Abstract
The overlap operator is a lattice discretization of the Dirac operator of quantum chromodynamics (QCD), the fundamental physical theory of the strong interaction between the quarks. As opposed to other discretizations, it preserves the important physical property of chiral symmetry, at the expense of requiring much more effort when solving systems posed with this operator. We present a preconditioning technique based on another lattice discretization, the Wilson-Dirac operator. The mathematical analysis precisely describes the effect of this preconditioning strategy in the case that the Wilson-Dirac operator is normal. Although this is not exactly the case in realistic settings, we show that current smearing techniques indeed drive the Wilson-Dirac operator towards normality, thus providing motivation for why our preconditioner works well in practice. Results of numerical experiments in physically relevant settings show that our preconditioning yields accelerations of more than an order of magnitude compared to unpreconditioned solvers.
Similar content being viewed by others
Notes
Physical space-time is a four-dimensional Minkowski space. We present the theory in Euclidean space-time since this version can be treated numerically. The two versions are equivalent, cf. [33].
For simplicity, we consider here the “naive” limit \(a \rightarrow 0\). In the full quantum theory one has \(\mathcal {D}_{\mathcal {L}}= D_W(m_0(a)) + \mathcal {O}(a)\) with the mass \(m_0(a)\) of order \(1/\log (a)\); see [33].
The domain wall discretization satisfies (8) approximately and, hence, has also been the focus of extensive research.
To represent a physically meaningful quantity, the Wilson gauge action is usually scaled. This is not relevant in the present context.
In all our experiments, setup never exceeded 2 % of the total execution time, so we do not report timings for it.
References
Albanese, M., Costantini, F., Fiorentini, G., Flore, F., Lombardo, M.P., Tripiccione, P.B.R., Fonti, L., Remiddi, E., Bernaschi, M., Cabibbo, N., Fernandez, L.A., Marinari, E., Parisi, G., Salina, G., Cabasino, S., Marzano, F., Paolucci, P., Petrarca, S., Rapuano, F., Marchesini, P., Giacomelli, P., Rusack, R.: Glueball masses and string tension in lattice QCD. Phys. Lett. B 192, 163–169 (1987)
Arnold, D.N.: Stability, consistency and convergence of numerical discretizations. In: Enquist, B. (ed.) Encyclopedia of Applied and Computational Mathematics. Springer (2016). preprint available under http://www.ima.umn.edu/~arnold//papers/stability.pdf
Babich, R., Brannick, J., Brower, R.C., Clark, M.A., Manteuffel, T.A., McCormick, S.F., Osborn, J.C., Rebbi, C.: Adaptive multigrid algorithm for the lattice Wilson-Dirac operator. Phys. Rev. Lett. 105, 201602 (2010)
Bergrath, T., Ramalho, M., Kenway, R. et al.: PRACE scientific annual report 2012, tech. rep., PRACE (2012). http://www.prace-ri.eu/IMG/pdf/PRACE_Scientific_Annual_Report_2012.pdf, p. 32
Berruto, F., Narayanan, R., Neuberger, H.: Exact local fermionic zero modes. Phys. Lett. B 489, 243–250 (2000)
Bonati, C., D’Elia, M.: A comparison of the gradient flow with cooling in SU(3) pure gauge theory. Phys. Rev. D 89, 105005 (2014)
Borsanyi, S., Delgado, Y., Durr, S., Fodor, Z., Katz, S.D., Krieg, S., Lippert, T., Nogradi, D., Szabo, K.K.: QCD thermodynamics with dynamical overlap fermions. Phys. Lett. B 713, 342–346 (2012)
Brannick, J., Brower, R.C., Clark, M.A., Osborn, J.C., Rebbi, C.: Adaptive multigrid algorithm for lattice QCD. Phys. Rev. Lett. 100, 041601 (2007)
