Consider the first order linear ODE

$F(x) + A x F'(x) + B = 0$, with initial condition $F(1) = 1$. Moreover, let $A,B \neq 0$, real, and $sgn(A) = sgn(B)$.*

*... if the latter matters.

Mathematica gives me the following solution, which is easy to verify:

$F(x) = (1+B)(x^{-1/A}) - B$.

My question is: is there any elementary way or theorem to establish that the above must be the unique global solution to the given ODE? Do I have to apply the Picard–Lindelöf theorem (if that works), or is there an easier argument?

I have not much knowledge about how to solve differential equations; the question arises from an applied economics paper I'm working on.


You can't expect "global" uniqueness (for all real $x$). The reason is that the differential equation provides no information about $F'(0)$ and thus the portions for $x > 0$ and for $x < 0$ "don't know about each other".

Here's how to prove uniqueness on $(0, \infty)$.Suppose you have two solutions $F$ and $G$ of the same problem. Set $f(x) = F(x) - G(x)$ for $x > 0$, then $Axf'(x) + f(x) = 0$ and $f(1) = 0$. Now set $g(x) = x^{1/A}f(x)$. Then you can check that $g'(x) = 0$ for $x > 0$ and $g(1) = 0$. Therefore $g(x) = 0$ for all $x >0 $ and therefore $F(x) = G(x)$ for all $x > 0$.

Here is an example for two solutions that agree for $x > 0$ but are different for $x < 0$. Consider the two functions $$ F(x) = \begin{cases} (1+B)x^{-1/A} - B \quad (x > 0)\\ B x^{-1/A} - B \quad (x < 0) \end{cases} $$ and $$ G(x) = (1+B)x^{-1/A} - B \quad (x \ne 0) \, . $$ Set $F(0) = G(0) = -B$. Both functions satisfy the differential equation and the condition $F(1) = 1$, but they differ for $x < 0$.

  • $\begingroup$ Many thanks for providing this elegant proof. I particularly like the trick of specifying $g(x) = x^{1/A)f(x)$ in order to exploit the properties of the product rule. Also, it's good to know that the solution is in fact unique on the domain $x > 0$, as that is all that matters for my problem. $\endgroup$ – Martin Jun 4 '13 at 15:40

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.