I5 : Gamma-ray instrumentation development

The project ended in December 2014.

  • Introduction

    Supernovae, the end point of the massive stars evolution, occur at a rate of 2-3 per century in a Galaxy such as the Milky Way. The radioactive elements synthesized in these explosive events decay subsequently and emits gamma-rays. In the days following the explosion, the gamma-ray photons loose energy through inelastic scatterings in the thick supernova envelope and emerge as visible light revealing the event. In the following months, as the envelope extents and get thinner, gamma-rays can escape. This gamma-ray line emission provides a direct measure of the amount of synthesized species.

    The atmosphere being opaque to gamma-rays, large space telescopes have been developed to measure the gamma-ray lines (0.5 – 1.3 MeV) emitted by the unstable nuclei forged during supernova explosions. However despite the numerous supernovae observed in the visible light during these high-energy observatory lifetimes, in only one occasion, gamma-ray lines from a supernova have been clearly detected by a telescope on board a satellite. All the other supernovae were too distant to allow for a detection of the emitted gamma-ray lines. One way forward to ensure these detections in a mission lifetime is to build more sensitive and more expensive detectors but, before funding them, a clear demonstration of their capabilities is required.

    SN_1987a_CO56_Matz_1988_0.pngOne can then conceive developing new gamma-ray detectors in a continuous R&D program and prove the capability and ruggedness of the products by performing observations from a high altitude (40 km) balloon. After each successful demonstration flight, the payload (or a copy) can be stored until a better detection payload has demonstrated its superiority and will replace the previous one. That way a qualified payload is always ready to fly and continuously improved. Should a supernova explode in our Galaxy or in the local group, a flight can readily be designed to observe the supernova a few months after the explosion with the best possible detector technology. In fact most of the detections of the 56Co line from a supernova (SN1987a) were made with modest gamma-ray detectors on board balloons (see figure aside). Should the progress of the detectors be such that it is proven (with margins) that any supernova in the Virgo cluster (the nearest one) will be clearly detected by a large satellite experiment, a proposal to develop such a mission can be submitted in response to the first ad hoc announcement of opportunity of a space agency. Meanwhile regular flights can be scientifically dedicated to specific objectives either in astrophysics or atmospheric physics.

    Since the Comptel experiment on board the CGRO mission, little progress has been made in the observation of the sky between 1 MeV and 50 MeV. The Compton scattering dominating the other photon-matter interactions up to 8 MeV, the experimental community has been very active to study various concepts of Compton telescopes using scintillators or/and semi-conductor detectors. However the exercise is very difficult and many of the bright ideas based upon Monte-Carlo simulations were revised after an unsuccessful balloon flight. Here clearly the balloons have proven to be very useful in avoiding the building of a large, expensive but inefficient satellite experiment. We think that the progress in microelectronics should permit today the building of an original concept, a “pure” silicon Compton telescope, without heavy calorimeter. Testing on a balloon a breadboard of such a telescope would be very useful. Moreover, an interesting astrophysical observation seems at reach for this breadboard. The linear polarization of the Crab nebula has been observed at visible, X-rays and gamma-ray wavelengths. The surprising result is that while the polarization angles observed in the visible and in gamma-rays (E>200 keV) agrees (~ 123°), that observed in X-rays (2–5 keV) differs significantly (156°). There should be transition spectral regions around the X-rays. Compton telescopes are naturally very good polarimeters. A measurement in the range 50 – 200 keV will refine the position of the change of regime between the X-rays and the gamma-rays and set constraints on the synchrotron nebula model.

    AIM and APC have in common a long history of space gamma-ray detectors (COS-B, GRANAT/SIGMA, INTEGRAL/ISGRI) and are still actively developing new ones (TARANIS/XGRE).

