Notes on Diffy Qs: Differential Equations for Engineers Study Guide and Chapter Summaries – Jiri Lebl (6th Edition) BS Math

BS Mathematics and engineering students taking a first course in Differential Equations can use this guide to study Notes on Diffy Qs: Differential Equations for Engineers by Jiri Lebl, under a dual Creative Commons licence. It is a complete course in ordinary differential equations that also reaches Fourier series, partial differential equations, the Laplace transform, power series, nonlinear systems and chaos, and it carries its own linear algebra appendix, so no separate prerequisite book is needed for the systems chapter.

The book grew out of Math 286 lecture notes at the University of Illinois and is written as a drop-in replacement for the big Edwards–Penney and Boyce–DiPrima texts. For exam preparation it covers the techniques papers actually ask for: separable and linear equations, constant-coefficient and nonhomogeneous equations, the eigenvalue method for systems, Fourier series with the heat and wave equations, and Laplace transforms. More than 700 numbered exercises, with answers printed for those numbered 101 and up, make it easy to practise before the test.

Book Overview

CourseDifferential Equations — a first course in ordinary differential equations with an introduction to PDEs, aimed at engineering students
Degree ProgramsBS Mathematics and engineering degrees — typically taken after the basic calculus sequence
LevelUniversity — lower-division (second-year) level
EditionEdition 6, version 6.11 — May 18, 2026 (466 pages); the author issues corrected minor versions once or twice a year
AuthorJiří Lebl, Oklahoma State University (the book began as his class notes for Math 286 at the University of Illinois at Urbana-Champaign in 2008–2009)
StructureIntroduction plus 8 chapters — First-order equations, Higher-order linear ODEs, Systems of ODEs, Fourier series and PDEs, More on eigenvalue problems, The Laplace transform, Power-series methods, Nonlinear systems — and two appendices (Linear algebra; Table of Laplace transforms)
ExercisesMore than 700 numbered exercises; answers or solutions to those numbered 101 and higher are printed in the back of the book
LanguageEnglish
LicenseDual licensed — Creative Commons Attribution-NonCommercial-ShareAlike 4.0 International (CC BY-NC-SA 4.0) or Creative Commons Attribution-ShareAlike 4.0 International (CC BY-SA 4.0), whichever suits your use
FormatFree PDF (466 pages, about 4 MB), a browsable HTML web version and the LaTeX source on GitHub; Sage demos, WeBWorK homework and video lectures on the author’s site

Chapter List

Introduction: Differential Equations and Their Classification

Difficulty: Easy · Key topics: ODE vs PDE, order, linear and nonlinear, homogeneous, constant coefficients, autonomous, general and particular solutions, mathematical modelling

The Introduction does three jobs. Section 0.1 is a note for instructors: the book grew out of Math 286 lecture notes at the University of Illinois, and the section shows which chapters depend on which and lists sample course plans. Section 0.2 defines a differential equation, checks by substitution that x = cos t + sin t + Ce^(-t) solves dx/dt + x = 2 cos t, walks through the whole modelling loop on bacteria growth, and lists four fundamental equations worth memorising. Section 0.3 sorts equations by ODE or PDE, order, linearity, homogeneity, constant coefficients and autonomy.

Key Points:

  • An ODE has one independent variable; a PDE has several and uses partial derivatives
  • The order is the order of the highest derivative that appears, so a PDE with a fourth x-derivative and a second t-derivative is fourth order
  • An equation is linear when the unknown and its derivatives appear only to the first power, are never multiplied together and never sit inside another function
  • A general solution carries arbitrary constants (as many as the order); an initial condition picks one particular solution
  • Four fundamental equations: y' = ky gives Ce^(kx); y' = -ky gives Ce^(-kx); y'' = -k^2 y gives C1 cos(kx) + C2 sin(kx); y'' = k^2 y gives C1 e^(kx) + C2 e^(-kx)
  • An autonomous equation has no explicit dependence on the independent variable, usually time

Memory Tip: Keep the book’s modelling loop in mind: real-world problem, simplify into a model, solve it, then interpret the answer. A solution is only as trustworthy as the assumptions you made in step two.

Common Mistake: Calling an equation linear because it “looks simple.” The equation x’ = x^2 has one tiny term, but the unknown is squared, so it is nonlinear (and its solutions blow up in finite time).

Important Questions:

  • What is the difference between an ODE and a PDE? An ODE has derivatives with respect to a single independent variable; a PDE involves partial derivatives with respect to several variables, like the heat equation u_t = u_xx.
  • Why does a second-order ODE need two initial conditions? Its general solution has two arbitrary constants (y” = 0 gives y = C1 x + C2), so two conditions, usually y(x0) and y'(x0), are needed to fix one solution.

