1 raghunath ganugapati(newt) preliminary exam 08/27/04 strategies for the search for prompt muons in...
TRANSCRIPT
1
Raghunath Ganugapati(Newt)Preliminary Exam
08/27/04
Strategies for the search for prompt muons in the downgoing atmospheric muon flux with the AMANDA Detector
2
Outline
• AMANDA detector
• Physics Goals of My Analysis
• Search strategies
3
The AMANDA Detector
19 strings
677 Optical Modules
Full 19 string version(AMANDA-II) operationalstarting in 2000
200 meters diameter
500 meters in height
4
Cherenkov Radiation
cosn
v/c, n= refraction index
We detect Cherenkov light obtained from the neutrino ice interaction as the muon travels faster than the speed of light in ice
5
Different Potential Event Origins
•Extra Terrestrial Neutrinos (E-2 spectrum)
•Backgrounds
Atmospheric Muons (E-3.7 spectrum)
Atmosphere Neutrinos and muons from conventional
mode of decay (π± , K± ) (E-3.7 spectrum)
Possible Atmospheric neutrino from charm
(E-2.7 spectrum)
6
Origin of Atmospheric Components
•The number of particles starts to increase rapidly as the shower moves downwards in the atmosphere on their way and in each interaction the particles loose energy.
• The number of particles reaching the earth depends on the energy and type of the incident cosmic ray and the ground altitude (sea level)
7
mc2
(MeV)
ct Ecritical
(GeV)
π ± 140 7.8m 115
K± 494 3.7m 855
D± 1870 317µm 3.8*107
Ecritical=(mc2/ct)*h0
h0=6.4Km
Interaction VS Decay
Ref:hep-ph/0010306 v3 19 Jan 2001
8
Prompt Muons
• Charmed particles decay before interacting hence muons from decays of charm are called prompt muon
• The flux of prompt muons differs qualitatively from ordinary muons (conventional π± , K± decay) in two ways
• The Energy spectrum is flatter (E-2.7) VS E-3.7 for conventional muons due to interaction
• The angular distribution is isotropic
9
Neutrino Fluxes
• The ZHV-D model of prompt neutrinos could not be constrained by looking at the neutrino data for one single year. We shall see towards the end if we could constrain this model of charm production from the stand point of looking at muon data.
• Can we do better with downgoing Muons?
AMANDA-II E-2
10
Neutrino Vs Muon Fluxes
The prompt muonic neutrino fluxes and the prompt muon flux are essentially the same at sea level. This result is independent of the charm production model and hence a constraint on a prompt muon is equivalent to a constraint on prompt neutrinos
Ref:GGV,hep-ph/0209111 v1 10 Sep 2002
11
Uncertainty in Prompt Muon Cross Sections
• The uncertainty spans three orders of magnitude. This is mainly because of the need to extrapolate accelerator data to very high energies and not much is known about p-p interactions. Note that the crossing between conventional to prompt muon fluxes happens between 40TeV and 3 PeV.
Ref: GGV,hep-ph/0209111 v1 10 Sep 2002
12
Charm Mechanism Vs π± , K± Mechanism
• The interaction of a high energy cosmic ray with air nuclei produces a D± which takes up most of the energy and momentum of the primary.
• Showering effect and the production of accompanying π± and K± is negligible when required to estimate the flux at the surface of the earth. Ref: Doctoral thesis of
Prof.Varieschi
13
Analysis Description
14
Signal Simulation Generation of single muons with an assumed energy spectrum of prompt muons(RPQM) and isotropic in zenith and azimuth angle at the surface of the earth; and propagate them through Earth and a detector response to these muons is obtained.
The conventional muons produced from and π± and K± decay will be a background to our detection of the charm muons. The program corsika 6.02 with the QGSJET model to simulate the hadron interactions and decay is used.
Signal ,Background Simulation and Data
Background Data
75 days life time worth data taken by the AMANDA detector during the year 2001 will be studied.
15
The distributions of various observables were studied to design our cuts to improve signal to background ratio and hence to improve our search for prompt muons.
