Recent results on the running coupling in QCD with two massless flavours
Extracted source text ยท arXiv hep-lat/0209023
%%%%%%%%%% espcrc2.tex %%%%%%%%%%
%
% $Id: espcrc2.tex 1.2 2000/07/24 09:12:51 spepping Exp spepping $
%
\documentclass[fleqn,twoside]{article}
\usepackage{espcrc2}
% change this to the following line for use with LaTeX2.09
% \documentstyle[twoside,fleqn,espcrc2]{article}
% if you want to include PostScript figures
\usepackage{graphicx}
% if you have landscape tables
\usepackage[figuresright]{rotating}
% put your own definitions here:
% \newcommand{\cZ}{\cal{Z}}
% \newtheorem{def}{Definition}[section]
% ...
\newcommand{\ttbs}{\char'134}
\newcommand{\AmS}{{\protect\the\textfont2
A\kern-.1667em\lower.5ex\hbox{M}\kern-.125emS}}
% add words to TeX's hyphenation exception list
\hyphenation{author another created financial paper re-commend-ed Post-Script}
% declarations for front matter
\title{
{
\vspace{-3.35cm} \normalsize \hfill
\parbox{30mm}{DESY 02-124\\Bicocca-FT-02-18\\MS-TP-02-2\\HU-EP-02/37
\\September 2002}
}\\[15mm]
\vspace{-0.15cm}
%\title{
Recent results on the running coupling in QCD with two massless
flavours\thanks{Talk~given~by~M.~Della~Morte.~Work supported in part by the
European Community's Human Potential Programme, contract
HPRN-CT-2000-00145, Hadrons/Lattice QCD.}}
\author{M.~Della Morte\address[Zeu]{DESY Zeuthen, 15738
Zeuthen, Germany}, R.~Frezzotti\address[Mi]{INFN-Milano and Universit\`a
Milano--Bicocca, Piazza della Scienza 3, 20126 Milano, Italy},
J.~Heitger\address[Mu]{WWU M\"unster, Institut f\"ur Theoretische Physik,
Wilhelm--Klemm-Str. 9, 48149 M\"unster, Germany},
F.~Knechtli\address[HU]{Institut f\"ur Physik, Humboldt--Universit\"at zu
Berlin, Invalidenstr. 110, 10115 Berlin, Germany}, J.~Rolf\addressmark[HU],
R.~Sommer\addressmark[Zeu] and U.~Wolff\addressmark[HU] \newline
(ALPHA Collaboration)}
\begin{document}
\begin{abstract}
We report on the latest results on
the running coupling of two flavour QCD in the Schr\"odinger
functional scheme. Results for the step scaling function are
obtained from simulations on lattices $L/a=8$ and $L/a=16$
which confirm the first results from
lattices $L/a=4,5,6$ presented one year ago by the ALPHA collaboration.
We also discuss some algorithmic aspects, in particular concerning the
occurrence of metastable states. A modified sampling, in order to estimate
the proper weight of
these states in the path integral, is proposed and tested.
%Extensions of this modified sampling to other problems (e.g. sampling
%of topological sectors) are conceivable.
\vspace{1pc}
\end{abstract}
% typeset front matter (including abstract)
\maketitle
\section{INTRODUCTION}
The relation between the short distance regime of QCD and its low energy
sector is a problem which can be addressed by lattice techniques. The $\Lambda$
parameter, characterizing the coupling at large energy, can be related to a
hadronic quantity by a recursive finite--size technique, avoiding in such a
way the multiple scale problem. This method has been successfully applied to
the zero flavour approximation of QCD~\cite{q_coup} and first results
extending it to QCD with two massless flavours have been presented
last year by our collaboration~\cite{coup_lett}.
Here we report about more recent results which allow for a better
estimate of systematic effects on our determination of the $\Lambda$
parameter in the two flavour theory. This result is essential also
in the computation of the running quark mass~\cite{run_mass}.
We consider the Schr\"odinger functional (SF), defined as the Euclidean
partition function of QCD on a cylinder of size $L^3 \times T$
\vspace{-0.12cm}
\begin{equation}
e^{-\Gamma}=\int_{T\times L^3} D[U,\psi,\bar{\psi}] e^{-S} \;,
\end{equation}
\vspace{-0.12cm}
with periodic (up to a phase) boundary conditions in the spatial directions
and Dirichlet boundary conditions at $x_0=0,T$.
