In pursuit of new physics with Bs,d0 → ℓ+ℓ−

Leptonic rare decays of Bs,d0 mesons offer a powerful tool to search for physics beyond the Standard Model. The Bs0 → μ+μ− decay has been observed at the Large Hadron Collider and the first measurement of the effective lifetime of this channel was presented, in accordance with the Standard Model. On the other hand, Bs0 → τ+τ− and Bs0 → e+e− have received considerably less attention: while LHCb has recently reported a first upper limit of 6.8 × 10−3 (95% C.L.) for the Bs0 → τ+τ− branching ratio, the upper bound 2.8 × 10−7 (90% C.L.) for the branching ratio of Bs0 → e+e− was reported by CDF back in 2009. We discuss the current status of the interpretation of the measurement of Bs0 → μ+μ−, and explore the space for New-Physics effects in the other Bs,d0 → ℓ+ℓ− decays in a scenario assuming flavour-universal Wilson coefficients of the relevant four-fermion operators. While the New-Physics effects are then strongly suppressed by the ratio mμ/mτ of the lepton masses in Bs0 → τ+τ−, they are hugely enhanced by mμ/me in Bs0 → e+e− and may result in a Bs0 → e+e− branching ratio as large as about 5 times the one of Bs0 → μ+μ−, which is about a factor of 20 below the CDF bound; a similar feature arises in Bd0 → e+e−. Consequently, it would be most interesting to search for the Bs,d0 → e+e− channels at the LHC and Belle II, which may result in an unambiguous signal for physics beyond the Standard Model.


