nd data/mc comparisons
DESCRIPTION
ND Data/MC Comparisons. Patricia Vahle University College London Fermilab Collaboration Meeting October, 2005. Outline: I. Things (I think) I understand II. Things I don’t understand. Cuts Applied. Beam Quality Cuts tor101 > = 0.1e12 pot/spill tortgt > = 0.1e12 pot/spill - PowerPoint PPT PresentationTRANSCRIPT
P. Vahle, Fermilab Oct 2005
1
ND Data/MC Comparisons
Patricia VahleUniversity College London
Fermilab Collaboration Meeting October, 2005
Outline:I. Things (I think) I understandII. Things I don’t understand
P. Vahle, Fermilab Oct 2005
2
Cuts Applied
Beam Quality Cuts• tor101 > = 0.1e12 pot/spill • tortgt > = 0.1e12 pot/spill • horizontal beam width < = 2.9 mm • vertical beam width < = 2.9 • -2 mm < = horizontal beam position at target < = 0 mm • 0 mm < = vertical beam position at target < = 2 mm • closest beam monitoring record witin 2 sec. of snarl
time • horn current > = 50 A
P. Vahle, Fermilab Oct 2005
3
Cuts Applied
• ND Data Quality Cuts• event vertex in "The Pittsburg Fid. Region"
– event vertex x position < 2.4 m – 0.6 m < event vertex z position < 3.56 m – 0.3 m < event vertex u position < 1.8 m – -1.8 m < event vertex v position < -0.3 m – event vertex 0.8 m away from coil hole
• trk.fit.pass==1 • track fit chi2/NDF < 10 • difference in planes between the u vertex and v vertex < 6 planes • number of tracks == 1 • absolute value of error in (q/p) divided by q/p < .2 if muon momentum is
obtained by curvature (not a stopper). • no other event with vertex time within 50 ns of event in question • 90% of shower energy must be in fully instrumented planes • muon charge < 1 • track vertex z position > 0.6 m
P. Vahle, Fermilab Oct 2005
4
Ely Energy Distributions
•Black—LE•Red—pME•Blue—pHE•Solid—data•Bands—MC with beam sys. errors
Data and MC in pME and pHE beams don’t match!
P. Vahle, Fermilab Oct 2005
5
Ely MC
• Old GMINOS executable, no target offsets to wfluk routine
• Black—Old ME MC• Pink—ME flux,
tgtshift=0 • All fixed in R1.18 MC• Only affects ME and
HE
P. Vahle, Fermilab Oct 2005
6
Eν Distributions Now
Much Better!
•Black—LE10•Red—pME•Blue—pHE•Solid—data•Bands—MC with beam sys. errors
P. Vahle, Fermilab Oct 2005
7
Events Accepted
Beam Config POT Events Accepted Events/1e17 POT %NC Contamination
LE10 Data 8.7e18 90531 1043.0
LE10 MC 2.4e18 25154 1061.8 8.6
pME Data 8.0e17 16457 2053.6
pME MC 3.4e18 74579 2169.1 8.8
pHE Data 5.7e17 17409 3034.6
pHE MC 1.8e18 56799 3112.5 9.6
P. Vahle, Fermilab Oct 2005
8
Eν Distributions Now
LE10
MEHE
Black—DataRed—MCBlue—true NC
Scaled to POT
There’s still an offset…Difference in means:•LE10—1.9%•ME—5.8%•HE—4.5%
P. Vahle, Fermilab Oct 2005
9
Eμ Distributions Now
LE10
MEHE
Black—DataRed—MC
Scaled to POT
More low E muons in dataMore high E muons in MC
P. Vahle, Fermilab Oct 2005
10
Eshw Distributions Now
LE10
MEHE
Black—DataRed—MC
Scaled to POT
P. Vahle, Fermilab Oct 2005
11
Lin. vs. Deweighted Eshw
LE10
MEHE
Black—DataRed—MC
•Default Shower energy Is now the “deweighted” version•Reco E_nu looks better using this version of the shower energy!
