Neutrino-Lepton Masses, Zee Scalars and Muon g − 2

Duane A. Dicus 1 , Hong-Jian He 1 , John N. Ng 2,3

1
The University of Texas at Austin, Austin, Texas 78712, USA
2
TRIUMF, 4004 Wesbrook Mall, Vancouver, BC V6T 2A3, Canada
3
National Center for Theoretical Science, P.O. Box 2-131, Hsinchu, Taiwan
Evidence for neutrino oscillations is pointing to the existence of tiny but finite neutrino masses. Such
masses may be naturally generated via radiative corrections in models such as the Zee model where
a singlet Zee-scalar plays a key role. We minimally extend the Zee model by including a right-
handed singlet neutrino νR . The radiative Zee-mechanism can be protected by a simple U (1)X
symmetry involving only the νR and a Zee-scalar. We further construct a class of models with a
single horizontal U (1)FN (à la Froggatt-Nielsen) such that the mass patterns of the neutrinos and
leptons are naturally explained. We then analyze the muon anomalous magnetic moment (gµ − 2)
and the flavor changing µ → eγ decay. The νR interaction in our minimal extension is found to
arXiv:hep-ph/0103126v2 17 Jul 2001

induce the BNL gµ − 2 anomaly, with a light charged Zee-scalar of mass 100 − 300 GeV.
PACS numbers: 14.60.Pq, 12.15.Ff, 13.15.+g, 13.40.Em [ hep-ph/0103126 ]

The standard model (SM) of the electroweak and by forbidding the mixings between the νR and the ac-
strong interactions has to be extended, in light of the ex- tive neutrinos. Such an extension is nontrivial since the
isting neutrino oscillation data [1] which provides strong successful embedding of a singlet νR requires the addi-
evidence for tiny but finite neutrino masses. The squared tion of a second singlet Zee-scalar in the minimal exten-
neutrino-mass differences are found to have two distinct sion. Next we note that the original Zee model neither
ranges, 10−11 eV2 ≤ δm2⊙ ≤ 10−5 eV2 and δm2atm ≃ predicts the size of the Zee-scalar Yukawa couplings nor
10−3 eV2 , in order to interpret the solar and atmospheric provides any insight on generating the lepton masses and
neutrino anomalies, respectively. The very small neu- their hierarchy. Though our simplest U (1)X in Type-
trino masses may be explained by invoking the seesaw I models protects the radiative Zee-mechanism, we can
mechanism [2] where a heavy right-handed Majorana sin- use this same U (1)X group as a horizontal symmetry
glet neutrino is introduced at the grand unification scale involving both neutrinos and leptons (called Type-II),
(1010−16 GeV). On the other hand, the radiative mech- and thus explain the mass patterns of the neutrinos and
anism for neutrino mass generation, such as advocated leptons in a natural way, à la Froggatt-Nielsen [9]. In
in the Zee model [3], provides an important alternative, both Type-I and -II models, the νR can interact with the
where the relevant new physics is expected to show up at right-handed muon and Zee-scalar with a natural O(1)
or near the weak scale. Such low scale models, besides Yukawa coupling, which has striking phenomenological
being able to explain some features of neutrino oscilla- consequences.
tions such as the bi-maximal mixing [4], are clearly of Finally, we apply the Type-I and -II models to ana-
phenomenological relevance. The existing precision data lyze the Zee-scalar-induced contributions to the muon
have put some nontrivial constraints on the models [4–6] anomalous magnetic moment gµ − 2 and the lepton-
and further tests will be available at forthcoming collider flavor-violation decay µ → eγ . We find that the re-
experiments. cent BNL gµ − 2 anomaly [10] can be explained with
The minimal Zee model [3] contains the three active a light charged Zee-scalar of mass around 100 − 300 GeV.
left-handed neutrinos of the SM and a bilepton singlet Our models also have the µ → eγ decay branching ratio
Zee-scalar which plays a key role for radiative genera- around or below the current experimental limit.
tion of their Majorana masses. There is no underlying Minimal Extension of the Zee Model
reason that forbids the existence of light right-handed
singlet neutrinos (νR ), as νR can also be naturally con- with a Right-handed Singlet Neutrino
tained in various extensions of the SM (such as the mod- The minimal Zee model [3] introduces one extra singlet
els with left-right symmetry or SO(10)). The introduc- charged scalar (S1± ) together with the usual two-Higgs-
tion of νR thus provides the simplest possible extension doublet sector. By assuming no right-handed νR , as in
of Zee model. However, a simple embedding of νR results the SM, this scalar only interacts with left-handed neu-
in the loss of the prediction of neutrino masses as they trinos and leptons. Thus, the Zee model contains the
are rendered arbitrary by the tree-level Dirac mass terms following additional Lagrangian,
[7,8,5,6]. In this work, we first build a class of minimally X fjj ′
∆L1 = e a H b S + +h.c.
ǫab Lcaj Lbj ′ S1+ + m3 ǫab H
extended Zee models including νR (called Type-I) and in- 2 1 2 1
j,j ′
voke a simple U (1)X symmetry (or its discrete subgroup 