Introduction
The decay B 0 s → µ + µ − belongs to the most interesting processes for testing the flavour sector of the Standard Model (SM). In this framework, B 0 s → µ + µ − emerges from quantum loop effects -penguin and box topologies -and is helicity suppressed. Consequently, this channel is strongly suppressed, and in the SM only about three out of one billion B 0 s mesons decay into the µ + µ − final state. Another key feature of the B 0 s → µ + µ − decay is that the binding of the anti-bottom quark and the strange quark in the B 0 s meson is described by a single non-perturbative parameter, the B 0 s -meson decay constant f Bs [1]. In scenarios of New Physics (NP), B 0 s → µ + µ − may be affected by new particles entering the loop topologies or may even arise at the tree level.
For decades, experiments have searched for the B 0 s → µ + µ − decay [2]. It has been a highlight of run 1 of the Large Hadron Collider (LHC) that the B 0 s → µ + µ − mode could eventually be observed in a combined analysis by the CMS and LHCb collaborations [3]. The corresponding experimental branching ratio is consistent with the SM prediction.
In addition to the branching ratio, B 0 s → µ + µ − offers another observable [4]. It is accessible thanks to the sizeable difference ∆Γ s between the decay widths of the B s mass eigenstates and is encoded in the effective B 0 s → µ + µ − lifetime. The LHCb collaboration has very recently reported a pioneering measurement of this observable, as well as the observation of B 0 s → µ + µ − in the analysis of their new data set collected at the ongoing run 2 of the LHC [5].
Furthermore, there is a CP-violating rate asymmetry which is generated through the interference between B 0 s -B 0 s mixing and decay processes [4,6]. It would be very interesting to measure this observable. However, such an analysis requires tagging information on the decaying B s meson, thereby making it more challenging than the untagged effective lifetime measurement.
In the SM, the key difference of B 0 s → µ + µ − with respect to B 0 s → τ + τ − and B 0 s → e + e − is due to the different lepton masses. In the case of the former decay, the large τ mass effectively lifts the helicity suppression, while it gets much stronger for the latter process due to the small electron mass. Consequently, we have SM branching ratios at the 10 −6 and 10 −13 level, respectively, while the corresponding B 0 s → µ + µ − branching ratio takes a value at the 10 −9 level [1].
From the experimental point of view, the analysis of B 0 s → τ + τ − is challenging because of the reconstruction of the τ leptons. Nevertheless, LHCb has recently presented the first upper bound for this channel of 6.8 × 10 −3 (95% C.L.) [7]. The B 0 s → e + e − decay has received surprisingly little attention, both from the experimental and theoretical communities, and has so far essentially not played any role in the exploration of flavour physics. The most recent upper bound on the B 0 s → e + e − branching ratio of 2.8 × 10 −7 (90% C.L.) was obtained by the CDF collaboration back in 2009 [8]. We have illustrated this situation in figure 1.
We will have a fresh look at the search for NP effects with the B 0 s → µ + µ − and B 0 d → µ + µ − channels in view of the new LHCb data [5], complementing the recent study by Altmannshofer, Niehoff and Straub [9]. However, the main focus of our discussion will be on the B 0 s → τ + τ − and B 0 s → e + e − decays (as well as their B 0 d counterparts), which were not considered in ref. [9]. The utility of the decays with τ + τ − and e + e − in the final states for probing NP effects was addressed in the literature before, for instance, in JHEP05(2017)156 refs. [10,11] and [12,13], respectively. The current key question is how much space for NP effects is left in these channels by the currently available data, in particular for the experimentally established B 0 s → µ + µ − mode. In order to explore this topic, which is in general very complex, and to illustrate possible NP effects, we consider a framework where the Wilson coefficients of the relevant four-fermion operators are flavour universal, i.e. do neither depend on the flavour of the decaying B 0 s or B 0 d mesons nor on the final-state leptons. We find that the corresponding NP effects are strongly suppressed in B 0 s → τ + τ − in this scenario. However, as the helicity suppression is lifted by new (pseudo)-scalar contributions, we may get a huge enhancement of the branching ratio of B 0 s → e + e − in this scenario, while still having the branching ratio of B 0 s → µ + µ − within the current experimental range. In particular, the branching ratio of B 0 s → e + e − may be enhanced to about 5 times the B 0 s → µ + µ − branching ratio, which is a factor of 20 below the CDF limit from 2009. Consequently, it would be most interesting to have a dedicated search for B 0 s → e + e − and B 0 d → e + e − , fully exploiting the physics potential of the LHC, where these decays will be interesting for ATLAS, CMS and LHCb, and the future Belle II experiment at KEK. In view of the theoretical cleanliness of these decays and the possible spectacular enhancement with respect to the SM, we may get an unambiguous signal for New Physics.
In figure 2, we have illustrated our NP analysis. The measured branching ratio of the B 0 s → µ + µ − channel allows us to constrain the corresponding short-distance functions, which are then converted into their counterparts for the B 0 s,d → τ + τ − and B 0 s,d → e + e − channels, having very different implications. The flowchart in figure 2 serves as a guideline for the following discussion.
The outline of this paper is as follows: we discuss the theoretical framework for our studies in section 2. In section 3, we have a closer look at the state-of-the-art picture following from the experimental results for the B 0 s,d → µ + µ − decays, while turning to the B 0 s,d → τ + τ − and B 0 s,d → e + e − modes in sections 4 and 5, respectively. Finally, we summarize our conclusions in section 6.

Low-energy effective Hamiltonian
Leptonic rare decays ofB 0 q mesons (q = d, s) are described by the following low-energy effective Hamiltonian [1,4,9]: where G F denotes the Fermi constant, V qq are elements of the Cabibbo-Kobayashi-Maskawa (CKM) matrix, and α is the QED fine structure constant. The heavy degrees of freedom have been integrated out and are described by the Wilson coefficients C q, ( ) 10 , C q, ( ) P and C q, ( ) S , which may depend both on the flavour of the quark q and on the flavour of the final-state leptons + − . However, in the SM and NP scenarios with "Minimal Flavour Violation" (MFV) [14], the short-distance functions are flavour universal.
The Wilson coefficients are associated with the four-fermion operators where m b denotes the b-quark mass and In the general Hamiltonian in eq. (2.1), we have only kept operators which give nonvanishing contributions toB 0 q → + − decays. In the SM, only the O 10 operator is present with a real coefficient C SM 10 . Concerning the impact of NP, the outstanding feature of theB 0 q → + − channels is their sensitivity to (pseudo)-scalar lepton densities entering the operators O (P )S and O (P )S , which have still largely unconstrained Wilson coefficients, thereby offering an interesting avenue for NP effects to enter. TheB 0 q → + − decay amplitude has the following structure [1]: where λ = L, R describes the helicity of the final-state leptons with η L = +1 and η R = −1. The quantities where M Bq and m are theB 0 q and masses, respectively, will play a key role in the following discussion. In general, the coefficients P q ≡ |P q |e iϕ Pq and S q ≡ |S q |e iϕ Sq have CP-violating phases ϕ Pq and ϕ Sq . In the SM, we obtain the simple relations JHEP05(2017)156

