Home » Nonlinear Differential Equations

Tag: Nonlinear Differential Equations

Find solutions of the Riccati equation y′ + y + y2 = 2

Consider the Riccati equation

    \[ y' + y + y^2 = 2. \]

This equation has two constant solutions. Starting with these and the previous exercise (linked above) find further solutions in the cases:

  1. If -2 \leq b < 1, find a solution on the interval (-\infty, +\infty) with y= b when x = 0.
  2. If b \geq 1 or b<-2, find a solution on the interval (-\infty, +\infty) with y = b when x = 0.

First, we find the two constant solutions. If u is constant then u' = 0 so

    \[ u' + u + u^2 = 2 \quad \implies \quad u^2+u-2 = 0 \quad \implies \quad u= 1 \text{ or } -2. \]

From Exercise 19 (linked above) we know we can obtain additional solutions y to the Riccati equation by

    \[ y = u + \frac{1}{v} \]

where v is a solution of

    \[ v' - v (P(x) + 2Q(x)u) = Q(x). \]

In the present case we have P(x) = Q(x) = 1, so v is the solution of either

    \[ v' -3v = 1 \qquad \text{or} \qquad v' + 3v = 1, \]

for the cases u = 1 and u=-2, respectively. Each of these is an first-order linear differential equation which we can solve using Theorem 8.3 (page 310 of Apostol). For the first one we have P(x) = 3, Q(x) = 1 and let a = 0, and c = v(0). Then we have

    \begin{align*}  v &= ce^{3x} + e^{3x} \int_0^x e^{-3t} \, dt \\  &= ce^{3x} + e^{3x} \frac{1}{3}( -e^{-3x} + 1) \\  &= ce^{3x} - \frac{1}{3} + \frac{1}{3}e^{3x} \\  &= \frac{e^{3x}(3c+1) -1}{3}. \end{align*}

This gives us the first solution

    \[ y_1 = u + \frac{1}{v} = \frac{e^{3x}(3c+1) + 2}{e^{3x}(3c+1) - 1}. \]

Evaluating the second differential equation, this time with P(x) = -3, Q(x) = 1, a = 0 and c = v(0) we have,

    \begin{align*}  v &= ce^{-3x} + e^{-3x} \int_0^x e^{3t} \, dt \\  &= ce^{-3x} + \frac{1}{3} - \frac{1}{3}e^{-3x} \\  &= \frac{3c-1+e^{3x}}{3e^{3x}}. \end{align*}

Therefore, the second solution is

    \[ y_2 = u + \frac{1}{v} = \frac{-2(3c-1) + e^{3x}}{(3c-1)+e^{3x}}. \]

Finally, for the specific cases in (a) and (b).

  1. We want -2 \leq b < 1, so we choose y_2. Then y_2(0) = b \implies c = \frac{1}{b+2}. (This follows since y(0) = -2 + \frac{1}{v(0)}.) Therefore, (3c-1) = -\frac{b-1}{b+2}, so,

        \begin{align*}  y_2 &= \frac{2 \left( \frac{b-1}{b+2} \right) + e^{3x}}{-\left( \frac{b-1}{b+2} \right) +e^{3x}} \\[9pt]  &= \frac{ \left( \frac{b+2}{b-1} e^{3x} + 2}{\left( \frac{b+2}{b-1} \right) e^{3x} -1} \\[9pt]  &= \frac{Ce^{3x} +2}{Ce^{3x} -1} \end{align*}

    where C = \frac{b-1}{b+2}.

  2. In this case we want b \geq 1 or b <-2, so we choose y_1. Since y_1(0) = b \implies c =\frac{1}{b-1} we have (3c+1) = \frac{b+2}{b-1}. Therefore,

        \begin{align*}  y_1 &= \frac{e^{3x} \left( \frac{b+2}{b-1} \right) + 2}{e^{3x} \left( \frac{b+2}{b-1} \right) -1} \\[9pt]  &= \frac{e^{3x} + 2 \left(\frac{b-1}{b+2}\right)}{e^{3x} - \left( \frac{b-1}{b+2} \right)} \\[9pt]  &= \frac{e^{3x} + 2C}{e^{3x} - C} \end{align*}

    where C = \frac{b-1}{b+2}.

Prove a formula for additional solutions of a Riccati equation given one solution

Consider a differential equation

    \[ y' + P(x) y + Q(x) y^2 = R(x) \]

called a Riccati equation. Prove that if u is a solution of this equation then there are additional solutions

    \[ y = u + \frac{1}{v} \]

where v satisfies a first-order linear differential equation.


