Beam Delivery Simulation � LHC Studies � L. Nevay , J. Snuverink, S. Boogert, � H. Garcia-Morales, S. Gibson, L. Deacon � R. Kwee-Hinzmann, S. Walker, A. Abramov � � 10 th December 2015 �
Beam Delivery Simulation - BDSIM � • Tracking code that uses Geant4 � • Open source C++ � • Automatically builds Geant4 model � • Uses MadX-like syntax for text input � • Mixes normal accelerator tracking & Monte Carlo particle physics � • Full showers of secondaries created by Geant4 processes � • Ability to simulate synchrotron radiation � BDSIM accelerator � • Simulate energy deposition and detector backgrounds � • Ability to import external geometry and LHC dipole � field maps � 2
Geometry & Tracking – Geant4 � • Write a C++ program to build geometry, generate particles, set physics models, record output. � • Compile and run program generating N events � ― Either as a command line program or interactive gui. � • Class library – you must write your own program � 3
Hello World Accelerator � 4
BDSIM Development � • BDSIM started ~2002 by G. Blair at RHUL � • BDSIM heavily developed since 2013 for LHC � • Complete review, refactorisation and moderisation � • Recent development followed 3 main themes: � Also: � • Geometry � • LHC Specific developments � • Documentation & general development � • Tracking � • Analysis tools & workflow � � • Physics processes � 5
Code & Software Develoment � • 60 000 lines of C++ � • Revised class hierarchy & factory patterns � • Increased factorisation – much easier to extend � • Consolidation of various development branches � • C++ 11 adoption & latest versions of Geant4, ROOT, CLHEP, � • Parser significantly improved by J. Snuverink � ― many unusual memory leaks, and problems fixed � ― highly object-orientated � • CTest test suite, CMake build system � ― much easier to use as compared to old configuration scripts � • CDash nightly and on demand automated building & testing � • Issue tracking & reporting � • Built in configuration for AFS � • Automated manual updates � • Regular release cycle � 6
Regression Testing � • Rapid development of BDSIM � • Occasionally, simple / basic things break � • Code too large to test all features yourself � • Automated build & testing system implemented � • Each example is also a test � • Reference histograms and results compared � • robdsim analysis tool used for comparison and testing � • 145 tests so far � • Runs nightly � • Hadronic & EM Shower development � • Tracking in each component � • Parser � • Geometry construction � • Geometry overlaps � • Many more…. � 7 EM Shower in Collimator �
Python Utilities Galore � • pymadx, pybdsim, pymad8, robdsim, pytransport, pylhc � • pymadx � ― loading and manipulation of Madx TFS files � ― range iterating, filtering, matching � ― PTC segments supported � ― use to plot a lattice above a graph – interactive too! � • pybdsim � ― conversion from Madx, Mad8, Transport etc � ― ASCII output analysis � ― programmatic model construction � • pylhc � ― utilities for parsing lhc model specific information � ― collimation files, aperture information (filtering, matching etc) � • Again, all open source and distributed with BDSIM � 8
Documentation � • New manual (html & pdf) automatically updated weekly � ― lots of syntax examples � ― www.pp.rhul.ac.uk/bdsim/manual � • Detailed Doxygen code documentation similarly � ― www.pp.rhul.ac.uk/bdsim/doxygen � 9
Public Git Access � • www.bitbucket.org/jairhul/bdsim � 300 – 500 commits per version � • Full open source development � 3 releases per year typically � • Issue tracking - (100 this year, 20 open) � • ~ 10 regular developers � • ~ 5 branches � Many developers working � at once without issue on � many versions � 10 A successful git branching model �
Geometry � • Previous geometry relatively simple cylinders � ― Adequate for conceptual studies � ― Great detail required for real machines � • Main geometry library rewritten � • Extensive use of factory pattern � ― Each factory represents a style and can make every type of say magnet � • 8 different aperture types (including detailed LHC) � • 6 different magnet styles (again with LHC style) � • 4 different tunnel styles � ― can generically follow the beam line � ― will be able to have external geometry and customise for certain ranges � • Most importantly all geometry works together � • Any beam pipe will work with any magnet! � • Very simple to extend with new geometry � ― guaranteed to work with all magnets � • BDSIM design provides this flexibility not inherent to Geant4 � 11
