Gain Measurements of a GridPix detector operated in Ar/iC 4 H 10 at different pressures

Gain Measurements of a GridPix detector operated in Ar/iC 4 H 10 at different pressures Analysis of data recorded at the Nikhef Detector R&D-group Eric Drechsler Universiteit van Amsterdam 16. January 2012 Abstract This report summarises the results of gain measurements in Ar/iC 4 H 10 performed at the Nikhef R&D-group. Photons with 5.9 kev from a 55 Fe-source were shot into a TPC with 6 cm drift gap at different pressures between 1.0 bar to 2.0 bar of the Ar/iC 4 H 10 (90/10) gas mixture. Primary electrons were collected with a GridPix detector. The maximum achieved single electron detection efficiency is 0.9 at p = 1.4 bar. The obtained gain curves decrease with increasing pressure and have maximum values around 10 4. 1

Contents 1. Introduction 3 2. GridPix Detector 3 3. Experimental Setup 5 3.1. General Setup.............................. 5 3.2. Trigger Setup and Signal Timing................... 7 3.3. Data Acquisition............................ 8 4. Analysis of Recorded Data 10 4.1. Occupancy and Time Spectra..................... 10 4.2. Cloud Shape............................... 11 4.3. Number of Detected Electrons..................... 13 4.4. Gain Curves............................... 15 5. Discussion 17 A. Plots 18 B. Dark Matter and WIMP Detection 19 B.1. Dark Matter............................... 19 B.2. Detection of WIMPs.......................... 19 References 21 2

1. Introduction Modern particle physics requires detectors at the frontier of current technologies. The detector components have to be robust and reliable. For a high spatial resolution very precise and sensitive materials are required. released A recent development is the GridPix detector, a micro pattern gaseous detector. A pixelated CMOS chip, the Timepix chip, provided with a mesh on top for charge amplification is used to detect ionisation produced in a gaseous gap by radiation. As readout for a Time Projection Chamber (TPC) the GridPix detector allows for a very high three dimensional resolution over a large volume. This technology is studied for example in the framework of a large TPC tracker for the ILC [1]. Likewise low statistics experiments, e.g. dark matter searches, investigate the use of GridPix technology. This report summarises the gain measurements and studies of single electron detection efficiency in Ar/iC 4 H 10 for different pressures and grid voltages performed at Nikhef. The measurements were performed in the framework of the dark matter WIMP search in noble liquids (DARWIN) consortium. Darwin has ambitious projects concerning the next generation of noble liquid detectors with increased sensitivity for the direct measurement of properties of dark matter - like mass and interaction strength - in Europe[2]. In the following a brief introduction to GridPix detectors is followed by a description of the experimental setup and the results of the data analysis. In the appendix a short discussion of dark matter and WIMP detection is provided. 2. GridPix Detector A GridPix detector is a Timepix chip with a thin metal foil on top. The Timepix is a development based on the Medipix2 -chip, which is a read-out chip using CMOS technology. It is divided into 256 256 pixels measuring 55 µm 55 µm each yielding to an overall sensitive area of 14 mm 14 mm. For collecting signals in gaseous detectors the chip is mounted under an amplification structure. In case of the GridPix detector shown in Fig.(1) this is a thin metal grid on top of insulating pillars with a pitch of 55 µm. Furthermore the chip includes a resistive layer to avoid damage from discharges. Between the grid and the chip a high voltage is applied - the grid voltage V Grid. When a drifting electron enters the amplification region electron avalanches are induced resulting in a signal amplification. This gain - the electron multiplication factor - depends on gas mixture, pressure and voltage. A scheme can be seen in Fig.(1). The Medipix2 includes a 14 bit counter originally intended for photon counting in X-ray imaging. The Timepix can use the 14 bit memory of each pixel not only for counting hits but also for counting clock pulses sent to the chip. Each pixel can be set to a specific threshold. The pixel is activated by a signal with adjustable length. This time window is called the shutter time. 3

