You are on page 1of 4


Flavor Changing Strings and Domain Walls

Gia Dvali and Goran Senjanovic*
International Centre for Theoretical Physics, Trieste, Italy
(Received 22 June 1993)
We consider the cosmological consequences of a spontaneous breaking of non-Abelian discrete sym-
metries, which may appear as a natural remnant of a continuous symmetry, such as a family symmetry.
The result may be a stable domain wall across which an electron would turn into a muon (or ve into v^)
or aflavoranalog of an Alice string-domain-wall structure with the same property.
PACS numbers: 98.80.Cq, 11.27.+d, 11.30.Qc

It is quite natural that the spontaneous breaking of a where 07 are Pauli matrices. Clearly, the eight-element
continuous symmetry leads at some stage to a discrete group Q belongs to SU(2) and equally clearly under Q,
one. If the original symmetry is non-Abelian, there is no mv is forbidden, while fiv is allowed. We focus here on
reason why the discrete subgroup should not be non- the lepton flavor version of this idea, where the basic dou-
Abelian as well. It is our aim in this paper to investigate blet representation of Q is (p and (^). To ensure that
the issue of topological defects in the case of a spontane- nie^m^ we need the scale of Q breaking AQ to satisfy
ously broken non-Abelian discrete symmetries, both as AQ ^ 10-100 MeV. On the other hand, the upper limit of
basic symmetries and also as a stage of symmetry break- AQ is also 10-100 MeV, otherwise the resulting domain
ing of some continuous symmetry. walls would carry too much energy density and distort the
In the case of pure non-Abelian discrete symmetry, we observed isotropy of the Universe [6].
find an interesting possibility of domain walls, light In other words, we predict
enough to be stable without spoiling the big-bang
scenario. They would have the spectacular property of A e = 10-100 MeV. (2)
changing an electron into a muon or ve into v^ as they In any case, for the mechanism of the suppression of mv
travel across the wall. If a discrete symmetry is embed- to work, AQ should be as low as possible, and the condi-
ded into a continuous one, one obtains a nice example of tion (2) is highly desirable. This is our claimed result of
nonintercommuting strings [1], which eventually become possibly stable domain walls in a realistic model of weak
boundaries of domain walls. interactions (it is possible, though, that gravity destabi-
The essential feature of our result is the fact that con- lizes the walls [7]).
tinuous symmetry could be almost any horizontal symme- Under the element ioi^Q, e—^jiive-^v^) and /i
try, in particular the popular family symmetry. The ~~* "~^(v/i—* ~"Ve) a lepton transversing this domain
nonintercommuting nature of the string-domain-wall wall would actually change its flavor. The astrophysical
structures is to be expected in a large class of such and cosmological consequences of such objects deserve
theories. In some cases the strings become flavor analogs further study and we will return to them in a future pub-
of Alice strings [2]; i.e., they may have the spectacular lication. Suffice it to mention that the usual argument
property of an electron turning into a muon as it travels [8] of percolation theory leading to the existence of a
around the string. large (stretching over the entire horizon) domain wall
We now turn our attention to phenomenologically may not be automatically applicable to the case of a
relevant examples of a non-Abelian discrete symmetry. non-Abelian discrete symmetry.
(I) Quaternion group and light domain walls.—A few (II) Embedding into a continuous symmetry.—As we
years ago, in an attempt to achieve a small neutrino mass mentioned before, quaternion group Q is a subgroup of
mv and a large magnetic moment /iv, Voloshin [3] sug- SU(2) and its phenomenological role was in fact inspired
gested an SU(2) global (or local) symmetry between v by this fact. Therefore, it is certainly natural to consider
and vc, which becomes even more appealing as a flavor the scenario with SU(2)// symmetry being broken down
symmetry [4] between ve and v^. Under such a symme- to Q at some higher energy. More generally, let us imag-
try, Hv, being antisymmetric, is a singlet, whereas the ine that your preferred continuous horizontal symmetry
mass is a triplet and thus forbidden. The trouble is that GH (global or local) breaks down to a non-Abelian
such SU(2)// symmetry must be broken above Mw and discrete symmetry Q at some scale A//. The most natural
thus cannot play its role at energies much below the weak candidate for G is a family symmetry.
scale relevant for neutrino mass considerations. Properties of the resulting topological structures de-
It was shown [5], however, that an equally important pend crucially on how the fermions transform under Q.
role can be played by a discrete subgroup of SU(2)//, This in turn determines the embedding of Q and its cov-
such as quaternion symmetry Q, ering continuous group SU(2)// in a family symmetry
e = {±l,±/a,,±/0r2,±/<T3}, (1) G„.
0031 -9007/93/72 (1 )/9(4)$06.00 9
© 1993 The American Physical Society

Here we study the two simplest and most natural possi- realized by the spin 1 scalars needed to give a mass to the
bilities. fermions. We should admit though that the presence of
(a) Regular embedding.— Imagine fermions belonging the S//=2 fields is not automatic and so we postpone its
to a two-dimensional representation of Q, i.e., spinor un- discussion for the case of the special embedding of
der SU(2)//. Therefore, SU(2)// is operating among two SO(3)# into SU(3) when we are forced to introduce it.
families, not a strange fact in view of the third family be- (b) Special embedding.— Let us now consider another
ing much heavier than the first two. This would simply natural embedding of the quaternion group Q in the fam-
amount to the original family symmetry SU(3)// [or ily symmetry, with fermions being vector rather than spi-
SU(A0//, TV > 3, if the number of families TV > 3] being nor representations of the covering group SU(2)#
hierarchically broken down to Q =SO(3)//. In this case family symmetry may be repre-
sented by SO(3)// itself or by a larger group SU(3)//
SU(3)„ — SU(2)„— 2 — 1 . (3)
containing SO(3)// as a special subgroup. By special
This picture reflects all the features of the Voloshin and subgroup we mean such an embedding of SO(3)// into
Chang et al. scenarios. The first stage of the breaking of SU(3)// when the fundamental triplet of SU(3)// coin-
SU(3)// can produce some interesting topological struc- cides with a triplet of SO(3)// subgroup.
tures [9]. For us the crucial point is to notice the possi- Note that the SU(3)// symmetry admits both vector
bility of a non-Abelian discrete channel which, as we em- and chiral realizations, depending on the transformation
phasized before, may play an important role in neutrino properties of left-handed and right-handed fermions. The
physics. concrete realizations were discussed already; here we
At the scale A// when SU(2)# breaks down to Q, ac- shall try to perform a maximally model independent
cording to the standard scenario, strings are produced. analysis.
These strings carry the important topological property of Let us focus first on the case of SO(3)// family symme-
being nonintercommuting; i.e., the two strings corre- try. Since quarks and leptons are triplets under SO(3)//,
sponding to the noncommuting elements of the homotopy the fermion mass terms transform as 3 x 3 = 5 4 - 3 + 1 , and
group cannot pass through each other [l]. As far as we so the Higgs fields which break SO(3)// and give masses
know, this is the first realistic example of such structures. to fermions must transform in the same manner. We do
Since normally the major mechanism that determines the not address here the detailed mechanism of the genera-
evolution of the string network is attributed to their inter- tion of fermion mass matrices, since we wish to keep our
commutativity [7], we expect the cosmology of the above analysis as model independent as possible.
system to be different from the conventional one. The bottom line in this approach is the fact that the
However, as discussed before the discrete symmetry Q horizontal family structure is projected in the fermion
must be broken at the scale KQ ^ 10 MeV, so that strings mass sector through higher-dimensional effective opera-
become boundaries of the domain walls [10]. Now the tors, since the family scale must lie much above the weak
flavor Alice effect will take place as the particle crosses scale. Of course, in general one can introduce extra
the wall. Of course, the string-wall system is no longer Higgs fields, decoupled from fermions, but such compli-
topologically stable, since it can decay through the hole cations appear neither necessary nor reasonable.
formation. This decay rate is exponentially suppressed by Notice that the diagonal mass terms can be generated
the unknown ratio AH/AQ. only by 5 and 1 representations (3 is antisymmetric).
One knows that independently of whether GH is global Neglecting the contribution of 3-plets first, let us consider
or local the scale A// must lie much above Mw> due to the a most general vacuum expectation value (VEV) of a 5-
flavor changing processes. In the local case this is trig- plet
gered by the exchange of horizontal gauge bosons with
A//^ 106 GeV, whereas in the global case the bound is
<5>- (4)
due to the possible emission of the resulting Goldstone
bosons. The astrophysical limit due to the production of ~(a+b)
Goldstone bosons in stellar objects is A//-£ 109 GeV [11]. which leaves unbroken group Q. In the case of a single
In other words, we can conservatively assume A / / ^ 1 0 6 5-plet, we can show that the most general solution is
GeV. In general we have no information on Ag, but if we a =6, in which the surviving symmetry is actually a semi-
take seriously the large-neutrino-magnetic-moment hy- direct product of U(l) and Z2. This would imply the ex-
pothesis AQ should not exceed 1 GeV or so. In this case, istence of monopoles and the well known Alice strings. In
of course, the string-wall structures would be stable for contrast with the grand unification case [12], these flavor
all practical purposes. monopoles are not necessarily cosmologically harmful.
The desired symmetry breaking SU(2)//—* AfiQ can be This is related to the fact that the scale A// may be as low
naturally achieved by the four-index symmetric represen- as 106 GeV in the local case, whereas in the global case
tation of SU(2)//, i.e., with SU(2)// family spin 5"//=2, rapid annihilation [13] of monopoles renders them harm-
while the next stage of symmetry breaking of Q should be less even for large A//. Of course, we must break the de-


generacy a = b by introducing at least one extra 5-plet. is common, in order to generate realistic mass matrices
Now, monopoles will become connected by the noninter- for the quarks, their off-diagonal elements must be
commuting strings and are no longer topologically stable. switched on. This in turn necessarily leads to the break-
As in the previous case, the stability of these structures ing of Z2XZ2 symmetry. The natural candidates for this
depends on the ratio of the VEV's of the two 5-plets. If effect are 3-plet Higgs fields or nondiagonal elements of
the respective Yukawa couplings are of the same order, 5-plets. Once again, we end up inevitably with a string-
then the electron-muon mass ratio indicates that these wall network. The measure of its stability is, of course,
VEV's must be of the same order. In this case one ex- the ratio of off-diagonal to diagonal elements of the fer-
pects that these structures can decay by nucleation into mion mass matrix, or in other words, the amount of mix-
monopole-antimonopole pairs which get rapidly annihilat- ing among different families. This means that for all
ed [14] and one is left with the topologically stable string practical purposes these structures could be considered
network. stable. Here and everywhere by stability we mean stabili-
In short the original symmetry is broken down to Q. ty under nucleation. Of course, these structures will tend
Actually, since there are no spinors, Q acts as Z2XZ2 on to disappear through collapse and a more careful and
the field context of the theory. The elements of Z2XZ2 model dependent analysis is needed to quantify this [17].
are It is the very existence of the wall that now allows ob-
servable flavor transitions. To see this, consider, for ex-
g\ - d i a g O , - 1, - 1), ^ 2 = = d i a g ( - 1,1, - 1), ample, a wall bounded by a #3 string [described by
(5) n 3 (0)]. This wall can be formed through the nonvanish-
g 3 = d i a g ( - 1 , - 1 , 1 ) , g4 = l . ing proportional triplet VEV (3) oc (1,0,0), since #3(3)
As before, the above symmetry breaking leads to the ^<3). The winding condition <3(0)) = n3(0)(3(O)> can-
formation of nonintercommuting strings. As one travels not thus be satisfied in the wall. Therefore, the fermion
around such a string, the VEV of the 5-plet winds con- mass matrix is modified nontrivially inside the wall and
tinuously by the SO(3)// group transformation this change is no longer a gauge artifact. In other words,
when a fermion is passing through the wall (nonadiabati-
<5(0)> = n r (0)(5(O)>n(0) (O<0<2;r), (6) cally) a flavor changing transition will necessarily take
where n(0) interpolates between different elements
gi € Z2XZ2 as 0 runs from 0 to In. Therefore, different Strictly speaking, the quark and lepton family groups
strings can be labeled by a group element of Z2XZ2 con- may be distinct. If there is no mixing in the lepton sector,
nected to the identity element by the path n(0) in ques- we would be left with the stable strings we described
tion [15]. For example, for the string #3, £li($) has the here.
form Similarly, we can consider a case of SU(3)// family
symmetry. The main difference from the previous case is
cos0/2 sin0/2 0 the new unshrinkable paths which connect elements of
a 3 (0) = -sin0/2 cos0/2 0 (7) Z 2 x Z 2 with the identity and belong to the SU(3)/SO(3)
0 0 1 manifold. For example, consider again a minimal string
corresponding to #3 =diag( — 1 , - 1 , 1 ) . This element can
which is just a rotation around a third axis in the flavor be obtained by an SO(3) rotation as well as the
space [in the same manner one obtains elements Cl\(0) SU(3)/SO(3) rotation n(0) =exp(/0/ 2 ^8), where A.8
and n 2 (0)]. =diag(l,l, —2) is a generator of SU(3)//. This provides
As in our previous discussion, these strings have a a new path with 0(0) = !, n(2;r)=£3- Which path
flavor changing property, but in a different manner. would correspond to the minimal energy depends on the
After a half a circle an electron turns into, say, a muon, pattern of symmetry breaking. For example, in the case
but at the end of the journey it is again an electron, with of the hierarchical breaking SU(3)//— SO(3)„— (?,
just an opposite sign of a wave function. This change clearly the minimal energy path is the one belonging to
cannot be observed by a local experiment, since all the the SO(3) manifold.
