Lattice calculation of $1^{-+}$ hybrid mesons with improved Kogut-Susskind fermions

Lattice calculation of 1 + hybrid mesons with improved
Kogut-Susskind fermions

C. Bernard

Department of Physics, Washington University, St. Louis, MO 63130, USA

T. Burch, E.B. Gregory, and D. Toussaint
I Department of Physics, University of Arizona, Tucson, AZ 85721, USA

O

^ ' C. DeTar and J. Osborn

> ■

O ' Physics Department, University of Utah, Salt Lake City, UT 8^112, USA

^ \ Steven Gottlieb

^ ■ Department of Physics, Indiana University, Bloomington, IN 47405, USA

(N

O ■ U.M. Heller

^ I

O ■ American Physical Society, One Research Road, Box 9000, Ridge, NY 11961-9000

O

X

R. Sugar

Department of Physics, University of California, Santa Barbara, CA 93106, USA

Oh!
CD .

^ , (Dated: February 1, 2008)

Abstract

We report on a lattice determination of the mass of the exotic 1 hybrid meson using an
improved Kogut-Susskind action. Results from both quenched and dynamical quark simulations
are presented. We also compare with earlier results using Wilson quarks at heavier quark masses.
The results on lattices with three flavors of dynamical quarks show effects of sea quarks on the
hybrid propagators which probably result from coupling to two meson states. We extrapolate the
quenched results to the physical light quark mass to allow comparison with experimental candidates
for the 1 ^ hybrid meson. The lattice result remains somewhat heavier than the experimental
result, although it may be consistent with the 7ri(1600).

PACS numbers: 11.15Ha,12.38.Gc

1

I. INTRODUCTION

The fact that gluons carry color charge suggests that they, hke quarks, could be "valence"
constituents of hadrons. In other words, we expect that the spectrum of QCD should contain
glueballs and hybrids, or particles with both quarks and gluons as valence constituents.
Hybrid mesons can have exotic quantum numbers, or J^^ combinations not possible with
a quark- ant iquark state. However, a state with exotic quantum numbers is not necessarily
a hybrid — it could be a qqqq state, realized either as a single "bag" containing four quarks
or as a "molecule" made of two qq mesons. Experimental evidence suggests the existence
of one or more mesons with exotic quantum numbers J^^ = namely the 7ri(1400)
and the 7ri(1600) j^. Analjd;ic and numerical methods to predict the mass of light hybrid
meson states include flux tube models |3|, the bag model

mm

, QCD spectral sum

rule s [slJ ll ,

12j |. relativistic WKB calculations and lattice QCD. Several lattice studies

Il5l | have used quenched Wilson or quenched Wilson-clover fermions to calculate the
masses of exotic hybrid states, although with quark masses much larger than the physical u
and d quark masses. Lacock and Schilling have done a calculation in two flavor QCD, again
with fairly heavy quarks jiol .

Here we report results of a lattice calculation of the mass of a 1 ^ hybrid meson using
improved Kogut-Susskind quarks. The use of Kogut-Susskind quarks allows us to work
at valence quark masses much smaller than were used in previous lattice calculations. In
addition, the "a^^^^" action that we use has leading lattice spacing errors of order a^g"^,
while the clover- Wilson action has errors of order a^. Our mass estimates in the quenched
approximation are consistent with earlier Wilson quark results, but extrapolation to the
physical light valence quark masses is under much better control. Preliminary results of this
calculation were reported in Ref. 17 1.

We have also calculated hybrid meson propagators including the effects of three flavors
of dynamical quarks, with light sea quark masses down to 0.4 times the strange quark mass.
We find that extracting mass estimates from the propagators in full QCD is difficult, and
we argue that this difficulty is due to mixing of the hybrid meson with two meson states —
the states into which it might decay.

2

II. 1+ HYBRID MESON OPERATOR

We can construct a l""*" hybrid meson operator as the cross product of a color octet 1
quark-antiquark (p meson) operator and the chromomagnetic field, which has J^*" = 1+^:
p X B With staggered quarks we have several choices of rho meson operators, but it

is convenient to choose the taste ^ singlet ps, with the spin ® taste structure 7i ® 1.

= 2 7/'"7,-®l^''i^f , (1)

where i, j and k are spatial indices and a and b are color indices. Each spin component of
the 1 ^ includes two terms, for example:

^-^ = PyB.-p.By , (2)

so if we had chosen a spin ® taste structure like 7^ ® 7^ the two components of l""*" would
have different tastes.

The Kogut-Susskind ps meson operator, with spin aligned in the k direction is xVkDkX^
jisl ] where x ^i^^id x ^i^^e the quark and antiquark fields respectively. The covariant symmetric
shift operator is given by

DM^) = ^[Ul{x - fi)q{x - fi) + U^{x)q{x + fi)] . (3)

We compute the field strength at each lattice point usin g th e four plaquettes in each

plane that have corners at this point, as described in Ref. In computing the field

strength, we use links that have been smoothed with 32 iterations of APE smearing in the
spatial directions only with relative weight of the staples set to 0.25 This smearing

removes short wavelength fluctuations in the gluon field, and reduces the noise in the hybrid
propagator. (The smeared links are only used in constructing F^^; the propagators are
computed using the original links.)