Figure 1: Photo of a GridPix detector (left) and scheme of the working principle (right). When a threshold exceeding signal arrives during this time the pixel counter is triggered. There are 3 different counting modes for each pixel: Medipix mode: single hit counting - gives information how often the pixel was hit during one shutter cycle. This is the basic Medipix-Chip mode. Time over Threshold (ToT): after the incoming charge exceeds the threshold the clock pulses are counted as sketched in Fig.(2). Since the signal has a certain length it will fall below the threshold again and thereby stopping the counting. The length of the signal-pulse increases with the charge. Therefore the ToT-mode allows a measurement of the collected charge. During one shutter cycle multiple signals are integrated. Figure 2: Scheme ToT-mode signal. When a signal arrives the number of clock cycles elapsed while the pulse was above threshold is stored. Multiple hits are summed up.[3] Time of Arrival (ToA): as sketched in Fig.(3) the counter is started at the very first signal which exceeds the threshold and then stopped by the shutter closing. Therefore large number of clock counts correspond to an early signal arrival and small numbers to a late arrival. After the counter is activated the pixel becomes insensitive for further signals during the same shutter cycle. 4

Figure 3: Scheme ToA-mode signal. When a signal arrives the number of clock cycles elapsed until the shutter closing is stored. Multiple hit detections during one shutter cycle are not possible.[3] For precise measurements the timewalk effect has to be considered. The signal is required to reach a certain threshold per pixel before the counter starts. The incoming charge produced by the avalanche in the amplification region produces a signal with an amplitude depending on the size of the gain. The time needed for reaching the signal peak is characterised by the electrostatical configuration of the system, i.e. independent of the signal height. Therefore a large signal has a steep rising slope and crosses the threshold earlier compared to a low signal with a softer slope. This introduces a fluctuation in measured time caused by differing gain. Another effect of GridPix detectors is crosstalk, electronic communication between nearby pixels. If the capacitance between pixels is high enough, the signal on one pixel can induce a signal with opposite sign on the neighbouring pixels. Furthermore if the gain is high one primary electron can lead to multiple pixel hits. In this case the avalanche produced in the amplification region spreads far enough to exceed the threshold on neighbouring pixels as well. 3. Experimental Setup 3.1. General Setup The setup shown in Fig.(4) consists of a TPC with a single GridPix detector as read-out. The TPC is inside a gas-tight vessel with gas inlet and outlet allowing for a gas flow. With a connected vacuum pump the vessel can also be evacuated. Voltage is supplied by signal cables connected with feedthroughs on the main flange. The TPC volume is defined by a cathode and a 6 cm distant guard electrode measuring 10 cm 10 cm each. The guard electrode has an opening with the size of the GridPix detector. The latter is mounted on a carrier board. The read-out of the Timepix chip is realised with the RelaxD-system. Furthermore the grid is connected to a pre-amplifier allowing for a measurement of the induced charge on the grid. Electrons drifting away from the grid towards the pixels and ions drifting back to the cathode induce a signal with opposite sign wrt. the one on the pixels. 5

Figure 4: Main parts of the experimental setup - the vessel with the TPC, the Pre-Amp and the source (left) and the trigger setup (right). The pre-amplifier delivers two signals - a fast signal used for triggering the shutter and a slower signal for measuring the charge produced in the avalanches. Between the cathode and the grid a drift field of 300 V cm 1 was applied corresponding to a voltage of 1800 V. In the amplification region a voltage between 300 V 450 V ( 80 kv cm 1 ) multiplies the incoming primary electrons. Since the guard plate is 1 mm above the grid an uniform drift field requires a correction of 30 V between the guard and the grid. The gas used in this setup was a mixture of Ar/iC 4 H 10 (90/10) at pressures between 1.0 bar to 2.0 bar. Electrons or photons can ionise and excite the Ar atoms. This can result in secondary ionisation since the freed electrons can ionise other Ar atoms. These processes can lead to an avalanche by charge multiplication giving rise to a detectable signal. The interaction between excited Ar-atoms leads to emission of ultraviolet photons. The energy of these photons is different than any atomic level of Argon and hence the Ar atoms are transparent for the UV-photons. This allows for a long mean free path of latter. The gas always carries a small fraction of easily ionisable impurities like polyatomic molecules. Photons from de-excitation can ionise these impurities thereby freeing further electrons. This can lead to avalanches at distinct locations in the the detector. To avoid such multiple signals a quencher gas like isobutane ic 4 H 10 is added. This quencher gas can absorb the emitted UV-photons. The mean free path of the UV-photon therefore stays short. Also the additional reaction between quencher and photon frees electrons and increases the charge signal at the initial location. The source of the events are 5.9 kev photons from an 55 Fe source aimed at the middle of the TPC. In Ar/iC 4 H 10 these photons can transfer all their energy by freeing an outer shell electron. The energy needed for ionisation is only a few ev hence the photons energy is transferred to the kinetic energy of the liberated 6

