there. The equations alone at a single point are not sufficient.
10.4 Analytic functions and harmonic functions
If f=u+iv is analytic with suitable second derivatives, u and v are harmonic: uₓₓ+uᵧᵧ=0 and vₓₓ+vᵧᵧ=0. They are harmonic conjugates. The complex derivative is f′=uₓ+ivₓ=vᵧ−iuᵧ.
10.5 Elementary complex functions
The exponential eᶻ=eˣ(cos y+i sin y) is entire and 2πi-periodic. Complex logarithm is multivalued; a branch requires a chosen domain and argument interval. Powers and roots inherit branch choices.
10.6 Contours and complex integration
For a parametrised contour z(t), ∫γ f(z)dz=∫ f(z(t))z′(t)dt. Integrals depend on orientation. A positively oriented simple closed curve has its interior on the left as it is traversed.
10.7 Cauchy's theorem and integral formula
If f is analytic on a simply connected domain, its integral around a closed contour is zero. Cauchy's integral formula gives f(a)=(1/2πi)∮ f(z)/(z−a) dz and derivative formulas for higher derivatives.
10.8 Taylor and Laurent series
Analytic functions admit Taylor expansions within the nearest singularity. Around an isolated singularity, a Laurent series includes positive and negative powers. The principal part determines the singularity type.
10.9 Singularities and residues
An isolated singularity is removable, a pole, or essential. The residue is the coefficient of (z−a)⁻¹. For a simple pole, Res(f,a)=lim(z→a)(z−a)f(z). The residue theorem converts contour integrals into 2πi times the sum of enclosed residues.
10.10 Conformal mappings
An analytic map with nonzero derivative is locally angle-preserving. Möbius maps (az+b)/(cz+d), ad−bc≠0, map generalized circles to generalized circles. Mapping domains can simplify boundary-value problems.
10.11 University assessment — original problems
Attempt each problem before revealing its worked solution. Marks reward mathematical reasoning, correct notation, justified steps and interpretation.
Question 10.1 — Solve z²=−4.
Solve z²=−4.
8 marksShow worked answer
Write −4=4e^(i(π+2kπ)). The square roots have modulus 2 and arguments π/2 and 3π/2, giving z=2i and z=−2i.
Question 10.2 — Test f(z)=z̄ for complex differentiability.
Test f(z)=z̄ for complex differentiability.
10 marksShow worked answer
Here u=x and v=−y. Cauchy–Riemann requires uₓ=1=vᵧ=−1, which fails everywhere. Thus f is nowhere complex differentiable.
Question 10.3 — Find an analytic function with real part u=x²−y².
Find an analytic function with real part u=x²−y².
8 marksShow worked answer
The Cauchy–Riemann equations give vᵧ=uₓ=2x and vₓ=−uᵧ=2y. Integrating yields v=2xy+C. Therefore f(z)=z²+iC for real constant C.
Poles at 0 and 1 lie inside. Residue at 0 is lim z→0 z(z²+1)/(z(z−1))=−1. Residue at 1 is lim z→1 (z−1)(z²+1)/(z(z−1))=2. Sum=1, so integral=2πi.
Undergraduate · Chapter 11
Chapter 11 — Abstract Algebra and Group Theory
Groups, subgroups, cyclic groups, homomorphisms, quotient groups, rings and fields. Abstract algebra studies algebraic structures through axioms and structure-preserving maps. Group theory formalises symmetry; rings and fields generalise arithmetic. Proofs require careful use of definitions and structural theorems.
Learning outcomes
By the end of this chapter, students should be able to define the central objects precisely, state and apply key theorems with their hypotheses, construct rigorous derivations, solve unfamiliar problems and interpret mathematical results in context.
11.1 Binary operations and groups
A group (G,*) has closure, associativity, an identity e and inverses. Commutativity is an additional property, not part of the group axioms. The order of an element is the least positive n with aⁿ=e, if such n exists.
11.2 Subgroups and cyclic groups
A nonempty subset H⊆G is a subgroup if it is closed under ab⁻¹. A group generated by g is ⟨g⟩={gⁿ:n∈ℤ}; cyclic groups are abelian. Every subgroup of a cyclic group is cyclic.
11.3 Permutation groups
Sₙ is the group of all permutations of n objects under composition, with |Sₙ|=n!. Disjoint cycle decomposition is unique up to cycle order. A k-cycle has sign (−1)^(k−1).
11.4 Cosets and Lagrange's theorem
Left cosets of H partition a finite group G into equal-sized sets. Thus |H| divides |G|. The order of any element divides |G|. The converse of Lagrange's theorem is not true for all groups.
11.5 Normal subgroups and quotient groups
H is normal if gH=Hg for every g∈G, equivalently ghg⁻¹∈H. Only normal subgroups yield quotient groups G/H with well-defined coset multiplication.
11.6 Homomorphisms and isomorphisms
A homomorphism φ:G→K satisfies φ(ab)=φ(a)φ(b). Its kernel is normal and image is a subgroup. The first isomorphism theorem states G/kerφ≅imφ.
11.7 Direct products and group actions
The direct product combines componentwise operations. A group action describes symmetries acting on a set. Orbit–stabiliser gives |Orb(x)|=[G:Stab(x)] for finite groups.
11.8 Rings and integral domains
A ring has an abelian additive group and associative multiplication distributing over addition. A commutative ring with identity and no zero divisors is an integral domain. Units are elements with multiplicative inverses.
11.9 Ideals and quotient rings
An ideal absorbs multiplication by ring elements and is an additive subgroup. Quotient rings R/I generalise quotient groups. In a commutative ring with identity, maximal ideals correspond to quotient fields.
11.10 Fields and polynomial structure
A field is a commutative ring in which every nonzero element is a unit. Polynomial division over a field yields quotient and remainder; irreducible polynomials generate field extensions via quotient constructions.
11.11 University assessment — original problems
Attempt each problem before revealing its worked solution. Marks reward mathematical reasoning, correct notation, justified steps and interpretation.
Question 11.1 — Show that the even integers form a subgroup of (ℤ,+).
Show that the even integers form a subgroup of (ℤ,+).
8 marksShow worked answer
The set is nonempty. If a=2m and b=2n, then a−b=2(m−n), which is even. The subgroup test applies.
Question 11.2 — Find the order of 8 in the additive group ℤ₁₂.
Find the order of 8 in the additive group ℤ₁₂.
10 marksShow worked answer
The order is 12/gcd(12,8)=12/4=3. Indeed 8,16≡4,24≡0 modulo 12.
Question 11.3 — How many elements of order 2 are in S₃?
How many elements of order 2 are in S₃?
8 marksShow worked answer
The three transpositions (12),(13),(23) each have order 2. The identity has order 1 and the two 3-cycles have order 3. Answer: 3.
Question 11.4 — Prove every subgroup of an abelian group is normal.
Prove every subgroup of an abelian group is normal.
10 marksShow worked answer
For h∈H and g∈G, ghg⁻¹=hg g⁻¹=h because gh=hg. Thus ghg⁻¹∈H, so H is normal.
Question 11.5 — For φ:ℤ→ℤ₆ given by φ(n)=n mod 6, find kernel and image.
For φ:ℤ→ℤ₆ given by φ(n)=n mod 6, find kernel and image.
8 marksShow worked answer
Kernel={n:6 divides n}=6ℤ. Every residue class occurs, so image=ℤ₆. The first isomorphism theorem gives ℤ/6ℤ≅ℤ₆.
Question 11.6 — Is ℤ₈ a field? Justify.
Is ℤ₈ a field? Justify.
10 marksShow worked answer
No. The nonzero element 2 is a zero divisor because 2·4=8≡0 mod 8. A field has no zero divisors.
Undergraduate · Chapter 12
Chapter 12 — Elementary and Analytic Number Theory
Divisibility, congruences, primes, arithmetic functions, Diophantine equations and analytic methods. Number theory studies the integers and their arithmetic structure. Elementary methods include divisibility, Euclid's algorithm, congruences and prime factorisation; analytic number theory studies arithmetic through series, products and estimates.
Learning outcomes
By the end of this chapter, students should be able to define the central objects precisely, state and apply key theorems with their hypotheses, construct rigorous derivations, solve unfamiliar problems and interpret mathematical results in context.
12.1 Divisibility and the Euclidean algorithm
For integers a,b with a≠0, a|b means b=ak for some integer k. The Euclidean algorithm repeatedly replaces (a,b) by (b,r), where a=qb+r and 0≤r<|b|; the final nonzero remainder is gcd(a,b). Back-substitution yields Bézout coefficients.
12.2 Primes and fundamental theorem of arithmetic
A prime p>1 has only positive divisors 1 and p. Every integer n>1 factors uniquely into primes up to order. Euclid's lemma states p|ab implies p|a or p|b. Infinitude of primes follows by considering one plus the product of any finite list.
12.3 Linear Diophantine equations
The equation ax+by=c has integer solutions exactly when gcd(a,b)|c. If (x₀,y₀) is one solution and d=gcd(a,b), all solutions are x=x₀+(b/d)t, y=y₀−(a/d)t.
12.4 Congruences and modular arithmetic
a≡b (mod m) iff m|(a−b). Congruences respect addition and multiplication. Cancellation by a is valid modulo m only when gcd(a,m)=1, unless additional divisibility conditions are established.
12.5 Chinese remainder theorem
For pairwise coprime moduli mᵢ, a system x≡aᵢ (mod mᵢ) has a unique solution modulo M=∏mᵢ. For non-coprime moduli, compatibility requires residues agree modulo each pairwise gcd.
12.6 Fermat and Euler theorems
If p is prime and p∤a, a^(p−1)≡1 (mod p). More generally, if gcd(a,n)=1, a^φ(n)≡1 (mod n), where φ is Euler's totient function.
12.7 Multiplicative arithmetic functions
A function f is multiplicative if f(mn)=f(m)f(n) when gcd(m,n)=1. Examples include φ(n), divisor count τ(n), and sum of divisors σ(n). For n=∏pᵃ, φ(n)=n∏(1−1/p), τ(n)=∏(a+1).
12.8 Quadratic residues
An integer a is a quadratic residue modulo odd prime p if x²≡a (mod p) has a solution. The Legendre symbol (a/p) records residue status. Euler's criterion gives a^((p−1)/2)≡(a/p) mod p for p∤a.