Decay observables
The B 0 s andB 0 s mesons show the phenomenon of B 0 s -B 0 s mixing, which leads to timedependent decay rates. Experiments actually measure the following time-integrated branching ratio [15]: Here the time-dependent untagged rate, where no distinction is made between initially, i.e. at time t = 0, present B 0 s orB 0 s mesons, takes the following form [4,6,9]: where the decay width difference ∆Γ s enters through the parameter [16] y s ≡ ∆Γ s 2Γ s = 0.0645 ± 0.0045, (2.10) with τ Bs = 1/Γ s denoting the B 0 s lifetime. Using the quantities introduced above, the observable A ∆Γs is given as follows [4,6]: Since it is challenging to determine the helicity of the final-state leptons experimentally, the rates in (2.9) are actually helicity-averaged. The observable A ∆Γs takes the SM value A ∆Γs | SM = +1, (2.12) but is essentially unconstrained when allowing for NP effects [4,6,9]. In view of the sizeable y s , we have to properly distinguish between the time-integrated branching ratio B(B s → + − ) measured at experiments and the "theoretical" branching ratio B(B s → + − ) theo , which corresponds to the decay time t = 0. These two branching ratios can be converted into each other through the following relation [17]: The physics information encoded in the effective lifetime is equivalent to the observable A ∆Γs [4], which can be determined with the help of

JHEP05(2017)156
Moreover, τ s allows us to convert the time-integrated branching ratio determined at experiments into the "theoretical" branching ratio with the help of the relation where all quantities on the right-hand side can be measured [4,17]. In the case ofB 0 d → + − decays, y d takes a value at the 10 −3 level. Consequently, the corresponding observable A ∆Γ d is experimentally not accessible in the foreseeable future.
In addition to these untagged observables, there are also CP-violating asymmetries which would be very interesting to measure, providing insights into possible new sources for CP violation encoded in the Wilson coefficients [4,6]. The experimental analysis of these observables would require tagging information, thereby making it more challenging than the exploration of A ∆Γs . However, it would nevertheless be very interesting to make efforts in the super-high-precision era of B physics to get also a handle on these quantities.
In order to search for NP effects by means of the branching ratio of the B 0 s → + − decays, it is useful to introduce the following ratio [4,6]: which takes by definition the SM value Using the expressions given above yields Current experimental information from B 0 s → J/ψφ and decays with similar dynamics gives the following results [16,18,19]: where we have used the SM value φ SM s = −2β s = −(2.12 ± 0.04) • . Similar quantities can also be introduced for the B 0 d → + − decays, in analogy to the expressions given above.

Scenario for the new physics analysis
A first analysis of the interplay between R s µµ and A µµ ∆Γs within specific models of physics beyond the SM was performed in ref. [6], giving also a classification of various scenarios. In view of the new LHCb results for the B 0 s → µ + µ − mode, a very recent study was performed in ref. [9], highlighting also the importance of measuring A µµ ∆Γs for the search and exploration of NP effects.
In order to illustrate NP effects, we shall consider a general scenario with no new sources of CP violation, i.e. real Wilson coefficients. This assumption could be explored with the help of the CP-violating observables discussed in refs. [4,6]. Moreover, we assume that we have flavour-universal Wilson coefficients, allowing us to introduce the notation

