Differential Equation Types

Why This Matters

Differential equations describe how things change: population growth, vibrating bridges, heat spreading through metal. In your Linear Algebra and Differential Equations course, you're tested on your ability to classify equations correctly and select appropriate solution methods. The connection to linear algebra runs deep: matrix methods, eigenvalues, superposition, and vector spaces all emerge naturally when solving these equations.

Understanding why each classification matters is more useful than memorizing definitions. The type of equation determines which tools you reach for: separation of variables, integrating factors, characteristic equations, or matrix exponentials. When you see a differential equation on an exam, your first job is always identification. Get that right, and the solution method follows.

---

Classification by Number of Variables

The most fundamental distinction is whether your unknown function depends on one variable or many. This determines whether you're working with ordinary or partial derivatives, which completely changes your solution toolkit.

Ordinary Differential Equations (ODEs)

• Single independent variable: the unknown function $y(x)$ or $x(t)$ depends on just one variable, so derivatives are unambiguous
• Foundation of this course: ODEs connect directly to linear algebra through matrix methods for systems and eigenvalue problems
• Explicit vs. implicit solutions: you may find $y = f(x)$ directly, or a relationship $F(x, y) = C$ that defines the solution implicitly

Partial Differential Equations (PDEs)

• Multiple independent variables: functions like $u(x, y, t)$ require partial derivatives $\frac{\partial u}{\partial x}$, $\frac{\partial u}{\partial t}$
• Beyond this course's scope: PDEs govern heat transfer, wave propagation, and quantum mechanics, but require more advanced techniques
• Separation of variables bridges both worlds: the technique you learn for ODEs extends to solving PDEs by reducing them to systems of ODEs

---

Classification by Order

The order of a differential equation is the highest derivative that appears. This tells you how many initial conditions you need and which solution methods apply.

First-Order Differential Equations

• Only $\frac{dy}{dx}$ appears: no second derivatives or higher, making these the simplest to solve
• Multiple solution techniques available: separation of variables, integrating factors, and exact equation methods all apply here
• Model rates of change directly: population growth ($\frac{dP}{dt} = kP$), radioactive decay, and Newton's law of cooling are classic applications

Second-Order Differential Equations

• Involves $\frac{d^2y}{dx^2}$: requires two initial conditions (position and velocity, for instance) to pin down a unique solution
• Characteristic equations connect to linear algebra: solving $ay'' + by' + cy = 0$ reduces to finding roots of $ar^2 + br + c = 0$
• Physical systems love second order: mass-spring systems ($m\ddot{x} + c\dot{x} + kx = F$) and RLC circuits naturally produce these equations

---

Classification by Linearity

Whether an equation is linear or nonlinear determines whether the powerful tools of linear algebra apply. Linear equations allow superposition: you can add solutions together to get new solutions.

Linear Differential Equations

A differential equation is linear if the dependent variable and all its derivatives appear only to the first power, and they aren't multiplied together or composed inside nonlinear functions. That means no $y^2$, no $\sin(y)$, no $(y')^2$, and no $yy'$.

• Superposition principle applies: if $y_1$ and $y_2$ solve the homogeneous equation, so does $c_1y_1 + c_2y_2$
• Standard methods always work: undetermined coefficients, variation of parameters, and matrix exponentials provide systematic solution approaches
• Coefficients can still be variable: an equation like $x^2 y'' + xy' + y = 0$ is linear even though the coefficients depend on $x$, because $y$ and its derivatives still appear to the first power

Nonlinear Differential Equations

• Dependent variable or derivatives appear nonlinearly: terms like $y^2$, $yy'$, or $\sin(y)$ break linearity
• No general solution methods exist: each equation requires individual analysis, and closed-form solutions are rare
• Chaos and complexity emerge here: the Lorenz system and logistic equation with harvesting exhibit sensitive dependence on initial conditions

---

Homogeneous vs. Inhomogeneous

This classification asks: is there an external forcing function, or is the system evolving on its own? The answer determines the structure of your general solution.

Homogeneous Differential Equations

• No forcing function: the right-hand side equals zero, like $y'' + 3y' + 2y = 0$
• Solutions form a vector space: the set of all solutions has the structure you studied in linear algebra (closure under addition and scalar multiplication)
• Fundamental solution sets: for an $n$th-order linear ODE, you need $n$ linearly independent solutions to span the solution space

Inhomogeneous Differential Equations

• External forcing present: the right-hand side is some nonzero function $g(x)$, like $y'' + 3y' + 2y = e^x$
• General solution = homogeneous + particular: find $y_h$ (complementary solution) and $y_p$ (particular solution), then combine as $y = y_h + y_p$
• Two main methods: undetermined coefficients works when $g(x)$ has a predictable form (polynomials, exponentials, sines/cosines); variation of parameters handles any continuous forcing function

---

Special First-Order Types

First-order equations come in several recognizable forms. Identifying the type immediately tells you which solution method to apply.