electron. Because of its high energy this electron frees more electrons in a narrow area. On average 221 electrons are freed in Ar/iC 4 H 10 [4]. The resulting liberated electrons form a spherical cloud and start to drift along the field lines. While drifting the diffusion spreads the distance between the electrons. Nevertheless the shape of the cloud projected on the anode is a circle. The 5.9 kev photon can also hit an electron from an inner shell. More energy - around 2.3 kev - is required to free the electron which has therefore 2.6 kev kinetic energy. Hence the electron from the inner shell ionises less atoms. Under emission of a photon with 2.3 kev the hole in the inner shell is filled by an outer shell electron. This photon can escape the detector hence preventing a detection of its energy. 3.2. Trigger Setup and Signal Timing Two read-out systems are used in this setup - the RelaxD Timepix read-out and a digital oscilloscope 1 for the grid signal. In order to trigger on individual events and to synchronise the two systems a trigger system was implemented. The connection schematics for this setup is shown in Fig.(5). Timers were used to achieve a proper synchronisation and veto signals. The RelaxD board is used in external trigger Figure 5: Trigger system for synchronisation and individual event triggering.[5] mode. The shutter is started as soon as a TTL pulse from high to low is registered and stopped when a pulse from low to high arrives. After the acquisition is finished the chip is read out and the board sends a busy signal to the trigger system. The oscilloscope is triggered by the fast signal of the pre-amplifier and records the waveforms of the fast and slow signals. During the data processing the oscilloscope sends a busy signal to the trigger system. A trigger is accepted only when the two 1 The oscilloscope is a LeCroy Waverunner6030. 7

Trigger Setup Using Relaxd ReadOut for Gain Measurements LOGIC SIGNAL CHART of TIMERS Accepting Triggers DAQ Read-Out Time Restart Trigger signal from Grid Timer 1 Shutter close started from trigger True False 100 us Timer 2 Shutter True False Shutter open Comparator (Output ECL) Timer 3 Busy True False NIM Signals Timer 4 Wait till the noise of the shutter is gone True False 1 us 0 V -0.8 V Timer 5 Veto True False Veto ON Veto OFF Timer 6 Read-Out Busy True False Oscilloscope Relaxd NIKHEF Amsterdam: Gijs Hemink & Matteo Alfonsi 25 February 2011 Figure 6: Logical signal chart showing the timing of the signals.[5] system are finished with processing of the earlier event. In Fig.(6) the timing of the different signals is shown. After both read-out systems are finished with processing the shutter is opened. To avoid collection of electronics noise from the shutter opening the trigger veto is released only 1 µs after. As soon as a signal arrives on the grid the oscilloscope triggers the shutter closing and the veto. The closing has a 100 µs delay to ensure the complete collection of the signal. Thereafter the shutter is closed and the data acquisition of the pixels stopped. The global busy signal is activated since the read-out starts. The RelaxD read-out takes place after the shutter closing and is faster than the oscilloscope read-out. The whole system is restarted after the processing is finished. The 40 MHz pulse frequency results in an maximum shutter time of 295.25 ns since the maximum storable count value of the pixels is 11810. For a successful photon event the 100 µs delay should yield to a mean of 3900 counts in the time distribution. 3.3. Data Acquisition The RelaxD-board is connected to a computer with a gigabyte Ethernet cable and connected to the carrier board with a 100 pin cable. The software for processing RelaxD-data is called RelaxDAQ. In this setup RelaxDAQ was used to store the binary RelaxD-data and to apply changes in the setting of the RelaxD-board. Another software used for threshold-equalisation and creating pixel masks was Pixelman. The oscilloscope is connected to the computer. During the data taking the oscilloscope stores the waveforms directly on the computer in binary format. A setting 8

on the oscilloscope allowed a fast processing of the data resulting in a maximum event frequency of 20 Hz limited by the speed of the oscilloscope. For each voltage 10.000 events were recorded. This took between 12 min to 25 min for a run without problems depending on the voltage. The binary data files were converted to text-files with conversion programmes. 9