12.9 Diophantine equations and continued fractions
Integer solutions may be constrained by parity, modular residues, factorisation and inequalities. Continued fractions provide rational approximations and are central to solving Pell equations x²−Dy²=1.
12.10 Analytic number theory: zeta function and primes
For Re(s)>1, ζ(s)=Σn≥1 n^(−s)=∏p(1−p^(−s))^(−1). The Euler product encodes unique prime factorisation. The prime number theorem states π(x)~x/log x as x→∞.
12.11 University assessment — original problems
Attempt each problem before revealing its worked solution. Marks reward mathematical reasoning, correct notation, justified steps and interpretation.
Question 12.1 — Use Euclid's algorithm to find gcd(252,198), then express it as a linear combination.
Use Euclid's algorithm to find gcd(252,198), then express it as a linear combination.
Modulo 10, powers of 7 cycle 7,9,3,1 with period 4. Since 100≡0 mod 4, 7¹⁰⁰≡1 mod 10. Last digit 1.
Question 12.6 — Show there are infinitely many primes.
Show there are infinitely many primes.
10 marksShow worked answer
Assume finitely many primes p₁,…,pₙ. Let N=p₁…pₙ+1. No pᵢ divides N. Any prime divisor of N is therefore absent from the list, contradiction.
Undergraduate · Chapter 13
Chapter 13 — Discrete Mathematics and Graph Theory
Logic, algorithms, relations, graphs, trees, connectivity, planarity and complexity. Discrete mathematics studies countable structures and finite processes. It underpins algorithms, computer science, networks, cryptography and combinatorial optimisation. Graph theory models pairwise relationships using vertices and edges.
Learning outcomes
By the end of this chapter, students should be able to define the central objects precisely, state and apply key theorems with their hypotheses, construct rigorous derivations, solve unfamiliar problems and interpret mathematical results in context.
13.1 Propositional and predicate logic
Logical equivalence, implication, quantifiers and inference rules formalise valid reasoning. A proof system must distinguish syntactic derivability from semantic truth. Quantifier negation reverses ∀ and ∃ and negates the predicate.
13.2 Sets, relations and functions
Relations are subsets of Cartesian products. Equivalence relations are reflexive, symmetric and transitive; partial orders are reflexive, antisymmetric and transitive. Functions may be injective, surjective or bijective.
13.3 Mathematical induction and recursion
Induction proves a base case and an implication from n to n+1. Strong induction assumes all earlier cases. Recursive definitions specify initial values and a rule generating later terms; termination and correctness require proof.
13.4 Algorithms and asymptotic analysis
An algorithm is a finite, unambiguous procedure. Big-O gives an asymptotic upper bound; Θ gives a tight bound and Ω a lower bound. Analyse worst-case, average-case or best-case complexity as specified.
13.5 Graphs and representations
A graph G=(V,E) has vertices and edges. Graphs may be directed, undirected, weighted, simple or multigraphs. Adjacency matrices encode edges; adjacency lists are often more space-efficient for sparse graphs.
13.6 Paths, connectivity and components
A walk may repeat vertices and edges; a trail does not repeat edges; a path does not repeat vertices. Connected components partition an undirected graph. In directed graphs, strong connectivity requires directed paths both ways.
13.7 Trees and spanning trees
A tree is connected and acyclic. A tree with n vertices has n−1 edges and a unique simple path between every pair of vertices. A spanning tree connects all vertices without cycles.
13.8 Euler and Hamilton paths
An Euler trail uses every edge exactly once; a connected graph has an Euler circuit iff every vertex has even degree, and an Euler trail iff exactly zero or two vertices have odd degree. A Hamilton path visits every vertex once; no equally simple general degree criterion exists.
13.9 Planar graphs and colouring
A planar graph can be embedded without edge crossings. For a connected planar graph, V−E+F=2. A simple planar graph with V≥3 has E≤3V−6. The four-colour theorem states every planar graph is 4-colourable.
13.10 Graph algorithms and complexity
Breadth-first search finds shortest paths in unweighted graphs; depth-first search supports connectivity and cycle detection. Dijkstra's algorithm requires nonnegative edge weights. Minimum spanning trees are found by Kruskal or Prim. P, NP and NP-completeness classify computational problems.
13.11 University assessment — original problems
Attempt each problem before revealing its worked solution. Marks reward mathematical reasoning, correct notation, justified steps and interpretation.
Question 13.1 — A graph has degrees 3,3,2,2,2. Is this degree sequence possible?
A graph has degrees 3,3,2,2,2. Is this degree sequence possible?
8 marksShow worked answer
The degree sum is 12, which is even, so the handshake lemma does not rule it out. A simple graph realisation exists: take a 5-cycle and add a chord joining two vertices at distance two; the degrees become 3,3,2,2,2.
Question 13.2 — How many edges does a tree with 25 vertices have?
How many edges does a tree with 25 vertices have?
10 marksShow worked answer
Every finite tree has n−1 edges. Therefore it has 24 edges.
Question 13.3 — Can a connected graph with exactly four odd-degree vertices have an Euler trail?
Can a connected graph with exactly four odd-degree vertices have an Euler trail?
8 marksShow worked answer
No. A connected undirected graph has an Euler trail only if it has zero or two odd-degree vertices. Four odd vertices violate the criterion.
Question 13.4 — A connected planar graph has V=10 and E=15. Find F.
A connected planar graph has V=10 and E=15. Find F.
10 marksShow worked answer
Euler's formula V−E+F=2 gives 10−15+F=2, hence F=7.
Question 13.5 — Give the asymptotic order of 4n²+7n log n+3.
Give the asymptotic order of 4n²+7n log n+3.
8 marksShow worked answer
The quadratic term dominates as n grows. The function is Θ(n²).
Question 13.6 — Why is Dijkstra's algorithm not generally correct with negative edge weights?
Why is Dijkstra's algorithm not generally correct with negative edge weights?
10 marksShow worked answer
Its greedy step permanently settles the currently smallest tentative distance. A later path using a negative edge can reduce that distance, invalidating the settled choice. Nonnegative weights are essential to the standard proof.
Undergraduate · Chapter 14
Chapter 14 — Combinatorics
Counting principles, permutations, combinations, recurrences, generating functions and extremal methods. Combinatorics counts finite structures and proves existence through counting arguments. It includes bijections, inclusion–exclusion, recurrences, generating functions and extremal bounds.
Learning outcomes
By the end of this chapter, students should be able to define the central objects precisely, state and apply key theorems with their hypotheses, construct rigorous derivations, solve unfamiliar problems and interpret mathematical results in context.
14.1 Sum and product principles
If disjoint alternatives have m and n possibilities, their union has m+n. If a process has m choices followed by n choices independently, it has mn outcomes. Avoid multiplying when choices are dependent without adjusting the counts.
14.2 Permutations and combinations
Ordered selections of r distinct objects from n number n!/(n−r)!. Unordered selections number C(n,r)=n!/[r!(n−r)!]. With repetition, ordered r-tuples from n symbols number nʳ.
14.3 Multinomial coefficients and distributions
The number of arrangements of n objects with multiplicities n₁,…,nₖ is n!/(n₁!…nₖ!). The number of nonnegative integer solutions to x₁+…+xₖ=n is C(n+k−1,k−1), by stars and bars.
14.4 Binomial theorem and identities
(x+y)ⁿ=Σ C(n,k)x^(n−k)yᵏ. Pascal's identity C(n,k)=C(n−1,k)+C(n−1,k−1) follows by separating subsets according to whether they contain a fixed element.
14.5 Inclusion–exclusion
For finite sets, |A∪B|=|A|+|B|−|A∩B|. For many sets, alternate sums of intersections correct repeated counting. The principle also counts permutations with forbidden fixed points.
14.6 Pigeonhole principle
If N objects are placed into k boxes, some box contains at least ceil(N/k) objects. Generalised forms prove that sufficiently large finite structures must contain repeated patterns.
14.7 Recurrence relations
A recurrence defines a sequence from earlier terms and initial conditions. Linear constant-coefficient recurrences are solved using characteristic polynomials; repeated roots introduce polynomial factors in n.
14.8 Generating functions
The ordinary generating function of (aₙ) is A(x)=Σaₙxⁿ. Products encode convolution and combinatorial constructions. Coefficients can be extracted algebraically or through partial fractions.
14.9 Catalan numbers and structures
Cₙ=(1/(n+1))C(2n,n) counts balanced parenthesis strings, binary tree shapes and several other recursively defined objects. Their recurrence is Cₙ=Σₖ₌₀ⁿ⁻¹ CₖCₙ₋₁₋ₖ.
14.10 Extremal and probabilistic counting
Extremal combinatorics asks how large or small a structure can be under constraints. Double counting evaluates one set of incidences in two ways. The probabilistic method proves existence by showing a random construction has positive probability of the desired property.
14.11 University assessment — original problems
Attempt each problem before revealing its worked solution. Marks reward mathematical reasoning, correct notation, justified steps and interpretation.
Question 14.1 — How many ways can 5 distinct books be arranged on a shelf?
How many ways can 5 distinct books be arranged on a shelf?
8 marksShow worked answer
There are 5!=120 permutations.
Question 14.2 — How many committees of 3 can be selected from 10 people?
How many committees of 3 can be selected from 10 people?
10 marksShow worked answer
Order is irrelevant, so C(10,3)=10·9·8/(3·2·1)=120.
Question 14.3 — How many nonnegative integer solutions satisfy x₁+x₂+x₃+x₄=9?
How many nonnegative integer solutions satisfy x₁+x₂+x₃+x₄=9?
8 marksShow worked answer
Stars and bars gives C(9+4−1,4−1)=C(12,3)=220.
Question 14.4 — Among 13 people, prove two share a birth month.
Among 13 people, prove two share a birth month.
10 marksShow worked answer
There are 12 months (boxes) and 13 people (objects). By the pigeonhole principle, one month contains at least ceil(13/12)=2 birthdays.
Question 14.5 — Solve aₙ=5aₙ₋₁−6aₙ₋₂, a₀=1, a₁=4.
Solve aₙ=5aₙ₋₁−6aₙ₋₂, a₀=1, a₁=4.
