%$Modified: Sun Nov 10 19:56:43 2002 by uwer $
\documentclass[twoside]{article}
\usepackage{fleqn,npb,pslatex,graphicx,cite}

\def\x#1{{x_{#1}}}
\def\nf{n_f}
\def\eps{\epsilon}
\def\e{\eps}
\def\qb{\bar q}
\def\as{\alpha_s}
\def\nn{\nonumber}
\def\R{{\rm R}}

\def\idotp#1,#2{(#1\cdot#2)}
\def\dotp#1{\expandafter\idotp#1}
\def\slash#1{{/\!\!\!\!\!#1}}
\def\osps(#1,#2,#3){\langle#1|\slash{#2}|#3\rangle}
\def\Polg{{\epsilon_g}}
\def\PolZ{{\epsilon_\gamma}}

\def\kq{{p_1}}
\def\kqb{{p_3}}
\def\kg{{p_2}}
\def\Li{{\rm Li}}
\def\Li(#1,#2){{{\rm Li}_{#1}(#2)}}
\def\Lii(#1,#2,#3,#4){{{\rm Li}_{#1,#2}\left(#3,#4\right)}}
\def\Liii(#1,#2,#3,#4,#5,#6){{{\rm Li}_{#1,#2,#3}\left(#4,#5,#6\right)}}
\def\z#1{{\zeta_{#1}}}
\def\RR#1{{\rm R_{#1}}}
\def\bare{\mbox{\scriptsize \rm bare}}
\def\ren{\mbox{\scriptsize \rm ren}}
\def\fin{\mbox{\scriptsize \rm fin}}

\title{
  QCD two-loop amplitudes for
  $e^+e^- \to$ 3 jets: the fermionic contribution
  }

