The same melody, one octave higher
You spent the last rung mastering the second-order linear equation a y'' + b y' + c y = 0, and you came away with three pillars: solutions superpose, a pair of independent solutions forms a basis, and the general solution is c1 * y1 + c2 * y2. Now we let the highest derivative climb. An nth-order linear equation simply allows y up to y^(n), the nth derivative, with a coefficient on each term. The remarkable news of this guide — the reason rung five is mostly a victory lap — is that nothing about the *structure* breaks. The theory does not get harder as the order rises; it just gets wider.
Write the general object plainly: a_n * y^(n) + ... + a_1 * y' + a_0 * y = 0, where the coefficients may be constants or functions of x. Two words still do all the work, exactly as before. Linear means y and every derivative appear only to the first power, never multiplied together, never wrapped in a sine — so each y^(k) carries a coefficient and the terms are added, nothing more. Homogeneous means the right-hand side is exactly zero. Hold onto those two words: they are the entire price of admission to the clean theory, at order 2 and at order 100 alike.
Superposition, untouched
Why does the structure survive? Because the one engine behind it — the superposition principle — never cared about the order in the first place. Suppose y1, y2, ..., yk all solve the homogeneous equation. Then any combination c1 * y1 + c2 * y2 + ... + ck * yk solves it too, for every choice of constants. The proof is the same single line you saw at order two: differentiation is linear, so plugging a combination into the left side just produces the same combination of the (zero) results, and a sum of zeros is zero. Order n changes how many derivatives you take; it changes nothing about why linearity lets solutions add.
And homogeneous still earns its keep for the same reason as before. If the right side were a nonzero g(x), adding two solutions would double it to 2 * g(x), and superposition would fail. Zero is the unique value that survives being scaled and added — so the homogeneous case is, once again, the foundation we lay first. The general solution structure you will rely on later, splitting any forced equation into 'complementary plus particular', rests entirely on this homogeneous backbone holding at order n.
Why exactly n dimensions
At order two the headline was that the solution set is a plane — a vector space of dimension two. The headline now writes itself: for an nth-order linear homogeneous equation the solution space has dimension n. The counting argument that gave you 'two' generalizes without a hiccup. An initial value problem for an nth-order equation needs n pieces of starting data — the value, the slope, the curvature, and so on: y(x0), y'(x0), up through y^(n-1)(x0). The existence-and-uniqueness theorem (carried over, with continuous coefficients and a nonzero leading term) guarantees exactly one solution for each such list of n numbers.
So choosing a solution is the same as choosing a list of n real numbers, and a space whose points correspond perfectly to such lists is n-dimensional. The slogan is worth memorizing because it is literally true: the order of the equation is the dimension of its solution space. A fifth-order linear equation has a five-dimensional space of solutions; you will need exactly five independent ingredients to describe them all, and five initial conditions to pin down one curve.
A basis of n functions and the Wronskian that certifies it
To describe an n-dimensional space you pick a basis of n independent vectors — and here the vectors are functions. A set of n solutions y1, ..., yn that are linearly independent is called a fundamental set of solutions. The moment you have one, you are finished: the general solution is y = c1 * y1 + ... + cn * yn, and every solution of the equation, without exception, is some choice of those n constants. Notice this is the order-two recipe with more seats at the table, not a new recipe.
At order two you could often check independence by eye — 'is one a constant multiple of the other?' — but with three, four, or five functions that hand-check becomes hopeless. The clean tool generalizes too: the Wronskian is now an n-by-n determinant, stacking each solution and its first n-1 derivatives into rows. If that determinant is nonzero at even a single point of the interval, the n solutions are independent and you genuinely have a fundamental set. Abel's identity still holds, so the Wronskian is either identically zero or never zero on an interval — one evaluation decides the whole question.
Equation: y''' - y' = 0 (third order, n = 3)
Three solutions: y1 = 1, y2 = e^x, y3 = e^(-x)
| 1 e^x e^(-x) |
W(x) = | 0 e^x -e^(-x) | = -2 (constant, never 0)
| 0 e^x e^(-x) |
Independent? W = -2 != 0 -> yes, a fundamental set
General sol.: y = c1*1 + c2*e^x + c3*e^(-x)How to actually find n solutions
Knowing the space is n-dimensional tells you *what* to hunt — n independent functions — but not how to catch them. For the workhorse case, constant coefficients, the order-two shortcut survives perfectly. Guess y = e^(rx); each derivative y^(k) brings down a factor r^k, so substituting turns the whole differential equation into a single polynomial equation in r of degree n. This is the characteristic polynomial, and by the fundamental theorem of algebra it has exactly n roots (counted with multiplicity). Calculus has, once more, been traded for algebra — only now the algebra is a degree-n polynomial instead of a quadratic.
The three root cases you learned at order two are still the entire vocabulary — they just appear in richer combinations. A simple real root r contributes e^(rx). A complex-conjugate pair contributes a cosine-and-sine pair via Euler's formula. The genuinely new wrinkle is multiplicity: a root repeated m times contributes not just e^(rx) but the whole family e^(rx), x * e^(rx), ..., x^(m-1) * e^(rx) — the extra powers of x are the higher-order echo of the lone x * e^(rx) partner you met for a repeated root at order two. Tally every root's contribution and you land on exactly n independent solutions, every time. The next guide is devoted to reading these root cases off the polynomial cleanly.
Two honest caveats close the loop. First, the e^(rx) gift is the reward of *constant* coefficients; for variable coefficients it generally fails. There is one classic exception you will meet in this rung — the Cauchy-Euler equation, whose x^k coefficients invite the guess y = x^r instead, recovering a degree-n polynomial in r by a different door. Second, even constant-coefficient algebra has teeth: a degree-n polynomial may have no nice closed-form roots once n is five or more, so in practice you sometimes find the roots numerically. The structure is exact; the arithmetic is not always kind.
Where this rung is heading
With the structure secured, the rest of the rung sharpens the *machinery*. A beautiful reframing helps: write D for d/dx, so D y = y', D^2 y = y'', and the whole equation becomes a polynomial in D acting on y. Then 'guess e^(rx)' and 'factor the characteristic polynomial' become the same act — you literally factor the operator. That algebra of operators powers the annihilator method for forced equations, gives variation of parameters its order-n form, and turns the Cauchy-Euler equation into a constant-coefficient one in disguise.
So hold the through-line of this whole rung in one breath: superposition makes the solutions a vector space, its dimension equals the order n, a fundamental set is a basis the Wronskian can certify, and for constant coefficients a degree-n characteristic polynomial hands you that basis root by root. Nothing here is new in spirit — it is the order-two theory you already trust, generalized. The remaining four guides simply hand you sharper tools to wield it: operators, annihilators, Cauchy-Euler, and variation of parameters at full generality.