The double mass hierarchy pattern: simultaneously understanding quark and lepton mixing
aa r X i v : . [ h e p - ph ] J a n TTP14-029
The double mass hierarchy pattern:simultaneously understandingquark and lepton mixing
Wolfgang Gregor Hollik ∗ and Ulises Jes´us Salda˜na Salazar ∗ , † ∗ Institut f¨ur Theoretische Teilchenphysik, Karlsruhe Institute of TechnologyEngesserstraße 7, D-76131 Karlsruhe, Germany † Instituto de F´ısica, Universidad Nacional Aut´onoma de M´exicoApdo. Postal 20-364, 01000, M´exico D.F., M´exico
Abstract
The charged fermion masses of the three generations exhibit the two strong hierarchies m ≫ m ≫ m . We assume that also neutrino masses satisfy m ν > m ν > m ν and derive the consequences ofthe hierarchical spectra on the fermionic mixing patterns. The quark and lepton mixing matricesare built in a general framework with their matrix elements expressed in terms of the four fermionmass ratios m u /m c , m c /m t , m d /m s , and m s /m b and m e /m µ , m µ /m τ , m ν /m ν , and m ν /m ν ,for the quark and lepton sector, respectively. In this framework, we show that the resulting mixingmatrices are consistent with data for both quarks and leptons, despite the large leptonic mixingangles. The minimal assumption we take is the one of hierarchical masses and minimal flavoursymmetry breaking that strongly follows from phenomenology. No special structure of the massmatrices has to be assumed that cannot be motivated by this minimal assumption. This analysisallows us to predict the neutrino mass spectrum and set the mass of the lightest neutrino well below0 .
01 eV. The method also gives the 1 σ allowed ranges for the leptonic mixing matrix elements.Contrary to the common expectation, leptonic mixing angles are found to be determined solely bythe four leptonic mass ratios without any relation to symmetry considerations as commonly used inflavor model building. Still, our formulae can be used to build up a flavor model that predicts theobserved hierarchies in the masses—the mixing follows then from the procedure which is developedin this work. Keywords: quark and lepton masses and mixing, CP violation
PACS: E-mail: [email protected] E-mail: [email protected] ontents × V thCKM and J q . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 133.3 Lepton sector . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 143.4 About precision . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16 Introduction
The Standard Model of particle physics (SM) describes the interactions among elementary particles athigh energies with great success. In spite of this, the setup of the SM lacks an explanation of the originof fermion masses and mixing. In particular, for the quark sector, one observes six masses, three mixingangles and one phase. It is a simple exercise to relate the quark mixing matrix to the fundamentalparameters of the theory, the Yukawa couplings. Generally, however, it is said that mixing angles aswell as the masses are independent free parameters. Is there really no functional relation between thequark masses and the corresponding mixing matrix elements? There are many models in the literaturethat want to give an explanation of the mixing matrix elements in terms of the masses [1–28]. Mostof them put assumptions on a specific texture in the original mass matrices. We shall show, bycontrast, that the pure phenomenological observation of strong hierarchies in the quark masses leadsto a functional description of the mixing matrix elements in terms of mass ratios. The consequencesin the mixing of this phenomenological observation have already been studied [15, 20, 26, 29–33]. Ourapproach differs from the previous ones in many aspects: i) we take the Singular Value Decompositionof the complex mass matrices as a starting point offering a generic treatment for both quarks andleptons; ii) by means of an approximation theorem we mathematically formulate the steps to buildthe reparametrization of the mixing matrix in terms of the singular values (fermion masses); iii)we rotate the mass matrices in all three planes of family space whereas before, the 1-3 rotation wasneglected; iv) as the two unitary rotations in the 2-3 and 1-3 plane involve an approximation ( m f, = 0and m f, = 0, respectively) we consider for the first time a modified method of perturbation theory toadd the effect of the terms neglected; v) we do not consider the complex CP phases as free parametersand show that a minimal choice is sufficient to explain CP data; vi) we provide explicit formulae forthe mixing angles in terms of only mass ratios.The applicability of this formulation to the leptonic mixing is not clear a priori . First, neutrinomasses do not show any strong hierarchy, at best a very mild one. Second, the leptonic mixing matrixexhibits large mixing, while the one in the quark sector is rather close to the unit matrix. Thispicture seems to suggest two quite different origins for the respective mixing matrices: quark massesstrongly dominating the mixing patterns, whereas geometrical factors found from symmetries shapingthe leptonic mixing, with only a weak intervention from the lepton masses [34, 35].Fermion masses, on the other hand, are also as puzzling as the mixing matrices: the top quarkmass is by far the largest among the charged fermions, there are six orders of magnitude separating thetop quark from the electron mass, six orders of magnitude separating the largest neutrino mass fromthe electron mass (assuming a neutrino mass scale of 0 . f = u, d, e ),the masses follow a hierarchy m f, ≫ m f, ≫ m f, , m u : m c : m t ≈ − : 10 − : 1 , m d : m s : m b ≈ − : 10 − : 1 ,m e : m µ : m τ ≈ − : 10 − : 1 , (1)while the two squared mass differences measured from neutrino oscillations obey a much weakerhierarchy, ∆ m : ∆ m ≈ − : 1 . (2)Quark masses plus mixing parameters sum up to ten arbitrary physical parameters in the SM.Consideration of neutrino masses, whether Dirac or Majorana, adds at least ten more parameters tothe count. Two more complex phases and a possibly arbitrary number of masses for sterile neutrinos2ppear in the more general cases including Majorana neutrinos [36]. The SM per se seems to lack acourse of action on how to relate the mixing matrix elements to the corresponding fermion masses.The first realization of a mixing angle in terms of the masses is commonly assigned to Gatto etal. [1] which is referred to as the Gatto-Sartori-Tonin relation. This relation is an expression of theCabibbo angle commonly written as θ q ≈ r m d m s , (3)where originally, the authors of [1] found a similar relation in terms of light meson masses from thedemand of weak self-masses being free from quadratic divergences. In a footnote, they break it downto an elementary discussion in a “naive quark model” and statetan θ = m n − m p m p = m n m λ , (4)where m n , m p , and m λ are the old notations of down-, up-, and strange-quark masses (moreover, thesecond equal sign was misleadingly written as a minus sign). The first work referring to [1] as origin of“tan θ = m n /m λ ” was [3] (even though with a typo in the abstract). For small angles, tan θ ≈ θ andwe are at Eq. (3). Since p m d /m s is an astonishingly good approximation for the Cabibbo angle, wewill show in the course of this paper how to rearrive at this expression in a formal way of parametrizingmixing matrices in terms of invariants.The work of [1] was followed by derivations of the same formula focused on the derivation in a moremodel-building related approach using left-right symmetric scenarios [1–5,12,37]. In the same decade,a model independent approach was initiated where mass matrices with different null matrix elements(“texture zeros”) were considered [38–43]; similar relations were then found for other mixing angles.Subsequently, horizontal or family discrete symmetries were used in order to relate the three familiesin a non-trivial fashion [11, 13, 14, 44–48]. In their initial stage, though, the experimental uncertaintyin the mixing angles and fermion masses was still too large as to build a stable model consistentwith the unstable phenomenology. This approach was vigorously resurrected in the last decade whenprecision measurements for neutrino oscillations started [35, 49, 50]. Relations between the neutrinomixing angles and lepton mass hierarchies were found [51, 52] where the values for the three neutrinomasses are compatible with what follows from our method, though θ was predicted too low (onlyabout 3 ◦ ). Nevertheless, up to now, no complex mass matrix with a well-motivated constrained setof parameters has been found to entirely and successfully postdict the Cabibbo-Kobayashi-Maskawa(CKM) quark mixing matrix or to predict the Pontecorvo-Maki-Nakagawa-Sakata (PMNS) matrix inthe lepton sector. In this work, we do not focus on a specific model predicting mixing angles, but giveexplicit relations following from a model independent treatment based on the observation of the twostrong hierarchies m ≫ m ≫ m in the charged fermion masses. Moreover, we dare to apply thesame fomulae to the neutrino mixing and derive the PMNS angles with astonishingly good agreement.This paper is organized in the following way: first, we start discussing the generic treatmentof mixing matrices following from hierarchical mass matrices in Section 2, where we focus on themathematical derivation of relations among fermion mass ratios and mixing angles. This result getsapplied to the phenomenological data in Section 3. Finally, we conclude. In the appendices, wereview the current status of input data, give a brief statement about the applicability of the methodelaborated in this work, comment on the hierarchical structure of the mass matrices as a consequenceof hierarchical masses and minimal flavor symmetry breaking, and provide the explicit, approximativeformulae that gave the results of Section 3. 3 Mass and mixing matrices
Let us extend the SM by three right-handed neutrinos to have a more symmetric treatment of theproblem in the quark and lepton sector. Dirac neutrinos alone still leave the question open why theYukawa couplings for neutrinos are so much smaller than for the charged fermions. Nonetheless, inthe description of fermion mixings in terms of fermion masses this assumption does not play a rˆoleand later we take an effective neutrino mass matrix without the need to specify whether neutrinosare Dirac or Majorana. The most general, renormalizable and gauge invariant construction of fermionmass matrices follows from the Yukawa Lagrangian − L Y = X f = d,e Y ijf ψ fL,i Φ ψ fR,j + X f = u,ν Y ijf ψ fL,i ( iσ Φ ∗ ) ψ fR,j + H.c. , (5)where i, j = 1 , , ψ f , where the left-handed fermions are grouped into SU(2) L doubletsand the right-handed ones are the usual singlets. The Higgs doublet is given by Φ = ( φ + , φ ) whereasits nonvanishing vacuum expectation value v = h φ i = 174 GeV. The spontaneous breakdown ofelectroweak symmetry gives rise to four Dirac mass matrices of the form M f = v Y f . (6)These mass matrices are 3 × D f = L f M f R f † , (7)where D f is a diagonal matrix with real and positive entries while L f and R f are two unitary matricesacting in family space on left- and right-handed fermions of type f respectively. Both transformations, L f and R f , correspond to the unitary matrices appearing in the Singular Value Decomposition of M f .These unitary matrices transform the sets of three left- or three right-handed fermion fields each fromthe interaction basis to the physical mass basis ψ ′ f,L = L f ψ f,L and ψ ′ f,R = R f ψ f,R . (8)The mass eigenstates are therewith ψ ′ f . In return, the diagonal weak charged current interactions areno longer diagonal, and mix different fermion families. This occurs as a consequence of the mismatchbetween the two different left unitary matrices acting inside the same fermion sector which results inthe observable mixing matrices in the charged current interactions V CKM = L u L d † and U PMNS = L e L ν † . (9) The singular values of the diagonal matrix D f in Eq. (7) are to be identified with the measured fermionmasses (see A). An interesting and not yet exploited fact is that the observed hierarchies in the masses(singular values) can be used to approximate the original mass matrices by lower-rank matrices asstated in the Schmidt-Mirsky approximation theorem [53–56]. The left and right unitary matrices, L f and R f are decomposed into the left and right singularvectors, l f,i and r f,i ( i = 1 , , L f † = [ l f, , l f, , l f, ] and R f † = [ r f, , r f, , r f, ]. Each It is often wrongly called the Eckart-Young-Mirsky or simply Eckart-Young theorem, see [57] for an early history onthe Singular Value Decomposition. m f,i . For square matrices when all threesingular values can be ordered as m f, > m f, > m f, ≥
