Download - LCFI Simulation and Physics Studies
LCUK Meeting, 26th September 2006 Sonja Hillert (Oxford) p. 1
LCFI Simulation and Physics Studies
LCUK Meeting, Durham, 26th September 2006
Sonja Hillert (Oxford)
on behalf of the LCFI collaboration
Overview
Sensitivity to detector design parameters
The LCFI Vertex Package
Summary and Outlook
LCUK Meeting, 26th September 2006 Sonja Hillert (Oxford) p. 2
Simulation and Physics Studies - Overview aim: optimisation of vertex detector parameters and evaluation of its performance
two main areas of work:
Development of high-level reconstruction tools:
• current focus: development of C++ based Vertex Package, interfaced to LCIO/MarlinReco
comprising vertex finder ZVTOP, NN-based flavour tag and quark charge sign selection
• aim for first release at the time of ECFA 2006 (Valencia)
• release will be followed by work on upgrades
Study of Physics channels:
• goal is to realistically quantify performance following vertex detector contribution to
physics analysis through entire analysis chain
in parallel: performance comparison of different detector designs using fast MC SGV,
to be replaced by full simulation and reconstruction framework in the future
LCUK Meeting, 26th September 2006 Sonja Hillert (Oxford) p. 3
Sensitivity to detector design parameters
left: probability 0 to reconstruct neutral
B hadron as charged
since vertex charge performance is
sensitive to multiple scattering need to
keep layer thickness small (target 0.1 % X0)
also strong dependence on momentum cut
(track selection) – this depends critically
on tracking performance:
• track finding capability
• background rates
• linking across subdetector boundaries
should push all these parameters to their limits, as all these effects will eventually
add up in the real detector
LCUK Meeting, 26th September 2006 Sonja Hillert (Oxford) p. 4
The ZVTOP vertex finderD. Jackson,
NIM A 388 (1997) 247
two branches: ZVRES and ZVKIN (also known as ghost track algorithm)
The ZVRES algorithm: very general algorithm
that can cope with arbitrary multi-prong decay topologies
• ‘vertex function‘ calculated from Gaussian
´probability tubes´ representing tracks
• iteratively search 3D-space for maxima of this function
and minimise 2 of vertex fit
ZVKIN: more specialised algorithm to extend coverage to b-jets with
1-pronged vertices and / or a short-lived B-hadron
• additional kinematic information
(IP-, B-, D-decay vertex approximately
lie on a straight line) used to find
vertices
• should improve flavour tag
LCUK Meeting, 26th September 2006 Sonja Hillert (Oxford) p. 5
Flavour tag and quark charge sign selection aim of flavour tag: distinguish between b-jets, c- jets and light-quark / gluon jets
heavy flavour jets contain secondary decays, generally observed as secondary vertices
NN-approach to combine inputs; most sensitive: secondary vtx Pt-corrected mass & momentum
for charged B-hadrons (40% of b-jets): quark sign can be determined from vertex charge:
need to find all stable tracks from B-decay chain
probability of mis-reconstructing vertex charge small for both charged and neutral cases
neutral vertices require ‘charge dipole’ procedure from SLD still to be developed for ILC
Klaus Desch/ Thorsten Kuhl
LCUK Meeting, 26th September 2006 Sonja Hillert (Oxford) p. 6
interface SGV to
internal format
interface LCIO to
internal formatinput to LCFI Vertex Package
output of LCFI Vertex Packageinterface internal
format to SGV
interface internal
format to LCIO
ZVRES ZVKIN
ZVTOP:
vertex information
track attachment
assuming c jet
track attachment
assuming b jet
track attachment
for flavour tag
find vertex-
dependent
flavour tag
inputs
find vertex charge