Chapter 1: First-Order Equations

Difficulty: Easy · Key topics: integrals as solutions, slope fields, Picard’s theorem, separable equations, integrating factor, Bernoulli equations, autonomous equations, Euler’s method, exact equations, transport equation

This chapter builds a toolbox for y’ = f(x,y). It starts with the case where plain integration solves the equation, then draws slope fields and states Picard’s theorem on existence and uniqueness, showing why y’ = 1/x and a square-root right-hand side misbehave. Next come separable equations (with implicit solutions), linear equations solved by an integrating factor, and substitutions including Bernoulli and homogeneous equations. Autonomous equations introduce critical points and stability through the logistic model. The chapter closes with Euler’s method, exact equations with a potential function, and a first PDE solved by characteristics.

Key Points:

  • Picard’s theorem: if f and its y-derivative are continuous near (x0, y0), a solution exists and is unique near that point, though it may exist only on a short interval (y’ = y^2 blows up)
  • Separable form y' = f(x)g(y): separate, integrate both sides, and check separately for constant solutions where g(y) = 0
  • First-order linear y' + p(x)y = f(x): multiply by r(x) = e^(∫ p dx); any antiderivative of p works because the constant cancels
  • Bernoulli equation y' + p(x)y = q(x)y^n: the substitution v = y^(1-n) turns it into a linear equation
  • Autonomous x' = f(x): critical points (f = 0) give equilibrium solutions, stable or unstable depending on the sign of f on each side
  • Exact equation M dx + N dy = 0 is exact when M_y = N_x; a potential F with F_x = M and F_y = N gives the implicit solution F(x,y) = C
  • Euler’s method follows the slope step by step: y_(i+1) = y_i + h f(x_i, y_i); the chapter also mentions improved Euler and Runge-Kutta

Memory Tip: When you meet a first-order equation, ask in this order: can I just integrate, is it separable, is it linear? Only then reach for substitution or exactness. The cheapest method that works is the right one.

Common Mistake: Dividing by y (or g(y)) to separate variables and quietly losing the constant solution y = 0, which the general formula may not contain.

Important Questions:

  • Does y’ = 2√|y| with y(0) = 0 have a unique solution? No. The right-hand side is not differentiable in y at 0, so Picard’s theorem does not apply, and both y = 0 and y = x^2 (for x ≥ 0) solve the problem.
  • How do you solve y’ + p(x)y = f(x)? Multiply by the integrating factor r(x) = e^(∫ p dx) so the left side becomes (r y)’, integrate both sides, then divide by r(x).

Chapter 2: Higher-Order Linear ODEs

Difficulty: Medium · Key topics: superposition, characteristic equation, complex roots and Euler’s formula, linear independence, mechanical vibrations, undetermined coefficients, variation of parameters, resonance

The chapter starts with second-order linear equations, the superposition principle and the existence and uniqueness theorem, using operator notation Ly. For constant coefficients it tries y = e^(rx), gets the characteristic equation, handles complex roots through Euler’s formula, and extends everything to order n with a Wronskian-free treatment of linear independence. Section 2.4 models mass-spring systems and RLC circuits with free damped and undamped motion. Section 2.5 solves nonhomogeneous equations as complementary plus particular solutions, by undetermined coefficients and variation of parameters. Section 2.6 studies forced oscillation, resonance and the steady periodic solution.

Key Points:

  • Superposition: any linear combination C1 y1 + C2 y2 of solutions of a linear homogeneous equation is again a solution
  • For ay'' + by' + cy = 0 try e^(rx): real distinct roots give C1 e^(r1 x) + C2 e^(r2 x); complex roots α ± iβ give e^(αx)(C1 cos βx + C2 sin βx)
  • Euler’s formula: e^(iθ) = cos θ + i sin θ, which is what turns complex exponentials into real sines and cosines
  • Damped vibration is decided by the sign of c^2 - 4km: overdamped (positive), critically damped (zero), underdamped (negative, oscillating with a decaying envelope)
  • The general solution of Ly = f is y = y_c + y_p; apply initial conditions only after both parts are added
  • Undetermined coefficients works when f has finitely many independent derivatives; variation of parameters handles cases like tan x
  • With no damping and forcing frequency equal to the natural frequency, a t sin(ωt) term appears and the response grows without bound

Memory Tip: If any term in your guess for y_p already solves the homogeneous equation, multiply the whole guess by x (or t) until the clash disappears. That one rule explains why resonance produces a t in front of the sine.

Common Mistake: Solving for C1 and C2 using the complementary solution alone and adding the particular solution afterwards. The initial conditions belong to the full solution y_c + y_p.

