Table Of Content

Trends in Pixel Detectors: Tracking and Imaging Norbert Wermes (Invited Paper) Abstract—For large scale applications, hybrid pixel detectors, withpresentdaysemiconductordetectors.Forapplicationswith 4 in which sensor and read-out IC are separate entities, constitute lower demands on the spatial resolution (PSF ∼5-10µm) 0 the state of the art in pixel detector technology to date. They 0 but with requirements on real time and time resolved data have been developed and start to be used as tracking detectors 2 acquisition, semiconductor pixel detectors are very attractive. andalsoimagingdevicesinradiography,autoradiography,protein n crystallographyandinX-rayastronomy.Anumberoftrendsand Verythindetectorassembliesaremandatoryforvertexdetec- a possibilities for future applications in these fields with improved torsatcolliderexperiments,in particularforthe plannedlinear J performance, less material, high read-out speed, large radiation e+e− collider. This imposes high demands on the detector 8 tolerance, and potential off-the-shelf availability have appeared development. While silicon is almost a perfect material for ] asenmdi-amreonmoloitmheicntaapriplyromacahteusrewdh.iAchmodnogntohtemreqaureiremcoonmolpitlhicicateodr particlephysicsdetectors,allowingtheshapingofelectricfields t hybridization but come as single sensor/IC entities. Most of bytailoredimpuritydoping,theneedofhighphotonabsorption e these are presently still in the development phase waiting to efficiency in radiological applications requires the study and d - be used as detectors in experiments. The present state in pixel use of semiconductor materials with high atomic charge, such s detectordevelopmentincludinghybridand(semi-)monolithicpixel as GaAs or CdTe. For such materials the charge collection n techniques and their suitability for particle detection and for i imaging, is reviewed. properties are much less understood and mechanical issues in . s particular those related to hybrid pixels are abundant, most c Index Terms—pixel detector, hybrid pixels, monolithic pixels, notably regardingthe hybridizationof detectors when they are i tracking, imaging s not available in wafer scale sizes. Finally, for applications in y highradiationenvironments,suchashadroncolliders,radiation h I. INTRODUCTION hard sensors are developed using either Si with a dedicated p [ Albeit being similar at first sight, the requirements for design or CVD-diamond. charged particle detection in high energy physics compared to 1 imaging applications with pixel detectors can be very differ- v II. HYBRID PIXELS:THE STATEOF THEART IN PIXEL 0 ent. In particle physics experiments charged particles, usually DETECTORTECHNOLOGY 3 triggered by other subdetectors, have to be identified with 0 high demands on spatial resolution and timing. In imaging In the hybrid pixel technique sensor and FE-chips are sepa- 1 applicationstheimageisobtainedbytheun-triggeredaccumu- ratepartsofthedetectormoduleconnectedbysmallconducting 0 lation (integrating or counting) of the quanta of the impinging bumps applied by using the bumping and flip-chip technology. 4 0 radiation,oftenalsowithhighdemands(e.g.>1MHzperpixel All LHC-collider-detectors [1], [2], [3] ALICE, ATLAS, and / incertainradiographyorCTapplications).Sipixeldetectorsfor CMS, LHCb for the RICH system [4], as well as some s c highenergychargeparticledetectioncanassumetypicalsignal fixed target experiments (NA60[5] at CERN and BTeV[6] at i charges collected at an electrode in the order of 5,000-10,000 Fermilab) employ the hybrid pixel technique to build large s y electronseventakingintoaccountchargesharingbetweencells scale(∼m2)pixeldetectors.Pixelareasizesaretypically50µm h and detector deterioration after irradiation to doses as high as ×400µm as for ATLAS or 100µm ×150µm as for CMS. The p 60 Mrad. In 3H-autoradiography, on the contrary, or in X- detectorsare arrangedin cylindricalbarrelsof 2−3layersand : v ray astronomy the amount of charge to be collected with high disks covering the forward and backward regions. The main i efficiency can be much below 1000e. The spatial resolution is purposeofthedetectorsatLHCis(a)theidentificationofshort X governedbytheattainablepixelgranularityfromafewtoabout livedparticles(e.g.b-taggingforHiggsandSUSYsignals),(b) r a 10µm at best, obtained with pixel dimensions in the order of patternrecognitionandeventreconstructionand(c)momentum 50µm to 100µm. The requirementsfrom radiologyare similar. measurement. The detectors need to withstand a total (10 yr) Mammography(∼80µmpixeldimensions)ismostdemanding particle fluence of 1015neq corresponding to a radiation dose with respect to space resolution. Some applications in autora- of about 50 Mrad. The discovery that oxygenated silicon is diography,however, require sub-µm resolutions, not attainable radiation hard with respect to the non-ionizing energy loss of protons and pions [7] saves pixel detectors at the LHC for N. Wermes is with Bonn University, Physikalisches Institut, Nussallee 12, whichtheradiationismostsevereduetotheirproximityto the D-53115Bonn,Germany,email:[email protected] Work supported by the German Ministerium fu¨r Bildung, Wissenschaft, interaction point. n+ electrode in n-bulk material sensors have Forschung und Technologie (BMBF) under contract no. 05HA8PD1, by been chosen to cope with the fact that type inversion occurs tWheestMfailnenisteurniudmer fcu¨orntrWacistsennos.chIaVftAu5nd−F1o0r6sc0h1u1ng98d,easndLabnydestheNoDrdeurhtsecinh–e afteraboutΦeq =2.5×1013cm−2.Aftertypeinversionthepn- Forschungsgemeinschaft DFG diode sits on the electrode side thus allowing the sensor to be operatedpartiallydepleted.Figure1(a) showsthe development The chip and sensor connectionis done by fine pitch bump- of the effective dopingconcentrationand the depletionvoltage ingandsubsequentflip-chipwhichisachievedwitheitherPbSn V as a function of the fluence expected for 10 years of (solder) or Indium bumpsat a failure rate of <10−4. Figure 3 dep operation at the LHC (see also [8]). Figures 1(b) and (c) show showsrowsof50µmpitchbumpsobtainedbythesetechniques. the ATLASpixelsensor wafer with 3 tiles [9] and the detector response after irradiation over two adjacent pixels showing a very homogeneouscharge collection except for some tolerable lossattheouteredgesandinanareabetweenthepixels.These areduetothebumpcontactandabiasgrid,respectively,which hasbeendesignedtoallowtestingofthesensorsbeforebonding them to the electronics ICs. Fig.2 (LEFT)INDIUM(PHOTOAMS,ROME)AND(RIGHT)SOLDER(PBSN, PHOTOIZM,BERLIN)BUMPROWSWITH50µMPITCH. Indium bumping is done using a wet lift-off technique applied on both sides (sensor and IC) [12]. The connection is obtained by thermo-compression. The Indium joint is com- paratively soft and the gap between IC and sensor is about 6µm. PbSn bumps are applied by electroplating[13]. Here the bump is galvanically grown on the chip wafer only. The bump is connected by flip-chipping to an under-bump metallization on the sensor substrate pixel. Both technologies have been successfully used with 8” IC-wafers and 4” sensor wafers. InthecaseofATLASamoduleof2.1x6.4cm2 areaconsists ofFE-chipsbump-connectedtoonesiliconsensor.TheI/Olines ofthechipsareconnectedviawirebondstoakaptonflexcircuit glued atop the sensor. The flex houses a module control chip (MCC) responsiblefor frontendtime/triggercontrolandevent building.Thetotalthicknessatnormalincidenceisinexcessof 2%X0. Figure3showsa241Am(60keVγ)sourcescan ofan ATLAS module with 46080 pixels and an amplitude spectrum obtained using the ToT analog information without clustering of pixels. Fig.1 The modules are positioned by dedicated robots on carbon- (A)EFFECTIVEDOPINGCONCENTRATIONANDDEPLETIONVOLTAGEASA carbonladders(staves)andcooledbyevaporationofafluorinert FUNCTIONOFTHEFLUENCE,(B)ATLASSILICONPIXELWAFERWITH3 liquid (C3F8) at an input temperature below −20oC in order TILES,AND(C)CHARGECOLLECTIONEFFICIENCYINTWOADJACENT to maintain the entire detector below −6oC to minimize the PIXELSAFTERIRRADIATION. damageinducedby radiation.This operationrequirespumping andthecoolingtubesmuststand16barpressureifpipeblock- ing occurs. All detector components must survive temperature Thechallengeinthedesignofthefront-endpixelelectronics cycles between −25oC and room temperature. The LHC pixel [10], [11] can be summarized by the following requirements: detectors are presently in production. A sketch of the final low power (< 50µW per pixel), low noise and threshold ATLAS pixel detector is shown in fig. 4. dispersion (together < 200e), zero suppression in every pixel, on-chip hit buffering,and small time-walk to be able to assign III. TRENDS IN HYBRID PIXELDETECTORS the hits to their respective LHC bunch crossing. The pixel ToimprovetheperformanceofHybridPixeldetectorsincer- groups have reached these goals in several design iterations tain aspects severalideas and developmentsare beingpursued. using first radiation-soft prototypes, then dedicated radhard designs, and finally using deep submicron technologies [11]. While CMS uses analog readout of hits up to the counting A. HAPS house, ATLAS obtains pulse height information by means of Hybrid(Active)PixelSensors(HAPS)[14]exploitcapacitive measuring the time over threshold (ToT) for every hit. couplingbetween pixels to obtain smaller pixel cells and pixel Fig.5 (LEFT)LAYOUTOFAHAPSSENSORWITHDIFFERENTPIXELAND Fig.3 READOUTPITCHES,(RIGHT)MEASUREDSPATIALRESOLUTIONSBY (TOP)SCANOFANATLASPIXELMODULE(46080PIXELS)WITHA241AM SCANNINGWITHAFINEFOCUSLASER. (60KEVγ)SOURCEAND(BOTTOM)THECORRESPONDINGAMPLITUDE SPECTRUMOBTAINEDUSINGTHESIGNALAMPLITUDEVIA TIME-OVER-THRESHOLD(TOT)WITHOUTCLUSTERING. andfromthemodulefront-endchips.Aninterestingalternative to the flex-kapton solution is the so-called Multi-Chip-Module TechnologydepositedonSi-substrate(MCM-D)[16].A multi- conductor-layerstructure is built up on the silicon sensor. This allows to bury all bus structures in four layers in the inactive area of the module thus avoiding the kapton flex layer and any wire bonding at the expense of a small thickness increase of 0.1% X0. In addition, the extra freedom in routing makes it possible to design pixel detectors which have the same pixeldimensionsthroughoutthesensor,i.e.withoutlargeredge pixelsusuallynecessarytoaccommodatethesensorareawhich remainsuncoveredbyelectronicsinbetweenchips.Figure6(b) illustrates the principle and fig. 6(c) shows scanning electron microphotographs(curtesyIZM,Berlin)ofaviastructuremade in MCM-D technology. Fig.4 C. Active edge 3-D silicon VIEWOFTHEATLASPIXELDETECTORINITSPRESENTDESCOPED The features of doped Si allow interesting geometries with VERSIONWITHTWOBARRELLAYERSAND2X3DISKS. respect to field shapes and charge collection. So-called 3D silicon detectors have been proposed [17] to overcome sev- eral limitations of conventional planar Si-pixel detectors, in particular in high radiation environments, in applications with pitchwith a largerreadoutpitchresultingin interleavedpixels. inhomogeneousirradiation and in applications which require a The pixel pitch is designed for best spatial resolution using large active/inactive area ratio such as protein crystallography charge sharing between neighbors while the readout pitch is [18]. A a 3D-Si-structure (fig. 7(a)) is obtained by processing tailored to the needs for the size of the front-end electronics the n+ and p+ electrodes into the detector bulk rather than cell. This way resolutions between 3µm and 10µm can be by conventional implantation on the surface. This is done obtained with pixel (readout) pitches of 100µm (200µm) as by combining VLSI and MEMS (Micro Electro Mechanical is shown in fig. 5(b) together with the layout of a prototype Systems) technologies. Charge carriers drift inside the bulk sensor in fig. 5(a). parallel to the detector surface over a short drift distance of typically50µm.Anotherfeatureis the factthatthe edgeof the B. MCM-D sensor can be a collection electrode itself thus extending the The present hybrid-pixel modules of the LHC experiments active area of the sensor to within about 10µm to the edge. use an additional flex-kapton fine-print layer on top of the Si- This results from the undepleted depth at this electrode. Edge sensor (fig. 6(a)) to provide power and signal distribution to electrodesalsoavoidinhomogeneousfieldsandsurfaceleakage Fig.6 (A)SCHEMATICVIEWOFAHYBRIDPIXELMODULESHOWINGICSBONDED VIABUMPCONNECTIONTOTHESENSOR,THEFLEXHYBRIDKAPTON LAYERATOPTHESENSORWITHMOUNTEDELECTRICALCOMPONENTS, ANDTHEMODULECONTROLCHIP(MCC).WIREBONDCONNECTIONSARE NEEDEDASINDICATED.