Natural quark mixing and inverse seesaw in a left-right model with an axion

We consider a minimal left-right model with a Peccei-Quinn symmetry, where generalised charge conjugation plays the role of the left-right symmetry. We show how the spontaneous breaking of the Peccei-Quinn symmetry by a scalar singlet can provide us with solutions not only to the strong CP and dark matter problems but can also help to generate naturally suppressed off-diagonal CKM elements and small neutrino masses via the inverse seesaw mechanism. For this, we make use of an economical scalar sector composed of a bi-doublet, two doublets and a singlet only. As a result of the new gauge bosons and neutrinos, the neutrinoless double beta decay receives new contributions which can, in principle, become relevant due to the low-scale nature of the inverse seesaw mechanism. The model can easily accommodate all the current data on fermion masses and mixing even if the left-right scale is only high enough to evade the current experimental constraints.


I. INTRODUCTION
A common way to tackle some of the open questions in the Standard Model (SM) is to extend its gauge structure. One of the simplest SM extensions, based on the gauge group is the left-right (LR) symmetric model. In its minimal versions, LR models also feature a discrete symmetry, either the generalised parity P or charge conjugation C, connecting both the SU (2) groups. Even though P is more commonly used as the LR symmetry and, in fact, it was originally chosen as such [1][2][3][4], the case for adopting C instead has the advantage of being compatible with the SO(10) grand unified theory. As a result of the imposition of such symmetries, three generations of right-handed neutrinos are necessarily introduced allowing, in principle, for the implementation of the (type-I) seesaw mechanism for neutrino mass generation.
In general, generating neutrino masses via the type-I seesaw mechanism requires a very high energy scale (v R ∼ Λ GU T ) from which the right-handed neutrinos get their masses. In the case of the LR model, the very same v R scale breaks the original symmetry down to the SM group, and searches for new gauge bosons can be used to put a lower bound on v R of only a few TeV [5,6].
While the type-I seesaw mechanism can be accommodated at such a "low" scale, the price to be paid is the introduction of unreasonably suppressed Yukawa couplings. Thus, other variants of the seesaw mechanism that can take place at the TeV scale without requiring unnaturally suppressed Yukawas, such as the inverse seesaw [7], become very attractive in this context. The implementation of the inverse seesaw in some of the different versions of the LR model has been explored e.g. in refs. [8][9][10][11][12][13][14].
When it comes to the quark sector, the minimal LR model provides quarks with masses due to their coupling to a scalar bi-doublet (and its charge conjugated) that acquires a nontrivial vacuum expectation value (vev). However, similar to the SM, the mass hierarchy among the different generations and the approximately diagonal structure of the quark mixing, CKM, matrix are left unexplained. In a recent work [15], the authors propose an elegant solution within the LR framework to the latter problem to what they refer as the "flavour alignment" puzzle. In their model, quark mixing is small as a result of being generated at one-loop in a LR model with a Peccei-Quinn (PQ) symmetry [16][17][18]. This happens when, in addition to the new global symmetry, new degrees of freedom are introduced in the scalar sector of the most popular LR model. With the spontaneous breaking of the PQ symmetry at a very high scale, the strong CP problem is solved and an invisible axion arises and plays the role of a dark matter candidate. Nevertheless, in order to reproduce the current neutrino data, they argue that the LR symmetry has to be broken at 50 TeV, which is above the most stringent constraints as well as outside the reach of current experiments. An alternative solution to the flavour alignment puzzle as well as the strong CP problem was later proposed in ref. [19], where the PQ symmetry is broken softly in such a way that the both the off-diagonal CKM elements and the strong CP phase are naturally suppressed for they are radiatively generated. In both ref. [15] and ref. [19], P is the LR symmetry.
In this paper, we face the above-mentioned problems by proposing a minimal LR model with a PQ symmetry, where C plays the role of the LR symmetry. In such a framework, we show that, when the global PQ symmetry is broken at a very high scale by a scalar singlet, the model provides solutions not only to the strong CP and dark matter problems but can also help to generate naturally suppressed off-diagonal elements for the quark mixing matrix and small neutrino masses via the inverse seesaw mechanism. The adopted scalar sector is compact. In addition to the PQ symmetry breaking scalar singlet σ, a SU (2) R doublet, η R , is required to break the LR symmetry down to the SM group. A SU (2) L doublet, η L , is also introduced to preserve the LR symmetry.
Finally, to perform the last step of the symmetry breaking process, a scalar bi-doublet, Φ, gets a vev at the electroweak scale. As the SU (2) R symmetry is broken by v R / √ 2, the vev of η R , the This work is written following the structure below. In Sec. II, we present our model, introducing all the relevant fields and showing how they transform under the symmetries considered. The scalar potential is then shown in Sec. III, followed by the derivation of the respective particle spectrum.
Sec. IV deals with the gauge sector and its spectrum. In Sec. V, fermion masses and mixing are studied in detail. For the quarks, we show how the structure of the CKM matrix is naturally generated in our model. While for the leptons, we show how the active neutrinos get sub-eV masses via the inverse seesaw mechanism. Section VI is devoted to the study of the most relevant contributions to 0ν2β. Finally, our conclusions are summarised in Sec. VII, and particular solutions for the quark and lepton sectors are displayed in the Appendices B and C, respectively.

