Chapter 2 · Introduction

Chapter 2 — Linear Ordinary Differential Equations

Linear equations of order two and higher: the structure of their solutions, how to find them, what their zeros do, and what changes when the data are given at two end points instead of one.

In 1743 Euler showed that an equation such as $y''+by'+cy=0$ can be solved completely by trying $y=e^{\lambda x}$: the differential equation collapses to the polynomial equation $\lambda^2+b\lambda+c=0$. A few decades later Lagrange, working on the perturbations of planetary orbits, found how to pass from the free equation to a forced one by letting the constants of the general solution vary. Together these two ideas — solutions of a linear equation form a vector space, and one particular solution of the forced equation is all that is missing — are the backbone of the theory of linear differential equations, which is the language of vibrations, circuits, beams, heat flow and quantum mechanics.

In the 1830s Charles Sturm and Joseph Liouville, studying heat conduction in a bar, turned the question around. Instead of fixing $y$ and $y'$ at one point, they prescribed conditions at both ends of an interval. Such boundary value problems may have no solution, exactly one, or infinitely many; for equations containing a parameter $\lambda$ they have nonzero solutions only for a discrete sequence of eigenvalues, and the corresponding eigenfunctions behave like the sines and cosines of Fourier series. The last section shows how to solve forced boundary value problems in one stroke with Green's function — the integral kernel that George Green introduced in 1828 for electrostatics. This chapter is the bridge from the first-order theory of Chapter 1 to the partial differential equations of Chapters 3 and 4, where separation of variables produces exactly these Sturm–Liouville problems.

Existence on all of J→n-dimensional solution space→Wronskian & Abel→eᵏˣ, xʳ, reduction of order→Sturm separation→y = y_c + y_p→Variation of parameters→Frobenius, Legendre, Bessel→Sturm–Liouville eigenvalues→Green's functions
🎯 By the end of this chapter you can…
  • state and use the existence and uniqueness theorem for linear equations, find the guaranteed interval from the normal form, and decide whether given functions can be solutions of a linear equation with continuous coefficients;
  • test linear independence with the Wronskian, compute Wronskians from Abel's formula without solving, and find a second solution by reduction of order;
  • solve constant-coefficient and Euler–Cauchy equations of any order, including repeated and complex roots, and read off boundedness, decay and oscillation;
  • use the Sturm separation theorem and the normal form $u''+Qu=0$ to locate and count zeros of solutions;
  • find particular integrals by undetermined coefficients, annihilators, the operator rules $1/f(D)$ and variation of parameters, and recognise resonance;
  • classify singular points, compute power-series and Frobenius solutions, and use the standard properties of Legendre polynomials and Bessel functions;
  • solve regular Sturm–Liouville problems, use orthogonality and the Rayleigh quotient, and expand functions in eigenfunctions (§2.3);
  • construct Green's functions for second-order boundary value problems and use them to solve forced problems (§2.4).
Teaser — One Initial Value Problem, Many Boundary Value Problems

For every $\lambda$ the initial value problem $y''+\lambda\,w(x)\,y=0$, $y(0)=0$, $y'(0)=1$ has exactly one solution on $[0,\pi]$ (§2.1). The boundary condition $y(\pi)=0$, however, is met only for special values of $\lambda$ — the eigenvalues. Press Play to increase $\lambda$ and watch the end value $y(\pi)$: each time it passes through $0$ (green flash) the curve gains one more zero inside $(0,\pi)$. With a non-uniform weight $w$ the eigenvalues are no longer the squares $1,4,9,\dots$, but the pattern of zeros is the same.

3.00
λ—
end value y(π)—
zeros inside (0, π)—
eigenvalues found so far—

How this is examined

Linear ODEs are among the most reliable sources of CSIR NET questions. In Part B you can expect to compute the Wronskian at one point from its value at another (Abel's formula), the general solution or a particular integral of a constant-coefficient or Euler equation, the correct trial form when the forcing resonates with a characteristic root, the largest interval of existence for a linear initial value problem, the indicial roots at a regular singular point, a value such as $\int_{-1}^1xP_n\,dx$ or $J_{1/2}$ in closed form, the eigenvalues of $y''+\lambda y=0$ under given boundary conditions, or the value of a Green's function at a point. Part C, where several options can be correct, tests statements: "these two functions can be solutions of a second-order linear equation", "the Wronskian can vanish at exactly one point", "between two zeros of one solution lies a zero of another", "every solution is bounded", "a logarithm must appear in the second Frobenius solution", "the boundary value problem has a unique solution", "the eigenfunctions are orthogonal with respect to this weight", "the Green's function is symmetric". The wrong options are almost always one of the standard counterexamples collected in the ⚠️ boxes of the four sections. GATE MA asks similar questions, often with a numerical answer.

Sections in this chapter

✅ Before you start

From Chapter 1 you need the existence and uniqueness theorem (§1.1) and, above all, the theory of linear systems in §1.3: reduction of an $n$th-order equation to a first-order system with the companion matrix, fundamental matrices, Liouville's formula and variation of parameters for systems — Chapter 2 builds directly on them. From the CSIR NET Unit 1 course: determinants and Cramer's rule, eigenvalues, power series and their radius of convergence, the intermediate value and Rolle's theorems, and the Gamma function. Comfort with complex numbers and Euler's formula $e^{i\theta}=\cos\theta+i\sin\theta$ is essential for complex characteristic roots. If this is your first visit, the welcome page explains how each section is organised.