Ordinary differential equation - Advanced Topics and Applications of ODEs

Systems of Ordinary Differential Equations Introduction So far, you may have studied single ordinary differential equations involving one unknown function. But in real applications—whether modeling population dynamics, circuit behavior, or mechanical systems—we often need to track multiple quantities simultaneously, and these quantities interact with each other. This is where systems of ordinary differential equations come in. A system allows us to write down equations where several unknown functions are linked together, each one's rate of change depending on all the others. What is a System of ODEs? A system of ordinary differential equations is a collection of equations involving multiple unknown functions and their derivatives. Each equation relates the derivatives of one function to the other functions and their derivatives. Formally, if we have $m$ unknown functions $y1(x), y2(x), \ldots, ym(x)$, we write them together as a vector function: $$\mathbf{y}(x) = \begin{pmatrix} y1(x) \\ y2(x) \\ \vdots \\ ym(x) \end{pmatrix}$$ A system of ODEs can then be written in the explicit vector form: $$\mathbf{y}^{(n)} = \mathbf{F}\left(x, \mathbf{y}, \mathbf{y}', \mathbf{y}'', \ldots, \mathbf{y}^{(n-1)}\right)$$ Here, $\mathbf{y}^{(n)}$ denotes the vector of $n$-th derivatives, and $\mathbf{F}$ is a vector-valued function that describes how the highest derivatives depend on the independent variable $x$, the unknown functions, and their lower-order derivatives. Example: Consider a simple system of two equations: $$\frac{dy1}{dx} = y2$$ $$\frac{dy2}{dx} = -y1$$ This can be written in vector form as: $$\frac{d}{dx}\begin{pmatrix} y1 \\ y2 \end{pmatrix} = \begin{pmatrix} 0 & 1 \\ -1 & 0 \end{pmatrix}\begin{pmatrix} y1 \\ y2 \end{pmatrix}$$ This is a first-order system of dimension 2 (two unknowns) and order 1 (first derivatives are the highest). Matrix Differential Equations When a system is linear with coefficients that may depend on the independent variable, we can write it as a matrix differential equation: $$\frac{d\mathbf{X}}{dt} = \mathbf{A}(t)\mathbf{X} + \mathbf{B}(t)$$ Here, $\mathbf{X}$ is a matrix-valued function, $\mathbf{A}(t)$ is a coefficient matrix that may depend on time, and $\mathbf{B}(t)$ is a known matrix-valued forcing term. The most important case is when $\mathbf{B}(t) = \mathbf{0}$ (the homogeneous case): $$\frac{d\mathbf{X}}{dt} = \mathbf{A}(t)\mathbf{X}$$ When $\mathbf{A}$ is a constant matrix (independent of $t$), solving this system involves finding the eigenvalues and eigenvectors of $\mathbf{A}$, which is a fundamental technique you'll encounter repeatedly. Phase Portraits for Two-Dimensional Systems Understanding the qualitative behavior of solutions is just as important as finding exact formulas. For a two-dimensional system (two unknown functions), we can create a phase portrait, which visualizes how the system behaves. A phase portrait is a plot in the $(y1, y2)$ plane—called the state space or phase plane—where each solution curve (trajectory) shows how the system evolves. The independent variable $x$ (or $t$, if it represents time) is not shown on the axes; instead, it's implicit in how the curve progresses. Key features of phase portraits: Trajectories are the curves traced out by solutions as time progresses Direction field (or vector field) shows small arrows at each point indicating the direction in which a trajectory moves Equilibrium points are special points where $\frac{d\mathbf{y}}{dt} = \mathbf{0}$—the system is stationary Saddle points, nodes, spirals, and other structures tell you about stability and long-term behavior Phase portraits are powerful because they immediately reveal whether solutions grow unbounded, shrink to zero, oscillate, or approach a steady state—without solving the system explicitly. Solving Systems: Key Techniques Method of Undetermined Coefficients The method of undetermined coefficients is used to find a particular solution of a non-homogeneous linear system with constant coefficients. The strategy is: Guess a form for the particular solution based on the form of the forcing term $\mathbf{B}(t)$ Substitute this guess into the differential equation Solve for the unknown coefficients Example: For the system $$\frac{d\mathbf{y}}{dt} = \mathbf{A}\mathbf{y} + \begin{pmatrix} e^t \\ 0 \end{pmatrix}$$ you might guess a particular