Introduction to High Energy Astrophysics: Hard X-Ray (10 to 500 keV) Detectors

We are still in the regime of photo-ionization processes:

We are now in a regime where it is difficult to "image" the source (at least with current X-ray mirror technology--go here for recent developments, and we will talk about Suzaku soon). Clever techniques and instrument design are necessary to provide source localization. Also note that we are now in a regime where the energies of the cosmic rays are similar to that of the photons, so energy deposition measurements are not good at screening out particle events! There are two types of detectors: solid state (like CCDs) and scintillators. Mostly scintillators have been flown, but NuSTAR has a solid state CCD-like device. The theory of operation for a scintillator is very straightfoward, a photon enters the scintillating material, liberates electrons, these generate light, and the light is detected by a photomulitplier tube (PMT):

Of course, a CCD camera could be used as the detector, but the readout time of CCDs is much longer than PMTs, and they are not as useful for event discrimination. Scintillators work by converting X-ray energy into visible light. We distinguish scintillators from phosphors by defining bulk crystalline materials such as NaI and CsI as scintillators, and thin granular layers of rare earth oxysulphides as phosphors.

The alkali halides NaI and CsI, which are usually activated by a small amount of thallium impurity, have been the scintillators of choice so far in X-ray astronomy. This has been true because these materials can be made into large area crystals, have good X-ray stopping power, and are efficient light producers. NaI(Tl) was first produced in the early 1950s, while CsI(Na) came along in the mid-1960s. Of the two materials, CsI(Na) is mechanically more robust and more immune to the ravages of moisture, it is now also commonly doped with Thallium (Tl). Other materials, such as plastics and the higher-Z bismuth germanate (BGO), have well defined roles in X-ray scintillators. BGO is still too difficult to make with large areas, and so, is used for small area detectors only and sometimes for anti-coincidence shielding. Here are some important properties for inorganic scintillators (note how good CsI is!):


Below 100 keV, X-ray photon interactions for both NaI and CsI are predominately through the photoelectric effect. The energy conversion efficiency (or fraction of the X-ray energy which appears as scintillation light) for NaI(Tl) is 0.12, for CsI(Na) is 0.10, and for CsI(Tl) is 0.05. These values are true at 20 degrees C, and are all highly temperature dependent. The decay constant for the optical emission lies in the microsecond range for most inorganic scintillators, and in the nanosecond range for plastics. Thus, by surrounding an alkali halide crystal with a plastic shield, and observing with a single photomultiplier tube, scientists can use pulse shape discrimination to determine whether the energy loss occurred in the shield or the main detector. This is an excellent method of background discrimination.

Plastics, thanks to their low efficiency in detecting lower energy X-rays, are used almost exclusively as anti-coincidence shields. One example is polystyrene. Here is some data for different types of polystyrene, with UPS-92S the choice for anti-coincidence shields due to its long decay time (allowing pulse shape rejection):


[Anthracene is another plastic used as a scintillator---apparently a very good one!]

With material thicknesses of 5 mm, for both NaI and CsI, the detection efficiency between 20-100 keV is essentially unity. The energy resolution of a scintillation counter is determined primarily by photoelectron statistics, i.e. the variation in the number of electrons detected at the PMT photocathode. If one assumes this variation is Poissonian, limits to the FWHM energy resolution of NaI(Tl) can be estimated as (ΔE)/E ~ 1.67/sqrt(E). In other words, not very good. It is clear that scintillators such as NaI and CsI are useful only for their large collecting areas and their high quantum efficiency above 20 keV.

