4
$\begingroup$

It is well known that the 2pt function of 2D free massless boson is given by \begin{equation} \label{s} \langle \phi(z,\bar{z})\phi(y,\bar{y})\rangle=-\frac{1}{4\pi g} \ln |z-y|^2\tag1 \end{equation} which is the eqn(2.10) in the CFT yellow book ("Conformal field theory" by P. Francesco et al). And this result is derived by Green function method.

I am trying to produce the above 2pt function via an alternative way, i.e. the mode expansion. For scalar field, the mode expansion is (6.54) in that book, which is \begin{equation} \phi(z,\bar{z})=\phi_0-\frac{i}{4\pi g}\pi_0 \ln(z\bar{z})+\frac{i}{\sqrt{4\pi g}}\sum_{n\neq 0}\frac{1}{n}(a_nz^{-n}+\bar{a}_n\bar{z}^{-n}) \end{equation} with the coefficients satisfying the commutators \begin{equation} [a_n,a_m]=n\delta_{m+n},~~[\bar{a}_n,\bar{a}_m]=n\delta_{m+n},~~a_0\equiv \bar{a}_0\equiv \frac{\pi_0}{\sqrt{4\pi g}} \end{equation} For vacuum state, we have \begin{equation} a_n |0\rangle=\bar{a}_n|0\rangle=0, ~(n\geq 0) \end{equation} Then using above commutators, we can compute the vacuum 2pt function as follows \begin{eqnarray} \langle\phi(z,\bar{z})\phi(y,\bar{y})\rangle&=&\phi^2_0-\frac{1}{4\pi g}\sum_{n>0}\frac{-1}{n^2}\langle(a_nz^{-n}+\bar{a}_n\bar{z}^{-n})(a_{-n}y^{n}+\bar{a}_{-n}\bar{y}^{n})\rangle\\ &=&\phi^2_0-\frac{1}{4\pi g}\sum_{n>0}\frac{-1}{n^2}n((y/z)^n+(\bar{y}/\bar{z})^n)\\ &=&\phi^2_0-\frac{1}{4\pi g}(\log(1-y/z)+\log(1-\bar{y}/\bar{z})),\tag2 \end{eqnarray} where we used \begin{equation} \sum_{n=1}\frac{x^n}{n}=-\log(1-x) \end{equation} However equation $(2)$ is not the standard form as in equation $(1)$. They are different by terms $\log z$ and $\log \bar{z}$. My question is what is wrong with equation $(2)$?

$\endgroup$
7
  • 2
    $\begingroup$ I believe the problem is that terms with phi0 and pi0 should also contribute. By the way, you missed that they also commute nontrivially. $\endgroup$ Commented Aug 20, 2021 at 9:34
  • $\begingroup$ Thanks for you comments. I think phi0 is just a0, the latter acting on $|0\rangle$ gives 0, so phi0 will not contribute. $\endgroup$ Commented Aug 20, 2021 at 9:58
  • 2
    $\begingroup$ That's not exactly right. As you wrote correctly in the body of the question, a0 should be identified with pi0, not phi0. phi0 does not need to annihilate the vacuum, so the term pi0 phi0 contributes to the correlation function. The vacuum average of phi0^2 is what defines the (necessary for 2d boson correlator) infrared cutoff and is also non-zero. $\endgroup$ Commented Aug 20, 2021 at 10:53
  • $\begingroup$ Sorry, I made a typo in my last comments. Indeed, $a_0$ is identified with $\pi_0$, then the term like this $\langle 0|\phi_0 \pi_0|0\rangle=0= \langle 0| \pi_0\phi_0|0\rangle$ will not contribute, since $\pi_0|0\rangle=0=\langle0|\pi_0$. $\endgroup$ Commented Aug 20, 2021 at 11:18
  • 3
    $\begingroup$ $\langle 0 \mid \phi_0 \pi_0\mid 0 \rangle$ and $\langle 0\mid \pi_0 \phi_0 \mid 0 \rangle$ can not both be zero, since $[\phi_0, \pi_0] \neq 0$. $\endgroup$ Commented Aug 20, 2021 at 11:22

0