(B)SCHEMATICLAYOUTOFAMCM-DPIXEL MODULE.(C)SEMPHOTOGRAPHOFAMCM-DVIASTRUCTURE(TOP) Fig.7 ANDACROSSSECTIONAFTERBUMPDEPOSITION(BOTTOM). (TOP)SCHEMATICVIEWOFA3DSILICONDETECTOR,(BOTTOM) COMPARISONOFTHECHARGECOLLECTIONIN3D-SILICONAND CONVENTIONALPLANARELECTRODEDETECTORS. currents which usually occur due to chips and cracks at the sensor edges. This is seen in fig. 7(b) which shows the two structures in comparison. The main advantages of 3D-silicon the characteristic systematic spatial shifts observed in CVD detectors come from the fact that the electrode distance is diamondduetothecrystalgrowth.Grainstructuresareformed short (50µm) in comparison to conventional planar devices at inwhichtrappedchargesattheboundariesproducepolarization the same total charge, resulting in a fast (1-2 ns) collection fields inside the sensor which cause the observed spatial shifts time, low (< 10V) depletion voltage and, in addition, a large of the order of 100-150 µm. active/inactive area ratio of the device. The technical fabrication is much more involved than for IV. HYBRID PIXELS IN IMAGING APPLICATIONS planar processes and requires a bonded support wafer and re- A. Counting Pixel Detectors for (Auto-)Radiography activeionetchingoftheelectrodesintothebulk.Acompromise between3Dandplanardetectors,socalledplanar-3Ddetectors A potentially new route in radiography techniques is the maintaining the large active area, use planar technology but counting of individual X-ray photons in every pixel cell. This with edge electrodes [19], obtained by diffusing the dopant approach offers many features which are very attractive for fromthedeeplyetchededgeandthenfillingitwithpoly-silicon. X-ray imaging: full linearity in the response function and a Prototype detectors using strip or pixel electronics have been principallyinfinite dynamic range, optimalexposuretimes and fabricated and show encouraging results with respect to speed agoodimagecontrastcomparedtoconventionalfilm-foilbased (3.5 ns rise time) and radiationhardness(> 1015 protons/cm2) radiography. It thus avoids over- and underexposed images. [20]. Counting pixel detectors must be considered as a very direct spin-off of the detector development for particle physics into biomedicalapplications.Theanalogpartofthepixelelectronics D. Diamond Pixel Sensors is in parts close to identicalto the onefor LHC pixeldetectors Diamond as a potentially very radiation hard material has whiletheperipheryhasbeenreplacedbythecountingcircuitry been explored intensively by the R&D collaboration RD42 at [23]. CERN [21]. Charge collection distances up to 300µm have The challenges to be addressed in order to be competitive been achieved. Diamond pixel detectors have been built with with integrating radiography systems are: high speed (> 1 pixel electronics developed for the ATLAS pixel detector and MHz) counting with a range of at least 15 bit, operation with operated as single-chip modules in testbeams [22]. A spatial very little dead time, low noise and particularly low threshold resolution of σ = 22µm has been measured with 50µm pixel operation with small threshold dispersion values. In particular pitch, compared to about 12µm with Si pixel detectors of the the last item is important in order to allow homogeneous same geometry and using the same electronics. This reveals imaging of soft X-rays. It is also mandatory for a differential Fig.9 (LEFT)2X2MULTI-CHIPCDTEMODULEWITHCOUNTINGPIXEL ELECTRONICSUSINGTHEMPECCHIPONAUSBBASEDREADOUTBOARD [27],(RIGHT)IMAGEOFACOGWHEELTAKENWITH60KEVX-RAYS. Fig.8 IMAGEOFA55FEPOINTSOURCETAKENWITHTHEMEDIPIX2COUNTING B. Protein Crystallography CHIPANDASILICONSENSOR(14X14MM2)[26]. Hybrid Pixel detectors as counting imagers lead the way to new classes of experiments in protein crystallography with synchrotron radiation [34]. Here the challenge is to image, with high rate (∼1-1.5 MHz/pixel) and high dynamic range, manythousandsofBraggspotsfromX-rayphotonsoftypically energymeasurement,realizedso farasa doublethresholdwith 12 keV or higher (corresponding to resolutions at the 1A˚ energywindowinglogic[24],[25]. A differentialmeasurement range),scattered off proteincrystals. The typicalspot size of a of the energy, exploiting the different shapes of X-ray spectra diffractionmaximumis100−200µm,callingforpixelsizesin behind for example tissue or bone, can enhance the contrast theorderof100−300µm.Thehighlinearityofthehitcounting performanceof an image. Finally,for radiographyhigh photon methodandtheabsenceofso-called”bloomingeffects”,i.e.the absorption efficiency is mandatory, which renders the not easy responseofnon-hitpixelsinthecloseneighborhoodofaBragg task of high Z sensors and their hybridization necessary. maximum, makes counting pixel detectors very appealing for Progress with counting pixel systems have been reported at protein crystallography experiments. A systematic limitation this conference[26], [27], [28]. The MEDIPIX chip developed and difficulty is the problem that homogeneous hit/count re- by the MEDIPIX collaboration [29], [25] uses 256 × 256, sponses in all pixels, also for hits at the pixel boundaries 55×55µm2 pixels fabricated in 0.25µm technology, energy or between pixels where charge sharing plays a role must windowing via two tunable discriminator thresholds, and a 13 be maintained by delicate threshold tuning. Counting pixel bitcounter.Themaximumcountrateperpixelisabout1MHz. developments are made for ESRF (Grenoble, France) [35] Figure 8 shows an image of a 55Fe point source obtained with and SLS (Swiss Light Source at the Paul-Scherrer Institute, the Medipix2 chip bonded to a 14x14 mm2, 300µm thick Si Switzerland) beam lines. The PILATUS 1M detector [36] at sensor [26]. A similar image with a single chip CdTe detector the SLS (217µm ×217µm pixels) is made of fifteen 16- and a higher energy 109Cd source (122 keV γ) has also been chip-modules each covering 8×3.5 cm2, i.e. a total area of taken and the capability to detect low energybeta decaysfrom 40×40cm2. Figure 10(a) shows a photograph of this detector tritium has been demonstrated [26]. with 15 modules and close to 106 pixels covering an area of 20×24cm2 [37]. Itis the first largescale hybridpixeldetector A Multi-Chip module with 2x2 chips using high-Z CdTe inoperation.Figure10(b)showsaBraggimageofaLysozyme sensorswiththeMPECchip[27]isshowninfig.9.TheMPEC with 10s exposure to 12 keV sychrotron X-rays [37]. Many chip features 32×32 pixels (200×200µm2), double threshold spots are contained in one pixel which is the limit of the operation, 18-bit counting at ∼1 MHz per pixel as well as achievable point spread resolution. low noise values (∼120e with CdTe sensor) and threshold In summary hybrid pixel detectors are the present state dispersion (21e after tuning) [30], [27]. A technical issue here of the art in pixel detector technology. For tracking as well is the bumping of individualdie CdTe sensors which has been as for imaging applications large (∼m2) detectors are being solved using Au-stud bumping with In-filling [31]. built or already in operation (PILATUS). On the negative side A challenge for counting pixel detectors in radiology is to are the comparatively large material budget, the complicated build large area detectors. Commercially available integrating assembling (hybridization) with potential yield pitfalls, and pixellated systems such as flat panel imagers [32], [33] consti- the difficulty to obtain high assembly yields with non-wafer tute the levelling scope. scalehigh-Zsensormaterials.Astrendswithinthistechnology, Linear e+e− Collider [38]. Very low (≪1% X0) material per detector layer, small pixel sizes (∼20µm×20µm) and a good rate capability (80 hits/mm2/ms) is required, due to the very intense beamstrahlung of narrowly focussed electron beams close to the interaction regionwhich produceelectron positron pairsinvastnumbers.Highreadoutspeedsof50MHzlinerate with 40µs frame time are necessary. Fig.11 (A)SKETCHOFAMONOLITHICPIXELDETECTORUSINGAHIGH RESISTIVITYBULKWITHPMOSTRANSISTORSATTHEREADOUT(AFTER [39]),(B)PRINCIPLEOFANMONOLITHICACTIVEPIXELSENSOR(MAPS) [41]TARGETINGCMOSELECTRONICSWITHLOWRESISTIVITYBULK MATERIAL.THECHARGEISGENERATEDANDCOLLECTEDBYDIFFUSIONIN THEFEWµMTHINEPITAXIALSI-LAYER. Among the developments of monolithic or semi-monolithic pixeldevicesperhapsthefollowingclassifyinglistcanbemade. - Non-standard CMOS on high resistivity bulk The first monolithic pixel detector, successfully operated in a particle beam already in 1992 [39], used a high resistivity p-type bulk p-i-n detector in which the junction had been created by an n-type diffusion layer. On one side, an array of ohmic contacts to the substrate served as Fig.10 (TOP)PHOTOGRAPHOFTHE20X24CM2LARGEPILATUS1MDETECTOR collection electrodes (fig. 11(a)). Due to this only simple pMOS transistor circuits sitting in n-wells were possible FORPROTEINCRYSTALLOGRAPHYUSINGCOUNTINGHYBRIDPIXEL in the active area. The technology was certainly non- DETECTORMODULES,(BOTTOM)IMAGEOFLYSOZYMETAKENWITH standard and non-commercial. No further development PILATUS[37],THEINSERTSHOWSTHATBRAGGSPOTSAREOFTEN emerged. CONTAINEDINONEPIXEL. - CMOS technology with charge collection in epi-layer Commercial CMOS technologies use low resistivity interleaved pixels and MCM-D promise better resolution and silicon whichis notsuitedforchargecollection.However, more homogeneous modules, while 3D-silicon detectors can in some technologies an epitaxial silicon layer of a few address applications in high radiation environments or those to 15µm thickness can be used [40], [41], [42]. The where a large active/inactive area ratio is mandatory. generatedchargeiskeptintheepi-layerbypotentialwells at the boundary and reaches an n-well collection diode V. MONOLITHICAND SEMI-MONOLITHICPIXEL by thermal diffusion (fig. 11(b)). The signal charge is DETECTORS therefore very small (<1000e) and low noise electronics Monolithic pixel detectors, in which amplifying and logic is a real challenge. As this development uses standard circuitryaswellastheradiationdetectingsensorareoneentity, processing technologies it is potentially very cheap. producedinacommerciallyavailabletechnology,arethedream Despite using CMOS technology the potential of full ofsemiconductordetectordevelopers.Thiswillnotbepossible CMOS circuitry in the active area is not available (only in the near future without compromising. Developments and nMOS) because of the n-well/p-epi collecting diode trends in this direction have much been influenced by R&D which does not permit other n-wells. for vertex tracking detectors at future colliders