0, the decomposition is unique up to a sharedcomplex phase for each pair of singular vectors. The number of non-zero singular values equals the rank of the mass matrix M f . The mass matrixcan be written in terms of its singular values with the respective left and right singular vectors as asum of rank one matrices, M f = (cid:20)(cid:18) l f, m f, m f, r † f, + l f, r † f, (cid:19) m f, m f, + l f, r † f, (cid:21) m f, . (10)Any hierarchy among the singular values is of major interest to us as it leads to a lower-rank approx-imation M rf ( r = rank[ M rf ] < m f, ≫ m f, ≫ m f, , Eq. 10 provides a powerful way to appreciate the double hierarchyof its singular values and the emerging relation to its rank by the use of Schmidt-Mirsky’s approxi-mation theorem. As both types of quarks and charged lepton masses satisfy those two hierarchies, weconclude, that their mass matrices can be safely approximated as either matrices of rank one or ranktwo, depending on how strong their double mass hierarchy pattern (DMHP) is.As illustrated in Eq. (10), this expression points also to the fact that the fermion mass ratios m f, /m f, and m f, /m f, play the dominant rˆole in determining the structure of the mass matrixwhereas m f, sets the overall mass scale. Only those two ratios will be necessary in the determinationof the mixing parameters, since the overall mass scale can be factored out. For later use, we abbreviateˆ m f, = m f, /m f, and ˆ m f, = m f, /m f, . In the following, the hat (ˆ) denotes the division by thelargest mass m f, . The four mass ratios parametrization
The fact, that only two mass ratios for each fermionspecies are independent parameters, gives four independent mass ratios in each sector (quarks and lep-tons). An important remark at this point is, that also four parameters are needed to fully parametrizethe mixing. This observation shall be used to build up the mixing matrix. In the standard parametriza-tion, those four values are three angles and one phase—additional phases are to be rotated away byredefinition of the fermion fields. The case of Majorana neutrinos does not allow to rotate away thephases for the neutrinos, so two “Majorana phases” are left. In the following, we will leave aside theissue of Majorana phases and only discuss the Dirac phases. We shall show that it is possible to usethe four mass ratios of each fermion sector to entirely parametrize the mixing without introduction ofnew parameters.It is interesting to note, that a complete parametrization of the fermion mixing in terms of thefermion mass ratios only works in the two- and three-family case. To completely parametrize themixing matrix, for n > n − mixing parameters. On the other hand, n − n − ≥ ( n − , this parametrization will be possible. In general, this only works out for two orthree families. In the case of degeneracy among some of the singular values, there is no longer a unique Singular Value Decompositionfor M f . This matters in the discussion of degenerate neutrino masses. .2 The lower-rank approximations Let us investigate the effect of neglecting the first generation masses. From now on we will work withthe singular values normalized by the largest one. In the ˆ m f, → O ( ˆ m f, ) we shall take into account all corrections of the sameorder later on to get a more precise result and reduce the error stemming from this approximation.The rank two mass matrices are then given byˆ M r =2 f = h l f, ˆ m f, r † f, + l f, r † f, i = m f ˆ m f m f ˆ m f . (11)In general, all the matrix elements should be different from zero. However, it is crucial to establisha connection between a lower-rank approximation and its origin to the Yukawa interactions. Thatis, ˆ m f, = 0 is equivalent to decoupling the first fermion family from the Higgs field, Y f j = 0 = Y fj .Effectively, thus, we are left with a 2 × m f, ≫ m f, to decouple the first generation masses.According to the lower-rank approximation theorem, the rank-two approximation differs in everyelement from the full rank matrix, whereas its norm, for any chosen one, only changes slightly. TheDMHP furthermore shows m f, ≫ m f, which can be exploited to further approximate the initialmass matrix by a rank-one matrix,ˆ M r =1 f = l f, r † f, = . (12)Successively reducing the rank of the mass matrices helps to simplify the parametrization withoutloosing track of the parameters. It is, however, not necessary to work in the very crude rank oneapproximation, but sufficient to consider as a starting point the rank two approximation.Eq. (12) reveals a left-over U(2) rotation in the 1-2 plane and one common U(1) factor for the thirdgeneration. We want to emphasize that the described picture of lower-rank approximations followswhat is discussed in the literature as minimally broken flavor symmetry [22, 58, 59]. In the limit ofvanishing Yukawa couplings, the SM exhibits a [U(3)] global flavor symmetry ([U(3)] if right-handedneutrinos are considered). Each individual U(3) flavor symmetry gets gradually brokenU(3) M −→ U(2) M −→ U(1) M −→ nothing , with M > M > M which simultaneously occurs in the up- and down sector and trivial U(1)s areleft out for readability. After the first symmetry breaking step at M , one global phase freedom is leftfor the third generation that is combined to a global U(1) for the second and third after the followingsymmetry breaking. There is one residual U(1) symmetry left for all fermions in each sector at theend which is either baryon or lepton number. It is not only safe to work with M ≫ M —where weare at the U(2) flavor symmetries of [22, 58, 59], but even M ≫ M which allows to work with therank-two approximation at a sufficiently low scale and perform the final symmetry breaking step atsay the electroweak scale.U(2) symmetric Yukawa couplings give a well-motivated and frequently used setup to study flavorphysics in supersymmetric [60, 61] and unified [22] theories and are still a viable tool to discuss recentresults in flavor physics [62, 63]. Application to lepton flavor physics was also considered [64–67],6ecently also in the context of [U(3)] breaking [68]. The implication of U(2) flavor symmetries whichcan be used in a weaker symmetry assignment [69], is the arrangement of the first two families into onedoublet whereas the third family transforms as a singlet under the flavor symmetry. This assignmentcan be achieved with the minimal discrete symmetry S [70–74] that was applied to neutrinos [75] aswell as quarks [28].The important point in the discussion of fermion mixings in terms of fermion masses via lower-rank approximations is, that we implicitly assume the maximal [U(3)] flavor symmetry broken witheach symmetry breaking step occurring simultaneously for each subgroup [U(3)] = U(3) Q × U(3) u × U(3) d × U(3) L × U(3) e × U(3) ν . Order of independent rotations
To parametrize the three-fold mixing, we follow the commonlyused three successive rotations depending on one angle and one phase each. The order of thesetransformations needs to follow the consecutive breakdown of the initial U(3) symmetry as implied bythe hierarchy in the masses. Therefore, L f = L f ( θ f , δ f ) L f ( θ f , δ f ) L f ( θ f , δ f ) , (13)where each individual rotation is parametrized by one angle θ fij and one phase δ fij . Note that this set of rotations diagonalize the mass matrices for each fermion type. The resultingmixing matrices are the product of all the individual rotations V CKM = L u L d † = L u L u L u L d † L d † L d † and U PMNS = L e L ν † = L e L e L e L ν † L ν † L ν † . By convention, up- and down-type rotations are exchanged for leptons. × It is instructive to first study the two-family limit in the rank-two approximation following fromˆ m f, ≪
1. The second hierarchy m f, ≪ m f, implies a 2 × m f = ˆ m fss ˆ m fsl ˆ m fls ˆ m fll ! , (14)with hierarchical elements | ˆ m fll | ≫ | ˆ m fsl | , | ˆ m fls | ≫ | ˆ m fss | and where we are now generically treatingtwo fermion families whose singular values obey the hierarchy, σ l ≫ σ s . In general, the matrix elementsare complex numbers. The labelling s and l refers to the corresponding smaller and larger singularvalue, respectively. It can be shown that the order of magnitude of ˆ m fss is about O ( | ˆ m fsl | ) (see C). Inthe following, we work with the approximation ˆ m fss = 0.Unlike most considerations, we take the outcome of the DMHP and minimal flavor symmetrybreaking to set the magnitudes of the off-diagonals equal—the phases are not constrained, such that | ˆ m fsl | = | ˆ m fls | not ˆ m fsl = ( ˆ m fls ) ∗ , as implied by the requirement of an Hermitian mass matrix. We only need normal mass matrices. In both cases (normal and Hermitian), the left and right Hermitian products are diagonalized by the Later, when reparametrizing the individual rotations in terms of the masses we will see that some of these six mixingparameters are unphysical while the rest can be expressed solely by two mass ratios. A matrix is normal if the left and right Hermitian products are the same: m m † = m † m . m f = | ˆ m fsl | e iδ fsl | ˆ m fsl | e iδ fls ˆ m fll ! . (15)As a self-consistency check, it is important to verify that the required hierarchy in all the massmatrix elements of the full-rank scenario actually is respected when expressing the matrix elements interms of the masses (singular values). Reparametrization in terms of the singular values
Due to our lack of knowledge of right-handed flavor mixing, the relevant object that determines our phenomenology is the Hermitian product n f = m f ( m f ) † , which exhibits two invariants: tr n f = σ f s + σ f l and det n f = σ f s σ f l . The small andlarge singular value are denoted by σ fs and σ fl , respectively. Through means of the two invariants, wefind | ˆ m fsl | = q ˆ σ fsl , and | ˆ m fll | = 1 − ˆ σ fsl , (16)where we have expressed for a generic treatment the normalized ratio of the small singular value overthe large one as ˆ σ fsl ≡ σ fs /σ fl .This reparametrization nicely shows the result of the Schmidt-Mirsky approximation theorem: onthe one hand, | ˆ m fll | ≫ | ˆ m fsl | , while on the other hand, | ˆ m fll | = 1 is the only non-vanishing matrixelement in the limit ˆ σ fs → L fsl (ˆ σ fsl , δ fsl ) = 1 q σ fsl e − iδ fsl q ˆ σ fsl − e iδ fsl q ˆ σ fsl . (17)This result has been already discussed previously by many authors [8, 9, 18, 26]. The mixing angle canbe obtained from tan θ fsl = q ˆ σ fsl . Note that this relation indeed is the Gatto-Sartori-Tonin result,see Eq.(3).