10
(2.23) as well as . (2.25) Using data for rare B → K ( * ) µ + µ − decays, the latter coefficient can be determined from experimental data (for a state-of-the-art analysis, see ref. [20]). Data for B → K ( * ) e + e − modes allows us also to take a possible violation of Lepton Flavour Universality into account. As a working assumption, we shall use C 10 = 1, which is consistent with the current rare B-decay data within the uncertainties, and corresponds to a picture of NP entering only through new (pseudo)-scalar contributions, which is the key domain for the B 0 q → + − decays. We obtain then the following expressions: which will serve as the basis for our following discussion of NP effects. In particular, we shall not assume any relation between the P q µµ and S q µµ coefficients, which typically arise in more general NP frameworks as well as in specific models [6,13].
In ref. [9], a scenario with heavy new degrees of freedom, which are linearly realized in the electroweak symmetry in the Higgs sector, and the feature of MFV was considered [13], including the Minimal Supersymmetric Standard Model (MSSM) with MFV violation. In the latter case, the coefficients C P , C S are suppressed by the mass ratio m q /m b , and the relation C S = −C P holds. Moreover, these coefficients are proportional to the lepton mass m (see also refs. [6,12]): where A q does not depend on the lepton flavour . If we neglect the m 2 /M 2 Bq term under the square root in (2.30), both P q and S q are independent of the lepton mass in this scenario, implying that the ratios of branching ratios of the various B 0 s,d → + − decays are given as in the SM, up to O(m 2 /M 2 Bq ) corrections. In the case of the B 0 s,d → τ + τ − decays, these effects may have a sizeable impact, as we will discuss in subsection 4.2.
The flavour-universal scenario introduced above offers an interesting general framework to explore NP effects in the B 0 q → τ + τ − and B 0 q → e + e − decays and to illustrate their potential impact. But before focusing on these modes, let us first discuss the picture for the B 0 q → µ + µ − channels following from the current data.

Experimental status
Using the results of ref. [1] and rescaling them to the updated parameters collected in table 1, we obtain the following SM branching ratios: On the experimental side, the LHCb collaboration has recently presented updated measurements of the B 0 s → µ + µ − and B 0 d → µ + µ − branching ratios [5]: The CP-averaged signal for B 0 s → µ + µ − has a statistical significance of 7.8 σ, These experimental results are consistent with the SM predictions within the uncertainties. In 2013, the CMS collaboration reported the following result [24]:  where the error is fully dominated by the huge uncertainty on the effective lifetime τ s µµ . As we have the model-independent relation it will be crucial to improve the experimental precision for this observable in the future data taking at the LHC.

General constraints on new physics
Let us first have a look at the B 0 s → µ + µ − decay observables. Using eqs.  (2), (3.14) which allows us to convert eq. (3.12) into a circular band in the |P s µµ |-|S s µµ | plane. The observable A µµ ∆Γs provides another constraint in this parameter space. Assuming real coefficients P s µµ and S s µµ , eq. (2.11) yields where we have also given the simplified expressions for cos φ NP s = 1. In figure 3, we illustrate the resulting situation in the P s µµ -S s µµ plane, showing both the circular band arising from the current experimental value of R s µµ and the impact of a future measurement of the A µµ ∆Γs observable.

