Differential-Equations Β· Unit 2 Β· Video 7 Β· Interactive Practice

The Wall the Equation Never Mentions: yβ€²=y2y' = y^2 and Its Private Singularities

IKey Formulas

FormulaNameWhat it tells you
yβ€²=y2y' = y^2The equationSeparable; no xx on the right
y=1cβˆ’xy = \dfrac{1}{c - x}General solutionOne curve for each constant cc
y(0)=y0β€…β€ŠβŸΉβ€…β€Šc=1y0y(0) = y_0 \;\Longrightarrow\; c = \dfrac{1}{y_0}Constant from the initial conditionWhich curve you are on
x=cx = cSingular point (the wall)Where the denominator vanishes and the solution ends

Key Insight: The equation contains no cc. The wall's position enters only through the initial condition, so a method that reads nothing but slopes cannot see it coming. The single exception is y≑0y \equiv 0, the solution lost when we divided by y2y^2 β€” the only one with no wall.

IISeparating the Variables

Separation costs one solution and produces a constant that turns out to locate a wall.

Step 1 β€” Write the derivative as a quotient

dydx=y2\frac{dy}{dx} = y^2

The right-hand side carries no xx: the slope at a point depends on its height alone, so the field is the same in every column.

IIIWhere the Wall Stands

The starting height alone fixes the constant, c=1/y(0)c = 1/y(0), and the solution ends at x=cx = c.

IVMarching Into the Wall

RK4 tracks the true solution to four decimals, then keeps marching where no solution exists.

xxRK4, h=0.01h = 0.01, from (0,1)(0,1)11βˆ’x\dfrac{1}{1-x}
00β€”11
0.50.5β€”22
0.90.9β€”1010
0.990.99β€”100100
1.001.00β€”undefined
1.011.01β€”βˆ’100†-100^{\dagger}
1.031.03β€”βˆ’33.3†-33.3^{\dagger}
2.002.00β€”βˆ’1†-1^{\dagger}
†^{\dagger} Past the wall these belong to a different solution; the one through (0,1)(0,1) ended at x=1x = 1.

πŸ’‘ Every method in this lecture adds hh times an average of values of y2β‰₯0y^2 \ge 0, so the computed values only climb: none of them can turn a positive yy negative, and the far branch below the axis stays out of reach. Refining the step makes the number reported at x=1x = 1 larger, not better β€” 163.98163.98 at h=0.05h = 0.05, 819.91819.91 at h=0.01h = 0.01, 8199.108199.10 at h=0.001h = 0.001.

VQuiz Questions

Problem 1 Β· Locating the Wall

Given: yβ€²=y2y' = y^2 with y(0)=4y(0) = 4 β€” find the value of xx at which this solution blows up.

βœ… Correct! c=1/y(0)=14c = 1/y(0) = \tfrac14, so the wall stands at x=14x = \tfrac14 and the solution exists only for x<14x < \tfrac14.
❌ The relation is inverted. Substituting x=0x = 0 into y=1cβˆ’xy = \frac{1}{c-x} gives y(0)=1cy(0) = \frac{1}{c}, so c=1/y(0)c = 1/y(0), not y(0)y(0).
❌ Not quite. The right-hand side really is smooth everywhere β€” that is exactly why the wall is invisible in the equation. The solution still ends at x=cx = c.
❌ Not quite. Start from y=1cβˆ’xy = \frac{1}{c-x}, put in x=0x = 0 and y=4y = 4, then ask where the denominator vanishes.
Show solution

Separating variables gives the family y=1cβˆ’xy = \dfrac{1}{c-x}.

Step 1 β€” Find cc from the initial condition:

4=y(0)=1cβˆ’0=1c⟹c=144 = y(0) = \frac{1}{c-0} = \frac{1}{c} \qquad\Longrightarrow\qquad c = \frac{1}{4}

Step 2 β€” Write the solution and find where the denominator vanishes:

y=114βˆ’x,14βˆ’x=0β€…β€ŠβŸΉβ€…β€Šx=14y = \frac{1}{\tfrac14 - x}, \qquad \tfrac14 - x = 0 \;\Longrightarrow\; x = \tfrac14

Starting higher means blowing up sooner: y(0)=4y(0) = 4 reaches its wall at x=14x = \tfrac14, while y(0)=1y(0) = 1 survives until x=1x = 1. The interval of existence is x<14x < \tfrac14.

Problem 2 Β· Asking for a Value That Is Not There

Given: yβ€²=y2y' = y^2 with y(0)=1y(0) = 1, so y=11βˆ’xy = \dfrac{1}{1-x} β€” what is y(2)y(2)?

βœ… Correct! The solution through (0,1)(0,1) lives only on x<1x < 1. Asking it for y(2)y(2) asks for something it does not have.
❌ That is the far branch. 11βˆ’2=βˆ’1\frac{1}{1-2} = -1 is a genuine point of y=11βˆ’xy = \frac{1}{1-x}, but on the piece with x>1x > 1 β€” a different solution of yβ€²=y2y' = y^2, not a continuation of this one across the wall.
❌ Not quite. ∞\infty is what a numerical march prints, not a value the solution takes. The solution simply stops at x=1x = 1.
❌ Not quite. Ask first whether x=2x = 2 lies inside the interval on which this solution exists.
Show solution

With c=1c = 1 the solution is y=11βˆ’xy = \dfrac{1}{1-x}, and its denominator vanishes at x=1x = 1:

y(0.5)=2,y(0.9)=10,y(0.99)=100,y(x)β†’+∞ asΒ xβ†’1βˆ’y(0.5) = 2, \qquad y(0.9) = 10, \qquad y(0.99) = 100, \qquad y(x) \to +\infty \text{ as } x \to 1^-

The formula 11βˆ’x\frac{1}{1-x} still returns numbers for x>1x > 1, and those numbers do lie on a solution curve β€” the branch running from βˆ’βˆž-\infty up toward the axis. But an initial-value problem asks for the one curve through (0,1)(0,1), and that curve is only the piece on x<1x < 1: integral curves cannot jump a singularity.