\author{%
  S. Moch\,\address{Deutsches Elektronensynchrotron DESY
    Platanenallee 6, D--15738 Zeuthen, Germany}%
  \thanks{{\tt moch@ifh.de}}, 
  P. Uwer\,\address{Institut f{\"u}r Theoretische Teilchenphysik,
    Universit{\"a}t Karlsruhe, 
    D-76128 Karlsruhe, Germany}\thanks{%
    {\tt uwer@particle.uni-karlsruhe.de}
    (speaker)},
  and S. Weinzierl\,\address{Dipartimento di Fisica, Universit\`a di Parma,
    INFN Gruppo Collegato di Parma, 43100 Parma, Italy}%
  \thanks{{\tt stefanw@fis.unipr.it}}}
\begin{document}
\begin{abstract}
\end{abstract}
\begin{abstract}
We review a new technique for the calculation of two-loop amplitudes
and discuss as an example the fermionic contributions to $e^+ e^- \to 
q\bar q g$.
\end{abstract}
\thispagestyle{empty}
\maketitle
\section{Introduction}
The construction of fully differential next-to-next-to-leading order
 (NNLO)  programs
is a challenging problem and urgently needed to match the accuracy
reached in today's collider experiments. While for inclusive
observables like for example the total hadronic cross section in $e^+e^-$
annihilation the step from NLO to NNLO is far
less demanding -- and has been taken a long time ago  -- the 
contrary is true for less inclusive quantities as for example jet
rates. Essentially one has to adress two problems in jet physics at
NNLO accuracy. The first one is related to the fact that one needs to treat
two-loop integrales with more than one scale. These integrals are much
harder to solve than one-scale integrals. The second problem which
needs to be solved is the combination of virtual and real corrections 
-- which are separately infrarot (IR) divergent -- into a IR finite
cross section.
Given these difficulties it is clear that only for specific
reactions NNLO calculations can be envisaged. Important reactions for
which one should go beyond the NLO approximation are for example: 
Bhabha scattering, $p p, p\bar p \rightarrow 2\;\mbox{jets}$, and
$ e^+e^- \rightarrow 3\; \mbox{jets}$. Among these examples the
$e^+e^-$ annihilation into 3 jets is of particular
interest. Historically this reaction was one of the first clean tests
of Quantum Chromodynamics (QCD) as the theory of strong
interaction. Today the 3-jet production in $e^+e^-$ annihilation is
still an excellent laboratory for precise tests of QCD and jet
physics. 
In particular $e^+e^-\to 3\; \mbox{jets}$ is also a perfect reaction
for measureing the QCD coupling $\as$ with high accuracy. It is exactly
here where the need for NNLO predictions for 3-jet production becomes
most prominent: at present the measurements of $\as$ are plagued by
scale uncertainties of the theoretical predictions
\cite{Bethke:2002rv}.  These uncertainties
arise from uncalculated higher order contributions. Although formally
of higher order these uncertainties can still be large. They can be 
reduced only by a NNLO calculation. The residual scale dependence of the
3-jet rate $f_3$ can be studied using the following formula
\begin{eqnarray}
   f_3\left(y,{\mu^2\over s}\right)
  &=& 
  {\as(\mu)\over 2\pi}  {A(y)}\nn\\
  &&\hspace{-1.5cm}
  +\left({\as(\mu)\over 2\pi}\right)^2\* 
  \left( {B(y)}
  +{1 \over 2} \*\beta_0\*\ln\left({\mu^2\over s}\right)\*{A(y)}\right)
  \nn\\
  && \hspace{-1.5cm}
  +\left({\as(\mu)\over 2\pi}\right)^3\* 
  \bigg( {C(y)} 
  +\beta_0\*\ln\left({\mu^2\over s}\right)\*{B(y)}\nn\\
  &&\hspace{-1.5cm}
  + {1 \over 4} \*\left[\beta_0^2\*\ln\left({\mu^2\over s}\right)^2
  +\beta_1\*\ln\left({\mu^2\over s}\right)\right]\*{A(y)}\bigg).
\end{eqnarray}
Here $\beta_0,\beta_1$ denote the coefficients of the QCD
$\beta$-function, $y$ is the resolution of the jet algorithm and 
$A$, $B$, and $C$ are the LO, NLO, NNLO coefficients at $\mu^2=s$.
The residual scale dependence is illustrated in fig. \ref{fig:scale}
for the Durham algorithm at $y=0.1$.
\begin{figure}[htbp]
  \begin{center}
    \includegraphics[width=7.cm]{scale.eps}
    \begin{picture}(0,0)
      \put(-110,-10.){$\mu/\sqrt{s}$}
      \put(-200,130.){$f_3$}
    \end{picture}
    \caption{Residual scale dependence of $f_3(y,{\mu^2\over s})$. 
      \label{fig:scale}}
  \end{center}
\end{figure}
For the NNLO curves three different scenarios have been assumed: one
scenario in which the NNLO corrections vanish for $\mu^2=s$, and two 
scenarios where a $\pm$ 5 \% correction at $\mu^2=s$ have been assumed.
It is clearly visible that the NNLO corrections improves the residual
scale dependence.
For a 3-jet prediction at NNLO accuracy several ingredients are needed:
First the matrix elements for $e^+e^- \to q \qb ggg,\; q \qb
q'\qb'\!g$ have to be evaluated. 
They are obtained from a leading-order calculation and are
known for a long time \cite{Berends:1989yn,Hagiwara:1989pp}. 
Second the one-loop matrix elements for 
$e^+e^- \to q \qb gg, q \qb q'\qb'$ are needed. They have been
calculated in 
refs. \cite{Bern:1997ka,Bern:1998sc,Glover:1997eh,Campbell:1997tv}. 
Finally one needs the NNLO amplitudes for 
$e^+e^- \to q \qb g$ \cite{Garland:2001tf,Garland:2002ak,Moch:2002an,Moch:2002wt,Moch:2002hm}. 
In the following the calculation of fermionic contribution of  the NNLO matrix
elements is reported. One should keep in mind that the combination of
these 3 ingredients into an infrared finite cross section is still a 
highly non-trivial -- and so far unsolved -- problem.


