Spatial representation of field commutator

blue2script
Messages
44
Reaction score
0
Hi all!

I worked for hours on this simple commutator of real scalar fields in qft:

\left[\Phi\left(x\right),\Phi\left(y\right) \right] = i\Delta\left( x-y \right)

where

\Delta\left(x\right) = \frac{1}{i}\int {\frac{{d^4 p}}<br /> {{\left( {2\pi } \right)^3 }}\delta \left( {p^2 - m^2 } \right)\operatorname{sgn} \left( {p^0 } \right)e^{ - ip \cdot x} }

The task is to solve this integral and to look at the cases x^2 = 0 and m\rightarrow 0.

But, what I get in my calculations is that for space like x the commutator is zero whereas the integral diverges for time-like x. Normally there should be a Bessel function involved in the end and I am just totally confused now. But maybe I should show the steps I took:

i) Spacelike x

The expression is Lorentz-invariant. For space-like x we go into a reference frame where x_0 is zero, then
\Delta \left( x \right) = \frac{1}<br /> {i}\int {\frac{{d^3 \vec p}}<br /> {{\left( {2\pi } \right)^3 }}\frac{1}<br /> {{2\sqrt {\vec p^2 + m^2 } }}\left( {e^{ - i\vec p \cdot \vec x} - e^{ - i\vec p \cdot \vec x} } \right)} <br />
and this is zero. What argument fails here? Calculating just one term of the bracket leads to a modified Bessel function. I followed Weinberg, p. 202 here - its strange...

ii) Timelike x

Then we go into a reference frame where \vex x = 0 and we get
\begin{gathered}<br /> \Delta \left( x \right) = \frac{1}<br /> {i}\int {\frac{{d^3 \vec p}}<br /> {{\left( {2\pi } \right)^3 }}\frac{1}<br /> {{2\sqrt {\vec p^2 + m^2 } }}\left( {e^{ - iEx_0 } - e^{iEx_0 } } \right)} \hfill \\<br /> = 4\pi \frac{1}<br /> {i}\int {\frac{{dp}}<br /> {{\left( {2\pi } \right)^3 }}\frac{{p^2 }}<br /> {{2\sqrt {p^2 + m^2 } }}\left( {e^{ - iEx_0 } - e^{iEx_0 } } \right)} \hfill \\ <br /> \end{gathered}<br />
This integral diverges quadratically.

So there has to be something wrong, otherwise this task wouldn't make any sense... I hope somebody can help me here, I am totally lost with this task. I worked hours on that problem and I don't know what to do any more.

A big thanks in advance to everybody!

Blue2script
 
Physics news on Phys.org
Consider doing the integral in terms of E instead of p?
 
Yes, sadly... same result :(
 
Peskin says
D(x-y) = \frac{1}{4\pi^2} \int_m^\infty\!dE\, \sqrt{E^2 - m^2} e^{-i E t}
from which I suspect you'll find your answer. It's in the lower bound of the E integral.
 
My bad. This might not be exactly your expression, which is due to the pole prescription in the Feynman propagator. In any event, I think you'll be able to get something that goes like e^{-m t}.
 
Last edited:
ah, ok, I see that on p. 26 is exactly what I wanted. Thanks for the hint! But then again: I can't see why there the integral 2.51 goes to e^{-imt] for t goes to infinity? I mean, sure, for big E the oscillation is too high and gives zero. But then on the lower end the prefactor \sqrt{E^2 - m^2 gives zero. On the other hand, if we don't take the m in the exponent too seriously, this is true.

A big Thanks to you, Ibrits! Do you have a better explanation for the limit t->\infinity?

Blue2script
 
blue2script said:
[...]a better explanation for the limit t->\infinity?
It's using the method of "stationary phase approximation".
See http://en.wikipedia.org/wiki/Stationary_phase_approximation

BTW, are you still trying to solve the spacelike case? If so, Scharf's
"Finite Quantum Electrodynamics", pp64-69, might help.
 
Last edited:
Notice that Weinberg's results are totally different as those from P&S and Zee.
(Weinberg is right, the others are wrong). The exact definition of the outside
the light-cone behavior of the Feynman propagator is given by:

\frac{1}{4\pi^2}~\frac{m}{\sqrt{r^2-t^2}}~K_1(m\sqrt{r^2-t^2})

This corresponds to Weinberg's equation (5.2.9). For small arguments we can
simply replace the Bessel function K_1(r) with 1/r, for instance:

K_1(0.001)~=~999.996238

Try it yourself http://functions.wolfram.com/webMathematica/FunctionEvaluation.jsp?name=BesselK". Since m is very small we can in general simply replace
Weinberg's (5.2.9) with:

\frac{1}{4\pi^2}~\frac{1}{r^2-t^2}

Which is independent of the mass m (!) and much more serious.

