Acknowledgments: I would particularly like to thank Heinz GRAAFSMA for giving me the opportunity to carry out my internship.

Transcription

1 Timothée ALLENET

2 Acknowledgments: I would particularly like to thank Heinz GRAAFSMA for giving me the opportunity to carry out my internship. I would also like to thank: At DESY in Hamburg: Julian BECKER Helmut HIRSEMANN Stefanie JACK Michael LOHMANN Alessandro MARRAS Björn NILSSON David PENNICARD Sabine SENGELMANN Sergej SMOLJANIN Feng TIAN Ulrich TRUNK Trixi WUNDERER At the Joseph Fourier University in Grenoble: Chantal GONDRAN Sylvie SPAGNOLI

3 Introduction 1. Context a. Deutsches Electronen SYnchrotron (DESY) Accelerators Particle Physics Photon science b. Storage-ring synchrotrons c. Free-Electron Lasers (FELs) How do they work? The Free electron LASer in Hamburg (FLASH) The European X-ray Free Electron Laser (EXFEL) FEL vs. Synchrotrons d. Photon Science-Detector Group (FS-DS group) 2. Background/detectors a. Introduction to detectors b. FS-DS detectors c. Solid-state detectors 3. Project Management a. Medipix3 and AGIP b. Task and objectives c. Theoretical approach d. Computer simulations Conclusion

4 Introduction This report accounts for a third year bachelor student in physics and chemistry. During the last year of this bachelor, students of the Joseph Fourier University in Grenoble are asked to do an internship with professionals to discover the working world, carry out a given project but also to confront their knowledge and skills outside of the university. This was an opportunity for me to work in the Deutsches Electronen Synchrotron in Hamburg in the field of photon science in a detector s division. Objectives: Investigate on the institute s activities Obtain general knowledge and understanding of synchrotrons and detectors Assist in the development of the division s detectors

5 1. Context a. Deutsches Electronen SYnchrotron (DESY) The Deutsches Elektronen Synchrotron is a centre in Hamburg for the investigation of structure of matter with an annual budget of 192 million. It was established in The 2000 employees at DESY develop, build and operate particle accelerators for two fields of research: photon science and particle physics. Accelerators One of the specialties of the DESY accelerator division is the construction and operation of superconducting accelerator technology (figure 1), needed for the European X-ray Free Electron Laser currently under construction in Hamburg as well as for the Future Linear Collider in Geneva. Figure 1: schematic cut of a superconductive accelerator cavity Photon science uses particle accelerators as light sources: by accelerating particles to reach high energies and then deflecting them, the electrons lose energy in their change of course which is emitted as radiation. This bright emission of highly collimated electromagnetic waves from electrons or positrons is what we call synchrotron light. It is the brightest known source of X-rays. Particle physics Accelerators are used as super-microscopes by particle physicists. The faster particles go the more energetic they are and the deeper they pierce into matter. The super electron microscope HERA on the DESY campus was shut down in 2007 after 15 years of operation, leaving particle physicists with accumulated measurement data to evaluate. These physicists also work in close collaboration with the CERN on the Large Hadron Collider in Geneva amongst other projects. Nowadays the trend is for High-Energy-Physics (HEP) worldwide to collaborate and work on a few large-scale facilities rather than smaller scale national ones. Photon science On the other hand, DESY has operating facilities for science with photons. Facilities include two synchrotrons and the new free-electron-laser prototype. All three instruments are varied and complementary light sources, which makes Hamburg one of the leading cities in science with photons. The new linear accelerator in construction: European X-Ray Laser (EXFEL) will add on to the choice of light sources in DESY. (See Appendix1: DESY Facilities)

6 b. Storage ring synchrotrons: In synchrotrons, particles are accelerated in segments and deflected by bending magnets. The seemingly circular path of electrons is made of multiple segments linked together: storage rings are in fact big polygons. Synchrotrons use these changes in beam trajectories to emit X- ray radiations. They also use undulator magnets (figure 2) on beam segments to generate more radiation. These are periodic arrangements of many short dipole magnets of alternating polarity. The beam crosses through the undulator which tends to make the electrons move in a wavelike curve, but there is no deflection of the beam before and after its way through an undulator. Figure 2: FLASH Undulator magnet schematics and picture Scientists then use the radiation emitted tangently to the electrons trajectory either by bending magnets or by undulators for experiments after processing the light. Processing includes control of the bandwidth, photon flux, beam dimensions, focus and collimation according to the experimental conditions wanted. Synchrotron light is use for three methods of sample reconstruction: Spectroscopy: determining energy of X-rays Imaging: determining intensity distribution Scattering: determining intensity as function of the angle c. Free-Electron Lasers (FEL) How do they work? Free-Electron Lasers are the fourth generation light sources for photon science. A FEL consists of a linear particle accelerating chain. Its components essentially consist of: o Injector (electron source): provides a high density beam of electrons o LINear ACcelerator (LINAC): accelerates the electrons to relativistic energies o Undulator magnets: generate the synchrotron radiation (photon beam) o Photon beamlines o Experimental stations In FELS, the beam is amplified in the undulator by using the Self-Amplified Spontaneous Emission (SASE) principle: The electrons in the beam start wiggling in the undulator and start emitting spontaneously. The photons emitted travel in a straight line, so they travel faster than the wiggling electrons and end up by overcoming them. These photons form an electromagnetic field which interacts with the electrons on their slalom path.

7 Electrons that are in phase with the transverse pulsating field get decelerated as it passes by, whilst out-of-phase electrons are accelerated. Gradually this process packs the electrons into micro-bunches which are overtaken by one photon wavelength for each oscillation (figure 3). Figure 3: Simulation of microbunching with the SASE principle As a result, all the electrons of the micro-bunches radiate in synchronization, producing an extremely short, intense and coherent pulse of X-rays. Altogether the beams from Free- Electron Lasers are extremely intense and collimated flashes of radiation: they have laser-like properties. The Free-Electron LASer in Hamburg FLASH: FLASH at DESY is a 315 meter long FEL facility. Operating since 2005 it was upgraded in 2007, and has produced 10-femtosecond pulses of ultraviolet and soft X-ray radiation with wavelengths from 60 down to 6.5nm (the record is 4.12nm) which allows users to observe structures down to 13 nm. In FLASH, the electron bunches are produced in a laser-driven photo-injector and accelerated to energies between 440MeV and 1.2GeV by a superconducting linac (record is 1.25GeV). In the injector, electrons are extracted from a solid cathode by a laser beam, focused and accelerated by an electron radio frequency (RF) gun and then directed towards the linac. The linac consists of eight 12m long accelerating modules, each containing eight superconducting cavities and of magnets for beam steering and focusing. The radiation is produced in a 27m long undulator, at the end of which a dipole magnet deflects the electron beam into a dump (figure 4), while the FEL radiation goes on into the experimental hall. Figure 4: Schematic view of FLASH facility

8 Along their acceleration, at intermediate energies of typically 125MeV and 370MeV the electron bunches are longitudinally compressed, which increases the peak current from the 50A output of the gun to several 1000A as required for the FEL operation. The peak current reaches more than 2500A in the end of the chain. The beam in FLASH consists of trains of 2700 pulses at a rate of 4.5MHz repeated with 10Hz. This results in pulses per second (figure 5). Figure 5: Temporal layout of FLASH beam The FLASH experimental hall starts 30 meters behind the last dipole magnet. The photon beam is delivered to one of 5 beamlines in the hall (figure 6). (See Appendix 2: Schematic view of the FLASH experimental hall) Beamlines BL1, BL2 and BL3 receive a direct non-monochromatized beam and each one of these three stations focuses it to a more or lesser degree: when it enters the experimental hall, the FEL beam is 3 to 5 mm wide and can be shrunk down to a spot size of approx 1μm. Stations PG1 and PG2 (Plane Grating) are equipped with a high-resolution monochromator which selects a narrow spectrum of the FEL pulse. Figure 6: FLASH experimental hall A new upgrade is underway to build a second tunnel with undulators and a second experimental hall with 6 measuring stations, which will be operated in parallel with those preexisting. The project also includes technical improvements to increase the beam quality, shorten pulse length and pull wavelength down to 4nm. It will also provide a choice of polarization for the light, which will lead to innovative experiments on magnetic materials.

9 The European X-ray Free Electron Laser (EXFEL): The European XFEL is a 3.4km FEL under construction financed by 13 European countries. This project takes active lessons from the FLASH facility in terms of design and experience. FLASH is often considered as an ice breaker for X-ray free electron science research and technology. The goal is to reach wavelengths of 0.1nm, to be able to detect down to 0.2nm (about the distance between two atoms in a chemical bond) (figure 7). EXFEL FLASH Minimum wavelength (nm) Energy for minimum wavelength (GeV) Number of installed accelerator modules Number of superconducting cavities Bunch charge (nc) 1 1 Bunch peak current (ka) Number of pulses per second Minimum pulse spacing (ns) Peak brilliance (photons/s/mm 2 /mrad 2 /0.1% BW) > Figure 7: Main parameters for EXFEL and FLASH accelerators In the EXFEL, electron bunches will leave the RF gun and enter the linac at 120 MeV. In a longer linac of a 1.6km the electrons will be accelerated to energies of up to 17.5GeV. Unlike FLASH, EXFEL will have several undulators (figure 8) in order to suit different experiments. EXFEL undulators have lengths exceeding 100m (SASE 3) and 200m (SASE 1 and SASE 2). (See Appendix 3: General layout of the European XFEL Figure 8: Schematic view of electron (black) and photon (red) EXFEL beamlines through different SASE and spontaneous emission undulators

10 FEL vs. Synchrotrons There are 3 principle reasons for which FELs are invested in rather than in synchrotrons. The first one is the pulse length: Free-Electron Lasers bring synchrotron light pulse duration from a few hundreds of picoseconds in synchrotrons down to tens of femtoseconds. The shorter the pulse the faster processes can be studied (comparable to short shutter times in fast photography). This enables the scientific community to study very rapid reactions like biomolecular processes that were until now only theoretical approaches. Another upgrade from synchrotrons to free electron lasers is the increase in intensity. (See Appendix 4: Compared peak brilliance of FLASH, future FELs and synchrotrons) Electrons accelerated in storage-rings are too far apart to be aware of each other as they naturally push each other away: different electrons emit independently. Therefore, the total energy produced by a bunch of electrons is proportional to the number of electrons in the bunch. In a FEL, many electrons radiate coherently so that the total energy radiated goes with the square of the number of electrons resulting in radiation powers far greater than in a synchrotron: FEL peak brilliance is approximately 100million orders of magnitude higher than synchrotron pulses. Increasing intensity leads to higher emission of particles when the pulse hits the sample. This renders very high ionization of matter possible: scientists can observe atoms stripped of there electrons. Combined shortness and brilliance of FEL pulses enables single shot acquisitions. Radiation damage interferes with atomic positions and atomic scattering factors but by taking a very intense and short image in time, FEL beams outrun the radiation damage of X-rays and cause enough scattered photons from the sample to have a statistically relevant diffraction pattern (figure 9) from which scientists reconstruct the sample s atomic structure. Figure 7: schematic of FEL scattering In terms, FEL scientists hope to solve the structure of proteins without having to crystallize them like they do in synchrotrons. FELs also have a higher degree of coherence than synchrotrons. This also works in favor of more precise results. In many ways you can compare synchrotrons to light bulbs and FEL to lasers for diffraction experiments. As a whole, 3rd generation storage ring synchrotrons are reaching their technical limits, whereas the two operational free electron lasers in the world (FLASH Germany and LCLS USA) are scouts in the field of research they made uncovered.

11 d. Photon Science-Detector Group (FS-DS group) During my internship at DESY I integrated the Photon science detector group. The group is a fairly new one so far composed of 13members: 4 scientists, 3 engineers, 2 technicians, 3 post-docs and a project coordinator. As detector group of the photon division, this group develops new photon detectors for DESY s synchrotrons (PETRA III and DORIS) and for FLASH and EXFEL. 2. Background/Detectors: a. Introduction to detectors Radiation detection relies on the principle of energy exchanges between an incident radiation and a layer of matter which will react with the radiation: the sensor or active volume. The impact of a radiation with the detector will result in the appearance of an electric charge or several within the active volume. By imposing an electrical field to this volume causing positive and negative charges to flow in opposite directions, we collect this resulting charge to create the basic output signal. The number of charges collected goes with the number of incident radiations over a period of time. To understand this we must look at the energy levels of an electron (figure 10). The periodic lattice of crystalline materials establishes allowed energy bands for electrons that exist within that solid. We call the valence-band the energy level of outer-shell electrons bound to specific lattice sites within a crystal. The conduction band is the energy level of electrons free to migrate through the crystal. The band-gap is the gap between the two. The energy one must provide to promote an electron from the conduction band to the vacuum level is the ionization energy. In a sense it is the energy an electron requires to escape from any atom s electromagnetic force. Absorbed radiation creates electrons in the conduction band and corresponding holes in the valance band. A hole is therefore the absence of a negative charged electron in a crystal. Sensors are made of gas or semiconductors because insulators have a band gap too big to allow any conduction, and conductors have no band gap which would result in collecting more charges than the ones emitted by radiation. With the advances made in manufacturing we can obtain semiconducting sensors almost exclusively rid of impurities with materials like silicon which make SSDs very stable and reliable detectors. For a photon detector, we will need a matter capable of interacting with incoming particles, and an electronic component to process the current generated by that interaction. Figure 8: Band-gap schematics

12 The ionization energy is also temperature dependent, according to Boltzmann s law. The probability of an electron-hole pair being thermally generated increases with the temperature. These electrons bias the measure. We call them dark current. There are two ways of processing the generated current from a detector: depending on the flux of incoming particles, we use counting or integrating detectors. Electronically speaking the detection of a particle comes with a dead time in which other incoming particles will not be seen. When the flux is low enough for the dead time not to overlap several signals we use counting detectors which treat single pulses each generated by one absorbed photon. If the flux is too high, we use integrating detectors which integrate a whole batch of charges over a period of time. Photon counting comes with no noise, as we simply observe the amount of signal that goes over a threshold when integrating detectors integrate noise and leakage current. Integrating detectors are however faster to read out. b. FS-DS detectors The photon science detector group in DESY has to provide for different sources and therefore develop different detectors: counting for their synchrotrons and integrating for FLASH and EXFEL amongst other projects. This group specializes in semiconductor detectors: detectors which use semiconducting matter as sensor.

13 3. Project Management My participation in the detector group concerns development of two detectors: the MEDIPIX3 and AGIPD. a. MEDIPIX3 and AGIPD MEDIPIX3 is a HPAD detector readout chip developed at CERN. Each pixel on a semiconductor sensor is bonded to a channel on the Medipix chip. The Medipix chip performs single photon counting: each X-ray hit on a pixel generates a signal pulse, and the pixel circuitry counts the number of hits over an adjustable energy threshold. The detector can achieve single-photon sensitivity, and can be read out at a high rate. The FS-DS group is in charge of designing larger-area Medipix3 modules, incorporating 12 Medipix3 chips to give an array of 1536 by 512 pixels. The Adaptive Gain Integrating Detector (AGIPD) project is a collaboration mainly between DESY and PSI in Switzerland. It is a project to build a pixellated SSD for EXFEL. The final goal is also to have a large area detector of 1000 by 1000 pixels. This will be achieved by assembling 16 modules, each incorporating 16 chips. AGIPD and Medipix3 are both Hybrid Pixel Array Detectors (HPAD s). They are composed of different layers (figure 11): Sensor: creates current signal according to incoming radiation Chips: they collect and process sensor current PCB: collects processed chip data and relays it to a computer The bump bonds and wire bonds relay electric signals between layers. Figure 9: Hybrid Pixel Array Detector Schematics The heat spreader is the most important component to limit constraints in the detector. It dissipates heat over a larger surface, and also prevents the chips and PCB from ungluing themselves due to material expansion when heated.

14 b. Task and objectives My job concerned the study of thermal conduction in these two large area modules. In an attempt to minimize dark current, we must cool detectors to avoid any thermally excited electrons, as explained above. Detector components are therefore heated by all the chips electronics on one side and cooled and their base (module holder), which generates thermal gradients. What s more, for mechanical reasons and for the detector to work, it is best: To avoid cracks in components due to large gradients within components To predict thermal expansion of components to make sure they will stick together even under constraint. The study of the heat spread is therefore crucial in the overall design and construction of detectors or other thermally challenged objects. Not paying attention to this would result in taking huge risks in terms of detector reliability and lifespan. First I made a theoretical approach to the heat spread before programming and running computer simulations. My first simulations were targeted towards a general understanding of the heat distribution in the two detectors by simulating a large batch of results with varying dimensions of detector parts which helped understand different heat profile trends. This showed what components were influent in heat spread and which were critical or negligible by measuring how much heat every layer blocked. Then my work came in close collaboration with all the other people working on these projects on a more or less locked design, to try different ideas for components and see if they would fit within the thermal constraints. In a sense thermal simulations help to lock the design of the detectors parts, and pointed out parts to work on. All of this comes with a lot of judgment on what is useful or not. The point is to improve the operation by making the cooling temperature easier to achieve (the lower you want to cool the more complicated it gets), and to have the heat spread sufficiently over the detector for it to work.

15 c. Theoretical approach With the design of these modules, I was able to estimate component resistance to heat by comparing the assembly to an electric circuit. Component resistance is calculated with matter properties and component dimensions. With the values of resistance and flux we can estimate temperature gradients. An example of results is shown in figure 12. Figure 10: Layer resistance for AGIPD 16 chip (right) and Medipix3 12 chip (left) modules It is important to note that the results shown are estimates along the z axis only. Some layers are non-isotropic as we have shown in the design. From these results we see that the glue between layers and the PCB are going to be the biggest thermal issues. This method is extremely repetitive and only considers heat spread as thin columns with homogenous heat at given heights: it is fully relevant for single-dimensional systems only. This doesn t take into account the horizontal heat spread in our design and therefore doesn t help study heat transfer at different parts of our assembly but gives us the average measures. It is possible to study X, Y and Z heat conduction precisely by considering a much more complicated electric circuit but at this point, using a dedicated computer simulations software becomes more appropriate. As a whole this method is useful to get a general idea of which layers and components hold heat and whether heat gradients will be critical for the use of these detectors, but not for precise heat conduction. To get a much more precise study and easily handle different design ideas, we have a computer software estimate all of this instead.

16 d. Computer simulations Programming The first step to running computer simulations was drawing the detectors into the simulation program: ANSYS (figure 13). This part was much like any other computer assisted technical drawing task. Some simplifications were made for the consecutive simulations to run more easily. For example the bump-bonds were not individually drawn but rather as a thin layer. Another example is the PCB which consists of different readout and ground layers: rather than drawing the exact layout of a PCB and changing for every simulation, we drew a large block. Figure 11: ANSYS 3D drawings of Medipix3 12 chip module; ANSYS runs simulations according to attributed thermal properties for different materials. Estimates were made for the bump-bond layer and PCB in the program to fit the conductivity of the real design. All the estimates were targeted for the worst case so as to only see improvement under real operation. (See Appendix5: Single layer heat) Present design discussion In terms of present design, the Medipix3 PCB is already manufactured: it is a thick ceramics PCB with silver ground layers and vertical silver thermal vias to leak heat through. The AGIPD PCB design on the other hand is still being discussed and many ideas are being tried. In terms of sensors, silicon will be used at first, but one of the goals of the FS-DS group is to move quickly towards Germanium sensors. Germanium has as few imperfections as silicon, but also has a higher radiation absorption ratio. This would lead to better detection but also to less detector damage due to radiation on the chips: if more photons are absorbed in the sensor, less reach and damage the chips. However Germanium must be operated at temperatures under -50 C. Silicon doesn t have to be cooled as much. This makes thermal conductivity a tricky part of the detector design for Germanium sensors. With Silicon sensors, the main concern is for the detector not to break, by avoiding gradients, but with germanium sensors, we must in addition make sure the cooling is sufficient for the sensor to reach -50 C. This implies either working on the detector s conductivity or working on techniques of cooling which gets tricky under -50 C. The cooling is applied at the bottom of the PCB for Medipix3 and the bottom of the module for the AGIPD modules. The cooling contact is a rectangular tube due to the connector hole in the middle, for the connector to reach the PCB. In terms of multi-module detectors like the AGIPD 1k*1k pixel detector, we do not consider radiated heat from one module to the other as they will be operated in vacuum.

17 Results For the analysis I chose to measure maximum and minimum temperatures of different component faces, to study overall heat and gradients. (See Appendix 6: example of measures taken) A first study with homogenous cooling on the bottom of the PCB was made to measure vertical gradients. Once that was done I studied the real design to have a look at horizontal temperature constraints (figure 14). Figure 12: AGIPD cut with global temperature Result analysis First of all the simulations show that the thickness of the sensor has a big effect on the heat differences in the sensor but not so much on the temperature difference from the top to the bottom of the detector. This confirms our expectations since the sensor isn t situated between the heat source and the cooling, and therefore has very little saying in vertical gradients. The simulations also show that the thicker the heat-spreader the better. In reality the thickness of this component is limited by the length of the wire-bonds. More importantly the simulations confirm the theoretical approach made: the heat blocking parts of the assembly are the glue and the PCB in both detectors. This means there is work to be done on a gluing protocol. The aim is to have the thinnest and most spread glue layer so as to spread heat the most, without overflowing on the side of the chips. Overflowing would create electrical contact between the chips and make the detector unusable. What s more all the chips on a module must be perfectly horizontal and exactly at the same height in order to get a proper image. However we cannot press the chips too hard to align them horizontally in case they might break. The FS-DS group made progress on this problem and developed a gluing station. The most likely idea for applying the glue is to use stencils to have the same amount and height of paint under each chip. Concerning the PCB, the simulations show it is the most constrained component of the detector, as it is only cooled on the sides but heated all over. There is work to be done on the AGIPD PCB. Different ideas are being considered, amongst very thin flexprint PCB s or thicker ceramic PCB s like the medipix3 ones which are very expensive. It is important to work on some kind of heat-spreading part of the PCB. These can be thermal vias or simply thermal conducting ground layers (copper, silver, etc )

18 Another way to minimize thermal gradients especially horizontal ones is to work on the cooling plate itself. By widening the contact surface between the cooling plate and the module the PCB can be cooled much more. The first way is therefore to manufacture the cooler with a hole just sufficient for the connector to go through. It would also be interesting to design this cooling plate with a slot to slide the modules in, in order to have not only the bottom surface cooled but also the sides. It is finally worth noticing that the high-pin connector (500 pins) will also provide a thermal path to the outside. The effectiveness of this depends, however, on the thermal conditions (cooled or not) of the electronics connected to the PCB. As this is yet undefined, this could not be included in the thermal simulations presented here. Conclusion In this report, only global conclusions and examples are presented, a detailed dossier with results has been deposited with the engineers of the FS-DS group. The simulation software developed is also an important and flexible tool that will be reused. Concerning the given project, I feel like my skills turned out very applicable to the tasks given. The project was understood, handled and successfully carried towards the results asked for. My work felt useful for the development of the detectors, I think it was a real contribution to these projects. This internship was extremely motivating to continue my studies. The motivation comes from the fact that there is a lot more to learn and to master by studying longer. However I think one month is too short for an internship and limits the impact one can have on the working environment, if the objective is to actually put into place a project rather than getting a glimpse of the working world. Finally I would like to emphasize how beneficial the fact that this internship was international. Working in this kind of environment makes communication trickier with other colleagues, especially in an internship like this one where there was a lot of contact with foreign institutes. Overall this internship was very confronting, interesting and motivating.

19 Appendix 1 DESY Synchrotrons and Free Electron

20 Appendix 2 Schematic view of the FLASH experimental hall

21 Appendix 3 General layout of EXFEL

22 Appendix 4 Compared peak brilliance of FLASH, future FEL s and synchrotrons

23 Appendix 5 Single layer heat measures sensor bumpbonds ASICs glue

24 heatspreader glue PCB module

25 Appendix 6 Example of measures taken Si HeatSpreader with 4*80microns kapton layers and 3*10microns copper PCB; 50% 1.37tc glue AGIPD\Si HS_KaptonCu PCB_50pc 1.6 glue Thermal coefficient width dimensions Tmax Tmin material (W/m.K) (mm) ( C) ( C) dthoriz Vertical Horizontal Sensor Si * Bumps Indium E * ASIC Si * Glue1 EPOTEK * HS Si * Glue2 EPOTEK * PCB Kapton+Cu * Glue3 EPOTEK Module Al AGIPD single module thermal simulation results Kapton+Cu PCB; ~50% glued ASIC with 1.6tc Max and min face temperatures T ( C) sensor top sensor btm Bump top Bump btm ASIC top ASIC btm Glue1 top Glue1 btm HS top HS btm Glue2 top Glue2 btm PCB top PCB btm Glue3 top Glue3 btm Module top Module btm 1