Japanese KamLAND Site


List of Members
OD Status Reports

Apartment Schedule

PMT's status

Outer Detector Installation




Status of KamLAND Veto Simulation at TUNL

R. Rohm
June 1999

We have implemented a simulation of the Cerenkov veto subsystem using the GEANT detector simulation package from CERN (version 3.21), adapting an earlier implementation put together by Y.F. Wang (Stanford). This simulation uses GEANT to track the muon and any charged secondaries (chiefly electrons), with the Cerenkov light tracked by a special-purpose routine for greater efficiency. The interior detector is not modelled except for tracking the muon through the sphere and buffer oil, also in the interest of efficiency. Other aspects of the simulation are summarized below.


Photons are produced and tracked by subroutines which are called at each tracking step of a charged particle in the buffer. For information on the average PE yield and similar statistics it is sufficient to track the photons in small bunches, accounting for attenuation in the water and at reflective surfaces until a PMT is encountered. (This may need to be modified to study fluctuations in events with a small total light yield.) The current implementation does the tracking with a slight simplification of the original geometry, and the tracking algorithm will be updated to the current geometry when that is incorporated.


With realistic tracking it is necessary to address other discrepancies between the simulation and the current detector specification, since these are now the dominant uncertainties. This is currently underway and should be soon completed; the updates are mainly in the overall detector dimensions and positioning of the veto PMTs. [The detector segmentation also needs to be incorporated into the tracking.]


Another aspect requiring closer scrutiny is the modelling of light absorption in the water. Originally the simulation treated the Cerenkov spectrum without consideration of the wavelength dependence, and the attenuation length was a fixed constant. For the production of the Cerenkov light this is reasonable, but because the attenuation length varies rapidly with wavelength the total photon flux does not obey a simple exponential law.

It appears that this effect can be modelled accurately (assuming the accuracy of the input data on absorption and the quantum efficiency of the PMTs). However, it is worth pointing out that the difference between the naive model and the realistic one can give a large margin of error in the PMT yield, as indicated by the initial results of including this in the simulation.


A more accurate simulation also makes it possible to ask more detailed questions, and makes it necessary to provide a means in the simulation to answer them. With realistic tracking of the Cerenkov light the distribution of the light among the PMTs can be modelled correctly, and give a better representation of the veto signal. Refining the analysis phase of the simulation is an ongoing process, and will be the next priority after updating the geometry and attenuation model.


At this stage the primary importance of the simulation is to provide interactive feedback to the design of the detector, at least so far as decisions remain to be made. This now seems to be a realistic goal, at least so far as questions of the average expected signal are concerned. There is also a reassuring correspondence between the simulation and a straightforward BBOE (beyond the back of the envelope) calculation, which gives an a priori estimate of the ratio of the number of photoelectrons produced to the Cerenkov light deposited close to that of the simulation.

Once modelling uncertainties are reduced, Monte Carlo models eventually face the dreaded "root N". Therefore, as soon as a stable update of the simulation is produced it is advantageous to begin to begin accumulating statistics as soon as possible. It is equally important to have a well-structured run plan. The best way to beat down the statistics is to focus attention on the areas of greatest interest and uncertainty; for instance, events which have a low-energy muon entering the side of the detector at a shallow angle near the waist of the sphere will certainly constitute a tiny fraction of the total muon events, but a much larger source of potential background. Beginning the systematic study and determining these "high-risk" scenarios is the next step in obtaining useful results from the simulation.

Last updated: August 20, 1999