Geometry � • 8 aperture models � ― circular, rectangular, elliptical, lhc (detailed), rectellipse, racetrack, octangonal � • Modelled on MadX aperture parameterisation � • Works with any other geometry � • 6 different magnet styles � Circular � LHC screen � SRF Cavities � ( S. Walker ) � LHC Style � RectEllipse � Poles circular yoke � Rectangular � / square � LHC detailed � Elliptical � Poles square yoke � 12
Tunnel Geometry � • Was only partially implemented previously � • Rewritten using factories � • Currently 4 different styles � • Can automatically follow beam line � • Can describe different styles for different sections* � • Can use external geometry for sections* � LHC arc before IP1 � 13 * under testing �
External Geometry? � • For when the generic components just won’t suffice � • Can import external geometry � ― SQL, Mokka, GDML, STL � • Can also overlay field maps and interpolate � ― 2D, 3D, etc. � • You can also export to GDML from BDSIM! � GDML LHCb � SQL Mokka example � 14
Tracking � • Quantitative comparison with PTC & SixTrack underway � • Very good agreement with PTC � ― Tracking & optical function calculation � • Factorising tracking into library � ― will reduce tracking time by order of magnitude for large machines � ― Will allow choice of integrators � A. Abramov � ― Will be able to use other tracking libraries shortly � ― Expected complete early 2016 � × 10 − 7 × 10 − 9 1 . 5 8 180 54 160 48 140 Residuals x (m) Residuals y (m) 42 120 36 Counts Counts 0 . 0 100 30 0 80 24 60 18 40 12 20 6 − 2 . 0 − 8 − 0 . 0006 0 . 0000 0 . 0006 − 0 . 00004 0 . 00000 0 . 00004 x(m) y(m) × 10 − 8 × 10 − 9 3 3 80 225 70 200 Residuals xp (rad) Residuals yp (rad) 60 175 50 150 Counts Counts 0 40 0 125 100 30 75 20 50 10 25 − 3 − 3 − 0 . 00010 0 . 00000 0 . 00015 − 0 . 00002 0 . 00000 0 . 00002 xp(rad) yp(rad) 15 Double Bend Achromat agreement with PTC �
Direct Injection � • Had the ability to read out in curvilinear coordinates – now in too � • Introduced ability to inject particles anywhere in lattice � • Any beam distribution as function of S � ― Interpolation of trajectory within arcs � ― Efficient look up of transforms � • Sixtrack loader written by R. Kwee � • Can therefore convert SixTrack hits to energy deposition � exits on 0,0 � reference particle � starts here � 16
Physics & Processes � • Benefit from regular Geant4 updates to many models � ― support latest Geant4 and one previous version � • Moved entirely to range cuts � ― thorough testing and many small bugs addressed � • Cut particles not on energy but on range to produce a secondary particle � • Much more accurate stopping location � ― and therefore energy deposition � • Improved physics accuracy for lower CPU usage � • Modular physics list implemented in Geant4 � ― can mix and add to physics processes very easily � • Remember, if it can be wrapped in C++, you can add the physical process � 17
Process Biasing � • Introduced interface to Geant4 process biasing � S. Boogert � • Any process for any particle can be biased for any volume or set of volumes � which processes � cross-section scaling � which particles � Define bias ‘object’ � primary, secondaries � or all � attach sets of biases to objects � • Extremely flexible interface � • Attach to vacuum or general accelerator material � • Previously required specially written wrapper class for each � 18
LHC Beam Loss Studies � 19
Machine Protection � Superconducting coil: T = 1.9 K, quench limit ~15 mJ cm -3 Factor 9.7 x 10 9 Proton beam: 145 MJ (design: 362 MJ) Fractional Loss Limit: 1 turn: 1x10 -9 Continuous: 1x10 -12 Damage: 1x10 -6 20 S.Redaeli Hi-Lumi Workshop 2013
(HL) LHC Beam Loss Studies � • Need accurate energy deposition in cryogenic magnets � • Today’s strategy: � ― proton (only) tracking with SixTrack – integer losses on aperture � ― FLUKA highly detailed model of small sections ( ~500m) � • The LHC works! What do we need that’s not there? � • Each step in energy and intensity presents more unknown losses and operation issues � • High Luminosity LHC (HL-LHC) will be upgrade to LHC � 21
Recommend
More recommend
