update on diffuse extraterrestrial neutrino flux search with 2000 amanda-ii data

Update on diffuse Update on diffuse extraterrestrial neutrino extraterrestrial neutrino

flux search with 2000 flux search with 2000 AMANDA-II dataAMANDA-II data

Jessica Hodges, Gary Hill, Jodi Cooley

This version of the presentation has been expanded to include work done after the meeting – where part of the nch problem was found to be related to no cross-talk cleaning in the cascade fit

Review of previous workReview of previous work

Jodi Cooley’s thesis work – presented Mons October 2003

Basic quality cuts 2-D cut against coincident muons 2-D cut against nch>50 and positive smoothness – likelihood ratio track/shower vs ldirc 2-D on cogz versus l.r. track/shower model rejection potential optimised on nch

Review of previous workReview of previous work

Cuts developed on 50% of data After nch>80 cut – 6 events on atmospheric background of 3.3 Second 50% of data yielded 4 events Combined 100% data nch>87 – 9 events on atmospheric background of 4.5 One event believed to be a coincident muon What does this excess mean?

New developmentsNew developments Reinvestigation of coincident muons using new dcorsika simulations

Discovery that the nch>50, positive smoothness cut isn’t justified (which was suggested by others earlier!)

Very recently started checking things with second 50% data, but only for nch less than 80 (less than where the cut would be)

The 50<nch<80, The 50<nch<80, positive smoothness positive smoothness

problemproblem

Jodi noticed a bump in the nch distribtion above 50 channels

The 50<nch<80, The 50<nch<80, positive smoothness positive smoothness

problemproblem

nch>50, big excess at positive smoothness

The 50<nch<80, The 50<nch<80, positive smoothness positive smoothness

problemproblem

nch>50, big excess at positive smoothness

First half of data set

ldirc versusl.r. track/shower

First half of data set

ldirc versusl.r. track/shower

Second half of data set

ldirc versusl.r. track/shower

Second half of data set

ldirc versusl.r. track/shower

Summary of the Summary of the 50<nch<80, positive 50<nch<80, positive smoothness problemsmoothness problem

Expect 14.1 atmospheric neutrinos First half data – 35 events Second half data – 15 events After cut – expect 8.2 atmos First half data – 9 events Second half data – 7 events

Summary of the Summary of the 50<nch<80, positive 50<nch<80, positive smoothness problemsmoothness problem

Data Atmospheric neutrinos

Total Pass Fail Total Pass Fail

1. 35 9 26 14.1 8.2 5.9

2. 15 7 8 14.1 8.2 5.9

F. 50 16 34 28.2 16.4 11.8

Summary of the Summary of the 50<nch<80, positive 50<nch<80, positive smoothness problemsmoothness problem

Data Atmospheric neutrinos

Total Pass Fail Total Pass Fail

1. 35 9 26 14.1 8.2 5.9

2. 15 7 8 14.1 8.2 5.9

F. 50 16 34 28.2 16.4 11.8

Summary of the Summary of the 50<nch<80, positive 50<nch<80, positive smoothness problemsmoothness problem

26 events on 5.9 is quite a fluctuation 34 events on 10.8 is as well…. ….even accounting for all the plots I’ve looked at in my life

Second half nch plotSecond half nch plot

First half nch plotFirst half nch plot

Full year nch plotFull year nch plot

Still a bump in the full year

Full year smoothness Full year smoothness plotplot

Still an excess at positive smoothness

Examine the 34 events Examine the 34 events

Look at distributions, both 1 and 2 dimensional See if any of the events lie in “bad” corners of the cut space – indicative of background

line fit velocityline fit velocity

Excess at slow velocities

smoothness, ndircsmoothness, ndircExcess at largest smoothness

Excess at largest ndirc!

likelihood ratios, likelihood ratios, up/down, track/showerup/down, track/shower

Maybe trend to lower ratios

Tendency to lower ratios as seen already

ldirc, ldirbldirc, ldirb

Nothing obviously bad

Cosine zenith, nchCosine zenith, nchMaybe less vertical than expected

Higher nch as observed already

Distributions of the 34 Distributions of the 34 eventsevents

Looks ok

Distributions of the 34 Distributions of the 34 eventsevents

Maybe a pileup toward low speed and high smoothness

Distributions of the 34 Distributions of the 34 eventsevents

No obvious disagreement

My conclusions about My conclusions about nch>50, positive nch>50, positive

smoothnesssmoothness Some of the 34 events are some kind of unsimulated background, some are good track eventsThe difference between the two data halves is certainly a fluctuation on these underlying distributions Hard to imagine a periodic signal that just happens to coincide with our data separation by file numbers