The two-family mixing matrix
Eq. (17) diagonalizes the mass matrix of one fermion type. Inthe weak charged current, an a -type fermion ( a = u, e ) meets a b -type fermion ( b = d, ν ), so we needtwo such diagonalizations to describe fermion mixing in the charged current interactions. Anyway,two unitary 2 × θ sl = θ asl ± θ bsl and δ = δ asl ± δ bsl . Explicitly, V sl = L asl L bsl † = diag(1 , e − iδ asl ) (cid:18) √ − λ e − iδ λe − iδ − λe iδ √ − λ e iδ (cid:19) diag(1 , e iδ asl ) , (18)where we factored out the phase δ asl . This choice is completely arbitrary, the same is true for δ bsl . Therelevant phases inside the matrix only depend on the difference . The mixing can then be obtained in Another solution can be found, that behaves wrongly in the limit ˆ σ fsl → θ fsl → ∞ instead of zero mixing. λ = sin θ sl = vuut ˆ σ asl + ˆ σ bsl − q ˆ σ asl ˆ σ bsl cos( δ asl − δ bsl )(1 + ˆ σ asl )(1 + ˆ σ bsl ) , (19)tan δ = ˆ σ bsl sin( δ asl − δ bsl )ˆ σ asl − ˆ σ bsl cos( δ asl − δ bsl ) , (20)tan δ = ˆ σ asl ˆ σ bsl sin( δ asl − δ bsl )1 + ˆ σ asl ˆ σ bsl cos( δ asl − δ bsl ) . (21)The functional dependence on the two initial complex phases is found to be only their difference. Fromthe hierarchies ˆ σ xsl = σ xs /σ xl ≪ x = a, b ) follow the new phases to be approximately given bytan δ ≈ − tan( δ asl − δ bsl ) and tan δ ≈
0. For the full-rank scenario, however, this simple conclusioncannot be drawn—it actually holds for the “initial” 2-3 rotation, but not anymore when subsequentrotations are added.
Comment on the complex phases
In general, the complex phases of the initial mass matrixelements are not constrained to a particular value. The employed matrix invariants only restrict themoduli of the matrix elements, the phases are unconstrained. There is nevertheless an ambiguityin those phases that is not necessary to set up a full parametrization of fermion mixing in the SM.The standard parametrization uses three successive rotations with θ ij ∈ [0 , π ] and one complex phase δ CP ∈ [0 , π ). These four parameters are sufficient to describe both mixing and CP violation in eachfermion sector (unless we want to include a description of Majorana phases for neutrinos). In contrast,we have four mass ratios—and the freedom to put either real or purely imaginary matrix elements.This last choice can be achieved by restricting all phases to be either maximal CP violating ( π/ π/
2) or CP conserving (0 or π ). Interestingly, at the end, there is no freedom in phase choices atall and we find that only the 1-2 phase is allowed to be maximally CP violating, which indeed followsfrom a symmetry argument. Working in the lower-rank approximations, we are neglecting the first generation mass ( ˆ m f, = 0) inthe 2-3 rotation and the second generation mass ( ˆ m f, = 0) while performing the 1-3 rotation. Thelast transformation that appears in Eq. (13) acting in the 1-2 plane needs no approximation. It affectsonly the upper left 2 × × × O ( ˆ m f, )terms in the 1-3 rotation means actually ignoring a large effect, because O ( ˆ m f, ) = O ( ˆ m f, ). Moreover,to include O ( ˆ m f, ) contributions in the 1-3 rotation following the initial rotation in the 2-3 plane,we first have to consider contributions of the same order that were missing in the initial rotation.Therefore, we briefly discuss how to consistently include corrections of missing pieces to improve theresult. Inclusion of corrections
We include the corrections as correcting (small) rotations. This procedureis crucial in view of the symmetry breaking chain from an enhanced flavor symmetry, as [U(3)] (corresponding to a rank-zero mass matrix), down to the least symmetry left over. Since each breakingstep is done by a small parameter, we do not disturb much by adding perturbations. Moreover, by9epeatedly applying rotations, this guarantees from the very beginning normalized eigenvectors, andfurthermore, an inclusion of formally higher order terms in perturbation theory. This can be seenfrom the following example of two real rotations, where ˆ ǫ . ˆ σ fsl : L fsl ( p =1) = L fsl (1) ( ± ˆ ǫ ) L fsl (0) (ˆ σ fsl ) = cos θ fsl ( p =1) sin θ fsl ( p =1) − sin θ fsl ( p =1) cos θ fsl ( p =1) ! , (22)and the new angle is given by sin θ fsl ( p =1) = q ˆ σ fsl ± √ ˆ ǫ q (1 + ˆ σ fsl )(1 + ˆ ǫ ) . (23)For real rotations, the requirement ˆ ǫ . ˆ σ fsl is irrelevant, because Ø(2) rotations commute. Therefore,there is also no need to specify any order in the addition of correcting rotations in each i - j plane.Inverting this procedure shows that it is equivalent to add the perturbation term − √ ˆ ǫ (cid:20) σ fsl ) − σ fsl + q ˆ ǫ ˆ σ fsl (ˆ σ fsl − (cid:21) (1 − ˆ ǫ ) (24)to the off diagonal matrix elements s - l and l - s .Continuing this, an arbitrary number of correcting rotations could be added in each 2 × θ fsl ( p = n ) = P nj =0 ( − δ j p ˆ a j + O (cid:16)(cid:2)p ˆ a i ˆ a j ˆ a k (cid:3) i = j = k (cid:17)p (1 + ˆ a )(1 + ˆ a )(1 + ˆ a ) · · · (1 + ˆ a n ) , (25)where we have denoted ˆ a ≡ ˆ σ fsl and ˆ a i> for the parameters of the following rotations. Each ( − δ i isthe orientation of the i -th rotation, which is either clockwise or counterclockwise (plus or minus). Weneglect in Eq. (25) all trilinear and higher products of ˆ a i , where no ˆ a i and no even products appear. Letus emphasize here, nevertheless, that these correcting rotations do not follow the traditional procedureof perturbation theory where we could naively think that the following new correcting rotation is apower of the previous one. Inclusion of new correcting rotations requires a careful treatment. Wehave found to be sufficient to include two correcting rotations to the mixing matrix parametrizationwhich are the contributions O ( ˆ m f, ), O ( ˆ m f, ), and O ( ˆ m f, · ˆ m f, ) which are of the same order as theneglected terms in each case. First rotation: The 2-3 sector
Starting from the rank-two approximation, we loose track of all p ˆ m f, contributions in the mass matrix. However, all correcting rotations have to be consistent withthe initial approximation ( ˆ m f, →
0) and, moreover, all “higher order” contributions ( ∼ ˆ m f, , ∼ ˆ m f, )are already covered as can be seen from (24). We therefore conclude, that all reasonable rotations inthe 2-3 plane can be expressed as L f p =2) = L f ( ˆ m f, · ˆ m f, ) L f ( ˆ m f, ) L f ( ˆ m f, ) . (26)Additionally, in principle, there is a freedom in the choice of the complex phase, which can be boileddown to the two different sign choices. The two signs reflect the freedom of choice for a clockwise or counterclockwise correcting rotation. econd rotation: The 1-3 sector What follows is the same procedure in the 1-3 sector after the2-3 rotations have been done. In this case, the p = 2 leading correcting rotations are L f p =2) = L f ( ˆ m f, · ˆ m f, ) L f ( ˆ m f, ) L f ( ˆ m f, ) . (27) Last rotation: The 1-2 sector
No approximation is left anymore, therefore the exact rotation isexpressed as L f = L f ( ˆ m f, ˆ m f, , δ f ) , (28)where we now explicitly put the phase δ f . This occurrence is very clear from the rank evolution: inthe rank-one approximation, there is the freedom of a U(2) rotation left in the 1-2 block. The initial2-3 and 1-3 rotations can always be taken real, the only possible phase then sits in the 1-2 rotation.The necessity of correcting rotations is very apparent from the flavor symmetry breaking chain:First, in the rank-two approximation we have X X X X L (0)23 −→ X
