English:Differential Equations

Introduction

A differential equation is an equation that connects an unknown function with one or more of its derivatives. Differential equations describe change: motion, growth, decay, heat flow, electric circuits, chemical reactions, population dynamics, control systems, and many other processes. In this university-level aiMOOC, you will learn to classify differential equations, solve important families of ordinary differential equations, analyze systems qualitatively, approximate solutions numerically, and connect ordinary differential equations with introductory partial differential equations.

You should be comfortable with differentiation and integration, algebra, complex numbers, and basic linear algebra. The central question throughout the course is not only “Can you solve the equation?” but also “What does the solution mean, how do you know it is valid, and what can you infer when no closed-form solution is available?”

Load video

YouTube

YouTube might collect personal data. Privacy Policy

Continue Dismiss

The image above shows a slope field. Each short segment displays the slope prescribed by a first-order differential equation at a point. Solution curves must follow these local directions. This geometric viewpoint is especially useful when an explicit formula is difficult or impossible to obtain.

Learning Goals

By the end of this aiMOOC, you should be able to classify differential equations by order, linearity, and type; solve standard first- and second-order ordinary differential equations; formulate initial-value and boundary-value problems; use Laplace transforms for linear initial-value problems; analyze planar systems using eigenvalues, phase portraits, and linearization; apply Euler and Runge–Kutta methods; recognize stiffness and numerical stability issues; explain the role of basic partial differential equations; and build, test, and critique mathematical models.

Foundations

Ordinary and Partial Differential Equations

An ordinary differential equation or ODE contains derivatives with respect to one independent variable. For example,

$\frac{dy}{dt}=ky$

models exponential growth or decay when $k$ is constant. A partial differential equation or PDE contains partial derivatives with respect to two or more independent variables. For example, the one-dimensional heat equation is

$u_t=\alpha u_{xx}$,

where $u(x,t)$ may represent temperature and $\alpha >0$ is a diffusivity constant.

The order of a differential equation is the highest derivative appearing in it. The equation $y^{″}+4y=0$ is second order. A differential equation is linear when the unknown function and its derivatives occur only to the first power and are not multiplied together. Thus $y^{\prime }+p(x)y=q(x)$ is linear, while $y^{\prime }=y^2-x$ is nonlinear.

An equation such as $y^{\prime }=f(y)$ that does not explicitly contain the independent variable is called autonomous. Autonomous equations are particularly useful for equilibrium and stability analysis.

Solutions, Initial Conditions, and Boundary Conditions

A solution is a function that satisfies the differential equation on a specified interval. A family such as $y=C{e}^{kt}$ is a general solution of $y^{\prime }=ky$. An initial condition such as $y(0)=y_0$ selects a particular member of the family.

An initial-value problem specifies data at one value of the independent variable. A boundary-value problem specifies conditions at different points, such as $u(0)=0$ and $u(L)=0$. Boundary-value problems appear naturally in steady heat flow, beam deflection, and eigenvalue problems.

For a first-order equation $y^{\prime }=f(t,y)$, continuity of $f$ near an initial point supports local existence of a solution. A suitable Lipschitz condition in $y$ provides local uniqueness. These ideas matter because a symbolic manipulation is meaningful only when the underlying problem actually has a well-defined solution on the interval being studied.

Verification and Dimensional Reasoning

You should verify a proposed solution by differentiating it, substituting it into the original equation, and checking all initial or boundary conditions. In applied problems, also check units. If $x$ is measured in meters and $t$ in seconds, then $x^{\prime }$ has units of meters per second. Dimensional inconsistency is often a warning that a model or algebraic step is wrong.

First-Order Ordinary Differential Equations

Separable Equations

A first-order equation is separable when it can be written in the form

$\frac{dy}{dx}=g(x)h(y).$

Where division is legitimate, rearrange it as

$\frac{1}{h(y)}dy=g(x)dx$

and integrate both sides. Be careful: dividing by $h(y)$ can discard equilibrium solutions satisfying $h(y)=0$. Always check those constant solutions separately.

For example, the logistic model

$P^{\prime }=rP(1-\frac{P}{K})$

has equilibria $P=0$ and $P=K$. For positive initial data and positive parameters, the model predicts growth that slows as the population approaches the carrying capacity $K$.

Linear First-Order Equations

A linear first-order equation has the standard form

$y^{\prime }+p(x)y=q(x).$

Its integrating factor is

$\mu (x)=e^{\int p(x)dx}.$

Multiplying the equation by $\mu$ converts the left-hand side into a product derivative:

$\frac{d}{dx}\left(\mu y\right)=\mu q.$

After integration,

$y=\frac{1}{\mu }(\int \mu qdx+C).$

This method is central to models involving mixing, Newtonian cooling, electric circuits, and linear response.

Load video

YouTube

YouTube might collect personal data. Privacy Policy

Continue Dismiss

Exact Equations and Useful Substitutions

An equation written as

$M(x,y)dx+N(x,y)dy=0$

is exact on a suitable region when there is a potential function $F$ satisfying $F_x=M$ and $F_y=N$. On a simply connected region, a standard test is $M_y=N_x$ when the relevant first partial derivatives are continuous. The implicit solution is then $F(x,y)=C$.

Other first-order forms can often be simplified by substitution. A Bernoulli equation $y^{\prime }+p(x)y=q(x)y^n$, with $n\ne 0,1$, becomes linear after the substitution $v=y^{1-n}$. A homogeneous first-order equation depending on $y/x$ may be reduced by setting $v=y/x$.

Second-Order Linear Equations

Homogeneous Equations with Constant Coefficients

Consider

$a{y}^{″}+b{y}^{\prime }+cy=0$, with $a\ne 0$.

Trying a solution $y=e^{rt}$ leads to the characteristic equation

$a{r}^2+br+c=0.$

If the roots $r_1$ and $r_2$ are distinct and real, then

$y=C_1{e}^{r_{1}t}+C_2{e}^{r_{2}t}.$

If there is a repeated real root $r$, then

$y=(C_1+C_{2}t)e^{rt}.$

If the roots are $\alpha \pm i\beta$, then

$y=e^{\alpha t}(C_1\cos \beta t+C_2\sin \beta t).$

The sign of $\alpha$ determines whether oscillations grow, remain neutral, or decay.

Mechanical Oscillation, Damping, and Forcing

A mass–spring–damper system is modeled by

$m{x}^{″}+c{x}^{\prime }+kx=F(t),$

where $m$ is mass, $c$ is damping, $k$ is stiffness, and $F(t)$ is an applied force.

When $F(t)=0$, the discriminant $c^2-4mk$ distinguishes overdamped, critically damped, and underdamped motion. With periodic forcing, the response has transient and steady-state parts. In weakly damped systems, forcing near the system's natural frequency can produce a large steady-state amplitude. This phenomenon is related to resonance.

Nonhomogeneous Equations

For

$a{y}^{″}+b{y}^{\prime }+cy=g(t),$

the general solution is $y=y_h+y_p$, where $y_h$ solves the homogeneous equation and $y_p$ is one particular solution. Methods such as undetermined coefficients work well when $g(t)$ is built from exponentials, polynomials, sines, and cosines. Variation of parameters is more general and uses a known fundamental set of homogeneous solutions.

The Wronskian helps test linear independence. For two functions $y_1$ and $y_2$,

$W=y_1{y_2}^{\prime }-{y_1}^{\prime }{y}_2.$

Linear Systems and Phase-Plane Analysis

Matrix Form and Eigenvalues

A system of first-order linear ODEs can be written as

${𝐱}^{\prime }=A𝐱.$

If $A$ has eigenpairs $({\lambda }_i,{𝐯}_i)$, then terms of the form $e^{{\lambda }_{i}t}{𝐯}_i$ describe basic solution modes. Eigenvalues therefore connect linear algebra directly with dynamics.

