Taylor Series Expansion Confusion

laser1
Messages
166
Reaction score
23
Homework Statement
description
Relevant Equations
.
Screenshot_1.png

For context, this is when deriving the Boltzmann distribution by using a canonical ensemble (thermodynamics).

omega is a function to represent number of microstates. According to wikipedia...

Screenshot_2.png

is the first order expansion around 0 (Maclaurin series).

My confusion: What are even differentiating with respect to? It seems like we are differentiating with respect to E, but then the epsilon looks like the "x" in the Maclaurin formula. I have more questions, but I'll leave it at this to avoid confusion for now. Thanks!
 
Physics news on Phys.org
Let's analyze it from behind. The Taylor series of a function ##f(x)## at a point ##a## is given by
$$
T(f,x,a)=\sum_{n\ge 0} \dfrac{f^{(n)}(a)}{n!}(x-a)^n=f(a)+f'(a)(x-a)+O((x-a)^2)
$$
For ##f(a)## we have ##f(a)=\ln \Omega(E),## i.e. ##a=E.## Then
$$
f'(a)(x-a)=f'(E)(x-E)=\dfrac{d\ln \Omega(E)}{dx}(x-E)
$$
which means the variable is ##x=-\varepsilon + E## and the second term of the series becomes
$$
f'(E)\cdot((-\varepsilon+E)-E)=\dfrac{d\ln \Omega(E)}{d(E-\varepsilon)}(-\varepsilon+E-E)=-\dfrac{d\ln \Omega(E)}{d(E-\varepsilon)}\varepsilon
$$
We have ##d(E-\varepsilon)=dE - d\varepsilon## and with ##\varepsilon## being a constant we have ##d\varepsilon = 0.##
Proof:
\begin{align*}
\dfrac{df(E)}{d(E-\varepsilon)}&=\dfrac{df((E-\varepsilon)+\varepsilon)}{d(E-\varepsilon)}=\dfrac{df(y+\varepsilon)}{dy}\\[6pt]
&=\dfrac{df(y+\varepsilon)}{d(y+\varepsilon)}\cdot \underbrace{\dfrac{d(y+\varepsilon)}{dy}}_{=1}\\[6pt]
&=\dfrac{df(E)}{dE}
\end{align*}
In total we get
$$
\ln \Omega (E-\varepsilon)=\ln \Omega(E)- \dfrac{d\ln \Omega(E)}{dE}\varepsilon + O(\varepsilon^2)
$$

laser1 said:
I have more questions, ...
Which?
 
Last edited:
laser1 said:
According to wikipedia...

View attachment 351220
is the first order expansion around 0 (Maclaurin series).

It looks like they are really expanding around ##E##, not around 0. The function and derivative are evaluated at ##E##. The Maclaurin series is too specialized (centered at ##x_0=0##) to give a good match to your equation. You should compare it to the more general Taylor series. ##f(a)+\frac {f'(a)}{1!}(x-a)+...##. Then ##E=a##, ##E-\epsilon=x##, and ##-\epsilon=x-a##.
laser1 said:
My confusion: What are even differentiating with respect to? It seems like we are differentiating with respect to E,
That is correct.
 
@fresh_42 Thank you, that makes sense. I also tried differentiating it with respect to epsilon around 0 and got this, but I am stuck at this step:
WhatsApp Image 2024-09-18 at 08.21.10.jpeg
 
laser1 said:
@fresh_42 Thank you, that makes sense. I also tried differentiating it with respect to epsilon around 0 and got this, but I am stuck at this step:

Apply the chain rule.
 
pasmith said:
Apply the chain rule.
WhatsApp Image 2024-09-21 at 21.30.38.jpeg


I did it a few days ago and was happy, but upon closer examination: I suspect dE/d(epsilon) = -1 only if the total energy of the ensemble is a constant.

Any idea where I have gone wrong with my second method?
 
laser1 said:
View attachment 351397

I did it a few days ago and was happy, but upon closer examination: I suspect dE/d(epsilon) = -1 only if the total energy of the ensemble is a constant.

Any idea where I have gone wrong with my second method?
I don't think that you made a mistake.
\begin{align*}
\left.\dfrac{d\Omega(E-\varepsilon)}{d\varepsilon}\right|_{\varepsilon=0}&=\left.\dfrac{d\Omega(E-\varepsilon)}{d(E-\varepsilon)}\right|_{E-\varepsilon=E}\cdot \left.\dfrac{d(E-\varepsilon)}{d\varepsilon}\right|_{\varepsilon=0}\\
&=\left.\dfrac{d\Omega(E-\varepsilon)}{d(E-\varepsilon)}\right|_{E-\varepsilon=E}\cdot\left\{\underbrace{\left.\dfrac{dE}{d\varepsilon}\right|_{\varepsilon=0}}_{=0}+\underbrace{\left.\dfrac{d(-\varepsilon)}{d\varepsilon}\right|_{\varepsilon=0}}_{=-1}\right\}\\
&=-\left.\dfrac{d\Omega(E-\varepsilon)}{d(E-\varepsilon)}\right|_{E-\varepsilon=E}
\end{align*}
The trick is now that it doesn't matter how you call your variables. We simply have
$$
-\left.\dfrac{d\Omega(E-\varepsilon)}{d(E-\varepsilon)}\right|_{E-\varepsilon=E}=-\left.\dfrac{d\Omega(x)}{dx}\right|_{x=E}=-\left.\dfrac{d\Omega(E)}{dE}\right|_{E}=-\dfrac{d\Omega(E)}{dE}
$$
 
Back
Top