4. Analysis of Recorded Data The data taking was performed at Nikhef in October/November 2011. For different pressures between p = 1.0 bar and p = 2.0 bar data in ToA mode was taken. Also data in ToT-mode for pressure between 0.9 bar and 1.6 bar was recorded. In the following only the analysis for the ToA data is reported. If not stated otherwise the following figures are for run with p = 1.4 bar and V Grid = 410 V. 4.1. Occupancy and Time Spectra As a first approach to understand the chip performance a two dimensional figure showing the occupancy per run for each pixel was investigated. Such a figure shows the effective detection area as well as possible dead columns and other distortions. The figure in Fig.(7) reveals that the shielding from the guard plate narrows the sensitive area. Furthermore the right upper corner which was broken and later fixed with glue is insensitive to incoming electrons [5]. Overall the effective detection area is between 25 and 220 for pixels in x and 40 and 225 for pixels in y. The pixel column at x = 88 is dead and was masked by means of the Pixelman software. The time spectrum in Fig.(8) shows the number of clock pulses stored on each Figure 7: Occupancy figure. The hits are required to be within 25 < x px < 220 and 40 < y px < 225. hit pixel. As soon as the signal on the pixel exceeds the threshold the pixel starts counting until the shutter is closed. Each pulse corresponds to 25 ns since the clock pulse frequency was set to 40 MHz. The mean of the distribution is given by the shutter closing which happens 100 µs after the signal is detected by the grid. The distribution in Fig.(8) is not symmetric but has a tail towards smaller times (later arrival). This is caused by the timewalk effect explained in Sec.(2). Longitudinal diffusion also influences the time spectrum. The signal on the grid has to reach a certain threshold to trigger the shutter closing. If the incoming charge gets spread in z-direction the first arriving electrons do not exceed the 10

threshold. Only when enough electrons arrive the shutter closing is triggered. Hence the pixels activated by the first electrons have a higher pulse count and vice versa the pixels hit by late electrons - after the shutter closing is triggered - will have a lower pulse count. In that way the longitudinal diffusion spreads the time spectrum. As a simplification a Gaussian was fitted to the distribution. The mean λ and the width σ were used to determine cuts on the time distributions by requiring the time count of each pixel to be within λ ± 5σ. With this cut all hits which do not belong to a certain event with well defined time information are discarded. Figure 8: Time spectrum with Gaussian fit. The width is determined by the longitudinal diffusion and statistical fluctuations. The tail towards lower pulse counts is caused by the timewalk effect. 4.2. Cloud Shape In Fig.(9) an example event in the effective detection area is shown. The expected circular cloud shape is visible. A first cut is obtained by requiring a minimum dimension of the cloud. Events generated very close to the grid get rejected by such a cut, since the drift distance is too small to obtain a sufficient spread. For such a cut the distribution containing the distance x between the maximum and the minimum hit pixel is shown in Fig.(10) for x and Fig.(App.18) for y. A minimum size of 5 per dimension is required for each event. The right figure in Fig.(10) reveals a decrease in the cloud size for increasing pressure. While the mean of x for p = 1.4 bar is around 40 pixels it decreases to 20 pixels for p = 1.6 bar. The same holds for the y-dimension. 11

Figure 9: An example event. The pixels are required to be in time and within the effective detection area. This behaviour can be explained by a decrease in the initial cloud size. Since a higher pressure corresponds to more particles in the volume the ionisation takes place in a smaller area. the decreasing mean free path of the electrons. coefficient gets smaller. The transversal diffusion Figure 10: The relative distance x for p = 1.4 bar (left) and p = 1.6 bar (right) with V grid = 410 V. The shape of the projection of the cloud on the anode can be used to classify the event. An estimator for the circularity c of the cloud is constructed by dividing the spread in x by the spread in y resulting in c x / y. Ideally c 1 but because of the statistical nature of the transversal diffusion small deviations have to be expected. This can be seen in Fig.(11). Large deviations could originate from events close to the border of the effective detection area since parts of the expected circle are cut away. The estimator is required to be 50 /77 < c < 77 /50. 12

