Mad Graph Mad Event Automated TreeLevel Feynman Diagram

  • Slides: 42
Download presentation
Mad. Graph + Mad. Event + Automated Tree-Level Feynman Diagram, Helicity Amplitude, and Event

Mad. Graph + Mad. Event + Automated Tree-Level Feynman Diagram, Helicity Amplitude, and Event Generation Tim Stelzer Fabio Maltoni

Exciting Time • Data from Tevatron now! – Gain experience with real data and

Exciting Time • Data from Tevatron now! – Gain experience with real data and analysis now. • Data from LHC soon! – Develop tools/skills necessary to get most out of LHC data when it comes on line • You will be doing the analysis for the next great discoveries! – Strong links between theory and experiment will be essential 10

Outline • Goal: Improve Mad. Graph/Mad. Event – Only 1 thing more useful •

Outline • Goal: Improve Mad. Graph/Mad. Event – Only 1 thing more useful • Introduction • Signal pp > h > tt~ bb~ – – – Generating/Checking Processes Calculating/Checking Cross Sections Plotting/Analyzing Events Running on your machines Decaying Particles • Backgrounds pp> tt~bb~ • Remarks 13

Standard Model News • Good News! – Every prediction is correct! rd Standa l�

Standard Model News • Good News! – Every prediction is correct! rd Standa l� e Mod • Bad News : ( – Limited ability to “solve” – QCD is problem! • Miracle that can predict so much 15

Parton Distribution Functions (Measured) Evolution +Splitting Hard Scattering Showering Fragmentation Hadronization e+e 18

Parton Distribution Functions (Measured) Evolution +Splitting Hard Scattering Showering Fragmentation Hadronization e+e 18

Factorization • Key Idea – Result = PDF * Evolution * hard_scatter *showering *

Factorization • Key Idea – Result = PDF * Evolution * hard_scatter *showering * hadronization – Each piece is “independent” • Really? – Soft gluons have “long” wavelength, basically can’t “see” hard scattering • Uncertainty – – PDF Factorization Scale Renormalization Scale Showering Scales 21

HERWIG + PYTHIA • Contain everything for many processes • Higher energy collisions require

HERWIG + PYTHIA • Contain everything for many processes • Higher energy collisions require new “Hard” processes • Large amount of time/talent devoted to event generation---Need to streamline! 25

Options • Prepackaged event generators – Fixed number of processes implemented – Usually well

Options • Prepackaged event generators – Fixed number of processes implemented – Usually well tested and packaged – Alpgen---Tree Level – MC@NLO --- One Loop • Create your own generator – “Any” process of interest – Comp. HEP --- trace – Mad. Graph/Mad. Event --- Helicity Amplitude 28

Big Picture • User Requests: – pp -> e+ ve jjj – QCD Order

Big Picture • User Requests: – pp -> e+ ve jjj – QCD Order = 3 – QED Order =2 Show Demo • Mad. Graph/Mad. Event Returns: – List all contributing sub processes – Feynman diagrams for each sub process – Self-Contained Fortran Package • Integrates cross section • Generates unit-weighted events • Les Houches / HERWIG ready 33

Steps • Generate Subprocesses/Diagrams • Calculate Benchmark Cross Section • Download to your machine

Steps • Generate Subprocesses/Diagrams • Calculate Benchmark Cross Section • Download to your machine – Events – Event Generator • Perform Analysis 35

Activity 1: Generate pp > h >ttbb~ Generate Process • Predict subprocess/diagrams – e.

Activity 1: Generate pp > h >ttbb~ Generate Process • Predict subprocess/diagrams – e. g. uu~ > bb~ tt~ Benchmark Cross section • Open http: //madgraph. hep. uiuc. edu/ Download Analysis • Sign up for account – Username: madgraph – Password: guest • Generate Process – pp > h > ttbb~� QED=2 QCD=2 • Verify Subprocesses and Diagrams 45

Activity 2 Cross Section + Plots Generate Process • Results and Event Database –

Activity 2 Cross Section + Plots Generate Process • Results and Event Database – Check Cross Section – Download events. dat – Plot events on your machine Benchmark Cross section Download Analysis 6 1 0. 3921316 E-04 0. 9118800 E+02 0. 7546772 E-02 0. 1270000 E+00 -2 2 6 -6 5 -5 0 0 1 1 0 0 2 2 0 502 0 503 0 501 0 503 -1 -1 1 1 -1 -1 1 0. 25012024368 E+02 0. 00000000000 E+00 0. 25012024368 E+02 2 0. 47489237575 E+04 0. 00000000000 E+00 -0. 47489237575 E+04 3 0. 32457365210 E+04 0. 35739564193 E+02 -0. 83428604245 E+02 -0. 32397820027 E+04 4 0. 57267081534 E+03 -0. 25236886174 E+02 0. 96458748768 E+02 -0. 53631164644 E+03 5 0. 54994426015 E+03 0. 31023829244 E+02 -0. 53539395070 E+02 -0. 54643174731 E+03 6 0. 40558418542 E+03 -0. 41526507264 E+02 0. 40509250547 E+02 -0. 40138633668 E+03 � 05

Let’s go do it!

