Issue 
A&A
Volume 580, August 2015



Article Number  A9  
Number of page(s)  11  
Section  Astrophysical processes  
DOI  https://doi.org/10.1051/00046361/201526344  
Published online  20 July 2015 
Theoretical uncertainties in extracting cosmicray diffusion parameters: the borontocarbon ratio
LAPTh, Université Savoie MontBlanc & CNRS, 9 chemin de Bellevue, BP 110, 74941 AnnecyleVieux, France
email: yoann.genolini@lapth.cnrs.fr
Received: 17 April 2015
Accepted: 7 May 2015
Context. PAMELA and, more recently, AMS02, are ushering us into a new era of greatly reduced statistical uncertainties in experimental measurements of cosmicray fluxes. In particular, new determinations of traditional diagnostic tools such as the borontocarbon ratio (B/C) are expected to significantly reduce errors on cosmicray diffusion parameters, with important implications for astroparticle physics, ranging from inferring primary source spectra to indirect dark matter searches.
Aims. It is timely to stress, however, that the conclusions obtained crucially depend on the framework in which the data are interpreted as well as on some nuclear input parameters. We aim at assessing the theoretical uncertainties affecting the outcome, with models as simple as possible while still retaining the key dependencies.
Methods. We compared different semianalytical, twozone model descriptions of cosmicray transport in the Galaxy: infinite slab(1D), cylindrical symmetry (2D) with homogeneous sources, cylindrical symmetry (2D) with inhomogeneous source distribution. We tested for the effect of a primary source contamination in the boron flux by parametrically altering its flux. We also tested for nuclear crosssection uncertainties. All hypotheses were compared by χ^{2} minimisation techniques to preliminary results from AMS02.
Results. We find that the main theoretical bias on the determination of the diffusion coefficient index δ (up to a factor two) is represented by the assumption that no injection of boron takes place at the source. The next most important uncertainty is represented by crosssection uncertainties, which reach ± 20% in δ. As a comparison, nuclear uncertainties are more important than the shift in the bestfit when introducing a convective wind of velocity ≲30 km s^{1}, with respect to a pure diffusive baseline model. Perhaps surprisingly, homogeneous 1D vs. 2D performances are similar in determining diffusion parameters. An inhomogeneous source distribution marginally alters the central value of the diffusion coefficient normalisation (at the 10%, 1σ level). However, the index of the diffusion coefficient δ is basically unaltered, as well as the goodness of fit.
Conclusions. Our study suggests that, differently for instance from the leptonic case, realistic modelling of the geometry of the Galaxy and of the source distribution are of minor importance to correctly reproduce B/C data at high energies and thus, to a large extent, for the extraction of diffusion parameters. The ansatz on the lack of primary injection of boron represents the most serious bias and requires multimessenger studies to be addressed. If this uncertainty could be lifted, nuclear uncertainties would still represent a serious concern; they degrade the systematic error on the inferred parameters to the 20% level, or three times the estimated experimental sensitivity. To reduce this, a new nuclear crosssection measurement campaign might be required.
Key words: astroparticle physics / cosmic rays / diffusion / methods: analytical
© ESO, 2015
1. Introduction
The pattern of relative abundances of nuclei in the cosmic radiation (CR) is roughly similar to the one of the solar system material, with some notable exceptions: fragile nuclei (with low binding energies) such as H or LiBeB are overrepresented in CR. This CR component is usually interpreted as the result of production by spallation of heavier species during the propagation of primary cosmic rays – whose injected abundance is assumed to closely trace that of the solar system – in the interstellar medium. The ratios of these secondary to primary fluxes have long been recognised as a tool for constraining CR propagation parameters, for some review see for instance Maurin et al. (2002) and Strong et al. (2007). The borontocarbon ratio (B/C), represents the most notable example among them. The key constraints on the diffusion parameters are inferred by its measurement, with the corresponding confidence levels (see for instance Maurin et al. 2001) widely used as benchmarks. It has also been recognised that datasets available one decade ago were still insufficient for a satisfactory measurement of these parameters (Maurin et al. 2001).
The current decade is undergoing a major shift, however, with experiments such as PAMELA (Adriani et al. 2014) and most notably AMS02^{1}, which are characterised by significantly increased precision and better control of systematics. The current and forthcoming availability of highquality data prompts the question of how best to exploit them to extract meaningful (astro)physical information. This new situation demands reassessing theoretical uncertainties, which will probably be the limiting factor in the parameter extraction accuracy. As a preliminary work, preceding the actual data analysis, we revisit this issue to determine the relative importance of various effects: some have already been considered in the past, some were apparently never quantified. We also found that the main theoretical biases or errors are related to phenomena that can be described in a very simple 1D diffusive model. We thus adopt it as a benchmark for our description, reporting the key formulae that thus have a pedagogical usefulness, too. In fact, we focus on determining the diffusion coefficient, which we parameterise as conventionally in the literature (see for example Ptuskin et al. 1997): (1)where D_{0} and δ are determined by the level and powerspectrum of hydromagnetic turbulences, ℛ is the rigidity, and the velocity β = v/c ≃ 1 in the highenergy regime of interest here (kinetic energy/nucleon ≳10 GeV/nuc). In fact, at lower energies numerous effects, in principle of similar magnitude, are present, such as convective winds, reacceleration, and collisional losses. At high energy, there is a common consensus that only diffusion and sourcerelated effects are important. We focus on the highenergy region since it is the cleanest to extract diffusion parameters, that is the least subject to parameter degeneracies. While adding lowerenergy data can lead to better constraints from a statistical point of view, the model dependence cannot but grow. Since our purpose is to compare theoretical with statistical uncertainties from observations, our choice is thus conservative: in a global analysis, the weight of the former with respect to the latter is probably larger. To deal with a realistic level of statistical errors of the data that will be available for the forthcoming analyses, we base our analyses on preliminary AMS02 data of the B/C ratio (Oliva et al. 2013).
This paper is organised as follows. In Sect. 2 we recall a simple 1D diffusion model providing our benchmark for the following analyses. This model certainly has pedagogical value, since it allows encoding the main dependences of the B/C ratio on input as well as astrophysical parameters in simple analytical formulae . At the same time, it provides a realistic description of the data, at least if one limits the analysis to sufficiently high energies. Relevant formulae are introduced in Sect. 2.1, while in Sect. 2.2 we recall the main statistical tools used for the analysis. In Sect. 3 we describe the main degeneracy affecting the analysis: the one with possible injection of boron nuclei at the sources. The next most important source of error is associated to crosssection uncertainties, to which we devote Sect. 4. In Sect. 5 we discuss relatively minor effects linked to modelling of the geometry of the diffusion volume, source distribution, or the presence of convective winds. In Sect. 6 we report our conclusions.
2. B/C fit with a 1D model
2.1. 1D diffusion model
The simplest approach to model the transport of cosmicray nuclei inside the Galaxy is to assume that their production is confined inside an infinite plane of thickness 2h, that is sandwiched inside an infinite diffusion volume of thickness 2H, symmetric above and below the plane. The former region stands for the Galactic disk, which comprises the gas and the massive stars of the Milky Way, whereas the latter domain represents its magnetic halo. A sketch of this model is given in Fig. 1. The boundary conditions fix the density of cosmic rays at the halo edges z = ± H to zero, while the condition h ≪ H (in practice, h is almost two orders of magnitude smaller than H) allows us to model the Galactic matter distribution as an infinitely thin disk whose vertical distribution is accounted for by the Dirac function 2hδ(z). Our focus on energies above 10 GeV/nuc allows us to neglect continuous (ionisation and Coulomb) energy losses, electronic capture, and reacceleration. These subleading effects cannot be truly considered as theoretical uncertainties, since they can be introduced by a suitable upgrade of the model. However, taking them into account at this stage would imply a significant loss in simplicity and transparency.
Fig. 1 Sketch of the 1D slab model of the Galaxy, with matter homogeneously distributed inside an infinite plane of thickness 2h sandwiched between two thick diffusive layers of thickness 2H. 