PMTs are the preferred detection devices, as they have relatively high Q.E. in the blue/green region where the scintillators emit most of their photons, and very fast response times (~ ns). They also require minimal cooling (you can use a Peltier cooler for example, so no cryogenics). These are the same kinds of PMTs used for optical photometry, e.g., the Hamamatsu R877:


Note the QE of PMT's remains low, ~ 25%. You can get "Avalanche Photo-Diodes" with QE's above 70%, but you must work out near 700 nm, and there are no good scintillators known to emit light out there. Let's look at a real design, that for the Burst And Transient Source Experiment (BATSE) on CGRO:


Module movie

Each BATSE module consisted of three PMTs that were mounted behind the scintillator, with a large conical light guide. The NaI scintillator was coated with a plastic anticoincidence shield. In fact, there were two actual detectors on each BATSE module, the Large Area Detector (LAD), and the Spectroscopy Detector (SD). The SD was smaller, but had a thicker crystal (above left), so it had better spectroscopic resolution. The LAD detector was an NaI(Tl) crystal 20 inches in diameter, and 0.5" thick (it was mounted on a quartz plate that was 0.75" thick). The SD was an NaI(Tl) crystal that was 5" in diameter, and 3" thick. Both detectors had 0.25" thick plastic scintillators for anti-coincidence discrimination. The SD had a thin beryllium window to improve low-energy response. Before we go too much further, we need to introduce the concept of effective area. We already encountered it when we were looking at the ROSAT telescope. While the physical area of an X-ray telescope or detector does not change with wavelength, its response to radiation varies with energy. That is, there is a certain efficiency for detecting photons of various energies depending on the detection scheme/material. Thus, the effective area (EA) of an X-ray detector is simply an efficiency multiplied by the physical area: EA = a*PA. In general, the EA drops with energy (but not always due to electron shell configurations). Here is the effective area for BATSE (compare to ROSAT):

Note that an effective area is usually measured at normal incidence as shown in this plot. As can be seen in the diagram above, there is NO collimation of the BATSE detector--it simply detects any photon that hits the scintillator! Even low-angle events were recorded (of course the efficiency in this case was poor!). So, how did you know where you were looking? In fact, it was extremely difficult. One of the clever solutions was the way the detectors were deployed, there were eight of them on CGRO, one mounted on each corner of the spacecraft (four are highlighted below):


Due to the angular response of the detectors, the maximum count rate occurred in the detector that was most favorably oriented toward the source. By reconstructing the flux ratios of the four detectors that could "see" a source at any one time, you could "localize" a source to about 4 degrees. To get to this precision, much lab-testing of the angular response of each detector occurred before launch. The pointing could then be reconstructed using these complicated response matrices. For persistent sources, however, there was an even better way to detect precise positions of sources through the "Earth occultation" measurement. This method uses the fact that in low-Earth orbit, a source is occulted by the limb of the Earth as the spacecraft orbits---as the source sets below the limb its flux disappears, and then reappears when it rises again on the other side of the orbit:

This "step flux" method was an excellent way to get a background-free flux for a source and a localization (note that the precession of the spacecraft's orbit and the Earth's orbit around the Sun changed the rise-set times to truly localize the source). See McNamara et al. (1998) for a 4.5 year light curve of Sco X-1 constructed with this method using the BATSE SDs (which had a much better low energy response):

The diagram showing Cyg X-1 above also shows the kind of spectral response delivered by BATSE: wide bands (and the large hard X-ray background!). BATSE had good sensitivity from about 25 keV up to about 500 keV--though above 300 keV, the thin NaI crystal had lower efficiency due to it becoming "transparent" to the hardest photons. Generally, BATSE "spectroscopy" was reported in four bands, or channels: 20 - 50 keV, 50 - 100 keV, 100 - 300 keV, and > 300 keV. For bright sources, the SDs had significantly better resolution, with about 20 true spectral channels available (though the LAD actually had 128 pulse-height spectral channels, and the SD had 256). CGRO carried a spectrometer that was based on almost identical technology called the Oriented Scintillation Spectrometer Experiment. As the name suggests, it could "point" the detectors. It also had a collimator. The layout is shown, below:


The collimator consisted of a grid of tungsten "slats" that limited the field of view to 3.8 degrees. Thus, to localize a source, OSSE would rock back and forth to "peak-up" on the source and figure out its location while at the same time acquiring a background---not too different from the way we take infrared data. Note that the entire device was surrounded by an "anti-coincidence" NaI shield, as in all other hard X-ray detectors, it could detect things it wasn't pointed at! Here's the overall design:


The detector was a phoswich constructed of both NaI and CsI, giving it a spectral response from 5 keV to 10 MeV, with the following resolution performance:


The total effective area was not especially impressive (multiply by four):


Note that the LADs had effective areas (per detector) of about twice this---thus, OSSE could only create useful spectra for the brightest sources in the sky. There is not a heck of a lot you can do to improve this without building some sort of hard X-ray telescope, or huge detectors! The next generation of detectors will mounted behind a hard x-ray telescopes and are Cadmium-Zinc-Telluride (CZT) "pixellated" detectors. CZT is a room temperature solid-state detector that has better energy resolution (10%) than CsI or NaI, and its high-Z means better stopping power for harder photons. This means thinner detectors, which reduces background. A CZT detector array is in operation as the BAT instrument on the Swift and is the detector on the NuStar mission:

The crystals themselves are only 4 x 4 mm (2mm thick). The BAT has 32,768 of these (it also employs an enormously complicated Coded-Aperture mask).

Here is the cross-section of a CZT detector:

"This figure shows a cross section of a detector using the principle of the drift-strip method. The drift-strip electrodes are biased in a way that makes the electrons move towards the anode strip. The structure consists of 14 drift-strip electrodes and 1 anode read-out strip. The drift-strips provide an electrostatic shield, which means that the movement of the positive carriers only induces a rather small signal at the anode strip [1]. The signal from the planar electrode is strongly influenced by the holes. Combining the signal of the planar electrode with the signal from the anode strip, it is demonstrated that it is possible to correct for any signals from the holes that might still influence the signal from the anode strip." CZT detectors apparently have many defects that trap the "holes", so they have to be circumvented, and apparently this method does that. And, gives pretty good spectral resolution:

(Right hand plot is the new drift-strip technology vs. old, single planar system.)

NuStar uses two grazing incidence telescopes, but here the grazing incidence angle is tiny (0.150 degree), thus it requires a very, very long focal length telescope (on a 10 meter mast):

(go here for more)


Introduction to High Energy Astrophysics: Gamma-Ray Detectors

MeV Detection: At MeV energies, photoionization is no longer an efficient photon stopping process (though not hopeless by any means, remember the response for OSSE), and we have to turn to Compton scattering for efficient photon detection:
Here is the diagram and formula for Compton scattering:

It is obvious from this diagram that source localization is going to be a challenge, since photon localization is a challenge! A detector system based on Compton scattering shares many of the properties and designs with the hard X-ray detectors, in fact one component of the system is merely a hard X-ray detector. To show you how it works, we need to look at COMPTEL, the only system ever flown that used the Compton scattering process. Here is a diagram of the "telescope":


Note that it is a "double-scatter" system. An incoming gamma-ray interacts with an electron in the scintillating material, in this case a liquid (NE 213A) inside a cylindrical container (8.5 x 27 cm). The electron is Compton scattered which ionizes the scintillator material causing a flash of light that is detected by an array of eight PMTs. What are liquid scintillators made of? Stuff that has lots of carbon and hydrogen atoms (lighter fluid, mineral oil, and such) because "The Compton scattering process, at sufficiently high photon energy, depends only on the number of electrons in the scintillator and not upon the nature of the nuclei. This process deposits only a fraction of the ionizing photon’s energy in the absorbing medium and thus the resulting scintillations are not a simple function of the initial energy." Here are the characteristics of NE 213A:

This stuff comprises the upper, Compton-scattering detector "D1" (composed of seven individual detector modules). A lower energy gamma-ray then passes on through D1 and travels to a NaI scintillator located 1.5 m below the upper one. The fourteen NaI scintillators used on "D2" are thick (7.5 cm), and each are viewed by seven PMTs. All of the detectors have plastic anticoincidence shields. The total geometric area of D1 was 4188 sq cm, and D2 was 8620 sq cm. To measure the energy of an incoming photon, the following scheme is used:
  1. The energy of the recoil electron of the Compton-scattered gamma-ray in the upper detector (E1), determined from the summed pulse heights of the eight PMTs associated with the triggered D1 module;
  2. The location of the interaction in the upper detector, determined from the relative pulse heights of the eight PMTs associated with the triggered D1 module;
  3. The pulse shape of the scintillation in the upper detector, provided by the output of a pulse-shape discriminator circuit;
  4. The energy loss in the lower detector ("E2", see figure below), determined from the summed pulse heights of the seven PMTs associated with the triggered D2 module;
  5. The location of the interaction in the lower detector, determined from the relative pulse heights of the seven PMTs associated with the triggered D2 module;
  6. The time-of-flight of the scattered gamma-ray from the upper to the lower detector
The energy of an incident cosmic gamma-ray is estimated by summing the energies deposited in the upper and lower detectors, assuming total absorption of the scattered photon. From the energy losses and the interaction locations recorded for both the upper and lower detectors the arrival direction of an incident gamma-ray is calculated by application of the Compton scattering formula. For a true Compton scatter event the arrival direction of a cosmic gamma-ray is known to lie on a so-called "event circle" on the sky. The center of the circle is the direction of the scattered gamma-ray within the telescope, and its angular radius is derived from the energy losses in both detectors:

In the ideal case of totally-absorbed Compton-scattered photons, the intersection of many such event circles on the sky yields the location of the gamma-ray source. In actual practice a detailed understanding of the properties of the scintillation detectors, and of the response of the telescope to background, multiple-interaction, and partial-absorption events, must be exploited to determine accurate source locations. Some statistics:

a. Effective area for gamma-rays: 20 - 50 cm2
b. Energy range: 0.8 - 30 MeV
c. Energy resolution: 5 - 8% (FWHM)
d. Angular resolution: 1.7 - 4.4 degree (FWHM)
e. Minimum source detectability at 3 sigma, 1-30 MeV (2-week observation) = 1.6 x 10-4 photon cm-2 s-1

COMPTEL actually had better source localization than BATSE, with some GRBs localized to about 1 degree. No currently existing or missions on the drawing board will have anything like COMPTEL on them (though I am told they are still planning a follow-on of some sort). One odd side benefit of COMPTEL was its ability to detect cosmic neutrons. In fact, COMPTEL produced an "image" of solar flare in the "light" of neutrons! The first non-electromagnetic astronomical images:

Energetic Gamma-Ray Detectors (30 MeV to 30 GeV)

In this energy regime, pair-production is the most efficient photon detection method:

pp          pprod

The pair production cross-section is dependent on Z:


So, high-Z materials are the best "stopping agents" when trying to induce pair production. Astronomical detectors at these energies share much in common with the ground-based spark chambers used in particle accelerators, and cosmic ray detectors. The main difference being the hope to image astronomical sources. So, let's look at the highly successful EGRET instrument flown on CGRO. Here is a cutaway:


EGRET was sensitive to to Gamma rays in the energy range from about 30 MeV to 30 GeV. In the mode used for most of the observations, the effective area of the telescope is about 1000 cm² at 150 MeV, 1500 cm² around 0.5-1 GeV, decreasing gradually at high energies to about 700 cm² at 10 GeV for targets near the center of the field of view. EGRET's effective area is maximum when the target is on axis and falls to approximately 50% of this value when the angular offset reaches 18°.

