Neutrino loops from neutrino mixing

Adam Latosiński1 , Krzysztof A. Meissner1,2 and Hermann Nicolai3

1
Faculty of Physics, University of Warsaw
Hoża 69, Warsaw, Poland
2
National Centre for Nuclear Research, Świerk, Poland
3
Max-Planck-Institut für Gravitationsphysik (Albert-Einstein-Institut)
Mühlenberg 1, D-14476 Potsdam, Germany

We show how neutrino mixing leads to a softening of the UV behavior of neutrino loops, and can
give rise to anomaly-like, but manifestly UV finite amplitudes. This mechanism may be of special
significance for the coupling of dark matter-like particles to the Standard Model, as we illustrate
with a one-loop example involving the coupling to W -bosons.
arXiv:1112.0134v1 [hep-ph] 1 Dec 2011

PACS numbers: 14.60.Pq,14.60.St,14.80.Va

In this letter we explain a mechanism specific to neu- terms (for the two-component formalism see e.g. [4]). In
trino loop diagrams in the minimally extended Stan- the SM these are supposed to arise via spontaneous sym-
dard Model (SM) with right-chiral neutrinos. This effect metry breaking, such that m = Yν hHi and M = YM hφi
not only improves the UV behavior of loop amplitudes with appropriate Yukawa couplings Yν and YM , and non-
but, more importantly, can lead to anomaly-like ampli- vanishing vacuum expectation values for the SM Higgs
tudes and effective couplings in processes involving dark field H(x) and a further complex (electroweak singlet)
matter-like particles. Although elementary (and used im- scalar field φ(x) [9]. We assume m and M real, as this
plicitly in our previous work [1]) this mechanism has not can be achieved by appropriate phase rotations.
been spelled out explicitly so far in the existing liter- The standard procedure to deal with (2) consists in
ature to the best of our knowledge. It relies crucially diagonalizing the mass matrix, introducing new fields
on neutrino mixing, and on the simultaneous presence of  ′   
Majorana- and Dirac-like mass and Yukawa terms for the ν cos θ − sin θ ν
= (2)
right-chiral fields. While reminiscent of Pauli-Villars reg- N′ sin θ cos θ N
ularization, it involves only real particles, and no ghosts
appear. Remarkably, this appears to be the only context in terms of which the kinetic terms become diagonal
where anomaly-like effects can arise from triangle dia- i  ′α 
grams without the linear UV divergences that one com- L = ν ∂/αβ̇ ν̄ ′β̇ + N ′α ∂
/αβ̇ N̄ ′β̇ + c.c − (3)
2
monly associates with anomalous amplitudes. m′ ′α ′
M′
ν να + ν̄α̇′ ν̄ ′α̇ − N ′α Nα′ + N̄α̇′ N̄ ′α̇


The relevant diagrams typically involve amplitudes −
2 2
with ‘dark matter’ (axion-like) particles on one side of the
diagram, and ‘visible’ SM particles on the other. Our re- A simple calculation gives
sults pertain in particular to scenarios where the ‘heavy’
neutrinos are not so heavy after all [1, 2], in contradis- 2m
tan 2θ = (4)
tinction to the more common see-saw scenarios with Ma- M
jorana mass scales of O(1010 GeV) or more [3]. For these and the mass eigenvalues m′ and M ′
we arrive at amplitudes which are not only UV and IR
finite, but also can explain in a perfectly natural manner sin2 θ cos2 θ
the extreme smallness of dark matter couplings to SM m′ = − M, M′ = M (5)
cos 2θ cos 2θ
particles, and this without any fine-tuning or the need to
adduce the kind of ‘shadow matter’ invoked by numerous Notice that m′ M ′ < 0, so m′ and M ′ always come with
other currently popular scenarios. opposite signs. The propagators then take the standard
For simplicity, we consider a one family model with diagonal form,
right-chiral neutrinos. Using Weyl spinors to express the
d4 p p/αβ̇
Z
4-component neutrino spinor N ≡ (NL , NR ) ≡ (να , N̄ α̇ ) ′ ′
hνα (x)ν̄β̇ (y)i = i e−ip (x−y) (6)
4 2
(2π) p − m′2
and its conjugate in terms of SL(2, C) spinors, the neu-
trino part of the kinetic Lagrangian reads d4 p m′ δαβ −ip (x−y)
Z
hνα′ (x)ν ′β (y)i = i e (7)
i  α  (2π)4 p2 − m′2
L = ν ∂/αβ̇ ν̄ β̇ + N α ∂
/αβ̇ N̄ β̇ + c.c. − (1)
2 with analogous expressions for the N ′ propagators af-
∗
M α M
−mν α Nα − m∗ ν̄α̇ N̄ α̇ − N Nα − N̄α̇ N̄ α̇ ter replacing m′ → M ′ . With the redefinitions (2) the
2 2 SM interaction vertices involving neutrinos are now off-
where we have included both Dirac and Majorana mass diagonal, and any Feynman diagram computation will
 2