Open with DEXTER 
The wellknown propagation equation for the (isotropic part of the gyrophaseaveraged) phase space density ψ_{a} of a stable nucleus a, with charge (atomic number) Z_{a}, expressed in units of particles cm^{3} (GeV/nuc)^{1}, takes the form (2)where the spatial diffusion coefficient D has been defined in Eq. (1). The crosssection for the production of the species a from the species b through its interactions with the interstellar medium (ISM) is denoted by σ_{b → a}, whereas σ_{a} is the total inelastic interaction (destruction) crosssection of the species a with the ISM. The fragmentation of the nucleus b takes place at constant energy per nucleon. v hence stands for the velocities of both parent (b) and child (a) nuclei. The surface density of the Galactic disk is denoted by μ, while m_{ISM} is the average mass of the atomic gas that it contains. The values of the production crosssections σ_{b → a} were calculated with the most recent formulae from Webber et al. (2003). The destruction crosssections σ_{a} were computed by the semiempirical formulae of Tripathi et al. (1997, 1999). The highenergy shapes of both crosssections exhibit a plateau that allows one to approximate them as constants in this energy range.
Solving the propagation Eq. (2) in the steadystate regime allows expressing the flux of the stable nucleus a inside the Galactic disk (z = 0) as (3)where The fluxes of the parent species are also taken at z = 0. Here Q_{a}, which stands for the source term, is homogeneous to a flux times a surface and is expressed in units of particles (GeV/nuc)^{1} s^{1} sr^{1}. It is related to q_{a} through (4)where N_{a} is a normalisation constant that depends on the isotope a. We assumed an injection spectrum with the same spectral index α for all nuclei. The value of N_{a} should be adjusted by fitting the corresponding flux to the measurements performed at Earth. However, these scarcely contain information on the isotopic composition of cosmic rays. Nuclei with the same charge Z are in general collected together, irrespective of their mass. More isotopic observations would be necessary to set the values of the coefficients N_{a} for the various isotopes a of the same element.
In our analysis, we assumed solar system values (Lodders 2003) for the isotopic fractions f_{a} of the stable species a that were injected at the sources. We then proceeded by computing the flux of each element Z at Earth. We fixed the normalisation N_{Z} for the total injection of all stable isotopes of the same charge Z by fitting the measured flux of that element. The normalisation entering in the calculation of Q_{a} is given by N_{a} = f_{a}·N_{Z}, the sum of the fractions f_{a} corresponding to the same element Z amounting to 1. The actual isotopic composition of the material accelerated at the source might be different from the solar system one, as is the case for neon (Binns et al. 2008). Our method might introduce a theoretical bias in CR element flux calculations. However, our main focus here is to extract the propagation parameters thanks to the different sensitivities between primary carbon and (a priori) secondary boron. The only isotopes that come into play in the B/C ratio are the stable nuclei ^{12}C, ^{13}C, ^{10}B, and ^{11}B, unstable ^{14}C plays a very minor role. The isotopes of either carbon or boron have similar rigidities and destruction crosssections. Varying the isotopic composition of carbon (and of boron, should it be partially primary) does not affect the ratio calculation. Futhermore, secondary boron is mainly produced by the fragmentation of one particular isotope of each heavier element. For example, the primary component of ^{12}C is two orders of magnitude larger than that of ^{13}C. This reduces the differences arising from the boron production crosssections. Althought most of the isotopes at stake are stable, radioactive nuclei were also taken into account in the calculation, and we obtained more complicated expressions for the fluxes, which are not displayed here for brevity. They are reported for instance in Appendix A of Putze et al. (2010). By defining the total flux of a nucleus of charge Z as the sum over all its isotopes a(5)and considering only the dominant contribution from stable nuclei, the B/C flux ratio can be written as (6)If we assume that there are no primary boron sources, that is, Q_{B} = 0, this expression simplifies into (7)The impact of relaxing this hypothesis is explored in Sect. 3 where the effect of a nonvanishing value for Q_{B} is considered.
2.2. Fitting procedure and benchmark values for this study
We used the AMS02 recent release of the B/C ratio (Oliva et al. 2013) to study the impact of systematics on the propagation parameters. As explained above, we limited ourselves to the highenergy subsample, above 10 GeV/nuc. The set of Eqs. (3) is of triangular form. The heaviest element considered in the network, which in our case is ^{56}Fe, can only suffer destruction. No other heavier species b enters in the determination of its flux , which hence is proportional to the injection term Q_{a}. Once solved for it, the algebraic relation yields the solution for the lighter nuclei, down to boron. We evaluated the cascade down to beryllium to take into account its radioactive decay into boron.
The primary purpose of our analysis is to determine the diffusion parameters D_{0} and δ from the B/C flux ratio . Another parameter of the model is the magnetic halo thickness H. As shown in Eq. (3), D_{0} and H are completely degenerate when only considering stable nuclei, which provide the bulk of cosmic rays. In the following, H is therefore fixed at 4 kpc for simplicity, although it should be kept in mind that, to a large extent, variations in D_{0} can be traded for variations in H. Finally, the injection spectral index α also enters in the calculation of the B/C ratio through the source terms Q_{a}. How strong its effect is on the bestfit diffusion parameters D_{0} and δ is one of the questions we treat in this section. To this purpose, we carried out a chisquare (χ^{2}) analysis of the B/C observations and minimised the function (8)where the sum runs over the data points i whose kinetic energies per nucleon are E_{k,i}, while and σ_{i} stand for the central values and errors of the measurements. The theoretical expectations also depend on the normalisation constants N_{a}, which come into play in the source terms Q_{a} of the cascade relations (3). To determine them, we first fixed the spectral index α and the diffusion parameters D_{0} and δ. We then carried out an independent χ^{2}based fit on the fluxes of the various elements that belong to the chain that reaches from iron to beryllium. The measured fluxes are borrowed from the cosmicray database of Maurin et al. (2014) from which we selected the points above 10 GeV/nuc. As explained above, this method yields the constants N_{Z} and eventually the values of N_{a} once the solar system isotopic fractions f_{a} are taken into account. The overall procedure amounts to profile over the normalisation constants N_{a} to derive as a function of α, δ and D_{0}. Minimisations were performed by MINUIT^{2}, a package interfaced in the ROOT programme^{3}.
Fig. 2 Relative variations of the bestfit propagation parameters (compared to the benchmark model of Table 1) with respect to the injection spectral index α. 

