1. Luminosity Measurement in the
Helsinki, November 2000
Reminder
what is LHCb ?
First studies for lumi
 pile-up vertex detector
Summary
Massimiliano Ferro-Luzzi, CERN
Massimiliano Ferro-Luzzi, CERN

2. CP violation and rare decays
A Large Hadron Collider beauty experiment for precision measurements of
CP violation and rare decays
(LHCbepmCPvrd… or just LHCb)
Measure asymmetries:
Af (t) =
e.g. (semi)leptonic
FCNC B-decays.
…always possible to
measure rates relative
to other channels
Rf (t) + Rf (t)
Rf (t) - Rf (t)
from four rates
Rf (t) = initial B decaying to final state f
(a bar stands for “CP conjugate”)
Massimiliano Ferro-Luzzi, CERN

3. Typical B event at LHCb B0  - D*+  D0 + B0  +- Primary vertex
Generated polar
angles of b and
b hadrons
(with PYTHIA)
B0  +-
10 mm
Aims for offline:
decay distance resolution  120 m
proper decay time resolution  0.04 ps
Massimiliano Ferro-Luzzi, CERN

4. LHCb Detector Acceptance: q  [10, 300] mrad h  [5.3, 1.9]
Massimiliano Ferro-Luzzi, CERN

5. LHCb Trigger total  100 mb beauty  0.5 mb To tape:
1.0
0.8
0.6
0.4
0.2
0.0
Bd  J/ (+-)Ks + tag
Bd  +- + tag
Bd  D K*
reconstructable events
Efficiency for
SICB monte-carlo
(PYTHIA)
total  100 mb
beauty  0.5 mb
To tape:
 3 million Bd evts/yr
 1 million Bs evts/yr
Clock
bunch crossing
with n>0 collision(s)
single collision
107
106
105
104
103
102
101
total
charm
beauty
Event rates [Hz]
L L L L3
e/h calo, -ch
ET , pT
VErtex LOcator
Impact param,
vertices
Trackers ,
Momentum
Full
reconstruction
Massimiliano Ferro-Luzzi, CERN

6. LHCb Si Vertex Detector Modules
etc.
R|f
-20 80
[cm]
towards
spectrometer Z
f|R
R|f
retract by  3 cm
during beam
filling/tuning mm
R-detector prototype
220 mm thick Si single-side
one module: R and f measuring planes
total of 220 k channels, analog, S/N=15
Design work ongoing for front-end chip
(DMILL and subm technologies)
Massimiliano Ferro-Luzzi, CERN

7. LHCb Vertex Detector Design
Exit
window
RF/Vacuum shield
Take into account:
static and dynamic vacuum effects,
wake field effects,
multiple scattering,
alignement, etc.
Silicon
module
(cooled)
Massimiliano Ferro-Luzzi, CERN

8. Problem for trigger: n>1 collisions
Level-1 is based on vertex finding and becomes problematic when there
is more than 1 collision in a bunch crossing
Level-0 (pT cut, output at 1 MHz) is less “beauty-selective” for events
with n>1 collisions.
Possible improvement: find and veto crossings with n>1 collisions !
Possible implementations: 1. Total energy in calorimeters
2. Time-of-flight clusters
our choice !  3. Look for primary vertices
Probability Pn to have n collisions in a bunch crossing:
Luminosity
Pn = e -m
m n n!
m = s L / f
with
Bunch crossing
frequency  30 MHz
Cross-section
Massimiliano Ferro-Luzzi, CERN

9. Pile-Up Detector: principle
N. Zaitsev’s PhD thesis
Univ. van Amsterdam, 11/2000 ZB ZA RB RA
ZPV
Silicon R-detectors
B A
RB (cm)
RB (cm)
RA (cm)
RA (cm)
If hits are from the same track:
peak P1
(peak size S1)
RA ZPV - ZA
RB ZPV - ZB =
k 
peak P2
(peak size S2)
ZA - k ZB
1 - k
or ZPV =
ZPV (cm)
 allows locating pp collisions
Massimiliano Ferro-Luzzi, CERN

