Milagro Technical Description
1 The Milagro Detector
Milagro is a water Cherenkov EAS detector located near Los Alamos, NM at 2630m above sea level, consisting of a ~5,000 m2 central (pond) detector surrounded by an array of 175 instrumented water tanks, (outriggers) that span an area of roughly 40,000 m2. Unlike scintillation arrays, the Milagro pond densely samples the EAS particles that reach the ground. Since the Cherenkov angle in water is 41o, an array of photomultiplier tubes (PMTs) placed at a depth of roughly ½ their spacing can detect nearly all of the particles that enter the water. In addition, at ground level the gamma rays in the EAS outnumber the electrons and positrons by a factor of ~4. Since the PMTs are placed below 4 radiation lengths of water these gamma rays can also be detected with high efficiency. These features give Milagro an unprecedented energy threshold for an EAS array. A second layer of PMTs under 16 radiation lengths of water is also sensitive to the hadronic component of cosmic-ray induced air showers. At present Milagro rejects roughly 90% of the background cosmic rays while retaining over 50% of the gamma-ray events. Improving this background rejection is one of our major technical objectives. A prototype instrument, Milagrito, was operated from 1997-1998. Milagrito had no capability to perform gamma-hadron separation and much lower sensitivity compared to Milagro. Construction on the central pond detector was completed in 1999 and since January 2000 has operated with a 95% duty cycle. The Milagro detector is just being completed with the addition of an array of 175 outrigger detectors.
2 The Central Detector
The Milagro pond is a 6-million gallon water reservoir, which measures 80m x 60m x 8m deep and is covered with a light-tight cover. The reservoir is instrumented with 723 20-cm PMTs on a 2.8m x 2.8m grid. The PMTs are deployed in two layers. The top layer of 450 PMTs is under 1.5 meters of water and the bottom layer of 273 PMTs is under 6 meters of water.The sides of the reservoir are sloped (2:1) so that the area of the bottom of the reservoir is smaller than the top, leading to the smaller number of PMTs in the bottom layer. The pond interior is shown in Figure 1 and an aerial view is shown in Figure 2.
3 The Outrigger Array
On average the PMTs in the pond will detect most of all electromagnetic particles that enter the pond. This sensitivity allows for the detection of extensive air showers with cores far from the pond (over 100 meters away). The shower front is not a plane, but is curved. Therefore, if the core of the air shower is outside the pond a fitted shower plane, using the wrong core position, will not be perpendicular to the true direction of the primary particle. This effect tends to degrade the angular resolution of Milagro. The outrigger array consists of 175 water tanks surrounding the Milagro pond. Each detector is a 1500 gallon water tank with an area of ~4.6m2 and 1m high. The tanks are lined with Tyvek (to reflect the light produced in the tank) and a PMT looks down into the tank. The tanks are distributed over ~40,000 m2 around the pond. The array deployment was completed in the summer of 2002 and integration into the reconstruction will be take place in early fall 2002.
By adding the outriggers we can determine
the core position when the core falls outside the pond. This not only improves
the angular resolution, it also allows us to make an estimate of the shower
energy for these events and improves our gamma hadron separation. Specifically,
our simulations predict an energy resolution of about 75% at 1 TeV improving
to about 50% at 10 TeV. (Our energy resolution is approximately lognormal, so
if we measure 10 TeV, 5 TeV is 1 sigma and 2.5 TeV is 2 sigma). The combined
effect of the outriggers is conservatively expected from simulations to improve
significance on a crab-like source by approximately a factor of two. This means
that we will get the same signal at least four times quicker.
4 Event Trigger
During most of the initial 2 years of data taking with the Milagro pond the event trigger was a simple multiplicity trigger, requiring 60 PMTs (in the top layer) to be hit within 180 ns. Milagro is capable of reconstructing gamma rays with high efficiency down to 10-15 PMTs. However, with the simple multiplicity trigger, the high flux of single muons caused the trigger rate to rise steeply below 60 PMTs, and overwhelm the 2000 Hz capability of the Milagro data acquisition system. Thus, the trigger and not the intrinsic capability of the detector limited Milagro’s sensitivity to the lowest energy gamma rays.
Since GRBs are known to be at cosmological
distances and the interstellar IR radiation fields attenuate the higher energy
photons they may emit, Milagro’s sensitivity to one of our most important
gamma-ray sources is limited by the effective area at the lower energies. In
order to increase our sensitivity to GRBs, the Maryland group has developed
a new custom VME based intelligent trigger system that lowers the trigger threshold
from 60 to 20 PMTs struck in the top layer. The new trigger works by using a
flash ADC and custom-programmed gate array system to examine the “risetime”
of the summed hit PMT signals in the trigger pulse. An air shower from near
zenith will have all of the PMTs that participate in the trigger struck nearly
simultaneously. A single muon, and the slower Cherenkov light, will have to
traverse a large distance across the pond to generate a trigger. Therefore by
requiring the risetime of the trigger to be short, one can remove single muons
at the trigger level. Figure 3 shows the relative increase in effective area
to the important low energy gamma rays with the implementation of the new trigger.
The new trigger has increased the effective area for low-energy (100 GeV) gamma
rays by a factor of 4, while keeping the trigger rate below 2000Hz. The effect
of this increased sensitivity at low energy on our physics capability for GRBs
is discussed in more detail in the GRB physics section.
5 Real-time Data Processing
Milagro records events at a rate of ~1700 Hz and, with compression, the raw data rate is ~2.5MB/sec (~80 TB/year). Given the cost and difficulties managing this quantity of data, all raw data must be reconstructed in real time. Reconstruction consists of determining the core position, incident direction, shower size, and parameters used in gamma-hadron separation. The results of the online reconstruction are archived for every event collected. Additionally, the raw data for events that reconstruct near a point of interest in the sky (Crab, AGNs, Moon, Sun) are saved for systematic study. The entire raw data set for the previous ~7 day period is buffered to disk and can be saved in response to information from other experiments.
The real time reconstruction performed
at the Milagro site requires significant processing capabilities. In early 2002,
the multiprocessor SGI computer system used for reconstruction was replaced
with an array of Linux-based workstations on a high-bandwidth network using
a single data server with multiple, inexpensive reconstruction clients. This
new system is expandable, allowing it to add clients as the reconstruction complexity
grows with development of new reconstruction algorithms that take advantage
of the outrigger array. The change to the new system was completed without significant
interruption of data collection at the experiment, and will result in substantial
savings on hardware service contracts. Los Alamos National Lab recently recognized
this system’s designer (F. Samuelson) with an award for outstanding work.