- #1
RedX
- 970
- 3
According to Wikipedia ( http://en.wikipedia.org/wiki/Propagator_(Quantum_Theory)#Feynman_propagator ) the propagator for the KG-field is:
[tex] \ \Delta_F(x,y)
\ = \lim_{\epsilon \to 0} \frac{1}{(2 \pi)^4} \int d^4p \, \frac{e^{-ip(x-y)}}{p^2 - m^2 + i\epsilon}
\ = \left \{ \begin{matrix} -\frac{1}{4 \pi} \delta(s) + \frac{m}{8 \pi \sqrt{s}} H_1^{(1)}(m \sqrt{s}) & \textrm{ if }\, s \geq 0 \\ -\frac{i m}{ 4 \pi^2 \sqrt{-s}} K_1(m \sqrt{-s}) & \textrm{if }\, s < 0. \end{matrix} \right [/tex]
where [tex]s:= (x^0 - y^0)^2 - (\vec{x} - \vec{y})^2. [/tex]
Since these are complicated functions, take the mass equal to zero, and using the properties of these functions for small arguments ( http://en.wikipedia.org/wiki/Hankel_function#Asymptotic_forms ) you get:
[tex] \ \Delta_F(x-y) =
\left \{ \begin{matrix} -\frac{1}{4 \pi} \delta(s) - \frac{i}{4 \pi^2 s}} & \textrm{ if }\, s \geq 0 \\ -\frac{i }{ 4 \pi^2 (-s)} & \textrm{if }\, s < 0. \end{matrix} \right[/tex]
This expression seems perfectly finite except at s=0. In a phi4 theory, there is only one vertex for the next order correction to the propagator, so call the coordinate of this vertex z, and then you'll get:
[tex]\Delta(x-y)=\Delta_F(x-y)+C*\int d^4z \Delta_F(x-z) \Delta_F(z-z) \Delta_F(z-y)=\Delta_F(x-y)+C*\Delta_F(0)\int d^4z \Delta_F(x-z) \Delta_F(z-y) [/tex]
for a constant C proportional to the coupling (this expression can be gotten from the Feynman diagram, with a line going from x to z, z to z, and z to y). My question is what's going on with this integral:
[tex]\int d^4z \Delta_F(x-z) \Delta_F(z-y)[/tex]
Qualitatively when the value of z goes to x in the integral, it can do so from two different directions - from a space-like direction, or a time-like direction. If the latter, then a delta function will be picked up, in addition to an infinity term 1/s (i.e., 1/(x-z)2).
But the infinity term is harmless, because the integral is of the order d4z which is greater than the denominator which is of order z2.
So the only trouble is [tex]\Delta_F(0) [/tex], and that's taken care of by a counter-term [tex]-\frac{C\Delta_F(0)}{2} \phi^2 [/tex] in the Lagrangian (hopefully that minus sign is correct and there are no terms of i: does anyone have a good way of memorizing all those little factors of i and -1?).
Now take a phi3 theory, forgetting for the moment that for such a theory we can't find the vacuum state (but we take it in 6 dimensions for it to be renormalizeable). For the next order correction in the propagator, there are two vertices z and q. The propagator is:
[tex]\Delta(x-y)=\Delta_F(x-y)+C*\int d^6z d^6q \Delta_F(x-z) \Delta_F(z-q) \Delta_F(z-q) \Delta_F(y-q)=
\Delta_F(x-y)+C*\Delta_F(0) \int d^6z d^6q \Delta_F(x-z) \Delta_F(z-q) \Delta_F(y-q)
[/tex]
Again, the divergent [tex] \Delta_F(0)[/tex] term, but my question is now about the integral next to it:
[tex]\int d^6z d^6q \Delta_F(x-z) \Delta_F(z-q) \Delta_F(y-q) [/tex]
This looks convergent, even if both z goes to x and q goes to y simultaneously in the integral.
So far, in all these cases, the infinities come from a [tex] \Delta_F(0)[/tex] term, and not the integral term next to it. Is it true to say that infinities in quantum field theory come only from [tex] \Delta_F(0)[/tex] terms, which correspond to self-interactions? That is, integrals of products of delta functions converge (which to me is not so obvious - with something like [tex]\frac{d^4x}{x^2}[/tex] I'm worried about infinity more than 0), but it is an overall [tex] \Delta_F(0)[/tex] that causes all the trouble?