Capitani, S., Durr, S., Hoelbling, C.: Rationale for UV-filtered clover fermions. JHEP 0611(2006), 028 (2006)
Cundy, N., van den Eshof, J., Frommer, A., Krieg, S., Schäfer, K.: Numerical methods for the QCD overlap operator. III: Nested iterations. Comput. Phys. Commun. 165, 221–242 (2005)
de Forcrand, P., Kurkela, A., Panero, M.: Numerical properties of staggered overlap fermions, PoS, LATTICE2010:080 (2010)
Del Debbio, L., Giusti, L., Lüscher, M., Petronzio, R., Tantalo, N.: QCD with light Wilson quarks on fine lattices (i): First experiences and physics results. JHEP 02(2007), 056 (2007)
Del Debbio, L., Giusti, L., Lüscher, M., Petronzio, R., Tantalo, N.: QCD with light Wilson quarks on fine lattices (ii): DD-HMC simulations and data analysis. JHEP 0702(2007), 082 (2007)
Dürr, S., Fodor, Z., Frison, J., Hoelbling, C., Hoffmann, R., Katz, S.D., Krieg, S., Kurth, T., Lellouch, L., Lippert, T., Szabo, K., Vulvert, G.: Ab initio determination of light hadron masses. Science 322, 1224–1227 (2008)
Edwards, R.G., Heller, U.M., Narayanan, R.: A study of practical implementations of the overlap Dirac operator in four-dimensions. Nucl. Phys. B 540, 457–471 (1999)
Frommer, A., Güttel, S., Schweitzer, M.: Convergence of restarted Krylov subspace methods for Stieltjes functions of matrices. SIAM J. Matrix Anal. Appl. 35, 1602–1624 (2014)
Frommer, A., Güttel, S., Schweitzer, M.: Efficient and stable Arnoldi restarts for matrix functions based on quadrature. SIAM J. Matrix Anal. Appl. 35, 661–683 (2014)
Frommer, A., Kahl, K., Krieg, S., Leder, B., Rottmann, M.: Adaptive aggregation based domain decomposition multigrid for the lattice Wilson Dirac operator. SIAM J. Sci. Comp. 36, A1581–A1608 (2014)
Ginsparg, P.H., Wilson, K.G.: A remnant of chiral symmetry on the lattice. Phys. Rev. D 25, 2649–2657 (1982)
Giusti, L., Hoelbling, C., Lüscher, M., Wittig, H.: Numerical techniques for lattice QCD in the epsilon regime. Comput. Phys. Commun. 153, 31–51 (2003)
Golub, G.H., Loan, C.F.V.: Matrix Computations, 2nd edition. The Johns Hopkins University Press, Baltimore, MD, USA (1989)
Guest, M., Aloisio, G., Kenway, R. et al.: The scientific case for HPC in Europe 2012–2020, tech. rep., PRACE, Oct (2012). http://www.prace-ri.eu/PRACE-The-Scientific-Case-for-HPC, p. 75
Hairer, E., Lubich, C., Wanner, G.: Geometric Numerical Integration. Springer Series in Computational Mathematics, vol. 31. Springer, Heidelberg (2010)
Hasenfratz, A., Hoffmann, R., Schaefer, S.: Hypercubic smeared links for dynamical fermions. JHEP 0705(2007), 029 (2007)
Hasenfratz, A., Hoffmann, R., Schaefer, S.: Localized eigenmodes of the overlap operator and their impact on the eigenvalue distribution. JHEP 0711(2007), 071 (2007)
Intel, Intel xeon x5570 (nehalem-ep) quad-core processors. http://ark.intel.com/products/37111/Intel-Xeon-Processor-X5570-8M-Cache-2_93-GHz-6_40-GTs-Intel-QPI
Jülich Supercomputing Centre, Juropa - Jülich research on petaflop architectures. http://www.fz-juelich.de/ias/jsc/EN/Expertise/Supercomputers/JUROPA/JUROPA_node.html
Kaplan, D.B.: A method for simulating chiral fermions on the lattice. Phys. Lett. B 288, 342–347 (1992)
Lüscher, M.: Exact chiral symmetry on the lattice and the Ginsparg-Wilson relation. Phys. Lett. B 428, 342–345 (1998)
Lüscher, M.: Local coherence and deflation of the low quark modes in lattice QCD. JHEP 07(2007), 081 (2007)
Lüscher, M.: Properties and uses of the Wilson flow in lattice QCD. JHEP 1008(2010), 071 (2010)
Lüscher, M.: Trivializing maps, the Wilson flow and the HMC algorithm. Commun. Math. Phys. 293, 899–919 (2010)
Montvay, I., Münster, G.: Quantum Fields on a Lattice. Cambridge University Press, Cambridge Monographs on Mathematical Physics (1994)
Morningstar, C., Peardon, M.J.: Analytic smearing of SU(3) link variables in lattice QCD. Phys. Rev. D 69, 054501 (2004)
Negele, J.W.: Instantons, the QCD vacuum, and hadronic physics. Nucl. Phys. Proc. Suppl. 73, 92–104 (1999)
Nepomnyaschikh, S.: Mesh theorems on traces, normalizations of function traces and their inversion. Soviet J. Numer. Anal. Math. Model. 6, 223–242 (1991)
Neuberger, H.: Exactly massless quarks on the lattice. Phys. Lett. B 417, 141–144 (1998)
Neuberger, H.: Bounds on the Wilson Dirac operator. Phys. Rev. D 61, 085015 (2000)
Niedermayer, F.: Exact chiral symmetry, topological charge and related topics. Nucl. Phys. Proc. Suppl. 73, 105–119 (1999)
Osborn, J.C., Babich, R., Brannick, J., Brower, R.C., Clark, M.A., Cohen, S.D., Rebbi, C.: Multigrid solver for clover fermions, PoS, LATTICE2010:037 (2010)
Oswald, P.: Preconditioners for nonconforming discretizations. Math. Comp. 65, 923–941 (1996)
Saad, Y.: Iterative Methods for Sparse Linear Systems, 2nd edn. SIAM, Philadelphia, PA, USA (2003)
Simoncini, V., Szyld, D.B.: Theory of inexact Krylov subspace methods and applications to scientific computing. SIAM J. Sci. Comput. 25, 454–477 (2003)
Smith, G.: Numerical Solution of Partial Differential Equations: Finite Difference Methods. Clarendon Press, Oxford Applied Mathematics and Computing Science Series (1985)
Sorensen, D., Lehoucq, R., Yang, C., Maschhoff, K.: PARPACK. http://www.caam.rice.edu/software/ARPACK, used version: 2.1, Sept (1996)
Toth, B.: QCD thermodynamics with dynamical overlap fermions, PoS, LATTICE2013:163 (2013)
van den Eshof, J., Frommer, A., Lippert, T., Schilling, K., van der Vorst, H.A.: Numerical methods for the QCD overlap operator. I: Sign-function and error bounds. Comput. Phys. Commun. 146, 203–224 (2002)
van den Eshof, J., Sleijpen, G.L.: Inexact Krylov subspace methods for linear systems. SIAM J. Matrix Anal. Appl. 26, 125–153 (2004)
Xu, J.: The auxiliary space method and optimal multigrid preconditioning techniques for unstructured grids. Computing 56, 215–235 (1996)
Acknowledgments
We thank the Budapest-Marseille-Wuppertal collaboration and Jakob Finkenrath for providing configurations. All numerical results were obtained on Juropa at Jülich Supercomputing Centre (JSC) through NIC grant HWU12.
Author information
Authors and Affiliations
Corresponding author
Additional information
This work was partially funded by the Deutsche Forschungsgemeinschaft (DFG) Transregional Collaborative Research Centre 55 (SFB/TRR55) and by the National Science Foundation under grant DMS:1320608.
Apppendix: The entries of \(D_W^HD_W-D_WD_W^H\)
Apppendix: The entries of \(D_W^HD_W-D_WD_W^H\)
To prove Proposition 2, we inspect the entries of \(D_W^HD_W-D_WD_W^H\). We use the notation \(\pi _\mu ^{\pm }\) for the matrices
The relations (3) between the \(\gamma \)-matrices show that each \(\pi _\mu ^\pm \) is a projection and that, in addition,
Considering all 12 variables at each lattice site as an entity, the graph associated with the nearest neighbor coupling represented by the matrix \(D_W\) is the 4d-torus, and similarly for \(D_W^H\). Table 4 repeats (6), giving the nonzero entries of a (block) row in \(D_W\) and \(D_W^H\) in terms of the \(12 \times 12\) matrices that couple lattice site x with sites x and \(x\pm \hat{\mu }\). We use m to denote \(m_0+4\), with \(m_0\) from (5).
The product \(D_W^HD_W\) represents a coupling between nearest and next-to-nearest lattice sites; the coupling \(12 \times 12\) matrices are obtained as the sum over all paths of length two on the torus of the product of the respective coupling matrices in \(D_W^H\) and \(D_W\). A similar observation holds for \(D_WD_W^H\). Table 5 reports all the entries of \(D_W^HD_W\), and we now shortly discuss all the paths of length 2 that contribute to these entries of \(D_W^HD_W\).
For the diagonal position (x, x), we have 9 paths of length 2, \((x,x) \rightarrow (x,x) \rightarrow (x,x)\) and \((x,x) \rightarrow (x,x\pm \hat{\mu }) \rightarrow (x,x), \mu = 0,\ldots ,3\). For a nearest neighbor \((x,x + \hat{\mu })\), we have the two paths \((x,x) \rightarrow (x,x) \rightarrow (x,x + \hat{\mu })\) and \((x,x) \rightarrow (x,x+\hat{\mu }) \rightarrow (x,x+\hat{\mu })\) and similarly in the negative directions. For a position \((x,x \pm 2\hat{\mu })\), there is only one path, \((x,x) \rightarrow (x,x \pm \hat{\mu }) \rightarrow (x,x \pm 2\hat{\mu })\), with the product of the couplings being 0 due to (28). Finally, for the other next-to-nearest neighbors, we always have two paths; for example, \((x,x) \rightarrow (x,x+\hat{\mu }) \rightarrow (x + \hat{\mu } - \hat{\nu })\) and \((x,x) \rightarrow (x,x-\hat{\nu }) \rightarrow (x + \hat{\mu } - \hat{\nu })\).
The coupling terms in \(D_WD_W^H\) are identical to those for \(D_W^HD_W\) except that we have to interchange all \(\pi _\mu ^+\) and \(\pi _\mu ^-\) as well as all \(\pi _\nu ^+\) and \(\pi _\nu ^-\).
This shows that, in \(D_W^HD_W-D_WD_W^H\), the only non-vanishing coupling terms are those at positions \((x,x+\hat{\mu } + \hat{\nu })\), \((x,x+\hat{\mu } - \hat{\nu })\) and \((x,x-\hat{\mu } - \hat{\nu })\) for \(\mu \ne \nu \). They are given in Table 6, where we used the identities
By rearranging the terms, we obtain the plaquettes from (20) and (21). We exemplify this for position \((x,x+\hat{\mu } + \hat{\nu })\):
Using the fact that, for the Frobenius norm, we have
we obtain the following for the squares of the Frobenius norms of all the coupling matrices from Table 6:
Finally, for any unitary matrix Q, we have
We now obtain \(\Vert D_W^HD_W-D_WD_W^H\Vert _F^2\) by summing the squares of the Frobenius norms of all coupling matrices. This is a sum over 24n coupling matrices, n being the number of lattice sites. As discussed before, groups of four of these coupling matrices refer to the same plaquette \(Q_x^{\mu ,\nu }\) up to conjugation in SU(3), so \(\mathrm {tr}\left( I-Q\right) \) is the same for these four plaquettes Q. We can thus “normalize” to only consider all possible “first quadrant” plaquettes \(Q_x^{\mu ,\nu }\) and obtain
Rights and permissions
About this article
Cite this article
Brannick, J., Frommer, A., Kahl, K. et al. Multigrid preconditioning for the overlap operator in lattice QCD. Numer. Math. 132, 463–490 (2016). https://doi.org/10.1007/s00211-015-0725-6
Received:
Revised:
Published:
Issue Date:
DOI: https://doi.org/10.1007/s00211-015-0725-6