6th February 2006 15:43 Proceedings Trim Size: 9in x 6in novikov˙gb75

PHOTON-NEUTRINO INTERACTION
OR
OPTICAL ACTIVITY OF INTERGALACTIC SPACE

VICTOR NOVIKOV
ITEP, Moscow, Russia

The birefringence of a neutrino sea is discussed in the Standard Model. We demon-

strate that the optical activity of a neutrino sea in the SM is dominated by the
contribution of the two-loop diagrams that are five order of magnitudes larger than
one-loop diagrams.

Vladimir Naumovich was fond of the Feynman diagram approach. For

him diagrams were not something like boring perturbative terms. Behind
the diagrams he saw a true space-time picture of the process. Diagrams
represented the bird’s-eye-view of the landscape of physics.
I am going to talk about a short paper where the clear understanding of
the space-time picture helps to find an enhancement factor of the order of
million. Though physical effects remain small even with this huge factor,
the theoretical observation seems to be rather nice. I believe that B.H.
would appreciate it.
My talk is based on a paper written in collaboration with G. Karl 1 .

1. Introduction. Preliminaries
The photon-neutrino interaction is very feeble, involving only electrically
neutral external particles. At a classical level this interaction vanishes (we
assume that the neutrino magnetic moment µν is zero). But in QFT the
corresponding amplitudes are non-zero even for µν = 0 due to virtual states
with charged particles. Thus in Fermi Theory the amplitude is described
by a triangle graph, with an electron running in the triangular loop (see
Fig. 1).
The cross sections are exceedingly tiny. Therefore this interaction can
only be of astrophysical interest. One of the possible applications involves

1
6th February 2006 15:43 Proceedings Trim Size: 9in x 6in novikov˙gb75

Figure 1. γν scattering in the four-fermion Model.

the propagation of light waves through a handed neutrino sea 2 which results
in the optical activity (birefringence) of the intergalactic space.

1.1. Prehistory
Optical activity was known from the time of Napoleon’s wars when Arago
(1811) and Biot (1812) discovered that polarization plane of linearly po-
larized light rotates when light propagates through active media. Later
(1817) Fresnel understood that this effect is connected with the fact that
eigenstates of propagation in media are the states with definite circular po-
larization and that the speed of light is slightly different for the left-handed
and right-handed polarizations of light, i.e. media are birefringent.

1.2. Modern History

In the last decade a number of authors raised a question on the possibility
of the violation of Lorentz and CP T invariance. To get a taste of this
activity, one can look at the paper written by Glashow and Coleman 3 .
The most interesting scenario is that the speed of light in vacuum (i.e. in
intergalactic space, in practice) is slightly different for different polarization
states of light, i.e. now the vacuum is birefringent. It is clear that such an
effect would indicate a small violation of Lorentz and CP T invariance at
the fundamental level.
There are not many choices to get Lorenz violation at the level of a
Lagrangian. If we wish to preserve gauge invariance and renormalizability
of QED there is more or less one way to modify the QED Lagrangian. That
is
1 1
L = − Fµν Fµν + ∆L, ∆L = µναβ bµ Aν Fαβ , (1)
4 2
6th February 2006 15:43 Proceedings Trim Size: 9in x 6in novikov˙gb75

where Aµ , Fαβ are the potential and the field-strength tensor for the elec-
tromagnetic field, and bµ is some new field. The condensation of the vector
field bµ in vacuum, i.e. a constant vector field in eq. (1), generates a viola-
tion of Lorentz and CP T symmetry.
One can check that in the ”rest” frame where bµ = (b0 , 0, 0, 0) the index
of vacuum refraction for left and right circularly polarized light n+,− is
different from unity and is different for different polarization states:

n+ = 1 + b0 /ω, n− = 1 − b0 /ω . (2)

Here ω is the light frequency.

For a linear polarized light this means that the plane of polarization
rotates at angle ∆φ when light propagates the distance l

∆φ = ω(n+ − n− )l . (3)

Thus the rotation power Φ = ∆φ(l)/l is

Φ = ω(n+ − n− ) = 2b0 . (4)

Non-zero b0 (vacuum condensation of vector field bµ ) leads to a birefrin-

gence of vacua. That is how people work with vacuum birefringence at a
fundamental level.
A less radical (and more solid) physical reason for the light to have
different speeds of propagation in the intergalactic space for different po-
larization states is a P -odd interaction of light with neutrino sea. In this
case the neutrino sea in the intergalactic space plays the role of optical
active media, i.e. of a left-handed sugar.
The idea that the intergalactic space is a birefringent medium for the
light due to the presence of a neutrino sea has been contemplated for a long
time. Some thirty years ago, Royer 2 estimated in a V − A theory an effect
of the order of GF

Φ ∼ αGF kF3 , (5)

where kF is the Fermi momentum of the neutrino sea. Later, Stodolsky a

noted that due to a theorem of Gell-Mann 5 there can be no such effect
with massless neutrinos and a point-like coupling, i.e.

