Flow Lab

Syllabus / Module 3: First-Order ODEs

Lesson 3.7

Existence & Uniqueness

When does an IVP have exactly one solution? The Picard–Lindelöf theorem (intuitively), Lipschitz conditions, non-uniqueness and blow-up.

Written byClaude Sonnet

🌊
Why this matters for simulation: Whether smooth solutions of 3D Navier–Stokes always exist is a $1M Millennium Prize problem.

By the end of this lesson you can…

  • State the existence and uniqueness theorem and read the Lipschitz condition as a bound on how fast the slope field tilts in y.
  • Show that y' = y^(1/3), y(0) = 0 has infinitely many solutions and locate where Lipschitz fails.
  • Explain why a smooth right side does not give global existence, and find a maximal interval of existence.
  • Use non-crossing of solutions to trap a solution between two others, and say where that breaks.
  • Carry out Picard iteration, and say what it and a numerical solver do at a non-unique IVP.
  • Say why the Lipschitz constant L controls how fast nearby solutions separate.
On this page
·
  1. What the question is
  2. Lipschitz: a speed limit on tilting
  3. The theorem, and what it does not say
  4. Non-uniqueness: y′=y1/3y' = y^{1/3}y′=y1/3
  5. The loose end from lesson 3.6
  6. Existence is local, even for a smooth fff
  7. Solutions cannot cross
  8. Picard iteration: building the solution
  9. What this means for a numerical solver
  10. Exercise: measuring L and running Picard
  11. Problems
  12. Key Takeaways

Every solver you have written assumes that the initial value problem has one answer, so that “the” solution is something Euler, RK4 and RK45 can all approximate. Nothing in the code checks it. You give the integrator ff and a starting point, and it returns numbers.

Two loose ends from earlier lessons are paid off here. In 3.3 Separable Equations the solution of y′=y2y' = y^2 reached infinity at a finite time, although y2y^2 is as smooth as a function can be. In 3.6 Exact Equations & Substitutions the equation y′+2y=3yy' + 2y = 3\sqrt y had two solutions from y(0)=0y(0) = 0, and a solver started at exactly 00 “sees slope 0” and never finds the second. One idea explains both: a speed limit on how fast the slope field may tilt.

▶ Video When an IVP has one solution: the Lipschitz condition, the cube-root equation with infinitely many solutions, finite-time blow-up, and Picard iterationcaptions only
Transcript
  1. A solver returns a number whatever you give it. Hand it y prime equals f of t and y, plus a start, and it never checks whether there is one answer, many, or one that stops early. When can we trust it?
  2. Fix t. The slope is a function of y. Lipschitz is a speed limit on how fast it tilts: through a point of the graph, lines of slope plus and minus L form a cone, and the graph must stay inside. The line y does. The cube root, vertical at zero, escapes.
  3. Take secants from the origin on grids of spacing 0.2, 0.02 and 0.002. Their slopes are 2.92, 13.57 and 62.996, and keep growing. Any L is eventually beaten: the cube root is not Lipschitz at zero.
  4. Now y prime equals cube root of y, from zero. Zero is a solution. So is a branch that waits until time c, then leaves: two thirds of t minus c, to the power three halves, either sign. At t equals one, c zero gives 0.5443, c one half gives 0.1925, and everything between plus and minus 0.5443 is reached.
  5. Existence can be local. y prime equals one plus y squared, from zero, has the solution tangent t, reaching infinity at pi over two, 1.5708, and backward at minus pi over two. That open interval is the maximal interval of existence.
  6. A smooth right side does this too. For y prime equals y squared the solution ends at one over y0: start higher, end sooner. The theorem never promised more than local.
  7. What does the theorem promise? On a box of half-height b, the slope is at most y0 plus b, squared, so the guaranteed length is b over that. For y0 one it peaks at b one, giving one quarter; in general one over four y0.
  8. Chain those guarantees along y equals one over one minus t. Each step lands on the curve, and the totals run 0.25, 0.4375, 0.5781, 0.6836, 0.7627. The full sum is exactly one over y0, the blow-up time. Local guarantees, chained, land on the wall.
  9. Solutions cannot cross. For y prime equals y times one minus y, zero and one are solutions, so a solution starting between them is trapped in the strip. Bounded, it cannot blow up: it exists for all time, and we never solved anything.
  10. Where Lipschitz fails, this changes. Run any cube-root branch backward: it reaches zero with slope zero and merges into the zero solution. Curves can merge exactly where f stops being Lipschitz.
  11. Picard iteration: feed a guess into one plus the integral of f. For y prime equals y, start at one, and get one plus t, then plus t squared over two, then the cubic term. These are the Taylor partial sums, converging to e to the t.
  12. Without Lipschitz, the machine picks a side. From a guess of zero it returns zero forever. From plus one thousandth, 200 passes give 0.5442 at t equals one, the c zero branch. One answer of infinitely many, silently.
  13. Nearby starts for y prime equals L y separate like e to the L t. From one and one plus ten to the minus eight, the gap at t equals two is ten to the minus eight times e to the two L. Euler's gap over that is 0.999, 0.975, 0.905 for L one, five, ten; for L ten it reached 4.39. Large L is a warning, and previews stiffness.
  14. Navier-Stokes has this shape. In three dimensions smooth solutions are known for a short time. For all time is open, a Millennium Prize worth one million dollars. In two dimensions, yes. As analogy only, y prime equals y squared is a local solution that ends.