Our zero momentum hybrid source and sink wave functions are constructed in Coulomb
gauge and consist of a product of quark and antiquark fields with phases and offsets appro-
priate to a color octet ps, as described above, and multiplied by the smeared field strength

^ We use the term "taste" to refer to the four types of quarks that are naturally present in the Kogut-
Susskind formulation, while "flavor" can also distinguish quarks with an additional externally imposed
label. For example, a meson with a source operator ® IV' but with disconnected diagrams not included
would be a taste singlet but flavor non-singlet, and would be a pion in the continuum limit.

3

symmetrized with respect to the positions of the quark or antiquark to form the required C
even combination, as illustrated in Fig. ^

The operator is summed over all spatial sites and a trace is taken over the color indices.

The algorithm for constructing the meson propagator starts in Coulomb gauge with a
quark "wall source", consisting of a unit color vector field in a spatially constant direction,
and applies the hybrid meson operator to form a source for the antiquark propagator. The
calculation of the meson propagator is completed by acting upon the resulting antiquark
propagator at an arbitrary time slice by the same hybrid operator and joining the resulting
color vector field with the quark field propagated from the same wall source, summing over
all sink spatial sites and color indices. The whole process is repeated, summing over the
three wall source colors.

III. SIMULATION AND MEASUREMENT

We measured the connected correlator of the 1^^ hybrid state on three sets of 28^ x 96
lattices generated with the "Otad" action [2^. To isolate the effects of dynamical quarks, we
used matched quenched and full QCD lattices with 10/(?^ = 8.40, mvaia = 0.016, 0.04, for the
quenched quarks, 10/(7^ = 7.18 for lattices with three degenerate flavors of dynamical sea
quarks at the strange quark mass (ma = 0.031) and 10/(7^ = 7.11 for lattices with m^^^ =
0.4ms (ma = 0.0124). These choices of 10/(7^ give approximately the same lattice spacing
(~ 0.09 fm) in the three cases. The corresponding choices of quark mass allow simulation at
roughly equivalent values of (mps/mv)^, the square of the ratio of the pseudoscalar to vector
meson masses. Table H] summarizes the simulation parameters and fit results for the l""*"
states, while Table |n] contains estimates for conventional hadron masses at these parameters.

The size of the datasets is comparable for quenched and full QCD runs. Successive full
QCD lattices are separated by six molecular dynamics trajectories, with each trajectory one
simulation time unit long. The full QCD lattices are not completely decorrelated but this
autocorrelation has negligible effects on the hybrid mass fittings, since hybrid propagators
have much larger statistical errors than, e.g. pion propagators. In particular, for the lightest
sea and valence quark mass, amq = 0.0124, we calculated the normalized autocorrelations of

(4)

4

the 1 propagators separated by six simulation time units at each Euchdean time separa-
tion, or distance between the wall source and sink. For propagation distances zero through
eight with the sample of 532 lattices we find 0.01, 0.13, -0.05, -0.00, 0.08, 0.01, -0.04, -0.16
and -0.08 respectively, instead of the uniformly positive autocorrelations that we would see
if the propagators were systematically correlated from one stored lattice to the next. Al-
though the statistical errors we quote come from the covariance matrix of the propagator,
we have also performed a jackknife error analysis of each fitted mass and found jackknife
error estimates to be consistent with errors from the covariance matrix. Varying the block
size from 1 to 10 had no significant effect on the jackknife error.

In a separate study wc have measured propagators of the pion, rho and nucleon. Sta-
tistical errors on these propagators are much smaller than for the 1~^ propagator, so some
effects of autocorrelations can be seen. For the nucleon at mass ami — 0.0124, which we use
for comparison with the hybrid propagators, the data was grouped in blocks of four lattices,
or 24 trajectories, before the covariance matrix was computed. Further blocking does not
significantly increase the error bars. The fact that the nucleon mass fits have good (in
fact, better than the quenched nucleon fits) is also evidence that this blocking has removed
most of the effects of the autocorrelations.

IV. RESULTS

We fit the measured correlators to the sum of oscillating and normal exponentials:

C{t) = Ae-^i-+* + A2(-l)*e-™^* + Asi-lYe-""'', (5)

where Mi-+ is the hybrid meson mass of interest and 1712 and 7713 are masses of non-exotic
parity partner states which have oscillating correlators in the Kogut-Susskind formulation.
In our case the oscillating parity partner is a l"*""*" (oi) state, which is lighter than the
l""^ hybrid, and the oscillating component dominates the correlator at large times. It is
therefore essential to include the oscillating state(s) in our fits. We performed both four
and five parameter fits. For the four parameter fits, we fix A3 = = 0, meaning that we
include one state of each parity. For the five parameter fits we fix m2 to an ai meson mass
determined from propagators with a standard qq source operator, and fit for A2, ma and
A3. We varied the range of the fit and tried to choose values for Mi-+ corresponding to

5

high-confidence fits that were insensitive to -Dmax and Djnm, the hmits of the fit range.

For the quenched lattices we were able to fit the propagators with reasonable confidence
levels (25-50%) for valence quark masses ma = 0.016 and ma = 0.040. Figure |21 shows the
measured propagator for ma = 0.016. Note the oscillating component due to parity partner
states. As expected, the oscillating component dominates at large distance, since the parity
partner has lower mass than the 1 ^. Figure El shows mass fits for the quenched lattices for
ma = 0.040 and -Dmax = 15, with both the two particle (four parameter) and three particle
(five parameter) fits. In the mass fit plots, we have included the small confidence level fits to
illustrate how adjusting the fit range produces more optimal fits. Figure IH shows the same
plot for ma = 0.016. In both plots the three particle fits exhibit a plateau with relatively
small error bars (< 1%), demonstrating the stability of the result with respect to variations
in the fit range. For the four-parameter fits, there is a slight oscillation of fitted values about
the same plateau. Furthermore, the range of fits with high confidence level and relatively
small errors is reduced. From plots like these, we picked a "best fit" , a value that met some
balance of the following criteria: insensitivity to fit range, high confidence level, reasonable
statistical errors. We can see that one might reasonably choose any one of several points
as a "best fit", and the range of resulting Mi-+ values is the basis of our estimate of the
systematic error coming from the presence of higher mass states in the propagators. In all
of these fit summary figures we include unused fits, that do not meet these criteria, say,
because of low confidence level, to help illustrate how we selected the optimal fits.

For lattices with three degenerate sea quarks at m^, we were also able to extract a value
for Mi-+ in reasonable agreement with the quenched result. Four and five parameter fits are
shown in Fig. The fits exhibit larger statistical errors than the quenched lattice fits, and
a slight dependence on range. The mass estimate in Table H] reflects this with significantly
larger statistical and systematic error bars than in the quenched case.

The lattices with ^ = OAmg proved more interesting and difficult. The 1 — l"*""*"
propagator for valence mass amq = 0.0124 for this ensemble is shown in Fig. IHl Fits to the
1 ^ mass for both valence masses are illustrated in Figs. [Z|and|Hl The fitted mass agrees
with those of the quenched and three-flavor results within two standard deviations, but with
larger systematic errors, estimated from the dependence on fit range.

In the case of the light valence quark {ma = 0.0124), we were unable to say much about
the 1 ^ hybrid mass with any confidence. It is apparent from visual examination of the

6

10/^2

-^configs

Range

aMi-+ c.l.

8.40

— 0.040 0.0499(5) 3.730(7)

416

4-15

1.062(12)(20) 0.27

8.40

— 0.016 0.0499(5) 3.730(7)

416

4-15

0.973(26)(20) 0.49

7.18

0.031 0.031 0.0405(7) 3.829(13)

509

5-15

0.986(30) (30) 0.83

7.11

0.0124, 0.031 0.031 0.0424(9) 3.708(14)

526

6-15

0.911(34)(100) 0.25

7.11

0.0124, 0.031 0.0124 0.0424(9) 3.708(14)

526

na

TABLE I: Summary of hybrid meson simulation parameters and results. All lattices have dimen-
sions 28^ X 96. The 1"+ (hybrid) mass fits are all three particle fits. The second error on the
hybrid mass estimates is an estimate of the possible systematic error from our choice of fit range.

10/^2 mscaO m^aia aMps aMy aM^ aM^ec

8.40 — 0.040 0.348 0.523(3) 0.771(2) 0.855(17)

8.40 — 0.016 0.223 0.468(3) 0.633(2) 0.749(18)

7.18 0.031 0.031 0.320 0.478(1) 0.699(1) 0.766(2)

7.11 0.0124, 0.031 0.031 0.326 0.479(2) 0.710(2) na

7.11 0.0124, 0.031 0.0124 0.206 0.414(2) 0.579(3) 0.692(4)

TABLE II: Preliminary values for conventional hadron masses at the hybrid mass simulation pa-
rameters. Statistical errors on the pseudo-scalar meson mass, aMps are smaller than the precision
shown. Pseudo-scalar and vector meson masses for the lO/gf^ = 8.4 quenched points were obtained
from interpolation or extrapolation from results at valence masses 0.015 and 0.030.

propagator (Fig. Ej) that there is a lessening of the overall slope, suggesting that the non-
oscillating piece may not be consistent with a single exponential. Indeed, the fits were very
range dependent. Together these factors indicate the presence of lighter 1^+ states, likely
to be the states of two mesons into which the hybrid can decay. However, with the statistics
available to us, we are unable to get convincing plateaus in the fits with more than one
exponential in the l""*" channel.

We performed a linear extrapolation in quark mass of the quenched results to the physical
value of {m-ps/mvY. Because the calculations at the two quark masses were done on the
same set of quenched configurations, they are highly correlated, and a single elimination
jackknife method was used to estimate the statistical error of the extrapolation.

