1. Dynamical Anisotropic-Clover Lattice Production for Hadronic Physics
J. Bulava, J. Foley, C. Morningstar, CMU
K. Orginos, College W&M
S. Cohen, J. Dudek, R. Edwards, B. Joo,
H.-W. Lin, D. Richards, C. Thomas, A. Thomas, JLab
S. Wallace, UMD
J. Juge, U. of Pacific
N. Mathur, Tata Institute
M. Peardon, S. Ryan, Trinity College

2. Physics Focus New generation of experiments: hadron spectroscopy
GlueX (JLab/HallD)
Panda (GSI/Fair)
BES III (Beijing)
Spectrum: (Anisotropic-Clover fermions)
Excited state baryon resonances (Hall B)
Conventional and exotic (hybrid) mesons (Hall D)
Ground and excited state form-factors (Hall B)
Critical need:
Hybrid meson spectrum/photo-coupling
Baryon spectrum

3. Anisotropic Actions
Anisotropic Symanzik gauge action: anisotropy =as/at
Bare gauge anisotropy °g
Anisotropic Clover: 3d-Stout-link smeared U’s (spatial).
Choose rs=1. No doublers
Bare fermion anisotropy °f
Clover coefficients ct and cs (Manke):
Consistent with non-perturbative improvement!

4. Anisotropy Tuning Results
Anisotropies
Nf=2+1, =3.5, as=0.1227(8)fm, at= (23)fm
Tuning results: arxiv:
Scale setting: determine as and ms in arxiv: gg gf
0 MeV
175 MeV
0 MeV
175 MeV

5. Nf=2+1 Anisotropic Clover: dynamical generation
Scaling based on actual (243) runs down to ~230 MeV
The entries are #traj, location[M-core-hours](m L)
Goal is ~ 11k traj
Blue uses NSF resources
MeV only a contingency – not proposed
323 proposed for USQCD, request available INCITE/ESP at ORNL,ANL, possibly multi-year.
as=0.1227fm

6. HMC Improvements Issues for 323x256
1-flavor light clover: problematic (via RHMC)
Recent major improvements/changes:
2-flavor Clover mass preconditioned by TWM
Switch to BiCG inverter
Switch to double precision
Tuning Omelyan integrator via Shadow Hamiltonian
Roughly factor >2X improvement

7. ORNL Allocations

8. ORNL Queues Queue Configuration Bin Cores Wall-Time Max Hrs
Aging Boost Days
Min
Max 1
17001
24.0 10 2
6001
17000 8 3
1025
6000
12.0 4
513
1024
4.0 5
257
512
2.5 6
256
1.5

9. ORNL Job Sizes

10. ORNL Running
Steps are wait period

11. Second Lattice Spacing?
Ratio method

12. Discretization effects
Isotropic clover action - small discretization effects
Quantify discretization effects:
Move to as=0.1fm
Compare ratio at m¼ ~ 500MeV (about 0.1 above)
Cost: 203x192, 5k traj ! ~1M core-hrs
Ratio method
BMW 2008

13. Distillation Correlation funcs Smear quark fields
Suggests (smooth) low-rank approximation for smearing
arxiv:

14. Distillation - mesons Smearing in correlator Correlator
Factorizes: operators and perambulators
A posteriori evaluation of C(2) after ¿
arxiv:

15. Demonstration for mesons
“½” spectrum (T1--) , m¼ = 383MeV, 163£ 128, 450 cfgs
Increasing vectors, better resolution
12x12 correlation matrix (derivatives source/sink) ! 5 states
4-th
~ 1.5% statistical error
3rd
2nd
1st
0-th
arxiv:

16. Demonstration for baryons
arxiv:
Nucleon (G1g), m¼ = 383MeV, 163£ 128, 316 cfgs
4x4 correlation matrix ! 2 states
Noise/signal: C(t) = a + b/Np ! p ~ Better than naïve scaling
2nd
~ 2.5% statistical error
0-th

17. Demonstration for two-mesons
arxiv:
Sump ¼(prel,t)¼(-prel,t)
m¼ = 383MeV, 163£ 128, 100 cfgs
5x5 correlation matrix (in p) ! 3 states
2nd
1st
0-th
Relative momentum at origin!

18. Distillation: annihilation diagrams
Two-meson creation op
Correlator
arxiv:

19. Distillation: 3pts Generalized perambulator Supports 3-pt functions tfX
arxiv:

20. Prospects Gauge production: Many improvements! Room for more… ORNL:
Heavily used: small fraction available for USQCD
Access to ANL resources requested
Distillation:
Looks promising!
Framework for multi-particle states
Flexible: useful for 2-pt and 3-pt

21. Distillation – baryons
Baryon op
Correlator
Perambulators
arxiv: