I stuck at the derivation of Eq. (12.5) in Peskin and Schroeder's QFT. The authors tried to from (12.3)

$$ Z = \int [\mathcal{D}\phi]_{\Lambda} \exp\left(-\int d^dx \left[\frac{1}{2} (\partial_{\mu} \phi)^2 + \frac{1}{2} m^2 \phi^2 + \frac{\lambda}{4!} \phi^4 \right]\right) \tag{12.3} $$ where $$ [\mathcal{D}\phi]_{\Lambda} =\prod_{|k|<\Lambda} d\phi(k) \tag{12.4} $$

The authors then introduced a new set of variables as $\phi = \phi + \hat{\phi}$, where new $\phi$ is the old one for $|k|< b \Lambda$, otherwise zero; and $\hat{\phi}$ equals the old one for $b\Lambda \leq |k| < \Lambda $, otherwise zero.

Eq. (12.3) was replaced as $$ Z = \int \mathcal{D}\phi \int \mathcal{D}\hat{\phi}\exp\left(-\int d^dx \left[\frac{1}{2} (\partial_{\mu} \phi +\partial_{\mu} \hat{\phi} )^2 + \frac{1}{2} m^2 (\phi + \hat{\phi})^2 + \frac{\lambda}{4!} (\phi+\hat{\phi})^4 \right]\right) \\ \cdots \tag{12.5} $$

I naively expect $\mathcal{D} \phi = \mathcal{D}\phi + \mathcal{D} \hat{\phi} $, as what happened in the Lagrangian. But, (12.5) looks like a multiplication, as $\int \mathcal{D} \phi \int \mathcal{D} \hat{\phi}$. My stupid question is, why here is the multiplication in the integral variables?

I may view as changing variables in multivariable calculus. Nevertheless, the difference in the old and new variables, as a Jacobi, may not be 1. I am unable to work out the Jacobi in the transformation.

  • 1
    $\begingroup$ Imagine that $\phi$ is periodic and therefore expandable as a discrete Fourier series. The high frequency plus low frequency decomposition of $\phi$ is a sum but the way to integrate over all choices of $\phi$ is to integrate over all the Fourier coefficients which is a product. $\endgroup$ Commented Jun 19, 2022 at 1:00
  • $\begingroup$ I am still stupid. Suppose I write $\phi = \sum_{k=0}^{100} c_k e^{ikx} + \sum_{k=101}^{200} c_k e^{ikx} $. $d\phi = \sum_{k=0}^{100} c_k de^{ikx} + \sum_{k=101}^{200} c_k de^{ikx} $, or what else :( $\endgroup$
    – RandomUser
    Commented Jun 19, 2022 at 1:05
  • $\begingroup$ This is a functional integral and functions are infinite dimensional vectors. When $\textbf{r} = x\hat{x} + y\hat{y} + z\hat{z}$, the volume element is $dx dy dz$ not $dr = dx \hat{x} + dy \hat{y} + dz \hat{z}$. $\endgroup$ Commented Jun 19, 2022 at 1:26
  • $\begingroup$ I miserably don't get how the infinite dimensional aspect related to the question here :(. In QM, one can get the path integral from $<x't'|x,t> = \int dx'' <x't'|x''><x''| x',t>$ and inserting more x''', x''''... etc. I may write $x'' = x'' + \hat{x}''$, that applies to each variable. I can so far only relate the question to how changing of variables should work, e.g., via Jacobi in multivariable calculus. E.g., from (x,y) to (u,v) $|\partial x/\partial u \partial y/ \partial u, \partial x / \partial v \partial y / \partial v|$, here $\phi$ to $\phi, \hat{\phi}$, and not sure how to proceed $\endgroup$
    – RandomUser
    Commented Jun 19, 2022 at 1:55
  • $\begingroup$ Seems to me the question is about how $d\phi$ to $d\phi$, $d\hat{\phi}$ $\endgroup$
    – RandomUser
    Commented Jun 19, 2022 at 2:00

1 Answer 1


It is given on Eq. (12.4). One has \begin{align} [\mathcal{D}\phi]_{\Lambda} &= \prod_{|k|<\Lambda} \mathrm{d}\phi(k), \\ &= \prod_{|k|<b\Lambda} \mathrm{d}\phi(k) \prod_{b\Lambda<|k|<\Lambda} \mathrm{d}\phi(k), \\ &= \prod_{|k|<b\Lambda} \mathrm{d}\phi(k) \prod_{b\Lambda<|k|<\Lambda} \mathrm{d}\hat{\phi}(k), \\ &= \mathcal{D}\phi \mathcal{D}\hat{\phi}. \end{align}

As mentioned in the comments, the matter is that it is an integral in many dimensions. What P&S are doing is pretty much making a change of variables such as $x \to x$, $y \to y$, $z \to w$, so the volume element transforms as $\mathrm{d}x\mathrm{d}y\mathrm{d}z=\mathrm{d}x\mathrm{d}y\mathrm{d}w$.

  • $\begingroup$ (1) (Edited notations, sorry for the confusing) Thanks for your answer. I think, the range under the $\Pi$, i.e., $|k|<\Lambda$, is not to 'count' $k$, such that each $k$ has a different value and one can split into two $\Pi$ as the second line. From the QM path integral, it looks like $\int dx′′dx′′′⋯<x,t|x′′t′′><x′′t′′|x′′′t′′′><x′′′t′′′|⋯|x′t′>=\int \mathcal{D} x...$ (2.5.49) in Sakurai's modern quantum mechanics or Peskin (9.12). Each $x$ runs over $−\infty$ to $+\infty$. In QFT case, each $|k|$ (like $x''$, $x'''$, in QM) runs over all value below $\Lambda$. $\endgroup$
    – RandomUser
    Commented Jun 19, 2022 at 5:24
  • $\begingroup$ (2) (Edited some wording, sorry for the confusing) The changing variable should include something like a Jacobi, here for $\phi \rightarrow \phi, \hat{\phi}$, the Jacobi may not 1. I am unable to find a 2*2 determinant, since from 1 to 2 variables, to get the value 1 as the result of changing variable. $\endgroup$
    – RandomUser
    Commented Jun 19, 2022 at 5:26
  • 1
    $\begingroup$ @RandomUser for each fixed $k$, the integral over $\mathrm{d}\phi(k)$ runs from $-\infty$ to $+\infty$. It is as if the $\phi(k)$ are playing the role of $x'$ in the Sakurai expression you wrote. As for the Jacobian, it is $1$ because P&S are just renaming some of the integration variables (those with $|k| > b \Lambda$) without really making any other transformations. $\endgroup$ Commented Jun 19, 2022 at 5:32
  • $\begingroup$ The trick in here is that the integration variables are the values of $\phi(k)$ themselves for each fixed $k$, just as in the Sakurai path integral it were the values $x(t)$ themselves for each fixed $t$. In particular, notice that the analogy between Sakurai and P&S is among $t$ and $k$ and among $x(t)$ and $\phi(k)$. $\endgroup$ Commented Jun 19, 2022 at 5:33
  • $\begingroup$ So far, the only way I can justify P&S is to regard $\mathcal{D} \phi \mathcal{D} \hat{\phi}$ as schematic, to integrate the original variable but in a truncated range... Maybe there is a better way to demonstrate here. $\endgroup$
    – RandomUser
    Commented Jun 19, 2022 at 5:56

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.