Figure 11: The ratio c between the cloud dimension in y and x. The chosen range is 50/77 < c < 77 /50. 4.3. Number of Detected Electrons The GridPix detector has a very high single electron detection efficiency. Since the transversal diffusion spreads the initial cloud an estimation of the number of primary electrons is possible by counting the number of hit pixels. The spread of the cloud should be high enough to resolve each primary electron with the GridPix (with a hole pitch of 55 µm). Fig.(12) shows the distribution of number of hit pixels per event. Two distinct Figure 12: The number of hit pixels. The first peak is the escape peak, the second the photo peak. The latter is expected to be around 221 but is shifted towards lower values. 13

peaks are visible. When the 5.9 kev photon from the 55 Fe source hits an outer shell electron of the Ar the photo peak is formed at high hit counts. The peak at lower hit counts is the escape peak from events with a liberated inner shell electron. The peaks in Fig.(12) are shifted towards lower numbers than expected, indicating a loss of primary electrons. As a cut the number of hit pixels is required to be within 5 to 350. Fig.(13) shows the integral values of the slow signal on the grid recorded by the Figure 13: The integral of the charge measured by the oscilloscope. The photo and the escape peak are visible. oscilloscope. The signal is caused by the electrons and the back drifting ions. It is proportional to the number of electrons after the amplification. Since both signals - the charge on the grid and the number of hit pixels - should be proportional a scatter figure between both signals is shown in Fig.(14). Two dense areas which are the photo and the escape peak are visible and connected by a straight line of entries. This shows that both signals are indeed proportional. The entries in between the dense areas are events near the border since neither pixels are hit nor charge is induced on the grid. By applying the cuts on the position and the shape of the cloud the straight connection between the two dense areas gets removed as well as some noisy hits. This can be seen in Fig.(15). Nevertheless an unexpected contribution of entries with oscilloscope signal in the photo/escape peak region but lower number of hit pixels is visible. These entries are events in which the charge seen by the grid does not correspond to the number of detected electrons in terms of hit pixels. A possible explanation is that two or more primary electrons hit the same pixel by entering the same hole on the grid. The charge induced on the grid is then the same as if two or more pixels got hit. The reason for this is the small diffusion coefficient. The electrons do not drift a sufficient transversal distance to be separated by the hole pitch of the GridPix. 14

Figure 14: A scatterfigure of the hit pixels and the charge integral for each event. The two dense areas are events contributing to the photo respectively the escape peak. Figure 15: The implemented requirements cut away the linear transition between the two dense areas. However a clear contribution of events with less hit pixels but high charge integral stays. 4.4. Gain Curves The previous discussion was repeated for every measured pressure. The results are used to create the efficiency curves in Fig.(16). These figures show the number of detected electrons in terms of hit pixels divided by the number of expected electrons Ne exp = 221 in dependence of the applied grid voltage. The left and the right figures show the curves at lower and at higher pressures respectively. The highest detection efficiency around 0.9 was achieved at 1.4 bar and V Grid = 410 V. Between 1.2 bar 1.6 bar the single electron detection efficiency is between 0.6 0.9 for the maximum grid voltages for each setting. The curves of 1.8 bar and 2.0 bar show a very small efficiency smaller than 0.1. For 1.0 bar - below atmospheric pressure - a maximum efficiency of 0.58 was achieved. The 15

overall maximum grid voltage with which a whole data set could be taken was at 1.6 bar V Grid = 450 V. Finally the achieved gain for each field configuration and Figure 16: Number of detected electrons divided by expected number of N e = 221. pressure can be calculated. Therefore the oscilloscope signal has to be calibrated to correlate the number of produced electrons in the amplification region to the recorded signal height. The number of electrons can be calculated from the integral of the oscilloscope signal Vosc int by n e = ( 1.23 ± 1.0) 10 5 int 13 Vosc + (3.696 ± 0.057) 10 Ne exp which is a result of the calibration of the detector realised in [5]. The gain curves are shown in Fig.(17). The highest gain was achieved at 1.2 bar with V Grid = 420 V Figure 17: Achieved gain curves for different pressures. followed by 1.6 bar with 450 V. In general the gain decreases with higher pressure at the same grid voltage. Also higher pressures allow for a higher maximum grid voltage the detector can be operated with. Furthermore the curves show an exponential like behaviour - disregarding 1.8 and 2.0 bar. 16

