Interplay of LQCD and Experiment - Home | …...JLab Future Trends 2016 Spectroscopy 14 PRL 113 182001 PRD 91 054008 PRD 92 094502 ARXIV: 1602.05122 • First glimpse of spectrum of

Interplay of LQCD and Experiment

Robert Edwards

JLab Future Trends 2016

Questions in Nuclear Physics

2

Hägler, Musch, Negele, Schäfer, EPL 88 61001

• What observable states does QCD allow? - What is the role of the gluons? What about exotic matter?- Focus of GlueX experiment

• How do nucleons arise?- how are quarks & gluons distributed in a proton or neutron?- Focus of JLab 12 GeV, Halls A, B & C, RHIC-spin and future EIC

• QCD must predict properties of light nuclei- predict nuclear reaction properties, connect to effective theories- $730M Facility for Rare Isotope Beam (FRIB) will investigate nuclear structure

and interactions

• How does QCD behave under extreme temperatures & pressures such as in supernovae or shortly after Big-Bang

- Studied in RHIC at BNL


USQCD National Effort

3

US Lattice QCD coordinated effort involving JLab, BNL & FNAL

Software:2001 - 2012: SciDAC I & II

2013 - 2017: NP + ASCR SciDAC2013 - 2017: HEP + ASCR SciDAC

Intensity Frontier Energy FrontierHEP

Cold nuclear physics Thermodynamics

NP

Science campaigns:

USQCD Exec. Comm. USQCD Scientific Program Comm.

USQCD Facilities:JLab: 276 Xeon nodes, 42 (*4) GPU, new 2016BNL: 2.6 racks of BG/Q

Leadership Comp. Facilities:Proposals to DOE + NSF programsResources: ORNL, ANL, NERSC, NSF

FNAL: 538 cluster (Xeon) nodes, 32 (*4) GPU


SciDAC

4

Architecture Algorithm

Application

LQCD software development effort to support Leadership Computing and USQCD facilities

JLab: Jie Chen, Balint Joo, Frank Winter


Many computing architectures

5

Clover SolvesTitan (K20X)

0 512 1024 1536 2048 2560 3072 3584 4096 4608Titan Nodes (GPUs)

0

50

100

150

200

250

300

350

400

450

TFLO

PS

BiCGStab: 723x256DD+GCR: 723x256BiCGStab: 963x256DD+GCR: 963x256

Strong Scaling, QUDA+Chroma+QDP-JIT(PTX)

B. Joo, F. Winter (JLab), M. Clark (NVIDIA)

Propagator speedTitan (K20X)

0 100 200 300 400 500 600 700 800 900CPUs or GPUs

0123456789

101112

Spee

dup

fact

or re

lativ

e to

CPU

CPUCPU + QUDA GPU solversAll GPU (QDP-JIT + QUDA )

Gauge generation speedupsTitan (K20X)

JLab + SciDAC Super Institute

Porting code: Just-In-Time compilationDomain specific language

• Opportunity to optimize for speed & cost

- Current & future leadership “capability” resources - gauge generation

- USQCD “capacity” resources - propagators & contractions

• Developments from JLab:

- NviDIA GPUs (B. Joo, F. Winter + NviDIA developers)

- Intel Many Core (Xeon Phi) (B. Joo, C. Watson + Intel Parallel Computing Labs)

- BlueGene/Q (F. Winter + ANL staff)

- Partnership with SciDAC Super Institute - “automatic” porting of code


Lattice QCD

6

• First-principles numerical approach to the field-theory

CUBIC LATTICE

- Evaluate correlation functions

via Monte-Carlo sampling of path-integral

on a finite cubic grid

» in principle recover physical QCD as

» practical calculations often use

» large scale computational problem ...

‘sum ‘field ‘probability

e.g

However, use finite L as a tool (see R. Briceno talk)


LQCD workflow

7


LQCD workflow

7

Generate the configurations

!  Leadership level !  60K cores, 10’s TF-yr


LQCD workflow

7



Analyze 100K copies 4 Kepler GPUs +

t=0 t=T

Propagators

Q[U ]�1Now also AMG!


LQCD workflow

7




t=0 t=T

Propagators

Q[U ]�1

t=0 t=T

Contract •  8 cores, CPUs Correlators

100K – 1M copies

Now also AMG!


