IC/92/432 UMDHEP 93-105 LMU-16/92 December 1992 PLANCK-SCALE PHYSICS AND SOLUTIONS TO THE STRONG 3 9 CP-PROBLEM WITHOUT AXION 9 1 n Zurab G. Berezhiani∗†, a J 5 Sektion Physik der Universit¨at Mu¨nchen, D-8000 Munich-2, Germany 1 Institute of Physics, Georgian Academy of Sciences, Tbilisi 380077, Georgia 2 v 8 Rabindra N. Mohapatra‡ 1 3 2 Department of Physics, University of Maryland, College Park, MD 20742, USA 1 2 9 and / h p Goran Senjanovi´c§ - p e h International Centre for Theoretical Physics, I-34100 Trieste, Italy : v i X Abstract r a We analyse the impact of quantum gravity on the possible solutions to the strong CPproblemwhichutilizethespontaneouslybrokendiscretesymmetries,suchasparity and time reversal invariance. We find that the stability of the solution under Planck scale effects provides an upper limit on the scale Λ of relevant symmetry breaking. This result is model dependent and the bound is most restrictive for the seesaw type models of fermion masses, with Λ< 106 GeV. ∗Alexander von Humboldt Fellow †E-mail: [email protected], 39967::berezhiani ‡E-mail: rmohapatra@umdhep, 47314::rmohapatra §E-mail: [email protected], vxicp1::gorans 1. Introduction. It is well known that the instanton effects bring about a periodic structure of QCD vacuum. This leads to CP-violation by strong interaction [1], character- ized by the Θ¯ parameter defined as Θ¯ = Θ +Θ (1) QCD QFD ˜ Here Θ is the coefficient of the P- and CP-violating gluonic anomaly term GG and QCD Θ =argDetmˆ mˆ is a fermionic contribution, where mˆ and mˆ are the up and down QFD u d u d quark mass matrices. This CP-violation manifests itself in an appearance of the neutron dipole electric moment, which is known experimentally [2] to be less than 10−25 e cm leading to a phenomenological upper bound on Θ¯ of about 10−9 [3]. However, in the standard model Θ¯ receives an infinite renormalization even if it is put to zero at tree level by hand [4]. Understanding the smallness of the Θ¯ without fine tuning of parameters is known as the strong CP-problem. There are two widely discussed approaches to solving this problem: i) The Peccei-Quinn mechanism [5] where the whole Lagrangian of QCD plus QFD is required to obey a global U(1) invariance with nonvanishing color anomaly, which PQ ¯ dynamically fixes Θ = 0. The spontaneous breaking of this symmetry leads to the existence of a pseudo-Goldstone boson - axion [6, 7]. ii) The discrete symmetry approach where a combination of discrete symmetries such ¯ as P or CP is used to set Θ = 0 naturally [8, 9, 10, 11, 12, 13, 14] at the tree level. In such a theory a finite Θ¯ arises at the higher loop level and one has to show that Θ¯ < 10−9. An essential common ingredient of both these approaches is the presence of global sym- ¯ metries, either U(1) or P/CP, that guarantee the smallness of Θ. However, these are not PQ dynamical symmetries. There is no physical ground to exclude that they are violated by higher dimensional effective operators, that couldoriginatefromnew interactions existing at some high scale M. These operators must be of non-renormalizable type so that at M → ∞ their effects disappear. The ultimate scale for such higher order operators can be regarded as a Planck scale M , where the gravity becomes as strong as other interactions. This Pl is a product of one’s experience with the quantum gravitational effects related to virtual black holes [15] or wormholes [16] which are likely not to respect global symmetries. It is 1 therefore important to include all higher dimensional operators, consistent with the local invariance, in the effective low energy theory before discussing whether the model solves 1 the strong CP-problem. Since in most models the extra global symmetries imposed on the Lagrangian are not automatic symmetries, one could argue that perhaps the renormalizable 2 terms in the low energy theory should also be allowed to break these symmetries. In the absence of detailed calculations of the non-perturbative quantum gravity effects it is hard to argue for or against this. In this paper (as also in refs. [22, 23]) we will assume that only the Planck scale induced non-renormalizable terms are relevant. Certainly, they also have the nice property of vanishing in the limit of zero