200 likes | 274 Views
GlueX Collaboration Meeting, Newport News, Oct. 3-5, 2013. HDGeant4 Development Report. Richard Jones – University of Connecticut. major hurdles overcome in geometry geometry validation complete particles now tracking in magnetic field remaining milestones.
E N D
GlueX Collaboration Meeting, Newport News, Oct. 3-5, 2013 HDGeant4 Development Report Richard Jones – University of Connecticut major hurdles overcome in geometry geometry validation complete particles now tracking in magnetic field remaining milestones
slide from collaboration meeting, May 2012 HDGeant4 project: work plan • Download and install a recent stable release of G4 • Create a template from one of the standard examples • Design classes for geometry import from hdds • Implement geometry import from hdds • Make a set of scripts to generate some standard views of GlueX • Debug the hdds geometry and validate using standard views • Implement classes for reading and stepping through magnetic fields • Configure a robust physics list for GlueX simulations • Implement the particle gun event generator • Implement Monte Carlo event input from hddm stream • Implement and test the internal cobrems generator • Create and document standard control macros for a few common scenarios • Set up mechanisms for verbose tracking output • Implement classes for hits, truth collection and output progress so far (5/2012) Richard Jones, GlueX Collaboration Meeting, Newport News, Oct. 3-5, 2012
HDGeant4 project: work plan • Download and install a recent stable release of G4 • Create a template from one of the standard examples • Design classes for geometry import from hdds • Implement geometry import from hdds • Make a set of scripts to generate some standard views of GlueX • Debug the hdds geometry and validate using standard views • Implement classes for reading and stepping through magnetic fields • Configure a robust physics list for GlueX simulations • Implement the particle gun event generator • Implement Monte Carlo event input from hddm stream • Implement and test the internal cobrems generator • Create and document standard control macros for a few common scenarios • Set up mechanisms for verbose tracking output • Implement classes for hits, truth collection and output progress so far (5/2012) (9/2013) Richard Jones, GlueX Collaboration Meeting, Newport News, Oct. 3-5, 2012
HDGeant4 project: work plan • Download and install a recent stable release of G4 • Create a template from one of the standard examples • Design classes for geometry import from hdds • Implement geometry import from hdds • Make a set of scripts to generate some standard views of GlueX • Debug the hdds geometry and validate using standard views • Implement classes for reading and stepping through magnetic fields • Configure a robust physics list for GlueX simulations • Implement the particle gun event generator • Implement Monte Carlo event input from hddm stream • Implement and test the internal cobrems generator • Create and document standard control macros for a few common scenarios • Set up mechanisms for verbose tracking output • Implement classes for hits, truth collection and output progress so far (5/2012) (9/2013) Richard Jones, GlueX Collaboration Meeting, Newport News, Oct. 3-5, 2012
slide from collaboration meeting, May 2012 HDGeant4 : first look with G4 vis Richard Jones, GlueX Collaboration Meeting, Newport News, Oct. 3-5, 2012
slide from collaboration meeting, May 2012 HDGeant4 : first look with G4 vis Richard Jones, GlueX Collaboration Meeting, Newport News, Oct. 3-5, 2012
slide from collaboration meeting, May 2012 HDGeant4 : more cutaway views Richard Jones, GlueX Collaboration Meeting, Newport News, Oct. 3-5, 2012
slide from collaboration meeting, May 2012 HDGeant4 : next steps • The geometry has accumulated defects since the original implementation in hdds • example: readout cables from CDC are completely outside of their mother volume! • Geant3 is largely blind to such defects, although the results are affected by them. • Geant4 gets very worried by inconsistencies in the geometry – and crashes. • Next step: work through geometry defects and fix them one by one – time consuming but essential. • Remaining steps: 5 – 10 days of work Richard Jones, GlueX Collaboration Meeting, Newport News, Oct. 3-5, 2012
debugging hdds geometry • Pretty 3D plots look great! • To see real problems, we need to examine planar cuts through the geometry. • G4 supports planar cut views through “sectionPlane” view. • First look at CDC in transverse cut near upstream end: problems! • Same geometry as we are using for geant3 !!! • Is this the hdds geometry? check in geant3 dcut view, no! • Is this the hdds geometry renderering in geant4 objects? check, no! • Is this a bug in the geant4 visualization system? major digging, yes! Richard Jones, GlueX Collaboration Meeting, Newport News, Oct. 3-5, 2012