8 marksShow worked answer
Characteristic equation r²−5r+6=0 gives roots 2,3. Thus aₙ=A2ⁿ+B3ⁿ. A+B=1 and 2A+3B=4 give A=−1,B=2. Therefore aₙ=−2ⁿ+2·3ⁿ.
Question 14.6 — Use inclusion–exclusion: in a class of 40, 24 study algebra, 19 calculus and 10 both. How many study at least one?
Use inclusion–exclusion: in a class of 40, 24 study algebra, 19 calculus and 10 both. How many study at least one?
10 marksShow worked answer
|A∪C|=24+19−10=33. Thus 33 study at least one and 7 study neither.
Undergraduate · Chapter 15
Chapter 15 — Probability Theory
Axioms, conditional probability, random variables, distributions, expectation, limit theorems and stochastic models. Probability theory provides a mathematical framework for uncertainty. It begins with a sample space and probability measure, then develops random variables, distributions, expectation, conditional models and asymptotic limit theorems.
Learning outcomes
By the end of this chapter, students should be able to define the central objects precisely, state and apply key theorems with their hypotheses, construct rigorous derivations, solve unfamiliar problems and interpret mathematical results in context.
15.1 Probability spaces and axioms
A probability space (Ω,F,P) consists of outcomes Ω, a sigma-algebra F of events and a probability measure P. Kolmogorov axioms: P(A)≥0, P(Ω)=1 and countable additivity for pairwise disjoint events. Consequences include P(Aᶜ)=1−P(A).
15.2 Counting probability and conditional probability
For equally likely finite outcomes, P(A)=|A|/|Ω|. Conditional probability P(A|B)=P(A∩B)/P(B) when P(B)>0. Independence means P(A∩B)=P(A)P(B), equivalently P(A|B)=P(A) when defined.
15.3 Bayes' theorem and total probability
If B₁,…,Bₖ partition Ω, P(A)=ΣP(A|Bᵢ)P(Bᵢ). Bayes' theorem reverses conditioning: P(Bⱼ|A)=P(A|Bⱼ)P(Bⱼ)/P(A). Base rates must be included when interpreting diagnostic evidence.
15.4 Random variables and distributions
A random variable X is a measurable function from Ω to ℝ. Its cumulative distribution function F(x)=P(X≤x) is nondecreasing, right-continuous, and has limits 0 and 1 at infinities. Discrete variables use probability mass functions; continuous variables use densities.
15.5 Expectation, variance and moments
E[X]=Σx p(x) or ∫xf(x)dx. Var(X)=E[(X−μ)²]=E[X²]−μ². For constants a,b, E[aX+b]=aE[X]+b and Var(aX+b)=a²Var(X). Covariance measures linear association.
15.6 Standard discrete distributions
Bernoulli(p) has mean p and variance p(1−p). Binomial(n,p) counts successes in n independent trials. Geometric(p) counts trials to first success. Poisson(λ) models counts with mean and variance λ under its assumptions.
15.7 Standard continuous distributions
Uniform(a,b), exponential(λ), normal(μ,σ²) and gamma distributions model different structures. The exponential distribution is memoryless. Normal standardisation uses Z=(X−μ)/σ.
15.8 Joint distributions and conditional expectation
Joint distributions describe dependence. Marginals sum or integrate out variables. Independence requires factorisation of the joint law. Conditional expectation E[X|Y] is a random variable and satisfies the tower property E(E[X|Y])=E[X].
15.9 Laws of large numbers and central limit theorem
Under suitable assumptions, sample averages converge to the population mean (law of large numbers). The central limit theorem states standardised sums of independent identically distributed variables with finite variance converge in distribution to N(0,1).
15.10 Transformations and generating functions
For a monotone differentiable transformation Y=g(X), density transformation uses the Jacobian: fY(y)=fX(g⁻¹(y))|d g⁻¹/dy|. Moment-generating functions, when finite near zero, encode moments and identify distributions under standard conditions.
15.11 University assessment — original problems
Attempt each problem before revealing its worked solution. Marks reward mathematical reasoning, correct notation, justified steps and interpretation.
Question 15.1 — A fair die is rolled. Find P(even or greater than 4).
A fair die is rolled. Find P(even or greater than 4).
8 marksShow worked answer
Even={2,4,6}; greater than 4={5,6}. Their union={2,4,5,6}, so probability=4/6=2/3.
Question 15.2 — A test has sensitivity 0.95, specificity 0.90 and disease prevalence 0.02. Find the probability of disease given a positive result.
A test has sensitivity 0.95, specificity 0.90 and disease prevalence 0.02. Find the probability of disease given a positive result.
10 marksShow worked answer
P(+|D)=0.95; P(+|Dᶜ)=0.10. P(D|+)=0.95·0.02/[0.95·0.02+0.10·0.98]=0.019/0.117≈0.1624, or 16.24%.
Question 15.3 — If X~Binomial(10,0.3), find its mean and variance.
If X~Binomial(10,0.3), find its mean and variance.
8 marksShow worked answer
E[X]=np=3. Var(X)=np(1−p)=10·0.3·0.7=2.1.
Question 15.4 — Let X have density f(x)=2x for 0
Let X have density f(x)=2x for 010 marksShow worked answer
Question 15.5 — If X and Y are independent with variances 4 and 9, find Var(2X−3Y).
If X and Y are independent with variances 4 and 9, find Var(2X−3Y).
8 marksShow worked answer
Independence gives zero covariance. Var(2X−3Y)=4Var(X)+9Var(Y)=16+81=97.
Question 15.6 — For X~Poisson(2), find P(X=0 or X=1).
For X~Poisson(2), find P(X=0 or X=1).
10 marksShow worked answer
P(X=0)+P(X=1)=e^(−2)+2e^(−2)=3e^(−2)≈0.4060.
Undergraduate · Chapter 16
Chapter 16 — Mathematical Statistics
Estimation, sampling distributions, hypothesis testing, regression, ANOVA and statistical inference. Mathematical statistics develops methods for learning about populations from random samples. It combines probability models with estimation, uncertainty quantification, testing and predictive modelling. The distinction between a population parameter and a sample statistic is fundamental.
Learning outcomes
By the end of this chapter, students should be able to define the principal concepts, state and apply relevant results with their assumptions, derive mathematical conclusions, solve multi-step problems and interpret results in their disciplinary context.
16.1 Populations, samples and sampling schemes
A population is the full collection under study; a sample is the observed subset. A statistic is a measurable function of sample data that contains no unknown parameters. Simple random sampling gives each sample of a fixed size equal selection probability. Stratified, cluster and systematic designs address different sampling structures.
16.2 Sampling distributions and standard errors
A sampling distribution is the probability law of a statistic under repeated sampling. The standard error is the standard deviation of that sampling distribution. For independent observations with variance σ², the sample mean has variance σ²/n and standard error σ/√n.
16.3 Point estimation and properties
An estimator is a statistic used to estimate a parameter. Bias is E[θ̂]−θ; variance measures dispersion; mean squared error is Var(θ̂)+Bias(θ̂)². Consistency concerns convergence in probability; efficiency compares variance among estimators under specified conditions.
16.4 Method of moments and maximum likelihood
Method-of-moments estimators equate sample moments to corresponding population moments. Maximum likelihood chooses parameter values maximising L(θ)=∏f(Xᵢ;θ), or log-likelihood ℓ(θ)=Σlog f(Xᵢ;θ). Check parameter constraints and boundary maxima.
16.5 Confidence intervals
A (1−α) confidence procedure has coverage 1−α under its model and sampling assumptions. A confidence interval is produced by a random procedure; after observing data, the parameter is fixed. For a normal mean with known σ, the interval is x̄±z₁₋α/₂ σ/√n.
16.6 Hypothesis testing and p-values
A test controls a Type I error probability α under the null. Type II error β is failure to reject a false null; power is 1−β. A p-value is the probability, assuming H₀, of a test statistic at least as extreme as observed. It is not P(H₀ is true).
16.7 Likelihood ratio and large-sample theory
Under regularity conditions, likelihood-ratio statistics often have asymptotic chi-square distributions. Maximum-likelihood estimators are commonly asymptotically normal with variance governed by Fisher information. Conditions matter, especially at parameter boundaries or under non-identifiability.
16.8 Comparing means and analysis of variance
Two-sample procedures compare population means under independent-sample assumptions. Paired designs analyse within-pair differences. One-way ANOVA decomposes total variability into between-group and within-group sums of squares; its F statistic compares mean squares.
16.9 Correlation and linear regression
The least-squares line minimises Σ(yᵢ−β̂₀−β̂₁xᵢ)². Under the classical linear model, errors have zero conditional mean and suitable variance and independence assumptions. Correlation measures linear association, not causation.
16.10 Nonparametric methods and diagnostics
Rank-based methods can reduce dependence on distributional assumptions. Residual plots assess nonlinearity, unequal variance and outliers; Q–Q plots assess distributional shape. Model diagnostics do not replace understanding of study design and data collection.
16.11 University assessment — original problems
Attempt each problem before revealing its worked solution. Marks reward mathematical reasoning, correct notation, justified steps and interpretation.
Question 16.1
A sample of size 36 has mean 50. Population standard deviation is 12. Find a 95% confidence interval for the population mean.
8 marksShow worked answer
Known σ: standard error=12/√36=2. Using z=1.96, margin=3.92. Interval is (46.08,53.92).
Question 16.2
For observations 2,4,6,8, find the unbiased sample variance.
10 marksShow worked answer
Mean=5. Squared deviations are 9,1,1,9, sum=20. Divide by n−1=3: s²=20/3≈6.667.
Question 16.3
A test has α=0.05 and β=0.20 at a specified alternative. Find its power.
8 marksShow worked answer
Power=1−β=0.80, or 80%.
Question 16.4
For Bernoulli observations with sample proportion 0.4, find the maximum-likelihood estimate of p.
10 marksShow worked answer
The likelihood is p^(Σxᵢ)(1−p)^(n−Σxᵢ). Maximising its log gives p̂=Σxᵢ/n=0.4.
Question 16.5
In a one-way ANOVA, SSB=40 with 3 groups and SSW=60 with 20 residual degrees of freedom. Find F.
8 marksShow worked answer
Between df=3−1=2; within df=20. MSB=40/2=20; MSW=60/20=3. F=20/3≈6.667.
