Is there a good reference for the distinction between projection operators in QFT, with an eigenvalue spectrum of $\{1,0\}$, representing yes/no measurements, the prototype of which is the Vacuum Projection Operator $\left|0\right>\left<0\right|$, which allows the elementary construction of a panoply of projection operators such as $$\frac{\hat\phi_f\left|0\right>\left<0\right|\hat\phi_f^\dagger}{\left<0\right|\hat\phi_f^\dagger\hat\phi_f\left|0\right>},\qquad \frac{\hat\phi_f\hat\phi_g\left|0\right>\left<0\right|\hat\phi_g^\dagger\hat\phi_f^\dagger}{\left<0\right|\hat\phi_g^\dagger\hat\phi_f^\dagger\hat\phi_f\hat\phi_g\left|0\right>},$$ or of higher degree; in contrast to (smeared) field operators such as $\hat\phi_f$, which have a continuous spectrum of eigenvalues? I see this as effectively the distinction between, respectively, the S-matrix and the Wightman field as observables.

I'm particularly interested in anything that considers the operational difference between these different classes of QFT observables in detail. It's obvious that the projection operators are nonlocal, insofar as they clearly don't satisfy microcausality, in contrast to the requirement of microcausality for the field operators. It also seems that the field operators cannot be used on their own to construct models for the detection of a particle, which is a yes/no event, without introducing the vacuum projection operator (but *is there* a way of constructing projection operators without introducing the vacuum projection operator? EDIT: yes, obviously enough, "is the observed value in the range $[a..b]$" is a yes/no observable, *etc., etc., ...*.)

This post has been migrated from (A51.SE)