In page 179 of Hansen and McDonald's book, Theory of Simple Liquids, 3rd edition, 2006, an identity of correlation functions was deduced.

enter image description here

Here $C_{AB}(t)$ is the time correlation function between dynamical variables. My question is how to get $\langle{\dot A}A^* \rangle=0$ after $\langle{\dot A}B^*\rangle=0$?

if $B=A$, $\langle{\dot A}A^* \rangle=-\langle A{\dot A^*}\rangle=-{\langle{\dot A}A^* \rangle}^*$ . So if $\langle{\dot A}A^* \rangle$ is a complex number, e.g., $a+bi$, then $a+bi=-(a+bi)^*$, what we get is only $a=0$, while $b$ can be any real number. This is different from $\langle{\dot A}A^* \rangle=0$.

I wonder if the statement that $A$ is a dynamical variable can give any restriction on the correlation function $\langle{\dot A}A^* \rangle$ that it should be real number.

  • 4
    $\begingroup$ you build the correlators so as to be real (because you identify a measure with one real number). The correlator of $\dot {A}A*$ is the derivative of a real function (the correlator of AA*) so it has no imaginary part. $\endgroup$
    – AoZora
    Apr 25, 2019 at 11:34
  • $\begingroup$ @france95 I see. Thanks a lot for your help! $\endgroup$
    – FaDA
    Apr 25, 2019 at 11:51
  • 1
    $\begingroup$ Please use '\langle' and '\rangle' ($\langle$ and $\rangle$) instead of '<' and '>' ($<$ and $>$). $\endgroup$ Apr 25, 2019 at 11:52
  • $\begingroup$ @france95 I checked it again, and I found that as $(\dot A A^∗)′=(AA^∗)+A(\dot A^∗)$, And since $(\dot A A^∗)=(A(\dot A^∗))^∗$,$ (\dot A A^∗)$ can still be a real number even if $(\dot A A^∗)$ is a complex number. Would you please give me more detail explanation? Thanks! $\endgroup$
    – FaDA
    Apr 25, 2019 at 12:55

1 Answer 1


I believe that there is an extra subtletly. Almost always we choose dynamical variables with a well defined signature $\varepsilon$ under time reversal. This is mentioned at the start of that chapter (in the 3rd edition). This means that (their eqn (7.1.9)) $$ \langle A(t) B^*(0) \rangle = \varepsilon_A \varepsilon_B \langle A(-t) B^*(0) \rangle = \varepsilon_A \varepsilon_B \langle A(0) B^*(t) \rangle . $$ It follows that the autocorrelation functions are both even and real functions of time (irrespective of whether $\varepsilon_A$ is $+1$ or $-1$). In other words, the above equation shows that $$ \langle A(t) A^*(0) \rangle = \langle A(0) A^*(t) \rangle \quad\Rightarrow\quad \langle \dot{A} A^* \rangle = \langle A \dot{A}^* \rangle $$ as well as the relation you have seen already from stationarity $$ \langle \dot{A} A^* \rangle = -\langle A \dot{A}^* \rangle . $$ Hence $\langle \dot{A} A^* \rangle=0$, as you wanted.

  • $\begingroup$ I got it this time. Thanks a lot for your help! $\endgroup$
    – FaDA
    Apr 25, 2019 at 15:54

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

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