Open with DEXTER 
To check the accuracy and robustness of our fitting procedure, a preliminary test is in order. A commonly accepted notion is that the B/C ratio does not depend, to leading order, on the spectral index α. There is indeed no dependence on α in the crosssection ratios of Eq. (7) in the pure diffusive regime where σ_{B} ≪ σ^{diff}. We have checked numerically that this behaviour holds by calculating the B/C bestfit values of the diffusion parameters at fixed spectral index α. The results are reported in Fig. 2, where D_{0} and δ are plotted, with their confidence limits, as a function of α. We scanned over the physical range that extends from −2.5 to −2 and observed that the relative variations of D_{0} and δ are 5% and 1%, respectively. The blue (D_{0}) and red (δ) bands are almost horizontal. An anticorrelation between D_{0} and δ is marginally noticeable and can be understood by the interplay of these parameters inside the diffusion coefficient D, the only relevant parameter that the B/C fit probes. We attribute the small variation of D_{0} with α to the different sensitivities of the normalisation constants N_{Z} of nitrogen and oxygen to the lowenergy data points as compared to carbon. This could result in fluctuations of the N_{N}/N_{C} and N_{O}/N_{C} ratios with respect to the actual values. In any case, the extremely small dependence of the B/C ratio on α confirms the naive expectations and suggests that it is useless and simply impractical to keep α as a free parameter.
Nonetheless, there is a particular value of the injection index that best fits the fluxes of the elements Z that come into play in the cascade from iron to beryllium. By minimising the χ^{2}function (9)we find α = −2.34 as our benchmark value. Applying then our B/C analysis yields the propagation parameters D_{0} and δ of the reference model of Table 1 which we used for the following analyses. The corresponding B/C ratio is plotted in Fig. 3 as a function of kinetic energy per nucleon and compared to the preliminary AMS02 measurements (Oliva et al. 2013). In what follows, we study how D_{0} and δ are affected by a few effects under scrutiny and gauge the magnitude of their changes with respect to the reference model. We could have decided to keep the injection index α equal to its fiducial value of –2.34, but we preferred to fix the spectral index γ = α−δ = −2.78 of the highenergy fluxes at Earth. Keeping α fixed would have little effect on the B/C ratio, but would degrade the goodness of the fits on absolute fluxes.
Benchmark bestfit parameters of the 1D/slab model, with respect to which comparisons are subsequently made.
Fig. 3 Preliminary AMS02 measurements of the B/C ratio (Oliva et al. 2013) are plotted as a function of kinetic energy per nucleon. The theoretical prediction of the 1D/slab reference model of Table 1 is also featured for comparison. 

Open with DEXTER 
Fig. 4 Relative variations of the bestfit propagation parameters, as compared to the benchmark model of Table 1, with respect to the lowenergy cutoff E_{cut} above which we carry out the B/C analysis. 

Open with DEXTER 
Another crucial test of our fitting procedure is to check how the results depend on the lowenergy cutoff E_{cut} above which we carried out our analysis. We set the flux spectral index γ to its benchmark value of Table 1 and determined the B/C bestfit values of the diffusion parameters as a function of E_{cut}, which was varied from 5 to 30 GeV/nuc. The results are plotted in Fig. 4 with the 1σ and 2σ uncertainty bands. As expected, the statistical errors increase when moving from a low E_{cut} to a higher value. That is why the reduced χ^{2} (dashed line) decreases steadily as the cutoff energy is increased. The higher the cosmicray energy, the fainter the flux and the scarcer the events in the detector. The widths of the blue (D_{0}) and red (δ) bands at E_{cut} = 10 GeV/nuc, however, are not significantly larger than for a cutoff energy of 5 GeV/nuc. This suggests that our estimates for the statistical errors are slightly pessimistic, which is acceptable and consistent with our purpose.
The other trend that we observe in Fig. 4 is a shift in the preferred value of δ to increasingly lower values as we limit the analysis to increasingly higher energies. This is no limitation of our procedure. On the contrary, it is a real feature that the data exhibit, as is clear in Fig. 3, where the tail of the B/C points does look flatter above 50 GeV/nuc. The anticorrelation between δ and D_{0} that we observe in Fig. 4 has already been explained by the interplay of these two parameters inside the diffusion coefficient D, to which the B/C ratio is sensitive. The increase of D_{0} is then generic and does not signal any new effect. At that stage, the statistical uncertainties are still of the same order as the systematic uncertainties generated by using different energy cuts. Should the decrease of δ with E_{cut} be confirmed with higher statistics, some intrinsic explanation might be necessary for the failure of a powerlaw fit. See for instance Sect. 3 for a possible explanation.
3. Primary boron?
Typical fits of the B/C ratio are based on the assumption that no boron is accelerated at the source, so that the term proportional to Q_{B} at the righthand side of Eq. (6) vanishes. However, this is just an assumption that need to be tested empirically. It is crucially linked to the hypothesis that the acceleration time is much shoter than the propagation time within the magnetic halo and that it occurs in a lowdensity environment. On the other hand, typical astrophysical accelerators such as supernova remnants might have the capability to accelerate up to TeV energies for of about t_{life} ~ 10^{5} years in an interstellar medium with n_{ISM} ~ 1 cm^{3}, or greater when surrounded by denser circumstellar material. The corresponding surface density n_{ISM}ct_{life} ~ 10^{23} cm^{2} easily leads to percentlevel probabilities for nuclei to undergo spallation in the sources. A factor of only a few times higher than this would certainly have dramatic consequences on the information inferred from secondarytoprimary ratios. More elaborate versions of this idea and related phenomenology have also been detailed as a possible explanation of the hard spectrum of secondary positron data (Blasi 2009; Blasi & Serpico 2009; Mertsch & Sarkar 2009), which was recently compared with the AMS02 data (Mertsch & Sarkar 2014).
Fig. 5 Left panel: variations of the bestfit propagation parameters D_{0} (blue) and δ (red) relative to the benchmark values of Table 1, as a function of the primary borontocarbon injection ratio. The reference model corresponds to the conventional no boron hypothesis for which N_{B}/N_{C} vanishes. Right panel: the theoretical value of the B/C ratio at 214 GeV/nuc (solid red curve) is plotted as a function of the primary borontocarbon injection ratio. The dashed black curve indicates the goodness of the B/C fit. As long as N_{B}/N_{C} does not exceed 13%, the theoretical B/C ratio is within 2σ from the AMS02 measurement (dasheddotted green curve). 

Open with DEXTER 
Apparently little attention has been paid to the bias introduced by the ansatz Q_{B} = 0. To the best of our knowledge, we quantify it here for the first time. As can be inferred from Eq. (6), in the presence of a primary source Q_{B}, the B/C ratio exhibits a plateau as soon as the crosssection ratio σ_{C → B}/ (σ^{diff} + σ_{B}) becomes negligible with respect to the primary abundances ratio N_{B}/N_{C}. This happens at sufficiently high energy since σ^{diff} increases with the diffusion coefficient D. The height of this highenergy B/C plateau is approximately given by the value of N_{B}/N_{C}. In the presence of this behaviour, the spectral index δ must increase to keep fitting the data at low energy, that is, here around 10 GeV/nuc. This also implies that D_{0} decreases with N_{B}/N_{C} as a result of the abovementioned anticorrelation between the diffusion parameters.
We have thus scanned the borontocarbon ratio at the source and studied the variations of the bestfit values of D_{0} and δ with respect to the reference model of Table 1. Our results are illustrated in Fig. 5, where the left panel features the confidence levels for δ (red) and D_{0} (blue) as a function of the N_{B}/N_{C} ratio. The B/C fit is particularly sensitive to the last few AMS02 points, notably the penultimate data point, around 214 GeV/nuc, for which the B/C ratio is found to be ~9%. In the right panel, the theoretical expectation for that point is plotted (solid red curve) as a function of the primary abundances ratio, while the dashed black curve indicates how the goodness of fit varies. It is interesting to note that a minor preference is shown for a nonvanishing fraction of primary boron, around 8%, due to the marginal preference for a flattening of the ratio already mentioned in the previous section. The N_{B}/N_{C} ratio is only loosely constrained to be below 13%. Such a loose constraint would nominally mean that a spectral index δ more than three times larger than its benchmark value would be allowed, with a coefficient D_{0} one order of magnitude smaller than indicated in Table 1. In fact, such changes are so extreme that they would clash with other phenomenological or theoretical constraints and should probably be considered as unphysical. A spectral index δ in excess of 0.9, corresponding to a relative increase of 100% with respect to our benchmark model, is already so difficult to reconcile with the powerlaw spectrum of nuclei and the present acceleration schemes that it would probably be excluded. The message is quite remarkable however. The degeneracy of the diffusion parameters with a possible admixture of primary boron is so strong that it dramatically degrades our capability of determining the bestfit values of D_{0} and δ, and beyond them the properties of turbulence, unless other priors are imposed.
4. Crosssection modelling
The outcome of cosmicray propagation strongly depends on the values of the nuclear production σ_{b → a} and destruction σ_{a} crosssections with the ISM species, mainly protons and helium nuclei. Some of these are measured, albeit in a limited dynamical range, while a significant number of them rely on relatively old semiempirical formulas, calibrated to the few available data points. In this section, we discuss how parametric changes in these inputs reflect on the B/C ratio. The effect of crosssection systematics was already studied by Maurin et al. (2010), who parameterised it in terms of a systematic shift with respect to the energy. Since we consider here only the highenergy limit, we simply allowed for a rescaling of the crosssections. However, we distinguished between two cases: a correlated (↗↗) or anticorrelated (↗↘) rescaling between the production σ_{b → a} and the destruction σ_{a} crosssections. These in fact are not affected by the same uncertainties. It is often the case that the latter are known to a better precision then the former since they rely on a richer set of data. A priori, it is conceivable that several relevant production crosssections might be varied independently. It is worth noting, however, that only a few nuclei – notably oxygen and carbon (~80%), and to a lesser extent nitrogen (~7%) – are in fact responsible for most of the produced boron, as shown in Fig. 6.
Fig. 6 Contribution of the various primary nuclear species to the secondary boron flux at 10 GeV/nuc, as estimated with the semiempirical code by Webber 03. 