For a two-dimensional system, negative real parts of all eigenvalues imply asymptotic decay toward the equilibrium for the linear system. Positive real parts produce growth. Opposite signs give a saddle. Complex eigenvalues produce rotation or spiraling, with the real part controlling growth or decay.

Load video

YouTube

YouTube might collect personal data. Privacy Policy

Continue Dismiss

Phase Portraits and Nonlinear Systems

A phase portrait shows trajectories in state space rather than plotting every variable against time. For nonlinear systems

${𝐱}^{\prime }=𝐟(𝐱),$

equilibria satisfy $𝐟({𝐱}_{\ast })=\mathrm{𝟎}$. Near an equilibrium, the Jacobian matrix

$J({𝐱}_{\ast })={\left[\frac{\partial f_i}{\partial x_j}\right]}_{𝐱={𝐱}_{\ast }}$

provides a linear approximation. If the equilibrium is hyperbolic, meaning no Jacobian eigenvalue has zero real part, the linearized system reveals the local qualitative stability type.

The Lotka–Volterra predator–prey model is a classic nonlinear system:

$x^{\prime }=\alpha x-\beta xy,y^{\prime }=\delta xy-\gamma y.$

The idealized model can produce closed trajectories around a coexistence equilibrium. Real ecological systems usually require additional effects such as carrying capacities, seasonal forcing, migration, or stochasticity, so model assumptions must be examined rather than accepted automatically.

Laplace Transform Methods

The Laplace transform converts certain differential equations in time into algebraic equations in the transform variable. For a suitable function $f(t)$,

$ℒ\{f(t)\}(s)={\int }_0^{\mathrm{\infty }}{e}^{-st}f(t)dt.$

For example,

$ℒ\{f^{\prime }(t)\}=sF(s)-f(0).$

This identity builds the initial condition directly into the transformed equation. Laplace transforms are especially useful for linear initial-value problems with discontinuous inputs, step functions, impulses, and piecewise forcing. After solving for $F(s)$, inverse transformation recovers $f(t)$.

Load video

YouTube

YouTube might collect personal data. Privacy Policy

Continue Dismiss

A disciplined workflow is: transform the equation, insert the initial data, solve the algebraic equation for the transformed unknown, simplify with partial fractions or known transform rules, invert, and verify the result in the original problem.

Numerical Methods

Euler's Method

Many differential equations do not have elementary closed-form solutions. Numerical methods approximate a solution at selected points. For the initial-value problem

$y^{\prime }=f(t,y),y(t_0)=y_0,$

Euler's method uses the recurrence

$y_{n+1}=y_n+hf(t_n,y_n),t_{n+1}=t_n+h.$

A smaller step size usually reduces discretization error but increases computational work. Euler's method is first order: under standard smoothness assumptions, its global error scales like $O(h)$ over a fixed interval.

Load video

YouTube

YouTube might collect personal data. Privacy Policy

Continue Dismiss

Runge–Kutta Methods and Error Control

Classical fourth-order Runge–Kutta, often called RK4, samples the slope several times within each step and combines those samples to achieve much higher accuracy than Euler's method for sufficiently smooth problems.

Modern ODE solvers often use embedded Runge–Kutta pairs or other adaptive schemes. They estimate local error and adjust the step size automatically. You should still inspect tolerances, conserved quantities, qualitative behavior, and convergence under refinement rather than treating a numerical answer as automatically correct.

Stability and Stiffness

Numerical stability concerns whether a discrete method reproduces the decay or growth behavior of the differential equation. For the test equation $y^{\prime }=\lambda y$, forward Euler gives

$y_{n+1}=(1+h\lambda )y_n.$

The method is stable for decaying modes only when the amplification factor lies inside its stability region. This requirement can force extremely small step sizes for stiff problems.

Stiff systems contain strongly separated time scales. Implicit methods such as backward differentiation formulas are often more efficient than explicit methods because their stability properties permit larger steps while still controlling rapidly decaying modes.

