• Introduction • General principles • Examples: quantitative nuclear theory • Predictive capability • Computing • Summary

Computational nuclear structure in the eve of exascale

Witek Nazarewicz (UTK/ORNL)

Nuclear Theory Seminar, Carnegie Mellon University, Nov. 15, 2012

• Introduction

• General principles

• Examples: quantitative nuclear theory

• Predictive capability

• Computing

• Summary

• A third rate theory forbids

• A second rate theory explains after the fact

• A first rate theory predicts

A. Lomonosov

Happy the man who has been able to discern the cause of things

Virgil, Georgica Theories

Models

The Nuclear Landscape

• Protons and neutrons formed 10^-6s-1s after Big Bang (13.7 billion years ago)

• H, D, He, Li, Be, B formed 3-20 min after Big Bang

• Other nuclei born later in heavy stars and supernovae The Nuclear Landscape and the Big Questions (NAS report)

• How did visible matter come into being and how does it evolve?

• How does subatomic matter organize itself and what phenomena emerge?

• Are the fundamental interactions that are basic to the structure of matter fully understood?

• How can the knowledge and technological progress provided by nuclear physics best be used to benefit society?

•Fundamental aspects (reduction)

• Nature of building blocks

• Nature of fundamental interactions

•Self-organization of building blocks (emergence)

• Nature of composite structures and phases

• Origin of simple patterns in complex systems

•Fundamental aspects (reduction)

• Nature of building blocks

• Nature of fundamental interactions

•Self-organization of building blocks (emergence)

• Nature of composite structures and phases

• Origin of simple patterns in complex systems

Nuclear structure Nuclear reactions

Hot and dense quark-gluon matter Hadron structure

Nuclear astrophysics New standard model

Applications of nuclear science Hadron-Nuclear interface

R es ol ut io n

Third Law of Progress in Theoretical Physics by Weinberg:

“You may use any degrees of freedom you like to describe a

physical system, but if you use the wrong ones, you’ll be sorry!”

E ff e ct iv e F ie ld T he o ry

DFT collective and

algebraic models

CI ab initio

LQCD

quark

models

Interfaces provide crucial clues Interfaces provide

crucial clues

dim ens ion of th e pr oble m

The nuclear landscape as seen by theorists …

11

Li

208

Pb

298

U

Physics of nuclei is demanding

Input

Forces, operators

• rooted in QCD

• insights from EFT

• many-body interactions

• in-medium renormalization

• microscopic functionals

• low-energy coupling constants optimized to data

• crucial insights from exotic nuclei

Many-body dynamics

• many-body techniques o direct schemes

o symmetry-based truncations o symmetry breaking and

restoration

• high-performance computing

• interdisciplinary connections

Open channels

• nuclear structure impacted by couplings to reaction and decay channels

• clustering, alpha decay, and fission still remain major challenges for theory

• continuum shell model, ab-initio reaction theory and microscopic optical model

• unified picture of structure and reactions

The Nuclear Many-Body Problem

Eigenstate of angular momentum, parity, and

~isospin

coupled integro-differential

equations in 3A dimensions

Illustrative physics examples

Ab initio theory for light nuclei and nuclear matter

Ab initio: QMC, NCSM, CCM,…

(nuclei, neutron droplets, nuclear matter)

 Quantum Monte Carlo (GFMC)

12C

 No-Core Shell Model ¹⁴F,

14C

 Coupled-Cluster Techniques

17F, ⁵⁶Ni, ⁶¹Ca

 Quantum Monte Carlo (GFMC)

12C

 No-Core Shell Model ¹⁴F,

14C

 Coupled-Cluster Techniques

17F, ⁵⁶Ni, ⁶¹Ca

Input:

•Excellent forces based on the phase shift analysis and few-body data

•EFT based nonlocal chiral NN and NNN potentials

•SRG-softened potentials based on bare NN+NNN interactions

NN+NNN interactions

NN+NNN interactions

Renormalization Renormalization Ab initio input

Many body method Many body

method

Observables Observables

• Direct comparison with experiment

• Pseudo-data to inform theory

GFMC: S. Pieper, ANL

1-2% calculations of A = 6 – 12 nuclear energies are possible excited states with the same quantum numbers computed