    One can note that the total duration of a balloon project is of the order of a few years, i.e. close to the nominal duration of a PhD. A student involved in the detector development, the instrument and mission design, the payload system aspects, the attitude stabilization and/or restitution, the data flow, the telemetry and the data analysis will develop the skills that are necessary to drive or contribute to scientific projects. In other words it is a very good Principal Investigator and Instrument Scientist school. On the technical side, students from engineering schools can be also involved in such a project, dealing with the mechanics, the thermal control, the electronics and the project management, all under the planning pressure. This is a way to educate future System Engineers and Project Managers.

    Description of the measurement technique

    Taking advantage of the Compton scattering formula to constrain the incoming photon direction was the original idea of Compton telescopes. Provided the scattered photon is absorbed, the incoming photon energy is the sum of the two energy deposits. The interaction positions (X1, Y1), (X2, Y2) give the scattered photon direction and θ the angle between the incoming and scattered photon directions is given by the energy deposits through the Compton formula. The incoming photon direction belongs therefore to a cone of opening angle θ and whose axis is the scattered photon direction. In a classical Compton telescope, two position sensitive spectrometers are used: a scatterer and a calorimeter.

    The scatterer should have a high scattering efficiency but should allow for only one interaction, otherwise the interaction positions and the energy deposits may be confused. This double requirement is a difficult one. Using detectors with a low atomic number helps reducing the probability of an absorption in the scatterer. The thickness is then an issue. A thin detector is not efficient enough and in a thick detector there will be always more than one interaction. One way out is to use a stack of thin detectors, thin enough to exclude the possibility of a second interaction in the same detector. The scatterer efficiency is then governed by the number of detectors and could be as large as resources allow. Ideally, we would like to have it large enough to ensure a full absorption, i.e. the scatterer would be itself the calorimeter.


    Figure 2: Schematics illustrating the principle of a Compton telescope made with a stack of thin detectors

    a) Case of a scattering followed by an absorption

    b) Case of three scatterings without absorption

    However, in space, electric power is limited and a compromise should be made implying probably the presence of a classical imaging calorimeter. The calorimeter must optimize the absorption and will use therefore high atomic number elements. It should also be a fast and good spectrometer and allow for a decent spatial resolution.

    Given the constraints of space experiments, fulfilling the scientific requirements is not an easy task and this explain why no other Compton telescope followed COMPTEL (1991-2000), the first Compton telescope onboard a satellite. However, technological progresses allow today the design of much more sensitive experiments.

    The scatterer

    The best spectrometer is high purity cooled germanium. However, the scatterer should minimize the dead mass, i.e. non-detecting material in which photon can interact without notice. The cooling requirement is not very good from this point of view. To stay with semiconductors, silicon has a lower Z, is a good spectrometer and do not require active cooling.

    It can be readily realized that for a square meter class instrument the number of pixels per layer may be very large (106) and that imply an excessive power consumption. To give an idea, lets assume a power consumption of 1 mW/channel, a hundred layers would require 100 kW. For that reason, the use of strip detectors with 2N independent channels for N2 pixels is required. Double Side Stripped Detectors (DSSDs) have been widely used in particle tracking experiments (e.g. CMS)


    Figure 2: Schematic of a Double Side Stripped Detector (DSSD)

    In space, Fermi uses couples of Single Side Stripped Detectors (SSSDs) for monitoring the electron and positron tracks because there is no need to measure the energy deposits. It is not possible to use these detectors if an energy deposit measurement is required as in a Compton telescope. SSSDs relates energy deposits with only one coordinate while DSSDs give the two coordinates related to an energy deposit.

    The R&D program

    The R&D program relies on two feet, simulation and experimentation. DSSDs are relatively sophisticated and expensive detectors requiring a significant production time (~1 year) and a reliable design should be made before ordering any part. This requires skills in simulating the electrical behavior of Silicon detectors (electron/hole pair creation, charge carrier transport, capacitance/resistance between electrodes, etc….). Once the simulation tools are mastered, the DSSDs can be designed and ordered. In parallel, test means must be developed. A test bench shall be set-up for measuring detector leakage currents and capacitances and a spectrometric test bench must be developped. When an assembly formed with a DSSD and its readout electronics (ASIC) will be satisfactory, a second set will be realized, synchronized (coincidence) and positioned above the first to form a minimal Compton telescope which can then be tested and characterized. A third layer will then be added to complete the prototype.
    the DSSDs electrical analysis and characterization activity will take place at APC and use a probing station placed in a black box in a clean environment. The spectrometric testbench will be developed at AIM and will use the IDEF-X, a space qualified ASIC developed at IRFU. The ASIC will be driven by an FPGA developed at APC.

