Laboratoire de physique des hautes énergies. SHAGARE project. Conception of the gamma ray detector. Master, second semester Laboratory work

Transcription

1 Laboratoire de physique des hautes énergies SHAGARE project Conception of the gamma ray detector Master, second semester Laboratory work Olivier Girard June 9, 213 Supervised by Prof. Aurelio Bay 1

2 Abstract The HAGARE project was defined last semester by three EPFL-students. This report focuses on the second phase of the project. The aim of the laboratory work was defined at the beginning of the semester as launching a first gamma ray detector prototype in the upper atmosphere. A PIN diode equipped with a CsI crystal was chosen for the sensor and an Arduino for the data handling and storage. All the electronics allowing for the read out of the diode had to be miniaturized as well as the interface between the sensor and the Arduino. After checking the capabilities of the device, the detector was finally launched with the help of Météo Suisse. The spectrum of gamma rays in the upper atmosphere was measured. Their flux was also estimated even though its variation with the altitude is debatable thanks to our measurement. CONTENTS 1 Introduction HAGARE and SHAGARE A gamma ray experiment Some gamma ray sources Choice for the sensor 6 3 Measurements with Arduino Uno Signals of different shapes Triggering pulses Interface between the sensor and the Arduino Testing the detector s component together 12 5 Assembling the detector Batteries and weight Tests Noise spectrum Pre-flight checks Low pressure Low temperature SPOT device Flight campaign Collaboration with Météo Suisse Launch decision Balloon flight O. Girard 2

3 8 Data analysis Rate of trigger and flux Energy spectrum Conclusion 31 References 32 A Cobalt spectrum 34 B Flight cancelled 35 C Spectrum during the flight 36 D Spectrum variation during the flight 37 E Pictures of the flight 38 O. Girard 3

4 1 INTRODUCTION This report explains the laboratory work that I did for the TP IV of my first year of Master. During one semester, I contributed to the conceptualization, the fabrication and the testing of a gamma ray detector that should be loaded on a balloon and fly in the upper atmosphere. First is a short introductory about the SHAGARE project followed by short section on the cosmic gamma rays. The next part shows the experiments that I set up in order to debug or test a component of the detector whereas the final part shows the results of the flight. The basic concepts of this project were posed by three EPFL-students on the previous semester. T. Kuntzer and E. van Schreven studied the concept of the HAGARE experiment that is to be discussed below and its feasibility whereas T. Humair investigated the physical goals of such an experiment and tested several options for a gamma ray sensor (see[1] and[2]). Moreover, D. Weiss, another EPFL-student, made simulations for the propagation of gamma rays in the upper atmosphere (see[3]). 1.1 HAGARE AND SHAGARE The High Altitude GAmma Ray Experiment project is born from a collaboration between the Swiss Space Center and the High Energy Physics Laboratory (LPHE) at the EPFL. This long term experiment aims to measure the energy spectrum and the direction of gamma rays of the low energy range (5 to 2 kev) in the upper atmosphere. The idea is to make profit of the Balloon-Borne Experiments for University Students (BEXUS) concept of the European Space Agency (ESA) that enables once a year an experiment designed by university students to fly from the North of Sweden. The selected experiment should weigh between 1 and 2 kg in total and avails of a four hour-flight at cruising altitude of 3 km. The students mentioned above who conceptualized this experiment proposed to carry out a smaller balloon-borne gamma ray experiment in order to gain experience with a balloon flight. The so-called Small HAGARE (SHAGARE) was first thought of as only testing the housekeeping systems and sensors[1]. However, the concept evolved at the beginning of the semester and the LPHE decided on its side to utilize SHAGARE to design, fabricate and test a gamma ray sensor prototype, small in size and weight so as to be loaded on a meteorological balloon. On the other side, the Processor Architecture Laboratory (LAP) joined the project. Joël Vallone, computer science student, designed, fabricated and programmed a tracking board for SHAGARE as a semester project (see[4]). As a physics student, my contributions to the fabrication of the detector were modest. In contrast, most of my work was its testing at the different stages of production. Furthermore, I programmed an Arduino (see[7]) in order to provide a safe interface between the analogue gamma ray sensor of the LPHE and the digital tracking board of the LAP. 1.2 A GAMMA RAY EXPERIMENT PERDaix, a former project selected by BEXUS, was carried out at the RWTH Aachen University in Germany with the participation of the EPFL. In 21, they launched the balloon in order to monitor the charged particle flux in the upper atmosphere (protons and electrons). HAGARE remains a particle detector although it will be designed to be sensitive to neutral particles, namely high energy photons. O. Girard 4

5 Photons do not interact with interstellar electromagnetic fields since they are neutral. Gamma rays are hence interesting in the sense that they point in principle directly to the location of their creation. Moreover, they originate from several high energy processes that are not yet very well understood. The atmosphere of the Earth prevents them to reach the ground, making thus necessary to load our detector on a satellite or a high altitude balloon. Simulations carried out by T. Humair ([2]) show that an altitude of 3 km is just sufficient to detect photons of energy in the range of 1 2 kev. This is where the idea of using a meteorological balloon originates. Météo Suisse launches balloons every day from Payerne to take measurements in the atmosphere up to an altitude of 3 35 km. We presented our project to Mr. J.-M. Clerc of Météo Suisse and he agreed to help us to launch our experiment. Very early, it was decided that the gamma ray sensor prototype will only consist of a single detecting part, enabling thus only to measure the flux and the energy of the particles. Its third property, the direction, will not be probed by the prototype. 1.3 SOME GAMMA RAY SOURCES This section mostly relies on the lectures on astroparticles given by Prof. A. Bay during the Spring semester and on T. Humair physical study of the HAGARE project (see[5]) and[2]). The energy spectrum of gamma rays is very diverse, including continuous and discrete parts, absorption and emission lines. Diffuse gamma ray background The diffuse gamma ray background is an important high energy photon source. Its origin involves many different physical processes. Cosmic particles interacting with the interstellar medium lead to the production of a diffuse gamma radiation. Yet, the annihilation of electrons and positrons into two or three photons yield a diffuse radiation with an energy continuum of maximum 511 kev. The origin of positrons can be multiple, such as the disintegration of pions into muons and then positrons. The Bremsstrahlung radiation of electrons in the Coulomb field of nucleus, the inverse Compton effect between a low energy photon and a relativistic electron and the synchrotron emission within a magnetic field also cause a diffuse radiation. In a balloon borne experiment, interaction between charged cosmic particles and the (upper) atmosphere should be considered too.γ-photons can indeed be produced secondarily due to a high energy particle interacting with the atmosphere. Gamma ray burst The observation of gamma ray flashes was confirmed only recently, in These events are very short (from a fraction of a second to some seconds) and seem to come from point-like sources. Their origin is not yet clearly understood. Solar flares The sun s surface often exhibits large instabilities called flares. In such phenomena, photons of a broad energy spectrum are emitted, including gammas. Point-like sources It was noticed that point-like objects such as supermassive black holes or pulsars produce gamma rays. T. Humair estimated the flux of gamma rays that one can expect in the upper atmosphere. It appears that the diffuse background yields a maximum rate of approximately.2 photons of energy between 2 and 2 kev within 1 s and 1 cm 2. This rate is calculated in the optimal O. Girard 5

6 situation of a zenith angle of and an altitude of 3 km. On the other hand, a gamma ray burst can cause a comparable rate assuming that the detector is pointed towards the source during the burst. 2 CHOICE FOR THE SENSOR During the previous semester, T. Humair measured precisely the properties of silicon photomultipliers (SiPMs) ([2]). This type of detector was first thought as being very likely to be used in the gamma ray detector of HAGARE. The tests had been carried out coupling the SiPMs to a bismuth germanium oxide (BGO) crystal, as well as CsI and NaI. The results on the resolution were disappointing. Consequently, the same tests were performed using the crystals with a PIN diode and a preamplifier. Since the resolution appeared to be better, we proposed restart the measurement of spectra with this type of diode at the beginning of the semester. PIN diodes are semiconductor detectors made of a middle intrinsic part surrounded by heavily p- and n-doped regions. They are almost not sensitive to temperature changes which constitutes a great advantage over SiPMs. The diode used, that has an active area of 1 cm 2, exhibits a maximum spectral response at some 9 nm (see[8] for more details). The quantum efficiency is approximately Q=8 % and constant for a wavelength between 4 and 1 nm. The slight temperature impact on this parameter is Q =.4 to.6 % per degree of temperature, depending on the wavelength of the photons. However, the CsI crystal that is coupled to the diode has Q a maximal emission for a wavelength of 55 nm for which the quantum efficiency should not vary; Q ( 5 nm) (see[8] and[9]). Hence, apart from the electronics, the sensor itself is Q expected to exhibit the same behaviour at different temperatures. We had two γ-photon sources available, one of cesium and one of cobalt. The procedure to obtain the energy spectrum of the 137 Cs source, which emits a kev-photon, is the following: the photon interacts through Compton scattering or photoelectric effect and makes the crystal scintillate. The more energetic it was the more scintillation photons are produced. Consequently, the integrated pulse eventually measured is larger for more energetic incoming photons. Nonetheless, we should remember that the sensor is also sensitive to charged particles since they might also interact with the crystal. In this case, the particle will not be totally absorbed but rather deposit a certain amount of its energy according to the Bethe- Bloch formula[6]; the scintillation photons reach the PIN diode and give rise to a very long pulse due to the very low capacitance of the diode. The latter is reversed biased (a voltage of 6 V was used here, slightly larger than the depletion voltage V dep = 35 V); the pulse is shaped and amplified (the gain was set to 2 for this manipulation) thanks to an integrator and differentiator. It is finally displayed on an oscilloscope whose trigger is set on an upward slope and whose threshold may be tuned. One can then manipulate the signal to measure the height of the signal which, in a good approximation, represents the energy of theγ-photon. The oscilloscope is then able to plot this figure as a histogram that constitutes the energy spectrum. We used the collected data to measure the resolution of the apparatus. It is defined as the full width at half maximum of the photoelectric peak divided by the position of the peak. Root was used to fit the peak to a Gaussian whose parameters led to the resolution. We proposed to tune the O. Girard 6

7 Int./Diff. Resolution [µs] [%] ± ± ± ±.2 Table 1: Resolution measured thanks to the photoelectric peak in the cesium disintegration spectrum for different integration/differentiation parameters. integration and differentiation parameters in order to see their influence on the resolution. The best result is shown in figure 1a Cs137 spectrum : resolution investigation Int/Diff parameters : 2µs Resolution is : (1.8±.16) % Photoelectric peak : (2.9±.21) V hist Entries 2 Mean 1.78 RMS Co6 spectrum : resolution investigation hist Entries 2 Mean RMS st 1 nd 2 Int/Diff parameters : 2µs peak : (1.6±.21) V, res.: (11.8 peak : (1.8±.17) V, res.: (7.83 ±.4) % ±.19) % (a) 137 Cs (b) 6 Co. Figure 1: Spectra of the two radioactive sources using a CsI crystal coupled to a PIN diode. The integration and differentiation parameters are set to 2µs. The first and highest peak is probably due to the intrinsic noise of the diode and the electronics but can also be caused by incoming photons being scattered by the surrounding of the detector. It is called the pedestal. It is followed by the Compton plateau. Finally, the photoelectric peak is clearly separable. A best resolution of 1.8 % was obtained with integration/differentiation parameters of 2µs whereas it was found to be over 2 % with parameters below 1µs. The other results are summarized in table 1. Since T. Humair obtained a good resolution using silicon multipliers with a BGO crystal, we proposed to change the crystal and test the PIN diode with the BGO. We measured the spectrum of the cesium source. Despite our effort, we did not manage to record any spectrum including a photoelectric peak. It turned out that only the pedestal, intrinsic to the PIN diode, was visible. It is likely that the coupling was not similar between the diode and the CsI and between the diode and the BGO. We set back the CsI crystal and measured the spectrum of another radioactive source, the 6 Co, which has two peaks at energies of kev and kev. However, their height is greatly reduced in this case. Furthermore, the first peak is superimposed to the Compton plateau making difficult to separate it. We obtained the spectrum shown in figure 1b. Fitting the two peaks to a Gaussian yields an estimate of the resolution. In our case, we obtained 11.8 and 7.8 %. O. Girard 7

8 Once the choice for the hardware of the sensor was made, we needed to think about how the signal will be sampled and recorded. Several options were available. For the flight of SHAGARE, it would be realistic to save the whole signal of the detector during the two hour flight. However, HAGARE may include hundred channels that will have to be sampled quickly when a particle passes. In order to match better with HAGARE s future requirements, we decided to trigger the reading of the sensor s analogue signal on the passage of a particle. Moreover, we chose to use the height of the signal as a reference for the energy of the incomingγ-photon rather than the integrated pulse because this parameter is more easily accessible and it is more likely that it will be taken as a reference for the energy in the HAGARE detector as well. 3 MEASUREMENTS WITH ARDUINO UNO 3.1 SIGNALS OF DIFFERENT SHAPES This section describes the different manipulations done to understand the operation of the Arduino Uno board that was purchased at the beginning of the semester. The electronic board was gradually incorporated into the detector itself with the aim of making it able to record the data coming from the sensor. As a first contact with the Arduino Uno board, we proposed to make sure that different known analogue signals can be handled correctly. We used a function generator available in the laboratory to make sine, triangle and noise signal. In all cases, the amplitude, frequency and offset can be tuned directly on the device. The signal went from the generator to the analogue input (A) of the Arduino board. A resistance of 5Ωwas set in parallel of the board. Programming it, we used the analogread() function which has a sampling rate of maximum ten thousand times per second. Wherefore, we set the signal frequency to a low value of 1 Hz and utilized the millis() function to stock the time reference of the board. The micros() function should be used if a higher frequency is set. Arduino test : signals from function generator Voltage [V] Sine signal Fit to Asin(ωt+φ)+B Triangle signal Noise signal Parameter Input Fit A[V].5.54 f [Hz] φ[rad].481 B[V] Time [ms] Figure 2: Test of the Arduino board inputting different signals of a frequency of 1 Hz. Table 2: Fitting parameters of the sine signal. The confidence levels of the fit are 1 4 or less, and therefore not mentioned here. The resulting signals are shown in figure 2. The output of the function generator was fixed to an amplitude (peak to peak) of 1 V, a frequency of 1 Hz and an offset of 4 V. Their shape was then O. Girard 8

9 changed. It can be seen that the signals are very well reproduced by the Arduino board. A fit was carried out on the sine signal and its parameters are in agreement with the output parameters of the generator (see table 2). 3.2 TRIGGERING PULSES After having managed reading rather smooth signals with the Arduino Uno, we wanted to investigate how to trigger the reading of an analog input of the board. Using the function generator, we produced a sinus of a frequency of 1 Hz that we split. One of them just entered the Arduino analogue pin A as before and the other one was used to generate a trigger pulse. It entered a discriminator whose threshold was tunable. The output of the discriminator was then set in a dual gate generator which was used to produce the trigger signal. Finally, the trigger pulse of which we could tune the width was plugged in the digital pin 2 of the Arduino that can be used as an external interrupt. The attachinterrupt() function enabled us to program the Arduino board for our purpose. We were able to order the board to sample the analogue signal once the digital trigger signal rises. An example of such a measurement is given in figure 3. In this example, for each trigger, we order the board to sample the analogue signal twenty times. As expected, we recover pieces of the original sinus. Arduino test : signals from function generator using an external trigger Voltage [V] Time [µs] 3 Figure 3: Test of the Arduino board inputting a sine signal of a frequency of 1 Hz and using an external trigger whose level is shown in red. The real signals which we want the Arduino to measure eventually in this project are not as simple as a smooth 1Hz-sine signal. Each incoming γ-photon on the detector will give rise to a pulse signal whose width is of the order of 1µs. We investigated whether the Arduino was able to sample such a short pulse. The setup used was the same but we generated a square signal instead of a sine one. The trigger pulse was shortened to a width of the order of 1µs. Assuming that the trigger and the signal reach the Arduino board at the same time, we were able with this setup to check how long the board takes to sample the analogue signal at the analogue pin when it receives an interrupt at the digital pin. The procedure was simply to vary the width of the pulse signal and the result is shown in figure 4. O. Girard 9

10 Arduino test : triggering pulses Voltage [V] Trigger (width =.65µs) Signal (width = 25.9µs) Signal (width = 33.27µs) Time [s] -6 Figure 4: Test of the Arduino board inputting a signal pulse of different widths (red and green) and using an external trigger (blue). We shortened the width of the signal pulse until the Arduino was not capable anymore to acquire the data, i.e. the amplitude of the pulse. Then, the pulses were kept in the configuration but set into the oscilloscope in order to measure them more precisely. In figure 4, the trigger is coloured in blue whereas two limiting case signals are in red and green. The red curve corresponds to a case where the Arduino was totally blind to the signal; it systematically sampled the pulse when it was already zero. The green curve represents the case where the Arduino was just able to sample one point on the signal pulse. In the region in between, i.e. between a pulse width of 25.9 and 33.27µs, the board was occasionally able to pick up one point on the pulse. As a consequence of this measurement, we can draw the conclusion that the Arduino takes some finite time to react from an interrupt. The analogue signal must be delayed or held during at least 33.27µs. For this purpose, the easier is to use a sample & hold circuit. 3.3 INTERFACE BETWEEN THE SENSOR AND THE ARDUINO The Arduino happened to be easy to program. When trying to incorporate it into the detecting device, the most difficult part was the interface between it and the sensor. As we have just seen, the interface should at least include a sample & hold chip. Furthermore, it must merge together the different signals from the gamma ray sensor and the Arduino board. On the first hand, the gamma ray sensor will continuously send two signals: an analogue one: the voltage signal; and a digital one: a trigger pulse sent whenever the latter exceeds a certain threshold. On the other hand, the Arduino receives the analogue signal and the corresponding trigger but it will take some time when processing to the data recording. Therefore, it must find a way to tell the detector that it cannot be triggered during a certain time. Moreover, as we saw before, the Arduino board takes more than 3 µs to respond to a trigger signal. The detector must thus hold the signal for at least this amount of time when it is triggered. O. Girard 1

11 Prof. Bay implemented all what was needed on the board shown in figure 5. It receives the analogue signal from the sensor in 1 and the digital trigger in 7. They are first delayed by 1.6µs using the delay line in 2. An analogue output similar to the input but delayed is available at the output 3. There are three connections to the Arduino: 8 (Arduino output): Arduino sends here a digital signal notifying that it is ready to record new data; 9 (Arduino input): the Master Gate trigger is the logic sum of the sensor trigger (γ-photon detected) and the Arduino trigger (connection 8, ready to record new data); 11 (Arduino input): the board allows to hold the signal for a certain time using a sample & hold chip. In 5 and 6, there are two useful adjustment devices. Thanks to them, we are able to tune the trigger pulse that is sent to the sample & hold chip. 5 delays slightly the trigger in order to hold the signal when it is at its maximum. 6 shortens or widens the trigger pulse so that the Arduino has enough time to sample the pulse height. 1 Analog input (signal from sensor) 7 Trigger input (from sensor) 2 Delay line (1.6µs) 8 Acknowledge (sent by Arduino) 3 Analog output 9 Trigger output (event occurring 4 Trigger input (from sensor, debugging) AND Arduino ready) 5 Trigger delay adjustment 1 Sample & hold offset adjustment 6 Trigger width adjustment 11 Sample & hold output Figure 5: Interface board made by Prof. A. Bay. It receives the signal and the trigger from the gamma ray sensor in 1 and 7 (or 4), the acknowledge from the Arduino in 8 and provides an analogue output delayed by 1.6µs in 3 plus the signal tracked and held in 11. We carried out a first test on this board with a function generator. We generated a square signal, which was used as a trigger in 7, and a simulated pulse in 1. An external power supply O. Girard 11

12 provided±6 V to the board. The Arduino recorded the data on an SD card thanks to an SD card shield. We used an oscilloscope and probes to monitor the operation of the board. In addition to the simulated pulses, we used a second analogue input of the Arduino. We set up the thermistor in order to record the temperature at each event. Figure 6 shows an example of the signal observed at the output of the sample & hold (11). Using the adjustment device 6, we shortened the holding of the signal and investigated again how far Arduino can go. The result is the same as before. The analogue signal must be held for at least(33±2)µs, otherwise the Arduino board is not able to sample the pulse. As another measurement, we probed the digital trigger sent by the Arduino to the board. This allowed us to determine how much time Arduino spends on a whole loop, including receiving the trigger, reading the analogue signals (sensor + temperature), saving them in a buffer with a time tag, sending a trigger pulse notifying it is ready for a new measurement. The time spent in this loop has been approximately measured with the oscilloscope to be(5±3)µs. Monitoring the sampled & held signal of the board Voltage [V] Original signal Signal held by the board Time [s] -3 Figure 6: Example of result. When the Arduino is not at work (blue), the signal is not held. When the code is uploaded (green), it notifies that it is ready, opening the Master Gate. The signal is then held for a certain time. 4 TESTING THE DETECTOR S COMPONENT TOGETHER Once the different components were working perfectly independently of each other, we were ready to carry out tests on the whole detecting device. Figure 7 shows a summary of the device that we developed and how the components were connected together. We will come back later to the choice done for the batteries. Anyway, for the first tests that we are discussing here, we used an external power supply. The Arduino was powered and controlled from the computer. O. Girard 12

13 Ø ØÓÖ ÁÒØ Ö Ó Ö Ö Ù ÒÓ Á ÖÝØ Ð ÈÁÆ Ó Ð Ö Ø ÓÒ ÈÖ ÑÔÐ Ö ½ ¾ Ö Ñ Ò ØÓÖ Ò ÐÓ Ò Ð ÌÖ Ö Ð Ý Ë ÑÔÐ ² ÀÓÐ Ò ÐÓ Ò Ð ÌÖ Ö Ò ÐÓ ÒÔÙØ Ø Ð ÒÔÙØ ÁÒØ ÖÖÙÔØ Ë Ö Å Ø Ð ÓÙØÔÙØ Å Ø Ö Ø ± Î + Î Ó ÈÁÆ Ó ØØ Ö ½¾Î ØØ Ö + Î ½º Î Î Î Figure 7: Simple schematics of the detecting device. Each part was mounted on a separate plate that we fixed together for the flight. The signals in 1,2 and 3 were monitored on the oscilloscope (figure 8). As a first check of the capabilities of the whole device, we proposed to inject a calibration pulse at the input of the detector provided for this purpose. The device was let working automatically for a while. We observed fluctuations in the height of the pulses in 1 even though the calibration pulse was set with a constant height at the input of the detector. The sampled signal was fluctuating in the same way in 3. These unwanted variations are thus probably induced by the preamplifier. Voltage [V] 1 8 Monitoring the signals sent to the Arduino Calibration pulses at the output of the detector (1) Trigger at the output of the detector (2) Signal sampled and held (3) Time [s] -6 Figure 8: Monitored signals of the device. This leads hence to an intrinsic resolution of our device which is not infinite. We made a rough estimate of this resolution by measuring the spectrum of the height the signal sampled by the Arduino. The result is shown in figure 9. The amplitude of the calibration signal was also changed in order to investigate the linearity of the detecting device. We fitted the spectra with a Gaussian and could then estimate the intrinsic resolution. O. Girard 13

14 Spectrum of the calibration signal Amplitude of the calibration signal 2, 25, 3, 35, 4mV hist2 Entries Mean 2.49 RMS.1938 Amp cal Resolution V peak [mv] [%] [V] ± ± ± ± ± Figure 9: Spectrum of the calibration pulse height. Table 3: Resolution obtained for calibration signals of different height at the input. In figure 9, we note that the width of the peaks, namely the absolute size of the fluctuations, does not depend on the height of the calibration signal. This leads to an intrinsic resolution which decreases with this parameter, as it can be seen in table 3. The mean value for each peak is also mentioned in the table. This enables to judge whether the device has a linear response. According to these few measurements, it is nicely linear. The last step in testing the device was to couple the PIN diode to the rest of the system and obtain the spectrum of some radioactive source. We measured the spectrum of 137 Cs and 6 Co (figure 1). Furthermore, we tuned the trigger level on the detector to allow for a better ratio between peak and pedestal. We can see that the final resolution is enough to separate easily the kev peak in the cesium spectrum (figure 1a), although it is worse than the first measurement we did with this source (see figure 1a, 24 vs 11 %). Calibrated as it is here, a 1.1 V signal corresponds to a kevγ-photon. Theγ-photon energy hence scales as 61.5 kev/v Cs spectrum Resolution : (24±.23) % Photoelectric peak : (1.1±.93) V hist Entries Mean 1.44 RMS Co spectrum hist Entries Mean RMS (a) The photoelectric peak is clearly visible for 137 Cs (b) Only the pedestal and the Compton plateau are visible in the case of 6 Co. Figure 1: Spectra recorded by the Arduino. The duration of each measurement was approximately 2 minutes. Regarding the spectrum of the cobalt source, the two peaks of energy around 1.2 MeV could not be separated (see figure 1b). This is a result of the worsened resolution due to the preamplifier. On previous measurements, we already remarked that the photoelectric peaks in the cobalt O. Girard 14

15 spectrum are much more difficult to distinguish with respect to the pedestal and the Compton plateau. Assuming that the photon energy scales linearly with the energy, we would obtain that the photoelectric peak at kev would lie at approximately 2.2 V which is compatible with figure 1b. 5 ASSEMBLING THE DETECTOR We chose to pile up the boards with the detector on the top. In the middle, we placed Prof. Bay s board in the reversed sense in order to be able to fine tune the parameters such as the width of the trigger in case we must change them. Under it, we mounted the Arduino fastened to a board where we implemented a little circuit with a LED used as a debugging tool. Moreover, the board includes a switch for all the components of the detector apart from the bias of the PIN diode. The batteries were put at the very bottom. A picture is shown in figure Sensor board 6 Arduino Uno 2 Interface board 7 SD card shield 3 Arduino board 8 Thermoresistance 4 Batteries (±9 V) 9 Main switch 5 Sensor (PIN diode coupled 1 Batteries for the bias of the to a CsI crystal) diode comprising a switch Figure 11: Picture of the detector. 5.1 BATTERIES AND WEIGHT We tested several types of batteries before making a final choice. The bias of the PIN diode was built with four batteries of 12 V. As the diode works in reverse bias, almost no current traverses the device. The batteries need therefore no large capacitance. In contrast, we measured the current that flows in the boards: it amounts to approximately 8 ma for the+9 V and 3 ma for the 9 V. The Arduino power input is flexible (between 7 and 12 V) and we decided to power it with O. Girard 15

16 9 V. Its current consumption was measured to be about 7 ma. As a first try, we therefore simply set 9V-batteries to power the electronics: one for the Arduino and two for the detector s and the interface boards. After half an hour, the voltage of all batteries had dropped of.5 to 1 V. They were hence of a too low capacitance. We changed for heavier batteries with larger capacitance. Six batteries of 1.5 V supplied the required 9 V for the boards whereas six others supplied the+9 V for the boards plus the Arduino. This setup was used for the test that is to be described below. We ran a measurement for seven hours and the batteries appeared to resist very well. The negative voltage had not dropped whereas the positive voltage had dropped of approximately.7 V. These batteries seemed to be a rather good choice for the power supply of the detector. However, they weight 7 g each, which made the whole detecting device very heavy (169 g), too heavy for a balloon flight. We reassembled the boards together using as much plastic instead of metal as possible which lightened the device by 15 g (1543 g). Since the negative voltage did not need as much current as the positive one, we changed the six 1.5V- for three 3V-batteries. This type of batteries is used daily by Météo Suisse and holds the time of a flight. This finally reduces the total weight to 122 g. To this must be added the weight of the SPOT device that will be used to track the balloon during the flight plus the ones of the silica gel, protecting from humidity condensation, and of the pieces of sagex used to wedge correctly the detector inside the box. Table 4 specifies the weight of the diverse components. Component Weight [g] Sensor board Detector Interface board Arduino board Battery holder 51 Batteries 9V : 3 3V 53 +9V : 6 1.5V 48 Tracking SPOT 6 Other Polystyrene box 241 Silica gel 6 Total 1321 Table 4: Mass budget of the gamma ray detector produced at the LPHE. All masses indicated here were measured. 5.2 TESTS As seen before, the final resolution of the apparatus was not as good as the one obtained at the beginning of the semester (see figure 1a, 1b and 1). As we discussed, the preamplifier induces fluctuations that can explain partly this worsened resolution. Another element able to influence the resolution is the optical coupling between the crystal and the PIN diode. Indeed, a single air bubble could capture photons on their way from the crystal to the diode and lead to a smaller photon yield. Hence, we extracted the crystal, cleaned it, rewrapped it in Teflon tape and coupled it to the diode with optical paste. This procedure directly appeared to have been beneficial because O. Girard 16

17 the signals were immediately five as high as before, stemming from a larger photon yield. We readjusted the amplification and ran new measurements Cs spectrum Resolution : (16.6±.15) % Photoelectric peak : (1.51±.1) V hist Entries Mean RMS Co spectrum st 1 nd 2 peak : (2.58±.99) V, res.: (19.1±1.3) % peak : (2.94±.74) V, res.: (13.1±.76) % hist Entries Mean RMS (a) The resolution is greatly improved thanks to a better optical coupling (b) The two high energy peaks in the spectrum of the cobalt source are visible. The measurement took seven hours. Figure 12: Spectra recorded by the Arduino and supplying the detector with batteries. A quick check of the capabilities of the device was carried out using the 137 Cs source. Figure 12a shows that the resolution improved of 8 %. As a next test, we measured the spectrum of the cobalt source during seven hours. The result is displayed in figure 12b. The two peaks in the γ energy range are visible which enables us to verify the linearity of the detector with respect to the energy of the incoming photon. A simple graph shows that the response seems linear. However, only three points are available. Table 5 summarizes the measurements for these energy points. Trusting in them and assuming a linear response, a fit gives: E[keV] = 472 x[v] 49 (1) Since the Arduino can take in analogue signal of maximum 5 V, we estimate the upper limit of the energy of a detectableγ-photon to be approximately E up MeV. Source E peak V peak Resolution [kev] [V] [%] [kev] 137 Cs ± ±1. 6 Co ± ± Co ± ±1.7 Table 5: Data deduced from the measured spectra of the twoγ-ray sources available. We exploited the very long measurement of the cobalt spectrum for a further analysis. We divided it into six spectra of one hour each (see figure 28 in appendix). This allowed us to look at the behaviour of the detector with time. As it can be seen, the measured spectrum changed considerably during the six hours. As mentioned previously, the positive voltage dropped during this time and this could have influenced the output of the detector. In addition, as the electronics dissipates heat, we marked that the temperature had significantly increased in the box (see figure 13). O. Girard 17

18 We note that the differences in figure 28 are more quantitative than qualitative. The gain does not change during the measurement as it is summarized in table 6. Note that the detection threshold given in volts was simply estimated visually. Temperature variation during the test Temperature [K] Time [min] Figure 13: Temperature increase during the 7h-measurement. The whole detector was put inside a polystyrene box and the measurement was started at room temperature. Hour Detection first second Total threshold peak peak # counts [V] [V] [V] [-] ± ± not visible 2.92 ± ± ± ± ± ± ± ± ± Table 6: Analysis of the long measurement of the cobalt spectrum. We note that nor the detection threshold, neither the position of the peaks seems to depend on the level of the batteries or on the temperature. Nevertheless, we observe a large drop in the total number of events in each hour. It reduces by a factor of three between the first and the third hour and then increases again until the end of the measurement. Figure 14 shows the average rate of triggers received by the Arduino. Different reasons can explain the rather bizarre variation. On the first hand, the number of triggers arriving to the Arduino could have varied, i.e. the flux of photons could have changed due, for example, to the electromagnetic noise around the setup changing during the day despite the "black box" (the lunch break occurred about one hour after the beginning of the measurement). On the second hand, the Arduino could have missed some trigger signals for an obscure reason. Moreover, the slight voltage drop in the batteries could also impair the operation of one of the electronic component. As we saw, the temperature changed during the seven hours but the rate variation seems to be not correlated at all to it. Anyway, in O. Girard 18

19 figure 28, we see that the height of the pedestal changes greatly during the measurement whereas the level of the two high energy peaks is less influenced. Rate of trigger <Rate> [s -1 ] Time [min] Figure 14: Rate of triggers received by the Arduino during the 7h-measurement. It was estimated by counting the triggers on a certain interval (here of 2 min). 5.3 NOISE SPECTRUM Noise spectrum in the lab hist Entries 213 Mean 1.38 RMS Figure 15: Noise spectrum recorded during thirty minutes. The detector was put in a closed polystyrene box, just as it will be for the flight. The data taken during the first and the last minutes are not included in order to exclude the noise events that occurred because of the daylight. A characterization of the noise produced either by external sources or by the electronics was required. We let the detector working inside the polystyrene box for half an hour. The noise spectrum is plotted in figure 15. It can be seen that 21 noise events were registered. This is significant compared to the number ofγ-photons that we expect to detect in the stratosphere. As we saw in O. Girard 19

20 section 1.3, the flux of gamma ray will certainly be of the order of.1 cm 2 s 1. Accounting for the facts that the size of the sensor is approximately 1 cm 2 and that the balloon will be at a higher altitude than 15 km during a time of the order of one hour, we expect to observe between 3 and 4 relevant events. Quantitatively, noise will hence probably dominate the signal in SHAGARE, that is, in this gamma ray sensor prototyping experiment. However, having a closer look to its spectrum, noise mainly dominates in the low energy range. Indeed, we observe only some 71 noise events with pulse height larger than 1 V, i.e. a fake γ-photon energy above E 45 kev. Pushing the argument still further, only 33 events occurs above 1.6 V, i.e. E 7 kev. On the other hand, the detection threshold, purely determined by the trigger level, is also visible at roughly.3 V corresponding to an energy of E threshold 1 kev. The noise, however, restricts our energy window to 7 23 kev, given the low gamma ray flux. 6 PRE-FLIGHT CHECKS The whole system comprising the Arduino was then ready for the flight. On LAP s side, the tracking board would be ready by the end of the semester. Hence, we carried out the last necessary checks on LPHE s detector. In the stratosphere, the pressure is as low as 1 mbar. It is not obvious that the electronics that we have developed resists to vacuum conditions. Some components may indeed outgas or overheat. In addition, the temperature in the stratosphere reaches 5 C. Even though the air is nevertheless very dry at high altitudes, humidity still represents a risk because some air of the ground level may condensate on the electronics if the temperature drops under C in the nacelle. This is the reason why the nacelle includes some silica gel. In order locate the balloon and recover the detector after the flight, it was decided to use a device external to the detector and fully autonomous. A commercial device such as a GPS SPOT is sufficient. Mr. R. Maag of MeteoLabor lent us a lightened SPOT which had already been used as tracking device during balloon flights. Although the detector as it was designed at the LPHE does not have any radio transmission system, it was to be ensured that it would not interfere with the GPS, forbidding it to notify its position. 6.1 LOW PRESSURE We decided hence to place the detector inside a vacuum chamber and measure the spectrum of a radioactive source. The vacuum was realized progressively in order to be more similar to the flight conditions. The resulting spectrum of the cesium source is available in figure 16a. We observe a large pedestal which stems from the fact that the vacuum chamber was not shielded from daylight. The photoelectric peak is slightly shifted towards a smaller value of the voltage than in previous tests. This could mean that the gain decreased under low pressure. However, it is probably not the case. This slight shift is likely to be due to the higher noise recorded in this measurement. Since the noise emerges just at smaller pulse heights than the photoelectric peak of cesium, the rising side of the peak is more affected by the noise and this shifts our fit to the left. On the other hand, the temperature was also recorded and it is shown in figure 16b. Although no air convection occurs under vacuum, the heating components could still thermalize (e.g. by O. Girard 2

21 Cs spectrum Resolution : (24.9±.38) % hist Entries Mean RMS.939 C] Temperature [ Temperature variation during the test under vacuum 2 Photoelectric peak : (1.43±.22) V (a) Cesium spectrum measured in vacuum. The pressure was dropped progressively before a thirty minutemeasurement was carried out Time [min] (b) Temperature variation in the vacuum chamber. Figure 16: Results of the test carried out under vacuum. radiation). This explains the increase in temperature which is comparable to the one observed in the first thirty minutes of the previous 7h-measurement (see figure 13); 3.5 C against 4 C after thirty minutes. Moreover, we see that at the end of the experiment, when air was let into the vacuum chamber, the temperature increased significantly. This does not arise from the external room with a maybe higher temperature, but rather from convection that can again happen. Some components could then diffuse their heat more easily. 6.2 LOW TEMPERATURE As a very primitive low temperature test, we put the detector in the fridge for half an hour. The noise spectrum does not change from the one available in figure 15. The temperature was measured and it is plotted in figure 17. We see that it dropped of some 6 C before it levels off. Dissipation is thus significant which bodes well for the flight. No temperature control system has indeed been integrated in the payload due to the weight limitation. Temperature variation during the test in the fridge Temperature [ C] Time [min] Figure 17: Temperature variation in the fridge. O. Girard 21

22 6.3 SPOT DEVICE The SPOT device was placed in the box, above the detector with the antenna pointing towards the sky. Once it is synchronized with the satellite, it emits its position every ten minutes. We simply turned on both devices and let them run during two hours. The SPOT device did not have any impact on the noise measured by the detector. Furthermore, no interference was to be mentioned since the position was updated on the net every ten minutes. 7 FLIGHT CAMPAIGN 7.1 COLLABORATION WITH MÉTÉO SUISSE A meeting with Mr. Clerc of Météo Suisse was organised in Payerne several days before the final tests on the detector were carried out. We explained in more details SHAGARE and its aims as well as worked out more precisely our collaboration. One of the major problems of the device that we developed at the LPHE is that no information concerning the altitude is recorded. Indeed, the SPOT device provides only the GPS coordinates and not the altitude. Although the detector remains a prototype, the flux of gamma rays as a function of the altitude is the physically interesting parameter that SHAGARE aims to estimate. Moreover, the SPOT device notifies its position only every ten minutes and it is not clear whether it will work properly at high altitude. A solution to both issues was to be found. Météo Suisse suggested to fly their daily atmospheric probe at the same time as our experiment, i.e. use only one balloon and fasten both boxes together. The probe contains a GPS and transmits in real time its position. This allows us to obtain the information about the altitude together with the outer temperature and pressure. Furthermore, it makes the retrieving of the balloon safer. I am very grateful to Mr. Clerc and his colleagues who provided us a crucial help in this project. 7.2 LAUNCH DECISION Mr. G. Romanens of Météo Suisse performed simulations for the balloon trajectory in order to select the best possible day for a launch. The predictions were based on the ECMWF model (see [1]) and assumed an ascent speed of 5.5 m/s and a burst altitude of 3 km. In order to take the decision of launching the balloon, we needed both a convenient trajectory and a good weather to have a reliable trajectory prediction. The bad weather in May did not help to take the decision. Figure 29 in appendix shows the prediction for a flight on Monday 27 May. For once that the weather was nice, the trajectory predicted for the balloon was not ideal. According to Mr. Clerc there was approximately 6 % probability that our experiment ends up in the lake. Hence, we decided to postpone the flight. On Thursday 3 May, the flight was finally possible. Next section summarizes the expedition which was not totally free of surprise. The details are only here to be comprehensive about how such a project is made possible. Some pictures of the launching and the recovery are shown in section E in appendix. O. Girard 22

23 7.3 BALLOON FLIGHT We met the Météo Suisse team in Payerne in the morning and carried out quick checks on the detector. It was then turned on as well as the SPOT device. Followed manipulations on the two boxes in order to fasten them together. They were tied to the edges of a bamboo branch using standard rope. The payload weighed approximately 2 kg in total. The balloon was filled with hydrogen and some 4 m was left between the balloon and the payload. At 12:32, we released the balloon. The ascent velocity was larger than expected ( 9 m/s). However, it followed nicely the predicted trajectory, as it can be seen in figure 18a. The GPS SPOT did not transmit its position during the whole flight. Its last message was communicated at 12:5 and, by cross-checking the position with Météo Suisse continuous GPS signal, we can say that the detector was then at an altitude of 7772 m. When descending, the SPOT device notified its first position at 14:31 at an altitude of m. Although the ascent velocity immediately appeared larger than expected, the balloon followed very well the predicted trajectory. The landing point was only 4.2 km away from the expected one. Figure 18b shows how we were lucky that the balloon did not end up in the gorge of Covatannaz where we would have had extreme difficulties to localize it. The GPS of Météo Suisse atmospheric probe and the SPOT agreed almost exactly on the final position, which limited our research zone to a small part of the forest along the road between Vuiteboeuf and Sainte-Croix. Mr. Clerc had a radio set on the frequency of his probe and, in just half an hour, we localized the two boxes at the very top of a 2 to 3-meter tree. The tree was dead which offered us mainly two possibilities. One the first hand, one could cut down the tree with the permission of the forester while, on the second hand, one could climb the tree or one next to it and retrieve the payload. We opted for the second option. However, since the weather was particularly rainy the day after the flight, namely on Friday, we were bound to wait till the next Monday. Following the advice of Mr. Clerc, we contacted "Vertige-Concept", a company based in Yverdon and specialized in work at a height. We met them on Monday 3 June and, after one hour and a half of non-trivial climbing, our boxes touched smoothly the ground. The box containing the gamma ray detector was partially filled with water. The detector was not operating anymore and we could smell that some electronic component had burnt. After inspection, we found that the reason for this was that the batteries were immersed in water and the 9 V induced electrolysis. The voltages dropped from+9.61 to+5.96 V and from 9.76 to 3.48 V. Fortunately, the SD card seemed not damaged. As a precaution, we waited the next day to plug it in the computer so that we were sure that it would be dry. Reading the SD card on the next day, we saw that instead of one single file, it contained a dozen of them with unreadable names. Although the data including the first minutes was missing, we were lucky and could find the important data, the one recorded during the flight. 8 DATA ANALYSIS The Météo Suisse team provided us the data of their atmospheric probe. Figure 19 shows the outer temperature and the pressure profiles measured during the ascent. The pressure decreases exactly exponentially with the altitude. The balloon burst at an altitude of m where a pressure of approximately 9 hpa prevails. In contrast, the temperature profile is not smooth and shows transitions between different atmospheric layers. O. Girard 23

24 (a) The predicted trajectory is in white whereas the real one in red. The orange path is its projection on the ground. The green labels are the positions notified by the SPOT device. (b) Zoom on the landing zone. The mismatch between the predicted and the actual trajectory is of 4.2 km. Figure 18: Comparison between the predicted and the real trajectory. As mentioned previously, the Arduino was turned on a few minutes before the launching. In order to match the time of the atmospheric probe and of the gamma ray detector, we first matched the inner and outer temperatures and then evaluated the rate of triggers due to cosmic particles. The latter should increase significantly when the balloon is launched. Figure 2 shows how the temperature varied throughout the flight. O. Girard 24

25 C] Temperature [ 1-1 Temperature profile (MeteoSuisse data) Pressure [hpa] 3 1 Pressure profile (MeteoSuisse data) Altitude [km] (a) Temperature profile Altitude [km] (b) Pressure profile. Figure 19: Météo Suisse measurements during the flight. C] Temperature [ Temperature variation during the flight Inside the box Outside (MeteoSuisse) Time [min] Figure 2: Inner and outer temperature variation during the flight. We observe that the inner temperature did not drop as substantially as the outer one. Yet, it dropped below C and the detector only warmed up after the landing. Although the behaviour of the sensor itself should not depend on the temperature, the electronics may be influenced in a yet unknown way. The test carried out in the fridge was indeed insufficient to draw any conclusion. However, the results which we will discuss in this section show that the detector was operating. As we discussed previously, the gamma ray detector did not measure any altitude during the flight. This information, crucial for our analysis, was obtained thanks to Météo Suisse data. We calibrated properly the Arduino time with respect to the atmospheric probe time (whose zero is the launching time). Then, each gamma ray signal was labelled with a time and an altitude. Figure 21 presents the result of this procedure. O. Girard 25

26 Altitude of the Arduino Altitude [km] Time [min] Figure 21: Altitude of the Arduino deduced by matching its time with the atmospheric probe time. 8.1 RATE OF TRIGGER AND FLUX As we did for the cobalt spectrum in figure 14, we calculated from the data the average rate of trigger received by the Arduino. This was simply done by counting the signals within time intervals of 1 min. However, we additionally applied a cut on the energy. Looking at the spectra that we obtained (see section 8.2), we divided our analysis in three portions: the pedestal (the noise is significant below 1.6 V, according to our discussion of section 5.3), a plateau and an increase at the saturation. The result is displayed in figure 22. We see clearly an increase when the balloon is launched with a maximum at time t 1 min which corresponds to the maximum altitude. Nevertheless, we see that the rate peaks also at an unexpected time before going down and up again. This bump appears as well in the rate of the noise. This additional peak is probably not a consequence of the low temperature or pressure, because no correlation is found between the rate of trigger and one of these parameters. It might rather be attributable to a temporary change in some property of the detector. O. Girard 26

27 Rate of trigger Rate of trigger <Rate> [s -1 ] <Rate> [s -1 ] Time [min] Time [min] (a) Pedestal: pulse height smaller than 1.6 V <Rate> [s -1 ] Rate of trigger (b) Plateau: pulse height between 1.6 and 4.5 V Time [min] (c) Saturation: pulse height larger than 4.5 V Figure 22: Rate of trigger received by the Arduino. A first hypothesis would be that the gain of the amplifier varied, pushing many noise events in the range of interest. However, it is invalidated by figure 31 in appendix. There, we plotted the spectrum of the pulses recorded during intervals of 25 min. We see that noise level fluctuates in the same way as the rest of the signal. Moreover, the position of the pedestal does not change meaning the gain was not varying. The level of the plateau increases between 25 and 5 min and this reflects exactly the increase in the rate of trigger. The efficiency of the sensor could explain this strange behaviour if it would have varied during the flight. However, this hypothesis cannot be verified since we have no mean to measure or calculate this parameter at any time. Combining figures 21 and 22, we can obtain the rate in terms of the altitude. Figure 23 exposes the result. The rate profiles at the ascent and at the descent appear very different. The descent one is closer to what we expected, according to[2]. O. Girard 27

28 Rate of trigger Rate of trigger <Rate> [s -1 ] Ascent Descent <Rate> [s -1 ] Ascent Descent Altitude [km] (a) Pedestal: pulse height smaller than 1.6 V. <Rate> [s -1 ] 3.5 Ascent 3 Descent Altitude [km] (b) Plateau: pulse height between 1.6 and 4.5 V. Rate of trigger Altitude [km] (c) Saturation: pulse height larger than 4.5 V. Figure 23: Rate of trigger received by the Arduino as a function of the altitude. Starting from figure 23, we can estimate the flux ofγ-photons as a function of the altitude. We can assume for a first estimate that the efficiency of the sensor is constant and equals to unity. However, we should remember that this assumption is not obvious as we have just seen. To determine the flux, we need the effective area of the sensor. Moreover, the flux depends on the zenith angle and no calibration of the sensor using a radioactive source put at different angles had been carried out. The orientation of the detector plays also a role and makes that the effective area changes continuously. We propose to obtain a rough estimate of the flux using only geometrical arguments. Our procedure is based on the assumption that only photons with a zenith angle below 6 can be detected and the sensor is always oriented towards the zenith. The assumption on the angle comes from[2] where T. Humair showed that the absorption is substantial for zenith angle above 6. Figure 24 schematizes the procedure. Knowing the dimensions of the crystal (H = 1.1 cm and l 1 =.93 cm, l 2 =.8 cm), we obtain an effective area of 5.5 cm 2. O. Girard 28

29 ÖÝ Ø Ð l 1,2 ¼ H Ó r =.5Ñ r eff Figure 24: Scheme explaining the procedure followed to determine the effective area of the sensor. The result is shown in figure 25. We omitted this time the rate of the pedestal. Due to our basic assumptions, the shape of the curve is the same as the rate of trigger. For particles of energy between The flux at the highest altitude appears to be approximately.2 cm 2 s 1 which is in the same order of magnitude as the prediction made by T. Humair. Estimated flux of particles Estimated flux of particles <Flux> [cm -2 s -1 ].6.5 Ascent Descent <Flux> [cm -2 s -1 ].6.5 Ascent Descent Altitude [km] (a) Plateau: pulse height between 1.6 and 4.5 V, i.e. energy between 7 and 2 kev Altitude [km] (b) Saturation: pulse height larger than 4.5 V, i.e. energy between 2 and 23 kev. Figure 25: Flux of particles of two different energy ranges detected by the sensor as a function of the altitude. 8.2 ENERGY SPECTRUM Until now, we only considered rates or fluxes. Here, we want to discuss the spectrum recorded by the detector. We proceeded in the same way as we did with laboratory measurements. The results are given in appendix in figure 3. We plotted the spectrum measured during the 15 minutes of the flight and the one measured only above 25 km. In order to plot the energy spectrum, we used the calibration in energy done during the semester. Thanks to relation (1), each bin in the histograms of figure 3 corresponds approximately to an energy bin. The energy spectrum is hence straightforward. It is presented in figure 26. O. Girard 29

30 Energy spectrum: the whole flight histe Entries Mean 1375 RMS Energy of γ-photons [kev] (a) Whole flight energy spectrum ( 15 min) Energy spectrum: above 25 km hist2e Entries 856 Mean 1519 RMS Energy of γ-photons [kev] (b) Energy spectrum measured above 25 km ( 3 min). Figure 26: Energy spectra measured during the flight. They are based on figure 3 and relation (1). Comparing this spectrum to the noise spectrum of section 5.3, it is clear that some signal is observed above 7 kev. It appears to be a flat spectrum which is consistent with our expectations. The diffuse gamma ray background dominates. In figure 31, we see that the plateau stayed flat during the whole flight. No signature of a gamma ray burst is visible. These events have a very broad duration and a statistical analysis of the spectrum variation in time should be carried out. However, given the small size of the sensor, not much statistics will be obtained. Furthermore, the probability that such an event occurred during the flight remains low. For the 4h-flight of HAGARE, T. Humair estimated this probability to only some 8 %[2]. An interesting feature is to be noted at the highest energies resolved by the detector. We observed numerous signals that saturated the detector (above 14 events of the 388 recorded during the flight). The spectrum exhibits a semi-peak of a width of E 1 to 2 kev at the saturation energy. This probably stems from charged particles that interact with the crystal. As O. Girard 3

31 we saw previously, the way photons and charged particles interact with matter is fundamentally different. A photon interacts with atoms through atom excitation or via the photoelectric effect and with electrons via the Compton scattering whereas a heavy charged particle such as a proton continuously interacts with electrons in the medium and deposits a small amount of energy at each interaction. The energy deposition is governed by the Bethe-Bloch formula.[6] Protons constitute a large part of the cosmic rays ([5]) and could be at the origin of the observed feature in the spectrum. The average rate of energy loss in CsI is de = MeV CsI dx min g 1 cm 2 for a medium density ofρ(csi)=4.51 g cm 3 leading to a mean rate of energy deposit of de CsI MeV/cm. Since the typical size of the crystal is 1 cm, the average energy dx min ρ(csi)=5.66 deposited by cosmic proton in the sensor is approximately 5.6 MeV. Hence, the increase in the number of particles detected at high energy is likely to be the tail of the charged particle spectrum. The tail overlaps with the energy range scanned by the detector and is further widened certainly due to the non-zero width of the noise of the electronics observed in figure 9. 9 CONCLUSION This laboratory work is a success in the sense that despite the time limitation, we made a prototype of gamma ray detector from A to Z. Thanks to the support of Météo Suisse, the detector was launched using an atmospheric balloon, resulting in the interesting outcome that we have discussed. We fast chose an option for the gamma ray sensor. The PIN diode combined with a CsI crystal has the great advantage over the SiPMs that its behaviour is less dependent on the temperature. However, the sensor should be further improved in many senses. As we saw, the electronics causes a non-negligible noise that must be imperatively lowered. Furthermore, its performance with temperature changes is not clearly known and the issues about low temperatures should not be forgotten. Towards the HAGARE project, all these pieces of electronics that made up the prototype will have to be embedded on smaller circuits in order to lighten each sensor and to be able to build a multichannel gamma ray detector. The lead shielding that will allow each sensor to look independently at different regions in the sky have not been considered during this first semester of prototyping. They should be studied carefully so as to optimize the weight. Charged particle Scintillators Photon Sensor Figure 27: Scheme explaining how one could discriminate between photons and charged particles. O. Girard 31

32 A solution has to be found concerning the issue on the detection of charged particles. They constitute the major part of the cosmic rays and we saw, thanks to a very simple calculation based on the Bethe-Bloch formula, that they can have a substantial impact in the energy range under consideration. Moreover, their passage in the PIN diode has not been accounted for. They will need to be properly discriminated from the γ-photons. One way of doing this is presented in figure 27. By surrounding the sensor with scintillators put in anticoincidence, we can exclude the charge particles. Indeed, their signature is very different from the signature of a photon since they interact continuously with electrons in matter. However, such an active shielding would certainly add some weight and take a rather large space. The electronics would have to be adapted. A new prototype of this kind could be developed using the present one as a basis. This would give a good insight into how such a system could be miniaturized. It could then be tested in the laboratory using cosmic muons or even in the upper atmosphere thanks to an additional flight. If a new flight is planned, the tracking board developed by the LAP ([4]) could be used through some changes. It enables to transmit the data by radio (for which no experience has been acquired during this semester) and to gather information from useful sensors that were not implemented for this first flight. The orientation of the payload would for example be determined. ACKNOWLEDGMENTS I am very grateful to the numerous persons who helped Joël Vallone and me in this project. Prof. Aurelio Bay was always very enthusiastic about the project and he even provided the interface board that we discussed. He had suggestions about the tests and supported me when analyzing the data. Raymond Frei did a huge work by constructing the sensor board itself. Moreover, he was available to answer my endless questions about understanding the electronics. I want to thank him, as well as his colleague Guy Masson for his help when I failed some soldering. I am also thankful to my assistant Ronald Brujin who was here on the first weeks of the semester and learned me useful features to program the Arduino and to code in root. I would like to thank likewise Mr. François Patthey. He provided the vacuum chamber in which we could check the capabilities of the detector. Let me also thank Federico Belloni from the Space Center who supported us with information about space conditions. Finally, I want to express my gratitude to Mr. Rolf Maag of MeteoLabor as well as the Météo Suisse team in Payerne composed of Mr. Jean-Michel Clerc, Mr. Gonzague Romanens and Mr. Gilbert Levrat. They showed a great interest for our project and, by lending us the GPS SPOT and by helping us to launch and recover our detector, they simply made this project possible. REFERENCES [1] T. Kuntzer, E. van Schreven, T. Humair. HAGARE Report Phase /A, High Altitude Gamma- Rays Experiment, EPFL, 212. [2] T. Humair. Balloon-Borne Gamma-Ray Experiment, EPFL, 212. [3] D. Weiss. Simulation of Gamma Ray Propagation in Upper Atmosphere, EPFL, 212. O. Girard 32

33 [4] J. Vallone. Tracking board hardware & software design for the SHAGARE project, EPFL, 213. [5] A. Bay. Introduction aux astroparticules, EPFL, 213. [6] G. Haefeli. Course in Particle Detection, EPFL, 213. [7] Arduino website. ØØÔ»»ÛÛÛº Ö Ù ÒÓº». [8] AME Horten Norway. UV enhanced photodetectors, features. [9] Saint-Gobain Crystals, Scintillation products. CsI(Tl), CsI(Na), Cesium Iodide, Scintillation Material, ØØÔ»»ÛÛÛº Ø ØÓÖ º Òع Ó ÒºÓÑ» Á Ìеº ÔÜ. [1] European Centre for Medium-Range Weather Forecasts. The ECMWF global atmospheric model, ØØÔ»»ÛÛÛº ÑÛ º ÒØ»ÔÖÓ ÙØ» ÓÖ Ø» Ù»Ì ÅÏ ÐÓ Ð ØÑÓ Ô Ö ÑÓ Ðº ØÑÐ. [11] habhub. Landing predictor written by CUSF, ØØÔ»» Ù ºÓÖ»ÔÖ Ø». [12] PDG Particle Physics Booklet, APS physics, 212. O. Girard 33

34 A COBALT SPECTRUM Co spectrum: first hour hist Entries Mean 1.86 RMS Co spectrum: second hour hist Entries Mean RMS st 1 peak : (2.56±.44) V, res.: (24±5.9) % 25 st 1 peak : (2.44±.2) V, res.: (35.7±19) % 2 nd 2 peak : (2.92±.23) V, res.: (14.5±2.2) % 2 nd 2 peak : (2.92±.25) V, res.: (14.8±2.5) % Co spectrum: third hour hist Entries Mean RMS Co spectrum: forth hour hist Entries 2844 Mean RMS st 1 peak : (2.54±.64) V, res.: (24.9±7.8) % 25 st 1 peak : (2.59±.14) V, res.: (14.1±1.6) % 2 nd 2 peak : (2.94±.21) V, res.: (13.1±2.2) % 2 nd 2 peak : (2.94±.12) V, res.: (9.87±.94) % Co spectrum: fifth hour hist Entries 2459 Mean RMS Co spectrum: sixth hour hist Entries 357 Mean 1.46 RMS st 1 peak : (2.6±.15) V, res.: (15.6±2.1) % 25 st 1 peak : (2.59±.19) V, res.: (18.1±2.8) % 2 nd 2 peak : (2.98±.71) V, res.: (9.34±.79) % 2 nd 2 peak : (2.9±.31) V, res.: (14±2.4) % Figure 28: Spectrum of cobalt divided into each hour of measurement. The detector was supplied with batteries. O. Girard 34

35 B FLIGHT CANCELLED Figure 29: Trajectory predicted by Météto Suisse simulation tool for Monday the 27 May at 12z. The flight had to be cancelled because of a too high probability that the balloon ends up in the lake. O. Girard 35

36 C SPECTRUM DURING THE FLIGHT Spectrum: the whole flight hist Entries Mean 2.97 RMS (a) Whole flight spectrum. Spectrum: above 25 km hist2 Entries 856 Mean RMS (b) Spectrum measured above 25 km. Figure 3: Spectra measured during the flight. O. Girard 36

37 D SPECTRUM VARIATION DURING THE FLIGHT 12 1 Spectrum: -25 min thist Entries 6395 Mean RMS Spectrum: 25-5 min thist Entries 1197 Mean 3.55 RMS Spectrum: 5-75 min thist Entries 7845 Mean RMS Spectrum: 75-1 min thist Entries 6716 Mean RMS Spectrum: min thist Entries 6211 Mean 2.91 RMS Spectrum: min thist Entries 1435 Mean RMS Figure 31: Spectrum recorded by the detector throughout the flight. O. Girard 37

38 TP IV LPHE E PICTURES SHAGARE project Conception of the gamma ray detector OF THE FLIGHT (a) The balloon a few seconds after the release. (b) Three hours later, we found the balloon at the top of a dead tree. (c) "Vertige-Concept", a company based in Yverdon and specialized in work at a height, helped us retrieve the payload. Figure 32: A selection of pictures. O. Girard 38