\section{A new approach to calculate NNLO amplitudes}
Due to the tremendous activities in the field enormous progress has
been made in the past. All the relevant double box scalar
integrals have been calculated 
\cite{Smirnov:1999gc,Tausk:1999vh,Smirnov:2000vy,Smirnov:2000ie}. 
For a few reactions these
results have been used to obtain NNLO scattering amplitudes 
\cite{Bern:2000dn,Anastasiou:2000kg,Anastasiou:2000ue,Anastasiou:2000mv,Anastasiou:2001sv,Glover:2001af,Garland:2001tf,Bern:2000ie,Bern:2001df,Bern:2002tk,Garland:2002ak,Anastasiou:2002zn}.
Today one can say that a {\it standard approach} to perform these
calculations exist. Essentially it consists of three steps:
\begin{enumerate}
\item Generation of relevant Feynman diagrams.
\item Reduction of tensor integrals to scalar integrals via
  Schwinger parametrization \cite{Tarasov:1996br,Tarasov:1997kx}.
\item Reduction of scalar integrals with raised powers and in higher
  dimensions to master integrals via partial integration
  and Lorentz invariance equations 
  \cite{'tHooft:1972fi,Chetyrkin:1981qh,Gehrmann:1999as}.
\end{enumerate}
The final result is than obtained in
terms of master integrals. They have to be calculated
analytically. The bottleneck of this approach is that one has to
create and to solve a huge system of equations to obtain the desired reduction 
to master integrals. In contrast to the naive expectation, topologies
which at first glance look not so difficult may involve much more 
work than the more complicated once. 
For example it is almost trivial to find the reduction scheme for the
planar double box although the master integral itself is one of the
most complicated once.
In ref. \cite{Moch:2001zr} we have proposed a different method. 
The basic idea is that
one applies only obvious reductions like for example 
the {\it triangle rule}. In particular one does not perform a complete
reduction to master integrals. Instead one calculates the scalar
integrals in higher dimensions and with raised powers of the
propagators directly. 
\begin{figure}[htbp]
  \begin{center}
    \parbox{3.7cm}{
      \begin{center}
        \includegraphics[scale=0.5]{figs.001}
      \end{center}
      }
    \parbox{3.7cm}{
      \begin{center}
        \includegraphics[scale=0.7]{figs.002}
      \end{center}
      }\vspace*{-0.4cm}
    \parbox[t]{3.7cm}{\centerline{a)}}\parbox[t]{3.7cm}{\centerline{b)}}
    \centerline{\includegraphics[width=7.3cm]{figs.003}}
    \centerline{c)}
    \caption{Tensor reduction of the general pentabox topology a) to
      scalar integrals in higher dimensions and raised powers of the
      propagators b). Application of the triangle rule yielding the C-topology
      $\mbox{C}(m={d\over 2}+\e,\nu_1,\nu_2,\nu_3,\nu_4,\nu_5)$ c).
      \label{fig:pentabox-a}}
  \end{center}
\end{figure}
To illustrate the method let us consider the so-called penta-box. 
A corresponding Feynman diagram is shown in fig. \ref{fig:pentabox-a}a. 
Applying Schwinger parameterization one arrives at the scalar topology
shown schematically in  fig. \ref{fig:pentabox-a}b. An obvious
application of the triangle rule allows the elimination of two
propagators \cite{Anastasiou:1999bn}
(cf. fig. \ref{fig:pentabox-a}c). 
The resulting topology is often called C-topology.
As mentioned earlier
the C-topology appears in higher dimensions $(d=2m-2\e,\; m\in N)$ and with 
raised powers of the propagators $(\nu_1,\ldots,\nu_5)$.
In the approach advocated here an analytic expression for the
C-topology is needed. Such an expression can be obtained in terms
of infinite sums \cite{Moch:2001zr}. The generic structure is shown in
eq.~(\ref{result_Ctopo}).
\begin{eqnarray}
  \label{result_Ctopo}
  &&\mbox{C}(m={d\over 2}+\e,\nu_1,\ldots,\nu_5)  \sim 
 \sum\limits_{i_1=0}^\infty
 \sum\limits_{i_2=0}^\infty
 \frac{x_1^{i_1}}{i_1!}
 \frac{x_2^{i_2}}{i_2!}\bigg[
 \nonumber \\&&\hspace{-0.7cm}
  x_1^{2m-2\eps-\nu_{1235}} x_2^{2m-2\eps-\nu_{2345}}
{\Gamma(i_1+2m-2\eps-\nu_{125})\over \Gamma(i_1+1+2m-2\eps-\nu_{1235})}
 \nonumber \\&&\hspace{-0.7cm}
 \times
 \frac{
   \Gamma(i_1+i_2+2m-2\eps-\nu_{235})
   \Gamma(i_2+2m-2\eps-\nu_{345}) 
   }{\Gamma(i_1+i_2+4m-4\eps-\nu_{12345}-\nu_5)}
 \nonumber \\
 & & \hspace{-0.7cm}
   \times\frac{
     \Gamma(i_1+i_2+3m-3\eps-\nu_{12345})}
   { \Gamma(i_2+1+2m-2\eps-\nu_{2345})} 
+ \ldots\bigg] %% \left( -s_{123} \right)^{2m-2\eps-\nu_{12345}},
\end{eqnarray}
It is obvious that such a representation is completely useless unless
one is able to evaluate the $\eps$-expansion in terms of known
functions. Fortunately this is possible using a few basic algorithms
for {\it nested sums} \cite{Moch:2001zr}. 
The basic idea here is that one rewrites the
sums using the relation
\begin{equation} \sum\limits_{i=0}^\infty \sum\limits_{j=0}^\infty 
  a_{ij}
  =   
  a_{00}
  + \sum\limits_{j=1}^\infty a_{0j}
  + \sum\limits_{i=1}^\infty a_{i0}
  + \sum\limits_{n=1}^\infty \sum\limits_{j=1}^{n-1} a_{j(n-j)}.
  \label{eq:nestedsums1}
\end{equation}
Furthermore one can expand the $\Gamma$-functions by the means of
\begin{eqnarray}
  {\Gamma(-n+\e)} &=& 
  {(-1)^n\over \e n!}{\Gamma(1+\e)}
  ( 1 + S_1(n)\e \nn\\
  &&
  \hspace{-2cm}+ S_{1,1}(n)\e^2 
  + S_{1,1,1}(n)\e^3 +
  S_{1,1,1,1}(n)\e^4+\ldots
  \label{eq:expandG}
\end{eqnarray}
and related identities \cite{Vermaseren:1998uu}.
Here $S_{1,\ldots}(n)$ denote the harmonic sums, 
see e.g. ref. \cite{Vermaseren:1998uu}. Inspecting
the sums which appear using eq.~(\ref{eq:nestedsums1}) and 
eq.~(\ref{eq:expandG}) one observes 
that the algorithms needed are just a generalization of the algorithms
studied in ref. \cite{Vermaseren:1998uu}. 
In particular defining the nested sums 
as a generalization of the harmonic sums by 
\begin{eqnarray}
  &&\hspace{-0.7cm}S(n;m_1,...,m_k;x_1,...,x_k)  
  = \sum\limits_{i_1=1}^n\sum\limits_{i_2=1}^{i_1}
 \cdots \sum\limits_{i_k=1}^{i_{k-1}} {{x_1^{i_1}\over i_1^{m_1}}
   \cdots {x_k^{i_k}\over i_k^{m_k}}}\nn\\
  &&= \sum\limits_{i=1}^n {{\frac{x_1^i}{i^{m_1}}}}
 S({{i}};m_2,...,m_k;x_2,...,x_k)  
\end{eqnarray}
the following four basic operations are sufficient to evaluate 
the $\eps$-expansion of the C-topology in terms of multiple
polylogarithms \cite{Goncharov}:
\begin{enumerate}
\item Multiplication
  \begin{equation}
    {{S(n;m_1,...;x_1,...)\,\times\, S(n;m_1',...;x_1',...)}}
  \end{equation}
\item Sums involving {{$i$}} and {{$n-i$}}:
  \begin{eqnarray}
    &&\sum\limits_{i=1}^{n-1}
    \frac{x_1^i}{i^{m_1}} S({{i}};m_2...;x_2,...)\nn\\
    &&\times
    \frac{{x_1'}^{n-i}}{(n-i)^{m_1'}} S({{n-i}};m_2',...;x_2',...)
  \end{eqnarray}
\item Conjugations:
  \begin{equation}
    \hspace{-0.6cm}\sum\limits_{i=1}^n {{\left( \begin{array}{c} n \\ i 
            \\ \end{array} \right) (-1)^i}} 
    \frac{x_0^i}{i^{m_0}} S(i;m_1,...,m_k;x_1,...,x_k)
  \end{equation}
\item Sums involving {{binomials}} and {{$n-i$}}:
  \begin{eqnarray}
    &&\sum\limits_{i=1}^{n-1}
    {{\left( \begin{array}{c} n \\ i \\ \end{array} \right)}}
    \left(-1\right)^i
    \frac{x_1^i}{i^{m_1}} S(i;m_2...;x_2,...)
    \nn\\&&
    \times \frac{{x_1'}^{n-i}}{(n-i)^{m_1'}} S({{n-i}};m_2',...;x_2',...)
  \end{eqnarray}
\end{enumerate}
