Feynman rules and vacuum polarization

Introduction to the physics of
highly charged ions
Lecture 11: Feynman rules and
vacuum polarization
Zoltán Harman
[email protected]
Universität Heidelberg, 27.01.2014
Recapitulation of propagator theory
Defining equation for the Feynman propagator for electrons and
positrons (in relativistic units: ~ = c = 1, and with Feynman slash
notation: A/ = γµ Aµ ):
/ 0 − eA/0 − m)GF (x0 , x; A) = δ (4) (x0 − x) .
(i∇
The free (A = 0) relativistic propagator satisfies the equation
0
/ − m)GF (x0 − x) = δ (4) (x0 − x) .
(i∇
Final result with a positive imaginary part in
the denominator:
Z
p+m
d4 p −ip(x0 −x)
/
0
GF (x −x) =
e
.
4
2
(2π)
p − m2 + i
The full propagator and S matrix
Iterative solution for the full propagator:
Z
0
0
/ 1 )GF (x1 − x)
GF (x , x; A) = GF (x − x) + e d4 x1 GF (x0 − x1 )A(x
Z
2
/ 1 )GF (x1 − x2 )A(x
/ 2 )GF (x2 − x) + . . .
+e
d4 x1 d4 x2 GF (x0 − x1 )A(x
Integral equation for the full wave function is:
Z
/
ψ(x) = ψ0 (x) + d4 yGF (x − y)A(y)ψ(y)
.
The S-matrix is given by:
Sfi = lim hψ0f |ψi i =
t→±∞
Z
lim hψ0f (x)|ψ0i (x)i + hψ0f (x)|
t→±∞
/
d4 yGF (x − y)eA(y)|ψ
i (y)i ,
yielding a perturbative expansion in powers of the elementary
charge e
Feynman rules for QED processes
Goal: formulation of a perturbative approach: writing down rules for
perturbation theory: Feynman rules – without rigorous derivation
Exemplified on the following processes:
Coulomb scattering of an electron
Electron-proton scattering
Electron-positron scattering
Coulomb scattering of an electron
For electrons (with positive energy):
Sfi = δfi − i
∞
X
en
Z
d 4 y1 . . .
Z
d 4 yn
n=1
(+E)
/ n )GF (yn − yn−1 )A(y
/ n−1 ) . . . GF (y2 − y1 )A(y
/ 1 )ψu(+E) (y1 ) .
ψ¯f
(yn )A(y
What are here the free wave functions? – Solutions of the free Dirac equation for an electron or
positron (r = ±1) with a well-defined four-momentum p:
ψ r = ω r (~p)e−ir px ,
with r=1,2,3,4 ;
~p
ω r (0) ,
ω r (~p) = ˆ
S −
E
with the Lorentz transformation matrix and the unit 4-spinors