gravity. Of course, no such apology is needed if the theory is automatically invariant under these global symmetries. Such a study for the PQ models was performed in recent papers [22]. It has been shown that barring an unnaturally high degree of suppression of the strength of the higher dimensional operators, the scale of U(1) symmetry, V , must be less than 100 GeV in PQ PQ the simplest theories of the invisible axion [7], whereas the lower bound on V coming PQ from various physical and astrophysical data is larger by many orders of magnitude [1]. This result considerably diminishes our belief in Peccei-Quinn symmetry as a solution to 3 the strong CP-problem. 1Some authors [17] put forward the idea that wormholes themselves may set Θ¯ to 0 or π dynamically, therebyavoidingthestrongCP-problem. However,asofnowthereisnouniversalagreementonthevalidity of this point. 2In a self-consistent picture the global symmetry should originate as an accidental symmetry of the theory at lower energies, being automatically respected by the renormalizable piece of the Lagrangian due tothecertainfieldcontent. Thewell-knownexamplesaretheleptonandbaryonnumberconservationinthe standardmodel. The quantum gravityeffects can violate them only throughthe d=5 and d=6 operators with M ∼M [18] - all renormalizable terms are automatically invariant under these symmetries. (In the Pl context of grand unification these operators can appear at the lower scale, with M ∼M [19].) Neither GUT U(1) nor P and CP are automatic in general, though there are a few attempts to introduce U(1) as PQ PQ an accidental global symmetry, at the price of enlarging the local symmetry of the theory [20]. It has been arguedrecently [21], that P or CP may also appear automatically as discrete gauge symmetries in theories with dimensional compactification. 3 It is amusingto notice that the originalPeccei-Quinnmodel [5] with low scaleaxion[6] is rather stable against Planck scale corrections. It is however ruled out by experimental data. 2 In this paper we consider the effects of gravity on the second class of solutions. The original idea [8] was to use discrete symmetries such as P or T in order to have Θ-term van- ishing at tree level and to keep it finite and calculable in perturbation theory. The challenge ¯ in this approach is to come out with a simple enough model which gives Θ sufficiently small. The original models [8, 9] suggested to illustrate the philosophy behind them ended up using some ad hoc further symmetries needed for the consistency of the program. One would not have expected these symmetries to exist for any other reason. Yet another difficulty of these models is that the Higgs sector involved in electroweak symmetry breaking is nonminimal, in which case the natural suppression of flavour-changing neutral currents (FCNC) [24] does not occur. A more realistic scheme was suggested by Nelson [10] and generalized by Barr [11], which utilize CP invariance to put Θ = 0, and special field content to achieve also QCD Θ = 0 at the tree level after spontaneous breaking of CP. The key ingredients of these QFD models, which also avoid a problem of FCNC, are: i) the presence of extra heavy fermions which are mixed with ordinary quarks, ii) the hypothesis that spontaneous CP-violationtakes place only inthese mixing terms. The Θ¯-term effectively arises only at the 1-loop level. It is less than 10−9 if the Yukawa coupling constants are sufficiently small, less than 10−3. However, the simplest possibility which utilizes the heavy fermions came in paper [12] and subsequently in papers [13, 14]. The idea of refs. [12, 13] is based on the universal seesaw mechanism [25, 26]: the quark and lepton masses appear due to their mixing with heavy fermions, in direct analogy with the well-known seesaw picture for neutrinos [27]. In this picture the solution to the strong CP-problem can be implemented through the spontaneous violation of P-parity only [12], as soon as one deals with left-right symmetric model SU(2) ⊗SU(2) ⊗U(1). No other additional symmetry is required. Alternatively, L R one can use a concept of CP-invariance (without P-parity) even in the context of the SU(2) ⊗ U(1) model [13]. This possibility, however, requires also some extra symmetries (e.g. horizontal