Φ ∼ 0. (6)

a This observation was recorded in a review 4 .

6th February 2006 15:43 Proceedings Trim Size: 9in x 6in novikov˙gb75

In the early 1980’s data on propagation of radio waves through the

intergalactic space put an upper bound on the possible optical activity of
the neutrino sea 6,7 .
More recently an evaluation was made for massless neutrino and for
on-shell photons 8 in one-loop approximation within the Standard Model
(SM). The effect is non-zero but extremely small
(kF ω)2
Φ ∼ αGF kF3 . (7)
m4W
Here mW is the mass of the W boson. Due to the Gell-Mann theorem there
is an additional suppression factor of the order (kF ω)2 /m4W to the naive
estimate of Royer b .
In my talk I demonstrate that these one-loop estimates for on-shell
photon-neutrino scattering are irrelevant. The main contribution to the
optical activity comes from the two-loop amplitude. The latter is larger
than the one-loop estimates by a factor of 105 or more. There is simple
physics behind this amazing result.

2. Photon-neutrino interaction
2.1. Gell-Mann theorem
According to Gell-Mann’s theorem 5 , point-like weak interactions and mass-
less neutrino leads to vanishing amplitude for photon-neutrino effective in-
teractions. The theorem is easy to understand in the cross channel, i.e. for
the annihilation of a ν ν̄ pair into γγ in the center of mass frame.
For point-like interaction the orbital moment of a pair of neutrinos is
exactly zero L=0. Thus for massless left-handed neutrino and right-handed
antineutrino the total angular momentum J=1. On the other hand, for two
on-shell photons the states with J=1 are forbidden by Landau’s theorem 10 .
As a result, the transition between ν ν̄ and γγ states is forbidden in a point-
like limit.
To escape Gell-Mann’s restriction, one needs non-local interactions in
order to include higher orbital momenta of the ν ν̄ pair into the annihilation
process. In the Standard Model, where the W and Z bosons mediate the
interaction, the probability amplitude for a neutrino pair to have a nonzero
orbital moment is proportional to some power of a small factor ∼ p/mW ,

b See also the calculations for off-mass shell photons in Ref. 9 .

6th February 2006 15:43 Proceedings Trim Size: 9in x 6in novikov˙gb75

where p is the neutrino momentum. The factor 1/mW measures the shortest
separation of two neutrinos during their interaction (non-locality) in one-
loop approximation. The loop changes from a triangle to a square, with
the weak boson providing one of the sides (see Fig. 2).

 
   
 







 


Figure 2. γν scattering in the Standard Model.

2.2. Effective Lagrangian Approach

The standard way to deal with low-energy scattering is to use the Effective
Lagrangian approach.
The simplest example of the Effective Lagrangian is the four-fermion
interactions of neutrinos ν with electrons e. For small fermion momenta
one can forget about degrees of freedom associated with W and Z boson
and can write the Effective Lagrangian only for fermionic fields:
GF
Lef f = √ (ν̄γα ν)(ēΓα e) , (8)
2

where Γα = gV γα + gA γα γ5 . In the SM gV = 32 − 2 sin2 θW , gA = 32 . For

momenta smaller than mW (or mZ ) the effective Lagrangian is as good as
the fundamental Lagrangian of the SM .
Consider now the process ν(p) + γ(k) → ν(p) + γ(k). The neutrino
has no electric charge and the interaction between neutrino and photons
takes place only at the level of loop diagrams. They are numerous and
a bit complicated. For small momenta (pk/m2W )  1 one can expand
ν(p) + γ(k) → ν(p) + γ(k) amplitudes in the power series in this small
parameter. The lowest terms of this expansion can be represented as a
matrix element of the appropriate operators, i.e. of the particular terms of
the Effective Lagrangian.
6th February 2006 15:43 Proceedings Trim Size: 9in x 6in novikov˙gb75

Each term of the Effective Lagrangian has to be a Lorenz-invariant

combination of the gauge-invariant electromagnetic field tensor Fµν and
the left-handed neutrino field νL = 21 (1 + γ5 )ν and their derivatives. The
operator may have high dimensions D. To preserve the correct dimension
of the Lagrangian [L] ∼ [m]4 the coefficients in front of this operator should
be proportional to the appropriate power of (1/m), where m is the scale of
mass that moves inside the loops. The actual calculation of the diagrams
gives the numerical coefficients.