Open with DEXTER 
First, we need to assess the reasonable range over which the various crosssections of the problem are expected to vary. For this, we compared our reference models for the destruction and production crosssections with those used in popular numerical propagation codes such as GALPROP (Strong & Moskalenko 2001) and DRAGON (Evoli et al. 2008)^{4}. The database implemented these two codes traces back to the GALPROP team work and is based on a number of references including – but not limited to – Nuclear Data Sheets and Los Alamos database Mashnik et al. (1998, see Moskalenko et al. 2001 and Moskalenko & Mashnik 2003 for a more complete list of references). In this work we compare the values given directly by the default crosssection parameterisations without any renormalisation (which can be implemented however).
In the case of the destruction crosssections σ_{a}, we compared our reference model (Tripathi et al. 1997) with the parameterisations of Barashenkov & Polanski (1994), Letaw et al. (1983) and Wellisch & Axen (1996). The last case only applies to elements with Z> 5, while the Letaw et al. (1983) modelling is conserved for lighter nuclei. Figure 7 shows the relative differences between our reference model and the three other semiempirical approaches and allows deriving an indicative lower limit on the systematic uncertainties for the destruction crosssections of roughly 2 to 10% for the B/C ratio. The systematic difference is at the 3% level for the channels (CNO) that contribute most to secondary boron production. The difference to our reference model is stronger for larger charges (Z> 10), but these nuclei have a negligible contribution to the B/C ratio.
Fig. 7 Relative differences between our reference model (Tripathi et al. 1997) for the destruction crosssections σ_{a} and the other parameterisations by Letaw et al. (1983), Wellisch & Axen (1996) and Barashenkov & Polanski (1994) are displayed as a function of the nucleus charge, at an energy of 10 GeV/nuc. Each bin is characterised by a given charge Z and encodes the arithmetic mean over the corresponding isotopes. Only the elements involved in the cascade from iron to beryllium are displayed. 

Open with DEXTER 
Fig. 8 2D histograms of the relative differences between various semiempirical models currently used to calculate the production crosssections σ_{b → a}. Our reference model is Webber et al. (2003) (Webber 03), and we compare it to the parameterisations from Webber et al. (1990) (Webber 93), Webber et al. (1998) (Webber 98) and Silberberg et al. (1998) (S&T 00). The charges of the parent and child nuclei are given on the vertical and horizontal axes, respectively. The relative difference in each bin is given by the arithmetic mean over the various isotopes of each element. A detailed view provides the most important channels for the B/C ratio studies. For a fragmentation of ΔZ< 4, we also give the first and second moments of the uncertainty distributions. 

Open with DEXTER 
For the production crosssections σ_{b → a}, one may chose between the semiempirical approach proposed by Silberberg et al. (1998), subsequently revised in 2000 and called here S&T 00, and the parameterisation provided by Webber et al. (1990, hereafter Webber 93) and its updates of 1998 (Webber et al. 1998) and 2003 (Webber et al. 2003). We selected the last set of values as our reference model, to which we have compared the other parameterisations to gauge the uncertainties that affect, on average, the values of σ_{b → a}. The relative differences between Webber 93, Webber 98, and S&T 00 with respect to Webber 03 are plotted in the form of the three histograms of Fig. 8. The charges of the parent and child nuclei are given on the vertical and horizontal axes. The most important reactions, whose crosssections are higher, correspond to a change of charge ΔZ not in excess of 3 during the fragmentation process and are located close to the diagonals of the 2Dgrids of Fig. 8. We first note that the Webber 93 and 98 production crosssections are on average larger than the values of the Webber 03 reference model. Most of the pixels on the diagonals of the corresponding histograms are red, and we measured an excess μ on the reactions for which ΔZ< 4 of 18% and 9.7% for Webber 93 and Webber 98 as compared to Webber 03. Furthermore, in both cases the dispersion of these differences is quite large and amounts to 31% for Webber 93 and 30% for Webber 98. A rapid comparison between S&T 00 and Webber 03 would also leave the impression that in the former case, the reactions in the upper left corner of the histogram have crosssections exceedingly larger than for the Webber 03 parameterisation. A close inspection along the diagonal indicates, on the contrary, that the S&T 00 values for ΔZ< 4 are on average 13% higher than for the reference model, with a dispersion σ of 28% similar to the other cases. The main production channels of secondary boron are listed in Table 2 and are also displayed in the expanded views of the small square regions that sit in the lower left corners of the histograms of Fig. 8. The most relevant reactions involve the stable isotopes of carbon, nitrogen, and oxygen fragmenting into ^{10}B and ^{11}B, and are indicated in boldface in Table 2. The largest contributor to secondary boron is ^{12}C. The three semiempirical models with which we compared our Webber 03 reference parameterisation tend to predict production crosssections that are 15% for S&T 00 to 25% for Webber 93 larger. In contrast, those models underpredict the spallation of ^{16}O by 10% in the case of Webber 93 and 98 to 18% for S&T 00. In the latter case, the production crosssection of ^{10}B from ^{14}N is 68% larger than for Webber 03. But nitrogen only contributes ~7% of the secondary boron, and this has no significant impact. To summarise this discussion, the production crosssections σ_{b → a} can be varied up or down by a factor of order 10–20% with respect to Webber 03.
Comparison between different crosssection parameterisations for the main production channels of secondary boron.
Varying the various production and destruction crosssections has an effect on the calculation of the B/C ratio and thus affects the determination of the propagation parameters D_{0} and δ. Before gauging this effect, we remark that secondary boron is essentially produced by CNO nuclei, as indicated in Fig. 6. These are essentially primary species for which is approximately given by the ratio Q_{b}/ (σ^{diff} + σ_{b}) and is proportional to the injection normalisation N_{b}. Furthermore, the relevant destruction crosssections σ_{C}, σ_{N} and σ_{O} being approximately equal to each other, with an effective value ranging from 290 to 317 mb, we conclude that the flux ratios are given by the corresponding ratios N_{b}/N_{C} of the injection normalisation constants, with the consequence that relation (7) simplifies to (10)As mentioned at the beginning of this section, we first rescaled in our code all production σ_{b → a} and destruction σ_{a} crosssections by the same amount κ, which ranges from 0 to 2, to study how D_{0} and δ are affected by this change. The results are summarised in the left panel of Fig. 9. The diffusion index δ does not suffer any change, whereas the diffusion normalisation D_{0} increases linearly with the rescaling factor κ. Multiplying both σ_{b → B} and σ_{B} by the same factor κ in Eq. (10) amounts to dividing the diffusion cross section σ^{diff} by κ. The B/C ratio depends then on the ratio σ^{diff}/κ, which scales as D_{0}/κ. The theoretical prediction on the B/C ratio is not altered as long as that ratio is kept constant, hence the exact scaling of D_{0} with κ displayed in the left panel of Fig. 9. The energy behaviour of the B/C ratio is not sensitive to the rescaling factor κ, which has been absorbed by D_{0}, and the fit yields the same spectral index δ irrespective of how much the crosssections have been changed. Despite the relatively modest alterations, the effect discussed here has two qualitatively interesting consequences. To commence, a systematic uncertainty on the central value of D_{0} at the 5 to 10% level seems unavoidable due to the current uncertainty level of about 10% on the nuclear crosssections. Then, fully correlated changes in both production and destruction crosssections can break the degeneracy between D_{0} and δ.
Fig. 9 Effect of rescaling nuclear crosssections for boron production channels and destruction ones: the left panel assumes correlated, the right panel anticorrelated rescaling. 