  • APC

    • François Lebrun (PI)
    • Philippe Laurent (scientist)
    • Walter Bertoli (project manager)
    • Youri Dolgorouki (experimentalist)
    • Mohamad Khalil (PhD)
    • Ronon Oger (electronics)
    • Nathan Bleurvacq (mechanics)


    • Olivier Limousin (Co-PI)
    • Diana Renaud (experimentalist)

  • Simulation

    Simulations tool

    An investigation of different software was conducted and the selection was made on SILVACO’s Tcad semiconductor simulation toolkit. Our work with SILVACO’s Tcad will include working with:

    • Atlas which is a physically-based 2D and 3D device simulator that predicts the electrical behavior of semiconductor devices at specified bias condition.
    • Deckbuild which is a runtime environment for Atlas where a text file collecting a sequence of commands corresponding to required bias conditions and control commands specified to select physical models and parameters.
    • Devedit which is a tool capable of defining the structure to be simulated in 2D and 3D.
    • Tonyplot which is a tool designed to visualize Tcad 1D, 2D and 3D (Tonyplot3D) structures provided by Silvaco.


    Simulation Goals

    Our goal in this simulation is not to reproduce in detail the behavior of DSSD sensors, but to extract tendencies and offer guidelines for the design of DSSDs to have better judgment of the capabilities of these detectors in terms of resolution and charge collection. However some vital input information to the simulation will remain unclear to us because it is hard to have access to such information (such as doping concentrations) due to the confidentiality of the manufacturing process, therefore we can neither know these parameters nor can we alter them should we decide to. If these parameters were set to constants which are close to what they represent in reality, it is then possible to study the effect of other parameters which we can decide such as the pitch, the thickness etc… if we find an optimal performance for a certain parameterization then we should expect a ‘close to optimal performance’ from the real DSSD even if its absolute value differs somehow from that given by the simulation.


    Simulation Technique

    Using C++ we have created an engine that can generate the DSSD structure requiring the number of strips, pitch, strip width, and size of the detector as main input. Other criteria can also be manipulated, such as doping concentrations, the thickness of all possible layers etc…We have built a parametric model of the DSSD that can be used to explore its behavior while changing the different characteristics of the model. By reducing the size of our DSSD (less than 10 electrodes) compared to the real one (64-128 electrodes) we obtain a problem that is easily resolved in a short computing time (few minutes to tens of minutes). This allows us to explore a large range of parameters in a reasonable simulation time.

    The following list states the simulation procedure and details the different characteristics of the sensors we anticipate to simulate and the data that can be extracted from the simulation results to obtain information on the possible behavior of real sensors.

    • Depletion voltage is an important parameter of semiconductor sensors. Under-depleted or undepleted strips will collect less or no charge when crossed by a charged particle.
    • The electric field shape inside the bulk of the sensor is an important parameter to determine its charge sharing behavior and its typical pulse shape.
    • The leakage, or dark, current, is given by the power supply applying the reverse bias voltage to the sensors. This current adds up to the signal when a particle is detected, increasing the noise and reducing the energy and position resolution of our sensor.
    • The AC analysis allows the reconstruction of the strip capacitance, one of the parameters determining the noise level of the readout electronics. It is given by the sum of two capacitances. The first is the body capacitance which derives from the coupling between the implant and the backplane layer. The second is the interstrip capacitance which is due to the coupling between the considered strip and its neighbors on the same side.
    • Link with GEANT4 to test single and multiple gamma ray interaction events where the output of GEANT4 would be the position and energy of the interaction and this will be used as input for SILVACO.


    Figure 1: Comparison of simulation results and measurements for a number of DSSDs whose bulk capacitance has been measured and published


    Figure 2: Comparison of simulation results and measurements for a number of DSSDs whose interstrip capacitance has been measured and published

    Figures 1 and 2 illustrate the accuracy attained with the simulation tool in predicting capacitance. The smallest total capacitance are obtained for the smallest strip width-to-pitch ratio. However for small strip width with regard to the pitch, on can fear a signignificant charge sharing. In addition dead zones may appear in between the electrodes due to the presence of a null electric field causing charges to stop drifting into the electrodes. It is therefore necessary to simulate all these effects before adopting a geometry for the strips.



    This activity on strip silicon detectors is new to AIM and APC. This implies an adaptation of existing experimental facilities and the development of new dedicated test benches. The experimental part of this work of R&D covers three parts:
    • Installation of a probe station for the characterization of detectors at APC
    • The implementation of a spectroscopic testbench at AIM
    • The design and realization of a mini Compton telescope

    Probe station

    We rely on various internal and external expertises to implement this bench at APC. In this context, our first measurements were carried out on the probe station of LPNHE (Paris), which was used for the characterization of strip silicon detectors for the tracker of the ATLAS experiment. The goal was to get some training in using a probe station to characterize DSSD detectors, a measure of the leakage current of a DSSD made by Hamamatsu (FoxSi) was performed. During this measurement campaign we have also developed a device to perform capacitance measurements as a function of voltage. We also contacted the IPHC (Strasbourg), which provided the electrical characterization of strip silicon detectors for the CMS experiment. In particular, this visit allowed us to overcome the difficulties related to the measurement accuracy and high applied voltages. Indeed, the range of the electric capacitances to be measured is of the order of 10 pF to a few tens of pF and the leakage currents of the order of a few micro-amps across the sensor to less than a nano-amp on a single strip. The depletion voltage of a 2 mm thick DSSD is of the order of 750 V, it falls to 420 V for a 1.5 mm DSSD. However, the RLC meter and pico-ampere meter at our disposal and allowing measurements with the required accuracy can sustend a voltage of the order of 40 to 50 V maximum. Thanks to feedback from the IPHC, we develop at APC a decoupling box that sustend the high voltage but is also also compliant with the required accuracy of the measurement. We could also define the electrical circuit diagrams for the electrical capacitance measurements between two strips, one on the ohmic side and the other on the junction side (“bulk”) and between two strips of the same side (“Interstrip”) with precautions to make measurements of a few picofarads. We were also able to map the circuit for measuring leakage currents of a few nano amperes on each strip.
    Figure 3 – Probe station being installed in a clean room laboratory APC
    Our goal now is to install these circuits on the experimental device shown in the picture above (Fig. 3).
    Our roadmap is as follows:
    • Capacitance measurements on a PCB simulating the hybrid board which will host the DSSDs specified for the mini Compton telescope – January 2013.
    • Electrical measurements (bulk capacitance and inter-strip and leakage currents) of “baby” DSSDs produced by SINTEF – January / February 2013.
    • Electrical measurements of “Musett” DSSDs- February 2013
    • Electrical measurements DSSD for the mini telescope Compton – upon receipt
    • DSSD BB7 of the IPN (Orsay) electrical measurements

    Spectrometric testbed

    In parallel with the configuration of the probe station, a spectrometric test bench is developed to test DSSDs at AIM. The vacuum chamber of the CEA / DSM / IRFU / SEDI / LDEF, we will adapt the electronic chain shown below (Fig 4). We can test and calibrate sensors using radioactive sources. As a first step, the MUSETT detector is connected to a 32 channels IDEF BD-X ASIC. We can therefore perform only a partial reading of the detector. This choice is motivated by the geometric similarities between the MUSETT DSSD and that we specify for the mini Compton telescope.
    The IDEF-X ASIC will be for the first time connected to a silicon DSSD. This measure will allow us to get a first impression of the performance of this system (DSSD silicon and acquisition chain IDEF-X described later).
    Figure 4 – Schematic of the experimental setup with the DSSD Musett ASIC IDEF-X.
    Our roadmap is as follows:
    Mounting in the vacuum chamber of the experimental device: 1 DSSD Musett ASIC IDEF-X. Currently mounting a functional chain in January 2013
    Mounting in the vacuum chamber of the experimental device can accommodate 3 DSSDs and 12 ASICs IDEF-X. This arrangement is however adapted to the choice of the third DSSD.
    DSSD performance specified for balloon flight
    Our primary scientific objective is to have sufficient resolution to be able to observe the polarization of the Crab Nebula between 100 keV and 300 keV. Megalib simulation (using Geant 4) powered by the instrument as shown CSNSM composed of a stack and a DSSDs LaBr3 calorimeter, an energy resolution of each of 3 keV DSSDs is compatible with a spectral resolution of the telescope May to September keV to 200 keV and an angular resolution of the order of 6 ° to 10 °. A resolution of 3 keV is achieved with the ASIC IDEF-X if the capacity at the inlet of the ASIC does not exceed 40 pF for a leakage current of less than 1 nA or to a leakage current of 10 nA a capacitance of less than 20 pF. Simulations SILVACO a DSSD 10 cm x 10 cm and 1.5 mm thick show that these values ​​are easily reached for a low ratio of track width on spacing. On the other hand, manufacturers contacted (or Micron Semiconductor SINTEF), we also confirm that these values ​​are technologically achievable and are ready to certify. Timely supply DSSDs are of the order of one year.
    Mini Compton Telescope
    Mini Compton telescope will consist of three floors stacked detection and synchronized them. One floor will consist of a DSSD mounted on a hybrid card. The hybrid card adapter spacing tracks will connect two ASICs IDEF BD-X 32 channels by faces (ie 4 ASICs by DSSD). Digitizing the signals from the 4 ASICs and the control thereof will be realized via a so-called map of integration, by a FPGA. The latter being referenced by a GPS system allow absolute dating of events in coincidence. Experiments onboard electrical power is limited, we would like to have thick detectors to provide strong stopping power without requiring a large number of playback channels. However, the depletion voltage varies as the square of the thickness detector and a thicker necessitate the use of a material of higher resistivity. A compromise performance / cost seems to be a reasonable thickness of about 1.5 mm.
    To develop this instrument, we will go through four intermediate steps each time increasing the level of complexity:
    Playing a DSSD Musett by ASIC IDEF-X. Currently mounting a functional chain in January 2013
    Playing a DSSD Mini Compton telescope with four ASICs IDEF-X. Currently supply. The order will be made in December 2012 for delivery of 2 DSSDs November 2013.
    Mounting two stages each comprising a Mini DSSD Compton telescope with four ASICs Idef X-and FPGA connection
    Previous configuration with an additional floor.
    Performance in angular and spectral resolution of the telescope mainly depend on the spectral resolution of each DSSDs. A good compromise to achieve our goals is DSSDs silicon with a thickness of 1.5 mm, an active area of ​​about 10 cm x 10 cm, a leakage current of less than 0 ° C nanoampere and capacity Electric under 20 pF per track. For now no manufacturer offers this type of detector in its catalog. An R & D on their part is necessary leading to a significant cost (about 40 k € for two sensors). Currently (December 2012), we are in the process of competition between different industrial sector and control DSSDs 2. The track width is not yet definitive (1 mm by default) but we can define in agreement with the provider within 3 months.
    The development plan of the electronic divided into three sets. The DSSD and hybrid map will be defined in collaboration with industrial and produced by it. Amplification and signal shaping will be performed by four ASICs IDEF-X BD, each connected to a voltage reference map. Map integration includes all the signals from the four ASIC and thus enables reading of the whole unified DSSD. These three cards are developed by AIM. The electronic service CPA is in charge of the control card of all ASICs.

    Acquisition chain

    Figure 5 – Wiring diagram of a floor of the Compton camera. Blue Zone: DSSD and hybrid map provided by industrial purple: Electronics developed and provided by AIM, green: e being developed to APC.