Question 16.6
A regression model gives β̂₁=2.5 and standard error 0.5. Give the large-sample 95% interval for the slope.
10 marksShow worked answer
2.5±1.96(0.5)=2.5±0.98. Interval (1.52,3.48). Its interpretation depends on the regression assumptions and sampling design.
Undergraduate · Chapter 17
Chapter 17 — Numerical Analysis and Scientific Computing
Floating-point arithmetic, error analysis, root finding, interpolation, quadrature and numerical ODEs. Numerical analysis studies algorithms for approximating mathematical problems and quantifies their accuracy, stability and computational cost. A computed answer is meaningful only when its error and conditioning are understood.
Learning outcomes
By the end of this chapter, students should be able to define the principal concepts, state and apply relevant results with their assumptions, derive mathematical conclusions, solve multi-step problems and interpret results in their disciplinary context.
17.1 Floating-point arithmetic and error
A floating-point system stores a finite number of significant digits and a bounded exponent. Rounding and cancellation create error. Absolute error is |x−x̂|; relative error is |x−x̂|/|x| for x≠0. Machine precision describes the scale of representable rounding errors.
17.2 Conditioning and stability
Conditioning is sensitivity of the mathematical problem to perturbations in input. Stability concerns how an algorithm propagates rounding and data errors. An ill-conditioned problem may defeat even a stable algorithm; an unstable algorithm may fail on a well-conditioned problem.
17.3 Bisection and fixed-point iteration
If continuous f changes sign on [a,b], bisection retains a sign-changing half-interval; after n steps its width is (b−a)/2ⁿ. Fixed-point iteration xₙ₊₁=g(xₙ) converges locally under suitable contraction conditions, such as |g′|≤k<1 on an invariant interval.
17.4 Newton and secant methods
Newton iteration is xₙ₊₁=xₙ−f(xₙ)/f′(xₙ), with quadratic local convergence near a simple root under regularity assumptions. The secant method approximates the derivative using two previous iterates and avoids explicit derivative evaluation.
17.5 Linear systems and iterative methods
Gaussian elimination solves dense systems directly; pivoting reduces numerical risk. Jacobi and Gauss–Seidel are iterative methods whose convergence depends on matrix properties. Residual r=b−Ax measures equation satisfaction but must be interpreted with conditioning.
17.6 Polynomial interpolation
Lagrange interpolation constructs the unique polynomial of degree at most n through n+1 distinct nodes. Newton divided differences allow incremental construction. High-degree equally spaced interpolation may oscillate; piecewise splines often behave better.
17.7 Numerical differentiation
Finite differences approximate derivatives. Forward difference has O(h) truncation error; centred difference typically has O(h²). Very small h amplifies rounding error, so a balance is required.
17.8 Numerical integration
Composite trapezoidal and Simpson rules approximate definite integrals. For sufficiently smooth f, composite trapezoidal error is O(h²) and composite Simpson error is O(h⁴). Adaptive quadrature refines intervals where estimated error is large.
17.9 Numerical solution of ODEs
Euler's method has global error O(h) under standard smoothness assumptions. Classical RK4 has global error O(h⁴). Stiff systems may require implicit methods; stability regions constrain allowable step sizes.
17.10 Reproducible scientific computing
A sound computational study records algorithm, tolerances, step size, precision, stopping criterion, software environment and validation tests. Compare against exact solutions, conserved quantities or independent methods where possible.
17.11 University assessment — original problems
Attempt each problem before revealing its worked solution. Marks reward mathematical reasoning, correct notation, justified steps and interpretation.
Question 17.1
How many bisection steps guarantee an interval width at most 10⁻³ from an initial width 1?
8 marksShow worked answer
Need 1/2ⁿ≤10⁻³, so 2ⁿ≥1000. Since 2⁹=512 and 2¹⁰=1024, n=10 steps suffice.
Question 17.2
Apply Newton's method once to f(x)=x²−2 from x₀=1.5.
Use the centred difference for f′(1) when f(x)=x³ and h=0.1.
8 marksShow worked answer
[f(1.1)−f(0.9)]/(0.2)=(1.331−0.729)/0.2=3.01. Exact derivative is 3; the O(h²) error is 0.01.
Question 17.4
Use the trapezoidal rule with one panel to approximate ∫₀²x² dx.
10 marksShow worked answer
h=2, T=h[f(0)+f(2)]/2=2(0+4)/2=4. Exact integral=8/3; approximation error is 4/3.
Question 17.5
For y′=y, y(0)=1, take one Euler step with h=0.1.
8 marksShow worked answer
y₁=y₀+h y₀=1+0.1=1.1. Exact value e^0.1≈1.10517.
Question 17.6
Explain the difference between conditioning and algorithmic stability.
10 marksShow worked answer
Conditioning measures sensitivity inherent in the problem; stability measures how an algorithm handles perturbations and rounding. A poorly conditioned problem can be solved by a stable method yet remain sensitive; an unstable method can magnify errors unnecessarily.
Undergraduate · Chapter 18
Chapter 18 — Optimisation and Operations Research
Linear programming, duality, network optimisation, integer programming, dynamic programming and decision analysis. Optimisation selects the best feasible decision according to a mathematical objective. Operations research applies optimisation and analytical methods to resource allocation, logistics, scheduling, production and decision-making.
Learning outcomes
By the end of this chapter, students should be able to define the principal concepts, state and apply relevant results with their assumptions, derive mathematical conclusions, solve multi-step problems and interpret results in their disciplinary context.
18.1 Mathematical formulation and feasible regions
A model specifies decision variables, an objective function and constraints. Feasible solutions satisfy all constraints; an optimum is a feasible solution with best objective value. Distinguish infeasibility, unboundedness and multiple optimal solutions.
18.2 Linear programming geometry
A linear programme has a linear objective and linear constraints. Its feasible region is a convex polyhedron. If a finite optimum exists, at least one optimal extreme point exists, although an entire face may be optimal.
18.3 Simplex method
The simplex method moves between adjacent basic feasible solutions while improving the objective. A tableau records basis, reduced costs and feasibility. Pivot selection, degeneracy, cycling and numerical tolerances are important implementation issues.
18.4 Duality and economic interpretation
Every linear programme has a dual. Weak duality bounds the primal objective; strong duality gives equal optimal values when finite optima exist. Dual variables often measure marginal value of relaxing constraints, subject to regularity and basis stability.
18.5 Sensitivity analysis
Sensitivity analysis studies changes in objective coefficients and right-hand sides. Shadow prices and allowable ranges are local to the current optimal basis; large changes may alter the basis and require re-optimisation.
18.6 Transportation and assignment problems
Transportation models allocate supply to demand at minimum shipping cost. Assignment is a special case with one-to-one allocation. Network structure permits specialised algorithms and integrality properties.
18.7 Integer and mixed-integer programming
Some decisions must be integer or binary. The LP relaxation provides bounds; branch-and-bound explores subproblems and cutting planes strengthen relaxations. Integer restrictions can make an otherwise convex-looking problem computationally difficult.
18.8 Network flows and shortest paths
A flow network has capacities and conservation at intermediate nodes. Max-flow min-cut equates maximum feasible flow with minimum cut capacity. Shortest-path algorithms require assumptions about edge weights; minimum-cost flow adds costs to flow decisions.
18.9 Dynamic programming
Dynamic programming decomposes a multistage problem using Bellman's principle of optimality. A recurrence relates the value of a state to feasible actions and successor states. State design controls computational tractability.
18.10 Queues, inventory and decision analysis
Queueing models analyse arrivals, service and waiting; inventory models balance ordering, holding and shortage costs. Decision analysis uses expected utility, scenarios and sensitivity to make uncertainty explicit.
18.11 University assessment — original problems
Attempt each problem before revealing its worked solution. Marks reward mathematical reasoning, correct notation, justified steps and interpretation.
Question 18.1
Maximise Z=3x+2y subject to x+y≤4, x≤2, y≤3, x,y≥0.
8 marksShow worked answer
Corner points include (0,0),(2,0),(2,2),(1,3),(0,3). Objective values are 0,6,10,9,6. Maximum is Z=10 at (2,2).
Question 18.2
Write the dual of max 3x+2y subject to x+y≤4, x≤2, y≤3, x,y≥0.
10 marksShow worked answer
For max with ≤ constraints and nonnegative variables, dual is min 4u+2v+3w subject to u+v≥3, u+w≥2, u,v,w≥0.
Question 18.3
A project has two activities in sequence taking 3 and 5 days. What is the minimum completion time?
8 marksShow worked answer
With strict sequential dependence, duration is 3+5=8 days. Parallelisation is not possible under the stated precedence.
Question 18.4
A network cut has capacities 4, 7 and 3 crossing from source side to sink side. What is its capacity and the max-flow upper bound?
10 marksShow worked answer
Cut capacity=4+7+3=14. Every feasible source-to-sink flow is at most 14 by the cut bound.
Question 18.5
An item has weight 2 and value 7; capacity is 5 and items are indivisible. With unlimited copies, how many should be selected?
8 marksShow worked answer
At most floor(5/2)=2 items fit. Two items weigh 4 and yield value 14, which is optimal because a third exceeds capacity.
Question 18.6
For an M/M/1 queue with arrival rate 2 per hour and service rate 3 per hour, find utilisation and mean number in system.
10 marksShow worked answer
ρ=λ/μ=2/3<1, so the steady state exists. L=ρ/(1−ρ)= (2/3)/(1/3)=2 customers.
Undergraduate · Chapter 19
Chapter 19 — Mathematical Modelling
Model formulation, dimensional analysis, scaling, dynamical systems, calibration, validation and uncertainty. Mathematical modelling translates a real system into variables, assumptions, equations and predictions. A model is an idealisation; its quality depends on purpose, data, assumptions, mathematical analysis and validation against observations.
Learning outcomes
By the end of this chapter, students should be able to define the principal concepts, state and apply relevant results with their assumptions, derive mathematical conclusions, solve multi-step problems and interpret results in their disciplinary context.
19.1 Model-building cycle
Identify the question, define system boundaries, select state and control variables, state assumptions, formulate equations, analyse or compute predictions, compare with data and revise. A model should be no more complex than its purpose requires.
