1. Preparation for ITS alignment
What? ITS detectors, target alignment precision
Why? Impact of misalignments
How? Strategy and methods
How well? First results from simulation
A. Dainese
(INFN – LNL)
for the ITS alignment group (CERN, LNL, NIKHEF, PD, TS)
Terzo Convegno sulla Fisica di ALICE - LNF, Andrea Dainese

2. Inner Tracking System (ITS)
Silicon Pixel Detector (SPD):
~10M channels
240 sensitive vol. (60 ladders)
Silicon Drift Detector (SDD):
~133k channels
260 sensitive vol. (36 ladders)
Silicon Strip Detector (SSD):
~2.6M channels
1698 sensitive vol. (72 ladders)
SSD
ITS total:
2198 alignable sensitive volumes
 d.o.f.
SPD
SDD
Terzo Convegno sulla Fisica di ALICE - LNF, Andrea Dainese

3. ITS detector resolutions & target alignment precisions
yloc
zloc
xloc
detector local c.s.: xloc~rfglob, yloc~rglob, zloc = zglob
SPD
(r = 4 & 7 cm)
SDD
(r = 14 & 24 cm)
SSD
(r = 39 & 44 cm)
nom. resolutions
xloc  (yloc)  zloc [mm3]
12  (0)  120
38  (0)  20
20  (0)  830
full mis. (shifts)
xlocyloczloc [mm3]
20  20  20
45  45  45
30  30  100
residual mis. (shifts)
10  10  20
15  15  100
rotations (mrad)
around xloc,yloc,zloc
0.3
Full: initial misalignments as expected from the mechanical imprecision after installation, actually set to mm at the sensor level, probably higher at the ladder or layer level (~100 mm), more later...
Residual: expected misalignment left after applying the realignment procedure(s). Target ~0.7resol.  ~20% degradation of the resolution
Terzo Convegno sulla Fisica di ALICE - LNF, Andrea Dainese

4. Impact of ITS misalignment on tracking performance
Effect of misalignment on d0 (and pt) resolutions studied by reconstructing misaligned events with ideal geometry
Estimated effect on D0Kp significance
Primary
Vertex B e X d0
rec. track
null
residual
full
full+
A.D, A.Rossi
Terzo Convegno sulla Fisica di ALICE - LNF, Andrea Dainese

5. Introduction of “realistic” misalignment
Why “realistic”?
misalignment should follow hierarchy of hardware structure; each level should be misaligned (and then realigned)
magnitude of initial misalignments should be realistic (input from hardware experts)
misalignments at the same hierarchical level should be correlated
E.g., for SPD: barrel / half-barrel / sector / half-stave / ladder
# sensitive volumes: / / / / 1
magnitude of misal. (mm): <1000 / ~ / ~ / / 5
(up to now only the ladder was misaligned)
Transition to realistic misalignment is in progress (requires changes to the ITS geometry)
Will provide:
better playground for preparation of realignment procedures
better estimate of effects of residual misalignments on performance
Terzo Convegno sulla Fisica di ALICE - LNF, Andrea Dainese

6. ITS alignment with tracks: general strategy
Data sets: cosmics + first pp collisions (and beam gas)
use cocktail of tracks from cosmics and pp to cover full detector surface and to maximize correlations among volumes
Start with B off, then switch on B  possibility to select high-momentum (no multiple scattering) tracks for alignment
General strategy:
start with layers easier to calibrate: SPD and SSD
good resol. in rf (12-20mm), worse in z ( mm)
ITS z resol. provided by SDD anode coord. (20mm)  easily calibrated  can be included from the beginning in alignment chain
global ITS alignment relative to TPC (already internally aligned)
finally, inclusion of SDD (drift coord: rf), which probably need longer calibration (interplay between alignment and calibration)
Two independent track-based alignment methods in preparation:
global: Millepede 1 (ported to ALICE for muon arm alignment)
local: iterative method based on residuals minimization
Terzo Convegno sulla Fisica di ALICE - LNF, Andrea Dainese

7. Preparation for cosmics data (1)  Cosmics Run in February 2008 
D.Elia
Expected muon rate through ITS inner layer (~200cm2, ~40m underground): ~0.02 Hz ( ~104 m/week)
Trigger with SPD layers (tracks with points):
Trigger with ACORDE (“peripheral” tracks with 4-8 points in SSD-SDD)
A side
FastOR (FO) of the 20 chips on 2 half-staves
For each half-barrel (A side, C side):
20 FOs outer layer, 10 FOs inner layer
Any logic combination of these 30 FOs
SPD
FastOR
C side
Option being considered for cosmics:
2Layers coinc. (≥2FOs inn layer & ≥2FOs out layer)
purity (fraction with 1 m with 4 SPD hits): ~97%,
inefficiency (fraction of lost m with 4 SPD hits): ~19%
Terzo Convegno sulla Fisica di ALICE - LNF, Andrea Dainese

8. Preparation for cosmics data (2)
Cosmics at ALICE: p > 10 GeV/c, <p>~20 GeV/c (Hebbeker, Timmermans, 2001)
Cosmics tracking in ITS:
modified stand-alone ITS tracking (cluster-grouping algorithm from Torino)
98% efficiency (12 points, 6inward+6outward) for muons that leave 12 hits, with B=0 and B=0.5T
high eff (~80-90%) also for muons crossing only outer layers
robustness tested with “extreme” misalignment scenarios
tracks prolongation from TPC to ITS being optimized for cosmics
Preparing first d0 resolution meas. by cosmics two-track matching
sd0=12mm
sd0=21mm no
misal.
full
A.D.,
A.Jacholkowski
Terzo Convegno sulla Fisica di ALICE - LNF, Andrea Dainese

9. Cosmics in the ITS SPD inner SDD inner SSD inner SPD inner 6 pts
50k m’s through inner ITS layer (~5 weeks of cosmics data)
Track multiplicity per module:
A.Rossi
SPD inner
SDD inner
SSD inner
103
102 10 1
Volume correlations. Number of modules correlated to a given module:
800
400
SPD inner
6 pts
tracks
12 pts
tracks
Terzo Convegno sulla Fisica di ALICE - LNF, Andrea Dainese

10. Track-based alignment in the barrel (ITS, TPC, TRD,
Track-based alignment in the barrel (ITS, TPC, TRD, ...) - Framework Overview -
C.Cheskov
Reconstruction
Reconstruction
Reconstruction
Reconstruction
Terzo Convegno sulla Fisica di ALICE - LNF, Andrea Dainese

11. ITS alignment with tracks: the global approach of Millepede
M.Lunardon, S.Moretto
Determine alignment parameters of “all” modules in one go, by minimizing the global c2 of track-to-points residuals for a large set of tracks (cosmics + pp)
Linear problem: the points residuals can be expressed as a linear function of the (global, di) alignment params and (local δi) tracks’ params
At the moment being tested with cosmics tracks and B=0
Strategy for tests:
validate algorithm with “fast simulation” (no detector response)
introduce full detector response
Terzo Convegno sulla Fisica di ALICE - LNF, Andrea Dainese

12. Millepede with fast simulation
SPD
SDD
SSD
5 weeks of cosmics
(50k in SPD1), B=0
Tracks with 12 pts
375 modules
(out of 2198)
xLOC
input
result
difference
yLOC
yloc
zloc
xloc
zLOC
Terzo Convegno sulla Fisica di ALICE - LNF, Andrea Dainese

13. Millepede with fast simulation RMS of (result-input)
As a function of the number of minimum # of points per track
5 weeks of cosmics
(50k in SPD1) , B=0
Also tracks
with <12 pts
SPD
x y z
5mm!
better with “peripheral tracks”
  
SDD similar
to SPD
Terzo Convegno sulla Fisica di ALICE - LNF, Andrea Dainese

14. Millepede with fast simulation RMS of (result-input)
As a function of the number of minimum # of points per track
5 weeks of cosmics
(50k in SPD1), B=0
Also tracks
with <12 pts
SSD
x y z
much better with “peripheral tracks”
  
  
Terzo Convegno sulla Fisica di ALICE - LNF, Andrea Dainese

15. Millepede: fast vs. full simulation
Deterioration of results with full detector simulation
This triggered investigations on the different steps of the simulation, with misalignment  found problems with overlaps that caused shift of ITS rec. points  will be solved in new ITS geometry
For the moment, continue alignment studies without misalignments
Terzo Convegno sulla Fisica di ALICE - LNF, Andrea Dainese

16. Millepede: full simul., no misal.
about 5 weeks of cosmics, B=0
663 modules ( 135 SPD SDD SSD )
PARAM
SPD
SDD
SSD
mean
RMS
x (mm)
-1.4
4.9
0.7
4.6
1.0
8.2
y (mm)
0.9
7.6
19.1
16.3
5.3
66.0
z (mm)
5.7
-0.3
-0.4
63.7
y (mdeg)
-0.9
15.6
-1.5
20.5
6.7
191
q (mdeg)
-0.7
9.2
0.0
5.6
-1.9
25.7
f (mdeg)
14.3
43.1
-1.6
3.7
152.5
M.Lunardon, S.Moretto
Terzo Convegno sulla Fisica di ALICE - LNF, Andrea Dainese

17. Millepede: full simul., no misal. SPD only, SPD+SSD
SPD: worse on the sides
 need pp tracks z f
top
bottom
about 5 weeks of cosmics, B=0
612 modules ( 192 SPD SSD )
PARAM
SPD RMS
SSD RMS
SPD only
SPD+SSD
x (mm) 8 6 14
y (mm) 19 12 71
z (mm) 13 11 60
y (mdeg) 36 28
199
q (mdeg) 10 7 20
f (mdeg) 50 48
161
M.Lunardon, S.Moretto
Terzo Convegno sulla Fisica di ALICE - LNF, Andrea Dainese

18. ITS alignment with tracks: iterative local method
C.Cheskov
A.Rossi
Determines alignment params by minimizing track-to-points residuals
Local: works on a module-by-module basis
Iterations are used to take into account correlations between the alignment params of different modules
Terzo Convegno sulla Fisica di ALICE - LNF, Andrea Dainese

19. Iterative local method: full simul., no misal.
about 5 weeks of cosmics, B=0
no iterations (not necessary without misalignment)
PARAM
SPD RMS
SDD RMS
SSD RMS
Iter
Millep
x (mm) 2 5 3 8
y (mm) 7 20 16 64 66
z (mm) 6 60
y (mdeg) 17 40 21 --
191
q (mdeg) 9 26
f (mdeg) 58 43
153
A.Rossi
Terzo Convegno sulla Fisica di ALICE - LNF, Andrea Dainese

20. Iterative local method: full simul. with misal., iterations
SPD inner: mean, RMS
iterationsconvergence
Dx (mm)
global z (cm)
global f
worse on the sides
 need pp tracks
A.Rossi
Terzo Convegno sulla Fisica di ALICE - LNF, Andrea Dainese

21. ITS-TPC relative alignment
Relative alignment of ITS and TPC (3 shifts + 3 angles) with straight tracks (including cosmics)
Alignment requirements: given by TPC resolutions:
shifts: ~100 mm
angles: ~0.1 mrad
Method (under development):
Assume that TPC and ITS are already
internally aligned and calibrated
Use independently fitted tracks
in the ITS and the TPC
Alignment params are estimated
by a Kalman filter algorithm
Proof-of-principle test with
“toy” tracks
M.Krzewicki
Terzo Convegno sulla Fisica di ALICE - LNF, Andrea Dainese

22. Summary
The ITS alignment challenge: determine 13,000 parameters with a precision of ~10 mm
Track-based alignment using cosmics and pp collisions
preparation for cosmics reconstruction in ITS
Two independent algorithms under development
Millepede (global)
local iterative method
Promising results even with cosmics only, should be much better with cosmics + pp tracks
ITS alignment relative to TPC also under study
Terzo Convegno sulla Fisica di ALICE - LNF, Andrea Dainese

23. EXTRA SLIDES
Terzo Convegno sulla Fisica di ALICE - LNF, Andrea Dainese

24. Fast Simulation A muon direction is generated
Intersection points with misaligned detectors are evaluated in local coordinate systems
large misal. order of 100 mm
Points smeared with given resolutions
Use tracks with 4-12 points
Advantages w.r.t. standard sim:
clean situation, without simulation/reconstruction effects
faster
Terzo Convegno sulla Fisica di ALICE - LNF, Andrea Dainese