So the interval of existence is βˆ’βˆž<x<1-\infty < x < 1, and y(2)y(2) is undefined for this solution.

Problem 3 Β· A Start Away from the Vertical Axis

Given: a solution of yβ€²=y2y' = y^2 passing through (1,12)\left(1, \tfrac12\right) β€” find y(2)y(2).

βœ… Correct! Here c=3c = 3, so y=13βˆ’xy = \frac{1}{3-x}, the wall stands at x=3x = 3, and x=2x = 2 is comfortably inside the interval of existence.
❌ That uses c=1/y0c = 1/y_0, which only holds when the start is at x0=0x_0 = 0. Substituting x=1x = 1, y=12y = \tfrac12 into y=1cβˆ’xy = \frac{1}{c-x} gives cβˆ’1=2c - 1 = 2, so c=3c = 3 and the wall is at x=3x = 3, not x=2x = 2.
❌ Not quite. Substitute the point into y=1cβˆ’xy = \frac{1}{c-x} to get cc first, then evaluate the resulting formula at x=2x = 2.
Show solution

Step 1 β€” Find cc from the point (1,12)\left(1, \tfrac12\right):

12=1cβˆ’1⟹cβˆ’1=2⟹c=3\frac{1}{2} = \frac{1}{c-1} \qquad\Longrightarrow\qquad c - 1 = 2 \qquad\Longrightarrow\qquad c = 3

Step 2 β€” Write the solution:

y=13βˆ’xy = \frac{1}{3-x}

Step 3 β€” Check that x=2x = 2 is legal, then evaluate:

The wall sits at x=c=3x = c = 3, and the start x=1x = 1 lies to its left, so this solution exists on x<3x < 3 β€” which contains x=2x = 2.

y(2)=13βˆ’2=1y(2) = \frac{1}{3-2} = 1

Verify: yβ€²=1(3βˆ’x)2=(13βˆ’x)2=y2y' = \frac{1}{(3-x)^2} = \left(\frac{1}{3-x}\right)^2 = y^2 βœ“, and y(1)=12y(1) = \frac{1}{2} βœ“.

Problem 4 Β· Refining the Step

Given: RK4 on yβ€²=y2y' = y^2 from (0,1)(0,1) with h=0.01h = 0.01 prints 819.91819.91 at x=1x = 1, where this solution has already ceased to exist β€” what does h=0.001h = 0.001 print at x=1x = 1?

βœ… Correct! The printed value scales like 1/h1/h: 163.98163.98 at h=0.05h = 0.05, 819.91819.91 at h=0.01h = 0.01, 8199.108199.10 at h=0.001h = 0.001. A smaller step gives a bigger fiction.
❌ Wrong direction. The number grows like 1/h1/h, not like hh: dividing the step by ten multiplies the printed value by about ten.
❌ It is a property of hh, not of the method. The same RK4 code prints 163.98163.98 at h=0.05h = 0.05 and 8199.108199.10 at h=0.001h = 0.001.
❌ Convergence needs something to converge to. The solution through (0,1)(0,1) has no value at x=1x = 1 at all, so no refinement of the step can approach one.
❌ Not quite. Compare the two reported values you were given, and ask how the number behaves as the step shrinks.
Show solution

The last true step is the one that reaches the wall. Near x=1x = 1 the exact solution behaves like yβ‰ˆ11βˆ’xy \approx \dfrac{1}{1-x}, so at the final grid point before the singularity the height is of size 1/h1/h β€” and that is the size of whatever the march is holding when it arrives at x=1x = 1:

h=0.05:Β 163.98h=0.01:Β 819.91h=0.001:Β 8199.10h = 0.05:\ 163.98 \qquad h = 0.01:\ 819.91 \qquad h = 0.001:\ 8199.10

Each tenfold refinement multiplies the printed value by about ten, so h=0.001h = 0.001 prints roughly 82008200.

Why the error orders do not apply. The statements "Euler's error β‰ˆCh\approx Ch", "improved Euler β‰ˆC2h2\approx C_2h^2", "RK4 β‰ˆC4h4\approx C_4h^4" all compare the computed value with the value of the solution. Past x=1x = 1 there is no such value, so there is nothing for the estimates to be estimates of. One step later the machine reports about 101310^{13}, two more and it overflows to ∞\infty, and from then on every step adds ∞\infty to ∞\infty: at x=2x = 2 the answer comes back ∞\infty.

The real lesson. Nothing in yβ€²=y2y' = y^2 mentions cc. The slope function is finite and smooth at every point of the plane, so no amount of slope-reading β€” at one point per step, or four β€” can announce in advance where the solution stops existing.

Solved: 0 / 4