A primary field $\mathcal{O}(w,\bar{w})$ with weight $(h,\bar{h})$ is defined by having the following OPEs with the stress tensor:

$$T(z)\mathcal{O}(w,\bar{w})=\frac{h\mathcal{O}}{(z-w)^2}+\frac{\partial \mathcal{O}}{z-w},\qquad \bar{T}(\bar{z})\mathcal{O}(w,\bar{w})=\frac{\bar{h}\mathcal{O}}{(\bar{z}-\bar{w})^2}+\frac{\bar{\partial} \mathcal{O}}{\bar{z}-\bar{w}}.$$

Let us consider a theory of a free boson defined by:

$$S = \int d^2 z\ \partial X\bar{\partial}X.$$

Then $\partial X$ is a $(1,0)$ primary field with and $\bar{\partial}X$ is a $(0,1)$ primary field (see e.g. Tong's string theory notes, section 4.3.3).

The field $X$ itself has the OPE

$$T(z)X(w,\bar{w})=\frac{\partial X}{z-w},\qquad\bar{T}(\bar{z})X(w,\bar{w})=\frac{\bar{\partial} X}{\bar{z}-\bar{w}}.$$

(See e.g. Polchiski eq. 2.4.6). So, by the definition above, it should be a primary field with weight $h=\bar{h}=0$.

However, I've looked through many textbooks and lecture notes (e.g. Di Francesco, Blumenhagen, Polchinski, Tong...) and it is never explicitly said that $X$ is primary. So... Is it primary or not?

  • If it is primary, then how come its descendants $\partial X$ and $\bar{\partial} X$ are also primary?

  • If it is not primary, then how come its OPE is exactly the correct OPE for a primary field of weight $(0,0)$?


2 Answers 2


I do not agree that $X$ is a primary field. Primary field is defined by its transformation properties under the conformal group (see e.g. yellow book). In particular, under scaling transformation, a correlation function involving primary operators, transforms as $$ \langle \mathcal{O}_1(\lambda x_1)\ldots\mathcal{O}_n(\lambda x_n)\rangle=\lambda^{-\Delta_1-\ldots-\Delta_n}\langle \mathcal{O}_1(x_1)\ldots\mathcal{O}_n(x_n)\rangle, $$ where $\Delta_i=h_i+\bar{h}_i$. It follows that for a primary field of dimension $(0,0)$ its two-point function must be constant. This is not true for $X$.

In fact, $X$ is not a local field in the sense of operator-state correspondence. The set of Virasoro primary operators in free boson CFT is $1,\partial X,\bar\partial X$ and the continuos family $e^{i\alpha X}$ for $\alpha\neq 0$.

  • $\begingroup$ In light of your answer I have edited mine.... Is the definition given in the original question then not the correct definition? $\endgroup$
    – Akoben
    Feb 12, 2016 at 13:18
  • $\begingroup$ Thank you for your answer! Could you please point me to a reference which explicitly states that $X$ is not a primary field? Polchinski (page 46-47, below eq. 2.4.17) states that the field $X$ "transforms as a tensor" and he defines a tensor operator and a primary field to be the same thing. I take it to mean that $X$ is a primary field, but it is stated in a confusing way. Is there an error in Polchinski? $\endgroup$
    – user90063
    Feb 13, 2016 at 3:20
  • $\begingroup$ Also, how does the 2-point function for $X$ transform exactly? It is my impression that $X(z)X(w)=\log(z-w)$ and this matches the general result that $A(z)B(w)=(z-w)^{-2h}$ if you take the limit $h \to 0$. $\endgroup$
    – user90063
    Feb 13, 2016 at 3:20
  • $\begingroup$ @Zoe, I'm not sure about a place where you can read that it is not a primary. From the two-point function of $X$ you can see two things. 1) In order to define it, one needs a scale, since $z$ and $w$ have dimension of length and so $\log|z-w|$ doesn't make sense without implicit assumption about a scale relative to which $|z-w|$ is measured. 2) The two-point function transforms inhomogeneously under rescalings, $\log(2|z-w|)=\log|z-w|+\log 2$. $\endgroup$ Feb 13, 2016 at 4:32
  • $\begingroup$ @Zoe,@Akoben I don't have Polchinski on me, but I'll try to look up your reference later. I am used to think that $X$ is somewhat pathological by itself. For example, its two-point function does not satisfy cluster decomposition property. It is not a part of state-operator correspondence. In my opinion it is better to think of $X$ as very different from other operators, and thus not to call it a primary. $\endgroup$ Feb 13, 2016 at 4:41

To see why the descendants are primary, you can use $$ \partial\left(T(z)X(w,\overline{w})\right) = T(z)\partial X(w,\overline{w}) = \frac{\partial^2 X}{z-w} + \frac{\partial X}{(z-w)^2} $$

And see that it is a primary field of weight $h = 1$, $\overline{h} = 0$, and similarly for the other field....


Your Answer

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