Important Questions:

  • What are the three cases of damped free motion? Overdamped (c^2 > 4km): two real roots, no oscillation; critically damped (c^2 = 4km): one repeated root; underdamped (c^2 < 4km): complex roots, oscillation with exponentially shrinking amplitude.
  • When does resonance occur, and what does it look like? In an undamped system when the forcing frequency equals the natural frequency; the particular solution contains t sin(ωt), so the amplitude grows without bound. With damping the amplitude peaks but stays finite.

Chapter 3: Systems of ODEs

Difficulty: Hard · Key topics: first-order systems, matrices and determinants, eigenvalue method, phase portraits, second-order systems, repeated eigenvalues, matrix exponential, nonhomogeneous systems

Coupled unknowns, such as salt in two connected tanks or two carts joined by springs, lead to systems of ODEs. Section 3.1 shows how any n-th order equation becomes a first-order system and extends Picard’s theorem to systems. Section 3.2 reviews matrices, determinants, row reduction and inverses; 3.3 writes the linear system as x’ = Px + f. The eigenvalue method (3.4) tries v e^(λt) with real and complex eigenvalues, 3.5 reads phase portraits from eigenvalues, and 3.6 finds normal modes of mass-spring systems. Sections 3.7 to 3.9 cover repeated eigenvalues, the matrix exponential e^(tA) and nonhomogeneous systems.

Key Points:

  • An n-th order equation becomes a first-order system with u1 = y, u2 = y', ..., un = y^(n-1)
  • For x' = Px, solutions are v e^(λt) where λ is an eigenvalue of P found from det(P - λI) = 0 and v is a matching eigenvector
  • Complex eigenvalues of a real matrix come in conjugate pairs; the real and imaginary parts of one complex solution give two independent real solutions
  • Phase-plane types: real positive (source), real negative (sink), opposite signs (saddle), purely imaginary (center), complex with positive or negative real part (spiral source or sink)
  • A defective eigenvalue has fewer independent eigenvectors than its multiplicity, so extra solutions are built from vectors satisfying (A - λI)^k v = 0
  • The matrix exponential e^(tA) solves x’ = Ax with x(0) = c, and e^(A+B) = e^A e^B only when AB = BA
  • Variation of parameters for systems: x_p = X(t) ∫ X(t)^(-1) f(t) dt with X a fundamental matrix

Memory Tip: The eigenvalue method is Chapter 2’s characteristic-equation trick in matrix form. Trying e^(rx) there gave a polynomial in r; trying v e^(λt) here gives det(P – λI) = 0. Same idea, bigger algebra.

Common Mistake: Assuming that a repeated eigenvalue always supplies as many independent eigenvectors as its multiplicity. When it does not (a defect), the plain eigenvalue method stops one solution short.

Important Questions:

  • How do you turn y”’ + a y” + b y’ + c y = f into a first-order system? Set u1 = y, u2 = y’, u3 = y”; then u1′ = u2, u2′ = u3, and u3′ = f – a u3 – b u2 – c u1.
  • What does a defective eigenvalue mean, and how do you still solve the system? Its geometric multiplicity is smaller than its algebraic multiplicity, so there are too few eigenvectors. The matrix exponential e^(tA) or vectors with (A – λI)^k v = 0 provide the missing solutions.

Chapter 4: Fourier Series and PDEs

Difficulty: Hard · Key topics: boundary value problems, eigenfunctions, orthogonality, Fourier series, Gibbs phenomenon, sine and cosine series, separation of variables, heat equation, wave equation, d’Alembert’s solution, Laplace equation, Poisson kernel

The chapter begins with boundary value problems, which unlike initial value problems can have no solution or infinitely many, and with three eigenvalue problems (Dirichlet, Neumann, periodic) whose eigenfunctions are orthogonal. Fourier series then appear as an eigenfunction decomposition, with convergence at jumps, term-by-term calculus and sine and cosine series from odd and even extensions. Periodic forcing and resonance return as an application. The second half solves PDEs by separation of variables: the heat equation, the vibrating string, d’Alembert’s travelling-wave formula, steady-state temperature in a rectangle, and the Dirichlet problem in a circle with the Poisson kernel.