family symmetry, as also in the model of Nelson [10]). One can show that in these models, with reasonable assumptions about the new scales and parameters, the 3 effective Θ¯ arising in loop effects is small enough (< 10−9). A different way to use the concept of parity was suggested in ref. [14]. The electroweak gauge symmetry of standard model was doubled by introducing the mirror world with new weak interactions being right-handed, which repeats the whole pattern of fermion masses of our left-handed world at some higher scale. This is a result of spontaneous violation of P-parity between ordinary and mirror worlds. The strong interactions are the same in both sectors, so the contributions of mirror fermions cancel the infinite renormalization of Θ¯ within the standard model and Θ¯ is guaranteed to be negligibly small, less than 10−19. In this paper we discuss the impact of the Planck scale effects on the models of refs. [12], [13] and [14], which in the following are referred to as BM, B and BCS models, respectively. We show that these effects provide an upper bound on the scale of relevant symmetry breaking (P or CP), with interesting phenomenological consequences. 2. The BM model. ThismodelisbasedonthegaugeSU(3) ⊗SU(2) ⊗SU(2) ⊗U(1) c L R symmetry with the quark fields in following representations: q (1/2,0,1/3), U (0,0,4/3), D (0,0,−2/3) Li Ri Ri q (0,1/2,1/3), U (0,0,4/3), D (0,0,−2/3) (2) Ri Li Li where the SU(2) isospins I and U(1) hypercharge Y are shown explicitly (the indices L,R L,R of the colour SU(3) are omitted), and i = 1,2,3 is the family index. The Higgs sector c consists of only two doublets H (1/2,0,1) L H (0,1/2,1) (3) R Obviously, the fields in first rows of eqs. (2)-(3) have the usual standard model content with respect to SU(2) ⊗U(1) whereas the fields in second rows form the analogous set of L SU(2) ⊗U(1).4 The most general Yukawa couplings, consistent with gauge invariance, are R 4By adding the obvious lepton fields to the quarks of eq. (2) the theory is free of gauge anomalies. As far as strong CP-problem is concerned, we do not consider them here. 4 essentially the standard model ones: L = Γij q¯ U H˜ +Γij q¯ D H +h.c. L Lu Li Ri L Ld Li Ri L L = Γij q¯ U H˜ +Γij q¯ D H +h.c. (4) R Ru Ri Li R Rd Ri Li R For the singlet quarks Q = U,D the mass terms MˆijQ¯ Q are also allowed, unless they Q Li Rj are suppressed by some additional symmetry. Imposing the discrete left-right symmetry P , which is essentially parity [28]: LR q ↔ q , Q ↔ Q , H ↔ H , Wµ ↔ Wµ (5) L R L R L R L R we have Γ = Γ = Γ (q = u,d), and the mass matrices Mˆ are forced to be hermitean. Lq Rq q Q The VEVs < H0 >= v and < H0 >= v , with v ≫ v = 174 GeV, violate the P L L R R R L LR invariance and break the gauge symmetry down to U(1) . As a result, the whole 6 × 6 em mass matrices of quarks take the form q Q R R M = q¯L  0 ΓvL  (6) Q¯  Γ†v Mˆ  L  R  where Γ = Γ and Mˆ = Mˆ for the up- and down-type quarks, respectively. Notice, u,d U,D that the DetM ∼DetÆà is real and therefore Θ = 0. Since the Θ is absent from QFD QCD the beginning due to parity invariance, we have Θ¯ = 0 naturally at tree level. Then all that remains to do is to identify properly the fermion mass eigenstates and show that the effective Θ¯ arising with radiative corrections is sufficiently small. In fact, the structure described above reflects the spirit of both BM and BCS models, which are in fact two limiting cases, corresponding to Mˆ ≫ v and Mˆ → 0, respectively. R However, the quantum gravitational effects can induce the higher dimensional operators violating explicitly the global P invariance and thereby effectively contributing to Θ¯. LR These operators should be cutoff by Planck scale M , so that their effects disappear at Pl M → ∞. The leading order terms allowed by gauge symmetry are the following: Pl 1 L5 = M q¯Li(αuijH˜LH˜R† +αdijHLHR†)qRj + h.c. (7) Pl 5 1 L′ = Q¯ Q (βij H†H +βij H†H )+h.c. (8) 5 M Li Rj RQ R R LQ L L Pl 1 L6 = M2 q¯Li(γLijuURjH˜L +γLijdDRjHL)HR†HR + (L ↔ R) + h.c. (9) Pl where the α,β and γ’s are in general the complex constants of the order of one. Notice, that for these operators to be P-invariant, the matrices α and β must be hermitean, and q Q † γ = γ . Since we expect that the Planck scale effects are not to