Open with DEXTER 
We now analyse the effects of an anticorrelated change of the production σ_{b → a} and destruction σ_{a} crosssections. Surprisingly, this has never been considered before, as far as we know, although the potential effect of this rescaling clearly is very strong. Multiplying σ_{b → B} by a factor κ while rescaling σ_{B} by a complementary factor of (2−κ) leads to the B/C ratio (11)Keeping the B/C ratio constant while increasing κ at a given energy translates into keeping the ratio (12)roughly constant, where C is a constant directly proportional to D_{0}. It can be immediately inferred that, when κ increases, C and D_{0} have to increase and thus δ has to decrease. This trend is confirmed in the right panel of Fig. 9. From realistic assessments of the minimum systematic uncertainties of about 10% derived from the different crosssection models, we estimate a systematic uncertainty of 10% on δ and of 40% on D_{0}.
5. Systematics related to CR propagation modelling
A significant effort has been made in recent years to provide increasingly sophisticated modelling of the CR diffusion environment, source distribution, and alternative forms of CR transport. In this section we discuss a perhaps surprising conclusion: these effects are less relevant for the prediction of B/C than the effects discussed previously (which are instead usually neglected)! The message is: although the efforts invested by the community in refining CR propagation modelling could have and have had important implications for other observables, for the mere purpose of fitting B/C to infer diffusion propagation parameters they are to a large extent unnecessary complications, until one can significantly reduce the biases previously discussed.
5.1. Geometric effects
The crude modelling of the diffusive halo as an infinite slab may appear too simplistic. In this section, we estimate the effects of a 2D cylindrical diffusion box, modelled as in Fig. 10. Furthermore, we assess the effect of adding a radial dependence in the injection term, as opposed to the uniform hypothesis. These can be seen as upper limits to reasonable systematics due to simplified description of the spatial dependence of the diffusion medium or source term: given our limited knowledge on this subject, even the most detailed modelling of the propagation medium and source term, in fact, may not be fully realistic.
Fig. 10 Cylindric model: the matter is homogeneously distributed inside a thin disk of thickness 2h and radius R_{gal} = 20 kpc. The solar system is at R_{⊙} ≃ 8 kpc from the Galactic centre. 

Open with DEXTER 
The formalism in such a situation is well known and we do not repeat it here (it has been summarised for instance in Putze et al. 2010). It suffices to say that to take advantage of the cylindrical symmetry, Eq. (2) can be projected on the basis of the zero order Bessel functions ensuring that the density vanishes on the edge of the cylinder of radius R_{gal} = 20 kpc. The flux of an isotope is then the sum over all its harmonic components (13)The results, reported in Table 3, allow us to draw a few conclusions:

the presence of a new escape surface atR_{gal} ≃ 20 kpc is basically irrelevant: the bestfit δ and its error remain the same, with a statistically insignificant, 2% modification of the bestfit value of D_{0};

perhaps more surprisingly, even the replacement of a uniform source distribution with a commonly assumed donut distribution of the form (Yusifov & Küçük 2004) (14)has minor effects, a mere 1% modification in the bestfit determination of δ, and a ~13% lowering of the bestfit value of D_{0}, still statistically insignificant (roughly a 1σ effect);

since the goodness of fit is similar, the B/C observable is essentially insensitive to these improvements. Unless they are justified by the goal of matching or predicting other observables, the complication brought by the 2D modelling of the problem are unnecessary in achieving a good description of the data.
Results on the propagation parameters fitted on the B/C for different geometries.
5.2. Convective wind
Although the highenergy CR propagation is mostly diffusive, the advection outside the Galactic plane (for instance due to stellar winds) has a nonnegligible effect, which we now quantify. We adopted the simplest model of constant velocity wind, directed outside the galactic plane, with magnitude u. Taking this effect into account this effect, the 1D, stationary propagation equation can be written as (15)The two new terms (second and third one on the LHS) account for the advection of the cosmicray density and the adiabatic losses, respectively. A characteristic time of these two processes can be estimated inside the thin disk of matter : (16)and (17)This means that adiabatic losses can be safely neglected compared to the typical diffusion time of (18)It is clear that our previous results provide a suitable firstorder approximation at least at high energy, with the leading correction at energies near 10 GeV/n given especially by the advection. The adiabatic energy loss, instead, is several times smaller and can be safely ignored in the following.
Fig. 11 Variations of the bestfit propagation parameters with respect to the velocity of the convective wind. 

Open with DEXTER 
Summary of the main systematics found in current analyses in determining the propagation parameters by fitting the B/C ratio.
The solution of Eq. (15) neglecting adiabatic losses has the same form of Eq. (3) for the flux of stable species, modulo the change (19)so that the behaviour of the solution smoothly interpolates between the convective timescale at low energy and the diffusive one at high energy: this can be simply checked by neglecting the exponential with respect to unity for a high value of its argument, or Taylorexpanding it to first order in the opposite limit. This formula also suggests that, if one fits the data by neglecting the convective wind, one biases its result towards a lower value of δ, and a corresponding higher value of D, so to reproduce a flatter dependence with energy at lowenergy as for the case described by Eq. (19), as illustrated in Fig. 11. Quantitatively, a variation of 15 km s^{1} in u is roughly similar to a 1σ shift in the benchmark parameters. Note, however, that the goodness of the fit worsens, or in other words, highenergy data are better described by a pure diffusive behaviour than by a convectivediffusive one. Overall, we conclude that these effects appear still somewhat less important in determining the diffusion parameters from highenergy data than the role of primary boron or even crosssection uncertainties. While convection, adiabatic losses, reacceleation, etc. are important to account for when extending the analysis down to very low energies (subGeV/nuc) or in global analyses, they do not currently constitute the main limitations to the determination of D_{0} or δ from highenergy data.
6. Conclusion
The highprecision measurements of cosmicray fluxes that have become available in recent years prompt the question of the theoretical uncertainties inherent to the models used to interpret them. In this article, we have compared the effect of different theoretical biases with statistical uncertainties in the determination of diffusion parameters from the borontocarbon flux ratio, or B/C. This is representative of a much broader class of observables, involving ratios of secondary to primary species, which have been recognised as key tools for diagnostics in cosmicray astrophysics. We adopted a pedagogical approach, showing and interpreting the results whenever possible within simple analytical models. We also used preliminary AMS02 data and limited the analysis to energies above 10 GeV/nuc, which gives a pessimistic – hence conservative – estimate of the statistical uncertainties that will eventually be available.
Our main results, summarised in Table 4, are the following:

The single most important effect that we quantified (to the best ofour knowledge, for the first time) is the degeneracy betweendiffusion parameters and a small injection of primary boron at thesource, finding at present even a statistically insignificantpreference for a small but finite value for a primary boron flux.This degeneracy cannot be removed by highprecisionmeasurements of B/C, but probably requires multimessengertests and certainly demands further investigations, in particular ifdata should manifest a significant preference for a highenergyflattening of secondarytoprimary ratios.

