3
$\begingroup$

How to solve this differential equation efficiently?

$$\begin{align} 2yy'' &= y'^2 + y^2, \\ y(0) &= 1,\\y'(0) &= -1. \end{align}$$

My method includes setting $p=y'$, $t=\dfrac{p}{y}$,$u=t^2$, which turns the original equation into linear equation $$ \frac{\text{d}u}{\text{d}y}+\frac1y u=\frac1y. $$

It requires so much substitutions.

Are there any other methods?

$\endgroup$

5 Answers 5

6
$\begingroup$

Hint

For the time being, do not consider the conditions and switch variables. The differential equation becomes $$2y\,x''(y)+x'(y) \left(y^2\, x'(y)^2+1\right)=0.$$ Reduction of order $p(y)=x'(y)$ leads to $$p(y)=\pm \frac 1 {\sqrt{y(y+c_1) }}=x'(y),$$ which is simple to integrate and to inverse.

Now, use the conditions (hoping that they are consistent).

Edit

Staying with the original equation, use the ansatz $$y(x)=z(x)\, e^{a x^b}$$ and search for $(a,b)$ which eliminate most of the terms. You should end with a very simple differential equation.

$\endgroup$
3
  • $\begingroup$ Thanks but I still do not understand how you change the original equation to $x(y)$-land, especially for $x''(y)$ part. I know how to change $\frac {\text{d}y}{\text{d}x}$ to $\frac {\text{d}x}{\text{d}y}$ by imagining $\text{d}x$ and $\text{d}y$ as separate parts, but not when larger order of derivatives appear. Could you explain that for me? $\endgroup$ Commented 2 days ago
  • 1
    $\begingroup$ @Harold Look at math.stackexchange.com/questions/566507/… $\endgroup$ Commented 2 days ago
  • $\begingroup$ @Harold note that you can accept the answer by clicking the check mark next to it $\endgroup$ Commented yesterday
6
$\begingroup$

We have

$$ 2\frac{y''}{y}=\left(\frac{y'}{y}\right)^2+1 $$

but

$$ \left(\frac{y'}{y}\right)'=\frac{y''}{y}-\left(\frac{y'}{y}\right)^2 $$

hence

$$ 2\left(\frac{y'}{y}\right)^2+2\left(\frac{y'}{y}\right)'=\left(\frac{y'}{y}\right)^2+1 $$

or

$$ 2\left(\frac{y'}{y}\right)'+\left(\frac{y'}{y}\right)^2-1=0 $$

Solving now

$$ 2u'+u^2-1=0\Rightarrow u = \frac{e^t-c_1}{e^t+c_1} $$

and finally

$$ y' = \frac{e^t-c_1}{e^t+c_1}y\Rightarrow y = c_2 e^{-t} \left(e^t+c_1\right)^2 $$

$\endgroup$
2
  • $\begingroup$ It's not clear in your solution how to determine $c_1$ and $c_2$ for the initial conditions $y(0)=1$ and $y'(0)=-1$. $\endgroup$ Commented yesterday
  • $\begingroup$ Thanks for the observation. I think now, after considering $c_1$ instead of $e^{2c_1}$ the initial conditions can be handled easier. $\endgroup$ Commented yesterday
5
$\begingroup$

If you wish to avoid substitutions you can integrate directly. Multiply by $y'/y^2$ to get an exact equation (when integrating I denote $y':=\mathrm dy(\xi)/\mathrm d\xi$ where $\xi$ is a dummy variable): \begin{align} \int_0^x\left( \frac{2y'}{y}y''-\frac{y'^3}{y^2}-y'\right) \mathrm d\xi=0,\\\\ \frac{y'^2}{y}-\frac{(-1)^2}{1}-y+1=\frac{y'^2}{y}-y=0, \end{align} and so \begin{align} y'&=-y,\\\\ \int_0^x\frac{y'}{y}\mathrm d\xi&=-\int_0^x \mathrm d\xi,\\\\ \log y-\log 1=- (x-0)&\implies y=e^{- x}. \end{align}

$\endgroup$
0
3
$\begingroup$

For this problem alone, there is an interesting short-cut. Try to guess what $y$ might be from the given condition. We know that, $y(0) = 1$ while $y'(0) = -1$. This means whatever the function is, the derivative of it yields the same value with the negative sign. The function that has the same value as its derivative is $e^x$. So, this function $y$ should be $e^{-x}$.

Do a little sanity check.

$y(0) = e^{-0} = 1$, $y'(0) = -e^{-0} = -1$, and $y'' = e^{-x}$

Substitute it in your equation,

$2yy'' = y'^2 + y^2$

$2e^{-x}e^{-x} = (-e^{-x})^2 + (e^{-x})^2$

$2e^{-2x} = e^{-2x} + e^{-2x} = 2e^{-2x}$

While you may need to add some constant $C$ to the function since you would be integrating it. But there you go. Probably the easiest way.

New contributor
Goldenprime is a new contributor to this site. Take care in asking for clarification, commenting, and answering. Check out our Code of Conduct.
$\endgroup$
1
  • 1
    $\begingroup$ This is a great way to deal with it! Actually the answer uses a similar guess after it derives $$2ytt'=1-t^2,$$ remember that $t=\dfrac p y$, where it finds $t=-1$. $\endgroup$ Commented 2 days ago
2
$\begingroup$

Another method using substitution (which I do not think I've seen mentioned above).

Since this ode is missing $x$, then let \begin{align*} \frac{dy}{dx}=p \end{align*} And now \begin{align*} y''&=\frac{dp}{dx}\\ &=\frac{dp}{dy} \frac{dy}{dx}\\ &=p' p \end{align*}

The original ode becomes \begin{align*} 2 y y'' &= (y')^2 + y^2\\ 2y p' p &=p^2 + y^2\\ p' &= \frac{p}{2y}+ \frac{y}{2p} \\ p' - \frac{p}{2y} &= \frac{y}{2p} \end{align*}

Which has the form $ p' + Q(y) p = R(y) p^n $ where here $Q(y)=-\frac{1}{2y}$ and $R(y)=\frac{y}{2}$ and $n=-1$. Hence it is standard first order Bernoulli ode which can be solved for $p$ using standard Bernoulli ode algorithm.

Once $p$ is known (which will be a function of $y$), then $y$ is solved from the first order ode $\frac{dy}{dx}=p$

So this method requires solving two first order ode's.

$\endgroup$
1
  • $\begingroup$ It's a good idea to see it as a Bernoulli equation there. Actually your method and mine are similar in the beginning. But after I got $p'-\frac p{2y}=\frac y{2p}$ I went straight to another substitution $t=\frac p y$ without identifying that it is actually a Bernoulli equation. Thanks for your observation! $\endgroup$ Commented 21 hours ago

You must log in to answer this question.

Start asking to get answers

Find the answer to your question by asking.

Ask question

Explore related questions

See similar questions with these tags.