Is it possible to solve $k = \frac{x}{\ln(x)}$ for $x$? My suspicion after a fruitless hour of manipulation is that it is not.
Is it possible to solve $k = \frac{x}{\ln(x)}$ for $x$?
141 Views Asked by Bumbble Comm https://math.techqa.club/user/bumbble-comm/detail AtThere are 2 best solutions below
On
Consider the function $f(x)=x-k\ln x$, defined for $x>0$. We have $$ \lim_{x\to\infty}f(x)=\infty $$ and $$ \lim_{x\to0}f(x)= \begin{cases} \infty & \text{if $k>0$}\\[4px] -\infty & \text{if $k<0$} \end{cases} $$
We already see that a solution for $f(x)=0$ (which is the same as your equation) exists when $k<0$, but this is not sufficient.
Consider the derivative $$ f'(x)=1-\frac{k}{x}=\frac{x-k}{x} $$
Case $k<0$
If $k<0$, the derivative is positive everywhere, so the function is increasing and we get a single solution.
Case $k=0$
There is no solution, because this would imply $x=0$, but the equation makes no sense for $x\le0$.
Case $k>0$
So let $k>0$. We see that the function has a minimum at $k$, with $$ f(k)=k(1-\ln k) $$ The number of solutions then depends on whether $1-\ln k$ is positive (no solution), zero (one solution) or negative (two solutions). Clearly, $1-\ln k>0$ if and only if $0<k<e$.
Summary
no solution if $0<k<e$;
one solution if $k=e$ or $k<0$
two solutions if $k>e$.
With numerical methods you can approximate the solutions to the required accuracy.
It can be solved with the Lambert W function.
First let $y=\frac1x$. Then let $z=\ln(y)$, and use the Lambert W function. Then substitute back.
\begin{align*} k &= \frac{x}{\ln x} \\ k &= \frac{\frac1y}{\ln \frac1y} \\ k &= \frac{1}{y\ln \frac1y} \\ k &= -\frac{1}{y\ln y} \\ -\tfrac{1}k &= y\ln y \\ -\tfrac{1}k &= e^zz \\ W\left(-\tfrac{1}k\right) &= z \\ W\left(-\tfrac{1}k\right) &= \ln(y) \\ e^{W\left(-\tfrac{1}k\right)} &= y \\ e^{W\left(-\tfrac{1}k\right)} &= \tfrac1x \\ e^{-W\left(-\tfrac{1}k\right)} &= x \end{align*}
Hence $x= e^{-W\left(-\tfrac{1}k\right)} $.
Note that the Lambert W function is only defined for $x\geq-\tfrac{1}e$. Therefore we need $k>e$ or $k<0$. Otherwise there is no solution.
When $k>e$, we need to consider both branches of W. Then $x= e^{-W_0\left(-\tfrac{1}k\right)} \vee e^{-W_{-1}\left(-\tfrac{1}k\right)}$.
Note that we need the Lambert W function. If $e^{-W\left(-\tfrac{1}k\right)}$ could be written with just elementary functions, then $W(x)$ can be written using just elementary functions. But it has been proven that it can't.