LQCD workflow

7




t=0 t=T

Propagators

Q[U ]�1

t=0 t=T


100K – 1M copies

Now also AMG!

1.0

1.5

2.0

2.5

3.0


LQCD workflow

7




t=0 t=T

Propagators

Q[U ]�1

t=0 t=T


100K – 1M copies

Few big jobs Few big files

Many small jobs Many big files

I/O movement

Now also AMG!

1.0

1.5

2.0

2.5

3.0


LQCD workflow

8




t=0 t=T

Propagators

Q[U ]�1

t=0 t=T


100K – 1M copies

~25%

> 5%

~75%

New analysis cost

Leadership level

Throughput mode Now also AMG!

1.0

1.5

2.0

2.5

3.0


Leadership facilities & USQCD facilities

9

Clover SolvesTitan (K20X)

• USQCD resources - capacity- Provide essential leverage of capability resources

- Systems cost optimized for USQCD physics program• SciDAC essential to exploit new computing architectures

- Virtuous circle - success at optimizing enables optimal use of future leadership systems

• Leadership resources - capability- Systems essential for USQCD gauge generation program

- Networks + compute nodes designed to support large node count jobs➡ Titan: exchange rate is 20M core-hours for $1M in integrated cost of project

- DOE + NSF programs (INCITE, ALCC, Petascale) available for high profile critical applications

- USQCD and associated projects quite successful

- Highlight: JLab spectroscopy project: 250M core-hour project still largest award to date @ ORNL


Spectroscopy: isovector meson spectrum

10

• Appears to be some qq-like near-degeneracy patterns_

PRL 103; PRD 82, 88


Spectroscopy: isovector meson spectrum

10

• Appears to be some qq-like near-degeneracy patterns_

PRL 103; PRD 82, 88


Meson spectrum from lattice QCD

11

EXOTIC MESONSMultiple exotic mesons within range of GlueX

PRL 103; PRD 82, 88


Meson spectrum from lattice QCD

12

PRL 103; PRD 82, 88

• ‘super’-multiplet of hybrid mesons roughly 1.2 GeV above the ρ

• these states have a dominant overlap onto


Chromo-magnetic excitation

13

• Subtract the ‘quark mass’ contribution

0.5

1.0

1.5

2.0

0.5

1.0

1.5

2.0

SU(3)F point


Chromo-magnetic excitation

13

• Subtract the ‘quark mass’ contribution

0.5

1.0

1.5

2.0

0.5

1.0

1.5

2.0

SU(3)F point

- Common energy scale of gluonic excitation


Spectroscopy

14

PRL 113 182001 PRD 91 054008 PRD 92 094502 ARXIV: 1602.05122

• First glimpse of spectrum of QCD light quark, strange quark, and charm quark mesons & baryons

• Results suggest rich pattern of states similar to non-rel. quark model + gluonic excitations

