Transcript Introduction to FTK

Introduction to FTK
More technical description: http://www.pi.infn.it/~orso/ftk/
March 9, 2005
Introduction to FTK - M. Shochet
1
What is the problem?
• In hadron colliders, the “easy” triggers are e, , , & generic jets.
• Is that good enough at the LHC?
– Opening a new energy frontier.
– SM  new physics, but what it is we don’t know
–  don’t ignore 3rd generation partons (b, )
• ex: EWK symmetry breaking
couples  mass  3rd generation important
– ex: bbH  bbbb or bb
March 9, 2005
Introduction to FTK - M. Shochet
2
Why is it hard to trigger on the 3rd generation?
• b : jet
Generic jets have production cross
•  : narrow jet
sections orders of magnitude larger.
• To identify b-jets and -jets, need tracks.
– b-jet: secondary vertex
– -jet: ≤ 3 tracks in narrow cone; no tracks around it
• 40 MHz interaction rate & 1 KHz level-2 accept rate
 selection of b-jets and -jets must be done quickly!
deadtime = L1 rate  L2 reject time
As high as possible to
lower PT threshold.
March 9, 2005
Introduction to FTK - M. Shochet
 make as short as
possible
3
Existing ATLAS (& CMS) Trigger
• Track pattern recognition & fitting done in Intel PC’s.
– 20-50 msec because of overhead of getting all of the raw silicon hits
into the processor and sorted for track finding.
TOO LONG!
March 9, 2005
Introduction to FTK - M. Shochet
4
FTK (Fast TracKer)
• Based on the CDF SVT
– Finds and reconstructs with near offline d0 precision all tracks of
PT > 1.5 GeV/c in ~20 sec.
sample for Bs mixing
March 9, 2005
Introduction to FTK - M. Shochet
Z → bb
5
How does it work?
• Complete pattern recognition on the fly as the silicon data
passes by using pre-stored bank of possible hit patterns.
The Pattern Bank
– CDF: 32K patterns/phi sector → 15M in ATLAS
made possible by advances in -electronics in past decade
(prototype: SVT upgrade)
March 9, 2005
Introduction to FTK - M. Shochet
6
Pattern recognition in action
AM = BINGO PLAYERS
PATTERN 5
PATTERN 4
PATTERN 1
PATTERN 2
PATTERN 3
PATTERN N
HIT # 1447
March 9, 2005
Introduction to FTK - M. Shochet
7
Track Fitting
• Very fast: In CDF, 200 nsec/track using FPGA.
– Narrow roads  linearized fit works well
• Linear equations (a few hardware adds and multiplies) with pre-stored
coefficients: curvature, d0, , 2.
Track impact parameter
 = 48m
March 9, 2005
Introduction to FTK - M. Shochet
8
Where does it fit in ATLAS?
SCT

Pixels 
R
O
D
s
on L1 accept
“Level 1.5”
FTK
R
O
B
s
R
O
B
s
silicon hits
silicon tracks
ask for ROI’s
Level 2
March 9, 2005
Introduction to FTK - M. Shochet
9
First Step: Making the Physics Case
• ATLAS wants a few specific channels studied so the
physics reach with and without FTK can be compared.
• In discussion with the ATLAS Physics Coordinators and
others, we came up with a “short list”.
March 9, 2005
Introduction to FTK - M. Shochet
10
Z  bb for measuring the b-jet response and resolution
(-jet & Z-jet balance have theory/exp problems)
for Mtop , MHiggs , …
Problem: L1 trigger rate  higher jet ET threshold
high threshold  high MJJ turn-on
Solution: high PT Z’s
qg  Zq  bbq
• 3-jet trigger
• highest ET jet is not tagged
• bb opening angle not fixed near 1800  lower
MJJ threshold
March 9, 2005
Introduction to FTK - M. Shochet
11
Higgs Physics
• bb H A
H  bb or H  
How much lower in tan can be reached?
• H  hh  bbbb
These jet PT’s are too low for existing triggers.
How much does FTK help?
March 9, 2005
Introduction to FTK - M. Shochet
12
 leptons
Need lower threshold for all  triggers.
l
How low can the threshold be?
• W   (Fabiola)
– for the high PT response of TileCal across the detector
– Existing  ET thresholds (35/45 GeV) too high for W’s
– (More general: isolated track triggers for calorimeter
calibration – rapid level-2 rejection  higher level-1
rate)
• (Ian) Existing  trigger is for 1-prong only. Polarization
information requires seeing other  decay modes.
•
•
March 9, 2005
Introduction to FTK - M. Shochet
13
More
•
(Michelangelo) B physics (if initial  is very low)
ex, B  
March 9, 2005
Introduction to FTK - M. Shochet
14
What has to be done?
What drives trigger decisions is background rejection, not
signal efficiency.
For signal, a few thousand events probably are enough.
But background studies require LOTS of events.
At 10 minutes/simulated event, …
 Use fully simulated events to parameterize
• jet response (with fluctuations)
• jet trigger efficiency vs ET ( especially at level 1)
• Erik Brubaker is working on this.
Use full simulation to parameterize FTK tracking:
• Pisa will get the FTKsim tools working with current ATLAS software
• then: tracking efficiency in jets, fake rate
Then can do convincing physics studies with fast simulation.
HOW?? (Iacopo)
March 9, 2005
Introduction to FTK - M. Shochet
15