1:
2:
3: \section 7. Concluding remarks
4:
5: In the coming years, simulations of lattice QCD with Wilson quarks
6: will no doubt rapidly progress towards
7: smaller quark masses and lattice spacings than are
8: reported here. In order to guarantee
9: the stability of the simulations [\ref{Stability}],
10: but also to keep
11: the finite-volume effects under control, the constraints
12: \equation{
13: \Mpi L\geq3, \qquad L\geq 2\,\fm,
14: \enum
15: }
16: should be respected in these computations. On a given
17: lattice, the bounds (7.1) set a lower limit on the
18: lattice spacings and pion masses that can be reached (see fig.~8).
19: Simulations of the two-flavour theory
20: may not be practical at all these points,
21: but the cost formula (2.1) is encouraging and suggests that
22: simulations at $a\leq0.08$ fm and $\Mpi\leq300$ MeV, for example,
23: can be performed already with the
24: computer resources available at present. However, to be able to sort
25: out the systematic errors, many lattices will have to be simulated
26: which may require a coordinated effort.
27:
28: \input figure8
29:
30: On the $A$, $B$ and $D$ series of lattices, the smallest values of
31: $\Mpi L$ are
32: $3.5$, $3.2$ and $3.6$ respectively, while the spatial sizes $L$ of the
33: lattices are estimated to be $1.72$, $1.67$ and $1.88$ fm, i.e.~somewhat
34: below the required minimum. Finite-volume effects
35: may not be totally negligible on these lattices
36: and will need to be investigated, extending the
37: studies by Orth et al.~[\ref{OrthEtAl}]
38: to smaller quark masses and lattice spacings.
39: So far we did not include the nucleons
40: in the physics analysis, because these are probably
41: even more sensitive to finite-volume effects
42: than the mesons.
43:
44: It may be somewhat surprising that no significant
45: lattice effects were seen in fig.~4, even though
46: O($a$) counterterms were only included in the $D$ series
47: of simulations.
48: The weak dependence on the lattice spacing could be
49: related to the fact, first noted by
50: Sharpe and Singleton [\ref{SharpeSingleton}], that
51: the O($a$) lattice effects amount to
52: an additive quark-mass renormalization
53: at leading order of chiral perturbation theory.
54: Since the quark masses that appear in
55: the PCAC relation already
56: include all additive re\-nor\-ma\-li\-zations, it follows that
57: the data points plotted in fig.~4 are insensitive
58: to these leading-order lattice effects.
59:
60: The mass dependence of the pseudo-scalar decay constant, on the other hand,
61: is a second-order effect in chiral perturbation theory.
62: At this order,
63: only some of the O($a$) lattice corrections
64: can be compensated by a renormalization of the
65: parameters in the chiral lagrangian,
66: and an accidental O($a$) improvement
67: is therefore not expected in this case.
68:
69: \vskip1.0ex
70: We wish to thank Rainer Sommer for technical discussions
71: and Gilberto Colangelo, Stephan D\"urr, J\"urg Gasser and
72: Heiri Leutwyler for some very helpful notes, summarizing
73: some relevant results of chiral perturbation theory.
74: The numerical simulations were performed on PC
75: clusters at CERN, the Centro Enrico Fermi, the Institut f\"ur
76: Theoretische Physik der Universit\"at Bern (with a contribution from
77: the Schweizerischer Nationalfonds) and on a CRAY XT3 at the Swiss
78: National Super\-computing Centre (CSCS).
79: We are grateful to all these institutions for the continuous support
80: given to this project.
81:
82: