# Finding amplitude of Fourier coefficient

I'm dealing with a confusing problem related to radiative transfer in atmospheres. In this problem, the solar flux on a planet it being modeled by $$S(t)=S_0+\sum_{n=1}^{\infty}S_n\,e^{in\omega t} \, ,$$ and temperature modeled by $$T(t)=T_0+\sum_{n=1}^{\infty}T_n\,e^{in\omega t-\phi} \, .$$

My question is a math one, not a thermodynamics one. By using the relation $$\sigma\,T^4(t)=(1-A)\,S(t) \, ,$$ I'm supposed to be able to find the amplitude of the Fourier coefficient $T_n$ to be $$T_n=\frac{1}{4}\frac{(S_n/S_0)T_0}{\sqrt{1+\tan^2(\phi)}} \, .$$

Working backwards, I've reduced this to $$\frac{T_n}{T_0}=\frac{1}{4}\frac{S_n}{S_0}\cos(\phi) \, ,$$ but I'm not sure what else I can do.

Potentially useful relation: $$\tan(\phi)=n\omega\tau, \qquad \tau=C_PP_0/4\sigma T_0^3g \, .$$

• There is a handy identity for $1+tan^2$....
– lux
Apr 18 '18 at 4:33
• Yes, that's how I got to the expression for $T_n/T_0$. Apr 18 '18 at 4:35
• Ah you mean you reduced the expected answer to that form? I thought you meant you had arrived at that point in your calculation of the coefficients
– lux
Apr 18 '18 at 4:36
• No, the tricky part still eludes me. I was just monkeying with it to try to find something more recognizable. Apr 18 '18 at 4:38
• Before throwing yourself in to the general proof have you tried checking it works for small values of n?
– lux
Apr 18 '18 at 4:40

Edit: I realized that, in your question, the Fourier series decomposition of the temperature should go as (please confirm the braces after $n$) $$T(t) = T_0 + \sum_{n=1}^\infty T_n e^{in (\omega t - \phi)}$$

Original answer: The only equation that contains any physical link between $T(t)$ and $S(t)$ is the one you have described: $$\sigma T^4(t) = (1-A)S(t).$$ Taking the time derivative and then dividing both sides by $\sigma T^4(t)$, you get \begin{align} 4\frac{\sigma T^3(t)}{\sigma T^4(t)} \frac{dT(t)}{dt} &= \frac{(1-A)}{\sigma T^4(t)} \frac{dS(t)}{dt}\\ & \\ \Rightarrow S(t) \frac{d T(t)}{dt} &= \frac{1}{4} T(t) \frac{dS(t)}{dt}, \end{align} where in the second equation, I have replaced $\sigma T^4(t)$ using the above equation and simplified further. Here, you can substitute the respective Fourier expansions and obtain \begin{align} S_0\left ( 1 + \sum_{m=1}^\infty \frac{S_m}{S_0} e^{im\omega t} \right ) \left( \sum_{n=1}^\infty T_n (i n \omega) e^{in(\omega t - \phi)} \right ) = \frac{1}{4}T_0 \left( 1 + \sum_{n=1}^\infty \frac{T_n}{T_0} e^{i n(\omega t - \phi)}\right) \\ \times \left( \sum_{m=1}^\infty S_m (i m\omega)e^{i m \omega t}\right ) \end{align}

\begin{align} \Rightarrow \left ( 1 + \sum_{m=1}^\infty \frac{S_m}{S_0} e^{im\omega t} \right ) \left( \sum_{n=1}^\infty \frac{T_n}{T_0} (i n \omega) e^{in(\omega t - \phi)} \right ) = \frac{1}{4} \left( 1 + \sum_{n=1}^\infty \frac{T_n}{T_0} e^{i n(\omega t - \phi)}\right) \\ \times \left( \sum_{m=1}^\infty \frac{S_m}{S_0} (i m \omega) e^{i m \omega t}\right ) \end{align} which, upon re-ordering the terms and comparing the coefficients for each unique Fourier mode, should give you the required result.

EDIT: The above expression can be re-written as $$\sum_{m,n=0}^\infty \frac{S_m}{S_0} \frac{T_n}{T_0} (i n \omega) e^{i(m+n)\omega t -in\phi} = \frac{1}{4} \sum_{m,n=0}^\infty \frac{S_m}{S_0} \frac{T_n}{T_0} (i m \omega) e^{i(m+n)\omega t -in\phi}$$ On the RHS, exchange $m$ with $n$ (they are merely dummy indices) and bring the expression to the LHS, you get $$\sum_{m,n=0}^\infty \left ( \frac{S_m}{S_0} \frac{T_n}{T_0} e^{-in\phi} - \frac{1}{4} \frac{S_n}{S_0}\frac{T_m}{T_0} e^{-im\phi} \right ) (i n \omega) e^{i(m+n)\omega t} = 0$$ You can further simplify by the following set of transformations \begin{align} k &= m + n\\ m &= k-n \end{align} and by requiring that the coefficient of $e^{ik\omega t}$ is ZERO $\forall k$, you can obtain the required relation. In fact, it will become $$\sum_{k=0}^\infty \left ( \sum_{n = 0}^k i n \omega \left [ \frac{S_{k-n}}{S_0} \frac{T_n}{T_0} e^{-in\phi} - \frac{1}{4} \frac{S_n}{S_0}\frac{T_{k-n}}{T_0} e^{-i(k-n)\phi} \right ]\right) e^{i k \omega t} = 0$$

$$\Rightarrow \sum_{n = 0}^k i n \omega \left [ \frac{S_{k-n}}{S_0} \frac{T_n}{T_0} e^{-in\phi} - \frac{1}{4} \frac{S_n}{S_0}\frac{T_{k-n}}{T_0} e^{-i(k-n)\phi} \right ] = 0$$ for every value of $k$. Beyond this point, it would be more convenient to put in the values and build a recursion relation. For instance, $k = 1$ gives $$\frac{T_1}{T_0} = \frac{1}{4} \frac{S_1}{S_0} e^{i \phi}$$ Take just the real part (or rather let the Fourier series index run from $-\infty$ to $+\infty$) and you get your expression (in terms of the $cos \phi$ which is equivalent to $1/\sqrt{\dots}$). Using the relation for $n=1$, proceed to $k = 2$ and so on. You'll start seeing a pattern in the recursive relations.

• Thank you! I follow you this far, but could you explain a bit more what you mean by "re-ordering the terms and comparing the coefficients for each unique Fourier mode"? Also what relation will get me from $e^{in\phi}$ to $1/\sqrt{1+tan^2(\phi)}$? Apr 18 '18 at 16:02
• This has been super helpful, but for the life of me I can't figure out how to get from exponentials to the $1/\sqrt()$... Apr 19 '18 at 0:31
• i have detailed out the steps. this approach gives you $cos \phi$ rather than $1/\sqrt{1 + \tan^2\phi}$ Apr 19 '18 at 2:36
• Okay, thank you! Sorry, it's been a while since I've dealt with series, and I've lost a lot of my intuition regarding them. Apr 19 '18 at 2:41