Proof. Let u be a function satisfying

    \[ u' + P(x) u + Q(x)u^2 = R(x). \]

Further, let y = u + \frac{1}{v} where v satisfies a first-order linear differential equation. Then,

    \[ y = u + \frac{1}{v} \quad \implies \quad y' = u' - \frac{v'}{v^2}. \]

Therefore,

    \begin{align*}  y' + P(x)y + Q(x) y^2 &= \left( u' - \frac{v'}{v^2} \right) + \left( u + \frac{1}{v} \right)P(x) + \left( u + \frac{1}{v} \right)^2 Q(x) \\[9pt]  &= u' - \frav{v'}{v^2} + P(x) u + P(x) \frac{1}{v} + Q(x) u^2 + 2Q(x) \frac{u}{v} + Q(x) \frac{1}{v^2} \\[9pt]  &= R(x) - \frac{v'}{v^2} + P(x)\frac{1}{v} + 2Q(x) \frac{u}{v} + Q(x) \frac{1}{v^2}. \end{align*}

Thus, if

    \[ -\frac{v'}{v^2} + P(x)v + 2Q(x) \frac{v}{u} + Q(x) \frac{1}{v^2} = 0, \]

then y = u+ \frac{1}{v} is a solution of our equation. But,

    \begin{align*}   &&-\frac{v'}{v^2} + P(x) \frac{1}{v} + 2Q(x) \frac{u}{v} + Q(x) \frac{1}{v^2} &= 0 \\[9pt]  \implies && -v' + P(x) v + 2Q(x) uv + Q(x) &= 0 \\[9pt]  \implies && v' - v  (P(x) + 2Q(x) u) &= Q(x). \end{align*}

This is a first-order linear differential equation (since u is a function of x). Hence, y = u + \frac{1}{v} is a solution of

    \[ y' + P(x) y + Q(x)y^2 = R(x) \]

where v is the solution of the first-order linear differential equation

    \[ v' - v (P(x) + 2Q(x) u) = Q(x). \qquad \blacksquare \]

Find solutions of 2xyy′ + (1+x)y2 = ex for given initial values

Find all solutions of the nonlinear differential equation

    \[ 2xyy' + (1+x)y^2 = e^x \]

on the interval (0, +\infty) satisfying the initial conditions

  1. y = \sqrt{e} when x = 1;
  2. y = -\sqrt{e} when x = 1;
  3. a finite limit as x \to 0.

    From a previous exercise (Section 8.5, Exercise #13) we know that a function y = f(x) which is never zero on an interval I is a solution of the initial value problem

        \[ y' + P(x) y = Q(x) y^n \qquad \text{on } I, \qquad \text{with } (f(a))^k = b \]

    if and only if (f(x))^k = g(x) on I where g(x) = v is the unique solution of the initial-value problem

        \[ v' + kP(x) v = kQ(x) \qquad \text{on } I, \qquad \text{with } g(a) = b. \]

    1. In the present problem we have the equation

          \[ y' + \frac{1+x}{2x} y = \frac{e^x}{2x} y^{-1} \qquad \text{on } (0, +\infty) \qquad \text{with } y = \sqrt{e} \text{ when } x = 1. \]

      Therefore, to apply the previous exercise we have

          \[ P(x) = \frac{1+x}{2x}, \quad Q(x) = \frac{e^x}{2x}, n = -1,  k = 1 - n = 2. \]

      Hence, f(x) is a solution to the given equation if and only if (f(x))^2 = g(x) where g(x) is the unique solution to

          \[ v' + \frac{1+x}{x}v = \frac{e^x}{x} \qquad \text{with } g(1) = e. \]

      We can solve this using Theorem 8.3 (page 310 of Apostol), first computing

          \[ A(x) = \int_1^x \frac{1+t}{t} \,dt = \log x + x - 1, \]

      giving us

          \begin{align*}  g(x) &= e \cdot e^{-\log x - x +1} + e^{-\log x - x +1} \int_1^x \frac{e^t}{t} e^{\log t + t - 1} \, dt \\[9pt]  &= \frac{e^2}{xe^x} + \frac{e}{xe^x} \int_1^x e^{2t-1} \, dt \\[9pt]  &= \frac{e^2}{xe^x} + \frac{1}{xe^x} \int_1^x e^{2t} \, dt\\[9pt]  &= \frac{e^2}{xe^x} + \frac{1}{xe^x} \left( \frac{1}{2} e^{2t} \Bigr \rvert_1^x \right) \\[9pt]  &= \frac{e^2}{xe^x} + \frac{1}{xe^x} \left( \frac{e^{2x}}{2} - \frac{e^2}{2} \right) \\[9pt]  &= \frac{e^x + e^2 - x}{2x} \end{align*}

