A first course in differential equations is where applied mathematics becomes recognisable to engineering and science students, and it is also where many of them discover that their calculus was less secure than they believed. The course assumes fluency in integration, in complex numbers, and in algebraic manipulation, and it uses all three simultaneously from the second week.

The distinctive difficulty is that the course is organised by technique rather than by idea. Students learn to solve separable equations, linear first-order equations, second-order constant-coefficient equations, and systems, and the organising principle that connects them is rarely stated.

What the Course Is Actually About

Differential equations are about modelling change, and the techniques exist to answer three questions: what behaviour does the model predict, how does that behaviour depend on parameters, and what happens in the long run.

Students who treat the course as a catalogue of methods can solve exercises and cannot answer a question that begins "a population grows according to." That question is the point of the subject.

Making the modelling explicit, with the same equation interpreted as a population, a cooling body, a discharging capacitor, and a loan, converts a technique into a tool.

First-Order Equations and the Two Cases

First-order equations divide into two cases that cover most of what an applied course requires.

Separable equations can be rearranged so that each variable appears on one side, and the solution follows from integrating both sides. This is the technique students master fastest, and it is also the one they over-apply, attempting separation on equations that do not separate.

Linear first-order equations have the form where the unknown and its derivative appear linearly, and they are solved with an integrating factor. The integrating factor is the part students memorise without understanding: it is a function chosen so that the left side becomes the derivative of a product.

Showing the derivation of the integrating factor once, from the product rule backwards, converts a formula into a construction.

The Qualitative Turn

The most valuable idea in the course is the one that requires the least computation: qualitative analysis of a differential equation without solving it.

A slope field shows the derivative at each point, and the solution curves follow the field. A phase line for an autonomous equation shows where the derivative is zero and whether those equilibria are stable. Long-run behaviour can be read off without a formula.

This matters practically because most differential equations that arise in applications cannot be solved in closed form, and the qualitative question is often the one that was actually asked. A student who can answer it has an advantage over one who can only integrate.

Second-Order Linear Equations

Second-order constant-coefficient equations are the centre of the applied course, because they model oscillation, damping, and resonance, which appear across physics and engineering.

The method is to assume an exponential solution and reduce the problem to a characteristic equation, whose roots determine the behaviour. Real distinct roots give exponential combinations, repeated roots require a factor of t, and complex roots give oscillation with exponential decay or growth.

The physics attached to the three cases is what makes them memorable: overdamped, critically damped, and underdamped. A student who has seen a door closer and a guitar string has a referent for the algebra, and the referent is what survives an examination.

Resonance and Why It Matters

Resonance deserves separate treatment because it is the phenomenon that makes the course matter outside the classroom.

A driven oscillator whose driving frequency matches its natural frequency produces unbounded amplitude in the undamped case and a large but bounded amplitude with damping. Every structural engineering student needs this; so does anyone who has watched a bridge or a building respond to wind or to an earthquake.

The mathematics is a particular solution with a factor of t that appears precisely when the forcing matches the homogeneous solution. That special case, which students treat as a technical annoyance, is the physically important one.

Systems and the Matrix Connection

Linear systems of differential equations are where the course connects to linear algebra, and the connection is direct: the solution is expressed through eigenvalues and eigenvectors.

A system written in matrix form has solutions that are combinations of exponentials in the eigenvalues, with directions given by eigenvectors. Real eigenvalues give growth or decay along fixed directions; complex eigenvalues give spirals.

Students who have studied linear algebra find this straightforward and satisfying. Students who have not find it opaque, and a short review of eigenvalues before the systems chapter is usually the highest-value hour in the course.

Numerical Methods and Why They Are Taught

Euler's method and its refinements are often taught briefly and treated as inferior to exact solutions, which misses the point.

The point is that most real differential equations have no closed-form solution, and numerical integration is how they are actually solved. Euler's method is the simplest illustration of the idea: follow the slope for a small step, then recompute the slope.

Showing the error behaviour, and why a smaller step reduces error at the cost of computation, introduces the trade-off that governs all practical computation. It also explains why the step size in a simulation is a modelling decision rather than a technical detail.

Where Students Lose Marks

Three patterns account for most lost marks in an applied differential equations course.

The first is integration errors. The course assumes integration is fluent, and it is not, for many students. A student who cannot integrate by parts reliably will lose marks in every linear equation problem.

The second is the constant of integration. In a general solution it must be carried, and in an initial value problem it must be determined. Dropping it, or determining it too early, produces answers that satisfy the equation and not the condition.

The third is domain and validity. A solution may fail at a point, or be valid only on an interval, and applied problems often require stating the interval. This is the mark that students leave on the table.

What Tutoring Should Do

The sequence that works: verify integration fluency, then establish the modelling frame, then the techniques, then the qualitative and numerical ideas.

Integration first because it is a prerequisite that the course does not teach. A two-week review, if needed, saves the term.

The modelling frame second because it makes the techniques purposeful. Every equation met should be attached to at least one physical situation.

Techniques third, each derived rather than memorised where the derivation is short. Qualitative and numerical ideas last, because they are the ones that survive after the techniques are forgotten.

The Worked Example Problem

A specific habit distinguishes students who do well in this course from those who do not, and it is worth naming.

The habit is to carry a single physical example through the entire term. One model, say a mass on a spring with damping, developed fully: the equation derived from Newton's second law, solved in the undamped case, then with damping, then with forcing, then examined qualitatively and numerically.

A student who has done this once has a referent for every technique and every vocabulary word in the course. Damping ratio, natural frequency, transient, steady state, resonance, amplitude and phase all become concrete, because they appeared in one developed example rather than as definitions in a chapter.

Textbooks tend to introduce a new example in each section for variety, which is good for coverage and bad for retention. The tutor's contribution is the opposite: one example, developed deeply, referred to constantly.

Using Software Without Losing the Idea

Most applied courses now use software, and the risk is that the software becomes the subject.

The legitimate uses are exploration and checking: plotting a slope field to confirm a qualitative analysis, solving numerically to see whether an exact solution behaves as predicted, varying a parameter to see how the behaviour changes.

The illegitimate use is producing an answer without a prediction. A student who types a differential equation into a solver and receives a solution containing a constant has learned nothing about the equation, and will be unable to say whether the solution is plausible.

The rule that keeps software useful is the same as in every other mathematical subject: predict first, then check. The prediction can be rough, even qualitative. It has to exist.

What to Do in the First Two Weeks

Differential equations courses tend to be survivable until the midterm and then suddenly not, which means the first fortnight determines the outcome.

Two things are worth doing immediately. Test integration fluency with ten items of the kind that appear in linear equation solutions: by parts, partial fractions, and simple substitution. And establish the modelling frame with one developed example, so that every subsequent technique has somewhere to attach.

A student who does both arrives at the midterm with the prerequisite secured and a mental structure for the material. A student who does neither spends the term accumulating techniques with no referent, which is the profile that fails.

An applied differential equations course is organised by technique but is really about modelling change, and students who treat it as a catalogue of methods cannot answer the questions the course exists to answer.

Teach separable and linear first-order equations with the integrating factor derived rather than given, and teach qualitative analysis early, since most real equations cannot be solved in closed form.

Attach the three second-order cases to physical referents, treat resonance as the central case rather than a technical annoyance, and connect systems to eigenvalues explicitly. Check integration fluency at the start, and insist on constants of integration and intervals of validity, which are where marks are most often left behind.