10. LHCb Pile-Up Detector: Implementation
N. Zaitsev’s PhD thesis
Univ. van Amsterdam, 11/2000
VD: add upstream two “R-only” modules
Read out for trigger level 0 (Vertex Detector enters in level 1)
Histogramming algorithm to distinguish single and multiple pp collisions.
Performance
multiple
Schematically:
(0) build a ZPV histogram
(1) search highest peak P1 (peak size S1)
(2) mask hits belonging to P1
(3) search next highest peak P2 (peak size S2)
(4) classify: if S2<Slimit then single,
else multiple
Slimit sets efficiency and contamination.
single
True single
True multiple
Slimit
peak size S2
Massimiliano Ferro-Luzzi, CERN

11. LHCb Pile-Up Veto: performance
N. Zaitsev’s PhD thesis
Univ. van Amsterdam, 11/2000
Occurrence rate of n-collision crossings
as a function of luminosity
Rate of “single-collision” B-events
as a function of luminosity
Before level-0
n =
After level-0: without and with Pile-up veto
 gain factor of at least 1.4 in B-event rate !!
Massimiliano Ferro-Luzzi, CERN

12. Pile-Up Detector for Lumi Monitoring
N. Zaitsev’s PhD thesis
Univ. van Amsterdam, 11/2000
Most subsystems of LHCb are readout with highly biased trigger
Pile-up: read out at level-0 (  unbiased )
large acceptance
high and stable efficiency
knows how many collisions in a bunch crossing
(distinguishes bunch crossings with n=0, n=1 and n>1 collisions)
Use of this feature increases statistical accuracy
but systematics can arise from osmose between the various classes
( mostly between classes n=1 and n>1 !!
Typical : (n=1)-efficiency  90% and
(n>1)-fraction in (n=1)-class 12% )
Massimiliano Ferro-Luzzi, CERN

13.  - Estimators The following estimators were studied:
(e0 , e1 , en>1= fractions of bunch crossings classified as having n= 0, n=1 and n>1 collisions)
1/0 ratio estimator m = e 1 / e 0
0-event estimator m = -ln(e 0)
non-0-event estimator
full sample estimator
Does not need to know “osmotic pressure”
between n=1 and n>1 !!
Uses “either n=0 or n>0 ” logic
 low systematic uncertainties
e1 - P en>1 - Pn>1 2
De Den>1
c 2 =
e0 - P e1 - P en>1 - Pn>1 2
De De Den>1
c 2 =
Massimiliano Ferro-Luzzi, CERN

14. 0-event  - Estimator P0,meas = P0 + S an Pn
“Osmotic flow” from class n>0 to class n=0 will cause a shift of m towards lower values.
Monte-carlo: a1 =  0.002
an>1 negligible at LHCb luminosities
mix event samples with 0, 1, …6 collisions (PYTHIA),
in Poisson proportions
add noise (to simulate beam-gas collisions, activity
of detectors, etc.)
fit distribution of largest peak size S1 to extract m
P0,meas = P0 + S an Pn
n>0
n=0 events
105
104
103
102
101
100
L [ 1032 cm-2 s-1 ]
5.0
2.5
0.5
Event Geometrical Cross
type Acceptance section
elastic % mb
diffractive % mb
inelastic % mb
peak size S1
Massimiliano Ferro-Luzzi, CERN

15. Summary
LHCb is a forward-but-not-really experiment (10 mrad < q < 300 mrad)
for studying CP-violation (asymmetries) and rare decays (relative to others)
in the B-meson sector
Preliminary studies for relative luminosity monitoring were carried out for
the Pile-up detector
- zero-event estimator is simplest, with lowest systematics
- precision limited by knowledge of acceptance to diffractive events
- we can expect a (systematic) uncertainty down to ~0.5 % for relative luminosity
- statistics are  infinite … enough to measure individual bunch-bunch luminosities
with ~0.5% statistical accuracy in about 20 seconds
absolute normalization will be done by using calibrated processes
and/or the TOTEM results (need monte-carlo in both cases)
measure interaction spot shape with vertex detector ?
Massimiliano Ferro-Luzzi, CERN