Ab initio: Examples

The ADLB (Asynchronous Dynamic Load-Balancing) version of GFMC was used to make calculations of ¹²C with a complete Hamiltonian (two- and three-nucleon potential AV18+IL7) on 32,000 processors of the Argonne BGP. These are believed to be the best converged ab initio calculations of

12C ever made. The computed binding energy is 93.5(6) MeV compared to the experimental value of 92.16 MeV and the point rms radius is 2.35 fm vs 2.33 from experiment.

The ADLB (Asynchronous Dynamic Load-Balancing) version of GFMC was used to make calculations of ¹²C with a complete Hamiltonian (two- and three-nucleon potential AV18+IL7) on 32,000 processors of the Argonne BGP. These are believed to be the best converged ab initio calculations of

12C ever made. The computed binding energy is 93.5(6) MeV compared to the experimental value of 92.16 MeV and the point rms radius is 2.35 fm vs 2.33 from experiment.

12C in GFMC: Pieper et al.

Epelbaum et al., Phys. Rev. Lett. 106, 192501 (2011)

Lattice spacing 1.97 fm

Examples: Ab Initio

Ab-initio description of medium-mass open nuclear systems

G. Hagen et al., Phys. Rev. Lett. 109, 032502 (2012)

½⁺ virtual state

• Strong coupling to continuum for neutron rich calcium isotopes

• Level ordering of states in the gds shell is

contrary to naïve shell model picture

• Strong coupling to continuum for neutron rich calcium isotopes

• Level ordering of states in the gds shell is

contrary to naïve shell model picture

RIKEN

Configuration interaction techniques

• light and heavy nuclei

• detailed spectroscopy

• quantum correlations (lab-system description)

NN+NNN interactions

NN+NNN

interactions RenormalizationRenormalization

Diagonalization

Truncation+diagonalization Monte Carlo

Diagonalization

Truncation+diagonalization Monte Carlo

Observables Observables

• Direct comparison with experiment

• Pseudo-data to inform reaction theory and DFT Matrix elements

fitted to experiment Matrix elements fitted to experiment

Input: configuration space + forces

Method

Fusion of Light Nuclei

Ab initio theory reduces uncertainty due

to conflicting data ^{ The n-3}H elastic cross section for 14 MeV neutrons, important for NIF, was not known precisely enough.

 Delivered evaluated data with required 5%

uncertainty and successfully compared to

measurements using an Inertial Confinement Facility

 ``First measurements of the differential cross sections for the elastic n-²H and n-³H scattering at 14.1 MeV using an Inertial Confinement Facility”, by J.A. Frenje et al., Phys.

Rev. Lett. 107, 122502 (2011)

http://physics.aps.org/synopsis-for/10.1103/PhysRevLett.107.122502

NIF

Computational nuclear physics enables us to reach into regimes where experiments and analytic theory are not possible, such as the cores of fission reactors or hot and dense

evolving environments such as those found in inertial confinement fusion environment.

16

Mean-Field Theory Density Functional Theory ⇒

• mean-field one-body densities⇒

• zero-range local densities⇒

• finite-range gradient terms⇒

• particle-hole and pairing channels

• Has been extremely successful.

A broken-symmetry generalized product state does surprisingly good job for nuclei.

Nuclear DFT

• two fermi liquids

• self-bound

• superfluid

Degrees of freedom: nucleonic densities

NN+NNN interactions

NN+NNN interactions

Density Matrix Expansion Density Matrix

Expansion Input

Energy Density Functional Energy Density

Functional

Observables Observables

• Direct comparison with experiment

• Pseudo-data for reactions and astrophysics

Density dependent interactions Density dependent

interactions

Fit-observables

• experiment

• pseudo data Fit-observables

• experiment

• pseudo data

Optimization Optimization

DFT variational principle HF, HFB (self-consistency)

Symmetry breaking DFT variational principle HF, HFB (self-consistency)

