Magnetic Monopoles As Agents of Chiral Symmetry Breaking in U (1) Lattice Gauge Theory
Magnetic Monopoles As Agents of Chiral Symmetry Breaking in U (1) Lattice Gauge Theory
Magnetic Monopoles As Agents of Chiral Symmetry Breaking in U (1) Lattice Gauge Theory
Magnetic Monopoles as Agents of Chiral Symmetry Breaking in U(1) Lattice Gauge Theory
Tom Bielefeld
Institut fr Theoretische Physik, Universitt Bremen, u a Postfach 330 440, D-28 334 Bremen, Germany
Simon Hands
Department of Physics, University of Wales Swansea, Singleton Park, Swansea, SA2 8PP, U.K.
John D. Stack
Department of Physics, University of Illinois at Urbana-Champaign 1110 West Green Street, Urbana, IL 61801-3080, U.S.A.
Roy J. Wensley
Department of Physics and Astronomy, Saint Marys College, Moraga, CA 94575, U.S.A.
Abstract
We present results suggesting that magnetic monopoles can account for chiral symmetry breaking in abelian gauge theory. Full U (1) congurations from a lattice simulation are factorized into magnetic monopole and photon contribu tions. The expectation is computed using the monopole congurations and compared to results for the full U (1) congurations. It is shown that excellent agreement between the two values of is obtained if the eect of photons, which dress the composite operator , is included. This can be estimated independently by measurements of the physical fermion mass in the photon background. 11.15.Ha, 11.30.Qc, 11.30.Rd Keywords: chiral symmetry breaking, magnetic monopoles
U(1) lattice gauge theory provides an ideal laboratory for studying the eects of monopoles. In U(1), all non-perturbative eects are known to be caused by monopoles [1]. Further, there is an ecient procedure for identifying the magnetic current of monopoles in the link angle congurations generated in a U(1) lattice gauge theory simulation [2]. Previously, two of the present authors have studied the heavy quark potential in U(1) lattice gauge theory, and shown that in the conned phase, the value of the string tension calculated from monopoles agrees quantitatively with the full U(1) string tension calculated directly from link angles [3]. In this paper, we turn our attention to the quenched chiral condensate in U(1) lattice gauge theory, . We expect (and nd) that this has a non-zero value in the conned phase, and vanishes in the deconned phase. Our interest is in how the value of calculated from the link angle congurations compares with the same quantity calculated from monopoles. Ascribing non-perturbative eects to monopoles would say that in the conned phase, a non-vanishing condensate will be induced by monopoles. Nevertheless, since a matrix element of a product of eld operators is involved, this value should be renormalized by short distance perturbative eects i.e. photons, so to make a quantitative comparison, we also need to calculate this renormalization. Our method for doing so, described in more detail below, is to compute the ratio of renormalized to bare masses for a charge moving solely in the eld of photons. Our principle result is that applying this renormalization factor to the monopole value of yields the full U(1) answer, to within statistical errors. Let us now briey describe the factorization of the link variables into photon and monopole parts. We start by resolving the U(1) plaquette angles into uctuating and monopole parts, = + 2m , where (, +]. From the monopole term m , we can dene two currents, one electric, the other magnetic. The electric current is dened by j = m . The magnetic current of monopoles is dened by m = + m , where m is the dual of m . The currents j (m ) reside on direct(dual) lattices. We will compute the chiral condensate using the Banks-Casher formula [4] (see below) which involves the density of eigenvalues of the Dirac operator. Since the Dirac particle is electrically charged, the eect of monopoles on it must be represented by an electric vector potential. This four-vector potential derives from the electric current j , and is analogous to the familiar three-vector potential A used for the case of a static monopole in the continuum. We have Amon (x) g
y
(1)
where g = 2/e is the magnetic unit of charge, and v the lattice Coulomb propagator in mon Feynman gauge, satisfying + v(x y) = x,y . By dening U (x) exp(ieAmon (x)), we dene the photon link by demanding that {U } be factorized on each link:
mon phot U (x) U (x)U (x).
(2)
We can now examine the chiral condensate separately on congurations {U }mon and {U }phot , as well as the full {U }. In the quenched approximation can be dened in the chiral limit via the eigenvalues of the Dirac operator [4]: 2
= ( = 0), V
(3)
where V is the lattice volume and ()d is the number of eigenvalues satisfying D|n = / in |n in the interval (, + d), with D[U ] the staggered lattice fermion kinetic operator / Dxy [U ] = / 1 2
(1)x1 ++x1 U (x)y,x+ U (y)y,x .