19.2 Dimensional analysis
Physical dimensions constrain valid equations. The Buckingham Π theorem reduces variables to dimensionless groups when dimensional homogeneity applies. Dimensional analysis can identify governing ratios and guide experimental design.
19.3 Scaling and nondimensionalisation
Introduce characteristic scales for variables and parameters to obtain dimensionless equations. Dominant dimensionless groups reveal regimes and simplify asymptotic analysis. Scaling should preserve the physical interpretation of boundary and initial data.
19.4 Compartment and conservation models
Compartment models track quantities in connected reservoirs. Conservation laws state rate of accumulation equals inflow minus outflow plus production. Transfers between compartments cancel in the total balance when no external source or sink exists.
19.5 Population and epidemic models
Exponential growth assumes per-capita rate constant; logistic growth includes density limitation. SIR models divide a population into susceptible, infectious and recovered compartments and encode transmission and recovery assumptions.
19.6 Chemical kinetics and reaction models
Mass-action kinetics relates reaction rates to reactant concentrations raised to stoichiometric powers under ideal assumptions. Reaction networks yield coupled ODEs and conservation relations. Empirical rate laws may differ from elementary-step mechanisms.
Calibration estimates unknown parameters from observations. Structural identifiability asks whether ideal noise-free observations uniquely determine parameters; practical identifiability concerns finite noisy data. Correlated parameters can be difficult to estimate separately.
19.9 Verification, validation and uncertainty
Verification asks whether equations are solved correctly; validation asks whether the model adequately represents the intended system. Sensitivity analysis studies response to inputs and parameters; uncertainty quantification propagates uncertain quantities to predictions.
19.10 Model comparison and limitations
Compare candidate models using predictive performance, residual structure, complexity and scientific plausibility. Extrapolation beyond observed conditions is risky. Report assumptions, uncertainty and domains where conclusions are credible.
19.11 University assessment — original problems
Attempt each problem before revealing its worked solution. Marks reward mathematical reasoning, correct notation, justified steps and interpretation.
Question 19.1
A population follows dP/dt=0.2P with P(0)=500. Find P(t) and doubling time.
8 marksShow worked answer
P(t)=500e^(0.2t). Doubling time satisfies e^(0.2t)=2, so t=ln2/0.2≈3.466 time units.
Question 19.2
For logistic growth dP/dt=0.4P(1−P/1000), identify carrying capacity and maximum growth rate.
10 marksShow worked answer
Carrying capacity K=1000. Growth rate is maximal at P=K/2=500; then dP/dt=0.4·500·0.5=100 population units per time.
Question 19.3
A tank has 100 L containing 10 kg salt. Brine enters at 2 L/min with 0.1 kg/L and leaves at 2 L/min. Form the salt balance.
8 marksShow worked answer
Let S(t) be kg salt. Inflow salt=2·0.1=0.2 kg/min; outflow concentration S/100 times 2=0.02S. Thus dS/dt=0.2−0.02S, S(0)=10.
Question 19.4
A spring-mass system has m=2 kg, c=4 Ns/m, k=8 N/m. Write the unforced equation and classify damping.
10 marksShow worked answer
2x″+4x′+8x=0. Critical damping coefficient is 2√(mk)=2√16=8. Since c=4<8, the system is underdamped.
Question 19.5
Explain why dimensional homogeneity is necessary but not sufficient for a correct physical model.
8 marksShow worked answer
Every additive term in a valid physical equation must have matching dimensions, so dimensional homogeneity is necessary. Many incorrect equations share the same dimensions; empirical evidence, mechanisms, boundary conditions and validation are also required.
Question 19.6
A parameter estimate is highly correlated with another parameter. What modelling concern arises?
10 marksShow worked answer
Practical identifiability may be poor: different parameter combinations can produce similar model outputs. Estimates may have large uncertainty. More informative measurements, experimental design or a simpler model may be needed.
Undergraduate · Chapter 20
Chapter 20 — Topology and Metric Spaces
Metric spaces, open and closed sets, continuity, compactness, connectedness and completeness. Topology studies properties preserved under continuous deformation, while metric spaces add a notion of distance. These ideas generalise convergence and continuity beyond the real line and provide foundations for analysis and geometry.
Learning outcomes
By the end of this chapter, students should be able to define the principal concepts, state and apply relevant results with their assumptions, derive mathematical conclusions, solve multi-step problems and interpret results in their disciplinary context.
20.1 Metric spaces and examples
A metric d:X×X→[0,∞) satisfies d(x,y)=0 iff x=y, symmetry and the triangle inequality. Euclidean, taxicab, maximum and discrete metrics provide different geometries on the same set.
20.2 Open balls, open and closed sets
An open ball B(x,r)={y:d(x,y)
20.3 Interior, closure and boundary
The interior is the largest open subset; closure is the smallest closed superset; boundary is closure minus interior. A set is dense if its closure is the whole space. Density depends on the ambient metric topology.
20.4 Convergence and Cauchy sequences
xₙ→x when d(xₙ,x)→0. A sequence is Cauchy when pairwise distances become arbitrarily small. Completeness means every Cauchy sequence converges in the space; ℚ with usual metric is not complete.
20.5 Continuity and homeomorphisms
f is continuous if preimages of open sets are open, equivalently ε–δ continuity in metric spaces. A homeomorphism is a bijection continuous in both directions; it preserves topological properties, not necessarily distances.
20.6 Compactness
A space is compact if every open cover has a finite subcover. In ℝⁿ with Euclidean topology, Heine–Borel says compact iff closed and bounded. Continuous images of compact spaces are compact; continuous real-valued functions on compact spaces attain extrema.
20.7 Connectedness and path connectedness
A space is connected if it cannot be split into two disjoint nonempty open subsets. Path connected means any two points can be joined by a continuous path. Path connected implies connected, but the converse need not hold.
20.8 Product and subspace topologies
A subspace inherits open sets by intersection with ambient open sets. Product topology is generated by products of open sets in the factors. These constructions allow local properties to be studied componentwise.
20.9 Contraction mappings
On a complete metric space, a contraction T with d(Tx,Ty)≤q d(x,y), 0≤q<1, has a unique fixed point. Iteration converges to it from any starting point. This theorem underlies many existence and numerical arguments.
20.10 Compactness, sequential properties and applications
In metric spaces, compactness is equivalent to sequential compactness and total boundedness plus completeness. These equivalences may fail in arbitrary topological spaces. Compactness supports uniform continuity and existence results.
20.11 University assessment — original problems
Attempt each problem before revealing its worked solution. Marks reward mathematical reasoning, correct notation, justified steps and interpretation.
Question 20.1
Show that d(x,y)=|x−y| is a metric on ℝ.
8 marksShow worked answer
Non-negativity and identity follow from absolute value; symmetry is |x−y|=|y−x|; triangle inequality |x−z|≤|x−y|+|y−z|. Hence it is a metric.
Question 20.2
In ℝ with usual metric, find the closure and boundary of (0,1).
10 marksShow worked answer
Closure is [0,1]. Interior is (0,1). Boundary is {0,1}.
Question 20.3
Is ℚ complete under the usual metric? Give a reason.
8 marksShow worked answer
No. A rational sequence of decimal truncations of √2 is Cauchy in ℚ but converges in ℝ to √2, which is not rational. Thus it has no limit in ℚ.
Question 20.4
Show [0,1] is compact in ℝ.
10 marksShow worked answer
It is closed and bounded. By the Heine–Borel theorem, it is compact.
Question 20.5
A map f:X→Y is a homeomorphism. Which properties does it preserve?
8 marksShow worked answer
It preserves topological properties such as compactness, connectedness and continuity structure. It need not preserve distances, angles or geometric lengths.
Question 20.6
For T(x)= (x+2)/3 on ℝ, find its fixed point and contraction factor.
10 marksShow worked answer
|T(x)−T(y)|=|x−y|/3, so q=1/3. Fixed point solves x=(x+2)/3, hence 3x=x+2 and x=1.
Undergraduate · Chapter 21
Chapter 21 — Geometry and Differential Geometry
Euclidean and affine geometry, curves, surfaces, curvature, manifolds and geodesics. Geometry studies spatial structure, while differential geometry uses calculus to investigate smooth curves, surfaces and manifolds. It connects local differential quantities such as curvature with global shape and physical applications.
Learning outcomes
By the end of this chapter, students should be able to define the principal concepts, state and apply relevant results with their assumptions, derive mathematical conclusions, solve multi-step problems and interpret results in their disciplinary context.
21.1 Euclidean and affine geometry
Euclidean geometry studies distances and angles under an inner product. Affine geometry retains incidence, collinearity and ratios along lines but not a preferred metric. Coordinate methods express geometric conditions as algebraic equations.
21.2 Vectors, inner products and transformations
The Euclidean inner product gives length and angle: cosθ=(u·v)/(||u||||v||). Orthogonal transformations preserve inner products and distances. Affine transformations preserve straight lines and ratios on a line.
21.3 Parametrised curves and arc length
A regular curve r(t) has r′(t)≠0. Its arc length is ∫||r′(t)||dt. Reparametrisation changes the parameter but not the geometric trace; arc-length parametrisation has unit speed.
21.4 Curvature and torsion
For a unit-speed space curve, curvature κ=||r″(s)||. For general parameter t, κ=||r′×r″||/||r′||³. Torsion measures departure from planarity and, when defined, τ=−B′·N.
21.5 Frenet–Serret frame
For a regular curve with nonzero curvature, tangent T, normal N and binormal B form an orthonormal frame. Derivatives with respect to arc length satisfy T′=κN, N′=−κT+τB, B′=−τN.
21.6 Surfaces and tangent planes
A regular surface may be parametrised r(u,v) with rᵤ×rᵥ≠0. Tangent vectors span its tangent plane; the normal is proportional to rᵤ×rᵥ. Implicit surfaces F(x,y,z)=0 have normal ∇F when nonzero.
21.7 First and second fundamental forms
The first fundamental form measures intrinsic lengths on a surface: E du²+2F du dv+G dv². The second fundamental form records normal curvature through coefficients e,f,g. Their combination defines principal curvatures and Gaussian curvature.
21.8 Gaussian and mean curvature
Principal curvatures k₁,k₂ yield Gaussian curvature K=k₁k₂ and mean curvature H=(k₁+k₂)/2. K is intrinsic by Gauss's Theorema Egregium; H depends on embedding and orientation convention.
