8

As the Lagrangian in classical mechanics corresponds to energy, it must be real. But is that the case in quantum field theory? I mean, it should still correspond to some sort of energy, but what about all the "$i$"s here and there, such as in the Dirac Lagrangian $i\bar{\psi}\gamma^\mu\partial_\mu \psi$ and the current density $J_\mu = ie[\dots]$ (see Griffiths for example)?

Another question is, how can it be hermitian, $\mathcal{L} = \mathcal{L}^\dagger$, when we have those "$i$"s? Wouldn't I get a minus sign if I complex-conjugated the interaction term and the Dirac field term? I'm really confused and hope someone can help

Qmechanic
  • 201,751
UserX
  • 83

1 Answers1

7

In quantum field theory, the Lagrangian density is an operator, not a number. So it doesn't make sense to say it has to be real; "real" is a term that applies to numbers, not operators.

What does have to be true is that $\mathcal{L}$ has to have real expectation values in all physical states, and that in turn means it has to be hermitian (what mathematicians call self-adjoint). But hermiticity is not just a matter of being real. You can have other non-hermitian factors besides $i$. In particular, the derivative $\partial_\mu$ in the Dirac Lagrangian is antihermitian, and so the combination $i\partial_\mu$ as a whole is hermitian.

David Z
  • 76,371
  • Thanks for clarifying :-) Could you explain what do you mean by operator here? I thought it is a function of fields and their derivatives. When you say operator, to my mind this means that it acts on something, but what does the Lagrangian density in QFT act on? – UserX Dec 11 '12 at 03:10
  • 1
    Fields are operators in QFT. And when you combine operators, you get another operator; that's why the Lagrangian, as a combination of fields and derivative operators, is an operator itself. The operators act on states of the universe, such as the vacuum state $\lvert 0\rangle$, or n-particle states made by applying the field operators to that state, e.g. $u\lvert 0\rangle$ is a state with one up quark. – David Z Dec 11 '12 at 03:19
  • I see, that makes sense. Sorry if I'm asking stupid questions but I didn't even start studying this stuff formally yet. – UserX Dec 11 '12 at 03:27
  • 2
    Dear @David, I would mostly disagree that the Lagrangian isn't a number. The only meaningful quantum formalism that uses the actions and Lagrangians is the Feynman path-integral approach and in that approach, they are c-number-valued functions of classical observables that are integrated over. ... Also, L is often unreal if we deal with odd-order terms in the Euclidean space (like Chern-Simons etc). – Luboš Motl Dec 11 '12 at 08:29
  • @Luboš I wouldn't say the path integral approach is the only meaningful one... at least, not in the sense that everything needs to be done that way. Yes, all of QFT does trace back to path integrals in the end, but in practice when doing perturbative QFT we use operator-valued Lagrangians directly, and that seems to be the level that the question is asked at. – David Z Dec 11 '12 at 19:09
  • @David, I'm with Lubos. You can construct a lot of operators from the "classical lagrangian", because the second derivative in the lagrangian formalism mixes the field over time in a way that you can build an operator representing the lagragian in terms of field variables. If you apply the Legendary transform you end up with the Hamilton, that you can represent in terms of operators and a term $pdq/dt$ that are ambiguous because the discretization of time and the time ordering assumed by the representation in hand. – Nogueira Dec 05 '15 at 19:07
  • @DavidZ Can you explain why the Lagrangian density has to real in classical field theory and it must have real expectation value in QFT? – SRS Jan 02 '17 at 12:42
  • @SRS see https://physics.stackexchange.com/q/127797/ – David Z Jan 02 '17 at 18:26
  • To use a concrete example - in Lattice QCD (or other numerical evaluations where you sum over a set of "paths" (configurations really in LQCD), does the action come out real or complex? The action that is then summed using e^S or e^iS depending on if you wick rotated or not. You sum theoretically over all possible configs, not just "real" configs. – BjornW Jan 18 '23 at 09:04