Separable Differential Equations

A separable equation is one you can rewrite so all $y$-terms are on one side and all $x$-terms are on the other: $f(y)\,dy = g(x)\,dx$.

Once separated, you integrate both sides independently: $\int f(y)\,dy = \int g(x)\,dx + C$. Always check separability first before trying more complex methods.

Exact Differential Equations

An equation in the form $M(x,y)\,dx + N(x,y)\,dy = 0$ is exact if it represents the total differential of some potential function $\phi(x,y)$.

• Exactness test: verify that $\frac{\partial M}{\partial y} = \frac{\partial N}{\partial x}$
• Solving it: find $\phi$ such that $\frac{\partial \phi}{\partial x} = M$ and $\frac{\partial \phi}{\partial y} = N$; the solution is $\phi(x,y) = C$
• If the test fails, sometimes an integrating factor (a function you multiply through) can make the equation exact

Bernoulli Differential Equations

These have the form $y' + P(x)y = Q(x)y^n$ and look nonlinear because of the $y^n$ term, but a substitution linearizes them.

• Substitution $v = y^{1-n}$ transforms the equation into a standard linear first-order equation in $v$
• Special cases: when $n = 0$ or $n = 1$, the equation is already linear; the Bernoulli technique only matters when $n \neq 0, 1$

---

Systems and Autonomy

When multiple quantities interact or when time doesn't appear explicitly, special structures emerge. These connect most directly to linear algebra through matrix methods.

Systems of Differential Equations

Multiple interrelated equations describe interacting quantities. For example, $\frac{dx}{dt} = ax + by$ and $\frac{dy}{dt} = cx + dy$ might model two competing populations.

• Matrix formulation $\mathbf{x}' = A\mathbf{x}$: eigenvalues and eigenvectors of $A$ determine solution behavior completely
• Solution via matrix exponential: the general solution is $\mathbf{x}(t) = e^{At}\mathbf{x}_0$, connecting differential equations to linear algebra's core concepts
• Eigenvalue types dictate behavior: real negative eigenvalues give decay (stable node), real positive give growth (unstable node), complex eigenvalues produce spirals or oscillations

Autonomous Differential Equations

• Independent variable absent: the equation $\frac{dy}{dt} = f(y)$ depends only on $y$, not explicitly on $t$
• Phase line analysis: equilibrium points occur where $f(y) = 0$; stability depends on the sign of $f'(y)$ at those equilibria (negative means stable, positive means unstable)
• Time-shift invariance: if $y(t)$ is a solution, so is $y(t - t_0)$ for any constant $t_0$

---

Problem Types: IVPs and BVPs

These classify problems by what conditions are specified, which determines whether a unique solution exists.

Initial Value Problems (IVPs)

• All conditions specified at one point: given $y(t_0) = y_0$ and possibly $y'(t_0) = v_0$ at the same starting time
• Existence and uniqueness: under reasonable continuity conditions (the Picard-Lindelöf theorem), exactly one solution passes through the initial point
• Forward evolution: you're finding how the system develops from a known starting state

Boundary Value Problems (BVPs)

• Conditions specified at different points: like $y(0) = a$ and $y(L) = b$ at the endpoints of an interval
• Solutions may not exist or may not be unique: this is a fundamental difference from IVPs
• Physical interpretation: BVPs often arise in spatial problems (temperature distribution in a rod, deflection of a beam) rather than time evolution

---

Quick Reference Table

---

Self-Check Questions

1. Given the equation $y'' + 4y = \cos(2t)$, classify it by order, linearity, and whether it's homogeneous or inhomogeneous. What solution method would you use?

2. Which two equation types both involve the superposition principle, and why does this principle fail for nonlinear equations?

3. Compare separable and exact differential equations: what must be true for each type, and which would you check for first when solving a first-order equation?

4. A system $\mathbf{x}' = A\mathbf{x}$ has eigenvalues $\lambda_1 = -2$ and $\lambda_2 = 3$. What does this tell you about the stability of the equilibrium at the origin? What type of equilibrium point is this?

5. You're modeling the temperature distribution in a metal rod of length $L$, with temperatures fixed at both ends. Is this an IVP or BVP? How would this differ from modeling the cooling of a cup of coffee over time?