find vertex-
independent
flavour tag
inputs
neural net flavour tag
LCUK Meeting, 26th September 2006 Sonja Hillert (Oxford) p. 7
Current Status
coding, validation and documentation far advanced;
remaining issues mainly to do with integration / system test
ZVTOP (Ben Jeffery): coding completed
• validation has focused on ZVRES branch:
performance at least as good as FORTRAN shown in detailed tests of increasing complexity;
• initial tests of ZVKIN look promising;
coding calculation of flavour-tag inputs (Erik Devetak)
and integration of NN-software into our package (Mark Grimes) progressing
internal working classes designed, largely implemented and tested:
• integration into MarlinReco beginning: test processors for ZVTOP & flavour-tag inputs exist
• interface to LCIO and to event-input from MarlinReco-based event reconstruction need
further work
LCUK Meeting, 26th September 2006 Sonja Hillert (Oxford) p. 8
Comparison of C++ & FORTRAN ZVRES
compare distributions of number of vertices and of tracks per vertex between
C++, FORTRAN and MC truth for 100 GeV b-jets (using identical input from fast MC SGV)
excellent agreement between the results from C++ and from FORTRAN
Ben Jeffery
LCUK Meeting, 26th September 2006 Sonja Hillert (Oxford) p. 9
Comparison of track-vertex association
tables show percentages of tracks from the primary, B- and D-hadron decay at MC level
that are assigned to primary, secondary (and tertiary) vertex or left isolated by ZVRES;
excellent agreement also seen in this detailed comparison of the two versions
results above were obtained for 100 GeV b-jets;
corresponding studies at 25 GeV and 250 GeV jet energy show equally good agreement
Mark Grimes
LCUK Meeting, 26th September 2006 Sonja Hillert (Oxford) p. 10
54% negative (C++ version better)
Ben Jeffery
Performance of decay length reconstructiondifference between the C++ and the FORTRAN deviation from MC in decay length
right: distribution of this difference, left: difference plotted vs MC decay length
fitter yields different
vertex position
differences in track-
vertex associations
C++ finds some vertices FORTRAN misses
and comes closer to MC truth than FORTRAN for entire range of decay lengths
LCUK Meeting, 26th September 2006 Sonja Hillert (Oxford) p. 11
LCIO persistency framework has been extended by dedicated vertex class to
accommodate the output of our software:
• each ReconstructedParticle points to one vertex from which it originated & to decay vertex
will provide MARLIN processors (modules) giving example code for
• running ZVTOP (one processor for each of the two branches ZVRES, ZVKIN)
• calculating neural net input variables from input to package & ZVTOP output
• training neural nets for flavour tag, obtaining NN outputs for pre-trained nets,
• vertex charge calculation
• combined processor: ZVRES + Hawkings flavour tag + vertex charge calculation
Interfacing the Vertex Package
Frank Gaede (DESY)
LCUK Meeting, 26th September 2006 Sonja Hillert (Oxford) p. 12
Summary and outlook
Development and validation of the LCFI Vertex Package are far advanced.
Excellent agreement is found between C++ and FORTRAN ZVRES branch of ZVTOP.
A new Vertex class has been introduced into LCIO. Integration of our package into
MarlinReco is beginning.
Interfacing to LCIO and to event-input from MarlinReco event reconstruction and
system tests need further work.
The first release of the code is planned for November.
Detailed comparisons with MarlinReco input and quantitative exploration of
improvements from the ghost track algorithm will be the next steps after the release,
in parallel to first studies of benchmarks processes sensitive to the detector design.