What the question is

For the IVP y′=f(t,y)y' = f(t,y), y(t0)=y0y(t_0) = y_0, three things can go wrong: no solution, several solutions, or exactly one that exists only for a short time. The direction field of 3.2 Direction Fields & Euler's Method Revisited suggests “one curve through every point”, the picture Euler’s method relies on. The theorem below says when it is trustworthy.

Lipschitz: a speed limit on tilting

Fix tt. The slope f(t,y)f(t,y) is a function of yy alone, and moving from height y1y_1 to y2y_2 tilts the arrow from f(t,y1)f(t,y_1) to f(t,y2)f(t,y_2). We say ff is Lipschitz in yy with constant LL on a region if

∣f(t,y1)−f(t,y2)∣≤L ∣y1−y2∣for all y1,y2 in the region.|f(t,y_1) - f(t,y_2)| \le L\,|y_1 - y_2| \quad\text{for all } y_1, y_2 \text{ in the region.}

The left side divided by ∣y1−y2∣|y_1 - y_2| is the slope of a secant line of the graph of f(t,⋅)f(t,\cdot). So no secant is steeper than LL: the arrows cannot tilt faster than linearly as you move in yy.

When fyf_y exists, the mean value theorem gives f(t,y1)−f(t,y2)=fy(t,ξ)(y1−y2)f(t,y_1) - f(t,y_2) = f_y(t,\xi)(y_1 - y_2) for some ξ\xi between them. So if ∣fy∣≤L|f_y| \le L on the region, ff is Lipschitz with constant LL, and a continuous fyf_y is bounded on any small closed box. Two cautions. A continuous fyf_y is sufficient, not necessary: f(y)=∣y∣f(y) = |y| has no derivative at 00, yet every secant slope lies in [−1,1][-1,1], so L=1L = 1. And Lipschitz is not the same as differentiable. Failure is always a secant slope that grows without limit.

The theorem, and what it does not say

Existence and uniqueness (Picard and Lindelof). Let ff be continuous on the box R={∣t−t0∣≤a, ∣y−y0∣≤b}R = \{|t - t_0| \le a,\ |y - y_0| \le b\} and Lipschitz in yy on RR (for instance, fyf_y continuous on RR). Let M=max⁡R∣f∣M = \max_R |f|. Then the IVP has exactly one solution on ∣t−t0∣≤h|t - t_0| \le h, where h=min⁡(a, b/M)h = \min(a,\ b/M).

The shape of hh has a reason: the slope never exceeds MM, so in time hh the graph moves at most Mh≤bMh \le b and stays in the box where the hypotheses hold. The theorem is local. It guarantees some interval around t0t_0 and says nothing about the solution on the rest of the box, or for all tt. Continuity alone gives existence (Peano) but not uniqueness, as the next section shows.

◆ Interactive Start a solution anywhere and read off the Lipschitz constant of a box around the start

Loading simulation…

  • y′=yy' = y and y′=∣y∣y' = |y|. One thick curve from any start; the readout gives L=1L = 1 for every box, even for ∣y∣|y| at y=0y = 0, where fyf_y does not exist. Tick also draw solutions from other y and the curves never meet.
  • y′=y2y' = y^2 and 1+y21 + y^2. LL grows with the box, since fy=2yf_y = 2y. The dashed lines mark where the thick solution ends; move the start and the end moves.
  • y′=y1/3y' = y^{1/3} and ∣y∣\sqrt{|y|}. Click near the tt-axis (it snaps onto y=0y = 0): LL is unbounded for every box and several solutions leave one point. Click slightly above the axis: a small box has a finite LL and one solution.

