46 M. Beneke et al. / Nuclear Physics B 891 (2015) 42–72
the Γ =0 result has been constructed for the ultrasoft contribution by extrapolation. Evaluating
the third-order potential corrections given in [25] for real energy E requires a substantial amount
of extra work, which we briefly discuss in the following.
The higher-order potential corrections to the Green function are expressed in terms of nested
harmonic sums, sums over gamma and polygamma functions, and generalised hypergeometric
functions [23,25]. The complex variable λ = (α
s
C
F
/2)
√
−m/(E +iΓ ) appears for example
in the argument of (poly-)gamma functions or as one of the parameters of the hypergeometric
functions. In our application, we have to evaluate the Green function for positive values of the
energy E, starting at E = 0. For vanishing width Γ , λ tends to +i∞ as E tends to zero. Thus,
we ha
ve to ensure that all expressions are well-defined in this limit and that their numerical
evaluation is possible.
In most cases it is possible to express the correction to the Green function in terms of harmonic
sums. This is in particular the case for most of the generalised hypergeometric functions, which
can be treated as described in Appendix A.1 of [34]. The harmonic sums can then be analytically
continued with the methods of [35,36]. Ho
wever, in some cases the correction to the Green
function is expressed in terms of single or even double sums which could not be expressed as
nested harmonic sums. For such sums it was often necessary to truncate the summation at some
(λ-dependent) value and construct suitable asymptotic e
xpansions to approximate the remainder.
In all cases we have checked that the numerical precision is sufficient for our extraction of the
bottom-quark mass, such that the numerical uncertainty can be neglected.
A relati
vely simple example of this procedure is given by the sum
∞
k=1
[(k −λ)(ψ(k −λ) −ψ(k)) + kλψ
(1)
(k)]
2
k
, (3.8)
which appears in the insertion of the Darwin term. Here ψ is the logarithmic derivative of the
gamma function and ψ
(1)
the first derivative of ψ. The sum converges only slowly when λ is
large, which makes the numerical evaluation difficult. Therefore, we introduce a cut-off Λ for
the summation and explicitly sum all terms up to this cut-off. Choosing Λ to be much larger than
|λ|, we can approximate the remainder by e
xpanding the summand in the limit k →∞. Note that
for ψ(k − λ) this is not simply an expansion in |λ|/k, but rather a double expansion for k 1
and k |λ|. In the first step the entire argument of the ψ function is considered large and the
terms of the resulting asymptotic series are further e
xpanded for |λ|/k → 0in the second step,
yielding
ψ(k −λ) = ln(k −λ) −
1
2(k −λ)
−
1
12(k −λ)
2
+
1
120(k −λ)
4
+O
1
(k −λ)
6
=ln k − (1 +2λ)
1
2k
−
1 + 6λ +6λ
2
1
12k
2
−
λ + 3λ
2
+2λ
3
1
6k
3
+
1 − 30λ
2
−60λ
3
−30λ
4
1
120k
4
+O
1
k
5
. (3.9)
After expanding the summand in (3.8), the sum over k from Λ +1to infinity can be evaluated
in terms of Hurwitz zeta functions (in more complicated cases we also encounter derivatives of
this function). The first three terms are
λ
4
4
ζ(3,Λ+ 1) +
λ
5
6
ζ(4,Λ+ 1) +
λ
4
36
−3 + 4λ
2
ζ(5,Λ+ 1), (3.10)
评论0
最新资源