It's hard for me to understand, now it has become more or less routine to make
anti-hydrogen in the laboratory, how one can maintain the idea that anti-matter
propagates backward in time... Going backwards in time we see:

Somehow gamma flashes mysteriously produce anti-protons and positrons at the
inside surface of a vacuum container. These instantaneously pair up to form
anti-hydrogen which turns out to be very cold (amazingly after such high energy
events) After a while the anti-hydrogen spontaneously ionizes (it's instable).
The constituents are then accelerated to ultrarelativistic speed and end their
lives in high energy synchrotron collisions..

Oh well...



Regards, Hans
 
Last edited by a moderator:
Hans de Vries said:
This corresponds to Weinberg's equation (5.2.9). For small arguments we can
simply replace the Bessel function K_1(r) with 1/r, for instance:

K_1(0.001)~=~999.996238

Try it yourself http://functions.wolfram.com/webMathematica/FunctionEvaluation.jsp?name=BesselK". Since m is very small we can in general simply replace
Weinberg's (5.2.9) with:

\frac{1}{4\pi^2}~\frac{1}{r^2-t^2}

Oops, no more time to edit over-impulsive posts. Rushing back home didn't help...
Of course m should be mc/hbar which is not small. So the 1/r behavior is indeed
below the Compton radius. Later on I want to say a few things more.


Regards, Hans
 
Last edited by a moderator:
  • #10
Hi Hans,

I just came back seeing your posts... thanks a lot for your answers! I am impressed... and will check everyting you said ;). But, well, that won't happen before tuesday, I am on a bike tour the next three days.

But please go on to continue writing, I am very interested! There are so many mysterious things in qft, especially for a, well somewhat, a beginner (I hear two courses on qft right now and already had courses in qed and qcd).

So, if I may please you, I am interested in everything related with this topic! A big thanks in advance for your couragement!

Have a nice weekend, I will check back monday evening!

Blue2script

PS: @strangerep Thanks for the hint of the stationary phase method, helped me a lot!
 
  • #11
strangerep said:
It's using the method of "stationary phase approximation".
See http://en.wikipedia.org/wiki/Stationary_phase_approximation
Could you explain how in more details? I don't see how to apply the stationary phase method because in this case the phase factor's 1st derivative is never 0.
 
  • #12
kof9595995 said:
Could you explain how in more details? I don't see how to apply the stationary phase method because in this case the phase factor's 1st derivative is never 0.
Sorry, I don't have enough time right now, but I'll try to remember to get back to this sometimes next week. (Since you're asking a necro-question I'll have to refresh my memory...)
 
  • #13
strangerep said:
It's using the method of "stationary phase approximation".
[...]
Well, well. I wrote that 3.5 years ago, and now have the benefit of that extra study time, so I must now post a correction...

The e^{-imt} (for t\to\infty is not obtained by the stationary phase method. Indeed the stationary phase method doesn't seem to work here at all -- neither by a stationary exponent (critical point of the 1st kind) nor the boundary points of the integration (critical points of the 2nd kind).

A much better reference for calculating these propagators is

G. Scharf, "Finite Quantum Electrodynamics -- The Causal Approach".

In the section titled "Discussion of the Commutation Functions", he calculates lots of these propagators exactly.

The answer is that the exact propagator is a Bessel function (actually a Hankel function in our case). The asymptotic behaviour of for large arguments approaches something like this:
<br /> H^{(2)}_0(z) ~\sim~ \sqrt{\frac{2}{\pi z}} \, e^{-iz} ~~~~~~[|z|\to\infty]<br />
(Ref: Abramowitz & Stegun.)

(In our case, z is mt.)

Imho, P&S really should have given a reference instead of simply quoting the asymptotic result (and should also have mentioned the factor of 1/\sqrt{t}).
 

Similar threads

Replies
3
Views
2K
Replies
5
Views
2K
Replies
4
Views
2K
Replies
15
Views
2K
Replies
4
Views
2K
Replies
6
Views
2K
Back
Top