(4)
For nite V , (0) vanishes; therefore an extrapolation to the 0 limit must be made from the spectrum density at non-zero . In practice () can be estimated by measuring the lowest O(25) eigenvalues of D per conguration using the Lanczos algorithm, and then / using a binning procedure [5]. Calculations of the chiral condensate were performed with U(1) congurations generated using the standard Wilson action on a 124 lattice. Three dierent values of the inverse coupling were chosen: = 1.005, 1.010, and 1.020. For each value of , 5000 lattice updates were performed before beginning measurements to allow for equilibration. The next 4000 congurations were used for measurements skipping every 20 lattice updates to reduce correlations. For values 1.010, our results exhibited the breaking of chiral symmetry (i. e. = 0). These two values of are known to correspond to the connement phase as well [6]. For = 1.020 the congurations were found to be in the chirally symmetric phase, which is coincident with the Coulomb phase. Our results for the chiral condensate calculations for the chirally broken phase are presented in Figs. (1) and (2). We show both the full U(1) calculations and the results from the monopole gauge eld congurations obtained using Eq. (1). Using a linear t and Eq. (3), the value of the chiral condensate was extracted from the 10 lowest values of (). The values for are shown in Table I. In Fig. (3) the results for the chiral condensate from full U(1) elds and monopoles are shown for = 1.020. For comparison, the results from the = 1.010 monopole congurations are also included. Linear ts to () yield a very small, but nite intercept. However, the values are found to be about a factor of 100 times smaller than the results in the broken phase. Thus, we are condent of being in the chirally symmetric phase for = 1.020. For completeness, the = 1.020 results are also included in Table I. In Fig. (4) we show the eigenvalue spectrum calculated using the {U }phot background from both broken and symmetric phases. There is no signal of chiral symmetry breaking (note that we choose anti-periodic boundary conditions for the fermions in the temporal direction to avoid zero modes from near-plane wave solutions). This supports our conjecture that chiral symmetry breaking in this model can be ascribed entirely to monopoles. Is it possible to account for the mismatch between U (1) and mon ? Since is a composite eld operator, we expect it to be modied by quantum corrections independently of whether it acquires a vacuum expectation value. In perturbation theory, UV uctuations in general result in the requirement for all eld operators to be renormalized, and for composite operators to have an additional renormalization. This consideration leads us to the following hypothesis: the mismatch between the two condensate measurements is due to the rescaling of the operator by the uctuations of the gauge elds contained in the {U }phot congurations; more concisely 3
mon U (1)
= ()tree = ()phot
mon mon
(5)
where ()phot is the dressed operator which takes into account photon-like uctuations, and the expectation value is taken in the monopole-only backgrounds. If the hypothesis is correct, then we can write ()phot = Z()tree , (6)
ie. the tree-level operator is multiplicatively enhanced by the photon uctuations, the condensate data suggesting that the factor Z 1.5. In support, we now describe an alternative and independent determination of Z, obtained by measuring the physical fermion mass mR in the {U }phot background. On the assumption that the uctuations in {U }phot are approximately Gaussian, then the dressing of the operator is given by a set of Feynman diagrams. Let (p) denote the complete set of 1PI diagrams describing corrections to the fermion two-point function. Then for the dressed fermion propagator we have
phot SF (p) =
1 ip + m0 (p) /
Z2 (a, ) , ip + mR /
(7)
where m0 , mR denote bare and physical fermion masses respectively, Z2 is a wavefunction rescaling which by analogy with perturbative QED we expect to be both gauge and cutodependent, and the symbol shows that the second equality holds only in the neighborhood of some subtraction point ip = . Equation (7) may be rearranged to read /
1 1 () = (1 Z2 Zm )m0 + (1 Z2 )
(8)
with mR = Zm (a, )m0 . Now, using the identity d 1 1 = , dm0 ip + m0 / (ip + m0 )(ip + m0 ) / / (10) (9)
we see that the operation of dierentiating with respect to m0 is equivalent to a zero mo mentum insertion of a operator in a Feynman diagram. Therefore d/dm0 is the set of 1PI diagrams, having one external and one external , which describe corrections to ie. , ()phot = Z2 1 ie.