In ref. \cite{Moch:2001zr} we have given explicit algorithms for 
the four basic 
operations. For their implementation we  used 
Form3 \cite{Vermaseren:2000nd} and Ginac 
\cite{Bauer:2000cp,Weinzierl:2002hv}.
Using these programs we calculated more than 300
C-topologies analytically. In particular we have checked that our
results for the two master integrals 
${\rm C}(2,1,1,1,1,1), {\rm C}(2,1,1,1,1,2)$  
agree with those given in ref. \cite{Gehrmann:2000zt}. 
All the simpler topologies can be calculated
in the same way.
\section{Results}
The amplitude ${\cal A}_\gamma$ for the process 
 $e^+e^-\to \gamma^\ast\to  q(\kq) g(\kg)\qb(\kqb)$
can be written as a leptonic part $L_\mu$ multiplied by a
hadronic part $H_\mu$:
\begin{eqnarray}
 {\cal A}_\gamma &\sim& {1\over s} \, H_\mu\, L^\mu \nn\\
 &&\hspace{-1.1cm}=
 {1\over s} \, g_s (H_\mu^{(0)} 
 + {\as\over 2\pi} H_\mu^{(1)}
 + \left({\as\over 2\pi}\right)^2 H_\mu^{(2)}+\ldots)\, L^\mu,
\end{eqnarray}
with $s=(\kq+\kg+\kqb)^2$ and $g_s =\sqrt{4 \pi \alpha_s} $.
Furthermore  the hadronic part can be decomposed according to the
colour structure. In particular we have
\begin{eqnarray}
  H_\mu^{(2)} &=& T^a_{q\bar q} \bigg( 
   N^2 H_{2,\mu}^{(2)} + H_{0,\mu}^{(2)} + {1\over N^2} H_{-2,\mu}^{(2)}
   \nn\\&+&
   { \nf N H_{\nf,1 ,\mu}^{(2)}
     + {\nf\over N} H_{\nf,-1,\mu}}^{(2)}
   + \nf^2 H_{\nf^2,\mu}^{(2)}\nn\\&+&
   \Sigma_f \left({4\over N}-N\right) H_{\Sigma_f,\mu}^{(2)} \bigg),
\end{eqnarray}
with $T^a_{q\bar q} $ the generator of the $SU(N)$ gauge group and
$\nf$ the number of massless quarks.
In this talk we discuss only the contributions proportional to
$ \nf N$ and ${\nf\over N}$. The hadronic contribution can also be
decomposed according to the spinor structure:
\begin{eqnarray}
  H_\mu
  &=& c_1 \*{1\over  s}\* \osps(\kq,\kg,\kqb) \* \Polg_\mu
  \nn\\&&\hspace{-0.3cm}
  +\ldots+
     c_4 \*{1\over s^2}  \* \osps(\kq,\kg,\kqb) \* \dotp{\Polg,\kq} \*
  \kq_\mu
  +\ldots
   \nn\\&&
  +
  c_{13}\*{1\over s^2}  \* \osps(\kq,\kg,\kqb) \* \dotp{\Polg,\kqb} \* 
  \kg_\mu 
\end{eqnarray}
where the functions $c_1-c_{13}$ depend only on the ratios 
$x_1=2(\kq\cdot \kg)/s$, $x_2=2(\kg\cdot \kqb)/s$. Due to various constraints arising from gauge invariance
or quark anti-quark antisymmetry only 4 of them are needed. All the
remaining ones can be expressed in terms of these 4 functions which 
we chose to be $c_2,\, c_4,\, c_6$ and $c_{12}$.
The perturbative expansion in $\alpha_s$ of the functions 
$c_i$ is defined through
\begin{equation}
  c_i = g_s \left(
    c_i^{(0)} \!
    +\left( \frac{\alpha_s}{2\pi}\right) c_i^{(1)} \!
    +\left( \frac{\alpha_s}{2\pi}\right)^2 c_i^{(2)}\!
    + O(\as^3) 
  \right)\!\!.
\end{equation}
Using
the methods described in the previous section we calculated all
the 13 functions and used the constraints as a check of
our calculation.
Furthermore we checked the structure of the ultraviolet as well as 
the infrared divergencies \cite{Catani:1998bh,Sterman:2002qn}. 
Note that the general structure of the IR divergencies as conjectured
in ref. \cite{Catani:1998bh} has been proven recently by Sterman and
Tejeda-Yeomans \cite{Sterman:2002qn}.
We have also compared our results
with results published recently \cite{Garland:2001tf,Garland:2002ak}. 
In refs. \cite{Garland:2001tf,Garland:2002ak} the result is written
in terms of 2-dimensional harmonic polylogarithms instead of the multiple
polylogarithms used here. Expressing the 2-dimensional 
harmonic polylogarithms in terms of multiple polylogarithms we found
complete agreement for the $ \nf N$ and ${\nf\over N}$ contribution
calculated here.

Following ref. \cite{Garland:2002ak} we used the general 
structure of the infrared
divergencies as described by Catani \cite{Catani:1998bh} to define
the finite parts of the coefficients $c_i$:
\begin{eqnarray}
   c_{i}^{(2),\fin} &=& c_{i}^{(1),\ren} - {\bf I}^{(1)}(\eps) c_{i}^{(1),\ren}
   - {\bf I}^{(2)}(\eps) c_{i}^{(0)} \, , 
  \label{eq:def-results2}
\end{eqnarray}
with the one- and two-loop insertion operators 
${\bf I}^{(1)}(\eps)$ and ${\bf I}^{(2)}(\eps)$ given in ref. 
\cite{Catani:1998bh}.
As an example, we present the result for the $\nf N$-contribution to 
the finite part $c_{4}^{(2),\fin}$ at two loops,
\begin{eqnarray}
  &&\hspace{-0.6cm}c_4^{(2),\fin}(\x1,\x2)\bigg|_{\nf N} =
%%START
%%c_4=
  \nf\*N\*\bigg(
  -{19 \over 18} \*{1\over \x1\*(1-\x2)}
  \nn\\&&
  \hspace{-0.6cm}
  + {1 \over 36} \*{\ln(\x2) \over (1-\x2)}\*
  \bigg[
  36\*{(1+\x1)\over (\x1+\x2)}-{59\over \x1}-{50\over \x1\*(1-\x2)}
  \nn\\&&
  \hspace{-0.6cm}
  +12\*{(1-3\*\x1)\*(1-\x1)\over \x1\*(1-\x1-\x2)}
  \bigg]
  +{(\x1-\x2)\over (\x1+\x2)^3}\*\RR1(\x1,\x2)
    \nn\\&&
    \hspace{-0.6cm}
 -\bigg[{5 \over 12} \*{(2-\x2)\over \x1\*(1-\x2)^2}
  +{2\over (\x1+\x2)^2}
  \bigg]\*\Li(2,1-\x2)
    \nn\\&&
 +{1 \over 4} \*{(2-\x2)\over \x1\*(1-\x2)^2}\*\bigg[
 \ln(\x2)^2+{1 \over 3} \*\z2\bigg]
  \nn\\&&
    \hspace{-0.6cm}
  -{1 \over 12} \*{(7\*\x1-5\*\x2-6\*\x1\*\x2+5\*\x2^2+\x1^2)
    \over (1-\x2)\*\x1\*(1-\x1-\x2)\*(\x1+\x2)}\*\ln(\x1)
  \nn\\&&
  \hspace{-0.7cm}
  -{\R(\x1,\x2)\over (1-\x1-\x2)\*(1-\x2)\*(\x1+\x2)}\*
  \bigg[{17 \over 6 }\*\x2-{35 \over 12} \*\x1
  \nn\\&&
  \hspace{-0.5cm}
  -{19 \over 12} +{1 \over 12} \*{(1+\x1)\over (1-\x2)}
  -{1 \over 3} \*{(1-3\*\x1)\*(1-\x1)\over (1-\x1-\x2)}
  \nn\\&&
  +{5\*\x2^2 \over 12\*\x1} -{5 \over 6} \*{\x2\over \x1}
  +2\*{\x1\*(1+\x1)\over (\x1+\x2)}\bigg]
  \nn\\&&
  \hspace{-0.5cm}
  -\bigg[
 {1 \over 12} \*{(2-\x2)\over \x1\*(1-\x2)^2}
  -{2\over (\x1+\x2)^2}
  \bigg]
  \*\Li(2,1-\x1)
  \bigg)
  \nn\\&&
  %%
  %%
  \hspace{-0.6cm}
  -i\*\pi\*\nf\*N\*\bigg(
  {(2-\x2)\*\ln(\x2)\over 2\*\x1\*(1-\x2)^2} 
  +{1 \over 2} \*{1\over \x1\*(1-\x2)}\bigg) %%:
%%STOP
.\label{eq:c4nn}
\end{eqnarray}
\def\Li{{\rm Li}}%
We have introduced the function $\R(\x1,\x2)$ defined in \cite{Ellis:1981wv}.
In addition, it is convenient, to define the symmetric function
$\RR1(\x1,\x2)$, which contains a particular combination of 
multiple polylogarithms \cite{Goncharov},
\begin{eqnarray*}
  \label{eq:RRfunctions}
  && \RR1(\x1,\x2) = \bigg(
  \ln(\x1) \Li_{1,1}\left({\x1 \over \x1\!+\!\x2},\x1\!+\!\x2 \right)
  \nn\\&&\hspace{-0.7cm}
  -{1 \over 2} \z2 \ln(1\!-\!\x1\!-\!\x2)
  -{1 \over 2}\ln(\x1) \ln(\x2) \ln(1\!-\!\x1\!-\!\x2) 
  \nn \\ & &\hspace{-0.7cm}
  -\ln(\x1) \Li_{2}(\x1\!+\!\x2)
  -\Li_{1,2}\left({\x1 \over \x1\!+\!\x2},\x1\!+\!\x2 \right)
  \nn \\ & &
  \hspace{-0.7cm}-\Li_{2,1}\left({\x1 \over \x1\!+\!\x2},\x1\!+\!\x2 \right)
   +\Li_{3}(\x1\!+\!\x2) \bigg)
   + (\x1\leftrightarrow\x2)\, .
\end{eqnarray*}
All the multiple polylogarithms have simple arguments. 
As a consequence it is straightforward to obtain the analytic
continuation, which can be used for the crossing of the amplitude.

