Skip to main content

Full text of "Improving the staggered quark action to reduce flavour symmetry violations"

See other formats



Improving the staggered quark action to reduce flavour symmetry 

J.-F. Lagae and D. K. Sinclair 3 '''' 

a HEP Division, Argonnc National Laboratory, 9700 South Cass Avenue, Argonne, IL, 60439, USA 

We investigate a class of actions for lattice QCD with staggered quarks aimed at reducing the flavour symmetry 
violations associated with using staggered fermions. These actions replace the gauge field link fields in the quark 
action with covariantly smeared fields. As such they are an extension of actions considered by the MILC collabo- 
ration. We show that such actions systematically reduce flavour symmetry violations in the weak coupling limit. 
Using the mass splitting between Goldstone and non-Goldstone pions as a measure of flavour symmetry violations 
we find that these actions have considerably less flavour symmetry violations than the standard staggered action, 
and represent an improvement on what can be achieved with the MILC action, on quenched configurations with 
P = 5.7. 







Improving the lattice QCD action using the 
method of Lepage et al. |jj and the perfect ac- 
tion method [Q] has allowed simulations with rel- 
atively large lattice spacings (~ 0.5/m). While 
these methods have proved successful for lattice 
QCD with Wilson § or non-relativistic quarks 
H| they have been less successful for staggered 
(Kogut-Susskind) quarks |4|||. 

Improved staggered quark actions such as 
the Naik(l+3-link) [0], aimed at improving the 
fcrmion dispersion relations to bring them closer 
to their continuum form, do provide some im- 
provement. However, they do not significantly 
decrease the flavour symmetry violations which 
plague the staggered quark action. The simplest 
of "perfect" actions also exhibit this deficiency 

The reason appears to be that these actions 
concentrate on improving the low momentum 
dispersion relations for the fermions associated 
with the "corners" of the Brillouin zone (and the 
gauge fields) and on making the interactions be- 
tween low momentum fields more continuum-like. 
Flavour symmetry violation occurs because the 
momentum transferred to the fermions by the 

gauge fields can change them from one flavour 
to another, and such transfers are not suppressed 
by such improvements. 

Recently the MILC collaboration have in- 
troduced an action where the gauge links in the 
staggered fermion action are replaced with a lin- 
ear combination of a single gauge link and a sum 
over the six 3-link "staples" joining the same sites. 
This does appreciably decrease flavour symmetry 
violations as measured by the splitting between 
the Goldstone and non-Goldstone pions. The rea- 
son for this is clear. Such a replacement effec- 
tively smooths the gauge field interaction with the 
fermions, thus reducing large momentum trans- 
fers at the vertex responsible for flavour symme- 
try breaking. 

We have improved on this scheme in a sys- 
tematic fashion. At tree level, our scheme sup- 
presses flavour symmetry breaking by an addi- 
tional power of a 2 . We assume that an action of 
the same form, but with different coefficients, is 
the appropriate choice beyond tree level. 



"Talk presented by D. K. Sinclair, at LATTICE'97, Edin- 
burgh, Scotland. 

tSupportcd by DOE contract W-31-109-ENG-38. 

The 4 flavours and 4 Dirac components in one 
staggered fermion are associated with the 16 poles 
of the free propagator. For massless quarks these 
occur at ~ 0, tt. Hence, at weak coupling, we 


can suppress flavour mixing to higher order in a if 
we can suppress coupling to gluons which couple 
the neighbourhood of one of these poles to any 
other. In particular we should suppress coupling 
to gluons with momentum components or 7T, 
but not all zero. 

To see how to do this let us ignore, for the 
moment, the complications of gauge invariance. 
Then the coupling of the gluon and quark fields 
would be of the form 

iip' (n)r) li (n)A (l {n)'ip(Ti + fi) + h.c. (1) 

We could then achieve our objective by making 
the substitution 

M n ) -» 7^(2 + Di + D_x)(2 + £> a + D_ 2 ) 

x(2 + D 3 + D_ 3 )(2 + D 4 + D^)A^n) (2) 

D ±v A M (n)=A li {n±v) (3) 
Then in momentum space 

AJk) -» — (1 + cosfci)(l +cosfc 2 ) 

(1 + cosfc 3 )(l + cos k 4 )Ap(k) (4) 

which vanishes when any component of k is 7r, and 
— > -^(fc) as fc — > 0. For fc having any component 
within O(a) of 7T, the coupling is suppressed by ~ 
a 2 . Thus this substitution would achieve precisely 
what we want. Note also that adding a fermion 
mass shifts the poles in the fermion propagator by 
0(a), and so does not invalidate these arguments. 