Non-uniqueness: y′=y1/3y' = y^{1/3}

Take y′=y1/3y' = y^{1/3}, y(0)=0y(0) = 0, with the real cube root. Dividing by y1/3y^{1/3} loses y≡0y \equiv 0, which solves the equation. For y≠0y \ne 0,

y−1/3 dy=dt  ⟹  32 y2/3=t−c  ⟹  y=±(2(t−c)3)3/2,t≥c.y^{-1/3}\,dy = dt \;\Longrightarrow\; \tfrac32\,y^{2/3} = t - c \;\Longrightarrow\; y = \pm\left(\frac{2(t-c)}{3}\right)^{3/2}, \qquad t \ge c .

Check: for t>ct \gt c, y′=±(2(t−c)3)1/2y' = \pm\left(\frac{2(t-c)}{3}\right)^{1/2}, and so is y1/3y^{1/3}. Now set y=0y = 0 for t≤ct \le c and the branch for t≥ct \ge c. Both pieces have value 00 and slope 00 at t=ct = c, so the glued function solves the equation. For every c≥0c \ge 0 and each sign this is a solution of y(0)=0y(0) = 0, plus y≡0y \equiv 0: infinitely many through one point. At t=1t = 1 the branch c=0c = 0 gives (2/3)3/2=0.5443(2/3)^{3/2} = 0.5443, c=0.5c = 0.5 gives (1/3)3/2=0.1925(1/3)^{3/2} = 0.1925, and y≡0y \equiv 0 gives 00; as cc runs over [0,1][0,1] and the sign over both choices, y(1)y(1) takes every value in [−0.5443, 0.5443][-0.5443,\ 0.5443].

Where Lipschitz fails. fy=13∣y∣−2/3f_y = \tfrac13|y|^{-2/3} is infinite at y=0y = 0. The secant between 00 and ε\varepsilon has slope ε−2/3\varepsilon^{-2/3}, which is 100100 at ε=0.001\varepsilon = 0.001 and unbounded as ε→0\varepsilon \to 0.

Why that lets a solution leave zero. Suppose f(0)=0f(0) = 0 and ∣f(y)∣≤L∣y∣|f(y)| \le L|y| for yy near 00. For a solution with y(0)=0y(0) = 0, while it stays near 00,

ddt(e−2Lty2)=e−2Lt(2y y′−2Ly2)≤0,\frac{d}{dt}\Big(e^{-2Lt}y^2\Big) = e^{-2Lt}\big(2y\,y' - 2Ly^2\big) \le 0 ,

since 2yy′≤2∣y∣∣f(y)∣≤2Ly22yy' \le 2|y||f(y)| \le 2Ly^2. A nonnegative quantity that starts at 00 and never increases stays at 00: a Lipschitz equation cannot lift a solution off an equilibrium. At distance ε\varepsilon the speed is at most LεL\varepsilon, too small to start. For y1/3y^{1/3} the speed is ε1/3\varepsilon^{1/3}, which beats LεL\varepsilon for small ε\varepsilon.

The loose end from lesson 3.6

The same thing explains y′+2y=3yy' + 2y = 3\sqrt y, with right side F(y)=3y−2yF(y) = 3\sqrt y - 2y. The part −2y-2y has derivative −2-2, harmless. The part 3y3\sqrt y has derivative 32y−1/2\tfrac32 y^{-1/2}, infinite at 00, so the theorem is silent at y(0)=0y(0) = 0, and there are two solutions, y≡0y \equiv 0 and y=94(1−e−t)2y = \tfrac94(1 - e^{-t})^2. The plainest case is y′=∣y∣y' = \sqrt{|y|}: both y≡0y \equiv 0 and y=(t−c)2/4y = (t-c)^2/4 for t≥ct \ge c solve it.

A solver sees slope 0. Euler at y=0y = 0 computes F(0)=0F(0) = 0 and does not move, at any step size. Start at 10−1210^{-12} for y′=∣y∣y' = \sqrt{|y|} and Euler with h=0.01h = 0.01 reaches 0.22660.2266 at t=1t = 1, on its way to the branch t2/4=0.25t^2/4 = 0.25. Starting values of 00 and of a trillionth give 00 and 0.22660.2266, and neither run warns you.