2.3. P-even scattering amplitude

Consider how all this works in the case of P -even νγ-scattering. In this case
the amplitude should be the same for right-handed photons and left-handed
photons. One has to construct appropriate Lorenz invariant operators from
the fields Fµν and νL . The simplest combination of the fields that satisfies
all these conditions looks as follows:
e4
Lef f ∼ [Fµα Fµβ ]ν̄γα ∂β (1 + γ5 )ν + h.c. . (9)
m4
It has the dimension D = 8. The matrix element of Lef f for forward
scattering gives the amplitude
e4
T ∼ (pk)2 (k)∗ (k) . (10)
m4
If we identify the parameter m in eq. (9) with the largest mass in the dia-
grams (i.e. with mW ), we reproduce the well-known result for the neutrino-
photon amplitude 11 up to the numerical constant
(pk)2
T ∼ GF α (k)∗ (k) . (11)
m2W
Of course it is important to calculate numerical coefficients. But even before
any calculations from this simple exercise we have learned two lessons.
First, the non-zero term for γν scattering appears only in the second
order in (pk)2 . It indicates immediately that any calculations that give
an effect of the order (GF α) ∼ α2 /m2W (such as in e.g. of Refs. 5,12 )
are erroneous. It is impossible to violate the Gell-Mann theorem in the
framework of the Effective Lagrangian approach. Appropriate operators
do not exist.
Second, the actual calculations for the Effective Lagrangian are sim-
plified enormously. Indeed now one expands every individual Feynman
diagram up to the order of (pk)2 . To calculate the numerical factor one can
6th February 2006 15:43 Proceedings Trim Size: 9in x 6in novikov˙gb75

put external momenta in remaining integrals to be equal zero, i.e. p = 0

and k = 0. For zero external momenta Feynman integration becomes trivial
and the whole problem reduces to a pure algebraic (though rather boring)
calculations of traces and products.

2.4. Optical activity. P-odd scattering amplitude

Now let us come back to P -odd effects in νγ scattering and find appropriate
operators in Lef f responsible for optical activity. The simplest Lagrangian
that depend on P -odd combinations of photons amplitudes looks as follows
1
Lef f ∼ [Fµα F̃µβ ][ν̄γα ∂β (1 + γ5 )ν] + h.c. , (12)
m4
where F̃µν = 12 µναβ Fαβ .
The surprise is that this operator does not work in our case. Indeed the
refraction index is proportional to the photon forward scattering ampli-
tude. One can check that the matrix element of Fµα F̃µβ between photons
with the same momenta and polarization (forward scattering) is identically
zero. Thus the operators of D = 8 do not contribute into P -odd forward
scattering and the effect is zero in (pk)2 /m4 order!
Now we have to look for operators of higher dimension D = 10. One of
these operators looks like
1
Lef f ∼
[Fµα (∂γ F̃µβ )][ν̄γα ∂β ∂γ (1 + γ5 )ν] + h.c. . (13)
m6
With this Lef f the forward scattering amplitude of a photon of momen-
tum k from a neutrino of momentum p is equal to
T = C(e4 /8π 2 )(pk/m2 )2 µναβ µ (k)∗ν (k)(pα kβ /m2 ) . (14)
This amplitude has different contributions to left-handed and to right-
handed photon scatterings: TLL = −TRR .

2.5. One-loop calculations of the P-odd effect

The actual calculation of the coefficient C has been done in one loop-
approximation in 8 with the results
T = C(e4 /8π 2 s2 )(pk/m2W )2 µναβ µ (k)∗ν (k)(pα kβ /m2W ) , (15)
where
C = 4/3(ln(m2W /m2 ) − 11/3) (16)
6th February 2006 15:43 Proceedings Trim Size: 9in x 6in novikov˙gb75

8
in the first reference in and
C = 4/3(ln(m2W /m2 ) − 8/3) (17)
8
in the second one . The reason for that discrepancy is unknown and
it would be interesting to understand whether there is a correct one-loop
result.
The message of this talk is that for a P -odd effect there is an enormous
enhancement factor in two-loop approximation. Thus any one-loop results
(correct or erroneous) are irrelevant.

3. Two-loop calculations
3.1. P-odd effect
The physical reason for the dominance of the two-loop diagrams over the
one-loop is simple.
To escape Gell-Mann’s restriction one needs non-local interactions in
order to include higher orbital momenta of the ν ν̄ pair into an annihila-
tion process. In the one loop approximation the only source of non-locality
is the W boson exchange. Thus the one-loop probability amplitude T (1)
for a neutrino pair to have nonzero orbital moment is ∼ p/mW . The fac-
tor 1/mW measures the shortest separation of two neutrinos in one-loop
approximation.
In the two-loop approximation there is a set of diagrams in which two
neutrinos are emitted at a separation of the order of the electron Compton
wavelength. This is due to e− e+ ν in the intermediate state (see Fig. 3).