5. Discussion In this report a short introduction to GridPix detectors was given, followed by a description of the experimental setup used to perform gain measurements of a GridPix detector in Ar/iC 4 H 10 at different pressures. The analysis of the collected data was summarised in Sec.(4) giving insights into detector performance and the physics processes in the detector for different pressures. The occupancy figure in Fig.(7) shows the sensitive area of the detector which is narrower than the chip itself. This loss is expected since the shielding of the guard plate yields to field distortions which are inevitable in this setup. For a future setup this could be optimised by using a field shaper. The time spectra of the recorded data was used to determine the window in which hits were in time. Therefore gaussians fits were performed to generate automatic upper and lower limits. The distributions showed an asymmetry caused by the timewalk effect and longitudinal diffusion. A more careful fitting procedure including the fitting of the asymmetric tail could give a measure for longitudinal diffusion and timewalk effect. The electrons freed by the 5.9 kev photon from the source form initially a spherical cloud. Hence the shape of the recorded hits was investigated. With changing pressure the minimum and maximum cloud spread in x and y changes. For increasing pressure and constant grid voltages the cloud sizes decreases due to a lower transversal diffusion coefficient and narrower initial clouds. Furthermore the circularity of the hits was used to cut away events on the edge of the sensitive area. By counting the number of hit pixels the number of freed electrons was determined. The photo and the escape peak are at lower counts than expected. This can be explained by multiple hits per pixel. Some of the primary electrons do not drift a sufficient transversal distance to be separated by the holes of the GridPix. The integrated grid signal is proportional to the number of electrons. A scatterfigure between the number of hit pixels and the integrated oscilloscope signal validates the multiple hit explanation. A cut on the tail would shift the position of the peaks towards higher values but also bias the results. The single electron efficiency curves in Fig.(16) vary with pressure. A maximum efficiency of 0.9 was achieved. Finally the gain curves in Fig.(17) show the achieved gain for each pressure. The operation of the system at 1.8 and 2.0 bar was difficult since voltage drops occured frequently and collection of noise was inevitable with the chosen threshold. Furthermore the gain is very small hence no clear signal can be detected. Nethertheless the GridPix detector showed an impressive capability of working at high pressures. This capability could be used to reduce background and noise in low statistics experiments like dark matter searches briefly discussed in the appendix. 17

A. Plots Cloud Size in y Figure 18: The relative distance y px = x max px 1.6 bar (right) with V grid = 410 V. x min px for p = 1.4 bar (left) and p = 18

B. Dark Matter and WIMP Detection One of the biggest secrets in modern astronomy, cosmology and particle physics is the nature of dark matter. Many research groups designed and constructed experiments with different technologies for direct dark matter searches. Noble liquid detectors are amongst the most promising designs in terms of sensitivity of direct WIMP searches. B.1. Dark Matter In 1933 the Swiss physicist Fritz Zwicky discovered a missmatch between the rotation velocity of galaxies and the prediction from the visible mass. This was the first indirect observation of dark matter (DM). An analysis of the energy distribution of our universe revealed distributions as shown in Fig.19. Figure 19: Measurements of the cosmic microwave background by the WMAP experiment unveil the composition of our universe 2. Different natures of DM were assumed, cold dark matter being a promising candidate. Latter requires new, yet unseen particles which undergo the gravitational and weak interaction but not the strong or electromagnetic force. Such a particle is called WIMP - weakly interacting massive particle. Candidates for WIMP particles require an extension of the Standard Model of elementary particle physics. In R-parity conserving supersymmetric models the lightest supersymmetric particle (LSP) is stable and hence gives a candidate for cold DM. The mass of the LSP is believed to be > 100 MeV and it could be detected by the LHC. B.2. Detection of WIMPs A WIMP candidate like the LSP could be directly observed by the LHC. Decay chains involving gluinos and other supersymmetric particles can lead to a final state with a certain amount of missing energy which could be an escaping WIMP. Furthermore indirect detection by observing cosmic decay products from WIMP annihilation could be possible. Since WIMPs interact gravitationally a locally higher density is expected in space areas with a high matter density like the centre 2 http://map.gsfc.nasa.gov/universe/uni_matter.html, retrieved on 06.02.2012 19

