2
$\begingroup$

If we consider a $D=0$ theory with the Lagrangian: $$\mathcal{L}[\phi]=g\phi^n+J\phi$$ And its Green functions: $$G_n=\langle\phi^n\rangle_{J=0}=\frac{1}{Z[0]}\frac{\delta^nZ[J]}{\delta J^n}|_{J\rightarrow0}.$$ An infinite change of the integration variable $\phi$ leads to the equation of motion: $$\langle d\mathcal{L}/d\phi\rangle_J=gn\langle\phi^{n-1}\rangle_J=\langle J\rangle_J.$$ Then we get $G_{n-1}=0$ when $J\rightarrow0$.

My question is: Do we have any explicit insight about why the $n-1$ point correlation function vanishes?

My understanding of the correlation function is about an amplitude between two states. I know that the one-point function $\langle\phi\rangle$ vanishes if the system has some symmetries. So does the $n-1$ point function vanish also for symmetries?

$\endgroup$
0

2 Answers 2

2
$\begingroup$

In the absence of space-time coordinates ($D=0$), the functional integral of the (Euclidean) "field" theory with the Lagrangian $g \phi^n$ reduces to an ordinary integral and the generating functional becomes the generating function $$Z(J) = \mathcal{N} \! \int\limits_{-\infty}^\infty \! d\phi \, e^{-g \phi^n+J\phi}, \quad g\gt 0,\tag{1} \label{1}$$ where the normalization constant $\mathcal N$ is determined by the condition $Z(0)=1$. It is obvious that the existence of the integral in \eqref{1} requires $n$ to be even. As a consequence, $$\langle \phi^{n-1} \rangle:= \frac{\partial^{n-1} Z(J)}{\partial J^{n-1}}{\Large|}_{J=0}= \mathcal{N} \! \int\limits_{-\infty}^\infty \! d\phi \, \phi^{n-1}\, e^{-g \phi^n} \tag{2} \label{2}$$ vanishes because of the trivial fact that the integrand $f(\phi) = \phi^{n-1} e^{-g \phi^n}$ is an odd function, $f(-\phi)=-f(\phi)$.

$\endgroup$
5
  • $\begingroup$ But maybe the function $G_{n-1}=0$ is true when $n$ is not necessarily a positive integer in non-Hermitian theories. $\endgroup$
    – Errorbar
    Commented Mar 19 at 7:32
  • $\begingroup$ For non-integer $n$, $\phi^n$ has a branch cut. In such a case you have to specify precisely how your integral is defined at all. $\endgroup$
    – Hyperon
    Commented Mar 19 at 7:45
  • $\begingroup$ @XinranSu By the way, the second equation in your question implied implicitely that $n$ was supposed to be a non-negative integer. (Otherwise the $n$-th derivative would not make sense.) $\endgroup$
    – Hyperon
    Commented Mar 19 at 8:10
  • 1
    $\begingroup$ Sorry, I have ignored the sources of my question. I have updated. $\endgroup$
    – Errorbar
    Commented Mar 19 at 9:18
  • $\begingroup$ @XinranSu It is difficult to shoot at a moving target... $\endgroup$
    – Hyperon
    Commented Mar 19 at 10:06
0
$\begingroup$
  1. More generally, for a complex scalar $\phi^n$ theory$^1$ where $n\in\mathbb{N}$ with a global $\mathbb{Z}_n$-symmetry $\phi\to e^{2\pi i/n}\phi$, then the $m$-point function $\langle\phi^m\rangle_{J=0}=0$ vanishes whenever $m\notin n\mathbb{N}_0$ is not a multiple of $n$, cf. e.g. my Phys.SE answer here.

  2. For a real scalar $\phi^n$ theory where $n$ is even, there is a global $\mathbb{Z}_2$-symmetry $\phi\to -\phi$, so that all odd $m$-point function $\langle\phi^m\rangle_{J=0}=0$ vanish by a similar argument.


$^1$ To make the action real, add a hermitian conjugate interaction term $\phi^{\ast n}$.

$\endgroup$
0

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.