- Possibly many states within energy range of GlueX

• Expt. verification requires measurement in many decay channels

- Need partial wave couplings and resonance parameters


Spectroscopy

14

0

30

60

90

120

150

180

400 500 600 700 800 900 1000

mp = 391 MeV

mR = 790 ± 2 MeV

g = 5.69 ± 0.07

mp = 236 MeV

mR = 855 ± 1 MeV

g = 5.70 ± 0.10

I=1 ππ scattering

PRL 113 182001 PRD 91 054008 PRD 92 094502 ARXIV: 1602.05122







Spectroscopy

14

0

30

60

90

120

150

180

400 500 600 700 800 900 1000

mp = 391 MeV

mR = 790 ± 2 MeV

g = 5.69 ± 0.07

mp = 236 MeV

mR = 855 ± 1 MeV

g = 5.70 ± 0.10

I=1 ππ scattering

PRL 113 182001 PRD 91 054008 PRD 92 094502 ARXIV: 1602.05122






2.0 2.1 2.32.2 2.4 2.5

2.0 2.1 2.32.2 2.4 2.5

2.0

0

050100

4.0

6.0


Spectroscopy

14

0

30

60

90

120

150

180

400 500 600 700 800 900 1000

mp = 391 MeV

mR = 790 ± 2 MeV

g = 5.69 ± 0.07

mp = 236 MeV

mR = 855 ± 1 MeV

g = 5.70 ± 0.10

I=1 ππ scattering

PRL 113 182001 PRD 91 054008 PRD 92 094502 ARXIV: 1602.05122






2.0 2.1 2.32.2 2.4 2.5

2.0 2.1 2.32.2 2.4 2.5

2.0

0

050100

4.0

6.0

-500

-300

-100

600 800 1000 1200 1400 1600

broad

narrow tensor resonance

vector bound-state

scalar virtual bound-

J=0 J=1 J=2

I=1/2 Kπ & ηπ scattering


Spectroscopy

14

• Have a tool for phenomenology directly connected to QCD

• Put ideas of gluonic excitations/hybrid hadrons on firmer footing

• Finite-volume gives access to physical observables

0

30

60

90

120

150

180

400 500 600 700 800 900 1000

mp = 391 MeV

mR = 790 ± 2 MeV

g = 5.69 ± 0.07

mp = 236 MeV

mR = 855 ± 1 MeV

g = 5.70 ± 0.10

I=1 ππ scattering

PRL 113 182001 PRD 91 054008 PRD 92 094502 ARXIV: 1602.05122






2.0 2.1 2.32.2 2.4 2.5

2.0 2.1 2.32.2 2.4 2.5

2.0

0

050100

4.0

6.0

-500

-300

-100

600 800 1000 1200 1400 1600

broad


vector bound-state


J=0 J=1 J=2



Hadron Structure

15

Orginos, Syritsyn + LHPC: 1404.4029, 1505.01803

» S. Syritsyn currently Nathan Isgur Fellow -> moving to Stony Brook

• LHPC: calc. @ mπ=149 MeV showing isovector GE(Q2) and GM(Q2) agree with expt. error band

- Needed is the isoscalar contribution

• First calculation of disc. light quark contributions to EM form factors: ~0.5% of connected result

• Strange EM form factors consistent with expt. and much smaller uncertainty

- Hierarchical Probing proved crucial to reduce noise by 10x and resolve the signal

• Challenge: to control systematic errors sufficiently to resolve proton radius puzzle


Hadron Structure

15



0.0

0.2

0.4

0.6

0.8

1.0

0.0 0.1 0.2 0.3 0.4 0.5

Gp�

nE

Q2

(GeV

2

)

fit to experiment

lattice data, m⇡ = 149 MeV

Electromagnetic Form FactorLHPC








Hadron Structure

15



0.0 0.2 0.4 0.6 0.8 1.0 1.2Q2 (GeV2)

�0.004

�0.002

0.000

0.002

0.004

0.006

0.008

0.010

0.012

GE

strangelight disconnected

Disconnected contribution to GE

0.0 0.2 0.4 0.6 0.8 1.0 1.2Q2 (GeV2)

�0.05

0.00

0.05

0.10

0.15

Gs E

+h

Gs M

G0HAPPEX

A4lattice

Disconnected contribution to GE

mp = 317 MeV

0.0

0.2

0.4

0.6

0.8

1.0

0.0 0.1 0.2 0.3 0.4 0.5

Gp�

nE

Q2

(GeV

2

)

fit to experiment

lattice data, m⇡ = 149 MeV

Electromagnetic Form FactorLHPC








Nuclear structure

16

Orginos & NPLQCD: 1206.5219, 1301.5790, 1409.3356, 1505.02422, 1506.05518

The binding of light nuclei and their properties • Binding of light nuclei

- Used to predict properties of larger nuclei through matching to Nuclear EFT • First calculations of an inelastic nuclear reaction (neutron capture cross section) • First calculations of the magnetic moments and polarizabilities of light nuclei ➡ Nuclear shell-model structure of light nuclei persists over range of quark masses ➡ Magnetic moments of nucleons are generically quark-model