      Therefore,

          \[ (f(x))^2 = g(x) \quad \implies \quad f(x) = \left( \frac{e^x+e^{2-x}}{2x} \right)^{\frac{1}{2}}. \]

    2. From part (a) we have g(x) = (f(x))^2 which implies f(x) = \pm (g(x))^{\frac{1}{2}}. Since f(1) = -\sqrt{e} we then have

          \[ f(x) = - \left( \frac{e^x + e^{2-x}}{2x} \right)^{\frac{1}{2}}. \]

    3. Let \lim_{x \to 0} f(x) = b where b is some finite real number. Then, (using the calculations we already completed in part (a) and just changing the initial-values a and b)

          \begin{align*}  v &= be^{-x-\log x + a + \log a} + e^{-x - \log x + a + \log a} \int_a^x \let( \frac{e^t}{t} \right) \left( e^{t + \log t - a - \log a} \right) \, dt \\[9pt]  &=\frac{bae^a}{xe^x} + \left( \frac{1}{xe^x} \right) \int_a^x e^{2t} \, dt \\[9pt]  &= \frac{bae^a}{xe^x} + \frac{e^{2x} - e^{2a}}{2xe^x} \\[9pt]  &= \frac{e^{2x} - c}{2xe^x}. \end{align*}

      Then,

          \[ \lim_{x \to 0} \frac{e^{2x} - c}{2xe^x} = L \quad \implies \quad \lim_{x \to 0} (e^{2x} - c) = 0 \quad \implies \quadd c = 1. \]

      Therefore,

          \[ y^2 = \frac{e^{2x}-1}{2xe^x} = \frac{e^x - e^{-x}}{2x} = \frac{\sinh x}{x}. \]

Find all solutions of xy′ + y = y2x2 log x for given initial values

Find all solutions of the nonlinear differential equation

    \[ xy' + y = y^2 x^2 \log x \]

on the interval (0, +\infty) satisfying the initial conditions

    \[ y = \frac{1}{2} \qquad \text{when} \qquad x = 1. \]


From a previous exercise (Section 8.5, Exercise #13) we know that a function y = f(x) which is never zero on an interval I is a solution of the initial value problem

    \[ y' + P(x) y = Q(x) y^n \qquad \text{on } I, \qquad \text{with } (f(a))^k = b \]

if and only if (f(x))^k = g(x) on I where g(x) = v is the unique solution of the initial-value problem

    \[ v' + kP(x) v = kQ(x) \qquad \text{on } I, \qquad \text{with } g(a) = b. \]

In the present problem we have the equation

    \[ y' + \frac{1}{x}y = y^2 x \log x \qquad \text{on } (0, +\infty) \qquad \text{with } y = \frac{1}{2} \text{ when } x = 1. \]

Therefore, to apply the previous exercise we have

    \[ P(x) = \frac{1}{x}, \quad Q(x) = x \log x, n = 2,  k = 1 - n = -1. \]

Hence, f(x) is a solution to the given equation if and only if (f(x))^{-1} = g(x) where g(x) is the unique solution to

    \[ v' - \frac{1}{x} v = -x \log x \qquad \text{with } g(1) = 2. \]

We can solve this using Theorem 8.3 (page 310 of Apostol), first computing

    \[ A(x) = \int_1^x -\frac{1}{t} \, dt = - \log x \]

giving us

    \begin{align*}  g(x) &= 2e^{\log x} + e^{\log x} \int_1^x (-t \log t) \cdot e^{-\log t} \, dt \\  &= 2x - x \int_1^x \log t \, dt \\  &= 2x - x (x \log x - x + 1) \\  &= x^2 + x - x^2 \log x. \end{align*}

Therefore,

    \[ (f(x))^{-1} = g(x) \quad \implies \quad f(x) = \frac{1}{x^2+x-x^2 \log x}. \]

Find solutions of xy′ – 2y = 4x3y1/2 for given initial values

Find all solutions of the nonlinear differential equation

    \[ xy' - 2y = 4x^3y^{\frac{1}{2}} \]

on the interval (-\infty, +\infty) satisfying the initial conditions

    \[ y = 0 \qquad \text{when} \qquad x = 1. \]


From a previous exercise (Section 8.5, Exercise #13) we know that a function y = f(x) which is never zero on an interval I is a solution of the initial value problem

    \[ y' + P(x) y = Q(x) y^n \qquad \text{on } I, \qquad \text{with } (f(a))^k = b \]

if and only if (f(x))^k = g(x) on I where g(x) = v is the unique solution of the initial-value problem

    \[ v' + kP(x) v = kQ(x) \qquad \text{on } I, \qquad \text{with } g(a) = b. \]