such as a FET transistor is implanted in every pixel on a sidewards depleted[56]bulk.Electronsgeneratedbyradiationin the bulk are collected in a potential minimum underneath the transistorchannelthusmodulatingitscurrent(fig.13).The bulk is fully depleted rendering large signals. The small capacitanceoftheinternalgateofferslownoiseoperation. Both together can be used to fabricate thin devices. The Fig.12 sensor technology is non-standard and the operation of (A)CROSSSECTIONTHROUGHAMONOLITHICCMOSONSOIPIXEL DEPFET pixel detectors requires separate steering and amplification ICs (see below). DETECTORUSINGHIGHRESISTIVITYSILICONBULKINSOLATEDFROMTHE LOWRESISTIVITYCMOSLAYERWITHCONNECTINGVIASINBETWEEN [15],[47],(B)CROSSSECTIONTHROUGHASTRUCTUREUSING AMORPHOUSSILICONONTOPOFSTANDARDCMOSVLSIELECTRONICS [48],[49]. - CMOS on SOI (non-standard) In order to exploit high resistivity bulk material, the authors of [47] develop pixel sensors with full CMOS circuitry using the Silicon-on-Insulator (SOI) technology. The connection to the charge collecting bulk is done Fig.13 by vias (fig. 12(a)). The technology offers full charge PRINCIPLEOFOPERATIONOFADEPFETPIXELSTRUCTUREBASEDONA collection in 200–300µm high resistive silicon with full SIDEWARDSDEPLETEDDETECTORSUBSTRATEMATERIALWITHAN CMOS electronics on top. At present the technology is IMBEDDEDPLANARFIELDEFFECTTRANSISTOR.CROSSSECTION(LEFT)OF however definitely not a commercial standard and the HALFAPIXELWITHSYMMETRYAXISATTHELEFTSIDE,ANDPOTENTIAL development is still in its beginnings. PROFILE(RIGHT). - Amorphous silicon on standard CMOS ASICs Hydrogenated amorphous-Silicon (a-Si:H), where the Because CMOS active pixels and DEPFET pixels are most H-content is up to 20%, can be put as a film on top advanced in their development, they are discussed in more of CMOS ASIC electronics. a-Si:H has been studied detail below. as a sensor material long ago and has gained interest again [48], [49] with the advancement in low noise, low A. CMOS Active Pixels power electronics. The signal charge collected in the <30µm thick film is in the range of 500-1500 electrons. The main difference of CMOS monolithic active pixel sen- A cross section through a typical a-Si:H device is shown sors [40] to CMOS camera chips is that they are larger in in fig. 12(b). From a puristic view it is more a hybrid area and must have a 100% fill factor for efficient particle technology,butthemaindisadvantageofthehybridization detection. Standard commercial CMOS technologies are used connection is absent. The radiation hardness of these which have a lightly doped epitaxial layer between the low detectors appears to be very high >1015cm−2 due to resistivity silicon bulk and the planar processing layer. The the defect tolerance and defect reversing ability of the epi-layer is – technologydependent– at most 15µm thick and amorphous structure and the larger band gap (1.8 eV). can also be completely absent. Collaborating groups around The carrier mobility is very low (µ = 2-5 cm2/Vs, µ = IReS&LEPSI[42],[43],RAL[44]andIrvine-LBNL-Ohio[45] e h 0.005cm2/Vs), i.e. essentially onlyelectronscontributeto use similar approaches to develop large scale CMOS active the signal.Basically anyCMOScircuitandICtechnology pixels also called MAPS (Monolithic Active Pixel Sensors) can be used and technology changes are uncritical for [41].Prototypedetectorshavebeenproducedin0.6µm,0.35µm this development.For high-Z applicationspoly-crystalline and 0.25µm CMOS technologies [54], [52]. HgI constitutes a possible semiconductor film material. In the MAPS device (fig. 11(b)) electron-hole pairs created 2 The potential advantages are small thickness, radiation by ionization energy loss in the epitaxial layer remain trapped hardness, and low cost. The development is still in its between potential barriers on both sides of the epi-layer and beginnings and – as for CMOS active pixel sensors – a reach, by thermal diffusion, an n-well/p-epi collection diode. real challenge to analog VLSI design. Thesensorisdepletedonlydirectlyunderthen-wellcollection diode where full charge collection is obtained; the charge - Amplification transistor implanted in high resistivity bulk collectionisincompleteinallotherareas.Withsmallpixelcells In so-called DEPFET pixel sensors [50] a JFET or MOS- collectiontimes in the orderof 100nsare obtained.The signal is small, some hundred to 1000e, depending on the thickness differenttechnologies.Itbecomesthinnerforsmallerprocesses of the epi-layer. A chip submission using a process with no and with it also the number of produced signal electrons epi-layer, but with a low doped substrate of larger resistivity vanishes. Only a few processing technologies are suited. With hasalsobeentried(MIMOSA-4)[53]whichprovedtofunction the rapid changeof commercialprocesstechnologiesthis is an with high detection efficiency for minimum ionizing particles. issue of concern. The fact that good detector performance has Thisrenderstheuncertaintyanddependenceonthefutureofthe been seen with non-epi wafers material provides some relief. epi-layer thickness in different technologies less constraining. Furthermore, in the active area, due to the n-well collection Matrix readout performed using a standard 3-transistor cir- diode,noCMOS(onlynMOS)circuitryispossible.Thevoltage cuit (line select, source-follower stage, reset) commonly em- signals are very small (∼mV), of the same order as transistor ployed by CMOS matrix devices, but can also include cur- threshold dispersions requiring very low noise VLSI design. rent amplification and current memory [54]. For an image The radiation tolerance of CMOS pixel detectors for particle two complete frames are subtracted from each other (CDS) detection beyond 1012 neq is still a problem although the which suppressesswitching noise. Noise figuresof 10-30eand achieved tolerance should be sufficient for a Linear Collider. S/N ∼20 have been achieved with spatial resolutions below Improved readout concepts and device development for high 5µm. The radiation hardness of CMOS devices is always a rate particle detection at a linear collider is underdevelopment crucialquestionforparticledetectioninhighintensitycolliders. [38].FortheupgradeoftheSTARmicrovertexdetectortheuse MAPS appear to sustain non-ionizing radiation (NIEL) to of CMOS active pixels is planned [54]. ∼1012n while the effectsof ionizing radiation damage (IEL) eq are at present still unclear [55]. Apart from this the focus of B. DEPFET Pixels further development lies in making larger area devices and The Depleted Field Effect Transistor structure (DEPFET) in increasing the charge collection performance in the epi- [50] providesdetectionandamplificationpropertiesjointlyand layer. For the former, first results in reticle stitching, that is hasbeenexperimentallyconfirmedandsuccessfullyoperatedas electrically connecting IC area over the reticle border, have singlepixelstructuresandaslarge(64x64)pixelmatrices[51]. been encouraging (fig. 14(a)) [52]. Full CMOS stitching over The DEPFET detector and implanted amplification structure is thereticleboarderstillhastobedemonstrated.Thepoorcharge showninfig.13.Sidewardsdepletion[56]providesaparabolic collectionbythermaldiffusionisanotherareacallingforideas. potential which has – by appropriate biasing and a so-called Approachesusingmorethanonen-wellcollectingdiode[44],a deep-n implantation – a local minimum for electrons (∼1µm) photo-FET[54],[42],[43],oraphoto-GATE[46]arecurrently underneath the transistor channel. The channel current can be investigated. The photo-FET called technique (fig. 14(b)) has steered and modulated by the voltage at the external gate and features similar to those of DEPFET pixels (see below). A - important for the detector operation - also by the deep- pMOS FET is implanted in the charge collecting n-well and n potential (internal gate). Electrons collected in the internal signalchargesaffectitsgatevoltageandmodulatethetransistor gate are removed by a clear pulse applied to a dedicated current. CLEAR contact outside the transistor region or by other clear mechanisms. The very low input capacitance (∼ few fF) and the in situ amplification makes DEPFET pixel detectors very attractive for low noise operation[57]. Amplificationvalues of 400 pA per electron collected in the internal gate have been achieved.Further amplificationenters at the second levelstage (current based readout chip). DEPFET pixelsare currentlybeingdevelopedfor three very different application areas: vertex detection in particle physics [58], [59],X-rayastronomy[60]andforbiomedicalautoradio- graphy [57]. Figure 15 summarizes the achieved performance innoiseandenergyresolutionwithsinglepixelsstructures(fig. Fig.14 15(a))andinspatialresolutionswith64x64pixelsmatrices(fig. (A)SEAMLESSSTITCHINGINBETWEENWAFERRETICLESTOOBTAIN 15(b)). With round single pixel structures noise figures of 2.2e LARGEAREACMOSACTIVEPIXELS,(B)SCHEMATICOFAPMOS at room temperature and energy resolutions of 131 keV for 6 PHOTO-FETTOIMPROVECHARGECOLLECTIONINMAPS. keVX-rayshavebeenobtained.Withsmall(20x30µm2)linear structures fabricated for particle detection at a Linear Collider the noise figures are about10e. The spacial resolution read off Aboveall,theadvantagesofafullyCMOSmonolithicdevice from fig. 15(b) in matrices operated with 50 kHz line rates is relate to the adoption of standard VLSI technology and its [59] resultinglowcostpotential(potentially∼25$percm2).Inturn the disadvantages also come largely from the dependence on σxy =(4.3±0.8)µm or 57mLPm for 22keVγ commercialstandards.Thethicknessof the epi-layervariesfor σ =(6.7±0.7)µm or 37LP for 6keVγ xy mm The capability to observe tritium in autoradiographical ap- Monolithicorsemi-monolithicdetectors,inwhichdetectorand plications is shown in fig. 15(c). readout ultimately are one entity, are currently developed in variousforms,largelydrivenbytheneedsforparticledetection at future colliders. CMOS active pixel sensors using standard commercialtechnologiesonlowresistivitybulkandSOIpixels, a-SI:H, and DEPFET-pixels, which try to maintain high bulk resistivity for charge collection, classify the different ways to monolithic detectors which are presently carried out. ACKNOWLEDGMENT The author would like to thank W. Dulinski, M. Caccia, P. Jarron, P. Russo, C. da Via’, and S. Parker for providing new material and information for this presentation. Fig.15 REFERENCES (A)RESPONSEOFASINGLEDEPFETPIXELSTRUCTURETO6KEVX-RAYS [1] Technical Design Report of the ATLAS Pixel Detector, FROMAN55FESOURCE,(B)IMAGEOBTAINEDWITH55FEANDATEST CERN/LHCC/98-13(1998); N.Wermes,DesignandPrototypePerformanceoftheATLASPixel CHART,THESMALLESTSTRUCTURESHAVEAPITCHOF50µMANDARE Detector, Nucl.Inst.Meth. A447(2002)121-128. 25µMWIDE,(C)3HLABELLEDLEAF. [2] CMS, the Tracker System Project, Technical Design Report, CERN/LHCC/98-006(1998); W. Erdmann, The CMS pixel detector, Nucl. Inst. Meth. A447 (2000)178-83. The verygoodnoisecapabilitiesof DEPFET pixelsare very [3] ALICETechnical DesignReport, CERN/LHCC99-12(1999); important for low energy X-ray astronomy and for autora- P. Chochula et al., The ALICE Silicon Pixel Detector, Nucl. Phys. diography applications. For particle physics, where the signal A715(2003)849-852. [4] T. Gys, The pixel hybrid photon detectors for the LHCb-RICH charge is large in comparison, this feature is used to design project, Nucl.Inst.Meth.A465(2001)240-246. verythindetectors(∼50µm)withverylowpowerconsumption [5] A.Baldit etal.,CERN-SPSC-2000-010(2000); whenoperatedasa row-wiseselected matrix[59]. Thinningof C.Lourenco,E.Radermacher, CERN-SPSC-2002-025(2002). [6] S. Kwan (BTeV Collaboration), The BTEV Pixel and Microstrip sensors to a thickness of 50µm using a technology based on Detectors, Nucl.Inst.Meth.A511(2003)48-51. deep anisotropic etching has been successfully demonstrated [7] ROSE Collaboration, G.Lindstro¨m et al., Nucl. Inst. Meth. A466 [61]. For the development of DEPFET pixels for a Linear (2001)308-326; G.Lindstro¨m,Radiation Damage inSilicon Detectors, Nucl. Inst. Collider matrix row rates of 50 MHz and frame rates for Meth.A512(2003)92. 520x4000pixelsof25kHz,read outattwo sides, are targeted. [8] O. Krasel et al., Measurement of the trapping time constants in Sensors with cell sizes of 20x30µm2 have been fabricated and Proton-Irradiated Silicon Pad Detectors, IEEE Nuclear Science SymposiumPortland 2003,theseproceedings. complete clearing of the internal gate, which is important for [9] M. S. Alam et al., The ATLAS Silicon Pixel Sensors, Nucl. Inst. highspeedon-chippedestalsubtraction,hasbeendemonstrated Meth.A456(2001)217-232. [59]. A large matrix is readout using sequencer chips for row [10] L.Blanquart etal. Analogfront-end cell designed inacommercial 0.25mprocessfortheATLASpixeldetector atLHC, IEEETrans. selectionand currentbasedchip forcolumnreadout[59]. Both Nucl.Sci.vol.49(4),(2002)1778-1782 chips have been developed at close to the desired speed for [11] L.Blanquartetal.,FE-I2:AFront-EndReadoutChipDesignedina a Linear Collider. The estimated power consumption for a commercial0.25µmProcessfortheATLASPixeldetectoratLHC, IEEENuclearScienceSymposiumPortland2003,theseproceedings. five layer DEPFET pixel vertex vertex detector is only 5W [12] C.Gemme,StudyofIndiumbumpingfortheATLASpixeldetector, rendering a very low mass detector without cooling pipes Nucl.Inst.Meth.A465(2001)200–203. feasible.ThemostrecentstructuresuseMOSFETsasDEPFET [13] G. Engelmann, J. Wolf, L. Dietrich, S. Fehlberg, O. Ehrmann and H. Reichl, Proc. 1999 Joint International Meeting of the Electro- transistors for which the radiation tolerance still has to be chemical Society,Hawaii 1999,Vol.99-2p.907; investigated. J. Wolf, PbSn60 Solder Bumping by Electroplating, Pixel2000 Workshop,Genova, Italy,June2000. [14] W. Kucewicz, G. Deptuch, A. Zalewska, M. Battaglia, K. Oster- VI. SUMMARY berg, R. Orava, M. Caccia, C. Meroni, G. Vegni, P. Grabiec, B. Driven by the demands for high spatial resolution and high Jaroszewicz, J. Marczewski, Capacitively coupled active pixel sen- sorswithanalogreadoutforfuturee+e-colliders,ActaPhys.Polon. rate particle detection in high energy physics, semiconductor B30(1999)2075–2083. pixel detectors have started to also become exploited for [15] M.Amati,M.Baranski,A.Bulgheroni,M.Caccia,K.Domanski,P. imaging applications. Hybrid pixel detectors, in which sensor Grabiec, M.Grodner, B.Jaroszewicz, W.Kucewicz, K.Kucharski, S.Kuta, W.Machowski, J.Marczewski, H.Niemiec, M.Sapor, D. and electronic chip are separate entities, connected via bump Tomaszewski,HybridActivePixelSensorsandSOIInspiredOption, bonding techniques represent today’s state of the art for both, Nucl.Inst.Meth.A511(2003)265–270. particledetectionandimagingapplications.Newtrendsinclude [16] P.Gerlach, C.Linder, K.H.Becks, Nucl.Inst.Meth. A473(2001) 102-106andreferences therein; interleaved pixels having different pixel- and readout-pitches, M. To¨pper, Intl. Conf. on High Density Packaging and MCMs, MCM-D structuresand 3D-silicon detectors with active edges. Denver, Colorado, April1999. [17] C.Kenney,S.Parker,J.Segal,C.Storment,Silicondetectorswith3- S.Watanabe,N.Wermes,AcountingCdTepixeldetectorforhardX- Delectrodearrays:fabricationandinitialtestresults, IEEETrans. rayandgamma-rayimaging, IEEETrans.Nucl.Sci.vol.48,(2001) Nucl.Sci.vol.46(4),(1999)1224–1236; 2401-2404. C.J. Kenney, S.I. Parker, B. Krieger, B. Ludewigt, T.P. Dubbs, H. [32] T. Ducourant, M. Michel, G. Vieux, T. Peppler, J.C. Trochet, R.F. Sadrozinski, Observation of beta and Xrays with 3-D-architecture Bastiaens,F.Busse, Optimizationofkeybuildingblocksforalarge- siliconmicrostripsensors, IEEETrans.Nucl.Sci.vol.48(2),(2001) arearadiographic and fluoroscopic dynamic digital X-raydetector 189-193; based on a-Si:H/CsI:Tl flat panel technology, Proc. SPIE 3977 C.J.Kenney,S.Parker, E.Walckiers, Resultsfrom3-Dsilicon sen- (2000)14-25; sorswithwallelectrodes:near-cell-edgesensitivitymeasurementsas N.Jungetal.,Proc.SPIE3336(1998)396-407. apreviewofactive-edgesensors, IEEETrans.Nucl.Sci.vol.48(6), [33] O.Tousignant,M.Choquette,Y.Demers,L.Laperriere,J.Leboeuf, (2001)2405. M.Honda,M.Nishiki,A.Takahashi,A.Tsukamoto,Progressreport [18] E. M. Westbrook, Pixels and Proteins: Better Detectors for Bi- on the performance of real-time selenium flat-panel detectors for ological Crystallography, Review given at IEEE Nuclear Science directX-rayimaging, Proc.SPIE4682(2002)503-510. Symposium,PortlandOct.2003,theseproceedings; [34] H. Graafsma, Detector Needs at Current and Future Synchrotron S.I. Parker, C.J. Kenney, J.D. Segal, E. Westbrook, J. Hasi, C. Da Sources,Review given atIEEENuclear Science Symposium,Port- Via’, E. Mandelli, G. Meddeleer, J. Morse, Direct X-ray Detector landOct.2003,theseproceedings. with 3D Silicon Pixel Sensors for Structural Molecular Biology, [35] P. Delpierre, J.F. Berar, L. Blanquart, B. Caillot, J.C. Clemens, IEEENuclearScienceSymposiumPortland2003,theseproceedings. C.Mouget, IEEETrans.Nucl.Sci.vol.48(4),(2001)987-991; [19] C.J. Kenney, E. Westbrook, J.D. Segal, S.I. Parker, J. Morse, J. P. Delpierre, J.F. Berar, L. Blanquart, N. Boudet, P. Breugnon, Hasi,C. DaVia’, FirstResults from Active-Edge, Planar radiation B. Caillot, J.C. Clemens, I. Koudobine, C. Mouget, R. Potheau, Sensors, IEEE Nuclear Science Symposium Portland 2003, these I. Valin, NSS-MIC/IEEE, San Diego (USA), Nov 4-10, 2001, proceedings. accepted forpublication inIEEETrans.Nucl.Sci.,Sept.2002. [20] S.I. Parker, A. Kok, C. Da Via’, J. Hasi, G. Anelli, P. Jarron, C.J. [36] C.Bro¨nnimannetal.,JournalofSyn.Rad.7(2000)301. Kenney,J.D.Segal,MeasurementsofFastRise-timePulsesfrom3D [37] E. Eikenberry, C. Broennimann, G. Huelsen, H. Toyokawa, M. Silicon Sensors, IEEENuclear Science Symposium Portland 2003, Suzuki,R.Horisberger,B.Schmitt,C.Schulze-Briese,T.Tomizaki, theseproceedings; The PILATUS X-ray Detector, 13th International Workshop on C.J. Kenney, E. Westbrook, J.D. Segal, E. Perozzielo, S.I. Parker, Room-Temperature Semiconductor X- and Gamma-Ray Detectors, J. Hasi, C. Da Via’, J. Morse, First Results from Active Edge 3D Portland2003. Architecture radiation Sensors, IEEE Nuclear Science Symposium [38] TESLATechnical DesignReport, Vol.IV,DESY-01-011(2001). Portland2003,theseproceedings. [39] W.Snoeys,J.Plummer,S.Parker,C.Kenney,Anewintegratedpixel [21] RD42 Collaboration, Diamond as a particle detector, Proc. Itatial detector for high energy physics, IEEE Trans. Nucl. Sci. vol.39, Phys.Socienty, Torino,vol.52(1996),105; (1992)1263–1269. R.Wedenigetal.,CVDDiamondPixelDetectors forLHCExperi- [40] B.Dierickx,G.Meynants,D.Scheffer,Near100%fillfactorCMOS ments,Nucl.Phys.B.(Proc.Suppl.)78(1999)497. active pixel, Proc. IEEE CDD&AIS Workshop, Brugge, Belgium, [22] M.Keiletal.,NewResultsonDiamondPixelSensorsusingATLAS June1997; FrontendElectronics, Nucl.Inst.Meth.A501(2003)153-159. G. Meynants, B. Dierickx, D. Scheffer, CMOS active pixel image [23] P.Fischer, Anareaefficient 128 channel counter chip, Nucl. Inst. sensor with CCD performance, Proceedings SPIE - Int. Soc. Opt. Meth.A378(1996)297–300. Eng.(USA),vol.3410(1998)68–76,9. [24] P. Fischer, A. Helmich, M. Lindner, N. Wermes, L. Blanquart, A [41] R. Turchetta, J.D. Bertst, B. Casadei, G. Claus, C. Colledani, W. PhotonCountingChipwithEnergyWindowing, IEEETrans.Nucl. Dulinski,Y.Hu,D.Husson,J.-P.Le-Normand,J.L.Riester,G.Dep- Sci.vol.47(3),(2000)881–884; tuch,U.Goerlach,S.Higueret,M.Winter,Amonolithicactivepixel M.Lindner,L.Blanquart,P.Fischer,H.Kru¨ger,N.Wermes,Medical sensor for charged particle tracking and imaging using standard X-ray imaging with energy windowing, Nucl. Inst. Meth. A465 VLSICMOS technology, Nucl. Inst. Meth. A458 (2001) 677-689; (2000)229–234. G.Claus, C.Colledani, W.Dulinski, D.Husson,R.Turchetta, J.L. [25] X.Llopart,M.Campbell,R.Dinapoli,D.SanSegundo,E.Pernigotti, Riester, D. Deptuch, G. Orazi, M. Winter, Particle tracking using Medipix2:A64-kpixelreadoutchipwith55-mumsquareelements CMOS monolithic active pixel sensor, Nucl. Inst. Meth. A465 working in single photon counting mode, IEEE Trans. Nucl. Sci. (2001)120–124. vol.49(5),(2002)2279–2283. [42] G. Deptuch, W. Dulinski, Y. Gornushkin, C. Hu-Guo, I. Valin, [26] G.Mettivier,M.C.Montesi,P.Russo,DigitalAutoradiographywith Monolithic Active Pixel Sensor with on-pixel amplification and a Medipix2 Hybrid Silicon Pixel Detector, Poster given at IEEE Double Sampling Operation, Nucl. Inst. Meth. A512 (2003) 299- Medical ImagingConference, Portland Oct.2003; 309; G. Mettivier, M.C. Montesi, P. Russo, Tritium digital autoradiog- G. Claus et al., Monolithic Active Pixel Sensors for a Linear raphy with a Medipix2 hybrid silicon pixel detector, submitted to Collider, Nucl.Inst.Meth.A473(2001)83–85. Nucl.Inst.Meth.A [43] G.Deptuchetal.,MonolithicActivePixelSensorwithIn-pixelDou- [27] M. Lo¨cker, P. Fischer, M. Kouda, S. Krimmel, H. Kru¨ger, M. ble Sampling and Columns-level Disrcrimination, Poster submitted Lindner, K. Nakazawa, T. Takahashi, N. Wermes, Single photon tothis conference. Counting X-ray Imaging with Si and CdTe Single Chip Pixel [44] R. Turchetta et al., Monolithic Active Pixel Sensors (MAPS) for Detectors and Multichip Pixel Modules, IEEE Nuclear Science Particle Physics and Space Science, Proceedings VERTEX 2003, SymposiumPortland 2003,theseproceedings. LakeWindermere, UK,September2003. [28] F.Edling,R.Brenner,N.Bingefors,K.Fransson,L.Gustafsson,L. [45] H. Matis, F. Bieser, S. Kleinfelder, G. Rai, F. Retiere, H. Ritter, delRiscoNorrlid,C.Ro¨nnqvist,AHybridPixelDetectorwithTwo K.Singh,S.Wurzel,H.Wiemann,E.Yamamoto,ChargedParticle Counters forX-rayImaging,PostergivenatIEEE-NSSymposium, Detection using a CMOS Active Pixel Sensor, IEEE Trans. Nucl. PortlandOct.2003,theseproceedings. Sci.vol.50,(()2003)1020-1025. [29] Medipix Collaboration, S.R. Amendiola et al., Nucl. Inst. Meth. [46] S. Kleinfelder, F. Bieser, Y. Chen, I. Kotov, H.S. MAtis, M. A466(2001)74-78. Oldenburg, F. Retiere, K.D. Singh, H.H. Wiemann, E. Yamamoto, [30] P. Fischer, J. Hausmann, A. Helmich, M. Lindner, N. Wermes, L. NovelIntegratedCMOSPixelStructuresforVertexDetectors,IEEE Blanquart, Acounting pixel readout chip andsensor system for X- NuclearScience SymposiumPortland 2003,theseproceedings. rayimaging,IEEETrans.Nucl.Sci.,Vol46,No.4(1999)1070; [47] J. Marczewski, K. Domanski, P. Grabiec, M. Grodner, B. L.Blanquart,V.Bonzon,G.Comes,P.Delpierre,P.Fischer,J.Haus- Jaroscewicz, A. Kociubinski, D. Tomaszewski, W. Kucewicz, S. mann,M.Keil,M.Lindner,S.Meuser,N.Wermes, PixelReadout Kuta,W.Machowski,H.Niemiec,MSapor,M.Caccia,SOIActive ElectronicsforLHCandBiomedicalApplications, Nucl.Inst.Meth. Pixel Detectors of Ionizing Radiation - Technology and Design A439(2000)80–90. Development, IEEE Nuclear Science Symposium Portland 2003, [31] P.Fischer,M.Kouda,H.Kru¨ger,M.Lindner,G.Sato,T.Takahashi, theseproceedings.

ebook img

Trends in Pixel Detectors: Tracking and Imaging PDF

1.2 MB·English
by  Norbert Wermes
#additional_collections #journals #arxiv
Save to my drive
Quick download
Download
Upgrade Premium
Most books are stored in the elastic cloud where traffic is expensive. For this reason, we have a limit on daily download.

