1 / 17

statistical power of mass hierarchy measurement (with ORCA) Aart Heijboer, Nikhef

statistical power of mass hierarchy measurement (with ORCA) Aart Heijboer, Nikhef. Computing oscillations probabilities. Oscillation probability given by P ab = |F ab | 2 with F the transition matrix is the flavour basis. 6. 5. 4. F = U V 6 V 5 V 4 V 3 V 2 V 1 U -1

nolen
Download Presentation

statistical power of mass hierarchy measurement (with ORCA) Aart Heijboer, Nikhef

An Image/Link below is provided (as is) to download presentation Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author. Content is provided to you AS IS for your information and personal use only. Download presentation by click this link. While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server. During download, if you can't get a presentation, the file might be deleted by the publisher.

E N D

Presentation Transcript


  1. statistical power of mass hierarchy measurement (with ORCA) Aart Heijboer, Nikhef

  2. Computing oscillations probabilities Oscillation probability given by Pab = |Fab|2 with F the transition matrix is the flavour basis 6 5 4 F= U V6 V5 V4 V3 V2 V1 U-1 The product of transition matrices computed for a matter of constant density ri. 3 2 • several options to compute exp in practice • Cayleigh-Hamilton formalism (hep-ph/9910546) • Diagonalize H • power series for exp. Cross-checked with published plots and 'Globes' package. all methods agree

  3. assumed zero Inputs for oscillations probabilities taken from most recent global-data fit Arxiv:1205.5254 note allowed ranges are not identical for NH and IH → taken into account in the following plots

  4. Oscillation probability (m → m) for straight up NH/IH difference largely degenerate with change in Dm2large beware: there are also plots around which accidentally exaggerate the NH/IH difference by comparing the central value or NH with a IH curve at the ~1 sigma edge of the allowed global fit.

  5. Other zenith angles... effect is fairly large, but... almost opposite for anti-neutrinos will be washed out by E- and angle resolutions next step: compute realistic event rates

  6. Interaction rates : ingredients neutrino fluxes (Bartol), include ne cross-sections from integrating DIS formula(approximates to ~20% genie and measurements) cross-sections

  7. Interaction rates

  8. Adding detector effects • This is where the guess-work starts; all results depend crucially on these assumptions • assume zenith angle muon is measured perfectly. resolution comes for angle between n and m • Neutrino energy resolution: assume 25% requires reconstruction of muon and hadronic shower • Acceptance: • neutrino vertex inside instrumented volume • require 15 hits(educated guess: need direction, E-resolution, • rejection of atm. muons.). assumed detector: (100m)3 =1Mton 6x6 strings 20 oms/string geant4 sim with full photon tracking

  9. Effect of detector on oscillogram: NH

  10. Effect of detector on oscillogram: IH

  11. after accounting for assumed detector resolutions: • maximum rate difference between NH and IH ~10% • to be compared to other uncertainties (e.g. on mixing angles and masses → next slide). difference between NH(central values from global fit) and IH (central values from global fit)

  12. effect of parameter uncertainties : NH each plot compares central values with +1 sigma variation in each parameter → cannot be neglected: need to deal with these (nuisance) parameters

  13. (Optimal) analysis to distinguish NH and IH Optimal observable to distinguish between NH and IH hypotheses = Maximum likelihood rato So this means, for each pseudo-experiment (data): 1. assume NH and find by maximizing (this involves computing many smeared-oscillograms for NH) 2. do the same for IH 3. compute The likehood contains a gaussian constraint representing the current knowledge from the global fit Dm221 and q21 are fixed in the fit to gain speed

  14. Results of parameter fit on (NH) peudo-experiments fitted value true value drawn from global-fit-allowed range ORCA can improve the current uncertainty for Dm2large & q23 already with 1 year of data s(q13) still dominated by other data(via Gaussian constraint in the likelihood) after 10 Mtonyr → ORCA not very sensitive to it.

  15. likelihood ratio distribution... ...for toy experiments in which the true hierarchy is normal or inverted. remember: results depend crucially on assumptions on resolution and detector layour and acceptance

  16. likelihood ratio distribution for toy experiments in which the true hierarchy is normal or inverted. remember: results depend crucially on assumptions on resolution and detector layour and acceptance. expressed in sigma's, separation betweenNH and IH = 3 sigma with 10 Mton x year.

  17. Conclusions / thoughts • Full toy analysis set up, including oscillation fit • good sensitivity to Dm2large & q23 before we can do MH • Determining MH is not an easy measurement • For assumed detector performance, need 10 Mton x year for 3 sigma • Of course, assumptions could be pessimistic • Still several sources of systematic to be accounted for • (earth density, rate normalization, cross-sections,....) • Determining the mass hierarchy with this type of detector requires Large instrumented volume (1 Mton is not enough ) • When we have full simulation and reconstruction, we should investigate sparse (or variable-density ) detectors • ( But demands on (energy) reconstruction quality are very high )

More Related