II. SYMMETRIES AND THE FIELD CONTENT
The continuous symmetry group of our LR model can be written as where U (1) P Q , the global Peccei-Quinn symmetry, is in parentheses to emphasise that it is not a gauged symmetry as the others. Additionally, as already mentioned, the model is invariant under a discrete C symmetry. In what follows, we present all the fields in our model and define how they transform under these symmetries.
The left-handed fermions transform as SU (2) L doublets, in the same way as in the SM. Contrary to the SM though, the right-handed fermions do not come as singlets of the weak gauge group but as SU (2) R copies of the SU (2) L doublets, that is the inverse seesaw, we also introduce three neutral fermion singlets, 1 with n an integer. As shown in ref. [20], when three left-handed and three right-handed neutrinos, ν αL and ν αR , are present and neutrino masses are generated via the inverse seesaw mechanism, three 1 Note that the fermion singlets SαR do not carry lepton number.
is also the minimum number of fermion singlets S αR required to reproduce the current neutrino data. At last, under the discrete C symmetry, the fermions transform according to 2 where the c on the right-hand side represents the usual charge conjugation transformation.
In the extended electroweak sector there exists a total of seven vector fields: W i Lµ , W i Rµ and B µ , with i = 1, 2, 3, associated with the gauge groups SU (2) L , SU (2) R and U (1) B−L , respectively.
These fields can be grouped into the covariant derivative acting on field multiplets. For example, the electroweak covariant derivative acting on SU (2) L(R) doublets, ψ L(R) , can be defined as where τ j are the Pauli matrices, g L = g R = g as a result of the LR symmetry, and P L(R) µ can be written in a 2 × 2 form as with √ 2W ± L(R)µ = (W 1 L(R)µ ∓ iW 2 L(R)µ ) and t = g B−L /g. The scalar sector is compact and contains the following singlet, SU (2) L and SU (2) R doublets and bi-doublet which transform under C as The Higgs mechanism takes place when the neutral components of such fields acquire non-trivial vevs, breaking spontaneously the initial symmetry in eq. (2) down to SU (3) c ⊗ U (1) Q in three steps. In the first step σ = v σ / √ 2 breaks the PQ symmetry. The second stage, i.e.
is achieved when η 0 R acquires a vev: v R / √ 2. Finally, the last step of the breaking process, is performed by the vevs of Φ: φ 0 We assume the following hierarchy In summary, in Table I we show the U (1) charges of the fermion and scalar fields, from which we can see that U (1) B−L is a gauged combination of U (1) B and U (1) L . It is also worth mentioning that, along the lines of refs. [21] and [22], discrete symmetries can be used to protected our axion solution from potentially dangerous U (1) P Q violating gravitational corrections.