1 ZZ2 = 1
(11)
d , dm0
(12)
where Z(a, ) is the rescaling factor we seek, and the factor Z2 is included because in the diagrammatic approach matrix elements are evaluated with dressed fermion propagators on the external legs. Combining (8) with (12), we nd Z = Zm = mR . m0 (13)
Hence a measurement of mR in the {U }phot background as a function of m0 gives an independent estimate of Z. This argument is essentially the same as that used to establish Z = m0 /mR in standard renormalized perturbation theory, where ()R = Z () is the renormalized operator whose Green functions with other renormalized elds and operators are nite [7]. Here we make no attempt to dene a renormalized operator, since bare quantities are evaluated in a lattice simulation, but rather use the same formalism to quantify the eects of operator enhancement by quantum uctuations. Note that we have not specied the subtraction point dening Z very precisely, and have assumed that Z is independent of m0 . From experience in perturbation theory, we expect that numerically the most signicant contribution to Z, of O(ln a), comes from short wavelength uctuations, and that Z is relatively insensitive to the details of the subtraction. To measure the physical fermion mass in the {U }phot background, we performed calculations of the fermion propagator starting with the same congurations used to compute (). Since the fermion propagator is not gauge invariant, it is rst necessary to x a gauge. Although a gauge transformation of {U }phot is not strictly a symmetry of the full theory, this procedure is justied since , and by hypothesis mR , are gauge invariant. In this work we have used Landau gauge [8] and extracted the lattice fermion timeslice propagator
phot
Cf (x4 ) = Re
x1 ,x2 ,x3 even
(D[U ] + m0 )1 / 0,x
(14)
using a conjugate gradient routine, for bare mass values m0 a = 0.1, 0.09, . . . , 0.04. The restriction to spatial sites an even number of lattice spacings from the origin in each direction improves the signal [9]. Note that we have not attempted to x the residual gauge freedom as in [9], but have instead relied on the uctuations in {U }phot being intrinsically small (see below). The physical mass mR can now be estimated by tting Cf to the following functional form: Cf (x4 ) = A (exp(mR x4 ) + (1)x4 exp(mR (L x4 ))) , (15)
where L is the lattice size in the time direction. We used 100 congurations from each of the three values previously studied. In order to take into account correlations between renormalized mass estimates due to using the same congurations for dierent bare masses we applied a bootstrap tting routine. To determine Z we plotted the mass obtained from Cf against the input bare mass, and tted the results to a linear form: mR (m0 ) = Z m0 + b. (16)
Fig. (5) shows a graph of mR (m0 ) from our simulation for = 1.010. The graphs for the other two values of are very similar. The results for Z and b from the linear ts for each 5
value of are shown in Table II. From the table it can be observed that the value of Z is a smooth function through the phase transition. Unexpected in our results were the small non-zero values of the intercept. This will be discussed briey below. Finally, we compare our value of Z calculated using {U }phot with the value observed in the chiral condensate calculations. We dene the renormalization factor from the chiral condensate to be Z =
U (1) mon
(17)
and present the comparison in Table III. The two independent determinations of the renormalization factor show excellent agreement. This answers the question asked earlier: Is it possible to account for the mismatch between U (1) and mon ? Our results demonstrate that the mismatch is simply operator renormalization due to the photons. As yet we have no satisfactory explanation for the non-zero intercepts in Table II, implying a small breaking of chiral symmetry in the photon-only congurations. It could be either a nite volume eect, or perhaps a residual gauge freedom associated with Gribov copies and the consequent non-uniqueness of the Landau gauge, or perhaps due to spatially constant gauge eld modes. The only eect that we have investigated quantitatively is that of constant modes. We repeated our previous calculations at = 1.005, this time performing a global gauge transformation to remove constant modes from the gauge elds. This was done in the following way: The full gauge elds were shifted by a direction dependent constant, A (x) = A (x) + c , (18)
where c is chosen so x A (x) = 0. The new gauge elds A were used to recompute Z and the chiral condensates. In repeating the calculations with this gauge xing we observed no measurable dierence in the results. Although we are not able to explain the small non-zero intercepts, the overall agreement of the slope with the value of Z required to explain the condensate results is extremely satisfying, and supports our hypothesis on the role of the residual photon uctuations enhancing the operator. An obvious direction in which to extend this analysis is the exploration of chiral symmetry breaking in non-abelian gauge theories following the identication of monopole networks after abelian projection. Studies of this kind have already appeared [10]; it is interesting to note that in the maximal abelian gauge used, there is a similar mismatch in the measured values of between full and monopole-only congurations. It would be interesting to check whether this eect could be accounted for by the operator renormalization described here. Finally note that even though this argument assumed a split of the background gauge eld into Gaussian uctuations, described by Feynman diagrams, and non-perturbative monopole-only congurations, the idea of classifying eects into those which rescale a local operator and those which lead to a non-vanishing expectation value for the operator may be generalized. For instance, it may in principle be possible to extend the analysis by factoring congurations in a scale dependent fashion, including only large monopole loops in the monopole part, and seeing if small monopole loops simply renormalize by comparing 6
the Z factors measured in two dierent ways. This may prove to be an eective probe of the scale at which non-perturbative eects manifest themselves. The scale dependence of monopole contributions to the string tension was studied in [11].