Key Points:

  • Eigenfunctions of X'' + λX = 0: Dirichlet (zero ends) gives sin(nπt/L); Neumann (zero slope) gives cos(nπt/L) plus a constant; periodic conditions give 1, cos and sin
  • Eigenfunctions for different eigenvalues are orthogonal under the inner product ∫ f g dt, just as eigenvectors of a symmetric matrix are
  • Fourier series of a 2L-periodic f: a0/2 + ∑ (a_n cos(nπt/L) + b_n sin(nπt/L)), with coefficients found by projecting f onto each eigenfunction
  • At a jump the series converges to the average of the left and right limits; near the jump partial sums overshoot (Gibbs phenomenon)
  • An odd extension gives a sine series (Dirichlet conditions); an even extension gives a cosine series (Neumann conditions)
  • Heat equation u_t = k u_xx (parabolic), wave equation y_tt = a^2 y_xx (hyperbolic), Laplace equation (elliptic); superposition keeps homogeneous side conditions
  • d’Alembert: y(x,t) = A(x - at) + B(x + at), two waves travelling in opposite directions, fitted using odd periodic extensions of the initial data

Memory Tip: Separation of variables is a four-beat recipe: assume u = X(x)T(t); split into two ODEs with a separation constant; let the boundary conditions pick X and the allowed values; then sum over n and match the initial data with a Fourier series.

Common Mistake: Using a sine series when the ends are insulated (Neumann). The boundary conditions decide whether you need the odd extension (sines) or the even extension (cosines and a constant term).

Important Questions:

  • What does a Fourier series converge to at a jump discontinuity? The average of the one-sided limits, (f(t-) + f(t+))/2; at points where f is continuous it converges to f(t) itself.
  • How is the heat equation solved on a wire with both ends held at zero? Separate u = X(x)T(t); X” + λX = 0 with X(0) = X(L) = 0 gives sin(nπx/L); each T_n decays like an exponential; expand the initial temperature in a sine series and sum the products.

Chapter 5: More on Eigenvalue Problems

Difficulty: Hard · Key topics: Sturm-Liouville problems, weight function, orthogonality, eigenfunction series, Fredholm alternative, vibrating beam, steady periodic solutions

Chapter 4’s eigenvalue problems turn out to be special cases of one framework, and Section 5.1 states it: the regular Sturm-Liouville problem (p y’)’ – q y + λ r y = 0 with p and r positive and mixed or Robin-type boundary conditions. The chapter shows that the eigenvalues form an increasing sequence tending to infinity, that eigenfunctions are orthogonal with weight r(x), and that a function can be expanded as a series in them. Section 5.2 treats a fourth-order beam equation with hinged ends, and Section 5.3 finds steady periodic solutions for a forced vibrating string and for seasonal temperature changes underground.

Key Points:

  • Regular Sturm-Liouville form: (p y')' - q y + λ r y = 0 with p > 0, r > 0 on the whole closed interval, including the endpoints
  • Eigenvalues satisfy λ1 < λ2 < ... with λn → ∞, one eigenfunction each up to a constant multiple
  • Eigenfunctions for different eigenvalues are orthogonal with respect to the weight: ∫ y_m y_n r dx = 0
  • Eigenfunction series: f = ∑ c_n y_n with c_n = <f, y_n> / <y_n, y_n>
  • The Fredholm alternative holds for regular problems: either a unique solution exists for every right-hand side, or the homogeneous problem has a nonzero solution
  • A steady periodic solution is independent of the initial conditions, and complex exponentials make finding it easier

Memory Tip: Think of this chapter as “symmetric matrices for functions.” Real, increasing eigenvalues, orthogonal eigenvectors and an expansion in the eigenbasis are exactly what you know from linear algebra, now with a weight r(x) in the inner product.

Common Mistake: Forgetting the weight function. Orthogonality and the coefficient formula c_n both use ∫ f y_n r dx, not the plain integral, so omitting r(x) gives wrong coefficients for Bessel-type problems.

Important Questions:

  • When is a Sturm-Liouville problem called regular? When p, p’, q and r are continuous on [a, b], p > 0 and r > 0 everywhere on the closed interval, q ≥ 0, and the boundary-condition constants are nonnegative and not both zero at either end.
  • How do you find the coefficients of an eigenfunction series? Use orthogonality: multiply f = ∑ c_n y_n by y_m r and integrate, giving c_m = <f, y_m> / <y_m, y_m> with the weighted inner product.

Chapter 6: The Laplace Transform

Difficulty: Medium · Key topics: Laplace transform, exponential order, inverse transform, transforms of derivatives, Heaviside function, transfer function, convolution, Dirac delta, impulse response, PDEs

The Laplace transform turns a differential equation into an algebraic one. Section 6.1 defines F(s) as the integral of e^(-st) f(t), states when it exists (functions of exponential order) and uses partial fractions for inverses. Section 6.2 shows that differentiation becomes multiplication by s, solves initial value problems this way, builds piecewise inputs with the Heaviside step, and introduces transfer functions, integral equations and periodic inputs. Section 6.3 defines convolution and the Volterra integral equation, Section 6.4 the Dirac delta and impulse response (with beam bending), and Section 6.5 applies the transform to PDEs.