Preview Trends in Pixel Detectors: Tracking and Imaging

Trends in Pixel Detectors: Tracking and Imaging Norbert Wermes (Invited Paper) Abstract—For large scale applications, hybrid pixel detectors, withpresentdaysemiconductordetectors.Forapplicationswith 4 in which sensor and read-out IC are separate entities, constitute lower demands on the spatial resolution (PSF ∼5-10µm) 0 the state of the art in pixel detector technology to date. They 0 but with requirements on real time and time resolved data have been developed and start to be used as tracking detectors 2 acquisition, semiconductor pixel detectors are very attractive. andalsoimagingdevicesinradiography,autoradiography,protein n crystallographyandinX-rayastronomy.Anumberoftrendsand Verythindetectorassembliesaremandatoryforvertexdetec- a possibilities for future applications in these fields with improved torsatcolliderexperiments,in particularforthe plannedlinear J performance, less material, high read-out speed, large radiation e+e− collider. This imposes high demands on the detector 8 tolerance, and potential off-the-shelf availability have appeared development. While silicon is almost a perfect material for ] asenmdi-amreonmoloitmheicntaapriplyromacahteusrewdh.iAchmodnogntohtemreqaureiremcoonmolpitlhicicateodr particlephysicsdetectors,allowingtheshapingofelectricfields t hybridization but come as single sensor/IC entities. Most of bytailoredimpuritydoping,theneedofhighphotonabsorption e these are presently still in the development phase waiting to efficiency in radiological applications requires the study and d - be used as detectors in experiments. The present state in pixel use of semiconductor materials with high atomic charge, such s detectordevelopmentincludinghybridand(semi-)monolithicpixel as GaAs or CdTe. For such materials the charge collection n techniques and their suitability for particle detection and for i imaging, is reviewed. properties are much less understood and mechanical issues in . s particular those related to hybrid pixels are abundant, most c Index Terms—pixel detector, hybrid pixels, monolithic pixels, notably regardingthe hybridizationof detectors when they are i tracking, imaging s not available in wafer scale sizes. Finally, for applications in y highradiationenvironments,suchashadroncolliders,radiation h I. INTRODUCTION hard sensors are developed using either Si with a dedicated p [ Albeit being similar at first sight, the requirements for design or CVD-diamond. charged particle detection in high energy physics compared to 1 imaging applications with pixel detectors can be very differ- v II. HYBRID PIXELS:THE STATEOF THEART IN PIXEL 0 ent. In particle physics experiments charged particles, usually DETECTORTECHNOLOGY 3 triggered by other subdetectors, have to be identified with 0 high demands on spatial resolution and timing. In imaging In the hybrid pixel technique sensor and FE-chips are sepa- 1 applicationstheimageisobtainedbytheun-triggeredaccumu- ratepartsofthedetectormoduleconnectedbysmallconducting 0 lation (integrating or counting) of the quanta of the impinging bumps applied by using the bumping and flip-chip technology. 4 0 radiation,oftenalsowithhighdemands(e.g.>1MHzperpixel All LHC-collider-detectors [1], [2], [3] ALICE, ATLAS, and / incertainradiographyorCTapplications).Sipixeldetectorsfor CMS, LHCb for the RICH system [4], as well as some s c highenergychargeparticledetectioncanassumetypicalsignal fixed target experiments (NA60[5] at CERN and BTeV[6] at i charges collected at an electrode in the order of 5,000-10,000 Fermilab) employ the hybrid pixel technique to build large s y electronseventakingintoaccountchargesharingbetweencells scale(∼m2)pixeldetectors.Pixelareasizesaretypically50µm h and detector deterioration after irradiation to doses as high as ×400µm as for ATLAS or 100µm ×150µm as for CMS. The p 60 Mrad. In 3H-autoradiography, on the contrary, or in X- detectorsare arrangedin cylindricalbarrelsof 2−3layersand : v ray astronomy the amount of charge to be collected with high disks covering the forward and backward regions. The main i efficiency can be much below 1000e. The spatial resolution is purposeofthedetectorsatLHCis(a)theidentificationofshort X governedbytheattainablepixelgranularityfromafewtoabout livedparticles(e.g.b-taggingforHiggsandSUSYsignals),(b) r a 10µm at best, obtained with pixel dimensions in the order of patternrecognitionandeventreconstructionand(c)momentum 50µm to 100µm. The requirementsfrom radiologyare similar. measurement. The detectors need to withstand a total (10 yr) Mammography(∼80µmpixeldimensions)ismostdemanding particle fluence of 1015neq corresponding to a radiation dose with respect to space resolution. Some applications in autora- of about 50 Mrad. The discovery that oxygenated silicon is diography,however, require sub-µm resolutions, not attainable radiation hard with respect to the non-ionizing energy loss of protons and pions [7] saves pixel detectors at the LHC for N. Wermes is with Bonn University, Physikalisches Institut, Nussallee 12, whichtheradiationismostsevereduetotheirproximityto the D-53115Bonn,Germany,email:[email protected] Work supported by the German Ministerium fu¨r Bildung, Wissenschaft, interaction point. n+ electrode in n-bulk material sensors have Forschung und Technologie (BMBF) under contract no. 05HA8PD1, by been chosen to cope with the fact that type inversion occurs tWheestMfailnenisteurniudmer fcu¨orntrWacistsennos.chIaVftAu5nd−F1o0r6sc0h1u1ng98d,easndLabnydestheNoDrdeurhtsecinh–e afteraboutΦeq =2.5×1013cm−2.Aftertypeinversionthepn- Forschungsgemeinschaft DFG diode sits on the electrode side thus allowing the sensor to be operatedpartiallydepleted.Figure1(a) showsthe development The chip and sensor connectionis done by fine pitch bump- of the effective dopingconcentrationand the depletionvoltage ingandsubsequentflip-chipwhichisachievedwitheitherPbSn V as a function of the fluence expected for 10 years of (solder) or Indium bumpsat a failure rate of <10−4. Figure 3 dep operation at the LHC (see also [8]). Figures 1(b) and (c) show showsrowsof50µmpitchbumpsobtainedbythesetechniques. the ATLASpixelsensor wafer with 3 tiles [9] and the detector response after irradiation over two adjacent pixels showing a very homogeneouscharge collection except for some tolerable lossattheouteredgesandinanareabetweenthepixels.These areduetothebumpcontactandabiasgrid,respectively,which hasbeendesignedtoallowtestingofthesensorsbeforebonding them to the electronics ICs. Fig.2 (LEFT)INDIUM(PHOTOAMS,ROME)AND(RIGHT)SOLDER(PBSN, PHOTOIZM,BERLIN)BUMPROWSWITH50µMPITCH. Indium bumping is done using a wet lift-off technique applied on both sides (sensor and IC) [12]. The connection is obtained by thermo-compression. The Indium joint is com- paratively soft and the gap between IC and sensor is about 6µm. PbSn bumps are applied by electroplating[13]. Here the bump is galvanically grown on the chip wafer only. The bump is connected by flip-chipping to an under-bump metallization on the sensor substrate pixel. Both technologies have been successfully used with 8” IC-wafers and 4” sensor wafers. InthecaseofATLASamoduleof2.1x6.4cm2 areaconsists ofFE-chipsbump-connectedtoonesiliconsensor.TheI/Olines ofthechipsareconnectedviawirebondstoakaptonflexcircuit glued atop the sensor. The flex houses a module control chip (MCC) responsiblefor frontendtime/triggercontrolandevent building.Thetotalthicknessatnormalincidenceisinexcessof 2%X0. Figure3showsa241Am(60keVγ)sourcescan ofan ATLAS module with 46080 pixels and an amplitude spectrum obtained using the ToT analog information without clustering of pixels. Fig.1 The modules are positioned by dedicated robots on carbon- (A)EFFECTIVEDOPINGCONCENTRATIONANDDEPLETIONVOLTAGEASA carbonladders(staves)andcooledbyevaporationofafluorinert FUNCTIONOFTHEFLUENCE,(B)ATLASSILICONPIXELWAFERWITH3 liquid (C3F8) at an input temperature below −20oC in order TILES,AND(C)CHARGECOLLECTIONEFFICIENCYINTWOADJACENT to maintain the entire detector below −6oC to minimize the PIXELSAFTERIRRADIATION. damageinducedby radiation.This operationrequirespumping andthecoolingtubesmuststand16barpressureifpipeblock- ing occurs. All detector components must survive temperature Thechallengeinthedesignofthefront-endpixelelectronics cycles between −25oC and room temperature. The LHC pixel [10], [11] can be summarized by the following requirements: detectors are presently in production. A sketch of the final low power (< 50µW per pixel), low noise and threshold ATLAS pixel detector is shown in fig. 4. dispersion (together < 200e), zero suppression in every pixel, on-chip hit buffering,and small time-walk to be able to assign III. TRENDS IN HYBRID PIXELDETECTORS the hits to their respective LHC bunch crossing. The pixel ToimprovetheperformanceofHybridPixeldetectorsincer- groups have reached these goals in several design iterations tain aspects severalideas and developmentsare beingpursued. using first radiation-soft prototypes, then dedicated radhard designs, and finally using deep submicron technologies [11]. While CMS uses analog readout of hits up to the counting A. HAPS house, ATLAS obtains pulse height information by means of Hybrid(Active)PixelSensors(HAPS)[14]exploitcapacitive measuring the time over threshold (ToT) for every hit. couplingbetween pixels to obtain smaller pixel cells and pixel Fig.5 (LEFT)LAYOUTOFAHAPSSENSORWITHDIFFERENTPIXELAND Fig.3 READOUTPITCHES,(RIGHT)MEASUREDSPATIALRESOLUTIONSBY (TOP)SCANOFANATLASPIXELMODULE(46080PIXELS)WITHA241AM SCANNINGWITHAFINEFOCUSLASER. (60KEVγ)SOURCEAND(BOTTOM)THECORRESPONDINGAMPLITUDE SPECTRUMOBTAINEDUSINGTHESIGNALAMPLITUDEVIA TIME-OVER-THRESHOLD(TOT)WITHOUTCLUSTERING. andfromthemodulefront-endchips.Aninterestingalternative to the flex-kapton solution is the so-called Multi-Chip-Module TechnologydepositedonSi-substrate(MCM-D)[16].A multi- conductor-layerstructure is built up on the silicon sensor. This allows to bury all bus structures in four layers in the inactive area of the module thus avoiding the kapton flex layer and any wire bonding at the expense of a small thickness increase of 0.1% X0. In addition, the extra freedom in routing makes it possible to design pixel detectors which have the same pixeldimensionsthroughoutthesensor,i.e.withoutlargeredge pixelsusuallynecessarytoaccommodatethesensorareawhich remainsuncoveredbyelectronicsinbetweenchips.Figure6(b) illustrates the principle and fig. 6(c) shows scanning electron microphotographs(curtesyIZM,Berlin)ofaviastructuremade in MCM-D technology. Fig.4 C. Active edge 3-D silicon VIEWOFTHEATLASPIXELDETECTORINITSPRESENTDESCOPED The features of doped Si allow interesting geometries with VERSIONWITHTWOBARRELLAYERSAND2X3DISKS. respect to field shapes and charge collection. So-called 3D silicon detectors have been proposed [17] to overcome sev- eral limitations of conventional planar Si-pixel detectors, in particular in high radiation environments, in applications with pitchwith a largerreadoutpitchresultingin interleavedpixels. inhomogeneousirradiation and in applications which require a The pixel pitch is designed for best spatial resolution using large active/inactive area ratio such as protein crystallography charge sharing between neighbors while the readout pitch is [18]. A a 3D-Si-structure (fig. 7(a)) is obtained by processing tailored to the needs for the size of the front-end electronics the n+ and p+ electrodes into the detector bulk rather than cell. This way resolutions between 3µm and 10µm can be by conventional implantation on the surface. This is done obtained with pixel (readout) pitches of 100µm (200µm) as by combining VLSI and MEMS (Micro Electro Mechanical is shown in fig. 5(b) together with the layout of a prototype Systems) technologies. Charge carriers drift inside the bulk sensor in fig. 5(a). parallel to the detector surface over a short drift distance of typically50µm.Anotherfeatureis the factthatthe edgeof the B. MCM-D sensor can be a collection electrode itself thus extending the The present hybrid-pixel modules of the LHC experiments active area of the sensor to within about 10µm to the edge. use an additional flex-kapton fine-print layer on top of the Si- This results from the undepleted depth at this electrode. Edge sensor (fig. 6(a)) to provide power and signal distribution to electrodesalsoavoidinhomogeneousfieldsandsurfaceleakage Fig.6 (A)SCHEMATICVIEWOFAHYBRIDPIXELMODULESHOWINGICSBONDED VIABUMPCONNECTIONTOTHESENSOR,THEFLEXHYBRIDKAPTON LAYERATOPTHESENSORWITHMOUNTEDELECTRICALCOMPONENTS, ANDTHEMODULECONTROLCHIP(MCC).WIREBONDCONNECTIONSARE NEEDEDASINDICATED.(B)SCHEMATICLAYOUTOFAMCM-DPIXEL MODULE.(C)SEMPHOTOGRAPHOFAMCM-DVIASTRUCTURE(TOP) Fig.7 ANDACROSSSECTIONAFTERBUMPDEPOSITION(BOTTOM). (TOP)SCHEMATICVIEWOFA3DSILICONDETECTOR,(BOTTOM) COMPARISONOFTHECHARGECOLLECTIONIN3D-SILICONAND CONVENTIONALPLANARELECTRODEDETECTORS. currents which usually occur due to chips and cracks at the sensor edges. This is seen in fig. 7(b) which shows the two structures in comparison. The main advantages of 3D-silicon the characteristic systematic spatial shifts observed in CVD detectors come from the fact that the electrode distance is diamondduetothecrystalgrowth.Grainstructuresareformed short (50µm) in comparison to conventional planar devices at inwhichtrappedchargesattheboundariesproducepolarization the same total charge, resulting in a fast (1-2 ns) collection fields inside the sensor which cause the observed spatial shifts time, low (< 10V) depletion voltage and, in addition, a large of the order of 100-150 µm. active/inactive area ratio of the device. The technical fabrication is much more involved than for IV. HYBRID PIXELS IN IMAGING APPLICATIONS planar processes and requires a bonded support wafer and re- A. Counting Pixel Detectors for (Auto-)Radiography activeionetchingoftheelectrodesintothebulk.Acompromise between3Dandplanardetectors,socalledplanar-3Ddetectors A potentially new route in radiography techniques is the maintaining the large active area, use planar technology but counting of individual X-ray photons in every pixel cell. This with edge electrodes [19], obtained by diffusing the dopant approach offers many features which are very attractive for fromthedeeplyetchededgeandthenfillingitwithpoly-silicon. X-ray imaging: full linearity in the response function and a Prototype detectors using strip or pixel electronics have been principallyinfinite dynamic range, optimalexposuretimes and fabricated and show encouraging results with respect to speed agoodimagecontrastcomparedtoconventionalfilm-foilbased (3.5 ns rise time) and radiationhardness(> 1015 protons/cm2) radiography. It thus avoids over- and underexposed images. [20]. Counting pixel detectors must be considered as a very direct spin-off of the detector development for particle physics into biomedicalapplications.Theanalogpartofthepixelelectronics D. Diamond Pixel Sensors is in parts close to identicalto the onefor LHC pixeldetectors Diamond as a potentially very radiation hard material has whiletheperipheryhasbeenreplacedbythecountingcircuitry been explored intensively by the R&D collaboration RD42 at [23]. CERN [21]. Charge collection distances up to 300µm have The challenges to be addressed in order to be competitive been achieved. Diamond pixel detectors have been built with with integrating radiography systems are: high speed (> 1 pixel electronics developed for the ATLAS pixel detector and MHz) counting with a range of at least 15 bit, operation with operated as single-chip modules in testbeams [22]. A spatial very little dead time, low noise and particularly low threshold resolution of σ = 22µm has been measured with 50µm pixel operation with small threshold dispersion values. In particular pitch, compared to about 12µm with Si pixel detectors of the the last item is important in order to allow homogeneous same geometry and using the same electronics. This reveals imaging of soft X-rays. It is also mandatory for a differential Fig.9 (LEFT)2X2MULTI-CHIPCDTEMODULEWITHCOUNTINGPIXEL ELECTRONICSUSINGTHEMPECCHIPONAUSBBASEDREADOUTBOARD [27],(RIGHT)IMAGEOFACOGWHEELTAKENWITH60KEVX-RAYS. Fig.8 IMAGEOFA55FEPOINTSOURCETAKENWITHTHEMEDIPIX2COUNTING B. Protein Crystallography CHIPANDASILICONSENSOR(14X14MM2)[26]. Hybrid Pixel detectors as counting imagers lead the way to new classes of experiments in protein crystallography with synchrotron radiation [34]. Here the challenge is to image, with high rate (∼1-1.5 MHz/pixel) and high dynamic range, manythousandsofBraggspotsfromX-rayphotonsoftypically energymeasurement,realizedso farasa doublethresholdwith 12 keV or higher (corresponding to resolutions at the 1A˚ energywindowinglogic[24],[25]. A differentialmeasurement range),scattered off proteincrystals. The typicalspot size of a of the energy, exploiting the different shapes of X-ray spectra diffractionmaximumis100−200µm,callingforpixelsizesin behind for example tissue or bone, can enhance the contrast theorderof100−300µm.Thehighlinearityofthehitcounting performanceof an image. Finally,for radiographyhigh photon methodandtheabsenceofso-called”bloomingeffects”,i.e.the absorption efficiency is mandatory, which renders the not easy responseofnon-hitpixelsinthecloseneighborhoodofaBragg task of high Z sensors and their hybridization necessary. maximum, makes counting pixel detectors very appealing for Progress with counting pixel systems have been reported at protein crystallography experiments. A systematic limitation this conference[26], [27], [28]. The MEDIPIX chip developed and difficulty is the problem that homogeneous hit/count re- by the MEDIPIX collaboration [29], [25] uses 256 × 256, sponses in all pixels, also for hits at the pixel boundaries 55×55µm2 pixels fabricated in 0.25µm technology, energy or between pixels where charge sharing plays a role must windowing via two tunable discriminator thresholds, and a 13 be maintained by delicate threshold tuning. Counting pixel bitcounter.Themaximumcountrateperpixelisabout1MHz. developments are made for ESRF (Grenoble, France) [35] Figure 8 shows an image of a 55Fe point source obtained with and SLS (Swiss Light Source at the Paul-Scherrer Institute, the Medipix2 chip bonded to a 14x14 mm2, 300µm thick Si Switzerland) beam lines. The PILATUS 1M detector [36] at sensor [26]. A similar image with a single chip CdTe detector the SLS (217µm ×217µm pixels) is made of fifteen 16- and a higher energy 109Cd source (122 keV γ) has also been chip-modules each covering 8×3.5 cm2, i.e. a total area of taken and the capability to detect low energybeta decaysfrom 40×40cm2. Figure 10(a) shows a photograph of this detector tritium has been demonstrated [26]. with 15 modules and close to 106 pixels covering an area of 20×24cm2 [37]. Itis the first largescale hybridpixeldetector A Multi-Chip module with 2x2 chips using high-Z CdTe inoperation.Figure10(b)showsaBraggimageofaLysozyme sensorswiththeMPECchip[27]isshowninfig.9.TheMPEC with 10s exposure to 12 keV sychrotron X-rays [37]. Many chip features 32×32 pixels (200×200µm2), double threshold spots are contained in one pixel which is the limit of the operation, 18-bit counting at ∼1 MHz per pixel as well as achievable point spread resolution. low noise values (∼120e with CdTe sensor) and threshold In summary hybrid pixel detectors are the present state dispersion (21e after tuning) [30], [27]. A technical issue here of the art in pixel detector technology. For tracking as well is the bumping of individualdie CdTe sensors which has been as for imaging applications large (∼m2) detectors are being solved using Au-stud bumping with In-filling [31]. built or already in operation (PILATUS). On the negative side A challenge for counting pixel detectors in radiology is to are the comparatively large material budget, the complicated build large area detectors. Commercially available integrating assembling (hybridization) with potential yield pitfalls, and pixellated systems such as flat panel imagers [32], [33] consti- the difficulty to obtain high assembly yields with non-wafer tute the levelling scope. scalehigh-Zsensormaterials.Astrendswithinthistechnology, Linear e+e− Collider [38]. Very low (≪1% X0) material per detector layer, small pixel sizes (∼20µm×20µm) and a good rate capability (80 hits/mm2/ms) is required, due to the very intense beamstrahlung of narrowly focussed electron beams close to the interaction regionwhich produceelectron positron pairsinvastnumbers.Highreadoutspeedsof50MHzlinerate with 40µs frame time are necessary. Fig.11 (A)SKETCHOFAMONOLITHICPIXELDETECTORUSINGAHIGH RESISTIVITYBULKWITHPMOSTRANSISTORSATTHEREADOUT(AFTER [39]),(B)PRINCIPLEOFANMONOLITHICACTIVEPIXELSENSOR(MAPS) [41]TARGETINGCMOSELECTRONICSWITHLOWRESISTIVITYBULK MATERIAL.THECHARGEISGENERATEDANDCOLLECTEDBYDIFFUSIONIN THEFEWµMTHINEPITAXIALSI-LAYER. Among the developments of monolithic or semi-monolithic pixeldevicesperhapsthefollowingclassifyinglistcanbemade. - Non-standard CMOS on high resistivity bulk The first monolithic pixel detector, successfully operated in a particle beam already in 1992 [39], used a high resistivity p-type bulk p-i-n detector in which the junction had been created by an n-type diffusion layer. On one side, an array of ohmic contacts to the substrate served as Fig.10 (TOP)PHOTOGRAPHOFTHE20X24CM2LARGEPILATUS1MDETECTOR collection electrodes (fig. 11(a)). Due to this only simple pMOS transistor circuits sitting in n-wells were possible FORPROTEINCRYSTALLOGRAPHYUSINGCOUNTINGHYBRIDPIXEL in the active area. The technology was certainly non- DETECTORMODULES,(BOTTOM)IMAGEOFLYSOZYMETAKENWITH standard and non-commercial. No further development PILATUS[37],THEINSERTSHOWSTHATBRAGGSPOTSAREOFTEN emerged. CONTAINEDINONEPIXEL. - CMOS technology with charge collection in epi-layer Commercial CMOS technologies use low resistivity interleaved pixels and MCM-D promise better resolution and silicon whichis notsuitedforchargecollection.However, more homogeneous modules, while 3D-silicon detectors can in some technologies an epitaxial silicon layer of a few address applications in high radiation environments or those to 15µm thickness can be used [40], [41], [42]. The where a large active/inactive area ratio is mandatory. generatedchargeiskeptintheepi-layerbypotentialwells at the boundary and reaches an n-well collection diode V. MONOLITHICAND SEMI-MONOLITHICPIXEL by thermal diffusion (fig. 11(b)). The signal charge is DETECTORS therefore very small (<1000e) and low noise electronics Monolithic pixel detectors, in which amplifying and logic is a real challenge. As this development uses standard circuitryaswellastheradiationdetectingsensorareoneentity, processing technologies it is potentially very cheap. producedinacommerciallyavailabletechnology,arethedream Despite using CMOS technology the potential of full ofsemiconductordetectordevelopers.Thiswillnotbepossible CMOS circuitry in the active area is not available (only in the near future without compromising. Developments and nMOS) because of the n-well/p-epi collecting diode trends in this direction have much been influenced by R&D which does not permit other n-wells. for vertex tracking detectors at future colliders such as a FET transistor is implanted in every pixel on a sidewards depleted[56]bulk.Electronsgeneratedbyradiationin the bulk are collected in a potential minimum underneath the transistorchannelthusmodulatingitscurrent(fig.13).The bulk is fully depleted rendering large signals. The small capacitanceoftheinternalgateofferslownoiseoperation. Both together can be used to fabricate thin devices. The Fig.12 sensor technology is non-standard and the operation of (A)CROSSSECTIONTHROUGHAMONOLITHICCMOSONSOIPIXEL DEPFET pixel detectors requires separate steering and amplification ICs (see below). DETECTORUSINGHIGHRESISTIVITYSILICONBULKINSOLATEDFROMTHE LOWRESISTIVITYCMOSLAYERWITHCONNECTINGVIASINBETWEEN [15],[47],(B)CROSSSECTIONTHROUGHASTRUCTUREUSING AMORPHOUSSILICONONTOPOFSTANDARDCMOSVLSIELECTRONICS [48],[49]. - CMOS on SOI (non-standard) In order to exploit high resistivity bulk material, the authors of [47] develop pixel sensors with full CMOS circuitry using the Silicon-on-Insulator (SOI) technology. The connection to the charge collecting bulk is done Fig.13 by vias (fig. 12(a)). The technology offers full charge PRINCIPLEOFOPERATIONOFADEPFETPIXELSTRUCTUREBASEDONA collection in 200–300µm high resistive silicon with full SIDEWARDSDEPLETEDDETECTORSUBSTRATEMATERIALWITHAN CMOS electronics on top. At present the technology is IMBEDDEDPLANARFIELDEFFECTTRANSISTOR.CROSSSECTION(LEFT)OF however definitely not a commercial standard and the HALFAPIXELWITHSYMMETRYAXISATTHELEFTSIDE,ANDPOTENTIAL development is still in its beginnings. PROFILE(RIGHT). - Amorphous silicon on standard CMOS ASICs Hydrogenated amorphous-Silicon (a-Si:H), where the Because CMOS active pixels and DEPFET pixels are most H-content is up to 20%, can be put as a film on top advanced in their development, they are discussed in more of CMOS ASIC electronics. a-Si:H has been studied detail below. as a sensor material long ago and has gained interest again [48], [49] with the advancement in low noise, low A. CMOS Active Pixels power electronics. The signal charge collected in the <30µm thick film is in the range of 500-1500 electrons. The main difference of CMOS monolithic active pixel sen- A cross section through a typical a-Si:H device is shown sors [40] to CMOS camera chips is that they are larger in in fig. 12(b). From a puristic view it is more a hybrid area and must have a 100% fill factor for efficient particle technology,butthemaindisadvantageofthehybridization detection. Standard commercial CMOS technologies are used connection is absent. The radiation hardness of these which have a lightly doped epitaxial layer between the low detectors appears to be very high >1015cm−2 due to resistivity silicon bulk and the planar processing layer. The the defect tolerance and defect reversing ability of the epi-layer is – technologydependent– at most 15µm thick and amorphous structure and the larger band gap (1.8 eV). can also be completely absent. Collaborating groups around The carrier mobility is very low (µ = 2-5 cm2/Vs, µ = IReS&LEPSI[42],[43],RAL[44]andIrvine-LBNL-Ohio[45] e h 0.005cm2/Vs), i.e. essentially onlyelectronscontributeto use similar approaches to develop large scale CMOS active the signal.Basically anyCMOScircuitandICtechnology pixels also called MAPS (Monolithic Active Pixel Sensors) can be used and technology changes are uncritical for [41].Prototypedetectorshavebeenproducedin0.6µm,0.35µm this development.For high-Z applicationspoly-crystalline and 0.25µm CMOS technologies [54], [52]. HgI constitutes a possible semiconductor film material. In the MAPS device (fig. 11(b)) electron-hole pairs created 2 The potential advantages are small thickness, radiation by ionization energy loss in the epitaxial layer remain trapped hardness, and low cost. The development is still in its between potential barriers on both sides of the epi-layer and beginnings and – as for CMOS active pixel sensors – a reach, by thermal diffusion, an n-well/p-epi collection diode. real challenge to analog VLSI design. Thesensorisdepletedonlydirectlyunderthen-wellcollection diode where full charge collection is obtained; the charge - Amplification transistor implanted in high resistivity bulk collectionisincompleteinallotherareas.Withsmallpixelcells In so-called DEPFET pixel sensors [50] a JFET or MOS- collectiontimes in the orderof 100nsare obtained.The signal is small, some hundred to 1000e, depending on the thickness differenttechnologies.Itbecomesthinnerforsmallerprocesses of the epi-layer. A chip submission using a process with no and with it also the number of produced signal electrons epi-layer, but with a low doped substrate of larger resistivity vanishes. Only a few processing technologies are suited. With hasalsobeentried(MIMOSA-4)[53]whichprovedtofunction the rapid changeof commercialprocesstechnologiesthis is an with high detection efficiency for minimum ionizing particles. issue of concern. The fact that good detector performance has Thisrenderstheuncertaintyanddependenceonthefutureofthe been seen with non-epi wafers material provides some relief. epi-layer thickness in different technologies less constraining. Furthermore, in the active area, due to the n-well collection Matrix readout performed using a standard 3-transistor cir- diode,noCMOS(onlynMOS)circuitryispossible.Thevoltage cuit (line select, source-follower stage, reset) commonly em- signals are very small (∼mV), of the same order as transistor ployed by CMOS matrix devices, but can also include cur- threshold dispersions requiring very low noise VLSI design. rent amplification and current memory [54]. For an image The radiation tolerance of CMOS pixel detectors for particle two complete frames are subtracted from each other (CDS) detection beyond 1012 neq is still a problem although the which suppressesswitching noise. Noise figuresof 10-30eand achieved tolerance should be sufficient for a Linear Collider. S/N ∼20 have been achieved with spatial resolutions below Improved readout concepts and device development for high 5µm. The radiation hardness of CMOS devices is always a rate particle detection at a linear collider is underdevelopment crucialquestionforparticledetectioninhighintensitycolliders. [38].FortheupgradeoftheSTARmicrovertexdetectortheuse MAPS appear to sustain non-ionizing radiation (NIEL) to of CMOS active pixels is planned [54]. ∼1012n while the effectsof ionizing radiation damage (IEL) eq are at present still unclear [55]. Apart from this the focus of B. DEPFET Pixels further development lies in making larger area devices and The Depleted Field Effect Transistor structure (DEPFET) in increasing the charge collection performance in the epi- [50] providesdetectionandamplificationpropertiesjointlyand layer. For the former, first results in reticle stitching, that is hasbeenexperimentallyconfirmedandsuccessfullyoperatedas electrically connecting IC area over the reticle border, have singlepixelstructuresandaslarge(64x64)pixelmatrices[51]. been encouraging (fig. 14(a)) [52]. Full CMOS stitching over The DEPFET detector and implanted amplification structure is thereticleboarderstillhastobedemonstrated.Thepoorcharge showninfig.13.Sidewardsdepletion[56]providesaparabolic collectionbythermaldiffusionisanotherareacallingforideas. potential which has – by appropriate biasing and a so-called Approachesusingmorethanonen-wellcollectingdiode[44],a deep-n implantation – a local minimum for electrons (∼1µm) photo-FET[54],[42],[43],oraphoto-GATE[46]arecurrently underneath the transistor channel. The channel current can be investigated. The photo-FET called technique (fig. 14(b)) has steered and modulated by the voltage at the external gate and features similar to those of DEPFET pixels (see below). A - important for the detector operation - also by the deep- pMOS FET is implanted in the charge collecting n-well and n potential (internal gate). Electrons collected in the internal signalchargesaffectitsgatevoltageandmodulatethetransistor gate are removed by a clear pulse applied to a dedicated current. CLEAR contact outside the transistor region or by other clear mechanisms. The very low input capacitance (∼ few fF) and the in situ amplification makes DEPFET pixel detectors very attractive for low noise operation[57]. Amplificationvalues of 400 pA per electron collected in the internal gate have been achieved.Further amplificationenters at the second levelstage (current based readout chip). DEPFET pixelsare currentlybeingdevelopedfor three very different application areas: vertex detection in particle physics [58], [59],X-rayastronomy[60]andforbiomedicalautoradio- graphy [57]. Figure 15 summarizes the achieved performance innoiseandenergyresolutionwithsinglepixelsstructures(fig. Fig.14 15(a))andinspatialresolutionswith64x64pixelsmatrices(fig. (A)SEAMLESSSTITCHINGINBETWEENWAFERRETICLESTOOBTAIN 15(b)). With round single pixel structures noise figures of 2.2e LARGEAREACMOSACTIVEPIXELS,(B)SCHEMATICOFAPMOS at room temperature and energy resolutions of 131 keV for 6 PHOTO-FETTOIMPROVECHARGECOLLECTIONINMAPS. keVX-rayshavebeenobtained.Withsmall(20x30µm2)linear structures fabricated for particle detection at a Linear Collider the noise figures are about10e. The spacial resolution read off Aboveall,theadvantagesofafullyCMOSmonolithicdevice from fig. 15(b) in matrices operated with 50 kHz line rates is relate to the adoption of standard VLSI technology and its [59] resultinglowcostpotential(potentially∼25$percm2).Inturn the disadvantages also come largely from the dependence on σxy =(4.3±0.8)µm or 57mLPm for 22keVγ commercialstandards.Thethicknessof the epi-layervariesfor σ =(6.7±0.7)µm or 37LP for 6keVγ xy mm The capability to observe tritium in autoradiographical ap- Monolithicorsemi-monolithicdetectors,inwhichdetectorand plications is shown in fig. 15(c). readout ultimately are one entity, are currently developed in variousforms,largelydrivenbytheneedsforparticledetection at future colliders. CMOS active pixel sensors using standard commercialtechnologiesonlowresistivitybulkandSOIpixels, a-SI:H, and DEPFET-pixels, which try to maintain high bulk resistivity for charge collection, classify the different ways to monolithic detectors which are presently carried out. ACKNOWLEDGMENT The author would like to thank W. Dulinski, M. Caccia, P. Jarron, P. Russo, C. da Via’, and S. Parker for providing new material and information for this presentation. Fig.15 REFERENCES (A)RESPONSEOFASINGLEDEPFETPIXELSTRUCTURETO6KEVX-RAYS [1] Technical Design Report of the ATLAS Pixel Detector, FROMAN55FESOURCE,(B)IMAGEOBTAINEDWITH55FEANDATEST CERN/LHCC/98-13(1998); N.Wermes,DesignandPrototypePerformanceoftheATLASPixel CHART,THESMALLESTSTRUCTURESHAVEAPITCHOF50µMANDARE Detector, Nucl.Inst.Meth. A447(2002)121-128. 25µMWIDE,(C)3HLABELLEDLEAF. [2] CMS, the Tracker System Project, Technical Design Report, CERN/LHCC/98-006(1998); W. Erdmann, The CMS pixel detector, Nucl. Inst. Meth. A447 (2000)178-83. The verygoodnoisecapabilitiesof DEPFET pixelsare very [3] ALICETechnical DesignReport, CERN/LHCC99-12(1999); important for low energy X-ray astronomy and for autora- P. Chochula et al., The ALICE Silicon Pixel Detector, Nucl. Phys. diography applications. For particle physics, where the signal A715(2003)849-852. [4] T. Gys, The pixel hybrid photon detectors for the LHCb-RICH charge is large in comparison, this feature is used to design project, Nucl.Inst.Meth.A465(2001)240-246. verythindetectors(∼50µm)withverylowpowerconsumption [5] A.Baldit etal.,CERN-SPSC-2000-010(2000); whenoperatedasa row-wiseselected matrix[59]. Thinningof C.Lourenco,E.Radermacher, CERN-SPSC-2002-025(2002). [6] S. Kwan (BTeV Collaboration), The BTEV Pixel and Microstrip sensors to a thickness of 50µm using a technology based on Detectors, Nucl.Inst.Meth.A511(2003)48-51. deep anisotropic etching has been successfully demonstrated [7] ROSE Collaboration, G.Lindstro¨m et al., Nucl. Inst. Meth. A466 [61]. For the development of DEPFET pixels for a Linear (2001)308-326; G.Lindstro¨m,Radiation Damage inSilicon Detectors, Nucl. Inst. Collider matrix row rates of 50 MHz and frame rates for Meth.A512(2003)92. 