solution of the form $\mathbf{y}p = e^t \begin{pmatrix} a \\ b \end{pmatrix}$ where $a$ and $b$ are constants to be determined. Reduction to Quadratures Sometimes a system can be solved by finding combinations of the equations that allow reduction to quadratures—expressing the solution in terms of known functions and integrals. Why it matters: For linear systems with constant coefficients, reduction to quadratures is always possible, and we use matrix methods to achieve it. However, for most nonlinear systems, this is impossible. This highlights an important principle: linear systems are special and solvable, while nonlinear systems generally require numerical or qualitative methods. <extrainfo> Fuchsian Theory and Frobenius Method When a linear ODE has regular singular points (points where solutions may blow up or become singular in a controlled way), we can use the Frobenius method based on Fuchsian theory. This method seeks power-series solutions of the form: $$y = \sum{k=0}^{\infty} ak x^{k+r}$$ where $r$ is chosen to make the series work at the singular point. This is a sophisticated technique that extends the power-series method to handle equations that aren't solvable by simpler means. </extrainfo> Boundary Value Problems Up to this point, you may have studied initial value problems, where you specify the value of the solution (and its derivatives) at a single point and solve forward. A boundary value problem is different: you specify conditions at two or more points. Example: Find $y(x)$ satisfying: $$y'' + \lambda y = 0, \quad y(0) = 0, \quad y(\pi) = 0$$ Here, the conditions are given at both endpoints ($x=0$ and $x=\pi$). This is fundamentally different from initial conditions and leads to very different behavior—sometimes there are no solutions, sometimes infinitely many. Sturm–Liouville Theory Sturm–Liouville theory addresses a special and important class of boundary value problems. A Sturm–Liouville problem has the form: $$\frac{d}{dx}\left(p(x)\frac{dy}{dx}\right) + q(x)y + \lambda w(x)y = 0$$ with appropriate boundary conditions, where $\lambda$ is a parameter to be determined (called an eigenvalue). Why this matters: For each eigenvalue $\lambdan$, there is a corresponding eigenfunction $yn(x)$. The key insight is that these eigenfunctions form a complete orthogonal set—they're perpendicular to each other (in a weighted sense) and can be used as a basis to expand arbitrary functions: $$f(x) = \sum{n=1}^{\infty} cn yn(x)$$ This is similar to Fourier series, but more general. Sturm–Liouville problems appear throughout applied mathematics: in vibrating strings, heat conduction, quantum mechanics, and more. Key points to remember: Eigenvalues are discrete (a countable set of values) Eigenfunctions are orthogonal with respect to the weight function $w(x)$ Together, the eigenfunctions form a complete basis for solving PDEs The Laplace Transform for Differential Equations The Laplace transform is a powerful tool for solving linear ODEs with constant coefficients. The transform converts a differential equation into an algebraic equation, which is much easier to solve. The method: Take the Laplace transform of both sides of the differential equation Solve the resulting algebraic equation for $\mathcal{L}\{y\}$ (the transform of the solution) Apply the inverse Laplace transform to recover $y(x)$ Why it's useful: Initial conditions are automatically incorporated Products of functions become convolutions (simpler to handle) Tables of transforms make the process efficient Works for equations that might be tedious with other methods Example: For $y'' + y = \sin(x)$ with $y(0) = 0, y'(0) = 0$, transforming gives: $$s^2 Y(s) + Y(s) = \frac{1}{s^2+1}$$ Solving for $Y(s)$ and inverting gives the solution directly. Summary Systems of ODEs are essential for modeling real-world phenomena where multiple quantities interact. The techniques you've learned—matrix methods, Sturm–Liouville theory, the Laplace transform, and phase portrait analysis—each provides a different lens for understanding and solving these systems. Linear systems, especially those with constant coefficients, have rich structure and can be solved systematically. Nonlinear systems are harder but can be understood qualitatively through phase portraits and equilibrium analysis. Master these tools, and you'll be able to tackle dynamics problems across engineering, physics, biology, and more.