Abstract. Because the time needed for a simulation in lattice QCD varies at a rate exceeding the fourth power of the lattice size, it is important to understand how small one can make a lattice without altering the physics beyond recognition. It is common to use a rule of thumb that the pion mass times the lattice size should be greater than (ideally much greater than) four (i.e., mπL ≫ 4). By considering a relatively simple chiral quark model we are led to suggest that a more realistic constraint would be mπ(L − 2R) ≫ 4, where R is the radius of the confinement region, which for these purposes could be taken to be around 0.8-1.0 fm. Within the model we demonstrate that violating the second condition can lead to unphysical behaviour of hadronic properties as a function of pion mass. In particular, the axial charge of the nucleon is found to decrease quite rapidly as the chiral limit is approached.
IntroductionOur current capacity to compute hadron properties in lattice QCD is constrained by the need to take many limits. For example, we need to take the spacing a → 0, the size of the lattice L → ∞ and the quark mass to around 5 MeV. Of course, these limits are not unconnected because the size of a region of space big enough to contain (say) a proton will need to grow with the Compton wavelength of the pion. At present, with lattice spacings that represent a reasonable approximation to the continuum limit and for full QCD with reasonable chiral symmetry, the time for a lattice simulation scales roughly like m −9 π and we are limited to pion masses larger than 0.4-0.5 GeV. This has led to much effort to explore the application of chiral perturbation theory as a tool for extrapolating hadron properties to the physical pion mass [1][2][3].In such an environment there is great interest in seeing whether one can lower the pion mass without increasing the size of the lattice, thereby saving a factor of m 4 π . Indeed, there have been many calculations for which even the rather optimistic rule of thumb that m π L > 4 has not been satisfied. Considerable attention has been devoted to studying such systems within the framework of effective field theory in order to understand how the relevant path integral might change as we go from one regime to another [4,5].The question we ask is somewhat different. We consider a simple chiral quark model upon which we impose boundary conditions which roughly approximate those on a lattice. Simple inspection of the solutions naturally leads one to conclude that the condition noted above is incorrect. Indeed, the pion cloud of the nucleon does not even begin until one is outside the region of space in which the valence quarks are confined. Within the bag model this is characterised by the bag radius, R, and within the cloudy bag model as well as the model