Flavor Non-universality Gauge Interactions and Anomalies in B-Meson Decays

Motivated by flavor non-universality and anomalies in semi-leptonic B-meson decays, we present a general and systematic discussion about how to construct anomaly-free $U(1)'$ gauge theories based on an extended standard model with only three right-handed neutrinos. If all standard model fermions are vector-like under this new gauge symmetry, the most general family non-universal charge assignments, $(a,b,c)$ for three-generation quarks and $(d,e,f)$ for leptons, need satisfy just one condition to be anomaly-free, $3(a+b+c)=-(d+e+f)$. Any assignment can be linear combinations of five independent anomaly-free solutions. We also illustrate how such models can generally lead to flavor-changing interactions and easily resolve the anomalies in B-meson decays. Probes with $B_s-\bar{B}_s$ mixing, decay into $\tau^\pm$, dilepton and dijet searches at colliders are also discussed.

Various models  in the literature, including extra Z , lepto-quark and loop-induced mechanisms, were proposed to address similar issues in the past.
In this paper we focus on Z models with an extra U (1) gauge symmetry. In the literature, usually just a specific charge assignment is chosen, without noting that many other options could be equally possible. Here, we provide a systematic investigation of general family U (1) gauge symmetry and illustrate how to choose charges consistently to get anomaly-free models without introducing new fermions, except for three right-handed neutrinos. For family universal models, there is only one non-trivial charge assignment, the well-known B − L symmetry. For family non-universal models, however, infinitely many solutions exist as linear combinations of five independent anomaly-free bases. We also show how some models can provide explanations for the anomalies in B-meson decays. This paper is organized as follows. In Section II, we discuss the consistent conditions for U (1) charge assignment first, then give an example to show how a realistic model can be constructed to match the observed fermion masses and mixings. In Section III, we exemplify one charge assignment in the context of anomalies in B-meson decays. Finally, we give our conclusion.

II. ANOMALY FREE FAMILY-NONUNIVERSAL U (1) MODELS
In this section, we give some general discussion about anomaly-free conditions for U (1) models, without introducing extra chiral fermions other than three right-handed neutrinos.
We denote the weak doublets and singlets as follows, ψ = u, d, e, ν, with P L = (1 − γ 5 )/2, P R = (1 + γ 5 )/2 and i = 1, 2, 3 as the family/generation index. The anomaly is proportional to the completely symmetric constant factor, T α is the representation of the gauge algebra on the set of all left-handed fermion and antifermion fields, and "tr" stands for summing over those fermion and anti-fermion species.
Note that the T s above may or may not be the same since they depend on the referred gauge groups and also the chiral fermions running in the loop of the triangle anomaly-diagram.
The anomaly free conditions for the theory are given by So far, the discussion has been standard and the solution space of the above equations is expected to be large since we have more variables than equations. Interestingly, one can easily check that the first four equations are satisfied automatically if fermions are vector-like under the new U (1) gauge symmetry, namely With vector-like charge assignment, we only need take care of the last two linear equations, which are actually reduced to just one, This equation is much easier to solve, but could have multiple solutions. For example, 1. Family universal model: which is the unique non-trivial solution, the well-known B − L gauge symmetry.
2. Family non-universal models: where z i are not identical. Since we have six variables but just one constraint, infinitely many solutions exist. For example, we are free to choose just one generation to be charged, the other two as singlets, or any assignments for quark sector with a proper choice of charges for leptons. Some models have been discussed in Refs. [60][61][62][63]. In general, we can have the charge assignment as in Table I, where a, b, c, d, e and f are arbitrary real numbers but satisfy As a special case, we could also imagine that anomalies are canceled separately in the quark and leptons sectors, namely Such a parametrization includes some well-studied models, such as a = b = c = 0 and The solution space for Eq. (9) is five-dimensional, so we can choose the following five independent solutions as the bases, As emphasized above, we are restricting ourselves to extended models with only three additional right-handed neutrinos. If more particles are to be introduced, requirements on the charge assignment should change correspondingly. For example, one could also introduce more SM-singlet Weyl fermions χ j with U (1) charge X j , in cases where SM fermions are vector-like in U (1) , giving Some fermion χ k actually could be a dark matter (DM) candidate. For instance, a Majorana mass termχ c k χ k would be induced after U (1) symmetry breaking by a SM-singlet scalar S with U (1) charge 2X k , since interactions likeχ c k χ k S † are allowed. Vector-like χ k is another popular scenario for DM where the Dirac mass termχ k χ k is allowed. In both cases, Z 2 symmetry can protect the stability of DM.
To build realistic models with correct SM fermion masses and mixings, we need to introduce some scalar fields H i to spontaneously break gauge symmetries. The scalar contents would be highly dependent on the charge assignments for these chiral fermions. In the most general cases, for the quark sector we can introduce several Higgs doublets with hypercharge Y = −1 and U (1) charges, a − b, a − c and b − c, to make renormalizable Yukawa interactions, giving the desired quark masses and CKM mixing matrix. In the lepton sector, Higgs doublets with U (1) charges, d − e, d − f and e − f , suffice to give lepton masses and neutrino mixing.
Below, we shall give an example with explicit charge assignment to illustrate how consistent models can be constructed [60]. Let us focus on the quark sector first. We shall use the following setup: The above symmetry can be regarded as 3(B u − B c ) + 6(B c − B t ), expanded in the five bases of Eq. 10. Some phenomenologies have been studied first in Ref. [60], and later in Ref. [35] along with L µ − L τ symmetry in the lepton sector. Here, this model is introduced just for illustration and will be referred to in comparison with the model for the B-anomaly in Section III.  (13) where y u,d ij are the Yukawa couplings. After H 1 gets a vacuum expectation value (VEV), the resulting mass matrices for u and d have the following form: This kind of mass matrix cannot give the correct CKM matrix, since the third generation will not mix with the other two. Now if we have two more Higgs doublets, H 2 with U (1) charge −3 and H 3 with +3, the following Yukawa term are allowed: When both H 2/3 get VEVs, these terms contribute to the mass matrices with Three-flavor mixing can still arise after diagonalization of M H 1 u,d + M H 2 u,d . One can easily discuss leptons, since similar physics appears. For example if z L i = (0, 1, −1), extra Higgs doublets with charges ±1 and/or ±2 would be able to achieve the required lepton masses and mixing.
Gauge bosons will get their masses through the Higgs mechanism. When H 2 and H 3 get VEVs, the U (1) gauge symmetry is also broken. If the U (1) gauge coupling is comparable to the electroweak coupling, the Z boson is expected to have a mass around the electroweak scale, which is highly constrained. To get a heavy Z boson, an electroweak singlet scalar S with U (1) charge z s can be introduced. Then the following vacuum configuration would break the gauge symmetries to U (1) em , The kinetic terms for scalars are where D µ is the covariant derivative. From this Lagrangian, the W ± mass can be simply read out, g 2 v 2 1 + v 2 2 + v 2 3 /2. Neutral gauge bosons, on the other hand, are generally mixed, but it is possible to make Z heavy when v s v i such that experimental constraints from Z − Z mixing can be safely evaded, since the mixing is proportional to v 2 i /v 2 s ; see Ref. [64] for a general review.
The interaction forψψZ can be obtained from gZ µ J µ Z , where g is the gauge coupling constant of U (1) and the current J µ Z in the gauge eigenstates is given by The above where Ψ = U, D, e, ν are the mass eigenstates. The CKM matrix is given by V CKM = V † U L V D L and the neutrino mixing matrix by V PMNS = V † e L V ν L . The rotation of fermion fields in Eq.(17) leads to We have used ψ ≡ ψ L = ψ R , since we are considering the vector-like charge assignment.
One can immediately notice that generally V † V ∝ I if ∝ I, namely family non-universal gauge interactions. In our previous examples, we have ψ ∝ diag (1, 1, −2) or diag (0, 1, −1), and we expect flavor-changing effects to arise. Since only V CKM or V PMNS is experimentally measured, the individual matrix V ψ L,R is unknown. Thus the resulting products V † are also unknown.

III. PHENOMENOLOGIES AND ANOMALIES IN B-MESON DECAYS
In this section, we discuss how the above framework can address recent anomalies in B physics. Since left-handed fermions have the same charges as the right-handed ones, we can where z U = (a, b, c), z L = (d, e, f ), and Flavor changing processes can happen when δ ψ = 0 or δB ψ L,R ij = 0. Note that elements in the matrix δB ψ L,R are not necessarily smaller than z ψ 1 for a general setup, since z ψ 1 can be zero if fermions in the first generation are U (1) singlets.