Zn ) to effectively protect the radiative Zee-mechanism = f12 νec µL − νµc eL + f13 (νec τL − ντc eL ) + (1)

1


f23 νµc τL −ντc µL + m3 (H10∗ H2− −H1− H20 ) S1+ +h.c. invariant. Other non-minimal variations of our Type-I
can be easily constructed.
where ℓj , ℓj ′ ∈ (e, µ, τ ) and Lj = (νj , ℓjL )T is the left-
handed doublet of the jth family. The (H1 , H2 ) are the Table 2. Summary of U (1) charges carried by the effective
usual two-Higgs-doublets with hypercharge (1/2, −1/2), operators in Type-I and -II models.
where H1 = (−H1+ , H10 )T , H2 = (H20 , H2− )T , and
He j = iτ2 H ∗ . The Yukawa sector can conserve total lep-
j
Operators U (1)X U (1)aFN U (1)bFN
ton number by assigning to S1± the lepton numbers ∓2. Lj ′ H1 ℓjR 0 yj yj − u j ′
Thus, the total lepton number is only softly violated by
e 2 ℓjR
Lj ′ H 0 yj −z yj −z −uj ′
the dimension-3 trilinear Higgs operator in Eq. (1). As
such, the small Majorana neutrino masses are radiatively e 1 νR
Lj ′ H x x′ x′ −uj ′
generated at one-loop and are automatically finite.
We minimally extend the Zee model by including a Lj ′ H2 νR x x′ + z x′ +z −uj ′
single right-handed Dirac neutrino νR with following νjc′ ℓjL S1+ 0 −z uj +uj ′ −z
Yukawa interactions,
νjc′ ℓjL S2+ −x −x−y uj +uj ′ − x−y
 c e + f νc µ + f νc τ
 +
∆L2 = f1 νR R 2 R R 3 R R S2 + h.c. (2) c ℓ S+
νR jR 1 x ′
x +yj −z x′ +yj −z
where S2± is a second singlet Zee-scalar. The nontrivial νR ℓjR S2+
c 0 x′ +yj −x−y x′ +yj −x−y
issue with embedding νR is to avoid arbitrary tree-level e 1 H2 S +
H 0 0 0
1
Dirac mass terms generated by the Yukawa interactions
Lj H2 νR and Lj He 1 νR (which mix νj and νR ), so that e 1 H2 S +
H −x z −x−y z −x−y
2
the predictive power of the radiative Zee-mechanism can T
νR νR 2x 2x ′
2x′
be effectively protected. We achieve this goal by not- + −
ing that the Yukawa sector (2) of νR possesses a global S1 S2 x x+y −z x+y −z
U (1)X symmetry, which can properly forbid the neutrino ǫab H1a H2b 0 z z
Dirac mass terms once a Zee-scalar S2± is included to-
gether with νR . It can be shown, by assigning the most
general U (1)X quantum numbers for the Zee-model with Neutrino Oscillations, Lepton Masses and
νR , that the unwanted tree-level neutrino Dirac masses Horizontal U (1)FN Symmetry
cannot be removed without S2± . We define our simplest
Type-I models with U (1)X in Table 1, where only νR and While the above Type-I models give the most economic
S2± carry U (1)X charges while all other fields are singlets embedding of νR with all the good features of the original
of U (1)X . Hence, the Type-I extension gives a truely Zee-model retained, they do not provide any insight on
minimal embedding of νR into the Zee model. two important issues: (i) There is no theory prediction on
the size of the Zee-scalar Yukawa couplings fjj ′ in Eq. (1),
Table 1. Quantum number assignments for Type-I and -II but the neutrino oscillation data requires the following
models. The hypercharge is defined as Y = Q − I3 . hierarchy [4]:
√
f12 m2τ f13 2δm2atm
Lj ℓjR νR H1 H2 S1+ S2+ S0 ≃ 2 ≃ 3 × 10 , 2
≃ ≃ 102 or 107 ,
f13 mµ f23 δm2⊙ (3)
U (1)Y −1/2 −1 0 1/2 −1/2 1 1 0
where f13 /f23 ≃ 102 (107 ) corresponds to the MSW large
U (1)X 0 0 x 0 0 0 −x − angle solution (vacuum oscillation solution). (ii) The
small lepton masses and their large hierarchy are not un-
U (1)aFN 0 yj x′ 0 z −z −x−y −1
derstood. Our goal is to construct this same U (1) group
U (1)bFN uj yj x′ 0 z −z −x−y −1 as a horizontal symmetry involving all the leptons so that
these two issues can be naturally explained à la Froggatt-
Nielsen (FN) [9]. [This U (1) will be called U (1)FN .]
Table 2 classifies all (dis-)allowed operators of Type- The basic idea is to consider a horizontal U (1)FN sponta-
I up to dimension-4. It shows that, as long as x 6= 0, neously broken by the vacuum expectation value hS 0 i of
the radiative Zee-mechanism is protected and the νR a singlet scalar S 0 . We can assign U (1)FN charges for rel-
remains massless. Such a massless νR does not con- evant fields such that different mass terms are suppressed
tribute to the invisible Z-width as it carries no weak by different powers of ǫ ≡ hS 0 i/Λ where Λ is the scale
charge. A special case of our Type-I is to consider its at which the U (1)FN breaking is mediated to the light
discrete subgroup Z4 under which νR and S2± transform fermions. For instance, a low energy effective operator
as, νR → iνR , S2± → ∓iS2± , while all other fields remain carrying a net U (1)FN charge q (either ≥ 0 or < 0) will