VEV's are simultaneously changing by the same group In this paper we studied the cosmological consequences
transformation. However, the flavor changing interaction of the spontaneous breaking of non-Abelian discrete sym-
of the string will manifest itself through the scattering of metries, both as basic symmetries and as a stage in sym-
the fermions on the string. This is due to the SO(3)// metry breaking of continuous symmetries. The non-
gauge field whose magnetic flux is trapped in the core (or Abelian discrete symmetry case is inspired by the quater-
the Nambu-Goldstone mode in the case of a global sym- nion group Q used to keep neutrino mass naturally small,
metry). This scattering amplitude can be strongly en- while allowing a large magnetic moment. The resulting
hanced due to the Aharonov-Bohm effect [16]. domain walls produced once the discrete symmetry gets
The rest of the story depends on whether the quarks broken can be stable, present today since the scale of the
and leptons share the same horizontal symmetry or not. breaking of Q, AQ can be as low as 1-100 MeV. The
In the more conventional scenario when the family group most interesting characteristic of these domain walls is

that the fermion changes its flavor when crossing the Schwarz and Y. S. Tyupkin, Nucl. Phys. B209, 427
wall; i.e., an electron would turn into a muon, an electron (1982).
neutrino into a muon neutrino, etc. The astrophysical [3] M. Voloshin, Yad. Fiz. 48, 804 (1988) [Sov. J. Nucl.
and cosmological consequences of such objects deserve Phys. 48, 512 (1988)1.
further study and we postpone them for a future publica- [4] K. J. Babu and R. N. Mohapatra, Phys. Rev. Lett. 63,
tion. 228 (1989).
[5] D. Chang, W.-Y. Keung, and G. Senjanovic, Phys. Rev.
Now, independent of how much you believe in non-
D42, 1599 (1990).
Abelian discrete symmetry, it can certainly result natu- [6] Ya. B. Zel'dovich, I. Yu. Kobzarev, and L. B. Okun, Zh.
rally in the process of a breaking of a continuous symme- Eksp. Teor. Fiz. 67, 3 (1974) [Sov. Phys. JETP 40, 1
try. We suggest that a large class of horizontal family (1974)]. It should be noted that AQ —100 MeV requires
symmetries do break at some stage to a non-Abelian a rather small Higgs field self-coupling; a more conserva-
discrete symmetry, such as the quaternion group studied tive limit would be ~ 1 0 MeV. We prefer, however, this
in this paper. The result is a string-wall system with a situation to having to postulate a large ( » l ) Yukawa
fermion changing its flavor in the wall. Clearly, the coupling for a muon.
effects associated with these systems are quite fascinating [7] B. Rai and G. Senjanovic, Phys. Rev. D (to be published).
and, stable or not, their cosmological and astrophysical [8] For a review, see A. Vilenkin, Phys. Rep. 121, 263
role is likely to be very important. (1985), and references therein.
[9] G. R. Dvali, Phys. Lett. B 265, 64 (1991).
Finally, we wish to mention that there are other non-
[10] T. W. B. Kibble, G. Lazarides, and Q. Shaft, Phys. Rev.
Abelian discrete subgroups of SU(3) which may be phe- D 26, 435 (1982); A. E. Everett and A. Vilenkin, Nucl.
nomenologically relevant [18], such as, e.g., a dicyclic Phys. B207, 43 (1982).
group which can be embedded into SU(2) x U ( l ) . [11] See, e.g., a review by J. E. Kim, Phys. Rep. 150, 1 (1987).
We appreciate discussions with B. Rai, V. Rubakov, [12] J. Preskill, Phys. Rev. Lett. 43, 1365 (1979).
and T. Kibble. G.D. would like to thank Professor Abdus [13] M. Bariola and A. Vilenkin, Phys. Rev. Lett. 63, 341
Salam, the International Atomic Energy Agency, (1989).
U N E S C O , and the International Centre for Theoretical [14] R. Holman, T. W. B. Kibble, and S. J. Rey, Phys. Rev.
Physics, Trieste, for support. Lett. 69, 241 (1992); Ben-Menahem and Cooper (Ref.
[15] D. Mermin, Rep. Prog. Phys. 51, 591 (1979).
[16] M. G. Alford and F. Wilczek, Phys. Rev. Lett. 62, 1071
*Electronic address: Goran@ictp.trieste,it (1989).
[1] G. Toulouse, J. Phys. (Paris), Lett. 38, L67 (1976); V. [17] E. P. S. Shellard, in The Formation and Evolution of
Poenaru and G. Toulouse, J. Phys. (Paris) 38, 887 Cosmic Strings, edited by G. Gibbons, S. W. Hawking,
(1977); T. W. B. Kibble, Phys. Rep. 67, 183 (1980); S. and T. Vachaspati (Cambridge Univ. Press, Cambridge,
Ben-Menahem and A. R. Cooper, Nucl. Phys. B388, 393 1990).
(1982). [18] D. Chang, W.-Y. Keung, S. Lipovaca, and G. Senjanovic,
[2] A. S. Schwarz, Nucl. Phys. B208, 141 (1982); A. S. Phys. Rev. Lett. 67, 953 (1991).