To make this approach gauge invariant, we re- 
place Ap by the standard gauge link variable 
= expiAp. The displacement operators D± v 
are replaced by covariant displacement operators, 
for example 

D u U„(n) = U v {n)Up{n + u)Ul{n + fi). (5) 

As one moves away from weak coupling, the co- 
efficients in this new action will develop g depen- 
dence, so that these coefficients are considered 
as free parameters to be adjusted to minimize 
flavour symmetry violations. Equation || then be- 

Up -> V„ = C{x + 2y Q 

+ ^2[x 1 D v + yi(D lu/ + D- lu/ )] 


+ J2i x sDupX + VsiD^px + D.^px)]}^ (6) 


where u, p and A are summed over ±1, ±2, ±3, 
±4 with \v\, \p\, \X\ all different, and C = 
1 / (x + 2y + 6xi + 12j/i + 1 2 x 2 + 2Ay 2 + Sx 3 + 16y 3 ) . 
The Z?'s acting on Up displace it in each of the 
subscript directions. Products of U's join the 
ends of the displaced link to the sites at the ends 
of the undisplaced Up so as to ensure gauge in- 
variance. To reduce the number of independent 
parameters, we symmetrize over all such minimal 
length paths joining the undisplaced sites to dis- 
placed sites. In general, we could assign different 
weights to such paths, provided we preserved lat- 
tice symmetries. 

The tree level values for these coefficients are 
a^o = 1, 2/o = xi = 1/2, yi = x 2 = 1/4, y 2 = x 3 = 
1/8, y 3 = 1/16. The MILC action has x = 1, 
xi = u, and x 2 = x 3 = yi = y 2 = y 3 = 0. 

The first check of this fermion action is to see 
that it does indeed suppress coupling to gluons 
with momentum components or n but not all 
zero, in the weak coupling limit when the coeffi- 
cients are set to their tree values. For this we use 
the weak coupling approximation 

Up^l + iAp. (7) 

and expand to first order in Ap. Since we 
are dealing with non-gauge-invariant quantities, 
we work in Landau gauge. In this approximation 
U p(k) — 1 vanishes at kp — 0,7r, except when 
k\ = k 2 = k 3 = &4 = 0, which is the required 


The true test of this new action is how it re- 
duces the flavour symmetry violation, as mea- 
sured by the tt 2 - n mass difference for values 
of P = Q/g 2 where this splitting is quite large for 
the standard action 0. 

We have measured the spectrum of local light 


hadrons on an ensemble of 203 quenched config- 
urations generated with the standard action at 
(3 — 5.7 on an 8 3 x 32 lattice. We have used 
the standard staggered quark action, the MILC 
action and our new improved action, for various 
values of their parameters. For the results pre- 
sented here, we used a 1 parameter subset of im- 
proved actions for which x n = x n , y n — x n+1 . 
For the MILC actions xo = 1, X\ = u, and 
X2 = X3 = Hi = yi = 2/3 = 0, as mentioned 

We plot the standard action results for and 
to^ in figure 1, along with those for MILC1 (the 
MILC action with u> = 1) and imp2 (our improved 
action with x = 1), which show the most improve- 
ment of those actions we have considered in each 




Figure 1. Comparison of standard, MILC and im- 
proved actions for masses of the Goldstone pion, 
7r and local non- Goldstone pion, -k-i- 

Since the p mass for the standard action, ex- 
trapolated to niq = 0, is ~ 15% larger than that 
for the MILC or improved actions, we should re- 
ally compare these with standard action results 
from a lattice with ~ 15% larger lattice spac- 

ing, where the improvement will be less dramatic. 
However, a reduction of 7r-7r 2 mass splitting of a 
factor of 2 for the MILC actions, as reported at 
[3 = 5.85 by the MILC collaboration seems likely. 

Since the p masses for the MILC and improved 
actions are very close, a direct comparison of 7r- 
7r 2 mass splittings is justified. Looking at an ex- 
trapolation to zero quark mass suggests that our 
improved action could reduce this mass splitting 
by around 40%. 


Actions which are designed to systematically 
suppress the coupling of high-momentum gluons, 
which give rise to flavour symmetry violations, 
reduce flavour symmetry violations in the pion 
mass spectrum by ~ 40% over the MILC action, 
and hence by a factor of ~ 3 over the normal stag- 
gered fermion action, on quenched configurations 
at (3 = 5.7. We are repeating these calculations 
on a larger lattice (16 3 x 32) to remove finite size 

Such improvements should be combined with 
an improved gauge action, and with a fermion 
action designed to improve the low-momentum 
dispersion relations (Naik or perfect action). 

More careful studies of such actions away from 
the weak coupling limit are needed. Finally we 
should relate these actions to those studied by 
Luo §. 


1. M. Afford et al., Phys. Lett. B361, 87 (1995); 
Nucl. Phys. B496, 377 (1997). 

2. P. Hasenfratz and F. Niedermayer, Nucl. 
Phys. B414 (1994). 

3. G. P. Lepage et al., Phys. Rev. D46 4052, 

4. C. Bernard et al., Nucl. Phys. B(Proc. 
Suppl.)53, 212 (1997). 

5. A. Peikert et al., poster presented at LAT- 
TICE'97 (1997). 

6. S. Naik, Nucl. Phys. B316, 238 (1989). 

7. T. Blum, et al, Phys. Rev. D55, 1133 (1997) 

8. Y. Luo, hep-lat e-print |hcp-lat/9702013