SiStrip Calibration, Local Reconstruction and Simulation meeting 27 km 100 m Status of Lorentz Angle Measurement Introduction Lorentz angle ? Problem

SiStrip Calibration, Local Reconstruction and Simulation meeting

= 27 km100 m

Status of Lorentz Angle Measurement

IntroductionLorentz angle ?

Problem :• E varies with the depth ~ average LA• Average LA varies with :

• Temperature• Drift velocity• Depth of collected charge readout mode (peak vs deconvolution)

8 December 2011

The charges drift at an angle qL

Cluster shift

S. Frosali

Th. Caebergs (UMons) – A. Nürnberg (KIT) – C.Vander Velde (ULB

IntroductionHow to measure the average LA ?Several methods developped by our precursors:• Look for minimum cluster width in # of strips (S. Frosali)

Sensitive to the higher cross-talk in deconvolution mode

Does not work for collision data (not enough w = 1 clusters)• Look for maximum probability of cluster with only 1 strip (prob1)• Look for minimum sqrt (variance) for w=2 and w=3 (avgv2, avgv3)• Look for minimum rms, for w = 2 and w = 3 (rmsv2, rmsv3)

The extrema are search for using a minimum c² symmetry fit.

All these methods give different results and the correct way to proceed is not yet understood!



8 December 2011

qL = qtrack for minimum width

(B. B

etchart)


= 27 km100 m


IntroductionMore difficulties• The difference between the methods are not the same for all modules,

even not for a category, TIB – TOB, parallel – stereo• They depend on the readout mode (charge depth influence)• They are different for cosmic runs and collision data (angular

distribution of tracks)

Question?Is it possible to measure « something » leading to a cluster position shift such that the alignment remains the same as long as the detector is not moved? Presently, we can just monitor the changes with time!

What do we monitor?The hole mobility, µ_H, versus the run #:

tg (qL) = µ_H B

This work was started by Y. Gotra.

8 December 2011


Comparison with former results

• Yuri Gotra looked at runs from April to mid-July 2010, last year

• We looked at the same runs, to check that we handle the analysis code in the right way

• We reproduced Yuri‘s results for most layers (all TIB, TOB 1-4). Example:

Yuri Gotra our analysis

8 December 2011


Different result for TOB layer 5 and 6• For some runs, we reproduce the results exactly (e.g. first 5

runs in the plots)• For the other runs, our analysis looks more stable• Differences to be understood (same files!)

Yuri Gotra our analysis

8 December 2011



Dispersion of mobility estimations• The dispersion of mobility estimations cannot be explained only by

temperature variations• There are unexplained variations with time

8 December 2011

Run #

April to mid-July2010



Dispersion of mobility estimations• ... and the values obtained can be very different for each method.

We started to investigate the reasons for those behaviours

8 December 2011

Run #



Too many high variance clusters at small angles

Typical patterns in TIB : the cowboy hatEx : TIB layer 1a – W=3

8 December 2011

Run 143657(22/08/2010)

Trying to adapt the variance methods for high variance spurious clusters

• As the average of clusters variance can be biased by the spurious clusters, we have tried other methods: the most probable value, the mean in the peak around the most probable value, the median No real improvement

Ex : TIB layer 1a (W=3) : mean most probable value



8 December 2011

tanθL - (dx/dz)reco = - 0.00101 +/- 0.00032

tanθL - (dx/dz)reco = -0.01532 +/- 0.00038

Run 143657(22/08/2010)

Trying to adapt the prob1 method for high variance spurious clusters

Change of normalisation : the TIB case• Improvement of the shape (here : TIB layer 2s)

• Fit result for tanθL - (dx/dz)reco = 0.0185 +/- 0.0006 0.0015 +/- 0.0007

Run 143657(22/08/2010)



8 December 2011

Nw=1 / Nclusters Nw=1 / (Nw=1 + Nw=2 )

tanθt - (dx/dz)reco tanθt - (dx/dz)reco

The maximum probability of w = 1 methodChange of normalisation : the TOB case• Also improvement of the shapes …• … but the statistics of W=1 clusters for TOB is too low, depleting the

center of the distribution.• and the Symmetry Fit to find the maximum seems sensitive to

asymmetry in statistics, even if symmetric in shape.

no improvement of the results (ex: TOB layer 2s)

SiStrip Calibration, Local Reconstruction and Simulation meeting 8 December 2011

Run 143657(22/08/2010)


Nw=1 / Nclusters Nw=1 / (Nw=1 + Nw=2 )

tanθL - (dx/dz)reco = 0.0059 +/- 0.0029

tanθL - (dx/dz)reco = - 0.0048 +/- 0.0032

tanθt - (dx/dz)recotanθt - (dx/dz)reco

Trying to rather clean the clusters






Cleaning is effective




Variance method




Maximum probability method


8 December 2011

Attempts to replace the Symmetry Fit• V-shaped fit function• Minimum or maximum of a fitted polynomial (order 3 or 4), around

the expected Lorentz angle


TIB 2aProb12

End 2010 Beginning 2011

CERN17 08.12.2011

Simulation originally written for lorentz angle measurement with laser in Karlsruhe