respect the P-invariance, Lq Rq we assume the above matrices to be arbitrary. Let us study now the impact of these operators on the Θ¯ parameter. It is convenient to assume that Mˆ ≫ v , in which case the ordinary light quarks are essentially q’s, whereas R Q’s form a heavy states mixed with the latter through the non-diagonal terms in eq. (6). The mass matrices of the q’s, induced due to this, so called universal seesaw mixing [25, 26], are the following: mˆ = v v ΓMˆ−1Γ† (10) L R The inter-family hierarchy (hierarchy between eigenvalues of mˆ) can be related either with corresponding hierarchy in Γ’s or with the inverted hierarchy [29] of the eigenvalues of Mˆ’s. ¯ As we have seen above, Θ is vanishing at tree level. It was shown in ref. [12] that a finite and small Θ¯ arises at the two loop level, whose magnitude can be less than 10−9 for the reasonable choice of parameters in the theory. However, the Planck scale operator (7) will change the mass matrix (6) to the form: M+∆M =  αvLvR/MPl ΓvL  (11)  Γ†v Mˆ   R  (Other contributions are neglected). Since the coefficients α are in general complex, the effective Θ¯ is induced: 1 v v Θ¯ ≃ Tr(αΓ†−1MˆΓ−1) = L R Tr(αmˆ−1) (12) M M Pl Pl Obviously, the dominant contribution in eq. (12) comes from the light quarks u and d with masses ∼few MeV. Then the condition Θ¯ < 10−9 constrains the scale of right-handed current v . Demanding that both the moduli and phases of the α’s are O(1) (certainly, R 6 one should not exclude the possibility of an order of magnitude suppression), and barring unforeseen conspiracies, we therefore conservatively estimate an upper limit on v of about R 106 GeV. As long as the seesaw formula (10) is assumed to be valid, i.e. Mˆ ≫ Γv , this R limit is rather independent of the details of the model. It equally applies to the original version of BM model [12], where the heavy Q fermion masses M are assumed to be of the same order and the inter-family hierarchy is related to the hierarchy of Yukawa couplings in eq.(4), as well as to the inverse hierarchy model [29], where all Γ’s are assumed to be O(1) and the inter-family hierarchy is originated from the hierarchy in Mˆ’s. 3. The BCS model. This model also utilizes the discrete P symmetry acting on LR the set of fermions as in eq. (2) and scalars as in eq. (3). However, the mass terms Mˆ Q of Q’s are put to zero due to additional axial symmetry U(1) . The U(1) hypercharges A A are defined as following: Y = Y for the fields of the first rows of eqs. (2) and (3), and A Y = −Y for the second rows. It is obvious that incorporating this symmetry, the theory A 5 remains free of gauge anomalies. It can be local or global. Forbiddingtheexplicitmassterms, theU(1) symmetryhasnothingagainsttheYukawa A couplings in eq. (4). As far as the fermion mass spectrum and mixing is concerned, this model completely operates with the parameters of the standard model. Two fermion sec- tors are completely decoupled in the mass matrix (6): mˆ = Γv is a mass matrix of the L ordinary quarks q and Q , whereas the mass matrix of the mirror ones q and Q is just L R R L rescaled by the factor v /v . At the tree level their contributions in Θ¯ cancel each other: R L Θ¯ =argDetÆà = 0, and the non-vanishing contribution to Θ¯ arising only at higher loops 5Theoriginalversion[14]oftheBCSmodelisbasedonthelocalsymmetrySU(3) ⊗[SU(2) ⊗U(1) ]⊗ c L L [SU(2) ⊗U(1) ], where SU(2) ⊗U(1) acting on the fields q ,Q and H corresponds to the standard R R L L L R L model of electroweak interactions and SU(2) ⊗ U(1) with the fields q ,Q and H corresponds to R R R L R the parallel mirror world, a complete replice of ours, but with new weak interactions being right-handed. These two worlds communicate only via the same colour SU(3) . No doubt that apart from nice ”mirror” c philosophy behind it, such a presentation is completely equivalent to that we consider above: U(1) ⊗ L U(1) =U(1)⊗U(1) with Y =Y +Y andY =Y −Y . Moreover,inourcaseU(1) canbe globalas R A L R A L R A well. It cannotserveus as aPeccei-Quinnsymmetry, being free ofgaugeanomalies. As we show below,the impact of the Planck scale physics for the case of global U(1) is different from the case of the local one. A 7 is extremely small. Indeed, it was shown by Ellis and Gaillard [4] that