00 0 X . After performing the symmetry breaking step to the full-rank matrix, we get contributions in all matrixelements not larger than O ( p ˆ m f, )—also in off-diagonal components that were already rotated away: ∗ ∗ ∗∗ X ∗∗ ∗ X . So, we indeed have to consider higher order corrections to the initial rotation. The correcting rota-tions also do not spoil the required hierarchy. After the successive 2-3 and 1-3 rotations there is acontribution shuffled into the 1-1 entry which is ∼ s m ∼ O ( ˆ m f, ) and therefore of higher ordercompared to O ( ˆ m / ˆ m ), the original 1-1 element. By building up the mixing matrices following the procedure of the previous section, there appears theimpression of an arbitrariness in the choice of complex phases. This arbitrariness can be attenuatedtaking into account some well motivated considerations. First, complex phases appear pairwise in theup- and down-type fermion sectors. We therefore have the freedom to keep track of them in onlyone sector and set all phases in the other one equal to zero. The charged current mixing matrix is11herefore constructed in the following way: V CKM = L u L d † ,L u = L u (cid:18) m u m c (cid:19) L u (cid:18) m u m c m t (cid:19) L u (cid:18) m c m t (cid:19) L u (cid:18) m u m t (cid:19) L u (cid:18) m u m c m t (cid:19) L u (cid:18) m u m t (cid:19) L u (cid:18) m c m t (cid:19) , (29) L d † = L d † (cid:18) m s m b , δ (0)23 (cid:19) L d † (cid:18) m d m b , δ (1)23 (cid:19) L d † (cid:18) m d m s m b , δ (2)23 (cid:19) × L d † (cid:18) m d m b , δ (0)13 (cid:19) L d † (cid:18) m s m b , δ (1)13 (cid:19) L d † (cid:18) m d m s m b , δ (2)13 (cid:19) L d † (cid:18) m d m s , δ (cid:19) , (30) U PMNS = L e L ν † ,L e = L e (cid:18) m e m µ (cid:19) L e m µ m τ ! L e (cid:18) m e m µ m τ (cid:19) L e (cid:18) m e m τ (cid:19) L e (cid:18) m e m µ m τ (cid:19) L e (cid:18) m e m τ (cid:19) L e (cid:18) m µ m τ (cid:19) , (31) L ν † = L ν † (cid:18) m ν m ν , δ (0)23 (cid:19) L ν † (cid:18) m ν m ν , δ (1)23 (cid:19) L ν † (cid:18) m ν m ν m ν , δ (2)23 (cid:19) × L ν † (cid:18) m ν m ν , δ (0)13 (cid:19) L ν † (cid:18) m ν m ν , δ (1)13 (cid:19) L ν † (cid:18) m ν m ν m ν , δ (2)13 (cid:19) L ν † (cid:18) m ν m ν , δ (cid:19) . (32)The method itself is not quite arbitrary at all. For the CKM mixing it gives well-separated regionsthat have to be entered with a specific choice for the phases (see Fig. 1). Since both quark masses aswell as CKM mixing matrix entries are rather well measured, this observations allows us to set thephases. We find only one distinct choice. Moreover, we make a minimal choice: on the one hand, weallow CP phases to be either maximally CP violating or CP conserving. On the other hand, we find,that the only maximally CP violating phase has to be in the 1-2 rotation of the down-type quarks orneutrinos, respectively. This can be seen from Fig. 2 where the three bands correspond to a phase δ = 0 , π and π .The previously derived subsequent rotations only depend on four mass ratios in each fermion sectorand have to be faced with phenomenological data. As input values we are using the quark and leptonmasses only (see A) and then give a prediction for the neutrino masses to be in agreement withobservations of neutrino mixing in this setup. The nature of the complex phases and its impact in the mixing matrix elements needs further inves-tigation. Giving a solution to this problem is, however, outside the scope of this work. We shall useour observation to distribute the CP violating phase properly and leave the origin of CP violation forlater work.A final comment can be done, though, that guarantees the uniqueness of the parametrization. InFig. 1, we show the maximally allowed ranges for the mixing matrix elements V ub and V cb . The amountof data points was constructed choosing the quark masses from their 1 σ regimes and randomly takingevery phase in the final paramerization from the set { , π , π } . It is sufficient to constrain oneself tothis set which gives the minimal and maximal allowed amount of CP violation [76]—and connectedto that minimal and maximal mixing. The latter can be seen from Eq. (19) for the two-generationsub-case: the phase difference δ asl − δ bsl controls the magnitude of the mixing angle between minimal( δ asl − δ bsl = 0) and maximal ( π ) mixing.The fact, that only one combination of phases survives, is astonishing: note that all possiblecombinations in Eq. (30) are generically 3 = 2187 choosing from { , π , π } . Still, after taking δ = π V c b V ub V c b V ub Figure 1: Distribution of allowed values in the V ub - V cb plane. The small red points show allowedregions where the masses were varied in their 1 σ regimes, the blue crosses show the values comingfrom the central values of the masses. Right: zoom into the phenomenological viable region. Thereare only three distinct phase choices leading to both small values for V ub and V cb .Table 1: The choice of phases in Eqs. (30) and (32) leading to the mixing matrices shown in (33) and(38). δ δ (0)13 δ (1)13 δ (2)13 δ (0)23 δ (1)23 δ (2)23 CKM π π π π π PMNS π π π π π π , 64 combinations are left. It is thereforenot a priori clear that the mass ratios alone give the right mixing. The functional dependence on themass ratios, however, is unique once the phases are set. We therefore use this description to determinethe position of the maximal CP phase, where in contrast the other phases give relative minus signs.The maximal CP violating phase in the neutrino 1-2 mixing is somehwat different to what was foundin connection with maximal atmospheric mixing [77]. V thCKM and J q Consideration of all the aforementioned prescriptions gives the following numbers for the magnitudeof the mixing matrix elements (see D for the explicit formulae of the mixing angles and the Jarlskoginvariant), | V thCKM | = . +0 . − . . +0 . − . . +0 . − . . +0 . − . . +0 . − . . +0 . − . . +0 . − . . +0 . − . . +0 . − . (33)and the following amount of CP violation as measured by the Jarlskog invariant, J q = Im( V us V cb V ∗ ub V ∗ cs ) = (2 . +1 . − . ) × − , (34)where all quantities here are seen to be in quite good agreement within the errors compared to theglobal fit result given by the PDG 2014 [78] (see A for present knowledge on masses and mixings).Note that generically, the amount of CP violation is much larger (Fig. 2) and a small value of V ub isconnected to a small J q , as expected. 13 ub -0.004-0.003-0.002-0.001 0 0.001 0.002 0.003 0.004 0 0.01 0.02 0.03 0.04 0.05 0.06 0.07 0.08 J q V ub Figure 2: The left plot shows the regions for the Cabibbo angle (more exactly V us —clearly the solution δ = π is favoured (which corresponds to the stripe in the center). On the right side, the rephasinginvariant J q is shown against V ub . Color code as in Fig. 1: red dots are points with masses varied inthe 1 σ regimes, blue crosses are the central values. Quark masses show a very strong hierarchy. Charged lepton masses also do. Neutrinos, though, donot do. Is it really viable to apply the DMHP also to lepton mixing? Leptonic mixing angles arelarge, this observation may hint to a different mechanism. However, mass ratios for neutrinos are alsolarge. The parametrization of fermion mixing in terms of mass ratios allows to also cope with largemixings by large mass ratios. Nevertheless, we have to include a solid examination of the errors inthis approximation and see whether the same procedure as for quarks is viable also for leptons.Are neutrino masses hierarchical? Neither the quasidegenerate solution nor the strong hierarchyare excluded yet. A hierarchical mass spectrum in any case predicts a very light lightest neutrino(it still can be exactly massless—in this case we would only have a rank two mass matrix), wheredegenerate masses are likely to be tested in the near future.The power of the mixing parametrization in terms of mass ratios lies in its invertibility: theformulae give us a unique description of the missing mass ratio once the mixing angle is measured.The pattern of neutrino masses brings us into the comfortable situation of nearly disentangling the 1-2from the 2-3 mixing, because ∆ m / ∆ m ≪
