Rivet - and (my) prospects for ALICE · 2016. 3. 22. · Rivet and (my) prospects for ALICE Jochen Klein CERN Seminar Bologna, Italy March 21st, 2016 This is an output file created

Rivetand (my) prospects for ALICE

Jochen Klein

CERN

SeminarBologna, Italy

March 21st, 2016

This is an output file created in Illustrator CS3

Colour reproductionThe badge version must only be reproduced on a plain white background using the correct blue: Pantone: 286 CMYK: 100 75 0 0 RGB: 56 97 170 Web: #3861AA

Where colour reproduction is not faithful, or the background is not plain white, the logo should be reproduced in black or white – whichever provides the greatest contrast. The outline version of the logo may be reproduced in another colour in instances of single-colour print.

Clear spaceA clear space must be respected around the logo: other graphical or text elements must be no closer than 25% of the logo’s width.

Placement on a documentUse of the logo at top-left or top-centre of a document is reserved for official use.

Minimum sizePrint: 10mmWeb: 60px

CERN Graphic Charter: use of the outline version of the CERN logo

Rivet pp physics tuning heavy-ion summary

outline

trying to bridge between technical and physics aspects

I Monte Carlo (QCD) event generators

I Rivet framework and analyses

I examples of relevant analyses(biased to scope of ALICE)

I parameters and tuning

I particularities of heavy-ion analyses

there will be code snippets to illustrate concepts,

but this is NOT meant to be a Rivet tutorial!

Jochen Klein (CERN) Rivet & (my) prospects for ALICE Bologna, Mar 2016 2 / 41


setting the stage

I data: measurements at hadron colliders

I goal: test QCD and/or QCD-inspired models

I involved analyseswith finite acceptance, trigger conditions, non-trivial cuts, . . .(normally corrected for detector effects)

I non-trivial algorithms on measured final statesto extract physics objects (e.g. jet finders)

I analyses published in paper,data values (often) published on HepData

comparison to theory often requires full final state

Monte Carlo event generators



Monte Carlo event generatorsevent generators serve various purposes:I provide explicit final state (at least those used with Rivet),

also with complex phase space

I test theory and models

I reproduce known/expected processesI detector performance studiesI efficiency evaluationI background subtractionI . . .

I some have lots of tuning freedom:good to reproduce data, limited theory insight

I some implement a well-constrained idea:good to test idea, limited possibility to adapt to data

Monte Carlo generators with different focus/goalJochen Klein (CERN) Rivet & (my) prospects for ALICE Bologna, Mar 2016 4 / 41


Monte Carlo modelling

consider pp collisions at LHC:

I assuming factorisation decoupling of short/long-range processes

I hard interaction (hard scattering, hard ISR/FSR)I hard scale pQCD descriptionI multi-parton interactions

I parton shower/cascadeI sequential branchings (DGLAP evolution)I initial/final state radiationI parameters: αs, intrinsic kt, Q

20 cut off for initial/final state radiation

I hadronisationI below cut-off scale:

phenomenological hadronisation model

I underlying event (everything but hard interaction)



general purpose generators (pp)

I PYTHIA 6/8 [hep-ph/0603175, 1410.3012]I fixed implementation of hard processesI multi-parton interactions

(controlled by shape and pT,0)I p⊥-ordered parton shower (earlier virtuality-ordered)I Lund string hadronisation

I Herwig(++) [hep-ph/0011363, hep-ph/0210213, 0803.0883]I automatic generation of hard processesI multi-parton interactionsI angular-ordered parton showerI cluster hadronisation

I Sherpa [0811.4622]I modular matrix element generationI multi-parton interaction similar to PYTHIAI dipole showerI cluster hadronisation


http://arxiv.org/abs/hep-ph/0603175http://arxiv.org/abs/1410.3012http://arxiv.org/abs/hep-ph/0011363http://arxiv.org/abs/hep-ph/0210213http://arxiv.org/abs/0803.0883http://arxiv.org/abs/0811.4622


further generators

I POWHEG BOX [hep-ph/0409146, 0709.2092, 1002.2581]I matching of NLO calculations with parton showers

by implementation of POWHEG method(Positive Weight Hardest Emission Generator)