Key Points:

  • Definition: F(s) = ∫(0 to ∞) e^(-st) f(t) dt; it exists for all large s when f is continuous and of exponential order
  • Derivatives become algebra: L{x'} = sX(s) - x(0) and L{x''} = s^2 X(s) - s x(0) - x'(0)
  • Heaviside step u(t - a) switches a function on at time a; its transform is e^(-as)/s
  • The transfer function is H(s) = 1/A(s) for Lx = f, so the output satisfies X(s) = H(s)F(s)
  • Convolution (f * g)(t) has transform F(s)G(s), giving a solution for any forcing function as an integral
  • The Dirac delta has transform L{δ(t - a)} = e^(-as) and models an instantaneous impulse; the response to it is the impulse response
  • Transform in one variable of a PDE in x and t and it becomes an ODE in the other variable

Memory Tip: Remember the pipeline: ODE in time, transform, solve an algebraic equation for X(s), partial fractions, inverse transform using the table in Appendix B. Every Laplace solution follows those five moves.

Common Mistake: Expecting the transform of a product to be the product of transforms. It is the convolution that matches the product: L{f * g} = F(s)G(s), while L{f g} is not F(s)G(s).

Important Questions:

  • How do you solve a linear constant-coefficient IVP with the Laplace transform? Transform both sides using the derivative rules with the initial conditions plugged in, solve the algebraic equation for X(s), then take the inverse transform, usually after partial fractions.
  • What is the impulse response and why does it matter? It is the solution of Lx = δ(t) with zero initial state. Convolving it with any input f gives the solution of Lx = f, so one response describes the system for every forcing.

Chapter 7: Power-Series Methods

Difficulty: Medium · Key topics: power series, radius of convergence, analytic functions, ordinary points, singular points, regular singular points, method of Frobenius, indicial equation, Bessel functions

When no closed-form trick works, this chapter assumes the solution is a power series and solves for its coefficients. Section 7.1 reviews convergence, the radius of convergence, the ratio and root tests, analytic functions and Taylor series, term-by-term differentiation, reindexing and series for rational functions. Section 7.2 solves y” + p y’ + q y = 0 near an ordinary point through a recurrence for the coefficients, with Airy’s and Hermite’s equations as examples. Section 7.3 handles singular points: plain series fail there, so the method of Frobenius tries x^r times a series, uses the indicial equation to find r, and leads to Bessel functions.

Key Points:

  • Every power series has a radius of convergence ρ (possibly 0 or infinity): absolute convergence inside, divergence outside
  • Power series can be differentiated term by term, and the radius of convergence stays the same
  • A point x0 is ordinary if P(x0) ≠ 0 in P y'' + Q y' + R y = 0; singular if P(x0) = 0
  • At an ordinary point, substitute y = ∑ a_k (x - x0)^k, reindex, and solve the recurrence for the a_k in terms of a0 and a1
  • Regular singular point: both limits (x - x0) Q/P and (x - x0)^2 R/P exist and are finite
  • Frobenius: look for y = x^r ∑ a_k x^k; the indicial equation gives the possible values of r
  • Bessel’s equation of order p has indicial equation (r - p)(r + p) = 0, so the roots are r = p and r = -p

Memory Tip: Treat the recurrence relation as the finish line. Plug in the series, shift indices so every sum starts at k = 0 with the same power of x, and set each coefficient to zero; the recurrence then writes itself.

Common Mistake: Trying a plain power series at a singular point. For the book’s example 2xy’ – y = 0 at x = 0, a plain series forces every coefficient to zero and returns only y = 0, while the real solution is a multiple of x^(1/2), so you must switch to the Frobenius form.

Important Questions:

  • What is the difference between an ordinary point and a regular singular point? At an ordinary point P(x0) ≠ 0 and a power series solution exists; at a regular singular point P(x0) = 0 but the limits (x – x0)Q/P and (x – x0)^2 R/P are finite, which makes the Frobenius method applicable.
  • What is the indicial equation used for? After substituting y = x^r ∑ a_k x^k into the ODE, the lowest power of x gives a quadratic in r. Its roots are the possible exponents of the Frobenius solutions.