of galaxies or the sun. Hence the higher annihilation rate would yield to escaping neutrinos with a high energy spectrum, i.e. above 100 MeV. These neutrinos can be detected on earth and by discriminating from the background a indirect observation of WIMPs could be achieved. A third option is the observation of an elastic scatter event between a WIMP and a nucleus. The scattering releases recoil energy which can be detected in different ways. By measuring the heat of lattice vibrations phonons from scatter events can be observed. This measurement requires the target to be cooled down to a few mk. Also ionisation is a possible result from the produced recoil energy. Latter can be measured by using semiconductors as target material. A third option is the production of photons by scintillation. For noble liquids, e.g. xenon or argon at low temperatures, the recoil produces ionisation and scintillation yielding to two measurable signals. Using noble liquids as detection material requires a high purity 1 ppb and a temperature between 87 K to 163 K in case of argon or xenon respectively [6]. Furthermore the energy transfer - hence the recoil energy - between nucleus and WIMP can be maximised by choosing the nucleus mass around the expected WIMP mass. Xenon is a good candidate for many WIMP models since its nucleus mass is m Xe N 122 MeV. Another important requirement of dark matter search experiments is a proper shielding. The rate of background events originating from e.g. natural radiation, electronics noise and cosmic radiation has to be minimised in order to measure rare WIMP scatter events. For this purpose, only very pure materials are used and the detectors are deep underground. The European consortium dark matter WIMP search in noble liquids (DARWIN ) investigates possible concepts for the measurement of properties of dark matter. Next generation dual phase noble liquid detectors are designed and studied to increase the sensitivity by three orders of magnitude [2]. Within this framework direct charge readout for WIMP scatter events is studied. A candidate is the GridPix-detector operated in a dual-phase noble liquid TPC. The detector has to be operated in the cryogenic environment of liquid Xe/Ar, which requires intense performance studies. A dual-phase Xenon dark matter detector like the Xenon100 detector at the LNGS is a TPC filled with liquid xenon. On top and bottom of this vessel several PMTs are placed. Furthermore a thin layer of gaseous xenon is maintained on top of the liquid xenon. If a WIMP scatter event occurs, the recoiling nucleus can excite and ionise the surrounding atoms. The excitation of the atoms leads to scintillation photons at 177.6 nm which is detected by PMTs almost immediately after the event. The electrons from the ionisation start to drift towards the anode and hence the gaseous layer. Once the electrons reach this layer, they excite gaseous atoms and thereby produce more UV-photons which are detected by PMTs. This results in two distinguishable signals S 1 and S 2 with different amplitudes. The distance of the event to the anode is determined by the time difference in arrival of the signals. The ratio of the signal amplitudes S 2 /S1 is furthermore a good discriminator of γ-background events. 20

References [1] T. Behnke, C. Damerell, J. Jaros, A. Myamoto, et al., ILC Reference Design Report Volume 4 - Detectors (2007), arxiv:0712.2356v1 [2] DARWIN Consortium, DARWIN - dark matter WIMP searches in noble liquids (02.03.2012), URL http://darwin.physik.uzh.ch/index.html [3] F. Kloeckner, Teststrahlmessung einer GEM-basierten TPC mit simultaner Datenauslese von acht Timepix-Chips, Diploma thesis, Physikalisches Institut, Universität Bonn (2010) [4] M. A. Chefdeville, Development of Micromegas-like gaseous detectors using a pixel readout chip as collecting anode, Ph.D. thesis, Universiteit Twente (2009) [5] G. Hemink, GridPix - A pixel sensor for noble liquid dark matter searches, Master s thesis, University of Twente (2011) [6] J. Jochum, Direct dark matter detection. International school in Astroparticle physics. (July 2011) 21