Nuclear structure

16

d nn 3He 4He nS L3 H L

3 He S3He L

4 He H-dib nX LL4 He

1+0+

12

+

0+

1+

12

+

32

+

12

+

32

+

0+

0+

0+

1+

0+

0+

s=0 s=-1 s=-2

2-Body 3-Body 4-Body

-160

-140

-120

-100

-80

-60

-40

-20

0

DEHMe

VL

Nuclear Structure

Orginos & NPLQCD: 1206.5219, 1301.5790, 1409.3356, 1505.02422, 1506.05518




Nuclear structure

16


3 He S3He L


1+0+

12

+

0+

1+

12

+

32

+

12

+

32

+

0+

0+

0+

1+

0+

0+

s=0 s=-1 s=-2


-160

-140

-120

-100

-80

-60

-40

-20

0

DEHMe

VL

Nuclear Structure

Refining Nuclear Forces

Orginos & NPLQCD: 1206.5219, 1301.5790, 1409.3356, 1505.02422, 1506.05518




Nuclear structure

16


3 He S3He L


1+0+

12

+

0+

1+

12

+

32

+

12

+

32

+

0+

0+

0+

1+

0+

0+

s=0 s=-1 s=-2


-160

-140

-120

-100

-80

-60

-40

-20

0

DEHMe

VL

Nuclear Structure


Orginos & NPLQCD: 1206.5219, 1301.5790, 1409.3356, 1505.02422, 1506.05518

Nuclear Reactions




Nuclear structure

16


3 He S3He L


1+0+

12

+

0+

1+

12

+

32

+

12

+

32

+

0+

0+

0+

1+

0+

0+

s=0 s=-1 s=-2


-160

-140

-120

-100

-80

-60

-40

-20

0

DEHMe

VL

Nuclear Structure


Orginos & NPLQCD: 1206.5219, 1301.5790, 1409.3356, 1505.02422, 1506.05518

Nuclear Reactions

p

n

d

3He

3H

-2

0

2

4

μ[��]

Magnetic Moments




Nuclear structure

16


3 He S3He L


1+0+

12

+

0+

1+

12

+

32

+

12

+

32

+

0+

0+

0+

1+

0+

0+

s=0 s=-1 s=-2


-160

-140

-120

-100

-80

-60

-40

-20

0

DEHMe

VL

Nuclear Structure


Orginos & NPLQCD: 1206.5219, 1301.5790, 1409.3356, 1505.02422, 1506.05518

Nuclear Reactions

p

n

d

3He

3H

-2

0

2

4

μ[��]

Magnetic Momentsp n nn d(Jz=±1) pp 3He 3H 4He

0

1

2

3

4

p n nn d(Jz=±1) pp 3He 3H 4He

β[M

N2(M

Δ-M

N)]

Nuclear Polarizabilities




Nuclear structure

16


3 He S3He L


1+0+

12

+

0+

1+

12

+

32

+

12

+

32

+

0+

0+

0+

1+

0+

0+

s=0 s=-1 s=-2


-160

-140

-120

-100

-80

-60

-40

-20

0

DEHMe

VL

Nuclear Structure

Near Future Program: • Properties of multi-neutron systems and the three-body forces

• Axial couplings for neutrino int. with nuclei

• Refinement of nuclear shell-model calculations of double beta-decay rates

• Nuclei and exotic nuclei at lighter pion masses


Orginos & NPLQCD: 1206.5219, 1301.5790, 1409.3356, 1505.02422, 1506.05518

Nuclear Reactions

p

n

d

3He

3H

-2

0

2

4

μ[��]

Magnetic Momentsp n nn d(Jz=±1) pp 3He 3H 4He

0

1

2

3

4

p n nn d(Jz=±1) pp 3He 3H 4He

β[M

N2(M

Δ-M

N)]

Nuclear Polarizabilities




Involvement with experimental program

17

Second phase of GlueX program with BaBar DIRC-s (approved)

Science case for JLab CLAS12 expt (approved)


Looking towards the future

18

Town Hall report providing input to next NSAC Long Range Plan

Recommendations endorsed at Joint Town Meeting on QCD in 2014, and appeared in LRP

With pre-exascale machines being deployed in 2017–2018, it is essential that the nuclear theory community be prepared to exploit these new architectures, requiring a concerted effort to port scientific codes and to optimize their performance. A significant software development workforce and close collaboration with computer scientists, applied mathematicians, and hardware vendors are required to successfully port our existing scientific code bases, in addition to expanding their scientific reach.


