Differential Equations on the Putnam

Study Sheet

Differential Equations on the Putnam

Separable and linear equations, uniqueness by integrating factor, blow-up, and Grönwall

The ODE Toolkit the Exam Assumes

Tip
Separable and first-order linear equations

Separable: y=f(x)g(y)dyg(y)=f(x)dxy' = f(x)g(y) \Rightarrow \int\tfrac{dy}{g(y)} = \int f(x)dx. Linear: y+P(x)y=Q(x)y' + P(x)y = Q(x) is solved by the integrating factor μ=eP\mu = e^{\int P}, which makes (μy)=μQ(\mu y)' = \mu Q.

The Putnam never asks you to solve a hard ODE; it asks you to recognize that a functional condition IS an ODE, solve the easy ODE, and then argue about existence, uniqueness, or a bound. y=y2y' = y^2 with y(0)=1y(0) = 1 gives y=11xy = \tfrac{1}{1-x} — and the point of the problem is the blow-up at x=1x = 1: no solution exists on all of R\mathbb R. Likewise f=1+f2f' = 1 + f^2 forces f=tanf = \tan, which cannot be defined on an interval longer than π\pi.

Tip
Uniqueness by integrating factor, and Grönwall

If f=ff' = f then (fex)=0(fe^{-x})' = 0, so f=Cexf = Ce^x. If fff' \le f and f(0)=1f(0) = 1, then fexfe^{-x} is non-increasing, so f(x)exf(x) \le e^x for x0x \ge 0 (Grönwall).

Multiplying by exe^{-x} converts a differential equation into "a derivative is zero" and a differential INEQUALITY into "a function is monotone" — the two-line arguments graders want in place of "obviously the solution is exe^x". For second-order equations the analogous tool is an energy: for y+y=0y'' + y = 0, E=y2+y2E = y^2 + y'^2 has E=2y(y+y)=0E' = 2y'(y + y'') = 0, so a solution with y(0)=y(0)=0y(0) = y'(0) = 0 has E0E \equiv 0 and is identically zero — uniqueness for free.

Tip
Constant-coefficient linear equations

y+py+qy=0y'' + py' + qy = 0 has solutions erxe^{rx} for the roots rr of r2+pr+q=0r^2 + pr + q = 0 (with xerxxe^{rx} for a double root, and eaxcosbxe^{ax}\cos bx, eaxsinbxe^{ax}\sin bx for r=a±bir = a \pm bi).

The solution space is two-dimensional (a solution is determined by y(0),y(0)y(0), y'(0) — that is the uniqueness theorem), so two independent solutions span everything. The same characteristic-root method solves linear recurrences, which is why the "Sequences" and "Differential Equations" units share a proof.

Side note
What the Putnam does with ODEs

Typical asks: show a function satisfying an f(x)=f(x+1)f'(x) = f(x + 1)-type condition has a specific form; bound f(2)f(2) given fff' \le f; show a solution of y=y2+xy' = y^2 + x cannot exist on all of R\mathbb R (compare with yy2y' \ge y^2 after some point and use blow-up); compute 01f\int_0^1 f for the unique ff with f=ff' = f, f(0)=1f(0) = 1. Existence and uniqueness (Picard) may be cited; blow-up must be argued.

Try it
Try it: an integrating factor

Solve y+1xy=xy' + \tfrac{1}{x}y = x with y(1)=1y(1) = 1 and evaluate y(2)y(2).

μ=x\mu = x: (xy)=x2(xy)' = x^2, so xy=x33+Cxy = \tfrac{x^3}{3} + C, C=23C = \tfrac23, and y(2)=12(83+23)=53y(2) = \tfrac12\left(\tfrac83 + \tfrac23\right) = \tfrac53.

Proofs & Why It Matters

Tip
Proof: Grönwall's inequality

If u(x)βu(x)u'(x) \le \beta u(x) on [0,T][0, T] with u(0)=u0u(0) = u_0, then u(x)u0eβxu(x) \le u_0e^{\beta x}.

Let v(x)=u(x)eβxv(x) = u(x)e^{-\beta x}. Then v=(uβu)eβx0v' = (u' - \beta u)e^{-\beta x} \le 0, so vv is non-increasing and v(x)v(0)=u0v(x) \le v(0) = u_0, i.e. u(x)u0eβxu(x) \le u_0e^{\beta x}. \blacksquare Significance: Grönwall is how uniqueness of ODE solutions is actually proved (apply it to the square of the difference of two solutions), and it is the model for every "differential inequality \Rightarrow growth bound" argument.

Tip
Proof: blow-up for yy2y' \ge y^2

If yy is differentiable with yy2y' \ge y^2 and y(0)=c>0y(0) = c > 0, then yy cannot exist on [0,1/c][0, 1/c].

Since yy20y' \ge y^2 \ge 0, yy is non-decreasing, so y>0y > 0 wherever it exists and (1y)=yy21\left(-\tfrac1y\right)' = \tfrac{y'}{y^2} \ge 1. Integrating from 00 to xx: 1y(x)+1cx-\tfrac{1}{y(x)} + \tfrac1c \ge x, i.e. 1y(x)1cx\tfrac{1}{y(x)} \le \tfrac1c - x, which is impossible once x1cx \ge \tfrac1c (the left side is positive). \blacksquare Significance: this comparison is the standard way to show that solutions of y=y2+(anything0)y' = y^2 + (\text{anything} \ge 0) have finite lifetime — a favorite Putnam "prove that no such function exists on all of R\mathbb R."

Formulas, Proofs & Tips

Tip
Separable differential equations
dydx=g(x)h(y)  dyh(y)=g(x)dx\frac{dy}{dx}=g(x)h(y) \ \Longrightarrow\ \int\frac{dy}{h(y)}=\int g(x)\,dx

What it means. Get all the yy's on one side and all the xx's on the other, then integrate both sides.

Example. dydx=xy\tfrac{dy}{dx}=xy: dyy=xdxlny=x22+C\int\tfrac{dy}{y}=\int x\,dx\Rightarrow \ln|y|=\tfrac{x^2}{2}+C.

Why it works. Dividing by h(y)h(y) and multiplying by dxdx separates the variables; integrating both sides is valid because each side is the derivative of the same quantity with respect to xx (a chain-rule substitution in reverse).

Tip. Apply the initial condition to find CC before simplifying — it is usually much less algebra.