I want to find a two-term asymptotic expansion of the equation above. For the small root, I know the first term may be $\epsilon$, but I don't know how to find the second term. Moreover, I don't know how to find the root tending to infinite either.
-
Welcome to MSE. Your question is phrased as an isolated problem, without any further information or context. This does not match many users' quality standards, so it may attract downvotes, or closed. To prevent that, please [edit] the question. This will help you recognise and resolve the issues. Concretely: please provide context, and include your work and thoughts on the problem. These changes can help in formulating more appropriate answers. – José Carlos Santos Dec 17 '22 at 09:38
-
Do you mean that you want the expansion of $x$ as a function of $\epsilon$ ? If this is the case, look at Lambert function – Claude Leibovici Dec 17 '22 at 09:40
-
1See the Lagrange inversion theorem. The article has the Lambert $W$ function as an example. – Jean-Claude Arbaut Dec 17 '22 at 10:03
1 Answers
(I was taught to do asymptotics the heuristic way, so please forgive the handwaviness below. There are more sophisticated ways to do things if you want more than two terms.)
Small root
We have seen that the first term is $x \sim \epsilon$. For the next term, suppose that $$ x \sim \epsilon + a \epsilon^n $$ where $n > 1$.
Noting that $x \mathrm{e}^{-x} \sim x (1 - x)$, we have
\begin{align} (\epsilon + a \epsilon^n) (1 - \epsilon - a \epsilon^n) &\sim \epsilon \\ \epsilon - \epsilon^2 + a \epsilon^n &\sim \epsilon, \end{align}
where we have discarded terms smaller than $\epsilon^n$. We see that $n = 2$ (which is indeed greater than $1$) and $a = 1$. Hence, to two terms,
$$ x \sim \epsilon + \epsilon^2. $$
Large root
The large root is not as straightforward. Taking logarithms we have $$ \log x - x = \log\epsilon. $$
Now, if $x$ is exceedingly large, then we may discard $\log x \ll x$, giving the very crude estimate $$ x \sim \log\frac{1}{\epsilon}. $$
For the next term, suppose that $$ x \sim \log\frac{1}{\epsilon} + x_1 $$ where $x_1 \ll \log(1/\epsilon)$. Substituting into the log expression, we get
\begin{align} \log \left( \log\frac{1}{\epsilon} + x_1 \right) - \log\frac{1}{\epsilon} - x_1 &\sim \log\epsilon \\ x_1 &\sim \log \left( \log\frac{1}{\epsilon} + x_1 \right) \\ &\sim \log \left( \log\frac{1}{\epsilon} \right), \end{align} which, for exceedingly exceedingly small $\epsilon$, is indeed much less than $\log(1/\epsilon)$.
Hence to two terms, $$ x \sim \log\frac{1}{\epsilon} + \log \log\frac{1}{\epsilon}. $$
- 276
-
Thank you very much! I just start to learn this subject and it helps me a lot. – ZYZmoonlight Dec 17 '22 at 10:55