In the present problem we have the equation

    \[ xy' - 2y = 4x^3 y^{\frac{1}{2}} \qquad \text{on } (-\infty, +\infty) \qquad \text{with } y = 0 \text{ when } x = 1. \]

Therefore, to apply the previous exercise we have

    \[ P(x) = -\frac{2}{x}, \quad Q(x) = 4x^2, n = \frac{1}{2},  k = 1 - n = \frac{1}{2}. \]

Hence, f(x) is a solution to the given equation if and only if (f(x))^{\frac{1}{2}} = g(x) where g(x) is the unique solution to

    \[ v' - \frac{1}{x}v = 2x^2 \qquad \text{with } g(1) = 0. \]

We can solve this using Theorem 8.3 (page 310 of Apostol), first computing

    \[ A(x) = \int_1^x -\frac{1}{t} \, dt = -\log x, \]

giving us

    \begin{align*}  g(x) &= e^{\log x} \int_1^x 2t^2 \cdot e^{-\log t} \,dt \\  &= x \int_1^x 2t \, dt \\  &= x \left( t^2 \Bigr \rvert_1^x \right) \\  &= x (x^2 - 1) \\  &= x^3 - x. \end{align*}

Therefore,

    \[ (f(x))^{\frac{1}{2}} = g(x) \quad \implies \quad f(x) = (x^3-x)^2. \]

Find solutions of y′ -y = -y2(x2 + x + 1) for given initial values

Find all solutions of the nonlinear differential equation

    \[ y' - y = -y^2 (x^2+x+1) \]

on the interval (-\infty, +\infty) satisfying the initial conditions

    \[ y = 1 \qquad \text{when} \qquad x = 0. \]


From a previous exercise (Section 8.5, Exercise #13) we know that a function y = f(x) which is never zero on an interval I is a solution of the initial value problem

    \[ y' + P(x) y = Q(x) y^n \qquad \text{on } I, \qquad \text{with } (f(a))^k = b \]

if and only if (f(x))^k = g(x) on I where g(x) = v is the unique solution of the initial-value problem

    \[ v' + kP(x) v = kQ(x) \qquad \text{on } I, \qquad \text{with } g(a) = b. \]

In the present problem we have the equation

    \[ y' - y = -y^2(x^2+x+1) \qquad \text{on } (-\infty, +\infty) \qquad \text{with } y = 1 \text{ when } x = 0. \]

Therefore, to apply the previous exercise we have

    \[ P(x) = -1, \quad Q(x) = -(x^2+x+1), n = 2,  k = 1 - n = -1. \]

Hence, f(x) is a solution to the given equation if and only if (f(x))^{-1} = g(x) where g(x) is the unique solution to

    \[ v' + v = x^2+x+1 \qquad \text{with } g(0) = 1. \]

We can solve this using Theorem 8.3 (page 310 of Apostol), first computing

    \[ A(x) = \int_0^x \,dt = x, \]

giving us

    \begin{align*}  g(x) &= e^{-x} + e^{-x} \int_0^x (t^2+t+1)e^t \, dt \\  &= e^{-x} + e^{-x} \left( e^t (t^2 - t + 2) \right) \Bigr \rvert_0^x \\  &= e^{-x} + e^{-x} ( x^2 e^x - xe^x + 2e^x - 2) \\  &= e^{-x} + x^2 - x + 2 - 2e^{-x} \\  &= x^2 - x + 2 - e^{-x}. \end{align*}

Therefore,

    \[ (f(x))^{-1} = g(x) \quad \implies \quad f(x) = \frac{1}{x^2-x+2-e^{-x}}. \]

Find solutions of y′ – 4y = 2exy1/2 for given initial values

Find all solutions of the nonlinear differential equation

    \[ y' - 4y = 2e^x y^{\frac{1}{2}} \]

on the interval (-\infty, +\infty) satisfying the initial conditions

    \[ y = 2 \qquad \text{when} \qquad x = 0. \]


From the previous exercise (Section 8.5, Exercise #13) we know that a function y = f(x) which is never zero on an interval I is a solution of the initial value problem

    \[ y' + P(x) y = Q(x) y^n \qquad \text{on } I, \qquad \text{with } (f(a))^k = b \]

if and only if (f(x))^k = g(x) on I where g(x) = v is the unique solution of the initial-value problem

    \[ v' + kP(x) v = kQ(x) \qquad \text{on } I, \qquad \text{with } g(a) = b. \]

In the present problem we have the equation

    \[ y' - 4y = 2e^x y^{\frac{1}{2}} \qquad \text{on } (-\infty, +\infty) \qquad \text{with } y = 2 \text{ when } x = 0. \]