To illustrate how it affects B meson decay, we exemplify the following anomaly-free charge This assignment can be expanded by the bases in Eq. (10), which is a nice example in the sense that it involves all five anomaly-free bases. If q µ = −3/2, the lepton sector has some kind of L µ + L τ symmetry. If |q µ | 1, only the third generation is effectively U (1) -charged. We should emphasize again that it is free to change the above assignment by adding any linear combinations of other anomaly-free solutions. For example, we could use z U = (1, 1, −1) which is just the sum of the above charges with (1, 1, −2) mentioned earlier. However, these two models give different signal strengths in experiments, such as LHC dijet events, therefore they are subject to different constraints.
The b → s transitions are usually analyzed in terms of the following effective Hamiltonian Here V is the CKM matrix and α = 1/137 is the fine-structure constant. Note that the coefficients C i and C i are scale-dependent, governed by the renormalization group equation.
They are first calculated at high scales and then run to a lower scale, which is usually taken as the bottom quark mass m b for decay processes. We just list some relevant operators for our later discussions: In general, all the above operators can be generated. Since anomalies are closely related to O 9 , we calculate the induced coefficient for O 9 by Z -mediated new physics To resolve the anomalies, C NP 9 should be around −1.1 [6,7], which can be translated into The above formula is generally applicable to any non-trivial charge assignment. In some cases, we can simplify it further. For instance, since B e L/R µµ are elements in the diagonal, we could expect B e L µµ ∼ q µ if |q µ | |3 + q µ |, or no rotation in the charged lepton sector (V PMNS = V ν L ), and Eq. (25) can be approximated as Now with the charge assignment as in Eq. (21), we explicitly have 33 . If the CKM matrix comes solely from the rotation of down quarks, we would have Other coefficients can also be calculated similarly. Also, if q µ = −3, we would expect new physics effects to show up in B → K ( * ) τ + τ − . Since we mainly focus on O 9 -related anomalies in B-meson decays to muons, we shall neglect other operators as long as the setup does not violate current limits. For example, we can freely choose B e L µµ = B e R µµ , which results in C NP 10 = 0 = C NP 10 . Z may also mediate B s −B s mixing in the above scenario, since the operator (sγ µ P L b) 2 is inevitably induced, which actually gives the most stringent limit at the moment. Current bounds [65] can be put on the following quantity: Comparing with Eq. (27), we can safely evade this constraint for |q µ | 3 and resolve B anomalies at the same time.
In Fig. 1, motivated   Since Z couples to both quarks and leptons, dilepton and dijet searches for heavy resonances at colliders can probe Z . The expected signal strength depends on where σ is the cross section for Z production, f and f are SM fermions, and Br denotes the decay branching ratio. For hadron colliders, we shall integrate the above quantity over the quark parton distribution functions (PDFs) (throughout our calculations, we have used MMHT2014 [66] PDFs). In the case of charge assignment for quarks, (0, 0, 1), hadron colliders such as LHC with energy √ s = 13 TeV have less discovery potential for Z , since Z would couple weakly to the first two generations through quark mixing only, but strongly to the third generation, which has small PDFs. A future 100 TeV hadron collider has a better chance because the production rate is increased thanks to the enhancement of the PDFs of bottom and top quarks. In Fig. 2(a), we give the ratio of Z production from bottom and top quarks in our model to that from light quarks if Z also couples to u and d. We 1 TeV would still be allowed, which can be inferred from Fig. 2(b), where we show dilepton searches for a Sequential SM (SSM) Z (SSM Z is identical to SM Z except for the mass) as the dashed black line. The region above the solid red curve is excluded by dilepton searches [67]. However, if the signal strength is reduced by 10 or 100, the exclusion limit would be shifted to 2.4 TeV (dashed blue) and 1.2 TeV (dotdashed purple), respectively. Since in the model discussed, the cross section is lowered by O(10 3 ), taking the branching ratio into account would give M Z O(600 GeV), with some dependences on q µ . Similarly, constraints from dijets are also weakened.
In comparison, the charge assignment (1, 1, −1), which is the linear combination of (1, 1, −2) in Section II and (0, 0, 1), will give different results. In such a case, Z can couple to light quarks and the cross section for production can be sizable.   Table I have to, and only need to, satisfy the condition given in Eq. (9). Generally, infinitely many anomaly-free family non-universal models exist, as linear combinations of five independent anomaly-free bases. If both bottom quark and muon couple to this new U (1), typically the anomalies in B-meson decays can be explained.
We have also discussed several other experimental searches for such models, including Z -mediated effects in B s −B s mixing, dilepton and dijet searches for heavy resonances at colliders. Some viable parameter space has already been probed by these searches. Future searches in colliders and other B-meson decay modes should be able to provide more powerful information on the physical parameters and test different scenarios for Z charge assignment.