The spatial links at $x_0=0,T$ are fixed to diagonal SU(3) matrices
specified in terms of $L$ and one angle $\eta$ while the
quark fields are given by Grassmann values $\rho$, $\bar{\rho}$ and
$\rho'$, $\bar{\rho}'$, which are used as sources and then set to zero.
We work with an O($a$) improved Wilson action and improved operators, employing
the non--perturbative $c_{\rm SW}$~\cite{csw} coefficient and one and
two--loop values for the
boundary coefficients $\tilde{c}_t$ and $c_t$~\cite{cA,ctctt}. The
improvement coefficient $c_{\rm A}$ is also set to its one--loop value. It
enters our definition of quark mass $m$ via the PCAC relation.
The mass is
required to be consistent with zero~\cite{coup_lett}.
For $T=L$, the SF coupling $\bar{g}^2$ is defined~\cite{q_coup,SFYM}
\vspace{-0.07cm}
\begin{equation}
\left. (\partial\Gamma/\partial \eta) \right|_{\eta=0} =
\left. \langle \partial S/\partial \eta \rangle \right|_{\eta=0} =
k\bar{g}^{-2}(L)
%{{\partial\Gamma}\over{\partial \eta}} \right|_{\eta=0} =
% {{k}\over{\bar{g}^2(L)}} \; ,
\end{equation}
\vspace{-0.07cm}
with $k$ fixed by the condition $\bar{g}^2=g_0^2+O(g_0^4)$.
Here $L$ plays the role of a renormalization scale.
%
\vspace{-0.1cm}
\section{NUMERICAL RESULTS}
\vspace{-0.1cm}
The main quantity in the present computation is the step scaling function (SSF)
\vspace{-0.02cm}
\begin{equation}
\sigma(u)=\bar{g}^2(2L) |_{\bar{g}^2(L)=u,m=0} \; ,
\end{equation}
which can be viewed as a non--perturbative integrated form of the
$\beta$--function. Once $\sigma(u)$ has been computed in the continuum limit
and for a large enough range of values, the equation
\begin{equation}
\sigma(\bar{g}^2(L/2))= \bar{g}^2(L)
\end{equation}
can be recursively solved $n$ times (for explicit equations see
e.g.~\cite{run_mass}) starting from the hadronic
scale $L_{\rm max}$ with $\bar{g}^2(L_{\rm max})=u_{\rm max}$.
In this way we obtain
values for $\bar{g}^2(2^{-n}L_{\rm max})$. For large $n$ this coupling is
perturbative and we compute the $\Lambda$ parameter in units of $L_{\rm max}$
making use of the three--loop $\beta$ function in the SF scheme~\cite{coup_lett,ctctt}
\begin{eqnarray}
&&\Lambda L_{\rm max} = 2^n (b_0\bar{g}^2)^{-b_1/2b_0^2}
\exp\left\{ -\frac1{2b_0\bar{g}^2} \right\} \nonumber\\
&&\times
\exp\left\{ -\int_0^{\bar{g}} dx \left[
\frac1{\beta(x)} + \frac1{b_0 x^3} -\frac{b_1}{b_0^2 x} \right]\right\}\
\label{Lambdapert}
\end{eqnarray}
Later we will have to relate $L_{\rm max}$ to a physical scale, e.g. by
determining $L_{\rm max}F_{\pi}$.
%
%Our simulations and results are summarized in Table~1.
%With respect to Ref.~\cite{coup_lett} we include here new results from
%simulations with $L/a=8$ and $L/a=16$. These allow for more reliable
%continuum limit extrapolations of the SSF. The simulations have been performed
% on the APEmille machines at DESY Zeuthen.