I relevant for cross sections at NLO precision

I DIPSY [http://home.thep.lu.se/DIPSY/]I generator based on ThePEG framework

(Dipoles in Impact Parameter Space and rapidity (Y))I rope hadronization

I Alpgen, MadGraph, MC@NLO, . . .

large variety of Monte Carlo generators + tunes need for reliable comparison to data

can be difficult to reproduce analysis purely from paper


http://arxiv.org/abs/hep-ph/0409146http://arxiv.org/abs/0709.2092http://arxiv.org/abs/1002.2581http://home.thep.lu.se/DIPSY/


history

I pre-LHC solution: HZToolanalysis library, focussed on H1 and ZEUS[http://hztool.hepforge.org]

I preparations for physics at LHC

I emergence of FastJetas “standard” tool for jet finding[http://fastjet.fr]

I efforts of the phenomenology community led to Rivet[http://rivet.hepforge.org]

Rivet

tools to calculate observablesand library of analyses


http://hztool.hepforge.orghttp://fastjet.frhttp://rivet.hepforge.org


Rivet



RivetRobust Independent Validation of Experiment and Theory:

I generator-agnostic analysis framework,co-evolved with fastjet

I reads input from Monte Carlo generator(from file/FIFO)

I runs one (or more) analyses on the input data

I produces plots corresponding to available measurementswith comparison MC/data

I distributed with (validated) analyses and corresponding data

in the following: focus on concepts

detailed documentation:Rivet user manual, tutorials


https://arxiv.org/abs/1003.0694


HepMC

need interface between generator and analysis:

I HepMC: standardized event recordcontains final state particles(with Monte Carlo information)

I direct output of (most) modern generators,for others: interface software (e.g. Agile)

de-facto standard for generator output(still evolving, e.g. for heavy-ion)

[http://lcgapp.cern.ch/project/simu/HepMC/]


http://lcgapp.cern.ch/project/simu/HepMC/


Rivet architecture

I frameworkI loading and steering of analysisI input/output of data and resultsI analysis tools, including interface to fastjetI plotting (now based on YODA, previously AIDA)

I library of analysis pluginsI about 300 distributed with Rivet (as of version 2.3.0)I new ones can be added locally

or contributed (once mature enough)

user interaction with Rivet through executables:

I rivet

I rivet-mkhtml, rivet-mkanalysis

I rivet-cmphistos

I make-plots



Rivet analysisRivet analysis comprisesI ID I

I title

I description:short summary of the analysis contentcollision system and energiesreference to paperstatus: preliminary, unvalidated, validated, obsolete, . . .

I algorithm:code to reproduce the analysis on Monte Carlo output

I data:measured data points and uncertainties

e.g. look at:rivet --show-analysis ALICE_2011_S8945144



Rivet implementationI user can easily provide additional analyses by

implementing C++ class inheriting from Rivet::Analysis

I one can start from a templaterivet-mkanalysis MC_MY_ANALYSIS

I requires implementation ofI init

definition of projections (final states), booking of histograms, ...I analyze

per-event analysis (whatever it may be), histogram filling, . . .I finalize

typically normalisation to cross section, . . .

I analysis built byrivet-buildplugin MC_MY_ANALYSIS.cc

I analyses loaded from library and run byrivet -a MC_MY_ANALYSIS output.hepmc



projections

abstraction of common physical entities projections(avoid repetition of computing-intense calculations)

I charged particlesconst ChargedFinalState cfs(-3.2, 3.2);

addProjection(cfs, "CFS");

I charged jetsFastJets fj(cfs, FastJets::ANTIKT, 0.4);

addProjection(fj, "Jets");

analysis typically iterates on such projections:const FinalState &cfs =

applyProjection (evt, "CFS");

foreach (const Particle &p, cfs.particles()) {

// fill histogram

}



histograms

I histogramming based on external package (YODA)

I book histogram manually:_hist = bookHisto1D("pt_h", 10, 0., 20.,

"ch. part.",

"$p_\\perp$", "counts");

I book histogram according to reference histogram:_hist = bookHisto1D("d01-x01-y01")

name corresponding to the one from HepData

I fill histograms with event weight:const double weight = event.weight();

_hist->fill(phi, weight);

for proper uncertainties (and merging)