Symmetry breaking

Symmetry restoration Multi-reference DFT (GCM) Time dependent DFT (TDHFB)

Symmetry restoration Multi-reference DFT (GCM) Time dependent DFT (TDHFB)

Nuclear Density Functional Theory and Extensions

• two fermi liquids

• self-bound

• superfluid (ph and pp channels)

• self-consistent mean-fields

• broken-symmetry generalized product states

Technology to calculate observables

Global properties Spectroscopy

DFT Solvers Functional form Functional optimization Estimation of theoretical errors

Mass table

Goriely, Chamel, Pearson: HFB-17 Phys. Rev. Lett. 102, 152503 (2009)

dm=0.581 MeV dm=0.581 MeV

Cwiok et al., Nature, 433, 705 (2005)

BE differences

Examples: Nuclear Density Functional Theory

Traditional (limited) functionals

provide quantitative description

Example: Large Scale Mass Table Calculations

 5,000 even-even nuclei, 250,000 HFB runs, 9,060 processors – about 2 CPU hours

 Full mass table: 20,000 nuclei, 12M configurations — full JAGUAR HFB+LN mass table, HFBTHO

0 4 8 12 16 20 24

S

( M e V )

Er

neutron number

80 100 120 140 160

experiment drip line

0 2 4

140 148 156 164

neutron number

0 4 8

58 62 66

proton number

N=76 _{154 162}

S2n (MeV) S2p (MeV)

FRDM HFB-21 SLy4 UNEDF1 UNEDF0 SV-min exp

Er

Description of observables and model-based extrapolation

• Systematic errors (due to incorrect assumptions/poor modeling)

• Statistical errors (optimization and numerical errors)

Erler et al., Nature (2012)

0 40 80 120 160 200 240 280

neutron number

0 40 80 120

p ro to n n um b er

to n d r ip line

t w o - n e u t r o n d r i p l i n e

232 240 248 256

n e u t r o n n u m b e r

proton number

90 110

100 Z=50

Z=82

Z=20

N=50

N=82

N=126

N=20

N=184

d r i p l i n e S V - m i n

k n o w n n u c l e i s t a b l e n u c l e i

N=28 Z=28

230 244

N=258

Nuclear Landscape 2012

S 2 n = 2 M e V

How many protons and neutrons can be bound in a nucleus?

Skyrme-DFT: 6,900±500

Skyrme-DFT: 6,900±500

Literature: 5,000-12,000

288

~3,000

Erler et al.

Nature 486, 509 (2012)

Asymptotic freedom ?

from B. Sherrill

DFT FRIB

current

Quantified Nuclear Landscape

0 40 80 120 160 200 240 280

neutron number

0 40 80 120

proton number

0 40 80 120

proton number

0 40 80 120 160 200 240 280

neutron number

0 40 80 120

proton number

-0.3 -0.2 -0.1 0 0.1 0.2 0.3 0.4

β₂

SkM* SkP

UNEDF1 SLy4

UNEDF0 SV-min

N=258

N=184

Z = 5 0

Z = 2 8 Z = 2 0 Z = 8 2

N=82

N=50

N=28

N=20 N=126

Quadrupole ground-state shape deformations

• Global behavior shows generic patterns and similar systematic trends

• Deviations from the usual behavior in Terra Incognita clearly seen

nuclear meson decay

Superallowed Fermi 0⁺ →0⁺ -decay studies

Kobayashi and Maskawa: … for

"the discovery of the origin of broken symmetry, which predicts the existence of at least three families of quarks in nature."

