Pixel B-layer Replacement Geometry Study

Track Impact Parameter Resolution

As the main goal is see the subtle differences between the different B-layer replacement geometries, the study was done using a standalone track error calculation. The software used is the LCDTRK package extended and maintained by Bruce Schumm from the original TRKERR package from Walt Innes. Track covariant error matrix is constructed from the specified detector geometry, material and hit resolution tp extrapolate the track parameter resolution at the interaction point.

Since the B-layer replacement is primarily an issue in the barrel, this initial study is for barrel only. The invariant detector description for beampipe, SCT and TRT are summarized below:

Detector Part	Description	Mean Radius (cm)	Orientation angle	Radiation length(%)	Spatial resolution (microns)
Beam pipe	800mm Be 800mm Be	2.94 3.42	0	0.227 0.227	-
Pixel support tube	Pixel outer carbon-fiber support tube	22.9	0	1.50	-
SCT Barrel 0	Si Strip pair	29.9	0, +40mrad	2.96	17
SCT Barrel 1	Si Strip pair	37.1	0, - 40mrad	2.96	17
SCT Barrel 2	Si Strip pair	44.3	0, +40mrad	2.96	17
SCT Barrel 3	Si Strip pair	51.4	0, - 40mrad	2.96	17
TRT layers	36 Straw tube layers	57.6-107.5	0	*360.40**	130

Note: Beam pipe description is based on current geometry DB used for simulation, but the real beampipe according to ATLAS detector JINST paper only has one 800mm Be shell, but wrapped by heater, kapton, aerogel insulation and aluminum foil, so that my own estimate of the actual radiation length is ~0.73%. An upgrade system might be able to reduce the beam pipe material, but in order to concentrate on the relative merits of the different pixel configurations, the beam pipe is kept the same throughout as in the current simulation.

For the pixel configurations, the variations also considered the possibility for a "new sensor" with lighter support/cooling arrangements to effectively halve the material per layer, and a 250mm pixel length.

Old sensor: 50mmx400mm pixels; 10mm rf resolution, 115mm z resolution; 3.1% radiation length including cooling/support/cables.
New sensor: 50mmx250mm pixels; 10mm rf resolution, 72mm z resolution; 1.5% radiation length including cooling/support/cables.

The pixel geometry options, together with the impact parameter resolution results for h=0:

Geometry	Sensor type	Mean Raidus (cm)	Impact parameter resolution (mm) at h=0 (rf, rz)
Geometry	Sensor type	Mean Raidus (cm)	P_T=1 GeV/c	P_T=10 GeV/c	P_T=100 GeV/c
Current ATLAS	L0 Old L1 Old L2 Old	5.05 8.85 12.25	106, 212	19.8, 102	9.0, 98
Insert old b sensor	L0' Old L0 Old (dead) L1 Old L2 Old	3.70 5.05 8.85 12.25	95, 189	17.4, 98	8.7, 95
Insert new b sensor	L0' New L0 Old (dead) L1 Old L2 Old	3.70 5.05 8.85 12.25	78, 141	16.6, 76	8.6, 73
Old sensor 2 layer	L0 Old L1 Old	3.70 8.85	79, 167	15.6, 103	9.2, 102
New sensor 2 layer	L0 New L1 New	3.70 8.85	58, 115	13.7, 71	9.2, 69
New sensor 3 layer	L0 New L1 New L2 New	3.70 5.05 8.85	64, 118	13.4, 60	7.6, 58

Another way to visualize the resolution variations is to use the current ATALS geometry as reference to see the ratio of resolution with respect to the reference:

Observations:

If the new B-layer can be placed at R=3.7cm, the performance will be better than current detector despite the extra material of the dead old B-layer.

Lowering B-layer radius from 5cm to 3.7cm and halving stave material has similar effect on rf impact parameter resolution in the intermediate momentum rage of a few GeV.

For the momentum range of around a few GeV, a 2-layer configuration may appear to be slightly better than 3-layers in rf resolution benefiting from the less material, but 3-layers does better at high momentum. However, is the 2-layer configuration really sufficient for pattern recognition in high energy jets and against high luminosity pileup ?

For the 3-layer configuration, the middle layer is preferably placed at close to B-layer to get the best high momentum impact parameter resolution (but it is not clear this is better for tracking pattern recognition).

Revisiting the effect of current B-layer (Sep/08)

It is most likely that by the time the new B-layer insertion is installed, the old B-layer will still be alive for sometime. Is then the new B-layer insertion somewhat of a waste other than some extra material added ? Once we have the new B-layer, would it be better to steal the old B-layer's services to make the job easier since the old B-layer may be just dead material anyway ? A specific comparison is made to answer these questions to compare the new B-layer insertion (with a more realistic material budget of 2% r.l.) with the old B-layer completely dead or still functional:

The conclusion is quite clear: we should keep the old B-layer running as long as we could. It's not just dead material and still brings some significant resolution gain to the table. The new B-layer insertion should plan for extra service path all the way and not steal the existing services from current B-layer.

Su Dong

Last update: Mon Apr 14 18:58:36 PDT 2008