21.9 Geodesics and shortest paths
Geodesics are curves with zero tangential acceleration, equivalently locally stationary length under suitable conditions. On a plane they are straight lines; on a sphere they are great circles. A geodesic need not be globally shortest between distant points.
21.10 Manifolds and coordinate charts
An n-dimensional smooth manifold is locally modelled on open subsets of ℝⁿ with compatible charts. Tangent spaces collect velocity directions; smooth maps have differentials. Coordinate changes allow local calculus on globally curved spaces.
21.11 University assessment — original problems
Attempt each problem before revealing its worked solution. Marks reward mathematical reasoning, correct notation, justified steps and interpretation.
Question 21.1
Find the curvature of a circle of radius 5.
8 marksShow worked answer
For a circle, curvature is constant κ=1/R=1/5.
Question 21.2
Find the arc length of r(t)=(3cos t,3sin t), 0≤t≤π.
10 marksShow worked answer
r′=(-3sin t,3cos t), speed=3. Length=∫₀^π3dt=3π.
Question 21.3
Find the curvature of r(t)=(t,t²,0) at t=0.
8 marksShow worked answer
r′=(1,2t,0), r″=(0,2,0). At 0, ||r′×r″||=2 and ||r′||³=1, so κ=2.
Question 21.4
Find the tangent plane to x²+y²+z²=9 at (1,2,2).
10 marksShow worked answer
F=x²+y²+z²−9, ∇F=(2x,2y,2z). At the point normal=(2,4,4). Plane: 2(x−1)+4(y−2)+4(z−2)=0, or x+2y+2z=9.
Question 21.5
State the Gaussian curvature of a sphere of radius R.
8 marksShow worked answer
Both principal curvatures have magnitude 1/R, so K=1/R².
Question 21.6
Explain why great circles are geodesics on a sphere.
10 marksShow worked answer
A great circle is the intersection with a plane through the sphere's centre. Its curvature vector is directed toward the centre, normal to the sphere, so its tangential acceleration on the surface vanishes; hence it is a geodesic.
Undergraduate · Chapter 22
Chapter 22 — Applied Mathematics and Mathematical Physics
Conservation laws, continuum models, waves, diffusion, potential theory, quantum and computational methods. Applied mathematics and mathematical physics use mathematical structures to describe physical systems. This chapter integrates differential equations, vector calculus, linear algebra, probability and numerical methods with conservation principles and physical interpretation.
Learning outcomes
By the end of this chapter, students should be able to define the principal concepts, state and apply relevant results with their assumptions, derive mathematical conclusions, solve multi-step problems and interpret results in their disciplinary context.
22.1 Modelling principles and conservation laws
A conservation law expresses rate of change of a conserved quantity as net flux plus sources. Integral balances can be converted to local differential equations using divergence theorems when fields are sufficiently smooth.
22.2 Continuum mechanics and balance equations
Continuum models represent matter by fields such as density, velocity and stress. Mass conservation yields a continuity equation; momentum balance yields equations involving stress divergence and body forces. Constitutive laws close the system.
22.3 Classical wave propagation
The wave equation uₜₜ=c²Δu models small-amplitude waves in idealised media. Wave speed depends on material parameters. Boundary conditions and initial displacement and velocity determine modal amplitudes.
22.4 Diffusion and heat transfer
The diffusion equation uₜ=κΔu describes smoothing and transport down concentration or temperature gradients. Its solutions dissipate high-frequency modes. Boundary conditions represent insulation, fixed temperature or heat exchange.
22.5 Potential theory and electrostatics
An electrostatic potential V satisfies E=−∇V. In source-free regions ΔV=0; with charge density ρ, Poisson's equation is ΔV=−ρ/ε₀ under the usual electrostatic convention. Boundary conditions determine the potential.
22.6 Fluid mechanics and incompressibility
For an incompressible Newtonian fluid, ∇·u=0 and the Navier–Stokes equations balance inertia, pressure, viscosity and body forces. Reynolds number compares inertial and viscous effects and guides regime analysis.
22.7 Lagrangian and Hamiltonian mechanics
The Euler–Lagrange equation d/dt(∂L/∂q̇)−∂L/∂q=0 follows from stationary action. For L=T−V, it yields classical equations. Hamiltonian mechanics uses canonical coordinates and equations q̇=∂H/∂p, ṗ=−∂H/∂q.
22.8 Fourier and spectral methods in physics
Fourier transforms decompose fields into spatial frequencies. Differential operators become multiplication by frequency variables under suitable conventions, simplifying linear constant-coefficient PDEs. Spectral methods can achieve rapid convergence for smooth solutions.
22.9 Quantum and mathematical operator models
In elementary quantum mechanics, states are vectors in a Hilbert space and observables are represented by self-adjoint operators. The time-dependent Schrödinger equation is iℏψₜ=Hψ. Boundary conditions and operator domains are mathematically significant.
22.10 Asymptotic methods and computational physics
Perturbation methods approximate solutions when a dimensionless parameter is small or large. Finite difference, finite element and spectral methods discretise governing equations. Convergence, stability, consistency and conservation properties guide numerical assessment.
22.11 University assessment — original problems
Attempt each problem before revealing its worked solution. Marks reward mathematical reasoning, correct notation, justified steps and interpretation.
Question 22.1
A one-dimensional wave travels with tension T=80 N and linear density μ=0.02 kg/m. Find its speed.
8 marksShow worked answer
For a stretched string c=√(T/μ)=√(80/0.02)=√4000≈63.25 m/s.
Question 22.2
For a one-dimensional heat mode with κ=0.5 and wave number k=4, find its temporal decay factor.
10 marksShow worked answer
A Fourier mode decays as e^(−κk²t)=e^(−0.5·16t)=e^(−8t).
Question 22.3
If electrostatic potential is V(x,y,z)=x²+y²+z², find the electric field.
8 marksShow worked answer
E=−∇V=−(2x,2y,2z).
Question 22.4
For an incompressible velocity field u=(2x,−2y,0), verify mass conservation.
10 marksShow worked answer
∇·u=∂(2x)/∂x+∂(−2y)/∂y+0=2−2=0. The field is divergence-free.
Question 22.5
For a one-dimensional harmonic oscillator with L=(1/2)m q̇²−(1/2)kq², derive its equation of motion.
8 marksShow worked answer
∂L/∂q̇=mq̇, so d/dt=mq̈. Also ∂L/∂q=−kq. Euler–Lagrange gives mq̈+kq=0.
Question 22.6
A particle has Hamiltonian H=p²/(2m)+V(q). State Hamilton's equations.
10 marksShow worked answer
q̇=∂H/∂p=p/m; ṗ=−∂H/∂q=−V′(q). Combining gives m q̈=−V′(q), Newton's equation.
Sample University Mathematics Examinations — Years 1 to 4
Eight original undergraduate examination papers, two per academic year. Each paper has exactly 30 questions and a maximum of 70 marks, with dedicated ruled answer areas.
General candidate instructions: Answer all questions. Show complete working and mathematical reasoning; state any theorem or assumptions used. Each paper contains twenty 2-mark questions and ten 3-mark questions (70 marks total). Use the answer spaces; attach and label extra sheets where necessary. Calculators may be used only as permitted by the examining institution.
Foundations, proof, algebra and introductory analysis
Time: 3 HoursMaximum: 70 Marks30 Questions
Instructions: Answer all questions. Give exact answers unless otherwise stated. Diagrams are not necessarily to scale. Unsupported answers may receive limited credit. Clearly number all workings.
Question 1. Prove by induction that 1+2+...+n=n(n+1)/2.
2 marks
Question 2. Negate: for every ε>0 there exists δ>0 such that...
2 marks
Question 3. Determine whether a relation defined by congruence modulo 5 is an equivalence relation.
2 marks
Question 4. Find the domain and range of f(x)=√(9−x²).
2 marks
Question 5. Prove that √3 is irrational.
2 marks
Question 6. Find the inverse of f(x)=(3x−2)/5.
2 marks
Question 7. For A={1,2,3,4}, B={3,4,5}, find union, intersection and difference.
2 marks
Question 8. Test validity of p→q, q→r, p ⊢ r using a truth table.
2 marks
Question 9. Determine whether f:Z→Z, f(n)=2n+1 is injective and surjective.
2 marks
Question 10. Prove that the sum of two odd integers is even.
2 marks
Question 11. Find the number of subsets of a 7-element set.
2 marks
Question 12. Find gcd(252,198) using the Euclidean algorithm.
2 marks
Question 13. Prove that if n² is even then n is even.
2 marks
Question 14. Determine whether ≤ on R is reflexive, symmetric and transitive.
2 marks
Question 15. Find g∘f for f(x)=2x+1 and g(x)=x².
2 marks
Question 16. Show that the union of two finite sets is finite.
2 marks
Question 17. Find all solutions of x²≡1 (mod 8).
2 marks
Question 18. Determine the power set of {a,b,c}.
2 marks
Question 19. Prove that the product of two rational numbers is rational.
2 marks
Question 20. Find the equivalence classes modulo 4.
2 marks
Question 21. Explain necessary and sufficient conditions with an example.
3 marks
Question 22. Prove that there is no greatest integer.
3 marks
Question 23. Express (p∧q)→r using ¬ and ∨ only.
3 marks
Question 24. Find A×B for A={a,b}, B={0,1,2}.
3 marks
Question 25. Determine whether the multiples of 3 form a subgroup of (Z,+).
3 marks
Question 26. Show that the intersection of two subgroups is a subgroup.
3 marks
Question 27. Find the composition of two specified bijections and its inverse.
3 marks
Question 28. Distinguish countable and uncountable sets with examples.
3 marks
Question 29. Prove that congruence modulo n is an equivalence relation.
3 marks
Question 30. State and prove a theorem about divisibility of integers.
3 marks
TOTAL: 70 MARKS
END OF PAPER
CBE NEXUS — First Year — Paper 2
UNDERGRADUATE MATHEMATICS
Calculus, sequences, series and linear algebra
Time: 3 HoursMaximum: 70 Marks30 Questions
Instructions: Answer all questions. Give exact answers unless otherwise stated. Diagrams are not necessarily to scale. Unsupported answers may receive limited credit. Clearly number all workings.
