S K
Atmospheric Neutrino Oscillations in SK-I
An Updated Analysis
Alec Habig, Univ. of Minnesota Duluthfor the Super-Kamiokande Collaboration
With much help from Masaki Ishitsuka & Mark Messier
28th ICRC, 2 Aug. 2003 , Tsukuba
Alec Habig Page 2
S KUpdated Analysis
• All “SK-I” data (April 1996-July 2001) reanalyzed (1489 live-days)– Ring selection, Particle ID, multi-ring fits improved– Up- reduction automated and fitting improved (1646
live-days)
• Monte Carlo predictions improved– New 2001 Honda 3D flux (was Honda 1995)– Fermi Momentum, Axial Mass changed to better match
K2K near detector interaction data • (pF now flat, MA for QE, single from 1.01.1)
– New calibs. improve Outer Detector, H2O parameters in detector simulation (GEANT 3 based)
28th ICRC, 2 Aug. 2003 , Tsukuba
Alec Habig Page 3
S KFlux Changes
• Honda 1995 1D to Honda 2001 3D– Absolute
normalization lower– “3D” enhancement
• At low energies • Near the horizon
• But at low E, following angle is large– Smears out the peak
near horizon– So 3D-ness changes
little for Super-K (see next slide…)
28th ICRC, 2 Aug. 2003 , Tsukuba
Alec Habig Page 6
S K
Sub-GeV Data
Key: Data MC (no osc.) MC (best fit)
e-like -like
Sub-GeV
(<1.33 GeV)
3353 (Data)
3013.9 (MC)
3227 (Data)
4466.9 (MC)
MC
data
e) /(e) /(
0.0510.0160.649
Sub-GeV(stat.) (syst.)
(note no “3D” horizon peak)No cos() shape informationat the lowest energies, only flavor ratio is useful
At higher energies, directionalitybetter preserved plusshorter L no longer oscillate:cos() shape information very useful
28th ICRC, 2 Aug. 2003 , Tsukuba
Alec Habig Page 7
S KMulti-GeV data
MC
data
e) /(e) /(
083.00.699 0.0320.030
Multi GeV+PC(stat.) (syst.)
e-like -like
Multi-GeV
+ PC
746 (Data)
700.4 (MC)
1562 (Data)
2098.0 (MC)
At even higher energies, flux up/down symmetric and low-L do not have time to disappear.
Key: Data MC (no osc.) MC (best fit)
baseline L: 12800 6200 700 40 15 km
Compare to Ae-like= -0.0200.0430.005 MC A-like= -0.0030.0050.009
Observed A-like 9.5 from no-oscillation prediction!
0.0040.0280.289NN
N-NA
likeμdownup
downup
(stat.) (syst.)
28th ICRC, 2 Aug. 2003 , Tsukuba
Alec Habig Page 8
S KMore Data
Key: Data MC (no osc.) MC (best fit)
Data C
Sub-GeV
Multi-ring 208 346.4
Multi-GeV Multi-ring 439 739.4
Up through going
Up stopping
11213 srscm100.351.57
11213 srscm100.020.020.41
11213 srscm100.140.61
Measured flux:
Theoretical calc:
Measured flux:
Theoretical calc:
11213 srscm100.040.021.70 (stat.) (syst.)
(stat.) (syst.)
(theo.)
(theo.)
E ~10 GeV
E~100 GeV
• More , different E and systematics
+Nn
+Nn
SK
SK
28th ICRC, 2 Aug. 2003 , Tsukuba
Alec Habig Page 9
S K
New Oscillation Results
• For oscillation:
• Best fit: sin2(2)=1.0, m2=2.0x10-3 eV2
– 2 = 170.8/170 dof
• 90% c.l. region:– sin2(2)>0.9– 1.3 < m2 < 3.0x10-3
eV2
Contours representoscillation hypotheseswhich fit the observed dataless well with a 2 corresponding to:
28th ICRC, 2 Aug. 2003 , Tsukuba
Alec Habig Page 11
S K
Preliminary
Difference from Previous Results
• Small improvements + the same data: – but the end result
has changed by more than you might expect
• What happened?– (Note this figure
is highly zoomed)
New result@2x10-3
90% CLregions
28th ICRC, 2 Aug. 2003 , Tsukuba
Alec Habig Page 12
S K
Effects of Improvements on Fit
• Changes each of which caused m2 region to move slightly down:– flux change (Honda 19952001)
– interaction model (pF flat, MA 1.01.1)
– Improved detector simulation (OD, H2O calib.)