Existence is local, even for a smooth ff

Now the other loose end. For y′=y2y' = y^2, fy=2yf_y = 2y is continuous everywhere, so the theorem gives a unique solution near every point. Yet the solution through y(0)=y0>0y(0) = y_0 \gt 0 is y=y0/(1−y0t)y = y_0/(1 - y_0 t), which blows up at t=1/y0t = 1/y_0. And y′=1+y2y' = 1 + y^2, y(0)=0y(0) = 0 has the solution y=tan⁡ty = \tan t, which exists on (−π/2,π/2)(-\pi/2, \pi/2) and ends at both sides. Through (t0,y0)(t_0, y_0) the solution is tan⁡(t−t0+arctan⁡y0)\tan(t - t_0 + \arctan y_0), so the interval has length π\pi whatever y0y_0 is.

The maximal interval of existence is the largest open interval around t0t_0 on which the solution is defined. It belongs to the solution, not the equation (lesson 3.3). A solution can end only by growing without bound or by reaching the edge of where ff is defined and locally Lipschitz; we use that standard fact without proof.

Why the theorem allows this. For y′=y2y' = y^2, y(0)=1y(0) = 1, the box ∣y−1∣≤b|y - 1| \le b has M=(1+b)2M = (1+b)^2, so the guaranteed length b/(1+b)2b/(1+b)^2 is largest at b=1b = 1: h=14h = \tfrac14, after which y=43y = \tfrac43. Repeat from there. With current value yky_k the best box has b=ykb = y_k and M=4yk2M = 4y_k^2, so the guaranteed step is 1/(4yk)1/(4y_k), after which the solution is 43yk\tfrac43 y_k. Values grow by 43\tfrac43 and steps shrink by 34\tfrac34, so the guaranteed time adds up to

14y0(1+34+(34)2+⋯ )=1y0,\frac{1}{4y_0}\left(1 + \tfrac34 + \left(\tfrac34\right)^2 + \cdots\right) = \frac{1}{y_0} ,

exactly the blow-up time (60 steps from y0=1y_0 = 1 sum to 0.999999970.99999997). The mechanism: L=2max⁡∣y∣L = 2\max|y| grows with height, so each application guarantees less, and the guarantees never sum past 1/y01/y_0.

The linear contrast. For y′+p(t)y=g(t)y' + p(t)y = g(t) the secant slope of g−pyg - py in yy is ∣p(t)∣|p(t)| at every height, so the box can be as tall as we like, and the formula of lesson 3.4 exists on every interval where pp and gg are continuous. Linear equations have global solutions; nonlinear ones in general do not.

Solutions cannot cross

Suppose two solutions u,vu, v of an equation satisfying the theorem meet at (t1,y1)(t_1, y_1). Both solve the IVP with that data, so they agree near t1t_1. The set where they agree is open, and closed by continuity, so they agree on their whole common interval. Two solution curves either never touch or coincide.

This is usable. For y′=y(1−y)y' = y(1-y), the constants y≡0y \equiv 0 and y≡1y \equiv 1 are solutions and fy=1−2yf_y = 1 - 2y is continuous. A solution with 0<y(0)<10 \lt y(0) \lt 1 starts between them and cannot cross either, so it is trapped in the strip for as long as it exists. Being bounded, it cannot blow up, so it exists for all tt. We learned this without solving anything. It is the seed of 3.8 Autonomous Equations, Equilibria & Stability, which turns equilibria and the strips between them into a theory of long-time behaviour.

Without uniqueness the picture changes. For y′=y1/3y' = y^{1/3}, every nonzero solution run backward reaches y=0y = 0 at its own cc, meeting the zero solution with the same slope 00, and merges into it (tick the other-starts box in the sim). Curves can touch and merge exactly where ff stops being Lipschitz.

✅ Concept check

Which right side fails to be Lipschitz in $y$ on every box containing $y = 0$?

✅ Concept check

For $y' = 1 + y^2$, $y(0) = 1$, at what time $t^*$ does the solution cease to exist? (Give a number.)

✅ Concept check

For $y' = \sin y$ (so $f_y = \cos y$ is continuous), two solutions satisfy $u(2) = v(2)$. What follows?