JHEP05(2017)156
In order to test the SM with the B 0 d → µ + µ − decay, it is advantageous to consider the ratio of its branching ratio and the one of B 0 s → µ + µ − [26]. We obtain the following general expression: (3.18) The CKM factor |V td /V ts | is required to utilize this ratio and has to be determined in a way that is robust with respect to the impact of NP effects. Assuming the unitarity of the CKM matrix, it can be extracted from the the length of the Unitarity Triangle (UT) as |V cb | = |V ts | + O(λ 2 ). Here λ ≡ |V us | is the Wolfenstein parameter [27], and (ρ,η) describes the apex of the UT in the complex plane [28]. Taking subleading corrections in λ into account and employing the UT side where γ is the angle between R b and the real axis.
Using pure tree decays of the kind B → D ( * ) K ( * ) [29][30][31], γ can be determined in a theoretically clean way (for an overview, see [32]). The current experimental value is given as follows [19]: In the future, thanks to Belle II [33] and the LHCb upgrade [34], the uncertainty for γ is expected to be reduced to the 1 • level. Concerning the R b side, it can be determined with the help of |V ub | and |V cb | extracted from analyses of exclusive and inclusive semileptonic B decays (for an overview, see the corresponding review in ref. [22]). The current status can be summarized as The determinations of γ and R b using pure tree decays are very robust with respect to NP effects. Consequently, they allow us to determine the ratio in eq. ; the narrow red sector illustrates the future improvement to 1 • precision. In the right panel, we add constraints from ∆M s /∆M d and CP violation in B 0 d → J/ψK S , and show the small black region for the apex following from the comprehensive fit of ref. [19]. that is also very robust concerning NP contributions, serving as the reference value for the analysis of eq. (3.21). The current data with the average value of R b in eq. (3.24) give V td V ts = 0.220 ± 0.010, (3.25) where the error may be reduced to 0.002 in the Belle II and LHCb upgrade era. We have illustrated the resulting situation for the UT in the complex plane in the left panel of figure 4. Thanks to the specific shape of the UT, we observe that the uncertainty of R t is fully governed by γ, while the uncertainty of R b has a minor impact. Consequently, also the discrepancy between the inclusive and exclusive determinations in eq. (3.23) has fortunately a negligible effect in this case. It is impressive to see the impact of the future extraction of γ, allowing a very precise determination of R t . For completeness, in the right panel of figure 4, we show other constraints in theρ-η plane following from ∆M s /∆M d and the determination of the CKM angle β = (21.6 ± 0.9) • through CP violation in B 0 d → J/ψK 0 S decays, taking penguin effects into account [18]. For comprehensive analyses of the UT, the reader is referred to refs. [19,35,36]. Using the CKM factor |V td /V ts | as determined through eq. (3.21), we may convert the measured ratio of the B 0 s,d → µ + µ − branching ratios into the following parameter:

JHEP05(2017)156
For NP models with MFV, which are characterized by universal short-distance functions, we have -with excellent accuracy -also a value of U ds µµ around one. A tiny difference may arise from the following small differences [22]: We shall return to this parameter within our general flavour-universal NP scenario in the next subsection (see eq. (3.38)). The current data give U ds µµ = 1.26 ± 0.49, (3.29) where the error is unfortunately too large to draw conclusions. At the end of the LHCb upgrade, corresponding to 50 fb −1 of integrated luminosity, LHCb expects to determine the with a precision at the 35% level [34]. Assuming a future measurement of τ s µµ , which determines A µµ ∆Γs through eq. (2.15), with a precision of 5% [4] and a reduction in the uncertainty of γ to 1 • would yield which would still not allow a stringent test in view of the significant uncertainty. We can straightforwardly generalize the observable U ds µµ defined in eq. (3.26) to neutral B 0 d,s decays with τ + τ − and e + e − leptons in the final state, as discussed below.
Should future measurements find a result for U ds µµ consistent with 1, thereby supporting the picture of the SM and models with MFV, we could extract the SU(3)-breaking ratio of "bag" parameters describing B 0 q -B 0 q mixing from the following relation (see also ref. [37]): allowing -in principle -an interesting test of lattice QCD. An agreement between experiment and theory would also support the lattice QCD calculation of the decay constants f Bq , which are key inputs for the SM branching ratios. However, even in the LHCb upgrade era, we would only get a precision for eq. (3.31) at the level of ±37%, while current lattice QCD calculations give the following picture [23]: In order to determine the ratio of the bag parameters using eq. (3.31) with the same relative error as the one achieved by the current lattice calculations in eq. (3.32), the measurement of B(B 0 d → µ + µ − )/B(B 0 s → µ + µ − ) should reach the 6% precision while having the 5% error in the measurement of the effective lifetime as for the LHCb upgrade. In this future scenario, we would be able to achieve a precision of ±0.06 for the determination of the observable U ds µµ , which would be interesting territory to search for signals of physics from beyond the SM.

New physics benchmark scenario
Let us now consider the situation in the general flavour-universal NP scenario introduced in subsection 2.3, which is characterized by eqs. (2.26) and (2.27), and assume that not only the ratio R s µµ but also the observable A µµ ∆Γs has been measured. Using eqs. (3.16) and (3.17), we may then determine the coefficients |P s µµ | and |S s µµ |, respectively, which allow us to extract the following ratios of short-distance coefficients: Since we can only determine the absolute values of P s µµ and S s µµ , which are real in our scenario, we have also to allow for negative values. We illustrate the corresponding situation in figure 5. As the current measurement of A µµ ∆Γs in eq. (3.10) does not yet provide a useful constraint, we vary this observable within its general range in eq. (3.11), yielding (3.35) Once the observable A µµ ∆Γs has been measured with higher precision, these allowed ranges can be narrowed down correspondingly. In ref. [13], constraints on similar coefficients were obtained.