\begin{table}[htbp]\label{table:1}
\vspace{-0.5cm}
\small{
\begin{center}
\begin{tabular}{cllll} \hline
$\!\!\!\!L/a\!\!\!\!$ & \multicolumn{1}{c}{$\!\!u$} &
\multicolumn{1}{c}{$\!\!\Sigma$}
& \multicolumn{1}{c}{$\!\!u$} & \multicolumn{1}{c}{$\!\!\Sigma$} \\ \hline
& & \\[-2ex]
% $\!\!\!\!4\!\!\!\!$ & $\!0.9793(7)$ & $\!1.0643(34) \rule{0ex}{1ex}$
% & $\!1.5031(12)$ & $\!1.720(5) $ \\
% $\!\!\!\!5\!\!\!\!$ & $\!0.9793(6)$ & $\!1.0721(39) $
% & $\!1.5033(26)$ & $\!1.737(10) $ \\
% $\!\!\!\!6\!\!\!\!$ & $\!0.9793(11)$ & $\!1.0802(44) $
% & $\!1.5031(30)$ & $\!1.730(12) $ \\
$\!\!\!\!8\!\!\!\!$ & $\!\!0.9807(17)$ & $\!\!1.0745(55) $
&$\!\!1.508(4)$& $\!\!1.716(14)$\\[0.5ex]
% $\!\!\!\!4\!\!\!\!$ & $\!1.1814(5)$ & $\!1.3154(55) $
% & $\!2.0142(24)$ & $\!2.481(17) $ \\
% $\!\!\!\!5\!\!\!\!$ & $\!1.1807(12)$ & $\!1.3287(59)$
% & $\!2.0142(44)$ & $\!2.438(19) $ \\
% $\!\!\!\!6\!\!\!\!$ & $\!1.1814(15)$ & $\!1.3253(67)$
% & $\!2.0146(56)$ & $\!2.508(26) $ \\
$\!\!\!\!8\!\!\!\!$ & $\!\!1.1818(29)$ & $\!\!1.3338(58) $
& $\!\!2.014(10)$ & $\!\!2.475(31) $ \\[0.5ex]
% $\!\!\!\!4\!\!\!\!$ & $\!1.5031(10)$ & $\!1.731(6) $
% & $\!2.4792(34)$ & $\!3.251(28) $ \\
% $\!\!\!\!5\!\!\!\!$ & $\!1.5031(20)$ & $\!1.758(11) $
% & $\!2.4792(73)$ & $\!3.336(50) $ \\
% $\!\!\!\!6\!\!\!\!$ & $\!1.5031(25)$ & $\!1.745(12) $
% & $\!2.4792(82)$ & $\!3.236(36) $ \\
$\!\!\!\!8\!\!\!\!$ & &
& $\!\!2.479(13)$ &$\!\!3.348(48)$\\[0.5ex]
$\!\!\!\!4\!\!\!\!$
% & $\!\!1.7319(11)$ & $\!\!2.058(7) $
& &
& $\!\!3.334(11)$ & $\!\!5.513(42) $ \\
% $\!\!\!\!5\!\!\!\!$ & $\!1.7333(32)$ & $\!2.086(21) $
% & $\!3.334(15)$ & $\!5.41(12) $ \\
$\!\!\!\!6\!\!\!\!$
% & $\!\!1.7319(34)$ & $\!\!2.058(20) $
& &
& $\!\!3.326(20)$ & $\!\!5.62(9) $ \\[0.5ex]\hline
% $\!\!\!\!8\!\!\!\!$ & &
% &$\!\!3.334(19) $ & \\[0.5ex]\hline
\end{tabular}
\end{center}
\caption{New data for the lattice SSF
$\Sigma(u,a/L)$. $c_{\rm t} (g_0)$ was set
to its one--loop value for the left part, and to two--loop for the two
rightmost columns.}
\vspace{-0.65cm}
}
\end{table}
%
Our new simulations and results
%compared to Ref.~\cite{coup_lett}
are summarized in Table~1.
These allow for more reliable
continuum limit extrapolations of the SSF.
%The simulations have been performed on the APEmille machines at DESY Zeuthen.
%
We extract the continuum limit values for the SSF using data for the two--loop
improved observable $\Sigma^{(2)}(u,a/L)$~\cite{coup_lett} from
$L/a=6$ and $L/a=8$.
The scaling of $\Sigma^{(2)}$ vs. $(a/L)^2$ is shown in Figure~\ref{fig_scal}
for two different values of $u$. For comparison we included in the plot the
continuum results in the quenched case (black squares) for the same
value of the couplings.
\begin{figure}[htb]
\vspace{-0.3cm}
\includegraphics[width=18pc]{Sigma_nf2.eps}
\vspace{-1.32cm}
\caption{Scaling plot for $\Sigma^{(2)}$.
Black squares indicate the continuum limit extrapolation for $N_{\rm f}=0$.}
\label{fig_scal}
\vspace{-0.7cm}
\end{figure}
%
The function $\sigma(u)$ is obtained interpolating the continuum
limit estimates
by a sixth--order polynomial with the first three coefficients constrained by
perturbation theory. Starting from $u_{\rm max}\!=\!3.3$ or $u_{\rm max}\!=\!5$ we
can then estimate the quantity $\Lambda L_{\rm max}$ as discussed above.
Results are summarized in Table~2.
%
\begin{table}[htbp]
\vspace{-0.4cm}
\begin{center}
\begin{tabular}
{ccc|ccc}
\hline
& \multicolumn{2}{c|}{$\bar{g}^2(L_{\rm max})=3.3$} &
& \multicolumn{2}{c}{$\bar{g}^2(L_{\rm max}')=5$} \\
$n$ & c.l. 6/8 & $L/a=5$ & $n$ & $\!\!6/8$ & $\!\!L/a=5$ \\
\hline
5 & 1.82(5) & 1.87 & 6 & 1.23(5) & 1.27 \\
6 & 1.84(6) & 1.89 & 7 & 1.25(6) & 1.28 \\
7 & 1.85(7) & 1.91 & 8 & 1.26(7) & 1.30 \\[0.5ex]
\hline
\end{tabular}
\vspace{0.15cm}
\caption{Values for $-\ln(\Lambda L_{\rm max})$ for two choices of
$L_{\rm max}$.}
\end{center}
\vspace{-1cm}
\end{table}
%
We use the difference between continuum values and data from $L/a=5$
to estimate our systematic uncertainty, which we linearly add to the
statistical errors. This yields
%\vspace{-0.05cm}
\begin{eqnarray}
\ln(\Lambda L_{\rm max}) &=& -1.85(13) \quad [\bar{g}^2(L_{\rm max})=3.3]
\nonumber \\
\ln(\Lambda L_{\rm max}') &=& -1.26(11) \quad [\bar{g}^2(L_{\rm max}')=5] \, .
\label{Lam5}
\end{eqnarray}
%\vspace{-0.05cm}
Comparing with the results in~\cite{coup_lett}, we see that the systematic
uncertainty on $\ln(\Lambda L_{\rm max})$ is nearly halved, while the central
values are fully consistent.
The non--perturbative evolution
of $\alpha(\mu)=\bar{g}^2(L)/4\pi$ $(\mu=1/L)$ starting from $\bar{g}^2=5$
is plotted in Figure~\ref{run_fig}.
Statistical and systematical errors (difference between continuum limit and
$L/a=5$) are too small to be visible in the plot.
The figure shows that differences between Monte
Carlo results and three--loop perturbation theory become appreciable for large
values of the coupling, while for high energies the perturbative
approximation very closely reproduces the non--perturbative results.
%
\begin{figure}[htb]
\vspace{-2.2cm}
\includegraphics[width=19pc]{alpha.eps}
\vspace{-2.4cm}
\label{run_fig}
\caption{Running of $\alpha\!=\!\bar{g}^2/4\pi$ in the SF scheme.}
\vspace{-0.7cm}
\end{figure}
%
\vspace{-0.15cm}
\section{METASTABLE STATES}
\vspace{-0.05cm}
For the largest values of the coupling in our study ($\bar{g}^2\simeq 3.3$ and
$\bar{g}^2 \simeq 5.5$) we observed a long tail towards negative values in the
distribution of
$\partial S/\partial \eta$. In addition, measuring by a cooling procedure
the pure gauge contribution to the action $S_{\rm G}^{cool}$,
and to the coupling $\partial S_{\rm G}^{cool} / \partial \eta $,
we observed occurrences of
metastable states in our Monte Carlo histories. The action $S_{\rm G}^{cool}$
for these states is consistent with the value for a secondary solution of the
field equations~\cite{SFYM}, given our choice for the boundary fields.
This solution appears to be a local minimum by numerical evidence.