r
~p
E+m
ˆ
S −
=
E
2m
1
 0

 pz
 E+m
p+
E+m
0
1
p−
E+m
−pz
E+m
pz
E+m
p+
E+m
1
0
p− 
E+m
−pz 
E+m 
 
 
1
0
0
 
 , ω 2 (0) = 1 , etc.
 , ω 1 (0) = 
0
0
0 
0
0
1
Solutions for incoming and outgoing e− in Coulomb scattering (normalization in volume V):
r
r
m
m
¯
ψi (x) =
u(pi , si )e−ipi x , ψ¯f (x) =
u(pf , sf )e+ipf x ;
Ei V
Ef V
with the notations u(p, s) = ω 1 (~p) ; u(p, −s) = ω 2 (~p). (Note: for positrons, the wave
functions would be v(p, s) = ω 3 (~p) ; v(p, −s) = ω 4 (~p).)
Coulomb potential: A0 (x) = A0 (~x) = −Ze/|~x|; ~A(~x) = ~0; thus the S-matrix element is
Z
m
1
1
(1)
¯
u(¯
pf , sf )γ0 u(pi , si ) d4 xei(pf −pi )x
Sfi = +iZe2 p
V Ei Ef
|~x|
The time component of the integral:
Z
dx0 e+i(Ef −Ei )x0 = 2πδ(Ef − Ei ) ,
expresses energy conservation. The 3-dimensional space-like integral with ~q = ~pf − ~pi :
Z
(1)
Sfi
= +iZe2
dx3
1 −i~q~x
4π
e
= 2 ;
~q
|~x|
so we obtain
1
m
4π
¯
p
u(¯
pf , sf )γ0 u(pi , si ) 2 2πδ(Ef − Ei ) .
V Ei Ef
|~q|
What is with 3-momentum conservation? – The infinitely heavy nucleus (source A0 ) does not
recoil, it can absorb an arbitrary amount of momentum from ~pi and change it into ~pf .
Electron-proton scattering
Previously: the nucleus was infinitely heavy, M = ∞; now we
describe the nucleus as a finite-mass particle (p+ ) with a wave
function. It is a spin- 21 particle just like the e− .
p+ current J µ (x): source of a 4-potential Aµ (x), which can be obtained solving Maxwell’s
equations; S-matrix as before:
Z
Sfi = −ie
/
d4 xψ¯f (x)A(x)ψ
i (x) .
Wave equation (D’Alembert equation) for the 4-potential of the proton current, written with
the quabla () operator:
Aµ (x) ≡ ∂ν ∂ ν Aµ (x) ≡
(This holds in the Lorentz gauge: ∂µ Aµ =
∂2
− ∇2
∂t2
∂
φ
∂t
Aµ (x) = 4πJ µ (x) .
− ∇~A = 0.)
Feynman Green’s function for this equation: DF (x − y)
DF (x − y) = 4πδ (4) (x − y) ,
(1)
in analogy to the e− e+ case before. We may also define the photon Green’s function in
momentum space (Fourier transform):
Z
DF (x − y) =
d4 q −iq(x−y)
e
DF (q) .
(2π)4
Writing on the RHS of Eq. (1) the Dirac-delta as δ (4) (x − y) =
Z
DF (x − y) =
d4 q
∂ν ∂ ν e−iq(x−y) DF (q) =
(2π)4
Z
R
d4 q −iq(x−y)
e
,
(2π)4
and with
d4 q
(−q2 )e−iq(x−y) DF (q) ,
(2π)4
we obtain the momentum representation of the photon propagator:
DF (q2 ) = −
4π
.
q2 + i
This is the Feynman (causal) Green’s function for a virtual photon. Since the photon is
chargeless, it is its own antiparticle, and we don’t have to care about backward-in-time
propagation.
The 4-potential induced by the 4-current of the proton can be written as
Aµ (x) =
Z
d4 yDF (x − y)J µ (y) .
Let us substitute this into the formula of the S-matrix:
(2)
Sfi
Z
= −i
d4 x eψ¯f (x)γµ ψi (x) DF (x − y)J µ (y) .
Expression in the [. . . ]: 4-current of the e− ; in analogy we construct the 4-current for the p+ :
s
Jfiµ (y) = ep ψ¯fp (y)γ µ ψip (y) ,
with
ψip (y) =
M
u(Pi , Si )e−iPi y .
Eip V
Thus the S-matrix can be further calculated as
s
s
e2
m2
M2 ¯
u(pf , sf )γµ u(pi , si )
= +i 2
V
Ef Ei Efp Eip
Z
Z
Z
d4 q −iq(x−y) −i(pf −pi )x i(Pf −Pi )y −4π ¯
e
e
e
u(Pf , Sf )γ µ u(Pi , Si ) .
d4 x d4 y
(2π)4
q2 + i
(2)
Sfi
Here, the integrals can be calculated easily (delta functions); we get
(2)
Sfi
=
−ie2
(2π)4 δ (4) Pf − Pi + pf − pi
V2
× ¯
u(pf , sf )γµ u(pi , si )
s
m2
Ef Ei
s
M2
Efp Eip
4π
¯
u(Pf , Sf )γ µ u(Pi , Si ) .
(pf − pi )2 + i
The delta function δ (4) (Pf − Pi + pf − pi ) expresses the conservation of 4-momentum
in the process
Symmetric expression in the electronic and protonic current (as expected):
e− feels the field induced by the p+ current ⇔ p+ feels the field induced by the e−
current
Feynman rules
Motivated by the calculations to the previous processes, we can write down general rules, using
which any process can be calculated to arbitrary order (in principle).
S-matrix element for a process 1 + 2 → 10 + 20 + 30 + . . . N:
Sfi
=
i(2π)4 δ (4)
p1 + p2 −
N
X
k=1
The transition amplitude Mfi =
P
(n)
n
Mfi
!
p0k
Mfi
2
Y
j=1
s
N
Nj Y
2Ej V k=1
s
Nk0
2Ek0 V
.
can be constructed according to the following rules:
1
In order n, draw all Feynman diagrams with n vertices, containing the right number of
initial and final particles
2
For external lines, write down the following expressions as factors:
3
For internal lines:
4
For all vertices, the following factor has to be included:
5
6
The
contracted with the index of the internal or external photon line:
P indexµ µ isP
µν
µ γµ or
µ γµ DF
P
P
At all vertices, 4-momentum conservation holds: a factor δ (4) ( j pj − k pk ); and
R d4 p
integration has to be performed over the undetermined momentum variable p: (2π)
4
The amplitudes of all contributing diagrams have to be added coherently, with the phase
factors
-1 for incoming positrons;
-1 for fermion exchange graphs;
-1 for all closed e− /e+ loops.
The S matrix at higher orders
1
So far: calculations at the lowest orders in e or α. Since α ≈ 137
, this
is often a useful approximation. However, with a good theory, one
should be able to corrections of higher order (at least in principle).
Also, many experiments are more accurate than the leading-order
theory only.
At first, the small higher-order corrections turned out to be infinitely
large. Renormalization is necessary to arrive to finite results.
Electron-positron scattering at higher orders
Corrections of order α2 :
Example: e− –e+ scattering. At
leading order (∝ α):
The vacuum polarization (VP) correction in diagram (i):
MfiVP
=
4πi
+e4 ¯
u(p01 , s01 )(−iγ µ )u(p1 , s1 ) − 2
q + i
" Z
#
4
d k
i
i
× Tr
(−iγ
)
(−iγ
)
ν
µ
/k − /
(2π)4 /k − m + i
q − m + i
4πi
v(p2 , s2 )(−iγ ν )v(p02 , s02 )
× − 2
q + i
Here, "Tr" stands for calculating the trace of the matrix (i.e. sum over diagonal elements). It
comes from the cyclicity of the fermionic loop:
Tr(GγGγ) =
X
Gαβ γβδ Gδ γα
αβδ
Counting the powers of k:
R
d4 k
k2
⇒ quadratic divergence, ∼ k2 : infinite expression!
Physical interpretation: the possibility of creating a virtual electron-positron pair influences the
properties of the photon propagator
iD0Fµν (q) = iDFµν (q) + iDFµλ (q)
iΠλσ (q)
iDFσν (q) ,
4π
with the vacuum polarization tensor
iΠλσ (q)
= −e2
4π
Z
"
#
1
d4 k
1
Tr
γ
γ
σ
λ
/k − m + i /k − /
(2π)4
q − m + i
Mathematically ill-defined due to the ultraviolet divergence. How to force convergence?
introduce an upper limit K on the integral – what is the physical meaning of K? What
should be the value of K? Why should be the results depend on K?
introduce a smooth damping factor in the integrand, e.g.
still problems with the physical interpretation
K2
k2 +K 2
⇒ finite expressions, but
One possible way of properly regularizing the expressions: Pauli-Villars renormalization.
Basic idea: subtract an integrand of the same asymptotic k → ∞ behavior to make the integral
convergent
Z
Πµν (q)
=
¯ µν (q)
Π
=
Z
d4 k fµν (q, k, m)
d4 k
⇒
fµ,ν (q, k, m) +
N
X
!
Ci fµν (q, k, Mi )
i=1
¯ is convergent. At the end, the limit
The constants {Ci , Mi } are to be chosen such that Π
Mi → ∞ has to be taken. Physical observables cannot depend on the {Ci , Mi }.
So, the regularized polarization tensor has the form:
¯ µν (q)
Π
=
=
(
Tr γµ (/k + m)γν (/k − /
q + m)
d4 k
(2π)4 (k2 − m2 + i)((k − q)2 − m2 + i)
)
N
X
Tr γµ (/k + Mi )γν (/k − /
q + Mi )
= ...
+
Ci 2
k − Mi2 + i (k − q)2 − Mi2 + i
i=1
Z
d4 k
kµ (k − q)ν + kν (k − q)µ − gµν (k2 − qk − m2 )
+ reg. .
16πie2
4
2
2
2
2
(2π)
(k − m + i) ((k − q) − m + i)
4πie2
Z
After some long calculation, one obtains for the regularized VP tensor:
2
2
¯ µν (q) = (q2 gµν − qµ qν ) − e ln Λ + ΠR (q2 ) ,
Π
3π m2
with the average cut-off momentum Λ and the regularization term ΠR (q2 ):
N
X
Mi2
m2
≡
− ln
ΠR (q2 )
=
2e2
π
Ci ln
i=1
Λ2
m2
Z 1
and
q2
dββ(1 − β) ln 1 − β(1 − β) 2
m
0
e2 q2
1
1 q2
+
−
+ ... .
π m2 15
140 m2
≈
For small q ("weak" scattering), ΠR (q2 ) → 0, and the 2nd-order amplitude can be written in
terms of the 1st-order amplitude as
(2)
Mfi
(1)
= Z3 Mfi
,
with
Z3 = 1 −
e2
Λ2
ln
.
3π m2
Thus, the effective, physical charge of the electron is renormalized (modified) to eR =
with e being the bare charge. Only eR is observable, and it only weakly depends on Λ.
√
Z3 e,
Effect of VP on an atomic electron
Modification of the Coulomb potential A0 (~x) = −Ze/|~x| in momentum space:
A00 (~q) = A0 (~q) + ΠR (−~q2 )A0 (~q) ,
or in real space
A00 (~x)
Z
=
≈
d3 q 0
A (~q) =
(2π)3 0
A0 (~x) −
Z
d3 q i~q~x 1 + ΠR (−~q2 ) A0 (~q)
e
(2π)3
e2
4 (3)
Ze
∇2 A0 (~x) = −
− e(Zα)
δ (~x) ,
15πm2
|~x|
15m2
using ∇2 (1/|~x|) = −4πδ (3) (~x). Non-relativistic energy shift of a bound hydrodenic state ψnl
in first-order perturbation theory:
VP
∆Enl
= hψnl |e∆A0 |ψnl i = −α(Zα)
4m
4
|ψnl (0)|2 = −
α(Zα)4 δl0 .
15m2
15πn3
Value for 2s electrons:
VP
∆E2s
= −1.122 × 10−7 eV = −27.1 MHz ↔ contradicts the experiment!