Partial Differential Equations

Three Classical PDEs

Differential equations involving several independent variables lead to PDEs. Three standard linear examples are:

Equation	Typical interpretation	Broad type
$u_t=\alpha u_{xx}$	Heat diffusion	Parabolic
$u_{tt}=c^2{u}_{xx}$	Wave propagation	Hyperbolic
$u_{xx}+u_{yy}=0$	Steady potential	Elliptic

The classification above refers to standard second-order linear PDEs and is connected with the signs of the principal coefficients. Initial and boundary conditions are essential because a PDE by itself usually describes a family of possible fields.

Heat Equation and Separation of Variables

For a rod of length $L$ with fixed zero temperature at both ends,

$u_t=\alpha u_{xx},u(0,t)=u(L,t)=0.$

Assuming a product solution $u(x,t)=X(x)T(t)$ leads to separated ODEs. Boundary conditions select discrete spatial modes such as $\sin (n\pi x/L)$. The time factors decay exponentially, so high-frequency temperature variations disappear faster than low-frequency ones.

Fourier series arise naturally because initial data can be expanded in eigenfunctions compatible with the boundary conditions.

Load video

YouTube

YouTube might collect personal data. Privacy Policy

Continue Dismiss

Mathematical Modeling with Differential Equations

A Modeling Workflow

A useful model begins with assumptions rather than with algebra. Identify the system boundary, choose state variables, define parameters with units, state mechanisms that determine rates of change, write the differential equations, supply initial or boundary data, solve or approximate the equations, compare predictions with observations, and revise the model when its assumptions fail.

Modeling situation	Representative equation	Main interpretation
Radioactive decay	$N^{\prime }=-\lambda N$	The decay rate is proportional to the amount present
Newtonian cooling	$T^{\prime }=-k(T-T_a)$	Temperature approaches an ambient value
Logistic growth	$P^{\prime }=rP(1-P/K)$	Growth slows as a carrying capacity is approached
RC circuit	$RC{V}^{\prime }+V=E(t)$	Voltage responds to an applied source
Damped oscillator	$m{x}^{″}+c{x}^{\prime }+kx=F(t)$	Inertia, damping, restoring force, and forcing interact

A model is not the same as reality. Parameters may vary, noise may matter, omitted variables may become important, and a model valid in one regime may fail in another. A strong analysis therefore includes sensitivity, validation, and a clear statement of limitations.

Scaling and Nondimensionalization

Nondimensionalization rewrites a model in terms of dimensionless variables. Suppose $t=T\tau$ and $x=Xu$, where $T$ and $X$ are characteristic scales. Substitution can reduce the number of independent parameters and reveal dominant balances. Dimensionless groups often show which physical effects matter most in a given regime.

This idea is valuable in fluid mechanics, chemical engineering, biomechanics, and numerical simulation because it separates the structure of an equation from arbitrary choices of units.

Common Errors and How to Check Your Work

Lost equilibrium solutions: Separating variables may require division by a factor that can be zero. Check constant solutions before dividing.

Missing constants: When integrating, include the constant of integration unless an equivalent constant has already been introduced.

Incorrect initial conditions: Apply all initial data after obtaining the general solution or directly within a transform method.

Characteristic-root mistakes: Distinguish distinct real roots, repeated roots, and complex-conjugate roots.

Confusing stability with boundedness: A bounded trajectory is not automatically asymptotically stable, and a center is not the same as a sink.

Ignoring units: Every term that is added in a physical equation must have compatible dimensions.

Trusting a single numerical run: Repeat the computation with stricter tolerances or smaller steps and compare results.

Ignoring the interval of validity: A formula can satisfy an equation locally yet fail across a singularity or outside the model's physical domain.