7

V. DISCUSSION AND CONCLUSIONS

There are several sources of systematic error to be estimated. The largest of these, namely
use of the quenched approximation, is inextricably mixed with the problem of determining
the overall scale, or lattice spacing, so we will discuss these issues together.

The first source of systematic error is due to the possibility of mixing of higher mass
states in the 1"^ propagators. As described above we estimate this by looking at the mass
range one might get by a reasonable variation of the fitting parameters.

We also have effects of finite lattice spacing. We obtained these results on lattices with
a ~ 0.09 fm. For the conventional hadrons, we have masses at both a ~ 0.13 fm and
a ~ 0.09 fm fFigs. 1^ and I10|). Since errors with this action are expected to be order a^g"^,
and the finer lattice spacing is about 1 / \/2 times the coarser lattice spacing, we expect that
the difference between a ~ 0.09 fm and a = masses is comparable to or slightly smaller
than the difference between a ~ 0.13 and 0.09 fm. For the quenched nipl^fa and niisi / ^
we see differences as large as 3% between the two lattice spacings, and a difference of about
2% in the ratio rriN/mp at the light quark mass. Differences are smaller at the heavier mass
— less than 1% in the nucleon to rho mass ratio. Therefore we expect effects of finite lattice
spacing on our results based on hadron mass ratios to be around 1% for strange quarks, and
we will use an estimate of 3% for light quarks.

The finite size of the entire lattice also introduces systematic error. The 28'^ x 96 lattice
corresponds to a box (2.5 fm)^ x 8.6 fm. In one case, three flavor QCD with light quark
mass about 0.2 times the strange quark mass with a lattice spacing of 0.13 fm, we have
calculated light hadron masses both in a 2.5 fm box and on a larger 3.6 fm spatial lattice.
The p, (j) and nucleon masses decrease by a barely significant 0.9(7)%, 0.25(25)% and 0.9(6)%
respectively as the lattice size goes from 2.5 to 3.6 fm. Since these effects are expected to
fall exponentially with lattice size, we can simply take these numbers as an estimate of the
effect of the 2.5 fm box size on the light hadron masses. However, hybrids are expected to
be rather extended objects and may feel the influence of a finite lattice more than smaller
particles, so we will use an estimate of 2% for this systematic error.

The largest systematic errors come from use of the quenched approximation, from the
choice of quantity used to set the lattice scale, and the necessity for an extrapolation to
the physical value of the valence quark mass. These effects are interrelated and so must be

8

discussed together.

The hybrid mass estimates obtained above are in units of the inverse lattice spacing a~^,
so to convert these to physical units we need to know a. The lattice spacing is determined
by calculating some quantity that is known from experiment. In other words, the simula-
tion actually produces the ratio of the hybrid mass to some other dimensionful quantity.
In a simulation with sea quark masses at their physical values, the choice of quantity to
fix the lattice spacing would be just a question of convenience. However, in the quenched
approximation, we will not get the real world values for ratios of masses, so there is an
important choice to be made. Because it is easily measured, and because it does not re-
quire an extrapolation in valence quark masses, the static quark potential is often used to
determine the lattice spacing. In particular, we may use the string tension, y/a ~ 440 MeV,
the coefficient of the linear term in V{r). We might also use tq ~ 0.50 fm or ri ^ 0.34 fm,
which are defined by r^F(ra.) = 1.65 or 1.00 respectively. However, the shape of the static
quark potential in quenched QCD differs from the shape with three dynamical fiavors {2^.
Hybrid mesons are expected to be large hadrons where the quarks are more likely to be
in the linear part of the static quark potential, where a is defined, rather than the region
of crossover between Coulombic and linear behavior, where tq and ri are defined. This
suggests that plotting results in units of the string tension might minimize (although by no
means eliminate!) effects of quenching. This expectation is borne out by calculations of
the conventional hadron spectrum with this same improved action, where using a to define
the lattic e sp acing produces better agreement of the quenched and three fiavor results than

2ll |. Figures El and E3 illustrate this with rho and nucleon masses plotted in units

usmg ri

of rf ^ and y/a respectively. Since one of our important goals is to compare quenched and
three fiavor results, we therefore plot our results in units of the string tension. We also
wish to compare our results with earlier results, and for this purpose the string tension in
other published simulations is either available or can be reasonably estimated. In Fig. ^2
we summarize our results along with the results of previous Wilson quark studies by the
MILC collaboration jl4|, the UKQCD collaboration Q, the SESAM collaboration Q,
as well as recent results from the Zhongshan University group using Wilson quarks on
an anisotropic lattice. We use the string tension a to establish the lattice length scale and
plot Mi~+ / yfa . Our results are consistent with the earlier results at heavier quark masses.
To compare with experiment, we need to convert Mnjyfo to physical units. Unfor-

