1. KM3NeT – ORCA Measuring Neutrino Oscillations and the Neutrino Mass Hierarchy in the Mediterranean Sea
J. Brunner
CPPM

2. Matter Effects and Neutrino Mass Hierarchy (NMH)
DM2 ~ A matter potential must be significant
L large enough – use atmospheric neutrinos !
Distinction between neutrinos and anti-neutrinos  cross-sections!

3. Method to Determine NMH
Both muon- and electron-channels contribute to net hierarchy asymmetry
electron channel more robust against detector resolution effects:
E, θ smearing
(kinematics
+ detector
resolution)

4. KM3NeT: Next generation Neutrino Telescope in the Mediterranean Sea
Distributed research infrastructure with 2 main physics topics:
Low-Energy studies of atmospheric neutrinos – High-Energy search for cosmic neutrinos 
Low-Energy
(ORCA)
( )
High-Energy
(ARCA)

5. The ORCA benchmark design
115 lines, 20m spaced,
18 DOMs/line 6m spaced
Instrumented volume ~3.8 Mt, 2070 OM
450 m
31 3” PMTs
Digital photon counting
Directional information
Wide angle of view
More photocathode than 1 ANTARES storey
Cost reduction compared to ANTARES
Poster : Ronald Bruijn

6. Ingredients for NMH measurement
Efficient and high purity trigger algorithm for neutrino and atmospheric muon events
Exploit excellent photon counting of multi-PMT DOMs
Use causality of direct photons  water almost scattering free for visible photons
Reconstruction of cascade and track topologies
High efficiency down to relevant energies
Good resolution in energy and zenith angle
Topology Identification (track  cascades)
Atmospheric muon rejection (no hardware veto)

7. Reconstruction Methods
Dedicated methods for tracks and cascades have been developed
8 parameters are determined
Time, position (3), direction (2), energy, inelasticity
Step by step procedure (no brute force fit)
Hit selection (similar to trigger)
Vertex & Directional fit (timing)
Energy & inelasticity fit (light yield & direction/vertex)

8. Performances ne (cascades)
Poster : Jannik Hofestädt
450 m
Excellent angular resolution
Dominated by kinematics
PINGU donne ses performances pour 40 strings; donc un détecteur plus dense que 20 strings.
MONOPOD est une boite niore qui donne l’energie de la gerbe et ignore la trace. Mais on peut penser que la plupart de la lumère vue vient de la gerbe?
E_nu  PINGU   ORCA
 3         0.7         1.0
 5         1.2         1.1
 7         1.6         1.9
 9         2.0         3.2
11        2.7         4.2
13        3.4         5.7
15        3.9         7.5
17        4.7         8.7
19        5.2         9.0
Energy resolution better than 25% in relevant range – close to Gaussian

9. Performances nm (tracks)
Poster : Salvatore Galata
450 m
Excellent angular resolution
Dominated by kinematics
PINGU donne ses performances pour 40 strings; donc un détecteur plus dense que 20 strings.
MONOPOD est une boite niore qui donne l’energie de la gerbe et ignore la trace. Mais on peut penser que la plupart de la lumère vue vient de la gerbe?
E_nu  PINGU   ORCA
 3         0.7         1.0
 5         1.2         1.1
 7         1.6         1.9
 9         2.0         3.2
11        2.7         4.2
13        3.4         5.7
15        3.9         7.5
17        4.7         8.7
19        5.2         9.0
Energy resolution better than 25% in relevant range – close to Gaussian

10. Effective Mass Above 10 GeV Meff close to instrumented volume
Similar for cascades and tracks

11. ORCA Layout Optimization
20m interline spacing imposed by line deployment (sea operations)
Vertical spacing open : 6,9,12,18m spacing inter-DOM shown
Median zenith res (°)
Median Frac. E res
Examples for cascades
Resolutions stable
Meff 5-10 GeV crucial
Effective mass (Mton)

12. Flavour (mis)-identification
KM3NeT/ORCA preliminary
Probabilities to classify as track-like
The different curves show the energy resolved classification performance for different lower cuts on the fraction of
trees in the forest that agree that the event belongs to the class of track like events.
As one can see a stricter cut on the tree fraction increases the classification purity on the expense of the rate
at constant energy. Clearly, the classification rate is much more sensitive to the event energy than the purity. This
indicates that the classification benefits significantly from the increase in light output and track length at higher
energies.
Discrimination between 2 classes of events: track-like ( nmCC ) and shower-like ( nNC, neCC )
Classification using „Random Decision Forest“ machine-learning algorithm.
Discrimination mainly due to event reconstruction observables.
Discrimination of track-like ( nmCC ) and cascade-like ( nNC, neCC ) events
Classification uses “Random Decision Forest”
Better than 80% above 10 GeV for all channels but nmCC