Looking towards the future - software

19

Proposals for coordinated software development for next generation of machines

NERSC (NESAP): Chroma Lattice QCD Code Suite, Balint Joo (Jefferson National Accelerator Facility)

Argonne (Early Science): The Hadronic Contribution to the Anomalous Magnetic Moment of the MuonCo-PI: Balint Joo (porting Chroma to “Theta” at ANL)



19




Exascale (ECP): Opportunities for HEP and NP



19








t=0 t=T

Propagators

Q[U ]�1

t=0 t=T


100K – 1M copies

1.0

1.5

2.0

2.5

3.0



19








t=0 t=T

Propagators

Q[U ]�1

t=0 t=T


100K – 1M copies

1.0

1.5

2.0

2.5

3.0

Few big jobs Few big files

Many small jobs Many big files

I/O movement

~25%

> 5%

~75%

New analysis cost

Leadership level

Throughput mode



19








t=0 t=T

Propagators

Q[U ]�1

t=0 t=T


100K – 1M copies

1.0

1.5

2.0

2.5

3.0

Specify correlators

Workflow coordinates tasks (compute, disk I/O, data transformations)



20





“Fat nodes”: Multiple memory hierarchies: GPU-like, CPU-like

Suggests using algorithms that exploit this hierarchy Exploit locality of data Expose more parallelism for execution

Collaboration/expertise in compiler/comms. Possibly influence emerging standards

Quite varied components in workflow Consider workflow languages in addition to C++

New I/O technologies, possibly in-situ like analysis


Looking towards the future - science program

21

• Spectroscopy

- Suggests rich spectrum of mesons & baryons - exotic & non-exotic hybrids • Next phase:

- More collaboration in S-matrix formalism & phenomenology

- tackle two-quark exotics & scalar sector

- N* and strange-baryon spectrum and decays

• Hadron structure

- First calculations at phys. pion mass of isovector quantities - charge radius, moments… • Next phase:

- direct calculation of isoscalar quantities - resolve proton charge radius “puzzle”

- full decomposition of spin from quarks and glue - relevant for 12GeV

- direct determination of quark distributions from pseudo-distributions (Ji’s method)

- structure of excited states gives insight into glue distribution - relevant to EIC

• Nuclear structure

- First predictions for larger nuclei through matching to nuclear EFT • Next phase:

- multi-neutron systems and three-body forces

- axial couplings for neutrino interactions & refinement of double beta-decay rates


Backup slides

22


LQCD group & collaborators

23

JEFFERSON LAB POSTDOCS & STUDENTS

Jozef Dudek (ODU) Robert Edwards Balint Joo Kostas Orginos (W&M) David Richards Frank Winter

Raul Briceno (ODU) Chris Bouchard (W&M) Arjun Gambhir (W&M) Sabbir Khan (ODU) Adithia Kusno (W&M) Sergey Syritsyn (JLab) Aaron Walden (ODU)

Christian Shultz (ODU) David Wilson (ODU)

COLLABORATORS

Mike Peardon (Trinity) Sinead Ryan (Trinity) Nilmani Mathur (Tata) Christopher Thomas (Cambridge) Steve Wallace (Maryland) Rajan Gupta (LANL) Christopher Monahan (Rutgers) Andreas Stathopoulos (W&M) Desh Ranjan (ODU) Momma Zubair (ODU)

“TECHNOLOGY”PRD79 034502 (2009) PRD80 054506 (2009) PRD85 014507 (2012)

MESON SPECTRUMPRL103 262001 (2009) PRD82 034508 (2010) PRD83 111502 (2011) JHEP07 126 (2011) PRD88 094505 (2013) JHEP05 021 (2013)

BARYON SPECTRUMPRD84 074508 (2011) PRD85 054016 (2012) PRD87 054506 (2013) PRD90 074504 (2014) PRD91 094502 (2015)

MATRIX ELEMENTSPRD91 114501 (2015) PRD90 014511 (2014) PRL115 242001 (2015)

DISTRIBUTIONS

PRD91 074513 (2015) OPE

NUCLEIPRD87 034506 (2013) PRD90 094507 (2014) PRL113 25001 (2014) PRD91 114503 (2015) PRD92 114502 (2015) PRL115 132001 (2015)

HADRON SCATTERINGPRD83 071504 (2011) PRD86 034031 (2012) PRD87 034505 (2013) PRL113 182001 (2014) PRD91 054008 (2015) PRD92 094502 (2015)


Excited states from correlators

24

• How to get at excited QCD eigenstates ?

