Skip to content
Tutorials Center
My learning

Topics

Lesson 3 of 39 ยท Getting Started

Initial Value Problems, Existence and Uniqueness

Learn what an initial value problem is, when a solution exists and is unique (Picard-Lindelof), and see non-uniqueness and finite-time blow-up demonstrated in Python.

A differential equation by itself describes a family of curves. To pick one, you supply where the system starts: y(t0) = y0. An equation together with such a starting value is an initial value problem (IVP):

y' = f(t, y), y(t0) = y0.

Physically this says "here is the law of motion, and here is the state right now; what happens next?" Before spending effort on a method you should ask two honest questions: does a solution exist at all, and if so, is there only one? A model with two different futures from the same present is useless for prediction. This lesson gives the theorem that answers both questions and two examples where the answer is surprising.

One equation, one constant

The general solution of y' = 2*y is y = C*e**(2*t). The initial condition y(0) = 3 gives C*e**0 = 3, so C = 3 and the particular solution is y = 3*e**(2*t). For a second-order equation you need two conditions, usually y(t0) and y'(t0): position and velocity. In general, an n-th order equation needs n initial values.

Python Solve an IVP and check the condition

Press Run (or Ctrl + Enter) to see the result here.

Program input ยท what you'd type while the program runs

Type the answers before you press Run, one per line. Each time the program reads a line (with input()), it gets the next line from this box.

The existence and uniqueness theorem

The standard result, due to Picard and Lindelof, states: suppose f(t, y) and its partial derivative df/dy are continuous on a rectangle around the point (t0, y0). Then the IVP has a solution on some interval around t0, and that solution is unique.

Three details deserve emphasis.

  1. Local, not global. The theorem promises a solution on some interval, possibly tiny. It says nothing about how far the solution extends.
  2. Continuity of f gives existence. (This is Peano's theorem; the extra condition on df/dy buys uniqueness.)
  3. Uniqueness explains non-crossing. If two solutions touched at a point, they would be two solutions of the same IVP, which the theorem forbids. This is the algebraic reason direction-field curves (see the previous lesson) never cross.

For linear equations y' + p(t)*y = g(t) the story is better: if p and g are continuous on an interval containing t0, the solution exists and is unique on that entire interval. Trouble can occur only where the coefficients themselves misbehave.

When uniqueness fails

Consider y' = 2*sqrt(y), y(0) = 0. Here f(y) = 2*sqrt(y) is continuous, but df/dy = 1/sqrt(y) blows up at y = 0, so the uniqueness hypothesis fails right at the starting point. And indeed there are many solutions: y = 0 for all t, but also y = t**2 for t >= 0, and even "wait until time a, then lift off", y = (t - a)**2 for t >= a, y = 0 before. Check the second: y' = 2*t and 2*sqrt(t**2) = 2*t.

Python Two different solutions, one IVP

Press Run (or Ctrl + Enter) to see the result here.

Program input ยท what you'd type while the program runs

Type the answers before you press Run, one per line. Each time the program reads a line (with input()), it gets the next line from this box.

All three satisfy the equation and the same initial condition. The model cannot decide when the "motion" starts. Physically, such an equation is missing information.

When existence is only local: blow-up

Now take y' = y**2, y(0) = 1. This f is perfectly smooth, so a unique solution exists. Separating variables gives y = 1/(1 - t). It exists for t < 1 and shoots to infinity as t approaches 1; nothing continues it past that. This is finite-time blow-up, and it can happen for nonlinear equations even when every coefficient is nice. Its interval of existence depends on the initial value: from y(0) = 2 the solution is y = 2/(1 - 2*t), which dies at t = 0.5.

Python Watch a solution blow up

Press Run (or Ctrl + Enter) to see the result here.

Program input ยท what you'd type while the program runs

Type the answers before you press Run, one per line. Each time the program reads a line (with input()), it gets the next line from this box.

Blow-up is a warning that numerical output past a certain time may be meaningless. In practice, always check whether your solution stays bounded.

What to remember when modeling

  • State your IVP completely: equation plus all initial values.
  • If your f is differentiable in y, you may trust uniqueness locally.
  • Non-differentiable f (square roots, absolute values at zero) can break uniqueness.
  • Existence on a short interval is not existence forever.

The proof of the theorem builds a sequence of approximations y_(k+1)(t) = y0 + integral of f(s, y_k(s)) ds (Picard iteration) and shows it converges. It is a handy by-hand method for simple equations, and a nice way to see why df/dy bounded matters.

Check your understanding

Quick check ยท 4 questions

of 4 answered You scored / 4 ยท perfect!

  1. 1. How many initial conditions does a third-order IVP need?

    An n-th order equation has n arbitrary constants, so n initial values are required, typically y, y' and y'' at t0.
  2. 2. For y' = 2*sqrt(y), y(0) = 0, why does the theorem not guarantee uniqueness?

    Uniqueness needs df/dy continuous near the starting point; here it blows up at y = 0, and indeed several solutions exist.
  3. 3. The solution of y' = y**2, y(0) = 1 is y = 1/(1 - t). Where does it stop existing?

    As t approaches 1 the denominator goes to 0 and y goes to infinity, so the solution blows up in finite time.
  4. 4. For a linear equation y' + p(t) y = g(t) with p and g continuous on an interval I containing t0, the solution exists on

    Linear equations have global existence on the whole interval where the coefficients are continuous.

Summary

  • An IVP is an equation plus initial values; the number of values equals the order.
  • If f and df/dy are continuous near (t0, y0), a unique solution exists on some interval.
  • Uniqueness explains why solution curves do not cross; it can fail when df/dy is unbounded, as for y' = 2*sqrt(y).
  • Nonlinear solutions may blow up in finite time, as y' = y**2 does.
  • Linear equations with continuous coefficients have solutions on the whole interval.

Next: Modeling with Differential Equations