Using down-going muons to evaluate cut parameters in IC22
Jon Dumm, Chad Finley, Teresa Montaruli - for ps wg phone call, Monday Nov 19, 2007
Currently, we are trying to evaluate cut parameters for IC22 in the context of a point-source analysis.
In order to demonstrate that cuts perform similarly on data and simulation we can use high quality down-going muons.
If these hq muon distributions look the same for data and sim, we can be reasonably confident that they are performing as expected on signal neutrinos.
For this study, we used a large sample of MinBias experimental data from Sept 11 and 12, 2007, Runs 109298 - 109301, representing about 48 hrs of data,
which gets a prescale factor of 1/200.
The simulation we used is IceSim 2.0 Corsika datasets 642, 645, 796, 800, and 801, representing 16 hr 45 min of livetime.
We also include coincident muons using datasets 618 and 630, representing almost 12 hrs of livetime (corrected by factor 2 from website).
These datasets are homogeneous, standard background generation, and more details can be found on the simulation webpage:
The MinBias filter selects one in every 200 triggered events, so for the simulation we keep all events and scale the rate down by 200.
At trigger level, we start with about 0.5M data events, 1M corsika events, and 1M coincident muon events.
All these data were processed up to a trial "level2", which is not final. More info about this filtering can be found here:
In brief, this level2 processing includes gulliver 1-iteration reconstruction, paraboloid, a bayesian zenith-weighted reconstruction forcing the event to be down-going, and some double-muon reconstructions which divide an event in two and reconstruct them separately.
For each quantity below, I show the distribution at trigger level and at a fairly tight cut level requiring that paraboloid give a sigma < 3 deg ( sigma = sqrt((pbfErr1/2)^2+(pbfErr2/2)^2) ) and gulliver finds at least 5 direct hits (-15,+75) ns.
Paraboloid Sigma < 3 deg
Ndir >= 5
Additionally, we require that the paraboloid fit succeeds when cutting on or showing a paraboloid result.
These are almost certainly looser than the final cuts will be.
Recall last year in IC9, we used simga<2.5 and Ndir>=9.
The experimental data is in black, the single-shower corsika is in blue, the coincident shower corsika is in red, and the total sim is in gray.
The second plot in each row is the ratio of experimental to simulated data.
Zenith (deg) - trigger
Zenith (deg) - cuts
Azimuth (deg) - trigger
Azimuth (deg) - cuts
log10(Sigma/(deg)) - trigger
Sigma (deg) - cuts
Ndir (-15,75)ns - trigger
Ndir (-15,75)ns - cuts
Reduced Likelihood - trigger
Reduced Likelihood - cuts
Reduced Likelihood - cuts + zoom
Double Muon reconstruction, Max Zenith of the two - trigger
Double Muon reconstruction, Max Zenith of the two - cuts
Double Muon reconstruction, first zenith vs second zenith (deg) - cuts
Bayesian (Zenith-weighted) reconstruction - trigger
*Really, to use down-going muons to understand how good, upgoing neutrinos look to a zenith-weighted cut, we would have to flip over the prior (zenith -> 180-zenith). This hasn't been done yet with a large number of events.
These plots merely show there is no pathological difference between sim and data.
Bayesian (Zenith-weighted) reconstruction - cuts