• Zenith Angle
•Energy
•Topology(single muon and a bundle of muons)
Defining Observables
Strategies for separation of Signal from Background
16
Zenith distribution
Cos(truezenith Angle)
• The true zenith distribution of signal is flat while the distribution of background is steep
BackgroundSignal
17
•The signal over background ratio tends to improve as we go towards the more horizontal region and hence we will likely increase our sensitivity by taking a cut on the Zenith angle
Ratios
•Further more a cut on the Zenith angle acts like a natural cut on the energy at the surface due to larger distance of propagation through the Earth
b/s
Cos(zenith)
True track
Reconstructed Track
Cut these out
18
Angular Resolution
• The angular reconstruction errors are large at this stage and the zenith angle distribution of background is much steeper than the signal
• A small error in the angular reconstruction for muons at large zenith angles translates to several kilometers propagation through Earth and incorrect energy losses
Angular Resolution(Δθ)
19
Quality Cuts
• Track Length(>120m) Distance between direct hits projected on to the length of the track • Smoothness(<0.26) Measure of how smoothly the hits are distributed along the track
• Reduced Chi square(<7.3) Chisquare computed using time residuals and divided by total number
of hits
• Cascade to track likelihood Ratio(<1) Tracks that have a sphericity in the pattern of timing like cascades are
hard to reconstruct(High energy muons with stochastic losses)
20
Angular Resolution
• For muons greater than 650 in zenith the angular resolution is ~70 before the quality cuts and ~3.50 after quality cuts
Angular Resolution(Δθ)
BackgroundBackground(after Q.C)
SignalSignal
(after Q.C)
21
Energy Spectra
log10(energy) GeVNumber of Hits Vs log10(energy) GeV
•The multiple muon background goes with the same slope as the signal so the signal will be masked in the fluctuations of the multiple muon background
• True muon energy correlates with energy released inside the detector and observed through parameters like number of optical module fired and number of hits
singlesmultiples
signal
22
Data Agreement
• Data seems to be in reasonable agreement with the simulation after quality cuts and zenith cut.
Number of Hits
2001(data)B.GSignal
23
•Idea1: Single Muons should have no early hits with greater than 3.5 photoelectron.
•Idea2: Truncated cherenkov cone timing pattern fits the multiple muon hypothesis better than the ordinary cone.
A new method to separate single muons from multiple muons using the hit topology information
24
Early Hit Illustration(Idea1)
snapshot
•Think of the cherenkov structure as propagating in time relative to the tracks.
•The hit at B is earlier by a time given by
length(AB)/cice
•Noise hits are random and can occur early as well and so the 3.5 photo electron cut is to ensure proximity to the track.
Muon1Muon1
Muon2
Early Hit
Reconstructed track
A
B
25
Muon1
Muon4
Muon2
Truncated Cone timing pattern
Muon3
Muon5
In the limit that the distance between two adjacent muons becomes zero the timing pattern fits a truncated cherenkov structure
Truncated cone illustration(Idea 2)
26
•If a hit is the first hit in an OM in the vicinity of the track(0-50m) and has a negative time residual(less than –15ns) and occurs with a large amplitude (> 3.5p.e.) then it means that it is more likely to be a multiple muon event by the method described previously. I call the number of such hits per each event as “earlyhits”.
•The 3.5 Photo Electron above is the expected adc in the vicinity of the track for hits produced by unscattered photons and thus is used as a benchmark for not cutting signal events which do have noise hits.
Early Hits
27
Limitations of Earlyhits method
timedelay(ns)
Data
B.G.
Signal
zoom
timedelay(ns)
28
Vertical Muons
• The time delay distribution
• For vertical muons(<30degrees) fits well; in the region on which early hits is defined but for horizontal muons we saw it is not so good?
Clue• Angular Resolution
Data
Background
timedelay(ns)
29
Time Delay Distribution by strings
timedelay(ns) timedelay(ns)
timedelay(ns)Strings1-4
Strings1-10 Strings 11-19
DataB.G. MCSignal
50m
100m
200m
30
Dust
Dust
Clear Ice
Reconstructed track in dataTrue track
Geometrical Effect
Reconstructed track in simulation
31
Earlyhits
Earlyhits(strings1-10) Earlyhits(strings1-10)
• Multiple muon events likely to have more early hits as compared to singles
• The data agrees with the simulation to a reasonable level
• The disagreement will be understood once we have a better simulation(Photonics).
• Angular resolution of tracks and Ice properties.
Cut these
SinglesMultiples
Signal
DataB.G.Signal
32
Difference of Earlyhits between truncated fit and the cherenkov fit
Truncated Cone Hypothesis
•Early hits characterize not so good reconstruction and likelihood function has a large penality on them.
•Truncated cone is better fit hypothesis for multiple muon compared with ordinary cherenkov cone.
•Data disagreement(to be understood) the same reasons discussed previously apply.
