Reconstruction with Key4hep

Overview of Key4hep reconstruction tooling

Juraj Smieško

for the Key4hep developers

CERN

First ECFA Workshop on e+e- Higgs / Electroweek / Top Factories

06 October 2022

Key4hep

  • Set of common software packages, tools, and standards for different Detector concepts
  • Common for FCC, CLIC/ILC, CEPC, EIC, …
  • Individual participants can mix and match their stack
  • Main ingredients:
    • Data processing framework: Gaudi
    • Event data model: EDM4hep
    • Detector description: DD4hep
    • Software distribution: Spack

Gaudi

  • Data processing framework
    • Stitches and steers various algorithms together
    • Controls event loop
    • Manages transient storage and I/O
  • Used by live experiments: ATLAS, LHCb
  • Allows concurrency, new developments: Gaudi::Functional
  • Key4hep started life by attempting to reuse algorithms already developed
  • Need for convertors/wrappers:
    k4MarlinWrapper, k4CLUE, k4Pandora, …

Hello World in Gaudi:

Source: Gaudi

Running the Reconstruction

  • Multiple possibilities how to run:
    • As part of the larger steering
      • Input from: k4SimGeant4, k4SimDelphes
    • As a separate step
      • Input from: … + DDSim
  • In both cases I/O needs to be handled (PODIO)
    • Reading event data from in EDM4HEP ROOT files
    • LCIO Conversion: k4LCIOReader
  • Algorithms developed for Marlin needs to be wrapped
  • Detector description and instantiation is done with DD4hep
    • XML + C++ code
  • Easiest way to access the stack is from CVMFS:

Digitization

  • The digitization could be done as
    • Last simulation step
    • First step of reconstruction
  • Detector specific, two main EDM4HEP datatypes
      SimTrackerHit, SimCalorimeterHit
  • Simple approach in case of k4RecCalorimeter: Sum hit in a cell
  • More involved solutions in MarlinReco
    • TrackDigi and CaloDigi
  • Recent highlight: Effort to add generic digitization components into DD4hep (DDDigi)
  • Recent highlight: Conversion of DDPlanarDigiProcessor

Tracking

  • Work on integrating ACTS into the Key4hep underway
    • State-of-the-art track reconstruction
    • Project was spawned from tracking code of ATLAS
    • The Key4hep wrapper: k4ActsTracking
    • Recent highlights:
      • Seamless loading of FCC detector models
      • Inclusion of EIC framework components
  • Available as Marlin processors:
    • Conformal tracking
    • Clupatra
    • ForwardTracking for the FTD

Clustering

  • CLUE: CLUstering of Energy
    • Intergrated in Key4hep as k4CLUE
    • Uses energy density to define ranking, seeding thresholds, …
  • LAr Calorimeter Reconstruction: k4RecCalorimeter
    • Sliding Window and TopoCluster based algorithms
    • Algorithm developed in proto Key4hep environment

Other Reconstruction algorithms

  • Through Marlin Wrapper are available also:
    • LCFIPlus for vertexing and flavour tagging
    • FastJet for jet clustering
    • KinematicFitting
    • Particle ID
    • Conditions
  • ACTS is also capable of vertexing

[arXiv:0802.1189]

Higher level reconstruction

  • PandoraPFA is a prime candidate for the integration in k4Pandora
  • It integrates multitude of pattern recognition algorithms
  • Developed for reimplementation of PFA at future e+e- linear collider
  • Ongoing efforts revolve around developing a direct Gaudi wrapper k4Pandora or use of two existing ones k4MarlinWrapper+DDMarlinPandora
  • Recent highlight: Dummy clustering achieved in k4MarlinWrapper+DDMarlinPandora

Example: LAr calorimeter

  • Sampling Calorimeter based on LAr/LKr + Pb/W
  • Simulation/Reconstruction fully steered in Gaudi
  • Several Gaudi based algorithms include
    • Sampling fraction determination
    • Upstream/Downstream energy correction
    • Adding noise to Calo Cells
    • Clustering: Sliding Window or TopoCluster based

Example: CLD

  • Uses DDSim to simulate events
  • Heavy use of the converters
  • The reconstruction consists of
    • Background Overlay, Digitization
    • Track Pattern Reconstruction (ConformalTracking), track fit
    • Particle Flow Reconstruction (PandoraPFA)
    • Vertexing and Flavour Tagging (LCFIplus)
    • Full CLD reconstruction in gaudi
  • Input and output are in EDM4hep

Conclusions

  • The reconstruction for the future colliders is slowly taking shape
  • Mostly thanks to the ability to integrate specialized projects
  • k4MarlinWrapper helps to bridge transitional period
  • Future algorithms have well defined environment to count on
  • Effort required to port reconstruction of already existing detector concepts to Key4hep

Backup

Summary

Reconstruction with Key4hep I

  • Key4hep: Gaudi, EDM4hep, DD4hep, Spack
  • Key4hep has ability to integrate other advanced reconstruction tools/frameweworks
  • k4MarlinWrapper helps to bridge transitional period
    • DDMarlinPandora, LCFIPlus, ConformalTracking, …
  • Integration of large frameworks underway
    • K4CLUE, k4Pandora, k4ActsTracking
  • Effort required to port reconstruction of already existing detector concepts to Key4hep

Reconstruction with Key4hep II

LAr Calorimeter

  • Sampling Calorimeter based on LAr/LKr + Pb/W
  • Simulation/Reconstruction fully steered in Gaudi
  • Several Gaudi based algorithms include
    • Sampling fraction determination
    • Upstream/Downstream energy correction
    • Adding noise to Calo Cells
    • Clustering: Sliding Window or TopoCluster based

CLD

  • Uses DDSim to simulate events
  • Heavy use of the converters
  • The reconstruction consists of
    • Background Overlay, Digitization
    • Track Pattern Reconstruction (ConformalTracking), track fit
    • Particle Flow Reconstruction (PandoraPFA)
    • Vertexing and Flavour Tagging (LCFIplus)
    • Full CLD reconstruction in gaudi
  • Input and output are in EDM4hep