Picard iteration: building the solution

The proof of the theorem is a construction, and the construction is an algorithm. Integrating the ODE gives the equivalent integral equation y(t)=y0+∫t0tf(s,y(s)) dsy(t) = y_0 + \int_{t_0}^t f(s,y(s))\,ds. The unknown is on both sides, so treat the right side as a machine: feed in a guess, get a better one.

yn+1(t)=y0+∫t0tf(s,yn(s)) ds.y_{n+1}(t) = y_0 + \int_{t_0}^{t} f\big(s, y_n(s)\big)\,ds .

Worked example: y′=yy' = y, y(0)=1y(0) = 1. Start from y0(t)=1y_0(t) = 1:

y1=1+t,y2=1+∫0t(1+s) ds=1+t+t22,y3=1+∫0t(1+s+s22)ds=1+t+t22+t36.y_1 = 1 + t, \qquad y_2 = 1 + \int_0^t (1 + s)\,ds = 1 + t + \frac{t^2}{2}, \qquad y_3 = 1 + \int_0^t \Big(1 + s + \frac{s^2}{2}\Big)ds = 1 + t + \frac{t^2}{2} + \frac{t^3}{6} .

By induction yn=∑k≤ntk/k!y_n = \sum_{k \le n} t^k/k!, since integrating tk/k!t^k/k! gives tk+1/(k+1)!t^{k+1}/(k+1)!. The iterates are the Taylor partial sums of ete^t: each pass buys one more term.

Why it converges. Let Dn(t)D_n(t) be the largest gap ∣yn+1−yn∣|y_{n+1} - y_n| on [t0,t][t_0, t]. The Lipschitz bound inside the integral gives ∣yn+1−yn∣(t)≤L∫t0tDn−1 ds|y_{n+1} - y_n|(t) \le L\int_{t_0}^t D_{n-1}\,ds, and repeating, Dn≤D0 (L∣t−t0∣)n/n!D_n \le D_0\,(L|t - t_0|)^n/n!, the general term of the series for eL∣t−t0∣e^{L|t-t_0|}. Terms of a convergent series go to zero, so the iterates converge. For L∣t−t0∣<1L|t - t_0| \lt 1 each pass shrinks the gap: the machine is a contraction, with exactly one fixed point. That is existence and uniqueness at once.

Without Lipschitz. For y′=y1/3y' = y^{1/3}, y(0)=0y(0) = 0, the guess y0≡0y_0 \equiv 0 gives y1=∫0t0 ds=0y_1 = \int_0^t 0\,ds = 0: the zero function is a fixed point, and the machine reports y≡0y \equiv 0 forever. Start from a positive constant such as 10−310^{-3} and it converges to a different fixed point: 200 passes on a 2000-point grid give y(1)=0.5442y(1) = 0.5442, matching the c=0c = 0 branch value 0.54430.5443 up to grid error. A negative start gives −0.5442-0.5442. Without a contraction, each fixed point attracts some guesses, and the output depends on where you begin.

What this means for a numerical solver

(a) A solver returns a number, always. Euler on y′=y1/3y' = y^{1/3} from 00 gives 0.00.0 at t=1t = 1 for h=0.01h = 0.01 and h=0.001h = 0.001. From 10−1210^{-12} it gives 0.51900.5190 and 0.54140.5414, and from −10−12-10^{-12} the negatives. The true values at t=1t = 1 fill the interval [−0.5443, 0.5443][-0.5443,\ 0.5443], and the solver reported one point with no flag. This is lesson 3.3’s RK4 printing finite values past a blow-up: a finite output is not evidence that the IVP has an answer.

(b) LL sets how fast errors can grow. For two exact solutions in a region where ff has constant LL, ddt12(u−v)2=(u−v)(f(t,u)−f(t,v))≤L(u−v)2\frac{d}{dt}\tfrac12(u-v)^2 = (u-v)\big(f(t,u) - f(t,v)\big) \le L(u-v)^2, so ∣u−v∣|u - v| grows at most like ∣u0−v0∣ eLt|u_0 - v_0|\,e^{Lt}. For y′=Lyy' = Ly the gap is exactly δeLt\delta e^{Lt}, so the bound is attained. Run Euler on y′=Lyy' = Ly from 11 and 1+10−81 + 10^{-8} with h=0.001h = 0.001: at t=2t = 2 the gap divided by δe2L\delta e^{2L} is 0.9990.999 for L=1L = 1, 0.9750.975 for L=5L = 5 and 0.9050.905 for L=10L = 10, and for L=10L = 10 the gap of 10−810^{-8} has become 4.394.39. A large LL means nearby solutions fly apart: the first warning of the stiffness met as h<2/λh \lt 2/\lambda in 3.2 Direction Fields & Euler's Method Revisited and treated in 4.5 Stability Regions & Stiffness. This bound is claimed only for exact solutions (proved above) and for Euler on the linear equation (computed below); error bounds for general methods are not derived here.