1. Additionally, the 1-2 mixing angle has the smallesterror in the global fit.
Predicted neutrino masses
We do not focus on a specific model behind the theory of neutrinomasses. It is sufficient to consider an effective neutrino mass matrix irrespective of the UV completionbehind. To embed our description into a theory of neutrino flavor, it definitely matters if neutrinosare Dirac or Majorana. The size of the masses, however, allows to neglect RG running in any case.Therefore, we also ignore the nature of the neutrino mass operator. Since we take the magnitudes ofthe Dirac masses symmetric for quarks, the only difference would be the off-diagonal phase. Havingthis similarity in mind, the 1-2 approximation for neutrinos follows directly from Eq. (18) and thedetermining equation for the missing mass ratio from Eq. (19) with obvious relabelings: | U e | ≈ s ˆ m eµ + ˆ m ν − p ˆ m eµ ˆ m ν cos( δ e − δ ν )(1 + ˆ m eµ )(1 + ˆ m ν ) , (35)14here the mass ratios are ˆ m eµ = m e /m µ and ˆ m ν = m ν /m ν . The three individual neutrino masses are obtained via the mass squared differences: m ν = q ∆ m / (1 − ˆ m ν ) ,m ν = q m ν − ∆ m ,m ν = q ∆ m − ∆ m + m ν . (36)In Eq. (35), there appears the phase difference δ e − δ ν . Although a twofold rotation shows no CPviolation, this phase has to be considered because it appears last in the order of successive rotations.Moreover, we observed a maximal CP phase in the quark 1-2 sector. Albeit there is no connectionbetween quark and lepton mixing at this stage, we shall keep the assignment δ e − δ ν = π and getˆ m ν = | U e | (1 + ˆ m e ) − ˆ m e − | U e | (1 + ˆ m e ) = 0 . . . . .
45 (37)using ˆ m e = 0 . | U e | = sin θ = 0 . . . . .
56. The masses are calculated as m ν = (0 . ± . ,m ν = (0 . ± . ,m ν = (0 . ± . . The errors were propagated from the ∆ m and added linearly to be more conservative. Within 3 σ ,the lightest neutrino can be massless. This prediction, however, will significantly improve with theimproved errors on ∆ m .The minimally and maximally allowed neutrino masses (corresponding to δ e − δ ν = 0 , π ) are veryclose: min (in eV) max (in eV) m ν = 0 . ± . m ν = 0 . ± . m ν = 0 . ± . m ν = 0 . ± . m ν = 0 . ± . m ν = 0 . ± . .
01 eV. U thPMNS as implied by the four leptonic mass ratios Albeit the hierarchy is not as strong asfor quarks and charged neutrinos, we dare to use the same description and show that indeed largemass ratios in the four mass ratio parametrization also lead to large mixing angles. The applicabilityof the whole method depends on hierarchical masses. In B we give a simple criterion parameter tocheck whether the lower-rank approximations are good approximations. Indeed, the deviation fromunity is only a few percent. Therefore, we safely use the previous described procedure.With the predicted neutrino masses (which only know about | U e | ) and the knowledge of thecharged fermion mass ratios, the leptonic mixing matrix exhibits the following numerical values | U thPMNS | = . +0 . − . . +0 . − . . ± . . +0 . − . . +0 . − . . ± . . +0 . − . . +0 . − . . ± . , (38) We are implicitly assuming normal ordering. Inverted ordering is excluded by construction because it is not hierar-chical in the minimal flavor symmetry breaking chain. .001 0.01 0.1 m0 (cid:144) eV0.0010.010.1m Ν (cid:144) eV 1 2 3 4 5 6 ∆ È Ue2 È Figure 3: Left: Evaluation of the three neutrino masses with the lightest mass ( m , in eV). In theregime m < . m ν /m ν basically does not change with decreasing m = m ν . Right: The value of | U e | independency from δ ν —the experimentally allowed 3 σ region (indicated by the horizontal red lines) iscompatible with the choice δ ν = π , while not with δ ν = 0 or π .whereas the implied amount of CP violation is displayed as J ℓ = Im( U e U µ U ∗ e U ∗ µ ) = 0 . +0 . − . . (39)We remark an astonishingly good agreement with the measured values (see A) and observe a close-to-maximal CP violation in the lepton sector! ( δ CP = 70 ◦ from the central values: J ℓ = J max ℓ sin δ CP ,the error on J max ℓ is nevertheless compatible with maximal CP violation, δ CP = 90 ◦ .) The goal of the presented work is not to be a precision analysis of quark and lepton mixing. Theprojected values of the mixing matrices are rather a rough-and-ready estimate compatible though verywell with experimental data. We wanted to show that the knowledge of fermion masses is sufficientto describe their mixing accepting a hierarchical nature.The errors that are presented in Eqs. (33) and (38) follow from the uncertainties in the masses.Better precision in the determination of quark masses leads to better discrimination in future whetherthe described procedure is valid. The estimates are not too bad, nevertheless, we ignored radiativecorrections to the mixing matrices and constrain ourselves on a tree-level discussion. One-loop cor-rections to the masses or Yukawa couplings would be suppressed by factors Y ij Y jk Y kl / (16 π ) and aretherefore in the range of the errors for the masses. Renormalization group running of the parametersis also negligible: quark mixing angles do basically not run. The running of fermion mixing parame-ters depends on a factor ( m i + m j ) / ( m i − m j ) which is small for the hierarchical spectra. Especiallyneutrino masses and mixings run only slightly in the scenario which is under consideration in thiswork. We investigated the long-standing question of understanding the functional description of the mixingmatrices in terms of the fermion masses. The pure phenomenological observation of strong hierarchiesamong the charged fermion masses m f, ≫ m f, ≫ m f, guides the way to a parametrization of fermionmixings in terms of mass ratios without further assumptions. By solely exploiting the mathematical16roperties of the mass matrices, namely their Singular Value Decomposition, and making use ofthe double mass hierarchy pattern (DMHP), we have shown that four mass ratios in each fermionsector and a maximal CP violating phase in the 1-2 rotation are sufficient to reproduce the numericalquantities of the fermionic mixing matrices. Hierarchical masses guarantee a unique decompositioninto singular vectors up to a complex phase shared by the respective pair of singular vectors of asingular value. This uniqueness theorem dissolves the common ambiguities found in the literatureoriginated in the freedom of weak bases. Schmidt-Mirsky’s approximation theorem has been used toapproximate the hierarchical mass matrices by lower-rank matrices that are the closest one to thegiven full rank matrix. The connection of each lower rank approximation to the nature of the Yukawainteractions, m f,i = 0 → Y fij = 0 = Y fji , helps to simplify the reparametrization of the mass matrixwithout loosing track of the parameters. This connection is established via the minimal breaking ofmaximal flavor symmetry [U(3)] → [U(2)] → [U(1)] → U(1) F in each fermion sector, where theremnant U(1) F symmetry is either baryon or lepton number. The approximation, however, neglectssizeable terms in the mass matrices that have been consistently added by use of correcting rotations.The arbitrariness of complex phases is reduced by requiring them to be either maximally CP violating( π/
2) or CP conserving ( π ). This assumption is motivated by the fact that the four mass ratios shouldbe enough to serve as mixing parameters in the unitary 3 × .
01 eV, while thelargest neutrino mass lies around 0 .
05 eV. We therefore conclude that, if also in the neutrino sector themixing is determined by the mass ratios without any further contribution, the electron neutrino massescapes its nearby measurement from tritium decay. Moreover, we give a prediction for the leptonicCP phase close to maximal, δ ν CP ≈ ◦ .Hence, contrary to the common expectation, leptonic mixing angles are found to be determinedsolely by the four leptonic mass ratios: m e /m µ , m µ /m τ , m ν /m ν , and m ν /m ν without any relationto the geometrical factors observed in most flavor models. Notwithstanding, we see a great powerof the described method in the application to flavor model building: once a model gives hierarchicalmasses, the mixing follows from this hierarchy. In contrast, our approach gives viable patterns andtextures for mass matrices in terms of the singular values (fermion masses). We explicitly leave thequestion of a model behind open. Likewise, the origin of CP violation stays unexplained, though ourobservation about the distribution of CP phases gives an important starting point. Acknowledgements
UJSS would like to acknowledge useful and detailed discussions about this idea in its initial stagewith A. Mondrag´on and E. Jim´enez. The authors want to thank the following people for their criticalassessment on this work during its different stages: M. Spinrath, U. Nierste, C. Wiegand, M. Zoller,J. Hoff, M. Hoeschele, and K. Melnikov. Also, the authors are indebted to U. Nierste for a carefulreading of the manuscript and his detailed comments on it. UJSS wants to acknowledge financialsupport from the Karlsruhe House of Young Scientists (KHYS) for his stay in Karlsruhe. WGH ac-knowledges support by the DFG-funded research training group GRK 1694 “Elementarteilchenphysikbei h¨ochster Energie und h¨ochster Pr¨azision”. For last, UJSS is grateful to the Institut f¨ur Theo-retische Teilchenphysik (TTP) for its warm hospitality during the realization and completion of thiswork. 17
State of the art in the fermion masses and mixing matrices
In this section, we collect the current knowledge about fermion mixing data and specify the inputvalues we use in the following for the masses.For all numerical evaluations made in this work, we stick to the updated values of the quark mixingmatrix [78], | V CKM | = . ± . . ± . . ± . . ± . . ± . . ± . . +0 . − . . +0 . − . . ± . , (40)with the Jarlskog invariant equal to J q = (3 . +0 . − . ) × − . In the standard parametrization by theParticle Data Group (PDG), the central values give the following mixing angles, θ q ≈ . ◦ , θ q ≈ . ◦ , θ q ≈ . ◦ . (41)The most recent update on the 3 σ allowed ranges of the elements of the PMNS mixing matrix aregiven by [79], | U P MNS | = . → .
