Mathematics Stack Exchange is a question and answer site for people studying math at any level and professionals in related fields. Join them; it only takes a minute:

Sign up
Here's how it works:
  1. Anybody can ask a question
  2. Anybody can answer
  3. The best answers are voted up and rise to the top

I'm going through some problems in Theorems, Corollaries, Lemmas, and Methods of Proof, and I'm stuck at a certain problem that seemed very interesting until I couldn't solve it for the life of me.

Let $\circ$ be defined on $\mathbb R^+$ as $a\circ b = a^b b^a$. One can show that $\circ$ is Abelian, that $\mathbb R^+$ is closed under $\circ$, and that there exists identity element $e=1$ such that $a\circ e = e \circ a = a$.

The next part of the problem asks to find $2^{-1}$, which means solving for $x$ where $2^x x^2 = 1$. Because $\circ$ is Abelian, there's no other way to arrange this expression. I tried taking logarithm of both sides to get

$$\begin{align} \log_2(2^x x^2) &= \log_2(1) \\ \log_2(2^x) + \log_2(x^2) &= 0 \\ x + 2\log_2(x) &= 0 \\ x &= -2\log_2(x) \\ 1 &= -2 \frac{\log_2(x)}{x} \\ -\frac{1}{2} &= \log_2(x^{\frac{1}{x}}) \\ 2^{-\frac{1}{2}} &= x^{\frac{1}{x}} \end{align}$$

This doesn't really get me anywhere. I tried other manipulations, but all expressions I got were quite ugly. WolframAlpha doesn't return any meaningful results for positive reals. I keep coming back to this problem, but it looks like I won't be able to solve it. Perhaps I'm missing some algebra I have to use to get the result.

share|cite|improve this question
Looks like a job for Lambert's W function. – marty cohen Jan 28 '12 at 19:05
Does one ask to find the inverse or to prove that it exists? – Did Jan 28 '12 at 19:07
@DidierPiau The book asked to find it, so I assume it exists. – Phonon Jan 28 '12 at 20:37
As Raymond Manzoni says, $-2$ works. I think it is worth noting, however, that some elements of $\mathbb{R}^+$ may not be invertible. Here's a plot of $(\frac{1}{3})^x x^{\frac{1}{3}}$ which suggests $\frac{1}{3}$ is one such element. – Dejan Govc Jan 28 '12 at 21:38
Since nobody mentioned it, let me add that an approximate value of the inverse of $2$ is $0.766665$. Let $u=\mathrm eW(1/\mathrm e)=0.756945...$ Every point in $(1,+\infty)$ has a unique inverse in $(u,1)$, the unique inverse of $\mathrm e$ is $u$ and vice versa, the unique inverse of $1$ is $1$, each point in $(u,1)$ has two inverses, one in $(1,\mathrm e)$ and the other in $(\mathrm e,+\infty)$ (for example $2$ and $4$ have the same inverse), and no point in $(0,u)$ has an inverse. – Did Jan 29 '12 at 14:05
up vote 3 down vote accepted

The equation $2^{-x}=x^2$ admits three real solutions : $x=-2,\;x=-4\;$ and a more sophisticated solution involving the LambertW function as returned by WolframAlpha (the only positive solution will be this last one).

These solutions may be obtained by writing $\,2^{-x}=x^2\,$ as $\,1=x\,e^{x\frac{\ln(2)}2}\,$ or $\,-1=x\,e^{x\frac{\ln(2)}2}\,$ that we may express as :
$\;\displaystyle \frac{\ln(2)}2=\left(x\frac{\ln(2)}2\right)\,e^{\left(x\frac{\ln(2)}2\right)}\;$ or $\;\displaystyle \frac{-\ln(2)}2=\left(x\frac{\ln(2)}2\right)\,e^{\left(x\frac{\ln(2)}2\right)}$

Since the LambertW function is defined implicitly by $\displaystyle z=W(z)e^{W(z)}$ we see that the answers are given by $\;W\left(\frac{\ln(2)}2\right)=x\frac{\ln(2)}2\;$ and $\;W\left(\frac{-\ln(2)}2\right)=x\frac{\ln(2)}2\;$ that is by
$$\tag{1}\boxed{\displaystyle x=\frac2{\ln 2} W\left(\frac{\ln 2}2\right)}$$ and $$\tag{2}\;\boxed{\displaystyle x=\frac2{\ln 2} W\left(\frac{-\ln 2}2\right)}$$

The subtle point is that this second solution $(2)$ will be split in two sub-solutions since the $\rm LambertW$ function admits two branches for $x \in \left(-\dfrac1e,0\right)\;$ (i.e. for $ W\left(\dfrac{-\ln 2}2\right)$) :

  • the upper one gives $W_0\left(\dfrac{-\ln 2}2\right)=-\ln 2\;$ (divided by $\ln 2$ and multiplied by $2$ returns the $x=-2\,$ solution) while
  • the lower branch gives the value noted $W_{-1}\left(\dfrac{-\ln 2}2\right)$ (after multiplication by $\dfrac2{\ln 2}$ this returns the $x=-4\;$ solution).

The solution $(1)$ is the only real and positive solution : $$\boxed{\displaystyle x=\frac2{\ln 2} W\left(\frac{\ln 2}2\right)\approx 0.766664695962123}$$ To clarify further : you probably had to find an approximation of the answer (possibly using the plot of $x\to 2^{-x}$ and $x \to x^2$ on the same graphics as shown on WolframAlpha) or find it numerically (say by iterations or with a Taylor series) or play with the LambertW function...

share|cite|improve this answer
OK, and thanks for the edit; I misread the answer missing the positive solution. – Marc van Leeuwen Oct 29 '13 at 5:05
@Marc: this was one of my early posts here and a fine occasion for an edit so thanks for taking care! Cheers, – Raymond Manzoni Oct 29 '13 at 8:02

Your Answer


By posting your answer, you agree to the privacy policy and terms of service.

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