Question 1. Find gcd(414,276) using the Euclidean algorithm.
2 marks
Question 2. Prove that if n² is even then n is even.
2 marks
Question 3. Determine whether ≤ on R is reflexive, symmetric and transitive.
2 marks
Question 4. Find g∘f for f(x)=2x+1 and g(x)=x².
2 marks
Question 5. Show that the union of two finite sets is finite.
2 marks
Question 6. Find all solutions of x²≡1 (mod 8).
2 marks
Question 7. Determine the power set of {a,b,c}.
2 marks
Question 8. Prove that the product of two rational numbers is rational.
2 marks
Question 9. Find the equivalence classes modulo 4.
2 marks
Question 10. Explain necessary and sufficient conditions with an example.
2 marks
Question 11. Prove that there is no greatest integer.
2 marks
Question 12. Express (p∧q)→r using ¬ and ∨ only.
2 marks
Question 13. Find A×B for A={a,b}, B={0,1,2}.
2 marks
Question 14. Determine whether the multiples of 3 form a subgroup of (Z,+).
2 marks
Question 15. Show that the intersection of two subgroups is a subgroup.
2 marks
Question 16. Find the composition of two specified bijections and its inverse.
2 marks
Question 17. Distinguish countable and uncountable sets with examples.
2 marks
Question 18. Prove that congruence modulo n is an equivalence relation.
2 marks
Question 19. State and prove a theorem about divisibility of integers.
2 marks
Question 20. Prove by induction that 1+2+...+n=n(n+1)/2.
2 marks
Question 21. Negate: for every ε>0 there exists δ>0 such that...
3 marks
Question 22. Determine whether a relation defined by congruence modulo 5 is an equivalence relation.
3 marks
Question 23. Find the domain and range of f(x)=√(9−x²).
3 marks
Question 24. Prove that √3 is irrational.
3 marks
Question 25. Find the inverse of f(x)=(3x−2)/5.
3 marks
Question 26. For A={1,2,3,4}, B={3,4,5}, find union, intersection and difference.
3 marks
Question 27. Test validity of p→q, q→r, p ⊢ r using a truth table.
3 marks
Question 28. Determine whether f:Z→Z, f(n)=2n+1 is injective and surjective.
3 marks
Question 29. Prove that the sum of two odd integers is even.
3 marks
Question 30. Find the number of subsets of a 7-element set.
3 marks
TOTAL: 70 MARKS
END OF PAPER
CBE NEXUS — Second Year — Paper 1
UNDERGRADUATE MATHEMATICS
Multivariable calculus, ODEs and vector calculus
Time: 3 HoursMaximum: 70 Marks30 Questions
Instructions: Answer all questions. Give exact answers unless otherwise stated. Diagrams are not necessarily to scale. Unsupported answers may receive limited credit. Clearly number all workings.
Question 1. Evaluate lim(x→0) sin(5x)/(2x).
2 marks
Question 2. Differentiate y=x²e^(3x).
2 marks
Question 3. Find and classify stationary points of f(x)=x³−6x²+9x.
2 marks
Question 4. Evaluate ∫(3x²−4x+1)dx.
2 marks
Question 5. Evaluate ∫₀¹ xe^(x²)dx.
2 marks
Question 6. Test convergence of Σ1/n².
2 marks
Question 7. Find the degree-3 Taylor polynomial for e^x at zero.
2 marks
Question 8. Evaluate lim(x→∞)(3x²−x+1)/(2x²+5).
2 marks
Question 9. Use first principles to differentiate x².
2 marks
Question 10. Find dy/dx if x²+xy+y²=7.
2 marks
Question 11. Evaluate ∫x ln x dx.
2 marks
Question 12. Test continuity of f(x)=x sin(1/x), f(0)=0.
2 marks
Question 13. Apply the mean value theorem to x² on [1,3].
2 marks
Question 14. Classify Σ(−1)^(n+1)/n as absolute, conditional or divergent.
2 marks
Question 15. Find radius and interval of convergence of Σx^n/n.
2 marks
Question 16. Evaluate ∫dx/(x²+4).
2 marks
Question 17. Classify extrema of f(x)=x⁴−4x².
2 marks
Question 18. Evaluate lim(x→0)(e^(2x)−1)/x.
2 marks
Question 19. Use integration by parts to evaluate ∫₀¹xe^x dx.
2 marks
Question 20. Determine convergence of ∫₁∞ dx/x^p.
2 marks
Question 21. Find the second derivative of ln(1+x²).
3 marks
Question 22. Approximate √1.04 by linearisation at 1.
3 marks
Question 23. Evaluate ∫₀^π x sin x dx.
3 marks
Question 24. Sum the geometric series Σ(3/4)^n, n≥0.
3 marks
Question 25. Find intervals of increase of x³−3x.
3 marks
Question 26. Evaluate lim(x→0)(1−cos4x)/x².
3 marks
Question 27. Find the Maclaurin polynomial for cos x through degree 4.
3 marks
Question 28. Apply the ratio test to Σ n!/n^n.
3 marks
Question 29. Evaluate ∫₀¹ dx/√x and classify the improper integral.
3 marks
Question 30. Apply the fundamental theorem to d/dx ∫₁ˣ ln(1+t²)dt.
3 marks
TOTAL: 70 MARKS
END OF PAPER
CBE NEXUS — Second Year — Paper 2
UNDERGRADUATE MATHEMATICS
Complex variables, PDEs, algebra and discrete mathematics
Time: 3 HoursMaximum: 70 Marks30 Questions
Instructions: Answer all questions. Give exact answers unless otherwise stated. Diagrams are not necessarily to scale. Unsupported answers may receive limited credit. Clearly number all workings.
Question 1. Test continuity of f(x)=x sin(1/x), f(0)=0.
2 marks
Question 2. Apply the mean value theorem to x² on [1,3].
2 marks
Question 3. Classify Σ(−1)^(n+1)/n as absolute, conditional or divergent.
2 marks
Question 4. Find radius and interval of convergence of Σx^n/n.
2 marks
Question 5. Evaluate ∫dx/(x²+4).
2 marks
Question 6. Classify extrema of f(x)=x⁴−4x².
2 marks
Question 7. Evaluate lim(x→0)(e^(2x)−1)/x.
2 marks
Question 8. Use integration by parts to evaluate ∫₀¹xe^x dx.
2 marks
Question 9. Determine convergence of ∫₁∞ dx/x^p.
2 marks
Question 10. Find the second derivative of ln(1+x²).
2 marks
Question 11. Approximate √1.04 by linearisation at 1.
2 marks
Question 12. Evaluate ∫₀^π x sin x dx.
2 marks
Question 13. Sum the geometric series Σ(3/4)^n, n≥0.
2 marks
Question 14. Find intervals of increase of x³−3x.
2 marks
Question 15. Evaluate lim(x→0)(1−cos4x)/x².
2 marks
Question 16. Find the Maclaurin polynomial for cos x through degree 4.
2 marks
Question 17. Apply the ratio test to Σ n!/n^n.
2 marks
Question 18. Evaluate ∫₀¹ dx/√x and classify the improper integral.
2 marks
Question 19. Apply the fundamental theorem to d/dx ∫₁ˣ ln(1+t²)dt.
2 marks
Question 20. Evaluate lim(x→0) sin(5x)/(2x).
2 marks
Question 21. Differentiate y=x²e^(3x).
3 marks
Question 22. Find and classify stationary points of f(x)=x³−6x²+9x.
3 marks
Question 23. Evaluate ∫(3x²−4x+1)dx.
3 marks
Question 24. Evaluate ∫₀¹ xe^(x²)dx.
3 marks
Question 25. Test convergence of Σ1/n².
3 marks
Question 26. Find the degree-3 Taylor polynomial for e^x at zero.
3 marks
Question 27. Evaluate lim(x→∞)(3x²−x+1)/(2x²+5).
3 marks
Question 28. Use first principles to differentiate x².
3 marks
Question 29. Find dy/dx if x²+xy+y²=7.
3 marks
Question 30. Evaluate ∫x ln x dx.
3 marks
TOTAL: 70 MARKS
END OF PAPER
CBE NEXUS — Third Year — Paper 1
UNDERGRADUATE MATHEMATICS
Real analysis, probability, statistics and numerical methods
Time: 3 HoursMaximum: 70 Marks30 Questions
Instructions: Answer all questions. Give exact answers unless otherwise stated. Diagrams are not necessarily to scale. Unsupported answers may receive limited credit. Clearly number all workings.
Question 1. Find gradient and directional derivative of f=x²y+yz² at (1,2,1) along (2,1,2).
2 marks
Question 2. Classify critical points of f=x²+y²−4x+6y.
2 marks
Question 3. Evaluate ∬R(x+y)dA on 0≤x≤1, 0≤y≤2.
2 marks
Question 4. Use Lagrange multipliers to maximise xy subject to x²+y²=1, x,y≥0.
2 marks
Question 5. Find eigenvalues and eigenvectors of [[2,1],[1,2]].
2 marks
Question 6. Test independence of (1,0,1),(0,1,1),(1,1,2).
2 marks
Question 7. Solve y'+2y=e^(−x), y(0)=1.
2 marks
Question 8. Solve y''−3y'+2y=0.
2 marks
Question 9. Compute divergence and curl of F=(x²y,yz,zx).
2 marks
Question 10. Evaluate ∮F·dr for F=(−y,x) around unit circle.
2 marks
Question 11. Find determinant and rank of [[1,2,3],[2,4,6],[1,1,1]].
2 marks
Question 12. Diagonalise [[4,1],[0,2]].
2 marks
Question 13. Find Fourier coefficients of f(x)=x on (−π,π).
2 marks
Question 14. Classify u_t=4u_xx and identify its standard type.
2 marks
Question 15. Solve u_x+u_y=0 by characteristics.
2 marks
Question 16. Find residue of (z+1)/(z²(z−1)) at z=0.
2 marks
Question 17. Evaluate ∮|z|=2 dz/(z−1).
2 marks
Question 18. Find gcd(414,276) and express it as a linear combination.
2 marks
Question 19. Find shortest path in a weighted graph with stated edge weights.
2 marks
Question 20. Compute expectation and variance of Bernoulli(p).
