We observe that for scales larger than 2 GeV, δE
US
10
is small and well inside uncer-
tainties. It is also comforting that the yellow error band in figure 1 (associated to the
uncertainty of m
b
) covers the difference between the NNLO and N
3
LO results in the RS
and in the RS’ scheme for the whole scale variation. This difference is often used as an
alternative determination of the error. Actually, in the RS’ scheme the distance between
the different orders is quite small, which may ask for future more aggressive determina-
tions of the bottom mass. We refrain from doing so until a more clear understanding of
the renormalization scale dependence of the observable at low scales and of the role played
by the ultrasoft scale is achieved.
As a final check, we have performed the fit to the Υ(1S) mass using the same setup.
We find a shift of almost +20 MeV, well within the uncertainties of eq. (2.6).
We now compare our result with previous determinations of m
b
(m
b
) from spectroscopy
or from sum rules. We show such comparison in figure 2. We find that all determinations are
perfectly consistent with each other. Determinations from sum rules with small number
of momentums (low n sum rules) seem to give slightly smaller values of the mass than
large n or spectroscopy determinations. We notice that the latter include the Coulomb
resummation. Therefore, the physics they are sensitive to is different. Nevertheless, at
this stage this difference is not statistically significant, and we are not in the position to
push this discussion further. It is also worth noting that, at present, partial NNLL sum
rules determinations produce slightly larger errors than other determinations. It would be
interesting to see if the complete result, or its combination with the large n N
3
LO sum
result, may lead to a more precise determination.
We now turn our discussion to the comparison between different spectroscopy determi-
nations. Our value is slightlty smaller than the determinations in [26–28] (but well within
each other’s uncertanties). Compared with the determinations of Ayala et al. [27] and Ma-
teu et al. [28], a part of the difference can be traced back to the fact that in the two latter
references the Υ plays a predominant role in the fit (which leads to a larger value for the
bottom mass than fits to the η
b
mass). In the first reference because the fit was only made
to the Υ, in the second reference because, even if both the vector and the pseudoscalar
were considered in the fit, their method weights the most the particle that produces the
smaller errror, which in their paper is the vector. It is worth mentioning that if one looks
at the individual fits in [28] to the η
b
mass one finds a smaller value of m
b
(m
b
), quite close
to ours. As at present all these determinations are consistent with each other, there is not
a clear reason to prefer one versus the other. As we have already discussed before, the
resummation of large logarithms may help to solve this problem. We hope to come back to
this issue in the future. The other fact that produces a different central value of m
b
(m
b
)
is the choice of the factorization scale µ to fix the central value. In this paper we have
chosen a larger value of the factorization scale than in [27, 28]. We somewhat feel that a
smaller value of the factorization scale leads to a region where the perturbative series does
not behave well enough. In this respect the factorization scales we choose are closer to the
factorization scales used in [26]. Actually, the difference with respect to the value obtained
in [26] can be traced back to the fact that in that reference an average of the determination
of the bottom mass from the Υ and the η
b
is made. Again pure determinations from the η
b
– 5 –