JHEP05(2017)156
2. × 10 -9 2.5 × 10 -9 3. × 10 -9 3.5 × 10 -9 4. × 10 -9 6. which is consistent with the LHCb result in eq. (3.4). Finally, we obtain 0.97 ≤ U ds µµ ≤ 1.00 (3.38) for the parameter U ds µµ introduced in eq. (3.26). As expected from the discussion in the previous subsection, this quantity shows a small difference from one due to the mass differences in eq. (3.28) within the flavour-universal NP scenario. It will be very interesting to get much better measurements of the B 0 d → µ + µ − decay and to see whether they will be consistent with the picture given above. 4 The decays B 0 s → τ + τ − and B 0 d → τ + τ −

Observables
As is evident from eq. (2.9), the helicity suppression of the SM rates of the B 0 s,d → τ + τ − channels is essentially lifted through the large mass of the τ leptons. Within the SM, we JHEP05(2017)156 obtain the following predictions: In order to calculate these results, we have employed the analysis of ref. [1], and have used the values of CKM and non-perturbative parameters given in table 1.
It is experimentally very challenging to reconstruct the τ leptons, in particular in the environment of the LHC. Nevertheless, the LHCb collaboration has recently come up with the first experimental upper limits for the corresponding branching ratios [7]: These results are in fact the first direct constraint for B 0 s → τ + τ − and the world's best limit for B 0 d → τ + τ − . The SM predictions for A τ τ ∆Γs and τ s τ τ take the same values as their B 0 s → µ + µ − counterparts: A τ τ ∆Γs | SM = +1, τ s τ τ | SM = τ Bs 1 − y s = (1.61 ± 0.06) ps. (4.5)

New physics benchmark scenario
Let us now have a look at the NP effects for the B 0 s,d → τ + τ − modes within the benchmark scenario introduced in subsection 2.3. Here we obtain the following coefficients: Consequently, the NP correction to C 10 and those proportional to P s µµ and S s µµ are strongly suppressed through the ratio of the muon and tau masses, which is given as follows [21]: The impact of NP in the B 0 d → τ + τ − decay is very similar to its B 0 s counterpart, with R d τ τ taking the same values as in eq. (4.9). Introducing a parameter U ds τ τ in analogy to eq. (3.26), we obtain 1.000 ≤ U ds τ τ ≤ 1.002 (4.10) In view of the challenges related to the reconstruction of the τ leptons, the NP effects arising in the general flavour-universal NP scenario cannot be distinguished from the SM case, unless there is unexpected experimental progress.

JHEP05(2017)156
It is interesting to have a quick look at the picture in the MSSM with MFV described by eqs. (2.29) and (2.30). As was pointed out in ref. [9], the measured value of the ratio R s µµ gives a twofold solutions for the parameter A q introduced in eq. (2.29). We find A s = 0.09 ± 0.09 ∨ 0.98 ± 0.09, (4.11) corresponding to the observables A µµ ∆Γs ∼ +1 (as in the SM) and A µµ ∆Γs ∼ −1, respectively. These solutions give R s τ τ | MFV MSSM = 0.84 ± 0.16 ∨ 0.47 ± 0.09, (4.12) and correspond to the branching ratios respectively. We observe that the large τ -lepton mass has a significant impact on these quantities, in particular in the case A s ∼ 1.
5 The decays B 0 s → e + e − and B 0 d → e + e −

Observables
The most recent SM predictions for the B 0 s,d → e + e − decays were given in ref. [1]. Using the updated input parameters in table 1, we obtain the following results: The extremely small values of these branching ratios with respect to their B 0 s,d → µ + µ − counterparts arise from the helicity suppression due to the tiny electron mass, corresponding to an overall multiplicative factor m 2 e in the expressions for B(B s,d → e + e − ). Consequently, within the SM, these decays appear to be out of reach from the experimental point of view, which seems to be the reason for the fact that these channels have so far essentially not played any role in the exploration of the quark-flavour sector.
Concerning the experimental picture, the CDF collaboration reported the following upper bounds (90% C.L.) back in 2009 [8]: Consequently, any attempt to measure the SM branching ratios for the rare decays B 0 s → e + e − and B 0 d → e + e − would require a future improvement by nearly six orders of magnitude. The LHC experiments have not yet reported any searches for these modes.