III. THE SCALAR SECTOR
With the scalar content and symmetries described so far, we can write down the most general renormalisable scalar potential as plus a non-Hermitian term where f is a dimensionful coupling andΦ = τ 2 Φ * τ 2 . In order to find the scalar spectrum, we replace the field decompositions below in the scalar potential From the first derivative of the potential, the constraints below follow A possible solution for the system of equations above is the asymmetric one with v L = 0 and v R = 0. In this case, we also have Similarly, it allows us to rewrite the parameters µ 2 σ , µ 2 and µ 2 Φ in terms of the vevs and the dimensionless couplings. From eq. (16), assuming the dimensionless couplings to be of order one, and the vevs: The scale of the vev v σ is taken here to have a value within the interval where the axion -the pseudo Nambu-Goldstone boson of the PQ symmetry breakdowncan be a cold dark matter candidate [23] (see below).
At this point it is worth stressing that without the PQ symmetry, the scalar potential would contain terms such as η † L Φη R , resulting in an effective vev for η L . As we shall see in the next sections, v L = 0 (or at least v L negligible when compared to the other vevs) is paramount for the inverse seesaw mechanism to take place. Therefore, the PQ symmetry is intrinsically associated with the neutrino mass generation mechanism in our model. Previous studies involving the inverse seesaw in LR models, such as in refs. [8,14], have dealt with this issue differently. While in ref. [8], the authors avoid such an issue by assuming an asymmetric LR model in which η L is absent, in ref. [14] the authors argue that by keeping the dimensionful parameter f small (f 100 keV for v R 100 TeV), the effective vev η 0 L will be small enough to make sure that the inverse seesaw is realised.
With this particular solution that minimises the potential, we can finally derive the scalar particle spectrum. The scalar sector initially contains eighteen degrees of freedom. After symmetry breaking, six of those are expected to be absorbed by the gauge sector and make the vector bosons massive. Therefore, we are left with twelve physical scalar degrees of freedom. For simplicity, we present here only the mass states and eigenvalues, while all the squared mass matrices are shown in the Appendix A.
While the following charged fields (4 degrees of freedom) are absorbed by the charged gauge bosons W ± L and W ± R , the remaining charged scalar fields become massive with the following squared masses The (complex) neutral field η 0 L does not mix with the other real fields, and gets the following mass term We consider now the CP-odd fields A R , A σ , A 1 and A 2 . The first one is a Goldstone boson, which is absorbed by the neutral gauge boson Z R , as defined in the next section. The other three are mixed, and after diagonalisation, we find where G 2 is absorbed by the gauge sector, making the neutral gauge boson Z L , defined in the next section, massive; A is a CP-odd scalar with mass and a is the axion field that gets its mass from non-perturbative effects Finally, we take a look at the CP-even fields S σ , S 1 , S 2 and S R . These four real fields mix among themselves according to a 4 × 4 symmetric mass matrix. Making use of the fact that v σ is much larger than any other vev in the model, we consider the limit where S σ decouples from the other real fields. In such a limit, S σ gets the following mass Thus, the original 4 × 4 becomes a 3 × 3 matrix. To make its diagonalisation process simpler, we take the smallest vev, v 2 , to be zero. Upon this simplification, we find the following eigenmasses which are associated with the physical fields whose main contribution come from the fields S 1 , S 2 and S R , respectively. The lighter among the three real scalar fields is therefore identified with the SM Higgs field, with mass m S 1 125 GeV, whereas the other two are heavier Higgses.

IV. THE GAUGE BOSON MASS SPECTRUM
The gauge bosons become massive via the Higgs mechanism once the scalar fields acquire nonvanishing vevs. Their masses can be obtained from the following terms Because σ is a gauge singlet, D µ σ → ∂ µ σ, it does not contribute to the gauge field masses.
When the scalar fields acquire vevs, as described in eq. (14) with v L = 0, the charged gauge bosonsW ± L(R) = (W 1 L(R) ∓ iW 2 L(R) )/ √ 2 mix with each other. The mass eigenstates, W ± L and W ± R , can be defined as the linear combinations below and the associated masses are In From the expressions above, we conclude that v = v EW ≈ 246 GeV and W L is identified with the SM charged vector boson. Furthermore, taking, for instance, v R = 15 TeV, v 1 ∼ v EW = 246 GeV and v 2 ∼ 1 GeV, we have that m W R 4.9 TeV, and the mixing angle is very small: |ζ| 2.2×10 −6 .
For these values, m W R is above the current lower bounds coming from new particle searches at the LHC [5,6].
Let us consider now the real gauge bosons W 3 Lµ , W 3 Rµ and B µ . They mix among themselves and give rise to one massless and two massive neutral gauge bosons. The massless field is the photon field A µ ; one of the massive fields is the SM Z boson, while the other is a heavy neutral gauge boson Z : where θ W is the Weinberg angle, for which sin 2 θ W 0.231. The masses are given by and the mixing angle between the massive fields is defined as From eqs. (29) and (31), we have the relation m Z 1.2 × m W R . When assuming, for example, v R = 15 TeV, we find that m Z 5.8 TeV and ϕ 1.79 × 10 −4 .