debugging hdds geometry G4 vis! Vertical sectionPlane: edge-on it should be a zero-thickness line, working for some volumes, not for others at all, still others are cut but shifted… Richard Jones, GlueX Collaboration Meeting, Newport News, Oct. 3-5, 2012
debugging hdds geometry G4 vis! • Careful search through slides by SLAC workshop slides shows admission of problems: • Examination of the G4 visualization code shows that all geometries are treated consistently, using the standard boolean solid logic that is a core component of the G4 geometry. • Browsing deep in the visualization code, I see that there are error messages that are suppressed by a compile-time switch in the standard build. • Turn in on, rebuild and rerun – many fatal errors reported! “sectionPlane works for some simple geometries, not reliable for more complex geometries.” Richard Jones, GlueX Collaboration Meeting, Newport News, Oct. 3-5, 2012
debugging G4 vis! page 16 page 31 page 52 Richard Jones, GlueX Collaboration Meeting, Newport News, Oct. 3-5, 2012
debugging G4 vis! … and 3 weeks later, problem found, fix generalized to most general geometry problem resolved, now on to overlaps… Richard Jones, GlueX Collaboration Meeting, Newport News, Oct. 3-5, 2012
debugging hdds geometry • Call CheckOverlaps in G4 • first pass: 3725 violations! • Many of these are repeats of the same geometry issue, so 95% were straight-forward to solve. • Last 5% were difficult: especially the CDC endplates. • Several false starts: • drill holes for the straws • segment the straws (5) • parallel worlds downstream end of CDC with endplate • New feature added to Geant4 in version 4.9.5 (2012) • Given the fancy name “parallel layered mass geometries” • Invented by medical simulation groups to describe human tissue Richard Jones, GlueX Collaboration Meeting, Newport News, Oct. 3-5, 2012
Layered mass geometries in parallel world (slide borrowed from talk by Makoto Asai, SLAC) Geometry IV - M.Asai (SLAC) 15 Richard Jones, GlueX Collaboration Meeting, Newport News, Oct. 3-5, 2012 Suppose you implement a wooden brick floating on the water. Dig a hole in water… Or, chop a brick into two and place them separately…
Layered mass geometries in parallel worlds (slide borrowed from talk by Makoto Asai, SLAC) Mass world Parallel world Geometry IV - M.Asai (SLAC) 16 Richard Jones, GlueX Collaboration Meeting, Newport News, Oct. 3-5, 2012 • Parallel geometry may be stacked on top of mass geometry or other parallel world geometry, allowing a user to define more than one worlds with materials (and region/cuts). • Track will see the material of top-layer, if it is null, then one layer beneath. • Alternative way of implementing a complicated geometry • Rapid prototyping • Safer, more flexible and powerful extension of the concept of “many” in Geant3
Layered mass geometries in parallel worlds (slide borrowed from talk by Makoto Asai, SLAC) Geometry seen by e+, e-, g Geometry seen by other particles Geometry IV - M.Asai (SLAC) 17 Richard Jones, GlueX Collaboration Meeting, Newport News, Oct. 3-5, 2012 • A parallel world may be associated only to some limited types of particles. • May define geometries of different levels of detail for different particle types • Example for sampling calorimeter: the mass world defines only the crude geometry with averaged material, while a parallel world with all the detailed geometry. Real materials in detailed parallel world geometry are associated with all particle types except e+, e- and gamma. • e+, e- and gamma do not see volume boundaries defined in the parallel world, i.e. their steps won’t be limited • Shower parameterization such as GFLASH may have its own geometry
debugging hdds geometry - complete new improved model of the straws in the endplate regions – no overlaps flagged! Richard Jones, GlueX Collaboration Meeting, Newport News, Oct. 3-5, 2012
tracking with particle gun generator Tracking of 10 300 MeV/c protons at ~20° without field (left) and with uniform field (right). Most material has been moved to an “invisible” layer – layers can be turned on/off. Richard Jones, GlueX Collaboration Meeting, Newport News, Oct. 3-5, 2012
HDGeant4 status and prospects • Still 5 weeks of work before first release – this fall • 50% error – assumes no major new G4 bugs! • First release will be alpha – not intended for actual studies! • Anticipated ~1 year shake-down transition period. • Implement Monte Carlo event input from hddm stream 8 days • Implement and test the internal cobrems generator 3 days • Standard control macros for a few common scenarios 1 day • Set up mechanisms for verbose tracking output 1 day • Implement classes for hits, truth collection and output 12 days Richard Jones, GlueX Collaboration Meeting, Newport News, Oct. 3-5, 2012