Let’s go do it!

Generating Events Generate Process Benchmark Cross section Download Estimate Uncertainty Analysis • Mad. Event

Generating Events Generate Process Benchmark Cross section Download Estimate Uncertainty Analysis • Mad. Event “Automatically” Integrates cross sections and generates unweighted events. • Understanding how, can provide important information. • A mini lesson on Monte Carlo 13

Monte Carlo Integration • Advantages – – Large numbers of dimensions Complicated cuts ONLY

Monte Carlo Integration • Advantages – – Large numbers of dimensions Complicated cuts ONLY OPTION Event generation • Limitations – Only works for function f(x) ≈ 1 16

Example: u ve -> g u ve • For strong cuts, |a 1+a 2|2

Example: u ve -> g u ve • For strong cuts, |a 1+a 2|2 ≈ 1 (Pt > 90, DR > 1. 0) 18

Adaptive M. C. (VEGAS) • Advantages – Grid adjusts to numerically flatten peaks –

Adaptive M. C. (VEGAS) • Advantages – Grid adjusts to numerically flatten peaks – Flexible • Limitations – Adjusting grid takes time – Peaks must lie on integration variable 22

Multi-Channel M. C. • Advantages – Allows for more complicated peaks – NEXTCALIBUR, WHIZARD

Multi-Channel M. C. • Advantages – Allows for more complicated peaks – NEXTCALIBUR, WHIZARD • Limitations – Need to calculate all gi values for each point. (slow) – Each phase space channel must be invertible – N coupled equations for a so only works for small number of channels. 25

Single Diagram Enhanced Mad. Event • Key Idea – Any single diagram is “easy”

Single Diagram Enhanced Mad. Event • Key Idea – Any single diagram is “easy” to integrate – Divide integration into pieces, based on diagrams • Get N independent integrals – – Errors add in quadrature so no extra cost No need to calculate “weight” function from other channels. Can optimize # of points for each one independently Parallel in nature • What about interference? – Never creates “new” peaks, so we’re OK 30

Mad. Event Example Vector Bosons 31

Mad. Event Example Vector Bosons 31

Mad. Event Example Heavy Quarks 32

Mad. Event Example Heavy Quarks 32

Not only gets an answer, gets the correct one! 35

Not only gets an answer, gets the correct one! 35

More Checks…. .

More Checks…. .

SDE Integration • It works! • Interference is included • Looking at diagrams contributions

SDE Integration • It works! • Interference is included • Looking at diagrams contributions – Gives information about distributions – Gives checks on integration (symmetry) 37

Activity 3 Run Check using SDE Generate Process Benchmark Cross section • Check Cross

Activity 3 Run Check using SDE Generate Process Benchmark Cross section • Check Cross Section Details – (10 minutes) – Show example Download Analysis • Keep going…. 47

Running at Home • Web Interface Provides – Intuitive Interface ? – Subprocesses and

Running at Home • Web Interface Provides – Intuitive Interface ? – Subprocesses and Diagrams – Basic results and events • Download Process to – – Customized cuts Large/Optimized runs Own the code! Portability 50

Activity 4: Run on your machine • Download the source code Generate Process –

Activity 4: Run on your machine • Download the source code Generate Process – madevent. tar. gz Benchmark • Edit important files – Source/setparam. f Cross section Download Analysis • Collider, Scale and PDF information – Sub. Processes/run_cuts. dat • Standard cuts • Generate events – bin/generate_events 5000 0 – Or use bin/survey and bin/refine Work/Break start again at 4: 30

Estimating Uncertainty Generate Process Benchmark Cross section Download Analysis • Sources NOT 0. 965!!