Chapter 8: Nonlinear Systems

Difficulty: Hard · Key topics: phase plane, critical points, linearization, Jacobian matrix, stability, conservative equations, pendulum, Lotka-Volterra, limit cycles, Poincaré-Bendixson, chaos

Because linear equations often approximate nonlinear ones locally, this chapter studies two-dimensional autonomous systems x’ = f(x,y), y’ = g(x,y). Section 8.1 finds critical points and linearizes with the Jacobian matrix. Section 8.2 defines isolated critical points and almost linear systems, classifies stability from the eigenvalues, explains why centers are tricky, and solves conservative equations through an energy integral. Section 8.3 applies this to the nonlinear pendulum and to predator-prey (Lotka-Volterra) models. Section 8.4 introduces limit cycles, the Van der Pol oscillator and two theorems about closed orbits. Section 8.5 closes with chaos: the Duffing equation and the Lorenz system.

Key Points:

  • Critical points solve f(x,y) = 0 and g(x,y) = 0 together; linearize by replacing the system with its Jacobian matrix at the critical point
  • A system is almost linear at an isolated critical point when the Jacobian there is invertible; eigenvalues then classify it (sink, source, saddle, spiral)
  • For a conservative equation x'' + f(x) = 0, an energy-type quantity is constant on every trajectory
  • The pendulum has centers at even multiples of π and saddles at odd multiples of π in the (θ, ω) phase plane
  • Lotka-Volterra: x' = (a - by)x, y' = (cx - d)y has a saddle at (0, 0) and an equilibrium at (d/c, a/b) where the eigenvalues are purely imaginary, so a conserved quantity is used to show the populations cycle in closed loops
  • Poincaré-Bendixson: a bounded solution without critical points in the region is periodic or approaches a periodic solution (two dimensions only)
  • Chaos means extreme sensitivity to initial conditions; the Duffing equation (forced) and the three-dimensional autonomous Lorenz system are the examples

Memory Tip: Linearization is the tangent line for systems. Near a critical point replace f and g by their Jacobian, read the behaviour from the eigenvalues, and trust the picture only close to that point.

Common Mistake: Trusting the linearization when it predicts a center. A center in the linear system does not guarantee closed orbits in the nonlinear one, because the nonlinear terms decide whether trajectories drift in or out.

Important Questions:

  • How do you classify a critical point of a nonlinear system? Compute the Jacobian matrix at the point and find its eigenvalues. If the point is isolated and the Jacobian is invertible, the eigenvalues give the type (sink, source, saddle, spiral) and its stability.
  • What does the Poincaré-Bendixson theorem say? If a solution stays forever in a closed bounded region of the plane with no critical points, then it is either periodic or tends towards a periodic solution (a limit cycle). It does not extend to three dimensions.

Appendix A: Linear Algebra

Appendix · Difficulty: Medium · Key topics: vectors and mappings, matrix algebra, elimination, rank, subspaces and kernel, rank-nullity, inner product, Gram-Schmidt, determinant

Appendix A is a self-contained linear algebra course in six sections, written for engineers. A.1 introduces vectors in R^n, linear mappings and matrices. A.2 builds matrix algebra from the idea of composing mappings, covering the identity, inverse, diagonal matrices and transpose. A.3 does elimination: row echelon form, inconsistent and non-unique systems, linear independence, rank and computing inverses. A.4 covers subspaces, basis, dimension, the kernel and the rank-nullity theorem. A.5 treats inner products, orthogonal projection and the Gram-Schmidt process, and A.6 the determinant. For a course combining linear algebra and ODEs it can replace the short review in Section 3.2.

Key Points:

  • Elementary row operations: swap two rows, multiply a row by a nonzero number, add a multiple of one row to another
  • A row of the form 0 = 1 means the system is inconsistent; free (non-pivot) variables mean infinitely many solutions
  • rank A = rank A^T, and rank-nullity says an n-column matrix satisfies n = rank + nullity
  • The kernel of a matrix equals the kernel of its reduced row echelon form, and a basis of it is read off the non-pivot columns
  • The inverse comes from row reducing [A | I] to [I | A^(-1)], which works exactly when A is invertible
  • Gram-Schmidt: at step k, subtract from v_k its projections onto the orthogonal vectors already built
  • det A ≠ 0 if and only if A is invertible; |det A| is the factor by which areas or volumes scale

Memory Tip: Match each linear algebra idea to its ODE twin: the kernel of a matrix corresponds to the solutions of a homogeneous linear ODE, a basis of the kernel to a fundamental set of solutions, and orthogonal projection to a Fourier coefficient.

Common Mistake: Cancelling a matrix as if it were a number. AB = AC does not imply B = C, even when A is not the zero matrix; that step needs A to be invertible.

Important Questions:

  • What does the rank-nullity theorem say? If a matrix has n columns, rank r and nullity k (the dimension of its kernel), then n = r + k.
  • How do you compute the inverse of a square matrix? Write the augmented matrix [A | I] and row reduce. If the left half becomes I, the right half is A^(-1); if it cannot, A is singular.