Interactive Tasks

Quiz: Test Your Knowledge

Which statement best defines the order of a differential equation? (The highest derivative that appears) (!The number of unknown constants) (!The number of initial conditions shown) (!The degree of the independent variable)

Which equation is linear in the unknown function y? (y prime plus x y equals sine x) (!y prime equals y squared) (!y y prime equals x) (!y prime plus sine y equals x)

What is the main purpose of an integrating factor for a first-order linear equation? (To create a product derivative) (!To remove all initial conditions) (!To turn every equation into a separable one) (!To replace differentiation by finite differences)

What does a slope field represent? (Local slopes prescribed by a first order equation) (!Exact values of every integration constant) (!Only equilibrium solutions) (!Eigenvectors of a coefficient matrix)

What does a saddle point in a linear planar system typically have? (Eigenvalues with opposite signs) (!Two negative real eigenvalues) (!Two positive real eigenvalues) (!Purely imaginary eigenvalues)

What is the characteristic equation used for? (Finding modes of a constant coefficient linear ODE) (!Computing a numerical step size) (!Checking whether a PDE is exact) (!Defining a Laplace transform)

Why is the Laplace transform useful for many linear initial-value problems? (It incorporates initial data into transformed derivative formulas) (!It makes every nonlinear equation linear) (!It removes the need for inverse transformation) (!It guarantees a closed form solution)

What is a key limitation of forward Euler on stiff problems? (Stability can require very small steps) (!It cannot approximate first order equations) (!It always gives an exact answer) (!It requires no initial value)

Which PDE is the standard model for diffusion of heat? (The heat equation) (!The wave equation) (!The eikonal equation) (!The transport identity)

What should you do after obtaining a symbolic solution to an initial-value problem? (Substitute it back and check the initial data) (!Assume the algebra guarantees correctness) (!Discard the domain of the solution) (!Remove all physical units)

Memory Game

Integrating factor	Multiplier that converts a linear first-order equation into a product derivative
Equilibrium	Constant state at which all time derivatives vanish
Eigenvalue	Scalar that controls a modal growth or decay rate in a linear system
Wronskian	Determinant used to study linear independence of solutions
Phase portrait	State-space picture of trajectories of a dynamical system
Stiffness	Property that can force explicit numerical methods to use very small steps
Resonance	Large response associated with forcing near a system frequency

Drag and Drop

Match the correct terms.	Topic
Separation of variables	Rearranges a suitable first-order equation so each variable can be integrated on its own side
Characteristic equation	Converts a constant-coefficient linear ODE into an algebraic polynomial in a trial exponent
Linearization	Uses a Jacobian matrix to approximate a nonlinear system near an equilibrium
Laplace transform	Converts differentiation in time into algebraic operations involving the transform variable
Runge–Kutta method	Approximates an ODE solution by combining several slope evaluations within a step

...

Crossword Puzzle

Separable	Which adjective describes an ODE whose variables can be placed on opposite sides before integration?
Equilibrium	What is a constant state where the derivative vanishes?
Wronskian	What determinant is used to test linear independence of solutions?
Eigenvalue	What scalar determines a modal rate in a linear system?
Stability	What concept asks whether nearby states remain near or approach an equilibrium?
Diffusion	What physical process is modeled by the heat equation?

Cloze Text

Complete the text.

A differential equation relates an unknown function to one or more of its . The highest derivative in an equation determines its . A first-order linear equation can often be solved using an . Constant states of an autonomous system are called . Eigenvalues help determine local behavior in a linear . Euler's method advances a numerical approximation using the current . The Laplace transform is especially useful for many linear . The heat equation is a standard model of .

Open-Ended Tasks

Easy

1. Slope Field Sketch: Draw a slope field for a first-order autonomous equation, mark its equilibria, and explain which equilibria appear stable or unstable.
2. Solution Verification: Choose a solved ODE example, differentiate the proposed solution, substitute it into the equation, and check every initial condition.
3. Cooling Experiment: Measure how a warm drink approaches room temperature, record a small data set, and discuss whether Newton's law of cooling is a reasonable model.
4. Differential Equation Explainer: Produce a one-page illustrated explanation or a short video showing the difference between an ODE, a PDE, an initial-value problem, and a boundary-value problem.