My conclusions about My conclusions about nch>50, positive nch>50, positive

smoothnesssmoothness Some of these events are very nice tracks with long direct lengths Eliminate the original cut – maybe look at cutting on line fit velocity vs smoothness Tighten other cuts Good events will survive to higher levels Bad events will get eliminated elsewhere See what happens in subsequent years

Passing rates with Passing rates with tightened cutstightened cuts

Start with cuts just described Tighten gradually until events disappear Examine passing rates of data and simulation

Passing rates with Passing rates with tightened cutstightened cuts

l.r. up to zenith-weighted downgoing fit - jkchi(12)-jkchi(11) = 35 - 53 ldirb(11) = 10 - 28 ndirc(11) = 155 - 200 | smootallphit(11) | = 0.275 - 0.05 10 cut sets defined (40 - 49)

First half data vs cut First half data vs cut levellevel

Second half data vs cut Second half data vs cut levellevel

Total data vs cut levelTotal data vs cut level

Limit setting potential Limit setting potential versus cut levelversus cut level

E2 (E) < 2.35 10-7

GeV-1 cm-2 s-1 sr-1

The cut against The cut against coincident muonscoincident muons

Original cut: slope 18 New cut: slope 4

The cut against The cut against coincident muonscoincident muons

Original cut: slope 18 MRF : 0.2348

The cut against The cut against coincident muonscoincident muons

New cut: slope 4 MRF : 0.2379

Outlook for the 2000 Outlook for the 2000 diffuse analysisdiffuse analysis

Some sort of weird effect was seen in the 1st 50%, but not in the 2nd

Eliminate the problematic cut Tighten cut against coincident muons

Explanation of one part Explanation of one part of the nch>50, positive of the nch>50, positive smoothness problemsmoothness problem

Clue (Doug Cowen) – cascade fit is very sensitive to noise hits If the Monte Carlo and data have slightly different noise, then a cascade fit comparison might not make sense

The cascade fit was The cascade fit was performed before the performed before the cross talk filter was cross talk filter was

applied!applied! Cascade fits on data have been done with all the cross talk present Monte Carlo contains no cross talk Likelihood ratios based on different hit selections make no sense

track/shower l.r. before track/shower l.r. before nch>50 cutnch>50 cut

Dramatic improvement in the agreement of this variable!

Bump in nch, correlates Bump in nch, correlates with excess at positive with excess at positive

smoothnesssmoothness

track/shower l.r., ldirc track/shower l.r., ldirc vs t/s l.r. for nch>50, vs t/s l.r. for nch>50, positive smoothnesspositive smoothness

Same plots, with Same plots, with consistent cross talk consistent cross talk

cleaningcleaning

Data and simulation now agree! (in shape)

Where they did not before

Place a new cut – data and atmos neutrinos are now both split about 50:50 about this cut

Place a new cut – data and atmos neutrinos are now both split about 50:50 about this cut

Place a different cut – excess now shows up in corner, but that’s where the E-2 is more prevalent…

Place a different cut – excess now shows up in corner, but that’s where the E-2 is more prevalent…

Examine the 22 events in Examine the 22 events in the cornerthe corner

Examine the 22 events Examine the 22 events in the cornerin the corner

Examine the 22 events Examine the 22 events in the cornerin the corner

Examine the 22 events Examine the 22 events in the cornerin the corner

Examine the 22 events Examine the 22 events in the cornerin the corner

Scanning of the events inside the green circle reveals some indication of residual cross talk – a classic “string 9/10 event”

New conclusionsNew conclusions

Once the cascade fit is correctly performed using the cross talk filtering selection, the excess of events at nch>50 and positive smoothness moves to the low l.r. / low ldirc corner of the 2D plot - which could be indicative of background However, this is the region where the E-2 signal prediction is expected to cluster We cannot cut this region away!