My question is about the perturbative expansion of the S-matrix using Dyson's expansion.

Let the Lagrangian density of the $\phi^4$ theory be \begin{equation} \mathcal{L} = \frac{1}{2}\left[\partial_\mu\phi(x)\right]^2 - \frac{m^2}{2}\phi(x)^2 - \frac{\lambda}{4!}\phi(x)^4, \end{equation} from which we find the Hamiltonian density \begin{equation} \mathcal{H}(x) = \underbrace{\frac{1}{2}\left[ \dot\phi^2 + (\nabla\phi)^2 +m^2\phi^2 \right]}_{\mathcal{H}_0} + \underbrace{\frac{\lambda}{4!}\phi^4}_{\mathcal{H}_1}. \end{equation} Now, to expand \begin{equation} S = T\mathrm{e}^{-i\int\mathrm{d}^4z\mathcal{H}_I(z)}, \end{equation} we clearly need to identify $\mathcal{H}_I$ in the interaction picture. This is where I start to get confused. The Hamiltonian $H_I$ is given by \begin{equation} H_I = \mathrm{e}^{iH_0t}H_1\mathrm{e}^{-iH_0t}, \end{equation} which reduces to $H_I=H_1$ only if $[H_0,H_1]=0$. However, all of the textbooks I've seen simply use $\mathcal{H}_1$ (as above) for $\mathcal{H}_I$ in computing the S-matrix.

Here is my question: It is not clear to me that $[H_0,H_1]=0$, which would surely require $[\mathcal{H}_0,\mathcal{H}_1]=0$ (?), which implies that $[\partial_\mu\phi,\phi]=0$. However, this latter commutator does not seem to give zero when the mode expansion is used. Furthermore, I am not entirely sure how the condition $[H_0,H_1]=0$ is inferred from the Hamiltonian density. Does one need to evaluate $[\mathcal{H}_0(x),\mathcal{H}_1(x)]$, IE at the same point in spacetime?

I know that this is a small point, but I hate not knowing the logic going from one step to the next, so if anyone could shed light on the justification for using $\mathcal{H}_1$ for $\mathcal{H}_I$, I would be very appreciative.

  • $\begingroup$ "All the textbooks I've seen simply use $\mathcal{H}_1$ for $\mathcal{H}_I$ in the S-matrix" - this isn't exactly right. $\mathcal{H}_I$ looks similar to $\mathcal{H}_1$, but it uses the interaction picture fields rather than the Schrodinger picture fields $\endgroup$ Jan 13, 2021 at 10:21
  • $\begingroup$ You're right - I shouldn't generalise! I refer mostly to Lancaster & Blundell Ch 19. $\endgroup$
    – dsfkgjn
    Jan 13, 2021 at 11:51
  • $\begingroup$ I actually guessed that you did, L&B are a bit sloppy with their notation for fields $\endgroup$ Jan 13, 2021 at 11:57

1 Answer 1


In general, $[H_0, H_1] \ne 0$ and $H_1 \ne H_I$. L&B describe the interaction picture in terms of state-vectors, which I'm not the biggest fan of, so we'll work from the point of the field operators instead. The fields in the interaction picture, $\phi_I$, are defined in terms of the Schrodinger-picture fields by $$ \phi_I(t, \vec{x})= e^{iH_0(t-t_0)} \phi_S(t_0, \vec{x})e^{-iH_0(t-t_0)} \tag{1} $$ In the Heisenberg picture, the fields evolve with the full Hamiltonian $H$: the above equation is defined using $H_0$ which is the first-order approximation to the full Hamiltonian when we expand around $\lambda = 0$, since we're working perturbatively. You shouldn't think of eq. $(1)$ as an equation for time evolution, but rather as a definition for convenience.

The mode expansion for $\phi_S$ is: $$ \phi_S(\vec{x})=\int\frac{d^3 k}{(2\pi)^3}\frac1{\sqrt{2\omega_k}}\left(a(\vec{k})e^{i\vec{k}\cdot\vec{x}}+a^\dagger(\vec{k})e^{-i\vec{k}\cdot\vec{x}}\right) $$ and consequently, $\phi_I$ is (note carefully the four-vector notation, in contrast to the previous equation) $$ \phi_I(x)=\int\frac{d^3 k}{(2\pi)^3}\frac1{\sqrt{2\omega_k}}\left(a(\vec{k})e^{-ikx}+a^\dagger(\vec{k})e^{ikx}\right) $$ where $x_0 = \Delta t = t - t_0$ and $k^\mu$ is on-shell, so $\phi_I$ enjoys a free mode expansion. It is these $\phi_I$ that the interaction Hamiltonian in the interaction picture (annoyingly, both these concepts have the same name) $H_I$ in the S-matrix is expanded in terms of, so while it looks similar visually to $H_1$ (which uses $\phi_S$), the interaction picture encodes some of the time-dependence and thus hides a lot of the details. When you set up the differential equation for $U(t, t_0)$: $$ i\frac{\partial U(t, t_0)}{\partial t} = H_I(t)U(t, t_0), $$

$H_I$ is defined by wedging $H_1$ in between the same $e^{\pm iH_0(t-t_0)}$ as the fields, so you can see that $$H_I = e^{iH_0(t-t_0)} H_1 e^{-iH_0(t-t_0)} = e^{iH_0(t-t_0)} \phi^4_S(t_0, \vec{x})e^{-iH_0(t-t_0)} \\= e^{iH_0(t-t_0)} \phi_S(t_0, \vec{x})e^{-iH_0(t-t_0)}e^{iH_0(t-t_0)}\phi_Se^{-iH_0(t-t_0)}...\phi_Se^{-iH_0(t-t_0)} = \phi_I^4,$$

with the obvious integrals and prefactors.


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

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