2
$\begingroup$

In a recent piece of homework I needed to solve an equation of the form $ae^x + bx +c = 0$ where $a,b$ and $c$ are constants. I could not do it; no matter how I tried I either went in circles or hit a brick wall. Can someone demonstrate a general method for solving equations of this form? Please show every step in your answers.

$\endgroup$
  • 1
    $\begingroup$ This article on exponential-linear equations focuses on these type of equations. $\endgroup$ – Arpan Apr 17 '15 at 7:10
  • $\begingroup$ See there : math.stackexchange.com/questions/479776/solving-aebxcxd-0?rq=1 Short answer, there are no easy solution. $\endgroup$ – Paul Picard Apr 17 '15 at 7:10
  • $\begingroup$ @PaulPicard Hmm. Well the solution can't be a complex number (since that wouldn't make sense in the context of the question), so I guess I've done something wrong. Thanks anyway :) $\endgroup$ – imulsion Apr 17 '15 at 7:13
  • 1
    $\begingroup$ See Lambert W function. $\endgroup$ – Lucian Apr 17 '15 at 7:33
  • 1
    $\begingroup$ @imulsion: Who said that there's only one solution ? And why, pray tell, can't at least some of them be complex ? And how on earth would you even know their nature, since the values of the three parameters are not even given ? $\endgroup$ – Lucian Apr 17 '15 at 7:41
3
$\begingroup$

\begin{align} a \exp(x)+b x+c&=0 \end{align} Substitute $bx+c=y$: \begin{align} a \exp((-c+y)/b)+y &= 0 \\ -y &= a\exp(-c/b)\exp(y/b) \\ -y/b\exp(-y/b) &= a/b\exp(-c/b) \end{align} The last equation is in the form $u\exp(u)=w$, which has a solution in terms of Lambert $W$ function: \begin{align} u&=W(w). \end{align} Hence \begin{align} -y/b&=W(a/b\exp(-c/b)) \\ y&=-b W(a/b\exp(-c/b)) \\ x&=-W(a/b\exp(-c/b))-c/b. \end{align} In particular, equation $88\exp(x)+12x-5=0$ with $a=88,b=12,c=-5$ has one real root, since the argument of $W$ is positive: \begin{align} x&=-W_0(22/3\exp(5/12))+5/12 \approx-1.3970513. \end{align}

Edit:

Lambert's W function is included in the Gnu Scientific Library (GSL) and is freely available by means of many computer systems and packages (wolframalpha, R, Asymptote, python scipy.special to name a few).

$\endgroup$
2
$\begingroup$

As said, there is an analytical solution in terms of Lambert function $$x=-W(d)-\frac{c}{b}$$ where $$d=\frac{a }{b}e^{-\frac{c}{b}}$$ If you do not want to (or cannot) use Lambert function, then numerical methods are the way to go.

Probably the simplest should be Newton method which, starting from a "reasonable" guess $x_0$, will update it according to $$x_{n+1}=x_n-\frac{f(x_n)}{f'(x_n)}$$ For the example you give in comments $$f(x)=5-12x - 88e^x$$ $$f'(x)=-12-88e^x$$ A look at the graph of the function shows that the solution is close to $-1.5$. So, let start iterating at $x_0=-1.5$; the method will generates the following iterates : $-1.393646353$, $-1.397047564$, $-1.397051301$ which is the solution for ten significant figures.

$\endgroup$
  • $\begingroup$ I've definitely done it wrong then - my solution has to be positive to make sense in the context of the question. thanks anyway though :) $\endgroup$ – imulsion Apr 17 '15 at 8:51
  • $\begingroup$ May be, there was a typo in the equation. In any manner, you are very welcome ! $\endgroup$ – Claude Leibovici Apr 17 '15 at 8:53

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.