520x4000pixelsof25kHz,read outattwo sides, are targeted. [8] O. Krasel et al., Measurement of the trapping time constants in Sensors with cell sizes of 20x30µm2 have been fabricated and Proton-Irradiated Silicon Pad Detectors, IEEE Nuclear Science SymposiumPortland 2003,theseproceedings. complete clearing of the internal gate, which is important for [9] M. S. Alam et al., The ATLAS Silicon Pixel Sensors, Nucl. Inst. highspeedon-chippedestalsubtraction,hasbeendemonstrated Meth.A456(2001)217-232. [59]. A large matrix is readout using sequencer chips for row [10] L.Blanquart etal. Analogfront-end cell designed inacommercial 0.25mprocessfortheATLASpixeldetector atLHC, IEEETrans. selectionand currentbasedchip forcolumnreadout[59]. Both Nucl.Sci.vol.49(4),(2002)1778-1782 chips have been developed at close to the desired speed for [11] L.Blanquartetal.,FE-I2:AFront-EndReadoutChipDesignedina a Linear Collider. The estimated power consumption for a commercial0.25µmProcessfortheATLASPixeldetectoratLHC, IEEENuclearScienceSymposiumPortland2003,theseproceedings. five layer DEPFET pixel vertex vertex detector is only 5W [12] C.Gemme,StudyofIndiumbumpingfortheATLASpixeldetector, rendering a very low mass detector without cooling pipes Nucl.Inst.Meth.A465(2001)200–203. feasible.ThemostrecentstructuresuseMOSFETsasDEPFET [13] G. Engelmann, J. Wolf, L. Dietrich, S. Fehlberg, O. Ehrmann and H. Reichl, Proc. 1999 Joint International Meeting of the Electro- transistors for which the radiation tolerance still has to be chemical Society,Hawaii 1999,Vol.99-2p.907; investigated. J. Wolf, PbSn60 Solder Bumping by Electroplating, Pixel2000 Workshop,Genova, Italy,June2000. [14] W. Kucewicz, G. Deptuch, A. Zalewska, M. Battaglia, K. Oster- VI. SUMMARY berg, R. Orava, M. Caccia, C. Meroni, G. Vegni, P. Grabiec, B. Driven by the demands for high spatial resolution and high Jaroszewicz, J. Marczewski, Capacitively coupled active pixel sen- sorswithanalogreadoutforfuturee+e-colliders,ActaPhys.Polon. rate particle detection in high energy physics, semiconductor B30(1999)2075–2083. pixel detectors have started to also become exploited for [15] M.Amati,M.Baranski,A.Bulgheroni,M.Caccia,K.Domanski,P. imaging applications. Hybrid pixel detectors, in which sensor Grabiec, M.Grodner, B.Jaroszewicz, W.Kucewicz, K.Kucharski, S.Kuta, W.Machowski, J.Marczewski, H.Niemiec, M.Sapor, D. and electronic chip are separate entities, connected via bump Tomaszewski,HybridActivePixelSensorsandSOIInspiredOption, bonding techniques represent today’s state of the art for both, Nucl.Inst.Meth.A511(2003)265–270. particledetectionandimagingapplications.Newtrendsinclude [16] P.Gerlach, C.Linder, K.H.Becks, Nucl.Inst.Meth. A473(2001) 102-106andreferences therein; interleaved pixels having different pixel- and readout-pitches, M. To¨pper, Intl. Conf. on High Density Packaging and MCMs, MCM-D structuresand 3D-silicon detectors with active edges. Denver, Colorado, April1999. [17] C.Kenney,S.Parker,J.Segal,C.Storment,Silicondetectorswith3- S.Watanabe,N.Wermes,AcountingCdTepixeldetectorforhardX- Delectrodearrays:fabricationandinitialtestresults, IEEETrans. rayandgamma-rayimaging, IEEETrans.Nucl.Sci.vol.48,(2001) Nucl.Sci.vol.46(4),(1999)1224–1236; 2401-2404. C.J. Kenney, S.I. Parker, B. Krieger, B. Ludewigt, T.P. Dubbs, H. [32] T. Ducourant, M. Michel, G. Vieux, T. Peppler, J.C. Trochet, R.F. Sadrozinski, Observation of beta and Xrays with 3-D-architecture Bastiaens,F.Busse, Optimizationofkeybuildingblocksforalarge- siliconmicrostripsensors, IEEETrans.Nucl.Sci.vol.48(2),(2001) arearadiographic and fluoroscopic dynamic digital X-raydetector 189-193; based on a-Si:H/CsI:Tl flat panel technology, Proc. SPIE 3977 C.J.Kenney,S.Parker, E.Walckiers, Resultsfrom3-Dsilicon sen- (2000)14-25; sorswithwallelectrodes:near-cell-edgesensitivitymeasurementsas N.Jungetal.,Proc.SPIE3336(1998)396-407. apreviewofactive-edgesensors, IEEETrans.Nucl.Sci.vol.48(6), [33] O.Tousignant,M.Choquette,Y.Demers,L.Laperriere,J.Leboeuf, (2001)2405. M.Honda,M.Nishiki,A.Takahashi,A.Tsukamoto,Progressreport [18] E. M. Westbrook, Pixels and Proteins: Better Detectors for Bi- on the performance of real-time selenium flat-panel detectors for ological Crystallography, Review given at IEEE Nuclear Science directX-rayimaging, Proc.SPIE4682(2002)503-510. Symposium,PortlandOct.2003,theseproceedings; [34] H. Graafsma, Detector Needs at Current and Future Synchrotron S.I. Parker, C.J. Kenney, J.D. Segal, E. Westbrook, J. Hasi, C. Da Sources,Review given atIEEENuclear Science Symposium,Port- Via’, E. Mandelli, G. Meddeleer, J. Morse, Direct X-ray Detector landOct.2003,theseproceedings. with 3D Silicon Pixel Sensors for Structural Molecular Biology, [35] P. Delpierre, J.F. Berar, L. Blanquart, B. Caillot, J.C. Clemens, IEEENuclearScienceSymposiumPortland2003,theseproceedings. C.Mouget, IEEETrans.Nucl.Sci.vol.48(4),(2001)987-991; [19] C.J. Kenney, E. Westbrook, J.D. Segal, S.I. Parker, J. Morse, J. P. Delpierre, J.F. Berar, L. Blanquart, N. Boudet, P. Breugnon, Hasi,C. DaVia’, FirstResults from Active-Edge, Planar radiation B. Caillot, J.C. Clemens, I. Koudobine, C. Mouget, R. Potheau, Sensors, IEEE Nuclear Science Symposium Portland 2003, these I. Valin, NSS-MIC/IEEE, San Diego (USA), Nov 4-10, 2001, proceedings. accepted forpublication inIEEETrans.Nucl.Sci.,Sept.2002. [20] S.I. Parker, A. Kok, C. Da Via’, J. Hasi, G. Anelli, P. Jarron, C.J. [36] C.Bro¨nnimannetal.,JournalofSyn.Rad.7(2000)301. Kenney,J.D.Segal,MeasurementsofFastRise-timePulsesfrom3D [37] E. Eikenberry, C. Broennimann, G. Huelsen, H. Toyokawa, M. Silicon Sensors, IEEENuclear Science Symposium Portland 2003, Suzuki,R.Horisberger,B.Schmitt,C.Schulze-Briese,T.Tomizaki, theseproceedings; The PILATUS X-ray Detector, 13th International Workshop on C.J. Kenney, E. Westbrook, J.D. Segal, E. Perozzielo, S.I. Parker, Room-Temperature Semiconductor X- and Gamma-Ray Detectors, J. Hasi, C. Da Via’, J. Morse, First Results from Active Edge 3D Portland2003. Architecture radiation Sensors, IEEE Nuclear Science Symposium [38] TESLATechnical DesignReport, Vol.IV,DESY-01-011(2001). Portland2003,theseproceedings. [39] W.Snoeys,J.Plummer,S.Parker,C.Kenney,Anewintegratedpixel [21] RD42 Collaboration, Diamond as a particle detector, Proc. Itatial detector for high energy physics, IEEE Trans. Nucl. Sci. vol.39, Phys.Socienty, Torino,vol.52(1996),105; (1992)1263–1269. R.Wedenigetal.,CVDDiamondPixelDetectors forLHCExperi- [40] B.Dierickx,G.Meynants,D.Scheffer,Near100%fillfactorCMOS ments,Nucl.Phys.B.(Proc.Suppl.)78(1999)497. active pixel, Proc. IEEE CDD&AIS Workshop, Brugge, Belgium, [22] M.Keiletal.,NewResultsonDiamondPixelSensorsusingATLAS June1997; FrontendElectronics, Nucl.Inst.Meth.A501(2003)153-159. G. Meynants, B. Dierickx, D. Scheffer, CMOS active pixel image [23] P.Fischer, Anareaefficient 128 channel counter chip, Nucl. Inst. sensor with CCD performance, Proceedings SPIE - Int. Soc. Opt. Meth.A378(1996)297–300. Eng.(USA),vol.3410(1998)68–76,9. [24] P. Fischer, A. Helmich, M. Lindner, N. Wermes, L. Blanquart, A [41] R. Turchetta, J.D. Bertst, B. Casadei, G. Claus, C. Colledani, W. PhotonCountingChipwithEnergyWindowing, IEEETrans.Nucl. Dulinski,Y.Hu,D.Husson,J.-P.Le-Normand,J.L.Riester,G.Dep- Sci.vol.47(3),(2000)881–884; tuch,U.Goerlach,S.Higueret,M.Winter,Amonolithicactivepixel M.Lindner,L.Blanquart,P.Fischer,H.Kru¨ger,N.Wermes,Medical sensor for charged particle tracking and imaging using standard X-ray imaging with energy windowing, Nucl. Inst. Meth. A465 VLSICMOS technology, Nucl. Inst. Meth. A458 (2001) 677-689; (2000)229–234. G.Claus, C.Colledani, W.Dulinski, D.Husson,R.Turchetta, J.L. [25] X.Llopart,M.Campbell,R.Dinapoli,D.SanSegundo,E.Pernigotti, Riester, D. Deptuch, G. Orazi, M. Winter, Particle tracking using Medipix2:A64-kpixelreadoutchipwith55-mumsquareelements CMOS monolithic active pixel sensor, Nucl. Inst. Meth. A465 working in single photon counting mode, IEEE Trans. Nucl. Sci. (2001)120–124. vol.49(5),(2002)2279–2283. [42] G. Deptuch, W. Dulinski, Y. Gornushkin, C. Hu-Guo, I. Valin, [26] G.Mettivier,M.C.Montesi,P.Russo,DigitalAutoradiographywith Monolithic Active Pixel Sensor with on-pixel amplification and a Medipix2 Hybrid Silicon Pixel Detector, Poster given at IEEE Double Sampling Operation, Nucl. Inst. Meth. A512 (2003) 299- Medical ImagingConference, Portland Oct.2003; 309; G. Mettivier, M.C. Montesi, P. Russo, Tritium digital autoradiog- G. Claus et al., Monolithic Active Pixel Sensors for a Linear raphy with a Medipix2 hybrid silicon pixel detector, submitted to Collider, Nucl.Inst.Meth.A473(2001)83–85. Nucl.Inst.Meth.A [43] G.Deptuchetal.,MonolithicActivePixelSensorwithIn-pixelDou- [27] M. Lo¨cker, P. Fischer, M. Kouda, S. Krimmel, H. Kru¨ger, M. ble Sampling and Columns-level Disrcrimination, Poster submitted Lindner, K. Nakazawa, T. Takahashi, N. Wermes, Single photon tothis conference. Counting X-ray Imaging with Si and CdTe Single Chip Pixel [44] R. Turchetta et al., Monolithic Active Pixel Sensors (MAPS) for Detectors and Multichip Pixel Modules, IEEE Nuclear Science Particle Physics and Space Science, Proceedings VERTEX 2003, SymposiumPortland 2003,theseproceedings. LakeWindermere, UK,September2003. [28] F.Edling,R.Brenner,N.Bingefors,K.Fransson,L.Gustafsson,L. [45] H. Matis, F. Bieser, S. Kleinfelder, G. Rai, F. Retiere, H. Ritter, delRiscoNorrlid,C.Ro¨nnqvist,AHybridPixelDetectorwithTwo K.Singh,S.Wurzel,H.Wiemann,E.Yamamoto,ChargedParticle Counters forX-rayImaging,PostergivenatIEEE-NSSymposium, Detection using a CMOS Active Pixel Sensor, IEEE Trans. Nucl. PortlandOct.2003,theseproceedings. Sci.vol.50,(()2003)1020-1025. [29] Medipix Collaboration, S.R. Amendiola et al., Nucl. Inst. Meth. [46] S. Kleinfelder, F. Bieser, Y. Chen, I. Kotov, H.S. MAtis, M. A466(2001)74-78. Oldenburg, F. Retiere, K.D. Singh, H.H. Wiemann, E. Yamamoto, [30] P. Fischer, J. Hausmann, A. Helmich, M. Lindner, N. Wermes, L. NovelIntegratedCMOSPixelStructuresforVertexDetectors,IEEE Blanquart, Acounting pixel readout chip andsensor system for X- NuclearScience SymposiumPortland 2003,theseproceedings. rayimaging,IEEETrans.Nucl.Sci.,Vol46,No.4(1999)1070; [47] J. Marczewski, K. Domanski, P. Grabiec, M. Grodner, B. L.Blanquart,V.Bonzon,G.Comes,P.Delpierre,P.Fischer,J.Haus- Jaroscewicz, A. Kociubinski, D. Tomaszewski, W. Kucewicz, S. mann,M.Keil,M.Lindner,S.Meuser,N.Wermes, PixelReadout Kuta,W.Machowski,H.Niemiec,MSapor,M.Caccia,SOIActive ElectronicsforLHCandBiomedicalApplications, Nucl.Inst.Meth. Pixel Detectors of Ionizing Radiation - Technology and Design A439(2000)80–90. Development, IEEE Nuclear Science Symposium Portland 2003, [31] P.Fischer,M.Kouda,H.Kru¨ger,M.Lindner,G.Sato,T.Takahashi, theseproceedings.

See more

The list of books you might like

Upgrade Premium
Most books are stored in the elastic cloud where traffic is expensive. For this reason, we have a limit on daily download.
DMCA & Copyright: Dear all, most of the website is community built, users are uploading hundred of books everyday, which makes really hard for us to identify copyrighted material, please contact us if you want any material removed.
Copyright @ PDFdrive.to, 2022 | We love our users