2
be suppressed by ǫ|q| . Though all mass terms are now where the allowed values of z are defined in Table 3.
allowed in the effective theory, we will build a class of Thus, Type-IIa suppresses fjj ′ couplings to O(10−2 −
FN-type models (called Type-II) in which the arbitrary 10−4 ). The models in Type-IIb1 (-IIb2), however, nicely
tree-level neutrino Dirac-mass terms are suppressed to accommodate the hierarchy (3) for the MSW large an-
a level much below the one-loop radiative Zee-masses, gle solution (vacuum oscillation solution), while the pre-
and thus the predictive power of the Zee-mechanism re- dicted size of f12 ∼ 10−3 − 10−6 is also of the right order
mains. The role of the FN-scalar S 0 is to provide the [4]. Finally, it is trivial to extend these models with more
spontaneous U (1)FN breaking and generate the relevant than one singlet νR (i.e., νRj with j = 1, · · · , NνR and
U (1)FN -invariant effective operators that will give the de- NνR = 3 for instance), by simply defining them to share
sired neutrino Yukawa couplings and lepton masses at the the same U (1) charges as in Tables 1 and 3.
weak scale. The heavy S 0 will be eventually integrated
Table 3. U (1)FN quantum number assignments for Type-IIa,
out from the low energy theory and our relevant particle
-IIb1 and -IIb2 models. [The defined range of z is |z| ∼ 3 for
spectrum of Type-II is the same as Type-I.
Type-IIa and 0 . z . 3 for Type-IIb1 and -IIb2.]
We provide two typical Type-II constructions, called
Type-IIa and -IIb, respectively. The Type-IIa is the sim-
plest extension of Type-I by further involving only the L1 L2 L3 eR µR τR νR H1 H2 S1+ S2+
right-handed weak-singlet leptons in the U (1)FN (cf. Ta- IIa 0 0 0 −5 −2 −1 x+1 0 z −z 1−x
ble 1). In the Type-IIb models, we further assign each
lepton doublet Lj a charge uj . So, the lepton masses are IIb1 −1 −3 −5 4 −1 −4 x 0 z −z 1−x
determined by ℓjR charges in Type-IIa, while Type-IIb
determines these masses by the charges of both ℓjR and IIb2 2 −5 −7 7 −3 −6 x 0 z −z 3−x
Lj . The low energy effective operators up to dimension-4
(with the heavy S 0 integrated out) are classified in Ta-
ble 2, from which we derive the general conditions for Zee Scalars, Muon g − 2 and µ → eγ
protecting the Zee-mechanism in Type-II,
The above minimally extended Zee-type models eco-
′
10 > |x | ∼ |x| ≫ 1, and |x − uj |, |x + y| ≫ 1, (4) nomically incorporate the νR and naturally explain the
mass patterns of the neutrinos and leptons. The Zee-
with xx′ > 0 and |y|, |z| ∼ O(1). For the explicit anal- scalar Yukawa couplings with the neutrinos/leptons also
ysis below, we choose a typical value of the suppression exhibit an interesting spectrum. Now we are ready to an-
factor ǫ ≃ 0.1. Thus, choosing leptons in mass-eigenbasis, alyze their phenomenological impact. The Brookhaven
we write their mass ratios as, E821 collaboration has announced a 2.6 standard devi-
ation in the muon anomalous magnetic moment aµ =
me : mµ : mτ ≃ ǫ 4 : ǫ 1 : ǫ 0 , (5) −10
(gµ −2)/2, i.e., ∆aµ ≡ aExp SM
µ −aµ = (42.6 ± 16.5) × 10
which require, [10], which gives a 90% C.L. range for new physics,

(y1 −u1 )−(y3 −u3 ) = ±4, (y2 −u2 )−(y3 −u3 ) = ±1. (6) 15.5 × 10−10 ≤ ∆aµ ≤ 69.7 × 10−10 . (9)