involve a sum of contributions for each vertex. We will re- contains the following vertices (cf. for example [6])
fer to this description (diagonal propagators, off-diagonal
gW µ gW − α µ β̇
vertices) as the ‘propagation picture’, because it is more Lint ⊃ − √ Wµ+ eα β̇
L σαβ̇ ν̄ − √ Wµ ν σαβ̇ ēL +
natural if we ask about the eigenstates of propagation. 2 2
As shown in our previous work [5], however, there is iM
− √ a(N α Nα − N̄α̇ N̄ α̇ ) (12)
another description (which we refer to as the ‘vertex pic- 2 2hϕi
ture’) where the vertices remain diagonal, while instead
the propagators are off-diagonal. This picture is better where the Yukawa interaction arises from a Majoron-like
adapted to the fact that the weak interactions in the SM coupling 21 YM (φN α Nα + c.c.) [7] upon expanding
involve only left-chiral particles. In that picture all di- √ i
agrams with both heavy and light neutrinos circulating


φ(x) = ϕ(x) exp ia(x)/ 2hϕi = hϕi + √ a(x) + · · · .
in the loops are manifestly UV finite, due to the unusual 2
fall-off properties of the propagator components for large (13)
momenta. More specifically, the propagators are deter- Consequently, the field a(x) carries the charge of a spon-
mined by inverting the kinetic matrix taneously broken global symmetry, and therefore couples
like a Goldstone boson (in [1] this symmetry is lepton
number symmetry). In this model, we now consider the
 
0 /αβ̇ mεαβ
i∂ 0
 ¯α̇β one-loop aW + W − amplitude, shown in Fig. 1a, which is
 i/∂ 0 0 −mεα̇β̇ 

A =  αβ (8) a sum of two diagrams with either νN or ν N̄ propagators
M εαβ i∂ /αβ̇ 

mε 0 in the loop in the ‘vertex picture’. It reads
0 −mεα̇β̇ i/∂¯α̇β −M εα̇β̇
−iMµν aW W (p, q) =
with the result (in momentum space) 2
gW M
Z
d4 k n µ
= √ σα β̇ hēβ̇L1 eα ν
L (k − p)iσα2 β̇2 ×
2

2 2hϕi (2π)4 1 1
hνα νβ i = iM m2 D(p)εαβ h
hνα ν̄β̇ i = i p2 − M 2 − m2 D(p)p/αβ̇ × hν̄ β̇2 N α3 (k)ihNα3 ν α1 (k + q)i +

hNα Nβ i = iM p2 D(p)εαβ
io
(9) − hν̄ β̇2 N̄α̇3 (k)ihN̄ α̇3 ν α1 (k + q)i (14)
hNα N̄β̇ i = i p2 − m2 D(p)p/αβ̇


ig 2 d4 k (k − p)λ
Z
hνα Nβ i = im p2 − m2 D(p)εαβ

= √W Tr{σ σ σ̄ µ σ λ σ̄ ν } ×
2 2hϕi (2π)4 (k − p)2 − m2e
hνα N̄β̇ i = −imM D(p)p/αβ̇ m2 M 2 [−kσ (2kq + q 2 ) + qσ (k 2 − m2 )]
−1 {(k 2 − m2 )2 − k 2 M 2 }{[(k + q)2 − m2 ]2 − (k + q)2 M 2 }
where D(p) := (p2 )2 − p2 (2m2 + M 2 ) + m4