In order to properly estimate the weight of these states
we enhanced their occurrence through a modified sampling~\cite{q_coup},
adding to the
HMC effective action a term
%\vspace{-0.02cm}
\begin{equation}
\left. \gamma{{\partial S_{\rm G}}\over{\partial \eta}} \right|_{\eta=0} +
{{1}\over{w_{\gamma}}}(\gamma-\gamma_0)^2 \;,
\end{equation}
%\vspace{-0.02cm}
where $\gamma_0$ and $w_{\gamma}$ are fixed values, while $\gamma$ is a
dynamical variable. The expectation values in the
original ensemble are then obtained by reweighting properly the observables.
Defining a quantity $q$ whose value is 1 for metastable states and 0
otherwise (so that $\langle \delta_{q1}\rangle +\langle \delta_{q0}\rangle=1$),
for an observable $O$ one can write
%\vspace{-0.05cm}
\begin{eqnarray}
\langle O\rangle &=& \langle \delta_{q1}O\rangle +
\langle\delta_{q0}O\rangle
%\over{\langle \delta_{q1}\rangle + \langle\delta_{q0}\rangle}}
\\ \nonumber
\vspace{0.15cm}
&=&\langle \delta_{q1}O\rangle + \langle O \rangle_{\bar{1}}
(1-\langle \delta_{q1}\rangle) \; ,
\end{eqnarray}
%\vspace{-0.1cm}
where $\langle O \rangle_{\bar{1}}=\langle \delta_{q0}O\rangle/
\langle \delta_{q0}\rangle$. If the main contribution to
$\langle O \rangle$ comes from the $q=0$
sector then a precise estimate of $\langle O \rangle$ just requires
a precise estimate of $\langle O \rangle_{\bar{1}}$, which can be
obtained by an algorithm which samples only the
$q=0$ sector, together with rough estimates of
$\langle \delta_{q1}\rangle$
and $\langle \delta_{q1}O\rangle$ which can be
obtained by the modified sampling.
By a set of simulations on $L/a=8$ and $12$ using this approach we
could estimate the effect of metastable states on $\bar{g}^2$ to be nearly
$0.10(2)\%$ for $\bar{g}^2 \simeq 3.3$.
This result is independent, within errors, from $L/a$.
Moreover, for this value of the coupling, we noticed a slightly better
behavior for the PHMC algorithm in comparison with HMC (smaller values of
$\gamma_0$ needed in order to generate a certain number of transitions
from/to a metastable state).
For $\bar{g}^2\simeq 5.5$, the effect of metastable states is
much larger, as expected, resulting in a $4$ to $5\%$ contribution.
However their sampling is also enhanced in this case using both PHMC or HMC
already in the original ensemble.
Indeed we repeated the $L/a=12$ simulation for $\bar{g}^2 \simeq 5.5$ using
ordinary PHMC as in Ref.~\cite{coup_lett} but disentangling the occurrence of
metastable states. This turned out to be around $6\%$ and the result
for $\bar{g}^2$ was fully consistent with the number in Ref.~\cite{coup_lett}.
{\bf Acknowledgments}. We thank NIC/DESY Zeuthen for allocating computer time
on the
APEmille machines to this project, and the APE group for their support.
%
\vspace{-0.15cm}
\begin{thebibliography}{9}
\vspace{-0.15cm}
\bibitem{q_coup} M. L\"uscher, R. Sommer, P. Weisz and U. Wolff,
Nucl. Phys. B413 (1994) 481.
\bibitem{coup_lett}
A. Bode et al., ALPHA collaboration,
Phys. Lett. B515 (2001) 49.
\bibitem{run_mass}
F.~Knechtli et al., ALPHA collaboration, these proceedings [hep-lat/0209025].
\bibitem{csw}
K.~Jansen and R.~Sommer, Nucl. Phys. B530 (1998) 185.
\bibitem{cA}
M.~L\"uscher and P.~Weisz, Nucl. Phys. B479 (1996) 429.
\bibitem{ctctt}
A.~Bode, P.~Weisz and U.~Wolff, Nucl. Phys. B576 (2000) 517 [2nd Erratum-ibid
hep-lat/9911018v3].
\bibitem{SFYM}
M. L\"uscher, R. Narayanan, P. Weisz and U. Wolff,
Nucl. Phys. B384 (1992) 168.
\end{thebibliography}
\end{document}