HepData interaction

I for published analyses:values can usually be obtained from HepData

I natural integration with HepDatauseful for retrieval of experimental data

I for published analysis (e.g. ALICE UE):rivet-mkanalysis ALICE_2011_I1080735

creates template and downloads data files

I for unpublished analysis:reference data must be fed manually



example

namespace Rivet {

class MC_MY_ANALYSIS : public Analysis {

public:

MC_MY_ANALYSIS() : Analysis("MC_MY_ANALYSIS") { }

void init() {

const ChargedFinalState cfs(-.9, .9);

addProjection(cfs, "CFS");

const FinalState fs_jet(-1., 1.);

addProjection(fs_jet, "FS_JET");

FastJets fj02(fs_jet, FastJets::ANTIKT, .2);

addProjection(fj02, "Jets02");

_h_pt_jet02 = bookHisto1D("pt_jet", 20, 0., 100.);

}

...



example (cont’d)void analyze(const Event& event) {

Jets alljets = applyProjection(event, "Jets02").

jetsByPt(0., 200., -.8, .8);

foreach (const Jet &jet, alljets)

_h_pt_jet02->fill(jet.momentum().pT(), event.weight());

}

void finalize() {

const double fac =

crossSection() / (millibarn * sumOfWeights() * 2 * pi);

scale(_h_pt_jet02, fac);

}

private:

Histo1DPtr _h_pt_jet02;

DECLARE_RIVET_PLUGIN(MC_MY_ANALYSIS);

}

}



mcplots

I availability of Rivet analyses allows forautomatized creation of comparison plots

I large number of generators and tunes

I mcplots project hosted at CERNI generate events

using voluntary computing (Test4Theory, LHC@home)I run analysesI provide interface to plot comparison

of selected generators/tunes to data

I good overview of quality of description by MC generators

extensive source of comparison plots:https://mcplots.cern.ch

[http://arxiv.org/abs/1306.3436]


https://mcplots.cern.chhttp://arxiv.org/abs/1306.3436


pp physics



applications for physics

simultaneous comparison to many observables,some interesting ones are:

I jets (spectra, fragmentation, shapes, . . . )

I identified particle spectra

I multiplicity distributions

I underlying event

I . . .

many of these plots are available from mcplots

wide-spread possibilities,biased selection of examples


https://mcplots.cern.ch


jets

(leading jet) [GeV]perp

p200 400 600 800

[pb/

GeV

]pe

rpds

igm

a/dp

-110

1

10

210

310

410

510

610ATLASHerwig++ (Def)Pythia 6 (Def)Pythia 8 (Def)Sherpa (Def)

7000 GeV pp Jets

mcp

lots

.cer

n.ch

100

k ev

ents

≥R

ivet

2.3

.0,

Herwig++ 2.7.1, Pythia 6.428, Pythia 8.212, Sherpa 2.2.0

ATLAS_2011_S9128077

(0.4))T

Jet Transverse Momentum (anti-k

200 400 600 8000.5

1

1.5Ratio to ATLAS

[mcplots.cern.ch]

I spectra [arXiv:1107.2092]I crucial check of hard interactionsI also critical as background

I fragmentation [ATLAS-CONF-2010-049]

I longitudinal momentum fractionprobes parton showerand hadronisation

I transverse shape [arXiv:1101.0070]I radial dependence

probes transverse evolutionI mostly sensitive to shower


http://arxiv.org/abs/1107.2092http://arxiv.org/abs/1101.0070


jets

{track}textz0.2 0.4 0.6 0.8 1

{tra

ck}

text

/dz

{tra

ck}

text

)dN

{jet}

text

(1/N

0

2

4

6

8

10

12

14

ATLASHerwig++ (Def)Pythia 6 (Def)Pythia 8 (Def)Sherpa (Def)

7000 GeV pp Jets

mcp

lots

.cer

n.ch

1M

eve

nts

≥R

ivet

2.3

.0,


ATLAS_2010_CONF_2010_049

= 4-6 GeV/c)T

(0.6), pT

Fragmentation (anti-k

0.2 0.4 0.6 0.8 10.5

1

1.5Ratio to ATLAS

[mcplots.cern.ch]









jets

r0 0.2 0.4 0.6

rho(

r)