0.9999(6)

Towner and Hardy 2010

Impressive experimental effort worldwide

Microscopic calculations of isospin-breaking corrections to superallowed -decay

W. Satuła et al.,Phys. Rev. Lett 106, 132502 (2011)

82

50 28

28

50 82

20 2 8

2 8

20

126

neutron stars

p ro to n s

neutrons

Extended Nuclear Landscape

Quest for understanding the neutron-rich matter on Earth and in the Cosmos

8 9 10 11 12 13 14 15 R (km)

0 1 2

M (Msolar)

2r⁰ 4r⁰

1.4 1.97(4)

NN

3r ⁰ 5r ⁰

0 . 3 0 f m 0 . 1 5 f m

r0 NN

+NNN

Gandolfi et al. PRC85, 032801 (2012)

Erler et al., 2012

Optimized Functionals Optimized Functionals

Numerical Techniques Numerical Techniques

Large-scale DFT Large-scale DFT

Confrontation with experiment; predictions Confrontation with experiment; predictions

Collective dynamics Collective dynamics

LACM, Fission: the ultimate challenge

Stability of the heaviest nuclei, r-process, advanced fuel cycle, stockpile stewardship…

PRC 78, 014318 (2008)

PRC 85, 024304 (2012) PRC 84, 054321(2011)

PRC 80, 014309 (2009)

PRC 80, 014309 (2009)

SF fission “Death Valley” in SHE

Trans-actinides

Superheavy nuclei

“critical” zones - 6

- 8 - 4 - 2 0 2 4 6 8 1 0 1 2 1 4 1 6 1 8

Z = 1 1 2

1 1 2 1 1 4 1 1 0 1 0 6

1 0 6 1 0 8 1 0 4

1 0 2 1 0 0

9 8 N= 1 26

N= 1 25 N = 18 4

1 4 0 1 4 5 1 5 0 1 5 5 1 6 0 1 6 5 1 7 0 1 7 5 1 8 0 1 8 5

N e u t r o n n u m b e r Log T (s)SF

T h e o r y Spontaneous fission half-lives

Spontaneous fission half-lives Actinides

48Ca-induced reactions

Oganessian et al.

A. Staszczak et al. arXiv:1208.1215

0 0.1 0.2 0.3 0.4 0.5

k_F [fm^-1]

0.4 0.5 0.6 0.7 0.8 0.9 1

E / E FG

QMC s-wave QMC AV4 Cold Atoms

0 2 4 6 8 10

- k_F a

0.4 0.5 0.6 0.7 0.8 0.9 1

E / E FG

Lee-Yang

Connections to Other Fields: Cold Atoms

Vortex Dynamics

Bulgac et al.,

Science, 332, 1288 (2011)

Exotic pairing phases

J. Pei et al., Phys. Rev. A 82, 021603(R) (2010)

Gezerlis and Carlson, Phys. Rev. C 77, 032801(R) (2008) Carlson, Gandolfi, Gezerlis, PTEP (2012)

Equation of State

http://www.physicstoday.org/resource/1/phtoad/v64/i8/p19_s1

Quality control

Uncertainty quantification

Quality Control

Verification and Validation

• Cross-check of different methods and codes

• Benchmarking

Uncertainty Quantification and Error Analysis

• Tools for correlation analysis to estimate errors and significance

• Uncertainty analysis Assessment

• Development and application of statistical tools

• Analysis of experimental data significance

Integral to any scientific project is the verification of methods and codes, the estimation of uncertainties, and assessment.

Earlier fit Earlier fit

Final fit Final fit

Quality Control (2)

P.G. Reinhard and WN, Phys. Rev. C 81, 051303 (R) (2010)

• G.F. Bertsch et al., Phys. Rev. C 71, 054311 (2005).

• M. Kortelainen et al., Phys. Rev. C 77, 064307 (2008).

• J. Toivanen et al., Phys. Rev. C 78, 034306 (2008).

• P. Klüpfel et al., Phys. Rev. C 79, 034310 (2009).

• P.-G. Reinhard and W. Nazarewicz, Phys Rev. C 81, 051303 (R) (2010)

• M . Kortelainen et al., Phys. Rev. C 82, 024313 (2010)

• J. Dudek et al., Int. J. Mod. Phys. E 19, 652 (2010).