New physics benchmark scenario
Let us now consider the flavour-universal NP scenario introduced in subsection 3.3. In this framework, we obtain the following coefficients: While we got a suppression of the NP effects in the B 0 s,d → τ + τ − decays through the large τ mass, (see (4.6) and (4.7)), we get now a huge enhancement thanks to the tiny electron mass [21]: It is particularly interesting to consider the ratio between the B 0 s → e + e − and the B 0 s → µ + µ − branching ratios: where we used cos φ NP s = 1 (see eq. (3.14)) and neglected the effects associated with y s and the tiny mass ratio m e /m µ ∼ 0.005. As the decay constants and CKM matrix elements  Figure 7. Illustration of the allowed range for the ratio R ee s,µµ as a function ofR s µµ in the flavouruniversal NP scenario with C 10 = 1. We show also contours for various values of A µµ ∆Γs . As in figure 6, the green and red bands correspond to positive and negative solutions for P s µµ , respectively. cancel in R ee s,µµ , this ratio is a theoretically clean quantity. Moreover, its measurement at the LHC is not affected by the ratio f s /f d of fragmentation functions [38], which is an advantage from the experimental point of view.
It is instructive to consider a situation with C 10 = 1 and (P s µµ ) 2 + (S s µµ ) 2 ≈ 1, (5.13) which corresponds to a B 0 s → µ + µ − branching ratio as in the SM and is consistent with the current experimental situation. We obtain then R ee s,µµ ≈ 2(1 − P s µµ ), (5.14) which yields 0 ∼ < R ee s,µµ ∼ < 4 (5.15) as P s µµ varies between −1 and +1 while moving on the unit circle in the P s µµ -S s µµ plane. It is also interesting to note that C 10 = 0 would result in R ee s,µµ ≈ 1 independently of the values of P s µµ and S s µµ on the unit circle, as can be seen in eq. (5.12). These simple considerations illustrate nicely the possible spectacular enhancement of the B 0 s → e + e − branching ratio with respect to the SM prediction.
In figure 7, we put these considerations on a more quantitative ground, showing the allowed region for R ee s,µµ as a function of R s µµ . We give also contours for various values of
In the case of the branching ratio of the B 0 s → e + e − channel, we obtain uncertainties from the decay constant f Bs , CKM factors and the B s lifetime τ Bs . The range in eq. In analogy to the B 0 s → τ + τ − case, it is instructive to have also a look at the MSSM with MFV, which is described by eqs. (2.28)-(2.30) and the solutions in eq. (4.11). In this framework, we get where we have neglected tiny m 2 µ /M 2 Bs and m 2 e /M 2 Bs corrections. Consequently, the ratio of the branching ratios is as in the SM, and we obtain from the measured B 0 s → µ + µ − branching ratio: It will be very interesting to search for the B 0 s,d → e + e − decays, in particular in view of the exciting situation that the CDF upper bound from 2009 is only about a factor of 20 above the upper bound (5.17) in the flavour-universal NP scenario. Should the B 0 s,d → e + e − decays actually be observed with hugely enhanced branching ratios, the MSSM with MFV would be ruled out.

Conclusions
Leptonic rare decays of B 0 s and B 0 d mesons play an outstanding role for testing the SM. The main actors have so far been the B 0 s → µ + µ − and B 0 d → µ + µ − modes, where the former decay is now well established in the LHC data and first signals for the latter channel were reported. Very recently, LHCb has presented the first measurement of the effective lifetime of B 0 s → µ + µ − , and upper bounds for the B 0 s → τ + τ − and B 0 d → τ + τ − modes.