LCUK Meeting, 26th September 2006 Sonja Hillert (Oxford) p. 13
Additional Material
LCUK Meeting, 26th September 2006 Sonja Hillert (Oxford) p. 14
ZVTOP - Progress
Initial aim: replace FORTRAN ZVRES in SGV for testing- allows comparison of intermediate algorithm states when working on identical tracks- new version can be verified to be at least as good as FORTRAN
SGV FORTRAN ZVRES Flavour Tagtracks vertices
SGV FORTRAN ZVRES Flavour Tag
C++ ZVRES
Current Status
Add ZVKIN: SGV C++ ZVRES Flavour Tag
C++ ZVKIN
LCIO C++ ZVTOP LCIOFinally:
LCUK Meeting, 26th September 2006 Sonja Hillert (Oxford) p. 15
The ZVTOP vertex finder
two branches: ZVRES and ZVKIN (also known as ghost track algorithm)
The ZVRES algorithm:
tracks approximated as Gaussian ´probability tubes´
from these, a ´vertex function´ is obtained:
3D-space searched for maxima in the vertex function that satisfy
resolubility criterion; track can be contained in > 1 candidate vertex
iterative cuts on 2 of vertex fit and maximisation of vertex
function results in unambiguous assignment of tracks to vertices
has been shown to work in various environments differing in
energy range, detectors used and physics extracted
very general algorithm that can cope with arbitrary multi-prong decay topologies
D. Jackson,
NIM A 388 (1997) 247
LCUK Meeting, 26th September 2006 Sonja Hillert (Oxford) p. 16
The ZVKIN (ghost track) algorithm
more specialised algorithm to extend coverage to b-jets in which one or both
secondary and tertiary vertex are 1-pronged and / or in which the B is very
short-lived;
algorithm relies on the fact that IP, B- and D-decay vertex lie on an approximately
straight line due to the boost of the B hadron
should improve flavour tagging capabilities
ZVRES
GHOST
SLD VXD3 bb-MC
LCUK Meeting, 26th September 2006 Sonja Hillert (Oxford) p. 17
tanh (MPt / 5 GeV)
joint probability
uds
c
b
Flavour tag Vertex package will provide flavour tag procedure developed by R. Hawkings et al
(LC-PHSM-2000-021) and recently used by K. Desch / Th. Kuhl as default
NN-input variables used:
• if secondary vertex found: MPt , momentum
of secondary vertex, and its decay length and
decay length significance
• if only primary vertex found: momentum and
impact parameter significance in R- and z for the
two most-significant tracks in the jet
• in both cases: joint probability in R- and z (estimator of
probability for all tracks to originate from primary vertex)
will be flexible enough to permit user further tuning of the input variables for the neural net,
and of the NN-architecture (number and type of nodes) and training algorithm
b
c
LCUK Meeting, 26th September 2006 Sonja Hillert (Oxford) p. 18
Towards completion of the Vertex Package
b
c
c (b bkgr)c
c (b bkgr)
b
Z peak ECM = 500 GeV
test of full chain of C++ ZVTOP with FORTRAN flavour tag and vertex charge imminent
pure FORTRAN results (from SGV) show below:
C++ code for calculation of inputs for flavour tag being written
Vertex charge reconstruction for c-jets under development
LCUK Meeting, 26th September 2006 Sonja Hillert (Oxford) p. 19
ECM = 100 GeV
b
c (b-bkgr)
c
open: FORTRAN
full: C++
First full test of C++ ZVRES in FORTRAN framework compare FORTRAN to C++ ZVRES when fed with identical input events
unrealistically large fixed track errors used in both versions for diagnostic purposes
identical FORTRAN code for flavour tag (left) and vertex charge (right)
flavour tag: good agreement between C++ and FORTRAN, with C++ slightly better for b-jets
for vertex charge C++ in this first test performs considerably less well than FORTRAN –
but could be more sensitive to fixed huge track errors
Sonja Hillert
LCUK Meeting, 26th September 2006 Sonja Hillert (Oxford) p. 20
Future plans beyond the Vertex Package release
more detailed comparisons with performance obtained when using MarlinReco –
in particular, compare results from ‘cheaters’ to full track reconstruction, for which algorithms
independent of the old BRAHMS code, are now under development performance of our tools
with different choices for the input can provide useful feedback for tracking code development
important for us: step towards full framework, ensure we understand later physics results
explore and quantify what can be gained for flavour tag and vertex charge
when using the ghost track (ZVKIN) algorithm
should feed into future update, so the community can profit from our studies
begin study of benchmark physics processes:
e+e- b bbar with extra dimensions model as studied by Riemann, Hewett
SUSY physics
further physics processes
LCUK Meeting, 26th September 2006 Sonja Hillert (Oxford) p. 21
Vertex charge study for c-jets
Victoria Martin
vertex charge (ZVTOP & L/D) and #(c decay tracks) for various decay channels (ECM = 500 GeV)
first results look very promising; next look at corresponding leakage rates and optimise L/D
LCUK Meeting, 26th September 2006 Sonja Hillert (Oxford) p. 22
from: Steve Magill, Digitization simulation using DigiSim and Marlin, talk at Cambridge ILC workshop, April 2006