Examples of some DFT- based work

Examples of some DFT- based work

Improved Density Functionals Neutron Drops, Masses, Fission,…

Derivative-free optimization, uncertainty quantification

http://www.deixismagazine.org/2011/03/cranking-up-the-speed-of-df/

http://www.deixismagazine.org/2011/03/pounding-out-atomic-nuclei/

http://www.mcs.anl.gov/news/detail.php?id=720

PHYSICAL REVIEW A 83, 040001 (2011): Editorial: Uncertainty Estimates

The purpose of this Editorial is to discuss the importance of including uncertainty estimates in papers involving theoretical calculations of physical quantities.

It is not unusual for manuscripts on theoretical work to be submitted without uncertainty estimates for numerical results.

In contrast, papers presenting the results of laboratory measurements would usually not be considered acceptable for publication in Physical Review A without a detailed discussion of the uncertainties involved in the measurements. For example, a graphical presentation of data is always accompanied by error bars for the data points. The determination of these error bars is often the most difficult part of the measurement. Without them, it is impossible to tell whether or not bumps and irregularities in the data are real physical effects, or artifacts of the measurement. Even papers reporting the observation of entirely new phenomena need to contain enough information to convince the reader that the effect being reported is real. The standards become much more rigorous for papers claiming high accuracy.

The question is to what extent can the same high standards be applied to papers reporting the results of theoretical calculations. It is all too often the case that the numerical results are presented without uncertainty estimates. Authors sometimes say that it is difficult to arrive at error estimates. Should this be considered an adequate reason for omitting them? In order to answer this question, we need to consider the goals and objectives of the theoretical (or

computational) work being done.

(…) there is a broad class of papers where estimates of theoretical uncertainties can and should be made. Papers presenting the results of theoretical calculations are expected to include uncertainty estimates for the calculations whenever practicable, and especially under the following circumstances:

1. If the authors claim high accuracy, or improvements on the accuracy of previous work.

2. If the primary motivation for the paper is to make comparisons with present or future high precision experimental measurements.

3. If the primary motivation is to provide interpolations or extrapolations of known experimental measurements.

These guidelines have been used on a case-by-case basis for the past two years. Authors have adapted well to this, resulting in papers of greater interest and significance for our readers.

Future: large multi-institutional efforts involving strong coupling between

physics, computer science, and applied math

“High performance computing provides answers

to questions that neither experiment nor analytic

theory can address; hence, it becomes a third

leg supporting the field of nuclear physics.”

1Teraflop=10¹² flops 1peta=10¹⁵ flops (today)

1exa=10¹⁸ flops (next 10 years)

Theoretical Tools and Connections to Computational Science Theoretical Tools and Connections to Computational Science

Tremendous opportunities

for nuclear theory!

•Funded for 5 years by DOE (NP/SC, NNSA, ASCR)

•9 universities and 7 national labs

•Junior scientists: 11 students, 19 postdocs/year

• ~50 researchers in



physics



computer science



applied mathematics

•International partners

EXAMPLE: Universal Nuclear Energy Density Functional

(other, smaller, collaborations exist: NuN, TORUS, PetaApps,…)

• Ab initio structure

• Ab initio functionals

• DFT applications

• DFT extensions

• Reactions

For a popular description of UNEDF, see:

• SciDAC Review Winter 2007

http://www.scidacreview.org/0704/pdf/unedf.pdf

• Nucl. Phys. News 21, No. 2, 24 (2011)

• Office of Science “Highlight Series”:

http://science.energy.gov/news/in-focus/2011/03-28-11-s/

Focus on:

• Predictive power

• Robust extrapolations

• Validation

• Guidance

Focus on:

• Predictive power

• Robust extrapolations

• Validation

• Guidance

UNEDF in 2007 UNEDF in 2011

! " #$%&' () *%' +, ( - %*. /0, ( 1" *#23*&$(

4 5. %' 6&5$%. (

! " #$%&' (- 17(

8 19(:. %$;<#3*. /. 0%*#, =(

>, ? ? %0' , (5' %&@/*+(

Nuclear Density Functional Theory

and Extensions

A &*, ( 53B, (

? %0C3B(

4 5. %' 6&5$%. (

- /&+3*&$/D&23*(

7' " *#&23*EB/&+3*&$/D&23*((

A 3*0%(F&' $3(

0' " *#&0%B(. G&#%(

4 5. %' 6&5$%. (

;" $$(. G&#%(

>, ? ? %0' , (' %. 03' &23*(

A " $2<' %;%' %*#%(- 17(:HFA =(

7/? %(B%G%*B%*0(- 17(:7- 8 19=(

I J KL ( 4 5. %' 6&5$%. (

Ab Initio Configuration Interaction

!"#$%

!"#$%

!"#$%

!"#$%

Compound Nucleus and Direct Reaction

Theory F3" G$%B(FC&**%$. ( - M9L (

+$35&$(

G' 3G%' 2%. (

4 5. %' 6&5$%. (

!"#$% #' 3. . (. %#23*. (

%N#/0%B(. 0&0%. ( B%#&, . ( O. . /3*(

. G%#0' 3. #3G/#(

/*;3' ? &23*(

+$35&$(G' 3G%' 2%. ( . G%#0' 3. #3G, (

. #&P %' /*+(

4 G2#&$(G30%*2&$. (

• As UNEDF has matured, it has been satisfying to see the increased coherence within the effort; the project has created and facilitated an increasing interplay among the major sections where none existed

previously. Each of the main physics areas now includes collaborations that cross over into other sectors.

• A fair number of applications are now

running on the biggest machines available, a vast change from when UNEDF started.

Started as a collection of researchers addressing five different aspects of nuclear science

As UNEDF has matured, it has been satisfying to see the increased coherence within the effort. Indeed, the project has created and facilitated an increasing interplay among the major sections where none existed previously. Each of the main physics areas now includes collaborations that cross over into other sectors.

Ab Initio ↔ DFT

• Neutron drops data provide constraints on density functionals

• Quantum Monte Carlo and DFT description of dilute fermions EFT ↔ DFT

• Novel functionals derived from chiral interactions

• Assessment of 3N forces

• Global characterization of functionals Ab initio ↔ Reactions

• Ab initio description of scattering and light-ion reactions DFT↔ Reactions

• QRPA states for reaction cross sections

• Microscopic optical model potential from functionals

• DFT and TDFT description of fission

• CI ↔ DFT

• Correlation energy in DFT from shell model

• Nuclear densities and reaction rates from CI and DFT PHY ↔ CS/AM (see below)

• High-performance computing

• New algorithms

• New frameworks

Coherence and Uniqueness

A fair number of applications are now running on the biggest machines available, a vast change from when UNEDF started. We can take advantage of the

movement towards exascale!

UNEDF Deliverables

250 publications:

1 Science, 1 Nature, 50 PRLs Over 30 codes and databases

UNEDF Deliverables

250 publications:

1 Science, 1 Nature, 50 PRLs Over 30 codes and databases

http://unedf.org/

! " #$ %&'( ) *( +, - ./ 012) , , &134 / 5678&

! &9.87.: 1) 7;&)1134 / 2., - 4 ( 7;, <&&)1- .( =( *&

;- +358- &2( ) *( +, - ./ 012) , , &134 / 5678&

" >?0.7.63&1) 2152) 637, &3@&A0BC&

" ! 7*( +, ;) 7*.78&2378&2.@( 64 ( &3@&A0BD&

" E .1+3, 13/ .1&1) 2152) 637, &3@&,( 2( 1;&&

&&&&4 ( *.54 04 ) , , &7512( .&

" F4 / +3=( *&(7( +8G0*( 7, .;G&@571637) 2&! " #$ %B&

! &91.$>A&1322) ?3+) 637&?( ;H( ( 7&)/ / 2.( *&4 ) ;- ( 4 ) 61.) 7, <&134 / 5;( +&,1.( 76, ;, <&

) 7*&7512( ) +&/- G, .1., ;, &2( ) *, &;3&(I 1.( 7;&562.J) 637&3@&2( ) *( +, - ./ 012) , , &134 / 5678&

+( , 35+1( , &@3+&7512( ) +&/- G, .1, &/+3?2( 4 , &

! &KLM&;3&NLM&3@&134 / 5678&+( , 35+1( , &&&&&&&&

5, ( *&);&2( ) *( +, - ./ 012) , , &,1) 2( &

!" #

$!!" #

%!!&# %!!' # %!$!# %!$$#

()" # %(" # (' " # ($" #

*!" # *+" # ((" # *)" #

%)" # ($" # %&" # %*" # ,#%!" # ,#%!" #- #,#+!" # .#+!" #

!#

$!#

%!#

(!#

*!#

)!#

+!#

/ !#

%!!&# %!!' # %!$!# %!$$# %!$%# %!$(#

012342567#

>2231) 637&

! , ) 8( &

!#

$!#

%!#

(!#

*!#

)!#

%!!&# %!!' # %!$!# %!$$# %!$%# %!$(#

012342567#

>2231) 637&

! , ) 8( &

89:2; <9=#>29=#

?; 6=2@A<#0BC8D7#?E8?FG#HI :AJ9I K; #

L9M49=#0NC8D7#?E8?FG#HI :AJ9I K; #

8OHPQK4=5#0RA::AK;57#8OHPQK4=5#0RA::AK;57#

Junior Scientists in UNEDF

POST-DOCTORAL ASSOCIATES (2010)

Christopher Calderon, LBNL (staff, Numerica co.) Joaquin Drut (Professor, UNC)

Stefano Gandolfi, LANL (staff, LANL) Kai Hebeler, OSU (TRIUMF)

Heiko Hergert, MSU (OSU) Jason Holt, UTK/ORNL

Eric Jurgenson, LLNL (staff, LLNL)

Markus Kortelainen, UTK (U. Jyväskylä) Plamen Krastev, UCSD (research, Harvard) Pieter Maris, ISU (Research Prof. ISU)

Eric McDonald, MSU (staff scientist, MSU) Gustavo Nobre, LLNL (BNL, NNDC)

Junchen Pei, UTK (Prof., Pekin U.) Nicolas Schunck UTK (staff, LLNL) Roman Senkov, CMU

Ionel Stetcu, UW (staff, LANL)

Jun Terasaki, UNC (staff, U. Tsukuba) Stefan Wild, ANL (staff, ANL)

Effect of UNEDF on workforce

Year-1: 9 students, 17 postdocs;

Year-2: 12 and 12;

Year-3: 10 and 18;

Year-4: 11 and 19

2010: Early Career Award

2011: Faculty UNC/Chapel Hill

2010:Staff LLNL

2012: Faculty Guelph

2012:

Harvard Research Computing

2010:

Math/CS Staff ANL

Relevant instruction (workshops, courses) is crucial for the future of the field

Prospects

Petaflop-Yrs on Task

Transport in QCD (quenched) Transport in QCD (quenched)

Isotope separator optimization

Isotope separator optimization Energy Recovery LinacEnergy Recovery Linac Nucleon Spin

Nucleon Spin Deuteron Deuteron

Alpha particle Alpha particle

10

1 10 10

Hot and Dense QCD Hot and Dense

QCD

Cold QCD Cold QCD

Nuclear Structure

Nuclear Structure

Nuclear Astrophysics

Nuclear Astrophysics

Accelerator Physics Accelerator

Physics

Excited hadron spectrum Excited hadron spectrum Nuclear force

Nuclear force

Neutron EDM Neutron EDM

10

Gluon distributions Gluon distributions

Light nuclei Light nuclei

Light ion reactions Light ion reactions

Triple a process Triple a process

0n  rates for ⁴⁸Ca 0n  rates for ⁴⁸Ca Neutron induced fission Neutron induced fission Weakly bound nuclei

Weakly bound nuclei

Dynamics of neutron star crust Dynamics of neutron star crust

3D supernova 3D supernova Global solar model

Global solar model

Precision nuclear network Precision nuclear network

Precision neutrino network Precision neutrino network Multienergy neutrino transport

Multienergy neutrino transport

QCD critical point QCD critical point High-T limit of QCD EOS

High-T limit of QCD EOS QCD at T>0QCD at T>0 Continuum extrapolated QCD EOS Continuum extrapolated QCD EOS Quarkonium spectroscopy