0

1

2

3

4

5

6


7000 GeV pp Jets

mcp

lots

.cer

n.ch

300

k ev

ents

≥R

ivet

2.3

.0,


ATLAS_2011_S8924791

(r/R)ρDifferential jet shape

0 0.2 0.4 0.60.5

1

1.5Ratio to ATLAS

[mcplots.cern.ch]









identified particle spectra

[GeV/c]perp

p0 1 2

-1 [G

eV/c

]pe

rpN

/dy/

dp2

d{e

vent

}te

xt1/

N

-210

-110

1

10

210

ALICEHerwig++ (Def)Pythia 6 (Def)Pythia 8 (Def)Sherpa (Def)

900 GeV pp Soft QCD

mcp

lots

.cer

n.ch

100

k ev

ents

≥R

ivet

2.3

.0,


ALICE_2011_S8945144

| < 0.5)η) (|+πpT(

0 1 20.5

1

1.5Ratio to ALICE

[mcplots.cern.ch]

I spectra of [arXiv:1101.4110]I pionsI kaonsI protons

I contributions at low and high p⊥

I important to constrainfragmentation and hadronisation

I collective effects (HI-like)modify spectra


http://arxiv.org/abs/1101.4110



[GeV/c]perp

p0 0.5 1 1.5 2

-1 [G

eV/c

]pe

rpN

/dy/

dp2

d{e

vent

}te

xt1/

N

-310

-210

-110

1 ALICEHerwig++ (Def)Pythia 6 (Def)Pythia 8 (Def)Sherpa (Def)

900 GeV pp Soft QCD

mcp

lots

.cer

n.ch

100

k ev

ents

≥R

ivet

2.3

.0,


ALICE_2011_S8945144

| < 0.5)η) (|+pT(K

0 0.5 1 1.5 20.5

1

1.5Ratio to ALICE

[mcplots.cern.ch]









[GeV/c]perp

p0.5 1 1.5 2

-1 [G

eV/c

]pe

rpN

/dy/

dp2

d{e

vent

}te

xt1/

N

-310

-210

-110

1


900 GeV pp Soft QCD

mcp

lots

.cer

n.ch

100

k ev

ents

≥R

ivet

2.3

.0,


ALICE_2011_S8945144

| < 0.5)ηpT(p) (|

0.5 1 1.5 20.5

1

1.5Ratio to ALICE

[mcplots.cern.ch]








multiplicity distributions

{ch}textN0 20 40 60

{ch}

text

text

dN/te

xtdN

-610

-510

-410

-310

-210

-110

1

10


7000 GeV pp Soft QCD

mcp

lots

.cer

n.ch

1M

eve

nts

≥R

ivet

2.3

.0,


ALICE_2010_S8625980

)T

| < 1.0, all pη > 0, |ch

Charged multiplicity (N

0 20 40 600.5

1

1.5Ratio to ALICE

[mcplots.cern.ch]

I charged particles [arXiv:1004.3514]around mid-rapidity

I addresses question ofparticle production

I differently well describedby Monte Carlo generators

I comparison toI different generatorsI different tunes of PYTHIA 8




multiplicity distributions

{ch}textN0 20 40 60

{ch}

text

text

dN/te

xtdN

-410

-310

-210

-110

1ALICEPythia 8 (Def)Pythia 8 (4C)Pythia 8 (4Cx)Pythia 8 (A2)Pythia 8 (Monash 13)

7000 GeV pp Soft QCD

mcp

lots

.cer

n.ch

1M

eve

nts

≥R

ivet

2.3

.0,

Pythia 8.212

ALICE_2010_S8625980

)T

| < 1.0, all pη > 0, |ch

Charged multiplicity (N

0 20 40 600.5

1

1.5Ratio to ALICE

[mcplots.cern.ch]

I charged particles [arXiv:1004.3514]around mid-rapidity

I addresses question ofparticle production

I differently well describedby Monte Carlo generators

I comparison toI different generatorsI different tunes of PYTHIA 8




underlying event

(leading track) [GeV]perp

p0 5 10 15 20

/det

adph

iGT

{chg

}te

xt N2

LTd

0

0.5

1

1.5

2

2.5

3

3.5


7000 GeV pp Underlying Event

mcp

lots

.cer

n.ch

100

k ev

ents

≥R

ivet

2.3

.0,


ATLAS_2010_S8894728

> 0.5 GeV/c)T

| < 2.5, pηAverage Charged Particle Density (TWRD) (|

0 5 10 15 200.5

1

1.5Ratio to ATLAS

[mcplots.cern.ch]