Cut these
SinglesMultipleSignal
DataB.Gsignal
Difference of Earlyhits between truncated fit and the cherenkov fit
33
Pass Rates plot After Topology Cuts
• We reject a lot of high energy multiple muons background and this comes at the cost of reduction in signal but still we reject more background compared with signal. The cuts were picked with eye so there is a possibility of doing better!!! Number Of Hits
Multiples SignalSingles
34
Data Agreement
• An overall reasonable agreement with the data has to ensured. The systematics really need to be grinded out.
Number of Hits
DataB.GSignal
35
Agreement of Few other Observables
Cos(Reco Zenith)
smoothness
Track Length(m)
Chisquare
Data
B.G.
Signal
36
●Apply all the cuts●nb=number of predicted background events
ns=number of predicted signal events f((,P) predicted flux
Probability of an event given detector response
Make your observation and find the limit on the number of signal events (Feldman&Cousins,1999). no=number of observed events
upper limit = µ90(no, nb)
Calculate your flux limit.
90= * (µ90/ns)
Limit Setting
37
Average Upper Limit
• We cannot know the actual upper limit until we look at the data, simulations can be used to calculate the average upper limit.
• “Average upper limit”( µ90
) = the sum
of expected upper limits, weighted by their Poisson probability of occurance and is done under the assumption of zero signal. This is the same as sensitivity.
90nobs 0
90 nobs , n b
nbnobs
nobs !exp nb
The average upper limit is calculated for each restriction on the number of hits per event
Integral spectrum
Number Of Hits
Background
Signal
38
Model Rejection Potential
• The “model rejection factor” is defined as
mrf= µ90/ns
• over an ensemble of experiments the optimal selection criteria minimize the “model rejection factor”.
• The sensitivity is then given by
90= * mrf Example of determining the mrf using this method.
Number Of Hits
Best MRF=0.39Signal there=25.20Background=11.2
Number Of Hits
39
MRF Cut(Nhits)
Signal B.G Data
ZENITH 5.1(last B.G)
510 2.03 19.8 20.0
Q.C 2.79 420 1.48 1.54 20.0
topology
0.39 300 25.6 11.8 63
MRF(RPQM)
Note: These MRF’S are computed using a 30% theoretical uncertainity on the background and 30% error on the systematics(detection efficiencies)
40
Average Upper Limit on ZHV-D model
Integral Spectrum
BackgroundSignal
Average Upper limit
Number Of Hits
• Since we cannot know the actual upper limit until we look at the data, simulations can be used to calculate the average upper limit.
41
MRF ON ZHV-D Model
• The model rejection factor on the ZHV-D model assuming a 30% systematic error(detection efficiency) and a 30% theoretical uncertainty on the theoretical background is 0.1(preliminary); this means that it could be constrained by an order of magnitude with just
75 days of statistics!!!!
Best MRF=0.10Signal there=81.87Background=11.2
Number Of Hits
42
AMANDA-II E-2
Constraining Charm Neutrino models by analysis of downgoing Muon Data
43
Conclusions And Future Work
• The capability to constrain prompt neutrino models by analyzing the downgoing muon data looks promising
• The systematic error calculations need to be done in detail
• The issue of Angular resolution has to be studied in detail for a range of ice properties and a more accurate simulation(Photonics) has to be looked into
• The capability to constrain various other prompt muon model has to be studied in detail
44
DATA DESCRIPTION FOR EXAMPLE 1
Track length is correlated with quality of the event.As seen from the previous plot events with short track length have poor quality.As can be seen the MC doesn’t describe the data too for these events.
The cut is Track Length>120
Data
BackgroundSignal
Cut these
Track Length
45
DATA DESCRIPTION FOR EXAMPLE 2
The chi square is a measure of how well the track fits the timing hypothesis and is a measure of the quality of the event.Large Chi square per hit means that is a poor quality event.
The cut is
Reduced Chisquare<7.3
DataBGSignal
Cut these
Chisquare
46
DATA DESCRIPTION FOR EXAMPLE 3
Smoothness is a measure of how regular the photon density is distributed along the track and so a well reconstructed muon track is more likely to have a higher smoothness.
The cut is
Smoothness<0.26
smoothness
Cut these
DataBG MCSG MC
47
DATA DESCRIPTION FOR EXAMPLE 4
This ratio represents if an event is more track like or cascade like. And is a measure of sphericity of timing.Good quality tracks look more track like.
The cut is
Ratio>0.0
Diff of Chi squares
Cut these
DataBGSignal
48
Given a track hypothesis we can calculate the expected photonarrival times from an unscatteredCherenkov cone.
The time residual is the difference between the actual arrival time and the expected arrival time using a Cherenkov geometry
Cherenkov Geometry