How to solve the following ordinary differential equation?
Consider the ODE $$y' + x = \sqrt{x^2+y}.$$I have no idea how to solve this equation. I have tried several ways but neither of them worked. Can anybody help me? Many thanks.
I have tried substituting $u=\sqrt{x^2+y}$, as well as squaring both sides of the ODE.
$\endgroup$ 82 Answers
$\begingroup$Let us assume $g(x)\geq 0$ and $y=x^2 g(x)^2-x^2$. We have $y'=2x g(x)^2+2x^2 g'(x)g(x)-2x$ and $\sqrt{x^2+y}=x g(x)$, hence:
$$2 g(x)^2+2x g'(x)g(x)-1 = g(x), $$$$x g'(x) = \frac{1}{2}+\frac{1}{2 g(x)}-g(x). $$Now let $h(t)=g(e^t)$. We have $h'(t) = e^{t} g'(e^t)$, so $g'(e^t)=h'(t)e^{-t}$ and by letting $x=e^t$ in the previous line$$ h'(t) = \frac{1}{2}+\frac{1}{2 h(t)}-h(t). $$This is a separable DE, which can be solved by finding the inverse function of $\int \frac{du}{\frac{1}{2}+\frac{1}{2u}-u}.$
The explicit solutions are a bit convoluted, since they depend on the roots of a cubic. One of them is
$$ y(x) = \frac{x^2}{4}\left[-4+\left(1+\frac{x}{\sqrt[3]{2-x^3+2\sqrt{1-x^3}}}+\frac{\sqrt[3]{2-x^3+2\sqrt{1-x^3}}}{x}\right)^2\right].$$
$\endgroup$ 1 $\begingroup$This is a heuristic derivation of the complete solution showing a possible motivation chain. Notice that Jack D'Aurizio has given a special solution before.
How can we simplify the ODE?
$$y' + x = \sqrt{x^2+y}\tag{1}$$
The first thing to do would be to get rid the square root. Hence the set $v^2 = x^2+y$ so that the r.h.s becomes just $v$. But this means that $y=v^2-x^2$ and $y'(x) = 2 v v'(x) - 2 x$ so that the the ODE becomes
$$2 v v' - x = v \tag{2}$$
Here we would like to get rid of $x$, hopefully by cancelling a factor $x$. To this end we try the natural substitution $v\to u x$ which gives
$$ 2 u x (u x)' -x = u x \implies 2 u (u x)' -1 = u \implies 2 u^2 +2 x u u' = u+1 \\\implies x u' = -u +\frac{1}{2}+\frac{1}{2u} $$
Now we can separate variables:
$$ x \frac{du}{dx} = -u +\frac{1}{2}+\frac{1}{2u} \implies \frac{du}{-u +\frac{1}{2}+\frac{1}{2u}} =\frac{dx}{x} $$
Integrating is elementary on both sides. Hence we find the following relation betwee $x$ and $u$
$$2 \left(-\frac{1}{3} \log (1-u)-\frac{1}{6} \log (2 u+1)\right)=\log (x) + const\tag{3}$$
Multiplying by $-3$ and simplifying the $\log$ gives the cubic equation for $u$
$$(1-u)^2(1+2u) = \frac{c^3}{x^3} \tag{4}$$
Here $c$ is a constant of integration.
Now $(4)$ can be solved by standard methods giving three solution for $u(x)$. We skip the explicit expression here for the time being as Jack has already given a special solution.
Finally, we have found that the ODE $(1)$ has three solutions given by
$$y(x) = x^2(u(x)^2-1)$$
explicitly
$$y(x,c) = \left\{x^2 \left(\frac{1}{4} \left(-\frac{g}{c x}-\frac{c x}{g}+1\right)^2-1\right),\\x^2 \left(\frac{1}{4} \left(\frac{g}{c x}+\frac{c x}{g}+1\right)^2-1\right),\\x^2 \left(-1+\frac{1}{16} \left(\frac{2 (-1)^{2/3} g}{c x}+\frac{c \left(x-i \sqrt{3} x\right)}{g}+2\right)^2\right)\right\} \tag{5a}$$
here
$$g = \sqrt[3]{c^3 x^3+2 \sqrt{1-c^3 x^3}-2}\tag{5b}$$
Here is a plot of the three solutions with three different parameters $c$
Notice the interesting cusp structure from which two of the three solutions emerge. The position of the cusp is given by $(x,y) = (\frac{1}{c} ,-\frac{1}{c^2})$
Discussion
- CAS
Mathematica solves the ODE with the command DSolve immediately, but it returns 6 solutions instead of 3. Three of these do not solve the ODE after explicit insertion. This confused my at first so I did not submit that solution. The reason is that Mathematica takes a an additional negative value of the square root into account, i.e. it simultaneously solves the ODE $y'+x = - \sqrt{x^2+y}$. The latter is obtained from $(1)$ by the substitution $x\to -x$.
- Modifications
We could simplify the ODE $(1)$ by subsequently modifying terms, as for instance
$$y'(x)=\sqrt{y(x)}\tag{m1}$$$$y'(x)=\sqrt{y(x)+x}\tag{m2}$$$$y'(x)=\sqrt{y(x)+x^2}\tag{m3}$$$$y'(x)+x=\sqrt{y(x)}\tag{m4}$$$$y'(x)+x=\sqrt{y(x)+x}\tag{m5}$$
All these ODEs can be transformed to a separable ODE which even can be integrated but the resulting implicit equations cannot be solved explicitly.
For example, the integral of $(m4)$ is
$$\frac{1}{2} \log \left(\frac{2 y(x)}{x^2}+\sqrt{\frac{y(x)}{x^2}}+1\right)-\frac{1}{\sqrt{7}}\text{arctan}\left(\frac{1+4 \sqrt{\frac{y(x)}{x^2}}}{\sqrt{7}}\right)=\log \left(\frac{c}{x}\right)\tag{i4}$$
$\endgroup$ 2More in general
"Zoraya ter Beek, age 29, just died by assisted suicide in the Netherlands. She was physically healthy, but psychologically depressed. It's an abomination that an entire society would actively facilitate, even encourage, someone ending their own life because they had no hope. Th…"