I UE contributes to particle yieldcritical for proper description

I particle yield [arXiv:1012.0791]in regions w.r.t. trigger particle

I towardsI awayI transverse

I ALICE results [arXiv:1112.2082]not in Rivet yet




underlying event

(leading track) [GeV]perp

p0 5 10 15 20

/det

adph

iGT

{chg

}te

xt N2

LTd

0

0.5

1

1.5

2

2.5

3


7000 GeV pp Underlying Event

mcp

lots

.cer

n.ch

100

k ev

ents

≥R

ivet

2.3

.0,


ATLAS_2010_S8894728

> 0.5 GeV/c)T

| < 2.5, pηAverage Charged Particle Density (AWAY) (|

0 5 10 15 200.5

1

1.5Ratio to ATLAS

[mcplots.cern.ch]








underlying event

[mcplots.cern.ch]








usage of Rivet-published datasome examples of Rivet usage by phenomenology community:I Systematic event generator tuning for the LHC

[0907.2973]

I Constraining MPI models using σeffand recent Tevatron and LHC Underlying Event data[1307.5015]

I Hydrodynamics and Jets in Dialogue[1407.1782]

I Origins of the di-jet asymmetry in heavy ion collisions[1512.08107]

I Sensitivities to PDFsin parton shower MC generator reweighting and tuning[1601.04229]

I many more,e.g. more than 200 Rivet-citing papers on inspire


http://arxiv.org/abs/0907.2973http://arxiv.org/abs/1307.5015http://arxiv.org/abs/1407.1782http://arxiv.org/abs/1512.08107http://arxiv.org/abs/1601.04229http://inspirehep.net/search?ln=en&ln=en&p=refersto%3Arecid%3A847552&of=hb&action_search=Search&sf=earliestdate&so=d&rm=&rg=25&sc=0


ALICE analysesrivet --list-analyses ALICE_

ALICE 2010 S8624100 Charged particle multiplicities at 0.9 and 2.36 TeV inthree different pseudorapidity intervals

ALICE 2010 S8625980 Pseudorapidities at three energies, charged multiplicityat 7 TeV

ALICE 2010 S8706239 Charged particle 〈pT〉 vs. Nch in pp collisions at 900GeV

ALICE 2011 S8909580 Strange particle production in proton-proton collisionsat

√s = 0.9 TeV with ALICE at the LHC

ALICE 2011 S8945144 Tranverse momentum spectra of pions, kaons and pro-tons in pp collisions at 0.9 TeV

ALICE 2012 I1181770 Measurement of inelastic, single- and double-diffractioncross sections in protonproton collisions at the LHCwith ALICE

ALICE 2014 I1300380 Production of Σ(1385)± and Ξ(1530)0 in proton-proton collisions at

√s = 7 TeV

ALICE 2015 I1357424 Tranverse momentum spectra of pions, kaons and pro-tons in pp collisions at 7 TeV



prospects for ALICE

I ALICE covers complementary aspects in pp physics(low p⊥ tracking, particle identification, . . . )

I relevant analysesI jet shapesI identified particles in jetsI underlying eventI strangeness productionI multiplicity dependenceI . . .

I on-going efforts to publish Rivet plugins for more analyses,both on-going and already published ones



tuning



tuningI so far considered MC generator and tune as given

I large parameter spaceat least for phenomenological descriptions

I parameters need to be tuned to experimental dataI manually by eye

good control of changesbut needs understanding of model and parameter effect

I brute-force scan of parameter spaceinfeasible for more than a few parameters

I some commonly used tunesI Pythia 6: Perugia 0, 2011, 2012 (manual)I Pythia 8: Monash 2013 (manual)I . . .

Professor [arXiv:0907.2973]


https://arxiv.org/abs/0907.2973


Professor

I problem: MC performance depends on choice of parameters

I goal: find optimal vector of parameters (P)

I criterion: minimal Goodness-of-Fit for deviationbased on comparison using Rivet analyses for judgement

I but: numerical minimisation requires many evaluations of GoF,each evaluation requires large generator run

with PROcedure For EStimating Systematic errORs(long form abandoned):

I parameterize per-bin responseby sampling of parameter space

I minimize GoF based on parameterized response

I validation



Professor tune - exampleI first production tuning using Professor:

systematic tune of Pythia 6.418 [0907.2973]

I tune in stages:I start from default valuesI tune flavour parameters

using precision data from LEP/SLACI tune shower and (non-flavour) hadronisation parameters

using LEP dataI tune UE and MPI parameters

using data from CDF and D0

I tune verificationby comparing parameterized and actual deviationsaround optimal parameters

established systematic tunealso as starting point for other tunes




impact on tuning

I usage of Rivet analysis much less error-pronethan re-implementation of analysis from paper

I results published for Rivet are used preferentiallyfor (automatized) comparison with generators

I Rivet analyses are typically used for tunes,either manually or automatized

I deviations from Rivet analyses can be spotted easilyand by everyone (e.g. on mcplots) good monitoring of MC performance and sensitivity

strong motivation for Rivet publicationsin order for the data to be used



heavy-ion



heavy-ion analyses

heavy-ion (motivated) analyses often differ from HEP:I classification of events

I centrality (impact parameter)I event planeI event shapeI . . .

I “event-global” analyses, e.g.I correlations in ∆ϕ, ∆η, p⊥, . . .

typically using event mixing for correctionsI fluctuationsI . . .

I post-processingI comparison Pb–Pb to ppI yield extractionI . . .



heavy-ion generators

generators focus on some aspects

I HIJING [nucl-th/9502021]I multi jet production

I AMPT [nucl-th/0411110]I HIJING for initial conditionsI parton cascadeI relativistic transport

I EPOS [hep-ph/0007198]I parton laddersI collectivity

jet quenching:I JEWEL [1311.0048]

I hard interaction fromPYTHIA

I parton showerwith medium interaction energy loss

I PYTHIA for hadronisation

I YaJEM [1009.3740]I parton shower with energy loss

no general-purpose heavy-ion generatornot all provide HepMC final state for usage with Rivet


http://arxiv.org/abs/nucl-th/9502021http://arxiv.org/abs/nucl-th/0411110http://arxiv.org/abs/hep-ph/0007198http://arxiv.org/abs/1311.0048http://arxiv.org/abs/1009.3740


prospects for heavy-ion analyses

I there is not one universal Monte Carlo generatorfocus on some aspects of heavy-ion physics

I Monte Carlo comparisons of heavy-ion generatorsnot (yet) as systematic and global as for HEP

I aim to do more global comparisons. . .



(current) limitations

Rivet was developped for HEP needs, not heavy-ions,e.g. post-processing of results not foreseen:

I combination of results from different collisions systems,e.g. Pb–Pb / pp (like RAA)

I fitsno integrated possibility for fitting,e.g. for determination of peak width

I yield extractionextraction of particle yield e.g. from invariant mass distributionrequires fitting or bin counting and background subtraction

not all analyses are straight-forward to implement,some technical aspects in Rivet need to be addressed



summary



summary

I Monte Carlo tools valuable tool for physics interpretation

I reliable comparisons of Monte Carlo generators to data needed,experiments providing Rivet analyses

I Rivet as a tool toI archive and publish measurements,

i.e. analysis and dataI compare different event generators

increasing efforts in ALICEto contribute Rivet analyses

both for pp and p–Pb/Pb–Pb


Backup


ALICE detector


ALICE coverage


related tools

I Agile, Sacrificehttp://agile.hepforge.org/

I YODAhttp://yoda.hepforge.org/

I AIDAhttp://aida.freehep.org/

I fastjethttp://fastjet.fr


http://agile.hepforge.org/http://yoda.hepforge.org/http://aida.freehep.org/http://fastjet.fr

Rivetpp physicstuningheavy-ionsummaryAppendix

Rivet - and (my) prospects for ALICE · 2016. 3. 22. · Rivet and (my) prospects for ALICE Jochen Klein CERN Seminar Bologna, Italy March 21st, 2016 This is an output file created

Documents