Obviously the entire instrument is surrounded by an anti-coincidence shield. An incoming photon encounters a series of closely space plates, and creates one (or more) particle pair(s). This particle is "tracked" using an imaging system---that is, the electron-positron pair has an expected path through the chamber, so it allows for both event discrimination, as well as directional information. The upper chamber consists of 27 thin tantalum foils, interleaved with "digital spark chambers". The pair production is expected to occur in the high-Z (181) tantulum. The ionizing track of the particle pair produces a spark in the spark chamber. If the incoming event triggers through the device correctly (downward path, correct time-of-flight), the "imaging" mode of the spark chamber was triggered to follow (reconstruct) the flight of the particle/anti-particle. This track is well defined, and thus the track gives angle of incidence information. At the bottom of the upper spark chamber is a large (4 x 4) array of plastic scintillators. There was a duplicate array at the bottom of the time-of-flight coincidence system. The correct path between these two arrays was the triggering device for a Gamma-ray event (this turned on the high voltage in the spark chamber to create the ionized path). Between the two plastic scintillators was a series of widely spaced spark chambers used to insure the correct particle paths were followed and to screen out other non-gamma events. At the bottom of the instrument there were 36 blocks, each 21 cm thick (for a total of 76 x 76 cm square array) of NaI(Tl) scintillator--the "Total Absorption Shower Counter" (TASC). Below this was a matrix of 16 PMTs, whose output went to an analyzing system.

A spark chamber shares much with a proportional counter, as well as the ROSAT PSPC we described earlier, it is a series of crossed grids of wires (or strips). High voltage is applied between these wires, and the ionized path through a gas made by a high velocity particle causes a spark to jump. By having a grid of wires, good x-y positional information is obtained, and the location of the event can be identified:


The technology is advancing, including solid state spark chambers that are used in GLAST. For the energetic Gamma-rays, most of the energy was deposited in the TASC. For the other events, the summation of the energy lost in the sparks, in the plastic scintillators, and in the TASC were combined to given you the total Gamma-ray energy. EGRET was a massive instrument, weighing-in at 1830 kg. It had a large effective area: 1500 sq. cm. It also had the best source localization of any CGRO instrument, ranging down to +/-5' for strong, hard sources (30' at worst). It certainly was the most productive of the CGRO instruments, and a next generation, the "Large Area Telescope" (LAT) device is the heart of GLAST.


From their website: "The GLAST LAT has a field of view about twice as wide (more than 2.5 steradians), and sensitivity about 50 times that of EGRET at 100 MeV and even more at higher energies. Its two year limit for source detection in an all-sky survey is 1.6 x 10−9 photons cm-2 s-1 (at energies> 100 MeV). It will be able to locate sources to positional accuracies of 30 arc seconds to 5 arc minutes. Yet, it is a relatively small and inexpensive mission, which will be able to be launched on a Delta II rocket."

"The two characteristics that the LAT will measure for each incoming gamma-ray are the energy of the photon and the angle at which the light ray hits the detector. These measurements will enable scientists to determine the location on the sky that produced the gamma-ray and the energy contained in that gamma-ray. The telescope consists of a tracker, followed by an energy-measuring calorimeter. The entire telescope is surrounded by anti-coincidence shielding, in order to eliminate signals which might be generated by background particles, such as cosmic rays. When a gamma-ray comes in contact with the converter material, it interacts to create a electron and positron pair. This interaction is called pair production. Each electron and positron then travels through the subsequent layers of tracking detectors and converters. For very energetic particles, further interactions occur, which produce additional pairs. The tracking detectors record information about the paths taken by the particles that are generated in the shower. The calorimeter records information about the energy of the particles in the shower. On-board analysis of the tracker and calorimeter data provides initial information about the energy and direction of the shower and helps filter out additional background signals. The on-board triggering system will reject over 100,000 background signals for each gamma-ray photon that it identifies. The remaining signals are telemetered to the ground where they are further processed to determine the energy and direction of the gamma-ray photon."