The tau Yukawa coupling itself can be estimated as Different authors [11] have interpreted this anomaly in
yτ ≃ (mτ /mt ) tanβ ≃ 10−2 tanβ ∼ ǫ1 (with tanβ = terms of supersymmetry, muon compositeness, extra Z ′ ,
hH2 i/hH1 i), in the typical range of tanβ ≃ 10 − 40, and leptoquarks and extended neutrino models. We attempt
this restricts the U (1)FN charges of τ as y3 − u3 = ±1. to explain it from the contribution of the Zee-scalars and
Table 3 summarizes three explicit realizations of Type-II the singlet νR in our minimal Type-(I, II) models.
models. From Table 3 and Eq. (2), the Yukawa couplings The Zee-scalars S1± and S2± in Type-I/II contribute to
of νR are predicted as, gµ − 2 via the Yukawa couplings f12,23 with (µL , νe,τ )
and f2 with (µR , νR ), respectively. Thus, we have,
Type IIa : (f1 , f2 , f3 ) ∼ (ǫ3 , 1, ǫ1 ) ; !
m2µ |f12 |2 + |f23 |2 |f2 |2
Type IIb1 : (f1 , f2 , f3 ) ∼ (ǫ5 , 1, ǫ3 ) ; (7) ∆aµ = 2 + 2
10 3
96π 2 M1 M S2
Type IIb2 : (f1 , f2 , f3 ) ∼ (ǫ , 1, ǫ ) .  2
100 GeV
From Table 3 and Eq. (1), we further predict the left- ≃ 11.8 × 10−10 × |f2 |2 , (10)
M S2
handed Yukawa couplings fjj ′ ,
2
with M 1 = (cos2 φ/MS2′ + sin2 φ/MH 2
′± )
−2
. Here (MS1′ ,
Type IIa : (f12 , f13 , f23 ) ∼ ǫ|z| ; 1
MH ′± ) are the mass-eigenvalues of the two charged
Type IIb1 : (f12 , f13 , f23 ) ∼ (ǫ4+z , ǫ6+z , ǫ8+z ) ; (8) scalars in Eq. (1) and φ is their mixing angle. Our mod-
Type IIb2 : (f12 , f13 , f23 ) ∼ (ǫ3+z , ǫ5+z , ǫ12+z ) ; els forbid or highly suppress the mixing between S1±