\section{Conclusion}
In this talk we have presented the calculation of a specific colour
structure of the QCD two-loop amplitude $e^+e^-\to q\qb g$. The
computation has been done by a completely new method \cite{Moch:2001zr}. 
The new approach provides a very efficient
method for two-loop calculations. 
The tools developed in this approach can also be
used for a systematic expansion of generalized hypergeometric
functions.
In addition the presented results provide an important cross check
on recently obtained results   \cite{Garland:2001tf,Garland:2002ak}.\\

%%\parindent0cm
%%{\bf Acknowledgment:} P.U. would like to thank the organizers for
%%the pleasant atmosphere at QCD02.
%%\bibliographystyle{h-elsevier2}
%%\bibliography{literatur}
\begin{thebibliography}{10}

\bibitem{Bethke:2002rv}
S. Bethke,
\newblock (2002), 
\newblock %%CITATION = ;%%.

\bibitem{Berends:1989yn}
F.A. Berends, W.T. Giele and H. Kuijf,
\newblock Nucl. Phys. B321 (1989) 39,
\newblock %%CITATION = NUPHA,B321,39;%%.

\bibitem{Hagiwara:1989pp}
K. Hagiwara and D. Zeppenfeld,
\newblock Nucl. Phys. B313 (1989) 560,
\newblock %%CITATION = NUPHA,B313,560;%%.

\bibitem{Bern:1997ka}
Z. Bern et~al.,
\newblock Nucl. Phys. B489 (1997) 3, 
\newblock %%CITATION = ;%%.

\bibitem{Bern:1998sc}
Z. Bern, L.J. Dixon and D.A. Kosower,
\newblock Nucl. Phys. B513 (1998) 3, 
\newblock %%CITATION = ;%%.

\bibitem{Glover:1997eh}
E.W.N. Glover and D.J. Miller,
\newblock Phys. Lett. B396 (1997) 257, 
\newblock %%CITATION = ;%%.

\bibitem{Campbell:1997tv}
J.M. Campbell, E.W.N. Glover and D.J. Miller,
\newblock Phys. Lett. B409 (1997) 503, 
\newblock %%CITATION = ;%%.

\bibitem{Garland:2001tf}
L.W. Garland et~al.,
\newblock Nucl. Phys. B627 (2002) 107, 
\newblock %%CITATION = ;%%.

\bibitem{Garland:2002ak}
L.W. Garland et~al.,
\newblock (2002), 
\newblock %%CITATION = ;%%.

\bibitem{Moch:2002wt}
S. Moch, P. Uwer and S. Weinzierl,
\newblock (2002), 
\newblock %%CITATION = ;%%.

\bibitem{Moch:2002an}
S. Moch, P. Uwer and S. Weinzierl,
\newblock (2002), 
\newblock %%CITATION = ;%%.

\bibitem{Moch:2002hm}
S. Moch, P. Uwer and S. Weinzierl,
\newblock (2002), 
\newblock %%CITATION = ;%%.

\bibitem{Smirnov:1999gc}
V.A. Smirnov,
\newblock Phys. Lett. B460 (1999) 397, 
\newblock %%CITATION = ;%%.

\bibitem{Tausk:1999vh}
J.B. Tausk,
\newblock Phys. Lett. B469 (1999) 225, 
\newblock %%CITATION = ;%%.

\bibitem{Smirnov:2000vy}
V.A. Smirnov,
\newblock Phys. Lett. B491 (2000) 130, 
\newblock %%CITATION = ;%%.

\bibitem{Smirnov:2000ie}
V.A. Smirnov,
\newblock Phys. Lett. B500 (2001) 330, 
\newblock %%CITATION = ;%%.

\bibitem{Bern:2000dn}
Z. Bern, L.J. Dixon and D.A. Kosower,
\newblock JHEP 01 (2000) 027, 
\newblock %%CITATION = ;%%.

\bibitem{Anastasiou:2000kg}
C. Anastasiou et~al.,
\newblock Nucl. Phys. B601 (2001) 318, 
\newblock %%CITATION = ;%%.

\bibitem{Anastasiou:2000ue}
C. Anastasiou et~al.,
\newblock Nucl. Phys. B601 (2001) 341, 
\newblock %%CITATION = ;%%.