9

tunately, although phenomenological estimates are available, the string tension is not a
parameter that is well known from experiment. The obvious workaround is to determine
the string tension from the lattice results for nip/y/a etc., which in the end means that
we are using the light hadron spectrum to set the length scale. Since ratios of quenched
hadron masses are not quite those of the real world, we will get different estimates of the
length scale depending on which hadron we choose. For the ss hybrid, the most reason-
able choice for setting the length scale is a hadron with valence quark masses at the same
value — the meson or Vt^ baryon, which means that we are essentially quoting Mh/M^
or Mh/Mq- with the quenched mass and Vt^ masses defined to be 1020 MeV and 1672
MeV. Estimating the masses of the conventional hadrons on our quenched lattices from a
linear extrapolation of results at airiq = 0.015 and 0.030, and setting the quenched string
tension from the (f) or Q~ gives ^/a^ = 436(4) or 437(9) MeV respectively. (This remarkable
agreement is surely coincidence, since other hadron mass ratios on these lattices differ by
much larger amounts from the real world.) To estimate the light quark hybrid mass in
MeV, we might use these estimates of ^/o^, or equally well argue that we should use light
quark hadrons for comparison. Using the linearly extrapolated or interpolated p, K*, N
or A masses to set the scale gives quenched y/a of 389(5), 410(4), 380(5) or 400(21) MeV
respectively, showing statistical errors only. These estimates are in reasonable agreement
with phenomenological estimates from potential models on charmonium and bottomonium
spectroscopy; for example ^/o' = 384 MeV or 427 MeV in Refs. Q and Q respectively.
Thus in estimating light quark hybrid masses in MeV we might consider a range of possible
values for the quenched ^/a from around 380 to 440 MeV.

We begin with estimates for the ss hybrid masses. As mentioned above, it seems most
consistent to use masses of hadrons made from strange quarks to set the lattice spacing in
this case. If we use the meson to set the length scale, using the results in Tables HI and ITTl
we find, with statistical error only

Systematic errors include fit choice, nonzero lattice spacing, finite spatial size, and effects of
quenching. The first three have been discussed above. Effects of quenching can be estimated
in part from the variation of our mass estimates among different ways of fixing the lattice
scale, and in part from differences of other hadronic ratios between full and quenched QCD,

10

as for example in Fig. QHl Here we include what we expect is a fairly conservative 5% error
for this effect, thus estimating

Mh,ss = 2071(26)(39)(1%)(2%)(5%) (7)
= 2071(120) MeV , (8)

where the errors are statistical, fit choice, lattice spacing, box size and quenching respectively.
A similar calculation using the Q~ mass to set the scale gives

/1.062(12)~

Mhss = 1672 ^-4

V0.855(17)^

= 2077(48) (39) (1%) (2%) (5%)

= 2077(129) MeV . (9)

We might also use the mass of a fictional octet baryon made from three quarks with the
mass of the strange quark, assigning it a mass of Mgss = rn^ + ^ (m^ — m^v) = 1507 MeV:

M„„- = 1507 fl™'
V 0.771(2)

= 2075(24) (39) (1%) (2%) (5%)

= 2075(119) MeV . (10)

These three estimates are in remarkably close, and doubtless partly fortuitous, agreement.

Repeating this calculation with the three flavor lattices with mu,d = with the 0, VL~
and sss baryon setting the scale produces

M„,, = 1020 (^^'
V0-4778(9)^

= 2105(64) (64) (1%) (2%) (3%)

= 2105(120) MeV (11)

M„.,- = 1672 (l«)
V0-7659(24) J

= 2152(66) (66) (1%) (2%) (3%)

= 2152(123) MeV (12)

M„„ = 1507 (l™ )

V0-6991(10) )

= 2125(65) (65) (1%) (2%) (3%)

= 2125(121) MeV (13)

11

respectively. Here we have assigned an error of 3% for the partial quenching, or the remaining
extrapolation of the sea quark masses to their physical values. Finally, we made an estimate
of the ss hybrid mass from the run with m„ ^ = 0.4ms. Although the error on this estimate,
mostly coming from the choice of fit range, is too large for it to be very useful, we include
it for completeness.

V0-4792(16) J
= 1939(73) (213) (1%) (2%) (2%)

= 1939(233) MeV . (14)

Since the sea quarks here are much lighter, we used 2% as our estimate of the systematic
error from partial quenching in this number. We can summarize this with an estimate of
2100 ± 120 MeV for the mass of the ss 1~+ hybrid meson.

To estimate the mass of a light quark 1~+ hybrid meson we use the jackknife extrapolation
of the quenched results to (mps/mv)'^ — 0.033, amn — 0.919(39). If we use the to set the
scale, this would correspond to a mass of

^ /0.919(39)\

= 1792(77) (36) (3%) (2%) (5%)

= 1792(139) MeV , (15)

with similar results using the D," or sss baryon. However, if we were to use the smaller
estimates of the string tension obtained from linear extrapolations of light quark hadron
masses to the physical light quark mass, we would obtain smaller values around 1600 MeV. As
discussed above we have assigned a larger 3% systematic error for the effect of nonzero lattice
spacing. We have also assigned a larger 5% error from quenching and chiral extrapolation.
One reason that a larger systematic error is required here is that we are estimating the lattice
spacing in large part from hadrons made up of strange quarks. Our strange quark mass was
fixed by tuning the pseudo-scalar to vector meson mass ratio, and would have come out
slightly different if we had used some other quantity. The effect of uncertainty in fixing the
strange quark mass mostly cancels from mass ratios of hadrons made up of strange quarks,
such as Mh,ss/M^, but will be present when quantities such as M<^ are used in estimating
the mass of light quark hadrons. More evidence that this larger systematic error is required
is seen in the extrapolations of conventional hadron masses to the physical light quark mass.

