The question is to find $x$ in:

\begin{equation*} x^2=e^x \end{equation*}

I know Newton's method and hence could find the approx as $x\approx -0.7034674225$ from

\begin{equation*} x_{n+1} = x_n - \dfrac{x_n^2-e^{x_n}}{2x_n-e^{x_n}} \end{equation*}

According to WolframAlpha:


They also say that $x=-2W(\dfrac{1}{2})$ which shows that it can be solved using some Lambert-W function...Can anyone tell me how to do this?

Thanks a lot!

P.S. - I studied a li'l bit of Lambert-W ... So i guess a detailed explanation would not be needed ... just the initial steps!

  • 1
    $\begingroup$ I edited the equation in order to have $x_n$ everywhere in the rhs. I hope you don't mind. $\endgroup$ – Claude Leibovici May 7 '15 at 10:11
  • $\begingroup$ @ClaudeLeibovici Haha...Thanks! Forgot it! $\endgroup$ – NeilRoy May 7 '15 at 10:12

$$ x^2=e^x\implies x/2=\pm\tfrac12e^{x/2}\implies-x/2\,e^{-x/2}=\pm\tfrac12 $$ Therefore, $$ x=-2\mathrm{W}\!\left(\pm\tfrac12\right) $$ Since $\mathrm{W}(x)$ is real only for $x\ge-\frac1e$, we only have one real solution: $$ x=-2\mathrm{W}\!\left(\tfrac12\right)=-0.70346742249839165205 $$

  • $\begingroup$ How did you know that $-2W(\dfrac{1}{2})$ approximates to $-0.703467...$? $\endgroup$ – NeilRoy May 7 '15 at 14:45
  • $\begingroup$ Mathematica has the LambertW function. I have also detailed an algorithm to compute $\mathrm{W}(x)$ in this answer. Five iterations of that algorithm give over $12$ places of accuracy. $\endgroup$ – robjohn May 7 '15 at 14:51

Write the equation as $x^2e^{-x} = 1$. Then $x^2e^{-x} = 4\left(\left(-\frac{x}{2}\right)e^{-\frac{x}{2}}\right)^2$.


The Lambert W function is the inverse of $xe^x$. We want to find the inverse of $e^x x^{-2}$, dividing by $x^2$.

Now: \begin{align} y &= x^{-2} e^x \\ y^{-0.5} &= x e^{-0.5x} &&\vee y^{-0.5} = -x e^{-0.5x} \\ -0.5y^{-0.5} &= -0.5x e^{-0.5x} &&\vee 0.5y^{-0.5} = -0.5x e^{-0.5x} \\ W(-0.5y^{-0.5}) &= -0.5x &&\vee W(0.5y^{-0.5}) = -0.5x \\ x &= -2W(-0.5y^{-0.5}) &&\vee x = -2W(0.5y^{-0.5}) \\ \\ \end{align}

Now we have y=1, so $x=-2W(-0.5)$ or $x=-2W(0.5)$. The first one is complex, so only the second one remains as real solution.


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.