The second most important theoretical uncertainty is associated to crosssections. In particular, anticorrelated modifications in the destruction and production crosssections with respect to reference values may also have an effect on the determination of the diffusion index δ, another effect discussed here for the first time. This should be kept in mind when comparing the outcome of data analyses relying on different databases for crosssections. The good news is that this problem is not due to intrinsic limitations in the astrophysical modelling or the lack of astrophysical data, but to the scarce laboratory measurements available. For the case of boron, experiments of production crosssections via spallation of oxygen, carbon and, to a minor extent, nitrogen, are essentially what would be needed to set the predictions on much firmer grounds.

Other effects we tested for are typically less important and are similar to or smaller than statistical uncertainties: effects such as those of convective winds, certainly important in more complete analyses including lowenergy data, appear unlikely to bring uncertainties large enough to compete with the abovementioned uncertainties. We also showed how the geometry of the diffusive box and the distribution of sources is virtually irrelevant, at least if only a B/C data analysis is concerned. More or less realistic radial distribution of sources, while it may marginally affect the determination of D_{0}, is still indistinguishable from the goodnessoffit point of view. Another outcome of this exercise is that at least at the 10% level, D_{0} is degenerate with a choice of geometry and source distribution, in addition to the already wellknown degeneracy with the diffusive halo height H.
In conclusion, we found that the main uncertainties in inferring diffusion parameters from B/C (and we expect from other secondarytoprimary ratios, too) depend on theoretical priors on sources (linked to sites and mechanisms of acceleration!) and, to a lesser extent, to nuclear crosssections. While exploring more complicated schemes and geometries for the diffusion may thus be important, we can anticipate that sensitivity to such effects will probably require fixing more mundane questions first! A multimessenger strategy, coupled to a new measurement campaign of nuclear crosssections, appears to be a next crucial step in that direction.
Updated version of these two codes can be found at: https://sourceforge.net/projects/galprop and http://www.dragonproject.org/Home.html, respectively
Acknowledgments
We would like to thank David Maurin for sharing useful data, notably Webber 2003 production crosssections and expertise on crosssections. Part of this work was supported by the French Institut universitaire de France, by the French Agence Nationale de la Recherche under contract 12BS050006 DMAstroLHC, and by the Investissements d’avenir, Labex ENIGMASS.
References
 Adriani, O., Barbarino, G. C., Bazilevskaya, G. A., et al. 2014, Phys. Rep., 544, 323 [NASA ADS] [CrossRef] [Google Scholar]
 Barashenkov, V. S., & Polanski, A. 1994, Electronic guide for nuclear crosssections: version 1994, Tech. Rep. E294417, JINRE294417, Joint Inst. Nucl. Res., Dubna [Google Scholar]
 Binns, W. R., Wiedenbeck, M. E., Arnould, M., et al. 2008, New A Rev., 52, 427 [NASA ADS] [CrossRef] [Google Scholar]
 Blasi, P. 2009, Phys. Rev. Lett., 103, 051104 [NASA ADS] [CrossRef] [PubMed] [Google Scholar]
 Blasi, P., & Serpico, P. D. 2009, Phys. Rev. Lett., 103, 081103 [NASA ADS] [CrossRef] [PubMed] [Google Scholar]
 Evoli, C., Gaggero, D., Grasso, D., & Maccione, L. 2008, in AIP Conf. Ser. 1085, eds. F. A. Aharonian, W. Hofmann, & F. Rieger, 380 [Google Scholar]
 Letaw, J. R., Silberberg, R., & Tsao, C. H. 1983, ApJS, 51, 271 [NASA ADS] [CrossRef] [Google Scholar]
 Lodders, K. 2003, ApJ, 591, 1220 [NASA ADS] [CrossRef] [Google Scholar]
 Mashnik, S. G., Sierk, A. J., Van Riper, K. A., & Wilson, W. B. 1998 [arXiv:nuclth/9812071] [Google Scholar]
 Maurin, D., Donato, F., Taillet, R., & Salati, P. 2001, ApJ, 555, 585 [NASA ADS] [CrossRef] [Google Scholar]
 Maurin, D., Taillet, R., Donato, F., et al. 2002, [arXiv:astroph/0212111] [Google Scholar]
 Maurin, D., Putze, A., & Derome, L. 2010, A&A, 516, A67 [NASA ADS] [CrossRef] [EDP Sciences] [Google Scholar]
 Maurin, D., Melot, F., & Taillet, R. 2014, A&A, 569, A32 [NASA ADS] [CrossRef] [EDP Sciences] [Google Scholar]
 Mertsch, P., & Sarkar, S. 2009, Phys. Rev. Lett., 103, 081104 [NASA ADS] [CrossRef] [PubMed] [Google Scholar]
 Mertsch, P., & Sarkar, S. 2014, Phys. Rev. D, 90, 061301 [NASA ADS] [CrossRef] [Google Scholar]
 Moskalenko, I. V., & Mashnik, S. G. 2003, International Cosmic Ray Conference, 4, 1969 [NASA ADS] [Google Scholar]
 Moskalenko, I. V., Mashnik, S. G., & Strong, A. W. 2001, International Cosmic Ray Conference, 5, 1836 [NASA ADS] [Google Scholar]
 Oliva, A. et al., on Behalf of the AMS Collaboration 2013, Precision Measurement of the Cosmic Ray BorontoCarbon Ratio with AMS, AMS collaboration, 33rd Intern. Cosmic Ray Conf. [Google Scholar]
 Ptuskin, V. S., Voelk, H. J., Zirakashvili, V. N., & Breitschwerdt, D. 1997, A&A, 321, 434 [NASA ADS] [Google Scholar]
 Putze, A., Derome, L., & Maurin, D. 2010, A&A, 516, A66 [NASA ADS] [CrossRef] [EDP Sciences] [Google Scholar]
 Silberberg, R., Tsao, C. H., & Barghouty, A. F. 1998, ApJ, 501, 911 [NASA ADS] [CrossRef] [Google Scholar]
 Strong, A. W., & Moskalenko, I. V. 2001, Adv. Space Res., 27, 717 [NASA ADS] [CrossRef] [Google Scholar]
 Strong, A. W., Moskalenko, I. V., & Ptuskin, V. S. 2007, Ann. Rev. Nucl. Part. Sci., 57, 285 [NASA ADS] [CrossRef] [Google Scholar]
 Tripathi, R. K., Cucinotta, F. A., & Wilson, J. W. 1997, Universal Parameterization of Absorption Cross Sections, Tech. rep., NASA Technical Paper 3621 [Google Scholar]
 Tripathi, R. K., Cucinotta, F. A., & Wilson, J. W. 1999, Nuclear Instruments and Methods in Physics Research B, 155, 349 [NASA ADS] [CrossRef] [Google Scholar]
 Webber, W. R., Kish, J. C., & Schrier, D. A. 1990, Phys. Rev. C, 41, 566 [NASA ADS] [CrossRef] [Google Scholar]
 Webber, W. R., Soutoul, A., Kish, J. C., et al. 1998, Phys. Rev. C, 58, 3539 [NASA ADS] [CrossRef] [Google Scholar]
 Webber, W. R., Soutoul, A., Kish, J. C., & Rockstroh, J. M. 2003, ApJs, 144, 153 [NASA ADS] [CrossRef] [Google Scholar]
 Wellisch, H. P., & Axen, D. 1996, Phys. Rev. C, 54, 1329 [NASA ADS] [CrossRef] [Google Scholar]
 Yusifov, I., & Küçük, I. 2004, A&A, 422, 545 [NASA ADS] [CrossRef] [EDP Sciences] [Google Scholar]
All Tables
Benchmark bestfit parameters of the 1D/slab model, with respect to which comparisons are subsequently made.
Comparison between different crosssection parameterisations for the main production channels of secondary boron.
Results on the propagation parameters fitted on the B/C for different geometries.
Summary of the main systematics found in current analyses in determining the propagation parameters by fitting the B/C ratio.
All Figures
Fig. 1 Sketch of the 1D slab model of the Galaxy, with matter homogeneously distributed inside an infinite plane of thickness 2h sandwiched between two thick diffusive layers of thickness 2H. 

