Gamma Ray Tracking at the Naval Research Laboratory

Gamma Ray Tracking

Astrophysics Tracking Detectors

Astrophysical gamma-ray instruments must detect single gamma rays in an environment with a low signal to background ratio.   There are a variety of telescope configurations, but a promising next-generation instrument is based on Compton imaging in the range from 0.3 – 30 MeV (see Figure 1).   The key performance parameters from which the ultimate sensitivity is derived are efficiency and background.  A space mission is constrained by both weight and power, which are the main cost drivers in a mission.  Thus, getting the best performance possible out of a detector is usually cost-effective. Gamma-ray tracking has tremendous potential both to enhance efficiency, as well as to reject background.   True cosmic gamma rays have a multiplicity of one.  In many but not all cases, the source position is unknown, and must be determined by the superposition of positions determined from many recorded events.  A single gamma ray is the simplest tracking problem possible, and one that any tracking instrument must be able to handle well.  We realized that a gamma-ray’s energy can be computed by applying the Compton scattering formula with knowledge of only the first two energy losses, and the second scatter angle [see Kurfess et al. AIP Proc 510, 789-793, 2000 ; N. Dogan and D.K. Wehe Nucl. Sci. Symp Vol 1, pp 269-273, 1993; or Kroeger et al., IEEE Trans Nucl Sci, submitted 2001 ].  Total energy efficiency can be greatly enhanced using this “3-Compton” technique , especially at higher energies (>0.3 MeV).

The key requirements for tracking and reconstruction of single gamma-rays (multiplicity=1) for astrophysical detectors, in rough order of importance are:
1.    Position resolution
2.    Multiple event resolution
3.    Detection threshold
4.    Energy resolution
5.    Passive materials/Dead layers (somewhat more important in silicon)
6.    Gaps

Arguments for the importance of position resolution at the top of the list are presented below (see section below, “Relative Importance of Position and Energy Resolution” ).

Tracking is also a background rejection issue.  Most events are from the activation of the spacecraft, passive materials, and the detectors themselves in the cosmic-ray/trapped radiation environment.   Single site events are rejected as they do not lend themselves to a Compton camera, and two site events are a special category most useful only at lower energies below 0.5 MeV.   Tracking provides a handle on identifying decays such as beta^--n gamma , beta ⁺-n gamma, e-capture -n gamma followed by pair annihilation, and n-capture gamma-rays, where n=0-many.  There are several 10s of activation isotopes that are of most concern.  However, the list of candidate activation isotopes we may expect to observe is on the order of 100.  Unlike the RIA problem, the initial position of the decay is unknown.  However, there should be a common point from which all decay products originate which may be in the active or passive volumes. This is an area of active interest in need of further development.  It will be critical if we are to succeed in developing the next-generation Compton telescope for space.

Additional references on astrophysics gamma-ray tracking detectors:
·    Aprile, E, et al., NIM A327 216, 1994.
·    Oberlack, U. et al., SPIE Vol 4141, 20 ,2000.
·    Boggs, S.E, and P. Jean A&AS 145, 311, 2000.
·    Boggs, S.E. and P. Jean, A&A 376, 1126, 2002.
 

Detector achievements: State of the art

The germanium strip detector (GSD) has proven to work reliably and provide many beautiful results with a minimum of effort.  Charge collection follows the simple planar electric field collection, x-y positions are identified by crossed electrodes, multiple interactions are identified by matching hole signals with electron signals, and depth (z) is determined by simple timing techniques.  More sophisticated pulse processing should provide enhanced performance in positions and very high rates, although strictly speaking simple analog electronics performs extremely well.

The germanium strip detector has successfully been used as a Compton camera in several configurations.  Most recently, Wulf et al. [IEEE Trans Nucl Sci, submitted] used a single 5x5x1 cm 3 to image sources at 511 and 662 keV.  These data were then processed to form a Compton image of each source using a simple “ring-sum” imaging algorithm (Figure 5).   Wulf quotes an angular resolution of the Compton camera of 7°, although this is misleading as this value represents the width of the image in his map.  In fact, the ring-sum image has a broad point-spread function, and the ring width of the individual gamma-ray cones used to make the image is much smaller than 7°. Earlier Compton work has also been reported where two germanium strip detectors [ Phlips et al. IEEE TNS 43 (1996) 1472 ] were used, one to scatter, and the second to absorb each gamma ray.  Phlips’ configuration was comparatively inefficient because of the geometry and event selection that was necessary at that time.  The efficiency of a strip detector Compton camera is potentially quite high with a compact design and 3-D readout.  Our Monte-Carlo simulations suggest that total-energy efficiencies as high as 20-40% at 0.3 to several MeV should be expected in a large instrument.  This includes allowances for tracking efficiency and passive materials [ Kroeger et al., IEEE TNS, submitted ].

The strip detector Compton camera has also been used as an efficient gamma-ray polarimeter requiring no complicated calibration or rotation to compensate for azimuthal  asymmetries [ Kroeger et al. NIM A 436, 165, 1999 ].  Results are shown in Figure 6. The polarimeter as tested did not have depth resolution implemented, but relied on the measured energy loss and known gamma-ray energy to select events that scattered in the range of 60° – 120° to maximize polarization sensitivity.  The modulation fraction (ratio of number scattered perpendicular to parallel to the plane of polarization) is measured to be 60% at 288 keV.  Adding depth resolution would make the polarimeter more efficient, and eliminate the need to know the gamma-ray energy in advance.  It’s worth noting that the incident energy is fully determined if the positions of the source, first, and second interactions, and the first energy loss are known.   This is a future projects planned by the NRL group.

Like the Compton camera, polarization required processing events with two interactions.  In both applications, events were selected with at least one non-triggered strip between the interactions so that they were clearly separated, effectively requiring a separation of ~2-4 mm in either x or y (not both) in order to clearly resolve the interactions.  This resolving power can be easily improved by going to a finer strip pitch.  However, Wulf has observed that interactions in adjacent strips can be resolved in depth, leading to the possibility that a resolving power equal to the strip pitch is a reasonable expectation.  The NRL device was tested as a polarimeter at 288 keV (90° scatter of 662 keV), and at 98 keV (90° scatter of 122 keV).  The latter experiment required detection of interactions less than 16 keV for the initial scatter in near proximity to the larger 82 keV photoelectric absorption of the second interaction.

Germanium Strip Detector – General

The x-y position resolution a strip detector is equal to the strip pitch, or possibly much better if pulse shape processing is used.  A San Francisco bay-area company (XIA) is also studying pulse shape processing to improve x-y position resolution.  This work is also going on at Argonne.  Certainly one can imagine 5 mm strip pitch with 1 mm spatial resolution using similar techniques to those being developed for the segmented coaxial devices.   Pulse shape processing may be desirable to resolve depths of two interactions that occur in adjacent in x- or y-strips.  This is an open question, and limited resolving capability in this situation has been demonstrated without it.  There could be a problem if the depth-timing mark from a small signal is masked by a large induced signal in a neighboring strip.  Clearly pulse shape processing could handle such a case.  NRL’s  efforts to date have been to avoid the additional complexity presented by digitizing what could become several thousand preamplifier signals.

On the other hand, fine strip pitch of 1 mm or even 0.5 mm pitch strips are easily fabricated, and these have several advantages over the wider strips: (1) simplicity of position readout, (2) lower probability of multiple interactions in the same pixel, and (3) sharper rising edge of the pulse used for depth sensing.  It is our expectation that a fine-pitch strip detector should easily determine the positions and energies of several simultaneous interactions without complicated data processing.

The size of a strip detector is limited by the boule size, although very large devices have yet to be demonstrated (Figure to left shows a boule being cut).  NRL is now having a 8x8x2 cm device made at LBNL.  Argonne has a similar-sized device made by Ortec with another segmentation technology.  In principle, the detector could be 3 cm thick (depth), 16 cm long, and about 8 or 9 cm wide.  A 16-cm long device would be, no doubt, be more expensive than a similar volume in coaxial detectors, as more of the boule is wasted in the processing. There is no reliable basis at this time for estimating the cost of mass-producing segmented strip devices vs. segmented coaxial devices.

Interest in the strip detector was once the almost exclusive domain of NRL and John Morse (ESRF, Grenoble) for the light source there.  It has expanded in the to include Argonne.  Interest has grown in the last two years, including U.C. Berkeley (astrophysical Compton imaging), SNM applications considered by LLNL, and detector development by two groups at LBNL.   Ortec in the U.S produces the Argonne detectors commercially.  NRL has obtained detectors from Eurisys (France), Ortec, and LBNL.  John Morse has devices from GSI/Juelich, Germany and from France.  These devices generally all use different contact and/or segmenting technologies.

Another important question is whether or not a germanium strip detector requires a guard ring.  The NRL attitude has been that the devices we have work, and they have guard rings.  NRL doesn’t make the detectors.  Detector developers at Eurisys argue the guard ring provides important handing capabilities and are more robust (better yield and tolerant of contamination).  LBNL developers have argued that the guard ring is not necessary, although their first successful devices all have guard rings.  The one device they made without a guard ring had serious field distortions due to what they describe as “surface channel” effects.  The feasibility of eliminating the guard ring has about an equal split of opinions either way among interested parties.  It is certainly a question worth further study.

Compton Camera (the simplest tracker)

Clearly one should also demonstrate the “trivial” case of characterizing single-site interactions as well.  Single site interactions can indicate non-uniformities and map performance through out the detector volume.  Dead or poorly performing volumes may not be apparent in the Compton camera experiment, where most events are discarded as “background” anyway.   These regions are best studied with single-site interactions.  Shadow-mask “images” are a quick way to study detector properties that might be missed by a few spots selected for a collimated beam.

A combination of the single-site and Compton camera experiments are strongly recommended as a detector benchmark.  At this stage, the Compton camera experiment has been demonstrated in a strip detector, but not in the segmented coaxial detector.

K. Vetter et al. [NIM A 452, 105-114, 2000] has demonstrated impressive position resolution in 3-D within a 36-fold segmented GRETA prototype coaxial detector.  Position sensitivity of 0.2 to 0.5 mm rms was shown at 374 keV for 12 positions within the detector.  Major questions remain: (1) can this performance be achieved with multiple interactions, (2) can good performance be obtained at a lower energy, and (3) can good performance be obtained throughout the full volume of the detector.  These questions are important in astrophysics and well as for GRETA, as these large coaxial detectors would make an excellent instrument in either application.

Greg Schmid et al. [NIM A 459 (2001) 565] demonstrated that the GRETA Prototype detector was not yet capable of Compton imaging at 244 keV (90 degree scatter --> 165 and 79 keV energy losses).  He concluded that the pulse-shape noise needed a seven-fold improvement in each segment, from 4 keV rms (9.4 keV FWHM), to 0.57 keV rms (1.34 keV FWHM) to perform the simple imaging task.   Even here, he restricted the valid events to those with a separation of at least 8 mm.  It is also important to note that this conclusion has not been experimentally validated; it is strictly a product of a simulation code.  However, this appears to me to be the most important next milestone for the GRETA detectors to achieve.

Relative Importance of Position and Energy Resolution

Energy resolution, though important, is not the main driver.  I.Y. Lee [NIM A 422 (1999) 195] correctly points out that energy resolution on the order of a few keV has a negligible contribution to the tracking uncertainties.  The dominant uncertainty is from position resolution when this is on the order of a few mm.  Improvements in position resolution are the most important detector property where we should be focusing our efforts.  Energy resolution on the order of 3-4 keV FWHM is not a tracking issue, though it may still be important to the physics.

Gamma Ray Interactions with Matter

High energy photons (gamma rays) interact with matter through one of four basic processes.  These are:

The photoelectric interaction, where a gamma ray is effectively absorbed by an electron bound to an atom.  The electron is stripped from the atom, carrying with it the momentum of the gamma ray, and the residual energy of the gamma ray minus the electron's original binding energy within the atom it had been bound.   The Compton interaction, where a gamma ray scatters from an electron.  The gamma ray transfers a fraction of its original energy and momentum to the electron (recoil electron).  The scattered gamma ray has a lower energy and a different direction from the original gamma ray, in what is called Compton scattering . The pair production interaction, where a gamma ray's energy is converted into the creation of an electron-positron pair.  A small fraction of the gamma ray momentum must be transferred to an atom to initiate the interaction.  The remainder of the momentum is carried away by the electron-positron pair.  The pair-production process requires a minimum gamma ray energy of 1.022 MeV before it can occur.  The positron will eventually annihilate with another electron (~10-100 ps time scale below 10 MeV), creating a pair of oppositely directed 0.511 MeV gamma rays.    Coherent scattering, where a gamma ray scatters from an electron bound to an atom imparting no significant energy.  The gamma ray changes direction, leaving no detectable energy loss at the interaction site.  Coherent scattering is generally significant at lower energies (below about 100 keV).   It is most easily observed lower-Z materials such as silicon, where the competing photo-electric process has a lower cross section.

Plot showing the gamma-ray cross-sections for germanium.