. In this
description the finiteness of the neutrino loop diagrams is For small axion momentum q µ , it suffices to retain only
a consequence of the fall-off properties of the propagators the leading terms in q µ and m′ , so we get
hνα νβ i and hνα N̄β̇ i which both decay faster than 1/p2 . 2
gW m′ M ′3
Alternatively, this effect can be understood in the −iMµν
aW W (p, q) ≈ −iq
ρ
√ Tr{σ σ σ̄ µ σ λ σ̄ ν } ×
‘propagation picture’, where the propagators have the 2 2hϕi
usual fall-off properties. Namely, the cancellations are d4 k (k − p)λ (2kσ kρ − gσρ k 2 )
Z
now due to the fact that off-diagonal vertices give rise to × (15)
(2π)4 (k − p)2 − m2e [k 2 (k 2 − M ′2 )]2
several propagator contributions that must be summed
over for a given diagram, and thus follow from the for- where we have now substituted the mass eigenvalues m′
mula relating the ν, N and ν ′ , N ′ propagators via (2), and M ′ for m and M . Since the integrand decays as
  ∼ |k|−7 for large k 2 the integral is UV convergent. There
hνα N̄β̇ i = sin θ cos θ −hνα′ ν̄β̇′ i + hNα′ N̄β̇′ i (10) is no IR divergence either, which justifies taking the limit
m′ → 0 in the denominator.
hνα ν β i = cos2 θhνα′ ν ′β i + sin2 θhNα′ N ′β i (11) The similarities with the standard anomalous triangle
diagram are evident, but we here arrive at a perfectly
Cancellation of the leading terms for large |p| in (10) is finite expression, without the need of any regularization.
obvious, while cancellation in (11) is due to the relation Because
m′ cos2 θ + M ′ sin2 θ = 0, that is, happens thanks to the
opposite signs of m′ and M ′ , cf. (5). In both cases the Tr{σ σ σ̄ µ σ λ σ̄ ν } =
cancellation is thus due to the fact that there are two = 2 η σµ η λν − η σλ η µν + η σν η µλ − iεσµλν


propagator components contributing with opposite signs
in the loop, just like in Pauli-Villars regularization. the result contains both ‘parity-even’ and ‘parity-odd’
Let us illustrate this mechanism with a simple ex- contributions, reflecting the fact that parity is maximally
ample. The SM with right-chiral fermions and sponta- broken in the SM. In particular, we do get a contribution
neously broken SU (2)w × U (1)Y symmetry in particular ∝ εσµλν mimicking the anomaly.
 3

Using Feynman parameters, we can perform the inte- also appear the manifestly non-gauge invariant contri-
gral over momenta and get butions displayed above. By contrast, for higher loop
diagrams with external gluons of the type considered in
2
gW m′ M ′3 [5], only the anomaly-like interaction can survive, be-
−iMµν
aW W (p, q) ≈ q
ρ
√ ×
32π 2 2hϕi cause SU (3)c remains unbroken, and the Chern-Simons-
µ ∼
Z 1 Z 1 like coupling ∂µ a JCS = a Tr G̃G is the only possibility
Tr{σ σ σ̄ µ σ λ σ̄ ν } × dy dx x(1 − x)y 3 × to reconcile gauge invariance with the Goldstone prop-
0 0 erty of a(x). Similar comments apply to the coupling of

[(−1 + 2y)pλ gρσ + (1 − y)(pρ gσλ + pσ gρλ )] a(x) to photons, which comes out proportional to aF̃ F .
× +
[−y(1 − y)p2 + (1 − y)m2e + yxM ′2 ]2 This explains why in both cases the amplitude acquires
2y(1 − y)2 pλ (−p2 gρσ + 2pρ pσ ) an ‘axion-like’ form.

+ We conclude with three remarks. First of all, similar
[−y(1 − y)p2 + (1 − y)m2e + yxM ′2 ]3
results and conclusions are obtained for the analogous
For small p2 we obtain (αW ≡ gW
2
/4π, ηe ≡ me /M ) diagram with external Z-bosons shown in Fig. 1b, where
the triangle loop is ‘purely neutrino’. Secondly, the mech-
−iMµν aW W (p, q) ≈ (16) anism exhibited here persists in more realistic scenarios
α W m ′ with three families, as is most easily seen in the ‘ver-
≈ −q ρ √ Tr{σ σ σ̄ µ σ λ σ̄ ν } × tex picture’. In this case, we have correspondingly more
48π 2hϕiM ′
complicated mass matrices,
(1 − 5ηe2 ) ln ηe2 − (1 + 3ηe2 )(1 − ηe2 )

× pλ gρσ +
(1 − ηe2 )3 0 0 m εαβ 0
 

(1 + ηe2 ) ln ηe2 + 2(1 − ηe2 )  0 0 0 −m∗ εα̇β̇ 

+(pρ gσλ + pσ gρλ )
(1 − ηe2 )3
(the singularity for ηe = 1 is spurious). For large p2
αW m′ M ′ ln(p2 /M 2
 