Open with DEXTER  
In the text 
Fig. 2 Relative variations of the bestfit propagation parameters (compared to the benchmark model of Table 1) with respect to the injection spectral index α. 

Open with DEXTER  
In the text 
Fig. 3 Preliminary AMS02 measurements of the B/C ratio (Oliva et al. 2013) are plotted as a function of kinetic energy per nucleon. The theoretical prediction of the 1D/slab reference model of Table 1 is also featured for comparison. 

Open with DEXTER  
In the text 
Fig. 4 Relative variations of the bestfit propagation parameters, as compared to the benchmark model of Table 1, with respect to the lowenergy cutoff E_{cut} above which we carry out the B/C analysis. 

Open with DEXTER  
In the text 
Fig. 5 Left panel: variations of the bestfit propagation parameters D_{0} (blue) and δ (red) relative to the benchmark values of Table 1, as a function of the primary borontocarbon injection ratio. The reference model corresponds to the conventional no boron hypothesis for which N_{B}/N_{C} vanishes. Right panel: the theoretical value of the B/C ratio at 214 GeV/nuc (solid red curve) is plotted as a function of the primary borontocarbon injection ratio. The dashed black curve indicates the goodness of the B/C fit. As long as N_{B}/N_{C} does not exceed 13%, the theoretical B/C ratio is within 2σ from the AMS02 measurement (dasheddotted green curve). 

Open with DEXTER  
In the text 
Fig. 6 Contribution of the various primary nuclear species to the secondary boron flux at 10 GeV/nuc, as estimated with the semiempirical code by Webber 03. 

Open with DEXTER  
In the text 
Fig. 7 Relative differences between our reference model (Tripathi et al. 1997) for the destruction crosssections σ_{a} and the other parameterisations by Letaw et al. (1983), Wellisch & Axen (1996) and Barashenkov & Polanski (1994) are displayed as a function of the nucleus charge, at an energy of 10 GeV/nuc. Each bin is characterised by a given charge Z and encodes the arithmetic mean over the corresponding isotopes. Only the elements involved in the cascade from iron to beryllium are displayed. 

Open with DEXTER  
In the text 
Fig. 8 2D histograms of the relative differences between various semiempirical models currently used to calculate the production crosssections σ_{b → a}. Our reference model is Webber et al. (2003) (Webber 03), and we compare it to the parameterisations from Webber et al. (1990) (Webber 93), Webber et al. (1998) (Webber 98) and Silberberg et al. (1998) (S&T 00). The charges of the parent and child nuclei are given on the vertical and horizontal axes, respectively. The relative difference in each bin is given by the arithmetic mean over the various isotopes of each element. A detailed view provides the most important channels for the B/C ratio studies. For a fragmentation of ΔZ< 4, we also give the first and second moments of the uncertainty distributions. 

Open with DEXTER  
In the text 
Fig. 9 Effect of rescaling nuclear crosssections for boron production channels and destruction ones: the left panel assumes correlated, the right panel anticorrelated rescaling. 

Open with DEXTER  
In the text 
Fig. 10 Cylindric model: the matter is homogeneously distributed inside a thin disk of thickness 2h and radius R_{gal} = 20 kpc. The solar system is at R_{⊙} ≃ 8 kpc from the Galactic centre. 

Open with DEXTER  
In the text 
Fig. 11 Variations of the bestfit propagation parameters with respect to the velocity of the convective wind. 

Open with DEXTER  
In the text 
Current usage metrics show cumulative count of Article Views (fulltext article views including HTML views, PDF and ePub downloads, according to the available data) and Abstracts Views on Vision4Press platform.
Data correspond to usage on the plateform after 2015. The current usage metrics is available 4896 hours after online publication and is updated daily on week days.
Initial download of the metrics may take a while.