V. FERMION MASSES AND MIXING
Considering the field content and symmetries presented in section II, the following renormalisable Yukawa terms can be written down with y l,q αβ = (y l,q αβ ) T . The first term provides all quarks with masses when the neutral fields in Φ acquire nonvanishing vevs. However, in this case both up-type and down-type quark mass matrices are proportional to y q αβ and, as a consequence, these matrices can be diagonalised simultaneously. Thus, the corresponding tree-level CKM matrix do not mix quark flavours, disagreeing with the experimental picture. This is a direct consequence of having an extra symmetry that distinguishes Φ from its charged conjugatedΦ and, as such, forbids the operator Q LΦ Q R from appearing in the tree-level Lagrangian above. In our case, this symmetry is U (1) P Q . Additionally, since the eq.
(33) does not present a Majorana mass term for the neutral fermion singlets S αR , some neutrinos remain massless.
These two issues can be dealt with by considering the following nonrenormalisable operators involving the scalar singlet σ with y l,q αβ = (y l,q αβ ) T , h αβ = h T αβ , n is an integer, kept arbitrary by now, associated with the PQ symmetry, see Table I, and Λ is a large mass scale suppressing the higher-dimensional operators. It is worth pointing out that since neither S αR nor σ carry lepton number, Lepton number violation occurs when η R gets a vev.

A. Quark masses: suppressed quark mixing
With the introduction of the higher-dimensional operators in eq. (34) in addition to those in eq.
(33), the up and down-type quark mass matrices, written according to Such mass matrices are no longer proportional to each other due to the contributions suppressed by . In this way our model predicts that the non-trivial mixing angles of the new CKM matrix vanish when the suppression factor approaches zero, providing thus a natural explanation to the "flavour alignment" puzzle.
The effective operators in eq. (34), leading to a small departure from alignment, are expected to come from a UV-complete theory upon the integration of heavy degrees of freedom. In ref.
There is enough freedom in the parameter space to easily accommodate all the quark masses and mixing. For instance, the values below can be used as benchmarks to fit the data: and the matrices D L , y q and h q are given in the Appendix B.

B. Lepton masses: inverse seesaw mechanism
For the charged leptons, we find We can take this matrix to be diagonal without loss of generality. Moreover, from eq. (36) we see that v 1 /v 2 1, and assuming, for simplicity, h l ij y l ij , we can take with When considering the solution in eq. (38), m D can also be taken diagonal and given in terms of the charged lepton masses Moreover, there is enough freedom to transform S αR to make µ diagonal For n large, the following hierarchy is obtained m R m D µ, which together with the texture of the mass matrix in eq. (39) allows for the inverse seesaw to take place.
Neutrino masses are then obtained by diagonalising the mass matrix in eq. (39) with the help of the unitary matrix U which can divided into two matrices: U = U 1 U 2 . Making use of the fact that m R m D µ, we can write U 1 as where a = m D (m R † ) −1 and b = µ(m R * ) −1 , and all the elements of the matrices above represent 3 × 3 matrices. When U 1 acts on M ν , the mass matrix is block-diagonalised where M ν is the mass matrix of the light neutrinos ν iL with masses m i , while the 3 × 3 entries M h ij form the 6 × 6 mass matrix of the heavy neutrinos N iR and S iR with masses M N i and M S i . At leading order, the lightest states, i.e. the active neutrinos, will have the following mass matrix Finally, for the last step, we define U 2 as where U is identified with the unitary PMNS mixing matrix that diagonalises M ν , while the 6 × 6 matrix formed by the V ij matrices diagonalises the heavy neutrino mass matrix so that 3 Considering v R = 15 TeV, v σ = 10 11 GeV, and the benchmark values in eq. (36), we have with d 1 = diag(2.8 × 10 −4 , 5.946 × 10 −2 , 1). Sub-eV neutrino masses follow naturally from eq.
(47) when n = 4 without the need for any unnaturally small coupling in y or h. On the other hand, for n < 4, sub-eV masses require h to be very small when compared to y; whereas for n > 4, y needs to be more and more suppressed with respect to h.
We set n = 4 and, in the Appendix C, provide a point in the parameter that shows that eq.
(47) can be easily used to fit the current neutrino data (masses and mixing). The solution presents normal mass ordering of the active neutrinos which is currently favoured over the inverted ordering case [25][26][27].
Finally, we would like to stress out the importance of the PQ symmetry for the generation of neutrino masses. The smallness of the active neutrino masses follow from their dependence on the PQ scale, v σ , through the suppression factor , rather than from a strong dependence on the largeness of the scale v R , as it is common when the type-I seesaw mechanism is in place. Therefore, all the current neutrino data can be successfully reproduced by our model even if the scale v R is only high enough to evade the constraints from new vector boson searches. This is in contrast to ref. [15], where the type-I seesaw mechanism is realised and the authors argue that only for values of v R ≥ 50 TeV can the current neutrino data be fitted.

VI. NEUTRINOLESS DOUBLE BETA DECAY
New contributions to 0ν2β arise in the LR model in processes where W R is exchanged instead of W L [28][29][30][31]. If the LR symmetry is broken at a very high scale, the new contributions can be neglected in favour of the standard one which is proportional to the Majorana effective mass m ee . In the present case, however, with v R = 15 TeV, W R as well as the heavy neutrinos will get masses around the TeV scale, so that the contributions to 0ν2β involving such particles may become relevant and therefore need to be checked.
In order to estimate the non-standard contributions, let us first write down, in the basis where the charged leptons are diagonal, the interactions in the lepton sector mediated by the charged gauge bosons W L and W R , where from the first to the second line, we have used eqs.  of the amplitudes below, respectively, where G F is the Fermi constant, m ee = i U 2 ei m i the effective Majorana mass, q 100 MeV the typical momentum exchange of the process, and M N i and M S i the masses of the heavy neutrinos N iR and S iR , respectively.
For the benchmark point in the parameter space given in the Appendix C, the effective Majorana mass is which is a typical value for m ee in the case of normal mass ordering. 4 By calculating and comparing all the contributions we have found that the standard is by far the dominant one, whereas among the non-standard contributions A λ is the largest, but it still very small when compared to A ν : |A λ /A ν | 1.26 × 10 −3 . Therefore, at least for this specific solution, since the dominant contribution is the standard one governed by small effective Majorana mass in eq. (50), no evidence for 0ν2β is expected to be observed in the next generation experiments which will scan the region | m ee | ∼ (0.01 − 0.05) eV [23,27].

VII. CONCLUSIONS
In this work, we have presented a minimal left-right model with a Peccei-Quinn symmetry and shown how it can help us to understand the origin of two intriguing features, namely, the approximately diagonal structure of the quark mixing (CKM) matrix and the smallness of neutrino masses.
The scalar sector is composed of a gauge singlet, a SU (2) R and a SU (2) L doublet, and a bidoublet: σ, η R , η L and Φ, respectively. Symmetry breaking occurs in three stages. Firstly, when the singlet acquires a vev, σ ∝ v σ = 10 11 GeV, the PQ symmetry is spontaneously broken.
TeV. At last, the SM group is broken by the two neutral components of Φ that acquire vevs satisfying the relation Step by step the Higgs mechanism is at play and, in the end, six would-be Goldstone bosons are absorbed by the gauge sector. The anomalous nature of the U (1) P Q symmetry, on the other hand, gives rise to a pseudo-Goldstone boson instead, the axion a, upon its spontaneous breaking by v σ . Furthermore, the strong CP problem is solved with the PQ mechanism, and the axion with mass given by eq. (23) becomes a good cold dark matter candidate.
Among the remainder eleven scalar degrees of freedom, there exist a very heavy CP-even field, S σ , with mass proportional to v σ , and S 1 , a 125 GeV Higgs boson, identified as the SM one. All the other massive scalars, including two charged fields, H ± 1 and H ± 2 , get masses proportional to the v R scale and can in principle be produced at colliders.
In the gauge sector of our LR model, we have twice as many massive gauge bosons as in the SM, as expected. With v R = 15 TeV, the new neutral and charged gauge bosons, with masses of 5.8 TeV and 4.9 TeV, respectively, are heavy enough to evade the current experimental limits.
In the fermion sector, we have shown that since the U (1) P Q symmetry distinguishes Φ fromΦ, with The active neutrino mass matrix in eq. (47) is diagonalised by the PMNS matrix (U) which can be parametrised in terms of three mixing angles: θ 12 , θ 13 and θ 23 , and three phases δ, α 21 and α 31 .
Below, we present a benchmark solution using the current best-fit values for the U parameters as given in ref. [23].