The primary interaction of photons in the GLAST energy range with matter is pair conversion. This process forms the basis for the underlying measurement principle by providing a unique signature for gamma rays, which distinguishes them from charged cosmic rays whose flux is as much as 105 times larger, and allowing a determination of the incident photon directions via the reconstruction of the trajectories of the resulting e+- e- pairs. Incident radiation first passes through an anticoincidence shield, which is sensitive to charged particles, then through thin layers of high-Z material called conversion foils. Photon conversions are facilitated in the field of a heavy nucleus. After a conversion, the trajectories of the resulting electron and positron are measured by particle tracking detectors, and their energies are then measured by a calorimeter. The characteristic gamma-ray signature in the LAT is therefore:

  • (1) no signal in the anticoincidence shield,
  • (2) more than one track starting from the same location within the volume of the tracker, and
  • (3) an electromagnetic shower in the calorimeter.

    The baseline LAT is modular, consisting of a 4 x 4 array of identical towers. Each 40 x 40 cm2 tower comprises a tracker, calorimeter and data acquisition module. The tracking detector consists of 18 xy layers of silicon strip detectors. This detector technology has a long and successful history of application in accelerator-based high-energy physics. It is well-matched to the requirements of high detection efficiency ( > 99%), excellent position resolution (<60 microns in this design), large signal:noise (>20:1), negligible cross-talk, and ease of trigger and readout with no consumables. The calorimeter in each tower consists of eight layers of 12 CsI bars in a hodoscopic arrangement, read out by photodiodes, for a total thickness of 10 radiation lengths. Owing to the hodoscopic (tracking) configuration, the calorimeter can measure the three-dimensional profiles of showers, which permits corrections for energy leakage and enhances the capability to discriminate hadronic cosmic rays. The anticoincidence shield, which covers the array of towers, employs segmented tiles of scintillator, read out by wavelength-shifting fibers and miniature phototubes.

    The LAT is self triggered; events that cause detector hits in three planes automatically trigger readouts of each tower and the anticoincidence system. Efficient rejection of the charged particle background, which is thousands of times more intense than the celestial gamma-ray radiation, is essential for GLAST to function. (The expected raw trigger rate in orbit will average a few kHz, and the rate of celestial gamma rays will be a few Hz.) The anticoincidence system is only the first line of defense in identifying cosmic rays that trigger the telescope. As described above, from simulations, other discriminators against charged particles have been developed to further reduce the background level. Some of the discriminators will be applied onboard to reduce the trigger rate to the ~30 Hz rate that can be stored and downlinked.

    Ultra-High Energy (100 GeV to 100 TeV) Gamma-Ray Detection: Photons that hurt!

    For the highest energy regime, we return to ground-based detection:


    The process is simple, a high-energy Gamma-ray enters our atmosphere, interacts with the atoms there, and pair-produces. These particles then interact, through bremsstrahlung and Compton scattering, giving up some of their energy to creating energetic photons, which in turn pair produce creating more electrons which then bremsstrahlung, etc. This creates an "air shower":

    For the lower-energy events, the shower is attenuated high in the atmosphere. But for UHE events, the particles can make it to the ground. The particles created in the process are traveling near the speed of light, and hence emit "Cerenkov radiation" (because they are in a medium with index of refraction n > 1):


  • [β is the ratio of the speed of the particle over that of light: v/c.]

    The light cone created by a photon is different from that created by cosmic rays, so particle events can be screened-out:


    So, an "image" of the event can discriminate between real Gamma-rays and cosmic rays. Here is some real data:

    How are these events observed? By large PMT arrays at the focus of big light collectors such as Whipple:

    The next generation systems (e.g., Veritas) put together an array of large light collectors for several reasons. The first is localization. The shower cone has a specific size, by imaging it you can better reconstruct the location of the incoming Gamma-ray. The other is simply due to rarity of events---the more physical area you cover, the more likely you are to see these events.

    Detecting the Cerenkov radiation is one way to "see" these UHE Gamma-rays, but of course, you can detect the particle shower too. One such experiment is Milagro, near Los Alamos:


    A huge pool of ultra-clear water filled with PMTs:



    Repair operations require unusual astronomical training:


    For more on Very High Energy Gamma-ray astronomy, click here