Quarkonium spectroscopy

Electron-cooling design Electron-cooling design

6D Vlasov 6D Vlasov

Martin Savage Martin Savage

Validated Nuclear Interactions Validated Nuclear

Interactions

Structure and Reactions:

Light and Medium Nuclei Structure and Reactions:

Light and Medium Nuclei

Structure and Reactions:

Heavy Nuclei Structure and Reactions:

Heavy Nuclei

Chiral EFT Ab-initio

Optimization Model validation

Uncertainty Quantification

Neutron Stars

Neutron Stars FissionFission

Neutrinos and Fundamental Symmetries

Neutrinos and Fundamental Symmetries

Ab-initio RGM CI

Load balancing Eigensolvers Nonlinear solvers Model validation

Uncertainty Quantification

DFT TDDFT

Load balancing Optimization Model validation

Uncertainty Quantification Eigensolvers

Nonlinear solvers Multiresolution analysis

Stellar burning Stellar burning

fusion fusion

NUclear Computational Low-Energy Initiative

Neutron drops EOS Correlations

TJNAF FRIB

LANL LLNL NIF

Majorana

LENP facilities

SNS

Theory is developing new statistical tools to deliver uncertainty

quantification and error analysis for theoretical studies as well as for the assessment of new experimental data. Such technologies are

essential as new theories and computational tools are explicitly intended to be applied to entirely

new nuclear systems and conditions that are not accessible to

experiment.

Future: large multi-institutional efforts involving strong coupling between physics, computer science, and applied math

http://unedf.org/

• The nuclear many-body problem is very complex, computationally difficult, and interdisciplinary.

• With a fundamental picture of nuclei based on the correct

microphysics, we can remove the empiricism inherent today, thereby giving us greater confidence in the science we deliver and predictions we make

• For reliable model-based extrapolations, we need to improve predictive capability by developing methods to quantify

uncertainties

• Large international coherent theory effort is needed to make progress

• New-generation computers will continue to provide unprecedented opportunities

• Collaboration with computer scientists and applied mathematicians is the key

Summary

Thank You

Thank You

“Load Balancing at Extreme Scale” – Ewing Lusk, Argonne National Laboratory

Impact Objectives

 Enable Green’s Function Monte Carlo calculations for ¹²C on full BG/P as part of UNEDF project

 Simplify programming model

 Scale to leadership class machines

 Demonstrate capabilities of simple programming models at petascale and beyond

 Show path forward with hybrid programming models in library implementation

Deployment of Detection Network

 Initial load balancing was of CPU cycles

 Next it became necessary to balance memory utilization as well

 Finally ADLB acquired the capability to balance message flow

 “More Scalability, Less Pain”

by E. Lusk, S.C. Pieper and R. Butler published in

SciDAC Review 17, 30 (2010)

Progress

ASCR- SciDAC UNEDF Computer Science Highlight

Improved Efficiency (compute time/wall time) with more nodes

Optimization Algorithms for Calibrating Extreme Scale Simulations

New Algorithm POUNDERS Typical Challenges

 Computational expense of simulation only allows for evaluating a few sets of parameter values

 Derivatives with respect to parameters can be unavailable or intractable to compute/approximate

 Experimental data incomplete or inaccurate

 Sensitivity analysis/confidence regions desired

 Exploits mathematical structure in calibration problems

 Benefits from expert knowledge

 data, weights, uncertainties, etc.

 Obtains good fits in minimal number of simulations

POUNDERS obtains better solutions faster

 Enables fitting of complex, state-of-the-art EDFs

• Optimization previously avoided because too many evaluations required to obtain desirable features

 Substantial computational savings over alternatives

 Using resulting EDF parameterizations, the entire

nuclear mass table was computed and is now distributed at www.massexplorer.org

 Nuclear Energy Density Optimization. Kortelainen et al., Physical Review C 82, 024313, 2010

 Three joint physics & optimization publications @ SciDAC11!

Energy density functionals (EDFs) for UNEDF