845 0 . → .
580 0 . → . . → .
517 0 . → .
699 0 . → . . → .
529 0 . → .
713 0 . → . . (42)Where the best fit points of the mixing angles are θ ℓ = 33 . ◦ , θ ℓ = 8 . ◦ , θ ℓ = 42 . ◦ . (43)The maximal value of the leptonic Jarlskog invariant is given by J max ℓ = 0 . ± .
010 and differentfrom zero at more than 3 σ —still, the proper J ℓ has first to be multiplied by sin δ CP and is supposedto be smaller.The study of the mixing matrices in terms of the masses is done at the scale of the Z boson mass.The input values for the numerical calculations are obtained using the experimental values of thequark masses as given by the PDG Review 2014 [78] and running them to the scale of the Z bosondetermining the electroweak scale. We include highest precision running in QCD by the virtue of theRunDec package [80]. For completeness, we show the input values and their uncertainties as well asthe resulting outputs in Table 2.The reported measured on-shell values in MeV for the charged lepton masses are, m e = 0 . , m µ = 105 . , m τ = 1776 . ± . , (44)where we have neglected the tiny experimental errors in the first two generation masses. The recentchanges of this values affect only the few last digits. Therefore, we safely trust the results of [81] fortheir values at the Z scale (in MeV): m e ( M Z ) = 0 . , m µ ( M Z ) = 102 . , m τ ( M Z ) = 1746 . +0 . − . . (45)The nine mass ratios are of essential use in the evaluation of the analytic formulae to describefermion mixing. We show our input values determined from Table 2 and Eq. (45) in Table 3.In the case of neutrinos, only two squared mass differences have been measured whose values aretaken from [79],NO: ∆ m = +2 . ± . × − eV , IO: ∆ m = − . ± . × − eV , ∆ m = 7 . +0 . − . × − eV , (46)where NO and IO stand for normal and inverted ordering, respectively.Still, the most recent direct bound on the neutrino mass scale stems from tritium beta decayexperiments: m ( ν e ) . Z boson mass scale by virtue of the RunDec package [80].The mass inputs correspond to the experimental measured values while the outputs, evaluated at the Z pole, include the resummation of higher order corrections from QCD by the RG running. RunDectakes properly into account the decoupling of heavy quarks below their scale. All masses are given inGeV. input output m u (2 GeV) = 0 . +0 . − . m u ( M Z ) = 0 . +0 . − . m d (2 GeV) = 0 . +0 . − . m d ( M Z ) = 0 . +0 . − . m s (2 GeV) = 0 . ± . m s ( M Z ) = 0 . ± . m c ( m c ) = 1 . ± . m c ( M Z ) = 0 . ± . m b ( m b ) = 4 . ± . m b ( M Z ) = 2 . ± . m t (OS) = 173 . ± . m t ( M Z ) = 172 . +1 . − . Table 3: Charged fermions mass ratios at the M Z scale. f m f, /m f, m f, /m f, m f, /m f, u . +0 . − . (7 . +2 . − . ) × − . ± . d . +0 . − . (9 . +1 . − . ) × − . ± . e B Applicability of the method
The Schmidt-Mirsky theorem relates the validity of the lower rank approximation to a measure ofbeing close to the full rank matrix. This measure has to be a scalar parameter and can be any norm.In the original formulation, the Frobenius norm was used, which is also the most natural choice since itis the square root over the sum of squared singular values and directly related to one of the invariantsof the mass matrix k M f k F = s X i =1 , , m f,i . (47)The use of this norm serves as a way to define a criterion which allows us to distinguish when thehierarchy is strong enough as to safely make an approximation. In this regard, we define the parameter x rf as, x rf ≡ q ( r − m f, + m f, k M f k F = vuut ( r − m f, + m f, m f, + m f, + m f, , (48)where r = rank[ M rf ] ∈ { , } . The approximation becomes better the closer x rf is to one and is exactin the x rf → x rf ≡ q [( r − m f, + m f, ] / ( m f, + m f, + m f, ), for thedifferent cases of the fermion masses, where x rf provides a measure of the applicability of the method.The fact that all cases here are sufficiently close to one guarantees the safe use of the lowest rankapproximations. Even for neutrinos, x ν is close to one, where we exploit the prediciton for neutrinomasses from Sec. 3.3. x rf u d e νr = 1 0.999993 0.999816 0.998274 0.978894 r = 2 0.999999 0.999999 0.999999 0.996773with the original norm. Hence, x rf is a measure of the applicability of the method. Table 4 shows thedifferent values obtained of x rf for the several charged fermion masses. The values in the rank oneapproximation, r = 1, for all practical purposes equal to one, though for both charged and neutralleptons deviate in the per mill and percent regime, respectively. From here we can already understandwhy the quark mixing matrix is so close to the unit matrix which is the trivial mixing matrix in therank one approximation. In a similar manner, the very mild hierarchy for neutrinos leads to a strongerdeviation from the rank one approximation and therefore larger mixing angles. C Hierarchical mass matrices
We show how to derive the hierarchical structure of the mass matrices by the use of the lower-rankapproximation theorem and the principle of minimal flavor violation. Let us consider the two-flavorcase and the mass matrix m = (cid:18) m ss m sl m ls m ll (cid:19) , (49)with the two singular values σ s and σ l respecting the hierarchy σ s ≪ σ l .We decompose the mass matrix in terms of the Singular Value decomposition L m R † = diag( σ s , σ l ) , (50)where the left and right unitary matrices diagonalize the Hermitian products L m m † L † = (cid:18) σ s σ l (cid:19) = R m † m R † . (51)Each Hermitian product can be expressed as a sum of rank one matrices with the components of L and R , m m † = σ s (cid:18) | L | L L ∗ L ∗ L | L | (cid:19) + σ l (cid:18) | L | L L ∗ L ∗ L | L | (cid:19) (52)and m † m = σ s (cid:18) | R | R R ∗ R ∗ R | R | (cid:19) + σ l (cid:18) | R | R R ∗ R ∗ R | R | (cid:19) . (53)Due to our lack of knowledge of right-handed flavor mixing, the relevant object that determines ourphenomenology is the left Hermitian product, m m † .20 pplying Schmidt-Mirsky’s approximation theorem Consider the rank-one approximation inEq. (52) by ˆ σ = σ s /σ l = 0 normalized with respect to the larger singular valueˆ m r =1 ( ˆ m r =1 ) † = (cid:18) | L | L L ∗ L ∗ L | L | (cid:19) . (54)The components of the left unitary matrix depend on ˆ σ . In the limit ˆ σ →
0, there is trivial mixingand the rank one left Hermitian product isˆ m r =1 ( ˆ m r =1 ) † = (cid:18) (cid:19) . (55)A small breaking of the [U(1)] symmetry for the massless fermions implies only a small deviationfrom the trivial mixing: | L | ∼ (cid:18) θθ (cid:19) . (56)The mixing angle is related to the parameter of symmetry breaking ˆ σ and it is an easy exercise toderive θ ∼ √ ˆ σ from Eq. (51).We then get an estimate on the magnitudes of each element in Eq. (52) | ˆ m ˆ m † | ∼ (cid:18) O ( θ ) O ( θ ) O ( θ ) 1 + O ( θ ) (cid:19) . (57)The explicit form of the mass matrix m stays unknown as long as we have no information about R . However, the minimal breaking of the maximal flavor symmetry applies to both chiralities simul-taneously and the argument from above is the same for the right Hermitian product. We thereforeknow that L and R have the same moduli and get the hierarchical structure of m :ˆ m = (cid:18) m ss m sl m ls m ll (cid:19) ∼ (cid:18) O ( θ ) O ( θ ) O ( θ ) 1 + O ( θ ) (cid:19) , (58)with | m sl | = | m ls | as a natural consequence of hierarchical masses and minimal flavor symmetrybreaking. The hierarchical structure for the mass matrix and its Hermitian product is the same.Hence, due to the strong hierarchy in the masses we can neglect the role of | m ss | ∼ θ in (57)working with the leading order contributions in θ and assume m ss = 0 as done in Eq. (15). This givescorrections to the Gatto-Satori-Tonin relation, tan θ = p σ s /σ l = √ ˆ σ , which are O ( θ ) = O (ˆ σ √ ˆ σ )and therefore neglected. D Explicit approximate formulae for the mixing angles and the Jarl-skog invariant
The explicit formulae for the distinct mixing matrix elements in terms of the mass ratios is ratherlengthy. We opt then, to show only the three mixing angles, used in the standard parametrization,with the corresponding Jarlskog invariant. This allows to express the mixing angles in terms of threemoduli of the mixing matrixsin θ f = q,ℓ = (cid:12)(cid:12)(cid:12) V f = q,ℓ (cid:12)(cid:12)(cid:12)r − (cid:12)(cid:12)(cid:12) V f = q,ℓ (cid:12)(cid:12)(cid:12) , sin θ f = q,ℓ = (cid:12)(cid:12)(cid:12) V f = q,ℓ (cid:12)(cid:12)(cid:12)r − (cid:12)(cid:12)(cid:12) V f = q,ℓ (cid:12)(cid:12)(cid:12) , sin θ f = q,ℓ = (cid:12)(cid:12)(cid:12) V f = q,ℓ (cid:12)(cid:12)(cid:12) . (59)21n the four mass ratios parametrization it is more natural to give not the formulae of the mixing anglesin terms of the masses but rather of the aforementioned moduli | V f = q,ℓ | ≈ s ˆ m a + ˆ m b (1 + ˆ m a )(1 + ˆ m b ) , (60) | V f = q,ℓ | ≈ ∓ p ˆ m a + q ˆ m b + p ˆ m a ∓ q ˆ m b + p ˆ m a ˆ m a ± q ˆ m b ˆ m b q (1 + ˆ m a )(1 + ˆ m b )(1 + ˆ m a )(1 + ˆ m b )(1 + ˆ m a ˆ m a )(1 + ˆ m b ˆ m b ) , (61) | V f = q,ℓ | ≈ ∓| V f = q,ℓ | s ˆ m a m a + p ˆ m a − q ˆ m b + p ˆ m a ˆ m a + q ˆ m b ˆ m b + ˆ m a + ˆ m b q (1 + ˆ m a )(1 + ˆ m a ˆ m a ) (1 + ( ˆ m a ) ) (1 + ˆ m a )(1 + ˆ m b )(1 + ˆ m b ˆ m b ) (cid:0) m b ) (cid:1) , (62)where we have denoted ˆ m a ( b ) ij = m a ( b ) i /m a ( b ) j , the upper and lower signs in Eq. 61 correspond to q and ℓ , respectively. The two fermion species of each sector are a = u, e and b = d, ν .The Jarlskog invariant is given by, J f = q,ℓ ≈ cos θ b sin θ b sin θ f = q,ℓ (cid:16) sin θ a sin θ f = q,ℓ + sin θ a − sin θ b (cid:17) , (63)wheresin θ a ( b )12 = vuut ˆ m a ( b )12 m a ( b )12 and sin θ a ( b )13 ≈ ± q ˆ m a ( b )13 + q ˆ m a ( b )13 ˆ m a ( b )23 + ˆ m a ( b )23 r(cid:16) m a ( b )13 (cid:17) (cid:16) m a ( b )13 ˆ m a ( b )23 (cid:17) (cid:16) m a ( b )23 ) (cid:17) . (64)The approximate relations here given differ from the complete one in ∼
1% order.
References [1] R. Gatto, G. Sartori, and M. Tonin, “Weak Selfmasses, Cabibbo Angle, and Broken SU(2) xSU(2),”
Phys.Lett.
B28 (1968) 128–130.[2] N. Cabibbo and L. Maiani, “Dynamical interrelation of weak, electromagnetic and stronginteractions and the value of theta,”
Phys.Lett.
B28 (1968) 131–135.[3] K. Tanaka and P. Tarjanne, “Cabibbo angle and selfconsistency condition,”
Phys.Rev.Lett. (1969) 1137–1139.[4] R. Oakes, “SU(2) x SU(2) breaking and the Cabibbo angle,” Phys.Lett.
B29 (1969) 683–685.[5] H. Genz, J. Katz, L. Ram Mohan, and S. Tatur, “Chiral symmetry breaking and the cabibboangle,”
Phys.Rev. D6 (1972) 3259–3265.[6] H. Pagels, “Vacuum Stability and the Cabibbo Angle,” Phys.Rev.
D11 (1975) 1213.[7] A. Ebrahim, “Chiral SU(4) x SU(4) Symmetry Breaking and the Cabibbo Angle,”
Phys.Lett.
B69 (1977) 229–230.[8] H. Fritzsch, “Calculating the Cabibbo Angle,”
Phys.Lett.
B70 (1977) 436.229] S. Weinberg, “The Problem of Mass,”
Trans.New York Acad.Sci. (1977) 185–201.[10] A. De Rujula, H. Georgi, and S. Glashow, “A Theory of Flavor Mixing,” Annals Phys. (1977) 258.[11] F. Wilczek and A. Zee, “Discrete Flavor Symmetries and a Formula for the Cabibbo Angle,”
Phys.Lett.
B70 (1977) 418.[12] R. N. Mohapatra and G. Senjanovic, “Cabibbo Angle, CP Violation and Quark Masses,”
Phys.Lett.
B73 (1978) 176.[13] F. Wilczek and A. Zee, “Horizontal Interaction and Weak Mixing Angles,”
Phys.Rev.Lett. (1979) 421.[14] D. Wyler, “The Cabibbo Angle in the SU(2) L × U(1) Gauge Theories,”
Phys.Rev.
D19 (1979) 330.[15] H. Fritzsch, “Hierarchical Chiral Symmetries and the Quark Mass Matrix,”
Phys.Lett.
B184 (1987) 391.[16] P. H. Frampton and Y. Okada, “Simplified symmetric quark mass matrices and flavor mixing,”
Mod.Phys.Lett. A6 (1991) 2169–2172.[17] J. L. Rosner and M. P. Worah, “Models of the quark mixing matrix,” Phys.Rev.
D46 (1992) 1131–1140.[18] S. Raby, “Introduction to theories of fermion masses,” arXiv:hep-ph/9501349 [hep-ph] .[19] T. Ito and M. Tanimoto, “The Unitarity triangle and quark mass matrices on the NNI basis,”
Phys.Rev.
D55 (1997) 1509–1514, arXiv:hep-ph/9603393 [hep-ph] .[20] Z.-z. Xing, “Implications of the quark mass hierarchy on flavor mixings,”
J.Phys.
G23 (1997) 1563–1578, arXiv:hep-ph/9609204 [hep-ph] .[21] S.-S. Xue, “Quark masses and mixing angles,”
Phys.Lett.
B398 (1997) 177–186, arXiv:hep-ph/9610508 [hep-ph] .[22] R. Barbieri, L. J. Hall, S. Raby, and A. Romanino, “Unified theories with U(2) flavorsymmetry,”
Nucl.Phys.
B493 (1997) 3–26, arXiv:hep-ph/9610449 [hep-ph] .[23] D. Falcone and F. Tramontano, “Relation between quark masses and weak mixings,”
Phys.Rev.
D59 (1999) 017302, arXiv:hep-ph/9806496 [hep-ph] .[24] A. Mondragon and E. Rodriguez-Jauregui, “The Breaking of the flavor permutationalsymmetry: Mass textures and the CKM matrix,”
Phys.Rev.
D59 (1999) 093009, arXiv:hep-ph/9807214 [hep-ph] .[25] A. Mondragon and E. Rodriguez-Jauregui, “The CP violating phase delta(13) and the quarkmixing angles theta(13), theta(23) and theta(12) from flavor permutational symmetrybreaking,”
Phys.Rev.
D61 (2000) 113002, arXiv:hep-ph/9906429 [hep-ph] .[26] H. Fritzsch and Z.-z. Xing, “Mass and flavor mixing schemes of quarks and leptons,”
Prog.Part.Nucl.Phys. (2000) 1–81, arXiv:hep-ph/9912358 [hep-ph] .2327] G. Branco, D. Emmanuel-Costa, and C. Simoes, “Nearest-Neighbour Interaction from anAbelian Symmetry and Deviations from Hermiticity,” Phys.Lett.
B690 (2010) 62–67, arXiv:1001.5065 [hep-ph] .[28] F. Gonz´alez Canales, A. Mondrag´on, M. Mondrag´on, U. J. Salda˜na Salazar, andL. Velasco-Sevilla, “Quark sector of S3 models: classification and comparison with experimentaldata,”
Phys.Rev.
D88 (2013) 096004, arXiv:1304.6644 [hep-ph] .[29] L. J. Hall and A. Rasin, “On the generality of certain predictions for quark mixing,”
Phys.Lett.
B315 (1993) 164–169, arXiv:hep-ph/9303303 [hep-ph] .[30] A. Rasin, “Diagonalization of quark mass matrices and the Cabibbo-Kobayashi-Maskawamatrix. ,” arXiv:hep-ph/9708216 [hep-ph] .[31] A. Rasin, “Hierarchical quark mass matrices,”
Phys.Rev.
D58 (1998) 096012, arXiv:hep-ph/9802356 [hep-ph] .[32] H. Fritzsch and Z.-z. Xing, “The Light quark sector, CP violation, and the unitarity triangle,”
Nucl.Phys.
B556 (1999) 49–75, arXiv:hep-ph/9904286 [hep-ph] .[33] Z.-z. Xing, “Model-independent access to the structure of quark flavor mixing,”
Phys.Rev.
D86 (2012) 113006, arXiv:1211.3890 [hep-ph] .[34] H. Ishimori, T. Kobayashi, H. Ohki, Y. Shimizu, H. Okada, et al. , “Non-Abelian DiscreteSymmetries in Particle Physics,”
Prog.Theor.Phys.Suppl. (2010) 1–163, arXiv:1003.3552 [hep-th] .[35] S. F. King and C. Luhn, “Neutrino Mass and Mixing with Discrete Symmetry,”
Rept.Prog.Phys. (2013) 056201, arXiv:1301.1340 [hep-ph] .[36] J. Schechter and J. Valle, “Neutrino Masses in SU(2) x U(1) Theories,” Phys.Rev.
D22 (1980) 2227.[37] H. Fritzsch, “Quark Masses and Flavor Mixing,”
Nucl.Phys.
B155 (1979) 189.[38] H. Fritzsch, “Weak Interaction Mixing in the Six - Quark Theory,”
Phys.Lett.
B73 (1978) 317–322.[39] P. Ramond, R. Roberts, and G. G. Ross, “Stitching the Yukawa quilt,”
Nucl.Phys.