Appendix B: Table of Laplace Transforms, Further Reading and Selected Solutions

Appendix · Difficulty: Easy · Key topics: Laplace transform table, Gamma function, Heaviside step, Dirac delta, further reading, answers to selected exercises

Appendix B is a one-page lookup table of Laplace transforms that Chapter 6 uses constantly. It lists powers of t (using the Gamma function for non-integer powers), exponentials, sine, cosine, sinh and cosh, the shifted Heaviside step and Dirac delta, products such as t sin(ωt) and t cos(ωt), damped sines and entries with the error function. The back matter then gives a short Further Reading list (Edwards and Penney, Boyce and DiPrima, Ince, Farlow, Berg and McGregor, Judson’s ODE project and Trench’s free text), followed by Solutions to Selected Exercises for every exercise numbered 101 or higher, and an index.

Key Points:

  • L{t^n} = n!/s^(n+1) and, for non-integer powers, L{t^p} = Γ(p+1)/s^(p+1)
  • L{e^(-at)} = 1/(s + a)
  • L{sin ωt} = ω/(s^2 + ω^2) and L{cos ωt} = s/(s^2 + ω^2)
  • L{sinh ωt} = ω/(s^2 - ω^2) and L{cosh ωt} = s/(s^2 - ω^2)
  • L{u(t - a)} = e^(-as)/s and L{δ(t - a)} = e^(-as)
  • L{t sin ωt} = 2ωs/(s^2 + ω^2)^2 and L{t cos ωt} = (s^2 - ω^2)/(s^2 + ω^2)^2
  • Answers are printed only for exercises numbered 101 and higher; lower numbers are meant for homework

Memory Tip: Read the table in both directions and match denominators first: s^2 + ω^2 signals sine or cosine, s^2 – ω^2 signals sinh or cosh, and a factor e^(-as) signals a shift in time.

Common Mistake: Expecting full worked solutions at the back. Most entries are short final answers, so you still have to supply and check the method yourself.

Important Questions:

  • What are the Laplace transforms of sin(ωt) and cos(ωt)? ω/(s^2 + ω^2) for sine and s/(s^2 + ω^2) for cosine.
  • Which exercises have answers printed in the book? Those numbered 101 and above in each section (for example 1.4.101); exercises numbered below 101 have no printed answers.

How to Study This Book

Start with the Introduction, but do not linger. Section 0.2 (what a solution is, the modelling loop and the four fundamental equations) deserves a careful read; Section 0.3 on classification is short. The author says the chapters should be covered more or less in order, with later chapters droppable: Chapters 1 and 2 are the foundation, and Chapters 3 and 4 both rest on them.

The author’s own core for single ODEs is Sections 0.2, 1.1–1.4, 1.6, 2.1, 2.2 and 2.4–2.6. If you only have a few weeks, learn exactly those until separable and linear equations, the characteristic equation, undetermined coefficients, and vibrations and resonance feel automatic. Sections 1.5 (substitution), 1.7 (Euler’s method), 1.8 (exact equations) and 1.9 (first-order linear PDEs) are marked in the book as safe to skip, but check your university syllabus first, because Bernoulli and exact equations are common exam topics.

Chapter 3 needs matrices and eigenvalues. Section 3.2 is a short review, and Appendix A is the full treatment that the author says can replace it in a combined linear algebra and ODE course. If you have never computed an eigenvector, read 3.2 and 3.4.1 slowly before continuing. Chapter 8 builds directly on Chapter 3, so if your course covers it the book suggests placing it right after Chapter 3.

Chapters 4 and 5 form the PDE block. Read Sections 4.1 to 4.4 before the heat and wave equations, because separation of variables is Fourier series applied to a PDE. Section 4.8 then gives d’Alembert’s solution as an alternative to the series method for the wave equation. Chapter 5 is best read right after Chapter 4, and the book notes that Chapter 3 can be dropped while still covering Chapters 4 and 5, apart from a few non-essential references to linear algebra.

Chapters 6 and 7 are the usual add-ons for students whose course skips systems and PDEs: the author’s three-hours-a-week plan without systems or PDEs is the Introduction, Chapter 1, Chapter 2, Chapter 6 and Chapter 7. Chapter 6 does not depend on Chapter 4 except for Section 6.5 on PDEs, which can be skipped. If time is short, the first two sections of Chapter 7 make a self-contained unit; Section 7.3 (Frobenius and Bessel functions) is the hardest part.

For a two-semester sequence the author suggests Semester 1: Introduction, Chapters 1, 2, 6 and 7; Semester 2: Chapters 3, 8, 4 and 5. The complete book takes about 76 lectures without Appendix A, or about 86 with Appendix A replacing Section 3.2, not counting exams and review.