- optimal operator for state :

for a basis of meson operators

- can be obtained (in a variational sense) from the matrix of correlators

- by solving a generalized eigenvalue problem

eigenvalues

- a large basis can be constructed using covariant derivatives :

‘diagonalize the correlation matrix’


Spectroscopy

25

PRL 113 182001 PRD 91 054008 PRD 92 094502 ARXIV: 1602.05122







Spectroscopy

25

0

30

60

90

120

150

180

400 500 600 700 800 900 1000

mp = 391 MeV

mR = 790 ± 2 MeV

g = 5.69 ± 0.07

mp = 236 MeV

mR = 855 ± 1 MeV

g = 5.70 ± 0.10

I=1 ππ scattering

PRL 113 182001 PRD 91 054008 PRD 92 094502 ARXIV: 1602.05122







Spectroscopy

25

0

30

60

90

120

150

180

400 500 600 700 800 900 1000

mp = 391 MeV

mR = 790 ± 2 MeV

g = 5.69 ± 0.07

mp = 236 MeV

mR = 855 ± 1 MeV

g = 5.70 ± 0.10

I=1 ππ scattering

-30

0

30

60

90

120

150

180

1000 1200 1400 1600

0.7

0.8

0.9

1.0 1000 1200 1400 1600


PRL 113 182001 PRD 91 054008 PRD 92 094502 ARXIV: 1602.05122







Spectroscopy

25

0

30

60

90

120

150

180

400 500 600 700 800 900 1000

mp = 391 MeV

mR = 790 ± 2 MeV

g = 5.69 ± 0.07

mp = 236 MeV

mR = 855 ± 1 MeV

g = 5.70 ± 0.10

I=1 ππ scattering

-30

0

30

60

90

120

150

180

1000 1200 1400 1600

0.7

0.8

0.9

1.0 1000 1200 1400 1600


-500

-300

-100

600 800 1000 1200 1400 1600

broad


vector bound-state


J=0 J=1 J=2

PRL 113 182001 PRD 91 054008 PRD 92 094502 ARXIV: 1602.05122







Spectroscopy

25

• Have a tool for phenomenology directly connected to QCD

• Put ideas of gluonic excitations/hybrid hadrons on firmer footing

• Finite-volume gives access to physical observables

0

30

60

90

120

150

180

400 500 600 700 800 900 1000

mp = 391 MeV

mR = 790 ± 2 MeV

g = 5.69 ± 0.07

mp = 236 MeV

mR = 855 ± 1 MeV

g = 5.70 ± 0.10

I=1 ππ scattering

-30

0

30

60

90

120

150

180

1000 1200 1400 1600

0.7

0.8

0.9

1.0 1000 1200 1400 1600


-500

-300

-100

600 800 1000 1200 1400 1600

broad


vector bound-state


J=0 J=1 J=2

PRL 113 182001 PRD 91 054008 PRD 92 094502 ARXIV: 1602.05122







Coupling to external currents

26

• what about production/decay featuring photons (e.g. GlueX)

BRICEÑO ET AL 2015• the finite-volume formalism is now in place

• first explicit application has appeared

2.0 2.1 2.32.2 2.4 2.5

2.0 2.1 2.32.2 2.4 2.5

2.0

0

050100

4.0

6.0

PRL 115 242001 (2015)


Scattering amplitudes from QCD

27

500 600 700 800 900 1000

0

0.2

0.4

0.6

0.8

1

1.2

0.18 0.20 0.22 0.24 0.26

0

30

60

90

120

150

180

0.18 0.20 0.22 0.24 0.26

0

0.4

0.8

1.2

1.6

2.0

2.4

0.26 0.28 0.30

PRD92 094502 (2015)

PRL113 182001 (2014) PRD91 054008 (2015)

arXiv:1602.05122

0.1

0.2

0.3

0.4

0.5

0.6

0.7

1000 1050 1100 1150 1200 1250 1300 / MeV

Documents

Interplay of LQCD and Experiment - Home | …...JLab Future Trends 2016 Spectroscopy 14 PRL 113 182001 PRD 91 054008 PRD 92 094502 ARXIV: 1602.05122 • First glimpse of spectrum of