Use T-CAD simulation of electric field

Place charge carrier pairs according to laser type

Track charge through sensor volume

Electric field

Magnetic field

Diffusion

…

Calculate induced signal per strip

Up to now, no readout electronic effects included

Take gaussian fit of charge cluster to calculate lorentz shift and hall mobility

Simulation of 300um CMS sensor at different temperatures with red laser

Expected temperature coefficient:-0.16e-3 (1/Tesla)/Kelvin at 300V, 4T

Andreas NürnbergSiStrip Calibration and Local Reconstruction meeting

Expected µH temperature dependence (simulation)

SIMULATION

CERN18 08.12.2011

Temperature effects: temperature & µH


µH - avgv2 – Run 161311

Silicon temperature – Run 161311

TIB & TOB Tracker module temperature quite inhomogeneous, even among modules of the same layer

Inner layers warmer than outer layersSeveral closed cooling loops clearly visible

Temperature distribution quite stable during a run Take average

Calculate µH for each module individuallyCheck for correlation between hall mobility and temperature

CERN19 08.12.2011

Average temperature depencence (from fit) in expected order of magnitude for this specific run (161311, 23 March 2011) for most layers

Expected value (0.16e-3 1/Tesla K) determined by simulation

TIB TOB

expected value

Temperature effects – Impact on Lorentz angle

There are temperature differences in a layer as large as 30°C for inner TIB layers. Such temperature differences should lead to µH differences and cluster shifts of the order of ~0.05 1/Tesla and ~5µm

A correction taking into account the module temperature is thus probably worth to be envisaged

SIMULATION


CERN20 08.12.2011

Depletion Voltage

Depletion voltage in TIB and TOB ranges from 100V up to 300V

Depletion voltage depends on bulk doping concentration (and radiation induced damage)

Different depletion voltage of the sensors may have an effect on the Lorentz angle as it affects the electric field distribution in the sensor bulk even at the same bias voltage

Backplane Strips

V_depl=300V

V_depl=100V

|E| [V/cm]

Electric field(VBias = 300V)

300

0


CERN21 08.12.2011

Depletion Voltage – Impact on Lorentz angle

Slopes mostly compatible to zero

No dependence of Lorentz angle to depletion voltage observed

Simulation shows no dependence as well

avgv3

expected value

avgv2

expected value


TIB TOB TIB TOB

CERN22 08.12.2011

Use simulation code written for hardware LA measurements

Induce charge by laser pulse from backside

Track charge carrier drift through sensor

Reconstruct cluster and calculate shift

No dependence on depletion voltage observed

Depletion Votage – Monte Carlo

300V bias

SIMULATIONSIMULATION


CERN23 08.12.2011

Parametrize electric field distribution in sensor (An algorithm for calculating the Lorentz angle in

silicon detectors, V. Bartsch et al., Nuclear Instruments and Methods in Physics Research A 497 (2003) 389–396)

Parametrize hall mobility / drift velocity depending on electric field and temperature (A REVIEW OF SOME CHARGE TRANSPORT PROPERTIES OF SILICON, C. JACOBONI et al., Solid-State

Electronics, 1977, Vol. 20, pp. 7749.)

Calculate „local“ shift in depth z

Integrate over sensor thickness total shift

Calculate tan(θL) = shift / thickness and divide by magnetic field (here fixed to 3.8T) to obtain µh

Parametrization of hall mobility


CERN24 08.12.2011

µh expectation vs. measurement

Tested on 2 runs 163817, 2nd May 2011

177139, 25th Sept. 2011

Calculate expected µh per module using measured silicon temperature and Vdepl

from database

Compare µh value obtained from data by different estimators to this expected value

Histograms should be centered around zero

The narrower the better

Run 163817, TOB


CERN25 08.12.2011

µh expectation vs. measurement

Measured µH is ~ 5-10% smaller than expectation

Possibly due to deconvolution mode of APV

Effective sensor thickness is reduced

Deco mode is not yet considered in calculation

avgv3 (filled symbols) seems to be closer to expectation than avgv2 (open symbols)

TOB

TIB


Runs from April to October 2010

• All methods quite stable results during the whole year for TIB1a

April May June July August Sept. Oct.



• Slight increase in avgv3 (black), slight decrease in avgv2 (blue) for TOB layer 2a• Drifts also in other layers and methods visible see backup for remaining layers




8 December 2011

Monitoring for 2011 – shift after bias scan

• Jump after bias scan (beginning 2011) in TOB (here : TOB 2a)




Conclusions• To be written

May 12th, 2011

CERN30 08.12.2011

Temperature dependence of µH is like it is expected, at least on average

However for the time being the methods to determine µH seem to produce dispersions not all explained by temperature differences

No impact of depletion voltage to Lorentz angle observed in data for Run 167674, 24 June 2011

Monte Carlo simulation shows no dependence from depletion voltage, as long as sensors are overdepleted (which is true for the strip tracker)

Implemented theoretical model to estimate Lorentz angle

Expected LA from model is 5-10% larger than measured oneProbably due to missing corrections for APV deconvolution mode

Conclusions


CERN31 08.12.2011

Investigate bias scan at beginning of 2011 run for further testing of model

Implement corrections for deco mode

Get automated access to temperature database (?)

Outlook


Backup















CERN40 08.12.2011

163817 – avgv2

TIB

TOB


CERN41 08.12.2011

163817 – avgv3

TIB

TOB


CERN42 08.12.2011

177139 – avgv2

TIB

TOB


CERN43 08.12.2011

177139 – avgv3

TIB

TOB


Documents

SiStrip Calibration, Local Reconstruction and Simulation meeting 27 km 100 m Status of Lorentz Angle Measurement Introduction Lorentz angle ? Problem