Dismiss Notice
Join Physics Forums Today!
The friendliest, high quality science and math community on the planet! Everyone who loves science is here!

Spatial representation of field commutator

  1. May 8, 2008 #1
    Hi all!

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

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


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

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

    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 [tex]x_0[/tex] is zero, then
    [tex]\Delta \left( x \right) = \frac{1}
    {i}\int {\frac{{d^3 \vec p}}
    {{\left( {2\pi } \right)^3 }}\frac{1}
    {{2\sqrt {\vec p^2 + m^2 } }}\left( {e^{ - i\vec p \cdot \vec x} - e^{ - i\vec p \cdot \vec x} } \right)}
    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 [tex]\vex x = 0[/tex] and we get
    \Delta \left( x \right) = \frac{1}
    {i}\int {\frac{{d^3 \vec p}}
    {{\left( {2\pi } \right)^3 }}\frac{1}
    {{2\sqrt {\vec p^2 + m^2 } }}\left( {e^{ - iEx_0 } - e^{iEx_0 } } \right)} \hfill \\
    = 4\pi \frac{1}
    {i}\int {\frac{{dp}}
    {{\left( {2\pi } \right)^3 }}\frac{{p^2 }}
    {{2\sqrt {p^2 + m^2 } }}\left( {e^{ - iEx_0 } - e^{iEx_0 } } \right)} \hfill \\
    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!!

  2. jcsd
  3. May 8, 2008 #2
    Consider doing the integral in terms of E instead of p?
  4. May 8, 2008 #3
    Yes, sadly... same result :(
  5. May 8, 2008 #4
    Peskin says
    [tex]D(x-y) = \frac{1}{4\pi^2} \int_m^\infty\!dE\, \sqrt{E^2 - m^2} e^{-i E t}[/tex]
    from which I suspect you'll find your answer. It's in the lower bound of the E integral.
  6. May 8, 2008 #5
    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 [tex]e^{-m t}[/tex].
    Last edited: May 8, 2008
  7. May 8, 2008 #6
    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 [tex]e^{-imt][/tex] 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 [tex]\sqrt{E^2 - m^2[/tex] 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?

  8. May 8, 2008 #7


    User Avatar
    Science Advisor

    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: May 8, 2008
  9. May 9, 2008 #8

    Hans de Vries

    User Avatar
    Science Advisor
    Gold Member

    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:


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


    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:


    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: Apr 23, 2017
  10. May 9, 2008 #9

    Hans de Vries

    User Avatar
    Science Advisor
    Gold Member

    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: Apr 23, 2017
  11. May 9, 2008 #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!


    PS: @strangerep Thanks for the hint of the stationary phase method, helped me a lot!
  12. Dec 29, 2011 #11
  13. Dec 30, 2011 #12


    User Avatar
    Science Advisor

    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...)
  14. Dec 31, 2011 #13


    User Avatar
    Science Advisor

    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 [itex]e^{-imt}[/itex] (for [itex]t\to\infty[/itex] 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:
    H^{(2)}_0(z) ~\sim~ \sqrt{\frac{2}{\pi z}} \, e^{-iz} ~~~~~~[|z|\to\infty]
    (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 [itex]1/\sqrt{t}[/itex]).
Share this great discussion with others via Reddit, Google+, Twitter, or Facebook