How is the path integral used to compute -point functions? The LSZ formula tells us that they are what we need to compute amplitudes. Once again I follow Ramond’s “Field Theory, a Modern Primer”.
Introduce a driving force or current coupled to the field,
Apply a convergence kludge to assuage any suspicions about convergence of the Gaussian integration
and use the Fourier transform
This is a unitary transformation, so . We can do the integrations in the exponent if and
(1)
Change variables
in which you recognize the Green function for the oscillator, then
and finally
(2)
Let’s get back into the time domain
(3)
Note that
This justifies it’s Green function name;
Heaviside step functions () have integral representations
obey if and is zero otherwise. You should prove that the integrals on the right sides of these formulas obey this relation, using the Cauchy theorem.
In the first integral we get a nonzero result iff we can complete the path of integration in the lower half-plane, requiring for , which happens if only. In the second integral we get a nonzero result iff we can complete the path of integration in the upper half-plane, requiring for , which happens if only. The Heaviside function is
Explicitly evaluate ;
You can show that , these are the usual oscillator retarded/advanced Green functions.
So what about ? The poles are at
You can show that , these are the usual oscillator retarded/advanced Green functions.\\
So what about ? The poles are at
We can complete the contour in the lower half-plane when so that as , contributing
We can complete the contour in the upper half-plane when so that as , contributing
resulting in a mix of retarded/advanced signals
Ramond (which I am following closely) makes an interesting statement regarding this; “This is a precursor of the Feynmann propagator which describes signal propagation from two sources: positive energy (particle) states moving in positive time and negative energy (antiparticle) states moving backward in time”.
Notation;
This is a generating function for -point functions.\\
-point functions. Starting with dividing into intervals , let be somewhere on this interval,
(4)
(5)
…from which it should be clear from the expansion on the right-side that it is time ordered; pick two {\bf arbitrary times} , then each occurs at some multiple of , ,
(6)
so the object on the right automatically produces time ordered operator product amplitudes.
Introduce variational derivatives
and returning momentarily to the oscillator
Let where is some force or current, then with and , , the variational derivatives will ignore all functions of for which in
(7)
from which you can see that the path integral generates time-ordered expectations of operator products for
At assume that the sources/drivers are zero, let be the ground/vacuum states at these times and we are using ;
(8)
Without sources/drivers . All of the following are equivalent
(9)
provided factors like are absorbed into .