(c) Near a non-Lipschitz point, convergence theorems stop applying. Their proofs contain LL. With L=∞L = \infty there is no bound, and (a) is the picture: which branch a solver selects depends on rounding noise.

🐍 PythonPlayground: Picard from different guesses, Euler on the cube-root equation, and error growth with L
⌘/Ctrl + Enter

Exercise: measuring L and running Picard

Implement a secant-slope estimate of the Lipschitz constant and the Picard machine on a grid, and use them on the lesson’s equations.

🧪 Coding exerciseExercise: Lipschitz estimate, Picard passes, and the gap between two Euler runs
⌘/Ctrl + Enter

Problems

✏️ Pen & paperReading a box●○○

For y′=t+y2y' = t + y^2, y(0)=0y(0) = 0, use the box ∣t∣≤1|t| \le 1, ∣y∣≤1|y| \le 1. (a) Find LL and MM. (b) What interval does the theorem guarantee? (c) RK4 gives y(0.5)=0.1266y(0.5) = 0.1266 and reaches 10610^6 near t=1.99t = 1.99. Does this contradict (b)?

Show worked solution

(a) fy=2yf_y = 2y, so L=2L = 2 works; ∣t+y2∣≤2|t + y^2| \le 2, so M=2M = 2.

(b) h=min⁡(1,1/2)=12h = \min(1, 1/2) = \tfrac12: one solution on ∣t∣≤12|t| \le \tfrac12.

(c) No. The theorem gives a lower bound, not a forecast; other boxes, or restarting as in the lesson, certify more.

✏️ Pen & paperTrapped, and where it ends up●●○

For y′=sin⁡yy' = \sin y, y(0)=1y(0) = 1: (a) name constant solutions that bracket y(0)y(0) and show the solution exists for all tt; (b) find lim⁡t→∞y\lim_{t\to\infty} y; (c) compute y(10)y(10) from the closed form ln⁡tan⁡(y/2)=t+ln⁡tan⁡(12)\ln\tan(y/2) = t + \ln\tan(\tfrac12).

Show worked solution

(a) y≡0y \equiv 0 and y≡πy \equiv \pi are solutions, and ∣fy∣=∣cos⁡y∣≤1|f_y| = |\cos y| \le 1 gives uniqueness, so the solution stays in (0,π)(0,\pi). Bounded, it cannot blow up: it exists for all tt.

(b) On (0,π)(0,\pi), sin⁡y>0\sin y \gt 0, so yy increases to a limit ℓ≤π\ell \le \pi. If ℓ<π\ell \lt \pi, then y∈[1,ℓ]y \in [1,\ell] for t≥0t \ge 0, where sin⁡y≥min⁡(sin⁡1,sin⁡ℓ)>0\sin y \ge \min(\sin 1, \sin \ell) \gt 0, so y′y' is bounded below by a positive constant and y→∞y \to \infty, a contradiction. Hence ℓ=π\ell = \pi.

(c) y(10)=2arctan⁡(tan⁡(12)e10)=3.141426y(10) = 2\arctan\big(\tan(\tfrac12)e^{10}\big) = 3.141426, matching RK4 with h=0.001h = 0.001. It is 0.0001660.000166 below π\pi, matching 2cot⁡(12)e−102\cot(\tfrac12)e^{-10}.

✏️ Pen & paperA crossing, and a range of futures●●●

Let y′=3y2/3y' = 3y^{2/3} (real cube root squared), y(0)=0y(0) = 0. (a) Show y=t3y = t^3 and y≡0y \equiv 0 both solve it. (b) They cross at the origin: does this contradict the lesson? (c) Show that y=0y = 0 for t≤bt \le b and y=(t−b)3y = (t-b)^3 for t≥bt \ge b solves it for each b≥0b \ge 0, and find all possible values of y(1)y(1).