Also, all this analysis is based on m=0. Would the propagator be radically different if m was not set equal to zero?
[tex] \ \Delta_F(x,y)
\ = \lim_{\epsilon \to 0} \frac{1}{(2 \pi)^4} \int d^4p \, \frac{e^{-ip(x-y)}}{p^2 - m^2 + i\epsilon}
\ = \left \{ \begin{matrix} -\frac{1}{4 \pi} \delta(s) + \frac{m}{8 \pi \sqrt{s}} H_1^{(1)}(m \sqrt{s}) & \textrm{ if }\, s \geq 0 \\ -\frac{i m}{ 4 \pi^2 \sqrt{-s}} K_1(m \sqrt{-s}) & \textrm{if }\, s < 0. \end{matrix} \right [/tex]
where [tex]s:= (x^0 - y^0)^2 - (\vec{x} - \vec{y})^2. [/tex]
Since these are complicated functions, take the mass equal to zero, and using the properties of these functions for small arguments ( http://en.wikipedia.org/wiki/Hankel_function#Asymptotic_forms ) you get:
[tex] \ \Delta_F(x-y) =
\left \{ \begin{matrix} -\frac{1}{4 \pi} \delta(s) - \frac{i}{4 \pi^2 s}} & \textrm{ if }\, s \geq 0 \\ -\frac{i }{ 4 \pi^2 (-s)} & \textrm{if }\, s < 0. \end{matrix} \right[/tex]
This expression seems perfectly finite except at s=0. In a phi4 theory, there is only one vertex for the next order correction to the propagator, so call the coordinate of this vertex z, and then you'll get:
[tex]\Delta(x-y)=\Delta_F(x-y)+C*\int d^4z \Delta_F(x-z) \Delta_F(z-z) \Delta_F(z-y)=\Delta_F(x-y)+C*\Delta_F(0)\int d^4z \Delta_F(x-z) \Delta_F(z-y) [/tex]
for a constant C proportional to the coupling (this expression can be gotten from the Feynman diagram, with a line going from x to z, z to z, and z to y). My question is what's going on with this integral:
[tex]\int d^4z \Delta_F(x-z) \Delta_F(z-y)[/tex]
Qualitatively when the value of z goes to x in the integral, it can do so from two different directions - from a space-like direction, or a time-like direction. If the latter, then a delta function will be picked up, in addition to an infinity term 1/s (i.e., 1/(x-z)2).
But the infinity term is harmless, because the integral is of the order d4z which is greater than the denominator which is of order z2.
So the only trouble is [tex]\Delta_F(0) [/tex], and that's taken care of by a counter-term [tex]-\frac{C\Delta_F(0)}{2} \phi^2 [/tex] in the Lagrangian (hopefully that minus sign is correct and there are no terms of i: does anyone have a good way of memorizing all those little factors of i and -1?).
Now take a phi3 theory, forgetting for the moment that for such a theory we can't find the vacuum state (but we take it in 6 dimensions for it to be renormalizeable). For the next order correction in the propagator, there are two vertices z and q. The propagator is:
[tex]\Delta(x-y)=\Delta_F(x-y)+C*\int d^6z d^6q \Delta_F(x-z) \Delta_F(z-q) \Delta_F(z-q) \Delta_F(y-q)=
\Delta_F(x-y)+C*\Delta_F(0) \int d^6z d^6q \Delta_F(x-z) \Delta_F(z-q) \Delta_F(y-q)
[/tex]
Again, the divergent [tex] \Delta_F(0)[/tex] term, but my question is now about the integral next to it:
[tex]\int d^6z d^6q \Delta_F(x-z) \Delta_F(z-q) \Delta_F(y-q) [/tex]
This looks convergent, even if both z goes to x and q goes to y simultaneously in the integral.
So far, in all these cases, the infinities come from a [tex] \Delta_F(0)[/tex] term, and not the integral term next to it. Is it true to say that infinities in quantum field theory come only from [tex] \Delta_F(0)[/tex] terms, which correspond to self-interactions? That is, integrals of products of delta functions converge (which to me is not so obvious - with something like [tex]\frac{d^4x}{x^2}[/tex] I'm worried about infinity more than 0), but it is an overall [tex] \Delta_F(0)[/tex] that causes all the trouble?
Also, all this analysis is based on m=0. Would the propagator be radically different if m was not set equal to zero?