13. Atmospheric muon rejection
Instrumental veto not mandatory
Atmospheric
n (E<20 GeV)
Atmospheric m
Few % contamination achievable without too strong signal loss
Poster : Luigi Fusco

14. Systematic Effects Various systematic effects taking into account
Oscillation parameters
Δm2, θ12 fixed; θ13 fitted within its error
ΔM2, θ23, δCP  fitted unconstrained
Flux, cross section , detector related (average fluctuation w.r.t. nominal)
Overall normalisation (2.0%)
ν/v ratio (4.0%)
e/μ ratio (1.2%)
NC scaling (11.0%)
Energy slope (0.5%)
Fitted unconstrained

15. Sensitivity to Neutrino Mass Hierarchy
Poster : Martijn Jongen
Dependence of sensitivity on time for fixed θ23 values
dCP fixed to zero for easy comparison with other experiments
Track vs shower event classification
Full MC detector response matrices including misidentified and NC events
Atmospheric muon contamination
Neutral current event contamination
Various Systematic uncertainties
Each type of events
(determined by neutrino flavour, interaction type and particle ID) now has
dedicated resolutions.

16. Sensitivity to Neutrino Mass Hierarchy
Dependency of sensitivity on θ23 for 3 years
NH easier to determine than IH
Second octant easier than first octant
When fixing dCP to zero sensitivity increases by ~0.5σ
Track vs shower event classification
Full MC detector response matrices including misidentified and NC events
Atmospheric muon contamination
Neutral current event contamination
Various Systematic uncertainties
Each type of events
(determined by neutrino flavour, interaction type and particle ID) now has
dedicated resolutions.

17. Sensitivity to Neutrino Mass Hierarchy
Dependency of sensitivity on θ23 and dCP for NH and 3 years
Best case : large θ23 and δCP = 0o
Worst case : small θ23 and δCP = 180o
Each type of events
(determined by neutrino flavour, interaction type and particle ID) now has
dedicated resolutions.

18. Sensitivity to PMNS parameters
ΔM2 – unconstrained fit in conjunction to mass hierarchy hypothesis testing
 Significant improvement of precision achievable
PDG 2014
The above figure shows the precision with which theta23 can be measured by
ORCA. Note that a 2 degree prior has been included, so this does not include a
determination of the octant.

19. Sensitivity to PMNS parameters
Θ23 – unconstrained fit in conjunction to mass hierarchy hypothesis testing
World best measurement after few years of data taking
The above figure shows the precision with which theta23 can be measured by
ORCA. Note that a 2 degree prior has been included, so this does not include a
determination of the octant.
PDG 2014

20. ORCA costs & timeline
Modular ring of up to 5-6 nodes via two cables to shore for up to
120 detection units + sea science instruments
Possibility to redirect the ANTARES cable to ORCA as second main cable
Phase 1 (funded) Deploy new main cable, junction box and a 6-7 string array in the ORCA configuration to demonstrate detection method in the GeV range.

21. ORCA costs & timeline Next Step -- 24-30 lines & second junction box
-- equipment for sea science

22. Conclusion ORCA started detector construction
Budget situation promising
Full simulation and reconstruction framework available
Systematic uncertainties under control
Further improvements possible using inelasticity
NMH determination feasible on unequalled time scale
Improvement of measurement precision for atmospheric oscillation parameters

23. END

24. Backup

25. Sensitivity studies 1) Fit parameters assuming NH
To optimally distinguish between IH and NH: likelihood ratio test with nuisance parameters
→ deal with degeneracies by fitting
Parameter estimates possibly
Using constraints
1) Fit parameters assuming NH
2) Fit parameters assuming IH
3) Compute DlogL = log( L(NH)/L(IH) )
q23 , Dm2large and normalization fitted from data

26. ORCA shower reconstruction (ne)
Res. (s): m

27. ORCA Sensitivity to Inelasticity
Use PDFs on the time residuals under the track (low-y) and shower (high-y) hypothesis
Select y-interval corresponding to highest likelihood
“total significance … may increase by
( )%, thus effectively increasing the
volume … by factor 1.5 – 2”
Ribordy & Smirnov PRD,
(muon channel only)
Work in progress
Should be further exploited
PID, NC rejection, neutrinos/anti-neutrinos…