What the book does not do: it does not cover the Fourier transform (the author says so in Chapter 6), and numerical methods are limited to Euler’s method with a short mention of improved Euler and Runge-Kutta in Section 1.7. It is written for engineers, so it emphasises methods and worked examples over proofs of existence theorems. Answers are printed only for exercises numbered 101 and higher, so use those for self-checking and the lower numbers for homework. The author’s website adds interactive Sage demos, WeBWorK homework sets and video lectures for many sections.


Used In These Programs

This book is used for the Differential Equations course in: BS Mathematics and engineering degree programs. Browse all Differential Equations books or all Mathematics books.

Who Should Read This

This book is written for engineering and BS Mathematics students taking a first course in differential equations straight after the calculus sequence; the author gives the basic calculus sequence as the prerequisite. Students who are still unsure about integration techniques and series should revisit Calculus Volume 2 first, since Chapters 1, 2 and 7 use them constantly. It suits courses that cover systems, Fourier series and PDEs in one semester, and it is a lighter alternative to the traditional thousand-page textbooks. It is not a proof-based theory text: for a second set of worked examples and a different explanation of the same ODE methods, pair it with Elementary Differential Equations by William Trench.


Applicable Universities

This book is useful for students at Pakistani universities offering BS Mathematics and engineering degrees, including Punjab University, Virtual University, COMSATS, FAST, UET, NUST, and other HEC-recognized institutions.

Official Copy of Notes on Diffy Qs

This book is free from its official source, author Jiri Lebl’s own site. Click below to open the official page, where you can download the current PDF (version 6.11), open the web version, or get the LaTeX source — no cost, no registration.

→ Visit Official Page

FAQs

Is Notes on Diffy Qs free to download?

Yes. The book’s own copyright page says it is dual licensed under the Creative Commons Attribution-NonCommercial-ShareAlike 4.0 International License (CC BY-NC-SA 4.0) and the Creative Commons Attribution-ShareAlike 4.0 International License (CC BY-SA 4.0), and you may assume whichever of the two suits your use. You can read, print, copy and share it at no cost from the author’s own site, with no registration. Derivative works must be clearly marked as such.

Which edition is this, and is it current?

Edition 6, version 6.11, dated May 18, 2026 on the author’s official page. Lebl has maintained the book continuously since 2008, fixing errata promptly and publishing a corrected minor version once or twice a year, and only raises the edition number for substantial changes. Version 6 added Appendix A on linear algebra and the PDE sections 1.9 and 6.5.

How many chapters and exercises does it have?

An Introduction, 8 chapters and 2 appendices (linear algebra, and a table of Laplace transforms). There are more than 700 numbered exercises, and answers or solutions for those numbered 101 and higher are printed in a Solutions to Selected Exercises section at the back, so you can self-check roughly a third of them.

Is it good for complete beginners in differential equations?

Yes for the differential equations themselves, since Chapter 1 starts from plain integration, but it does assume the basic calculus sequence. Chapters 3 and 4 also need matrices and integration skill; Chapter 3 includes a short matrix review and Appendix A is a full linear algebra course if you need more.

Does it cover Laplace transforms, Fourier series and PDEs?

Yes. Chapter 6 covers the Laplace transform, convolution, the Dirac delta and impulse response, and solving PDEs with it. Chapter 4 covers Fourier series and the heat, wave and Laplace equations by separation of variables and d’Alembert’s solution, and Chapter 5 generalises to Sturm–Liouville problems. It does not cover the Fourier transform.

Can it replace Boyce–DiPrima or Edwards–Penney for my course?

That is what the author designed it for. He describes it as more or less a drop-in replacement for those two textbooks, and his website gives a section-by-section correspondence to both. It is shorter and lighter on proofs, so if your instructor assigns problems by number from a particular textbook, check the correspondence table first.

Related Books

Notes on Diffy Qs: Differential Equations for Engineers is this project’s engineering-oriented differential equations text — first-order and higher-order ODEs, systems, Fourier series and PDEs, Laplace transforms, power series and nonlinear systems in one free book. Browse more Mathematics books for the rest of your degree.

Notes on Diffy Qs: Differential Equations for Engineers, Edition 6 (version 6.11, May 18, 2026), by Jiří Lebl. Copyright 2008–2026 Jiří Lebl. Free, dual licensed under the Creative Commons Attribution-NonCommercial-ShareAlike 4.0 International License (CC BY-NC-SA 4.0) or the Creative Commons Attribution-ShareAlike 4.0 International License (CC BY-SA 4.0). Access for free at https://www.jirka.org/diffyqs/

Leave a Comment