Home

Charakterisierung hochbestrahlter polykristalliner Diamantsensoren

1. C V 500 kHz 250 mV 500 ms E Ramping Hl 2 E LU x 2 MI Im X down g 10 ji o Z IR x E x 1h Ir ai H x x Pre fr ae n I el Mi 5 L x 1 ni FL ie N mer L kl Manni IN 3 nm MX hi N re 7 Ki Loi S 10 m De IRIS E x The 3 E N x 2 H x 5 ae Ra Rx rila ss iii ile ind O I O N N i LI LO AO NO O N RZ 1500 1000 500 0 500 1000 1500 voltage V b Voltage sweep in negative direction Figure 5 11 Uncompensated charges with respect to the bias voltage A theoretical change of the sign of the uncompensated charge is indicated by vertical lines uncompensated charge It is even orders of magnitude higher than the typical doping 012 16 charges cm concentration in Silicon of 1 Most probably this is caused by the heavy irradiation which creates vacancies and inter stitials which function as charge traps These defects can contribute as uncompensated charge carriers The filling and defilling of those traps could explain the overall behaviour of the capacitance with respect to the voltage especially the hysteresis if the cross sec tion of the processes changes differently with voltage However this does not explain the bump and thus the theoretical change of the sign of uncompensated charge carriers A measurement with the diamond being pumped which is illustrated in Figure in dicates that the bump may not be caused by the charge traps In this case the diamond 38 5 2 Ex
2. 10000 5000 E ridi liv Ds 1 5 0 4 0 3 0 2 0 1 0 0 1 0 2 0 3 0 4 0 5 signal V Figure A 13 Noise after the shaper with diamond sample 1 connected and a bias voltage of 500 V The shaper was set to a shaping time of 2 us and a nominal gain of 1000 12 A 3 CCD measurement A 3 3 CCD with respect to the bias voltage CCD vs time Data E w CCD um 9 65 0 03 n zu 8 F _ L_T Epp 05 ru I rat 9 ul a gg ge 75 6 time h Figure A 14 Measurement of CCD of diamond sample 1 over time at a bias voltage of 200 V Only statistical errors are shown CCD vs time TE a 5 13 6 CCD um 13 15 0 03 QA pr 7 w N o N N SSER EHEN SURE N GENS RES ERE NERE ERE yt ER SO p p ani 44 4 nd cae yp oN N gt D N N a time h Figure A 15 Measurement of CCD of diamond sample 1 over time at a bias voltage of 400 V Only statistical errors are shown 73 A Appendix CCD vs time Data Th CCD um E N oa ee eee a 15 8 B a AT ge t_ 15 2 a opty oo N gt o 12 time h Figure A 16 Measurement of CCD of diamond sample 1 over time at a bias voltage of 600 V Only statistical errors are shown CCD vs time
3. CCD vs time ve m Bong a age o F ae O C pe ape 18 a C me 16 a C aw ER od 14 Pa n af 12 m r ll we I an 1 jH E wo ae er I 0 20 440 60 80 100 120 t h Figure 6 5 Measurement of CCD of diamond sample 2 over time at a bias voltage of 700 V Before the measurement the diamond was completely unpumped An increase of CCD which is interpreted as pumping is visible Only statistical errors are shown for the first hour of the measurement is given in Figure The CCD raises in time at the beginning steeper than after four days A similar result is obtained for diamond sample 1 This is shown in Figure Even though the increase of CCD with time for this sample is not as steep as for sample 2 it also does not reach saturation after 3 days Pumping the DUT during a measurement is very slow as Figure shows Even after five days the CCD does not saturate Thus to determine the maximum CCD the DUTs are pumped before the measurements by removing the collimator A very interesting result is given in Figure 6 8 In this case the diamond sample 2 was held at a bias voltage of 1000 V while being pumped for over 24 hours Thus it is expected that the DUT is fully pumped and the CCD stable in time However in this case the CCD actually decreases in the first few hours This can be explained by a short detrapping time of certain states at this high voltage which leads to an overall decrease of filled t
4. Data CCD um 17 55 0 12 i wo a CCD um _ N ie F A N x dl time h Figure A 17 Measurement of CCD of diamond sample 1 over time at a bias voltage of 800 V Only statistical errors are shown CCD vs time Data E sf CCD um 20 16 0 35 A L S L 247 221 H i i H eo 1 0 1 2 in h Figure A 18 Measurement of CCD of diamond sample 1 over time at a bias voltage of 1000 V Only statistical errors are shown 74 CCD vs time A 3 CCD measurement 15 CCD um 14 8 14 6 14 4 14 2 14 13 8 Data abit 1 tH bh Lt a 7 TT fl T 10 time h Figure A 19 Measurement of CCD of diamond sample 1 over time at a bias voltage of 400 V Only statistical errors are shown CCD vs time E 17 2 Q 6 17 oO 16 8 D oa TTT IRLSRABRRARZRNANRLENRERE A a avi a D i CCD um 16 25 0 05 Data Fit _ n ec N qui TTT time h Figure A 20 Measurement of CCD of diamond sample 1 over time at a bias voltage of 600 V Only statistical errors are shown CCD vs time OHR T E Fit CCD um 18 34 0 17 a 27 o o ttt L 1 18E r Prt EF ss LE I I ee E
5. However the power threshold for graphitisation in the bulk of this sample can even be a little higher for two reasons Single hot i e high energetic electrons are needed as a seed for the avalanche ionisation causing the phase transition Due to the incomplete crystal lattice at the surface and at grain boundaries it might need less energy to create these seed electrons Additionally all these spots were produced on the growth side of the diamond which is not even This can be seen e g in Figure The rough surface might cause some reflections and thus increase the fluence locally This increased local fluence can create small nano sized graphite grains at grain boundaries At these graphite spots it takes much less energy to create a seed electron as graphite has no band gap Thus once a nano sized graphite grain is created it grows with each shot even at energies below the nominal threshold For a power setting of 545 5 pW nearly complete graphitisation of the spot size at 59 7 Graphitisation of diamond using a femtosecond laser Figure 7 6 Graphite spots for different times of exposure to the laser beam From left to right 584 6 pW for 3 s 482 5 pW for 60 s 40 s and 20 s no transformation visible the surface is observed For the graphite spot produced with a power of 584 6 pW an even larger fraction of the surface is graphitised In this case the shadow caused by the plasma is visible again 7 2 4 Va
6. 146 144 142 10 11 distance from lens cm Figure 7 2 Power profile of the laser beam for an attenuator setting of 2300 au The power is averaged over one second direction perpendicular to the beam with an actuator which is also controlled via a PC A shutter has to be controlled manually This makes it impossible to have a shorter duration of exposure of the DUT than 1 s 7 2 Experimental results 7 2 1 Calibration For the graphitisation it is important to know the intensity of the laser beam As the intensity is controlled by different settings of the attenuator the power of the laser beam has to be measured for each setting This is done by putting a power meter into the beam Due to the pulsing of the laser the sensor averages the power over one second The power profile for a certain attenuator setting is illustrated in Figure It is stable with distance to the lens up to distances greater than two times the focal length of 5 cm As the sensor has only a limited active area this may be due to the widening of the beam Thus it can be concluded that in the focal region there is no dependence for the power with respect to the distance to the lens The power of the laser beam for different settings of the attenuator is shown in Figure 7 3 and summarised in Table 7 1 All measurements were taken in the focal plane The attenuator shows no linear behaviour For calculating the intensity the beam spot size is needed For an est
7. GEORG AUGUST UNIVERSITAT Fakult t f r AZ G TTINGEN Physik D Master s Thesis Charakterisierung hochbestrahlter polykristalliner Diamantsensoren f r ionisierende Strahlung Characterisation of highly irradiated polycrystalline diamond sensors for ionising radiation prepared by Lars Graber from Helmstedt at the II Physikalischen Institut Thesis number II Physik UniG6 MSc 2011 01 Thesis period 14th March 2011 until 13th September 2011 Supervisor Dr Jens Weingarten First referee Prof Dr Arnulf Quadt Second referee Priv Doz Dr J rn Gro e Knetter Zusammenfassung Hochbestrahlte Diamanten wurden auf ihre Eignung im Bezug auf die Verwendung als Sensormaterial fiir Spurdetektoren untersucht Hierf r wurden zwei Teststationen f r C V und CCD Messungen aufgebaut Es zeigt sich dass fiir hochbestrahlte Diamanten eine Abhangigkeit der Kapazitat von der angelegten Spannung besteht Messungen der CCD sind sowohl bei sehr niedriger CCD als auch fiir einen groBen Spannungsbereich mit hoher Pr zision m glich Graphitisierung von Diamant mittels eines Femtosekundenlasers wurde zur Herstellung von Elektroden im Material fiir 3D Sensoren untersucht Stichworter Hochenergiephysik Diamant Halbleitersensoren Abstract Highly irradiated diamonds are studied for their use as sensors for tracking detectors Two different setups for C V and CCD measurement were built For highly irradiated diamonds a change of the capaci
8. Picture of the two spots from the side The substrate side is on the top and the growth side at the bottom Figure 7 8 Two graphite spots produced with a power of 584 6 pW at the edge of the diamond sample The phase transformation is only induced at the surface and not in the bulk of the diamond 62 8 Conclusion amp Outlook The results presented in this thesis show that diamond can be used as a sensor material even after high radiation doses The measurements of the capacitance with respect to the bias voltage show that for highly irradiated diamonds the capacitance shows hysteresis effects This is caused by the charge traps which are a result of the irradiation Thus the CV measurement can be used to determine the quality of irradiated diamond samples The CCD measurements show that samples with very low CCD at low voltages can be measured with a high statistical accuracy Even from samples which cannot be measured with other setups a CCD can be extracted The setup was designed to have a very low noise level as the voltage signal in diamond is very small The only remaining problem is the trigger efficiency of the scintillator because it has direct influence on the measured value of the CCD As it was not possible to tune the whole setup with a diamond sample with a high known CCD the systematic uncertainty is still quiet high The influence of pumping was observed in both measurements As the pumping fills the charge traps
9. in this example 3 8 101 protons cm they have the same CCD as the unirradiated pCVD sample This hints that processes of the damage of the bulk from irradiation is for both types the same Especially scCVD diamond is not radiation harder than pCVD diamond So scCVD dia monds have a satisfying CCD when unirradiated but at high doses they suffer under the same low values as pCVD diamond 19 4 Bulk segmentation l Preliminary Summary of Proton Irradiation 500 450 400 u i H 350 X i Open Circles scCVD x shifted by 3 8 i Solid Squares pcvD i Curve c d ccd0 1 k phi ccd0 300 250 200 150 100 50 0 charge colection distance um 0 5 10 15 20 25 Fluence x10415 p cm 2 Figure 4 1 Measured CCD before and after irradiation with 24 GeV protons for scCVD open circles and pCVD solid squares diamond For both types the CCD at zero fluence is normalised to 220 um The data for scCVD diamond is shifted to the left to match the damage curve of the pCVD diamond This shows that both types have the same damage constant 16 The more promising approach to have full charge collection efficiency even at a high ra diation dose is to reduce the space between the electrodes Most modern hybrid tracking detectors use a planar electrode design e g the ATLAS Pixel Detector L7 This means that reducing the space between electrodes results in thinner sensor material which is often unfavourable as the depo
10. of trapping this does not yield an efficiency of 100 as the CCD is only the mean distance a charge carrier travels before being trapped The aim for diamond sensor production is to exceed the distance between the electrodes with the CCD as much as possible This can be achieved by using diamonds with larger CCDs or by reducing the distance between the electrodes Diamonds with larger CCD are generally scCVD diamonds due to the absence of grain boundaries Some good pCVDs have also large CCDs but these are expensive and have a slow production rate So both cases would significantly increase the costs for a diamond detector compared to using cheap pCVD diamonds Additionally the production of sc CVD diamonds is much slower and more complicated Polycrystalline diamond can be grown in full wafer size whereas areas of only about 1 cm are achieved with singlecrys talline diamond Another reason to not just rely only on the higher CCD of scCVD diamonds is the de crease of CCD in both types of diamond with increasing radiation dose An experimental result is illustrated in Figure 4 1 It can be concluded that the CCD decreases exponen tionally with respect to the radiation dose for both types Moreover they even have the same damage constant This is shown by a shift of the scCVD data to lower fluencies so that it fits to the damage curve of the pCVD samples for negative fluencies Thus scCVD diamonds have a higher initial CCD but after irradiation of
11. used Both signals are recorded by an oscilloscope and transferred to a computer shaping times are 100 ns 500 ns 2 us and 8 us Additionally it also amplifies the signal with a nominal gain of 10 100 or 1000 The preamplifier has a decay time of 140 us The signal for each event after the shaper has only a full width at half maximum FWHM of 2 4 times the shaping time and thus reduces the likelihood of pile up effects significantly The scintillator is used as an external trigger It has a surface of approximately 1 cm and thus covers the whole solid angle seen by the diamond sample With a thickness of 1 cm it is assumed that it has a detection efficiency of nearly 100 if the rate is not too high for the photomultiplier 6 1 1 Data aquisition amp analysis To read out the signal a Tektronix DSP 4104 oscilloscope is used It is set to trigger on the signal from the scintillator The read out is realised via USB and controlled by a self written LabWindows program This is either done for each single hit or for the mean value of 128 hits For the measurement of the CCD the oscilloscope is set to average over 128 triggered hits This event is read out and stored with a time stamp As the averaging is done on the oscilloscope the information of the single triggered hits is lost This makes this type of data acquisition vulnerable to inefficiencies of the scintillator If the scintillator does 43 6 Measurement of CCD not trigger a
12. 1127 Dia 1 n 108 10 6 T 10 4 1022 EL Sie 0 1 2 3 4 6 time h Figure 6 10 Measurement of CCD of diamond sample 1 over time at a bias voltage of 200 V Prior to this measurement the diamond was fully pumped at a higher electric field During the first two hours an increase of CCD is visible Only statistical errors are shown for at least 24 hours After the CCD is measured for this voltage for several hours the voltage is decreased The removal of the collimator to pump the samples can cause a systematic uncertainty as the metal box and the scintillator might be moved during the removal Thus the measurements at the lower voltages are done without further pump ing As more charge traps can be filled at a lower electric field a slight pumping effect for these measurements is observed An example is given in Figure 6 10 The increase of CCD during the first two hours of the measurement hints that the detrapping time for charge traps might decrease with increasing electric field For the determination of the saturation CCD the time of pumping is excluded from the linear fit Plots of the individual measurements with the fit are given in Figure to A 32 No clear saturation for the CCD with respect to the voltage is observed up to 1000 V which corresponds to an electric field of 2 5 V um and 1 9 V um for diamond sample 1 and 2 respectively The increase of the CCD is steeper at lower voltages Even at fairly low volta
13. 445 1958 K L Choy Chemical vapour deposition of coatings Progress in Materials Science 48 2 57 2003 C Bauer et al Radiation hardness studies of CVD diamond detectors Nucl Inst and Meth A 367 1 3 207 1995 Proceedings of the 7th International Wire Chamber Conference M Capeans et al ATLAS Insertable B Layer Technical Design Report Technical Report CERN LHCC 2010 013 ATLAS TDR 019 CERN Geneva 2010 O S Briining et al LHC Luminosity and energy upgrade A Feasibility Study Tech nical Report LHC Project Report 626 CERN LHC Project Report 626 CERN Geneva 2002 G Lutz Semiconductor radiation detectors device physics Springer Verlag 2007 81 Bibliography 14 D K Schroder Semiconductor material and device characterization Wiley IEEE 15 16 17 18 19 24 25 26 82 Press 2006 H Spieler Semiconductor Detector Systems Oxford University Press 2005 D Asner et al Diamond pixel modules Nucl Inst and Meth A 636 1 Supple ment 1 S125 2011 7th International Hiroshima Symposium on the Development and Application of Semiconductor Tracking Detectors T Rohe et al Sensor design for the ATLAS pixel detector Nucl Inst and Meth A 409 224 1998 S I Parker et al 3D A proposed new architecture for solid state radiation detectors Nucl Inst and Meth A 395 3 328 1997 Proceedings of the Third International Workshop on Semicondu
14. The capacitance C of a fully depleted p n junction can be approximated by a parallel plate capacitor Its capacitance is given by A C Goer 3 6 A denotes the area of the electrodes The derivative of the external voltage with respect to the depletion width can be calculated via dvix d Vii Vez dVi w 0m w 3 7 The variation of the internal bias voltage Vp with respect to the depletion width is very small and can be neglected So combining Equation 3 7 with Equation 3 4 yields dV r _ _ w l dW EoEr 3 8 Using this result and Equation 3 6 the derivative in Equation can be solved to dC __e0A4 W dis WeNp 0 r 0 r A WeNp C3 0 rA42eNp oa E 3 9 Ener A2e dV on The last equation gives information about the overall doping concentration at a specific depletion depth W For a full profile of the concentration it has to be measured at differ ent voltages These voltages can be transfered into the depletion depth Unirradiated diamond sensors are neither doped nor have intrinsic uncompensated charge centers The expected result for Np or N4 is zero This means that the change of capac itance with respect to the bias voltage has to be infinite according to Equation This shows that the model does not describe diamond well The assumption that the depletion depth depends on the bias voltage is not valid for a perfect diamond crystal which is free of intrinsic charges For a diam
15. a 0 greater than 1000 losses due to Bremsstrahlung are dominating For a Gy lower than 0 1 energy is lost due to nuclear reactions An important aspect to note is that dE dx only gives the mean energy loss For thin sensors the energy loss follows a Landau distribution So the mean value does not equal the most probable value which is significantly lower An example for this is given in Figure for a minimum ionising particle m i p in silicon The minimum of the Bethe Bloch formula is located at a Gy of approximately four As the energy loss increases only slightly for higher Gy before reaching a plateau all particles with Gy above four are called m i p Their energy corresponds to the minimal loss of energy due to ionisation effects So a m i p generates the lowest signal in a detector Electrons lose energy mainly through Bremsstrahlung The loss is directly proportional to the energy E of the incoming particle dE E de Xo 2 1 Energy loss of particles L T a za u on Cu l 100 gt E 4 Bethe Bloch Radiative I c L Anderson J Py Ziegler EE 6 3S 080 2p EH Radiative 4 J Ro E Minimum effects PA losses 3 amp t ionization reach 1 ERRE ENuclear E N eee Ale en i Golose E Cc Ve eee N ee j Without 1 0 001 0 01 0 1 1 10 100 1000 104 10 106 By l 0 1 1 10 100 jl 10 100 1 10 100 MeV c GeV c TeV c Muon mo
16. detector can be very low This 13 3 Diamond as a sensor material for particle detectors is important for environments with high particle flux such as for the ATLAS upgrade project of a new innermost layer of the Pixel detector IBL LI or sLHC 12 Due to the high breakdown field of 10 V cm of diamond it is possible to apply high voltages to such a sensor At high electric fields the saturation velocities of the charge carriers can be reached and the overall charge collection benefits from this see Equation 3 1 The roughly 40 longer radiation length of diamond compared to silicon means that less energy per distance is deposited in diamond This and the larger band gap result in a much smaller signal from a particle in diamond for equally thick sensors But in combination with only minimal or no cooling for the sensor itself the material budget of a diamond detector can be very low However a major disadvantage is charge trapping on e g grain boundaries When the charges are trapped they can no more contribute to the signal Thus trapping decreases the signal Using single crystalline diamonds and other improvements can increase the mean free path of charge carriers and thus limit charge loss The energy needed to remove an atom permanently from its place in the crystal lattice the displacement energy is at least two times higher in diamond than in silicon This contributes to the radiation hardness of diamond as damages of this ty
17. holes with an energy level close to the valence band The electrons and holes recombine near the interface This creates a space charge region called depletion zone It is typically expanded through the whole sensor via an external reverse bias voltage Charges in this depleted region are collected and thus it is important to know what bias voltage has to be applied to completely deplete the sensor The width W of the depleted region can be calculated via 13 W ee Np Vii u Ve 3 3 eNaNp o and are the vacuum and relative permittivity respectively e is the elementary charge N and Np are the concentration of acceptors and donors respectively Vp is the potential caused by the depleted region in thermal equilibrium and V is the external bias voltage For the case of reverse biasing it is Vez lt 0 Typically one type is much heavier doped than the other The advantage is that the depleted region propagates nearly exclusively into the weaker doped side Assuming a heavily doped p type and a weakly doped n type Equation 3 3 simplifies to 2 0Er eNp Voi Vex 3 4 To calculate the width of the depletion region the doping concentration of the weakly doped material has to be known This concentration can be derived from the variation of the capacitance with respect to the applied bias voltage The Ansatz for this calculation is dC dC dW Ver dVes dW 3 5 16 3 4 C V measurement
18. low ones namely 5 MHz and 1 kHz respectively The systematic uncertainty also depends on the frequency For low and high values it is significantly higher than for values between 100 kHz and 1 MHz Within this range the measured values for the capacitance are also nearly equal For very low frequencies the measured capacitance is smaller and for very high ones significantly larger Thus for all following measurements a frequency of 500 kHz is chosen Another variable is the amplitude of the applied oscillating voltage The result for differ ent levels of the amplitude is given in Table The mean values are nearly independent 33 5 C V measurement of the oscillator level but the spread for individual measurements rises with lower ampli tudes Thus the statistical error for an amplitude of 32 mV is an order of magnitude larger than for 250 mV An oscillator level of 250 mV is chosen for all following measurements 5 2 2 Voltage sweep As stated in Chapter 3 4 a change of the capacitance with respect to the applied voltage can yield some information about intrinsic charge carriers in the diamond For an insula tor like unirradiated diamond no free charge carriers and thus no change of capacitance is expected For the measurements the voltage is swept from negative to positive bias voltages and vice versa As any change of the capacitance with respect to the bias voltage is small this sweep is repeated several times to reduce the stati
19. pass the DUT It has a thickness of 1 cm and a hole of 1 mm This reduces the hit rate in the diamond to 3 83 kBq For a detailed calculation see Appendix A 3 1 For the read out of the DUT the clamp is connected with a SHV cable to a charge sensitive preamplifier CSP10 manufactured by Fast ComTec GmbH The cable has to be as short as possible to reduce the noise pick up through this connection To further reduce the noise from the preamplifier its bias voltage is supplied by a low noise power supply A sketch of the read out chain is shown in Figure The preamplifier has a nominal amplification of 1400 mV per 1 pC To reduce the noise a CSA 4 shaper from Fast ComTec GmbH is used Its available 41 6 Measurement of CCD a Picture of the plastic tube with source and b Picture of the fixture for the CCD measure metal box inserted and the preamplifier Not visi ment The samples are held by a Copper clamp ble are the collimator and the scintillator under the A hole in the clamp ensures that the material bud box get of the fixture is as low as possible Figure 6 1 Pictures of the setup Plastic tube Figure 6 2 Cut through the setup 42 6 1 Experimental setup Sr90 Collimator ea ADP Preamplifier Oscilloscope Shaper Figure 6 3 Illustration of the read out chain The signal from the diamond is processed in a charge sensitive preamplifier and a shaper As an external trigger a scintillator is
20. pm Sample 3 Unirradiated diamond This diamond sample was never exposed to huge amounts of irradiation The physical size of this pCVD diamond is rather small with the gold electrodes having an area of 0 133 0 005 cm each It has a thickness of 507 5 pm and is of optical quality 5 2 Experimental results 5 2 1 Calibration To determine the capacitance a periodic voltage pulse has to be applied to the device under test DUT i e the diamond sample It is important to know if the capacitance varies with respect to this frequency and its amplitude Thus at a bias voltage of 0 V it is measured for different frequencies For better statistics the measurement is repeated 1001 times with an interval of 500 ms for each frequency As the program of the B1505A does not allow repeated measurements at the same bias voltage the voltage is swept from 0 V to 1 mV with a nominal interval of 1 uV However this sweep can be considered as a measurement at fixed bias voltage as Figure shows in which all measured bias 31 5 C V measurement Distribution of bias voltage 500 kHz 250 mV zZz E Entries 1001 g 60 Constant 59 25 2 30 Mean 0 00163 0 00044 3 sol Sigma 0 01301 0 00030 C o L 40 30 20 10 8 65 0 04 0 03 0 02 0 01 0 004 0 02 vil 0 03 0 04 0 05 bias voltage V Figure 5 6 Distribution of the measu
21. semiconductor The potential to completely deplete a sensor is the depletion voltage The positive and negative space charge induce an electric field in the depletion zone If an electron hole pair is created in this region the charge carriers will move in opposite directions This prevents their recombining Thus deposited charge can only be collected in this depleted zone It is therefore crucial to deplete the whole sensor Radiation can damage the crystal lattice and generate e g vacancies and interstitials These defects may introduce new energy states in the band gap and thus contribute to the effective doping of a sensor As more p doping like defects are created than n doping like the effective doping changes with increasing radiation dose For a n doped material this can lead to a type inversion As the doping concentration increases the depletion voltage increases too At high radiation doses it might be so high that the sensor cannot be fully depleted anymore This leads to a loss of efficiency The p n junction significantly increases the resistivity of the sensor and thus decreases the leakage current and shot noise to an acceptable level For diamond being an insulator no doping is needed as its intrinsic resistivity is very high Thus undoped diamond is used as sensor material A sketch of a diamond detec tor can be seen in Figure Essentially it is a drift chamber For semiconductors the 2 Principles of solid state particle det
22. the short shaping times of 0 1 us and 0 5 us the noise level is four and two 45 6 Measurement of CCD times higher than for a shaping time of 2 us respectively At a longer shaping time of 8 ps the noise level increases again Accordingly for the measurements a shaping time of 2 us is chosen Another parameter of the shaper is the internal gain with nominal settings of 10 100 and 1000 The noise levels for the different gains are measured with the same method as the noise levels of the shaping time The results are presented in Table 6 1 and the individual results in Figure A 12 and A 13 For a nominal gain of 10 the noise level is approximately 1 5 times larger than for higher gains No significant difference in the noise levels for a gain of 100 and 1000 is measured As for the highest gain the shaper saturates for large pulses an internal gain of 100 is used for the measurements The internal gain can be adjusted via a screw in the shaper and thus this is also calibrated using a function generator Again the oscilloscope computes the mean of 512 events and the voltage difference For a nominal gain of 100 an actual gain of 101 6 2 is measured This yields a systematic uncertainty of the shaper of approximately 2 The last component to calibrate is the scintillator The trigger level of the oscilloscope and the supplied voltage of the scintillator are adjusted such that without a source roughly two events per ten seconds are recorded
23. the thickness d of the sensor another equivalent definition is possible Cop d 3 2 do This is a handy definition to measure the CCD if the induced charge qo is known The aim for the CCD is to exceed the thickness of the sensor as only in this case the full charge can be measured This is important as only few charges compared to silicon are produced in the beginning So a further decrease is very disadvantageous Trapping of charges is often associated to grain boundaries so the CCD of an scCVD diamond should be longer than that of a pCVD diamond In the past only scCVD diamonds could have CCDs which exceed the thickness of the sample This is now also possible for pCVD samples but these have to be of high quality To improve the CCD several methods are available A higher bias voltage increases the charge carrier velocity and thus the CCD up to the saturation of the velocity Irradiation with a strong source can also increase the CCD This so called pumping fills the charge traps and thus reduces the trapping of signal charges It is naturally reversed over time However this detrapping is very slow for most traps and takes at least several days Heating the sample and exposure to UV light also reverse immediately the pumping So careful handling of a pumped sensor is necessary Compared to silicon diamond is very radiation hard up to sLHC doses This is due to the displacement energy which is at least twice as high as for silicon Also
24. to be influenced by pumping The difference be tween the highest capacitance of the bump and the lowest capacitance at negative voltage 39 5 C V measurement is in both cases around 3 fF and 5 fF for a voltage sweep in positive or negative direc tion respectively This bump might be caused by charge traps which have a very short detrapping time or by other effects induced by the high irradiation dose Another effect which indicates that the high irradiation dose causes uncompensated charges in the diamond is the capacitance itself The measured and theoretical values of the unirradiated diamond sample 3 match quiet well as shown before For the highly irradiated diamond sample 1 the theoretical value is 3 66 0 09 pF The larger error on the capacitance of this sample compared to sample 3 is mainly caused by the higher uncertainty on the thickness of the sample From the measurements an average value of 3 94 0 03 pF can be obtained A deviation of these two is noticeable This might be caused by using the relative permittivity e of natural diamond to calcu late the theoretical value for this sample The voltage sweep shows that this diamond sample has some uncompensated charges These charges can cause a polarisation and thus increase the value of the relative permittivity The turn on behaviour at the begin ning of the voltage sweep hints at such a polarisation In contrast this behaviour is not seen for the unirradiated sample T
25. was 1 3 5 10 and 30 B3l of 30 results in a diameter of approximately 3 5 pm This diameter is small enough for a sensor electrode as the spacing between them would be around 50 um This value can be decreased down to a diameter of 1 zm 84 The maximum length of such a tube was 680 um which is more than the expected thickness of a diamond sensor In this case the length was only limited by the thickness of the sample As one sensor needs at least a few ten thousands of those electrodes the time needed for fabrication is also very important Considering a growth rate of 30 and a sensor thickness of about 400 um one electrode takes approximately 14 s to produce As the positioning can be automated moving the sensor to the next electrode is fast Using more than one laser or splitting the beam of a high power laser can further decrease the pro duction time From these facts it can be concluded that using a femtosecond laser may be a very pow erful way to fabricate 3D diamond sensors However there are a few restrictions The pulses of the laser have to be in the range of femtoseconds as e g 300 ps pulses produce highly frayed out structures 32 Also the quoted thresholds for the graphitisation may vary between samples especially comparing scCVD and pCVD diamond As the later one has grain boundaries which can reflect parts of the laser beam this may increase the fluence locally and cause a phase transformation below the
26. 0 60 70 t h Figure 6 7 Measurement of CCD of diamond sample 1 over time at a bias voltage of 700 V Before the measurement the diamond was completely unpumped An increase of CCD and thus a pumping effect is visible Only statistical errors are shown 48 6 2 Experimental results N N CCD um N O N a Y N o A N N H tit N N N o mm N A li li li foira l reie 6 8 10 12 14 16 18 20 Figure 6 8 Measurement of CCD of diamond sample 2 over time at a bias voltage of 1000 V Before the measurement the diamond was pumped for more than 24 hours A slight decrease of the CCD during the measurement is visible Only statistical errors are shown CCD vs bias voltage E 3 L iu L x te x 20 x L x x u x 15 L L x u F Sample 1 positive u x 10 53 Sample 1 negative L A Sample 2 x Eee ge de ine Se ei er pt Gioni ae n 0 0 5 1 1 5 2 2 5 electric field V um Figure 6 9 CCD for completely pumped samples with respect to the electric field For diamond sample 1 a positive and negative bias voltage is applied to collect electrons and holes respectively Only statistical errors are shown 49 6 Measurement of CCD CCD vs time Data EE Fit wee CCD um 11 45 0 03 8 11 6 i ETA 14
27. 1 Calibration 02 6 2 2 Evolution of CCD with time 6 2 3 CCD with respect to the bias voltage 7 Graphitisation of diamond using a femtosecond laser 7 1 Experimental setupl kata eee eB N We a eni 7 2 2 Variable distance to the focal lens z eens pe 8 Conclusion amp Outlook vi A 1 Pictures of the diamond A 2 C V measurement i A 3 CCD measurement 2 222 222 2 2 A 3 1 Calculation of activity after collimator A 3 2 Shaper calibration A 3 3 CCD with respect to the bias voltage 55 59 56 56 58 58 60 61 63 Nomenclature Acronyms abbreviation m i p DUT CCD CVD scCVD pCVD meaning minimum ionising particle device under test charge collection distance chemical vapor deposition singlecrystalline diamond polycrystalline diamond vu 1 Introduction Tracking detectors are an important part of every multipurpose particle physics detector especially at hadron colliders for example the Large Hardon Collider LHC As the num ber of particles increases with the center of mass energy the requirements on detectors have increased too They have to have a high granularity which means better spatial resolution Silicon has been the material of choice for most tracking detectors for the past decades but may be at its limits Increasing particle flux causes significant damage in the sensor and limi
28. 3 2 La tT4 Tk rigen 6 time h Figure A 25 Measurement of CCD of diamond sample 2 over time at a bias voltage of 300 V Only statistical errors are shown CCD vs time Data Fit al a ty a T ei CCD um 16 6 0 03 4 4 Hp pui 2 time h Figure A 26 Measurement of CCD of diamond sample 2 over time at a bias voltage of 400 V Only statistical errors are shown CCD vs time Data CCD um p oa gt CCD um 18 26 0 04 Fit At 4 p N Ss Kt p i Bad T H w pa 6 time h Figure A 27 Measurement of CCD of diamond sample 2 over time at a bias voltage of 500 V Only statistical errors are shown 77 A Appendix CCD vs time Data FR Ao N omy pr H t AM LU Figure A 28 Measurement of CCD of diamond sample 2 over time at a bias voltage of 600 V Only statistical errors are shown 7 time h E CCD um 21 28 0 03 m E pat tI a int ye ttt T Figure A 29 Measurement of CCD of diamond sample 2 over time at a bias voltage of 700 V Only statistical errors are shown CCD vs time aE p me Hat Ati Prp i Figure A 30 Measurement of CCD of diam
29. 6 4 Voltage difference of the output pulse of the preamplifier against input pulse at the test input Note that the input pulse is negative shaping time us nominal gain noise level mV 0 1 100 32 0 0 5 100 15 6 2 100 7 5 8 100 11 9 2 10 11 7 2 1000 7 5 Table 6 1 Noise levels for different settings of the shaper Note that the levels for a gain 10 and 1000 are normalised to a gain of 100 a 400 pm thick sample only takes 4 17 ns 5 Holes are reported to have an even higher saturation velocity As the feedback capacitor takes 140 ps to discharge the ballistic deficit can be neglected For an effective noise suppression the choice of the shaping time is crucial To test the noise levels for the different shaping times diamond sample 1 is put in the fixture and a voltage of 500 V is applied The scintillator is used as an external trigger To prevent hits in the DUT the scintillator is not placed far away from the diamond To trigger it the source is placed directly above it The signals from the diamond should now be pure noise and are recorded by the oscilloscope and processed by a PC To determine the noise level the voltage is histogrammed for a period of 10 us around the trigger i e 1000 data points from each event At least 500 triggered events are used for the histogramms for each shaping time These histogramms are fitted with a Gaussian distribution The results are given in Table and the individual plots in Figure to For
30. However the height of the pulses from the scintillator are altered with time This is especially drastic in the first 24 hours after changing or turning on the supplied voltage Accordingly the scintillator is recalibrated after one day and then left unchanged for the measurements Also removing the metal box and placing it again in the plastic tube without touching or moving the scintillator can in some cases alter its performance Thus the systematic uncertainty resulting from the scintillator inefficiency is at least on the order of 5 um CCD due to the averaging on the oscilloscope It is possible to reduce this uncertainty by using a diamond sample with a high known CCD like e g diamond sample 3 and using it as a reference Unfortunately the sample was only available for a short time and then the preamplifier broke Thus this tuning was not possible 6 2 2 Evolution of CCD with time While measuring the CCD of a sample the radiation from the source pumps the diamond Thus an increase of the CCD in time is observed To ensure that the samples are com pletely unpumped when starting a measurement they are exposed to a UV light source for two minutes The experimental result for diamond sample 2 at a bias voltage of 700 V is given in Figure 6 5 For analysis the events are split into time intervals of one hour The CCD is calculated with the method described in Section An example of the calculated mean signal 46 6 2 Experimental results
31. O pe I pe ea E 0 1 2 3 4 5 6 time h Figure A 21 Measurement of CCD of diamond sample 1 over time at a bias voltage of 800 V Only statistical errors are shown 75 A Appendix CCD vs time JE Data Lai Fip H E WB AE TAE PAET D BE SE LN A TEA DE BR RE TS SES E ER SC S SAS SEE OA EN ZA O oO 1 8 time h Fit CCD um 18 63 0 45 CCD um N gt N N N ee da oa gt N o Satyr tao Figure A 22 Measurement of CCD of diamond sample 1 over time at a bias voltage of 1000 V Only statistical errors are shown CCD vs time Data E 7 CCD um 6 51 0 01 a i 8 Kuh ie Mic E E A ni So pg sE FF pt 3 4 5 3 10 12 7 Hi 16 time h Figure A 23 Measurement of CCD of diamond sample 2 over time at a bias voltage of 100 V Only statistical errors are shown CCD vs time TURE g 11 2 Fit EJ CCD um 10 87 0 04 8 n a o ue i SE o 68 A O H EA cs o N ON ES age E EE CS nn n n nni 5 time h Figure A 24 Measurement of CCD of diamond sample 2 over time at a bias voltage of 200 V Only statistical errors are shown 76 A 3 CCD measurement Data CCD vs time E 146 CCD um 14 22 0 04 a S uu kit u i ad gt 1
32. a point indicates the systematic uncertainty The error on the capacitance is the statistical error 67 A Appendix C V 500 kHz 250 mV 500 ms fe 1 03 s IH x 4 Il o i x uncompensated charge carriers cm 102 k al gt b _ si x RU IM ie te LIST iL HAT ul t Te aio T Tit H LI a Hik ste feh MIR L x l x He 1 Pi R 10 PL 1 Ta ate H E pi Ea Cel N N 1500 1000 500 0 500 1000 1500 voltage V Figure A 6 Uncompensated charges as a function of the bias voltage A change of the sign of the uncompensated charge according to the theoretical model is indicated by vertical lines in the colour of the direction of the voltage sweep 68 A 3 CCD measurement A 3 CCD measurement A 3 1 Calculation of activity after collimator To be able to calculate the efficiency of the DUT it is important to know the particle flux after the collimator The nominal activity of the source of Anom 13 8 MBq is emited from a small grain of Strontium The task is to calculate the fraction of particles which pass the collimator For the calculation it is assumed that the source is point like A sketch of the source and the collimator is given in Figure The grain does not directly face the collimator but is located in a recess at a distance of a 5 mm from it The collimator itself has a thickness of D 10 mm and the circular hole a diameter of d 1 mm T
33. atistical errors are shown the same behaviour of the capacitance with respect to the change of the voltage The difference of the capacitance compared to the measurement in the original configuration at the same electric field is caused by handling of the diamond and a resulting change of its position inside the fixture rather than by material effects There is no significant difference for the capacitance between ramping the voltage up and down below a voltage of 700 V For high positive voltages up to 1500 V the hysteresis is not zero A slight turn on behaviour is visible for a sweep in positive direction at 1500 V compared to the behaviour of the capacitance at the same field configuration in Figure This is due to a polarisation of the diamond It is lost at the beginning of each sweep due to the inability of the B1505A to hold the bias voltage between two sweeps In Figure 5 9 this 35 5 C V measurement C V 500 kHz 250 mV 500 ms 3 948 Ramping o X down Es wo A D capacitance pF 3 944 3 942 3 94 3 938 3 936 india diana 1500 1000 500 0 500 1000 1500 voltage V E 0 003 3 a 0 002 0 001 0 001 0 002 0 003 prio I li ri ao e do _rg aL si i dr lil ss i 150 1000 500 0 500 1000 1500 voltage V 0 004 Figure 5 9 Voltage sweep from 1500 V to 1500 V for the highly irradiated diamond sample 1 The sweep was repeated 122 times The uppe
34. ce on the number of extrema and the charge spec trum The resulting charge spectrum for a measurement with diamond sample 1 at 700 V is given in Figure The Gaussian distribution of the noise is visible The signal is not separated from the noise 51 6 Measurement of CCD i n rejected 7 mi maximum recorded voltage No signal maximum nimum Figure 6 12 Example of the influence of the minimum voltage difference the local extrema have to have on the recorded charge spectrum In this example the threshold is higher than AV and lower than AV The algorithm searches from left to right Thus the local maximum with a voltage difference of AV is rejected If the threshold is lower than AV the algorithm would record the local maximum and start searching for a minimum So in this case the Entries 144065 o Choise 9059 e wE Hnoise 147 r E Noise 605 4 g E Csignal 3620 n MPV 649 9 5 an E O Landau 187 4 E O Gauss 5919 i o o o o second maximum would not be recorded TTIMM TITTI UT TITTY o dla f 8000 10000 collected charge e pren li Pia MN I 2000 4000 6000 Figure 6 13 Spectrum of diamond sample 1 at a bias voltage of 700 V The fit function is given in Equation 52 6 2 Experimental results To analyse this spectrum in the first step a Gaussian distribution is fitted to the noise spectrum The fit range is restricted to negative values as this r
35. cies ranging from 1 kHz to 5 MHz To determine the impedance it uses an auto balancing bridge method Figure shows a scheme of such a bridge measurement It consists of three known impedances Z of which at least one can be varied The impedances are adjusted such that no current flows through D i e balancing the bridge If this is the case the ratio of 28 5 1 Experimental setup Figure 5 3 Measurement of impedance Z using the bridge method The three known impedances Z are adjusted such that no current flows in the detector D 35 Figure 5 4 Simplified sketch of the measurement of impedance Z using the auto balancing bridge method A virtual ground is generated at point L thus the same current flows through the unknown impedance Z and the known resistor R The impedance is calculated from the potential over the resistor R and over Z 85 the impedances Z and Z equals those of Z and Z3 Zi Zr a Zi Sig Rz 3 A The auto balancing bridge method also uses the current to determine the impedance but unlike the bridge method the current through the unknown impedance is mirrored in a known resistor and measured directly A sketch of this method is given in Figure To have the same current in the impedance and the resistor point L in Figure is kept fixed at ground potential To achieve this a detector at this point controls the not ideal amplifier so that no potential is detected From the measurement o
36. ctor Pixel Detectors for Particles and X rays C D Via et al Advances in silicon detectors for particle tracking in extreme radiation environments Nucl Inst and Meth A 509 1 3 86 2003 Proceedings of the 4th Internatonal Workshop on Radiation Imaging Detectors S I Parker C J Kenney Performance of 3 D architecture silicon sensors after intense proton irradiation IEEE Transactions on Nuclear Science 48 5 1629 2001 M Mathes Development and characterization of diamond and 8D silicon pixel de tectors with ATLAS pixel readout electronics Ph D thesis Universit ts und Landes bibliothek Bonn 2008 G Pellegrini et al Technology development of 38D detectors for high energy physics and imaging Nucl Inst and Meth A 487 1 2 19 2002 J Qian et al Partial graphitization of diamond crystals under high pressure and high temperature conditions Journal of Applied Physics 90 3 1632 2001 J Qian et al Graphitization of diamond powders of different sizes at high pressure high temperature Carbon 42 12 13 2691 2004 V R Howes The Graphitization of Diamond Proceedings of the Physical Society 80 3 648 1962 S Talapatra et al lon irradiation induced structural modifications in diamond nanoparticles Nanotechnology 17 1 305 2006 27 28 29 30 31 32 33 35 36 37 38 39 Bibliography D Saada et al Computer simulation of damage in diamond due t
37. culated by the uncertainty on the baseline and the uncertainty on the peak value 6 2 Experimental results 6 2 1 Calibration The charge sensitive preamplifier is calibrated by using a test input A negative voltage pulse is generated by a function generator and applied to the test input of the preamplifier as well as to the oscilloscope The oscilloscope triggers on this pulse and also records the output of the preamplifier It automatically computes the mean of the last 512 events The pulse heights of both pulses are measured using the build in measurement function of the oscilloscope The result is presented in Figure For a nominal amplification of 1400 mV pC the test input should be amplified by 6 5 B7 As the test input is only amplified by 6 304 0 002 see Figure this results in a total amplification of 1358 1 mV pC The resulting systematic uncertainty is less than 0 1 and can thus be neglected Another systematic uncertainty to consider is the ballistic deficit of the preamplifier With a saturation velocity for electrons of at least 9 6 10 cm s the charge collection in 44 6 2 Experimental results Gain of preamplifier test input Data Fit 3000 output voltage mV a Ss N o 1500 1000 test input gain 6 304 0 002 500 amplifier gain mV pC 1358 1 si i loi i 10 5 la ag ia la si 100 200 300 0 500 input voltage mV Figure
38. de a large chunk typically a few hundred micrometers is removed To produce a singlecrystalline scCVD diamond a diamond substrate is needed This substrate has to be a singlecrystal itself As the Carbon atoms will bond to an already evenly oriented diamond structure all seed spots have the same orientation This prevents the development of grain boundaries when the seed spots merge After production the diamond is cut from the substrate with a laser The need for a singlecrystalline substrate to produce scCVD diamonds significantly limits the size of the produced diamonds to approximately 1 cm compared to full wafer sized pCVD diamonds However this value has increased in the last years and probably will improve even further 12 3 2 Fundamental properties of diamond property diamond silicon band gap eV 5 47 1 12 breakdown field V cm 10 3 10 resistivity Q cm gt 10 2 3 10 intrinsic carrier density cm lt 10 1 5 10 mass density g cm 3 52 2 33 atomic charge 6 14 dielectric constant 5 7 11 9 displacement energy eV atom 43 13 20 energy to create e h pair eV 13 3 6 radiation length cm 12 2 9 4 avg signal created um e 36 89 avg signal created 0 1 rad length Xo e 4400 8400 Table 3 1 Comparison of properties of diamond and silicon 5 3 2 Fundamental properties of diamond So far the most commonly used sensor material in high energy physics is silicon But with the increasing
39. e CCD at high voltages have a higher statistical uncertainty The values for the CCD have an overall low statistical error of less than 1 The sys tematic uncertainty neglecting the influence of the scintillator is estimated to be below 2 um This uncertainty is due to the handling of the diamond samples and the metal box As stated above the uncertainty induced by the scintillator is at least on the order of 5 um Furthermore the values for diamond sample 2 deviate with a factor of 0 5 from values measured at SiLab University of Bonn 88 If this is also true for diamond sam ple 1 cannot be said as it was not possible to measure any CCD for this sample at SiLab Thus the setup has to be tuned with another good diamond sample for which the CCD is known for a reference measurement However with the current setup it is possible to measure very low values of CCD with a high statistical accuracy For diamond samples with a high CCD it is possible to analyse single event data The data is recorded for each event from the oscilloscope to obtain the full information For further analysis the local extrema of all events are converted into charges and histogrammed These local extrema are defined by a minimum voltage difference to the previous one An example is shown in Figure This parameter has to be chosen such that no fluctua tions between two points are seen as extrema but the whole noise spectrum is recognized Thus the value has significant influen
40. e caused by a plasma which is ignited when the laser graphitises the diamond on its surface The graphite absorbs the beam much better than the diamond and is therefore much more heated This heat might ignite the plasma Due to a slight tilt of the diamond with respect to the beam direction the plasma is not uniform around the beam So the shadow extends more to the left than to the right of the graphite spot 7 2 2 Variable distance to the focal lens To grow graphite pillars in the diamond the focal plane needs to be inside the diamond bulk However as diamond sample 1 is opaque the beam is scattered in the diamond This reduces the intensity of the beam so that the fluence threshold for graphitisation cannot be reached inside the diamond Thus for this sample only graphitisation at the surface of the diamond is possible This also happens only in the focal region as the intensity of the laser beam is otherwise to low 7 2 3 Variable power of the laser beam To determine the intensity threshold for phase transition on the surface of the diamond the sample is illuminated at different settings of the attenuator The result is shown in Figure At a power of 367 4 pW no transformation is visible The first visible graphite spot is produced at a power of 482 5 pW This corresponds to a power of 4 0 0 8 kW per pulse The large error is caused by the uncertainty on the duration of each pulse which is assumed to be at least 20 Alt
41. e statistical error for the mean value The results for different frequencies are given in Table Plots for the individual mea 32 5 2 Experimental results 13 95 C t 0 V 500 kHz 250 mV x10 Data Mean Capacitance 3 94028E 12 5 1E 16 Boop E de n n E E 0 100 200 300 400 time s Figure 5 7 Measurement of capacitance at a constant bias voltage of 0 V and a frequency of 500 kHz The error for each data point indicate the systematic uncertainty The error on the capacitance is the statistical error amplitude mV capacitance pF statistical error pF systematic error pF 32 3 940 0 004 0 009 64 3 940 0 002 0 008 125 3 940 0 001 0 008 250 3 940 5 1074 0 007 Table 5 2 Mean values of capacitance for 1001 measurements at a bias voltage of 0 V and a frequency of 500 kHz for different oscillator levels surements can be found in the Appendix Figures A 2 to The systematic errors for this and all following measurements are calculated using data from the official data sheet from Agilent The statistical error of the capacitance is for all frequencies signif icantly lower than the systematic uncertainty This shows that the measurement is very stable and high statistics is not needed for a precise determination of the capacitance However it also gives some indication that the statistical spread of the values depends on the frequency and is especially high for very high and
42. ectors Amplifier Charged Particle VISTE Electrodes Wa I Figure 2 3 Schematic illustration of a diamond detector with the first part of the electronic read out As the bias voltage is connected to the same electrode as the read out the capacitor decouples the amplifier from it 5 concept is the same but with a p n junction This junction is reverse biased to create a depletion zone for charge collection In diamond such a special zone is not needed as there are no intrinsic free charge carriers Thus it does not matter in which direction the voltage is applied which means that both types of charge carriers electrons and holes can be collected with the same sensor by just switching the sign of the electric field Due to the fact that no depleted zone has to be created by the potential the bias voltage can be the same even after irradiation without any drop in efficiency 2 3 Read out chain To read out a sensor the charges have to be collected For this an electric field has to be applied to the diamond Because of it the electrons and holes drift to the electrodes This drift induces a charge on the electrodes The instantaneous induced charge i is given by Ramo s theorem 6 in qUEw 2 2 v denotes the drift velocity of the charges q the total deposited charge and Ey the weighting field which can be derived from the geometry of the electrodes and the applied voltage To obtain the total collected charge Q t
43. eep and shows thus a hysteresis The change of capacitance is also not monotonously but a bump is visible The difference between ramping the bias voltage up and down is negligible for values higher than 700 V Flipping the diamond i e applying the high voltage on the opposite electrode shows that the hysteresis of the capacitance is caused by the diamond as other parts of the setup e g the fixture are not changed In Figure such a measurement is presented It has 34 5 2 Experimental results C V 500 kHz 250 mV 500 ms pri ao a capacitance pF w oa be kx ee 2 474 cite le TH 1 359 n wo a a oa 1 3589 1 35885 1 3588 UM VALLI ATTERSEE 1 35875 gt gt I sr r o La 400 200 0 200 600 voltage V Difference of capacitance between ramping up down IL E S 02 E i oi li Cet h Ai i N na Tun al In m Tall i C IN LK un of RIMA inant n m 1 ri HA s 1 i C A SULA ET m Na ial Ih I Il i 0 1 HM da UNI LITI I N il i Mi Hi I ii 0 2 C 1 l f 1 f f 1 l 1 1 1 I 1 f f l 600 400 200 0 200 400 voltage V Figure 5 8 Voltage sweep from 500 V to 500 V for the non irradiated diamond sample 3 The sweep was repeated 235 times The upper plot shows the measured capacitance for each step with a step size of 1 V The lower indicates the difference for the values between ramping up and down Only st
44. egion contains no signal The whole spectrum is fitted with a convolution of Gaussian and Landau distributions CNoise Gauss UNoise O Noise Csignal Landau MPV O Landau 7 Gauss MPV O signa 6 1 The parameter noise and ONoise are obtained from the noise fit and are fixed For the signal the Landau distribution is convoluted with another Gaussian distribution This is done as the Landau distribution describes the spectrum only for infinitely thin detectors The CCD can be calculated via 1 22 CCD SE IL MPV oise 36 e um HNoise 27 0 0 8 um This value is approximately 10 um larger than the value obtained from the mean values at the same bias voltage As the signal is not clearly separated from the noise spectrum this method has problems with such low CCDs The fit of the signal is very unstable with small fluctuations of the noise e g varying the threshold for the extrema Figure 6 13 is one of the few examples where the fit seems reasonable But for samples with a higher CCD this method works very well as only signal events contribute to the signal fit This eliminates the dependence on the efficiency of the scintillator as its inefficiency only leads to a higher noise peak but does not influence the Landau distribution Thus this method can be used as a cross check for the efficiency of the scintillator 53 7 Graphitisation of diamond using a femtosecond laser As described in Chapter 4 2 it is possib
45. electrodes In both cases the same amount of charge is deposited by the incoming particle as it has to pass the full sensor thickness But for the 3D electrodes these charges are generated much closer to the electrodes As the spacing between them is lower for semiconductors the depletion voltage decreases 19 two sides of the sensor and maintain a gap to the other end of the sensor approximately equal to the spacing of the electrodes Now even if a particle passes through an electrode it deposits some amount of charge in the sensor bulk Furthermore the whole modules can be mounted not perpendicularly but slightly tilted with respect to the interaction point This reduces the probability of such an event It is also possible to increase the resolution of detectors with 3D sensors as the distance in silicon between the electrodes can be as low as 50 pm in all directions 20 As this forces the read out electronics to be at this size t00 current testing with 3D sensors is done by connecting several electrodes of the same type on the surface Thus the current ATLAS Pixel read out chip the FE I3 can be used PI 4 2 Graphitisation of diamond For operation with planar electrodes a diamond sample can be metallised with e g pixel lated electrodes These metal contacts can easily be stripped off and remade as the diamond is not damaged during this process This is not as simple for 3D electrodes as they are deposited inside the diamond bulk Sev
46. eral techniques to drill holes for the electrodes into Silicon are available 22 But most of these e g etching will not work with diamond due to its unique chemical characteristics To build an electrode in diamond it is not needed to drill an actual hole in the material As diamond consists of Carbon a conversion of the diamond sp lattice to graphite sp 21 4 Bulk segmentation lattice is enough for a conducting electrode To convert the bonds several methods are available but not all capable of producing long graphite pillars in the diamond bulk Using high pressure up to 2 GPa and heating the diamond to 1300 K causes a controlled conversion of around 30 of the diamond bonds 24 So far this only works with thin diamonds of 30 40 um For thicker diamonds much higher pressures are needed Furthermore the process takes 30 minutes per spot These points rule out this method for the fabrication of 3D sensor electrodes Simply heating the whole diamond also works but this graphitisation process cannot really be controlled 25 The conversion usually starts at impurities which have to be created carefully before heating the sample But heating single atoms of the diamond is also possible by irradiating it with ions For nano sized diamonds full graphitisation can be achieved 26 Computer simulations show that a minimum transfered energy for the heated atoms is needed for the diamond to convert to graphite 28 These thresholds differ
47. erwise the charges from multiple hits would add up The time constant of this process is a crucial parameter as it has influence on the voltage output of the amplifier and thus on the measured charge It should be long enough to discharge the feedback capacitor as less as possible during charge collection This error is called ballistic deficit But the time constant also has to be short enough to prevent the pile up of several hits 2 Principles of solid state particle detectors The signal is often further processed through a filter This band pass shapes the signal and reduces the noise by attenuating low and high frequencies Thus it is crucial to know the charge collection time of the sensor and choose the passband accordingly For digitisation the next step is a discriminator As long as the signal exceeds a threshold the output is a digital one otherwise it is zero If the threshold is high enough this step further reduces the noise For adjustment it can be controlled globally as well as for each pixel cell individually The output can either be read out directly from a buffer or sampled with a clock to determine the start and end of a signal 10 3 Diamond as a sensor material for particle detectors 3 1 Production of CVD diamond Using natural diamonds for large scale particle detectors is impossible due to their shape weight and price Natural diamonds found in mines or elsewhere tend to be small and light i e less than a
48. es of diamond if used as a sensor The principle of 3D sensors and the options to manufacture them in diamond are explained in Chapter 4 Measurements of the capacitance of irradiated diamond samples with respect to the bias voltage are presented in Chapter 5 The results of the CCD measurement are given in Chapter 6 In Chapter 7 the first results of the graphitisation of diamond are 1 Introduction described 2 Principles of solid state particle detectors 2 1 Energy loss of particles Particles traversing matter interact with it and thus loose energy This energy loss happens for a heavy charged particle mainly in three ways 1 e ionisation or excitation e Cherenkov radiation e transition radiation in inhomogeneous materials For solid state detectors such as silicon or diamond sensors ionisation is the main process Ionisation in a solid state detector means the creation of electron hole pairs The amount Ne n Of pairs can be calculated via AE nen mw where AF is the total deposited energy and W the average energy needed to create one electron hole pair The energy depends on the band gap between the highest band which is filled with electrons and the lowest empty band called valence and conduction band respectively This is an intrinsic parameter of semiconductors However only for direct semiconductors the band gap equals the energy needed to create an electron hole pair Direct means that an electron can switch f
49. es the systematic uncertainty The error on the capacitance is the statistical error C t 0 V 100 kHz 250 mV x10 Data cf sla L Mean 3 95 3 xe Tl atid a Lo c S oO a 8 945 3 94 3 935 Capacitance 3 94078E 12 4 2E 16 3 93 i lu 500 time s Figure A 3 Measurement of capacitance at a constant bias voltage of 0 V and a frequency of 100 kHz The error for each data point indicates the systematic uncertainty The error on the capacitance is the statistical error 66 A 2 C V measurement x C t 0 V 1 MHz 250 mV 10 r Data Mean 93 956 S 3 954 8 952 o 3 95 TT 3 948 3 946 3 944 z Capacitance 3 942 3 94 3 94789E 12 3 0E 16 3 938 700 200 300 700 500 time s Figure A 4 Measurement of capacitance at a constant bias voltage of 0 V and a frequency of 1 MHz The error for each data point indicates the systematic uncertainty The error on the capacitance is the statistical error F gt N a C t 0 V 5 MHz 250 mV x10 Data Mean Tl i capacitance gt N N 4 21 4 2 4 19 Capacitance 4 20387E 12 4 9E 16 4 18 tu PR VA a CO ge Pg i 0 100 200 300 400 500 time s Figure A 5 Measurement of capacitance at a constant bias voltage of 0 V and a frequency of 5 MHz The error for each dat
50. f the voltage over the 29 5 C V measurement impedance U and the voltage Ur over the resistor R the impedance Z can be calculated Us i Be UR _ 1 Ur eC Inf RU The model described above is simplified as for a measurement of the impedance the vector components of the voltage needs to be known i e active and reactive voltage Thus the amplifier and the measurement of the two voltages have to be expanded with components to measure the vector components of the voltage As the expected capacitance of the diamond is very low an important thing is to com pensate the stray capacities and inductances These result from cables and the setup in general For low values of the unknown capacitance the stray capacities can have a significant influence on the measured value The correction is done via the short and open compensation For both compensation measurements everything is connected as for a measurement but the diamond is removed from the fixture During the short com pensation the electrodes are connected directly to each other i e shorting them This theoretically leads to zero impedance The measured difference to this value can be at tributed to stray capacities and is subtracted from the measurements The open compensation is just the other way round The electrodes have no contact and thus the impedance should be infinite Again if a value for the capacitance can be measured this is due to stray capacities and can be c
51. for equally thick samples the deposited energy is lower as the radiation length is about 30 longer compared to silicon This is an advantage as it reduces the material budget Due to the longer radiation length multiple scattering occurs less often For measurements of the CCD the diamond sample is irradiated by a radioactive source The resulting spectrum is then compared to the expected If it is shifted to lower charges the CCD does not exceed the sensor thickness Experimental results show that for secCVD diamonds not only the CCD is longer but also the energy resolution is better than for pCVD diamonds 5 This can be attributed to the absence of grain boundaries in scCVD diamonds as charges are easily trapped at these points Trapping is a statistical process and thus more trapping decreases the resolution 15 3 Diamond as a sensor material for particle detectors 3 4 C V measurement As the leakage current and thus the shot noise of a detector made of an intrinsic semicon ductor e g silicon would be too high to detect any particles a p n junction is used This junction consists of two differently doped semiconductors The n type semiconductor is doped with donors which means that the doped atoms have five valence electrons The additional electron is only weakly bound and has an energy level close to the conduction band For a p type semiconductor it is the other way round The doped atoms are missing an electron and thus introduce
52. ges and thus small signals precise values for the CCD are obtained For diamond sample 1 both types of charge carriers electrons and holes are collected by applying a positive or negative voltage respectively In diamond holes have a higher mo bility and thus the CCD is expected to be larger at negative voltages 5 With exception of the measurement at 1000 V this is confirmed by the data The larger statistical error at high voltages is caused by the increasing leakage current Its behaviour for the unpumped diamond sample 1 with respect to the bias voltage is illustrated in Figure 6 11 For this measurement each voltage is applied for ten minutes before measuring the leakage current 300 times This data is histogrammed and fitted with a Gaussian distribution to obtain the mean value and its statistical error 50 6 2 Experimental results Leakage current vs bias voltage 20 0 leakage current nA 20 40 60 80 5 p UA N I N GE BEE I N A EEE N N EEE EEE BEE BER BERN BERN GER 1000 500 0 500 1000 bias voltage V Figure 6 11 Leakage current as a fuction of the bias voltage for diamond sample 1 For voltages up to 500 V the leakage current increases linearly with the voltage This is the behaviour of an ohmic resistor For higher voltages especially negative the cur rent rises drastically As the shot noise is proportional to the square root of the leakage current measurements of th
53. gram Their shape is irregular and roughly spherical For a particle detector the sensor material needs to be thin in one direction and reasonably thick in the other two Because of its lower price only synthetic diamonds are an option The most rudimentary approach for manufacturing diamonds is exposing Carbon atoms to high pressure One of the first to synthesise diamonds with this method was Hall in 1954 7 8 However with sizes of a few millimeters these diamonds are too small for detector purposes Today the most common way to produce large diamond samples is chemical vapor de position CVD A gas containing Carbon often methane is ionised and the non carbon atoms removed This removal of the non carbon atoms in case of methane Hydrogen is done by free radicals i e unbound atoms such as atomic Hydrogen As Hydrogen is normally only found in a bound state a plasma is created The free Carbon atoms settle on a substrate They can move on its surface as the substrate is heated At a seeding spot the Carbon atoms will either build graphite or diamond crystals At low pressure and high temperature graphite sp bonds are energetically favoured over the diamond sp bonds However the free Hydrogen atoms will remove some Carbon from the substrate The rate of this process depends on the type of bond Graphite bound Carbon atoms are removed approximately a hundred times more frequent than diamond This means that virtually no graphite is left on
54. haper was set to a shaping time of 0 5 us and a nominal gain of 100 70 gntries 0000 25000 20000 15000 10000 5000 05 A 3 CCD measurement Entries Constant Mean Sigma 0 02 0 01 0 0 01 0 02 0 04 0 03 0 03 625000 3 324e 04 0 001793 0 007484 0 04 0 05 signal V Figure A 10 Noise after the shaper with diamond sample 1 connected and a bias voltage of 500 V The shaper was set to a shaping time of 2 us and a nominal gain of 100 Entries 503000 R Constant 1 683e 04 2 C Mean 0 00177 6000 Sigma 0 01189 14000 12000 10000 8000 6000 4000 2000 E a AR A UO A E 1 8605004 0 03 0 02 0 01 0 0 01 0 02 0 03 0 04 0 05 signal V Figure A 11 Noise after the shaper with diamond sample 1 connected and a bias voltage of 500 V The shaper was set to a shaping time of 8 us and a nominal gain of 100 71 A Appendix Entries 502000 Constant 1 708e 04 Mean 0 0004828 Sigma 0 001172 Pei tbe ee re ee a P005 0 004 0 003 0 002 0 001 0 0 001 0 002 0 003 0 004 0 005 signal V Figure A 12 Noise after the shaper with diamond sample 1 connected and a bias voltage of 500 V The shaper was set to a shaping time of 2 us and a nominal gain of 10 Entries 521000 z Constant 2 772e 04 2 C Mean 0 01856 Sigma 0 07469 25000 20000 15000
55. his equation has to be integrated until 2 3 Read out chain Filter Discriminator Bump Bonding Pad Figure 2 4 Scheme of the electronics of a typical hybrid pixel cell The sensor is represented by C pet charge collection is complete Q fia In electronic circuits this is done by the current charging a capacitor the so called feedback capacitor It is connected parallel to a charge sensitive amplifier The potential difference on its in and output is amplified This resulting voltage can be further progressed A scheme of such a read out chain is illustrated in Figure This is the typical design for a pixel cell of a hybrid pixel detector Hybrid means that the sensor and the read out electronics are on two different chips The sensor and the read out chip are connected via a bump bond for each pixel cell The advantage of this approach is the ability to design the sensor and electronics independently and to connect each pixel cell to individ ual read out electronics This and the low manufacturing costs due to separated sensor and electronic chips make this design the common choice for pixel detectors at hadron colliders The collected charge from the detector is gathered in the feedback capacitor Crg The resulting voltage difference at a parallel connected charge sensitive amplifier is amplified To discharge the feedback capacitor either a resistor or a current source is connected in parallel This discharging is needed as oth
56. hough only very small spots of graphite are visible this is more than twice the power at which graphite pillars can be 58 7 2 Experimental results Figure 7 5 Graphite spots for different powers of the laser beam From left to right 545 5 pW 482 5 pW slightly visible 367 4 pW no transformation visible and 584 6 pW All spots were produced with an exposure time of 10 s except the last spot which was exposed 20 s to the beam grown in monocrystalline diamond which is 210 pW 33 The intensity for each pulse the fluence at this setting of the attenuator is 19 1 This value is more than an order of magnitude higher than the upper threshold for continuous growth of graphite which is 1 2 0 2 32 Thus phase transition from diamond to graphite is expected at much lower power There are several explanations for the observed higher threshold On the one hand all results from publications were produced using monocrystalline and not polycristalline di amond like sample 1 However due to grain boundaries in polycrystalline diamond the threshold for graphitisation is expected to be lower On the other hand the monocrys talline diamonds were of optical quality which may support graphitisation at lower powers The measured power threshold does not differ that much from the expectation compared to the measured fluence This may indicate that the focused beam might be wider than the graphite spots indicate
57. hus the angle a can be calculated via d 2 D a a 2arctan To determine the ratio of particles which are emitted within the opening angle of a with respect to the full solid angle of 47 the surface of the spherical cap Asur has to be known It can be calculated via Asur 277 1 COS A 2 r is the radius of the sphere With the surface of the spherical cap its solid angle Q is determined by Sr90 source Figure A 7 Sketch of the source facing the collimator The opening angle is indicated with a 69 A Appendix Thus the effective activity Aes after the collimator is Anori d 2 Aeff 5 1 COS arctan D i 3 83 kBq A 3 2 Shaper calibration Entries 501000 R Constant 5 553e 04 2 Mean 0 01134 5o00 Sigma 0 03198 50000 40000 30000 20000 10000 0 3 signal V Figure A 8 Noise after the shaper with diamond sample 1 connected and a bias voltage of 500 V The shaper was set to a shaping time of 0 1 us and a nominal gain of 100 In this case two different sources of noise are visible The dominating source has a much lower noise level than the suppressed source Entries 509000 Constant 1 291e 04 94000 Mean 0 001852 C Sigma 0 01564 412000 10000 s000 sooo 4000 2000 85 a pg oe oe ee oe signal V Figure A 9 Noise after the shaper with diamond sample 1 connected and a bias voltage of 500 V The s
58. hus the irradiation might have changed the relative permittivity of diamond sample 1 40 6 Measurement of CCD For the measurement of the CCD a new setup was built As the deposited charge in diamond is small compared to Silicon a very low noise level of this setup is required to obtain a signal 6 1 Experimental setup The setup to measure the CCD of the diamond samples is shown in Figure a and a schematic cut through in Figure 6 2 The samples are put into a fixture inside a metal box This box works as a Faraday cage to reduce the noise of the setup The fixture is a Copper clamp which holds the DUT A photograph of this clamp is given in Figure 6 1 b During the measurements the box is placed under a plastic tube This tube holds the radioactive source and a collimator The tube ensures that the source is in the same position with respect to the DUT A plastic scintillator is located underneath the box It is important to reduce the radiation length of this setup to ensure that as many particles as possible traverse the diamond and reach the scintillator Thus the metal box has two holes above and below the clamp which are covered with thin Aluminium foil Additionally a hole is drilled through the Copper clamp as the Copper would otherwise shield the diamond from the radiation A Sr90 source with an activity of 13 8 MBq is used The plastic collimator ensures that every particle which hits the scintillator also had to geometrically
59. imation of this the size of a typical fully graphitised area like in Figure is taken It has a rectangular shape with a short edge of 23 2 um and a long edge of 113 4 pm This yields a 56 7 2 Experimental results Power of the laser beam 5 cm for different attenuator settings power uW oa o 400 300 200 FH x SI Tr Br pIE EEE ERETI je g i a E n n 2400 2600 2800 3000 3200 3400 attenuator au N N Figure 7 3 Power of the laser beam for different attenuator settings at the focal point The power is averaged over one second Shadow Figure 7 4 Typical graphite spot after irradiation with the laser beam The graphitised area has a rectangular shape and a size of 2 6 0 2 10 cm The shadow left of the spot is most probably due to plasma ignited at the diamond surface 57 7 Graphitisation of diamond using a femtosecond laser Attenuator Setting au Power uW Intensity W cm 2300 156 2 6 0 0 6 2550 367 4 14 1 1 3 2800 482 5 18 5 18 3050 845 5 21 0 220 3300 584 6 22 0 2 1 Table 7 1 Power and intensity of the laser beam for different settings of the attenuator The measurements were taken in the focal plane spot size of 2 6 0 2 107 cm The resulting intensity for each attenuator setting is given in Table In Figure 7 4 next to the graphitised area a slight shadow is visible on the surface of the diamond This may b
60. in the two publications from 416 eV resulting in a fraction of 62 sp bonds found by Saada et al and 2 58 eV causing 34 5 sp bonds found by Sorkin et al The difference can be explained with the fact that in the first publication twelve atoms are heated to this energy but one at a time whereas in the second four layers with a total of 16 atoms are heated at once The minimum percentage of graphit bound Carbon atoms for a conducting layer is determined by Sorkin et al to be 45 These results clearly show the limitation of this process It is not enough to excite single Carbon atoms to produce graphite bonds Instead several at close range have to be hit by the ions Heavy ions can transfer these energies in single collisions but due to their high stopping power they do not penetrate deeply into the diamond Even assuming a typical proton from a tandem accelerator with 6 MeV penetrates the diamond only ap proximately 200 um deep This means that the diamond has to be flipped and irradiated from both sides which can cause problems But such protons may not pass enough energy per collision to cause sp bonds Thus irradiation with ions is not really suitable for the formation of graphite pillars in the diamond bulk However the transformation works at the surface using heavy ions BI All methods described above heat the Carbon atoms in some way Another approach is to excite the electrons from the valence into the conduction band with a femtosec
61. le to induce a phase transformation from diamond to graphite using a femtosecond laser This method was tested with diamond sample 1 As it is opaque only graphitisation of the surface is possible The experimental setup is described in Section and the results in Section 7 1 Experimental setup A sketch of the experimental setup is shown in Figure The laser has a wavelength of 800 fm and is pulsed with a frequency of 1 kHz Its nominal pulse duration is 30 fs However after passing the final lens a pulse duration of 120 ns due to dispersion is expected The laser has a nominal power of 3 W which results in a power of 25 MW per pulse It is splitted several times so that not the full power of the beam is available To lower the intensity of the beam an attenuator is used which can be controlled via a PC To further reduce the intensity the beam passes through a 2 plate and a polariser These are only used for weakening the beam and not for their primary use The beam is focused with a lens with a focal length of 5 cm The diamond can be moved in one A 2 Lens DUT Laser Attenuator Polariser Figure 7 1 Schematic setup for the graphitisation of diamond The attenuator is controlled via a PC as well as the horizontal position of the diamond The focal length of the lens is 5 cm 59 7 Graphitisation of diamond using a femtosecond laser Power profile for 2300 160 158 power uW 156 154 152 150 148
62. licon and diamond is rather long see Table 3 1 their stopping power is very weak Especially compared to those of e g Iron with a radiation length of 1 76 cm 3 Thus semiconductors are used to build tracking detectors and not calorime ters The purpose of a tracking detector is not to measure the full energy of a particle but to determine its trajectory through a magnetic field With this information the ratio of charge to momentum of the particles can be calculated Electrons of an atom have discrete energy levels In a crystal lattice these energy levels cannot have the same value for each atom due to the Pauli exclusion principle Thus the energy levels for each atom shift to slightly lower or higher values compared to a single atom This creates the so called energy bands in a solid state body The energy difference between the highest filled and the lowest unfilled band is called band gap If there is no band gap the solid state body is a conductor Semiconductors have values below 4 eV Materials with higher band gaps are insulators Silicon has a band gap of 1 12 eV and is thus a semiconductor Due to the small band gap at room temperature single electrons in the valence band have enough energy to cross the gap This creates an electron in the conduction band and a hole in the valence band Both of them can be used as charge carriers Thus the resistivity for intrinsic Silicon is 2 2 Sensor design rather low with 2 3 10 Qcm If a
63. lthough the diamond sample is hit by a particle it only reduces the rate but does not influence the measured value It does influence the measurement if the diamond sample has no signal from a particle and the scintillator triggers In this case a pure noise hit is included in the average and thus decreases the mean signal To decrease the statistic uncertainty these events are separated into time intervals For each of these intervals the mean signal is calculated Each event has a resolution of 1000 data points which correspond to a time relative to the trigger i e 5 us before and after the trigger For every data point all values from the events within one interval are histogrammed and fitted with a Gaussian distribution This yields the mean signal from these events and its statistical uncertainty An example is shown in Figure To calculate the height of this signal the non signal region i e between 5 us and 1 ps before the trigger is fitted for each interval with a constant to determine the baseline level As the signal is a negative voltage pulse this value is subtracted from the lowest measured voltage in the signal region i e between 2 us and 7 us after the trigger From this voltage the collected charge is calculated This value is divided by the average number of charges created by a traversing m i p per micrometer i e 36 electrons 1m see Table B 1 to obtain the CCD of the sample The statistical uncertainty of the CCD is cal
64. mall a good shielding of the whole setup is needed to reduce the electro magnetic pick up and thus the overall noise level of the system Thus the metal box acts as a Faraday cage but also protects the diamond from light sources This is important as UV light unpumps the diamond 5 1 2 The B1505A Semiconductor Device Analyzer Unlike e g current or voltage the capacitance of a device in an electric circuit cannot be measured directly But from the impedance Z of the device the capacitance C can be calculated using i aa on fC 1 eC FZ 5 1 The frequency f has to be chosen so that it affects the measurement as little as pos sible For high frequencies the inductance of the cables can become dominant for low 27 5 C V measurement Figure 5 1 Photograph of the B1505A On top of it lies the metal box including the fixture for the diamonds and a bias tee to connect it with the B1505A Figure 5 2 Photograph of the open fixture with a diamond sample inserted The upper electrode of the diamond is connected via a spring the lower via a round rivet frequencies the measured value of the unknown impedance may be too high for an accu rate measurement Therefore too high or too low frequencies might change the value for the capacitance Thus to measure the capacitance one needs to measure the impedance at a specific frequency The B1505A can measure the impedance of a DUT up to a bias voltage of 3 kV and with frequen
65. mentum Figure 2 1 Energy loss of muons in Copper 3 S 0 25 E 0 18 0 161 z 0 141 0 12 E 0 1 0 081 0 06 0 04 0 02 pu ae fee ere L fitta rr 00 200 300 400 500 600 700 AE d eV um Figure 2 2 Example for the Landau distribution L A of the energy loss of a m i p in silicon with thickness d The energy loss has been normalised to the sensor thickness d The mean value and most probable values are indicated 4 2 Principles of solid state particle detectors Xo is a material specific parameter called radiation length It describes the mean distance at which an electron has only 1 e of its original energy left Photons interact with matter in different ways These are photo effect e Compton effect e pair production The cross section of each process depends on the energy of the photon and on the material For a typical high energy photon in tracking detectors only pair production is important In this process the photon transforms into an electron positron pair This means that the whole energy of the photon is deposited in this process and thus in the detector at once Uncharged massive particles cannot interact with matter through the electromagnetic force due to their missing charge Their only interaction process is scattering In this process electrons or atoms of the sensor material may be excited and their relaxation can be measured 2 2 Sensor design As the radiation length of Si
66. mes Only statistical errors are shown Clow is the mean of ten consecutive data points and Chig is the mean of the next ten data points Viow and Vhign are the corresponding mean voltages The uncertainty on the two mean capacities is quiet high as they cover each a voltage interval of 30 V Thus the errors on the uncompensated charges which are mostly caused by the derivative are high too Also the uncertainty on the size of the electrodes contributes According to Equation 3 9 the bump results from a sudden change in the sign of the uncompensated charges which seems unphysical However the model was developed for semiconductors with a depleted region so that it does not describe the situation in diamond properly The number of uncompensated charges does not vary a lot especially in the region above 700 V Here between 2 102 charges cm and 7 10 charges cm are measured This value is quite high considering that diamond should be an insulator and thus have no 37 5 C V measurement C V 500 kHz 250 mV 500 ms 5 Ramping EL ir a up 23 2 5 10 E k o JE D C L s I _ o Toh A 4 3 1022 E H _ B A 5 E L nia a get a 2 E u _ ing a ea 2 ie m Hy getti A E Fr My la er 8 LA i 2 yon ak ate E H Tapt u E Hn La athe Jd en ee ons fe er on ee E 1500 1000 500 0 500 1000 1500 voltage V a Voltage sweep in positive direction
67. nominal threshold The transformation also does not depend only on the fluence but also slightly on the 24 4 2 Graphitisation of diamond number of laser pulses This dependence may be due to stable nano sized defects caused at lower fluence which only grow slowly with each pulse Cracking of the diamond due to the graphite tubes seems not to be a problem Although diamond is more than twice as dense as graphite layers with graphite pillars of 1 um diameter with a gap of only 1 pm have been produced without damaging the diamond bulk 34 Small cracks directly at the electrodes might also act as charge traps 25 5 C V measurement 5 1 Experimental setup An Agilent B1505A Semiconductor Device Analyzer is used to measure the capacitance of the diamond samples The diamond is put in a metal box which acts as a Faraday cage to reduce electro magnetic pick up and protects the sample from light A picture of the whole setup can be seen in Figure 5 1 1 The diamond fixture Figure shows a picture of the open fixture used to hold the diamond sample The upper electrode is connected to the diamond via a spring At the end the spring has a small bump such that a contact to the diamond electrode is established Below the spring is a copper plate In this plate a round rivet is embedded to contact the lower electrode The contacts are mounted on a plastic block to insulate them from the metal box As the amount of measured charges is very s
68. o ion impact and its annealing Phys Rev B 59 10 6650 1999 A Sorkin et al Computer simulations of damage due to passage of a heavy fast ion through diamond Phys Rev B 70 6 064110 2004 S Prawer R Kalish on beam induced transformation of diamond Phys Rev B 51 22 15711 1995 B Miller et al Patterned Electrical Conductance and Electrode Formation in lon Implanted Diamond Films Journal of The Electrochemical Society 141 4 L41 1994 J Krauser et al Jon track lithography and graphitic nanowires in diamondlike carbon J Vac Sci Technol 26 6 2468 2008 T Kononenko et al Microstructuring of diamond bulk by IR femtosecond laser pulses Applied Physics A Materials Science amp Processing 90 645 2008 M Neff et al Femtosecond laser writing of buried graphitic structures in bulk dia mond Applied Physics A Materials Science amp Processing 97 543 2009 T Kononenko et al Three dimensional laser writing in diamond bulk Diamond and Related Materials 20 2 264 2011 K Okada T Sekino Agilent Technologies Impedance Measurement Handbook July 2006 July Agilent Technologies 2006 Agilent Technologies Agilent B1500A Semiconductor Device Analyzer User s Guide 7 edition 2008 Fast ComTech GmbH CSP10 User Manual 2009 J W Tsung private communication I N Bronstein et al Handbuch der Mathematik Verlag Harri Deutsch 2008 83 Danksagung Ich m chte Prof Dr Arn
69. ompensated 5 1 3 Diamond samples Sample 1 Highly irradiated diamond This sample is a pCVD diamond which was highly irradiated Such a high dose alters the appearance of the diamond as it has lost its optical quality and is opaque A picture of it is shown in Figure a b Because this diamond sample is very old it has a smaller grain size and thus more grain boundaries than newer pCVD samples as these profit from improved production processes The seed and growth side can be clearly distinguished which shows that the diamond was not polished Its surface is quadratic and measures approximately 1 cm It has a thickness of 400 10 um The electrodes are made of gold and have a circular shape and a diameter of 6 079 0 001 mm 30 5 2 Experimental results a Diamond sample 1 b Diamond sample 1 c Diamond sample 2 growth side substrate side Figure 5 5 Pictures of the diamond samples Sample 2 Irradiated diamond This sample was irradiated with 3 101 25 MeV protons per square centimeter Never theless this pCVD diamond has still optical quality as can be seen in Figure 5 5 c The electrodes are separated on both sides in two parts For measurements only the bigger electrode was used Also the electrodes on both sides are not made from the same ma terial The side used for collecting the charges and applying the high voltage is made of gold and on the other side of TiW The diamond has a thickness of 520 5
70. ond laser 34 Single very hot i e high energetic electrons can cause an avalanche ionisation If the density of the excited electrons is large enough a phase transition from diamond to graphite is induced Although the band gap of diamond of 5 47 eV is too large to excite an electron with a single photon from an 800 nm laser other effects occur with 22 4 2 Graphitisation of diamond plane Discrete structure Continuous structure Laser beam Figure 4 3 Image of graphite impurities in diamond produced by multiple shots using a 120 fs laser The separation into two types of structures is visible 32 femtosecond lasers These effects are nonlinear photoionisation and multiphoton ionisa tion and are caused by the high electric field of the laser Depending on the frequency and intensity of the laser one or the other effect dominates For an 800 nm laser both effects are shown to contribute equally 32 In Figure 4 3 the graphitisation of diamond using an 800 nm laser with pulses of 120 fs and a total deposited energy of 320 nJ can be seen The damaged region lies upstream of the focal plane It can be seen that the transformation is separated into two regions One with discrete structures and one with continuous growth This shows that the fluence F has significant impact on the structure of the graphitisation At the boarder of these two regions the fluence is calculated to be Fj 1 2 0 2 A fluence higher than this thre
71. ond sample 2 over time at a bias voltage of 800 V Only statistical errors are shown 78 CCD vs time A 3 CCD measurement CCD um 23 57 0 06 Data Fit de N o gt IPT RES RES RZSE RENE E 22 N N Ir N d om o a are 3 5 4 time h Figure A 31 Measurement of CCD of diamond sample 2 over time at a bias voltage of 900 V Only statistical errors are shown Ra CCD um 24 59 0 02 m ni a ha Mi ii Tl Figure A 32 Measurement of CCD of diamond sample 2 over time at a bias voltage of 1000 V Only statistical errors are shown 79 Bibliography 1 2 13 K Kleinknecht Detektoren f r Teilchenstrahlung Teubner 2005 W R Leo Techniques for Nuclear and Particle Physics Experiments A How to Approach Springer Verlag 1994 K Nakamura et al The Review of Particle Physics J Phys G 37 075021 2010 J Gro e Knetter Verter Measurment at a Hadron Collider The ATLAS Pixel Detector Habilitation Bonn University 2008 H Pernegger High mobility diamonds and particle detectors Physica Status Solidi a 203 13 3299 2006 S Ramo Currents Induced by Electron Motion Proceedings of the IRE 27 9 584 1939 F P Bundy et al Man Made Diamonds Nature 176 4471 51 1955 H T Hall Ultrahigh Pressure Research Science 128 3322
72. ond the whole sensor is free of space charge even if no external voltage is applied So what is expected in this case is that the capacitance does not change at all with respect to the applied voltage But through a high irradiation dose traps are induced in the diamond If these traps are 17 3 Diamond as a sensor material for particle detectors ionised they act like doping atoms The behaviour of the trapping centers with respect to the bias voltage depends on the energy level of the respective trap Different traps can be neutralised or ionised depending on their distance from the valence and conduction band Therefore the occurance of traps is noticeable by a change of capacitance with respect to the bias voltage although it should be very small For semiconductor sensors it is known that high irradiation doses change the effective doping concentration and even invert the type of doping 15 So the C V measurement may be a method to quantify the damage of the irradiation to the diamond However as the diamond has no p n junction and thus no clear majority charge carrier this description is not perfect 18 4 Bulk segmentation 4 1 Motivation for new types of electrodes The efficiency of a diamond sensor depends on the charge collection distance CCD and thus also on the distance between the electrodes As mentioned in Chapter 3 3 the CCD for good samples can be larger than their thickness But due to the statistical behaviour
73. pe will occur less often As this is the main process inflicting damage in the sensor material of tracking detectors it is a major advantage of diamond over silicon Thus diamond will probably not fully replace silicon tracking detectors But the low material budget of a full detector its radiation hardness and low noise level are good reasons to build at least the inner most layer of a tracking detector for a hadron collider of diamond sensors 3 3 Charge collection distance One of the main problems of diamond sensors is charge loss through trapping This means that charges induced by a particle do not propagate through the whole sensor to the electrodes but are trapped on the way However due to Ramo s theorem see Equation a current is still induced on the electrodes As stated before to measure the full charge the carriers have to reach the electrodes In case they are trapped before only a fraction qm of the full charge gu is measured The ratio of these two values is an intrinsic parameter for each diamond and has to be known to interpret the signals correctly A derived figure of merit is the charge collection distance CCD which equals 14 3 3 Charge collection distance the mean distance electrons and holes drift apart before being trapped It is defined as CCD per unt E 3 1 He denotes the mobility of electrons and holes Tep their lifetime and E the applied electric field Assuming that the CCD does not exceed
74. perimental results C V 500 kHz 250 mV 500 ms T amp gt 3 936 2 Ramping up i 3 935 X down g 2 3 934 i 3 933 3 932 3 931 3 93 3 929 3 928 A I UN N I DU N DD DD DV DD O N DD E 1500 1000 500 0 500 1000 1500 voltage V E 0 002 lt 0 001 o TITTI TITTI UTT 0 001 0 002 0 003 p Mer gie Sr ALI DO O a VE O CON LO FO CI I O PO ai 1500 1000 500 0 500 1000 1500 voltage V Figure 5 12 Measurement with the same parameters as in Figure p 9 but with pumped diamond The sweep was repeated 185 times Only statistical errors are shown was pumped for six hours with a Sr90 source with an activity of 13 8 MBq to neutralise and thus decrease the number of charge traps The difference between the measured capacitance at the highest and the lowest voltage of the sweep decreases compared to the unpumped measurement With a pumped diamond this difference is 7 54 0 15 fF compared to 10 26 0 20 fF with an unpumped sample This indicates that the change of capacitance with respect to the voltage is indeed influenced by the charge traps Some of them are filled for a pumped sample and thus the number of active charge traps is reduced The hysteresis itself has a smaller amplitude with 5 fF for the pumped measure ment compared to 6 fF for the unpumped sample This further hints that the hysteresis is caused by the charge traps The amplitude of the bump seems not
75. potential is applied this leads to a significant leakage current A high leakage current is disfavoured as the shot noise of a sensor is proportional to the square root of it For intrinsic silicon this noise is easily in the order of the detected signal Thus for semiconductors p n junctions are used which consist of two differently doped semiconductors Doping means the inclusion of foreign atoms in the crystal lattice of an intrinsic semiconductor These atoms either have an electron more or less on the outermost orbit than the semiconductor They are called donators or acceptors and the materials n or p doped respectively Donators introduce new energy states with slightly lower energies than the conduction band in the band gap Acceptors also generate new energy states but with energies slightly higher than the valence band In thermal equilib rium at room temperature these states are completely ionised For a p n junction these two doped semiconductors are merged At the border of these materials the positive and negative charges cancel out This results in an effective space charge in this region It is either positive or negative for p or n doped material respec tively This region is called depletion zone The thickness of this zone can be expanded by reverse biasing the junction Typically the doping concentration of the two materials differs by orders of magnitude so that the depletion region expands nearly exclusively into the lower doped
76. r plot shows the measured capacitance for each step with a step size of 3 V The lower indicates the difference for the values between ramping up and down Only statistical errors are shown could not be seen as it is only a small effect and the capacitance changes more drastically at the begining of the voltage sweep According to Equation a change of capacitance with respect to the voltage yields information about the number of uncompensated charge carriers in the diamond This is illustrated in Figure for a sweep in positive direction and in negative direction A combined plot is given in the Appendix in Figure For this calculation the mea surement with the non flipped diamond was used see Figure The derivative of the capacitance with respect to the voltage was calculated using dC zu Chigh Ciow dV u Vhigh Viow l 36 5 2 Experimental results C V 500 kHz 250 mV 500 ms 3 93 Ramping X down 3 928 capacitance pF 3 926 3 924 3 922 3 92 EN RENE RENE ERE RARE BARS A hr N DD A 1500 1000 500 0 500 1000 1500 voltage V 3 918 T 2 Q 0 003 0 002 0 001 0 0 001 0 002 TTT rrr 0 003 i lu s 1 SENE EE E O O 1 1500 1000 500 0 500 1000 1 voltage V Figure 5 10 Measurement with the same parameters as in Figure 5 9 but with flipped diamond This means that the high voltage is applied on the other electrode The sweep was repeated 139 ti
77. raps for the reduced particle flow with collimator This increases the effective number of traps in the diamond and thus decreases the CCD 6 2 3 CCD with respect to the bias voltage As the velocity of the charge carriers increases with the electric field a higher field should yield a larger CCD see Equation B 1 This increase is not unlimited as the velocity saturates at high fields This can be seen in Figure The values for the CCD are obtained from fully pumped samples To measure these values both samples are set to a bias voltage of 1000 V and pumped 47 6 Measurement of CCD Mean signal around trigger 0 002 gnal SI 0 001 0 001 0 002 0 003 Signal Trigger pP na e n Events 1078 Peak over ground V 0 004752 Collected charge e 303 2 CCD um 8 422 po Po ie RR At s Figure 6 6 Mean values of all 1078 events from the first hour of the measurement in Figure Note that each event already contains the mean values of 128 single events For each point of the curve the corresponding data of the 1078 events are histogrammed and fitted with a Gaussian distribution The mean signal of the trigger is printed in blue CCD vs time E 18 m a PE hp E E ph F 3 17 u TE E papt BEE F 16 Apr E agg m PE 156 Lau E at Sa 147 wa 13 E re 12 lE 1 H TE A EN TE BEN KT EE RE RT je Al Sig SEE TS ST ES DUO ge ig te 0 10 20 40 5
78. red bias voltage during one capacitance measure ment at 0 V The uncertainty of the voltage supply exceeds the uncertainty induced by a voltage sweep from 0 V to 1 mV frequency kHz capacitance pF statistical error pF systematic error pF 1 3 691 0 004 0 244 100 3 941 4 107 0 008 500 3 940 5 10 0 007 1000 3 948 3 10 0 007 5000 4 204 5 10 0 021 Table 5 1 Mean values of capacitance for 1001 measurements at a bias voltage of 0 V and an oscillator level of 250 mV for different frequencies voltages during one sweep are histogrammed The distribution has a Gaussian shape and the mean value is 1 6 0 4 mV The shape indicates that the deviation of the bias voltage is not due to the sweep but rather to the intrinsic uncertainty of the high voltage supply This explains why the mean value is higher than the nominal start and end voltage of the voltage sweep The sigma of 13 0 0 3 mV is more than ten times higher than the amplitude of the voltage sweep So it is fair to assume that the influence of the sweep on the measurement is negligible An example of such a measurement at constant bias voltage is given in Figure This and all following measurements for calibration are performed using the highly irradiated diamond sample 1 As the bias voltage is set to 0 V the capacitance is stable in time The mean capacitance is calculated by histogramming all measured data values and fitting a Gaussian distribution Its sigma gives th
79. requirements for new detectors silicon may reach its limits Especially its radiation hardness is considered critical for applications like the sLHC Diamond on the other hand has proven to be very radiation hard 10 Despite being an electrical insulator it can be used for solid state detectors A comparison of the properties of diamond and silicon is summarised in Table Diamond being an insulator has a very high resistivity This significantly decreases the leakage current compared to Silicon As the shot noise is proportional to the square root of the leakage current it is very low in diamond sensors The band gap in diamond is roughly five times larger than in silicon This increases the energy needed to create an electron hole pair in diamond to 13 eV compared to 3 6 eV in silicon So the signals in diamond are smaller for the same deposited energy But the larger band gap is also a strength of diamond as it decreases thermal noise significantly It is possible to run diamond detectors with little to no cooling without encountering problems associated with noise Also due to the high thermal conductivity cooling is very effective Due to its lower dielectric constant the capacitance of diamond is lower than for silicon So the capacitive load on the read out electronics and thus the noise level is lower as well Another positive aspect of diamond is the high mobility and saturation velocity of electrons and holes Thus the response time of the
80. riable duration of exposure The initial formation of graphite depends entirely on the fluence of each single shot As the transformation happens on the timescale of one shot it is not possible for effects from consecutive shots to pile up The number of shots and thus the duration of exposure to the beam can only control the size of the graphitised volume This dependence is illustrated in Figure It is clearly visible that with increasing duration of exposure of the diamond the graphite spots grow larger at a power of 482 5 pW While after 20 s no graphite can be spotted the graphite grains grow larger for 40 s and 60 s This slow growth indicates that at this power level the threshold for phase transition at the surface is barely reached Furthermore this measurement hints that the threshold might even be a little higher As explained above due to reflections a higher local fluence can create nano sized graphite spots at grain boundaries The spot with 20 s of exposure is located on a relative big diamond grain such that no nano sized graphite spot was created This would explain why in this case no graphite is visible but in Figure 7 5 for 10 s some graphite is produced As expected for lower laser power no phase transition even after an exposure time of more than 2 min is observed Going to higher powers significantly speeds up the formation of graphite In Figure different durations of exposure for a power of 584 6 pW are shown No
81. rom the valence band to the conduction band without having to change its momentum If the band gap is indirect like in silicon and diamond the highest possible state in the valance band and the lowest possible state in the conduction band require different momenta A change of the momentum of the electron increases the energy needed to create an electron hole pair The average of the deposited energy AF can be calculated using the Bethe Bloch for mula 2 It describes the mean energy loss of heavy charged particles i e all charged 2 Principles of solid state particle detectors velocity of the incident particle speed of light u c a classical electron radius electron mass Avogadro s number mean excitation potential atomic number of the absorbing material atomic weight of the absorbing material density of the absorbing material charge of the incident particle max maximum energy transfer in a single collision sito Er Table 2 1 Explanation of symbols used in Equation particles except electrons and high energy muons For these Bremsstrahlung dominates dE Z2 2m VW maz _ de Na ma i ee 29 2 1 The symbols used in this equation are explained in Table 2 1 It is possible to express this formula as a function of the relativistic parameters 8y An example is given in Figure 2 1 for the energy loss of muons in Copper It shows that the Bethe Bloch formula cannot describe the energy loss for every Gy of the muon At
82. shold yields a discrete structure lower fluence results in a continuous graphitisation If the fluence is high F gt Fj several distinct graphitised spots are created These grow so fast that the one furthest away from the focal plane i e the first one in the beam absorbs the beam Thus the other seeds stop growing For a lower fluence the spots grow significantly slower so that a continuous structure is generated This structure will grow until the fluence drops below a threshold of Fj 0 35 0 05 Below this thresh old no growth is observed at all If the fluence is always lower than F an exclusively continuous growth is realised As the structure stops growing when F lt F2 the focal plane has to be moved through the thickness of the sample to continue the growth of the graphite pillars The continuous region is very interesting for diamond sensors as this may provide a tech nique to grow graphite electrodes in the sensor bulk Important for this are the length and diameter of such pillars The length of the continuous growth region can be extended by just moving the sample away from the focal plane of the laser Depending on the speed of this movement the diameter can also be controlled as illustrated in Figure 4 4 A speed 23 4 Bulk segmentation Figure 4 4 Graphite tubes in diamond produced with an 800 nm laser which is pulsed with 140 fs at a repetition rate of 1 kHz From top to bottom the speed of the moving sample
83. significant difference is visible between these spots It can be concluded that at this power level the full graphitisation of the focal area takes less than a second 60 7 2 Experimental results Figure 7 7 Graphite spots for different times of exposure to the laser beam From left to right 584 6 uW for 10 s 1 s 2s and 5s 7 2 5 Depth of graphite spots For three dimensional electrodes the graphite has to be produced within the diamond bulk So an interesting question is how deep the graphite grows even when the focus is on the surface of the diamond sample To investigate this the laser is focused at the edge of the test sample The resulting graphite spots on the surface at a power of 584 6 pW are shown in Figure a As only a part of the laser spot is on the diamond when looking from the side the depth of the graphite spot can be measured This is illustrated in Figure b The graphite is nearly exclusively produced on the surface Its depth is below the resolution of the microscope This result is expected as the diamond is optically dense at the wavelength of the laser and thus even if the focus is inside the diamond bulk graphitisation can only happen on the surface Additionally once graphite is produced on the surface it absorbs the laser further reducing the intensity in the diamond bulk 61 7 Graphitisation of diamond using a femtosecond laser a Look on the surface ot the substrate side b
84. sited charge and thus the signal decreases and the sensor is more vulnerable to mechanical stress An alternative to this design are 3D sensors 13 In this case the electrodes are not on the surface of the sensor material but penetrated inside the bulk This concept is illustrated in Figure It has several general advantages com pared to planar sensors like lower depletion voltage for semiconductors and faster charge collection while still retaining the full signal In diamond sensors the main advantage is the decreased distance between the electrodes which reduces the influence of trapping Two general designs of 3D sensors are available In the first design the electrodes com pletely penetrate the sensor This has the advantage that both electrodetypes can be contacted from the same side On the other hand this can cause problems if the density of the electrodes is very high as in this case it is more difficult to just connect one type of electrode with each other The major disadvantage of this design is also a general problem of 3D sensors If a particle passes perpendicular through the sensor it may deposit all its energy in one of the electrodes As these charges cannot be collected the efficiency drops at the electrodes To tackle this in another design the electrodes are implemented from the 20 4 2 Graphitisation of diamond particle particle uo1ia dap depletion Figure 4 2 Sensor with 3D left and planar right
85. stical uncertainty After changing the bias voltage the device waits for 500 ms before measuring the capacitance The mea sured values for each voltage step are fitted with a Gaussian distribution to calculate the mean value and its statistical error As the bias voltage is set to 0 V at the end of each measurement the B1505A is set to apply the first bias voltage for at least five seconds before starting the next measurement Using a non irradiated diamond sample like sample 3 the measurements show the ex pected behaviour A voltage sweep from 500 V to 500 V is shown in Figure The linear fit shows with a slope of 2 9 3 8 V fF that the capacitance of this DUT is stable over the whole voltage range This means that there are no free charge car riers in the diamond For the calculation of the mean value all measured values are histogrammed and fitted with a Gaussian distribution This yields a capacitance of 1 3589 0 0005 stat 0 0039 syst pF It is in good agreement with the theoretical value of 1 3215 0 0529 pF obtained from the physical dimensions of the electrodes and the thickness of the diamond For a highly irradiated diamond the assumption of no uncompensated charge carriers in the bulk does not hold true A characteristic result is given in Figure It is clearly visible that the capacitance changes with respect to the bias voltage Additionally the measured value for the capacitance depends on the direction of the voltage sw
86. tance with respect to the bias voltage is observed Mea surements of very low values of CCD are possible over a wide voltage range with high statistical precision Graphitisation of diamond using a femtosecond laser is tested for the manufacturing of electrodes for 3D sensors Keywords high energy physics diamond semiconductor sensors ill Contents 1 Introduction 2 Principles of solid state particle detectors 2 1 Energy loss of particles i Pea A DERG 2 2 Sensor design 2 3 Read out chain 3 Diamond as a sensor material for particle detectors 3 1 Production of CVD diamonds 2 seh ac we ash a rin 3 2 Fundamental properties of diamond oa a a a a e 3 3 Charge collection distance lt lt subite ie 3 4 C V measurement 4 Bulk segmentation 4 1 Motivation for new types of electrodes 00 000004 4 2 Graphitisation of diamond oe cade E SLEALE Sese 5 C V measurement 5 1 Experimental setup 5 1 1 The diamond fixturel 22 2 oo oo nn 5 1 2 The B1505A Semiconductor Device Analyzer 5 1 3 Diamond samples o 0 a a a ba He ea 5 2 Experimental results 5 2 1 Calibration 5 2 2 Voltage een ci ui LE ea a Le Pe 6 Measurement of CCD 6 1 Experimental setup 6 1 1 Data aquisition amp analysis xe e ee an i 6 2 Experimental results 11 11 13 14 16 19 19 21 27 27 21 27 30 31 31 34 41 41 43 44 Contents 6 2
87. the hysteresis of the capacitance with respect to the bias voltage decreases and the CCD increases An indication that a high particle flux is needed to fully pump a highly irradiated sample was found The shortest detrapping time in this measurement was on the order of hours The graphitisation process of diamond using a femtosecond laser for three dimensional sensors was tested successfully As an opaque diamond sample was used only graphitisa tion on the surface was observed The threshold of surface graphitisation for this sample was determined to be at least 482 5 pW For graphitisation inside the diamond bulk a sample of optical quality is needed as well as a mechanism to move the sample in all three dimensions Also a shutter to control the duration of exposure has to be installed 63 A Appendix A 1 Pictures of the diamond Substrate sid Figure A 1 Picture of one edge of diamond sample 1 see Section 5 1 3 The growth side top is clearly rougher than the substrate side bottom 65 A Appendix A 2 C V measurement C t 0 V 1 kHz 250 mV x10 E x 8 em Ss 5 F 2 4 1 3 1 al 3 9 3 8 Capacitance Ft 3 96063E 12 3 8E 15 3 7 Bu EEE EEE SN RPE RS OE LEERE ER BE See ler ge oe e en 0 100 200 300 400 500 600 time s Figure A 2 Measurement of capacitance at a constant bias voltage of 0 V and a frequency of 1 kHz The error for each data point indicat
88. the substrate Crucial parts of this process are the plasma generation and the mixture of a Carbon containing gas and Hydrogen in an Argon atmosphere Most commercial suppliers for CVD diamonds use microwaves to generate the plasma A sketch of such a CVD unit is presented in Figure 11 3 Diamond as a sensor material for particle detectors Microwave generator 2 45Hz Plasma Wave guide Precursor gases CH4 H2 Ar Viewing port Loading door Substrate heating stage Exhaust to vacuum pump Figure 3 1 Schematic of an apparatus for producing CVD diamonds The plasma is generated using a microwave generator 9 Another important aspect is the type of the substrate If a non diamond substrate is used i e silicon the diamond growth will start in several spots distributed over the substrate surface This forms separate singlecrystalline grains with slightly different orientations When they grow further and eventually merge grain boundaries will arise from these different orientations As growth in certain directions is favoured the number of grains will decrease with the thickness of the diamond Such diamonds are called polycrystalline pCVD diamonds To improve the quality of the diamond i e reduce the number of grains the diamond is polished from both sides But as mentioned above the number of grains reduces in direction of the growth side So this side is only polished to receive a flat surface From the substrate si
89. ts its lifetime Diamond sensors are an alternative for the harshest environments Diamond is very radi ation hard due to the high displacement energy and has similar properties compared to silicon The amplitude of the signal is lower than in Silicon as it has a larger band gap But due to this larger band gap the overall noise level is lower Thus diamond sensors are an option for the innermost layer of a future tracker at the upgrade of the LHC the so called super LHC sLHC During operation the sensors will be exposed to very high radiation doses This will dam age the bulk and especially decrease the charge collection efficiency Thus their behaviour at the end of their lifetime has to be studied In this thesis studies for highly irradiated samples are presented For information about the damage inside the bulk the change of capacitance with respect to the bias voltage is determined The charge collection distance CCD a figure of merit which is connected to the charge collection efficiency is measured for different samples For this measurement a setup is built which has a very low noise level To increase this efficiency new sensor concepts like 3D sensors are an option First steps for the fabrication of electrodes of such sensors using a femtosecond laser to graphitise the diamond are presented In Chapter 2 an overview of the mode of operation of solid state tracking detectors is presented Chapter 3 explains the special properti
90. ulf Quadt f r die M glichkeit danken in seiner Arbeitsgruppe diese Masterarbeit anfertigen zu k nnen Ebenso gilt mein Dank meinem Betreuer Dr Jens Weingarten der mich stets unterst tzt hat und mir hilfreich zur Seite stand Auch Prof Dr Claus Ropers m chte ich f r den Zugang zu einem Femtosekundenlaser danken sowie Max Gulde f r die technische Unterst tzung bei dieser Messung F r das Korrekturlesen dieser Arbeit danke ich Jens und Nina Zum Schluss gilt mein Dank meinen Eltern die mir ein sorgenfreies Studium erm glicht haben 85 Erklarung nach 18 8 der Priifungsordnung fiir den Bachelor Studiengang Phy sik und den Master Studiengang Physik an der Universitat Gottingen Hiermit erkl re ich dass ich diese Abschlussarbeit selbst ndig ver fasst habe keine anderen als die angegebenen Quellen und Hilfsmittel benutzt habe und alle Stellen die w rtlich oder sinngem aus ver f fentlichten Schriften entnommen wurden als solche kenntlich gemacht habe Dar berhinaus erkl re ich dass diese Abschlussarbeit nicht auch nicht auszugsweise im Rahmen einer nichtbestandenen Pr fung an dieser oder einer anderen Hochschule eingereicht wurde G ttingen den 8 September 2011 Lars Graber

Charakterisierung hochbestrahlter polykristalliner Diamantsensoren

Contents

Download Pdf Manuals

Related Search

Related Contents