In this paper I describe the research I participated in during the 8 week Research Experience for Teachers (RET) program at the Santa Cruz Institute for Particle Physics (SCIPP).
The goal of my project was to use a computer simulation to distinguish between gamma rays and background radiation detected by the Gamma-ray Large Area Space Telescope (GLAST) and to develop approaches to eliminating background events from the data.
GLAST is an instrument designed to detect 20 MeV to 300 GeV gamma rays in the universe. It will be launched into orbit in the year 2005. The satellite telescope consists of three main parts: the particle tracker, the Anticoincidence Detector (ACD), and the calorimeter. There is also an information readout system and spacecraft components.
The particle tracker makes up the main body of the detector. It is composed of layers of lead sandwiched between silicon strip detectors. The silicon strip detectors respond to the passage of charged particles through them by generating a current in the nearest strips. By "connecting the dots" between hits in adjacent layers of silicon, the path of a charged particle may be recreated. The strips themselves are separated by 200 um. Detectors in each layer measuring x and y position are separated by about 2 mm.
The ACD is composed of segmented tiles, which cover the top and sides of the silicon tracker. It responds to the passage of charged particles through it.
The calorimeter is located beneath the tracker. It is made up of 8 layers crystal cells oriented perpendicular (x-y) to each other. The energy deposited in these cells may be measured and some low-resolution tracking may be attempted.
There are basically four types of particles we will consider: Gamma rays, protons, electrons and albedos.
Gamma rays are photons in the energy range of 20 MeV to 300 GeV. These are the particles we are interested in detecting and tracking. Since photons are uncharged, an individual photon is undetected until it collides with an atomic nuclei (usually in a lead layer) generating an electron-positron pair. This pair creates a track in the silicon layers resembling (in the most simple case) an upside-down "V", also known as a vertex. This vertex should point back along the original photon trajectory. For all but the highest energy photons, subsequent scattering can alter the path of the charged particle, reducing the pointing resolution of the tracks. Furthermore, the electronpositron pair may in turn generate more particles that ultimately deposit most of their energy in the calorimeter in a characteristic electromagnetic shower.
Protons are massive charged particles often referred to as charged cosmic rays. Cosmic rays have a flux that is as much as 105 times larger than that of gamma rays. Thus, effective elimination of these particles is essential. Protons interact with the ACD and the silicon creating a track. Due to the proton's high momentum, the paths are usually straight with no interaction with the detector. Occasionally protons collide with atomic nuclei and generate numerous particles (strong interaction), sometimes resulting in very messy patterns in the tracker. These events often deposit significant energy in the calorimeter but with a shower unlike the electromagnetic shower produced by gamma rays.
Electrons are smaller charged particles. At 4 GeV their flux is 1% of the proton flux. They interact with the ACD and create tracks in the silicon. Electrons may create bremsstrahlung photons that in turn convert into electron-positron pairs. Showers produced in the calorimeter will resemble those of photons with the same energy.
Albedos are photons reflected from the earth's atmosphere. They usually enter the detector from the bottom or sides. Since they are uncharged, they do not interact with the ACD, and only leave tracks as the result of a conversion to an electron-positron pair. Since albedos have much less energy than gamma rays, their conversion products have very low pointing resolution.
We use a detailed computer simulation of the detector, with all relevant parameters defined. We then apply input data resembling the appropriate particles using a Monte Carlo generator. Based on the well understood interactions of gammas, protons and electrons with the materials of the detector, the reaction of the detectors to the particles can be simulated.
We used two sets of data, one consisted of 34,460 gamma events with a range in energy, the other was an assortment of 21,464 background events (32,217 protons, 194 electrons and 2,049 albedos) in proportions resembling those existing at the proposed detector location. The detector response to each of these events was simulated and the events in each category were pooled for analysis.
We then applied existing cuts to the data. The cuts can be categorized into two basic types: Cosmic cuts and Point Spread Function (PSF) cuts. Cosmic cuts are designed to cut background events, and PSF cuts are designed to improve the quality of the tracks in terms of their ability to point back to their source.
We initially looked at four cosmic cuts; the Veto_DOCA, Surplus_Hit_Ratio, CsI_Fit_errNrm, and the CsI_Xtal_Ratio cuts. [See table 1] We found that the Veto_DOCA cut was highly effective at reducing protons and electrons. In its absence 1191 protons and 33 electrons survived. The CsI_Xtal_Ratio cut was effective at reducing albedos. Without this cut, 314 albedos were accepted. The other two cosmic cuts were much less effective and could probably be replaced. [See table 1 for cut rates]
We should note that the Veto_DOCA cut, while effectively reducing charged background particles, also eliminates a considerable number of desirable gamma events. Gamma events which trigger the Veto_DOCA cut, are conversions to electron-positron pairs which take place near the sides of the tracker. A better cut might look at more than simply the distance between a track and a lit ACD tile.
The PSF, or tracker cuts, we focused on were Fit_Type, fst_Hit_Count, Diff_1st_XY_Lyr, and Active_Dist. We basically kept these four cuts since they are key to a decent PSF.
After applying the Fit_Type, Veto_DOCA, and CsI_Xtal_Ratio cuts to the background we found that 16 events survived. At this point we wanted to look at the surviving events on a case by case basis to identify patterns in these events which could be used to create more effective cuts.
Nine of the sixteen surviving events were side-entering protons. Side entering protons resulted in a lit ACD tile. But these events failed to trigger the Veto_DOCA. This is due to the fact that they did not create a track for some distance into the tracker because their horizontal path failed to encounter a conversion layer. At the point where the proton encounters tracker material, it usually creates a response in several adjacent silicon strips. Side-entering protons are characterized by lit ACD tiles on opposite sides of the detector, usually in line with a significant event recorded in the tracker. This gives us a good pattern around which to design a cut. [See figure 3]
Three of the surviving events were side entering albedos. Side entering albedos are very tricky. Since they are uncharged, no ACD response is triggered. Interactions with the calorimeter can create crossed paths resembling electron-positron vertexes. Furthermore, since albedos have lower energy than gamma rays, they point back to their source with less reliability. The vertexes may appear to point upward even though the photon entered from below. We see no strong pattern around which to design a cut. But it seems that some of the apparent vertexes have "tails", that is, they look more like an upside down "Y" than a "V". Also, the energy deposited in the calorimeter is usually low for these events. [See figure 4]
The remaining four events were protons which converted below the calorimeter sending an assortment of particles into the detector through the calorimeter. No ACD response is triggered since the particles enter through the calorimeter. Proton conversions may create multiple particles whose tracks may cross creating a deceptive vertex. These patterns may be difficult to distinguish from desirable gammas. Varying amounts of energy may be deposited in the calorimeter. Since the particles enter from the bottom, they may deposit a disproportionate amount of energy in the bottom layers of the calorimeter. A gamma generated event, on the other hand, is more likely to deposit most of its energy higher in the calorimeter, and less in the lower layers. This suggests that a cut may be made which uses the ratio of energy deposited in the bottom layers to the overall or upper layer deposited energy. Alternatively, if the calorimeter could identify the direction of a particle moving through it, a cut for these events would be simple. In the article, Studies of Background Rejection: I. GLAST 5 x 5 Configuration, a directionality filter (CsI_E_moment) is referred to. This filter would identify upward movement through the calorimeter. Since it is not yet included as an official filter, we were unable to test its effectiveness on these events. [See figures 1 and 2]
In summary, by investigating background events which survive existing cosmic and PSF cuts, we have been able to suggest additional cuts to eliminate them. After designing these cuts they should be tested on new sets of background and gamma rays to determine their effectiveness in eliminating the former while retaining the latter.
Anti Coincidence Detector Requirements and Implications for the OLAST Trigger and Rates, Jonathan F. Ormes with Cathie A. Meetre, Alexander A.
Moiseev, Jay P. Morris, Steven M. Ritz, Jeff M. Silvis and David J. Thompson, June 2,1999
Exploring Nature's Highest Energy Processes with the Gamma-ray Large Area Space Telescope, A Science Document from the GLAST Facility Science Team
Gamma-ray Large Area Space Telescope: Quest for the Ultimate Sources of Energy in the Universe, A Summary by the GLAST Facility Science Team and NASA/Goddard Space Flight Center, January 1998
Ntuple Description Table, Dan Suson
Studies of Background Rejection: I. GLAST 5 x 5 Configuration, Jay P. Norris, Cathie A. Meetre, and Jeff M. Silvis, May 26, 1999