Figure 3. The sample of two-loop contribution into γν scattering

Thus in two-loop approximation the amplitude T (2) for a neutrino pair

to have nonzero orbital moment is ∼ p/me . Moving to the next order
in electro-weak interaction we loose a small factor α/2π , but win a large
6th February 2006 15:43 Proceedings Trim Size: 9in x 6in novikov˙gb75

factor (m2W /m2e )! The net effect is

(T (2) /T (1) ) ∼ (α/2π)(m2W /m2e ) ∼ 107 . (18)
The actual calculation is rather lengthy. The result is
T (2) = (13/27)(G2 e2 /64π 4 )(gV2 + gA
2
)(pk)2 µναβ µ (k)∗ν (k)(pα kβ /m2e ) .
(19)
The enhancement factor is
T (2) /T (1) = (α/64πsin2 θW )(m2W /m2e )(13/27)(gV2 + gA
2
)/C ∼ 105 , (20)
where C is the one-loop coefficient from eq. (15). We have lost two orders
of magnitude compared with the naive estimate. This happened mainly
due to the large logarithmic factor in the one-loop coefficient C. Still, the
effect is very large ∼ 105 !

3.2. Two-loop estimates of P-even effect

It is interesting to understand whether a similar enhancement factor ap-
pears for the main P-even amplitude, e.g. for the cross section of ν ν̄ anni-
hilation into two photons. The answer is negative.
The arguments are the following. Consider two-loop amplitudes with
+ −
(e e ν) in the intermediate state between two external neutrino vertices.
In local four-fermion approximation we expect that these diagrams are pro-
portional to G2F α/m2e . Thus to preserve the correct dimension in an effec-
tive Lagrangian, we need an operator of dimension D = 10. The appropri-
ate effective Lagrangian is
G2F α
Lef f ∼ [Fµα (∂γ Fµβ )][ν̄γα ∂β ∂γ (1 + γ5 )ν] + h.c. (21)
m2e
For this Lef f the scattering amplitude is of the third order in (pk)
G2F α
T ∼C (pk)3 (k)∗ (k). (22)
m2e
Thus for P-even scattering second order loops give a correction of the
2
order (GF MW )(pk/m2e ), i.e. a small correction to the one-loop result.

4. Numerical estimates
In spite of a huge missing factor 105 in the one-loop estimates of optical
activity of the neutrino sea 13 we find that the physical effect is still tiny.
It is rather unlikely that direct measurements of such a small rotation of
6th February 2006 15:43 Proceedings Trim Size: 9in x 6in novikov˙gb75

10

the plane of polarization of light when it travels in the intergalactic space

are ever possible. But may be correct estimates of the P-odd effect give
hints for more subtle experiments. In any case it is nice to have a correct
estimate.

5. Summary
To conclude:
If there is a neutrino sea, the Universe is slightly birefringent.

6. Acknowledgments
I would like to thank the organizers of ”Gribov-75” Conference and espe-
cially Julia Nyı́ri, Yuri Dokshitzer, Peter Levai and Yulik Khriplovich for
invitation, for their hospitality and for the excellent conference.
This research was partly supported by RFBR grant 05-02-17203.

Bibliography
1. G. Karl and V. Novikov, Pis’ma v ZhETF 81, 313 (2005).
2. J. Royer, Phys. Rev. 174, 1719 (1968).
3. S. Coleman and S. Glashow, Phys. Rev. D59, 116008 (1999);
R. Jackiw and V.A. Kostelecky, Phys. Rev. Lett. 82, 3572 (1999) and refer-
ences therein.
4. G. Karl, Canadian Journal of Physics 54, 568 (1976).
5. M. Gell-Mann, Phys. Rev. Lett. 6, 70 (1961).
6. J.N. Clarke, P.P. Kronberg and M. Simard-Normandin, Mon. Not. Roy. 190,
205 (1980).
7. J.N. Clarke, G. Karl and P.J.S. Watson, Canadian Journal of Physics, 60,
1561 (1982).
8. G. Karl and V. Novikov, hep-ph/0009012;
A. Abbasabadi and W. Repko, Phys. Rev. D64, 113007 (2001); ibid. D67,
073018 (2003).
9. S. Mohanty, J.F. Nieves and P. Pal, Phys. Rev. D58, 093007 (1998) 093007.
10. L.D. Landau, Dokl. Akad. Nauk USSR 60, 207 (1948);
C.N. Yang, Phys. Rev. 77, 242 (1950).
11. M.J. Levine, Nuovo Cimento A48, 67 (1967).
12. S.G. Matinyan and N.N. Tsilosuni, Zurn. Eksp. Teor. Fiz. 41, 1691 (1961);
Sov. Phys. JETP 14, 1195 (1961).
13. A. Abbasabadi and W. Repko, Phys. Rev. D67, 073018 (2003).