Trying to understand the Langevin Equation. In particular, this passage from a Wikipedia article has me confused (section: "Thermal Noise in an Electrical Resistor"):

$\frac{dU}{dt} =-\frac{U}{RC}+\eta \left( t\right),\;\;$ $\left\langle\eta \left( t\right) \eta \left( t^{\prime }\right)\right\rangle = \frac{2k_{B}T}{RC^{2}}\delta \left(t-t^{\prime }\right).$

This equation may be used to determine the correlation function

$\left\langle U\left(t\right) U\left(t^{\prime }\right) \right\rangle =\left( k_{B}T/C\right) \exp \left( -\left\vert t-t^{\prime }\right\vert /RC\right) \approx 2Rk_{B}T\delta \left( t-t^{\prime}\right),$

Here's my derivation of a rather different result:

$\displaystyle \frac{dU(t)}{dt} =-\frac{U(t)}{RC}+\eta \left( t\right)$

$\displaystyle U(t+t^\prime)\frac{dU(t)}{dt} =-\frac{U(t+t^\prime)U(t)} {RC}+U(t+t^\prime)\eta(t)$

Now concentrate on LHS:

$\displaystyle U(t+t^\prime)\frac{dU(t)}{dt} = \frac{d}{dt}(U(t+t^\prime)U(t))-U(t)\frac{d}{dt}U(t+t^\prime)= \frac{d}{dt}(U(t+t^\prime)U(t))-U(t)\frac{d}{dt^\prime}U(t+t^\prime)$ $\displaystyle=\frac{d}{dt}(U(t+t^\prime)U(t))-\frac{d}{dt^\prime}(U(t)U(t+t^\prime))$

now do ensemble averaging and assuming system is statistically stationary. Then the first term on the new LHS vanishes as does the second term on the RHS, leaving:

$\displaystyle -\frac{d}{dt^\prime}\langle U(t+t^\prime)U(t)\rangle = -\frac{\langle U(t+t^\prime)U(t)\rangle} {RC}$

which gives the obviously unphysical

$\displaystyle \langle U(0)U(t^\prime)\rangle=\langle U(0)U(0) \rangle \exp(+t^\prime / RC)$

Can anyone tell me where I went wrong and supply the derivation of the Wikipedia result?


1 Answer 1


You start by solving the differential equation. It is a first order, linear differential equation with constant coefficients. So the solution of the homogenous system is quite simple: $U(t) = c\cdot e^{-t/RC}$.

Now we solve the particular system with variation of the constant $c$, which means we try the Ansatz $U(t) = c(t)\cdot e^{-t/RC}$. This give the differential equation $c'(t) = \eta(t)e^{t/RC}$, which can simply integrate to $c(t) = \int\limits_{0}^{t} \eta(\hat{t})e^{\hat{t}/RC}d\hat{t} + U_0$.

So the solutions of the langevin equation is

$U(t) = e^{-t/RC}\left(\int\limits_{0}^{\hat{t}} \eta(t)e^{\hat{t}/RC}d\hat{t} + U_0\right)$.

Now we can calculate all moments. The first moment is pretty strait forward

$\langle U(t)\rangle = \left\langle e^{-t/RC}\left(\int\limits_{0}^{t} \eta(\hat{t})e^{\hat{t}/RC}d\hat{t} + U_0\right) \right\rangle = U_0e^{-t/RC}$.

The second moment needs a little more effort.

$\langle U(t)U(t')\rangle = \left\langle e^{-(t+t')/RC}\left(\int\limits_{0}^{t}\int\limits_{0}^{t'} \eta(\hat{t})\eta(\hat{t'})e^{(\hat{t}+\hat{t'})/RC}d\hat{t}d\hat{t'} + 2U_0\int\limits_{0}^{t} \eta(\hat{t})e^{\hat{t}/RC}d\hat{t} + U_0^2\right)\right\rangle % =e^{-(t+t')/RC}\left(\int\limits_{0}^{t}\int\limits_{0}^{t'} \frac{2k_BT}{RC^2}\delta(\hat{t}-\hat{t}') e^{(\hat{t}+\hat{t'})/RC}d\hat{t}d\hat{t'} + U_0^2\right)$

To be correct, we have to treate two different cases, one for $t<t'$ and one for $t>t'$. But let's focus on $t>t'$ at the moment. Then the next step would be

$\langle U(t)U(t')\rangle =e^{-(t+t')/RC}\left(\frac{2k_BT}{RC^2}\int\limits_{0}^{t}e^{2\hat{t}/RC}d\hat{t} + U_0^2\right) % =e^{-(t+t')/RC}\left(\frac{k_BT}{C}e^{2t/RC} + U_0^2\right) % =\frac{k_BT}{C}e^{(t-t')/RC} + U_0^2e^{-(t+t')/RC}$.

For $t<t'$ you'll get

$\langle U(t)U(t')\rangle =\frac{k_BT}{C}e^{(t'-t)/RC} + U_0^2e^{-(t+t')/RC}$.

So we can combine both results to

$\langle U(t)U(t')\rangle =\frac{k_BT}{C}e^{-|t-t'|/RC} + U_0^2e^{-(t+t')/RC}$.

If you choose the initial voltage $U_0 = 0$, this gives you the exakt result from wikipedia.

  • $\begingroup$ Much obliged. But I don't think I am really going to understand things until I know why my derivation failed. Any ideas about that? $\endgroup$ Commented Sep 17, 2015 at 21:21
  • $\begingroup$ You set $\langle U (t)\eta (t)\rangle =0$ in your third step. Now, if you look at the solution of the langevin equation, this is clearly not the case, since $U (t) $ itself contains a term proportional to $\eta (t) $. $\endgroup$
    – manthano
    Commented Sep 18, 2015 at 6:02
  • $\begingroup$ Dang, if it had been a snake, it'd a bit me! Thanks! $\endgroup$ Commented Sep 18, 2015 at 13:18
  • $\begingroup$ Suggestion: You can use $\$\$\mbox{[equation]}\$\$$ on your main equations. So they will be centralized in the answer, and in nicer form. $\endgroup$ Commented Oct 17, 2015 at 21:03

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

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