Scaled to POT
P. Vahle, Fermilab Oct 2005
12
Lin. vs. Deweighted Eshw
LE10
MEHE
Black—DataRed—MC
But. . .There’s a problem with the deweighted shower energy
Scaled to POT
Differences between the two methods suggests an underlying data/MC difference
P. Vahle, Fermilab Oct 2005
13
Lin. vs. Deweighted Eshw
Deweighted shw. energy bigger than linear shw. energy
Deweighted shw. energy smaller than linear shw. energy
DataMC
pME beam
P. Vahle, Fermilab Oct 2005
14
Track Strip Pulseheight
LE10
MEHE
Black—DataRed—MC
Scaled to # Events
P. Vahle, Fermilab Oct 2005
15
Shower Strip Pulseheight
LE10
MEHE
Black—DataRed—MC
Excess of <1PE hits in Data showers
1 PE ~= 100 sigcor
Scaled to # Events
P. Vahle, Fermilab Oct 2005
16
Shower strip times
LE10
ME HE
Black—DataRed—MC
Scaled to # Events
P. Vahle, Fermilab Oct 2005
17
Shower Strip times
LE10
ME HE
Black—DataRed—MC For shower hits > 2.5 pe
Scaled to # Events
P. Vahle, Fermilab Oct 2005
18
Number of Shower Strips
LE10
ME HE
Black—DataRed—MC
Scaled to # Events
P. Vahle, Fermilab Oct 2005
19
Number of Shower Strips
LE10
MEHE
Black—DataRed—MC For shower hits > 2.5 pe
Scaled to # Events
•Data and MC means agree within 1 strip•Mean in data > mean from MC•Few entries in the zero bin
P. Vahle, Fermilab Oct 2005
20
Ave Shower pulseheight
LE10
MEHE
Black—DataRed—MC
Scaled to # Events
P. Vahle, Fermilab Oct 2005
21
Ave Shower pulseheight
LE10
MEHE
Black—DataRed—MC Only counting hits > 2.5 pe
Scaled to # Events
•Data and MC means differ by < 1pe•Data bigger than MC
P. Vahle, Fermilab Oct 2005
22
Max. Shower pulseheight
LE10
ME HE
Black—DataRed—MC
Scaled to # Events
Few pe differences in meansData has bigger max shower pulse height than MC
P. Vahle, Fermilab Oct 2005
23
Total Shower pulseheight
LE10
MEHE
Black—DataRed—MC
Scaled to # Events
•Data mean is bigger than MC mean•But in GeV, Data<MC?•Needs a closer look at sigcor->sigmap->mip conversion
X10^3
X10^3 X10^3
P. Vahle, Fermilab Oct 2005
24
Total Shower pulseheight
LE10
MEHE
Black—DataRed—MCOnly counting hits > 2.5 pe
Scaled to # Events
X10^3
X10^3 X10^3
P. Vahle, Fermilab Oct 2005
25
Clean Timing Cut
•Time between accepted event and the event closest in time•Designed to get rid of “runt events”•Used to cut out first bin (<50 ns)•Intensity dependent
LE10
MEHE
Black—DataRed—MC
Scaled to POT
P. Vahle, Fermilab Oct 2005
26
Runt EventsT
ime
(ns)
Z (m)
Z (m)
Z (m)
Time (ns)
U (
m)
V (
m)
•Red—accepted event•Green—hits within 60 ns of event, in same slice but not in event•Pink—hits within 60 ns of event, but not in same slice•Blue—all hits from the slice that contains pink
P. Vahle, Fermilab Oct 2005
27
Event Timing
LE10
MEHE
Black—DataRed—MC
Higher intensity in MC meansThe clean timing cut throws away many more good events in the MC than in data
P. Vahle, Fermilab Oct 2005
28
Event Timing
LE10
MEHE
Black—DataRed—MC
Events in data last longer
P. Vahle, Fermilab Oct 2005
29
Event Timing
LE10
MEHE
Black—DataRed—MC Only considering hits > 2.5 pe
P. Vahle, Fermilab Oct 2005
30
Shower E vs. Duration
Reco. E shw (GeV)
Tim
e du
ratio
n (n
s)
LE10data
ME data
HEdata
MC
MC
MC
Lower energy showers last longer
P. Vahle, Fermilab Oct 2005
31
Cut on time duration
LE10
MEHE
Black—DataRed—MC
Scaled to POT
There’s still an offset…Difference in means:•LE10—1.2%•ME—5.2%•HE—3.5%
Require time duration<200 ns
P. Vahle, Fermilab Oct 2005
32
Summary
•Bug found in Ely vintage MC. Old ME and HE spectra from MC do not have wfluk reweighting•New MC looks better, but there’s still an offset•Using deweighted shower energy calculation mitigates the shift, but is this masking an underlying difference?•Excess of <1 pe hits in showers•Other low level shower quantities don’t look crazy•Timing quantities differ between data and MC
•Time between events—different intensities•Time length of events
•Future investigations•Higher levels of calibration chain?•Different slicing parameters?