Estimating Uncertainty Generate Process Benchmark Cross section Download Analysis • Sources NOT 0. 965!! – PDF – Factorization Scale – Hard-Scattering (loops) • as (renormalization scale) 38

Decay_Events Generate Process Benchmark Cross section Download Analysis • Mad. Event has decay utility

Decay_Events Generate Process Benchmark Cross section Download Analysis • Mad. Event has decay utility – Includes spin information – Keeps correct cross section – Includes Les. Houches info for passing to HERWIG/Pythia** 44

Activity 5: Decay top • Decay t -> bjj – What is cross section

Activity 5: Decay top • Decay t -> bjj – What is cross section – Plot jj mass • Decay t~ > b~ lv – What is cross section – Plot lv mass 58

Backgrounds • “Backgrounds are measured not calculated…. . ” Tony Liss • Assignment: Write

Backgrounds • “Backgrounds are measured not calculated…. . ” Tony Liss • Assignment: Write down some backgrounds for tt. H. 01

Activity 6: Backgrounds • • • Predict Subprocesses/diagrams Generate Diagrams Check results Download Events

Activity 6: Backgrounds • • • Predict Subprocesses/diagrams Generate Diagrams Check results Download Events Determine cross section – 90 < mbb < 150 20

Mad. Event Summary • Advantages – Can handle “any” problem – Web interface is

Mad. Event Summary • Advantages – Can handle “any” problem – Web interface is convenient/fast – Expandable to MSSM (out this year) – Many Checks – Works for Theorists and Experimentalists • Limitations – W+5 jets (7, 000 diagrams w/ hundreds subprocesses) – No loops (yet…) 25

Background pp>tt~bb~ QED • Generate Diagrams – Pp > tt~bb~ /h tells it not

Background pp>tt~bb~ QED • Generate Diagrams – Pp > tt~bb~ /h tells it not to allow Higgs • Check results – dd~ = d~d – uu~ = u~u <> dd~ – gg • – Plots (find Z peak)

Status • http: //madgraph. physics. uiuc. edu – 500 MHz Pentium w/ 40 Gigabyte

Status • http: //madgraph. physics. uiuc. edu – 500 MHz Pentium w/ 40 Gigabyte disk – Creates Event Generators to download – Lots of “hits” on cite • FNAL, CERN, DESY, U. W. , U. Chicago…. . • CDF use – W/Z + jets analysis – Photon + jets analysis – Virtual Collider • Future looks promising!

Vision • Add Web based Integration – User inputs cuts/parameters – Code runs in

Vision • Add Web based Integration – User inputs cuts/parameters – Code runs in parallel on cpu farm – Returns cross section and events • Advantages – – Reduces overhead to getting results Reduce computing cost by load leveling Events can easily be shared/stored Centralize maintenance • Source Code will ALWAYS be Available

To Do: • Add Computers • Add MSSM to Mad. Graph II • Improve

To Do: • Add Computers • Add MSSM to Mad. Graph II • Improve Speed – Phase Space Grid Guess – Helicity Sums – Color Sums • Add Factorization – 10 jet production

Signal pp>h>tt~bb~ Generate Process Benchmark Cross section Download Estimate Uncertainty Analysis • http: //madgraph.

Signal pp>h>tt~bb~ Generate Process Benchmark Cross section Download Estimate Uncertainty Analysis • http: //madgraph. hep. uiuc. edu/ • Process= pp > h > tt~ bb~ – QCD = 2 QED = 2 • Username = yeti • Password = yeti

Signal pp>h>tt~bb~ Generate Process Benchmark Cross section Download Estimate Uncertainty Analysis • Generation –

Signal pp>h>tt~bb~ Generate Process Benchmark Cross section Download Estimate Uncertainty Analysis • Generation – setcuts. f (run_cuts. dat) • Et, DR, Mij – setparams. f • PDF, Scales, Collider • Analysis – plot_events. f • Easy secondary cuts – Decay • Decay unstable particles in events. dat

Activity 4: Download and Run Event Generator Generate Process Benchmark Cross section Download Analysis

Activity 4: Download and Run Event Generator Generate Process Benchmark Cross section Download Analysis • Download Code madevent. tar. gz csh> mkdir pp_h_ttbb csh > cd pp_h_ttbb csh > mv. . /madevent. tar. gz. csh> gunzip madevent. tar. gz csh> tar –xvf madevent. tar csh> emacs Sub. Processes/run_cuts. dat csh> emacs Source/setparam. f csh> bin/generate_events

Background • Download events • Select cuts to improve Signal/Background – 90 < Mbb

Background • Download events • Select cuts to improve Signal/Background – 90 < Mbb < 150 • emacs Source/plot_events. f

Ideas From Tony • Technical How to run – Parameters – Uncertainties pdf, systematic

Ideas From Tony • Technical How to run – Parameters – Uncertainties pdf, systematic etc – Limitations/assumptions • Backgrounds don’t come from MC come from Data – Built tools, not analysis – Join CDF • Matrix Element and Shower+Matching • Different generators, LO vs NLO