in standard model Θ¯, once put to zero at tree level, arises only at the 3-loop level and is about 10−19. The divergent contributions appear only at the 6-loop level. However, in BCS scenario these are cancelled by contributions of mirror quarks: v , as the scale of the mirror (or parity) R symmetry breaking provides the natural cutoff. This scale can be arbitrarily large and so leaves us with a little hope of detecting the mirror fermions. The Planck scale operators, however, provide an upper bound on v . Let us consider R first the case of U(1) symmetry being local, as it was suggested in the original version of A the BCS model. In this case the effective d = 5 operators of eq. (8) are forbidden by local symmetry and the dominant contributions to Θ¯ come from the d = 6 operators of the eq. (9). One has: 2 v Θ¯ ≃ R Tr(γΓ−1) (13) 2 M Pl Then, by considering the contributions of the light quarks being dominant, the condition Θ¯ < 10−9 constrains v to be less than about 1012 −1013 GeV. R In the case of U(1) symmetry being global both d = 5 operators (7) and (8) are active. A Then the fermion mass matrices take the form: M+∆M =  αvLvR/MPl ΓvL  (14)  Γ†v βv2/M   R R Pl  so that the condition 2 v Θ¯ ≃ R Tr(αΓ†−1βΓ−1) < 10−9 (15) M2 Pl implies v < 109−1010 GeV.6 This limit makes mirror world accessible at SSC/LHC, since R in this case the mirror partner of electron cannot be heavier than about 10 TeV. 4. The B model. This model is also based on the field content of eqs. (2)-(3), but in addition it utilizes also concept of local horizontal symmetry SU(3) [30]: the fermions of H thefirst row ineq. (2) transformas triplets ofSU(3) andof the second row as anti-triplets, H while the scalars in eq. (3) are SU(3) singlets. In fact, this is exactly the field content H 6Obviously, the same limit applies to the general case of the model without U(1) symmetry when the A mass terms of Q’s are allowed but are assumed to be less than v . R 8 of ref. [25] where the universal seesaw mechanism was suggested for the generation of the quark and charged lepton masses. Clearly, SU(3) is free of gauge anomalies. The matrices H Γ of the Yukawa coupling constants are forced now to be SU(3) singlets (i.e. proportional H totheunit3×3matrix). The explicit massterms Mˆ areforbiddenbyhorizontalsymmetry, Q but they appear due to Yukawa couplings G Q¯ Q ξ , where ξ (n = 1,2,..), are some nQ L R n n scalar fields in representations 3 and ¯6 of SU(3) , introduced for the breaking of horizontal H symmetry. Therefore, the mass matrices of the heavy fermions Q = U,D have the form: Mˆ = G < ξ > (16) Q X nQ n Provided that Mˆ > v , mass matrix of the ordinary quarks q appears due to their seesaw Q R mixing withQ’s. SincetheYukawa couplingsΓ arethesame foreachfamily, theinter-family hierarchy between q’s is necessarily related to the inverse hierarchy of the masses of Q’s, which, on the other hand, reflects the hierarchy of the horizontal symmetry breaking. The presence of chiral horizontal symmetry SU(3) makes it unnatural to impose the H left-right parity. However, CP-invariance can be imposed, which implies that allthe Yukawa couplings can be taken to be real. The spontaneous CP-violationoccurs in a sector of heavy fermionsduetorelativephasesoftheVEVs< ξ >[13],andistransferedtothemassmatrix n mˆ of q’s due to seesaw mechanism. However, Θ¯ remains vanishing at tree level. It appears in radiative corrections and can be rendered to be less than 10−9 under certain assumptions on the parameters of the theory. Let us include now the Planck scale effects. Instead of d = 5 operators (7), which are forbidden by horizontal symmetry, one has to consider the d = 6 operators 1 q¯ (αnH˜ H˜† +αnH H†)q ξ† + h.c. (17) M2 L u L R d L R R n Pl Accounting for these operators, after the similar considerations as in BM model one can 10 deduce the limit v < 10 GeV. This constraint holds also true if instead of SU(3) one R H considers the left-right horizontal symmetry SU(3) ⊗SU(3) , with the scalars ξ being HL HR n inrepresentations(¯3,3). TheP parityisnaturalinthiscase, underwhichξ ↔ ξ†. Thenthe n n strong CP-problem can be solved due to P-parity only, without imposing CP-invariance, since the tree-level features of the model are essentially the same as in BM model. 9