Standard

1. Numerical Method Comparison: Implement Euler's method and one Runge–Kutta method for the same initial-value problem, compare errors under step refinement, and present your results in a graph or table.
2. Mass Spring Model: Build or simulate a mass–spring–damper system, estimate parameters, and explain how changing damping affects the predicted motion.
3. Model Interview: Interview a scientist, engineer, economist, or other professional who uses rate-based models and summarize where differential equations enter their work.
4. Phase Portrait Project: Create a phase portrait for a two-dimensional linear system, classify its equilibrium from eigenvalues, and compare the algebraic prediction with the plotted trajectories.

Advanced

1. Nonlinear Dynamics Investigation: Analyze a nonlinear two-variable model, find equilibria, compute the Jacobian, classify hyperbolic equilibria, and test the local predictions numerically.
2. Stiff Solver Study: Construct or select a stiff initial-value problem, compare an explicit method with a suitable implicit solver, and explain the difference using numerical stability.
3. Heat Equation Project: Solve or numerically approximate a one-dimensional heat equation with specified initial and boundary data, then create an animation or sequence of plots that explains diffusion over time.
4. Applied Modeling Visit: Visit or virtually tour a laboratory, engineering facility, research group, or technical workplace where dynamic models are used, then produce a report connecting observed processes to differential-equation assumptions and limitations.

Learning Assessment

1. Model Selection Assessment: Given three real-world scenarios, formulate suitable differential equations, define each variable and parameter with units, and justify every modeling assumption.
2. Method Choice Assessment: For a mixed set of ODEs, decide whether separation, integrating factors, characteristic roots, Laplace transforms, qualitative analysis, or numerical approximation is most appropriate and defend each choice.
3. Stability Reasoning Assessment: Analyze a planar linear system from its matrix, predict the phase portrait from eigenvalues, and explain how the prediction changes if a parameter crosses a stability boundary.
4. Numerical Reliability Assessment: Compare numerical solutions obtained with different step sizes or tolerances, identify evidence of convergence or instability, and recommend a trustworthy computational setup.
5. Transfer to PDEs Assessment: Explain how separation of variables turns a simple heat-equation boundary-value problem into ODE eigenvalue problems and connect the resulting modes with Fourier series.
6. Model Critique Assessment: Evaluate a differential-equation model against data or a realistic scenario, identify at least two limitations, and propose a mathematically meaningful extension.

Evidence of Learning

Knowledge

You can explain the meaning of order, linearity, autonomy, initial and boundary data, equilibrium, stability, eigenvalues, transforms, numerical error, stiffness, and PDE classification.

Skills

You can solve representative first- and second-order ODEs, verify solutions, analyze linear and nonlinear systems, construct phase portraits, apply Laplace transforms, implement numerical methods, and interpret numerical convergence.

Products

Strong evidence includes a correct analytic solution portfolio, a numerical notebook or program, a phase-plane analysis, a validated applied model, and a visual or video explanation that communicates mathematics clearly.

Transfer

You can choose an appropriate method for an unfamiliar problem, connect model assumptions to equation structure, judge when a numerical method is trustworthy, and transfer ODE ideas to PDE separation, engineering systems, biological dynamics, or other applied contexts.

OERs on the Topic

Additional freely accessible resources include MIT OpenCourseWare Differential Equations and OpenStax Direction Fields and Numerical Methods.

Linked Learning Areas

Differential Equations Ordinary differential equation Partial differential equation Initial value problem Boundary value problem Slope field Separation of variables Integrating factor Laplace transform Linear dynamical system Phase portrait Eigenvalue Numerical analysis Euler method Runge–Kutta methods Heat equation Fourier series

Differential equations connect Calculus, Linear algebra, Numerical analysis, Dynamical systems, Mathematical modeling, Physics, Engineering, Biology, Economics, and Computer science. At university level, these links help you move between symbolic methods, qualitative reasoning, computational experiments, and domain-specific interpretation.