12

If a naive linear extrapolation is made, and the resulting masses used to set the scale for the
1^+ hybrid mass, the close agreements of the scales from various conventional hadrons that
we found when using hadrons made from strange quarks is no longer present, as seen in the
string tension estimates above.

Given the systematic errors from quenching and chiral extrapolation, our estimate for
the mass of the light quark l""*" meson is not inconsistent with the experimental candidate
7ri(1600). In Fig. [TT]we include the l""*" experimental candidates 7ri(1400) and 7ri(1600) at
the physical value of (mps/my)^ = {mT,/mpY = 0.033. These particles are represented by
vertical bars, offset slightly to the left or right for clarity, representing the range of values
for the quenched string tension from 380 to 440 MeV.

The rriu^d = 0.4ms data illustrates that dynamical quarks introduce new and significant
processes that contribute to the l""*" propagator. On this same set of lattices, mass fits
for stable hadrons, even with ma = 0.0124 valence quarks, display plateaus as functions of
minimum included distance, -Dmin with fixed maximum distance, -Dmax- The plateaus are
similar for quenched and full QCD. In contrast, for the 1 ^, the full QCD fits do not show
even the shorter plateau found in the quenched fits. We illustrate this by comparing fit
plots for quenched and full QCD hybrids and nucleons in Figure IT^ Fit plots for nucleon
and quenched hybrids show a plateau, indicating the propagator has a single exponential
form in the region -Dmin to -Dmax- The full QCD hybrid fit plot deviates from a plateau in a
significant manner — at minimum distance five, in the range which we have generally used
for our quoted mass estimate, the low mass full QCD fits drop to a smaller value. Though
quenching often introduces a systematic effect in the mass, this propagator is different in a
way that suggests mixing of more than one exponential, representing propagators of different
states with J^'-' = 1"+. Our hybrid propagators with light, dynamical quarks show features
that are not evident either in hybrid propagators with heavier or quenched quarks, or in
stable hadron propagators even with light dynamical quarks.

Four-quark states, molecular states of two mesons, or two independent mesons can have
J^*^ = 1 ^ without the gluonic excitations. For example the combination of 6i + vr can
give 1^^ with 1 = 1, and as the sum of these masses is less than the predicted mass of
the lowest 1~^ hybrid, we expect that dynamical quarks introduce the possibility of the
hybrid decaying into this two-meson state. In fact, at the values of the quark masses that
we used the l""*" energies found in our dynamical simulations, while similar to the quenched

13

hybrid masses, are also very close to the expected decay channel masses. For the run with
three degenerate sea quarks at rus, our 1~+ mass is amn = 0.97(3)(3), very close to the
sum of the 'Vand "61" masses: am^ + arrihi = 0.32 + 0.68 = 1.00. For the run with
rriu^d = 0. Arris, we would expect decays into a pseudo-scalar K and a P-wave strange meson
- a Ki. Again, our estimated mass for the ss 1 ^, aruH = 0.90(4)(10), is close to the sum
arriK + arriKi = 0.27 + 0.63 = 0.90.

We now have ahead of us the task of understanding these contributions so that we can
make useful predictions of the 1 ^ hybrid mass in the presence of dynamical quarks. It is
clear from our results with dynamical quarks that it will not be sufficient to simply do the
same analysis that was done on the quenched gauge configurations, simply replacing them
with full QCD configurations. One obvious avenue that may shed some light is to measure
cross-correlators between the px B operator and the two-meson state, as was explored with
Wilson quarks in Ref. 1J|. A more detailed study along these lines in the static quark
(heavy quark) limit has been done by the UKQCD collaboration {2^. It may also be useful
to study the dependence of the exotic energy as a function of valence quark mass (possibly
with fixed sea quark mass) to look for an avoided level crossing as the decay threshold is

crossed, as was done for the non-exotic 0^^ meson in Ref.

2l|.

ACKNOWLEDGEMENTS

Computations for this work were performed at the San Diego Supercomputer Cen-
ter (SDSC), the Pittsburgh Supercomputer Center (PSC), Oak Ridge National Labora-
tory (ORNL) and the National Energy Resources Supercomputer Center (NERSC). This
work was supported by the U.S. Department of Energy under contracts DOE - DE-FG02-
91ER-40628, DOE - DE-FG02-91ER-40661, DOE - DE-FG02-97ER-41022 and DOE - DE-
FG03-95ER-40906 and National Science Foundation grants NSF - PHY99-70701 and NSF
- PHYOO-98395.

[1] D. Aide et al, Phys. Lett. B 205 (1988) 397; D. R. Thompson et ai, Phys. Rev. Lett. 79
(1997) 1630; S.U. Chung et al, Phys. Rev. D 60 (1999) 092001; A. Abele et al, Phys. Lett.
B 423 (1998) 175; A. Abele et al, Phys. Lett. B 446 (1999) 349.