Show worked solution

(a) For y=t3y = t^3, y1/3=ty^{1/3} = t, so 3y2/3=3t2=y′3y^{2/3} = 3t^2 = y' for either sign of tt. And 0=3⋅00 = 3\cdot 0.

(b) No. fy=2y−1/3f_y = 2y^{-1/3} is unbounded at 00, so the hypotheses fail there and crossing is allowed there only.

(c) For t>bt \gt b, y′=3(t−b)2=3y2/3y' = 3(t-b)^2 = 3y^{2/3}; at t=bt = b both pieces have value and slope 00. Since f≥0f \ge 0, every solution is nondecreasing, so y(1)≥0y(1) \ge 0. While y>0y \gt 0, ddty1/3=y′/(3y2/3)=1\frac{d}{dt}y^{1/3} = y'/(3y^{2/3}) = 1, so y(1)1/3≤1y(1)^{1/3} \le 1. The family gives y(1)=(1−b)3y(1) = (1-b)^3 for 0≤b≤10 \le b \le 1, filling out [0,1][0,1], which is therefore the set of possible values.

✏️ Pen & paperPicard on the tangent equation●●○

For y′=1+y2y' = 1 + y^2, y(0)=0y(0) = 0, start from y0≡0y_0 \equiv 0 and compute y1,y2,y3y_1, y_2, y_3. Compare with tan⁡t=t+13t3+215t5+17315t7+⋯\tan t = t + \tfrac13 t^3 + \tfrac{2}{15}t^5 + \tfrac{17}{315}t^7 + \cdots: through which power does each iterate agree?

Show worked solution

y1=ty_1 = t. y2=∫0t(1+s2) ds=t+13t3y_2 = \int_0^t (1 + s^2)\,ds = t + \tfrac13 t^3. y3=∫0t(1+(s+13s3)2)ds=∫0t(1+s2+23s4+19s6)ds=t+13t3+215t5+163t7y_3 = \int_0^t \big(1 + (s + \tfrac13 s^3)^2\big)ds = \int_0^t \big(1 + s^2 + \tfrac23 s^4 + \tfrac19 s^6\big)ds = t + \tfrac13 t^3 + \tfrac{2}{15}t^5 + \tfrac1{63}t^7.

y1y_1 agrees with tan⁡t\tan t through t1t^1, y2y_2 through t3t^3, y3y_3 through t5t^5: its t7t^7 coefficient is 163=0.0159\tfrac1{63} = 0.0159, not 17315=0.0540\tfrac{17}{315} = 0.0540. In general yny_n is correct through t2n−1t^{2n-1} (exact fractions confirm this up to n=5n = 5). The series has radius π/2\pi/2, the blow-up time, so these polynomials approximate the solution only inside its interval of existence.

Key Takeaways

  • The IVP has exactly one solution near t0t_0 if ff is continuous and Lipschitz in yy (for instance fyf_y continuous), on an interval of guaranteed length min⁡(a,b/M)\min(a, b/M). It is local.
  • Lipschitz means no secant of f(t,⋅)f(t,\cdot) is steeper than LL. A bound on ∣fy∣|f_y| suffices, but ∣y∣|y| is Lipschitz without being differentiable, and a continuous ff can fail it.
  • y′=y1/3y' = y^{1/3}, y(0)=0y(0) = 0 has infinitely many solutions. A Lipschitz equation cannot lift a solution off an equilibrium; y1/3y^{1/3} and y\sqrt y can, which is why y′+2y=3yy' + 2y = 3\sqrt y had a second solution.
  • Existence is local even for smooth ff: y′=y2y' = y^2 ends at 1/y01/y_0 and y′=1+y2y' = 1 + y^2 gives tan⁡t\tan t. Linear equations are global.
  • Solution curves cannot cross where uniqueness holds, so a solution between two others stays between them (lesson 3.8). Where Lipschitz fails, curves can merge.
  • Picard iteration is a contraction on a short interval; for y′=yy' = y it gives the Taylor series of ete^t.
  • A solver always returns numbers, and at a non-unique IVP it silently picks a branch. LL governs how fast nearby solutions separate, like eLte^{Lt}.
  • 3D Navier-Stokes: short-time smooth solutions are known, global ones are open. The same local-versus-global gap as y′=y2y' = y^2, as an analogy only.
Finished this lesson?

Mark it complete to track your progress (saved in this browser). Reading every section does this for you.

Where this leads

These later lessons build directly on this one: