Consider a quantum field $\phi(x)$ (say, a scalar) at a spacetime location $x$. It turns out that $$\langle\phi(x)\rangle_0\equiv\langle 0|\phi(x)|0\rangle=0.\tag{1}$$ This is easily obtaied by using the Fourier exapansion $$\phi(x)=\int\frac{d^3\textbf{p}}{(2\pi)^{3/2}\sqrt{2E_{\textbf{p}}}}(a_{\textbf{p}}e^{-ip\cdot x}+a^\dagger_{\textbf{p}} e^{ip\cdot x}).\tag{2}$$

Now in quantum mechanics, any expectation value is computed w.r.t a probability distribution $p(x)$. For example, $$\langle x\rangle_\psi\equiv\langle \psi|x|\psi\rangle=\int\limits_{-\infty}^{\infty}xp(x)dx\tag{3}$$ where $p(x)=\psi^*(x)\psi(x)$.

Which probability distribution $p[\phi(x)]$ is used to compute (1)? Does the path-integral quantization provide any answer?


The idea is exactly the same as in single particle QM. Take the expectation value $\langle 0 | \hat\phi(x) | 0 \rangle$ and insert a complete set of field eigenstates $|\varphi\rangle$; these are the states that satisfy $\hat\phi(x) |\varphi\rangle = \varphi(x)|\varphi\rangle$. We do this with a path integral:

$$\langle 0 | \hat\phi(x) | 0 \rangle = \int \mathcal{D}\varphi\ \langle 0 | \hat\phi(x)|\varphi\rangle \langle\varphi | 0 \rangle = \int \mathcal{D}\varphi\ \langle 0 |\varphi\rangle \langle\varphi | 0 \rangle \varphi(x).$$

So your probability distribution is $p[\varphi] = |\langle\varphi | 0 \rangle|^2$, just like in regular QM. I'm not sure yet how to calculate this, but by analogy with the harmonic oscillator it would probably be along the lines of $\exp(-\varphi^2)$.

Edit: I think I've managed to calculate it, though I make no guarantees of this being correct. In particular I'm pretty sure there should be some $2\pi$'s around.

We proceed by analogy with the harmonic oscillator. You can find the ground state wavefunction by using the fact that $\langle x | \hat a | 0 \rangle = 0$; this gives you a differential equation for $\langle x | 0 \rangle = 0$. In much the same way, we write $\langle \varphi | \hat a_\mathbf{p} | 0 \rangle = 0$. We then express $\hat a_\mathbf{p}$ in terms of $\hat \phi$ and $\hat \pi$, and use the fact that acting on a wavefunctional $\hat \pi$ is $-i \delta / \delta \varphi(x)$. I did it following Peskin and Schroeder's conventions for normalizations and working in the Schrödinger picture, and arrived at the following equation (kx is the dot product $\mathbf{k} \cdot \mathbf{x}$):

$$ \int \mathrm{d}^3x\ e^{-ikx} \left( \omega_k \varphi(x) \langle\varphi | 0 \rangle + \frac{\delta}{\delta \varphi(x)}\langle\varphi | 0 \rangle \right) = 0$$

We will eventually express everything in terms of the Fourier transform $\hat \varphi$ of $\varphi$ (who knew, right?), so we will need to use the fact that

$$ \int \mathrm{d}^3x\ e^{-ikx} \frac{\delta}{\delta \varphi(x)}\langle\varphi | 0 \rangle = \frac{\delta}{\delta \hat \varphi} \langle\varphi | 0 \rangle \bigg|_{-k}$$

so our equation becomes

$$\frac{\delta}{\delta \hat \varphi (k)} \langle\varphi | 0 \rangle = - \omega_k \hat \varphi (-k) \langle\varphi | 0 \rangle$$

and we can guess the solution

$$\langle\varphi | 0 \rangle = \exp \left( -\frac12 \int \mathrm{d}^3k\ \omega_k \hat \varphi(k) \hat \varphi(-k) \right).$$

You could rewrite this in position space but the Fourier transform of $\omega_k$ involves Bessel functions so I don't know if it's worth it.


I think the path integral approach provides an answer. In terms of the normalized generating functional $\mathcal{Z}[J]=\frac{Z[J]}{Z[0]}$, the expectation value is given by $$\langle 0|\phi(x)|0\rangle=\frac{\delta \mathcal{Z}[J]}{\delta J(x)}\Bigg|_{J=0}=\int\mathcal{D}\phi\phi(x)\exp\Big[\int i(\mathcal{L}+j\phi)d^4x\Big]\Big|_{J=0}\tag{1}$$ which when compared with the average of any function w.r.t a normalized probability distribution given by $$\overline{f(u)}=\int \limits_{a}^{b}duf(u)p(u)$$ tells that the relevant probability distribution in this case (1) is $$p[\phi(x)]=\exp(iS)=\text{complex}!!!$$ But what bugs me here is that $p[\phi(x)]$ is complex, and therefore, the name probability distribution may not be justified.

  • 2
    $\begingroup$ QFT is not necessarily relativistic. If you look at the Euclidean theory, then you get a perfectly nice $\mathrm{e}^{-S}$, which is the main reason for the better convergence behaviour there. $\endgroup$ – ACuriousMind May 29 '17 at 17:17
  • 1
    $\begingroup$ Minor detail: there should be an integral inside the exponential, and you forgot to set $J=0$. Your "complex probability distribution" is then $\exp(iS)$. $\endgroup$ – Javier May 30 '17 at 18:18

You may as well ask which $|\psi\rangle$ to use. We simply use the vacuum state, viz. $\langle 0|a_\mathbf{p}|0\rangle =0$. Thus $p=|\langle 0|x\rangle |^2$.

  • $\begingroup$ What do you mean by $|x\rangle$ in quantum field theory? $\endgroup$ – SRS May 29 '17 at 14:37
  • $\begingroup$ The label $x$ couldn't denote a single particle's position as it does in QM, as you've probably noticed. It would have to be another label of a complete basis, e.g. a $4$-momentum. $\endgroup$ – J.G. May 29 '17 at 14:50

protected by Qmechanic May 30 '17 at 4:40

Thank you for your interest in this question. Because it has attracted low-quality or spam answers that had to be removed, posting an answer now requires 10 reputation on this site (the association bonus does not count).

Would you like to answer one of these unanswered questions instead?

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