µν
−iMaW W (p, q) ≈ √ 1+O
8π 2hϕi p2
2 
× 2 pµ q ν + pν q µ − η µν p · q − iεµνρλ pρ qλ

p
where we have written out the trace. The result contains
the manifestly non-gauge invariant pieces corresponding
to the first three terms in brackets, which are expected in
accordance with the fact that the electroweak SU (2)w ×
U (1)Y symmetry is broken [10]. In addition, there is an
anomaly-like contribution ∝ a εµνρσ ∂µ Wν+ ∂ρ Wσ− . The
full amplitude falls off linearly in pµ , whence this ‘form
factor’ will also improve the UV behavior of higher loop
diagrams when inserted as a subdiagram [5].
Because a(x) is a Goldstone boson, one momentum
factor can always be pulled out of the diagram. With
unbroken gauge invariance, this would only leave the
‘anomaly-like’, but in this case manifestly UV finite am-
plitude, whereas with external W or Z bosons there

M=
mT εαβ αβ
 (18)
0 Mε 0 
† α̇β̇ α̇β̇
0 −m ε 0 −M ε
where m is now a complex 3-by-3 matrix, and M a real

diagonal 3-by-3 matrix. With the usual see-saw assump-
tion of a large hierarchy between m and M the heavy
neutrino masses are given approximately by the eigenval-
(17) ues of M , while the squared mass eigenvalues of the light
neutrinos follow by diagonalizing the hermitean matrix
M −1/2 m† mM −1/2 (for a recent discussion of corrections
to these formulae see e.g. [8]).
Finally, in [1] we have proposed to actually identify
the field a(x) (usually referred to as the ‘Majoron’) with
the axion, in which case the smallness of various axion
couplings can be explained very naturally, and without
the need to invoke any large intermediate mass scales.
This proposal is based on the fact that the effective cou-
plings of a(x) to gluons and photons assume the standard
axionic form. In this Letter, we have provided a direct
argument explaining why this happens. We hope that
the new effect exhibited here will lend more credence to
this proposal, which in our opinion possesses attractive
features in comparison to other axion models.

[1] K.A. Meissner and H. Nicolai, Phys. Lett. B648 (2007)
312; Eur.Phys. J. C 57 (2008) 493;
[2] M. Shaposhnikov, arXiv:0708.3550 [hep-th] and refer-
ences therein.
[3] P. Minkowski, Phys. Lett. B67 (1977) 421; T. Yanagida,
Progress of Theoretical Physics 64 (1980) 11031105; M.
Gell-Mann, P. Ramond, and R. Slansky in Supergrav-
ity ed. by P. van Nieuwenhuizen and D. Z. Freedman,
North Holland, 1979; S. L. Glashow, in Proceedings of
the 1979 Cargèse Summer Institute on Quarks and Lep-
tons, Plenum Press 1980; R. N. Mohapatra and G. Sen-
janovic, Phys.Rev.Lett. 44 (1980) 912.
[4] J. Bagger and J. Wess, Supersymmetry and Supergravity,
Princeton University Press, 1984.
[5] A. Latosinski, K.A. Meissner, H. Nicolai, Axions without
Peccei-Quinn Symmetry, arXiv:1010.5417 [hep-ph]
[6] S. Pokorski Gauge Field Theories, Cambridge Univ.
Press, 2nd edition (2000).
 4

[7] Y. Chikashige, R.N. Mohapatra, R.D. Peccei,
Phys. Lett. 98 (1981) 265
[8] H. Hettmansperger, M. Lindner, W. Rodejohann, JHEP
1104 (2011) 123.
[9] Recall that the neutrino mass hierarchy can be imple-
mented by taking Yν ∼ O(10−5 ) [1, 2], that is, without
introducing a large intermediate mass scale by hand. In
this case, the heavy neutrinos typically assume mass val-
ues in a range below O(1 TeV).
[10] This link is somewhat indirect, but readers should keep in
mind that without spontaneous symmetry breaking this
diagram would be absent altogether.

p+q p+q
Wµ+ ν/N Zµ0 ν/N
k+q k+q
k−p e k−p ν
k q k q
Wν+ ν/N Zν0 ν/N
p p

Fig 1. a) a-W-W diagram b) a-Z-Z diagram