14

[2] G. S. Adams et al, Phys. Rev. Lett. 81 (1998) 5760; E. I. Ivanov et al, Phys. Rev. Lett. 86
(2001) 3977; Y. P. Gouz et al, In *Dallas 1992, Proc, HEP, vol. 1* 572-576.

T. Barnes, F. E. Close and E. S. Swanson, Phys. Rev. D 52, 5242 (1995)
M. Flensburg, C. Peterson and L. Skold, Z. Phys. C 22, 293 (1984)
T. Barnes, F. E. Close and F. deViron, Nucl. Phys. B 224, 241 (1983).
M. Chanowitz and S. R. Sharpe, Nucl. Phys. B 222,211 (1983).
T. Barnes and F. E. Close, Phys. Lett. B 116, 365 (1982).
T. Barnes, Nucl. Phys. B 158, 171 (1979).

K. G. Chetyrkin and S. Narison, Phys. Lett. B 485, 145 (2000).
J. M. Cornwah and S. F. Tuan, Phys. Lett. B 136, 110 (1984).
J. I. Latorre, P. Pascual and S. Narison, Z. Phys. C 34, 347 (1987).
S. Narison, Nucl. Phys. A 675, 54C (2000).

P. Lacock, C. Michael, P. Boyle and P. Rowland, Phys. Lett. B 401 (1997) 308.

C. Bernard et al, Phys. Rev. D 56 (1997) 7039; C. Bernard et al, Nucl. Phys. (Proc. Suppl.)

60A (1998) 61.

[3]
[4]
[5]
[6]
[7]
[8]
[9

[lo:
[11

[12
[13
[14

[15

[16
[17

[18] M. F. Golterman, Nucl. Phys. B 273, 663 (1986).

[19] M. Falcioni, M.L. Paciello, G. Parisi and B. Taglienti, Nucl. Phys. B 251 (1985) 624; M. Al-

banese et al, Phys. Lett. B 192 (1987) 163.
[20] K. Orginos and D. Toussaint, Phys. Rev. D 59 (1999) 014501; K. Orginos, D. Toussaint and

R. L. Sugar, Phys. Rev. D 60 (1999) 054503; K. Orginos, R. Sugar and D. Toussaint, Nucl.

Phys. (Proc. Suppl.) 83 (2000) 878 ; |arXiv:hep-l at799C)9087 . G. P. Lepage, Phys. Rev. D 59

(1999) 074502, |arXiv:hep-lat /9809157| .
[21] C. Bernard et al, Phys. Rev. D 64 (2001) 054506; C. Bernard et al, arXiv:hep-lat /0208041] ,

to appear in Nucl. Phys. B(Proc. Suppl.); C. Bernard et al, in progress.
[22] J.L. Richardson, Phys. Lett. 82B, 272 (1979).
[23] E. Eichten et al, Phys. Rev. D 21, 203 (1980).

[24] C. McNeile, C. Michael and P. Pennanen (UKQCD Collaboration), Phys. Rev. D 65 (2002)

15

Z.H. Mei and X.Q. Luo, |arXiv:hep-la t/0206012 ; to appear in Nucl. Phys. B (Proc.
Suppl.) hep-lat/0209049j.

P. Lacock and K. Schilling (SESAM cohaboration), Nucl. Phys. (Proc. Suppl.) 73 (1999) 261.
C. Bernard et a/., |hep-lat/0209097[ to appear in Nucl. Phys. B(Proc. Suppl.).

1 «^ —

X

x+ ^

/-

x-ti

X

/

— *

/

F(x-

FIG. 1: Chromomagnetic field measured at the site of the antiquark (left) and the quark (right).

10"

10"

10-* r-

10'

<> T I<>T

<><>

10 20
At

30

FIG. 2: Propagator for quenched lattice with W/g'^ = 8.40, ma=0.016. Octagons represent positive
values, diamonds represent negative values.

094505.

16

1.5

1.0

cd

0.5

0.0

O: 2 particle fits
□ : 3 particle fits

Sea

m

valence

Ul

strange

i

J I I L

D

4

mm

6

FIG. 3: aMi-+ vs. D^^^ for 10/g'^ = 8.40 quenched lattices

o^Tivaience — 0.040. The octagons are
four parameter fits, with one mass and ampUtude of each parity, and the squares are five parameter
fits with one 1"^"^ mass fixed to the ai mass, as described in the text. All these fits used a maximum
distance -Dmax = 15. The four parameter fit points are shifted slightly to the right for clarity. The
symbol size is proportional to the confidence level of the fit, with the symbol size in the labels
corresponding to 50%.

17

1.5

1.0

cd

0.5

0.0

IQ []0

O: 2 particle fits
□ : 3 particle fits

m

m

Sea

valence

0.4 m

strange

J I \ I I I I I L

D

4

mm

6

FIG. 4: aMi-+ vs. -Dmin for lO/g'^ = 8.40 quenched lattices with amvaience = 0.016, using
15. Notation is the same as in Fig. |21

18

1.5

1.0

+
I

0.5

"1 1 — I 1 1 1 — I 1 1 — I — I — I 1 1 — I — I 1 r

EI m.