B406 (1993) 19–42, arXiv:hep-ph/9303320 [hep-ph] .[40] G. Branco, D. Emmanuel-Costa, and R. Gonzalez Felipe, “Texture zeros and weak basistransformations,”
Phys.Lett.
B477 (2000) 147–155, arXiv:hep-ph/9911418 [hep-ph] .[41] R. Roberts, A. Romanino, G. G. Ross, and L. Velasco-Sevilla, “Precision test of a fermion masstexture,”
Nucl.Phys.
B615 (2001) 358–384, arXiv:hep-ph/0104088 [hep-ph] .[42] H. Fritzsch and Z.-z. Xing, “Four zero texture of Hermitian quark mass matrices and currentexperimental tests,”
Phys.Lett.
B555 (2003) 63–70, arXiv:hep-ph/0212195 [hep-ph] .[43] M. Gupta and G. Ahuja, “Flavor mixings and textures of the fermion mass matrices,”
Int. Jour. Mod. Phys. A, (2012) 1230033, arXiv:1302.4823 [hep-ph] .[44] S. Pakvasa and H. Sugawara, “CP Violation in Six Quark Model,” Phys.Rev.
D14 (1976) 305.2445] S. Pakvasa and H. Sugawara, “Discrete Symmetry and Cabibbo Angle,”
Phys.Lett.
B73 (1978) 61.[46] E. Derman, “Flavor Unification, τ Decay and b Decay Within the Six Quark Six LeptonWeinberg-Salam Model,”
Phys.Rev.
D19 (1979) 317–329.[47] Y. Yamanaka, H. Sugawara, and S. Pakvasa, “Permutation Symmetries and the Fermion MassMatrix,”
Phys.Rev.
D25 (1982) 1895.[48] R. Yahalom, “Horizontal Permutation Symmetry, Fermion Masses and Pseudogoldstone Bosonsin SU(2) L × U(1),”
Phys.Rev.
D29 (1984) 536.[49] G. Altarelli, “Status of Neutrino Mass and Mixing,”
Int.J.Mod.Phys.
A29 (2014) 1444002, arXiv:1404.3859 [hep-ph] .[50] S. F. King, A. Merle, S. Morisi, Y. Shimizu, and M. Tanimoto, “Neutrino Mass and Mixing:from Theory to Experiment,”
New J.Phys. (2014) 045018, arXiv:1402.4271 [hep-ph] .[51] H. Fritzsch and Z.-z. Xing, “Lepton mass hierarchy and neutrino mixing,” Phys.Lett.
B634 (2006) 514–519, arXiv:hep-ph/0601104 [hep-ph] .[52] H. Fritzsch and Z.-z. Xing, “Relating the neutrino mixing angles to a lepton mass hierarchy,”
Phys.Lett.
B682 (2009) 220–224, arXiv:0911.1857 [hep-ph] .[53] E. Schmidt, “Zur Theorie der linearen und nichtlinearen Integralgleichungen - I. Teil:Entwicklung willk¨urlicher Funktionen nach Systemen vorgeschriebener,”
MathematischeAnnalen no. 4, (1907) 433–476.[54] G. Eckart and G. Young, “The approximation of one matrix by another of lower rank,” Psychometrika (1936) 211–218.[55] L. Mirsky, “Symmetric gauge functions and unitarily invariant norms,” Quart. J. Math. (1960) 50–59.[56] G. Golub, A. Hoffman, and G. Stewart, “A generalization of the Eckart-Young-Mirsky matrixapproximation theorem,” Linear Algebra and Its Applications no. C, (1987) 317–327.[57] G. Stewart, “On the early history of the singular value decomposition,”
SIAM Review no. 4,(1993) 551–566.[58] R. Barbieri, G. Dvali, and L. J. Hall, “Predictions from a U(2) flavor symmetry insupersymmetric theories,” Phys.Lett.
B377 (1996) 76–82, arXiv:hep-ph/9512388 [hep-ph] .[59] R. Barbieri, L. J. Hall, and A. Romanino, “Consequences of a U(2) flavor symmetry,”
Phys.Lett.
B401 (1997) 47–53, arXiv:hep-ph/9702315 [hep-ph] .[60] A. Crivellin and U. Nierste, “Supersymmetric renormalisation of the CKM matrix and newconstraints on the squark mass matrices,”
Phys.Rev.
D79 (2009) 035018, arXiv:0810.1613 [hep-ph] .[61] A. Crivellin, L. Hofer, U. Nierste, and D. Scherer, “Phenomenological consequences of radiativeflavor violation in the MSSM,”
Phys.Rev.
D84 (2011) 035030, arXiv:1105.2818 [hep-ph] .[62] R. Barbieri, D. Buttazzo, F. Sala, and D. M. Straub, “Flavour physics from an approximate U (2) symmetry,” JHEP (2012) 181, arXiv:1203.4218 [hep-ph] .2563] A. J. Buras and J. Girrbach, “On the Correlations between Flavour Observables in Minimal U (2) Models,”
JHEP (2013) 007, arXiv:1206.3878 [hep-ph] .[64] C. D. Carone and L. J. Hall, “Neutrino physics from a U(2) flavor symmetry,”
Phys.Rev.
D56 (1997) 4198–4206, arXiv:hep-ph/9702430 [hep-ph] .[65] M. Tanimoto, “Neutrino mass matrix with U(2) flavor symmetry and its test in neutrinooscillation experiments,”
Phys.Rev.
D57 (1998) 1983–1986, arXiv:hep-ph/9706497 [hep-ph] .[66] L. J. Hall and N. Weiner, “U(2) and maximal mixing of muon-neutrino,”
Phys.Rev.
D60 (1999) 033005, arXiv:hep-ph/9811299 [hep-ph] .[67] R. Barbieri, P. Creminelli, and A. Romanino, “Neutrino mixings from a U(2) flavor symmetry,”
Nucl.Phys.
B559 (1999) 17–26, arXiv:hep-ph/9903460 [hep-ph] .[68] G. Blankenburg, G. Isidori, and J. Jones-Perez, “Neutrino Masses and LFV from MinimalBreaking of U (3) and U (2) flavor Symmetries,” Eur.Phys.J.
C72 (2012) 2126, arXiv:1204.0688 [hep-ph] .[69] A. Aranda, C. D. Carone, and R. F. Lebed, “U(2) flavor physics without U(2) symmetry,”
Phys.Lett.
B474 (2000) 170–176, arXiv:hep-ph/9910392 [hep-ph] .[70] L. J. Hall and H. Murayama, “A Geometry of the generations,”
Phys.Rev.Lett. (1995) 3985–3988, arXiv:hep-ph/9508296 [hep-ph] .[71] J. Kubo, A. Mondragon, M. Mondragon, and E. Rodriguez-Jauregui, “The Flavor symmetry,” Prog.Theor.Phys. (2003) 795–807, arXiv:hep-ph/0302196 [hep-ph] .[72] S. Morisi, “S(3) family permutation symmetry and quark masses: A Model independentapproach,” arXiv:hep-ph/0604106 [hep-ph] .[73] F. Feruglio and Y. Lin, “Fermion Mass Hierarchies and Flavour Mixing from a Minimal DiscreteSymmetry,”
Nucl.Phys.
B800 (2008) 77–93, arXiv:0712.1528 [hep-ph] .[74] T. Teshima and Y. Okumura, “Quark/lepton mass and mixing in S invariant model andCP-violation of neutrino,” Phys.Rev.
D84 (2011) 016003, arXiv:1103.6127 [hep-ph] .[75] R. Jora, J. Schechter, and M. Naeem Shahid, “Perturbed S(3) neutrinos,”
Phys.Rev.
D80 (2009) 093007, arXiv:0909.4414 [hep-ph] .[76] I. Masina and C. A. Savoy, “Real and imaginary elements of fermion mass matrices,”
Nucl.Phys.
B755 (2006) 1–20, arXiv:hep-ph/0603101 [hep-ph] .[77] I. Masina, “A Maximal atmospheric mixing from a maximal CP violating phase,”
Phys.Lett.
B633 (2006) 134–140, arXiv:hep-ph/0508031 [hep-ph] .[78]
Particle Data Group
Collaboration, K. Olive et al. , “Review of Particle Physics,”
Chin.Phys.
C38 (2014) 090001.[79] M. Gonzalez-Garcia, M. Maltoni, and T. Schwetz, “Updated fit to three neutrino mixing: statusof leptonic CP violation,” arXiv:1409.5439 [hep-ph] .[80] K. Chetyrkin, J. H. Kuhn, and M. Steinhauser, “RunDec: A Mathematica package for runningand decoupling of the strong coupling and quark masses,”
Comput.Phys.Commun. (2000) 43–65, arXiv:hep-ph/0004189 [hep-ph] .2681] Z.-z. Xing, H. Zhang, and S. Zhou, “Updated Values of Running Quark and Lepton Masses,”
Phys.Rev.
D77 (2008) 113016, arXiv:0712.1419 [hep-ph] .[82]
Troitsk Collaboration
Collaboration, V. Aseev et al. , “An upper limit on electronantineutrino mass from Troitsk experiment,”
Phys.Rev.
D84 (2011) 112003, arXiv:1108.5034 [hep-ex] .[83]
KATRIN Collaboration
Collaboration, A. Osipowicz et al. , “KATRIN: A Next generationtritium beta decay experiment with sub-eV sensitivity for the electron neutrino mass. Letter ofintent. ,” arXiv:hep-ex/0109033 [hep-ex]arXiv:hep-ex/0109033 [hep-ex]