1 hbd update itzhak tserruya dc meeting, april 9, 2008 april 9, 2008

22
1 HBD Update Itzhak Tserruya DC meeting, April 9, 2008 April 9, 2008

Upload: alexia-iris-knight

Post on 30-Jan-2016

215 views

Category:

Documents


0 download

TRANSCRIPT

Page 1: 1 HBD Update Itzhak Tserruya DC meeting, April 9, 2008 April 9, 2008

1

HBD Update

Itzhak Tserruya

DC meeting, April 9, 2008

April 9, 2008

Page 2: 1 HBD Update Itzhak Tserruya DC meeting, April 9, 2008 April 9, 2008

2

HBD West arm status: March 12, 2008.

HBD West fully refurbished April 9, 2008

Page 3: 1 HBD Update Itzhak Tserruya DC meeting, April 9, 2008 April 9, 2008

April 9, 2008 3

Gain vs time: installed stacks

VGEM = 511 V

Typical GEM failure rate: 2/39

Page 4: 1 HBD Update Itzhak Tserruya DC meeting, April 9, 2008 April 9, 2008

April 9, 2008

CsI evaporation CsI is evaporated onto 4 GEMs in one shot. For each GEM 3 measurements are taken at 160 nm across X axis. QE looks great

4

Page 5: 1 HBD Update Itzhak Tserruya DC meeting, April 9, 2008 April 9, 2008

5

Lab tests plans of HBD West

Since March 12: - N2 gas flowing inside the glove box is heated to speed up degassing Temperature inside glove box at ~ 38o - Work on the CF4 recirculation gas system (from LEGS) for lab tests almost ready.

While inside the glove box:1. HV test of each GEM individually, up to 400 V in N2. Done2. Install shades and light cube. Planned for this week3. Close vessel, flow CF4 and perform HV test on each module up to a HV corresponding

to a gain of 5000. Next Week

Take vessel out from the glove box:1. Install HV dividers2. Complete grounding and tubing3. Endurance HV test with CF4.4. Install in the IR

Outlook: Progress has been slow over the last month, but we are on track for installation of

both arms well before the start of run-9.April 9, 2008

Page 6: 1 HBD Update Itzhak Tserruya DC meeting, April 9, 2008 April 9, 2008

6

On-line QE monitoring

• We always wanted to have some way of monitoring (at least relatively) the QE of the CsI photocathodes.

• The method used in run 7 with a UV lamp was not at all reliable.

• After run 7, we found that the QE did not deteriorate at all in spite of the tough conditions of the run and was as good as in a freshly evaporated photocathode.

• In spite of that, we still think it would be good to have an on-line monitoring of the QE. We also want to have some means of monitoring the gain when there is no beam.

• These two goals are nicely realized with a little gadget developed at BNL that we call the light cube or the scintillation cube.

April 9, 2008

Page 7: 1 HBD Update Itzhak Tserruya DC meeting, April 9, 2008 April 9, 2008

The Scintillation Cube

241AmSemi-collimated

Spectralon Diffuse UV Reflector

SB

D

-T

rigger

Scint. Light Poisson Distr.

PMT signal

SBD signal

55Fe source

(calibrated UV light source )

April 9, 2008 7

Page 8: 1 HBD Update Itzhak Tserruya DC meeting, April 9, 2008 April 9, 2008

Cube#pe’s (zero-

method, using areas under fits)

#pe’s (zero-method, counting

bins directly)

#pe’s (using arith. mean, PA calib & PMT gain)

Spectralon2: 2.283 2.042 1.464

Black: 2.185 2.116 1.865

Lucite1: 3.020 2.967 2.247

Lucite2: 3.018 2.888 2.097

Results:

• probability of zeros in pulse height spectrum: <#pe’s> = - ln(P(0))

• using arithmetic mean of pulse height spectrum in mV, pre-amp calibration and PMT gain.

Calibrating the p.e. yield

8

Page 9: 1 HBD Update Itzhak Tserruya DC meeting, April 9, 2008 April 9, 2008

Light Cube tests on a triple GEM stack

• The number of photoelectrons in the light signal was calculated in two ways.

1. Using the 55Fe signal, whose electron output is well known, to calibrate the light signal in electrons.

1. Using the shape of the light signal and what we know about statistics to determine its mean number of electrons.

There is good agreement between the two methods and our expectation.

p.e.PMT *QE(GEM)

QE(PMT)

*mesh *GEM *

5.6MeV E si(SBU)

5.6MeV ESi(BNL)

* Containment

3.14 *27.5%

8.2%

* 0.90* 0.82*

5.6 1.41

5.6 1.0

*1.0 7.1

55Fe method gives 5.1 Shape gives ~6.1

(Systematic errors are not yet taken into account)

Expectation:

)()()(

10955

accidentalVlightVFeV

9

Page 10: 1 HBD Update Itzhak Tserruya DC meeting, April 9, 2008 April 9, 2008

Shape Fitting Method• The light signal is compared with a Monte Carlo simulation which uses a Poisson distribution around a <photoelectrons> convoluted with an exponential to simulate gain.

•The MC is run for multiple values of <pe>.

•The ratio of the MC and data is fit to a flat line of 1 and the chi2 of this fit is minimized to find the <pe> of the light signal.

•The good agreement implies that the MC is a robust description of the data!

Page 11: 1 HBD Update Itzhak Tserruya DC meeting, April 9, 2008 April 9, 2008

<pe>=15 <pe>=25

Cut is made on the total spectrum at 5 pe on the left, and at a fixed percentage of the singles on the right. E is the fraction of the solo spectrum accepted, is the fraction of the double spectrum rejected.

pe pe

E E

11

Page 12: 1 HBD Update Itzhak Tserruya DC meeting, April 9, 2008 April 9, 2008

Cut on Solo Spectrum

Scale the fraction of tagged doubles,, by 90% and plot against <pe> (the input for each MC) on the x axis. Remember that is not the quantity to be optimized because, our signal drops like the efficiency =E2.

Jitter due to finite binning

CDRRun 7

April 9, 2008 12

Page 13: 1 HBD Update Itzhak Tserruya DC meeting, April 9, 2008 April 9, 2008

Background Estimation•Estimate the combinatorial background using the cocktail.

•Usually the cocktail is used to simulated physics, i.e. particles that come from the same vertex are paired.

•Here I make random pairs to simulate combinatorial background (following Axel’s method from a few years ago).

•EXODUS generates electrons with kinematics, parent id, etc weighted appropriately.

•If the parent is a pion or an eta apply a rejection of

• Add in hadrons, charm, and then make pairs for the background.

•The ratio of the background without the HBD (=0) to with it operating (>0) is R.

April 9, 2008 13

Page 14: 1 HBD Update Itzhak Tserruya DC meeting, April 9, 2008 April 9, 2008

A Slight Digression

•The combinatorial background from the cocktail is too low compared to Run 4 data.

• Some calculations for the Run 7 BUP attributed this difference to hadron contamination in the data.

• In the best case the difference is indeed due to hadron contamination, and (almost) all of this contamination would be rejected by improved e-id using the HBD. This would improve the background reduction (R) by a factor of 2.29.

• In the worst case there would be no background reduction due to improved e-id.

April 9, 2008 14

Page 15: 1 HBD Update Itzhak Tserruya DC meeting, April 9, 2008 April 9, 2008

15

Single electron spectra

April 9, 2008

West East

• Analysis restricted to the HBD working modules acceptance, namely WS5, WN2, WN3, WN5 in the West Arm and ES1, ES2, ES3, EN3, EN4, EN5 in the East arm. • 40% most peripheral events are selected. • This analysis clearly demonstrates that HBD rejects not only most of the background it generates but also most of the background electron tracks in the Central Arms.

1

3

2

1/3

3/21/2

Page 16: 1 HBD Update Itzhak Tserruya DC meeting, April 9, 2008 April 9, 2008

• But R(<pe>) is not the quantity of merit because of course it will improve as E decreases but our signal degrades.• In the best case of improvement due to eid R->R * 2.29

Background reduction

April 9, 2008 16

Page 17: 1 HBD Update Itzhak Tserruya DC meeting, April 9, 2008 April 9, 2008

Improvement of Significance

With a Run 4 sized data set (f=1)

CDRRun 7

Seff is the background free equivalent signal ( 1/√Seff = ΔS/S)

April 9, 2008 17

Page 18: 1 HBD Update Itzhak Tserruya DC meeting, April 9, 2008 April 9, 2008

Best Case ScenarioCDR

Run 7

Best case: there was hadron contamination in Run 4 and the HBD gets rid of almost all of it. (Run 4 sized data set f=1) April 9, 2008 18

Page 19: 1 HBD Update Itzhak Tserruya DC meeting, April 9, 2008 April 9, 2008

Improvement with a larger data set

f is the increase in statistics compared to Run 4April 9, 2008 19

Page 20: 1 HBD Update Itzhak Tserruya DC meeting, April 9, 2008 April 9, 2008

20

Backup

April 9, 2008

Page 21: 1 HBD Update Itzhak Tserruya DC meeting, April 9, 2008 April 9, 2008

21

Gain derived from scintillation (I)

Scintillation hit identification:• single pad hits not belonging to any track in peripheral events

Gain determination:• Fit the range (10-50) ADC channels with an exponential function

• 1/slope increase with event multiplicity

• 1/slope = Gain . <m>

(where <m> = avrg nr of scintillation photons in a fired pad)

Assuming the nr of scintillation photons per pad follows a Poisson distribution:

A fired pad measures:

<m> =

P(0) = probability to have no hit in a pad =

<m> =

)0(1)(

)(

1

1

PnP

nnP

n

n

n!eμ μn

P(n)

e

2/)]0(ln[12/11

Pe

1

/slo

pe

April 9, 2008

Page 22: 1 HBD Update Itzhak Tserruya DC meeting, April 9, 2008 April 9, 2008

22

Gain derived from scintillation (II)

P(0) is not measured

Determine the probability P(0,th) of not firing a pad for a given threshold and extrapolate to a zero threshold

P(0,th) = 1 – [nr of fired pads (A>th)] / [total nr of pads]*

* The large pads are excluded in this analysis i.e. the total nr of pads = 93 or 94.

1/slope

April 9, 2008