3
and S2± [cf. Table 2 and Eq. (4)]. Note that the f12,23 lies slightly above the limit by a factor of 2 − 3; given the
terms in (10) are negligible compared to the f2 term uncertainty of the parameters, it can be easily adjusted
for Type-I and Type-II (with |z| ≥ 1), cf. Eqs. (7)-(8). to stay within the bound. Also, a much weaker bound
The precision bound from µ → νµ eν e decay gives [4], on f3 can be derived from τ → µγ decay, i.e., |f3 | .
f12 /M1 < 0.18/TeV, which, combined with Eq. (3), also 0.06 − 0.16 ∼ O(0.1) at 90% C.L., for 1.1 . |f2 | . 3,
renders the f12,23 terms irrelevant for the gµ − 2 anomaly. which is consistent with the Type-II predictions in (7).
Hence, the original Zee-model cannot resolve the gµ − 2 Finally, if we include NνR (≥ 2) singlet νRj with the same
anomaly. From (9) and (10), we deduce, Yukawa coupling f2 , the upper [lower] bound in Eqs. p (11)
M S2 and (16) [Eq. (12)] will be relaxed by a factor of NνR .
41.1 GeV ≤ ≤ 87.2 GeV . (11)
|f2 | In summary, the Zee model naturally generates small
neutrino Majorana masses by radiative corrections, but it
Since the LEP2 direct search for charged particles re- neither predicts the Zee-scalar Yukawa couplings nor pro-
quires MS2 & 100 GeV, |f2 | is constrained as,
vides any insight on the lepton mass hierarchy. We have
1.1 . |f2 | . O(1) , (12) constructed a class of minimally extended Zee-models
with the right-handed neutrino νR embedded, where a
where the upper bound is imposed by perturbativity. U (1) symmetry protects the radiative neutrino masses
Eq. (12) is in the predicted range of f2 for our Type- while generating the lepton mass hierarchy, the hierarchy
II models [cf. Eq. (7)]. Thus, combining (11) and (12), of the Zee-scalar Yukawa couplings required by the neu-
we conclude that, to fully accommodate the BNL gµ − 2 trino oscillation data, the hierarchy of Zee-scalar Yukawa
data, the Zee-scalar S2± in our Type-I and -II models has couplings necessary for consistency with the µ → eγ
to be generically light, around 100 − 300 GeV. This leads bound, and the size of the Zee-scalar Yukawa coupling
to the possibility of discovering the light charged Zee- needed for the BNL gµ − 2 anomaly. Furthermore, a
scalar at the Tevatron Run-2, the LHC or a future high light Zee scalar S2± is predicted in our models, with a
energy Linear Collider. Since f2 ≫ f1,3 for Type-II, the mass around 100 − 300 GeV.
S2± → µ±R νR decay has a large branching ratio, and dom-
inates over the e± ± This research was supported in part by the U.S. De-
R νR and τR νR modes. Though Type-I
models do not predict f1,3 , they allow f2 = O(1) while partment of Energy under Contract No. DE-FG03-93ER
f1,3 ≪ f2 is forced by the µ → eγ and τ → µγ bounds 40757 and by the Natural Science and Engineering Re-
below. We have the partial decay width, search Council of Canada. We thank Sasha Kopp and
f2 Ernest Ma for discussions.
Γ[S2± → µ± R νR ] ≃ MS2 ≃ O(1) GeV, (13)
16π 2
for MS2 ≃ 100 − 300 GeV. Hence, S2± is a very nar-
row spin-0 resonance. The predicted branching ratio [1] E.g., S. Fukuda, et al., Phys. Rev. Lett. 85, 3999 (2000).
Br[S2± → µ± ± [2] M. Gell-Mann, P. Ramond and R. Slansky, in proceed-
R νR ] ≃ 1 in Type-(I, II) suggests that S2
can be best detected via muon plus missing energy. ings of the Workshop on Supergravity, p.315, 1979; T.
Yanagida, proceedings of the Workshop on Unified The-
Our models have further implications for the flavor-
ories and Baryon Number in the Universe, p.79, 1979.
violating rare decay µ → eγ, whose branching ratio is [3] A. Zee, Phys. Lett. B93, 389 (1980); B161, 141 (1985).
bounded by, Br[µ → eγ] < 1.2 × 10−11 , at the 90% C.L. [4] C. Jarlskog, M. Matsuda, S. Skadhauge, M. Tanimoto,
The Type-(I, II) models give the following contributions, Lett. B449, 240 (1999); P. H. Frampton and S. L.
! Glashow, Phys. Lett. B461, 95 (1999).
αem v 4 |f1 f2 |2 |f13 f23 |2 [5] A. Yu. Smirnov and M. Tanimoto, Phys. Rev. D55, 1665
Br[µ → eγ] = + (1997).
384π MS42 M1
4
[6] G. C. McLaughlin and J. N. Ng, Phys. Lett. B455, 224
αem 4 (1999); B464, 232 (1999).
≃ |f1 f2 |2 (v/MS2 ) , (14)
384π [7] W. Wolfenstein, Nucl. Phys. B173, 93 (1980).
√ [8] M. Fukugita and T. Yanagida, Phys. Rev. Lett. 58, 1807
where v = ( 2GF )−1/2 ≃ 246 GeV. Thus, we derive, (1987).
|f1 f2 | < 2.3 × 10−4 (MS2 /100 GeV)2 . (15) [9] C. D. Froggatt and H. B. Nielsen, Nucl. Phys. B147, 277
(1979).
Combining this with the gµ − 2 bound in (11), we find, [10] H. N. Brown, et al., Phys. Rev. Lett. 86, 2227 (2001).
[11] E.g., A. Czarnecki and W. J. Marciano, Phys. Rev. D
|f1 /f2 | < 1.8 × 10−4 , ⇒ f1 . (2 − 6) × 10−4 . (16) 64, 013014 (2001); K. Lane, hep-ph/0102131; L. Ev-
erett, et al., hep-ph/0102145; J. L. Feng, et al., hep-
Comparing this with Eq. (7), we see that Type-IIb1 has ph/0102146; D. Chakraverty, et al., hep-ph/0102180;
f1 just below the current bound while Type-IIb2 is well T. Huang, et al., hep-ph/0102193; E. Ma and M.
below it. On the other hand, the f1 coupling in Type-IIa Raidal, hep-ph/0102255; Z. Z. Xing, hep-ph/0102304.