Therefore, to apply the previous exercise we have

    \[ P(x) = -4, \quad Q(x) = 2e^x, n = \frac{1}{2}, k = 1 - n = \frac{1}{2}. \]

Hence, f(x) is a solution to the given equation if and only if (f(x))^{\frac{1}{2}} = g(x) where g(x) is the unique solution to

    \[ v' - 2 v = e^x \qquad \text{with } g(0) = \sqrt{2}. \]

We can solve this using Theorem 8.3 (page 310 of Apostol), first computing

    \[ A(x) = \int_0^x -2 \,dt = -2x, \]

giving us

    \begin{align*}  g(x) &= \sqrt{2} e^{2x} + e^{2x} \int_0^x 2e^t \cdot e^{-2t} \, dt \\  &= \sqrt{2} e^{2x} + e^{2x} (-e^{-x} + 1) \\  &= \sqrt{2} e^{2x} + e^{2x} - e^x. \end{align*}

Therefore,

    \[ (f(x))^{\frac{1}{2}} = g(x) \quad \implies \quad f(x) = \left( \sqrt{2} e^{2x} + e^{2x} - e^x \right)^2. \]

Find all solutions of a given initial value problem

Let the function v = g(x) be the unique solution of the initial-value problem

    \[ v' + kP(x)v = kQ(x) \]

with g(a) = b where k \neq 0 is a constant, P(x) and Q(x) are continuous functions on an interval I, and a \in I with b any real number. If n \neq 1 and k = n-1 prove that the function y = f(x) for which f(x) \neq 0 for all x \in I is a solution of

    \[ y' + P(x) y  = Q(x) y^n \qquad \text{on } I, \qquad \text{with } f(a)^k = b \]

if and only if (f(x))^k = g(x) for all x \in I.


Proof. (\Rightarrow) Assume y = f(x) is a solution of y' + P(x)y = Q(x) y^n. Let v = y^{1-n} = y^k. Then,

    \[ v' = ky^{k-1} y' \quad \implies \quad y' = \frac{v'}{ky^{k-1}}. \]

(This division is allowed since y \neq 0 on I implies y^{k-1} \neq 0 on I.) Therefore,

    \begin{align*}  y' + P(x) y = Q(x) y^n && \implies && \frac{v'}{ky^{k-1}} + P(x) y &= Q(x) y^n \\[9pt]  && \implies && \frac{v'}{ky^{k-1+n}} + P(x) y^{1-n} &= Q(x) \\[9pt]  && \implies && \frac{v'}{k} + P(x) y^k &= Q(x) &(k = 1-n) \\[9pt]  && \implies && v' + kP(x) v = kQ(x). \end{align*}

Thus, if y = f(x) is a solution of y' + P(x)y = Q(x)y^n then g(x) = v = y^k is the unique solution of

    \[ v' + kP(x)v = kQ(x) \qquad \text{with} \qquad g(a) = (f(a))^k = b. \]

(\Leftarrow) Conversely, if (f(x))^k = g(x) on I then

    \[ g'(x) = k(f(x))^{k-1} f'(x) \quad \implies \quad f'(x) = \frac{1}{k} g'(x) (f(x))^{1-k}. \]

Therefore,

    \[  f'(x) + P(x) f(x) = \frac{1}{k} g'(x)(f(x))^{1-k} + P(x) f(x) \]

Then, since we know by hypothesis that v = g(x) is the unique solution of v' + kP(x) v = kQ(x) we have

    \[ g'(x) = kQ(x) - kP(x) g(x). \]

Substituting this into the above equation (and noting that our assumption that (f(x))^k = g(x) implies (f(x))^{1-k} = \frac{f(x)}{g(x)}) we have

    \begin{align*}  f'(x) + P(x) f(x) &= \frac{1}{k} (kQ(x) - kP(x)g(x)) (f(x))^{1-k} + P(x) f(x) \\[9pt]  &= Q(x) (f(x))^{1-k} - P(x) g(x) (f(x))^{1-k} + P(x)f(x) \\[9pt]  &= Q(x) (f(x))^{1-k} - P(x) g(x) \frac{f(x)}{g(x)} + P(x) f(x) \\[9pt]  &= Q(x) (f(x))^{1-k} - P(x) f(x) + P(x) f(x) \\[9pt]  &= Q(x) (f(x))^{1-k} \\  &= Q(x) (f(x))^n. \end{align*}

Therefore y = f(x) is a solution of y' + P(x)y = Q(x)y^n with (f(a))^k = g(a) = b. \qquad \blacksquare