ACKNOWLEDGEMENTS
SJH is supported by a PPARC Advanced Fellowship. The work of JDS and RJW was supported in part by the U.S. National Science Foundation under grants PHY-9412556 and PHY-9403869. JDS and RJW also thank the Higher Education Funding Council for Wales for support in the early stages of this work. We thank John Mehegan for help in analyzing the fermion propagator.
REFERENCES
[1] [2] [3] [4] [5] T. Banks, R. Myerson, and J.B. Kogut, Nucl. Phys. B129 (1977) 493. T.A. DeGrand and D. Toussaint, Phys. Rev. D22 (1980) 2478. J.D. Stack and R.J. Wensley, Nucl. Phys. B371 (1992) 597. T. Banks and A. Casher, Nucl. Phys. B169 (1980) 103. I.M. Barbour, P. Gibbs, J.P. Gilchrist, H. Schneider and M. Teper, Phys. Lett. 136B (1984) 80; I.M. Barbour, P. Gibbs, K.C. Bowler and D. Roweth, Phys. Lett. 158B (1985) 61; S.J. Hands and M. Teper, Nucl. Phys. B347 (1990) 819; G. Salina and A. Vladikas, Nucl. Phys. B348 (1991) 210. R. Gupta, M. Novotny, and R. Cordery, Phys. Lett. B 172 (1986) 86. J.C. Collins, Renormalization, ch. 6 (Cambridge University Press) (1984). P. Coddington, A. Hey, J. Mandula, and M. Ogilvie, Phys. Lett. B 197 (1987) 197. M. Gckeler, R. Horsley, P.E.L. Rakow, G. Schierholz and R. Sommer, Nucl. Phys. o B371 (1992) 713. F.X. Lee, R.M. Woloshyn and H.D. Trottier, Phys. Rev. D53 (1996) 1532; R.J. Wensley, Nucl. Phys. Proc. Suppl. 53 (1997) 538. J.D. Stack and R.J. Wensley, Phys. Rev. Lett. 72 (1994) 21.
TABLES
TABLE I. The values of = 1.005 1.010 1.020
V (0)
U (1) 0.165(3) 0.138(2) 0.002(1) TABLE II. Results of the ts mR (m0 ) (cf. (16))
TABLE III. Comparison of our two determinations of the renormalization factor Z; The column labeled Zm comes from the photon congurations via Eq. (16), and the column labeled Z comes from the chiral condensates via Eq. (17) 1.005 1.010 Zm 1.570(4) 1.561(5) Z 1.570(30) 1.550(40)
FIGURES
0.06
0.04 () 0.02
0 0 0.01
FIG. 1. The spectral density function for = 1.005 from full U (1) elds (solid squares) and monopoles (open squares).
0.02
0.03
10
0.06
0.04 () 0.02
0 0 0.01
FIG. 2. The spectral density function for = 1.010 from full U (1) elds (solid squares) and monopoles (open squares).
0.02
0.03
11
0.04
()
0.02
0 0
FIG. 3. The spectral density function for = 1.020 from full U (1) elds (solid squares) and monopoles (open squares). For comparison the results from monopoles for = 1.010 (open circles) are also plotted.
0.01
0.02
12
0.02
()
0.01
0 0.05
FIG. 4. The spectral density function from the photon background {U }phot for = 1.005 (solid squares) and = 1.020 (open squares).
0.1
0.15
13
0.2
0.15
mR
0.1
0.05
0 0 0.05
FIG. 5. The renormalized mass as a function of input mass for = 1.010 calculated using the photon propagator as dened in Eqs. (14) and (15).
m0
0.1
0.15
14