\bibitem{Anastasiou:2000mv}
C. Anastasiou et~al.,
\newblock Phys. Lett. B506 (2001) 59, 
\newblock %%CITATION = ;%%.

\bibitem{Anastasiou:2001sv}
C. Anastasiou et~al.,
\newblock Nucl. Phys. B605 (2001) 486, 
\newblock %%CITATION = ;%%.

\bibitem{Glover:2001af}
E.W.N. Glover, C. Oleari and M.E. Tejeda-Yeomans,
\newblock Nucl. Phys. B605 (2001) 467, 
\newblock %%CITATION = ;%%.

\bibitem{Bern:2000ie}
Z. Bern, L.J. Dixon and A. Ghinculov,
\newblock Phys. Rev. D63 (2001) 053007, 
\newblock %%CITATION = ;%%.

\bibitem{Bern:2001df}
Z. Bern, A. De~Freitas and L.J. Dixon,
\newblock JHEP 09 (2001) 037, 
\newblock %%CITATION = ;%%.

\bibitem{Bern:2002tk}
Z. Bern, A. De~Freitas and L. Dixon,
\newblock JHEP 03 (2002) 018, 
\newblock %%CITATION = ;%%.

\bibitem{Anastasiou:2002zn}
C. Anastasiou, E.W.N. Glover and M.E. Tejeda-Yeomans,
\newblock Nucl. Phys. B629 (2002) 255, 
\newblock %%CITATION = ;%%.

\bibitem{Tarasov:1996br}
O.V. Tarasov,
\newblock Phys. Rev. D54 (1996) 6479, 
\newblock %%CITATION = ;%%.

\bibitem{Tarasov:1997kx}
O.V. Tarasov,
\newblock Nucl. Phys. B502 (1997) 455, 
\newblock %%CITATION = ;%%.

\bibitem{'tHooft:1972fi}
G. 't~Hooft and M.J.G. Veltman,
\newblock Nucl. Phys. B44 (1972) 189,
\newblock %%CITATION = NUPHA,B44,189;%%.

\bibitem{Chetyrkin:1981qh}
K.G. Chetyrkin and F.V. Tkachov,
\newblock Nucl. Phys. B192 (1981) 159,
\newblock %%CITATION = NUPHA,B192,159;%%.

\bibitem{Gehrmann:1999as}
T. Gehrmann and E. Remiddi,
\newblock Nucl. Phys. B580 (2000) 485, 
\newblock %%CITATION = ;%%.

\bibitem{Moch:2001zr}
S. Moch, P. Uwer and S. Weinzierl,
\newblock J. Math. Phys. 43 (2002) 3363, 
\newblock %%CITATION = ;%%.

\bibitem{Anastasiou:1999bn}
C. Anastasiou, E.W.N. Glover and C. Oleari,
\newblock Nucl. Phys. B575 (2000) 416, 
\newblock %%CITATION = ;%%.

\bibitem{Vermaseren:1998uu}
J.A.M. Vermaseren,
\newblock Int. J. Mod. Phys. A14 (1999) 2037, 
\newblock %%CITATION = ;%%.

\bibitem{Goncharov}
A.B. Goncharov,
\newblock Math. Res. Lett. 5 (1998) 497.

\bibitem{Vermaseren:2000nd}
J.A.M. Vermaseren,
\newblock (2000), 
\newblock %%CITATION = ;%%.

\bibitem{Bauer:2000cp}
C. Bauer, A. Frink and R. Kreckel,
\newblock J. Symbolic Computation 33 (2002), 
\newblock %%CITATION = ;%%.

\bibitem{Weinzierl:2002hv}
S. Weinzierl,
\newblock Comput. Phys. Commun. 145 (2002) 357, 
\newblock %%CITATION = ;%%.

\bibitem{Gehrmann:2000zt}
T. Gehrmann and E. Remiddi,
\newblock Nucl. Phys. B601 (2001) 248, 
\newblock %%CITATION = ;%%.

\bibitem{Catani:1998bh}
S. Catani,
\newblock Phys. Lett. B427 (1998) 161, 
\newblock %%CITATION = ;%%.

\bibitem{Sterman:2002qn}
G. Sterman, M.E. Tejeda-Yeomans,
\newblock (2002), 
\newblock %%CITATION = ;%%.

\bibitem{Ellis:1981wv}
R.K. Ellis, D.A. Ross and A.E. Terrano,
\newblock Nucl. Phys. B178 (1981) 421.
\newblock %%CITATION = NUPHA,B178,421;%%.

\end{thebibliography}
\end{document}