2 marks
Question 21. Find MLE of θ for density θe^(−θx), x≥0.
3 marks
Question 22. Perform two Newton iterations for x²−2=0 from x₀=1.5.
3 marks
Question 23. Approximate ∫₀²x²dx by trapezoidal rule with n=2.
3 marks
Question 24. Formulate an LP maximising 5x+4y subject to 2x+y≤8, x+2y≤8.
3 marks
Question 25. Find least-squares line for (0,1),(1,2),(2,2).
3 marks
Question 26. Determine completeness of R with its usual metric.
3 marks
Question 27. Find curvature of r(t)=(a cos t,a sin t,bt).
3 marks
Question 28. Use divergence theorem for F=(x,y,z) on sphere radius a.
3 marks
Question 29. Derive Euler–Lagrange equation for L=½y'²−V(y).
3 marks
Question 30. State and prove a result concerning eigenvalues of a real symmetric matrix.
3 marks
TOTAL: 70 MARKS
END OF PAPER
CBE NEXUS — Third Year — Paper 2
UNDERGRADUATE MATHEMATICS
Algebra, graph theory, complex analysis and modelling
Time: 3 HoursMaximum: 70 Marks30 Questions
Instructions: Answer all questions. Give exact answers unless otherwise stated. Diagrams are not necessarily to scale. Unsupported answers may receive limited credit. Clearly number all workings.
Question 1. Diagonalise [[4,1],[0,2]].
2 marks
Question 2. Find Fourier coefficients of f(x)=x on (−π,π).
2 marks
Question 3. Classify u_t=4u_xx and identify its standard type.
2 marks
Question 4. Solve u_x+u_y=0 by characteristics.
2 marks
Question 5. Find residue of (z+1)/(z²(z−1)) at z=0.
2 marks
Question 6. Evaluate ∮|z|=2 dz/(z−1).
2 marks
Question 7. Find gcd(414,276) and express it as a linear combination.
2 marks
Question 8. Find shortest path in a weighted graph with stated edge weights.
2 marks
Question 9. Compute expectation and variance of Bernoulli(p).
2 marks
Question 10. Find MLE of θ for density θe^(−θx), x≥0.
2 marks
Question 11. Perform two Newton iterations for x²−3=0 from x₀=2.
2 marks
Question 12. Approximate ∫₀²x²dx by trapezoidal rule with n=2.
2 marks
Question 13. Formulate an LP maximising 5x+4y subject to 2x+y≤8, x+2y≤8.
2 marks
Question 14. Find least-squares line for (0,1),(1,2),(2,2).
2 marks
Question 15. Determine completeness of R with its usual metric.
2 marks
Question 16. Find curvature of r(t)=(a cos t,a sin t,bt).
2 marks
Question 17. Use divergence theorem for F=(x,y,z) on sphere radius a.
2 marks
Question 18. Derive Euler–Lagrange equation for L=½y'²−V(y).
2 marks
Question 19. State and prove a result concerning eigenvalues of a real symmetric matrix.
2 marks
Question 20. Find gradient and directional derivative of f=x²y+yz² at (1,2,1) along (2,1,2).
2 marks
Question 21. Classify critical points of f=x²+y²−4x+6y.
3 marks
Question 22. Evaluate ∬R(x+y)dA on 0≤x≤1, 0≤y≤2.
3 marks
Question 23. Use Lagrange multipliers to maximise xy subject to x²+y²=1, x,y≥0.
3 marks
Question 24. Find eigenvalues and eigenvectors of [[2,1],[1,2]].
3 marks
Question 25. Test independence of (1,0,1),(0,1,1),(1,1,2).
3 marks
Question 26. Solve y'+2y=e^(−x), y(0)=1.
3 marks
Question 27. Solve y''−3y'+2y=0.
3 marks
Question 28. Compute divergence and curl of F=(x²y,yz,zx).
3 marks
Question 29. Evaluate ∮F·dr for F=(−y,x) around circle of radius 2.
3 marks
Question 30. Find determinant and rank of [[1,2,3],[2,4,6],[1,1,1]].
3 marks
TOTAL: 70 MARKS
END OF PAPER
CBE NEXUS — Fourth Year — Paper 1
UNDERGRADUATE MATHEMATICS
Advanced analysis, topology, geometry and applied mathematics
Time: 3 HoursMaximum: 70 Marks30 Questions
Instructions: Answer all questions. Give exact answers unless otherwise stated. Diagrams are not necessarily to scale. Unsupported answers may receive limited credit. Clearly number all workings.
Question 1. Prove uniform limit of continuous functions on a compact interval is continuous.
2 marks
Question 2. Prove every bounded real sequence has a convergent subsequence.
2 marks
Question 3. Determine uniform convergence of f_n(x)=x^n on [0,1].
2 marks
Question 4. Evaluate ∫₀∞e^(−ax)dx for a>0.
2 marks
Question 5. Prove continuous functions on compact metric spaces are uniformly continuous.
2 marks
Question 6. Find Laurent expansion of 1/[z(z−1)] about zero and annulus.
2 marks
Question 7. Evaluate ∮|z|=2 e^z/z³ dz.
2 marks
Question 8. Classify groups of order 6 up to isomorphism, with justification.
2 marks
Question 9. Prove every subgroup of a cyclic group is cyclic.
2 marks
Question 10. Solve simultaneous congruences using the Chinese remainder theorem.
2 marks
Question 11. Prove there are infinitely many primes congruent to 1 modulo 4.
2 marks
Question 12. Derive inclusion–exclusion for three finite sets.
2 marks
Question 13. Use generating functions to count compositions using 1- and 2-unit steps.
2 marks
Question 14. Prove a connected graph on n vertices has at least n−1 edges.
2 marks
Question 15. Find chromatic polynomial of C₄.
2 marks
Question 16. Derive normal approximation to binomial and continuity correction.
2 marks
Question 17. Find covariance of independent exponential random variables.
2 marks
Question 18. Derive likelihood equations for normal model with unknown mean and variance.
2 marks
Question 19. Construct a 95% confidence interval for a mean when σ is known.
2 marks
Question 20. State Gauss–Markov assumptions and explain BLUE property.
2 marks
Question 21. Prove quadratic local convergence of Newton's method near a simple root.
Question 23. Establish explicit Euler stability condition for y'=λy.
3 marks
Question 24. Derive KKT conditions for differentiable constrained optimisation.
3 marks
Question 25. Formulate the dual of a specified primal linear programme.
3 marks
Question 26. Derive diffusion equation from conservation and Fick's law.
3 marks
Question 27. Use dimensional analysis to derive dimensionless drag parameters.
3 marks
Question 28. Prove contraction mapping theorem and apply it to an integral equation.
3 marks
Question 29. Compute Gaussian curvature of a surface of revolution.
3 marks
Question 30. Derive Euler–Lagrange equations for a two-coordinate mechanical system.
3 marks
TOTAL: 70 MARKS
END OF PAPER
CBE NEXUS — Fourth Year — Paper 2
UNDERGRADUATE MATHEMATICS
Advanced statistics, numerical analysis, optimisation and mathematical physics
Time: 3 HoursMaximum: 70 Marks30 Questions
Instructions: Answer all questions. Give exact answers unless otherwise stated. Diagrams are not necessarily to scale. Unsupported answers may receive limited credit. Clearly number all workings.
Question 1. Derive inclusion–exclusion for three finite sets.
2 marks
Question 2. Use generating functions to count compositions using 1- and 2-unit steps.
2 marks
Question 3. Prove a connected graph on n vertices has at least n−1 edges.
2 marks
Question 4. Find chromatic polynomial of C₄.
2 marks
Question 5. Derive normal approximation to binomial and continuity correction.
2 marks
Question 6. Find covariance of independent exponential random variables.
2 marks
Question 7. Derive likelihood equations for normal model with unknown mean and variance.
2 marks
Question 8. Construct a 95% confidence interval for a mean when σ is known.
2 marks
Question 9. State Gauss–Markov assumptions and explain BLUE property.
2 marks
Question 10. Prove quadratic local convergence of Newton's method near a simple root.
Question 12. Establish explicit Euler stability condition for y'=λy.
2 marks
Question 13. Derive KKT conditions for differentiable constrained optimisation.
2 marks
Question 14. Formulate the dual of a specified primal linear programme.
2 marks
Question 15. Derive diffusion equation from conservation and Fick's law.
2 marks
Question 16. Use dimensional analysis to derive dimensionless drag parameters.
2 marks
Question 17. Prove contraction mapping theorem and apply it to an integral equation.
2 marks
Question 18. Compute Gaussian curvature of a surface of revolution.
2 marks
Question 19. Derive Euler–Lagrange equations for a two-coordinate mechanical system.
2 marks
Question 20. Prove uniform limit of continuous functions on a compact interval is continuous.
2 marks
Question 21. Prove every bounded real sequence has a convergent subsequence.
3 marks
Question 22. Determine uniform convergence of f_n(x)=x^n on [0,1].
3 marks
Question 23. Evaluate ∫₀∞e^(−ax)dx for a>0.
3 marks
Question 24. Prove continuous functions on compact metric spaces are uniformly continuous.
3 marks
Question 25. Find Laurent expansion of 1/[z(z−1)] about zero and annulus.
3 marks
Question 26. Evaluate ∮|z|=2 e^z/z³ dz.
3 marks
Question 27. Classify groups of order 6 up to isomorphism, with justification.
3 marks
Question 28. Prove every subgroup of a cyclic group is cyclic.
3 marks
Question 29. Solve simultaneous congruences using the Chinese remainder theorem.
3 marks
Question 30. Prove there are infinitely many primes congruent to 1 modulo 4.
3 marks
TOTAL: 70 MARKS
END OF PAPER
These are original sample practice papers, not official examinations issued by a university. Before formal adoption, departments should map each paper to their approved year-specific course outline and assessment regulations.
University reference materials
These are supplementary open university materials. The CBE Nexus exercises above are independently written and are not reproduced examination questions.
Reference links are supplementary and do not imply affiliation or endorsement. Follow the syllabus and notation prescribed by your university.
Full undergraduate University Mathematics pathway
This chapter establishes the language and proof techniques used throughout the 22-chapter programme. The remaining chapters are retained in the official sequence and require their own full lecture notes and assessments.