O: 2 particle fits
□ : 3 particle fits

■'^sea ■'^strange
■'^valence ■'^strange

Q Q I I I I I I I I I I I I I I I I I I L

2 4 6

Dmin

FIG. 5: aMi-+ vs. -Dmin for W/g^ = 7.18 with three degenerate dynamical quarks with mass
arusea. = a^^vaience = 0.031, using Dmax = 15. Notation is the same as in Fig. |SI

19

FIG. 6: Propagator for three flavor lattice with 10/^^ = 7.11, ma=0.0124. Octagons represent
positive values, diamonds represent negative values.

20

1.5

1.0

+
I

0.5

"1 1 — I 1 1 1 — I 1 1 — I — I — I 1 r

"1 1 r

O: 2 particle fits
□ : 3 particle fits

^sea ~ ^■'^ ^strange '^strange
^valence ~ ^strange

Q Q I I I I I I I I I I I I I I I I I I L

2 4 6

Dmin

FIG. 7: aMi-+ vs. Dmin for = 7.11 with three dynamical quarks with masses amiight =

0.0124 and amheavy = 0.031. The valence quark mass is arrivaience = 0.031. Notation is the same
as in Fig. |31

21

1.5

1.0

+
I

0.5

"1 1 — I 1 1 1 — I 1 1 — I — I — I 1 1 — I — I 1 r

[]

O: 2 particle fits
□ : 3 particle fits

^sea ~ ^■'^ ^strange '^strange
^valence ~ ^strange

Q Q I I I I I I I I I I I I I I I I I I L

2 4 6

Dmin

FIG. 8: aMi-+ vs. Dmin for = 7.11 with three dynamical quarks with masses amiight =

0.0124 and amheavy = 0.031. The valence quark mass is amvaience = 0.0124. Notation is the same
as in Fig. |31

22

3.0

2.5

m
>

: 2.0

1.5

1.0

1 1 1 1

1 1

III

-

CD

m —
ill

_

_ (D @

m □ ^-v-

□

3 — flavor _

1 1 1 1

O:

2 — flavor

III

0.0

0.2 0.4

(mpg/mv)^

FIG. 9: Vector meson ('V') and octet baryon ("i?") masses in units of ri, which is defined from
the static quark potential by r^F(ri) = 1.0. This graph contains points from quenched simulations
with a ~ 0.13 fm (octagons) and 0.09 fm (crosses), and from simulations with three flavors of
dynamical quarks (two light and one strange quark) at a ~ 0.13 fm (squares) and 0.09 fm (bursts).
The diamond is from a two flavor simulation with a ~ 0.13 fm. Points above the dashed line are
baryon masses, and those below the dashed line vector meson masses.

23

m
>

1 1 1 1

1 1

III

-

□
X °

-

^ X

1 — 1

□ X

Lt_JJJ ' '

m

: quenched

: 3 — flavor -

1 1 1 1

O: 2 — flavor

1 1 1 1 1

0.0

0.2

(mpg/mv)

0.4

FIG. 10: Vector meson ("y") and octet baryon ("i?") masses in units of the square root of the
string tension. The meaning of the symbols is the same as in Fig. |^

24

8

6

o
□

o

X

✓ IN

+

nf=0 imp. stag, Symanzik
nf=3 imp. stag, Symanzik
nf=2+l imp. stag, Symanzik
MILC Conventional Wilson, plaq
MILC Clover Wilson, plaq
UKQCD (Clover Wilson)
ZSU (Clover Wilson, aniso.)
SESAM {n^=2 Wilson)

-^Ti(1600)

77i(l400)

5C

0.0

J I I L

J I I L

J I I L

J I I L

J I I L

0.2

0.4 0.6

(mps/my)^

0.8

1.0

FIG. 11: Summary of 1 hybrid meson mass predictions as a function of (mps/jny)^. The bold
octagon represents the linear extrapolation of = data to (mps/?raT/ )^ = 0.033. The improved

staggered points are from this work, while the earlier data is from Refs (13 ,

14

15

3

25

1.0

;z;

cd

+
I

<^ 0.5

0.0

"1 — I — I — r

©' A]

1 — n — I — r

[]

5K *

"1 — I — I — r

"1 — I — r

O: quenched hybrid
□ : full QCD hybrid

0*W
^1^ ^1' ^1' ^j' ^1' M' M/" SI/

/IN ;k -I- -1- -[- TF^

5 I

O: quenched nucleon
full QCD nucleon

J I I L

J I I L

J I I L

J I L

10

15

D

min

FIG. 12: Hybrid and nucleon mass fits in quenched and full QCD with light dynamical quark
mass amiight ~ 0.4ms . The valence quark mass is about 0.4?7?.5, which, is o^'/^vaience

= 0.016 for the

quenched case and 0.0124 for the three flavor case.

26