– Improved event reconstruction (Particle ID, ring selection, up- fitting)
• Net effect on 2 surface of several small changes in same direction is larger
28th ICRC, 2 Aug. 2003 , Tsukuba
Alec Habig Page 14
S K
Sub-Sample Consistency
Note open-ended“swoosh” shape of a one-parameterflavor ratio fit to two osc. parameters(lowest E eventsub-sample)
• Check oscillation fits using different classes of data independently – allowed regions all overlap best fit
• The low energy sub-sample’s only handle on oscillations is the /e flavor ratio– Used to be high (alone!), is now consistent with other sub-samples
28th ICRC, 2 Aug. 2003 , Tsukuba
Alec Habig Page 15
S KUnusual Models
• Ways to make disappear without , flavor oscillations include:– Lorentz inv. violation– decay, decoherence
• Fits using all available SK data strongly constrain many such models– Hard for model to get
good fit over 5 orders of mag. in E and 4 in L
– Long decay and decoherence disfavored but not eliminated
Mode Best Fit 2 P(2) 2
-
sin22sin2(1.27m2L/E)
sin22=1.00
m2=1.9x10-3 eV2
189 50% 0.0 0
-e
~sin22sin2(1.27m2L/E)
sin22=0.98
m2=4.2x10-3 eV2
304 0% 111 10.5
-s
~sin22sin2(1.27m2L/E)
sin22=0.93
m2=2.5x10-3 eV2
231 2% 42.2 6.5
LxE (L.I. violation)
sin22sin2(LxE)
sin22=0.89
=5.1x10-4 GeV/km
329 0% 103 10.1
decay (short )
sin4+cos4(1-e-L/E)
cos2=0.49
=3.2x10-3 GeV/km
287 0% 98.1 9.9
decay (long )
(sin2+cos2 e-L/2E)2
cos2=0.33
=9.8x10-3 GeV/km
207 19% 18 4.2
decoherence
0.5sin22(1-e-L/E)
sin22=0.98
=6.6x10-3 GeV/km
198 33% 9.4 3.1
Null Hypothesis 469 0% 280 16.7
(FC+PC (cut into 2 samples @Evis = 5 GeV)+NC+multiring+up-, 195 bins, 190 d.o.f.)
Data Used:(diff. from std.)
28th ICRC, 2 Aug. 2003 , Tsukuba
Alec Habig Page 16
S K to sterile?
• High energy experience matter effects which suppress oscillations to sterile – Matter effects not seen in
up- or high-energy PC data– Reduction in neutral current
interactions also not seen
– constrains s component of disappearance oscillations
• Pure s disfavored
– s fraction < 20% at 90% c.l.
28th ICRC, 2 Aug. 2003 , Tsukuba
Alec Habig Page 17
S KCPT Violation
• Do oscillate differently than ?
• SK cannot tell the difference between and event-by-event– But we see the sum of the two– One behaving very differently
would show up in the total
28th ICRC, 2 Aug. 2003 , Tsukuba
Alec Habig Page 20
S KSummary
• oscillations fit the data better than other means of making disappear – Best fit value is (m2 = 2.0x10-3 eV2, sin2(2) = 1.0)– 1.3 < m2 < 3.0x10-3 eV2, sin2(2) > 0.9 @ 90% c.l.
• Analysis improvements to– interaction & flux models– Detector simulation– Event reconstruction
• No one improvement drove the changes to the final fit– Each contributed a little in the same direction– All data sub-samples now individually consistent with the
overall fitThe presenter gratefully acknowledges support for this presentation from the National Science Foundation via its RUI grant #0098579, and from The Research Corporation’s Cottrell College Science Award